Antibiotic Resistance Protocols, Fourth Edition Springer, 2024

Methods in
Molecular Biology 2833
Stephen H. Gillespie Editor
Antibiotic
Resistance
Protocols
Fourth Edition
METHODS IN MOLECULAR BIOLOGY
Series Editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, UK
For further volumes:

http://www.springer.com/series/7651
For over 35 years, biological scientists have come to rely on the research protocols and
methodologies in the critically acclaimed Methods in Molecular Biology series. The series was
the first to introduce the step-by-step protocols approach that has become the standard in all
biomedical protocol publishing. Each protocol is provided in readily-reproducible step-by
step fashion, opening with an introductory overview, a list of the materials and reagents
needed to complete the experiment, and followed by a detailed procedure that is supported
with a helpful notes section offering tips and tricks of the trade as well as troubleshooting
advice. These hallmark features were introduced by series editor Dr. John Walker and
constitute the key ingredient in each and every volume of the Methods in Molecular Biology
series. Tested and trusted, comprehensive and reliable, all protocols from the series are
indexed in PubMed.
Antibiotic Resistance Protocols
Fourth Edition
Edited by
Stephen H. Gillespie
Division of Infection and Global Health, School of Medicine, University of St Andrews, St Andrews, Fife, UK
Editor
Stephen H. Gillespie
Division of Infection and Global
Health, School of Medicine
University of St Andrews
St Andrews, Fife, UK
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-0716-3980-1 ISBN 978-1-0716-3981-8 (eBook)
https://doi.org/10.1007/978-1-0716-3981-8
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Science+Business Media, LLC, part
of Springer Nature 2024
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and
retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter
developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations
and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to
be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been
made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer
Nature.
The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.
If disposing of this product, please recycle the paper.

Preface
Much has changed in the world of antibiotics resistance research since the first volume of this
protocol series was published in 2001. The most important factor is that the tide of
antibiotic resistance is rising steadily [1]. Fortunately, recognition of the scale of the
problem has also occurred with the publication of national and international reports that
set out the significant danger to human health posed by this problem [2]. International
organizations to fight antibiotic resistance, such as the Global Antibiotic Research and
Development Partnership (GARDP), have been established with the aim of developing
five new treatments by 2025 [3]. Similarly, academic groups and consortia have been
established to address antibiotic resistance from fundamental science through to implemen-
tation science.
When the first edition of this book was published, it took perhaps as much as a year to
sequence and assemble a genome. Now with the availability of high-throughput platforms,
sequencing of multiple strain has become commonplace, and it is now starting to take its
place in routine microbiology practice [4]. There remain many questions about how samples
are collected, strains isolated and sequenced, and the results integrated in the microbiolo-
gical workflow. Several of our chapters address these issues. Sequencing is also a powerful
tool to understand the development, emergence, and spread of resistance, as is discussed in
this volume. The challenge of integrating sequencing methodologies is not trivial, and this
subject is addressed in the current edition.
We now have molecular methods to detect and quantify the number of live bacteria that
will aid in the monitoring of treatment response in difficult infections caused by slow-
growing organisms such as Mycobacterium tuberculosis [5]. The sensitivity and specificity
of such a technique can be harnessed to make a diagnosis from specimens other than
sputum, such as stool [6], that may make it easier to diagnose tuberculosis in children.
Despite the increasing role of sequencing, growing microorganisms will probably
remain a major function of microbiology laboratories. The question is, however, can we
do it more rapidly and more cleverly? We present a number of techniques in this volume that
unequivocally answer that question in the affirmative with novel methods to test resistance
and interactions between antibiotics, physiological conditions or using innovative tools like
the hollow fiber, which can mimic human pharmacology well, or Raman spectroscopy that
enables changes in the bacteria to be measured in real time.
Sometimes the problem of antibiotic resistance can seem very large and only really be
encompassed numerically. In this volume, we present mathematical models that can describe
resistance within host [7].
v
vi Preface
The main purpose of this volume is to present a range of different techniques that the
reader can use to address important questions in antibiotic resistance. In the coming years, I
believe several of the techniques presented here will play an important role in answering
questions as to how to control antibiotic resistance, how to develop new agents, and how to
address the problems posed by microbes that have become resistant to our antibiotics.
St Andrews, Fife, UK Stephen H. Gillespie
References
1. Collaborators AR, Murray CJ, Ikuta KS, et al (2022) Global burden of bacterial antimicrobial
resistance in 2019: a systematic analysis. Lancet Lond Engl 399:629–655
2. O’Neill J (2016) Tackling drug resistant infections globally. Final report and recomendations, HM
Government, London
3. Global Antibiotic Research and Development Partnership. https://gardp.org/
4. Parcell BJ, Gillespie SH, Pettigrew KA, et al (2020) Clinical perspectives in integrating whole genome
sequencing into the investigation of healthcare and public health outbreaks – hype or help? J Hosp
Infect 109:1–9
5. Sabiiti W, Azam K, Farmer ECW, et al (2020) Tuberculosis bacillary load, an early marker of disease
severity: the utility of tuberculosis molecular bacterial load assay. Thorax 75:606–608
6. Musisi E, Sessolo A, Kaswabuli S, et al (2022) High mycobacterium tuberculosis bacillary loads
detected by tuberculosis molecular bacterial load assay in patient stool: a potential alternative for
nonsputum diagnosis and treatment response monitoring of tuberculosis. Microbiol Spectr 10:
e02100-21
7. Bowness R, Chaplain MAJ, Powathil GG, et al (2018) Modelling the effects of bacterial cell state and
spatial location on tuberculosis treatment: Insights from a hybrid multiscale cellular automaton model.
J Theor Biol 446:87 100
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1 Using the Zebrafish Larval Model of Infection to Investigate Antibiotic
Efficacy and Combination Treatments Against Staphylococcus aureus. . . . . . . . . . . 1
Amy K. Tooke and Simon J. Foster
2 Three-Dimensional Rotary Culture to Model Mycobacterial
Biofilms in Low-Shear Detergent-Free Liquid Suspension . . . . . . . . . . . . . . . . . . . 11
Daire Cantillon and Simon J. Waddell
3 Rapid Gene Silencing Followed by Antimicrobial Susceptibility
Testing for Target Validation in Antibiotic Discovery . . . . . . . . . . . . . . . . . . . . . . . . 23
Chris Daniel, Sam Willcocks, and Sanjib Bhakta
4 Modified HT-SPOTi: An Antimicrobial Susceptibility Testing
to Evaluate Formulated Therapeutic Combinations
Against Bacterial Growth and Viability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Anushandan Navaratnarajah, Chris Daniel, and Sanjib Bhakta
5 Investigating Combination Therapy as a Means
to Enhance Activity and Repurpose Antimicrobials . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Robert J. H. Hammond
6 Investigating Photoactive Antimicrobials as Alternatives (or Adjuncts)
to Traditional Therapy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Robert J. H. Hammond and Marianna Leite De Avellar
7 Using Hollow Fiber to Model Treatment
of Antimicrobial-Resistant Organisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
8 A Microtiter Plate Assay at Acidic pH to Identify Potentiators
that Enhance Pyrazinamide Activity Against Mycobacterium tuberculosis. . . . . . . . 65
Christopher William Moon, Eleanor Porges,
Stephen Charles Taylor, and Joanna Bacon
9 Within-Host Mathematical Models of Antibiotic Resistance . . . . . . . . . . . . . . . . . . 79
Aminat Yetunde Saula, Gwenan Knight, and Ruth Bowness
10 Use of Individual-Based Mathematical Modelling
to Understand More About Antibiotic Resistance Within-Host. . . . . . . . . . . . . . . 93
Aminat Yetunde Saula, Christopher Rowlatt, and Ruth Bowness
11 Monitoring Live Mycobacteria in Real-Time Using
a Microfluidic Acoustic-Raman Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Mingzhou Chen, Vincent Baron, Björn Hammarström,
Robert J. H. Hammond, Peter Glynne-Jones,
Stephen H. Gillespie, and Kishan Dholakia
12 Phylogenetic Survival Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Arturo Torres Ortiz and Louis Grandjean
vii
viii Contents
13 Rapid Drug Susceptibility Testing to Preserve Antibiotics. . . . . . . . . . . . . . . . . . . . 129

Stephen H. Gillespie and Robert J. H. Hammond
14 Quantifying Viable M. tuberculosis Safely Obviating
the Need for High Containment Facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Bariki Mtafya, Emmanuel Musisi, Paschal Qwaray, Emanuel Sichone,
Natasha Walbaum, Nyanda Elias Ntinginya,
Stephen H. Gillespie, and Wilber Sabiiti
15 Improved Diagnosis and Treatment Monitoring of Tuberculosis
Using Stool and the Tuberculosis Bacterial Load Assay (TB-MBLA) . . . . . . . . . . 153
Emmanuel Musisi, Bariki Mtafya, William Saava Wambi,
Josephine Zawedde, Abdulwahab Sessolo, Willy Ssengooba,
Natasha Walbaum, Nyanda Elias Ntinginya,
Stephen H. Gillespie, and Wilber Sabiiti
16 Application of Pathogen Genomics to Outbreak Investigation. . . . . . . . . . . . . . . . 161
Benjamin J. Parcell, Kerry A. Pettigrew, and Katarina Oravcova
17 Use of Whole Genome Sequencing for Mycobacterium tuberculosis
Complex Antimicrobial Susceptibility Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Vanessa Mohr, Lindsay Sonnenkalb, Christian Utpatel, Ivan Barilar,
Margo Diricks, Viola Dreyer, Stefan Niemann,
Thomas A. Kohl, and Matthias Merker
18 Use of Whole Genome Sequencing for Mycobacterium tuberculosis
Complex Antimicrobial Susceptibility Testing: From Sequence
Data to Resistance Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Viola Dreyer, Lindsay Sonnenkalb, Margo Diricks, Christian Utpatel,
Ivan Barilar, Vanessa Mohr, Stefan Niemann,
Thomas A. Kohl, and Matthias Merker
19 Analysis of Whole Genome Sequencing Data for Detection
of Antimicrobial Resistance Determinants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Marie Anne Chattaway
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Contributors
JOANNA BACON • Discovery Group, UK Health Security Agency, Porton Down, Salisbury, UK
IVAN BARILAR • Molecular and Experimental Mycobacteriology, Research Center Borstel,
Borstel, Germany
VINCENT BARON • Division of Infection and Global Health, School of Medicine, University of
St Andrews, St Andrews, UK
SANJIB BHAKTA • Mycobacteria Research Laboratory, Institute of Structural and Molecular
Biology, School of Natural Sciences, Birkbeck, University of London, London, UK
RUTH BOWNESS • Department of Mathematical Sciences, University of Bath, Bath, UK
DAIRE CANTILLON • Department of Global Health and Infection, Brighton and Sussex
Medical School, University of Sussex, Brighton, UK; Department of Tropical Biology,
LiverpoolLiverpool, UK
MARIE ANNE CHATTAWAY • Gastrointestinal Bacteria Unit, United Kingdom Health Security
Agency, London, UK
MINGZHOU CHEN • School of Physics, University of St Andrews, St Andrews, UK
CHRIS DANIEL • Mycobacteria Research Laboratory, Institute of Structural and Molecular
Biology, School of Natural Sciences, Birkbeck, University of London, London, UK
MARIANNA LEITE DE AVELLAR • School of Physics, University of St Andrews, St Andrews, UK
KISHAN DHOLAKIA • School of Physics, University of St Andrews, St Andrews, UK; Centre of
Light for Life, University of Adelaide, Adelaide, SA, Australia
MARGO DIRICKS • Molecular and Experimental Mycobacteriology, Research Center Borstel,
Borstel, Germany
VIOLA DREYER • Molecular and Experimental Mycobacteriology, Research Center Borstel,
Borstel, Germany
SIMON J. FOSTER • School of Biosciences, Krebs Institute, and Florey Institute, University of
Sheffield, Sheffield, UK; Newcastle University Biosciences Institute, Newcastle upon Tyne,
UK
STEPHEN H. GILLESPIE • Division of Infection and Global Health, School of Medicine,
University of St Andrews, St Andrews, UK
PETER GLYNNE-JONES • Faculty of Engineering and Physical Sciences, University of
Southampton, Southampton, UK
LOUIS GRANDJEAN • Department of Infection, Immunity and Inflammation, Institute of
Child Health, University College London, London, UK
BJÖRN HAMMARSTRÖM • Biophysics, KTH Royal Institute of Technology, Stockholm, Sweden
ROBERT J. H. HAMMOND • Division of Infection and Global Health, School of Medicine,
University of St Andrews, St Andrews, UK
GWENAN KNIGHT • Department of Infectious Disease Epidemiology, London School of Hygiene
and Tropical Medicine, London, UK
THOMAS A. KOHL • Molecular and Experimental Mycobacteriology, Research Center Borstel,
Borstel, Germany
MATTHIAS MERKER • Evolution of the Resistome, Research Center Borstel, Borstel, Germany
VANESSA MOHR • Molecular and Experimental Mycobacteriology, Research Center Borstel,
Borstel, Germany
ix
x Contributors
CHRISTOPHER WILLIAM MOON • Discovery Group, UK Health Security Agency, Porton Down,
Salisbury, UK
BARIKI MTAFYA • National Institute for Medical Research – Mbeya Medical Research Centre,
Mbeya, Tanzania
EMMANUEL MUSISI • Infectious Diseases Research Collaboration, Kampala, Uganda; Adroit
Biomedical & Bioentreprenuership Research Services, Kampala, Uganda; Division of
Infection and Global Health, School of Medicine, University of St Andrews, St Andrews, UK
ANUSHANDAN NAVARATNARAJAH • Mycobacteria Research Laboratory, Institute of Structural
and Molecular Biology, School of Natural Sciences, Birkbeck, University of London, London,
UK
STEFAN NIEMANN • Molecular and Experimental Mycobacteriology, Research Center Borstel,
Borstel, Germany
NYANDA ELIAS NTINGINYA • National Institute for Medical Research –, MbeyaMbeya,
Tanzania
KATARINA ORAVCOVA • School of Biodiversity, One Health and Veterinary Medicine,
University of Glasgow, Glasgow, UK
BENJAMIN J. PARCELL • Clinical Senior Lecturer and Honorary Consultant, Population
Health and Genomics, School of Medicine, University of Dundee, Dundee, UK
KERRY A. PETTIGREW • School of Biomedical Science, School of Health and Life Sciences,
Teesside University, Middlesbrough, UK
ELEANOR PORGES • Discovery Group, UK Health Security Agency, Porton Down, Salisbury,
UK
PASCAL QWARAY • National Institute for Medical Research –, MbeyaMbeya, Tanzania
CHRISTOPHER ROWLATT • Department of Mathematical Sciences, University of Bath, Bath,
UK
WILBER SABIITI • Division of Infection and Global Health, School of Medicine, University of St
Andrews, St Andrews, UK
AMINAT YETUNDE SAULA • Department of Mathematical Sciences, University of Bath, Bath,
UK
ABDULWAHAB SESSOLO • Infectious Diseases Research Collaboration, Kampala, Uganda
EMANUEL SICHONE • National Institute for Medical Research – Mbeya Medical Research
Centre, Mbeya, Tanzania
LINDSAY SONNENKALB • Molecular and Experimental Mycobacteriology, Research Center
Borstel, Borstel, Germany
WILLY SSENGOOBA • School of Biomedical Sciences, College of Health Sciences, Makerere
University, Kampala, Uganda
STEPHEN CHARLES TAYLOR • Pathogen Immunology Group, UK Health Security Agency,
Porton Down, Salisbury, UK
AMY K. TOOKE • School of Biosciences, University of Sheffield, Sheffield, UK; Newcastle
University Biosciences Institute, Newcastle upon Tyne, UK
ARTURO TORRES ORTIZ • Department of Infection, Immunity and Inflammation, Institute
of Child Health, University College London, London, UK
CHRISTIAN UTPATEL • Molecular and Experimental Mycobacteriology, Research Center
Borstel, Borstel, Germany
SIMON J. WADDELL • Department of Global Health and Infection, Brighton and Sussex
Medical School, University of Sussex, Brighton, UK
NATASHA WALBAUM • Division of Infection and Global Health, School of Medicine, University
of St Andrews, St Andrews, UK
Contributors xi
WILLIAM SAAVA WAMBI • School of Biomedical Sciences, College of Health Sciences, Makerere
University, Kampala, Uganda
SAM WILLCOCKS • Mycobacteria Research Laboratory, Institute of Structural and Molecular
Biology, School of Natural Sciences, Birkbeck, University of London, London, UK; Division
of Biosciences, College of Health, Medicine and Life Sciences, Brunel University London,
Uxbridge, UK
JOSEPHINE ZAWEDDE • Infectious Diseases Research Collaboration, Kampala, Uganda
Chapter 1
Using the Zebrafish Larval Model of Infection to Investigate

Antibiotic Efficacy and Combination Treatments Against
Staphylococcus aureus
Amy K. Tooke and Simon J. Foster
Abstract
There is an increasing need for new treatment regimens to combat antibiotic-resistant strains of bacteria.
Staphylococcus aureus is a clinically important, opportunist pathogen that has developed resistance to a range
of antibiotics. The zebrafish larval model of systemic disease has been increasingly utilized to elucidate
S. aureus virulence mechanisms and host-pathogen interactions. Here, we outline how this model can be
used to investigate the effects of different antibiotics alone and in combination against S. aureus.
Key words Zebrafish, Antibiotics, Antimicrobials, Drug screening, Staphylococcus aureus, In vivo
model, Infection, Combination treatment
1 Introduction
Bacterial infections are commonly treated with antibiotics, that

have underpinned human healthcare for over 80 years. However,
with the global rise of antimicrobial resistance, such infections are
expected to be one of the largest causes of death by 2050 [1]. In
2019, methicillin-resistant Staphylococcus aureus (MRSA) alone was
responsible for over 100,000 deaths [2]. Thus, there is urgent
demand for new antibiotics to be developed alongside the refine-
ment and repurposing of existing treatments as even “last resort”
antibiotics often proved to be ineffective. For example, existing
drugs can be given as combination treatments, in order to decrease
the likelihood of resistance developing as quickly as if a monother-
apy was used [3]. Antibiotics used in combination may display
additive or synergistic effects, or undesirably, antagonistic ones;
therefore, it is important to determine their combined effects
before further clinical development. Funding for developing new
antimicrobials is limited, so further utilization of existing com-
pounds is advantageous [4]. Also, microorganisms do not always
Stephen H. Gillespie (ed.), Antibiotic Resistance Protocols, Methods in Molecular Biology, vol. 2833,
https://doi.org/10.1007/978-1-0716-3981-8_1,
© The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer Nature 2024
1
2 Amy K. Tooke and Simon J. Foster
respond to antibiotics in an in vivo context as they do in vitro.

Therefore, there is a need to be able to efficiently and economically
screen compounds, both individually and in combination, for their
efficacy against infectious organisms in vivo.
Zebrafish have been used as a model developmental organism
for decades, but in more recent years increasingly utilized as an
infection model for a range of bacteria, including Gram positives,
Gram negatives and mycobacteria, as well as fungal pathogens and
viruses [5, 6]. Zebrafish larval models offer a range of advantages,
including high throughput. The natural transparency of the
embryos and larvae additionally means that they can be used for
intravital microscopy, to visualize fluorescently labeled infectious
agents. Also due to their genetic tractability, host cells of interest
can also be labeled. Embryonic zebrafish quickly develop an innate
immune system including macrophages and neutrophils within
24 and 36 h post fertilization (hpf) respectively, with functional
similarity to their mammalian counterparts [5, 7]. This is particu-
larly advantageous for disease models where the innate immune
system plays a more significant role than the adaptive response, such
as with S. aureus. Larval zebrafish with labeled phagocytes can also
be used to investigate intracellular pathogens. Finally, zebrafish
larvae can be used to screen new drugs for toxicity and for efficacy
against different pathogens [5].
Here, we outline how to use zebrafish larvae infected with
S. aureus to determine the effects of antibiotic interventions.
Figure 1 outlines the range of effects that antibiotics tested using
this system can have on zebrafish larvae.
2 Materials
1. E3 10× stock: 50 mM NaCl, 1.7 mM KCl, 3.3 mM CaCl2,

3.3 mM MgSO4 (see Note 1).
2. 3.0% w/v methylcellulose solution in E3 buffer that has been
autoclaved and cooled to 70 °C (see Note 2).
3. Tricaine (3-amino benzoic acid ester): Stock is made to 0.4%
w/v in 20 mM Tris-HCl pH 7.0 and stored in the dark at 4 °C.
To anesthetize embryos, this is added to E3 at a 1/20 dilution
to give a final concentration of 0.02% w/v (200 mg/L).
4. Stereomicroscope with transmitted illumination, up to 50×
magnification.
5. Injection rig (micromanipulator and pneumatic injector).
6. Glass capillary tubes (no filament), needle puller machine.
7. Capillary loading tips (Eppendorf microloader 20 μL).
8. Two pairs of Dumont fine tweezers (#5).
9. Incubator set to 28.5 °C.
Antibiotic Efficacy Testing in Zebrafish 3
Fig. 1 A range of outcomes can occur after exposing zebrafish larvae to antibiotics. If the antibiotic is toxic
against uninfected larvae at a particular concentration it cannot be used in assays against infected larvae. If
the antibiotic is non-toxic, it can be either curative or non-curative. Non-curative doses do not prevent
bacterial growth and larval survival is not increased in the treated group. Curative doses can have a
bactericidal or bacteriostatic effect, which is established by determining the bacterial numbers in sampled
larvae before and after antibiotic treatment is given
10. Antibiotics of interest (see Note 3).

11. Sterile Petri dishes (10 mm).
12. Sterile 96-well plates with lids.
13. 50 mL centrifuge tubes.
14. 1.5 mL microcentrifuge tubes.
15. 250 mL conical flasks.
16. 3 mL Pasteur pipettes.
17. Tryptone soy broth (TSB) (30 g/L in distilled water).

18. Bacteriological agar (add to TSB to 1.5% w/v) to make TSA
(tryptone soy agar). Sterilize by autoclaving.
19. Automated plate reader.
20. GraphPad Prism software: for plotting Kaplan-Meier survival
curves and statistical analysis.
3 Methods
These protocols are intended to be used with 33 hpf larvae infected

with S. aureus as outlined in [8]. Briefly, larvae are infected with
S. aureus at a standard dose of approximately 1500 colony forming
units (CFU) (depending on strain background), resulting in 50%
survival by 92 h post infection (hpi) without treatment. This sur-
vival rate means that if a treatment significantly increases or
decreases survival, it can be easily identified—if an untreated
group had a particularly high or low survival rate, it will be difficult
to identify if a treatment increases or decreases survival, respectively.
3.1 Choice of Dug An in vitro minimum inhibitory concentration (MIC) assay carried
Concentrations Based out over 24 h of growth in a sterile 96 well plate is necessary to
on In Vitro Minimum determine the range of concentrations that should be tested against
Inhibitory infection in the zebrafish larvae. This method can give 3 biological
Concentration repeats for 2 different strains in one plate (see Note 4).
1. Grow an overnight culture of the strain(s) of S. aureus of
interest in 10 mL TSB in a 50 mL centrifuge tube at 37 °C,
shaking at 200 revolutions per minute (rpm) with antibiotics
for selection where appropriate. Grow one overnight culture
per biological repeat.
2. In a sterile 96-well plate (with lid) set up serial 2× dilutions of
your antibiotic(s) of interest in TSB, leaving columns 11 and
12 with TSB with no antibiotic, to use as controls (see Fig. 2).
3. Inoculate each well (columns 1–11) with 10 μL overnight
bacterial culture back diluted to OD600 0.01, leaving column
12 as a blank.
4. Seal the lid onto the plate with parafilm (see Note 5).
5. Incubate the plate overnight at 37 °C (see Note 6).
6. After 24 h of growth, read the OD600 of each well in an
automated plate reader, using column 12 as a blank. Column
11 should have a high bacterial density, which decreases
moving across the plate back to column 1. There should be
no growth in column 12; if there is, the media is contaminated.
7. The MIC can be interpreted as the lowest concentration of
antibiotic at which there is no bacterial growth.
Fig. 2 Layout of the MIC assay. Serial dilutions of antibiotics of interest are carried out across columns 1–10,
with column 11 a positive control with bacteria added with no antibiotic, and column 12 a blank with no
antibiotic or bacteria added (just TSB). Grey wells can be left empty. This method gives 3 biological repeats of
2 strains each per 96-well plate
3.2 Toxicity Assays Before an antibiotic is tested against infected larvae, it must be
and Choice of Solvent determined whether the drug itself is toxic to the zebrafish. This
is carried out by titrating a range of concentrations of the test
compound diluted in E3, which the larvae are then immersed
in. Survival is monitored over 92 h compared to an untreated
control. The highest concentration in the range tested where sur-
vival is 100% is the maximum of what can be used to treat infected
embryos, as toxic concentrations negate any potential antibiotic
curative effect.
1. Dechorionate larvae at approximately 24–28 hpf.
2. At 31 hpf, prepare a range of dilutions of the compounds of
interest in E3 buffer. Prepare them at 6× the final concentration
and add 50 μL antibiotic to each well containing 250 μL with a
larva in it (otherwise the concentration will be diluted when the
larva is transferred from its dish of E3 (see Note 7)).
3. Repeat step 2 with equivalent concentrations of solvent that
each antibiotic is dissolved in. For example, if adding 50 μL of
antibiotic in 50% v/v ethanol, add 50 μL of 50% v/v ethanol
minus the antibiotic (see Notes 8 and 9).
4. Finally, keep a group of larvae that are untreated.

5. Monitor survival of embryos twice a day until 92 hpt (hours
post treatment) and plot results as a Kaplan-Meier survival
curve.
3.3 Antibiotic To determine whether the antibiotic of interest degrades over the
Stability Assay time course of an experiment (e.g., up to 92 h) in E3, compounds
at the concentration used in in vivo experiments with the zebrafish
larvae are incubated in E3 at 28.5 °C for a range of timepoints
corresponding to the duration of the infection studies (e.g.,
0–92 hpi). Then bacteria are inoculated into the E3, incubated
overnight and live cells enumerated by plating. If cells grow this
means that the antibiotics have degraded throughout the time
course, whereas if growth is inhibited, this confirms the stability
of the compounds under the experimental conditions.
For a range of timepoints, incubate antibiotics in E3 for up to
92 h at the concentration to be tested with zebrafish larvae (see
Table 1). Additionally set up the same conditions with PBS as a
control to rule out any interactions between the E3 components
and the antibiotic.
1. Set up an overnight culture of your strain of interest in the
afternoon/evening of day 4 by inoculating a single colony into
10 mL of TSB (with antibiotics for selection where necessary)
and incubating overnight at 37 °C with shaking at 200 rpm.
2. On day 5, dilute the overnight culture to OD600 0.01 and add
15 μL to 135 μL of each antibiotic in E3 in a 96-well plate.
Additionally, dilute each antibiotic E3 mix 10 and 100 times
and inoculate these as well (see Fig. 3).
Table 1
Timepoints for setting up each condition in the antibiotic stability assay
Total incubation time of E3

Setup time (hours) Day/time with antibiotics (hours)
0 Day 1/16:00 92
18 Day 2/10:00 74
24 Day 2/16:00 68
42 Day 3/10:00 50
48 Day 3/16:00 44
66 Day 4/10:00 26
72 Day 4/16:00 20
92 Day 5/10:00 0
Fig. 3 Layout of the 96-well plates and corresponding TSA plates for antibiotic stability assay. Bacteria are
incubated with antibiotics that have been incubated in PBS or E3 for predetermined amounts of time overnight
before plating out spots from each well on a TSA plate to establish whether the antibiotics have degraded or
are still effective against the bacteria
3. Repeat step 2 with the PBS controls.

4. Incubate the plates overnight at 37 °C.
5. Spot 10 μL from each well onto TSA (tryptone soy agar) and
allow to dry (see Fig. 3).
6. Incubate the plate overnight at 37 °C.
7. The following day, inspect the plate. Growth inhibition of
bacteria in the E3 indicates stability of the antibiotic in E3
over the time course of the in vivo experiment. If growth of
bacteria decreases from the 92 h to 0 h incubation in the
diluted conditions, this indicates that some antibiotic degrada-
tion was occurring that would not be noticeable at the neat
concentration. The PBS control can be used to determine
whether E3 components are involved in antibiotic degradation.
3.4 Titration of 1. Infected larvae are immersed in E3 containing the antibiotic

Antibiotic Doses (s) of interest at an appropriate concentration and placed into
(Survival Curves)— individual wells of a 96-well plate (see Notes 4 and 10).
Infected Fish 2. Survival of larvae is monitored twice a day until 92 hpi (hours
post infection). Death is defined by absence of a heartbeat/
visible circulation.
3.5 Determination of Larvae can be sampled from a population over the course of an
Live Bacterial experiment to determine the live bacterial numbers in the fish at
Numbers With/Without that given timepoint. This allows monitoring of whether antibiotic
Antibiotics treatments that appear to “cure” fish by increasing survival
compared to untreated controls are exhibiting a bactericidal or

bacteriostatic effect. This provides a deeper insight into growth
dynamics as opposed to just observing if the fish are alive or dead.
1. Larvae are infected and given antibiotic treatment as required.
2. At each timepoint, 5 living larvae are sampled at random and
each homogenized in 100 μL of E3 (using either a pellet pestle
or mechanical homogenizer with sterilized ceramic beads) (see
Notes 11 and 12).
3. Serial dilutions of homogenate in PBS are carried out (20 μL of
homogenate is added to 180 μL PBS) in a 96-well tissue culture
plate with a multichannel pipette.
4. Three dilutions per timepoint for each sampled plate are plated
out on TSA and incubated, in order to determine the CFU per
larva. This can be done by plating out 100 μL per dilution on
1 agar plate, or to minimize plates used, spot out 5× 10–20 μL
spots onto an agar plate per dilution. This way a total of
15 spots (5 spots for 3 dilutions for 1 fish) per agar plate.
5. Plates are allowed to dry then incubated at 37 °C overnight.
6. Colonies are counted and CFU per larva can be calculated.
4 Notes
1. Preparation of working solution: 50 mL 10× E3 stock is made

up to a working concentration with 450 mL deionized water.
One drop of methylene blue is added as an antifungal agent to a
final concentration of 0.00005% w/v. The solution is auto-
claved to sterilize.
2. The solution is stirred, frozen, and defrosted multiple times to
ensure complete solubilization. Stocks are aliquoted into
20 mL syringes and stored at -20 °C, then defrosted in a
28 °C incubator 24 h before use.
3. Antibiotic stocks for bacterial selection are often made at
1000× the working concentration (in MilliQ water, or
50–100% ethanol) and stored at -20 °C. For these zebrafish
studies generally we used the same stock solutions of antibiotics
as for bacterial culture, then diluted them to the appropriate
concentrations in E3 (also see Note 4).
4. Generally a concentration in the E3 buffer around 50–100×
the in vitro MIC is an ideal starting point to test for a curative
effect in the fish [9].
5. Placing the plate in a plastic box with a lid containing paper
towels dampened with PBS helps to prevent desiccation of the
wells if your incubator is particularly drying (e.g., if it has a fan).
6. At such low volumes shaking is not necessary if it is not

available.
7. An alternative method is to fill a whole Petri dish (approx.
25 mL) with the antibiotic E3 solution and transfer larvae
into this from their original dish with as little E3 as possible.
However, this requires the use of more of the compound which
is particularly pertinent to avoid if it is a novel compound that
you only have a small amount of. To transfer larvae from one
dish to another with minimal buffer, once larvae are in the
Pasteur pipette hold the pipette vertically rather than horizon-
tally; larvae will sink toward the bottom of the pipette then
eject as little E3 containing larvae out as possible.
8. Throughout the course of the experiment observe the mor-
phology of larvae and compare the treated group to an
untreated control, in order to determine if there are any effects
of the drug on the larvae that do not kill them. It is important
to consider whether the drugs of interest can diffuse into the
fish. Not all antibiotics are able to diffuse through the skin of
the fish embryos to reach the bacteria in the bloodstream, and
additionally not all antibiotics that diffuse into the bloodstream
can enter phagocytes that can potentially contain the bacteria.
It is important to closely observe the larvae for any morpho-
logical changes that may occur as a result of treatment—even if
they survive, as some antibiotics can still have negative effects.
Additionally, some antibiotics may be observed to precipitate
out onto the skin of larvae, indicating that they may be unable
to diffuse into the fish and are therefore unsuitable for this
assay.
9. A further key control experiment is to titrate the solvent that
the antibiotic is carried in in the E3 with the larvae to rule out
any toxicity effects observed being from the solvent rather than
the antibiotic. For example, if an antibiotic is stocked in ethanol
and this is added to the E3 at higher concentrations it will kill
the larvae, but at lower percentages (up to 0.5% v/v) the fish
are unharmed. To minimize the risks associated with poten-
tially toxic solvents, antibiotics should be dissolved at as high a
concentration as possible in their respective solvent so that a
minimal concentration of solvent is added to the E3.
10. If addition of antibiotics is not to be carried out immediately
after infection, it can be achieved by putting embryos into their
wells in a known volume of E3 so that a concentrated aliquot of
E3 containing the antibiotic of interest can be easily added at
any time throughout the experiment without having to transfer
the larvae to new buffer containing the drug. For example,
keep the larvae each in 250 μL then add 50 μL of the antibiotic
in E3 at 6 times the final required concentration. For ease, this
can be done with a multichannel pipette.
11. It is important that where possible, fish are rinsed in fresh E3 to

remove external antibiotics present in their E3, as once the fish
is homogenized, the newly exposed bacteria come into contact
with the antibiotic during this process. This results in an arte-
factual result whereby it appears that there are no/low bacterial
numbers in the larvae when the homogenate is plated out,
when in fact at the point of sampling, there were indeed living
bacteria inside the larvae. It must be said that this is only
possible with living or very recently deceased larvae—when
the larvae die they very quickly lose structural integrity and
disintegrate into their buffer, meaning it is not possible to carry
out this washing procedure.
12. When labeling the tubes that individual larvae are homoge-
nized in, include the timepoint, if the fish is alive or dead, and a
number (1–5 for living fish). The fish can then be stored at -
20 °C after homogenization has taken place. This means that if
inappropriate dilutions are plated out, individual larvae’s
homogenate can be replated instead of the whole group. It
should be noted that we have observed that more than 1–2
freeze-thaw cycles result in degradation of bacteria so should
be minimized where possible.
References
1. O’Neill J (2016) Tackling drug-resistant infec- 5. Torraca V, Mostowy S (2018) Zebrafish infec-
tions globally: final report and recommendation: from pathogenesis to cell biology. Trends
tions. Review on Antimicrobial Resistance. Cell Biol 28(2):143–156. https://doi.org/10.
Wellcome Trust and HM Government. 1016/j.tcb.2017.10.002
https://wellcomecollection.org/works/ 6. Renshaw SA, Trede NS (2012) A model 450 mil-
thvwsuba lion years in the making: zebrafish and vertebrate
2. Murray CJ, Ikuta KS, Sharara F et al (2022) immunity. Dis Model Mech 5:38–47. https://
Global burden of bacterial antimicrobial resis- doi.org/10.1242/dmm.007138
tance in 2019: a systematic analysis. Lancet 7. Herbomel P, Thisse B, Thisse C (1999) Ontog-
399:629–655. https://doi.org/10.1016/ eny and behaviour of early macrophages in the
S0140-6736(21)02724-0 zebrafish embryo. Development 126(17):
3. Righi E, Scudeller L, Chiamenti M et al (2020) 3735–3745. https://doi.org/10.1242/dev.
In vivo studies on antibiotic combination for the 126.17.3735
treatment of carbapenem-resistant Gram-nega- 8. Prajsnar TK, McVicker G, Williams A et al
tive bacteria: a systematic review and meta- (2018) Use of larval zebrafish model to study
analysis protocol. BMJ Open Sci 4:e100055. within-host infection dynamics. Methods Mol
h t t p s : // d o i . o r g / 1 0 . 1 1 3 6 / b m j o s - Biol 1736:147–156. https://doi.org/10.
2019-100055 1007/978-1-4939-7638-6_14
4. Klug DM, Idiris FIM, Blaskovich MAT et al 9. McVicker G, Prajsnar TK, Williams A et al
(2021) There is no market for new antibiotics: (2014) Clonal expansion during Staphylococcus
this allows an open approach to research and aureus infection dynamics reveals the effect of
development. Wellcome Open Res 6:146. antibiotic intervention. PLoS Pathog 10:
https://doi.org/10.12688/wellcomeopenres. e1003959. https://doi.org/10.1371/journal.
16847.1 ppat.1003959
Chapter 2
Three-Dimensional Rotary Culture to Model Mycobacterial

Biofilms in Low-Shear Detergent-Free Liquid Suspension
Daire Cantillon and Simon J. Waddell
Abstract
In vitro biofilm models have allowed researchers to investigate the role biofilms play in the pathogenesis,
virulence, and antimicrobial drug susceptibility of a wide range of bacterial pathogens. Rotary cell culture
systems create three-dimensional cellular structures, primarily applied to eukaryotic organoids, that better
capture characteristics of the cells in vivo. Here, we describe how to apply a low-shear, detergent-free rotary
cell culture system to generate biofilms of Mycobacterium bovis BCG. The three-dimensional biofilm model
forms mycobacterial cell aggregates in suspension as surface-detached biomass, without severe nutrient
starvation or environmental stress, that can be harvested for downstream experiments. Mycobacterium bovis
BCG derived from cell clusters display antimicrobial drug tolerance, presence of an extracellular matrix, and
evidence of cell wall remodeling, all features of biofilm-associated bacteria that may be relevant to the
treatment of tuberculosis.
Key words Mycobacterium, RCCS, Microgravity, Low-shear suspension, Three-dimensional

(3D) culture system, Cell aggregate, Biofilm, Tuberculosis
1 Introduction
Biofilms are communities of bacteria encased in an extracellular

matrix. These structures afford the inhabitants protection from
environmental insults and the immune system, as well as from
antimicrobial drugs [1, 2]. Importantly, bacteria within a biofilm
show increased antimicrobial drug tolerance, meaning that they are
harder to eradicate with antimicrobial chemotherapy strategies
[3]. The role that biofilm-like growth might play in the pathogen-
esis of tuberculosis (TB) is less well explored compared to other
pulmonary bacterial infections. Like many mycobacteria, Mycobac-
terium tuberculosis (M. tb) grows as cords or clumps in liquid media
without the addition of detergent (usually Tween 80 or Tyloxapol).
These clusters or aggregations of cells likely increase the phenotypic
heterogeneity of the bacterial population and may limit penetration
https://doi.org/10.1007/978-1-0716-3981-8_2,
11
12 Daire Cantillon and Simon J. Waddell
of antimicrobial drugs. Corded biofilm-like growth of M. tb was

observed in human lungs by Georges Canetti [4], and has since
been reported in animal models of TB disease [5–7].
To elucidate the functional significance of aggregated growth
in TB, several in vitro models have been developed. These models
often rely on the formation of a mycobacterial pellicle at the liquid–
air interface and attachment of the biofilm to a solid surface.
Mycobacteria in biofilm-like pellicles have been shown to be
transcriptionally distinct from planktonic mycobacteria, harbour
drug-tolerant bacilli, and contain free mycolic acids in an altered
extracellular matrix—replicating some phenotypes observed in vivo
[8–13].
The Rotary Cell Culture System (RCCS) was designed by the
National Aeronautics and Space Administration (NASA) to investi-
gate the impact of microgravity on cellular responses. It was repur-
posed to culture organoids of human origin in low-shear
conditions, as well as investigate bacterial pathogen-host interac-
tions [14–17]. An overview of the RCCS system is shown in Fig. 1.
The culture vessels, with large semipermeable membranes, rotate at
Fig. 1 Overview of the Rotary cell culture system. (a) RCCS 10 mL High Aspect Ratio Vessel (HARV). (i) Large
outlet port for harvesting biofilms; (ii) fill port butterfly valve; (iii) exit port butterfly valve; (iv) fill port screwcap
for addition of media; (v) exit port screwcap for removal of air bubbles; (vi) RCCS attachment point to the
vertical support. The semi-permeable membrane is visible on the back of the vessel. (b) RCCS-4DQ vertical
support and control panel, where the rotation of each vessel is independently controlled. RCCS vertical support
platform with 4 10 mL HARV mounted. (c) Centrifugal force (Fcf) is controlled through rotation of the vessels,
enabling biomass to accumulate in a low-shear liquid suspension. Gravity (g) collects biomass to the bottom of
non-rotating RCCS vessels
Generation of Mycobacterial Biofilms in Suspension 13
a controllable speed to keep the growing cell aggregations in sus-

pension. This allows the biofilms to grow without attachment to a
solid substrate, facilitating the formation of complex microarchi-
tectures. Cell culture in three-dimensional systems also enables
enhanced gaseous, waste, and nutrient exchange from all sides
compared to traditional culture methods [18]. The RCCS system
was used to model aggregated Mycobacterium bovis BCG growth in
detergent-free low-shear liquid suspension. Mycobacteria derived
from these cell clusters were more tolerant to isoniazid and strep-
tomycin but not rifampicin, a drug-tolerant phenotype that was lost
after passage in drug-free media [8]. This chapter describes the
generation of mycobacterial cell aggregates, modeling biofilm-like
growth suspended in liquid media, using the RCCS system.
2 Materials
2.1 Culture Media 1. Middlebrook 7H9 broth base (Merck).

2. Tween 80 (Merck).
3. Middlebrook ADC (albumin, dextrose, catalase) growth sup-
plement (see Note 1).
4. Ultrapure water.
5. Magnetic stir bars.
6. 1 L Duran autoclavable glass bottles.
7. Water bath set at 52 C.
8. Middlebrook 7H10 agar powder (Merck).
9. Middlebrook OADC (oleic acid, albumin, dextrose, catalase)
growth supplement (see Note 2).
10. Glycerol (Merck).
11. Plastic Petri dishes 90 mm (Fisher Scientific).
12. 1 L glass beaker.
13. 1 L graduated cylinder.
14. Monobasic potassium phosphate (Merck).
15. Magnesium sulfate (Merck).
16. L-asparagine (Merck).
17. Ferric ammonium citrate (Merck).
18. Citric acid (Merck).
19. Zinc sulfate solution (1% in molecular grade water) (Merck).
20. Sodium hydroxide (10 N) (Merck).
21. pH meter.
2.2 RCCS Culture 1. RCCS-4DQ bioreactor system (Synthecon Ltd).

System 2. 3D rotating disposable High Aspect Ratio Vessels (HARV)
10 mL (Synthecon Ltd).
3. Sterile single-use cotton swabs.
4. 70% ethanol.
5. 25 mL serological pipette.
6. Phosphate Buffered Saline (PBS), sterile tissue culture grade
pH 7.4.
7. Humidified incubator set at 37 C.
8. Spectrophotometer capable of measuring OD600nm.
9. Single-use plastic cuvettes (1.5 mL) (Fisher Scientific).
10. Universal tubes (30 mL) (Fisher Scientific).
11. Single-use sterile 5 mL syringes (Fisher Scientific).
3 Methods
All work with Mycobacterium bovis BCG should be conducted in a

hazard level 2 containment facility working in a class II biosafety
cabinet (see Note 3).
3.1 Preparation 1. For 1 L Middlebrook 7H9 ADC with 0.05% Tween 80 liquid
of Media culture media—Add 4.7 g Middlebrook 7H9 broth powder to
a 1 L Duran bottle with 900 mL ultrapure water and a mag-
netic stirring bar. Then add 500 μL Tween 80 and stir on a
magnetic plate until dissolved. Autoclave at 121 C for 15 min,
then cool to 52 C in a water bath. Add 100 mL Middlebrook
ADC growth supplement and mix well (see Note 1). Store at
4 C, warm to 37 C before use.
2. For 1 L Middlebrook 7H10 OADC with 0.5% glycerol agar
plates (approximately 25 plates)—Add 19.5 g Middlebrook
7H10 agar powder to a 1 L Duran bottle with 900 mL ultra-
pure water and a magnetic stirring bar. Add 5 mL glycerol, then
stir until dissolved. Autoclave at 121 C for 15 min, then cool
to 52 C in a water bath. Add 100 mL OADC and mix well (see
Note 2). Pour 30 mL per 90 mm Petri dish, allow to set, and
then store at 4 C. Warm to 37 C before use.
3. For 1 L Sauton media with 0.05% Tween 80—Add approxi-
mately 800 mL ultrapure water to a 1 L glass beaker with a
stirring bar. With the bar spinning, add 0.5 g (monobasic)
potassium phosphate, 0.5 g magnesium sulfate, 4.0 g L-aspar-
agine, 0.05 g ferric ammonium citrate, and 2.0 g citric acid.
Once dissolved, add 60 mL glycerol followed by 100 μL 1%

zinc sulfate solution. If Tween 80 is required, add 500 μL of
the detergent and mix well. Adjust the pH to 7.2 with 10 N
sodium hydroxide, then correct the final volume to 1 L with
ultrapure water using a graduated cylinder. Transfer to a 1 L
Duran bottle and autoclave at 121 C for 15 min. Store at 4 C,
warm to 37 C before use.
4. For 1 L Sauton media without Tween 80 (for use in the RCCS
system)— repeat as in Subheading 3.1, step 3. above, leaving
out the Tween 80. Make up ~1.5 total culture volume for the
planned RCCS experiment to ensure media volume is in excess
of the volume required.
3.2 RCCS Inoculum 1. Prepare the mycobacterial inoculum by thawing a glycerol

Preparation stock of M. bovis BCG at room temperature. Once defrosted,
add 1 mL (corresponding to ~108 CFU/mL) to 9 mL Mid-
dlebrook 7H9 ADC Tween 80 media in a 30 mL universal tube
and incubate at 37 C statically for 1 week.
2. Passage 1 mL of this mycobacterial culture into 9 mL Sauton
medium with 0.05% Tween 80 in a 30 mL universal tube and
incubate at 37 C statically for 1 week (see Note 4).
3. After 7 days this culture should be turbid and in mid-late log
phase growth. On the day of RCCS inoculation, measure the
OD600nm (see Note 5) and dilute the culture to OD600nm 0.005
(corresponding to ~105 CFU/mL) with Sauton no Tween
80 media (see Note 6).
4. Plate a serial dilution of this inoculum onto solid Middlebrook
7H10 OADC plates to confirm the starting number of myco-
bacterial colony forming units (CFU) (see Note 7).
3.3 Inoculation of the 1. RCCS vessels are sterile, single-use bioreactors that do not
RCCS Culture System need to be autoclaved (see Note 8). The day before inoculation
of the RCCS, pre-treat each of the RCCS vessels, adding 10 mL
sterile PBS through the fill port of the vessels. Attach to the
RCCS vertical support platform and place inside a humidified
tissue culture incubator at 37 C overnight (see Note 9).
2. The following day, remove the PBS from the vessels using the
exit port with a 5 mL syringe and discard the liquid (see
Note 10).
3. Using a 20 mL syringe, add 10 mL Sauton no Tween 80 myco-
bacterial inoculum through the fill port while simultaneously
removing air with a second 20 mL syringe from the exit port.
Ensure no air bubbles remain by drawing excess inoculum
through the fill and exit ports to remove air from the system.
Clean the necks of the fill and exit ports after closing with a
swab soaked in 70% ethanol to surface disinfect.
4. Once all four vessels have been inoculated, attach to the RCCS
vertical support platform and place inside the humidified tissue
culture incubator at 37 C. Run the two-meter flat RCCS cable
from the incubator to the external RCCS interface controls (see
Note 11). The rotation of each vessel is controlled indepen-
dently; set two vessels to rotate at 15 rpm, the other two vessels
at 0 rpm as static controls (see Note 12).
3.4 Establishing and 1. Check the cultures daily for air bubbles in the vessels. Air
Maintaining Cell bubbles can disrupt the formation of biofilms and should be
Aggregates in the removed. To maintain the humidified environment of the incu-
RCCS System bator and reduce evaporation from the vessels, only open the
incubator once per day (see Note 13).
2. Remove air bubbles by adding Sauton no Tween 80 media
through the fill port with a 5 mL syringe, while drawing air
from the exit port with a second 5 mL syringe. Remove/add
the same culture volume to all vessels to maintain nutrient
balance across replicates or conditions (see Note 14).
3. Clean the necks of the fill and exit ports after closing with a
swab soaked in 70% ethanol (see Note 15) to maintain sterility.
4. Visible clusters of M. bovis BCG that are free-falling in suspen-
sion will be macroscopically visible by day 7. As the cellular
aggregates form and increase in size, decrease the speed of
rotation of the RCCS vessels to 10 rpm (see Note 16).
5. Media constituents, cells, macromolecules, chemicals, and anti-
microbial drugs may be added to the vessels through the fill
port to measure the impact on bacterial biofilm formation as
required, or aggregates may be allowed to form without inter-
vention for analysis in downstream applications.
3.5 Harvesting 1. Harvest mature M. bovis BCG biofilms after 21 days (see Fig. 2
Biofilms from the and Note 17). Carefully transfer the RCCS vessel to a class II
RCCS biosafety cabinet. Surface disinfect the external surfaces of the
vessel with a swab soaked in 70% ethanol to prevent contami-
nation being introduced when removing the mycobacterial
biomass.
2. Remove the cap from the large outlet-sample removal port and
loosen the fill port to prevent the vessel becoming air-locked.
3. Use a 25 mL serological pipette to gently remove the entire
vessel contents to sterile containers for downstream applica-
tions, such as drug sensitivity testing or biochemical analyses
(see Note 18).
Fig. 2 Representative images of M. bovis BCG cell aggregates cultured in the RCCS system. Mycobacterial
biofilm formation in the RCCS system with rotation at day 7 (a), day 14 (b), and day 21 (c). Replicate vessels
without rotation as control cultures at day 7 (d), day 14 (e), and day 21 (f). Control vessels (d, e, f) were
agitated briefly for the purposes of imaging the biomass settled at the bottom of the vessels. Black arrows
mark mycobacterial biomass grown in these 10 mL HARV RCCS vessels that are 6.5 cm in diameter
4 Notes
1. Middlebrook ADC (albumin, dextrose, catalase) growth sup-

plement can be purchased or made. To make this supplement,
dissolve 10 g bovine albumin fraction V (Pan Biotech) in
approximately 150 mL ultrapure water in a 500 mL glass
beaker with magnetic stirrer. Once dissolved, add 4 g dextrose
(Merck), 1.7 g sodium chloride (Merck), 8 mg catalase from
bovine liver (Merck), and 112.3 μL oleic acid (Merck). Once all
components have dissolved, make up to 200 mL final volume
in a graduated cylinder with ultrapure water. Filter through a
0.22 μM filter unit using a vacuum pump into a sterile 250 mL
Duran bottle. Store at 4 C.
2. Middlebrook OADC (oleic acid, albumin, dextrose, catalase)
growth supplement can be purchased or made. To make this
supplement, dissolve 10 g of bovine albumin fraction V (Pan
Biotech) in approximately 150 mL ultrapure water in a 500 mL
glass beaker with magnetic stirrer. Once dissolved, add 4 g
dextrose (Merck), 1.7 g sodium chloride (Merck), 8 mg cata-
lase from bovine liver (Merck), and 112.3 μL oleic acid
(Merck). Once all components have dissolved, make up to
200 mL final volume in a graduated cylinder with ultrapure
water. Filter through a 0.22 μM filter unit using a vacuum
pump into a sterile 250 mL Duran bottle. Store at 4 C.
3. This protocol has been developed for use with the hazard
group 2 bacterium, Mycobacterium bovis BCG. The method
should be fully risk-assessed and amended appropriately if
applying to other mycobacteria/bacteria. For example, addi-
tional secondary containment will be required if using Myco-
bacterium tuberculosis or other hazard group 3 pathogens.
4. This inoculum train is used to maximize the recovery and
growth phenotype of the bacterium. Middlebrook 7H9 ADC
0.05% Tween 80 is nutrient-rich and facilitates the growth and
recovery of M. bovis BCG after storage at 80 C. Sauton
media with Tween 80 is then used to condition M. bovis BCG
to the altered nutrient profile of this simple media before
transfer to the RCCS system. Sauton media with Tween 80 is
included here to prevent M. bovis BCG clumping, which may
lead to erratic OD readings and unnecessary variation in cell
numbers when setting up replicate experiments. Sauton media
without Tween 80 is used in the RCCS system to allow myco-
bacteria to grow without the influence of detergent.
5. It is recommended to estimate mycobacterial colony forming
units (CFU) from OD600nm prior to starting the experiment, as
correlations can vary between instruments and laboratories.
6. Ensure an excess of culture inoculum, for example, 50 mL to

inoculate 4 10 mL vessels.
7. Use sterile PBS to serially dilute the inoculum and then plate
using the Miles Misra method onto solid media, incubate at
37 C for 3–4 weeks.
8. Synthecon Ltd. 3D rotating disposable High Aspect Ratio
Vessels (HARV) come in 10 mL or 50 mL culture volumes.
This method describes the use of 4 10 mL HARV vessels.
9. Pre-incubation with PBS is necessary to equilibrate the vessel’s
semipermeable membrane. Without this step, there will be
greater media loss through the semipermeable membrane in
the first 48 h leading to the formation of bubbles in the vessel.
10. Larger volume syringes may be used, however the 5 mL syringe
allows for more precise measurement of air removed from, or
media added to, the vessel. Smaller volume syringes are less
likely to leave air in the system when disconnected.
11. The RCCS cable is flat so it will fit through a closed incubator
door without disrupting the seal of a humidified incubator.
12. Depending on the experiment design, non-rotating cultures
may be required as negative controls. In this instance, set two
vessels rotating and leave two vessels static as comparators (see
Fig. 2). The growth of bacterial clusters may be variable
between vessels, thus this system is best suited to generating
bacterial aggregated biomass for downstream analyses.
13. Take measures to reduce evaporation from the vessels, which
have large semipermeable membranes to enable gaseous
exchange. Pre-warm media, humidify the incubator, and resist
frequently opening the inner incubator door.
14. Air bubbles may disrupt the formation of biofilms and should
be removed as soon as possible.
15. This is to remove residual media that will dry around the fill
and exit ports. This could lead to external surface contamina-
tion that could be introduced to the RCCS vessel when next
using the ports.
16. The speed of rotation to maintain the cell clusters in suspension
will depend on the size of the bacterial aggregates. Start rotat-
ing at 15 rpm, then reduce to 10 rpm as the clusters get larger,
adjust as necessary to prevent the aggregates from coming into
contact with the vessel walls that will disrupt the clumped
bacterial structures. Reduction of rpm balances the centrifugal
forces exerted by rotation on the aggregates that would force
the cell clusters to the rim of the vessel, exposing them to
potentially destructive shear forces.
17. The timeframes indicated in this protocol are for culturing

cellular aggregates of M. bovis BCG, these will need to be
adapted for other mycobacteria/bacteria.
18. A 25 mL serological pipette is most suitable to remove the
entire vessel contents with cellular aggregates intact. Other
thin- or wide-mouthed pipettes may be more appropriate
depending on the size of the bacterial aggregates generated,
and whether the biofilms will be used intact or disrupted into
single-cell suspensions for downstream experiments.
Acknowledgments
S.J.W. would like to acknowledge funding from The Royal Society

(research grant RG110191). D.C. was funded by a University of
Brighton PhD studentship. Figure 1 was created with BioRender
(www.biorender.com).
References
1. Gonzalez JF, Hahn MM, Gunn JS (2018) 8. Cantillon D, Wroblewska J, Cooper I, New-
Chronic biofilm-based infections: skewing of port MJ, Waddell SJ (2021)
the immune response. Pathog Dis 76(3): Three-dimensional low shear culture of Myco-
fty023 bacterium bovis BCG induces biofilm forma-
2. Sharma D, Misba L, Khan AU (2019) Antibio- tion and antimicrobial drug tolerance. NPJ
tics versus biofilm: an emerging battleground Biofilms Microbiomes 7(1):12
in microbial communities. Antimicrob Resist 9. Ojha AK, Baughn AD, Sambandan D, Hsu T,
Infect Control 8:76 Trivelli X, Guerardel Y et al (2008) Growth of
3. Ciofu O, Moser C, Jensen PO, Hoiby N Mycobacterium tuberculosis biofilms containing
(2022) Tolerance and resistance of microbial free mycolic acids and harbouring drug-
biofilms. Nat Rev Microbiol 20(10):621–635 tolerant bacteria. Mol Microbiol 69(1):
4. Canetti G (1955) Tubercle bacillus in the pul- 164–174
monary lesion of man: histobacteriology and 10. Flores-Valdez MA, Aceves-Sanchez MJ, Peter-
its bearing on the therapy of pulmonary tuber- son EJR, Baliga N, Bravo-Madrigal J, De la
culosis, vol 231. Springer, New York, p 480 Cruz-Villegas MA et al (2020) Transcriptional
5. Chakraborty P, Bajeli S, Kaushal D, Radotra portrait of M. bovis BCG during biofilm pro-
BD, Kumar A (2021) Biofilm formation in duction shows genes differentially expressed
the lung contributes to virulence and drug tol- during intercellular aggregation and substrate
erance of Mycobacterium tuberculosis. Nat attachment. Sci Rep 10(1):12578
Commun 12(1):1606 11. Trivedi A, Mavi PS, Bhatt D, Kumar A (2016)
6. Basaraba RJ, Ojha AK (2017) Mycobacterial Thiol reductive stress induces cellulose-
biofilms: revisiting tuberculosis bacilli in extra- anchored biofilm formation in Mycobacterium
cellular necrotizing lesions. Microbiol Spectr tuberculosis. Nat Commun 7:11392
5(3):10.1128/microbiolspec.TBTB2- 12. Ackart DF, Hascall-Dove L, Caceres SM, Kirk
0024-2016. NM, Podell BK, Melander C et al (2014)
7. Lenaerts AJ, Hoff D, Aly S, Ehlers S, Expression of antimicrobial drug tolerance by
Andries K, Cantarero L et al (2007) Location attached communities of Mycobacterium tuber-
of persisting mycobacteria in a Guinea pig culosis. Pathog Dis 70(3):359–369
model of tuberculosis revealed by r207910. 13. Keating T, Lethbridge S, Allnutt JC, Hendon-
Antimicrob Agents Chemother 51(9): Dunn CL, Thomas SR, Alderwick LJ et al
3338–3345 (2021) Mycobacterium tuberculosis modifies
cell wall carbohydrates during biofilm growth functions of three-dimensional hepatocyte

with a concomitant reduction in complement aggregates. Biomaterials 35(7):2162–2171
activation. Cell Surf 7:100065 17. Schwarz RP, Goodwin TJ, Wolf DA (1992)
14. Unsworth BR, Lelkes PI (1998) Growing tis- Cell culture for three-dimensional modeling
sues in microgravity. Nat Med 4(8):901–907 in rotating-wall vessels: an application of
15. Barrila J, Crabbe A, Yang J, Franco K, Nydam simulated microgravity. J Tissue Cult Methods
SD, Forsyth RJ et al (2018) Modeling host- 14(2):51–57
pathogen interactions in the context of the 18. Edmondson R, Broglie JJ, Adcock AF, Yang L
microenvironment: three-dimensional cell cul- (2014) Three-dimensional cell culture systems
ture comes of age. Infect Immun 86(11): and their applications in drug discovery and
e00282-18 cell-based biosensors. Assay Drug Dev Technol
16. Chang TT, Hughes-Fulford M (2014) Molec- 12(4):207–218
ular mechanisms underlying the enhanced
Chapter 3
Rapid Gene Silencing Followed by Antimicrobial

Susceptibility Testing for Target Validation in Antibiotic
Discovery
Chris Daniel, Sam Willcocks, and Sanjib Bhakta
Abstract
Mycobacterium tuberculosis is the main causative agent of tuberculosis (TB)—an ancient yet widespread
global infectious disease to which 1.6 million people lost their lives in 2021. Antimicrobial resistance
(AMR) has been an ongoing crisis for decades; 4.95 million deaths were associated with antibiotic resistance
in 2019. While AMR is a multi-faceted problem, drug discovery is an urgent part of the solution and is at
the forefront of modern research.
The landscape of drug discovery for TB has undoubtedly been transformed by the development of high-
throughput gene-silencing techniques that enable interrogation of every gene in the genome, and their
relative contribution to fitness, virulence, and AMR. A recent advance in this area is CRISPR interference
(CRISPRi). The application of this technique to antimicrobial susceptibility testing (AST) is the subject of
ongoing research in basic science.
CRISPRi technology can be used in conjunction with the high-throughput SPOT-culture growth
inhibition assay (HT-SPOTi) to rapidly evaluate and assess gene essentiality including non-essential,
conditionally essential (by using appropriate culture conditions), and essential genes. In addition, the
HT-SPOTi method can develop drug susceptibility and drug resistance profiles.
This technology is further useful for drug discovery groups who have designed target-based inhibitors
rationally and wish to validate the primary mechanisms of their novel compounds’ antibiotic action against
the proposed target.
Key words Mycobacteria, CRISPRi, Gene essentiality, Antimicrobial susceptibility testing, Target
validation
1 Introduction
Antimicrobial resistance is one of the largest global threats to health

today [1, 2], discovering new antibiotics is crucial in order to
combat resistance and save lives. The need is especially urgent to
treat tuberculosis, caused by the pathogen, Mycobacterium tubercu-
losis, as drug resistance is emerging even against newer drugs such
as bedaquiline [3]. In order to screen and assess prospective
https://doi.org/10.1007/978-1-0716-3981-8_3,
23
24 Chris Daniel et al.
antimicrobials and their mechanisms of action, a combination of

rapid gene silencing and traditional gold-standard susceptibility
testing can be used to investigate drug-target interactions accu-
rately and efficiently at the cellular level.
Clustered regularly interspaced short palindromic repeat inter-
ference (CRISPRi) relies on a catalytically deactivated Cas9 enzyme
coupled with a single-guide RNA (sgRNA) to silence gene tran-
scription in a sequence-specific manner [4–9]. The technique is
currently limited by the necessity for the target site to occur adja-
cent to protospacer adjacent motifs (PAM), which occur through-
out the genome. Within the last few years, sgRNA sequences, along
with their knockdown strength have been published for every gene
in the genome for M. tuberculosis and M. smegmatis [6].
In contrast to traditional methodologies such as deletion muta-
genesis, CRISPRi is ideally suited to the study of core essential
fitness genes. This is facilitated by virtue of partial gene silencing,
allowing continued survival of the bacterial cell, while still resulting
in phenotypic knockdown.
Using CRISPRi in conjunction with antimicrobial susceptibil-
ity testing (AST) can: (a) quantitatively establish the contribution
of the target to bacterial fitness in the absence of antimicrobial
stress, (b) provide confidence in target identification by altered
MIC upon exposure to the experimental compound, (c) aid in the
identification of novel compounds to be used as part of a multi-
target combinatorial therapeutic strategy, and (d) validate hypo-
thetical AMR mechanisms, including those involving sites outside
the coding regions of the genome [6, 10, 11].
This chapter describes the creation of silenced genes using
CRISPRi, investigating gene essentiality and evaluating antimicro-
bial susceptibility using an optimized HT-SPOTi method [12–
15]. This methodology is summarized in Fig. 1. The method is
an adaptation of previously published work by Jeremy M. Rock
[4]. It is economic, efficient, and can gain fast results for slow-
growing species of mycobacteria. We describe this method using
Mycobacterium abscessus as a model species.
2 Materials
2.1 Golden Gate 1. Golden Gate Cloning (NEBridge® Golden Gate Assembly Kit,
Assembly [16] BsmBI-v2, NEB).
2. pIRL2 plasmid (Addgene #163631).
3. Oligonucleotides.
4. PCR buffer (containing MgCl2).
5. Molecular biology grade water, nuclease free (e.g.,
HyClone™).
6. Competent E. coli cells (e.g., E. coli DH5α).
Rapid Gene Silencing Followed by Antimicrobial Susceptibility Testing. . . 25
Fig. 1 Workflow for rapid gene silencing. (Created with BioRender.com)

7. Super Optimal broth with Catabolite repression (S.O.C)

medium.
8. LB (Luria-Bertani) agar.
9. LB broth.
10. Kanamycin (50 μg/mL).
11. 0.2 mL PCR tubes.
12. Heating block.
2.2 Colony PCR 1. Mycobacterium abscessus colonies.

2. 1 μL plastic loops.
3. Master Mix (e.g., Qiagen HotStarTaq Master Mix).
4. Molecular grade water.
6. Thermal cycler.
2.3 Agarose Gel 1. Loading dye (e.g., GelRed®).

Electrophoresis 2. 10Kb Ladder.
3. Agarose.
4. 1 × Tris-Acetic Acid-EDTA (TAE) buffer.
5. Electrophoresis tank, gel tray and comb.
6. Power supply.
2.4 Plasmid DNA 1. Plasmid DNA extraction kit (e.g., Qiagen QIAprep Spin
Extraction and Miniprep Kit).
Sequencing 2. 1.5 mL Microcentrifuge tubes.
2.5 Transformation 1. Middlebrook 7H9 Broth (M7H9) (Difco™).

into Mycobacterium 2. Albumin-Dextrose-Catalase (ADC) growth supplement.
abscessus
3. 0.2% and 0.5% Glycerol.
4. 0.05% Tween80.
5. Middlebrook 7H10 Agar (M7H10) (Difco™).
6. Oleic Acid- Albumin-Dextrose-Catalase (OADC) growth
supplement.
8. Electroporator.
2.6 Validation of 1. Middlebrook 7H9 Broth (M7H9) (Difco™).

Gene Silencing 2. Albumin-Dextrose-Catalase (ADC) growth supplement.
3. Total RNA extraction kit (e.g., Qiagen RNeasy kit) inc.
DNAse.
4. cDNA synthesis kit (e.g., ReadyScript cDNA synthesis mix,
Merck).
5. Master Mix for real-time quantitative PCR.

6. Molecular grade water.
8. Thermal cycler.
2.7 Gene Essentiality 1. Reagent Dispenser (e.g., Multidrop Combi Reagent

Testing Using HT- Dispenser).
SPOTi 2. Middlebrook 7H10 Agar (M7H10) (Difco™).
3. Oleic Acid-Albumin-Dextrose-Catalase (OADC) growth
supplement.
4. 0.5% Glycerol.
6. Anhydrotetracycline (1, 10, 100, 1000 ng/mL).
7. 96-well plate with lid.
8. 1.5 mL Microcentrifuge tubes.
3 Methods
3.1 CRISPRi Mutant 1. Select the gene of interest for silencing. Identify a PAM
Generation Using sequence on the non-coding strand using sequences recog-
Golden Gate Assembly nized and identified by the Sth1dCas9 [4, 6]. There are differ-
ent strategies for selecting and designing sgRNA using either
the coding or the non-coding strand [4], the method we
describe has been optimized for this specific dCas9 system (see
Note 1).
2. Immediately upstream of the PAM, select 20–25 bp to be the
sgRNA. inhA from M. tuberculosis H37Rv is used as an exam-
ple below [8].
inhA (Rv1484) PAM 5′ CCAGAAT 3′
sgRNA 5′ GTCGGTGATGATTCCGCTAA 3′
5′ AACGGATTCTGGTTAGCGGAATCATCACCGACTC 3′
3′ TTGCCTAAGACCAATCGCCTTAGTAGTGGCTGAG 5′
3. Perform a BLASTN search of the sgRNA and the PAM against

the nr database (http://www.ncbi.nlm.nih.gov/BLAST/) to
ensure gene specificity.
4. Purchase the oligonucleotides with overhangs of GGGA on the
forward oligo sequence (the sgRNA) and AAAG on the reverse
oligo sequence on the 5′. These are specific to the plRL2
plasmid sequence when digested by BsmB1 v2. You will also
need to purchase a reverse primer on the plasmid for PCR.
A primer sequence for this has been provided by Rock

lab (5′ TTCCTGTGAAGAGCCATTGATAATG 3′) approxi-
mately 160 bp away from the BsmBI cut site for amplification.
5′ GGGAGTCGGTGATGATTCCGCTAA 3′.
5′ AAAGTTAGCGGAATCATCACCGAC 3′.
5. Make a 100 μM stock of the oligonucleotides using molecular
grade water.
6. Anneal the two oligonucleotides (9 μL of each) with 2 μL of
standard PCR buffer (containing MgCl2) at 95 °C for 5 min,
then leave to cool at room temperature for 90 min.
7. Make a 100 × dilution of the annealed primers using molecular-
grade water.
8. Clone the annealed dsDNA into pIRL2 using the Golden Gate
Assembly Kit. Place the following into a PCR tube:
Component (see Note 2) Volume

pIRL2 plasmid 6 μL
Annealed dsDNA 1 μL
T4 DNA ligase buffer 2 μL
Molecular grade water 10 μL
BsmBI-v2 and T4 DNA ligase enzyme 1 μL
Run the reaction mix in a thermal cycler at 42 °C for

10 min, then 60 °C for 5 min. Add 2 μL of the mix to
100 μL of DH5α competent E. coli and leave on ice for 5 min.
9. Transform using a thermomixer at 42 °C for 30 s exactly before
placing on ice for 1 min. Then add 250 μL of S.O.C. medium
and leave to recover in rotating incubator at 180 rpm, 37 °C for
90 min. Spread the recovered E. coli on a LB plate supplemen-
ted with 50 μg/mL of kanamycin and leave overnight.
10. Once the colonies have grown, perform a colony PCR. Take a
single colony from the plate using a 1 μL inoculation loop and
add to 15 μL of sterile molecular-grade water in a PCR tube.
11. Make up the PCR solution using the HotStarTaq Master Mix
as per the volumes outlined in the manufacturer’s instructions:
Component Volume
HotStar Taq master mix 2x 25 μL
Molecular-grade water 15 μL
2 μM of sgRNA and pIRL2 primer 5 μL
Bacterial suspension 5 μL
12. Mix the reaction gently by tapping, or gently pipetting. Iden-

tify annealing temperature of the primers provided by the
manufacturer of the oligonucleotides (see Note 3). Then run
a standard PCR program; 95 °C for 15 min, 94 °C for 30 s,
annealing temperature for 30 s, 72 °C for 1 min. Repeat from
step 2 25–30 times, then a final extension of 72 °C for 10 min
and 10 °C for 1 h.
13. Take the remaining bacteria left in the suspension, grow single
colonies in separate 50 mL falcon tubes containing 10 mL of
LB supplemented with 50 μg/mL of kanamycin overnight.
3.2 Agarose Gel 1. Prepare a 1% mix of agarose and 1 × TAE buffer for the gel.
Electrophoresis Dissolve 1 g of agarose in 100 mL of 1 × TAE in a microwave
for 2 min, gently swirling every 30 s. Let the solution cool, add
1 μL of GelRed® (10,000 × concentrated) and pour into a gel
tray. Add an appropriate comb ensuring no bubbles are in the
gel, leave to cool and completely set.
2. Once cooled, carefully remove the comb and the gel from the
casting tray. Place the gel tray into electrophoresis unit and
cover with 1 × TAE.
3. Add the appropriate volume (see recommendation by manu-
facturer) of ladder to the first well, mix 2 μL of loading dye to
the PCR product and add 10 μL of the final solution to
each well.
4. Connect the power supply in the correct orientation (negative
to positive), and run the gel for approximately 40 min at 100 V.
5. Once run, disconnect the power supply and visualize the gel
using a gel imaging system. If the cloning is successful, based
on the reverse pIRL2 primer provided a band should be visible
at approximately 160 bp.
3.3 Plasmid DNA 1. When the cultures have grown, make glycerol stocks using
Extraction and 500 μL of sterile 50% glycerol and 500 μL of culture.
Sequencing 2. Using the remaining culture, extract the plasmid DNA accord-
ing to the manufacturer’s instructions using the Qiagen QIA-
prep Spin Miniprep Kit.
3. Once extracted, measure the concentration, A260/280,
A260/230 using a nanodrop. Sequence the purified plasmid
to verify correct cloning of the sgRNA.
3.4 Transformation 1. Grow an M. abscessus culture to OD600 ~ 0.4 (in 100 mL

into M. abscessus M7H9 supplemented with sterile 10% ADC, 0.2% glycerol,
0.05% Tween80), leave on ice for 45 min. Wash the culture
three times, with sterile, ice-cold 10% glycerol stock and cen-
trifuged at 3000× g at 4 °C for 10 min. Every wash, the volume

of glycerol was reduced (25 mL, 10 mL, 5 mL, finally suspend
in 1 mL).
2. Take 2 μL of extracted plasmid DNA and gently add to 100 μL
competent M. abscessus cells in a cooled cuvette [17]. Electro-
porated at 2.5 kv, 25 μF, 1000 Ω (single pulse) using an
electroporator. Once pulsed, transfer the bacteria immediately
to 10 mL of M7H9 (with no antibiotics) for recovery, over-
night. The following day, spread 100 μL of the recovered
bacteria onto an M7H10 agar plate supplemented with
50 μg/mL of kanamycin, 10% OADC and 0.5% glycerol, to
select for transformants.
3.5 Validation of 1. It is important to evidence that successful, on-target transcrip-

Gene Silencing tional silencing has been achieved before moving on to pheno-
typic characterization. Begin by performing a total RNA
extraction from the aTc-induced and un-induced control
CRISPRi strain, as well as the pIRL2 empty strain that lacks
sgRNA. Include a gDNA digest step using a standard commer-
cial kit such the Qiagen RNeasy kit (see Note 4). Note that the
culture conditions should be such that the gene of interest is
known to be expressed. It is also recommended to perform this
assay including a time-course so that a knockdown kinetic can
be measured prior to choosing the conditions for subsequent
phenotypic experiments.
2. RNA quality and quantity may be analyzed using the Nano-
drop as described above.
3. Perform a cDNA first-strand synthesis followed by an RNA
digest using ReadyScript cDNA synthesis mix (Merck), follow-
ing the manufacturer’s recommended protocol.
4. Gene expression may be measured using real-time quantitative
PCR, with primers designed to amplify the gene of interest, as
well as the dCas9 sequence.
5. Successful gene silencing should be confirmed by observation
of a reduction in transcript signature for the gene of interest in
the aTc-induced condition, with no reduction observed in the
un-induced or in the pIRL2 empty plasmid control strains.
Gene silencing should coincide with an upregulation of
dCas9 transcription following aTc induction.
3.6 HT-SPOTi: 1. Make 200 μL of M7H10 agar according to the manufacturer’s

Investigating Gene instructions supplemented with 50 μg/mL of kanamycin.
Essentiality Grow an M. abscessus culture to OD600 ~ 1.0
(in 10 mL M7H9).
2. Prepare the reagent dispenser to dispense 200 μL into each well

of the 96-well plate. Clean the cassettes by priming with 70%
ethanol, then hot sterile ddH2O.
3. While still hot (~50 °C), dispense 200 μL of M7H10 agar
through the cassette into the 96-well plate. Upscale or down-
scale the volume depending on how many well plates you need.
Allow the plate to cool to room temperature. Clean the reagent
dispenser immediately by priming with hot sterile ddH2O then
70% ethanol.
4. Once the plate is cooled, add 2 μL of increasing concentrations
of aTc on each row (see Note 5). For example, in row A, add
0 ng/mL aTc. In row B, add 1 ng/mL; in row C add 10 ng/
mL, etc. A range between 0 ng/mL and 100 ng/mL can be
used, however we have tested growth up to 1000 ng/mL (see
Note 6). Shake for 5 s in the reagent dispenser to allow
adequate coverage across the well. Allow to air dry for approxi-
mately 5 min.
5. From the original M. abscessus culture make ten-fold serial
dilutions of your bacteria—e.g., 1 mL of culture in an Eppen-
dorf, then take 100 μL of that culture and add to 900 μL of
M7H9 (with 50 μg/mL kanamycin) to make your first
dilution—repeat until you have the desired number of
dilutions.
6. Take 100 μL from each microcentrifuge tube and add to a new,
clean 96-well plate on one row, e.g., row A, column 1 will
contain 100 μL from the 1 mL culture. Row A, column 2 will
contain 100 μL from the first dilution.
7. Using a multichannel pipette, take 2 μL from across the whole
row and carefully add it to each row of the 96-well agar plate.
Be careful not to pierce the agar with the pipette tips.
8. Leave the culture to air dry for approximately 5 min to absorb
into the culture medium. When complete, place the lid on the
plate and parafilm well. Leave in a static incubator at 37 °C and
record results daily. If a gene is essential, at highest induction
(e.g. 100 ng/mL aTc) you should expect to see no spot growth
on the plate.
9. For antimicrobial susceptibility testing with pure compounds,
please refer to the HT-SPOTi method [12, 13]. For formu-
lated combinations, please refer to Chap. 4.
4 Notes
1. The scores of the PAM sequences can be found below pub-

lished by Bosch et al. (2021).
Rock et al
(PMID:
Feature 28165460) Latest scores
type Feature value Coefficient Standard Error scores (v4)
pam NNAGAAG -0.475817348 0.008734433 1 3
pam NNAGAAT -0.488880689 0.009939129 2 1
pam NNAGAAA -0.483132276 0.009819937 3 2
pam NNGGAAG -0.447069048 0.006684094 4 6
pam NNAGAAC -0.435568754 0.00696453 5 7
pam NNGGAAA -0.462225545 0.008070677 6 5
pam NNAGCAT -0.311428851 0.007047131 7 11
pam NNAGGAG -0.295406866 0.010454085 8 13
pam NNAGGAT -0.349796506 0.006605494 9 8
pam NNAGCAA -0.343255004 0.007165109 10 9
pam NNGGAAC -0.3390862 0.006176493 11 10
pam NNGGAAT -0.464826227 0.007108415 12 4
pam NNAGCAG -0.205407769 0.005037596 13 15
pam NNAGGAA -0.307675194 0.007352754 14 12
pam NNAGGAC -0.221139096 0.007400517 15 14
pam NNGGGAG -0.131023581 0.008648108 16 19
pam NNGGGAT -0.205029516 0.006042356 17 16
pam NNGGGAA -0.17913605 0.006986338 18 17
pam NNAGCAC -0.133758901 0.004883204 19 18
pam NNGGGAC -0.086574295 0.006839874 20 21
pam NNGGCAT -0.108215289 0.005737048 21 20
pam NNGGCAG -0.02998046 0.005098359 22 23
pam NNGGCAA -0.080786224 0.00579652 23 22
pam NNGGCAC 0 N/A 24 24
2. Other publications have varied volumes.

3. Alternatively, you can use a calculator available such as New
England Biolabs or Integrated DNA Technologies to deter-
mine the annealing temperature of your primers based on the
Taq polymerase you are using.
4. Mycobacterial RNA can be challenging to extract due to the

cell wall. Alternative to using a kit, a manual extraction method
can be utilized to extract higher yields with greater purity.
5. The volume has been optimized to 2 μL—increased volumes
make the agar too wet when adding the bacteria and will give
inconsistent results.
6. We have observed with M. abscessus using higher concentra-
tions of aTc may be required due to lower repression levels
which has also been observed by another group [18]. Optimi-
sation using this plasmid should be species specific.
References
1. WHO (2022) Global Tuberculosis 11. Faulkner V, Cox AA, Goh S et al (2021) Re-
Report:2022 sensitization of Mycobacterium smegmatis to
2. Murray CJ, Ikuta KS, Sharara F et al (2022) Rifampicin Using CRISPR Interference
Global burden of bacterial antimicrobial resis- Demonstrates Its Utility for the Study of
tance in 2019: a systematic analysis. Lancet Non-essential Drug Resistance Traits. Front
6736:02724-0 Microbiol 11:619427
3. Andries K, Villellas C, Coeck N et al (2014) 12. Evangelopoulos D, Bhakta S (2010) Rapid
Acquired resistance of mycobacterium tuber- methods for testing inhibitors of mycobacterial
culosis to bedaquiline. PLoS One 9:e102135 growth. In: Gillespie S, McHugh T (eds) Anti-
4. Rock JM, Hopkins FF, Chavez A et al (2017) biotic resistance protocols. Humana Press, pp
Programmable transcriptional repression in 193–201
mycobacteria using an orthogonal CRISPR 13. Danquah CA, Maitra A, Gibbons S et al (2016)
interference platform. Nat Microbiol 2:16274 HT-SPOTi: a rapid drug susceptibility test
5. Rock J (2019) Tuberculosis drug discovery in (DST) to evaluate antibiotic resistance profiles
the CRISPR era. PLoS Pathog 15:1–10 and novel chemicals for anti-infective drug dis-
covery. Curr Protoc Microbiol 40:17.8.1–
6. Bosch B, DeJesus MA, Poulton NC et al 17.8.12
(2021) Genome-wide gene expression tuning
reveals diverse vulnerabilities of M. tuberculo- 14. Abrahams GL, Kumar A, Savvi S et al (2012)
sis. Cell 184:4579–4592.e24 Pathway-selective sensitization of mycobacte-
rium tuberculosis for target-based whole-cell
7. Li S, Poulton NC, Chang JS et al (2022) CRIS- screening. Chem Biol 19:844–854
PRi chemical genetics and comparative geno-
mics identify genes mediating drug potency in 15. Kurepina N, Chen L, Composto K et al (2022)
Mycobacterium tuberculosis. Nat Microbiol CRISPR inhibition of essential peptidoglycan
7:766–779 biosynthesis genes in mycobacterium abscessus
and its impact on b-lactam susceptibility. Anti-
8. McNeil MB, Keighley LM, Cook JR et al microb Agents Chemother 66:e0009322
(2021) CRISPR interference identifies vulner-
able cellular pathways with bactericidal pheno- 16. Bird JE, Marles-Wright J, Giachino A (2022) A
types in Mycobacterium tuberculosis. Mol user’s guide to golden gate cloning methods
Microbiol 116:1033–1043 and standards. ACS Synth Biol 11:3551–3563
9. Rock JM, Hopkins F, Chavez A et al (2016) 17. Parish T, Brown AC (2009) Mycobacteria Pro-
Programmable transcriptional repression in tocols Second Edition Humana Press Totowa
mycobacteria using an orthogonal CRISPR NJ
interference platform. Nat Microbiol 2:16274 18. Gupta R, Rohde KH (2023) Implementation
10. El Bakali J, Blaszczyk M, Evans JC et al (2023) of a mycobacterial CRISPRi platform in Myco-
Chemical validation of mycobacterium tuber- bacterium abscessus and demonstration of the
culosis phosphopantetheine adenylyltransferase essentiality of ftsZ. Tuberculosis 138:102292.
using fragment linking and CRISPR interfer- https://doi.org/10.1016/j.tube.2022.
ence. Angew Chem Int Ed Engl 62: 102292
e202300221
Chapter 4
Modified HT-SPOTi: An Antimicrobial Susceptibility Testing

to Evaluate Formulated Therapeutic Combinations Against
Bacterial Growth and Viability
Anushandan Navaratnarajah, Chris Daniel, and Sanjib Bhakta
Abstract
Antimicrobial resistance (AMR) poses a serious threat to global health, potentially causing 10 million
deaths per year globally by 2050. To tackle AMR, researchers from all around the world have generated a
selection of various formulated (viz. nanoparticulate, liposomal) therapeutic combinations to be evaluated
for new antimicrobial drug discovery. To meet the urgent need for accelerating new antibacterial drug
development, we need rapid but reliable whole-cell assay methods and models to test formulated therapeu-
tic combinations against several pathogens in different in vitro conditions as models of actual infections.
Over the past two decades, high-throughput spot-culture growth inhibition assay (HT-SPOTi) has been
demonstrated to be a gold-standard drug susceptibility method for evaluating novel chemotherapeutic
entities and existing drugs against various microbes of global concern. Our modified HT-SPOTi method
serves the purpose of evaluating drug combinations against Gram-positive/negative microorganisms as well
as acid-fast bacilli. The newly developed and modified HT-SPOTi assay builds upon the limitations of our
previously published method to incorporate antimicrobial susceptibility testing with formulated therapeutic
combinations. The modified HT-SPOTi is compared with a range of other antimicrobial susceptibility
testing methods and validated using a library of existing antibiotics as well as formulated therapeutic
combinations. The modified HT-SPOTi assay can serve as an efficient and reliable high-throughput drug
screening platform to discover new potential antimicrobial molecules, including as part of therapeutic
formulations.
This chapter describes the generation of drug susceptibility profile for formulated therapeutic combina-
tions using modified HT-SPOTi in a semi-automated system.
Key words Antimicrobial resistance (AMR), ESKAPE pathogens, Formulated therapeutic combina-
tions, Antimicrobial susceptibility testing (AST), Combination therapy
1 Introduction
Antimicrobial resistance (AMR) poses a significant threat to global

health economy in the twenty-first century. If this is not tackled
immediately, by 2050 AMR could potentially cause 10 million
deaths a year [1]. Global researchers create diverse therapeutics to
https://doi.org/10.1007/978-1-0716-3981-8_4,
35
36 Anushandan Navaratnarajah et al.
combat AMR and accelerate drug discovery, emphasising rapid and

reliable assays to test formulations against multiple pathogens [2, 3].
Over two decades, HT-SPOTi has become a gold standard for
evaluating drugs against global microbial threats [4–7]. AMR has
been identified as a global public health concern by organizations
like the World Economic Forum, Infectious Diseases Society of
America, and World Health Organization [8]. The need for devel-
oping new drugs with novel modes of action to tackle AMR is
urgently required [9–11].
Liposome technology has led to the development of numerous
drug formulations for human use, many of which are currently in
clinical trials [12, 13]. Encapsulation in liposomes improves thera-
peutic indices by altering pharmacokinetics and pharmacodynam-
ics. Liposomes can encapsulate drugs, and adjunctive chemical
entities including nanoparticles with different solubilities, such as
hydrophobic and hydrophilic drug-like molecules [14]. They are
powerful drug carriers that protect substances from degradation,
extend half-lives, control drug release, and provide biocompatibil-
ity and safety [15]. They selectively deliver drug loads to diseased
sites, reducing side effects, increasing tolerated doses, and improv-
ing therapeutic benefit [16]. However, exposure to high tempera-
tures can break down the liposomal structure, limiting its use as a
carrier [17]. HT-SPOTi has proven to be the gold standard for
evaluating drugs and chemotherapeutics against a wide range of
microorganisms [4, 18]. Nevertheless, exposing formulated drug
(s)/inhibitor(s) with hot agar (50–60 °C) in this HT-SPOTi
method limited its use for evaluating liposomal or nano-particulate
formulations in vitro.
In this chapter, we present a rapid, high-throughput, solid
agar-based method for determining the inhibitory concentration
of formulated therapeutic combinations [19] against bacteria. The
in vitro spot culture growth inhibition assay described here uses a
96-plate format [4] and is therefore a very economical, high-
throughput method that can be performed in a general microbiol-
ogy laboratory with a basic configuration. This is summarized in
Fig. 1. The test is based on the type of bacilli growth, with growth
being measured by generated spots arising from approximately
1000 viable bacterial cells at the center of the wells and containing
a range of different concentrations of formulated compounds
[4]. In this modified HT-SPOTi method, the serially diluted for-
mulated therapeutic compounds are evenly spread on the surface of
the agar medium and allowed to diffuse within the agar. This allows
optimal access to the growing bacteria in order to affect the bacte-
rial growth or viability. We describe this method using Staphylococ-
cus aureus (Gram +ve bacteria), Escherichia coli (Gram -ve bacteria),
Mycobacterium abscessus (Acid-fast bacteria) in this chapter, but this
Modified HT-SPOTi: An Antimicrobial Susceptibility Testing to Evaluate. . . 37
Fig. 1 Workflow of modified HT-SPOTi. (Created with BioRender.com): On the 96-well plate, different test
compounds or formulations were added into row A to G. From column 1 to column 11 show two-fold
decreasing concentrations of the respective compound/formulation. A dark circle around the well is to indicate
the minimum inhibitory concentration of the compound/formulation tested. Abbreviations: ZN Ziehl–Neelsen,
DMSO Dimethyl sulfoxide, MIC Minimum inhibitory concentration)
method could also be used effectively for a range of other microbial

species including fast-growing environmental bacteria as well as
slow-growing intracellularly surviving pathogens such as Mycobac-
terium tuberculosis.
2 Materials
2.1 Growth Medium 1. LB liquid medium (for Gram +ve/-ve bacteria).

2. Middlebrook 7H9 (M7H9) broth with 0.2% Glycerol and
0.05% Tween-80 (for mycobacteria).
3. Albumin–Dextrose–Catalase (ADC) growth supplement.
4. 50 mL falcon tube.
2.2 Drug Testing 1. Formulated therapeutic combinations to be tested in liquid

form (see Note 1).
2. Known antibiotic to be used as a control.
3. Dimethyl sulfoxide (DMSO), double-distilled water (ddH2O)
or appropriate solvent.
4. 96 well plate (V bottom).
2.3 In Vitro Modified 1. Bacterial cell cultures at the mid-exponential phase

Spot Growth Inhibition (OD600 ~ 1.0).
Assay 2. Reagent dispenser (e.g., Multidrop Combi Reagent
Dispenser).
3. LB agar medium (for Gram +ve/-ve bacteria).
4. Middlebrook 7H10 (M7H10) agar medium with 0.5% Glyc-
erol (for mycobacteria).
5. Growth supplement for bacteria where needed; e.g., Albumin–
Dextrose–Catalase (ADC)/Oleic acid–Albumin–Dextrose–
Catalase (OADC) for fast/slow growing mycobacteria
respectively.
6. 96-well plate (flat bottom).
3 Methods
3.1 Preparation of 1. Culture Mycobacterium abscessus in 10 mL of Middlebrook

Bacterial and 7H9 (M7H9) liquid medium supplemented with 10% (vol/-
Mycobacterial vol) Albumin–Dextrose–Catalase (ADC) 0.2% of glycerol and
Cultures 0.05% of Tween-80 at 37 °C in 50 mL falcon tube in a rotating
incubator at 180 rpm.
2. Culture Escherichia coli and Staphylococcus aureus in 10 mL of

LB liquid medium at 37 °C in 50 mL falcon tube in a rotating
incubator at 180 rpm.
3. Grow the bacterial cultures until the mid-exponential phase
when the OD600 ~ 1.0.
4. For the quality control, stain the culture with a relevant stain,
for example, Ziehl–Neelsen (ZN) acid fast staining or Gram
staining.
5. Dilute the bacterial culture with growth medium to obtain a
final concentration of ~1000 cells per well in a volume of 2 μL
(see Note 2).
3.2 Preparation of 1. Natural extract or compound to be used as a control is dis-

Master Plate for Drug solved in 100% DMSO, sterile double distilled water (ddH2O),
Testing or any other appropriate solvent to make 50 mg/mL stock (see
Note 3).
2. Use a 96-well plate (V bottom) and perform two-fold serial
dilutions of the drug stocks with an appropriate solvent to give
a wide concentration range (see Note 4).
To do this:
(a) Add a chosen volume (e.g., 50 μL) of solvent to columns
2–11 of the plate.
(b) Add double the chosen volume (e.g., 100 μL) of the
original stock of the compounds in column 1.
(c) Now, take the chosen volume (e.g., 50 μL) of the drug
suspension from column 1 and add to column 2 mixing
the solution by gently pipetting.
(d) Carry on in the same manner until column 11.
3. For formulated therapeutic combinations, step 2 is carried out
using an appropriate solvent.
3.3 Agar Plate 1. After autoclaving the Middlebrook 7H10 (M7H10) agar
Preparation medium supplemented with 0.5% glycerol, keep the agar hot
(~50 °C) in a steam bath and add 10% (v/v) OADC to it. Place
it in a water bath kept at a temperature of 55° to 60 °C to
prevent it from solidifying.
2. Sterilize the dispensing cassette and tubing by running warm
70% ethanol through followed by priming it with hot (~65 °C)
sterile water (see Note 5).
3. Program the dispenser to suit the plate type (96-well, flat
bottom) and the volume of agar to be dispensed (e.g.,
200 μL) (see Note 6).
4. Dispense the agar medium and leave the plates in the safety
cabinet undisturbed to solidify (~5 min). Once set, cover the
plates with lid (see Note 7).
3.4 Adding Drugs to 1. Using a multi-channel pipette, transfer 2 μL of serially diluted

Prepared Agar Plate formulated therapeutic combinations or natural extract or drug
from your master plate column 1–11 on to the prepared agar
plate column 1–11 (see Note 8).
2. Repeat with this step until row A to G of prepared agar plate.
3. Transfer 2 μL of 100% DMSO or sterile ddH2O (or other
appropriate solvent) in wells of row H1 to H6 and column
A12 to D12 as a control.
4. Place plate in the holder of the equipment and shake for 1 min
(see Note 9).
5. Leave the plates in the safety cabinet undisturbed to air-dry for
5 min.
3.5 Spotting 1. Sterilize the dispensing cassette and tubing by running warm
Bacterial Cultures 70% ethanol through followed by priming it with hot (~65 °C)
sterile water.
2. Program the dispenser to suit the plate type (96-well, F bot-
tom), volume of bacteria to be dispensed (2 μL), and the wells
in column 1–11 (see Note 10).
3. Dispense the bacteria on growth medium. Leave the plates in
the safety cabinet undisturbed for 5 min.
4. Seal the plates with parafilm, and leave in a static incubator at
37 °C and record spots observed daily.
4 Notes
1. Liposomal formulations are phospholipid bilayers formed

through hydration, prepared using film hydration method.
They are prepared by dissolving phospholipids in organic sol-
vents (such as chloroform, methanol, ether), drying, and
hydrating with aqueous media (such as water, buffers, or saline
solution) [20].
2. The number of initial cells has been optimized to ~1000 cells
per well—the presence of more cells prevents the observation
of single spots.
3. DMSO at the final concentration of 0.1% does not cause any
effect on the growth of bacterial cultures [21].
4. The plates can be sealed with parafilm and stored in a 4 °C
fridge for 2 days for formulated combinations [22].
5. The reagent dispenser has a “Prime” button, which can be used

to prime the cassette for 1–2 min to run warm 70% ethanol or
hot sterile water. Ensure uniform dispensing by checking the
nozzles. In the case of agar blocking the nozzles, a small pin can
be used to clear the agar and re-prime with hot sterile water.
6. The reagent dispenser has a “Full plate” function which can be
used to select the full plate for dispensing. It is important to
select and check the correct volumes for dispensing (e.g.,
200 μL for agar or 2 μL for bacteria).
7. The reagent dispenser has a “Start” function, which can be
used to dispense the growth medium. The plates can be sealed
with parafilm and stored in a 4 °C fridge for 2 days prior to the
experiment [23]. At this stage, they can also be transferred to
other laboratories if required.
8. The volume has been optimized to 2 μL—increased volumes
make the agar too wet for when adding the bacteria and will
give inconsistent results. Use only A to G row for testing, and
leave row H for drug-free control.
9. The reagent dispenser has a “Shake” function, which can be
used to ensure uniform distribution of the drug on the growth
medium.
10. The reagent dispenser has a “Select column” function, which
can be used to select column 1–11, instead of full plate for
medium dispensing if required.
Acknowledgments
Formulated therapeutic combinations were kindly gifted to us by

Dr. Satyanarayana Somavarapu from the UCL School of Pharmacy.
We thank Dr. Veronique Seidel from the University of Strathclyde
for critically reviewing the chapter.
References
1. Murray CJ, Ikuta KS, Sharara F et al (2022) 4. Evangelopoulos D, Bhakta S (2010) Rapid
Global burden of bacterial antimicrobial resis- methods for testing inhibitors of mycobacterial
tance in 2019: a systematic analysis. Lancet growth. Methods Mol Biol 642:193–201
399:629–655 5. Daraee H, Etemadi A, Kouhi M et al (2016)
2. Saifullah B, Chrzastek A, Maitra A et al (2017) Application of liposomes in medicine and drug
Novel anti-tuberculosis nano delivery formula- delivery. Artif Cells Nanomedicine Biotechnol
tion of ethambutol with graphene oxide. Mole- 44:381–391
cules 22(10):1560 6. Danquah CA, Maitra A, Gibbons S, et al
3. Wang Y, Grainger DW (2022) Regulatory con- (2016) HT-SPOTi: a rapid drug susceptibility
siderations specific to liposome drug develop- test (DST) to evaluate antibiotic resistance pro-
ment as complex drug products. Front Drug files and novel chemicals for anti-infective drug
Deliv 2:901281 discovery. 17.8.1–17.8.12
7. Collins LA, Franzblau SG (1997) Microplate 16. Sercombe L, Veerati T, Moheimani F et al

alamar blue assay versus BACTEC 460 system (2015) Advances and challenges of liposome
for high-throughput screening of compounds assisted drug delivery. Front Pharmacol 6:286
against Mycobacterium tuberculosis and Myco- 17. Amin M, Huang W, Seynhaeve ALB et al
bacterium avium. Antimicrob Agents Che- (2020) Hyperthermia and temperature-
mother 41:1004–1009 sensitive nanomaterials for spatiotemporal
8. WHO Global antimicrobial resistance and use drug delivery to solid tumors. Pharmaceutics
surveillance system (GLASS) report, https:// 12:1–23
w w w. w h o . i n t / p u b l i c a t i o n s / i / i t e m / 18. Rizi K, Murdan S, Danquah CA et al (2015)
9789240062702 Development of a rapid, reliable and quantita-
9. Mayorga-Ramos A, Zúñiga-Miranda J, tive method – “SPOTi” for testing antifungal
Carrera-Pacheco SE et al (2023) CRISPR- efficacy. Microbiol Methods 117:36–40
Cas-based antimicrobials: design, challenges, 19. Mursaleen L, Chan SHY, Noble B et al (2023)
and bacterial mechanisms of resistance. ACS Curcumin and n-acetylcysteine nanocarriers
Infect Dis 9(7):1283–1302 alone or combined with deferoxamine target
10. Slater KB, Kim D, Chand P et al (2023) A the mitochondria and protect against neuro-
current perspective on the potential of nano- toxicity and oxidative stress in a co-culture
medicine for anti-tuberculosis therapy. Trop model of Parkinson’s disease. Antioxidants
Med Infect Dis 8:100 (Basel) 12:130
11. Daniel C, Bhakta S (2022) Immunobiology of 20. Lee MK (2020) Liposomes for enhanced bio-
tubercle bacilli and prospects of immunomod- availability of water-insoluble drugs: in vivo
ulatory drugs to tackle tuberculosis (TB) and evidence and recent approaches. Pharmaceutics
other non-tubercular mycobacterial infections. 12:264
Immunobiology 227:152224 21. Guzman JD, Gupta A, Evangelopoulos D et al
12. Ho Yi Chan S, Sheikh K, Zariwala MG et al (2010) Anti-tubercular screening of natural
(2023) Dry powder formulation of azithromy- products from Colombian plants:
cin for COVID-19 therapeutics. J Microencap- 3-methoxynordomesticine, an inhibitor of
sul 40:1–37 MurE ligase of Mycobacterium tuberculosis. J
13. Bulbake U, Doppalapudi S, Kommineni N et al Antimicrob Chemother 65:2101–2107
(2017) Liposomal formulations in clinical use: 22. Šturm L, Ulrih NP (2021) Basic methods for
an updated review. Pharmaceutics 9:12 preparation of liposomes and studying their
14. Nsairat H, Khater D, Sayed U et al (2022) interactions with different compounds, with
Liposomes: structure, composition, types, and the emphasis on polyphenols. Int J Mol Sci
clinical applications. Heliyon 8:e09394 22:6547
15. Liu P, Chen G, Zhang J (2022) A review of 23. Mialon M, Tang Y, Singh AK et al (2012)
liposomes as a drug delivery system: current Effects of preparation and storage of agar
status of approved products, regulatory envir- media on the sensitivity of bacterial forward
onments, and future perspectives. Molecules scattering patterns. Open J Appl Sci 01:26–35
27:1372
Chapter 5
Investigating Combination Therapy as a Means to Enhance

Activity and Repurpose Antimicrobials
Abstract
Current clinical practice assumes that a single antibiotic given as a bolus or as a course will successfully treat
most infections. In modern medicine, this is becoming less and less true with drug-resistant, multi-drug-
resistant, extensively drug-resistant, and untreatable infections becoming more common. Where single-
drug therapy (monotherapy) fails, we will turn to multi-drug therapy. Alternatively, combination therapy
could be useful to prevent the emergence of resistance. Multi-drug therapy is already standard for some
multi-drug resistant infections and is the standard for the treatment of some pathogens such as Mycobacte-
rium tuberculosis.
The use of combination therapy for everyday infections could be a clear course out of the current AMR
crisis we are facing. With every additional drug added to a combination (n + 1) the likelihood of the
pathogen evolving resistance drops exponentially.
Many generic antibiotics are cheap to manufacture as they have fallen out of patent protection but are less
effective at pharmacologically effective doses due to overuse in the past. Combination therapy can combine
these generic compounds into cocktails that can not only treat susceptible and resistant infections but can
also reduce the risk of new resistances arising and can resuscitate the use of antimicrobials once thought
defunct.
In this chapter, we will summarize theory behind combination therapy and standard in vitro
methodologies used.
Key words Antibiotic resistance, Combination therapy, Therapeutics, Chequerboard analysis, Syn-
ergy, Fractional inhibition concentration
1 Introduction
Antimicrobial resistance (AMR) has become a major public health

issue [1–4]. Traditional single-drug therapy (monotherapy) has
contributed to the development of AMR alongside other issues
such as overprescribing and lax stewardship policies [5–9]. Combi-
nation therapy has emerged as a potential solution to address AMR
[10–12]. In this chapter, we will explore the current methodologies
https://doi.org/10.1007/978-1-0716-3981-8_5,
43
44 Robert J. H. Hammond
for 2-drug combination therapy and the mathematics underpin-

ning these that can be used to evaluate synergistic treatment
options.
1.1 Definition and Combination therapy refers to the use of two or more antimicrobial
Types of Combination agents in a single treatment regimen. These agents can be from
Therapy similar or different classes with similar or highly dissimilar mechan-
isms of action. Indeed, as we will see, where combinations include
agents with dissimilar but complementary mechanisms of action
the overall effect can be enhanced [12]. Synergistic, additive, indif-
ferent, and antagonistic interactions can occur between agents in a
combination therapy regimen as set out below [11, 13–15].
1.2 Synergy Synergy is defined as the combination of two or more entities

whose outcome is greater than the sum of their individual parts
[16]. In the case of antimicrobial therapy this can have a range of
meanings. Where bacterial population has become resistant to a
given antibiotic (X) a synergistic response would be to use agent X
in combination with another agent(s) and have the resistant popu-
lation destroyed, where the second agent would not have had this
effect alone. Other effects include increased efficacy, decreased
likelihood of resistance development, and, subsequently, improved
patient outcomes. These come about due to the mechanisms
behind synergy. These are multifactorial but briefly; prevention or
slowing of the development of resistance due to mutation (RATE/
FREQUENCY) [17], increasing the rate of bacterial kill [18, 19],
avoiding or circumnavigating known resistance mechanisms [20],
and the capacity to address multiple targets within a single bacte-
rium [11, 12].
1.3 Definition of Synergy can be defined mathematically. A combination of drugs

Synergy (2 only) is seen to be synergistic if it has a Fractional Inhibitory
Concentration Index (FICI) value of 0.5.
In microbiology we normally define the effectiveness of a single
agent using the minimum quantity of drug required to inhibit
growth: the minimum inhibitory concentration (MIC). The mini-
mum quantity required to kill a microbial population is the mini-
mum bacterial concentration (MBC). In combination therapy, a
different mathematical calculation is required. The Fractional
Inhibitory Concentration Index (FICI) is calculated thus:
A B
þ ¼ FI C A þ FI C B ¼ FICI
MI C A MI C B
where A and B are the MIC of each antibiotic in combination, and
MICA and MICB are the MIC of each drug individually.
In order for the FICI (also known as the ΣFIC) to be 0.5 the
MICs of the individual agents must be active in their combination
at ¼ of the monotherapy values, that is to say, 0.25 MIC. Where
Evaluating Antibiotic Combinations 45
the FICI falls above the crucial value of 0.5 other language is
necessary to define the reaction. An FICI of 0.5–2 will indicate an
additive reaction, 2–4 is defined as an indifferent reaction and an
FICI of >4 is an antagonistic reaction [21, 22] where the antibiotics
are actually inhibiting each other’s activity and causing a poorer
outcome than monotherapy.
In an in vitro case study presented below a combination of only
2 agents is used. This is the standard use-case of synergy in medi-
cine, however there are exceptions. The standard intensive phase
therapy for tuberculosis treatment includes 4 antimicrobials used in
tandem and cases of 3 or more agents used in MDR infections are
not uncommon [23]. Where this occurs, the mathematics for
defining synergy no longer work and will need to be redefined in
the future.
2 Materials
2.1 Bacterial Growth 1. Mueller Hinton cation adjusted broth (Sigma-Aldrich, UK)
(see Note 1).
2. Resistant organism, in this case a KPC Klebsiella pneumonia
(ATCC BAA 1705) (see Note 2).
3. Antimicrobials of choice, in this instance colistin and Zidovu-
dine (Sigma-Aldrich, UK) (see Notes 3 and 4).
4. Phosphate Buffered Saline (PBS) (Sigma-Aldrich, UK).
5. Sterile water.
6. 96- or 384-well plates (Nunc, UK) (see Note 5).
7. Appropriate multichannel pipettor (Gilson, or similar) (see
Notes 6 and 7).
8. Incubator (Genlab (Fisher Scientific)).
9. Spectrophotometer (BioChrom WPA Co8000, UK).
10. 96- or 384-well plate reader with the ability to detect at
600 nm absorbance (Clariostar Pro).
11. Virkon solution (2%) for decontamination (Sigma-
Aldrich, UK).
12. Microcentrifuge tubes (Eppendorf tubes®, (Sigma-
Aldrich, UK)).
3 Methods
3.1 Broth To establish the minimum inhibitory concentration (MIC) of two

Microdilution (BMD) drugs to a strain of Klebsiella pneumoniae (e.g.) a 96-well broth
microdilution (BMD) methodology is used as follows.
1. Weigh antimicrobials in their powered state and dilute in sterile

deionized water to a stock concentration of 10 mg/mL
(10 g/L)
2. Make drug dilutions at 10 final concentration in a series of
eleven 1.5 mL micro-centrifuge tubes by dilution with sterile,
deionized water (see Note 7).
3. Fill a final tube with sterile, deionized water only and use this as
an additive for the growth control (GC).
4. Add 20 μL of these 12 solutions in triplicate to a sterile 96-well
plate.
5. Add 160 μL sterile media (Muller Hinton cation adjusted) into
the same wells that contain the drugs (see Note 8).
6. Measure the optical density of the overnight culture of organ-
ism in a spectrophotometer at OD600.
7. Dilute this culture to approximately 5 106 cfu/mL
(OD600 ~ 0.01) (see Note 9).
8. This culture was added to the 96-well plate at a volume of
20 μL diluting the culture to a final concentration of
~5 105 cfu/mL and making the final volume in each well
200 μL.
9. 96-well plate was incubated overnight (~16 h) and read in a
96-well plate reader, data presented as OD600.
3.2 Checkerboard A similar assay was conducted to establish if any synergistic effect
existed when CSS and AZT were combined. This checkerboard
assay can be thought of as a two-dimensional BMD.
1. Dilute the experiment drugs in the same way as noted above in
Subheading 3.1, step 2.
2. Add the drug solution to the 96-well plate in the manner
displayed in Fig. 1. Add sterile media and bacteria as noted
above to make a final volume of 200μL and bacterial concen-
tration of ~5 105 cfu/mL
3. Measure the OD for all experimental case results using the
appropriate plate reader at 600 nm.
CSS mg/L
4 2 1 0.5 0.25 0.125 0.063 0.031 0.016 0.008 0
64 324 194 129 97 80 72 68 66 65 65 64
32 164 98 65 49 40 36 34 33 33 32 32
16 84 50 33 25 20 18 17 17 16 16 16
AZT 8 44 26 17 13 10 9 9 8 8 8 8
mg/L 4 24 14 9 7 5 5 4 4 4 4 4
2 14 8 5 4 3 2 2 2 2 2 2
1 9 5 3 2 2 1 1 1 1 1 1
0 4 2 1 1 0 0 0 0 0 0 0
Fig. 1 Schematic representation of the layout of checkerboard assay for AZT + CSS
4 Notes
1. Mueller Hinton cation adjusted broth is the recommended

media for AST testing as it is specifically formulated to not
affect the performance of antibiotics [24].
2. Like any other benchtop experiment be mindful of the specific
requirements of your organism of interest. If using a fastidious
organism that requires blood, include that in the media as you
would for a conventional culture. If using a mycobacterium use
Middlebrook or similar to support their fastidious needs.
3. The antimicrobials used can be ones to which the organism of
choice is resistant to depending on the results you are attempt-
ing to attain. “Rescuing” an antibiotic is possible with the right
combination of synergistic therapies.
4. Alternatively, the antimicrobials used can be ones to which your
organism of choice is susceptible to if the aim of your experi-
ment is to, for example, lower the concentration of a less
efficacious antibiotic that is frequently prescribed at high
doses, such as nitrofurantoin (EUCAST MIC 64 mg/L).
5. These are the most applicable formats for medium- and high-
throughput broth microdilution assays. Using a 384-well plate
evaporation can become an issue. Make sure to work in a sterile
and humid environment such as a class II biosafety cabinet to
minimize both the risks of contamination and evaporation
upsetting the volumetric calculations.
6. The pipettor will need to match the plate used, a 384-well
plate-based set up will require a 24-channel pipettor and a
96-well plate-based set up will require a 12- or 8-channel
pipettor.
7. Be aware of the limitations of multichannel pipettors; they are
notoriously inaccurate and can become more so as the heat of
your hand warms them. Pipette tops must be loaded with equal
force across all channels. Using the correct tips, as prescribed by
the manufacturer, is essential to minimize inter-channel
variation.
8. The antimicrobials are made to 10 final concentration so that
they can be added to the BMD in small volumes and not overly
dilute the media. The positive control is diluted with the same
volume of water for this reason.
9. It is important to note the maximum volume for the plates you
are using. For most standard 96-well plates the maximal vol-
ume is 200 μL. Hence the 20 μL of antimicrobial solution used
in 4. is diluted 10x as in Note 7.
10. An OD600 of 1.0 is approximately 8 108 cfu/mL based on

the specific light-absorbing characteristics of E. coli. OD600
values can be different based on different bacteria and it is
worthwhile creating a cfu vs. OD standard curve for your
organism of choice before beginning the experiment proper.
References
1. Gajdacs M, Abrok M, Lazar A, Burian K 9. Soothill G, Hu Y, Coates A (2013) Can
(2020) Increasing relevance of gram-positive we prevent antimicrobial resistance by using
cocci in urinary tract infections: a 10-year anal- antimicrobials better? Pathogens 2(2):
ysis of their prevalence and resistance trends. 4 2 2 – 4 3 5 . h t t p s : // d o i . o r g / 1 0 . 3 3 9 0 /
Sci Rep 10(1):17658. https://doi.org/10. pathogens2020422
1038/s41598-020-74834-y 10. Siriyong T, Murray RM, Bidgood LE, Young
2. Vasala A, Hytonen VP, Laitinen OH (2020) SA, Wright F, Parcell BJ et al (2019) Dual
Modern tools for rapid diagnostics of antimi- β-lactam combination therapy for multi-drug
crobial resistance. Front Cell Infect Microbiol resistant Pseudomonas aeruginosa infection:
10:308. https://doi.org/10.3389/fcimb. enhanced efficacy in vivo and comparison with
2020.00308 monotherapies of penicillin-binding protein
3. Galata V, Laczny CC, Backes C, Hemmrich- inhibition. Sci Rep 9(1):9098. https://doi.
Stanisak G, Schmolke S, Franke A, Meese E, org/10.1038/s41598-019-45550-z
Herrmann M, von Müller L, Plum A, Müller R, 11. Coates ARM, Hu Y, Holt J, Yeh P (2020)
Stähler C, Posch AE, Keller A et al (2019) Antibiotic combination therapy against resis-
Integrating culture-based antibiotic resistance tant bacterial infections: synergy, rejuvenation
profiles with whole-genome sequencing data and resistance reduction. Expert Rev Anti-
for 11,087 clinical isolates. Genomics Proteo- Infect Ther 18(1):5–15. https://doi.org/10.
mics Bioinformatics 17(2):169. https://doi. 1080/14787210.2020.1705155
org/10.1016/j.gpb.2018.11.002 12. Hu Y, Liu Y, Coates A (2019) Azidothymidine
4. O’Neill J (2016) Tackling drug-resistant infec- produces synergistic activity in combination
tions globally: final report and recommenda- with Colistin against antibiotic-resistant Enter-
tions. Government of the United Kingdom obacteriaceae. Antimicrob Agents Chemother
5. Raymond B (2019) Five rules for resistance 63(1). https://d oi.o rg/10.1 128 /aac.
management in the antibiotic apocalypse, a 01630-18
road map for integrated microbial manage- 13. Deshpande D, Srivastava S, Nuermberger E,
ment. Evol Appl 12(6):1079–1091. https:// Pasipanodya JG, Swaminathan S, Gumbo T
doi.org/10.1111/eva.12808 (2016) Concentration-dependent synergy and
6. Wilkinson A, Ebata A, MacGregor H (2018) antagonism of linezolid and Moxifloxacin in
Interventions to reduce antibiotic prescribing the treatment of childhood tuberculosis: the
in LMICs: a scoping review of evidence from dynamic duo. Clin Infect Dis 32016:S88–S94
human and animal health systems. Antibiotics 14. Dutta NK, Pinn ML, Karakousis PC (2014)
(Basel) 8(1). https://doi.org/10.3390/ Reduced emergence of isoniazid resistance
antibiotics8010002 with concurrent use of thioridazine against
7. Manyi-Loh C, Mamphweli S, Meyer E, Okoh acute murine tuberculosis. Antimicrob Agents
A (2018) Antibiotic use in agriculture and its Chemother 58(7):4048–4053. https://doi.
consequential resistance in environmental org/10.1128/AAC.02981-14
sources: potential public health implications. 15. Tripathi RP, Tewari N, Dwivedi N, Tiwari VK
Molecules 23(4). https://doi.org/10.3390/ (2005) Fighting tuberculosis: an old disease
molecules23040795 with new challenges. Med Res Rev 25(1):
8. Bebell LM, Muiru AN (2014) Antibiotic use 93–131. https://doi.org/10.1002/med.
and emerging resistance: how can resource- 20017
limited countries turn the tide? Glob Heart 16. Bonapace CR, White RL, Friedrich LV, Bosso
9(3):347–358. https://doi.org/10.1016/j. JA (2000) Evaluation of antibiotic synergy
gheart.2014.08.009 against Acinetobacter baumannii: a comparison
with Etest, time-kill, and checkerboard meth- and dominance. Dominance and antibiotic
ods. Diagn Microbiol Infect Dis 38(1):43–50. combination]. Pathol Biol (Paris) 35(5):
https://doi.org/10.1016/s0732-8893(00) 495–500
00163-2 22. Rakholiya KD, Kaneria MJ, Chanda SV (2013)
17. Long H, Miller SF, Strauss C, Zhao C, Chapter 11 – medicinal plants as alternative
Cheng L, Ye Z et al (2016) Antibiotic treat- sources of therapeutics against multidrug-
ment enhances the genome-wide mutation rate resistant pathogenic microorganisms based on
of target cells. Proc Natl Acad Sci U S A 113: their antimicrobial potential and synergistic
E2498. https://doi.org/10.1073/pnas. properties. In: Rai MK, Kon KV (eds) Fighting
1601208113 multidrug resistance with herbal extracts,
18. Kloprogge F, Hammond R, Copas A, Gillespie essential oils and their components. Academic,
SH, Della PO (2019) Can phenotypic data San Diego, pp 165–179
complement our understanding of antimyco- 23. Umemura T, Kato H, Hagihara M, Hirai J,
bacterial effects for drug combinations? J Anti- Yamagishi Y, Mikamo H (2022) Efficacy of
microb Chemother 74:3530. https://doi.org/ combination therapies for the treatment of
10.1093/jac/dkz369 multi-drug resistant gram-negative bacterial
19. Kloprogge F, Hammond R, Copas A, Gillespie infections based on meta-analyses. Antibiotics
SH, Della PO (2019) Can phenotypic data (Basel) 11(4). https://doi.org/10.3390/
complement our understanding of antimyco- antibiotics11040524
bacterial effects for drug combinations? J Anti- 24. Koeth LM, King A, Knight H, May J, Miller
microb Chemother 74(12):3530–3536. LA, Phillips I et al (2000) Comparison of
https://doi.org/10.1093/jac/dkz369 cation-adjusted Mueller–Hinton broth with
20. Lewis K (2008) Multidrug tolerance of bio- Iso-Sensitest broth for the NCCLS broth
films and persister cells. Curr Top Microbiol microdilution method. J Antimicrob Che-
Immunol 322:107–131 mother 46(3):369–376. https://doi.org/10.
21. Drugeon HB, Dellamonica P, Caillon J (1987) 1093/jac/46.3.369
[Synergy, addition, indifference, antagonism
Chapter 6
Investigating Photoactive Antimicrobials as Alternatives

(or Adjuncts) to Traditional Therapy
Robert J. H. Hammond and Marianna Leite De Avellar
Abstract
Photodynamic therapy (PDT) is an established therapy used for the treatment of cutaneous skin cancers and
other non-infective ailments. There has been recent interest in the opportunity to use aPDT (antimicrobial
PDT) to treat skin and soft tissue infections. PDT utilizes photosensitizers that infiltrate all cells and
“sensitize” them to a given wavelength of light. The photosensitizer is simply highly absorbent to a given
wavelength of light and when excited will produce, in the presence of oxygen, damaging oxygen radicals
and singlet oxygen. Bacterial cells are comparatively poor at combatting oxidative stress when compared
with human cells therefore a degree of selective toxicity can be achieved with aPDT.
In this chapter, we outline methodologies for testing aPDT in vitro using standard lab equipment.
Key words Antibiotic combination, Photodynamic therapy, Drug development, MDR therapy,
Photosensitizers, Susceptibility
1 Introduction
Antimicrobial resistance (AMR) is a global concern that poses a

significant threat to public health [7, 14, 18, 20]. To address this,
the development of new and alternative antimicrobial agents is
crucial [4, 10, 11, 16, 20]. Photoactive antimicrobials are a new
class of compounds that exhibit antimicrobial activity upon expo-
sure to light [2, 3, 5, 12, 13]. This chapter describes a protocol for
investigating the photoactivity of new antimicrobial compounds
against a panel of bacterial pathogens.
Photodynamic therapy (PDT) is a form of treatment that uses a
light source and a chemical compound (photosensitizer) which,
when activated by light in the presence of oxygen, generates reac-
tive oxygen species. Photodynamic therapy is currently applied in
the treatment of skin cancer, early stages of esophageal and lung
cancer, acne, and some oral infections [9].
https://doi.org/10.1007/978-1-0716-3981-8_6,
51
52 Robert J. H. Hammond and Marianna Leite De Avellar
For antimicrobial applications, PDT there have been promising

results in pre-clinical and early clinical studies with effective patho-
gen killing and resolution of the infection with low side effects
[12, 13], including infections caused by bacteria, fungi, viruses,
and leishmaniasis [1, 5, 15, 19, 21]. One clinical example is the
2013 study by Fu and colleagues [6] in which they report success
with the use of methylene blue in clearing MRSA from the anterior
nares of patients in hospitals in China.
Treatment of infected chronic wounds, such as diabetic foot
ulcers, where blood flow and immune responses are limited, can be
complicated and long lasting and sometimes including several
courses of systemic antibiotics [17]. Application of topical antimi-
crobial PDT in this type of lesion benefits from direct local action,
avoiding systemic side effects, and inducing immune response and
expression of growth factors locally. Organic light emitting diodes
[OLEDs] are electroluminescent devices that can be produced as
small, thin, and lightweight wearable devices, and still provide the
homogeneous and efficient light output to be used as the light
source for PDT. Application of OLEDs instead of the light sources
currently used, which are both large and rarely available in health
centers presently, could facilitate treatment for patients, decreasing
time inside health care facilities and allowing ambulatory
treatment [12].
PDT causes non-selective oxidative damage to structures, act-
ing in multiple sites in vicinity to the location of the photosensitizer
[2, 3, 9]. Resistance induced by PDT has not been observed in
bacteria, and is, therefore a promising treatment in an era of anti-
microbial resistance and ongoing necessity to develop new antimi-
crobial treatments.
2 Materials
2.1 Growth of 1. Mueller Hinton cation adjusted broth (Sigma-Aldrich, UK)

Microorganisms (see Note 1).
2. A range of organisms including resistant strains (MRSA) and
susceptible, wild type or lab strains (MSSA ATCC 25923) (see
Note 2).
3. 96-well plates (Nunc, UK) (see Note 3).
4. Appropriate multichannel pipettor (Gilson, or similar) (see
Note 4).
(a) Be aware of the limitations of multichannel pipettors; they
are notoriously inaccurate and can become more so as the
heat of your hand warms them. Pipette tips must be
loaded with equal force across all channels.
(b) Using the correct tips, as prescribed by the manufacturer,
is essential to minimize inter-channel variation.
Investigating Photoactive Antimicrobials 53
2.2 aPDT 1. Photosensitizer of choice, in this case methylene blue (see

Note 5).
3. Sterile water.
5. 96-well plate reader with the ability to detect at 600 nm and
450 nm absorbance (Clariostar Pro).
(a) The use of two wavelengths is crucial here as the 600 nm
reading will give an industry-standard value for the num-
ber of bacteria present (OD600) but will activate the meth-
ylene blue. Therefore, if measuring the status of the
experiment during the incubation period only the
450 nm absorbance measurement should be used.
(b) The 450 nm reading also allows for the fine calibration
of the concentration of methylene blue present in each
well and can help quantify photobleaching in longer
experiments.
Aldrich, UK).
7. Microcentrifuge tubes (Eppendorf tubes®, (Sigma-
Aldrich, UK)).
8. Light emitter (see Note 6).
(a) Make sure to use a wavelength of light in the peak absorp-
tion range of the photosensitizer. For example, methylene
blue has a peak absorbance of ~665 nm therefore a
red-light source as close to 665 nm as possible is used.
3 Methods
3.1 Bacterial Strains 1. Bacterial culture S. aureus (ATCC 25923), S. epidermidis, and
and Growth Conditions (MRSA/S. pyogenes) from glycerol stock (50%) grown in
Mueller-Hinton cation adjusted broth (MHB2) (Sigma-
Aldrich) at 37 C for 16 h.
2. The optical density of the culture should be measured on a Cell
density meter (Biochrome WPA CO8000, UK) at wavelength
600 nm (OD600) and diluted to be used as final OD600 equal
to 0.001 (approximately 105 cfu/mL) or 0.1 (approximately
107 cfu/mL).
3.2 Photosensitizer The photosensitizer solution was prepared by diluting Methylene

Solution blue 1.5% (Sigma-Aldrich, UK) in filtered (syringe filter 0.22 μm)
Phosphate Buffered Saline (PBS—Sigma-Aldrich, UK) to make
5 mM or 1 mM stocks and stored in the dark.
3.3 Toxicity of 1. Toxicity of methylene blue in the dark was measured on

Methylene Blue in the S. aureus, S. epidermidis, and S. pyogenes at ~105 cfu/mL,
Dark using two-fold decreasing concentrations of methylene blue
from 160 μM to 0.3125 μM.
2. Samples were set as 200 μL on 96-well clear plates in triplicates,
positive controls as the correspondent bacteria and negative
control as MHB2.
3. Absorbance of the samples was measured on ClarioStar plate
reader (BMG LABTECH, Germany) at 450 nm endpoint and
then incubated in the dark for 6 h at 37 C.
4. After incubation, absorbance was measured again with the
same settings.
5. A viability test was performed after the incubation period.
6. Samples were plated on Muller-Hinton Agar in triplicates and
incubated at 37 C overnight.
7. For S. epidermidis and S. pyogenes, the experiment was repeated
with higher concentrations of methylene blue, from 640 μM to
1.25 μM. Toxicity test on S. aureus can also performed on
other substrates such as 384-well plates, cuvettes, or any
light-permissible labware. In these cases, bacteria are exposed
to two-fold dilutions of methylene blue from 15.63 μM to
0.12 μM. Samples are set as 1.5 mL in (for example) plastic
cuvettes in duplicates with S. aureus ~105 cfu/mL, positive
controls used as the S. aureus at the same concentration and
negative control as MHB2.
3.4 Photodynamic Antimicrobial photodynamic therapy in vitro tests described below

Therapy Tests were conducted in 96-well plate format.
1. Experiments were set up as in Toxicity of Methylene blue in the
dark.
2. PDT samples contained S. aureus at ~105 cfu/mL, Methylene
blue at different concentrations (2.5 μM and 0.25 μM) and
photoactivated by the OLEDs for 6 h, with Irradiance reaching
the samples at 5 mW/cm2, providing a total power of 108 J/
cm2 (see Note 7).
3. The control samples used are:
– Dark control (containing methylene blue and bacteria, with-
out exposure to light from the OLEDs).
– Positive control (bacteria in MHB2) and negative control
(MHB2).
Investigating Photoactive Antimicrobials 55
4 Notes

2. Like any other benchtop experiment be mindful of the specific
requirements of your organism of interest. The medium can be
adjusted if required. For example, if using a fastidious organism
that requires blood, include that in the media as you would
using a conventional culture. If using a cutaneous mycobacte-
rium (for example) use Middlebrook or similar.
3. This is the most applicable format for medium- and high-
throughput assays. Make sure to work in a sterile and humid
environment such as a class II biosafety cabinet to minimize
both the risks of contamination and evaporation upsetting the
volumetric calculations.
4. The pipettor will need to match the plate used, a 384-well
plate-based set up will require a 24-channel pipettor and a
96-well plate-based set up will require a 12- or 8-channel
pipettor.
5. Methylene blue has been approved as a photosensitizer by the
FDA for use in humans. Other photosensitizers are also being
investigated so do not limit your work to a single treatment
option. For example, toluidine blue is the other most common
PS for aPDT. Others include: Dimethyl methylene blue, New
methylene blue, Rose Bengal, Curcumin, hypericin, titanium
dioxide, and Fullerenes. These are the most applicable format
for medium- and high-throughput assays.
6. This can be almost any radiation source but checking local
guidelines is essential.
7. These values will be different for each individual experiment.
Knowing them in your own set up is crucial.
References
1. Alamon-Reig F, Marti-Marti I, Loughlin CR, 3. Boltes Cecatto R, Siqueira de Magalhães L,
Garcia A, Carrera C, Aguilera-Peiro P (2022) Fernanda Setúbal Destro Rodrigues M,
Successful treatment of facial cutaneous leish- Pavani C, Lino-dos-Santos-Franco A, Teixeira
maniasis with photodynamic therapy. Indian J Gomes M et al (2020) Methylene blue
Dermatol Venereol Leprol 88(5):667–670. mediated antimicrobial photodynamic therapy
https://doi.org/10.25259/IJDVL_1175_ in clinical human studies: the state of the art.
2021 Photodiagn Photodyn Ther 2020(31):
2. Benov L (2015) Photodynamic therapy: cur- 101828. https://doi.org/10.1016/j.pdpdt.
rent status and future directions. Med Princ 2020.101828
Pract 24(Suppl 1):14–28. https://doi.org/ 4. Coates ARM, Hu Y, Holt J, Yeh P (2020)
10.1159/000362416 Antibiotic combination therapy against
resistant bacterial infections: synergy, rejuve- (2006) Clinical and research applications of
nation and resistance reduction. Expert Rev photodynamic therapy in dermatology: experi-
Anti-Infect Ther 18(1):5–15. https://doi. ence of the Scottish PDT Centre. Lasers Surg
org/10.1080/14787210.2020.1705155 Med 38(5):403–416. https://doi.org/10.
5. Dai T, Huang YY, Hamblin MR (2009) Pho- 1002/lsm.20369
todynamic therapy for localized infections-- 14. O’Neill J (2016) Tackling drug-resistant infec-
state of the art. Photodiagn Photodyn Ther tions globally: final report and recommenda-
6(3–4):170–188. https://doi.org/10.1016/j. tions. Government of the United Kingdom.
pdpdt.2009.10.008 15. Ozlem-Caliskan S, Ilikci-Sagkan R, Karakas H,
6. Fu XJ, Fang Y, Yao M (2013) Antimicrobial Sever S, Yildirim C, Balikci M et al (2022)
photodynamic therapy for methicillin-resistant Efficacy of malachite green mediated photody-
Staphylococcus aureus infection. Biomed Res namic therapy on treatment of cutaneous
Int 2013:159157. https://doi.org/10.1155/ Leishmaniasis: in vitro study. Photodiagn
2013/159157 Photodyn Ther 40:103111. https://doi.org/
7. Gajdacs M, Abrok M, Lazar A, Burian K 10.1016/j.pdpdt.2022.103111
(2020) Increasing relevance of gram-positive 16. Raetz C, Guan Z, Ingram B, Six D, Song F,
cocci in urinary tract infections: a 10-year anal- Wang X et al (2009) Discovery of new biosyn-
ysis of their prevalence and resistance trends. thetic pathways: the lipid a story. J Lipid Res 50
Sci Rep 10(1):17658. https://doi.org/10. (Suppl):S103–S108. https://doi.org/10.
1038/s41598-020-74834-y 1194/jlr.R800060-JLR200. R800060-
8. Koeth LM, King A, Knight H, May J, Miller JLR200 [pii]
LA, Phillips I et al (2000) Comparison of 17. Stevens DL, Bisno AL, Chambers HF, Everett
cation-adjusted Mueller–Hinton broth with ED, Dellinger P, Goldstein EJC et al (2005)
Iso-Sensitest broth for the NCCLS broth Practice guidelines for the diagnosis and Man-
microdilution method. J Antimicrob Che- agement of Skin and Soft-Tissue Infections.
mother 46(3):369–376. https://doi.org/10. Clin Infect Dis 41(10):1373–1406. https://
1093/jac/46.3.369 doi.org/10.1086/497143
9. Kofler B, Romani A, Pritz C, Steinbichler TB, 18. Galata V, Laczny CC, Backes C, Hemmrich-
Schartinger VH, Riechelmann H et al (2018) Stanisak G, Schmolke S, Franke A, Meese E,
Photodynamic effect of methylene blue and Herrmann M, von Müller L, Plum A, Müller R,
low level laser radiation in head and neck squa- Stähler C, Posch AE, Keller A et al (2019)
mous cell carcinoma cell lines. Int J Mol Sci Integrating culture-based antibiotic resistance
1 9 ( 4 ) . h t t p s : // d o i . o r g / 1 0 . 3 3 9 0 / profiles with whole-genome sequencing data
ijms19041107 for 11,087 clinical isolates. Genom Proteom
10. Laxminarayan R, Duse A, Wattal C, Zaidi AK, Bioinform 17(2). https://doi.org/10.1016/j.
Wertheim HF, Sumpradit N et al (2013) Anti- gpb.2018.11.002
biotic resistance-the need for global solutions. 19. Varzandeh M, Mohammadinejad R,
Lancet Infect Dis 13(12):1057–1098. https:// Esmaeilzadeh-Salestani K, Dehshahri A,
doi.org/10.1016/s1473-3099(13)70318-9 Zarrabi A, Aghaei-Afshar A (2021) Photody-
11. Mantravadi PK, Kalesh KA, Dobson RCJ, namic therapy for leishmaniasis: recent
Hudson AO, Parthasarathy A (2019) The advances and future trends. Photodiagn Photo-
quest for novel antimicrobial compounds: dyn Ther 36:102609. https://doi.org/10.
emerging trends in research, development, 1016/j.pdpdt.2021.102609
and technologies. Antibiotics (Basel) 8(1). 20. Vasala A, Hytonen VP, Laitinen OH (2020)
h t t p s : // d o i . o r g / 1 0 . 3 3 9 0 / Modern tools for rapid diagnostics of antimi-
antibiotics8010008 crobial resistance. Front Cell Infect Microbiol
12. Moseley H, Allen JW, Ibbotson S, Lesar A, 10:308. https://doi.org/10.3389/fcimb.
McNeill A, Camacho-Lopez MA et al (2006) 2020.00308
Ambulatory photodynamic therapy: a new con- 21. Vital-Fujii DG, Baptista MS (2021) Progress in
cept in delivering photodynamic therapy. Br J the photodynamic therapy treatment of Leish-
Dermatol 154(4):747–750. https://doi.org/ maniasis. Braz J Med Biol Res 54(12):e11570.
10.1111/j.1365-2133.2006.07145.x h t t p s : // d o i . o r g / 1 0 . 1 5 9 0 / 1 4 1 4 -
13. Moseley H, Ibbotson S, Woods J, 431X2021e11570
Brancaleon L, Lesar A, Goodman C et al
Chapter 7
Using Hollow Fiber to Model Treatment

of Antimicrobial-Resistant Organisms
Abstract
The use of animal models is still widespread in science but there is a movement away from this manner of
experimentation. One option approved by the FDA for human-like studies is the hollow fiber bioreactor
(HFS). HFSs are highly controllable, self-contained systems that allow for the modeling of individual
tissues and disease phenotypes. Oxygen, drug concentration & half-life, and immune cell invasion are all
scalable to human and veterinary conditions using a HFS. There are drawbacks to the systems including cost
and contamination so the use of these systems must be carefully managed.
With these limitations in mind, the scope of the technology is great. Antimicrobial susceptibility testing
(AST) is possible with greater accuracy and clinical validity than classical in vitro techniques making minimal
inhibitory concentration (MIC) data generated on the bench more translatable to the clinic.
In this chapter, we will outline the background of the HFS and some typical uses.
Key words Hollow fiber, Antibiotic resistance, Pharmacokinetic, Pharmacodynamics, Regimen

development
1 Introduction
Antimicrobial resistance (AMR) is a growing concern in the health-

care industry, with increasing numbers of bacterial and fungal
infections becoming resistant to commonly used antibiotics [1–
4]. In order to study the treatment of these resistant organisms,
researchers have developed in vitro models that closely mimic the
conditions found in the human body. One such model is the use of
hollow fiber bioreactors, which provide a controlled environment
for the growth and treatment of microorganisms [5–8].
1.1 Understanding A hollow fiber bioreactor is a system consisting of a bundle of small,

Hollow Fiber semi-permeable fibers encased in a larger tube (see Fig. 1). The
Bioreactors fibers are made of a material that allows for the transfer of nutrients
and oxygen to the microorganisms growing inside, while also
https://doi.org/10.1007/978-1-0716-3981-8_7,
57
B
1
D
E
F
A
Fig. 1 Schematic diagram of the hollow fiber system fully constructed. In our hollow fiber experiments, media
was supplied via input A under motive force from pump 1 (Masterflex C/L). The FiberCell Duet pump (2) then
draws the media along tube B to tube C and supplies it to the cartridge. The media passes through the
cartridge and flows down tube D to re-join the media in the central reservoir. At the same time pump 1 is also
drawing media out of the central reservoir (at the same rate as it is being introduced from tube A) along tube E
and into tube F and finally into the elimination reservoir where it is discarded
allowing for the removal of waste products [6, 7]. The fibers are
connected to a fluid source, such as a culture medium, which
provides a suitable environment for the growth of microorganisms.
1.2 Advantages of The principal advantage of hollow fiber bioreactors is that the
Using Hollow Fiber experimenter can control many of the critical parameters. Hollow
Bioreactors fiber bioreactors allow for precise control of the culture environ-
ment, including temperature, pH, drug delivery, and oxygen levels.
This means that the conditions can be reproduced accurately
throughout the course of the experiment and between
experiments.
No less important is the scalability of the design. Bioreactor
design allows for the cultivation of large quantities of microorgan-
isms, making it a suitable model for large-scale studies. 1011 cfu/
mL of some bacteria is possible in a volume of 20 mL of culture
medium, far outstripping conventional high-density
methodologies.
Using Hollow Fiber to Model Treatment of Antimicrobial-Resistant Organisms 59
Hollow fiber bioreactors have been approved by the FDA as a

means to mimic human pharmacology [9]. The bioreactor closely
resembles the conditions found in the human body, providing a
more accurate representation of the effects of antimicrobial
treatments.
All medical and research technologies should be easy to use and
the hollow fiber bioreactor is no different. The bioreactor is easy to
set up and use, making it accessible to researchers with varying
levels of experience.
1.3 Applications of Hollow fiber bioreactors have myriad uses and some examples are
Hollow Fiber detailed below.
Bioreactors in In a world where AMR is becoming increasingly important and
Antimicrobial research into it is picking up pace novel or updated methods of
Resistance Studies antibiotic susceptibility testing are crucial. Hollow fiber bioreactors
can be used to determine the susceptibility of microorganisms to
various antibiotics, providing valuable information for the selection
of appropriate treatments. In addition, they can be used to establish
treatment efficacy. Researchers can use the bioreactor to study the
efficacy of various antibiotics and combination therapies on resis-
tant microorganisms (see Chap. 5).
While considering AMR it is vital to understand how a patho-
gen resists the antibiotic in question. Understanding the mechan-
isms of resistance allows researchers to gain insights into potential
strategies to overcome pathogen resistance or develop new
therapies.
1.4 Limitations and Hollow fiber bioreactors are relatively complex. At their fullest
Considerations extent they can contain meters of tubing and several bottles of
media. This complexity of design means that understanding and
successfully running a hollow fiber bioreactor requires a certain
level of technical expertise to set up and use. It also means that
there is considerable opportunity for contamination if mitigation
strategies are not put into place before the experiment begins.
The expense of the hollow fiber can also be prohibitive. The
consumable element is the hollow fiber cartridge itself and these
can cost hundreds of pounds each. The other aspects of the set up
are more long-lasting but are still costly. This can be considered a
significant restrictive factor as unfortunately its accessibility will be
limited to some researchers.
While the FDA has approved the hollow fiber bioreactor for
in vivo-like studies it may not fully capture the complexities of the
human body limiting its scope as a research tool. This, in turn, leads
to limitations in the conclusions that can be drawn from any results
generated when compared with in vivo data and especially when
comparing with first-in-human trials.
1.5 Initial The use of hollow fiber bioreactors provides a valuable tool for
Conclusions researchers studying the treatment of antimicrobial-resistant organ-
isms. The controlled environment and scalability of the bioreactor
make it a useful model for studies on antibiotic susceptibility,
treatment efficacy, and mechanisms of resistance. However, it is
important to keep in mind the limitations and considerations
when interpreting the results of studies using this model.
In the below example of an AST testing method, the hollow
fiber assay was used to simulate in vivo conditions without the use
of human or animal test subjects. As a highly manageable test
system, it allows control of O2 concentration, drug half-life (T1/
2), maximal concentration of drug (Cmax), and time over which
maximal concentration of drug is reached (Tmax) as well as other
factors. This makes it ideal for testing novel antibiotics or antibiotic
combination therapies.
2 Materials
As an example, AST testing:

1. Mueller Hinton cation adjusted broth (Sigma-Aldrich, UK)
(see Note 1).
2. Resistant organism, in this case a KPC Klebsiella pneumonia
(ATCC BAA 1705) (see Notes 2 and 3).
3. Antimicrobials of choice, in this instance colistin and Zidovu-
dine (Sigma-Aldrich, UK) (see Note 4).
5. Sterile water.
6. Appropriate Hollow fiber cartridge (such as the C3008) (Fiber-
cell Systems, USA) (see Note 5).
7. Appropriate cartridge pump (such as the Fibercell Duet)
(Fibercell Systems, USA) (see Note 6).
8. Feed and waste bottle connectors with associated silicone tub-
ing and luer lock attachments.
9. Feed and waste bottles able to accept connectors (plastic,
500–1000 mL).
10. 0.22 μm filters.
11. As required, appropriate pump for motivating feed and waste
medium (such as a peristaltic; Watson Marlow 120 U/DM2,
Wolf Labs, UK).
12. As required, appropriate pump for drug delivery, such as a
syringe driver (Aladdin Programmable Syringe Pump AL
1000-220).
(a) To deliver antibiotics in a controlled manner over a period

of time a device such as a syringe driver is vital. If
modeling PK/PD curves, being able to closely emulate
human-like Cmax and Tmax values will mean more accurate
experiments with more robust data.
(b) If a single dose of drug is required and the time compo-
nent is immaterial then delivering a bolus of drug by hand
is entirely justified.
Aldrich, UK).
16. Microcentrifuge tubes (Eppendorf tubes® (Sigma-Aldrich)).
3 Methods
3.1 Hollow Fiber The overall shape of the rig used is influenced by the experiment
Bioreactor being undertaken. In the below example an AST test is being done
wherein components required are as in the material section above.
Briefly: C3008 cartridge, media “feed” bottle, empty “waste” bot-
tle, connectors, drug infusing pump (syringe driver), 3-way stop-
cock, silicone tubing.
1. Prepare Hollow fiber cartridges as per the manufacturer’s
instructions and attached to the overall rig, aseptically, as is
shown in Fig. 1 (see Note 7).
2. The attachment is best undertaken in a sterile cabinet with all
tubing ends that are going to meet being sprayed liberally with
70% ethanol.
3. Introduce bacteria into the cartridge via the sampling ports,
indicated in Fig. 1, and allowed to settle for 1 h (see Note 8).
4. After this time, activate the Duet pump to ensure flow through
the cartridge is smooth and to eliminate bubbles.
5. Antibiotics are introduced into the system at the red X (see
Fig. 1) using a 3-way stopcock and a syringe driver (Aladdin
Programmable Syringe Pump AL 1000-220).
6. Set the pump to deliver drug such that the final concentration
at the end of dosing was equal to the Cmax of the drug over the
time Tmax.
7. Stop pump 1 for the duration of drug delivery at a predeter-
mined time (Tmax).
Calculate the specific Cmax and Tmax parameters for each
drug and drug combination used based on the drug’s half-life
in vivo, the flow rate of the system, and other factors [7].
8. Once the drug infusion is finished the system automatically

restarts; pumps supplying both drug-free media and eliminat-
ing from the reservoir functioning at the required rate until the
next drug infusion.
9. Design the dosing schedules to mimic the exposure to clinical
doses of each antibiotic.
3.2 Sampling 1. Sample the bacteria at regular intervals throughout the experi-
mental period (7 days) to address the pharmacology of the
studied drug.
2. Select the sampling schedule that best assesses changes in bac-
terial numbers in response to the administered drugs and mea-
sure the associated drug concentration.
3. In order to sample the bacteria from the cartridge Switch flow
pump (FiberCell Duet (2)) is switched off temporarily, 5 mL of
medium is removed as described by the manufacturer
(FiberCell).
4. Remove the filled syringe from the assembly. Spray the Luer
lock with ethanol and add a new sterile syringe to the vacated
Luer lock.
5. Restart the flow pump and continue the experiment until the
next sampling interval.
3.3 Sample Depending on the initial conditions for the experiment and the
Examination reason for the experiment, the samples produced and the generated
data will vary.
In the exemplar above an AST test was attempted against a
resistant organism. As this is an in vivo-like system the results
generated here might be different to those found in a static assay
like a broth microdilution, disc diffusion, or static time kill assay.
Examining these samples requires understanding of the patho-
gen used and an understanding of common culture techniques.
Where sampling has taken place throughout a dosing regimen
bacteria might be in different growth phases, stress responses, or
stages of death. These are valuable information and can be collected
in a variety of ways; growth kinetic experiments, RNA sequencing,
and viability assays to name a few.
Generally the most compelling piece of information, and the
most common, is the simplest; are the cells alive or not? This data
can, again, be extracted in a variety of ways such as viability plating,
colorimetric assays, and DNA sequencing.
4 Notes

Like any other benchtop experiment be mindful of the
specific requirements of your organism of interest. If using a
fastidious organism that requires blood, include that in the
media as you would using a static rig. If using a mycobacterium
use Middlebrook or similar.
2. Because this is a dynamic system some requirements can be
loosened. For example, the inclusion of Tween 80 (or similar
detergent) to reduce mycobacterial clumping can be dispensed
with as the bacteria are under mild sheer stress that will dis-
courage clumping.
3. Some bacteria are prone to developing biofilms, especially in
drug-induced, stressful conditions. In the hollow fiber, these
circumstances need to be taken into account and the activity of
any antibiotics used (for example) need to be controlled for in
the face of high cell densities forming robust biofilms.
4. The antimicrobials used can be ones to which the organism of
choice is resistant to depending on the results you are attempt-
ing to attain. “Rescuing” an antibiotic is possible with the right
combination of synergistic therapies. Alternatively, the antimi-
crobials used can be ones to which your organism of choice is
susceptible to if the aim of your experiment is to, for example,
lower the concentration of a less efficacious antibiotic that is
frequently prescribed at high doses, such as nitrofurantoin
(EUCAST MIC 64 mg/L).
5. The cartridge that the individual researcher wishes to use
should be dictated by the pore size of the particular fibers.
The pore size is what will decide the passage of given substances
(proteins, lipids, carbohydrates, vitamins) into and out of the
extracellular space where the bacteria are. This will, in turn,
dictate what the researcher collects in their waste bottle if
attempting a protein pull-down or similar.
6. In this context appropriate means suitable for your system.
Usually, the manufacturer chosen will supply both cartridges
and the correct pumps simultaneously.
7. Different manufacturers will instruct different methods for
preparing their cartridges. Generally, one or more hours of
“cleaning” with sterile PBS is required where the PBS in
pumped cyclically through the cartridge followed by 1+ hours
of “conditioning” with the media of choice pumped through in
a similar manner.
8. 1 h is manufacturer-specific guidance but is a roughly accurate

figure. The hour allows the bacteria to settle into place in the
fibers so when the cartridge pump is activated the bacteria are
not all forced under sheen stress to one end of the cartridge;
they remain evenly distributed.
References
1. Gajdacs M, Abrok M, Lazar A, Burian K 6. Kloprogge F, Hammond R, Copas A, Gillespie
(2020) Increasing relevance of gram-positive SH, Della PO (2019) Can phenotypic data
cocci in urinary tract infections: a 10-year anal- complement our understanding of antimyco-
ysis of their prevalence and resistance trends. bacterial effects for drug combinations? J Anti-
Sci Rep 10(1):17658. https://doi.org/10. microb Chemother 2019:3530. https://doi.
1038/s41598-020-74834-y org/10.1093/jac/dkz369
2. O’Neill J (2016) Tackling drug-resistant infec- 7. Kloprogge F, Hammond R, Kipper K, Gillespie
tions globally: final report and recommenda- SH, Della PO (2019) Mimicking in-vivo expo-
tions. Government of the United Kingdom sures to drug combinations in-vitro: anti-
3. Galata V, Laczny CC, Backes C, Hemmrich- tuberculosis drugs in lung lesions and the hol-
Stanisak G, Schmolke S, Franke A, Meese E, low fiber model of infection. Sci Rep 9(1):
Herrmann M, von Müller L, Plum A, Müller R 13228. https://doi.org/10.1038/s41598-
(2019) Integrating culture-based antibiotic 019-49556-5
resistance profiles with whole-genome 8. Storm MP, Sorrell I, Shipley R, Regan S,
sequencing data for 11,087 clinical isolates. Luetchford KA, Sathish J et al (2016) Hollow
Genom Proteom Bioinform 17(2):169–182. fiber bioreactors for in vivo-like mammalian
https://doi.org/10.1016/j.gpb.2018.11.002 tissue culture. JoVE 111:e53431
4. Vasala A, Hytonen VP, Laitinen OH (2020) 9. Romero K, Clay R, Hanna D (2015) Strategic
Modern tools for rapid diagnostics of antimi- regulatory evaluation and endorsement of the
crobial resistance. Front Cell Infect Microbiol hollow fiber tuberculosis system as a novel
10:308. https://doi.org/10.3389/fcimb. drug development tool. Clin Infect Dis 61
2020.00308 (suppl_1):S5–S9
5. Deshpande D, Srivastava S, Nuermberger E, 10. Koeth LM, King A, Knight H, May J, Miller
Pasipanodya JG, Swaminathan S, Gumbo T LA, Phillips I et al (2000) Comparison of
(2016) Concentration-dependent synergy and cation-adjusted Mueller–Hinton broth with
antagonism of linezolid and moxifloxacin in Iso-Sensitest broth for the NCCLS broth
the treatment of childhood tuberculosis: the microdilution method. J Antimicrob Che-
dynamic duo. Clin Infect Dis 32016:S88–S94 mother 46(3):369–376. https://doi.org/10.
1093/jac/46.3.369
Chapter 8
A Microtiter Plate Assay at Acidic pH to Identify Potentiators

that Enhance Pyrazinamide Activity Against Mycobacterium
tuberculosis
Christopher William Moon, Eleanor Porges, Stephen Charles Taylor,
and Joanna Bacon
Abstract
Pyrazinamide (PZA) is a key component of chemotherapy for the treatment of drug-susceptible tuberculo-
sis (TB) and is likely to continue to be included in new drug combinations. Potentiation of PZA could be
used to reduce the emergence of resistance, shorten treatment times, and lead to a reduction in the quantity
of PZA consumed by patients, thereby reducing the toxic effects. Acidified medium is required for the
activity of PZA against Mycobacterium tuberculosis. In vitro assessments of pyrazinamide activity are often
avoided because of the lack of standardization, which has led to a lack of effective in vitro tools for assessing
and/or enhancing PZA activity.
We have developed and optimized a novel, robust, and reproducible, microtiter plate assay, that centers
around acidity levels that are low enough for PZA activity. The assay can be applied to the evaluation of
novel compounds for the identification of potentiators that enhance PZA activity. In this assay, potentiation
of PZA is demonstrated to be statistically significant with the addition of rifampicin (RIF), which can,
therefore, be used as a positive control. Conversely, norfloxacin demonstrates no potentiating activity with
PZA and can be used as a negative control. The method, and the associated considerations, described here,
can be adapted in the search for potentiators of other antimicrobials.
Key words Mycobacterium tuberculosis, Pyrazinamide, Potentiation, Microtiter plate, Acidic pH,
Chequerboard, Rifampicin, Norfloxacin, Highest single agent
1 Introduction
An important aim for improving tuberculosis (TB) treatment is to

shorten the period of antibiotic therapy without increasing relapse
rates or encouraging the development of antibiotic-resistant strains.
Pyrazinamide (PZA) is a key component of front-line chemother-
apy against Mycobacterium tuberculosis. It plays an essential role in
the shortened 6-month treatment course [1, 2] due to its ability to
act upon the non-replicating/slow-growing or antibiotic-resistant
https://doi.org/10.1007/978-1-0716-3981-8_8,
65
66 Christopher William Moon et al.
organisms that emerge following treatment with the other front-

line drugs, isoniazid (INH) and rifampicin (RIF). Although new
combinations of antibiotics with novel modes of action are being
evaluated, for example within the SimpliciTB clinical trial, which is
evaluating the BPaMZ regimen, consisting of bedaquiline (B),
pretomanid (Pa), moxifloxacin (M), and pyrazinamide (Z), optimal
dosing and treatment duration can be investigated further for
existing antibiotics. Potentiation of frontline drugs is an approach
that can extend the life of existing antibiotics, such as PZA.
In the context of the method described here, a potentiator is
defined as a substance that does not necessarily have bactericidal
activity on its own, but when used at a low concentration in
combination with an antibiotic, will enhance its activity. Some
progress has been made in the identification of potentiators for
PZA. Weak acids and energy inhibitors work synergistically with
PZA [3] and mutations in energy production and ion homeostasis
pathways can enhance PZA activity [4]. Other conditions shown to
promote the activity of PZA are anaerobiosis [5] co-incubation
with electrochemical un-couplers [6] or efflux inhibitors
[7]. There are likely to be multiple different mechanisms of poten-
tiation and compounds that enhance PZA which could shed light
on the mechanism(s) of PZA activity. Certainly, it seems that some
enhancers (weak acids and fatty acids) are likely to be perturbing the
membrane affecting membrane potential, allowing higher cytoplas-
mic concentrations of pyrazinoic acid (POA), the activated product
of PZA. Existing antibiotics, RIF [8], valinomycin [3], and beda-
quiline [9], have also been shown to potentiate PZA. Despite
having bactericidal activity at high concentrations, RIF serves as a
reproducible, positive control, at low concentrations, in our assay,
to find further compounds that enhance PZA activity.
PZA targets trans-translation [10] (a process found in all bac-
teria, which contributes to the release of ribosomes that are stalled),
which is required by all bacteria to respond to stress [11]. Based on
this mode of action for PZA, it follows that actively growing
organisms could be targeted by PZA, particularly if PZA activity
was enhanced by potentiators that create stress/reduce metabolism
[12, 13]. Growth inhibition by bacteriostatic antibiotics is asso-
ciated with suppressed cellular respiration, whereas cell death from
exposure to bactericidal antibiotics causes accelerated respiration
[9]. A reduced cellular respiration induced by a potentiator would
allow PZA to target non-replicating bacteria.
PZA is a pro-drug that is hydrolyzed to its active form of
pyrazinoic acid (POA) by M. tuberculosis pyrazinamdidase/nicoti-
namidase (PncA) in the bacterial cytoplasm, which is excreted from
the bacteria by an unidentified POA efflux mechanism. POA dif-
fuses passively back in through the cell wall as the protonated
molecule (HPOA) when the pH of the bacterial environment is
acidic. Investigations as to how PZA activity can be maximized
Enhancing Pyrazinamide Activity Against Mycobacterium tuberculosis 67
upon have often been avoided because of the complexities and lack
of standardization of in vitro assay conditions, which has led to a
lack of effective in vitro tools for assessing/enhancing PZA activity.
The optimal pH for the activity of PZA in vitro has previously been
shown to be pH 5.6 [14]. Although acidification of the growth
medium is considered to be important for PZA activity, previous
studies have demonstrated that in certain situations, PZA is active
at a neutral pH [15]. Further to this, metabolic potentiation of
PZA activity has been discovered independently of environmental
pH [16]. However, a key component of the assay, described here, is
the growth of M. tuberculosis in a defined, suitably buffered
medium, at pH 5.5, followed by exposure to drug, under the
same conditions in the microtiter plate. This is designed to reflect
the acidic conditions that the organism will be encountering intra-
cellularly in the macrophage and in the granuloma, during disease;
inflammatory lung lesions have shown to be pH 5.5—pH 6.0
[17]. Variations in in vitro parameters such as inoculum size and
medium components can also alter the activity of PZA [7] and
these have been optimized in our method. In this assay, potentia-
tion of PZA is demonstrated to be statistically significant with the
addition of rifampicin (RIF), which is used as a positive control.
Conversely, norfloxacin (NOR) demonstrates no potentiating
activity with PZA and serves as a negative control.
In summary, robust and reproducible in vitro assay conditions
are required for the discovery of potentiators that can boost PZA
activity. We have successfully addressed these issues and have devel-
oped an in vitro microtiter plate chequerboard assay that can be
used to find compounds that will potentiate PZA activity against
M. tuberculosis. The method, and the associated considerations,
described here, can be adapted in the search for potentiators of
other antimicrobials.
2 Materials
2.1 Growth Medium 1. Middlebrook 7H9 medium supplemented with oleic acid, dex-
trose, and catalase (OADC) enrichment (referred to as “Mid-
dlebrook 7H9”). The ingredients required for 1 L of
Middlebrook 7H9: Middlebrook 7H9 broth powder (4.7 g),
glycerol (5 mL), Tween-80 (2 mL), and oleic albumin dextrose
catalase (OADC; (Beckman Dickinson, UK—212240))
(100 mL). Add the Middlebrook 7H9 broth powder to
500 mL of distilled water, and then add the glycerol and
Tween-80. Stir the solution to dissolve all the ingredients and
make the volume up to 900 mL using distilled water. Autoclave
this medium at 121 °C, 15 psi for 15 min. Once cooled to
room temperature, add 100 mL of OADC. Store at room
temperature. For best results, use freshly made medium (see
Note 1).
2. Middlebrook 7H9 medium supplemented with OADC enrich-

ment, adjusted to pH 5.5 (referred to as “acidified Middleb-
rook 7H9”). The ingredients required for 1 L of acidified
Middlebrook 7H9: Middlebrook 7H9 broth powder (4.7 g),
glycerol (5 mL), Tween-80 (2 mL), sodium phosphate dibasic
(0.037 g), potassium phosphate monobasic (1.77 g), and oleic
albumin dextrose catalase (OADC; Beckman Dickinson; UK—
212240) (100 mL). Add the Middlebrook 7H9 broth powder
to 500 mL of distilled water, and then add the glycerol, Tween-
80, sodium phosphate dibasic, and potassium phosphate
monobasic. Stir the solution to dissolve all the ingredients
and make the volume up to 900 mL using distilled water.
Autoclave this medium at 121 °C, 15 psi for 15 min. Once
cooled to room temperature, add 100 mL of OADC. Adjust
the pH to 5.5 using hydrochloric acid (3 M HCl). Store at
room temperature. For best results, use freshly made medium
(see Note 1).
2.2 Viability 1. Resazurin (0.04% w/v): the ingredients required for 50 mL of

Detection Reagents resazurin are sodium phosphate dibasic (1.82 g), potassium
phosphate monobasic (0.06 g), and resazurin sodium salt
(0.02 g). Add all ingredients to a sterile empty tube, add
distilled water to a final volume 50 mL and stir to dissolve.
Addition of the described phosphate buffers will result in a
pH 8.2 solution (see Note 2). Filter sterilize the solution
using a 0.2 μm filter. Store at room temperature, wrap in foil
to protect the solution from light and use within 6 months.
2.3 Additional 1. Formaldehyde (40% w/v).

Reagents 2. Phosphate buffered saline (PBS), pH 7.4.
3. Dimethyl sulfoxide (DMSO).
2.4 Equipment 1. Erlenmeyer cell culture vented flasks, 125 mL, sterile (Corning,
USA—431143).
2. Shaking incubator, 37 °C, 200 rpm.
3. Screw cap, O-ring, 2 mL tubes.
4. UV/vis spectrophotometer.
5. Cuvettes.
6. Reagent reservoirs.
7. Nunc 96-well, black plates (Thermo Fisher Scientific, UK—
237105) (see Note 3).
8. SealPlate film (Excel Scientific).
9. MicroAmp Optical Adhesive film (Thermo Fisher Scientific,
UK—4311971).
10. Plate scraper (Applied Biosystems).

11. GloMax Discover Microplate Reader (Promega UK, Ltd) (see
Note 4).
12. D300e Digital Dispenser (Tecan, Switzerland) (see Note 5).
3 Methods
3.1 Preparation of M. 1. Derive mycobacterial stocks, used for the inoculations, from
tuberculosis Inoculum the growth of M. tuberculosis H37Rv, in a defined and standar-
dized way. The preferred method is by continuous culture in a
chemostat as described by Bacon et al. in 2004 [18] and Bacon
and Hatch in 2009 [19] (see Note 6).
2. Store culture stocks as 1 mL volumes at -70 °C, until use (see
Note 7).
3.2 Growth of M. 1. Add 500 μL of the M. tuberculosis stock, as described in Sub-

tuberculosis in heading 3.1, to three Erlenmeyer flasks (one flask for each
Middlebrook 7H9 for biological replicate), each containing 25 mL of Middlebrook
5 Days 7H9.
2. Incubate the flasks for 5 days, at 37 °C, shaking orbitally at
200 rpm (see Note 8) (see Fig. 1).
3.3 Growth of M. 1. Sample the cultures, described in Subheading 3.2, for optical
tuberculosis in density measurements. Fix the cells by the addition of 4%
Acidified Middlebrook formaldehyde (final concentration v/v) to a tenfold dilution
7H9 for 7 Days of the bacterial culture, in PBS pH 7.4, in a screw cap tube with
an O-ring. Invert the tube and leave for 30 min, to fix the
bacteria (see Note 9).
2. Place the formaldehyde-fixed mycobacterial suspension into a
plastic cuvette.
3. Read the optical density at 540 nm (OD540nm) against water as
the blank.
Fig. 1 Schematic representation of assay workflow

4. If each of the three cultures meets the growth requirements of a

minimum OD540nm of 2.0 (see Note 10) sample 1 mL of
culture from each of the flasks and transfer into three sterile
Erlenmeyer flasks (one flask for each biological replicate) each
containing 25 mL of acidified Middlebrook 7H9 (see Note 1).
5. Incubate the flasks for 7 days, at 37 °C, shaking orbitally at
200 rpm (see Fig. 1).
3.4 Growth of M. 1. Prepare six 96-well microtiter plates; three biological repeat
tuberculosis in plates for the positive control drug (RIF) and three biological
Acidified Middlebrook repeat plates for the negative control drug (NOR).
7H9 for 10 Days 2. Add 100 μL of acidified Middlebrook 7H9 to all wells of rows
(Microtiter Plates) in A-G and wells H11 and H12 of each 96-well plate.
the Presence of 3. Add 190.44 μL of acidified Middlebrook 7H9 to wells
Antibiotic H1-H10 of each 96-well plate.
Combinations
4. Add 9.56 μL PZA to wells H1-H10 in all plates from a stock
3.4.1 Preparation of PZA- concentration of 20 mg/mL prepared in acidified Middleb-
RIF and PZA-NOR rook 7H9 to achieve a concentration of 1600 μg/mL in a final
Combinations in 96-Well assay volume of 110 μL per well.
Plates (Chequerboard) 5. Perform a twofold serial dilution of PZA across each plate, by
transferring and mixing 100 μL from wells H1-H10 through to
wells B1-10 in all plates, to achieve concentrations of
1600–25 μg/mL in rows B-H and columns 1–10 (see Fig. 2).
6. To three of the 96-well plates create a RIF twofold dilution
series from a stock concentration of 0.5 mg/mL prepared in
dimethyl sulfoxide (DMSO) to achieve concentrations of
50–0.195 ng/mL in rows A-H and columns 2–10 in a micro-
titer plate (see Fig. 2a).
7. To the remaining three 96-well plates create a norfloxacin
twofold dilution series from a stock concentration of 2 mg/
mL prepared in DMSO to achieve concentrations of
32–0.125 μg/mL in rows A-H and columns 2–10 in a microti-
ter plate (see Fig. 2b) (see Note 12).
(a) Addition of PZA and RIF, or NOR on opposing axis in the
microtiter plate creates a dose matrix of PZA and RIF or
NOR with eight concentrations of PZA and ten concen-
trations of RIF or NOR. No antibiotics are added to well
A1.
8. To columns 1–10, rows A-H add sufficient DMSO to raise and
normalize the DMSO concentration of each well to 3% (v/v)
(see Note 13).
9. To column 11, rows A-H add 3.4 μL DMSO to achieve a final
concentration of 3% (v/v) (see Note 14).
10. No antibiotic or DMSO additions are made to column
12 (medium-only control) (see Note 15).
Fig. 2 Example plate layout for creating a 10 × 8 dose matrix of rifampicin and
pyrazinamide (a), and norfloxacin and pyrazinamide (b). Concentrations of
rifampicin are 50, 25, 12.5, 6.25, 3.13, 1.56, 0.781, 0.391, and 0.195 ng/mL,
concentrations of norfloxacin are 32, 16, 8, 4, 2, 1, 0.5, 0.25, and 0.125 μg/mL,
concentrations of pyrazinamide are 1600, 800, 400, 200, 100, 50, and 25 μg/mL
3.4.2 Inoculation of 1. Sample the cultures that have been growing for 7 days in
Antibiotic Dose Matrix acidified medium (described in Subheading 3.3) for optical
Microtiter Plates with M. density measurements. Fix the cells by the addition of 4%
tuberculosis formaldehyde (final concentration v/v) to a tenfold dilution
of the bacterial culture, in PBS pH 7.4, in a screw cap tube with
an O-ring. Invert the tube and leave for 30 min, to fix the
bacteria.
2. Place cell suspension into a plastic cuvette.
3. Read the optical density at 540 nm (OD540nm) against water.
4. If cultures meet growth requirements of a minimum OD540nm
of 1.0 (see Note 16), dilute the three cultures to an OD540nm of
0.55, using acidified Middlebrook 7H9 (see Notes 17 and 18).
5. Add 10 μL of the prepared M. tuberculosis culture (at 0.55
OD540nm) to columns 1–11 of all six 96-well plates (see
Notes 15 and 19).
6. Add 10 μL of acidified Middlebrook 7H9 to column 12 of the
six 96-well plates (see Note 15).
7. Apply a plate seal (SealPlate film, Excel Scientific) to all plates
(see Note 20).
8. Incubate the six 96-well plates in a contained shaking incubator
for 10 days, at 37 °C, orbitally shaking at 200 rpm (see Note 8)
(see Fig. 1).
3.5 Endpoint 1. Remove the plate seals from the microtiter plates and discard.
Detection/Final 2. Add 10 μL of resazurin (that has been adjusted to pH 8.2 as
Viability Assessment described in Subheading 2.2) to all wells of the microtiter
plates and incubate at room temperature, for 6 h (see Note 21).
3. Add a MicroAmp Optical Adhesive film (Thermo Fisher Scien-
tific) plate seal, to each plate, ensuring a tight, complete seal (see
Notes 22 and 23).
4. Image the microtiter plates using a microplate reader with fluo-
rescent detection (Ex: 525 nm Em: 580–640 nm) (see Note 24).
3.6 Data Analysis 1. Biological repeat averaging—Calculate the average of the resor-
ufin fluorescence data collected from each biological replicate
in each well of the microtiter plates.
2. Baseline correction—Calculate the average of column
12 (medium-only control). Subtract this value from all other
values (columns 1–11), to generate baseline-corrected data.
3. Normalization—Calculate the average of column
11 (M. tuberculosis with 3% DMSO addition). Divide all
baseline-corrected values by this average value for column
11, then multiply by 100 (columns 1–10), to generate normal-
ized data compared to M. tuberculosis growth with no antibi-
otic addition (see Note 25).
Fig. 3 Example baseline corrected and normalized pyrazinamide and rifampicin dose matrix data (a) and
pyrazinamide and norfloxacin dose matrix data (c). Data values are colored so that growth of 100% that of the
3% DMSO control M. tuberculosis growth (column 11 average) is green and 0% is red. Example of highest
single agent determination of antibiotic combination synergy of pyrazinamide and rifampicin (b) and pyrazi-
namide and norfloxacin (d). Synergistic effects (higher positive values) are colored blue and antagonistic
effects (higher negative values) are colored red
4. Data visualization—Visualize all normalized values in a tabular

format using GraphPad Prism (see Fig. 3a).
5. Synergy assessment using highest single agent method—Calculate
if antibiotic pairings are synergistic, additive, or antagonistic by
inputting values for each paring into the highest single agent
(HSA) formula (see Fig. 3b) (see Eq. 1) (see Note 26). Calculate
the HSA average from the individual HSA scores at each com-
bination of antibiotic concentrations.
Max E A, E B
HSA = ð1Þ
E AB
4 Notes
1. Reproducibility of mycobacterial growth can be improved by

storing the medium in the fridge and adding the OADC on the
morning the medium is to be used.
2. A solution of resazurin that has been adjusted to pH 8.2 is
required to neutralize the effects of the acidity of the Middleb-
rook 7H9 medium that has been adjusted to pH 5.5. This
neutralization allows for the generation of the fluorescent sig-
nal through the bacterial metabolism of resazurin to resorufin.
3. Black walled microtiter plates are required for optimal detec-
tion of fluorescent signals without spillover of signal between
wells.
4. A plate reader capable of fluorescent measurement of resazurin
(excitation peak at 571 nm and emission peak at 584 nm) is
required. The Promega GloMax Discover is an example of such
a plate reader.
5. A digital dispenser capable of pipetting nL volumes of liquid is
required and the Tecan D300e is an example of such a
dispenser.
6. M. tuberculosis stocks do not have to be generated from con-
tinuous culture in chemostats, but M. tuberculosis stocks should
be generated in a standardized way to ensure reproducible
results.
7. No more than one freeze-thaw cycle per M. tuberculosis H37Rv
stock vial is recommended.
8. Erlenmeyer flasks containing M. tuberculosis culture must be
shaken and not incubated statically, to achieve the required
growth parameters needed for growth in Middlebrook 7H9.
9. A final concentration of 4% (v/v) formaldehyde is added to
M. tuberculosis cultures and left for 30 min to conform with
laboratory fixation and inactivation requirements.
10. An OD540nm of 2.0 or above is needed to ensure reproducible
growth of M. tuberculosis in Middlebrook 7H9.
11. The acidity of acidified Middlebrook 7H9 should always be
checked on the day that it is being used for propagating
M. tuberculosis. If required, adjustment of the pH to pH 5.5
must be performed through addition of HCl or NaOH.
12. A Tecan D300e digital dispenser is used to create a dilution
series of RIF and NOR across the microtiter plates.
13. A Tecan D300e digital dispenser is used to add a consistent
concentration of DMSO in each well to a value of 3% (v/v).
The concentration of DMSO per well is ensured not to exceed
3% as DMSO concentrations greater than 3% decrease the
M. tuberculosis viability.
14. A DMSO-only control is required on each microtiter plate to

determine viability of M. tuberculosis in the absence of antibio-
tics, this allows for normalization of antibiotic-induced inhibi-
tion of M. tuberculosis growth during data analysis.
15. A medium-only (no M. tuberculosis addition) control is
required on each microtiter plate to allow for baseline correc-
tion during data analyses.
16. An OD540nm of 1.0 or above is needed to ensure reproducible
growth of M. tuberculosis in acidified Middlebrook 7H9 in the
96-well microtiter plates.
17. When 10 μL of this culture is added to 100 μL medium within
the 96-well plates (see Subheading 3.4), a final OD540nm of
0.05 is achieved with a final assay volume of 110 μL.
18. A total of 0.88 mL of diluted cells are required to inoculate
each microtiter plate replicate and an excess (10 mL) is
prepared to ensure that sufficient volume remains if a reagent
reservoir is used to aid microtiter plate inoculation.
19. A reagent reservoir and multichannel pipette are recommended
to add bacterial inoculum to 96-well plates.
20. Plate seals are added to the microtiter plates to prevent evapo-
ration and changes in acidity in the wells during the 10-day
incubation period.
21. A 6-h incubation time is required to allow sufficient time for a
measurable fluorescent signal to be achieved from the reduc-
tion of resazurin to resorufin.
22. MicroAmp Optical Adhesive film seals are added to all microti-
ter plates to conform with laboratory containment require-
ments regarding the movement of microtiter plates around
the laboratory.
23. Ensure that microtiter plates are cooled to room temperature
before adding MicroAmp Optical Adhesive film seals so that
condensation does not form on the underside of the seal lid,
preventing accurate fluorescence measurements.
24. The resazurin excitation peak is at 571 nm and the emission
peak is at 584 nm. To read the fluorescent emission spectra
using a Promega GloMax Discover, excitation wavelength of
525 nm and emission wavelength range of 580–640 nm
are used.
25. For conditions in which M. tuberculosis growth was inhibited
by the presence of an antibiotic, or antibiotic pairing, low
percentage growth values are expected.
26. The HSA model assumes that if a combination effect (EAB)
exceeds those of its constituents’ effects (EA and EB), there
must be some interaction. The model values are calculated at
each dose matrix point using Eq. (1). HSA values greater than
10 represent synergistic combinations, values between 10 and
-10 represent additive combinations and values less than -10
represent antagonistic combinations. HSA is only defined on a
dose-by-dose basis with no model fit, it is sensitive to the dose
range selected.
Acknowledgments
This work was funded by the Department of Health, and Public

Health England Pipeline fund, UK. The views expressed in this
chapter are those of the authors and not necessarily those of the
Department of Health or Public Health England. The authors also
express their gratitude to Michelle Finney for the provision of
equipment required for the assay.
References
1. British T (1976) Short-course chemotherapy 9. Ibrahim M et al (2007) Synergistic activity of
in pulmonary tuberculosis. Lancet (London, R207910 combined with pyrazinamide against
England) 2:1102–1104 murine tuberculosis. Antimicrob Agents Che-
2. Mitchison DA (1985) The action of antituber- mother 51:1011–1015
culosis drugs in short-course chemotherapy. 10. Shi W et al (2011) Pyrazinamide inhibits trans-
Tubercle 66:219. https://doi.org/10.1016/ translation in Mycobacterium tuberculosis. Sci-
0041-3879(85)90040-6 ence 333:1630–1632
3. Wade MM, Zhang Y (2006) Effects of weak 11. Keiler KC (2008) Biology of trans-translation.
acids, UV and proton motive force inhibitors Ann Rev Microbiol 62:133–151
on pyrazinamide activity against Mycobacte- 12. Rodionova IA et al (2014) Metabolic and bac-
rium tuberculosis in vitro. J Antimicrob Che- tericidal effects of targeted suppression of
mother 58:936–941 NadD and NadE enzymes in mycobacteria.
4. Zhang Y, Shi W, Zhang W, Mitchison D MBio 5:10-1128
(2014) Mechanisms of pyrazinamide action 13. Cook GM, Greening C, Hards K, Berney M
and resistance. Microbiol Spectr 2:10–1128 (2014) Energetics of pathogenic bacteria and
5. Wade MM, Zhang Y (2004) Anaerobic incuba- opportunities for drug development. Adv
tion conditions enhance pyrazinamide activity Microb Physiol 65:1–62
against Mycobacterium tuberculosis. J Med 14. Gumbo T, Dona CSWS, Meek C, Leff R
Microbiol 53:769–773 (2009) Pharmacokinetics-pharmacodynamics
6. Zhang Y, Wade MM, Scorpio A, Zhang H, Sun of pyrazinamide in a novel in vitro model of
Z (2003) Mode of action of pyrazinamide: tuberculosis for sterilizing effect: a paradigm
disruption of Mycobacterium tuberculosis mem- for faster assessment of new antituberculosis
brane transport and energetics by pyrazinoic drugs. Antimicrob Agents Chemother 53:
acid. J Antimicrob Chemother 52:790–795 3197–3204
7. Zhang Y, Permar S, Sun Z (2002) Conditions 15. Thiede JM et al (2022) Pyrazinamide suscepti-
that may affect the results of susceptibility test- bility is driven by activation of the SigE-
ing of Mycobacterium tuberculosis to pyrazina- dependent cell envelope stress response in
mide. J Med Microbiol 51:42–49 Mycobacterium tuberculosis. MBio 13:
8. Sbarbaro JA, Iseman MD, Crowle AJ (1992) e0043921
The combined effect of rifampin and pyrazina- 16. den Hertog AL et al (2016) Pyrazinamide is
mide within the human macrophage. Am Rev active against Mycobacterium tuberculosis cul-
Respir Dis 146:1448–1451 tures at neutral pH and low temperature.
Antimicrob Agents Chemother 60:4956. 18. Bacon J et al (2004) The influence of reduced
https://doi.org/10.1128/aac.00654-16 oxygen availability on pathogenicity and gene
17. Lanoix J-P et al (2016) Selective inactivity of expression in Mycobacterium tuberculosis.
pyrazinamide against tuberculosis in C3HeB/ Tuberculosis 84:205–217
FeJ mice is best explained by neutral pH of 19. Bacon J, Hatch KA (2009) Continuous culture
Caseum. Antimicrob Agents Chemother 60: of mycobacteria. Methods Mol Biol 465:153–
735–743 171
Chapter 9
Within-Host Mathematical Models of Antibiotic Resistance

Aminat Yetunde Saula, Gwenan Knight, and Ruth Bowness
Abstract
Mathematical models have been used to study the spread of infectious diseases from person to person. More
recently studies are developing within-host modeling which provides an understanding of how pathogens—
bacteria, fungi, parasites, or viruses—develop, spread, and evolve inside a single individual and their
interaction with the host’s immune system.
Such models have the potential to provide a more detailed and complete description of the pathogenesis
of diseases within-host and identify other influencing factors that may not be detected otherwise. Mathe-
matical models can be used to aid understanding of the global antibiotic resistance (ABR) crisis and identify
new ways of combating this threat.
ABR occurs when bacteria respond to random or selective pressures and adapt to new environments
through the acquisition of new genetic traits. This is usually through the acquisition of a piece of DNA from
other bacteria, a process called horizontal gene transfer (HGT), the modification of a piece of DNA within a
bacterium, or through. Bacteria have evolved mechanisms that enable them to respond to environmental
threats by mutation, and horizontal gene transfer (HGT): conjugation; transduction; and transformation. A
frequent mechanism of HGT responsible for spreading antibiotic resistance on the global scale is conjuga-
tion, as it allows the direct transfer of mobile genetic elements (MGEs). Although there are several MGEs,
the most important MGEs which promote the development and rapid spread of antimicrobial resistance
genes in bacterial populations are plasmids and transposons. Each of the resistance-spread-mechanisms
mentioned above can be modeled allowing us to understand the process better and to define strategies to
reduce resistance.
Key words Mathematical modeling, Antibiotic resistance, Within-host modeling, Differential equa-
tions, Prediction
1 Introduction
As early as the twentieth century, several mathematical models have

been used to study the spread of infectious diseases from person to
person. However, more studies are now focusing on within-host
modeling which provides an understanding of how pathogens—
bacteria, fungi, parasites, or viruses—develop, spread, and evolve
inside a single individual and their interaction with the host’s
https://doi.org/10.1007/978-1-0716-3981-8_9,
79
80 Aminat Yetunde Saula et al.
immune system. In some cases, where additional control mechan-

isms such as the administration of drugs are incorporated, these
models also describe the relationship between drug concentrations
and the effect on the infection within a host.
Although there are many ways to study the dynamics of infec-
tious diseases, given appropriate data, the use of mathematical
models has the potential to provide a more detailed and complete
description of the pathogenesis of diseases within-host and identify
other influencing factors that may not be detected when other
research methods are adopted [1]. In a more specific context to
bacterial infection, many mathematical models exist that intend to
aid understanding of the global antibiotic resistance (ABR) crisis, in
terms of helping to explain why it is happening, gaining insight into
how we can limit its impact, and proposing new ways of treating
infections: both susceptible infections, helping to minimize future
resistance emergence, and resistant infections by modeling new
compounds or novel treatment strategies.
ABR occurs when bacteria respond to random or selective
pressures and adapt to new environments through the acquisition
of new genetic traits. This is usually through the acquisition of a
piece of DNA from other bacteria, a process called horizontal gene
transfer (HGT), the modification of a piece of DNA within a
bacterium, or through mutations [2–4]. Naturally, bacteria have
evolved mechanisms that enable low mutation rates, usually in the
range of 10-6 to 10-9 per nucleotide per bacterial generation
[5]. However, under stressful conditions, bacteria have mechanisms
that induce or enhance mutagenic processes, thus, increasing muta-
tion rates [6]. With HGT, the movement of genetic material occurs
from a parent organism to another organism that is not its offspring
[7]. This allows bacteria to respond and adapt swiftly to their
environment by amassing (large) DNA sequences from another
bacterium in one transfer. This occurs through three main mechan-
isms, namely, conjugation; transduction; and transformation
[8, 9]. A frequent mechanism of HGT responsible for spreading
antibiotic resistance on the global scale is conjugation, as it allows
the direct transfer of mobile genetic elements (MGEs) between two
adjacently located bacteria, i.e., from a donor bacterial species to
different recipient species through cell-to-cell contact
[10, 11]. Although there are several MGEs, the most important
MGEs which promote the development and rapid spread of anti-
microbial resistance genes in bacterial populations are plasmids and
transposons [12].
By modeling within-host, the heterogeneity of bacteria and
immune cells can be easily represented, and the mechanisms men-
tioned above can be modeled.
Within-Host Mathematical Models of Antibiotic Resistance 81
2 Within-Host Modeling
In the construction of within-host models, it is possible to describe

the evolution of the host-bacteria interaction over space and time.
This could include the incorporation of bacterial replication, the
onset of host immunity, the inhibition/death of bacteria in a spe-
cific environment/space, damage to host tissues/cells, etc. [13]. In
aiding the elimination of bacteria, the immune system, which con-
sists of a complex network of different cells, constitutes a form of
heterogeneity to the host’s environment. Also, these cells usually
cluster in different organs, thus, creating a spatial structure. Mod-
eling within-host provides the best possible means to adequately
capture the biological processes involved in the evolution of
bacteria from susceptible to resistant strains. We have seen several
within-host studies [14–17] where these processes have been com-
prehensively embodied and have described the emergence, selec-
tion, or evolution of ABR either by horizontal gene transfer,
microbial fitness, or by mutations. Further, more within-host mod-
els have shown that during the course of treatment, improper or
non-optimal use of antibiotics, such as treating below the minimum
inhibitory concentration (MIC), can lead to bacteria adapting to
become resistant to these compounds. It is worth noting that the
concentration of drugs may vary between organs, thereby, compli-
cating the prediction about the evolution of bacteria that are resis-
tant to treatment [18, 19]. Thus, when there is a low concentration
of drugs at the site of infection, this could give rise to the mutant
selection window, which lies above the MIC of the sensitive strain
but below the MIC of the maximally resistant strain, which is also
called the mutant prevention concentration (MPC) [20]. There-
fore, the spatial heterogeneity of drug concentration in the host’s
body, which significantly contributes to the antibiotic resistance
mechanisms can be captured with within-host mathematical
models.
Previous studies [21–23] have postulated that upon infection,
many bacteria rapidly evolve. However, this simplification assumes
that the within-host evolution occurs immediately, which may not
be accurate. Therefore, it becomes important to understand how
changes in the size and/or the genetic composition of the bacterial
population could occur over the course of an infection [24]. An
example is the study carried out on 6 patients infected with Myco-
bacterium tuberculosis (M.tb.) which shows that during the treat-
ment of the disease, M.tb. susceptible strains independently evolved
into drug-resistant strains in the lungs. It was also shown that there
is a diversity of niches (i.e., the environmental conditions) within
the lungs driven by immunological processes and that the occur-
rence of ABR is more plausible in niches with high bacterial growth
caused by the clonal expansion of the bacterial population (i.e., the

dramatic increase in the number of bacteria that arise from the
replication of a bacteria of the same trait) [25].
It is important to emphasize that the evolution of antibiotic
resistance is not limited to the process by which mutants emerge
and their fixation but involves more complex dynamics that factor
in the heterogeneity in the host population [26]. Within-host
mathematical models may help to understand more about the
mechanisms at play during these processes. With these models, it
is possible to record susceptible bacterial populations, complete
replacement of susceptible populations with resistant strains, and
the coexistence of susceptible and resistant bacterial populations
driven by the within-host competition between the two strains of
bacteria, as well as the immune response. Many within-host studies
[27–29] provide examples where the host environment can main-
tain both resistant and sensitive bacterial populations, with coexis-
tence attributed to many processes such as the host population
structure, the niche formed by treated hosts, as well as balancing
selection.
When modeling bacterial infection and subsequently ABR
within-host, the dynamics to be incorporated in the model is very
important. This usually involves key factors such as the immune
response and the treatment to be administered. Therefore, a clear
description of how the growth of bacteria will be limited either by
the immune cells or the administration of antibiotics must be made
[30]. Inevitably, without the immune response and optimal admin-
istration of drugs (i.e., correct dose, time of administration, length
of treatment), it is expected that the probability of the emergence of
resistance will increase. In a more complex turn, during the treat-
ment of bacterial infection, the host microbiome as well as the
associated immune function can be accidentally disrupted through
the administration of antibiotics [31–33]. Although this mecha-
nism is rarely modeled, how the microbiome and variabilities in the
system may affect ABR should also be considered [34, 35]. By
modeling within-host, it is possible to examine the competition
between bacteria and commensal microbiota, while assessing the
effect of the competition in antibiotic selection for resistance. For
instance, the within-host model of [36] provides more insight into
how microbiome pathogen interaction influences ABR, through
colonization resistance, resource competition, or ecological release.
Clearly, when modeling the evolution of bacterial infection
within-host, environmental conditions have been shown to signifi-
cantly influence disease outcomes. Mathematical models can track
individual entities in the system and the changes that could occur
for these individuals following any changes within an environment
over time. This provides the population-level behavior that could
emerge from these interactions, while presenting explicitly the
history of network structures that arise among the individuals in a
system [37]. Hence, it becomes a useful tool for providing an

update on the changes that occur for each individual (e.g., immune
cells), as well as changes in bacterial population within an environ-
ment in a particular space and time. For instance, [38] using an
individual-based model showed that with resource limitation
within an environment, the competition between sensitive and
resistant bacteria intensifies, as a reduction in the growth rate of
sensitive bacteria increases the density of resistant bacterial strains.
Also, in an environment where treatment has been administered,
resistant bacteria fare better, as the susceptible bacteria succumb to
the drugs. Nonetheless, it is known that resistant strains of bacteria
often (but not always) have lower fitness than antibiotic-susceptible
strains. Therefore, with the use of a mathematical model, it was
possible to observe in the shape of the evolution of resistant strains
that frequency-dependent selection plays a significant role and
bacterial strain selection outcome depends on the balance of ben-
efits and costs, which is a result of the specific bacterial types and the
antibiotics used. Although the overall results of this work are
intuitive, with mathematical models we are able to investigate this
cost-benefit balance in a more thorough way, analyzing specific
antibiotics to understand more about how dosing, for example,
could be modified to avoid ABR.
Another area where within-host models can glean new insights
is in studying how mixed infections impact population-level resis-
tance. In [39] a model studying mixed-strain M. tuberculosis infec-
tions found that these mixed-strain infections may promote the
coexistence of drug-sensitive and resistant strains in two ways.
First, a strain that has specific advantages within a co-infected host
may outcompete another strain despite perhaps having a lower
basic reproductive number. Second, some individuals that previ-
ously suffered from mixed strain infection, but whose disease now
has progressed to phenotypically drug-sensitive tuberculosis (TB),
may retain small subpopulations of drug-resistant bacteria that can
flourish once the host is treated with antibiotics. The impact of
these within-host mechanisms can be used to gain insight into the
resulting population-level spread of resistance, highlighting
another use of these mathematical models.
Another important focus of mathematical modeling ABR is in
understanding more about the spatial structure of bacterial popula-
tions and the spatial heterogeneity in drug concentrations. This
spatial heterogeneity of antibiotics can facilitate drug resistance and
mathematical models can be used to investigate these impacts
[40, 41].
Understanding the factors that could influence the evolution of
ABR does not necessarily imply its incorporation into a model will
be straightforward. However, various mathematical modeling tech-
niques can be used to investigate ABR emergence within-host. We
have seen several studies [42–44] where compartmental or
individual-based modeling is adopted for within-host modeling of

bacterial infection. Usually, the adoption of compartmental mod-
els, in which the total quantities characterizing a system are split
into a number of interacting homogeneous subsystems or compart-
ments, leads to the use of equation-based models (EBMs) as tools.
Thus, the formulation of the compartmental models based on
EBMs may be in the form of ordinary differential equations
(ODEs) or partial differential equations (PDEs). However, the
evaluation of the EBMs of these compartmental models is done at
an aggregate level, which is in contrast to individual-based models
(IBMs). IBMs are simulation-based models that are able to account
for heterogeneity in a system, while capturing individual behaviors
that make up a system [45].
3 Types of Models: Compartmental Models and IBMs
As mentioned earlier, we might choose to model ABR within-host

using either compartmental models or IBMs. However, the con-
struction of both models is different, and there are many advan-
tages and disadvantages to their adoption.
Compartmental models use mathematical formulations that
describe the transition or exchange of elements and material
between compartments, i.e., here bacteria (considering various
strains), and may include host responses and antibiotic effects.
The aim is to optimally describe the underlying dynamics of the
biological system under consideration to account for what is
expected or known to occur realistically. Here, the interaction
(e.g., death of a population induced using antibiotics/growth inhi-
bition by immune cells) and transition rates (e.g., from susceptible
to resistant populations) between compartments must be
accounted for, and these are known as parameters of the model.
With an adequate representation of the parameters of the compart-
mental models, it is possible to track the population size of each
compartment through time.
There are many ways by which compartmental models are
developed. One could choose to build on an already existing
model by adding more compartments or complexities to the origi-
nal underlying system dynamics or construct an entirely new
model. Where modelers start with a simple formulation for the
evolution of bacteria, the number of compartments provides the
basic understanding that is needed to understand the problem
under consideration. Although further expansion of the model
would make it more complex, since there will be more interacting
subpopulations of bacteria, it is expected that the evolution of
bacteria should follow a similar profile to the simple model formu-
lation. An example of compartmental models starting with a simple
mathematical formulation of a system of ODEs followed by an
Fig. 1 Schematic illustration of a model of resistance to one antibiotic. This shows the evolution of bacteria
population following a logistic growth from the initial state of susceptibility to an antibiotic to the decrease in
the population of susceptible bacteria attributed to the actions of the immune cells (γ), the bactericidal effect
of the antibiotic (α) on the susceptible population, and the host’s carrying capacity K indicating scarcity of
resources which limits growth rate (i.e., vN/K). The subsequent growth of resistant bacteria follows fitness
cost, with a decrease in the population of resistant bacteria caused by γ and the carrying capacity K (i.e., v1N/
K ). The antibiotic does not affect resistant bacteria. The movement from susceptible to resistant compartment
(highlighted with a blue arrow) describes the proportion q of resistant bacteria to the antibiotic, that emerge
during the reproduction of susceptible bacteria
expansion into a complex framework was seen in [46] model of

antibiotic resistance. The system of ODEs captures the populations
of susceptible and resistant bacteria while accounting for popula-
tion growth, their immune response, and the use of antibiotics.
Initially, the model was constructed for antibiotic resistance to a
single drug and then expanded to account for resistance to two
different antibiotics. In some cases, these mathematical formula-
tions can be represented with schematic diagrams to aid visualiza-
tion of the models, allowing a clear depiction of how a certain
proportion of bacteria evolve to different forms of resistant mutants
of specific antibiotics during the reproduction of bacteria (either
susceptible or resistant bacteria). See Figs. 1 and 2 for examples of
such schematics.
Although we have provided an example where a compartmental
model is used to describe bacterial populations, in the context of
treatment modeling or more specifically antibiotic resistance, these
models are useful for describing the dynamics of drug distribution
for different infection sites. Usually, when the site of infection of a
host has a complex structure and poor vascularization, it becomes
difficult to optimize antibiotic treatment, which could encourage
the development of resistance. Thus, by adopting compartmental
Fig. 2 Schematic illustration of a model of resistance to one antibiotic. This shows the evolution of bacteria
population following a logistic growth from the initial state of susceptibility to antibiotics to the decrease in the
population of susceptible bacteria due to the actions of the immune cells (γ), the bactericidal effect of both
antibiotics (αi, i = 1, 2) on the susceptible population and the host’s carrying capacity K (i.e., vN K ). The
subsequent growth of resistant bacteria populations Ri, (i = 1, 2, 3) indicates the acquisition of resistance to
drug 1, drug 2, and both drugs respectively with growth rates vi, (i = 1, 2, 3) following fitness cost. The
reduction in the populations of resistant bacteria is due to γ, the bactericidal effect of drug 2 on R1, the
bactericidal effect of drug 1 on R2, and the carrying capacity K (i.e., viN/K ). Both antibiotics do not affect R3
population due to acquired resistance to both drugs. The movement between compartments (highlighted with
blue arrows) describes the proportion qi, (i = 1, 2) of resistant bacteria to drug 1, drug 2, or both drugs that
emerge during the reproduction of susceptible bacteria
models, the relationship between antibiotic effect and disease out-

come could be established. This is often called a compartmental
pharmacokinetic/pharmacodynamic model. We see an example of
this in the work of [41] where an assessment of the distribution and
penetration of drugs at sites of mycobacterial infection in the lungs
of tuberculosis patients influences the evolution of TB bacteria.
This compartmental model enabled the assessment of drug pene-
tration based on different tissues, where lesions with low plasma
levels had sub-optimal concentrations, thereby, leading to the
emergence of lesion-specific acquired resistance. Thus, emphasiz-
ing that spatial organization plays a huge role in determining
treatment outcomes and must be taken into account when model-
ing treatment and the emergence of ABR.
As stated earlier, for compartmental modeling, each compart-
ment is used to describe the changes in subpopulation size based on
known dynamics of the biological processes involved. In some
cases, especially when modeling diseases, modelers may choose to
reduce the number of parameters in the system for simplification
through non-dimensionalization of the state variables, i.e., the
defined densities of each population (e.g., S for susceptible, Ri
(i = 1,.., N) for resistant bacteria). The simplification of the models

formulated, either as PDEs or ODEs, could enable further investi-
gation into the problem under investigation, through mathematical
analysis of the system of equations of the model. This could provide
qualitative results on key information that contributes to antibiotic
resistance such as the conditions that influence the net reproductive
rate of susceptible bacteria; total elimination of susceptible bacteria;
time of generation of resistance; coexistence of susceptible and
resistant bacteria; fitness cost associated with mutation, etc.
A disadvantage of using compartmental models is that they are
typically deterministic, even though stochastic systems can be con-
structed such as in [47]. Deterministic models do not reflect natu-
ral variation within hosts/bacterial populations. Also, while the
adaption of compartmental PDE models can be used to model
the spatial movement and dynamics of densities of bacterial cells
over time, similar to ODEs, these models cannot capture spatial
and temporal heterogeneity of individual cells. However, with the
adoption of an individual-based model, this limitation does not
pose a problem.
While compartmental models make use of mathematical equa-
tions, individual-based models use an algorithm for simulating the
dynamics of the biological system under investigation. When using
individual-based models, the actions and interactions of each ele-
ment that makes up the system are simulated with some pre-set
rules in a specific spatial environment. Usually, IBMs can capture
unique individual behaviors of the elements in the system, as well as
any adaptive behavior that may occur as a result of interaction with
other elements or the conditions of the environment in which they
reside. When modeling antibiotic resistance using IBMs, it is possi-
ble to simulate how the bacterial population competes for
resources. This is an important attribute since competition has
been shown to be an important feature of the within-host popula-
tion structure. An example of a model that focuses on this is the
stochastic individual-based model of [48], which shows that the
coexistence of sensitive and resistant bacterial populations is driven
by the within-host competition between the two strains of bacteria.
Although stochastic events can be simulated using both compart-
mental models and IBMs, IBMs provide more flexibility when
accounting for variation due to chance [49, 50]. This is a feature
that is especially of interest when modeling the evolution of bacteria
within a host.
There are also limitations with IBMs: they often contain many
parameters, which are hard to estimate, especially in ABR research
where there is little experimental data available. The models also
typically take a long time to run. However, despite their drawbacks,
with heterogeneity inherent in many biological processes, the adop-
tion of IBMs for simulating the heterogeneous dynamics of bacte-
rial infection and its evolution is becoming popular. Moreover, in a
general context, IBMs are a progressively useful tool for under-

standing infectious diseases, assessing targeted intervention mea-
sures, and improving policy-making [51]. While we have briefly
introduced IBMs in this chapter, more details on IBMs, their
construction, use, and limitations are presented in the next chapter.
In more recent studies, in addition to just employing mathe-
matical models such as EBMs, IBMs, or other stochastic simulation
models, EBMs are now more frequently being integrated into an
IBM framework. This leads to hybrid models that are able to give
an adequate representation of all the processes involved in a system,
thereby giving individual-level and system-level information. These
hybrid models are spatio-temporal in nature and have been shown
to provide numerous insights into host-pathogen dynamics and
antimicrobial effects [52, 53]. For example, [54] uses a hybrid
framework of ODEs and PDEs integrated into an individual-
based model to track the acquisition of bacterial strains that are
resistant to antibiotics, as well as compensatory mutations. Here
they found that the concentration of antibiotics impacts the pro-
portion of resistant bacteria that may emerge, and a relatively
moderate increment in the dose of antibiotics could allow efficient
penetration of antibiotics that could significantly alter the time of
the emergence and the percentage at which resistance is acquired.
Other spatial modeling methods, which include models that
utilize PDEs and metapopulation models, have been used in
within-host modeling of infection in general [55–58], there is little
evidence of their use in modeling within-host antibiotic resistance.
There is a spectrum of mathematical modeling types that can all
be useful to increase understanding of ABR, from simple compart-
mental models right up to complex hybrid individual-based models
that, as well as discrete processes, can also include integrated differ-
ential equations systems. There is a trade-off between these model-
ing approaches—as the complexity increases so do the number of
parameters that need to be estimated, as well as the computational
time required to run simulations. This must be taken into account
before deciding which modeling approach to take.
4 Future Developments: More Data Is Needed
We have discussed and reviewed some of the many mathematical

models that study antibiotic resistance. However, there is still a
massive lack of knowledge regarding the underlying mechanisms
of ABR. This fact limits the impact of mathematical modeling in
this area. There is also a need for more rigorous model develop-
ment and testing in this area [34]. New focused research is urgently
needed in order to provide experimental data that mathematicians
can then use to parameterize and validate their models. Only then
will meaningful and useful insights be seen from these models.
References
1. Awofisayo-Okuyelu A, Pratt A, McCarthy N, 16. Huddleston JR (2014) Horizontal gene trans-
Hall I (2019) Within-host mathematical mod- fer in the human gastrointestinal tract: poten-
elling of the incubation period of salmonella tial spread of antibiotic resistance genes. Infect
typhi. R Soc Open Sci 6:182143 Drug Resist 7:167
2. Martinez J, Baquero F (2000) Mutation fre- 17. McCarthy AJ, Loeffler A, Witney AA, Gould
quencies and antibiotic resistance. Antimicrob KA, Lloyd DH, Lindsay JA (2014) Extensive
Agents Chemother 44:1771–1777 horizontal gene transfer during staphylococcus
3. Boto L, Mart´ınez JL. (2011) Ecological and aureus co-colonization in vivo. Genome Biol
temporal constraints in the evolution of bacte- Evol 6:2697–2708
rial genomes. Genes 2:804–828 18. Lipsitch M, Levin BR (1997) The population
4. Pang Z, Raudonis R, Glick BR, Lin TJ, Cheng dynamics of antimicrobial chemotherapy. Anti-
Z (2019) Antibiotic resistance in pseudomonas microb Agents Chemother 41:363–373
aeruginosa: mechanisms and alternative thera- 19. Holt RD, Barfield M (2006) Within-host path-
peutic strategies. Biotechnol Adv 37:177–192 ogen dynamics: some ecological and evolution-
5. Chevallereau A, Meaden S, van Houte S, Wes- ary consequences of transients, dispersal mode,
tra ER, Rollie C (2019) The effect of bacterial and within-host spatial heterogeneity. In: Dis-
mutation rate on the evolution of crispr-cas ease evolution: models, concepts, and data ana-
adaptive immunity. Philos Trans R Soc B 374: lyses, vol 71, pp 45–66
20180094 20. Drlica K (2003) The mutant selection window
6. Foster PL (2005) Stress responses and genetic and antimicrobial resistance. J Antimicrob
variation in bacteria. Mutat Res/Fundam Mol Chemother 52:11–17
Mech Mutagen 569:3–11 21. Thompson JN (1998) Rapid evolution as an
7. Burmeister AR (2015) Horizontal gene trans- ecological process. Trends Ecol Evol 13:329–
fer. Evol Med Public Health 2015:193 332
8. Bushman F (2002) Lateral DNA transfer. Cold 22. Thomas CM, Nielsen KM (2005) Mechanisms
Spring Harbor Laboratory Press of, and barriers to, horizontal gene transfer
9. Von Wintersdorff CJ, Penders J, Van Niekerk between bacteria. Nat Rev Microbiol 3:711–
JM, Mills ND, Majumder S, Van Alphen LB 721
et al (2016) Dissemination of antimicrobial 23. Culyba MJ, Van Tyne D (2021) Bacterial evo-
resistance in microbial ecosystems through lution during human infection: Adapt and live
horizontal gene transfer. Front Microbiol 7: or adapt and die. PLoS Pathog 17:e1009872
173 24. Alizon S, Luciani F, Regoes RR (2011) Epide-
10. Andam CP, Fournier GP, Gogarten JP (2011) miological and clinical consequences of within-
Multilevel populations and the evolution of host evolution. Trends Microbiol 19:24–32
antibiotic resistance through horizontal gene 25. Kaplan G, Post FA, Moreira AL,
transfer. FEMS Microbiol Rev 35:756–767 Wainwright H, Kreiswirth BN, Tanverdi M
11. Clewell DB (2013) Bacterial conjugation. et al (2003) Mycobacterium tuberculosis
Springer Science & Business Media growth at the cavity surface: a microenviron-
12. Munita JM, Arias CA (2016) Mechanisms of ment with failedimmunity. Infect Immun 71:
antibiotic resistance. Microbiol Spectr 4:4–2 7099–7108
13. Coombs D, Gilchrist MA, Ball CL (2007) 26. Read AF, Huijben S (2009) Perspective: evolu-
Evaluating the importance of within-and tionary biology and the avoidance of antimi-
between-host selection pressures on the evolu- crobial resistance. Evol Appl 2:40–51
tion of chronic pathogens. Theor Popul Biol 27. Boni MF, Feldman MW (2005) Evolution o
72:576–591 antibiotic resistance by human and bacterial
14. Andersson DI, Patin SM, Nilsson AI, Kugel- neche construction. Evolution 59:477–491
berg E (2007) The biological cost of antibiotic 28. Cobey S, Baskerville EB, Colijn C, Hanage W,
resistance. In: Enzyme-mediated resistance to Fraser C, Lipsitch M (2017) Host population
antibiotics: mechanisms, dissemination, and structure and treatment frequency maintain
prospects for inhibition, pp 339–348 balancing selection on drug resistance. J R
15. SchulzzurWiesch P, Engelstadter J, Bonhoeffer Soc Interface 14:20170295
S (2010) Compensation of fitness costs and 29. Blanquart F, Lehtinen S, Lipsitch M, Fraser C
reversibility of antibiotic resistance mutations. (2018) The evolution of antibiotic resistance in
Antimicrob Agents Chemother 54:2085–2095 a structured host population. J R Soc Interface
15:20180040
30. Handel A, Margolis E, Levin BR (2009) within-host spatiotemporal dynamics of inva-

Exploring the role of the immune response in sive bacterial disease. PLoS Biol 6:e74
preventing antibiotic resistance. J Theor Biol 43. Kirschner D, Pienaar E, Marino S, Linderman
256:655–662 JJ (2017) A review of computational and math-
31. Buffie CG, Pamer EG (2013) Microbiota- ematical modeling contributions to our under-
mediated colonization resistance against intes- standing of mycobacterium tuberculosis
tinal pathogens. Nat Rev Immunol 13:790– within-host infection and treatment. Curr
801 Opin Syst Biol 3:170–185
32. Kim S, Covington A, Pamer EG (2017) The 44. Chen HH, Stringer A, Eguale T, Rao GG,
intestinal microbiota: antibiotics, colonization Ozawa S (2019) Impact of antibiotic resistance
resistance, and enteric pathogens. Immunol on treatment of pneumococcal disease in
Rev 279:90–105 Ethiopia: an agent-based modeling simulation.
33. Chatterjee A, Modarai M, Naylor NR, Boyd A J Trop Med Hyg 101:1042
SE, Atun R, Barlow J et al (2018) Quantifying 45. Parunak DHV, Savit R, Riolo RL (1998)
drivers of antibiotic resistance in humans: a Agent-based modeling vs. equation-based
systematic review. Lancet Infect Dis 18:e368– modeling: a case study and users’ guide. In:
e378 International workshop on multi-agent sys-
34. Birkegård AC, Halasa T, Toft N, Folkesson A, tems and agent-based simulation. Springer, pp
Græsbøll K (2018) Send more data: a system- 10–25
atic review of mathematical models of antimi- 46. Alavez-Ramirez J, Castellanos JRA, Esteva L,
crobial resistance. Antimicrob Resist Infect Flores JA, Fuentes-Allen JL, Garcıa-Ramos G
Control 7:1–12 et al (2007) Within-host population dynamics
35. Knight GM, Davies NG, Colijn C, Coll F, of antibiotic-resistant M. tuberculosis. Math
Donker T, Gifford DR et al (2019) Mathemat- Med Biol 24:35–56
ical modelling for antibiotic resistance control 47. Wang X, Wang J (2017) Modeling the within-
policy: do we know enough? BMC Infect Dis host dynamics of cholera: bacterial–viral inter-
19:1–9 action. J Biol Dyn 11:484–501
36. Smith DR, Temime L, Opatowski L (2021) 48. Davies NG, Flasche S, Jit M, Atkins KE (2019)
Microbiome-pathogen interactions drive epi- Within-host dynamics shape antibiotic resis-
demiological dynamics of antibiotic resistance: tance in commensal bacteria. Nat Ecol Evol 3:
A modeling study applied to nosocomial path- 440–449
ogen control. elife 10:e68764 49. Willem L, Verelst F, Bilcke J, Hens N, Beutels P
37. DeAngelis DL, Grimm V (2014) Individual- (2017) Lessons from a decade of individual-
based models in ecology after four decades. based models for infectious disease transmis-
F1000prime Reports 6 sion: a systematic review (2006-2015). BMC
38. Wale N, Sim DG, Jones MJ, Salathe R, Day T, Infect Dis 17:1–16
Read AF (2017) Resource limitation prevents 50. Funk S, Camacho A, Kucharski AJ, Eggo RM,
the emergence of drug resistance by intensify- Edmunds WJ (2018) Real-time forecasting of
ing within-host competition. Proc Natl Acad infectious disease dynamics with a stochastic
Sci 114:13774–13779 semi-mechanistic model. Epidemics 22:56–61
39. Sergeev R, Colijn C, Cohen T (2011) Models 51. Roche B, Duboz R (2017) Individual-based
to understand the population-level impact of models for public health. In: Handbook of
mixed strain m. tuberculosis infections. J Theor statistics, vol 37. Elsevier, pp 347–365
Biol 280:88–100 52. Marino S, El-Kebir M, Kirschner D (2011) A
40. Kepler TB, Perelson AS (1998) Drug concen- hybrid multi-compartment model of granu-
tration heterogeneity facilitates the evolution loma formation and t cell priming in tubercu-
of drug resistance. Proc Natl Acad Sci 95: losis. J Theor Biol 280:50–62
11514–11519 53. Bowness R, Chaplain MA, Powathil GG, Gille-
41. Strydom N, Gupta SV, Fox WS, Via LE, spie SH (2018) Modelling the effects of bacte-
Bang H, Lee M et al (2019) Tuberculosis rial cell state and spatial location on
drugs’ distribution and emergence of resistance tuberculosis treatment: Insights from a hybrid
in patient’s lung lesions: a mechanistic model multiscale cellular automaton model. J Theor
and tool for regimen and dose optimization. Biol 446:87–100
PLoS Med 16:e1002773 54. Pienaar E, Linderman JJ, Kirschner DE (2018)
42. Grant AJ, Restif O, McKinley TJ, Sheppard M, Emergence and selection of isoniazid and
Maskell DJ, Mastroeni P (2008) Modelling rifampin resistance in tuberculosis granulomas.
PLoS One 13:e0196322
55. Ganguli S, Gammack D, Kirschner DE (2005) 57. Hao W, Schlesinger LS, Friedman A (2016)
A metapopulation model of granuloma forma- Modeling granulomas in response to infection
tion in the lung during infection with myco- in the lung. PLoS One 11:e0148738
bacterium tuberculosis. Math Biosci Eng 2:535 58. Pitcher MJ, Bowness R, Dobson S, Eftimie R,
56. Golubchik T, Batty EM, Miller RR, Farr H, Gillespie SH (2020) Modelling the effects of
Young BC, Larner-Svensson H et al (2013) environmental heterogeneity within the lung
Within- host evolution of staphylococcus on the tuberculosis life-cycle. J Theor Biol
aureus during asymptomatic carriage. PLoS 506:110381
One 8:e61319
Chapter 10
Use of Individual-Based Mathematical Modelling

to Understand More About Antibiotic Resistance
Within-Host
Aminat Yetunde Saula, Christopher Rowlatt, and Ruth Bowness
Abstract
To model complex systems, individual-based models (IBMs), sometimes called “agent-based models”
(ABMs), describe a simplification of the system through an adequate representation of the elements.
IBMs simulate the actions and interaction of discrete individuals/agents within a system in order to discover
the pattern of behavior that comes from these interactions. Examples of individuals/agents in biological
systems are individual immune cells and bacteria that act independently with their own unique attributes
defined by behavioral rules. In IBMs, each of these agents resides in a spatial environment and interactions
are guided by predefined rules. These rules are often simple and can be easily implemented. It is expected
that following the interaction guided by these rules we will have a better understanding of agent-agent
interaction as well as agent-environment interaction. Stochasticity described by probability distributions
must be accounted for. Events that seldom occur such as the accumulation of rare mutations can be easily
modeled.
Thus, IBMs are able to track the behavior of each individual/agent within the model while also obtaining
information on the results of their collective behaviors. The influence of impact of one agent with another
can be captured, thus allowing a full representation of both direct and indirect causation on the aggregate
results. This means that important new insights can be gained and hypotheses tested.
Key words Mathematical modeling, Antibiotic resistance, Agent-based modeling, Differential equa-
tions, Prediction
1 Introduction
In the previous chapter, we have shown that modeling the evolu-

tion of antibiotic resistance is complex, and as such, it is best to start
with simple models on which to build on. Although equation-
based models utilizing ordinary differential equations (ODEs)
and partial differential equations (PDEs) have been traditionally
used when modeling mathematical problems, there are limited
actions these methods can include when modeling complex sys-
tems. For complex systems that are intrinsically difficult to model,
https://doi.org/10.1007/978-1-0716-3981-8_10,
93
individual-based models (IBMs), sometimes called “agent-based

models” (ABMs), describe a simplification of the system through
an adequate representation of the elements of the system and
therefore can be useful types of models.
IBMs simulate the actions and interaction of discrete indivi-
duals/agents within a system, to discover the pattern of behavior
that comes from these interactions [1] Examples of individuals/
agents in biological systems are individual immune cells and bacte-
ria. Usually, each individual element acts independently but has its
own unique attributes, such as their state and basic behavioral rules.
In IBMs, each of these individuals resides in a spatial environment
that is explicitly represented. The interaction between these indivi-
duals within a local environment is guided by predefined rules.
These rules are often simple and can be easily implemented. It is
expected that following the interaction guided by these rules, we
have a better understanding of agent-agent interaction as well as
agent-environment interaction. In a general sense, we can catego-
rize mathematical models based on how much randomness they
have, i.e., the model can be deterministic (giving the same answer
each time the model is run) or stochastic (outputting slight varia-
tions for each model run). For most biological processes, noise and
stochasticity are inherent attributes, and thus, a certain level of
unpredictability described by probability distributions must be
accounted for. Therefore, it is to be expected that when individuals
interact and move within a spatial environment in which they
reside, an element of randomness will occur. While many ODEs
and PDEs are deterministic and not able to incorporate this attri-
bute, IBMs are stochastic, where we can include random character-
istics of individuals, which are vital for adequate representation of
interacting systems. Thus, when investigating bacterial infection,
by using IBMs, phenomenon such as bacterial replication rate
which occurs at the cellular level can be simulated from probability
distributions. This enables the tracking of changes in the bacterial
population within an environment over time. In addition, events
that seldom occur such as the accumulation of rare mutations can
be easily modeled using IBMs due to their ability to generate both
the average outcome of a system, as well as the distribution of
possible outcomes [2].
In contrast to equation-based models, IBMs are able to track
the behavior of each individual/agent within the model while also
obtaining information on the results of their collective behaviors.
Often, when representing biological systems mathematically, the
influence of an individual element on another might not be direct
[3]. This could be due to the mediation by another individual
element (e.g., an immune cell or a bacterium) following a set
rule, thereby indirectly instigating an action that influences the
overall outcome of the model. With IBMs, this can be captured,
thus, allowing a full representation of both direct and indirect
Use of Individual-Based Mathematical Modelling to Understand More About. . . 95
causation on the aggregate results. An example is the immune cell

response to TB bacteria, where active T cells kill the bacteria. The
activation of T cells primarily occurs after the interaction with
macrophages, which enables cytotoxic activity on bacteria. The
importance of this ability cannot be over-emphasized, as it is quite
easy to get lost trying to understand the aggregate outcome or
behavior of a complex system. However, with the behaviors of the
individuals/agents emerging from predefined rules, the explora-
tion of complex systems becomes easier.
Several TB models [4–7] have indicated that when modeling
bacterial infection, the state and location of bacterial cells signifi-
cantly impact the disease outcome. Also, when treating bacterial
infections, how the concentration of antibiotics is distributed over
space and time has a significant impact on the evolution of bacteria,
for instance, the emergence of drug resistance. Hence, many of the
problems we encounter have an environmental and location-
dependent element to them that needs to be incorporated. There-
fore, through the adaptation of IBMs, it is possible to model the
explicit spatial structure of cellular processes that occur in a
biological system, which usually have an impact on evolution
[8, 9]. Also, it is expected that the interaction among individuals
will not always take place at the same time, and as such, we need to
account for the interactions occurring at different times. This could
bring about an adjustment in the original behavior of the indivi-
duals, allowing the state and the actions of individuals, as well as
their interaction with other individuals and their local environment
to change in time and space. This adjustment does not present a
problem or further complication to individual-based models as it
can capture the spatio-temporal heterogeneity of individual
entities.
The ability of IBMs to capture the adaptive behavior of indivi-
duals in a system is a desirable feature to modelers [10]. Hence, in
the context of within-host antibiotic resistance, when investigating
cell competition and fitness effect (e.g., between sensitive and
resistant bacteria), the use of IBMs is insightful. Further, while
the rules guiding biological phenomenon are constructed by clin-
icians and experimental biologists based on biological intuition and
experimental data obtained from in vivo and in vitro experiments,
the use of IBMs allows mathematicians to transcribe these rules into
a computational framework, thus facilitating interdisciplinary col-
laboration. This makes it possible to give iterative feedback to
carefully assess parameters and observe the system behavior beyond
the original observation range, thus, bringing about new experi-
ments that test the predictions generated by the model designed
[11, 12].
Due to their numerous advantages, IBMs are gaining ground
in within-host modeling. This has been seen for several models of
bacterial infections such as in TB modeling [13–15], the modeling
of Helicobacter pylori [16] and Mycobacterium avium complex

[17]. We have also seen modeling efforts elsewhere for viral and
parasitic infections such as in HIV-1 [18], Leishmania major [19],
malaria [20], and SARS-CoV-2 [21]. Although the IBMs adopted
for these studies were developed for different purposes, they are all
able to account for the heterogeneity of their various environ-
ments, as well as the response and characteristics of the individuals
within these environments.
IBMs, of course, also have their limitations. The cost of com-
putation in terms of computer storage capacity and execution time,
which arises because of tracking individual entities, is the main
limitation of IBMs. However, this is easily overcome by under-
standing the number of individual entities to be represented and
how much detail the representation should have [10]. An example
to combat this issue is the work of [22], where “tuneable resolu-
tion” was adopted for the IBM of lung granuloma in tuberculosis.
This enabled the reduction of nine ODEs describing the change of
soluble TNF concentration inside an IBM compartment having
either macrophages or T cells into a single equation, while ensuring
sensitive relationships are maintained.
In biological applications, parametrization is also a limitation to
using IBMs. This is due to the high level of requirements that must
be met to obtain information relevant to the biological process. In
addition, the flexibility IBMs provide usually involves a high num-
ber of parameters, bringing about parameter uncertainty. Never-
theless, the problem of parametrization can be approached through
empirical knowledge via probabilistic rules and allowing a certain
range of parameter variation [23, 24]. Conducting sensitivity anal-
ysis and validating the model helps tackle this limitation. The test
for parameter uncertainty and sensitivity is usually recommended
when modeling biological phenomena, thus, is not specific to
IBMs. Several studies such as [25] and [26] have discussed the
many approaches to conducting these tests, giving intuition into
when and how to use specific sensitivity measures based on the
relationship exhibited by the parameters and model outputs.
More specifically, uncertainty and sensitivity analysis methods to
use when working with IBMs are discussed in detail in [27].
With careful construction of the IBM, the benefits usually
outweigh the limitations. The applicability of IBM from simple to
highly complex systems and its ability to predict a broad range of
realistic system-level characteristics from a model in which indivi-
duals follow a few simple decision rules is remarkable. More specifi-
cally, its adoption at the within-host level will allow a
comprehensive understanding of the host system and the evolution
of bacteria to resistant strains with or without antibiotic exposure
while capturing the heterogeneity and spatial interactions of the
individual entities.
2 Constructing IBMs
Although we have mentioned different scenarios where IBMs have

been used in within-host infection, limited studies have been done
for antibiotic resistance. We hope that by detailing a step-by-step
approach to developing an IBM, more understanding of how local
interactions of the individual elements influence the emergent
behavior of a system will make its construction easier. Depending
on the purpose for which we intend to develop an IBM, several
decisions have to be made for its construction. This usually ranges
from a clear definition of the environment where interaction
occurs, the definition of the individuals that will make up the
model, the declaration of the variables characterizing individual
states, individual and environmental processes involved, general
attributes that could be observed in the system, to more specific
assumptions made about the behavioral rules of the individuals. In
what follows, we present guidelines indicating the steps to be taken
when constructing an IBM. We start by describing and highlighting
the first things that should come to mind when developing an IBM,
and then proceed to more implementation-based questions about
model initialization, input, and output, and discuss some program-
ming languages where ABM implementation and execution are
preferred.
2.1 Defining the The first step in IBM construction is to clearly define the individuals
Environment and the that make up the system, the environment which serves as a habitat
Individual Elements for the individuals, and the spatial and temporal extent of the
model. Most IBMs, especially when modeling within-host antibi-
otic resistance, are spatially explicit. This implies that they capture
behavioral shifts of individuals under varying conditions in their
local environment (such as a section of lung tissue, a section of
intestinal epithelium, etc.) at different spatial and temporal scales.
In the following steps, more information will be provided about the
behavioral changes that could occur. For now, we shall focus on the
many questions that revolve around our definition of the environ-
ment and the processes or actions that could be carried out by the
individuals within our defined environment.
When defining an environment, different types of space can be
adopted during the construction of an IBM. This could be through
the use of arbitrary graphs, discrete or continuous spaces. A discrete
(grid-like) space is commonly utilized for within-host IBMs due to
its simplicity, visual appeal, and its easy tracking of the position of
cells over time. With a grid-like space, it is easy to observe the
removal, replication/recruitment of agents distributed over cells,
as well as the events that occur at each time step—which is an
important attribute when modeling the evolution of bacteria.
Thus, we present some of these questions as guidelines for consid-
eration when defining the individuals that make up the system and
Fig. 1 Simple illustration of possible environment and individuals for a within-host model of antibiotic
resistance
their environment, emphasizing the use of grid-like spaces. We give

an example defining this step in Fig. 1.
(a) How do we adapt a grid-like space within our environment?
This could be how the neighborhood in which agents can
move or perform some specific action is defined—i.e., there
are different neighborhoods that can be adopted in these
models, ones that take into account the eight neighboring
cells or those that take into account just the four neighboring
cells (above, below, immediately left and right).
(b) Should we specify that only a unique type of individual may
occupy the same grid, or can we accommodate more than one
type of individual on the same grid cell?
(c) What is the size of the grid cell to be used, why have we chosen
this dimension, and what grid spacing should we use?
(d) Which individuals are relevant to the model?
2.2 Declaration of A crucial step in the appropriate construction of an IBM is the

Initial State Variables explicit specification on possible states (and traits) each individual
(and Traits) of the element may have, and how a change of state (and trait) could occur
Individual Entities over time and space. The state of a particular individual refers to the
current condition that the individual is in at a specific time. We can
also go further to define if an individual, say, an inactive individual
has a distinguishing quality or characteristic, which we refer to as a
trait. We must ensure that only key individuals or individual-level
Fig. 2 An example of possible initial state (and traits) variables. Green box represents the individuals, blue
indicates the state variables, and red indicates the trait variables
characteristics (curbing excessive detail) needed for the problem

should be included, while ensuring adequate representation of the
system. For example, when modeling bacterial infection how much
complexity in the immune response is necessary? We know that, for
example, T cells can differentiate into CD4 or CD8 T cells and that
the function of T cells includes activating other immune cells,
producing cytokines and regulating the immune response, and
directly killing infected host cells. If we deem it appropriate, we
may choose to ignore the differentiation of the T cells into two
types, and just have a type of T cell, with different states that are
assigned for specific functions at a given probability or rate. In the
past it was assumed that there was usually only one bacterial phe-
notype, but we are increasingly identifying more complexity. For
example, Mycobacterium tuberculosis undergoes significant pheno-
typic change in response to various environmental pressures
[28]. Figure 2 presents an illustrative example with a clear declara-
tion of the initial possible states (and traits) that could be exhibited
by bacteria (with more than one phenotype), neutrophils, macro-
phages, T cells, as well as other cells which we collectively term,
generic cells. At this stage, we want to ask two significant questions.
(e) What is the initial set of variables that characterizes the indivi-
duals in our model (i.e., defining the elementary properties of
the individuals)?
(f) What choices are possible for each characteristic variable? For
example, if the characteristic variable is a state, then the choices
could be resting or active.
2.3 Defining We have defined the initial states and traits of the individuals in the
Individual Activities/ second step, as a basis for what contributes to state updates. The
Actions and traits of the individual entities and their behavioral rules organize
Environmental their actions and how they interact. In step 2, we have not defined
Attributes Influencing the activities that could occur and the outcome of interactions
State (and Traits) between individuals. However, it is important to note that indivi-
Updates duals have different objectives. Hence, we need to carefully think
through the actions that are important to address these objectives,
again, ignoring irrelevant ones. Hence, for each type of individual, a
rule must be defined that allows the implementation of possible
activities that could be carried out in a particular situation. Usually,
these activities influence a change in the state of the individuals as
well as the environment. Hence, the actions to be incorporated and
their implications must be clearly stated.
In some cases, individuals may make decisions in response to
some form of stimuli, (e.g., the initiation of cell recruitment
through the production of cytokines in response to bacterial infec-
tion, movement of immune cells toward the site of infection,
induced by the high concentration of chemokine molecules)
and/or other criteria (e.g., random/pseudo-random migration).
These activities or actions usually contribute to the overall outcome
observed and may take different forms. Some of these activities
occur randomly, in a hierarchy, concurrently, or could be fixed.
Hence, it becomes important to assign rates or probabilities,
where appropriate, to the activities or actions carried out. Examples
of these actions or decisions taken by individuals include the
recruitment of T cells at a particular density of bacteria or the supply
of antibiotics at a particular time step to inhibit bacterial growth or
induce mortality ( fixed); the death of bacteria after the movement
of active macrophages toward the infection site (hierarchy); the
migration of T cells toward infected macrophages to induce the
mortality of macrophages and internalized bacteria at the same time
(concurrent); and the phagocytosis of bacteria upon contact with
inactive macrophages (random). In the examples given, following
these actions, a state and trait update will occur for bacteria and
macrophages. While the interaction between individuals affects the
environment, environmental attributes such as ventilation, perfu-
sion, and drainage could also influence a state update of individuals,
thus, affecting the overall outcome of our result [29]. A good
example is the response of bacteria to surplus or deprivation of
oxygen and/or different concentrations of antibiotics in an envi-
ronment, where an oxygen deficit environment could reduce the
replication rate of bacteria.
Following clear definitions of the variables characterizing the
state (and trait) of the individuals in the system as well as the
activities that could be carried out, we begin to ask questions of
the form “where, when, and how?” For instance, in modeling
antibiotic resistance, we might seek to clearly define “when” and

“how” mutants could be acquired, as well as “where” it could
occupy in our defined local environment. We might choose, for
example, to define “when”—as the period the antibiotic treatment
is below the MIC; “how”—might be done such that at each time
step, a prescribed probability is set in place, that a bacterium on a
grid cell could become resistant if treatment stays below the MIC
for a particular duration of time. We have given these examples to
widen our understanding of the relevance of the questions that
should come to mind when developing IBMs for a particular
research problem. More succinctly, in this very crucial step, we
want to address the following:
(g) What activity do we want the individuals in our model to
exhibit? The activities could range from migration, reproduc-
tion, recruitment, and growth inhibition, to mortality and
many more.
(h) Are we restricting some activity to a particular location or
time step? This could be that replication of bacteria could
only take place within a particular neighborhood, or migra-
tion of an individual to a neighboring grid cell occurs with a
particular probability at a given time step.
(i) In what form should the processes/actions taken by indivi-
duals be carried out and in what order should the state vari-
ables be updated?
(j) How should individual interactions and activities be executed
in our simulations? This implies a clearly defined plan of the
processes involved in the model. For example, which activity
occurs as a result of interactions between individuals, and
which occurs at fixed time intervals?
(k) What scale should be used for the model parameters describ-
ing the actions and interactions of the individuals?
(l) Have we clearly defined the probabilities attached to each
activity that could occur, as well as the rates for state updates
for the individuals in our model?
(m) Have we thought of the possible relevant environmental attri-
butes that could be integrated into the model and how this
could influence the state (and traits) of the individuals in our
system?
(n) What temporal resolution (i.e., length of simulation time) will
be able to capture the dynamics and temporal distribution of
the bacterial infection under investigation?
(o) Is it appropriate to assign lifespan to some individuals in the
system?
2.4 Declaration of In step 3, we mentioned that various activities could be carried out
Assumptions About by each individual element in a system. However, we are yet to
Behavioral Rules of discuss how individual characteristics or actions could change in an
Individuals environment over time and how the behavior of an individual
might change as a consequence of the behavior exhibited by other
individual(s). It is important to have a clear thought on the possi-
bility that some of the individuals in our system might exhibit
adaptive behavior which could influence their form or actions,
and their interaction with other individuals and the environment
in which they reside. A good example is a probability that some
drug-susceptible bacteria in a particular environment can develop
mechanisms against antibiotics, thereby improving their fitness,
and bringing about resistant bacteria. In this case, how the drug
affects susceptible bacteria will be different from its effect on
mutants. Hence, we need to declare how the adaptability of indivi-
duals to their environment might impact the changes to an indivi-
dual’s behavior as well as the overall outcome of the research
problem. The integration of this behavior into the IBM con-
structed will not pose a problem, as one of the classical properties
of IBMs is the ability to integrate dynamics and feedback within the
model [30].
Further, the interactions between individuals within a spatial
environment might occur randomly. However, we are yet to incor-
porate the possibility that some individuals in an environment
could detect changes in their environment, which could ultimately
influence their decisions. An example is the migration of immune
cells through space toward the site of infection invoked by a signal-
ing protein such as chemokine molecules. Also, a clear definition of
the outcome of the interaction between individuals of different
types should be made. Thus, it is important to state some underly-
ing assumptions of behavioral rules and attributes that could
emerge in our model. This could be stochasticity, the ability to
detect or sense signals, emergence (i.e., the behavior/outcome that
emerges only after the interaction with others), and other attributes
we consider to be significant for our model. Again, we present some
guidelines to the questions that could guide our model construc-
tion when declaring attributes that might occur over time in our
model.
(p) Do we have individual(s) in our model that are able to adapt to
the environment?
(q) How does adaptive behavior influence the actions, and fitness
of individuals, as well as their interaction with other indivi-
duals and the environment?
(r) What level of randomness (stochasticity) is to be included in
the behavioral rules? For example, how much variation is there
in the replication rate of particular bacteria?
(s) Do we have individual(s) that can detect signals, and how does
this attribute contribute to the model to be constructed?
(t) Are there any assumed interactions among the individuals in
our system, and what should the impact of these
interactions be? E.g., should we assume that the interaction
between an active T cell and bacteria will always result in the
death of the bacteria?
In Algorithm 1, we present an example of a step-wise represen-
tation that could be adapted when modeling the evolution of
antibiotic resistance using the steps described above. This includes
some possible activities/actions of the individuals, the order and
situation in which they could be carried out, the subsequent effect
on their state (and trait) updates, as well as how a behavior change
could occur. For clarity, we have emphasized the state (and traits)
updates in red and detailed the processes/actions carried out as
comments. Here, we have included an option for more than one
phenotype of bacteria. Where a single phenotype is modeled, the
process of state change of bacteria should be ignored. Although it is
not explicitly represented, it is assumed that regulated cytokines
production is involved in the recruitment of both innate and adap-
tive immune cells. Figure 3b gives a visualization to Algorithm
1 (found in Fig. 3a) in the form of a flowchart. It is important to
note that the immune system consists of a complex network of cells,
with many processes involved in their ability to perform various
functions that aid in the containment of bacterial infection. Thus,
the flowchart presented should be taken as an example with
assumptions/simplifications to the response of the immune cells
and the treatment of bacterial infection using antibiotic(s).
2.5 Defining the For IBM simulations, one must think of the initial values to assign
Initialization of State to some (or all) of the state and (traits) variables of the individuals in
(and Trait Variables) the model when starting the simulation. In within-host modeling
of antibiotic resistance; this could be an initial deposit of a user-
defined number of bacteria and/or tissue-resident immune cells
having the state (and traits) defined in step 2. To ensure our model
is fully representative of possibilities that could arise, simulations
should be carried out many times. So, we must also decide if we
intend to vary the initial values or use the same values for all
simulations.
2.6 Defining The use of appropriate parameter values during the simulation of
Parameter Input the IBM is as significant as the model design and construction. This
is because the overall outcome of the bacterial infection under
investigation will be in response to the parameter input of the
model constructed. Hence, the use of accurate parameter values
ensures we simulate the model with biologically relevant outcomes.
The choice of parameter inputs is often driven by experimental
Algorithm 1: An example of a model design with particular assumptions and simplifications.
Require: initial deposit of user-defined bacteria in a grid cell in the environment
if oxygen conc. is surplus then
update bacteria state from slow- to fast-growing l> State change
else if oxygen conc. is deficit then
update bacteria state from fast- to slow-growing l> State change
while replication of bacteria takes place do
recruitment of active neutrophils through blood vessels l> At a specific rate
if active neutrophils ←→ bacteria then
l> Phagocytosis of bacteria, followed by bacteria mortality at a specific rate
update bacteria state to dead l> State change
end if
end while
recruitment of resting macrophages (Mr ) through blood vessels l> At a specific rate
if chemotactic signal = 0 then
healthy Mr moves randomly. l> At a specific rate
Mr ←→ bacteria l> Phagocytosis/inhibition of bacteria growth
update Mr trait from healthy to infected/chronically infected l> Trait change
else if chemotactic signal /= 0 then
Ma moves chemotactically l> At a specific rate
Ma ←→ bacteria l> Mortality of bacteria at a specific rate
end if
while replication of bacteria persist do l> User-defined bacteria
density recruitment of active T cells through blood vessels l> At a specific
rate if active T cells ←→ Mr then l> Activation at a specific
rate update Mr to Ma l> State change
else if active T cells ←→ bacteria then
l> Mortality of bacteria at a specific rate
else if active T cells ←→ infected/chronically infected macrophages then
l> Mortality of bacteria and macrophages at different rates
update macrophages and bacteria states to dead l> State change
Fig. 3 (a): An example of an algorithm that could be adapted when modeling the evolution of antibiotic
resistance. (b): An example of a model design with particular assumptions and simplifications made. The
dashed lines with bullet points indicate activities/actions being carried out. The solid lines with arrows show
the outcome following the activity and the result of the interactions between individuals after some specific
actions
end if
recruitment of generic cells through blood vessels l> At a specific rate
if generic cells ←→ bacteria then l> Mortality at a specific rate
end if
end while
while replication of bacteria persist do l> User-defined bacteria
density supply antibiotic(s) l> At user-defined
time
if antibiotic(s) conc. ≥ MIC then
inhibit bacteria growth.
else if antibiotic(s) conc. < MIC then
l> For a particular duration of time
update bacteria state from drug-susceptible to resistant
l> Proportion of mutants at a specific rate (behavioral change)
else if antibiotic(s) conc. ≥ MBC then
kill bacteria l> At a specific rate
end if
end while
end while
end while
end if
Fig. 3 (continued)
Initial deposit of user-defined drug-susceptible

Replication bacteria (fast and slow-growing)
Yes No
Is oxygen concentration
surplus?
Update bacteria state from Update bacteria state from

slow-to fast-growing fast- to slow-growing
Replication of bacteria Yes
persist (at user-defined
Neutrophils move
rates)? Recruit active neutrophils
chemotactically and
Recruitment through the blood vessels Phagocytosis
interact with bacteria
at a specific rate
at a specific rate
Recruit resting (healthy)
Cytotoxicity
macrophages through the
Resting (healthy)
blood vessels at a specific Update bacteria state to
macrophages move
rate dead
randomly and interact
with bacteria
Yes
Update macrophages state Phagocytosis
Is chemotactic signal = 0?
Replication of bacteria from resting to active
rates)? No
Yes Activation Update macrophages trait
Active macrophages move
from healthy to
chemotactically and infected/chronically infected
Active T cells move
interact with bacteria (kills
randomly and interact
bacteria at a specific rate)
with resting (healthy)
Recruitment macrophages
Cytotoxicity Update bacteria state to

dead
Recruit active T cells through
the blood vessels at a
specific rate
Active T cells move Active T cells interact with

randomly and interact with infected/chronically infected Update bacteria and
bacteria (kills bacteria at a macrophages (kills bacteria Cytotoxicity macrophages state to dead
specific rate) and macrophages at a specific
rates)
Cytotoxicity Update bacteria state to

Replication of bacteria dead
rates)?
Yes Recruit other generic cells
through the blood vessels at
a specific rate
Recruitment
Supply antibiotic(s) at Antibiotic Concentration >= MIC

Replication of bacteria Yes specific times Yes No
rates)?
Growth Adaptive
Inhibition bahaviour
Antibiotic Concentration >= MBC Inhibit bacteria growth Update bacteria behaviour from
drug-susceptible to resistant
Kill drug-susceptible bacteria
Toxicity (at different rates for slow-and Update bacteria state to
fast-growing) dead
Fig. 3 (continued)
literature, or it could be estimated. To ensure the reproducibility of

results, the parameter values used should be clearly stated (ideally
using a table) and referenced.
2.7 Decision In an IBM simulation, the length of the simulation may be long
on Result Display with short time steps, hence we need to decide the interval at which
the output should be captured. With this, it will be possible to see
how the system evolves (e.g., the emergence of resistant strains
from an original population of drug-susceptible bacteria), the activ-
ities carried out by individuals over time (i.e., the process outputs
such as replication, growth inhibition, death), and the emergence
that arise from the behavior of individuals as well as the environ-
mental attributes. In addition to displaying the snapshots of the
outcome of the simulation over space and time, due to the inherent
stochasticity in biological processes, to get an accurate representa-
tion of the system, the simulations should be run many times, after
which the mean values and the variances could be plotted to see the
summary outcomes.
Following these steps and providing suitable and well-thought
answers to the questions posed is a basis for the development of a
good IBM, which will make implementation and execution easier.
In addition, if these steps are followed appropriately, we expect that
the location of the individuals in our model, the individual and
environmental attributes, and the spatial relationships that occur
over time to be captured.
Although there is a wide variety of software environments for
IBMs such as Netlogo (http://ccl.northwestern.edu/netlogo),
MASON (https://cs.gmu.edu/~eclab/projects/mason/), Agents.jl
(https://juliapackages.com/p/agents), however, we would encour-
age the use of object-oriented programming (OOP) languages such
as Julia, C++, Python, and Java. There are numerous advantages of
OOP, including the facilitation of code maintenance and reusability,
easier troubleshooting and collaborative development, flexibility
through polymorphism, and many more.
References
1. Macal CM (2016) Everything you need to 4. Bowness R, Chaplain MA, Powathil GG, Gille-
know about agent-based modelling and simu- spie SH (2108) Modelling the effects of bacte-
lation. J Simul 10:144–156 rial cell state and spatial location on
2. Bozic I, Antal T, Ohtsuki H, Carter H, Kim D, tuberculosis treatment: insights from a hybrid
Chen S et al (2010) Accumulation of driver and multiscale cellular automaton model. J Theor
passenger mutations during tumor progres- Biol 446:87–100
sion. Proc Natl Acad Sci 107:18545–18550 5. Strydom N, Gupta SV, Fox WS, Via LE,
3. Wilensky U, Rand W (2015) An introduction Bang H, Lee M et al (2019) Tuberculosis
to agent-based modeling: modeling natural, drugs’ distribution and emergence of resistance
social, and engineered complex systems with in patient’s lung lesions: a mechanistic model
NetLogo. MIT Press and tool for regimen and dose optimization.
PLoS Med 16:e1002773
6. Hult C, Mattila JT, Gideon HP, Linderman JJ, 19. Dancik GM, Jones DE, Dorman KS (2010) An
Kirschner DE (2021) Neutrophil dynamics agent-based model for leishmania major
affect mycobacterium tuberculosis granuloma infection. In: Unifying themes in complex sys-
outcomes and dissemination. Front Immunol tems. Springer, pp 243–250
12:712457 20. Whitlock AO, Juliano JJ, Mideo N (2021)
7. Joslyn LR, Linderman JJ, Kirschner DE (2022) Immune selection suppresses the emergence
A virtual host model of mycobacterium tuber- of drug resistance in malaria parasites but facil-
culosis infection identifies early immune events itates its spread. PLoS Comput Biol 17:
as predictive of infection outcomes. J Theor e1008577
Biol 539:111042 21. Rowlatt CF, Chaplain MA, Hughes DJ, Gille-
8. Feder AF, Pennings PS, Hermisson J, Petrov spie SH, Dockrell DH, Johannessen I et al
DA (2019) Evolutionary dynamics in (2022) Modelling the within-host spread of
structured populations under strong popula- sars-cov-2 infection, and the subsequent
tion genetic forces. G3: genes. Genom Genet immune response, using a hybrid, multiscale,
9:3395–3407 individual-based model. part i: Macrophages.
9. Monaco H, Liu KS, Sereno T, Deforet M, Tay- bioRxiv
lor BP, Chen Y et al (2022) Spatial-temporal 22. Kirschner DE, Hunt CA, Marino S, Fallahi-
dynamics of a microbial cooperative behavior Sichani M, Linderman JJ (2014) Tuneable res-
resistant to cheating. Nat Commun 13:1–11 olution as a systems biology approach for
10. Railsback SF, Grimm V (2019) Agent-based multi-scale, multi-compartment computa-
and individual-based modeling: a practical tional models. Wiley Interdiscip Rev Syst Biol
introduction. Princeton University Press Med 6:289–309
11. Anderson AR, Quaranta V (2008) Integrative 23. Radchuk V, Kramer-Schadt S, Berger U,
mathematical oncology. Nat Rev Cancer 8: Scherer C, Backmann P, Grimm V (2021) Indi-
227–234 vidual- based models. Demographic methods
12. West J, Robertson-Tessi M, Anderson AR across the tree of life. Oxford University Press,
(2022) Agent-based methods facilitate integra- p 213
tive science in cancer. Trends Cell Biol 33:300– 24. Breckling B (2002) Individual-based model-
311 ling potentials and limitations. Sci World J 2:
13. Segovia-Juarez JL, Ganguli S, Kirschner D 1044–1062
(2004) Identifying control mechanisms of 25. Marino S, Hogue IB, Ray CJ, Kirschner DE
granuloma formation during m. tuberculosis (2008) A methodology for performing global
infection using an agent-based model. J Theor uncertainty and sensitivity analysis in systems
Biol 231:357–376 biology. J Theor Biol 254:178–196
14. Wong EA, Joslyn L, Grant NL, Klein E, Lin 26. Renardy M, Hult C, Evans S, Linderman JJ,
PL, Kirschner DE et al (2018) Low levels of Kirschner DE (2019) Global sensitivity analysis
t cell exhaustion in tuberculous lung granulo- of biological multiscale models. Curr Opin
mas. Infect Immun 86:e00426–e00418 Biomed Eng 11:109–116
15. Joslyn LR, Renardy M, Weissman C, Grant 27. Hamis S, Stratiev S, Powathil GG (2021)
NL, Flynn JL, Butler JR et al (2021) Temporal Uncertainty and sensitivity analyses methods
and spatial analyses of tb granulomas to predict for agent- based mathematical models: an
long-term outcomes. In: Complex systems and introductory review. In: The physics of cancer:
computational biology approaches to acute research advances, pp 1–37
inflammation. Springer, pp 273–291 28. Lipworth S, Hammond RJH, Baron VO,
16. Joseph IM, Kirschner D (2004) A model for Hu Y, Coates A, Gillespie SH (2016) Defining
the study of helicobacter pylori interaction with dormancy in mycobacterial disease. Tuberculo-
human gastric acid secretion. J Theor Biol 228: sis 99:131–142
55–80 29. Pitcher MJ, Bowness R, Dobson S, Eftimie R,
17. Weathered C, Pennington K, Escalante P, Pie- Gillespie SH (2020) Modelling the effects of
naar E (2022) The role of biofilms, bacterial environmental heterogeneity within the lung
phenotypes, and innate immune response in on the tuberculosis life-cycle. J Theor Biol
mycobacterium avium colonization to infec- 506:110381
tion. J Theor Biol 534:110949 30. Auchincloss AH, Garcia LMT (2015) Brief
18. Zarrabi N, Mancini E, Tay J, Shahand S, Sloot introductory guide to agent-based modeling
PM (2010) Modeling hiv-1 intracellular repli- and an illustration from urban health research.
cation: two simulation approaches. Procedia Cad Saude Publica 31:65–78
Comput Sci 1:555–564
Chapter 11
Monitoring Live Mycobacteria in Real-Time

Using a Microfluidic Acoustic-Raman Platform
Mingzhou Chen, Vincent Baron, Björn Hammarström,
Robert J. H. Hammond, Peter Glynne-Jones, Stephen H. Gillespie,
and Kishan Dholakia
Abstract
Tuberculosis (TB) is the most common cause of death from an infectious disease. Although treatment has
been available for more than 70 years, it still takes too long and many patients default risking relapse and the
emergence of resistance. It is known that lipid-rich, phenotypically antibiotic-tolerant, bacteria are more
resistant to antibiotics and may be responsible for relapse necessitating extended therapy. Using a micro-
fluidic system that acoustically traps live mycobacteria, M. smegmatis, a model organism for M. tuberculosis
we can perform optical analysis in the form of wavelength-modulated Raman spectroscopy (WMRS) on the
trapped organisms. This system can allow observations of the mycobacteria for up to 8 h. By adding
antibiotics, it is possible to study the effect of antibiotics in real-time by comparing the Raman fingerprints
in comparison to the unstressed condition. This microfluidic platform may be used to study any microor-
ganism and to dynamically monitor its response to many conditions including antibiotic stress, and changes
in the growth media. This opens the possibility of understanding better the stimuli that trigger the lipid-
rich downregulated and phenotypically antibiotic-resistant cell state.
Key words Raman spectroscopy, Tuberculosis, Dormancy, Single cell, Antibiotic resistance, Real-time
monitoring, Rapid diagnosis
1 Introduction
Bacteria change constantly as they respond to the physiological

conditions within the host, immune attack, and antibiotic action.
All of these are of considerable interest to microbiologists and
clinicians alike. Most approaches involve repetitive sampling and
experimental methods can change the organisms reducing the rele-
vance of the results obtained. Moreover, point sampling provided a
mere snapshot whereas we are often interested in changes in the
composition of microorganisms continuously over time. Such con-
https://doi.org/10.1007/978-1-0716-3981-8_11,
109
110 Mingzhou Chen et al.
tinuous measurements can provide important insights into the

organism’s pathogenesis, its resistance profile, the risk of resistance,
its ability to survive in a challenging environment, or its interaction
with other microorganisms. The procedure that is described in this
chapter provides way to capture, hold bacteria and to interrogate
using a method that does not, itself, change the cell state of the
bacteria.
Although most bacteriological methods depend on discrete
sampling, many technologies use real-time monitoring of bacterial
populations to indicate when sampling should occur. The methods
employed include fluorescent oxygen sensor in BACTEC MGIT
[1], electrical sensors [2], and continuous measurement of the
biomass [3]. However, these surrogate markers only give a limited
view of changes in a complex cell content and provide little insight
into the underlying physiological processes that underpin growth
and development or response to stresses.
In the case of the hollow fiber, a tool used to mimic pharmaco-
logical exposure to antibiotics must be samples at given time points.
This sample-intense approach does allow the physiological state of
bacteria to be matched to the concentration of antibiotics to which
they are exposed [4–6]. Sequential sampling allows bacterial gene
expression to be measured in a discontinuous way. This typically
changes the culture and is usually destructive. These techniques,
therefore, provide only intermittent snapshots of variables that
change continuously over time.
Acoustic trapping can produce and maintain suspended clusters
of bacteria [7]. This opens the possibility of monitoring a viable
population of suspended bacteria over time, and to probe their
response to stresses, including drugs and a changing environment.
In contrast to optical trapping, acoustic trapping levitates cells over
periods of many hours or weeks, with little heating or impact on cell
viability [8]. It may be simply implemented using piezoelectric
transducers, operating typically at megahertz frequencies. Impor-
tantly, a suitably designed trap can also facilitate real-time optical
interrogation of the trapped bacteria.
In this technique, we couple acoustic trapping with Raman
spectroscopy, which means that we can collect biochemical infor-
mation from bacteria over time in a label-free, all-optical manner.
Importantly, this is less likely to change the physiology of the
trapped bacteria in a significant way. We have shown previously
that wavelength-modulated Raman (WMR) spectroscopy is a
promising non-destructive methodology to study myco-bacterial
cell content for cells plated onto coverslips [9]. Raman measure-
ments of small samples typically include a background signal cre-
ated by the surface upon which they reside. A key advantage of the
approach we describe here is the significant reduction in back-
ground signal achieved by holding the bacterial sample away from
Monitoring Live Mycobacteria in Real-Time Using a Microfluidic Acoustic. . . 111
device surfaces with acoustic levitation. This creates a new platform

for real-time interrogation of bacteria and allows the biochemical
effect of dynamic changes in nutrients and antibiotics to be studied.
Acoustic trapping has been used previously to trap microparti-
cles [10], levitated droplets [11], and infrared spectroscopy of
ultrasonically trapped particles in bioreactors has been successful
[10]. Trapping bacteria, which are much smaller particles, in con-
tinuous flow systems, is challenging due to competition with acous-
tic streaming forces [12]. More recently, automatic sorting of
isotopically labeled microbial cells was demonstrated with a combi-
nation of microfluidics, optical tweezers, and Raman
spectroscopy [13].
We have used the technique described in this chapter to address
an important clinical question: how can we improve tuberculosis
therapy? The main challenge to shortening antituberculosis therapy
is the problem of relapse which occurs long after the patient has
been rendered culture-negative and is not fully understood [14–
16]. There is increasing evidence supporting the clinical signifi-
cance of the presence of lipid inclusions or lipid bodies in
M. tuberculosis cells [17]. A higher risk of poor TB treatment
outcome correlates with a higher proportion of mycobacteria with
lipid inclusions in patients’ sputum after 3 and 4 weeks of treatment
[18]. It is thought that lipid-rich mycobacteria with intracellular
inclusions of non-polar lipids can be up to 40 times more resistant
to first-line antibiotics compared with lipid-poor mycobacteria
(those with an absence of intracellular inclusions of non-polar
lipids) [19]. Since current susceptibility testing is based on cells
that have a low proportion of lipid body-positive cells the results
generated in clinical laboratories may not represent the minimal
inhibitory concentration at the site of the lesion where stresses that
are associated with the presence of lipid bodies are present. Thus,
this phenomenon may play an important role in patients’
relapse [20].
We have used the system described to study mycobacterial lipid
bodies in real-time in the presence and absence of antibiotic stress
[21]. Using this technique, it is possible to target a given bacterial
cell population and observe the effect of culture conditions and
drugs on the cells in real-time allowing a number of complex
questions in bacteriology to be addressed, such as the response to
chemotherapy or differing nutritional components. The Raman
spectra provide detailed qualitative and quantitative information
of important chemical components of the bacteria cells, notably
lipids and nucleic acids, based on their unique vibrational charac-
teristics (fingerprints). Although we have presented a method here
for examination of mycobacteria there is no reason in principle why
this technique could not be applied to any other bacterial species
capable of growing in the bacterial system. Similarly, although we
have used isoniazid as our model antibiotic it could be applied to

any other antimicrobial. The microfluidic system could also be
adjusted to investigate non-antibiotic stresses such as changes in
pH or differences in the nutritional content of media opening a
range of new research opportunities.
2 Materials
2.1 Bacteria, Culture 1. Middlebrook 7H9 broth and agar powder (Sigma-Aldrich UK)
Conditions, and was used to make sterile broth and plates.
Contamination Control 2. Middlebrook 7H11 (Sigma Aldrich, UK).
3. The medium should be supplemented with Tween 80 (0.05%
(v/v), Fischer Scientific) and with 2 mL of glycerol for 450 mL
of 7H9 broth.
4. M. smegmatis (NCTC 8159) National Type Culture Collection
(UK Health Security Agency) (see Note 1).
5. Benchtop centrifuge that accepts 20 mL containers and spins at
approximately (find 4500 g).
6. Phosphate Buffered saline.
7. Brain Heart Infusion Agar (Oxoid).
8. Isoniazid (Sigma-Aldrich).
9. Virkon solution.
10. Sterile 96-well plates (Nunc, DK).
11. Incubator (add details).
12. Spectrophotometer (Biochrom UK).
13. 96-well plate reader with the ability to detect at 600 nm absor-
bance (BMG Labtech).
2.2 Microfluidic 1. A microfluidic chamber and acoustic trap system is constructed

Chamber as illustrated in Fig. 1. A channel height ~ 120 μm was laser-cut
into double-sided transfer taper (3 M 9629PC). This is then
bonded on one side by a 1-mm-thick quartz glass plate
(25 × 25 mm, SPI Supplies) into which two 1-mm holes are
drilled through it for fluidic inlet/outlet ports.
2. Aquartz coverslip (150-μm-thick) forms the bottom surface
(and the acoustic reflector layer) of the device.
3. A transparent transducer was formed from a 400-μm-thick
piece of z-cut lithium-niobate (Roditi International Corpora-
tion Ltd) with 200-nm indium thin oxide (ITO) electrodes
deposited in-house.
4. The back electrode is “wrapped around” with silver conductive
paint to a section of the front electrode that had been scored to
electrically isolate it.
Fig. 1 Outline of the overall arrangement of the acoustic trap and application of the Raman laser
5. Silver epoxy is used to make the electrical connection to both

terminals.
6. Attach the transducer to the 1-mm quartz glass plate using
epoxy (Epotek 301).
7. Fluidic connections are made to the ports by attaching a laser-
cut acrylic mounting plate (with double-sided transfer tape) to
hold a short length of silicone tubing, into which PTFE (poly-
tetrafluoroethylene) tubing was pushed with a friction fit.
2.3 Raman 1. We use a Raman system similar to that described in our previ-
Spectroscopy System ous work (refs) with the exception that the fluorescence imag-
ing function was not required. The various components of the
system are noted below.
2. A Nikon microscope (Nikon TE2000-E).
3. A tunable Ti:Sa laser (M2 SolsTis lasers, 1 W@785 nm).
4. A spectrometer system (Andor Shamrock SR303i + Newton
camera).
2.4 An 1. It is critical to maintain a stable environment for the Raman

Environmental Control experiments.
Enclosure 2. To achieve this, we enclosed the Raman setup (See Diagram) in
a customized manufactured enclosure (see Note 1). The tem-
perature inside the enclosure was controlled by an incubator
Temperature Gauge (Solent Scientific).
2.5 Precision Syringe 1. Medium is pumped into the chamber at a very slow but con-
Pump trolled speed into the microfluidic chamber by a precision
syringe pump (AL2000, World Precision Instruments), due
to the small volume of the chamber.
3 Methods
3.1 Determining MIC For experiments depending on antibiotic stress it is essential to

establish an accurate MIC as set out below.
1. Inoculate a sample of M. smegmatis (NCTC 8159) taken from
glycerol stock and cultivated for 3–7 days in Middlebrook 7H9
(Sigma Aldrich, UK) and on Middlebrook 7H11 (Sigma
Aldrich, UK) and culture until it achieves a reading 0.01
OD600. (see Note 2).
2. Starting from a stock solution of Isoniazid 1280 μg·mL-1 make
doubling dilutions eightfold to 10 μg·mL-1 (8× serial
dilutions).
3. Put an aliquot of 20 μL of each of these solutions in triplicate
into a sterile 96-well plate. Then put 6 wells containing sterile
water alongside (negative & positive controls) for a total of
30 wells. 160 μL of sterile Middlebrook 7H9 media was added
to each of the 30 wells.
4. Put 20 μL of M. smegmatis culture into 27 of the wells except-
ing the negative control wells which makes a final concentra-
tion estimated to be 5 × 105 cfu/mL. To the negative control
wells add another 20 μL of sterile Middlebrook 7H9 (see Note
3).
5. With all the components added the total volume of each well
should be 200 μL and the INH added had been diluted 10×
making an appropriate range of final concentrations.
6. A 1:10 dilution of the original M. smegmatis stock is further
diluted 8× serially and plated onto a Middlebrook 7H11 agar
plate for cfu counting in the Miles and Misra fashion [22]. This
was to confirm the cfu estimate made previously.
7. After 3–5 days, or until confluent growth is evident in the
positive control wells and no growth was seen in the negative
control wells, the 96-well plate was removed from the incuba-
tor. The plate was then analyzed in a 96-well plate reader at
600 nm absorbance. These steps were repeated 4 times to yield
a total n number of 12 for each dilution of INH. After
3–5 days, or until confluent growth was evident in the positive
control wells and no growth was seen in the negative control
wells, the 96-well plate was removed from the incubator. The
plate was then analyzed in a 96-well plate reader at 600 nm
absorbance. These steps were repeated 4 times to yield a total n
number of 12 for each dilution of INH. The MIC was found to
be between 32 μg·mL-1 with 32 μg·mL-1 yielding the MIC90
and 16 μg·mL-1 yielding the MIC50 (see Note 4).
3.2 Optimization of To obtain consistent results it is essential that the experimental

Experimental parameters are optimized as this will provide a stable trap, a good
Parameters Raman signal-to-noise ratio and a controlled temperature.
1. It is necessary to investigate the amplitude of the acoustic
trapping on the temperature as this may alter both the growth
rate of the studied organism and metabolism. Set the tempera-
ture to 35.5 °C inside the system and measured by sensors
located on top of the chamber (see Note 5).
2. The bacterial concentration also has to be high as more bacteria
facilitate and reinforce the trapping. A stable trap is achieved by
using 4 mL of 7-day-old M. smegmatis re-suspended, after
centrifugation and removal of the supernatants, in 500 μL of
medium (8 times concentrated). Trap the bacteria using 7 Vpp
for 2–3 min, generating large aggregates of bacteria then
reduce the amplitude to 3 Vpp to maintain the trap throughout
the experiment and the Raman measurement. This helps to
create a stable trap as once the bacteria were aggregated a
lower trapping force was needed to maintain the trap. The
laser power must be optimized to not disturb the trapping. In
our experiments, bacterial concentration ranged from
3.1 × 108 CFU mL-1 to 1.0 × 109 CFUmL-1 (see Note 6).
3. Use a resonance frequency of 8.07 MHz (frequency deter-
mined by observation of test bead movement across a range
of test frequencies), a half-wavelength standing wave is setup in
the channel below the transducer, causing particles to be both
levitated at the channel half-height (against gravity) and
trapped in the lateral direction against flow. The acoustic pres-
sure amplitude inside the capillary for a given drive voltage is
estimated by balancing the weight of a 10-μm fluorescent
polystyrene bead against the acoustic radiation force in the
manner described previously (see Note 7). This forms a thin
layer of bacteria levitated in the center of the microfluidic
channel.
4. Control the laser power and the temperature in the chamber
monitoring the critical parameters, every hour, for all experi-
ments and time points to ensure that the power remains close
to the initial value set at T = 0 h (in this case close to 100 mW).
3.3 Raman 1. This is the critical part of the experimental process when Raman
Spectroscopy of spectra are taken from the test organism under differing con-
Trapped Bacteria ditions over time. Careful consideration should be taken as to
how the conditions are to be changed. In the instance
described here, we can envisage following the changes in the
trapped culture over time when there is no stress. This is
important to ensure that the experimental setup is not
contributing artifactual changes. In the demonstration experi-

ment noted here the bacterial culture is exposed to different
concentrations of an antibiotic (in this case isoniazid). With
minor modifications, the antibiotic can be changed to study the
effects of differing mechanisms of action.
2. The Raman spectrum of the bacteria is measured at different
time points over a period of 8 h during which both the temper-
ature and the laser power are controlled and monitored
every hour.
3. The WMR spectra from three biological replicate experiments
should be averaged for each time point and for each physiolog-
ical condition.
4. These data can be graphed to demonstrate the dramatic
changes that occur during the time course of the
experiment [21].
5. For our experiments on M. smegmatis the analysis of spectra
was focused in the fingerprint region between 600 cm-1 and
1800-1. With different organisms and different applied stres-
ses, other parts of the spectrum would become more relevant.
6. Assign the Raman peak positions as shown in Table 1 (see
Note 8).
Table 1
Raman peaks with their assignments and overall trends using three experiments for each condition
over time (up to 8 h of measurement)
Raman peak (cm-1) Associated chemical bonds Potential Raman peak assignments
635 C—C, C—S Tyrosine
783 Nucleic acids
1007 C—C Phenylalanine
1040 Carbohydrates
1080 C—C, C—O—H, C—N, C—O Lipids, carbohydrates, proteins
1130 C—N, C—C, C—O Carbohydrates, proteins, lipids
1150 C—C Carotenoids
1303 CH2 Lipids
1443 CH2, CH3 Lipids, proteins
1523 C〓C Carotenoids
1606 C〓C Tyrosine, phenylalanine
1658 C〓C, C〓O Lipids, proteins, amide I
1750 C〓O Lipids
4 Notes
1. The temperature must be close to 37 °C throughout the

experiments as it is the optimal growth temperature for
M. smegmatis. Experimenters should, however, adjust the envi-
ronmental temperature to adapt to the growth characteristics
of the organism studied.
This was to confirm the cfu estimate made previously.
2. Working cultures should be taken from a single source, directly
from a stock grown from the original type of culture or defined
clinical isolate that is the focus of the investigation. This source
should be used thereafter without further sub-culturing steps
to reduce the risk of further variability.
3. With all the components added the total volume of each well is
200 μL. As the INH is added at 10× the final concentration
range is between 1 and 128 mg/L.
4. The organisms studied must be controlled for purity by
sub-culture on, for example, brain heart infusion (BHI) agar
plates as in the case of M. smegmatis. The bacterial suspensions
should also be controlled for purity after the experiment by
plating on, for example, BHI agar. The content of the control
plates will need to be adjusted to accommodate the growth
requirements of the organism studied. Between experiments,
the chamber was cleaned and incubated with a Virkon solution
and rinsed using sterile phosphate-buffered saline (PBS).
5. Figure 1 shows an example of the relationship between the
amplitude in Vpp and the temperature in degree Celsius. It is
important to use the lowest Vpp possible to minimize the
trapping force on bacteria as this may create unexpected
changes. Our experience suggests that 3 Vpp produces a stable
trap and an increase in the observed temperature by
1.65 ± 0.19 (1 SD)°C.
6. With a laser power of 100 mW targeting trapped cells as
described above we observed that the laser force applied did
not break the trap and permitted to acquire wavelength-
modulated Raman (WMR) spectroscopy measures. The laser
itself induced a small temperature increase of about 0.5 °C. The
acquisition time was optimized using those conditions to
obtain good quality Raman spectra S12 with good signal-to-
noise ratio. A 50 s per spectrum of acquisition time was vali-
dated and produced reproducible results.
7. Acoustic pressure is related to the drive voltage applied to the
transducer by a factor of 18.2 kPa/Vpp ± 30%. Initial trapping
of the bacteria (2–3 min) is achieved with a pressure amplitude
of 128 kPa (7 Vpp). This enabled formation of a thin layer of
bacteria levitated in the center of the channel. Once the aggre-

gate formed, stabilized by secondary forces [21], it was found
possible to reduce the amplitude to 54.6 kPa (3 Vpp) thereaf-
ter, which helped reduce acoustic streaming. The time between
the re-suspension in fresh broth and the first Raman spectrum
was kept as short as possible (<5 min).
8. Using this approach it is possible to see an evolution of Raman
peaks over time in no-stress and INH-stress conditions Raman
peaks at 635 cm-1 and 1606 cm-1.
References
1. Tortoli E, Cichero P, Piersimoni C et al (1999) microparticles in suspension and reaction mon-
Use of BACTEC MGIT 960 for recovery of itoring using Raman microspectroscopy. Anal
mycobacteria from clinical specimens: multi- Chem 79:7853–7857
center study. J Clin Microbiol 37:3578–3582 11. Tuckermann R, Puskar L, Zavabeti M et al
2. Zhang X, Jiang X, Yang Q et al (2018) Online (2009) Chemical analysis of acoustically levi-
monitoring of bacterial growth with an electri- tated drops by Raman spectroscopy. Anal Bioa-
cal sensor. Anal Chem 90:6006–6011 nal Chem 394:1433–1441
3. Godin M, Delgado FF, Son S et al (2010) 12. Hammarström B, Laurell T, Nilsson J (2012)
Using buoyant mass to measure the growth of Seed particle-enabled acoustic trapping of bac-
single cells. Nat Methods 7:387–390 teria and nanoparticles in continuous flow sys-
4. Gumbo T, Louie A, Deziel MR et al (2004) tems. Lab Chip 12:4296–4304
Selection of a moxifloxacin dose that sup- 13. Lee KS, Palatinszky M, Pereira FC et al (2019)
presses drug resistance in Mycobacterium An automated Raman-based platform for the
tuberculosis, by use of an in vitro pharmacody- sorting of live cells by functional properties.
namic infection model and mathematical mod- Nat Microbiol 4:1035–1048
eling. J Infect Dis 190(1642):1651 14. Gillespie SH, Crook AM, McHugh TD et al
5. Gumbo T, Dona CSWS, Meek C et al (2009) (2014) Four-month moxifloxacin-based regi-
Pharmacokinetics-pharmacodynamics of pyra- mens for drug-sensitive tuberculosis. New
zinamide in a novel in vitro model of tubercu- Engl J Med 371:1577–1587
losis for sterilizing effect: a paradigm for faster 15. Merle CS, Fielding K, Sow OB et al (2014) A
assessment of new antituberculosis drugs. Anti- four-month gatifloxacin-containing regimen
microb Agents Chemother 53(3197):3204 for treating tuberculosis. New Engl J Med
6. Gumbo T, Angulo-Barturen I, Ferrer-Bazaga S 371:1588–1598
(2015) Pharmacokinetic-pharmacodynamic 16. Jindani A, Harrison TS, Nunn AJ et al (2014)
and dose-response relationships of Antituber- High-dose rifapentine with moxifloxacin for
culosis drugs: recommendations and standards pulmonary tuberculosis. New Engl J Med
for industry and academia. J Infect Dis 211: 371:1599–1608
S96–S106 17. Mukamolova GV, Mukamolova GV, Turapov
7. Gutiérrez-Ramos S, Hoyos M, Ruiz-Suárez JC O et al (2010) Resuscitation-promoting factors
(2018) Induced clustering of Escherichia coli reveal an occult population of tubercle bacilli in
by acoustic fields. Sci Rep 8:4668 sputum. Am J Resp Crit Care 181(174):180
8. Bazou D, Kearney R, Mansergh F et al (2011) 18. Sloan DJ, Mwandumba HC, Garton NJ et al
Gene expression analysis of mouse embryonic (2015) Pharmacodynamic modeling of bacil-
stem cells following levitation in an ultrasound lary elimination rates and detection of bacterial
standing wave trap. Ultrasound Med Biol 37: lipid bodies in sputum to predict and under-
321–330 stand outcomes in treatment of pulmonary
9. Baron VO, Chen M, Clark SO et al (2017) tuberculosis. Clin Infect Dis 61:1–8
Label-free optical vibrational spectroscopy to 19. Hammond RJH, Baron VO, Oravcova K et al
detect the metabolic state of M. tuberculosis (2015) Phenotypic resistance in mycobacteria:
cells at the site of disease. Sci Rep 7:9844 is it because I am old or fat that I resist you? J
10. Ruedas-Rama MJ, Domı́nguez-Vidal A, Radel Antimicrob Chemother 70:2823–2827
S et al (2007) Ultrasonic trapping of
20. Phillips PPJ, Mendel CM, Burger DA et al mycobacteria with a microfluidic acoustic-
(2016) Limited role of culture conversion for Raman platform. Commun Biol 3:236
decision-making in individual patient care and 22. Billington OJ, McHugh TD, Gillespie SH
for advancing novel regimens to confirmatory (1999) Physiological cost of rifampin resistance
clinical trials. BMC Med 14:19 induced in vitro in mycobacterium tuberculo-
21. Baron VO, Chen M, Hammarstrom B et al sis. Antimicrob Agents Chemother 43:1866–
(2020) Real-time monitoring of live 1869
Chapter 12
Phylogenetic Survival Analysis

Arturo Torres Ortiz and Louis Grandjean
Abstract
Going back in time through a phylogenetic tree makes it possible to evaluate ancestral genomes and assess
their potential to acquire key polymorphisms of interest over evolutionary time. Knowledge of this kind
may allow for the emergence of key traits to be predicted and pre-empted from currently circulating strains
in the future. Here, we present a novel genome-wide survival analysis and use the emergence of drug
resistance in Mycobacterium tuberculosis as an example to demonstrate the potential and utility of the
technique.
Key words Genome-wide survival analysis, Phylogenetics, Ancestral States, Hazard, Acquisition of
drug resistance, Evolutionary time, Traits, Phenotypes, Prediction
1 Introduction
Whole-genome sequencing has transformed our understanding of

the evolution of microorganisms. Reductions in sequencing costs
together with increased throughput have enabled the generation of
comprehensively sampled large genomic datasets that are available
in public repositories. Mycobacterium tuberculosis, the world’s sin-
gle most deadly microorganism, once thought to be genetically
monomorphic (lacking in genomic variation), has been shown to
harbor considerably more phenotypic variation than was antici-
pated [1]. This has led to significant improvements in our under-
standing of the classification, evolution, and diagnosis of drug
resistance in M. tuberculosis [2].
Genetic diversity among microbes often results in phenotypic
differences. Understanding the genetic basis of many pathogen
traits, such as drug resistance, virulence, transmission, or
immune-evasion is important for clinical management and public
health control measures. Genome-wide association studies (GWAS)
compare genetic sequences from multiple isolates and systemati-
cally search for associations between each polymorphic position in
the genome and the phenotype of interest. However, clonal
https://doi.org/10.1007/978-1-0716-3981-8_12,
121
122 Arturo Torres Ortiz and Louis Grandjean
expansion of pathogens produces a strong population structure

that confounds causal relationships between phenotype and geno-
types with shared ancestry. Moreover, genome-wide genetic linkage
often precludes the separation between causal genetic variants and
non-causal linked mutations [3]. Phylogenetic analysis can correct
for some of these confounders, as the phylogenetic tree reveals the
population structure and the evolutionary relationships between
the genotype and the phenotype, and thus they have been sug-
gested as valuable tools for pathogen GWAS [4].
M. tuberculosis does not undergo horizontal gene transfer or
intergenomic recombination [1], and it is therefore particularly
amenable to robust phylogenetic reconstruction of evolutionary
pathways and events. Both Bayesian and Maximum Likelihood
phylogenetic reconstruction methods output the most likely phy-
logenetic tree under a given model of evolution. This output
includes both the genomic sequences at the tips (the time of
sampling) and the inferred genomic sequences of the most recent
common ancestors (MRCAs).
Unlike viruses, bacteria, particularly M. tuberculosis, have a
relatively slow mutation rate [5]. As a consequence, to establish a
statistically significant time signal (a steady accumulation of muta-
tions over time) a large longitudinal sampling window is often
required [6]. Demonstrating a statistically significant time signal
in the data allows for the generation of a “time-stamped” phylog-
eny. This is a phylogenetic tree in which each ancestral state can be
dated and each branch in the tree has an associated timescale in
months or years.
Phylogenetic analysis can not only aid finding causal genetic
variants associated with a phenotype, but also elucidate the evolu-
tionary dynamics of traits of interest. For instance, independent
acquisitions of mutations (homoplasies) associated with drug resis-
tance can indicate the effects of natural selection, and have been
used to uncover causal associations with antibiotic resistance in
M. tuberculosis [7]. In this chapter, we present a phylogenetic
method to understand genotypes that do not yet present a specific
phenotype of interest but rather are more likely to acquire it in the
future. With a view to predicting and pre-empting the emergence
of future phenotypic traits of interest, we present a novel ancestral
state genome-wide survival analysis using M. tuberculosis as an
example. We used this method to identify genetic polymorphisms
in drug-susceptible genotypes that are more likely to acquire anti-
biotic resistance in the future (pre-resistance) [8]. The time
between ancestral states is considered in the framework of a survival
analysis with censorship being “no event” between nodes and an
event being the emergence of the causative polymorphism in the
descendent node. This approach can be summarized in the follow-
ing steps (Fig. 1):
Phylogenetic Survival Analysis 123
Fig. 1 Phylogenetic survival analysis workflow. The proposed approach starts with a genome alignment as
input, from which a phylogenetic tree is inferred. Different phenotypes are then determined for each isolate at
the tips of the tree. For representation, the phenotype of interest is characterized by red color. The ancestral
states for each phenotype are then inferred. To identify genetic signatures of genotypes leading to the
phenotype of interest, we select the last genotype without the phenotype (the pre-resistance node in the
example) and perform a time-to-event analysis where the time is represented by the branch length between
nodes and the event by the presence or absence of the phenotype of interest. Once the phenotype has evolved
within a branch, the rest of nodes are not counted for that clade
1. Genome alignment.
2. Phylogenetic inference.
3. Ancestral state and sequence reconstruction.
4. Survival analysis.
2 Materials
Access to high-performance computing facilities is recommended,

particularly if assembling and aligning raw genomic data prior to
creating an alignment of variant sites. Our entire bioinformatic
pipeline from raw genomic data, through genome-wide survival
analysis and presentation of results, is available via our GitHub
page (github.com/arturotorreso/mtb_pre-resistance). Phyloge-
netic and genome-wide survival analysis was undertaken using R.
2.1 Computing The phylogenetic analysis is performed using R [9]. This is a

platform-independent software and therefore can be used in any
standard computer. However, the method can be adapted to any
programming language.
2.2 Software The following R packages are recommended to run the analysis:
(a) Ape [10].
(b) Phangorn [11].
(c) BactDating [12].
(d) Survival [13].
In addition, multiple programs are required to prepare the
data for analysis and for phylogenetic inference. Recom-
mended software is mentioned in later sections.
3 Methods
The proposed workflow takes raw reads as input, and it outputs the
association between each polymorphism to the phenotype of inter-
est. All the scripts are located in our GitHub page (github.com/
arturotorreso/mtb_pre-resistance). The recommended software
and the choice of programming language are optional.
3.1 Pre-processing Before preparing the alignment, the raw reads need to be
pre-processed as outlined in the pseudoseq_pipeline.sh using Trim-
momatic in order to remove illumina adaptors and low-quality
bases [14].
3.2 Genome As genomes are compared at each genomic position, all sequences
Alignment need to be in the same genomic coordinates. Therefore, all
pre-processed reads need to be aligned to the same reference
genome. In order to increase the accuracy of the final alignment,
we recommend preparing the genome alignment using the
pre-processed reads as shown in pseudoseq_pipeline.sh by using
the following software:
(a) Spades (https://github.com/ablab/spades) [15].
First, short-reads from each isolate are assembled into
longer contigs in order to increase the probability that a contig
is properly mapped to the reference genome.
(b) Minimap2 (https://github.com/lh3/minimap2) [16].
Contigs are then mapped against the reference genome
using Minimap2. The output is in the Sequence Alignment/
Map format (SAM/BAM)
(c) SAMtools (https://github.com/samtools/samtools) [17].
The alignment can be sorted by genomic position and

processed to facilitate the following steps.
(d) BCFtools (https://github.com/samtools/bcftools) [18].
Using the alignment from the previous step, genetic var-
iants (positions at which the isolate differs from the reference
genome) are identified and filtered to remove low-confidence
variants. The output is in the Variant Calling Format (VCF),
from which a complete sequence can be derived using the
command “bcftools consensus” or the accompanying script
vcf2pseudoseq.py by adding the genetic variants to the
provided reference genome. This sequence will be used as a
sample-specific reference genome to increase the mapping
accuracy of the short-reads.
(e) BWA (https://github.com/lh3/bwa) [19].
Short-reads are aligned against the sample-specific refer-
ence genome using bwa mem.
(f) Sambamba (https://github.com/biod/sambamba) [20].
PCR and optical duplicates can be removed from the
alignment file using sambamba markdup.
(g) GATK (https://github.com/broadinstitute/gatk) [21].
Short-reads can be re-aligned around indels to increase
indel detection accuracy using GATK’s IndelRealigner. From
this post-processed alignment, a complete sequence can be
inferred using the command “bcftools consensus” or the
accompanying script vcf2pseudoseq.py as outlined before.
(h) BEDtools (https://github.com/arq5x/bedtools2) [22].
Hypervariable sites can be masked (switched to Ns) so
low-quality and uncertain regions won’t be used in down-
stream phylogenetic analysis.
3.3 Phylogenetic A phylogenetic tree can be inferred using a wide variety of methods.
Inference For this workflow, we propose using a time-calibrated phylogeny,
where the branch length is measured in calendar units. This can be
performed in one step using methods such as BEAST [23],
although this method is limited by the number of isolates in the
analysis and it can become unfeasible with a large sample size.
Alternatively, time calibration can be performed in two steps by
inferring a phylogenetic tree and then time-calibrating it using
methods such as those implemented in BactDating [12], TreeDater
[24], or LSD [25], which allow to obtain a time-calibrated phylog-
eny even in large studies. Instructions about how to run BactDat-
ing can be found at runBactDat.R. When a time-calibrated
phylogenetic tree cannot be accurately inferred, the branch length
of the tree in units of genetic distance can be used as the time
variable within the phylogenetic survival model.
3.4 Ancestral State To understand the evolutionary dynamics of the phenotype of

Reconstruction interest along evolutionary time, we infer the ancestral states of
the internal nodes of the phylogenetic tree. This is achieved both
for continuous or discrete variables using Maximum Likelihood,
Bayesian methods, or Maximum Parsimony as implemented in
multiple R packages such as Ape [10], PhyTools [26], or
Phangorn [11].
3.5 Ancestral The ancestral sequences of internal nodes of the phylogenetic tree
Sequence can be inferred using similar methods as with the ancestral state
Reconstruction reconstruction previously described (see Note 1).
To perform ancestral sequence reconstruction using Phangorn,
see Note 2.
3.6 Phylogenetic Once the ancestral reconstructions have been finalized, the survival
Survival Analysis analysis can be performed as described in the R scripts within the
GitHub repository survival_analysis.R and survTree_functions.R.
In short, they are summarized as follows:
1. The phylogenetic tree is scanned from the root to the tips.
2. For each branch, the branch length and the phenotypes of the
internal and external nodes are recorded (e.g., 0 for absence
and 1 for presence, or continuous phenotype).
3. Once a phenotype has evolved in a specific branch, subsequent
branches with no phenotype change are not considered.
The time-to-event analysis is performed within R using the
Survival package. A time-to-event model can be fit using a combi-
nation of the Surv function, which creates a survival object, and the
survfit function. Moreover, a Cox Proportional-Hazards Model can
be fitted using the coxph function to compute the instantaneous
hazard of acquiring the phenotype of interest.
Using these functions, the model can be fitted to calculate
differences in survival rates between groups. In our analysis, we
computed differences in drug resistance acquisition between clades,
or between mono-resistant clades in comparison to fully susceptible
groups.
Additionally, a genome-wide association analysis is performed
by fitting a Cox model by selecting the internal node for each
branch where a phenotype change has occurred and the remaining
nodes where the phenotype is not present. This association can be
repeated for each polymorphic site of interest, as indicated in the
scripts gwas1_alignment.R and gwas2_analysis.R.
The Survival and Cox models can be substituted by a lin-
ear model, a logistic regression model or any other suitable
model to compute the association without using the branch length
or time within the model. This can be useful to calculate odds ratios
or to run the association when the branch lengths of the tree are
unreliable.
As only the first occurrence of the phenotype is used in the

analysis, the bias due to population structure is partially corrected
for. However, further correction can be performed by adding an
eigen decomposition of a kinship matrix as covariates in the model
as described in gwas1_alignment.R and gwas2_analysis.R. The
number of eigenvalues to incorporate as covariates will depend on
the genetic structure of the dataset.
3.7 Output The final output of the analysis consists of the fitted parameters of
the survival and the Cox model, including a p-value reflecting
statistical significance of the differences between groups, both in
the case of clade differences or genotypes.
4 Notes
1. For our purposes, we employ the R package Phangorn since it

allows the user to determine ambiguous bases and additional
characters to distinguish different polymorphisms such as dele-
tions or structural variants.
2. This method is described in our GitHub repository (anc_seq_-
recons.R).
Acknowledgments
This work was supported by the Wellcome Trust (201470/Z/16/

Z), by the National Institute of Allergy and Infectious Diseases of
the National Institutes of Health under award number
1R01AI146338, and by the GOSH/ICH Biomedical Research
Centre. We would also like to acknowledge the UCL Computer
Science Technical Support Group (TSG) and the UCL Department
of Computer Science High Performance Computing Cluster.
References
1. Gygli SM, Borrell S, Trauner A, Gagneux S methods using simulated genomes and pheno-
(2017) Antimicrobial resistance in Mycobacte- types. Microb Genom 6:e000337. https://doi.
rium tuberculosis: mechanistic and evolution- org/10.1099/mgen.0.000337
ary perspectives. FEMS Microbiol Rev 41:354– 4. Collins C, Didelot X (2018) A phylogenetic
373. https://doi.org/10.1093/femsre/ method to perform genome-wide association
fux011 studies in microbes that accounts for popula-
2. Walker TM, Kohl TA, Omar SV et al (2015) tion structure and recombination. PLoS Com-
Whole-genome sequencing for prediction of put Biol 14:e1005958. https://doi.org/10.
Mycobacterium tuberculosis drug susceptibil- 1371/journal.pcbi.1005958
ity and resistance: a retrospective cohort study. 5. Gagneux S (2018) Ecology and evolution of
Lancet Infect Dis 15:1193–1202. https://doi. Mycobacterium tuberculosis. Nat Rev Micro-
org/10.1016/S1473-3099(15)00062-6 biol 16:202. https://doi.org/10.1038/
3. Saber MM, Shapiro BJ (2020) Benchmarking nrmicro.2018.8
bacterial genome-wide association study
6. Menardo F, Duchêne S, Brites D, Gagneux S 3094–3100. https://doi.org/10.1093/bioin

(2019) The molecular clock of Mycobacterium formatics/bty191
tuberculosis. PLoS Pathog 15:e1008067. 17. Li H, Handsaker B, Wysoker A et al (2009)
https://doi.org/10.1371/journal.ppat. The sequence alignment/map format and
1008067 SAMtools. Bioinformatics 25:2078–2079.
7. Farhat MR, Shapiro BJ, Kieser KJ et al (2013) https://doi.org/10.1093/bioinformatics/
Genomic analysis identifies targets of conver- btp352
gent positive selection in drug-resistant Myco- 18. Danecek P, Auton A, Abecasis G et al (2011)
bacterium tuberculosis. Nat Genet 45:1183– The variant call format and VCFtools. Bioin-
1189. https://doi.org/10.1038/ng.2747 formatics 27:2156–2158. https://doi.org/10.
8. Torres Ortiz A, Coronel J, Vidal JR et al 1093/bioinformatics/btr330
(2021) Genomic signatures of pre-resistance 19. Li H, Durbin R (2010) Fast and accurate long-
in Mycobacterium tuberculosis. Nat Commun read alignment with Burrows–Wheeler trans-
12:7312. https://doi.org/10.1038/s41467- form. Bioinformatics 26:589–595. https://
021-27616-7 doi.org/10.1093/bioinformatics/btp698
9. R Core Team, R Foundation for Statistical 20. Tarasov A, Vilella AJ, Cuppen E et al (2015)
Computing (2021) R: a language and environ- Sambamba: fast processing of NGS alignment
ment for statistical computing formats. Bioinformatics 31:2032–2034.
10. Paradis E, Claude J, Strimmer K (2004) APE: https://doi.org/10.1093/bioinformatics/
analyses of phylogenetics and evolution in R btv098
language. Bioinformatics 20:289–290. 21. McKenna A, Hanna M, Banks E et al (2010)
https://doi.org/10.1093/bioinformatics/ The genome analysis toolkit: a mapreduce
btg412 framework for analyzing next-generation
11. Schliep KP (2011) phangorn: phylogenetic DNA sequencing data. Genome Res 20:
analysis in R. Bioinformatics 27:592–593. 1297–1303. https://doi.org/10.1101/gr.
https://doi.org/10.1093/bioinformatics/ 107524.110
btq706 22. Quinlan AR, Hall IM (2010) BEDTools: a
12. Didelot X, Croucher NJ, Bentley SD et al flexible suite of utilities for comparing genomic
(2018) Bayesian inference of ancestral dates features. Bioinformatics 26:841–842
on bacterial phylogenetic trees. Nucleic Acids 23. Bouckaert R, Heled J, Kühnert D et al (2014)
Res 46:e134–e134. https://doi.org/10. BEAST 2: a software platform for bayesian
1093/nar/gky783 evolutionary analysis. PLoS Comput Biol 10:
13. Therneau TM, Grambsch PM (2000) Model- e1003537. https://doi.org/10.1371/journal.
ing survival data: extending the cox model. pcbi.1003537
Springer, New York 24. Volz EM, Frost SDW (2017) Scalable relaxed
14. Bolger AM, Lohse M, Usadel B (2014) Trim- clock phylogenetic dating. Virus Evol 3:
momatic: a flexible trimmer for Illumina vex025. https://doi.org/10.1093/ve/vex025
sequence data. Bioinformatics 30:2114–2120. 25. To T-H, Jung M, Lycett S, Gascuel O (2016)
https://doi.org/10.1093/BIOINFORMAT Fast dating using least-squares criteria and
ICS/BTU170 algorithms. Syst Biol 65:82–97. https://doi.
15. Bankevich A, Nurk S, Antipov D et al (2012) org/10.1093/sysbio/syv068
SPAdes: a new genome assembly algorithm and 26. Revell LJ (2012) phytools: an R package for
its applications to single-cell sequencing. J phylogenetic comparative biology (and other
Comput Biol 19:455–477. https://doi.org/ things). Methods Ecol Evol 3:217–223.
10.1089/cmb.2012.0021 https://doi.org/10.1111/J.2041-210X.
16. Li H (2018) Minimap2: pairwise alignment for 2011.00169.X
nucleotide sequences. Bioinformatics 34:
Chapter 13
Rapid Drug Susceptibility Testing to Preserve Antibiotics

Stephen H. Gillespie and Robert J. H. Hammond
Abstract
Antibiotic resistance is a global challenge likely to cost trillions of dollars in excess costs in the health system
and more importantly, millions of lives every year. A major driver of resistance is the absence of susceptibility
testing at the time a healthcare worker needs to prescribe an antimicrobial. The effect is that many
prescriptions are unintentionally wasted and expose mutable organisms to antibiotics increasing the risk
of resistance emerging. Often simplistic solutions are applied to this growing issue, such as a naı̈ve drive to
increase the speed of drug susceptibility testing. This puts a spotlight on a technological solution and there
is a multiplicity of such candidate DST tests in development. Yet, if we do not define the necessary
information and the speed at which it needs to be available in the clinical decision-making progress as
well as the necessary integration into clinical pathways, then little progress will be made. In this chapter, we
place the technological challenge in a clinical and systems context. Further, we will review the landscape of
some promising technologies that are emerging and attempt to place them in the clinic where they will have
to succeed.
Key words Drug susceptibility testing, Rapid diagnostics, Microbiology automation, Microbiology
systems, Limit of detection (LoD), Photonics, Genome sequencing, Polymerase chain reaction
1 Introduction
The clinical bacteriological diagnostic paradigm has remained stub-

bornly wedded to culture, although many new rapid techniques
have been described. This is often justified because of an uncon-
scious fallacy about culture: that it is likely to capture all the infect-
ing organisms and allows accurate drug susceptibility testing (DST)
to be performed. In the reality of clinical practice patients often
attend their physician and receive antibiotics, use an old prescrip-
tion, or purchase them on the grey market. This is unlikely to result
in cure but will render bacteriological samples negative at worst or
unreliable at best. Importantly, the fallacy also assumes that there is
a specimen to culture and that the patient can produce it. Even
https://doi.org/10.1007/978-1-0716-3981-8_13,
129
130 Stephen H. Gillespie and Robert J. H. Hammond
when the specimen is produced and sent for culture, loss of viability
in transit, especially true for anaerobes, results in false negative
result and no material for DST.
Thus, in many circumstances, culture is not necessarily the
optimal way of identifying the susceptibility pattern of a causative
pathogen. Although bacteria grow rapidly, dividing in as little as
once every 20 min, the high limit of detection (approx. 108 cfu/mL
Vs. circulating pathogens at ~<100 cfu/mL) means that a positive
result takes many hours to signal positive. This limitation also
applies when these systems are used to perform drug susceptibility
tests. In most circumstances defining the species of the infecting
organism will provide a strong treatment indication to the clinician
based on prior knowledge. For example, detecting the presence of
Staphylococcus aureus suggests a limited number of antibiotics that
will be useful. Recent developments using MALDI-TOF on pri-
mary cultures have proved very useful in generating a report in a
single working shift [1–3]. This is illustrated in Table 1, which
summarizes the time to result for common rapid susceptibility
testing platforms that depend on culture. But, perhaps, the greatest
drawback for many of these systems is the need to start with a pure
culture of the causative organism in most circumstances. This
means that there is usually an initial culture stage imposing a
significant delay of between 8 and 24 h. Despite the recent focus
on antibiotic stewardship, the importance of rapid diagnostics is
often overlooked in the discussions [12].
Table 1
A summary of selected phenotypic drug susceptibility test methods
Time to
positive
Technique Methodology result References
Microscan Lyophilized fluorogenic substrates and pH indicators 16 h [4]
Phoenix Detects change in redox using a colorimetric oxidation- 20 h [5]
reduction indicator
Vitek 2 Identification of biochemical reactions and nutrient usage ~10.4 h [6]
(3.3–17.5 h)
BacT- A colorimetric sensor and reflected light monitor production 17 h [7]
ALERT of carbon dioxide (CO2) that is dissolved in the culture
medium changing the color of the medium
BACTEC Increases in CO2 production due to microbial growth are 12.5 h [7]
detected by non-invasive fluorescence
VersaTREK Detects pressure changes due to gas consumption (O2) ~23.7 h [8–11]
and/or production (CO2, N2, H2, etc.) by microorganisms (11–36.4 h)
in a BC bottle.
Rapid Drug Susceptibility Testing to Preserve Antibiotics 131
2 What Do We Mean by Fast?
To answer this question, we need to consider the clinical context of

diagnosis. Out-patient practice usually requires a diagnosis to be
available within hours. The decision must be made rapidly before
the patient goes home and is beyond close supervision. In addition
to the need to identify severe infection that requires speedy inter-
vention, it is necessary to provide a therapeutic answer quickly.
Delay may mean that the patient, who has left the clinic, will need
to be recalled. This adds expense to the consultation process to
both the patient and the hospital system and patients are more likely
to default from further care.
Once a specimen needs to travel to a different site or even
different department, hours of delay are implied, making point-
of-care testing an ideal objective. Thus, we can assume that any
successful technique will be able to deliver a result at the point of
consultation. But how long should the test take from consultation
to an actionable result? What are the time limits and what are the
organizational constraints? From the point at which a specimen is
taken we would want the result to be available and actionable on the
same day. This gives a theoretical 8-h window so that the requesting
physician who has seen the case and ordered the test can interpret it
and make a therapeutic judgment. Less time than that should be
available as patients presenting later in a clinic space will also need to
be diagnosed in a timely manner. Perhaps, it might be better to
think of this window as being within a work shift, which is a 4-h
window. In this time, the specimen must be taken, prepared for
examination, tested, and the results reported. This means that a
“rapid” test must take no more than a maximum of 2 h. These are
the reasons that underpin the drive to point of care (POC) testing
[12, 13]. A “mix and match” approach is now recommended such
that healthcare providers can select tests and implementation pro-
cesses that best fit their clinical needs and best fit the technique to
the question at hand [12]. This is an important consideration as it
places the focus on the clinical need and the place of testing in the
patient pathway rather than focusing on the technology in
isolation [14].
If we examine the clinical consultation and laboratory interac-
tion more closely, there are, in effect, three questions to be
addressed, as noted in Fig. 1.
1. Is there an infection present?
If the presence of an infection can be categorically ruled out this
is of major benefit as other possible diagnoses can be explored.
Question 1
Patient Clinical No infection Alternative

presentation consultation diagnosis
Question 2
Infection with species

detection and
susceptibility testing
Infection Species Treatment
likely identity by protocola Unexpected
resistance
causes
treatment
Question 3
failure
Deliver
effective
Identify treatment
susceptibilities
Fig. 1 An overview of the interaction between clinical questions and the microbiological diagnostic process
2. What is the infecting organism?

This provides critical data. Identification often allows treatment
to begin, especially in those infections where treatment is usually
directed by local, national, or globally agreed protocol.
3. What are the organism’s susceptibilities?
This provides important infection control advice preventing
further spread of a resistant infection and enabling reporting to
health protection authorities. It is critically important that DST
continues to be performed to provide a predictive database that
can inform clinical decision-making when laboratory results are not
available and to follow trends in the evolution of resistance. This is
all secondary in the case of a critically ill patient where timely DST
will lead to life-saving targeted care.
3 Speeding the Laboratory Process
There are different broad methods to address this issue: detect the
growth of bacteria or detect a drug susceptibility characteristic
directly from the specimen. Despite the difficulties of defining
what is meant by “fast” there is almost universal agreement of the
need to speed up the current laboratory methodologies [15]. These
can be generally divided into three broad categories: detecting the
metabolic effects of growth, detecting growth directly, or detecting
growth because of novel sensors. It is important to consider that

determination of susceptibility often needs to be planned or inter-
preted in the light of the identity of the organism. This is because
innate resistance must be considered and the achievable concentra-
tions can vary between Gram-positive and -negative species. For
most purposes, international guidelines require the identification of
the organism to accurately propose the most effective antibiotic
[16, 17].
4 Indirect Detection of Bacterial Growth by Detecting Metabolic Change
Delay in culture-based systems centers on their inherent limit of

detection. Large laboratories generally have high throughput auto-
mated systems for susceptibility testing. Most of the current tech-
niques include species identification as this has an important impact
on the interpretation of susceptibility data. Here we review some of
the well-established systems (see Table 1). Of note: identification
generally occurs via agar growth from overnight cultures and fre-
quently on chromogenic agar plates. These techniques take a mini-
mum of 16 h for confidence.
4.1 The BD Phoenix The Phoenix platform determines growth by detecting both the
Automated turbidity of growth and a fluorescent oxidation-reduction indica-
Microbiology System tor. The MIC is determined by using doubling dilutions of
pre-prepared antimicrobials in manufacturer-prepared cards.
There is a degree of flexibility in the content of the cards available
that can be adapted to the antibiotic stewardship program. CLSI-
based breakpoints are employed and interpreted by an expert sys-
tem [18]. The onboard system can be customized to take account
of local reporting algorithms. It provides susceptibility testing
across a wide range of organisms encountered in routine microbio-
logical practice with a simple workflow [19–21].
4.2 Vitek 2 This system is a broth microdilution minimum inhibitory concen-

tration system that is incorporated into a self-contained card. It
detects growth using trade-marked colorimetric system to produce
species identification by detecting biochemical reactions and nutri-
ent usage. It relies on continuous turbidity measurements to deter-
mine the MIC of bacterial isolates and is calculated using multiple
analyses including raw data, growth change (%), the shape of slope,
AUC, and slope change (%). A real-time comparison is made
between the current isolate and the standard curves stored in the
Vitek 2 database [9]. Susceptibility testing results are interpreted
using its onboard expert system [18].
4.3 Microscan Microscan uses a card-based broth dilution methodology with

lyophilized fluorogenic substrates and pH indicators. This is said
to provide 10–100 times the sensitivity of conventional assays
[22–24].
Supported by large-scale equipment manufacturers, these sys-
tems are competitive and are widely used. They have been exten-
sively tested and compared with broadly similar results [19, 20, 25–
27]. Novel resistance mechanisms are constantly being evaluated
and modifications to the methodology and expert systems are made
to enable detection of the changing resistance patterns [18, 24,
27]. This consideration is especially important for routine labora-
tories as this helps to keep up to date with the changing epidemiol-
ogy of resistant organisms, availability of antibiotics, and ensuring
that adequate laboratory methods are in place to detect the chang-
ing landscape of resistance.
5 Direct Bacterial Detection
It could be argued that the reason DST takes so long is the low
sensitivity of the methods to detect and quantify the number of
bacteria present in a sample. Here we summarize some of the
innovative approaches to improve the detection and quantification
of bacteria that are being applied to testing susceptibility. Light-
based solutions (or photonics) have been seen as a way of increasing
the sensitivity of detection of bacterial growth and there is a multi-
plicity of such methods (Table 2). These can detect metabolic
changes, detect reduced or inhibited growth, detect morphological
changes, detect reduced motility, or identify drug-induced death.
5.1 Detecting Detecting metabolic changes has been used to determine suscepti-
Metabolic Components bility by monitoring metabolic changes as bacteria are exposed to
antibiotics. Raman spectroscopy in its many forms has the capacity
to interrogate the chemical components of bacteria [34]. A MIC
methodology based on this approach has been described [35]
although time to result was at least 4 h. Trapping cultures in a
microfluidic acoustic chamber and examining the culture with
wavelength-modified Raman spectroscopy could be used to moni-
tor metabolic changes in real time detecting changes at a molecular
level. This is a promising technique reviewed in this volume but
could be cumbersome to deploy in its current configuration. Its
ability to interrogate the chemistry of an organism means that it
would be able to address problematic organisms such as mycobac-
teria and could be an important methodology for further research
at this time [31].
Table 2
Summary of selected photonics drug susceptibility test methodologies
Time to
Technique positive
name Principle result Target References
Bacterioscan Forward-angle laser light scattering 3h Urinary tract [28, 29]
infection
screening
SLIC Light-scattering and an integrating 10 min UTI, rapid [30]
chamber DST
Raman Monitoring changes in Raman spectrum tbd Research [31]
spectroscopy in bacteria exposed to antibiotic technique
oCelloscope Digital time-lapse angled field <3 h DST [32, 33]
microscopy to detect changes in
number or morphology
Accelerate- Morphokinetic analysis of bacterial 2–7 ha [8–10]
pheno growth and integrated interpretation
a
The lower figure represents time identification and the latter time to AST.
5.2 Detecting Growth Forward angle light scattering is used to detect change in urine
Through Bacterial samples as a means of detecting bacteriuria. The application pro-
Scattering vides a screening system that prevents unnecessary cultures and,
potentially unneeded antibiotic prescription in children and adults
[28, 29]. This being brought to market by Astrego focused on
detection of urinary tract infection.
Another approach the methodology, “Scattered light integrat-
ing collector” (SLIC) has been used to significantly decrease the
limit of bacterial detection to as few as 25 cfu mL-1 [30]. This is
achieved by using an integrating chamber, which captures all the
light of an incident laser beam. As the number of organisms
increases through growth, the amount of scattering increases
making a highly sensitive readout of the number of particles present
in a fluid. Using this technique identifies another important
finding: it is easier to determine whether a bacterium is resistant
than it is to confirm that it is sensitive. When the strain is resistant,
the growth curve follows that of the control, whereas in the case of
susceptibility it is necessary to run the culture until it is certain that
no growth is going to take place.
5.3 Optical These methods depend on detecting a change in the bacteria or

Morphological counting the bacteria by direct optical methods. This may include
Reading Methods detecting the cell division or the volume enclosed by bacteria in
microfluidic systems. These methods, like other photonics
approaches, have the advantage of being able to monitor the
response to the addition of antibiotic in real time [32, 36].
5.4 Camera Systems Using a complex microfluidic and time-lapse camera system mea-
suring bacterial length in real time, Baltekin et al. detected the
action of antibiotics on E. coli in approximately 15 min [36] .
Digital time-lapse angled field microscopy has been applied to
bacteria growing multi-well plates in the presence and absence of
antibiotic. Changes in size or morphology are plotted in real time
with 95% of results available within 3 h [32, 33]. Choi et al. used
bacterial imaging and microfluidic systems to measure the area
occupied by immobilized microbes. Differing morphological pat-
terns could be used to provide susceptibility results in 3–4 h [37]. A
commercial integrated system directed toward speeding up blood
culture diagnosis detects bacteria on a cartridge and records cell
division, growth patterns, and other changes are analyzed in real
time. This is given the name morphokinetic analysis and delivers
results for identification in 2 h and AST in 7 h [8, 9, 38, 39].
5.5 Atomic Force Atomic force microscopy (AFM) is a powerful technique to image
Microscopy many types of surface including bacteria and fungi. It is capable of
measuring adhesion strength, magnetic forces, and mechanical
properties using a tip 10–20 nm in diameter attached to a cantile-
ver. When the tip moves in response to interaction with the surface
it is measured by a laser. Atomic force microscopy generates a 3d
profile. Samples do not require specialized treatment that would
change the structure of the sample making drug susceptibility
testing possible. Atomic force microscopy has been used in a
research application to monitor the change in the cell wall of
bacteria and fungi while being treated with antibiotics opening
the possibility of using this technique to define susceptibility
[40, 41]. Its application is likely to remain most used in researching
the underlying mechanisms of resistance and response to antibiotics
rather than a diagnostic test [42].
6 Detecting Volatile Compounds (Electronic Nose)
For many years sensing the production of volatile compounds has

been used in microbiology to detect the presence of growth. Such a
method is now entering clinical use as detection of volatiles has now
been applied to the determination of DST in a large-scale commer-
cial assay format: Vitek Reveal. A DST result is achieved in, on
average, 4.6 h. Initial comparison with Vitek 2 and a Sensititre
system was favorable [43].
7 Direct Detection by Molecular Means
An alternative way of reducing the time to result is lowering the

limit of detection of the assay, or by detecting resistance without the
need for culture. As we have increasingly understood the genetic
coding of resistance this becomes an attractive approach to resis-

tance detection. There are now a wide range of mechanisms to
identify resistance by detecting the genes that encode it. They can
be broadly summarized into those that detect the presence of a
gene associated with resistance or determination of the gene
sequence of the whole organism (whole-genome sequencing or
guided to specific genes (targeted sequencing)). This works best
for organisms where resistance has occurred due to the acquisition
of a new gene as in the case of methicillin-resistant Staphylococcus
aureus (MRSA) or carbapenemase-producing Enterobacterales
(CPE). Note in this summary we are agnostic about the amplifica-
tion method that is used whether PCR, ligase chain reaction, or
loop-mediated isothermal amplification, as the same strengths and
weaknesses apply [44, 45]. Such methods are attractive as they are
the essentially molecular culture of a specific gene. Considerable
investment has been made in adapting these techniques for the
needs of clinical microbiological practice. For many large labora-
tories, it is essential that such methods fit on their existing high-
throughput diagnostic platforms [15, 46, 47]. These are linked to
systems that provide comprehensive pathogen diagnostics, for
example, for respiratory tract infection, where the number of speci-
mens is large. There is strong support from the manufacturers with
rapid resolution of problems that arise and the likelihood that they
are able to be integrated with the electronic patient record and
results reporting system. They also mean that the investment
made will result in efficiencies that improve quality and reduce
costs. As well as this the flexibility of PCR-based systems and
their careful design means that they should be a “poka-yoke”
system, from the Japanese for avoid blunders (those that minimize
the possibility of operator error).
Nucleic acid amplification test (NAAT)-based susceptibility
testing has the most clinical impact when there are a limited num-
ber of well-characterized resistance determinants, when the treat-
ment is syndromic or by protocol, when the organism grows
poorly, slowly, or is hazardous [48]. In some situations, such as
N. gonorrhoeae, NAAT has largely replaced culture in first-line
diagnosis [49]. To implement such an approach, it is essential that
the genomic epidemiology of resistance is comprehensively under-
stood as this can aid the application of a single target NAAT in other
complex situations. This can not only identify rapid treatment but
can indicate patients that require rapid isolation. A successful exam-
ple is MRSA. NAAT has proved valuable to detect this troublesome
infection control threat rapidly [45, 50]. The specificity means that
it can be applied, for example, to the diagnosis of MRSA in blood
culture where it proves very effective [51].
Tuberculosis is a critically important infectious disease where
the detection of infection and the presence of resistance are impor-
tant. This is heightened by the very slow conventional diagnostics
that take at least 2 months to deliver a susceptibility result

[52, 53]. Since 95% strains with rifampicin resistance have muta-
tions in the M. tuberculosis rpoB gene, this marker can be used as a
surrogate for resistance [54]. Strains that are rifampicin-resistant
are at highest risk of being multidrug-resistant. A commercial assay,
GenXpert™, can generate a result in 2 h with minimal sample
processing. The availability of a simple-to-use NAAT-based assay
has been revolutionary making it possible to provide rapid diagno-
sis in low- and middle-income countries [55]. With support from
global partners this technique has been rolled out widely [56]. With
the increase in multiple and extremely drug-resistant (MDR/XDR)
tuberculosis, additional resistance markers have been included in
Xpert MTB/XDR that includes detection for resistance markers for
isoniazid, ethionamide, fluoroquinolones, amikacin, and capreo-
mycin [57]. A further updated version of the assay, using a
10-color reflex methodology, is suitable for POC application to
detect isoniazid, fluoroquinolone, and second-line injectable drug
resistance [58]. This steady progression in the capability of a com-
mercially produced assay with a focused goal shows the role that
NAAT-based DST could play. On the other hand, at a time when
the capability to detect resistance in second-line injectable anti-
tuberculosis drugs becomes available, it is likely that their use will
decline rapidly as new highly active anti-tuberculosis drugs such as
bedaquiline and pretomanid will have proved their use and start to
become available [59]. A more skeptical view might be that NAAT-
based testing tells the physician which antibiotic drugs cannot be
used whereas he or she wants to know what can be used. This,
however, is to misunderstand the diagnostic process. A positive
organism identification allows protocol-based treatment but more
importantly means that the cause of disease is confirmed and there
is no need to look for an alternative diagnosis (see Fig. 1). This
would save scarce resources and time. Moreover, we can anticipate
that, as our knowledge of antibiotic resistance epidemiology and
molecular genetics improves, more comprehensive molecular tests
will be developed. This aspect is developed further in the section on
genome sequencing where more comprehensive diagnosis is
available.
8 Whole Genome Sequencing for Drug Susceptibility Testing
Resistance to antibiotics is defined in the genome and it is axiomatic

that sequencing the genome of an organism will reveal the full
susceptibility pattern of that organism. In addition to that, it will
allow the epidemiology of resistant organisms to be tracked by
identifying the relationship between strains isolated in each envi-
ronment [60, 61]. Whole genome sequencing is reducing steadily
in price and bringing this powerful technique into the ambit of the
clinical laboratory. The data generated can be analyzed on powerful

computers hosting efficient programs that assemble the genomes
and can make comparisons with the databases of sequenced gen-
omes. As such it already has a wide application in a public health
setting [62]. More user-friendly systems such as the Oxford nano-
pore system make sequencing more accessible, especially if linked
with efficient analysis algorithms [63, 64]. Yet it is likely that WGS
systems will remain in the public health and research environment
for some time yet except where their application provides consider-
able advantages in speed and cost.
One area where genome sequencing is leading the way is the
identification and susceptibility testing for M. tuberculosis. As men-
tioned previously, M. tuberculosis poses many challenges in the
clinical microbiology laboratory; for example, it grows slowly in
comparison to contaminating organisms that can destroy a sample
[65]. The organism is classified as hazard group 3 and must be
cultivated in a high containment laboratory requiring a large capital
expenditure to establish and a significant recurrent revenue com-
mitment. Rapid testing is useful but provides incomplete knowl-
edge upon which clinicians can act. M. tuberculosis is a clonal
organism which means that resistance to antibiotics arises by
point mutations. This means the sequence can be translated into a
phenotypic result once the epidemiology is fully established
[66]. The clinical diagnostic requirement is to detect and identify
that the patient’s sample has M. tuberculosis, or the organism
isolated from that sample contains that pathogen. Subsequently, it
is necessary to determine the phenotypic susceptibility of the strain,
but this process is necessarily slow due to the slow growth of the
organism and the complexity of growth-based susceptibility meth-
ods [52]. Whole genome sequencing can deliver both results
[67]. The question is whether a practical system can be established
to do this quickly and economically.
Over the past number of years, international consortia have
sequenced very large numbers of strains of M. tuberculosis with
phenotypic susceptibility and WGS. As well as illuminating the
epidemiology and molecular evolution of resistance, these data
make it possible to predict phenotypic resistance from the genotype
with a high degree of confidence [68, 69]. These data have been
incorporated into a comprehensive catalog that has been endorsed
by the World Health Organisation [48].
It is now possible to sequence directly from patient samples.
This process uses targeted, amplicon-based deep sequencing which
selectively amplifies targeted gene regions before sequencing. This
lowers the limit of detection and reduces interference from unre-
lated DNA sequences. Targeted sequencing can improve the depth
of coverage [70] and can be applied to portable sequencing devices
such as the Oxford Nanopore [71]. Thus, whole genome sequenc-
ing is well established as a method to deliver identification and
susceptibility results much faster that with conventional methods.

It should now be a standard in high-income countries. With further
innovation in methodology and reduction in cost, sequence-based
methods could be rolled out to low- and middle-income countries
where the need is greatest.
9 Summary and Conclusion
It is encouraging to see that there is a strong research drive to speed

up antibiotic susceptibility testing and a range of innovative tech-
niques have been described. In this review, we have summarized
some of the established and new innovative techniques that could
become clinically available in the next few years. Yet, if real progress
is to be made it will be important that these developments are
evaluated in a clinical environment so that the benefits of rapid
drug susceptibility testing can be realized. Molecular diagnostics
are maturing and finding their place in the clinical diagnostic path-
way. Genome sequencing has started to find its place in viral diag-
nostics and to improve the speed of tuberculosis diagnosis, but
there is progress still to be made. Further investment will be
required to bring some of the more innovative and rapid techniques
to market such as the photonic techniques described. Rapid sus-
ceptibility testing is improving and there is a portfolio of these
technologies applicable to a specific organism, thus improving
patient outcomes and reducing the risk of resistance emerging.
References
1. Dubourg G, Lamy B, Ruimy R (2018) Rapid gram-positive bloodstream infections: speed

phenotypic methods to improve the diagnosis and accuracy of Alfred 60AST. BMC Microbiol
of bacterial bloodstream infections: meeting 19:268
the challenge to reduce the time to result. 6. Ling TKW, Liu ZK, Cheng AFB (2003) Evalu-
Clin Microbiol Infect 24:935–943 ation of the VITEK 2 system for rapid direct
2. Idelevich EA, Becker K (2019) How to accel- identification and susceptibility testing of
erate antimicrobial susceptibility testing, vol gram-negative bacilli from positive blood cul-
25. Clin Microbiol Infect, p 1347 tures. J Clin Microbiol 41:4705–4707
3. Idelevich EA, Becker K (2021) Matrix-assisted 7. Butler-Laporte G, Yansouni CP, Paquette K
laser desorption ionization–time of flight mass et al (2020) Real-word time-to-positivity of
spectrometry for antimicrobial susceptibility two widely used commercial blood culture sys-
testing. J Clin Microbiol 59:e01814–e01819 tems in patients with severe manifestations of
4. Jacobs MR, Good CE, Abdelhamed AM et al sepsis: an analysis of the FABLED study. Open
(2022) Accuracy of direct antimicrobial suscep- Forum Infect Dis 7:ofaa371
tibility testing of gram-negative bacteria from 8. Zalas-Wie˛cek P, Bogiel T, Gospodarek-
positive blood cultures using MicroScan system Komkowska E (2022) The accelerate
and value of using expert rules for β-lactam Pheno™ system—a new tool in microbiologi-
agents. Antimicrob Agents Chemother 66: cal diagnostics of bloodstream infections: a
e02148–e02121 pilot study from Poland. Pathogens 11:1415
5. Anton-Vazquez V, Adjepong S, Suarez C et al 9. Patel YA, Kirn TJ, Weinstein MP et al (2021)
(2019) Evaluation of a new rapid antimicrobial Systematic evaluation of the accelerate pheno
susceptibility system for gram-negative and system for susceptibility testing of gram-
negative bacteria isolated from blood cultures. 21. Junkins AD, Arbefeville SS, Howard WJ et al
Microbiol Spectr 9:e01836–e01821 (2010) Comparison of BD phoenix AP work-
10. Chapot V, Effenberg L, Dohmen-Ruetten J flow with Vitek 2. J Clin Microbiol 48:1929–
et al (2021) Evaluation of the accelerate 1931
pheno system for rapid identification and anti- 22. McGregor A, Schio F, Beaton S et al (1995)
microbial susceptibility testing of positive The MicroScan WalkAway diagnostic microbi-
blood culture bottles inoculated with primary ology system – an evaluation. Pathology 27:
sterile specimens from patients with suspected 172–176
severe infections. J Clin Microbiol 59(5): 23. York MK, Brooks GF, Fiss EH (1992) Evalua-
e02637-20 tion of the autoSCAN-W/A rapid system for
11. Samuel LP, Pimentel JD, Tibbetts RJ et al identification and susceptibility testing of
(2011) Comparison of time to positivity of gram-negative fermentative bacilli. J Clin
the VersaTREK® REDOX 80-mL and the Microbiol 30:2903–2910
REDOX EZ draw 40-mL blood culture bottles 24. Bondi A, Curtoni A, Peradotto M et al (2023)
for common bacterial bloodstream pathogens. Performance evaluation of BD Phoenix and
Diagn Microbiol Infect Dis 71:101–105 MicroScan WalkAway plus for determination
12. Hays JP, Mitsakakis K, Luz S et al (2019) The of fosfomycin susceptibility in enterobacterales.
successful uptake and sustainability of rapid Antibiotics 12:1106
infectious disease and antimicrobial resistance 25. Stefaniuk E, Baraniak A, Gniadkowski M et al
point-of-care testing requires a complex ‘mix- (2003) Evaluation of the BD Phoenix auto-
and-match’ implementation package. Eur J mated identification and susceptibility testing
Clin Microbiol Infect Dis 38:1015–1022 system in clinical microbiology laboratory prac-
13. van Belkum A, Bachmann TT, Lüdke G et al tice. Eur J Clin Microbiol Infect Dis 22:479–
(2019) Developmental roadmap for antimicro- 485
bial susceptibility testing systems. Nat Rev 26. Barry J, Brown A, Ensor V et al (2003) Com-
Microbiol 17:51–62 parative evaluation of the VITEK 2 advanced
14. Holmes AH, Moore LSP, Sundsfjord A et al expert system (AES) in five UK hospitals. J
(2016) Understanding the mechanisms and Antimicrob Chemother 51:1191–1202
drivers of antimicrobial resistance. Lancet 27. Khan A, Arias CA, Abbott A et al (2021) Eval-
387:176–187 uation of the Vitek 2, Phoenix, and MicroScan
15. Smith KP, Kirby JE (2019) Rapid susceptibility for antimicrobial susceptibility testing of steno-
testing methods. Clin Lab Med 39:333–344 trophomonas maltophilia. J Clin Microbiol 59:
16. Testing EC on AS European guidelines for e00654–e00621
breakpoint tables for interpretation of MICs 28. Davaro E, Tomaras AP, Chamberland RR et al
and zone diameters Version 11.0, valid from (2020) Evaluation of a novel light scattering
2021-01-01, e: https://www.eucast.org/ methodology for the detection of pathogenic
fileadmin/src/media/PDFs/EUCAST_files/ bacteria in urine. J Appl Lab Med 5:370–376
Breakpoint_tables/v_11.0_Breakpoint_ 29. Hassan F, Bushnell H, Taggart C et al (2019)
Tables.pdf Evaluation of BacterioScan 216Dx in compari-
17. Clinical and Laboratory Standards Institute son to urinalysis as a screening tool for diagno-
(2022) Performance standards for antimicro- sis of urinary tract infections in children. J Clin
bial susceptibility testing, Clinical and Labora- Microbiol 57:e00571-19
tory Standards Institute Wayne 30. Hammond RJH, Falconer K, Powell T et al
18. Winstanley T, Courvalin P (2011) Expert sys- (2022) A simple label-free method reveals bac-
tems in clinical microbiology. Clin Microbiol terial growth dynamics and antibiotic action in
Rev 24:515–556 real-time. Sci Rep-UK 12:19393
19. Carroll KC, Glanz BD, Borek AP et al (2006) 31. Baron VO, Chen M, Hammarstrom B et al
Evaluation of the BD phoenix automated (2020) Real-time monitoring of live mycobac-
microbiology system for identification and teria with a microfluidic acoustic-Raman plat-
antimicrobial susceptibility testing of entero- form. Commun Biol 3:236
bacteriaceae. J Clin Microbiol 44:3506–3509 32. Fredborg M, Rosenvinge FS, Spillum E et al
20. Carroll KC, Borek AP, Burger C et al (2006) (2015) Rapid antimicrobial susceptibility test-
Evaluation of the BD phoenix automated ing of clinical isolates by digital time-lapse
microbiology system for identification and microscopy. Eur J Clin Microbiol Infect Dis
antimicrobial susceptibility testing of staphylo- 34:2385–2394
cocci and enterococci. J Clin Microbiol 44: 33. Fredborg M, Rosenvinge FS, Spillum E et al
2072–2077 (2015) Automated image analysis for
quantification of filamentous bacteria. BMC LAMP quantification in clinical samples. Sci

Microbiol 15:255 Transl Med 9:eaal3693
34. Singh S, Verma T, Khamari B et al (2023) 46. Vasala A, Hytönen VP, Laitinen OH (2020)
Antimicrobial resistance studies using Raman Modern tools for rapid diagnostics of antimi-
spectroscopy on clinically relevant bacterial crobial resistance. Front Cell Infect Microbiol
strains. Anal Chem 95:11342–11351 10:308
35. Novelli-Rousseau A, Espagnon I, Filiputti D 47. van Belkum A, Burnham C-AD, Rossen JWA
et al (2018) Culture-free antibiotic-susceptibil- et al (2020) Innovative and rapid antimicrobial
ity determination from single-bacterium susceptibility testing systems. Nat Rev Micro-
Raman spectra. Sci Rep 8:3957 biol 18:299–311
36. Baltekin Ö, Boucharin A, Tano E et al (2017) 48. Walker TM, Miotto P, Köser CU et al (2022)
Antibiotic susceptibility testing in less than The 2021 WHO catalogue of mycobacterium
30 min using direct single-cell imaging. Proc tuberculosis complex mutations associated
Natl Acad Sci 114:9170–9175 with drug resistance: a genotypic analysis. Lan-
37. Choi J, Yoo J, Lee M et al (2014) A rapid cet Microbe 3:e265–e273
antimicrobial susceptibility test based on 49. Mohammed H, Ison CA, Obi C et al (2014)
single-cell morphological analysis. Sci Transl Frequency and correlates of culture-positive
Med 6:267ra174 267ra174 infection with Neisseria gonorrhoeae in Eng-
38. Cenci E, Paggi R, Socio GVD et al (2020) land: a review of sentinel surveillance data. Sex
Accelerate Pheno™ blood culture detection Transm Infect 91:287–293
system: a literature review. Future Microbiol 50. Laurent C, Bogaerts P, Schoevaerdts D et al
15:1595–1605 (2010) Evaluation of the Xpert MRSA assay
39. Sze DTT, Lau CCY, Chan T-M et al (2021) for rapid detection of methicillin-resistant
Comparison of novel rapid diagnostic of blood Staphylococcus aureus from nares swabs of
culture identification and antimicrobial suscep- geriatric hospitalized patients and failure to
tibility testing by accelerate Pheno system and detect a specific SCCmec type IV variant. Eur
BioFire FilmArray blood culture identification J Clin Microbiol Infect Dis 29:995–1002
and BioFire FilmArray blood culture identifica- 51. Parkes-Smith J, Bergh H, Harris PNA (2023)
tion 2 panels. BMC Microbiol 21:350 Assessing the performance of the Cepheid
40. Villalba MI, Rossetti E, Bonvallat A et al Xpert in identifying and differentiating
(2023) Simple optical nanomotion method methicillin-susceptible Staphylococcus aureus
for single-bacterium viability and antibiotic and methicillin-resistant Staphylococcus aureus
response testing. Proc Natl Acad Sci 120: from blood culture bottles. Pathology 55:113–
e2221284120 116
41. Quilès F, Accoceberry I, Couzigou C et al 52. Kontsevaya I, Cabibbe AM, Cirillo DM et al
(2017) AFM combined to ATR-FTIR reveals (2023) Update on the diagnosis of tuberculo-
Candida cell wall changes under caspofungin sis. Clin Microbiol Infect 17(6):806
treatment. Nanoscale 9:13731–13738 53. Heyckendorf J, Reimann M, Marwitz S et al
42. Grzeszczuk Z, Rosillo A, Owens Ó et al (2020) (2021) Pathogen-free diagnosis of tuberculo-
Atomic force microscopy (AFM) as a surface sis. Lancet Infect Dis 21:1066
mapping tool in microorganisms resistant 54. Telenti A, Imboden P, Marchesi F et al (1993)
toward antimicrobials: a mini-review. Front Detection of rifampicin-resistance mutations in
Pharmacol 11:517165 mycobacterium tuberculosis. Lancet 341:647–
43. Tibbetts R, George S, Burwell R et al (2022) 651
Performance of the reveal rapid antibiotic sus- 55. Moure R, Muñoz L, Torres M et al (2011)
ceptibility testing system on gram-negative Rapid detection of mycobacterium tuberculo-
blood cultures at a large urban hospital. J Clin sis complex and rifampin resistance in smear-
Microbiol 60:e00098–e00022 negative clinical samples by use of an integrated
44. Hellyer TJ, Desjardin LE, Teixeira L et al real-time PCR method. J Clin Microbiol 49:
(1999) Detection of viable Mycobacterium 1137–1139
tuberculosis by reverse transcriptase-strand dis- 56. Chakaya J, Khan M, Ntoumi F et al (2021)
placement amplification of mRNA. J Clin Global tuberculosis report 2020—reflections
Microbiol 37(518):523 on the global TB burden, treatment and pre-
45. Schoepp NG, Schlappi TS, Curtis MS et al vention efforts. Int J Infect Dis 113:S7
(2017) Rapid pathogen-specific phenotypic 57. Naidoo K, Dookie N (2022) Can the GeneX-
antibiotic susceptibility testing using digital pert MTB/XDR deliver on the promise of
expanded, near-patient tuberculosis drug-
susceptibility testing? Lancet Infect Dis 22: mycobacterium tuberculosis from genome
e121–e127 sequence data with Mykrobe. Wellcome Open
58. Cao Y, Parmar H, Gaur RL et al (2021) Xpert Res 4:191
MTB/XDR: a 10-color reflex assay suitable for 65. Sabiiti W, Azam K, Farmer ECW et al (2020)
point-of-care settings to detect isoniazid, fluo- Tuberculosis bacillary load, an early marker of
roquinolone, and second-line-injectable-drug disease severity: the utility of tuberculosis
resistance directly from mycobacterium molecular bacterial load assay. Thorax 75:
tuberculosis-positive sputum. J Clin Microbiol 606–608
59(3):e02314–20 66. Gillespie SH (2002) Evolution of drug resis-
59. Conradie F, Diacon AH, Ngubane N et al tance in mycobacterium tuberculosis: clinical
(2020) Treatment of highly drug-resistant pul- and molecular perspective. Antimicrob Agents
monary tuberculosis. N Engl J Med 382:893– Chemother 46:267–274
902 67. Gygli SM, Keller PM, Ballif M et al (2019)
60. Parcell BJ, Gillespie SH, Pettigrew KA et al Whole genome sequencing in mycobacterium
(2021) Clinical perspectives in integrating tuberculosis. Ann Transl Med 0:e02175–
whole genome sequencing into the investiga- e02118
tion of healthcare and public health 68. Consortium TCr (2022) A data compendium
outbreaks—hype or help? J Hosp Infect J associating the genomes of 12,289 mycobacte-
109:1–9. https://doi.org/10.1016/j.jhin. rium tuberculosis isolates with quantitative
2020.1 resistance phenotypes to 13 antibiotics. PLoS
61. Parcell BJ, Oravcova K, Pinheiro M et al Biol 20:e3001721
(2017) Pseudomonas aeruginosa intensive 69. Project TCrC and the 100000 G (2018) Pre-
care unit outbreak: winnowing of transmissions diction of susceptibility to first-line tuberculo-
with molecular and genomic typing. J Hosp sis drugs by DNA sequencing. N Engl J Med
Infect 98:282–288 379(1403):1415
62. Köser CU, Ellington MJ, Cartwright EJP et al 70. Jouet A, Gaudin C, Badalato N et al (2021)
(2012) Routine use of microbial whole Deep amplicon sequencing for culture-free
genome sequencing in diagnostic and public prediction of susceptibility or resistance to
health microbiology. PLoS Pathog 8: 13 anti-tuberculous drugs. Eur Respir J 57:
e1002824 2002338
63. Bradley P, Gordon NC, Walker TM et al 71. Cabibbe AM, Spitaleri A, Battaglia S et al
(2015) Rapid antibiotic-resistance predictions (2020) Application of targeted next-
from genome sequence data for Staphylococ- generation sequencing assay on a portable
cus aureus and mycobacterium tuberculosis. sequencing platform for culture-free detection
Nat Commun 6:10063 of drug-resistant tuberculosis from clinical
64. Hunt M, Bradley P, Lapierre SG et al (2019) samples. J Clin Microbiol 58(10):e00632–20
Antibiotic resistance prediction for
Chapter 14
Quantifying Viable M. tuberculosis Safely Obviating

the Need for High Containment Facilities
Bariki Mtafya, Emmanuel Musisi, Paschal Qwaray, Emanuel Sichone,
Natasha Walbaum, Nyanda Elias Ntinginya, Stephen H. Gillespie,
and Wilber Sabiiti
Abstract
Mycobacterium tuberculosis is an infectious pathogen that requires biosafety level-3 laboratory for handling.
The risk of transmission is high to laboratory staff, and to manage the organism safely, it is necessary to
construct high containment laboratory facilities at great expense. This limits the application of tuberculosis
diagnostics to areas where there is insufficient capital to invest in laboratory infrastructure. In this method,
we describe a process of inactivating sputum samples by either heat or guanidine thiocyanate (GTC) that
renders them safe without affecting the quantification of viable bacteria. This method eliminates the need
for level 3 containment laboratory for the tuberculosis molecular bacterial load assay (TB-MBLA) and is
applicable in low- and middle-income countries.
Key words Heat inactivation, GTC, Containment level III, TB-MBLA, Tuberculosis, Diagnosis,
Accessibility
1 Introduction
Tuberculosis (TB), caused by Mycobacterium tuberculosis (M. tb), is

a major public health threat globally [1] and requires level III
laboratory containment for handling [2, 3]. About 10 million indi-
viduals fall sick and 1.5 million die of TB annually [4]. Moreover,
over 40% of cases are left undiagnosed making the goal of tubercu-
losis eradication more difficult as patients must be treated on the
basis of clinical suspicion even with the availability of rapid diag-
nostic tools [3].
Healthcare workers such as laboratory staff have a higher risk of
TB infection compared to staff in other healthcare occupations
[5]. Procedures involving manipulation of TB specimens such as
culture and extraction of nucleic acids have high risk for generating
aerosols which are the main source of TB infections [6]. The World
https://doi.org/10.1007/978-1-0716-3981-8_14,
145
146 Bariki Mtafya et al.
Health Organization (WHO) recommends such procedures be

carried out in a biosafety level 3 (BSL3) laboratory equipped with
Class II biological safety cabinets and a negative pressure system
[6]. However, such infrastructures are very expensive to establish
and maintain, and as a result they are not available in most of the
resource-poor countries.
Among rapid diagnostic techniques, molecular tests such as the
Xpert MTB/RIF Assays use a closed cartridge system with chao-
tropic buffers and can be performed at a primary microscopic health
facility without BSL3 laboratory containment. Lack of BSL3 labo-
ratory infrastructure slows the uptake of novel TB tests such as the
TB-MBLA, LPA, and standard culture [7]. To achieve the WHO
End TB strategy with the ambitious target to reduce TB incidences
by 90% in 2035, efforts are needed to improve access to rapid and
accurate TB diagnostics in healthcare settings with minimal bio-
safety requirements. Scale up of novel diagnostic tools that obviates
the need for complex and expensive bio-safety level 3 containment
will increase and equalize access for novel TB diagnostic services in
routine practice laboratories.
There have been significant improvements with new methods
for rapid diagnosis of tuberculosis. Yet even while using such tests,
it is many weeks before the healthcare staff can be sure that the
patient is responding, and this opens the possibility that an ineffec-
tive treatment will be used. In such a circumstance there is an
increased risk of the infecting organism acquiring additional resis-
tance which become more difficult to treat with the standard TB
medicines.
Here, we describe a method that renders TB samples safe for
TB-MBLA application in routine practice laboratories without bio-
safety level 3 containment [8]. This will speed TB diagnosis,
enabling ascertaining response to therapy rapidly and opportunity
for clinicians to intervene timely. By reducing or eliminating the
need for biosafety laboratory 3 facilities, it will improve access to
better TB diagnostics and therapeutic monitoring [9, 10].
2 Materials
2.1 Equipment 1. Exhaust protective Biological Safety Cabinet (Class II)

(see Note 1).
2. Vortex mixer.
3. Blotting paper.
4. Tissue paper.
5. Water bath.
6. Heating block Disposable waste containers.
Quantifying Viable M. tuberculosis Safely Obviating the Need. . . 147
2.2 Materials 1. Micropipettes; P20, P200, P1000.

2. Micropipette tips; for 20 μL, 200 μL and 1000 μL.
3. 15 mL centrifuge tubes.
4. 1.5–2 mL microcentrifuge tubes.
5. Sputum specimens.
6. Sterile Pasteur pipettes.
7. Gloves.
8. TB disinfectant, e.g., Jik, Dismozon Pour or Tristel.
9. Ethanol; 96–100%.
10. Disposable waste bags.
3 Methods
3.1 Sputum 1. Collect sputum samples from TB patients in a clean container

Specimens or follow your local sputum collection procedures.
2. Homogenize sputum using a sterile magnetic stirrer for 30 min
at room temperature or use your in-house procedure.
3.1.1 Heat 1. Aliquot 1 mL of sputum in 1.5 mL microcentrifuge tube.

Inactivation (see Note 4) 2. Heat the sample at 80 C for 20 min in water bath.
3. After heating, allow the samples to cool at room temperature
for 30 min.
4. Add 100 μL of extraction control (Vitalbacteria, UK) (see
Note 2).
5. Centrifuge 1.5 mL tubes at 20,000 g for 20 min at room
temperature.
6. Remove the supernatant and leave at least 100 μL in the tube.
7. Add 950 μL of lysis buffer (RNA Pro solution, MP biomedi-
cals) and proceed with RNA extraction Subheading 3.2 (see
Note 3).
3.1.2 GTC 1. Mix 1 mL of homogenized sputum with 4 mL of GTC.

Inactivation (see Note 4) 2. Incubate the mixture at room temperature for 15 min.
3. Add 100 μL of extraction control (Vitalbacteria, UK).
4. Centrifuge 15 mL falcon tubes at 3000 g for 30 min.
5. Remove the supernatant and leave at least 100 μL in the tube.
6. Add 950 μL of lysis buffer (RNA Pro solution, MP biomedi-
cals) and proceed with RNA extraction Subheading 3.2.
3.2 RNA Extraction 1. Vortex the samples and transfer the whole liquid into lysing
matrix.
2. Transfer the tubes to a homogenizer (Precellys 24—PeQlab or
FastPrep—MP Biomedicals) (see Note 5).
3. Set the program 6.0 for 40 s on FastPrep or use 6000 rpm for
40 s, if using the Precellys 24 machine for homogenization of
the sample.
4. After homogenization, centrifuge the tubes for 5 min at
12,000 g at room temperature.
5. Leave the tube to stand for 5 min following centrifugation.
6. Label fresh 1.5 mL screw cap tubes (RNase free) with the
corresponding numbers and add 300 μL chloroform into
each clean tube.
7. From the homogenization tubes, carefully transfer the whole
liquid part to the chloroform-containing tube using a fine tip
pipette (be careful not to transfer bits of sample debris or lysing
matrix).
8. Vortex the tubes for 10 s each and incubate at room tempera-
ture for 5 min.
9. Centrifuge at 12000 g at room temperature for 5 min.
10. Label new 1.5 mL micro-centrifuge tubes with the
corresponding numbers.
11. Carefully transfer the upper aqueous phase to the fresh tubes
using 200 μL filter tips, being careful not to transfer any of the
interphase or lower layer.
12. Add 500 μL of 100% ice-cold ethanol to each tube.
13. Mix the contents by inverting the tubes 5 times.
14. Transfer the tubes containing samples to the 20 C freezer
overnight (or to the 80 C freezer for 15 min, if you are
doing the whole extraction on the same day).
15. Chill the microfuge to at least 12 C prior to centrifugation (set
it at 4 C).
16. Centrifuge the samples at 13,000 g for 20 min.
17. Discard the supernatant using a fine pipette tip.
18. Add 500 μL 70% ice-cold ethanol to each tube.
19. Centrifuge as above for a further 10 min.
20. Discard the supernatant using a Pasteur pipette with fine tip.
Use a new pipette for every sample.
21. Dry the RNA at 50 C in a heat hot block (approximately
20–30 min).
22. Add 100 μL RNase-free water to each tube, leave at room

temperature for 5 min and resuspend the RNA by vortexing
for 5 s.
23. Store RNA extract at 80 C until the day of DNA removal or
proceed directly to Subheading 3.3.
3.3 DNA Removal 1. This is done using Turbo DNase (Turbo DNA-free kit,
Ambion AM1907) or RNase-Free DNase Set (Qiagen).
2. Make a master mix of the Turbo DNase I; 10 buffer (blue
top) and DNase I enzyme (white top) for the number of
samples plus 2 extra. Per sample, add 10 μL buffer and 2 μL
DNase enzyme to each sample. For example, a 10-sample
master mix would contain 120 μL of 10 buffer and 2 μL of
DNase enzymes.
3. Mix by vortexing and then pipette 12 μL into each tube con-
taining your RNA extract.
4. Spin briefly (5–10 s at 13000 g) to bring everything to the
bottom of the tube. This step eliminates droplets hanging on
the tube wall and ensures that all RNA extract is in contact with
DNase enzyme.
5. Incubate at 37 C for 30 min in the hot block or incubator.
6. After 30 min, add an additional 1 μL of DNase enzyme to each
tube. Ensure all enzyme goes into the sample by pipetting up
and down (3) and then use the tip to stir around and mix the
enzyme with the rest of the mixture.
7. Incubate at 37 C for a further 30 min in the hot-block or
incubator.
8. Thaw the DNase inactivation reagent (white milky substance)
at least 10 min prior the finish of DNase incubation and keep in
the fridge. Resuspend by vortexing.
9. Add 10 μL of DNase inactivation reagent into each RNA
extract and incubate for 5 min at room temperature.
10. Vortex three times during the 5-min incubation step at room
temperature.
11. Centrifuge at 13,000 g for 2 min.
12. Label 1.5 mL screw cap tubes with sample ID and
extraction date.
13. Set your pipette at 105 μL and carefully pipette off the RNA to
a fresh 1.5 mL RNase-free tube without touching any of the
inactivation matrix.
14. Store the RNA at 4 C until use if running qPCR on the same
day or next day.
15. For long-term use, store the RNA at 80 C.
3.4 Preparation of 1. A ready-to-go master mix containing the probes, primers, and
Master Mix Quantitec master mix is supplied in TUBE 1 and RT enzyme is
supplied in TUBE 2. Mix the content from TUBE 1 and
2 following the instructions included in the pack insert to
obtain the RT+ mix or follow the manufacturer’s instructions
if using a different kit (e.g., Life Arc kits).
2. The RT- master mix is supplied as “ready to use”, no additional
manipulation is required.
3. Add 16 μL of the master mix solution to each PCR tube
(if using tubes for the RotorGene Q) or in the 96/384 well
plate other PCR platforms, e.g., Quantistudio, CFX96, or ABI
7500 (see Note 6).
4. Add 4 μL of the RNA sample to each well containing 16 μL of
the master mix (total reaction volume will be 20 μL) as indi-
cated in Table 1.
5. Each sample and controls should be run in duplicate for RT+
reactions (containing the RT enzyme) and in a single reaction
for the master mix without reverse transcriptase (RT-).
6. Seal the tube with PCR tube capes and the plate with the
protective film, respectively.
7. Centrifuge the plate at 1500 rpm for 1 min to ensure all liquids
are at bottom of wells and transfer the plates into the CFX96,
ABI 750, or Quantistudio instrument (see Note 7).
8. No need to centrifuge the tubes if using the Rotorgene. The
machine will spin all the tubes when running.
3.5 Thermal Cycling 1. Hold 50 C, 30 min [this is the reverse transcription].

Conditions 2. Hold 95 C, 15 min [this activates the Taq polymerase].
3. Cycling, 40 cycles of: 94 C, 45 s not acquiring, 60 C, 60 s
acquiring at Green and Yellow for the Rotorgene thermal
cycler.
4. Ensure the correct channels and targets are appropriately
selected if using a different PCR machine.
Table 1
Quantities of the master mixes required and samples
RT+ mix per reaction (μL) RT mix per reaction (μL)

Ready-to-go master mix 16 16
Sample (RNA template) 4 4
Total volume (including sample) 20 20
4 Notes
1. An exhaust protective cabinet is still required for processing of

samples due to the likely presence of live M. tuberculosis in the
samples. Once heat inactivated we could not detect live organ-
isms in our experiments.
2. The method presented here is optimized for the VitalBac-
teria™ kit. Alternative kits are becoming available and can be
used with a little adaptation.
3. Alternative extraction materials can be used but careful evalua-
tion is required to ensure equivalence efficiency.
4. Guanidine thiocyanate, GTC (Promega) is prepared using the
following steps: 100 g of GTC powder with 120 mL of molec-
ular biology grade water (Sigma). Incubate the mixture at
37 C overnight and in the next morning, add 20 mL of Tris-
HCL and shake the mixture. Add 2 mL of β-mercaptoethanol
to the mixture and top with Molecular biology grade water to a
final volume of 200 mL. Mix the solution well and aliquot
4 mL of GTC into 15 mL large centrifuge tubes. Freeze at
80 C until use. Use the aliquots within 6 months and avoid
freeze and thawing.
5. Ensure that the tubes are fully pushed in, that the plastic tube
holder is engaged with the metal pin and the spokes are located
above each tube. If you are using the FastPrep method, screw
the cap on the top and then close the machine lid.
6. The final concentration of primer and probe are 0.1 μM and
0.2 μM per reaction, respectively.
7. TB MBLA is adaptable to different real-time PCR platforms.
The method presented here can be adapted readily.
References
1. Lange C, Aarnoutse R, Chesov D, van 5. Joshi R, Reingold AL, Menzies D, Pai M
Crevel R, Gillespie SH, Grobbel HP et al (2006) Tuberculosis among health-care work-
(2020) Perspective for precision medicine for ers in low- and middle-income countries: a
tuberculosis. Front Immunol 11:566608 systematic review. PLoS Med 3(12):
2. Siddiqi SH, Rüsch-Gerdes S (2006) MGIT 2376–2391
Procedure Manual: For BACTEC™ MGIT 6. World Health Organization (2012) Tubercu-
960™ TB System. 2006;(July):1–52 losis laboratory biosafety manual. World
3. World Health Organization (2021) Technical Health Organization Publication. ISBN
Report on critical concentrations for drug sus- 978 92 41504638, Geneva
ceptibility testing of isoniazid and the rifamy- 7. Allen V, Nicol MP, Tow LA (2016) Sputum
cins (rifampicin, rifabutin and rifapentine), processing prior to mycobacterium tuberculo-
1–86 p sis detection by culture or nucleic acid amplifi-
4. World Health Organisation (2022) Global cation testing: a narrative review. Res Rev J
tuberculosis report. Geneva Microbiol Biotechnol 5(1):96–108
8. Mtafya B, Qwaray P, John J, Sichone E, bacillary load, an early marker of disease sever-
Shoo A, Gillespie SH et al (2023) A practical ity: the utility of tuberculosis molecular bacte-
approach to render tuberculosis samples safe rial load assay. Thorax 75:606–608
for application of tuberculosis molecular bacte- 10. Ntinginya NE, Bakuli A, Mapamba D,
rial load assay in clinical settings without a bio- Sabiiti W, Kibiki G, Minja LT et al (2022)
safety level 3 laboratory. Tuberculosis 138 Tuberculosis molecular bacterial load assay
(November 2022):102275 reveals early delayed bacterial killing in patients
9. Sabiiti W, Azam K, Farmer ECW, Kuchaka D, with relapse. Clin Infect Dis 76(3):990–994
Mtafya B, Bowness R et al (2020) Tuberculosis
Chapter 15
Improved Diagnosis and Treatment Monitoring

of Tuberculosis Using Stool and the Tuberculosis Bacterial
Load Assay (TB-MBLA)
Emmanuel Musisi, Bariki Mtafya, William Saava Wambi,
Josephine Zawedde, Abdulwahab Sessolo, Willy Ssengooba,
Natasha Walbaum, Nyanda Elias Ntinginya, Stephen H. Gillespie,
and Wilber Sabiiti
Abstract
The diagnosis and monitoring of tuberculosis treatment is difficult as many patients are unable to produce
sputum. This means that many patients are treated on the basis of clinical findings and consequently some
will be exposed to anti-tuberculosis drugs unnecessarily. Moreover, for those appropriately on treatment
and unable to produce a sputum sample, it will be impossible to monitor the response to treatment. We
have shown that stool is a potential alternative sample type for diagnosis of tuberculosis. Currently, available
protocols like the Xpert MTB/RIF use DNA as a target to detect Mycobacterium tuberculosis in stool but
DNA survives long after the organism is dead so it is not certain whether a positive test is from an old or a
partially treated infection. The TB MBLA only detects live organisms and thus, can be used to follow the
response to treatment. In this chapter, we describe a protocol for TB-MBLA, an RNA-based assay, and
apply it to quantify TB bacteria in stool.
Key words Tuberculosis, Treatment, Drug resistance, Antibiotics, Molecular diagnostics
1 Introduction
Rapid detection of M. tuberculosis and treatment initiation is imper-

ative particularly among vulnerable groups like children and people
living with advanced HIV disease yet diagnosis of TB in these
subpopulations of individuals is challenging due to difficulty in
obtaining adequate sputum. Usually, this leads to low case detec-
tion and high mortality rates [1, 3–5]. Collection methods for
alternative samples like gastric and nasopharyngeal aspirates,
bronchoalveolar lavage, and induced sputum are unpleasant, inva-
sive, and have low diagnostic yield. Besides, such sampling
https://doi.org/10.1007/978-1-0716-3981-8_15,
153
154 Emmanuel Musisi et al.
procedures are not normally available in resource-constrained set-

tings where the TB burden is high. Normally people swallow
sputum, which ends up in the gut, and hence, can be found in the
stool. A stool sample has been suggested as an alternative or addi-
tional sample type for bacteriological confirmation of pulmonary
TB. Recently, WHO rolled out the use of stool samples for diagno-
sis of TB on Xpert MTB/RIF ultra (Xpert ultra) platform. The
Xpert test is a sensitive measure of M. tuberculosis DNA and has
proved itself in diagnosis worldwide. It does, however, detect both
live and dead organisms as the DNA persists for an extended
period after cell death [2, 3].
The tuberculosis molecular bacterial load assay (TB-MBLA)
was developed to utilize the abundant cellular 16S rRNA as a
marker for the detection and quantification of viable Mycobacte-
rium tuberculosis bacilli in sputum samples [4, 5]. In a similar way
to viral-load monitoring in HIV-positive individuals, TB-MBLA
can chart the TB treatment response by measuring changes in the
M. tuberculosis viable bacillary load over the course of treatment.
The TB-MBLA, therefore offers both rapid diagnosis and the
opportunity for treatment response monitoring advantages in real
time. This would be an important advance for healthcare staff and
would inform clinical decision-making. Although TB-MBLA may
detect and quantify M. tuberculosis in stool samples, its utility in
stool has not been assessed. Having a sensitive quantitative test that
can be used on an easily accessible sample may improve TB diagno-
sis and treatment response monitoring in patients who do not
provide suitable sputum samples easily [6].
2 Materials
2.1 Materials for Stool samples can be collected using a flushable Hystool® stool
Stool Collection collection device (https:/www.hystool.co.uk) (see Note 1).
2.2 General 1. Class I Exhaust protective cabinet (Biosafety cabinet).

Laboratory 2. Tissue paper (Kitchen towel).
Requirements
3. Permanent marker.
4. Laboratory record (see Note 2).
5. Freezer (20 C).
6. Freezer (80 C).
7. Chemical waste discard jar.
8. Safety goggles that are chemical resistant.
9. 500 mL plastic containers.
10. Racks for 1.5 mL and 2 mL microtubes and for 15 mL centri-
fuge tubes.
Stool TB MBLA Treatment Monitoring 155
11. Electronic pipette with matching pipette tips (10 mL tips).

12. Vortex (e.g., Megafuge 16, Heraeus).
13. Sterile pasture pipettes (2 mL).
2.3 M. tuberculosis 1. Fresh or stored samples (see Note 3).

RNA Extraction 2. Thawed internal controls (see Note 4).
3. DNAse enzyme, Turbo DNA-free, AM1907M (Invitrogen).
4. Chloroform.
5. RNAse away (10666421, Fisher Scientific) (see Note 5).
6. Absolute ethanol (99–100%).
7. Ice cold 70% ethanol.
8. TB disinfectant (Tristel Fuse).
9. Molecular grade water (RNAse and DNAse free, e.g.,
13138533, Fisher Scientific).
10. Sterile RNase-free microcentrifuge tubes (1.5 mL).
11. FASTRNA Pro blue kit with homogenization beads from MP
biomedicals.
12. Refrigerated microcentrifuge for 1.5 mL and 2 mL homogeni-
zation tubes.
13. Homogenizer (Precelleys 24, Peqlab set at 6000 rpm, 40 s).
14. Thermomixer with thermal block for 1.5 mL microcentrifuge
tubes.
15. Pipettes and sterile filtered pipette tips that are DNAse and
RNAse free, Range: P1000, P200, P10, and P2.
3 Methods
3.1 Collecting the 1. Attaching the stool collection bag on the toilet.
Stool Sample Peel the backing paper off the adhesive pads on the corners
of the stool collection bag and attach it to the toilet seat using
the adhesive pads.
2. Utilizing the toilet.
Fold the piece of the toilet paper and place it at the bottom
of the stool bag to prevent direct contact between the bag and
stool. Loose stool could burst the bag. Instruct the patient to
sit on the toilet seat and use it normally. Care should be taken
for stool not to pass through the bag. Try as much as possible
not to pee in the stool because this could contaminate the
sample or dissolve it.
3. Pick a portion of the stool (about two scoops) into the sample
container using a scoop which is attached to the cap of the stool
container.
4. Unpeel the sticky pads from the toilet seat allowing the gab,
sticker, and contents to fall into the toilet and flush normally.
The stool bag used will dissolve in water and the sticker paper
was biodegradable.
3.2 Processing Stool 1. Place about ~6.0 g sterile stool into a pre-labeled stool con-
Samples tainer and homogenize in 10 mL phosphate buffered saline
(PBS). Incubate at room temperature for 15 min.
2. Centrifuge twice the resulting suspension at 3000 g for
20 and 10 min, respectively. Re-suspend the pellet in 6 mL of
phosphate-buffered saline and aliquot portions of 1 mL.
3.3 Preparation of 1. Primers and probes are lyophilized and stable at room temper-
Probes and Primers ature as set out in Table 1.
Notes for the TB-MBLA 2. Make stock primer per the manufacturer’s instructions.
Study
3. Make stock probe per the manufacturer’s instructions.
3.4 RNA Extraction 1. Thaw samples and extraction control (EC) and place on ice
until used.
2. Label same number of homogenization tubes with glass beads.
Label both lids and tubes.
3. Add 100 μL of EC directly into the sample, mix by inverting
20 times. Change tips between samples.
4. Spin at 3000 g for 30 min, carefully tip off GTC into 15 mL
centrifuge tubes.
5. Use a fine pipette tip to remove all the GTC. Avoid the pellet as
much as possible.
Table 1
Preparation of primer mix for the internal control (total volume = 200 μL)
Given Use the formula; V1C1 = V2C2

Stock concentration ¼ 100 μM Y · 100 μM ¼ 200 μL · 10 μM
Final concentration ¼ 10 μM 200 μL · 10 μM
Y¼
20 μL
Final volume ¼ 200 μL
Stock volume of the internal control primers ¼ Y
Y = 20 μL-------<------- Stock volume
So; add 20 μL of internal control stock forward primer
20 μL of internal control stock reverse primer
160 μL of RNAse-free molecular grade water
6. Resuspend pellet in 950 μL lysis buffer using a P1000

micropipette.
7. Transfer the resuspended pellet to lysing matrix beads using a
P1000 micropipette.
8. Bead beat at 6000 rpm for 40 s (Homogenization).
9. Spin at 12,000 g for 5 min at RT.
10. Incubate tube at room temperature for 5 min.
11. Transfer all the supernatant to 300 μL of chloroform in micro-
centrifuge tube. Use fine pipette.
12. Vortex for 10 s.
13. Incubate at room temperature for 5 min.
14. Spin at 12,000 g for 5 min.
15. Transfer supernatant to labeled microcentrifuge tubes using a
200 μL filter tip.
16. Add 500 μL of ice-cold absolute ethanol.
17. Keep at 20 C for overnight.
18. Chill the microcentrifuge at 4 C.
19. Spin at 13000 g for 20 min.
20. Discard the supernatant using a fine pipette tip.
21. Add 500 μL of ice-cold absolute 70% ethanol.
22. Spin at 13000 g for 10 min at 4 C.
23. Discard the supernatant.
24. Dry RNA at 50 C in a heat-hot block for 30 min.
25. Dissolve the extracted RNA in 100 μL of RNAase-free water.
26. Re-suspend by vortexing for 5 s.
27. This material can be stored at 80 C if necessary.
3.5 DNAse Treatment 1. Prepare the master mix by adding 10 μL (10X) DNase buffer,
and 2 μL DNAse enzyme to a sterile microcentrifuge tube (see
Note 6).
2. Add 12 μL of DNAse master mix to each sample.
3. Vortex for 5 s.
4. Spin at 13,000 g for 10 s.
5. Incubate at 37 C for 30 min in the heat hot block.
6. Add 1 μL of DNAse enzyme to each tube. Pipette up and down
three times.
7. Incubate at 37 C for 30 min in the heat hot block.
8. Thaw DNase inactivation buffer (10 min before).
9. Add 10 μL of DNAse inactivation reagent.
Table 2
Quantities required for the master mix
Volume per RT+ Volume per RT Total vol. of RT+ Total vol. of RT
Master mix reaction μL reaction μL reactions μLa reactions μLa
Quantitec mix 10 10
(PCR mix)
Mtb 16S primer 0.4 0.4
mix (F + R)
IC primer mix 0.4 0.4
(F + R)
Mtb 16S-FAM 0.2 0.2
probe
IC probe 0.2 0.2
RT enzyme 0.2 0
RNAs-free water 4.6 4.8
Total volume 16 16
a
Columns left blank to allow calculations to be inserted
10. Vortex three times every after 5 min.

11. Spin at 13,000 g for 2 min.
12. Transfer 110 μL of the supernatant to a new clean
centrifuge tube.
13. Store at 4 C for same day, or 80 C for long-term.
3.6 Preparation of 1. This mix contains reverse transcriptase enzyme.

RT-qPCR Master Mix 2. Aliquot 200 μL of the mix into 1.5 mL RNAse-free microcen-
(Use QuantiTec trifuge tubes, store at 25 C See Table 2.
Multiplex RT-PCR NR
Kit (QT))
3.7 Reverse 1. For unknown samples, dilute all RNA extracts to be used in a 1:
Transcriptase qPCR 10 ratio in RNase-free water.
2. Mix well by vortexing for 5 s and briefly spin down to remove
any droplets or air bubbles.
3. For standard samples for a standard curve, take the Mtb and EC
RNA standards from the 80 C freezer and thaw at room
temperature. Make seven and six 10-fold dilutions of Mtb and
EC standard samples respectively (see Note 7).
4. Master mix preparation NOTE: Master mix (MM) is a solution
of PCR reagents sufficient to amplify all samples, standards, and
water for a no template control (NTC). The water used as NTC
should be the same water used in the extraction and for
preparing the MM. Ensure that the standards, each RNA sam-
ple, and its decimal dilution are amplified 2 for the reverse
transcriptase positive (RT+) reaction and 1 for the reverse
transcriptase negative (RT) reaction. The RT reaction is a
control to determine the efficiency of DNA removal (see
Note 8).
5. Transfer 16 μL of MM into each PCR reaction tube.
6. Add 4 μL of RNA extract into each RT+ and RT reaction tube
and water into the NTC reaction tubes.
7. Load the reaction tubes into a real-time PCR machine and set
the PCR conditions as follows: 50 C for 30 min, 95 C for
15 min, 40 cycles at 94 C for 45 s, and 60 C for 1 min with
acquisition with fluorophores that absorb in green and yellow
channels.
4 Notes
1. Alternative stool collection devices can be used.

2. Effective recording of samples and results is essential to ensure
that issues of M. tuberculosis cross-contamination, details of the
process and quality assurance are captured [7]. Increasingly,
computer-based laboratory information systems can improve
the traceability of samples [8].
3. Ideally samples should be processed as quickly as possible. They
should be transported to the laboratory at between 4 and 8 C
with a record of transit temperature kept. If this is not possible,
or there is a need to batch the samples there is no loss of signal
if the specimen is preserved in GTC and frozen at 80 C
[5, 9].
4. Internal controls can be obtained from VitalBacteriaTM
(https://www.vitalbacteria.com).
5. RNase Away reagent is crucial for getting rid of RNases from
work surfaces and apparatus.
6. We have tested many of the alternative products for this section
and found that the results are equivalent. Thus researchers
should investigate, validate, and use alternative similar
products.
7. For example, for 20 samples, the master mix will contain
200 μL (10X) buffer and 40 μL DNase enzyme.
8. Standard samples are supplied by VitalbacteriaTM with the
TB-MBLA kit. It is essential to change the tips before transfer-
ring the mixture from one tube to another.
References
1. Schumacher SG, Denkinger CM (2019) The 6. Musisi E, Ssesolo A, Sloan DJ et al (2022)
impact of Xpert MTB/RIF—do we have a final Detection and quantification of viable mycobac-
answer? The Lancet Global Health 7:e161–e162 terium tuberculosis bacilli in saline-processed
2. Kennedy N, Gillespie SH, Saruni AO et al stool samples by tuberculosis molecular bacterial
(1994) Polymerase chain reaction for assessing load assay: a potential alternative for processing
treatment response in patients with pulmonary stool. Microbiol Spectr 10:e00274–e00222
tuberculosis. J Infect Dis 170(713):716 7. Ruddy M, McHugh TD, Dale JW et al (2002)
3. Friedrich SO, Rachow A, Saathoff E et al (2013) Estimation of the rate of unrecognized cross-
Assessment of the sensitivity and specificity of contamination with mycobacterium tuberculosis
Xpert MTB/RIF assay as an early sputum bioin London microbiology laboratories. J Clin
marker of response to tuberculosis treatment. Microbiol 40:4100–4104
Lancet Respir Med 1(462):470 8. Mekota A-M, Gillespie SH, Hoelscher M et al
4. Honeyborne I, McHugh TD, Phillips PPJ et al (2022) Building sustainable clinical trial sites in
(2011) Molecular bacterial load assay, a culture- Sub-Saharan Africa through networking, infra-
free biomarker for rapid and accurate quantifica- structure improvement, training and conducting
tion of sputum mycobacterium tuberculosis clinical studies: the PanACEA approach. Acta
bacillary load during treatment. J Clin Microbiol Trop 238:106776
49(3905):3911 9. Mtafya B, Sabi I, John J et al (2022) Systematic
5. Sabiiti W, Azam K, Farmer ECW et al (2020) assessment of clinical and bacteriological mar-
Tuberculosis bacillary load, an early marker of kers for tuberculosis reveals discordance and
disease severity: the utility of tuberculosis molec- inaccuracy of symptom-based diagnosis for
ular bacterial load assay. Thorax 75:606–608 treatment response monitoring. Front Med 9:
992451
Chapter 16
Application of Pathogen Genomics to Outbreak

Investigation
Benjamin J. Parcell, Kerry A. Pettigrew, and Katarina Oravcova
Abstract
Outbreaks are a risk to public health particularly when pathogenic, hypervirulent, and/or multidrug-
resistant organisms (MDROs) are involved. In a hospital setting, vulnerable populations such as the
immunosuppressed, intensive care patients, and neonates are most at risk. Rapid and accurate outbreak
detection is essential to implement effective interventions in clinical areas to control and stop further
transmission. Advances in the field of whole genome sequencing (WGS) have resulted in lowered costs,
increased capacity, and improved reproducibility of results. WGS now has the potential to revolutionize the
investigation and management of outbreaks replacing conventional genotyping and other discrimination
systems. Here, we outline specific procedures and protocols to implement WGS into investigation of
outbreaks in healthcare settings.
Key words Outbreak, Typing, Whole-genome sequencing, Short-read sequencing, Bacterial

pathogens
1 Introduction
The transmission of healthcare-associated pathogens, especially

multidrug-resistant organisms (MDROs) in a hospital is a major
threat to patient, staff, and visitor health. Healthcare outbreaks can
arise from breakdown in infection prevention and control (IPC)
measures such as hand hygiene, inadequate hospital environment,
e.g., aging facilities, limited isolation rooms, poor ventilation,
defective water supplies, and contamination of medical equipment.
Outbreaks can cause major disruption for hospital services resulting
in closure of wards, cancellations of services, environment contam-
ination, increased costs for cleaning, and use of personal protective
equipment (PPE) along with anxiety for staff and patients.
Rapid and accurate outbreak detection is essential to imple-
ment effective targeted interventions to control further transmis-
sion. The Centers for Disease Control and Prevention (CDC)
https://doi.org/10.1007/978-1-0716-3981-8_16,
161
162 Benjamin J. Parcell et al.
define an epidemic as an “increase, often sudden, in the number of

cases of a disease above what is normally expected in that popula-
tion in that area” [1]. They define that an outbreak carries “the
same definition of epidemic, but is often used for a more limited
geographic area” and a cluster is an “aggregation of cases grouped
in place and time that are suspected to be greater than the number
expected, even though the expected number may not be known”.
In the United Kingdom, National Services Scotland (NSS) guid-
ance advises the infection prevention and control team (IPCT) to
assume there could be an healthcare-associated infection (HAI)
outbreak when there are “two or more linked cases with the same
infectious agent associated with the same healthcare setting over a
specified time period” or when there is a “higher-than-expected
number of cases of HAI in a given healthcare area over a specified
time period” [2].
Outbreak management hinges on microbiology services
providing results in a timely manner. Microbiology laboratories
have digital reporting systems called laboratory information man-
agement systems (LIMS), also referred to as a laboratory informa-
tion system (LIS). In addition to storing data such as sample type,
date, location, organism’s identification, and susceptibility profile
these computer systems can be used to produce microbiology
reports and audit turn-around times and service quality. This data
can also be used to identify organisms of concern or clusters of
similar organisms in specific sample types or patients from any
location [3]. These systems can be automated and set up so that
reports are created to detect trends of infection and resistance for
daily review as defined by the local IPCT. They may also be linked
to surveillance software such as ICNET Clinical Surveillance Plat-
form which alerts the IPCT to potential outbreaks [4]. Automated
outbreak detection tools which are statistically supported have been
found to be beneficial in terms of streamlining work for IPCTs
[5]. Novel applications have been developed in which routine IPC
data such as ward admission, date of samples, and pathogen geno-
mic information can be visualized. An example is Shiny, an open-
source web application for statistical software R [6]. Additionally,
there are platforms providing tools for comprehensive epidemio-
logical investigation such as Outbreak Toolkits which serve the
purpose of storing and visualizing data [7]. Another web-based
platform Microreact is also freely available and can be utilized for
interactive data visualization and genomic epidemiology [8, 9]..
Generally, outbreak identification methodology relies on iden-
tifying the same microorganism of concern from at least two differ-
ent patients in the same location within a close time frame. As part
of this, phenotypic tests, such as antibiotic sensitivity patterns
(a surrogate for relatedness of bacterial isolates) are used, but they
can lack sufficient discriminatory power to differentiate between
two strains. Once potential outbreak isolates are identified, they are
Application of Pathogen Genomics to Outbreak Investigation 163
usually sent to a central reference laboratory for further typing.

Current typing techniques involving single or multiple gene
sequence-based analysis approaches, such as staphylococcal protein
A (spa) genotyping for Staphylococcus aureus, M protein gene
(emm) typing of Group A Streptococcus, multilocus sequence typing
(MLST), or amplification across the genome such as variable-
number tandem repeats (VNTR) typing of Gram-negatives, and
whole genome restriction analysis such as pulsed field gel electro-
phoresis (PFGE) for Enterococcus spp., can lack granularity and
discrimination making it difficult to identify whether bacterial iso-
lates are related or not and whether transmission is likely to have
occurred. Of note, current genotyping methods do not typically
take the spread of extrachromosomal mobile genetic elements
(MGEs) such as plasmids which can carry resistance genes for
antibiotics, disinfectants, and heavy metals along with additional
virulence determinants into account. Additionally, WGS offers the
opportunity to also investigate these further.
Reliable pathogen genotyping data should be detailed enough
to allow the discrimination of dissimilar strains isolated from hosts
or their environment, and therefore allowing transmission events to
be ruled in or out. It has been reported that pathogen sequencing
data can provide greater levels of confidence for users. [10] Newer
genotyping methods such as whole genome sequence data analysis
can identify whether the same bacteria have spread between differ-
ent patients by aligning and mapping the sequences of interest and
identifying DNA differences such as single nucleotide polymorph-
isms (SNPs) and acquisition of accessory genes. If bacteria are
closely related, they will have a low number of SNPs suggesting
evidence of likely transmission between patients or a common
source of infection. Sequencing can be applied to outbreak investi-
gation in different scenarios. For instance, it can be used reactively
when epidemiology suggests there is an outbreak. Alternatively, it
may be used proactively as part of genomic surveillance in a high-
risk setting. This may involve sequencing multidrug-resistant or
hypervirulent organisms from patient samples or the environment
and/or prospectively sequencing a select population of patients
who may be vulnerable to infection or who may be from a critical
location such as an intensive care unit (ICU).
Different WGS techniques exist, and traditional short-read
sequencing (SRS) is the most used. More recent long-read
sequencing (LRS) also known as single-molecule or third-
generation sequencing can be used to sequence bacterial genomes
including the regions that cannot easily be sequenced and assem-
bled by SRS, thus enabling the retrieval of much longer
(>10,000 bp) sequencing reads than SRS systems (75–300 bp).
[11]The market leaders in LRS are Pacific Biosciences (PacBio) and
Oxford Nanopore Technologies (Nanopore). LRS also offers the
additional advantage of portability with the Oxford Nanopore
Technologies (ONT) MinION device designed to be as small as a

memory (USB) stick. The LRS MinION device has been applied in
the field during Ebola, Zika, and Coronavirus outbreaks and has
also been used in space [12–15]..
Analyzing high-throughput sequencing (HTS) data can pose a
challenge for researchers and clinicians lacking a background in
bioinformatics. Traditionally, bioinformatic support has always
been required. Bioinformatics brings together the storage of
sequencing data and the application of computational algorithms
to analyze this data [12]. A suitable suite of tools or a pipeline, a set
of scripts, should be selected based on what output is required from
WGS. This depends on whether the aim is to focus on pathogen
identification, further characterization including, virulence factors,
antibiotic resistance, or whether metagenomic approaches are
being untaken.
Comparative genomic approaches are essential to answer geno-
mic molecular epidemiological questions and understand the trans-
mission dynamics of pathogens [13]. After a clinical sample is taken
and cultured, DNA is extracted from bacterial isolates, and a DNA
library is created for sequencing. In principle, there are three ways
to generate microbial genomes: de novo assembly, reference
mapping, and assembly and alignment-free methods [16]. It is
important that quality of sequencing reads is assessed for all
sequencing platforms in terms of read quality, trimming poor qual-
ity and adaptor sequences, and demultiplexing the sample into
individual isolate read sets. [17] Quality control processes for
short-read sequence data analysis should detect miscalled bases as
sequencing machines carry out base calling in which “raw” signals
are converted to reads for each nucleotide [18]. Software tools such
as Fast QC can assess quality of data. Phred-like quality scores
(Q scores) can be used to measure the probability that bases have
been correctly called. In general, a score of Q25 and above is an
accepted score to continue processing the data [18]. Following this
quality can be assessed by filtering and pre-processing sequence
reads before the genome is assembled. A diagram of workflow for
assessing the performance and quality of data and processing in a
WGS service for outbreak detection can be seen in Fig. 1.
In de novo assembly, a priori knowledge of a reference organ-
ism is not required as overlapping reads are tiled or assembled into
longer fragments, contigs; contigs, not necessarily overlapping, are
assembled into scaffolds often aided by the paired-end sequencing,
and scaffolds can be assembled into a chromosome after
gap-closing or genome finishing process. This method may be
preferred for organisms that are difficult to characterize due to
high genomic plasticity and extrachromosomal sequences such as
plasmids. [17] Using short-read data can lead to challenges, partic-
ularly in the assembly of compound repeat regions, as it can be
more difficult to resolve such regions accurately [17]. Genome
Fig. 1 Recommendations for performance and quality assessment of data and processing in WGS service for
outbreak detection. Quality assessments of data are displayed in dark grey-colored boxes
assembly by de novo methods is generally more accurate and easier

with larger fragments from long-read sequencing such as sequences
produced by Pacific Biosciences and Nanopore instruments or by
combining short- and long-read data in a hybrid assembly. [17]
SPAdes is one of the most common de novo assemblers. [19, 20]
Other software packages include CLC Genomic Workbench (Qia-
gen), and Velvet [21, 22].
Assembly-free methods are a new development in comparative
microbial genomics. Usually, k-mer or hash-based methods are
used to examine query sequences [17]. They can be used for
rapid molecular analysis such as SNP-based methods with kSNP.
[23] SRST2 typing algorithm utilizes k-mers for sequence typing to
identify sequence types, including determining the presence or
absence of specific alleles or genes such as antimicrobial resistance
or virulence determinants [24].
Reference mapping, also known as guided assembly, is ideal in
situations when there is a well characterized reference genome and
organisms have low population diversity and when mobile genetic
elements do not need to be investigated [17]. Algorithms such as
Burrows-Wheeler Alignment tool (BWA) and bowtie include align-
ment and can be used for this. [25, 26] As part of alignment a
reference genome is used for mapping reads from the isolate being
investigated. Researchers look for differences between the two and
this forms the basis for SNP calling. [17] Occasionally certain reads
will not map to the reference genomes due to high degree of
variability, such as containing novel regions, and if this occurs
filtering is carried out. Pipelines in which reference mapping is
carried out can contain variant calling options such as GATK and
VarScan [27, 28]. Gene annotation can also be completed with
BreSeq or Prokka [29, 30]. Phylogenetics and prediction of resis-
tance can also be included. Free web-based interface programs are
now available online which can be used for inferring phylogenetics
directly from next-generation sequencing (NGS) data. Programs
such as CSI Phylogeny (Center for Genomic Epidemiology),
snpTree, NDtree, and Reference sequence Alignment-based Phy-
logeny builder (REALPHY) can be used [31–34].
A number of the above-mentioned and new scripts and pipe-
lines have been incorporated into Galaxy, a web-based platform
using open-source software to analyze sequencing data [36]. One
of the recently emerged pipelines is for instance SUPRI,
asequenced-based ultra-rapid pathogen identification from meta-
genomic data [35, 36]. Microbial identification and characteriza-
tion through read analysis (MICRA) is an automatic pipeline with a
web interface for fast characterization of microbial genomes
through read analysis [37]. By mapping against reference genomes
microorganism-specific genes and determinants of antibiotic
resistance can be identified. TORMES represents a simplified
way to perform WGS analysis [38]. It is a pipeline for
non-bioinformatician users which can be run using very simple

commands for the analysis of bacteria from any origin generated
by HTS on Illumina platforms. It automates sequence quality
filtering, de novo analysis, draft genome ordering against a refer-
ence, annotation, multi-locus sequence typing (MLST), and detec-
tion of antibiotic resistance and virulence genes [38]. Results can be
seen in an interactive web-like report open for sharing and
revising [38].
There are other useful resources available such as the virulence
factor database (VFDB) first published in 2005. [39] VFDB is a
powerful search engine which can be used for easy extraction of
information from its database by text search, BLAST search, or
virulence factors (VFs) function category search [39, 40].
For antibiotic resistance a number of tools are available
including ResFinder, ARG-ANNOT (Antibiotic Resistance Gene-
ANNOTation), and CARD (Comprehensive Antibiotic Resistance
Database). [41–43] Further information on establishing sequenc-
ing services for outbreak investigation and bioinformatics can be
found in the following references. (44, 45)
At present, there is no official centralized cloud or virtual
computing system for clinical laboratories to carry out pathogen
genomic analysis. Local computing clusters need to be established
to be sustainable and have the capacity for demand. Potentially, this
could be arranged with data centers such as the European Nucleo-
tide Archive (ENA) or Genomics England data center [18].
Researchers tend to use existing University-based high-perfor-
mance computing services, however issues may arise when compu-
tational capacity demand increases. Recently, guidance was
developed by the World Health Organization (WHO) detailing
the principles for pathogen genome data sharing. These cover the
main areas of capacity development, collaboration, high-quality,
reproducible data, global and regional representativeness, timeli-
ness of results, intellectual credit, equitable access along with the
themes of publicly accessible platforms, interoperability, transpar-
ency, and consistency with applicable law and ethical regulations
[46]. A systematic scoping review published in 2020 identified that
although pathogen sequencing can be transformative for public
health and outbreak investigation, various key ethical issues can
arise mainly in relation to how pathogen sequence data is used.
The authors suggested that ethical standards would naturally be
informed by the views and values of the public, and for this area to
progress there needs to be a strong commitment to values of
justice, including global health equity. [47].
Time and resources also need to be allocated to develop staff
skills in carrying out WGS and for the interpretation of results. It
would be essential that training would become integrated in curri-
cula for microbiology, infection, public health, and IPC training.
There are also residential courses such as “Genomics and Clinical

Microbiology” available at the Wellcome Genome Campus, Hinx-
ton, and online virtual courses that can be accessed. [48].
One of the key elements of implementing WGS into outbreak
investigation is the communication of results. This can often be
overlooked, and it is essential that sequencing results be conveyed
to IPCTs in a rapid, meaningful, and actionable way to aid outbreak
and patient management. Various papers have included evidence-
based design and evaluation steps for creating reference microbiol-
ogy laboratory reports. [49, 50] In Fig. 2, we have given an
example of a report that could be communicated to IPCTs.
2 Materials
The following section describes the wet laboratory work involved in

sequencing bacterial isolates for outbreak investigation. This is
based on the work carried out by the Scottish Healthcare Asso-
ciated Infection Prevention Institute (SHAIPI) which formed a
clinical WGS service to confirm or refute outbreaks in hospital
settings from across Scotland. Information on the benefits and
barriers to establishing a WGS outbreak service have been described
[44]. Detailed protocols are described below and it is good practice
to refer to the manufacturers’ most recent version of DNA extrac-
tion, library preparation, and sequencing protocols. We used Mas-
terPure Gram Positive DNA Purification Kit (Cambio, UK) and
Nextera XT (Illumina) library construction kits; however, others
are available. Additionally, users need to become familiar with
health and safety precautions while in the laboratory, such as
being particularly aware of safely working with buffers containing
formamide.
2.1 DNA Extraction • Cell pellet(s) of the bacterial isolate(s) of interest, based on
single colony pick, grown overnight or to late exponential
growth phase if longer in liquid growth medium
(107–109 CFU/mL), obtained by centrifugation (see Note 1).
• DNA extraction kit and associated materials. In this protocol for
the purposes of DNA extraction from Staphylococcus spp. we
suggest the MasterPure Gram Positive DNA Purification Kit
(Cambio), which requires:
– MasterPure Gram Positive DNA Purification Kit (Cambio,
UK).
– TE Buffer.
– Lysostaphin.
– Isopropanol.
– 70% ethanol (freshly prepared).
Fig. 2 An example of a WGS report. The report gives a score on the likelihood of an outbreak and an
explanation of the phylogeny and its clinical implications. Note that for case 1 the patient’s bacteriology
sample isolate was not stored by the microbiology laboratory as at that point in time an outbreak was not
suspected
– Vortex.
– Incubator at 37 °C.
– Ice bucket.
2.2 DNA Quality • Qubit Broad Range Assay Kit (ThermoScientific, UK).
Assessment and • Qubit High Sensitivity Assay Kit (ThermoScientific).
Quantification
• Qubit Assay Tubes (ThermoScientific).
• NanoDrop spectrophotometer (ThermoScientific).
• Qubit fluorometer (ThermoScientific).
• Optional—agarose and gel electrophoresis equipment, TAE
Buffer (Tris-acetate-EDTA), high molecular weight DNA lad-
der, DNA loading buffer.
2.3 Library • Nextera XT DNA library preparation kit (Illumina Inc., San
Preparation Diego, CA, USA).
• Nextera XT index kit (Illumina).
• PhiX (Illumina).
• Magnetic DNA-binding beads (such as AMPure XP Beads,
Beckman Coulter, UK).
• Absolute ethanol for molecular biology.
• NaOH.
• PCR grade water.
• 96-well thermal cycler with heated lid.
• Heat block for 1.5 ml centrifuge tubes.
• High-speed microplate shaker.
• Magnetic stand, 96-well format.
• Microplate centrifuge.
• Vortex.
• 96-well PCR plates.
• Multichannel pipettes, single channel pipettes, and sterile
filtered tips.
• 96-well storage plates, round well, 0.8 ml (deep well plate).
• Microseal “B” adhesive seals (BioRad).
• RNase/DNase-free multichannel reagent reservoirs, disposable.
• TruSeq Index Plate Fixture.
• Optional—2100 Bioanalyzer or Tape Station (Agilent, UK).
• Optional—Repeater pipette, 125 ml and 1250 ml syringes.
2.4 MiSeq Sequencer • 25 mL pipettes.

Preparation • Pipette filler.
• Tween 20.
• Lint-free lens wipes.
• Distilled water.
2.5 Illumina • MiSeq platform (Illumina Inc., San Diego, CA, USA).
Sequencing • MiSeq Reagent Kit (v2 or v3, Illumina).
3 Methods
All procedures are performed at room temperature unless other-

wise specified [51].
3.1 DNA Extraction Following the MasterPure Gram Positive DNA Purification Kit
(Cambio) protocol (see Note 2):
1. Centrifuge 1.0 mL of an overnight Gram-positive bacterial
culture to create a pellet. Discard the supernatant.
2. Add 150 μL of TE Buffer and vortex to resuspend the cell
pellet.
3. Add 1 μL of Ready-Lyse Lysozyme to each resuspended pellet.
4. Incubate at 37 °C for 30 min.
5. Make a mastermix of 1 μL of Proteinase K (50 μg/μL) and
149 μL of Gram-Positive Lysis Solution for each pellet. Add
150 μL of the Proteinase K/Gram Positive Lysis Solution to
each sample and mix thoroughly.
6. Incubate at 65–70 °C for 15 min, vortexing briefly every 5 min.
7. Cool samples to 37 °C and place on ice for 3–5 min.
8. Add 175 μL of MPC Protein Precipitation Reagent to 300 μL
of lysed sample. Vortex for 10 s to mix vigorously. Centrifuge at
4 °C for 10 min at >10,000 × g in a microcentrifuge to pellet
the precipitate. Transfer the supernatant to a clean microcen-
trifuge tube and discard the pellet.
9. Add 1 μL of RNase A (5 μg/μL) to each sample and mix
thoroughly.
10. Incubate at 37 °C for 30 min.
11. Add 500 μL of isopropanol, invert 30–40 times and then
centrifuge at 4 °C for 10 min at >10,000× g in a microcen-
trifuge. Remove the isopropanol without dislodging the DNA
pellet.
12. Rinse the pellet with 70% ethanol. Briefly re-centrifuge if the
pellet becomes dislodged.
13. Resuspend the DNA pellet in 35 μL of TE Buffer.
3.2 DNA Quality 1. Take a blank reading using 1.5 μL TE buffer.

Assessment and 2. Pipette 1.5 μL sample onto the cleaned pedestal of a Nano-
Quantification Drop spectrophotometer and read the DNA absorbance (see
3.2.1 DNA Purity Check
Note 3).
on a NanoDrop 3. Wipe the pedestal clean with a lint-free wipe after use.
Spectrophotometer 4. Pure DNA is indicated by absorbance (260 nm/280 nm) ratio
of >1.8–2.0 and a 260/230 absorbance ratio of 2.0–2.2.
3.2.2 DNA Molecular 1. Prepare a 1% agarose gel in 1 × TAE.

Weight and Fragmentation 2. Load the recommended volume of a high molecular weight
Assessment by Agarose DNA ladder (e.g., 23 kb or greater) to one well.
Gel Electrophoresis
3. To each well, load 10 μL DNA mixed with 2 μL gel loading
buffer.
4. Under UV, determine visually the integrity of the DNA sample
(see Note 4). There should be a large, e.g., >30 kbp molecular
weight band with minimal smearing.
3.2.3 DNA Quantity 1. Prepare Qubit™ solution for samples and 2 standards (1 μL
Qubit reagent and 199 μL Qubit buffer per sample and stan-
dard) and label the tops of the Qubit 0.5 mL tubes accordingly.
2. Distribute the Qubit™ solution into each tube: 190 μL for
tubes containing each standard and 198 μL for sample tubes.
3. Add 10 μL of the standard to its tube and 2 μL of each sample
to its tube.
4. Mix each tube by vortexing for 3–5 s.
6. Read the standards by inserting the tubes into the fluorometer
and after selecting the appropriate procedure (dsDNA BR for
broad range or dsDNA HS for High Sensitivity Qubit kit) press
Read standard.
7. Specify sample volume (2 μL) and units (ng/μL) and read the
fluorescence for each sample.
8. If negative controls are more than 1 ng/μL, repeat the previous
steps to address contamination.
9. We recommend first quantifying each DNA sample using the
Qubit Broad Range Assay Kit, diluting each DNA appropri-
ately if the concentration is too high, then re-quantifying using
the Qubit High Sensitivity Assay Kit. Each DNA sample should
be carefully diluted, in PCR grade water, to 0.4 ng/μL prior to
library preparation (see Note 5).
3.2.4 Sample and For WGS on a MiSeq instrument, it is important to determine how
Sequencing Kit Selection many samples of a given species should be included on a sequencing
run to obtain the maximum amount of quality data. A prepared
library will be sequenced using a MiSeq Reagent Kit (most up-to-
date version), containing a set of reagents in a pre-filled cartridge.

Reagent kits are designed to run a given number of cycles, and the
number of cycles is proportional to the number of reads generated.
Thus, a 500-cycle MiSeq v2 Reagent kit has the capacity to generate
sufficient sequence data to perform WGS on approximately 60 Mb
of genome. This equates to 20–24 Staphylococcus aureus samples
(genome of ~2.8 Mbp), or 12 samples (genome of ~5 Mbp).
Accordingly, a 600-cycle MiSeq v3 Reagent kit has capacity for
significantly more. The expected capacity of the reagent kit can be
used to plan how many samples of each species can be feasibly
included in a sequencing run.
3.3 Library The Nextera XT library preparation kit uses a single transposase
Preparation enzymatic reaction where sample DNA is simultaneously fragmen-
ted and tagged with sequencing adapters. The library is then ampli-
fied with sample-specific indices, cleaned up, and normalized before
pooling the indexed samples and adding the PhiX control for
sequencing.
3.3.1 Tagmentation 1. Thaw TD and ATM tubes on ice, gently invert 3–5 times and
(see Note 6) spin briefly in microcentrifuge.
2. Pipette 10 μL TD buffer to each well of a labeled 96-well PCR
plate.
3. Using a multichannel pipette, add 5 μL DNA (at 0.4 ng/
μL = 2 ng total) to each sample well of the plate.
4. Add 5 μL ATM mastermix to each well using a multichannel
pipette, and mix gently 5 times by pipette, changing tips
between samples.
5. Cover plate with a PCR seal and centrifuge at 280× g at 20 °C
for 1 min.
6. Incubate in a thermocycler (with heated lid) at 55 °C for 4 min,
followed by a hold step at 10 °C. Once the plate reaches 10 °C,
proceed immediately to the neutralization step.
3.3.2 Neutralization A. Briefly centrifuge the plate, carefully remove the PCR seal, and
add 5 μL chilled NT Buffer to each well using a multichannel
pipette. Pipette gently 5 times to mix, avoiding cross-
contamination.
B. Apply a new PCR seal, briefly centrifuge (280× g at 20 °C for
1 min) and incubate at room temperature for 5 min.
3.3.3 PCR Amplification 1. Thaw NPM and index primers on the bench at room tempera-
(see Note 7) ture, then mix gently by inverting the thawed tubes 3–5 times
and centrifuging them using 1.7 mL microcentrifuge tubes as
adaptors.
2. For the 24 libraries, arrange the index primers in the TruSeq

Index Plate Fixture, with index 1 (i7) primers (orange caps)
arranged in order horizontally (N701 in column 1, N706 in
column 6), and index 2 (i5) primers (white caps) in order
vertically (S517 in row A, S502 in row B, S503 in row C,
S504 in row D).
3. After a brief centrifugation, place the PCR plate in the TruSeq
Index Plate Fixture, remove the seal and add 15 μL NPM to
each well containing index primers using a multichannel
pipette.
4. Remove and discard the white and orange caps of the index
tubes. To avoid index cross-contamination, pipette tips should
be changed between samples. Add 5 μL of index 2 primers to
each column of the plate using a multichannel pipette, then add
5 μL of index 1 primers to each row. Re-close the tubes with
fresh caps. Pipette gently to mix, avoiding cross-
contamination. Apply a new PCR seal and centrifuge at
280× g at 20 °C for 1 min.
5. Place in a thermocycler and run PCR (with heated lid) under
the following conditions: 72 °C for 3 min and 95 °C for 30 s,
followed by 12 cycles of 95 °C for 10 s, 55 °C for 30 s, 72 °C
for 30 s, then a final step of 72 °C for 5 min and a hold at 10 °C.
The plate can remain on the cycler overnight or be stored at
2–8 °C for up to 2 days.
3.3.4 Post-PCR Clean-Up 1. Remove the magnetic beads from the fridge 30 min prior to
use, to allow them to warm to room temperature (see Note 8).
2. Centrifuge the PCR plate at 280× g at 20 °C for 1 min, remove
the seal and transfer 50 μL of PCR product to a labelled deep-
well plate.
3. Vortex the magnetic beads thoroughly for 30 s until fully
resuspended. Using a repeater pipette, immediately pipette
25 μL of beads into each well of the deep well plate (see Note
9). Aim for the sides of the well to avoid splashing.
4. Shake the deep-well plate on a microplate shaker at 1800 rpm
for 2 min, then incubate at room temperature without shaking
for 5 min.
5. Place the deep well plate on a magnetic stand for 2 min or until
the supernatant has cleared. Using a multichannel pipette,
carefully remove and discard the supernatant (changing tips
between each sample).
6. Keeping the plate on the magnet, pipette 200 μL freshly
prepared 80% ethanol (see Note 10) to each well using a
multichannel pipette. Leave the plate on the magnetic stand
for 30 s, then carefully pipette off and discard the supernatant.
7. Repeat step 6.
8. To remove any visible residual ethanol, use a P20 multichannel
pipette with fine tips.
9. Keeping the plate on the magnet, allow the pellet to air-dry for
15 min.
10. Remove the plate from magnet and pipette 45 μL RSB into
each well using a repeater pipette. Place the plate on a micro-
plate shaker at 1800 rpm for 2 min.
11. After shaking, incubate the plate at room temperature for
2 min, before returning it on the magnet for a further 2 min
or until the supernatant is clear.
12. Transfer 50 μL of supernatant (this is the cleaned-up library)
from the deep-well plate to a new labeled PCR plate using a
multichannel pipette. Seal the plate and store at -15 to -25 °C
for up to 1 week.
3.3.5 Manual Library 1. Quantify the cleaned libraries using the Qubit High Sensitivity
Normalization and Pooling Assay as previously described in Subheading 3.2.3, to provide
(see Note 11) the concentration of each library in ng/μL.
2. Run 1 μL of the cleaned library on an Agilent High Sensitivity
DNA chip on Bioanalyzer, following standard protocols (see
Note 12), to provide the average fragment size in bp.
3. Calculate the molarity of each library using the following
formula:
Molarity ðnMÞ = ðC=M LÞ 1e6
where
C = concentration (ng/μL, from Qubit)
M = average molar mass (660 g/bp)
L = average fragment length (bp, from Bioanalyzer)
4. Calculate the average concentration of the libraries and pool
them so that each library has a pooled concentration equal to
the average (see Note 13). The following formula is used for
this calculation:
V ind = ðMav 5 μLÞ=Mind
where
Vind = volume of individual library to add to pool (μL)
Mav = average library molarity (nM)
Mind = molarity of individual library (nM, from step C)
5. Keep pooled library in fridge until ready for sequencing. Do
not denature the pooled library until you are ready to begin the
sequencing run (see Note 14).
3.4 Preparing PhiX 1. Dilute 10 nM PhiX library to 4mN: In a 1.5 μL microtube,

Control (see Note 15) combine 5 μL 10 nM PhiX library and 3 μL 10 mM Tris-Cl,
pH 8.5 with 0.1% Tween 20.
2. Denature PhiX library: In a 1.5 μL microtube, combine 5 μL
4 nM PhiX library and 5 μL 0.2 N NaOH (freshly prepared).
Vortex briefly and centrifuge.
3. Incubate this 2 nm PhiX library for 5 min at room temperature
to denature.
4. Immediately dilute 10 μL of denatured 2 nm Phi-X library in
990 μl chilled HT1 buffer.
5. This 20 pM PhiX library can be stored for up to 3 weeks at -
15° to -25 °C (see Notes 15 and 16).
6. Denatured and diluted PhiX library will be spiked into the
denatured sequencing library prior to loading on the reagent
cartridge (see Subheading 3.6 for details).
3.5 Preparing the 1. Thaw the reagent cartridge in a basin of lukewarm water. When
MiSeq Instrument immersing the cartridge, do not exceed the “max water level”
mark on the cartridge.
2. In MiSeq Control Software, select “Manage Instrument”,
select “Reboot”.
3. Perform Maintenance Wash: three washes taking approxi-
mately 20 min each.
(a) Prepare fresh wash solution by diluting 25 mL Tween
20 (10%) in 475 mL MilliQ or Elga water.
(b) Add 6 mL wash solution to each reservoir of the wash tray
and transfer the remaining wash solution to the wash
bottle.
(c) In MiSeq Control Software, select “Perform Wash”, select
“Maintenance Wash” and follow the instructions.
4. Open Illumina Experiment Manager, select “Create Sample
Plate” and enter sample IDs and indices. Save and select “Cre-
ate Sample Sheet”. Select “Small Genome Resequencing”,
enter reagent cartridge barcode, and run ID. Set cycle number.
5. Make sure that there are at least 100Gb of free space in the
instrument memory.
6. Prepare fresh 0.2 N NaOH by diluting 200 μL 1 N NaOH in
800 μL sterile MilliQ water.
7. Remove a new flow cell from fridge/cold room. Remove the
flow cell from plastic container, retaining the storage solution
in the container. Flush the flow cell with MilliQ water to wash
off the storage solution, dry carefully with lint-free tissue. Add
several drops of 70% ethanol to lint-free tissue and wipe the
glass part of the flow cell to remove any residue or smears.
3.6 Denaturing the 1. In a 1.5 μL microtube, add 5 μL pooled library and 5 μL fresh
Pooled Library 0.2 N NaOH.
2. Briefly vortex and centrifuge.
4. Add 990 μL chilled HT1 buffer.
5. Invert several times to mix, then pulse centrifuge.
6. If PhiX control is to be used, either spike in at 1% or at 5% for
low diversity libraries:
Most Libraries (1%):
Denatured and diluted PhiX control 6 μL
Denatured and diluted sample library 594 μL
Low Diversity Libraries (≥ 5%).
Denatured and diluted PhiX control 30 μL
Denatured and diluted sample library 570 μL
7. Pierce the foil seal of the indicated reservoir of the reagent
cartridge with a sterile 1 mL pipette tip and load 600 μL of
denatured library combined with the PhiX library.
3.7 Beginning the 1. Open MiSeq Control Software, select “Sequence” and follow
Sequencing Run instructions to load the flow cell, reagent cartridge, and incor-
poration buffer.
2. Before the run can start, the MiSeq performs several checks:
wait until these have all been performed successfully (indicated
by a green tick), then select “Start Run”.
3. Vibration can negatively affect a MiSeq run. Avoid using any
centrifuges, plate shakers, or vortexes on the same bench as the
MiSeq instrument during the run.
4 Notes
1. After subculturing clinical isolates and picking a single colony

for overnight growth in the appropriate broth, such as Brain
Heart Infusion, a small amount of bacterial suspension (1 mL)
is centrifuged to produce a cell pellet which can be processed
immediately or stored at -20 °C.
2. The isolation of plentiful, high-quality DNA is paramount to
successful downstream applications, and whole genome
sequencing is no exception. The selection of a DNA extraction
protocol depends on a range of factors including bacterial
species, requirements of quantity and quality of DNA, and
time and cost considerations. When establishing our real-time
NGS pipeline in 2015, we compared several commercial DNA
extraction kits and found that, while the Promega Wizard Kit
was the least expensive of the kits tested and yielded high
molecular weight DNA with good levels of purity, the Cambio
MasterPure Gram Positive DNA Purification kit gave the high-
est yields for all routine bacterial species. This was particularly
beneficial for time-dependent sequencing of mixed batches of
clinical isolates, containing samples of different species.
3. A quick assessment of the purity of DNA in solution can be
conducted with minimal DNA wastage using a microvolume
spectrophotometer, such as the NanoDrop. The ratio of absor-
bance at 260 and 280 nm provides an insight into the relative
quantity of nucleic acids and proteins, which have absorbance
maxima at 260 nm and 280 nm, respectively. A 260/280 ratio
of approximately 1.8–2.0 is considered an acceptable indicator
of DNA purity. While protein absorbs maximally at 280 nm,
other carryover contaminants such as phenol or guanidine
absorb at 230 nm. Thus, the 260/230 ratio is also checked;
in this case, a ratio of 2.0–2.2 is optimal.
4. Assessing the molecular weight of extracted DNA is of less
importance in short-read sequencing than long-read methods,
where it is a possible determinant of read length. However,
when selecting a DNA extraction method, it is worthwhile
checking the approximate length of extracted DNA fragments.
This can be done cheaply and easily by electrophoresis using a
low percentage agarose gel and identifying excessive degrada-
tion visually under UV light. Alternatively, specialist instru-
ments are available for measuring fragment length of both
genomic DNA and sequencing libraries, such as TapeStation
(Agilent).
5. Quantification of DNA is particularly important for library
preparation. For a sequencing run to yield data that is repre-
sentative of the whole batch of input DNA, all samples must be
of a uniform concentration. Furthermore, the enzymatic pro-
cesses of library preparation require an optimized amount of
input DNA to be present. For these reasons, DNA quantifica-
tion is a major part of the sequencing preparation pipeline.
While spectrophotometry provides a good assessment of
DNA purity, it is strongly advisable to use a DNA-binding
fluorescence-based method for DNA quantification. These
methods use a fluorescent dye which binds to the minor groove
of double-stranded DNA, giving a more specific, sensitive, and
reproducible quantification of DNA concentration. DNA can
then be diluted in sterile PCR-grade water to a uniform and
optimal concentration.
6. As the name suggests, this is a process that combines fragmen-
tation and tagging in a single step. Transposons randomly
fragment the DNA to approximately 200-500 bp lengths,
while adding adapters to the 5′ and 3′ ends. The adapter

contains primer binding sites, index sequences, and hybridiza-
tion sequences. This is a short and comparatively simple step in
the library preparation process, but its importance cannot be
underestimated. The first consideration is that it is very sensi-
tive to the concentration of input DNA. If a DNA sample is too
diluted, the transposon enzymes present in ATM will over-
fragment the DNA, leading to too-short sequence reads for
that sample. Conversely, if a DNA sample is too concentrated,
there will be insufficient enzyme in the reaction to perform
tagmentation, and that sample will be under-represented in the
output sequence data. For this reason, careful fluorescence-
based quantification, and subsequent dilution to the optimal
concentration, is vital. However, depending on the precise
laboratory setup, the optimal concentration may not necessar-
ily be exactly that specified in the protocol, and it may be
worthwhile to test and compare slight modifications. For
example, we found that we obtained better results by using
double the recommended concentration of DNA for tagmen-
tation; we used an input DNA concentration of 0.4 ng/μL
rather than the manufacturer’s concentration of 0.2 ng/μL.
7. The amplification step is the stage where index primers are used
to simultaneously amplify and barcode each individual tagmen-
ted DNA sample. Up to 384 individual DNA samples can be
sequenced on a single MiSeq run, with the combination of
multiple barcode kits. Given that prepped samples are pooled
prior to loading, it is vital that index primer 1 (i5) and index
primer 2 (i7) are selected, recorded, and pipetted, as these are
the instrument’s only means of identifying each individual
sample during the sequencing run. Cross-contamination and
errors in pipetting are to be avoided. A specialist loading rack,
the TruSeq Index Plate Fixture, is provided with the
instrument.
8. There are many alternative brands of magnetic beads. For
laboratories with a very restricted budget, it is advisable to
test and compare samples of magnetic beads of various brands.
For the purposes of this chapter, we will describe our protocol
for AMPure bead clean-up.
9. AMPure are beads added to a given ratio, depending on the
purpose of the cleanup. Magnetic beads bind preferentially to
longer DNA fragments, so the higher the bead-to-sample ratio,
the smaller the fragments that will be bound. For Nextera XT
library prep, 25 μL of beads are added to 50 μL of library.
10. Ethanol for washing must be freshly prepared at 80%. A more
dilute ethanol will remove DNA from the beads along with any
impurities, and the library will be lost.
11. As previously mentioned, it is important to ensure that all

libraries are of a uniform concentration prior to pooling, oth-
erwise the sequence data will not have an even representation
of all the samples included. Because successful Illumina
sequencing is highly dependent on loading an optimal number
of DNA molecules to the flow cell—as they hybridize to the
flow cell, producing clonal clusters during bridge amplification,
and over- and under-clustering is a major reason for poor run
performance—it is necessary to calculate the molarity of each
library. This involves measuring both the concentration of
double-stranded DNA in solution (ng/μL), and the average
fragment size (bp).
12. The details of the method can be found at https://www.
agilent.com/cs/library/usermanuals/public/2100_Bio
analyzer_HSDNA_QSG.pdf
13. The conventional method is to dilute all samples in HT1 buffer
to the concentration of the most dilute sample. However, this
means that low-concentration samples must be excluded from
the sequencing run if they are below the required concentra-
tion. By pooling the libraries to the average concentration,
rather than the lowest concentration, it is possible to minimize
the exclusion of low-concentration libraries. Instead of diluting
libraries in HT1, the most dilute libraries act as the diluent for
the most concentrated libraries. There is still a lower limit, as
some libraries are too dilute to contribute equally to the overall
pool within the required volume limits. As well as enabling
moderately poor-yield libraries to be included, maximizing the
concentration of the normalized pooled library allows for
greater flexibility in optimizing the cluster density. While the
Illumina protocol specifies 4 nM as the optimal loading molar-
ity for a library https://support.illumina.com/content/dam/
illumina-support/documents/documentation/system_docu
mentation/miseq/miseq-denature-dilute-libraries-guide-1
5039740-10.pdf, we found that loading slightly higher molar-
ity libraries improved clustering on our instrument. In our own
laboratory, 4.5 nM was optimal to give a cluster density of one
million per mm2, and quality and quantity of data was maximal
at up to 6 nM. We recommend adjusting loading molarities
slightly if cluster density is not optimal at the specified molarity.
14. Denaturing the library is the point of no return for MiSeq.
Once the library is denatured, it must be loaded as soon as
possible, and the run begun. For that reason, it is important to
ensure that the reagent cartridge is thawed, the instrument is
washed and ready, and that there is sufficient data storage space
(at least 100Gb) for the run to complete.
15. Using the PhiX library as an internal quality control is optional

for WGS; however, if you are performing targeted sequencing
with PhiX spike-in, the run will fail. The reason for this is that,
while WGS sequences random fragments across the genome,
targeted sequencing focuses on a much narrow range of
sequences, and the relative proportions of the nucleotides are
more likely to be imbalanced. As a result, targeted sequencing
is “low diversity” and requires the addition of PhiX to com-
pensate for this. Illumina recommends a low-concentration
PhiX control spike-in at 1% for most libraries. For
low-diversity libraries, increase the PhiX control spike-in to at
least 5%.
16. After 3 weeks of storage cluster numbers tend to decrease.
Acknowledgments
This work was supported by the Wellcome Trust ISSF award [grant
number 097831/Z/11/Z]; and the Chief Scientist Office through
the Scottish Infection Research Network [SIRN10].
Thanks to Voòng Vı̃nh Phát (OUCRU, Ho Chi Minh City,
Vietnam) for several library preparation optimization suggestions.
References
1. U.S. Department of Health and Human Ser- disease outbreak analysis using the R software.
vices and Centers for Disease Control and Pre- Epidemics 7:28–34
vention (CDC) (2012) Principles of 8. Wellcome Sanger Institute MicroReact
Epidemiology in Public Health Practice, 9. Gladstone RA, Lo SW, Goater R et al (2020)
Third Edition: An Introduction to Applied Visualizing variation within Global Pneumo-
Epidemiology and Biostatistics, coccal Sequence Clusters (GPSCs) and country
2. National Services Scotland (NSS) Chapter 3 – population snapshots to contextualize pneu-
Healthcare Infection Incidents, outbreaks and mococcal isolates. Microb Genom 6:e000357
Data Exceedance. National Infection Preven- 10. Woolhouse M (2017) Quantifying
tion and Control Manual. https://www.nipcm. Transmission 5
hps.scot.nhs.uk/chapter-3-healthcare-infec
tion-incidents-outbreaks-and-data- 11. Johnson E, Raza S University of Cambridge –
exceedance/ What is long read sequencing? https://www.
phgfoundation.org/briefing/what-is-long-
3. Mayhall CG (2011) Hospital Epidemiology read-sequencing
and Infection Control
12. Quick J, Loman NJ, Duraffour S et al (2016)
4. Baxter ICNET Clinical Surveillance Software. Real-time, portable genome sequencing for
https://www.icnetsoftware.com/ Ebola surveillance. Nature 530:228–232
5. Baker MA, Yokoe DS, Stelling J et al (2020) 13. de Jesus JG, Giovanetti M, Rodrigues Faria N,
Automated outbreak detection of hospital- Alcantara LCJ (2019) Acute vector-borne viral
associated pathogens: value to infection pre- infection: Zika and MinION surveillance.
vention programs. Infect Control Hosp Epide- Microbiol Spectr 7(4). https://doi.org/10.
miol 41:1016–1021 1128/microbiolspec.AME-0008-2019.
6. Smith CM, Allen DJ, Nawaz S et al (2019) An PMID: 31400093.
interactive data visualisation application to 14. Cheng VCC, Ip JD, Chu AWH et al (2022)
investigate nosocomial transmission of infec- Rapid spread of severe acute respiratory syn-
tions. Wellcome Open Res 4:100 drome coronavirus 2 (SARS-CoV-2) Omicron
7. Jombart T, Aanensen DM, Baguelin M et al subvariant ba.2 in a single-source community
(2014) OutbreakTools: a new platform for outbreak. Clin Infect Dis 75:e44–e49
15. Bai N University of California San Francisco- 28. Koboldt DC, Chen K, Wylie T et al (2009)
NASA’s DNA Sequencing in Space is a Success, VarScan: variant detection in massively parallel
UCSF Researchers Confirm. https://www. sequencing of individual and pooled samples.
ucsf.edu/news/2016/08/404006/nasas- Bioinformatics 25:2283–2285
dna-sequencing-space-success-ucsf- 29. Deatherage DE, Barrick JE (2014) Identifica-
researchers-confirm#:~:text=The%20pocket% tion of mutations in laboratory-evolved
2Dsized%20MinION%20employs,to%20 microbes from next-generation sequencing
decode%20the%20DNA%20sequence data using breseq. Methods Mol Biol 1151:
16. Fan H, Ives AR, Surget-Groba Y et al (2015) 165–188
An assembly and alignment-free method of 30. Seemann T (2014) Prokka: rapid prokaryotic
phylogeny reconstruction from next- genome annotation. Bioinformatics 30:2068–
generation sequencing data. BMC Genomics 2069
16:522 31. Center for Genomic Epidemiology CSI Phy-
17. MacCannell D (2019) Platforms and analytical logeny. https://cge.food.dtu.dk/services/
tools used in nucleic acid sequence-based CSIPhylogeny/
microbial genotyping procedures. Microbiol 32. Leekitcharoenphon P, Kaas RS, Thomsen MCF
Spectr 7:AME-AME-0005-2018 et al (2012) snpTree--a web-server to identify
18. Luheshi L, Raza S, Moorthie S, et al (2015) and construct SNP trees from whole genome
PHG foundation- pathogen genomics into sequence data. BMC Genomics 13(Suppl 7):S6
practice 33. Gustavsson P, Syberfeldt A (2018) A new algo-
19. Deurenberg RH, Bathoorn E, Chlebowicz MA rithm using the non-dominated tree to
et al (2017) Application of next generation improve non-dominated sorting. Evol Comput
sequencing in clinical microbiology and infec- 26:89–116
tion prevention. J Biotechnol 243:16–24 34. Bertels F, Silander OK, Pachkov M et al (2014)
20. Bankevich A, Nurk S, Antipov D et al (2012) Automated reconstruction of whole-genome
SPAdes: a new genome assembly algorithm and phylogenies from short-sequence reads. Mol
its applications to single-cell sequencing. J Biol Evol 31:1077–1088
Comput Biol 19:455–477 35. Naccache SN, Federman S, Veeraraghavan N
21. Zeden MS, Gründling A (2023) Bacterial et al (2014) A cloud-compatible bioinformatics
whole-genome-resequencing analysis: basic pipeline for ultrarapid pathogen identification
steps using the clc genomics workbench soft- from next-generation sequencing of clinical
ware, vol 2023. Cold Spring Harb Protoc, samples. Genome Res 24:1180–1192
p pdb.prot107901 36. Galaxy. https://usegalaxy.org/
22. Zerbino DR, Birney E (2008) Velvet: algo- 37. Caboche S, Even G, Loywick A et al (2017)
rithms for de novo short read assembly using MICRA: an automatic pipeline for fast charac-
de Bruijn graphs. Genome Res 18:821–829 terization of microbial genomes from high-
23. Gardner SN, Slezak T, Hall BG (2015) throughput sequencing data. Genome Biol
kSNP3.0: SNP detection and phylogenetic 18:233
analysis of genomes without genome align- 38. Quijada NM, Rodrı́guez-Lázaro D, Eiros JM
ment or reference genome. Bioinformatics 31: et al (2019) TORMES: an automated pipeline
2877–2878 for whole bacterial genome analysis. Bioinfor-
24. Inouye M, Dashnow H, Raven L-A et al matics 35:4207–4212
(2014) SRST2: rapid genomic surveillance for 39. Chen L, Yang J, Yu J et al (2005) VFDB: a
public health and hospital microbiology labs. reference database for bacterial virulence fac-
Genome Med 6:90 tors. Nucleic Acids Res 33:D325–D328
25. Li H, Durbin R (2009) Fast and accurate short 40. Altschul SF, Madden TL, Sch€affer AA et al
read alignment with Burrows-Wheeler trans- (1997) Gapped BLAST and PSI-BLAST: a
form. Bioinformatics 25:1754–1760 new generation of protein database search pro-
26. Langmead B, Trapnell C, Pop M et al (2009) grams. Nucleic Acids Res 25:3389–3402
Ultrafast and memory-efficient alignment of 41. Florensa AF, Kaas RS, Clausen PTLC et al
short DNA sequences to the human genome. (2022) ResFinder – an open online resource
Genome Biol 10:R25 for identification of antimicrobial resistance
27. McKenna A, Hanna M, Banks E et al (2010) genes in next-generation sequencing data and
The genome analysis toolkit: a mapreduce prediction of phenotypes from genotypes.
framework for analyzing next-generation Microb Genom 8
DNA sequencing data. Genome Res 20: 42. Gupta SK, Padmanabhan BR, Diene SM et al
1297–1303 (2014) ARG-ANNOT, a new bioinformatic
tool to discover antibiotic resistance genes in 48. Wellcome Connecting Science Courses and
bacterial genomes. Antimicrob Agents Che- Conferences -Genomics and Clinical Microbi-
mother 58:212–220 ology. https://coursesandconferences.
43. Alcock BP, Huynh W, Chalil R et al (2023) wellcomeconnectingscience.org/event/geno
CARD 2023: expanded curation, support for mics-and-clinical-microbiology-2024-
machine learning, and resistome prediction at 20240114/
the comprehensive antibiotic resistance data- 49. Crisan A, McKee G, Munzner T et al (2018)
base. Nucleic Acids Res 51:D690–D699 Evidence-based design and evaluation of a
44. Parcell BJ, Gillespie SH, Pettigrew KA et al whole genome sequencing clinical report for
(2021) Clinical perspectives in integrating the reference microbiology laboratory. PeerJ
whole-genome sequencing into the investiga- 6:e4218
tion of healthcare and public health outbreaks – 50. Ferdinand AS, Kelaher M, Lane CR et al
hype or help? J Hosp Infect 109:1–9 (2021) An implementation science approach
45. Parcell BJ, Oravcova K, Pinheiro M et al to evaluating pathogen whole genome
(2018) Pseudomonas aeruginosa intensive sequencing in public health. Genome Med
care unit outbreak: winnowing of transmissions 13:121
with molecular and genomic typing. J Hosp 51. MasterPure Gram Positive DNA Purification
Infect 98:282–288 Kit Manual. https://cambio.co.uk/library/
46. (2022) WHO guiding principles for pathogen images/html_images/2022%20protocols/
genome data sharing. Presented at the manual_MGP04100_MasterPure-Gram-
47. Johnson S, Parker M (2020) Ethical challenges Positive-DNA.pdf
in pathogen sequencing: a systematic scoping
review. Wellcome Open Res 5:119
Chapter 17
Use of Whole Genome Sequencing for Mycobacterium

tuberculosis Complex Antimicrobial Susceptibility Testing
Vanessa Mohr, Lindsay Sonnenkalb, Christian Utpatel, Ivan Barilar,
Margo Diricks, Viola Dreyer, Stefan Niemann, Thomas A. Kohl,
and Matthias Merker
Abstract
Whole genome sequencing (WGS) is becoming an important diagnostic tool for antimicrobial susceptibility
testing of Mycobacterium tuberculosis complex (MTBC) isolates in many countries. WGS protocols usually
start with the preparation of a DNA-library: the critical first step in the process. A DNA-library represents
the genomic content of a DNA sample and consists of unique short DNA fragments. Although available
DNA-library protocols come with manufacturer instructions, details of the entire process, including quality
controls, instrument parameters, and run evaluations, often need to be developed and customized by each
laboratory to implement WGS technology effectively. Here, we provide a detailed workflow for a
DNA-library preparation based on an adapted Illumina protocol optimized for the reduction of reagent
costs.
Key words WGS, Whole genome sequencing, DNA-library, Mycobacterium tuberculosis complex
1 Introduction
Tuberculosis is a chronic and long-lasting disease affecting more

than ten million patients globally, and requires a combination
therapy of at least four different antibiotics [1]. Rapid identification
of existing or emerging antimicrobial resistances is essential to
design an optimal personalized therapy to ensure high cure rates.
This is especially true for patients with multidrug-resistant tuber-
culosis (MDR-TB), i.e., resistance against both isoniazid and rifam-
picin. The current gold standard to determine antimicrobial
susceptibility are in vitro phenotypic assays; however, it can take
weeks to report results to the clinic due to the slow growth of
Mycobacterium tuberculosis complex (MTBC) strains. Therefore,
targeted rapid molecular tests are also endorsed by WHO as a
diagnostic option to rule-in resistance against selected antibiotics
https://doi.org/10.1007/978-1-0716-3981-8_17,
185
186 Vanessa Mohr et al.
[2]. In 2021, the WHO presented the first catalog of mutations and
their association with antimicrobial resistance which can be inter-
rogated with next generation sequencing (NGS) technology [3].
The first step in the NGS workflow for WGS is the preparation
of a DNA-library. For short-read NGS applications, DNA quality
such as fragmentation is of less concern, but DNA molecular con-
centration needs to be determined accurately, e.g., by fluorometric
assays (see Materials below). A DNA-library is a collection of short
(~500–1200 bp) random DNA fragments representing the geno-
mic content of a biological DNA sample. Although the DNA-
library preparation is a crucial process that can influence the quality
of the sequencing data, there is usually little detailed information in
scientific publications on quality controls, instrument parameters,
and run evaluations.
In fact, there are several library-preparation protocols available,
with state-of-the-art kits requiring as little as 1 ng of input DNA.
For most protocols, the initial DNA undergoes a random fragmen-
tation which can occur via mechanical or enzymatic shearing.
Enzymatic shearing is often the method of choice as it usually
requires less input material, no additional laboratory equipment,
has a significantly shorter turnaround time, and low-cost protocols
can be applied [4]. After or during the shearing procedure, short
oligonucleotides are ligated/attached to the DNA fragments.
These short flanking sequences function as primer binding sites
for the subsequent PCR amplification during which indices (for
sample tracking) and adapters (for attachment to the flow cell, and
sequencing reaction) are integrated into the final DNA-library.
As a showcase method for this chapter, we will present a previ-
ously published library preparation protocol based on Illumina
reagents [4], offering excellent performance at relatively low
costs. Likewise, the NGS sequencing run and run evaluation will
be exemplarily described using the Illumina NextSeq 500 instru-
ment. Most of the workflow principles can be easily transferred to
other common library preparation solutions. For a list of abbrevia-
tions and definitions, see Table 1.
This chapter should be seen in combination with Chap. 18
which presents the workflow for bioinformatic analysis of raw
sequencing data (fastq files) and the detection of mutations impli-
cated in antimicrobial resistance.
2 Materials
2.1 DNA-Library 1. Fluorometer with at least 0.1 ng/μL sensitivity.

Preparation Adapted 2. Reagents for DNA measurements and 1 dsDNA HS assay kit
from Baym et al. [4] from Thermo Fisher Scientific.
3. Thermocycler.
Use of Whole Genome Sequencing for Mycobacterium tuberculosis. . . 187
Table 1
Abbreviations and definitions
Abbreviation Comment
Fastq Raw sequencing data in text format with base quality values
DNA-library Random collection of DNA fragments
Index Unique sequence used to label a NGS library for multiplexing
Run Multiplexing approach on a DNA sequencing machine
NGS Next generation sequencing
WGS Whole genome sequencing (an NGS application)
MDR-TB Multidrug-resistant tuberculosis
MTBC Mycobacterium tuberculosis complex
Tagmentation The process of tagging and enzymatic fragmentation of genomic DNA
TDE1 Tagment DNA enzyme
TD buffer Tagment DNA buffer
UD index Unique dual index
EB Elution buffer
KAPA A ready-to-use PCR master mix
HT1 Hybridization buffer
PhiX A ready-to-use control library of phage DNA
4. Vortex.
5. Centrifuge for 96-well plates and 1,5 mL microcentrifuge
tubes.
6. Tag DNA Enzyme & Buffer Large Kit (Illumina) (includes
TDE1 and TD buffer). This should be stored at 20 C.
7. IDT for Illumina—Nextera DNA UD Index Set A, B, C, D,
stored at 20 (each kit includes 96 unique index combina-
tions for altogether 192 samples).
8. KAPA HiFi Library Amplification Kit (Roche), stored in
1.5 mL aliquots at 20 C. Serves as PCR mastermix.
9. Magnetic rack like DynaMag™-PCR Magnet (Thermo Fisher
Scientific) and magnetic beads, e.g., MagSi-NGS Prep Plus
(Steinbrenner).
10. EB (5 mM Tris/HCL, pH 8.5) or molecular biology grade
water for elutions and dilutions. Avoid buffers with EDTA, this
will negatively interact with the sequencing chemistry (see
Note 1).
11. Capillary-Gel-Electrophoresis device, e.g., fragment analyzer

or bioanalyzer (Agilent Technologies) and analysis kit to mea-
sure the quality and quantity of fragments up to 6000 bp, e.g.,
Standard Sensitivity NGS Fragment Analysis Kit, 1 bp–
6000 bp.
12. 80% molecular grade ethanol.
13. Common consumables like 96-well plates, pipet tips with filter,
1.5 mL microcentrifuge tubes, 8-well-stripes, reservoirs, seal-
ing mats or adhesive film for 96-well plates. Preferably all of
low DNA binding materials.
14. Genomic DNA preferably with the following quality and quan-
tity metrics: at least 20 μL with a concentration of 20 ng/μL
(photometer) or 2 ng/uL (fluorometer), OD 260/280 1.8–2,
OD 230/260 2. Deviating DNA qualities/quantities have
severely reduced chances to be successfully sequenced.
2.2 DNA Sequencing 1. NGS instrument, e.g., NextSeq 500 (see Note 2).
2. Sequencing reagents suitable for the sequencer and sample
number, e.g., NextSeq v2.5 Mid Output 300 Cycles or Next-
Seq v2.5 High Output 300 Cycles.
3. 200 mM Tris-HCL pH 7.
3 Methods
3.1 DNA-Library 1. Dilute 96 genomic DNA samples to ~2 ng/μL using molecular

Preparation Adapted grade water or EB.
from Baym et al. [4] 2. Verify the concentration by Qubit.
3. Transfer 1 μL of 2 ng/μL DNA-dilution into a new 96-well
PCR plate.
4. Prepare tagmentation mix by combining 156 μL TD buffer and
31.2 μL TDE1 enzyme in a new tube. Distribute 22.5 μL in
each well of an 8-well PCR stripe. This enables the use of a
multichannel pipet for further distribution of 1.5 μL to each
well in the 96-well plate.
5. Close the PCR plate with a sealing mat to make it airtight.
6. Mix by vortexing briefly, spin down the sample plate for several
seconds and start the tagmentation program in the thermocy-
cler as seen in Table 2 (see Note 3).
7. Proceed immediately with the remaining protocol after the
thermocycler has reached 10 C and keep the samples on ice
or in a cooling rack.
Table 2
Tagmentation program on a PCR machine
55 C Hold

55 C 10 min
10 C Cool down
Table 3
PCR program adding indices and adapters to finalize the DNA-library
72 C 3 min
98 C 5 min
98 C 10 s 14 cycles
63 C 30 s
72 C 30 s
72 C 5 min
10 C Cool down
8. Distribute 145 μL KAPA PCR Mix to each well of an 8-well

PCR stripe. This enables the use of a multichannel pipet for
further distribution of 11 μL to each sample well of the 96-well
PCR plate.
9. Transfer 8 μL Nextera DNA UD Index from the index plate to
each sample (see Note 4).
10. Close the PCR plate with a sealing mat to make it airtight, mix
by vortexing briefly, spin down the sample plate for several
seconds, and start the PCR program in the thermocycler
(Table 3) (see Notes 5 and 6).
11. For the following magnetic bead clean-up, the beads need to
be at room temperature, well mixed and dissolved. Pour the
beads in a sterile reservoir for liquids and use a multichannel
pipette to add 13,5 μL beads to each sample and mix well (see
Note 7).
12. Incubate 5 min at room temperature before placing the 96-well
plate on a magnetic rack. Wait 1–2 min until the solution
becomes clear again and the beads have collected on the well
side with the magnet.
13. Carefully remove the supernatant without touching the beads
(see Note 8).
14. With the plate still on the magnet, add 200 μL of 80% ethanol
to each sample well (see Note 9). Incubate for 1 min, then
remove and discard the supernatant.
15. Repeat steps 13 and 14.

16. Carefully remove all residual liquid (see Note 10).
17. Dry the beads for approximately 2–5 min at room temperature
(see Note 11).
18. Remove the 96-well plate from the magnetic rack and suspend
beads in 26 μL EB buffer (see Note 7).
19. Incubate 5 min at room temperature before placing the 96-well
plate on a magnetic rack. Wait 1–2 min until the solution
becomes clear again and the beads have collected on the well
side with the magnet.
20. Transfer 25 μL of the supernatant, which contains the finished
library, without beads, to a new 96-well plate.
21. Finished libraries can be stored at 20 C.
22. Proceed with quality control via capillary gel electrophoresis to
determine the concentration in ng/μL and the fragment
length.
23. Calculate the concentration in nM, using the following for-
mula [5]:
concentration in ng=μL
106
ð660 g=mol average library size in bpÞ
¼ concentration in nM
3.2 DNA Sequencing 1. Plan the sequencing run by selecting the libraries to be multi-
on the NextSeq 500 plexed (see Note 12).
Platform 2. Thaw reagent cartridge in a water bath at room temperature
(~2–3 h).
3. Prepare the sample sheet in the NextSeq 500 local run manager
(see Note 13).
4. Dilute all libraries to 2 nM with EB (see Note 14).
5. 30 Minutes prior to starting, bring the flow cell to room
temperature.
6. Combine 2.5 μL of each library with the same molar concen-
tration in a new tube.
7. Combine in a new tube 10 μL of 2 nM Pool and 10 μL of 0.2 M
NaOH and incubate for 5 min at room temperature to achieve
denaturation.
8. Add 10 μL of 200 mM Tris-pH 7.0 and 970 μL HT1.
9. Dilute the pool to its final loading concentration and add PhiX
(see Note 15).
1.3 pM 1.4 pM 1.5 pM

20 pM denatured pool 84.5 μL 91 μL 97.5 μL
HT1 buffer 1215.5 μL 1209 μL 1202.5 μL
Denatured PhiX library 20 pM 2 μL 2.1 μL 2.2 μL
10. Before starting a run perform a complete power cycle of the

sequencer.
11. Rotate the reagent cartridge a couple of times to mix the
reagents.
12. Pierce well 10 on the cartridge with a clean pipette tip and
dispense the whole volume (1300 μL) of final pool in it (see
Note 16).
13. Follow instructions on the instrument and verify the para-
meters (run number, read length, paired end, index reads).
14. Wait until all system checks are done and the sequencer starts
the run.
3.3 Run Evaluation of In this section one ensures that the products from the run have
a NextSeq 500 Run good quality and meet the pre-defined expected specifications.
Evaluating run performance and sequence data quality is a complex
endeavor. In the following, we list some crucial steps especially for
runs on Illumina instruments.
1. Load the run into the Illumina Sequencing Analysis Viewer
(SAV).
2. On the first tab (Analysis), review the Q30 value. This value
should be within specifications.
3. Open the third tab (Summary). The total yield should be
within specifications for the run chemistry, i.e., for the NextSeq
500 100–120 Gb for a High-Output kit and 32.5–39 Gb for a
Mid-Output kit.
4. Also in the third tab, check whether the density on the flow cell
is in the recommended range, i.e., between 170 and 220 K/
mm2 for the NextSeq 500.
5. Still in the third tab, the value of PF % indicates the number of
flow cell clusters passing quality filters.
6. Open the fourth tab (Indexing). The bars indicate the amount
of sequencing data attributed to each input library. When
DNA-libraries were loaded with equal molarities, number of
reads/bp per sample should be evenly distributed. Obvious
outliers indicate errors during the dilution step to 2 nM (see
Subheading 3.2, step 4).
4 Notes
DNA-Library preparation adapted from Baym et al. [4].

1. The DNA sample should not contain EDTA as it interferes
with the tagmentation enzyme.
2. Which sequencer to choose depends on the number and type
of samples, desired sequencing data, and of course instrument
availability.
3. The tagmentation enzyme is thermoactive, which makes this
step time critical. Incubating too long would result in shorter
fragments, while incubating too short can result in too long
fragments. It is advised to include a preheating step in the
thermocycler program.
4. The index combination is a unique nucleotide sequence that
links the sequence data to the sample. Be very careful at this
step, pipetting errors at this point can lead to a sample mix-up
after sequencing.
5. This step can run overnight. In this case make sure that the
cooldown step to 4 C is set to infinity.
6. The non-purified PCR reaction can be stored up to 2 days at
4 C.
7. Added magnetic beads need to be well-mixed if one is to
achieve a good sample recovery. Therefore, when adding the
beads to the sample or the other way around pipette at least
10 times up and down and try to soak up the library-bead mix
from the bottom of the well and dispense on top. An alternative
would be, to add the beads to the library close the PCR plate,
mix by vortexing until the beads are incorporated and then spin
down the plate very briefly to collect the liquid at the bottom of
the well. By varying the ratio of PCR amplification to beads,
one can influence the size distribution of purified libraries.
8. The PCR product is bound to the beads, disturbing the pellet
or discarding beads, will lower the output concentration of the
library.
9. Do not flush the beads from the wall. Try to add the ethanol to
the opposite side of the well.
10. There are two satisfactory methods that can be used: removing
the large amount of liquid with 200 μL tips and afterward with
10 μL tips or first spin down the plate very briefly to collect all
residual ethanol at the bottom and then pipette of with a
200 μL tip.
11. Don’t over dry the pellet! Over drying can cause a loss of yield.
The beads should still have a light shimmer.
12. Selecting libraries to be multiplexed is a crucially important and

complex process. Libraries need to be compatible, most impor-
tantly having unique index combinations and ideally indices of
the same length, and libraries should ideally have a similar
length distribution. For non-randomly fragmented input
material, special considerations apply. The number of libraries
that can be multiplexed in one run depends on the total
sequence data produced by the run, the genome size of the
individual samples, and the desired average genome coverage,
i.e., sequence data per sample.
13. The local run manager is the Illumina software steering the
respective instruments, and will give a warning if for some
reason an index combination is doubled, therefore it can be
helpful to prepare the sample sheet prior to dilutions and
pooling.
14. Dilutions, higher or lower than 2 nM, are also possible. The
key is to dilute all libraries, which are supposed to be on one
sequencing run, to the same nM concentration to ensure equal
distribution of sequencing data.
15. Final concentration depends on the individual Illumina
machine. We recommend to load a lower molarity of the
pooled DNA-library, evaluate the run, i.e., clusters per mm2,
Q30 passing filter, total output, and subsequently adjust the
input concentration of the pooled DNA-library.
16. Check if any air bubbles are present, if so remove them by
flicking the well.
References
1. WHO consolidated guidelines on tuberculosis: mycobacterium-tuberculosis-complex-and-

module 4: treatment: drug-susceptible tubercu- their-association-with-drug-
losis treatment. https://www.who.int/ resistance. Accessed 17 Nov 2022
publications-detail-redirect/9789240048126. 4. Baym M, Kryazhimskiy S, Lieberman TD et al
Accessed 17 Nov 2022 (2015) Inexpensive multiplexed library prepara-
2. WHO consolidated guidelines on tuberculosis: tion for megabase-sized genomes. PLoS One
module 3: diagnosis: rapid diagnostics for tuber- 10:e0128036. https://doi.org/10.1371/jour
culosis detection, 2021 update. https://www. nal.pone.0128036
who.int/publications-detail-redirect/ 5. Converting ng/μL to nM when calculating
9789240029415. Accessed 17 Nov 2022 dsDNA library concentration. https://knowl
3. WHO releases the first catalogue of mutations in edge.illumina.com/library-preparation/dna-
Mycobacterium tuberculosis complex and their library-prep/library-preparation-dna-library-
association with drug resistance. https://www. prep-reference_material-list/000001240.
who.int/news/item/25-06-2021-who- Accessed 4 Apr 2023
releases-the-first-catalogue-of-mutations-in-
Chapter 18
Use of Whole Genome Sequencing for Mycobacterium

tuberculosis Complex Antimicrobial Susceptibility Testing:
From Sequence Data to Resistance Profiles
Viola Dreyer, Lindsay Sonnenkalb, Margo Diricks, Christian Utpatel,
Ivan Barilar, Vanessa Mohr, Stefan Niemann, Thomas A. Kohl,
and Matthias Merker
Abstract
Whole genome sequencing of Mycobacterium tuberculosis complex (MTBC) isolates has been shown to
provide accurate predictions for resistance and susceptibility for many first- and second-line anti-tubercu-
losis drugs. However, bioinformatic pipelines and mutation catalogs to predict antimicrobial resistances in
MTBC isolates are often customized and detailed protocols are difficult to access. Here, we provide a step-
by-step workflow for the processing and interpretation of short-read sequencing data and give an overview
of available analysis pipelines.
Key words Mycobacterium tuberculosis complex, Resistance prediction, Bioinformatics pipeline,

Mutation catalog
1 Introduction
Strains of Mycobacterium tuberculosis complex (MTBC) are highly

clonal and resistance against antibiotics is solely mediated by muta-
tion of the bacterial “chromosome”. Thus, the detection of
resistance-mediating mutations, e.g., single nucleotide polymorph-
isms (SNPs), and insertions or deletions (indels), can rule-in a
drug-resistant phenotype. In this regard, whole genome sequenc-
ing (WGS) offers the highest resolution and can interrogate all
genes implicated in antibiotic resistance [1]. A major breakthrough
in clinical TB research was the landmark study of the CRyPTIC
Consortium and the 100,000 Genomes Project showing that WGS
in combination with an interpretation database can also confidently
predict susceptibility (as opposed to rule-in resistance) against first-
line antituberculosis drugs [2]. An interpretation database is a
https://doi.org/10.1007/978-1-0716-3981-8_18,
195
196 Viola Dreyer et al.
catalog of mutations implicated in resistance and susceptibility to

individual TB antibiotics backed by phenotypic antibiotic suscepti-
bility testing. Catalogues can also include rulesets, i.e., a rule that
states any insertion and deletion in gene A or B should be consid-
ered as resistance determining. Over the years, these knowledge
databases have been refined through the accumulation of large
collections of clinical and laboratory-generated strains [2–5], and
can be also employed to guide multidrug-resistant (MDR)-TB
therapies [5]. Unfortunately, there is not yet a consensus of which
mutations provide “enough” evidence to be considered as
resistance-associated. The first standard reference for the interpre-
tation of mutations conferring antibiotic resistance against all first-
line and many second-line drugs was released by the WHO in 2021
[6]. The main limitation of this new WHO reference database is the
lack of phenotypic antibiotic susceptibility testing results for clinical
isolates resistant against the newest MDR-TB drugs such as pre-
tomanid, bedaquiline, and delamanid, or repurposed and newly
endorsed MDR-TB drugs such as clofazimine and linezolid
[6, 7]. In order to compensate this knowledge gap, literature
reviews and novel in vitro assays have been performed to generate
and compile new phenotypic and genotypic evidences for these
important MDR-TB drugs [7–10]. In the future, large-scale pro-
spective studies will be needed to demonstrate the applicability of
next-generation sequencing for rapid clinical diagnostics. There-
fore, bioinformatic pipelines require standardized parameters and
reporting forms [11]. This includes thresholds for mutation detec-
tion and quality criteria for whole genome sequence data, such as a
minimum average genome coverage depth, detection of contami-
nants such as human or other microbial reads, or the detection of
distinct clonal MTBC populations within a sample.
In this chapter, we provide a description of the basic workflow
for the bioinformatic analysis and quality control of Illumina short-
read data for the prediction of antimicrobial resistances for clinical
isolates of MTBC and non-tuberculosis mycobacteria (NTMs). We
list several available software solutions providing a pre-defined
analysis and interpretation pipeline, and provide a more detailed
customizable solution employing the MTBseq pipeline with an
example of a standardized reporting form for drug resistance deter-
minants in clinical MTBC/NTM isolates [12].
1.1 Overview At present, WGS of MTBC strains is routinely performed using

Bioinformatics next-generation sequencing (NGS) technologies, resulting in
Analysis of MTBC NGS millions of short reads. Due to the exceptionally clonal nature of
Data MTBC strains, mapping of NGS data from clinical isolates to an
MTBC reference genome (usually M. tuberculosis H37Rv) will
cover >99% of the reference sequence. Therefore, the relatively
simple and robust reference mapping approach followed by the
detection of variants, most importantly SNPs and short indels,
has become the de facto standard for NGS data analysis of MTBC
strains [13]. Reaching a final resistance profile from read data
requires a dedicated bioinformatics analysis covering several key
steps:
• Sequence data preprocessing and quality control
• Reference mapping and variant detection
• Annotating detected variants with a resistance marker database
and reporting
Initial sequence data can optionally be subjected to rigorous
quality controls before proceeding with the main analysis pipeline.
These quality controls can include detection of contaminating
non-MTBC sequence information, overall sequence data accuracy,
and overrepresentation of specific oligomers in the tails of the reads.
As modern reference mapping tools can deal with low-sequence
quality regions in the start and end tails of the reads, quality
trimming to cut those regions is usually not necessary.
In the reference mapping step, a suitable program such as BWA
[14] aligns raw read data (fastq files) to a given reference genome.
While there are some potential problems such as reads from highly
repetitive genomic regions, reads of low sequencing accuracy and
hybrid or contaminated reads, this step is relatively simple and
modern algorithms are suitably sophisticated. Routine quality
metrics should also be determined for the resulting reference
mapping. Important quality metrics are average coverage depth,
i.e., the number of reads mapped per position of the genome, and
coverage breadth, i.e., the amount of the reference genome covered
by sequence data (see Note 1). Comparing the sequence of mapped
reads against the reference sequence, e.g., with GATK [15], allows
for the detection of mutations including short indels. It is impor-
tant to note that these are called relative to the reference genome,
and only analysis results using the same reference genome are
therefore directly comparable. For inferring resistance profiles espe-
cially for MDR-MTBC strains, variant detection needs to be as
comprehensive as possible. Pipelines should report a confident
estimate of detected variants and indicate problems with variant
calling caused by insufficient data quality, at least for resistance-
associated regions of the genome. Special consideration must be
given to the detection of heterozygous/mixed variants present in a
minority fraction of the reads at a given position. These can arise
from co-infections with two distinct strains, or from clonal subpo-
pulations evolving in patients during the course of the treatment.
To allow a valid interpretation of resistance-associated variants,
considerable efforts are ongoing to build up databases containing
validated resistance mutations that can then be used to generate
resistance reports from WGS data.
WGS data analysis pipelines should use the most up-to-date

version for data interpretation which includes the detection of
benign mutations that are not associated with drug resistance [1],
and possibly quantify the resistance level, i.e., the MIC associated
with certain mutations [16]. Lastly, the analysis results should be
extracted and summarized in a plain language report with an inter-
pretation and a confidence grading for each detected mutation.
The general workflow can be implemented using freely avail-
able tools or commercial software suites, but it is important to
carefully consider appropriate analysis parameters for accurate
WGS data analysis. While numerous tools for automated resistance
prediction from MTBC WGS data have been developed, none of
them have yet to be implemented with the described functionality
in full.
2 Methods
2.1 Identifying Drug The following protocol explains how to install MTBseq [12] via
Resistance Conda/Miniconda and provides details for individual data proces-
Determinants with sing steps on a computer running the Linux operating system
MTBseq (tested for Ubuntu 16.04 LTS).
1. Install Conda or Miniconda hereafter install MTBseq with:
conda install -c bioconda mtbseq
MTBseq expects the files to be named in a specific way:

[SampleID]_[LibID]_[*]_[Direction].f(ast)q.gz (see Note 2).
MTBseq will create its own working environment of output
folders in the directory in which the program is executed and
use all FastQ files following the required naming convention
within that directory as input data. Therefore, data from differ-
ent projects can be easily placed into separate folders. Allowed
FastQ file extensions are fastq.gz or fq.gz. The whole workflow
is module-based, so that all basic steps can be done one after
the other. The basic steps are represented in Fig. 1 and
described in detail below:
2. Read Mapping:
MTBseq --step TBbwa
This step maps next-generation sequencing reads to a ref-

erence genome, using BWA. This is the first step within the
pipeline. Depending on the library type (single-end or paired-
Fig. 1 Schematic representation of the MTBseq workflow. Modules encapsulating specific functionality are
shown in blue boxes
end), this module will align the reads of one FastQ-file with
single-end reads or two fastq-files with paired-end reads to a
reference genome. (see Note 3)
Input:
[SampleID]_[LibID]_[*]_[Direction].fastq.gz
Output:
Bam/[SampleID]_[LibID]_[*].bam
Bam/[SampleID]_[LibID]_[*].bai
Bam/[SampleID]_[LibID]_[*].bamlog
3. Refinement of the initial mapping:
To improve the initial mapping, MTBseq employs GATK
for a realignment around small insertions and deletions and a
base call recalibration with default parameters (see Note 4)
[15]:
For base call recalibration, MTBseq employs a set of known
variants to be excluded. The calibration list is stored in the
MTBseq github directory “var/res/MTB_Base_Calibration_-
List.vcf” of the package.
Input:
Output:
GATK_Bam/[SampleID]_[LibID]_[*].gatk.bam
GATK_Bam/[SampleID]_[LibID]_[*].gatk.bai
GATK_Bam/[SampleID]_[LibID]_[*].gatk.bamlog
GATK_Bam/[SampleID]_[LibID]_[*].gatk.grp
GATK_Bam/[SampleID]_[LibID]_[*].gatk.intervals
4. Summarize base calls for mapped reads
MTBseq --step TBpile
This step creates so-called “pileup files” (.mpileup) from

the refined mapping files (.gatk.bam), using the program SAM-
TOOLS (see Note 5). The pileup format describes the base-pair
information at each chromosomal position. Each line consists
of the chromosome number (1 for bacterial genomes), the
genome position, the reference base, the number of reads
covering the site, and base qualities.
Input:
GATK_Bam/[SampleID]_[LibID]_[*].gatk.bam
Output:
Mpileup/[SampleID]_[LibID]_[*].gatk.mpileup
Mpileup/[SampleID]_[LibID]_[*].gatk.mpileuplog
5. Create position list:
MTBseq --step TBlist
This step transforms the pileup format (.gatk.mpileup) into

a table. The position list files capture the essential information
from mapping in a table format with 21 columns, containing
the nucleotide counts in mapped reads for each position of the
reference genome.
Input:
Mpileup/[SampleID]_[LibID]_[*].gatk.mpileup
Output:
Position_Tables/[SampleID]_[LibID]_[*].gatk_position_ta-
ble.tab
6. Variant detection/annotation:
MTBseq --step TBvariants --all_vars

This step performs the variant detection based on the

generated position lists (step 5). Variant calling can be run in
different detection modes using the OPTIONS --all_vars, --
snp_vars, and --lowfreq_vars (for details please refer to citation
[17]). As per default, a variant is called with a minimum cover-
age of four reads in both forward and reverse direction, a
minimum of four reads calling the allele with a phred score of
at least 20, and a minimum allele frequency of 75% (see Note
6). Positions fulfilling these thresholds will be considered as
“unambiguous” in the subsequent analysis.
Input:
ble.tab
Output:
Called/[SampleID]_[LibID]_[*].gatk_position_uncovered_
[mincovf]_[mincovr]_[minfreq]_[minphred20]_[all_vars]
[snp_vars][lowfreq_vars].tab
Called/[SampleID]_[LibID]_[*].gatk_position_variants_
[mincovf]_[mincovr]_[minfreq]_[minphred20]_[all_vars]
[snp_vars][lowfreq_vars].tab
7. Lineage classification:
MTBseq --step TBstrain
This step makes a lineage classification based on a set of

phylogenetic SNPs [18–20]. This module creates a tabular
delimited file within the “Classification” directory. Each entry
also gives an indication of the data quality for the positions used
to infer the phylogenetic classification. The column indicating
the quality contains the label “good” if all phylogenetic posi-
tions contained in the set are covered at least tenfold, and show
a frequency of at least 75%. A “bad” quality is indicated if any
phylogenetic position does not meet these thresholds, and
“ugly” is indicated if any phylogenetic position does not have
a clear base call.
Input:
ble.tab
Output:
Classification/Strain_Classification.tab
8. Calculate quality control measures
MTBseq --step TBstats
During this step an overview of mapping quality and

detected variants for a dataset is generated, using the SAM-
TOOLS flagstat program. This step creates or updates a tabular
delimited file “Mapping_and_Variant_Statistics.tab”. This file
stores all sample statistics for the analyzed datasets present in
the working environment (see Note 7). The columns of the
output are shown in Table 1.
Table 1
Column names with explanations for the file: “Mapping_and_Variant_Statistics.tab”
Column Explanation
Date The date of MTBseq execution
SampleID The sample ID
LibraryID The library ID
FullID Complete dataset name
Total reads The total amount of sequenced reads
Mapped reads (%) The percentage of reads mapped to the reference genome
Genome size The size of the reference genome
Genome GC The GC content of the reference genome
(Any) Total bases (%) Percentage of the reference genome covered by reads
(Any) GC-content GC content of the reference genome covered by reads
(Any) coverage mean Mean coverage depth
(Any) coverage median Median coverage depth
(Unambiguous) Total bases (%) Percentage of the reference genome covered unambiguously
(Unambiguous) GC-content GC content of the reference genome covered unambiguously
(Unambiguous) coverage mean Mean coverage depth of unambiguously covered positions
(Unambiguous) coverage median Median coverage depth of unambiguously covered positions
SNPs Number of detected SNPs
Deletions Number of detected deletions
Insertions Number of detected insertions
Uncovered Positions of the reference genome not covered by a read
Substitutions (including stop codons) Number of substitutions within genes
Input:
ble.tab
Output:
Statistics/Mapping_and_Variant_Statistics.tab
9. Cleaning the data set:
After all samples have been processed the user should check
the Mapping_and_Variant_Statistics.tab file. As it is a tabular
file, it is possible to open it, e.g., with Microsoft Excel or any
other office tools. The most important values are (see also
Note 1):
• Mapped Reads (%), i.e., low percentage can indicate
contaminations
• (Any) Coverage mean, i.e., the overall sequencing depth
• (Unambiguous) Total Bases (%), i.e., the overall sequencing
breadth
10. Extract mutation implicated in drug resistance
Open the variant file in the “Called” folder from the
low-frequency variant detection workflow (point 6, [lowfreq_-
vars].tab) for the sample of interest and apply the following
filters:
(a) Filter the column “ResistanceSNP” for entries, including
positions marked as phylogenetic or resistance mediating
based on the provided input catalog (see Note 8).
(b) Filter the column “InterestingRegion” for additional (yet
unknown) mutations putatively implicated in antimicro-
bial resistance but with an unclear phenotype.
2.2 Generating In Fig. 2, we provide an example for an antimicrobial resistance

Resistance Reports report form (see Note 9). Here, the presence of a resistance muta-
tion from the MTBseq catalog (column “ResistanceSNP”) is a
strong indicator for phenotypic resistance against the indicated
antibiotic and is annotated as “R” in the interpretation column of
the report. The absence of a mutation from the catalog, however,
should always be confirmed by a phenotypic drug susceptibility test
to eventually confirm drug susceptibility “S”. Comments can be
given, e.g., for low-level resistance mediating mutations, alternative
interpretations, or phenotypic confirmations. Mutations in genes
implicated in antimicrobial resistance (column “Interesting
region”) but absent from the interpretation catalog and with an
unclear phenotype should be listed in a second table. This may help
to resolve discrepancies between different interpretation databases,
differences between phenotypes and genotypes, and could reveal
novel resistance mediating mutations. Currently, we parse these
Fig. 2 Example for an antibiotic resistance report form for Mycobacterium tuberculosis complex isolates
considerations and catalog rule-sets from the resulting MTBseq

into a plain language report with customized in-house scripts. In
future MTBseq versions, it is planned to compute similar reports
automatically.
2.3 Comparative The MTBseq tool can be also employed for comparative analysis of
Genomic Analysis multiple datasets. However, we have focused here on the steps to
generate a resistance report. The entire pipeline can be run with the
following command:
MTBseq --step TBfull
2.4 Alternative Tools In recent years, several software solutions have been developed for a
for MTBC mostly automated data analysis (with pre-specified variant calling
Antimicrobial parameters) of MTBC NGS data for resistance inference. They
Resistance Prediction differ in their implementation, with some tools representing local
installed software suites and others built as cloud-based solutions.
In the following, we represent a short overview listing several freely
available tools. A crucial point to keep in mind here is which
mutation catalog is used by the software for determining resistance
patterns as this is the major factor for differences in the resulting
drug resistance report [21]. Below, we summarize some of the
advantages and disadvantages.
1. Mykrobe (Predictor TB; https://www.mykrobe.com/):
• No registration necessary.
• Runs offline on a standard Desktop PC or Laptop.
• Desktop Apps available for Mac and Windows 64-bit.
• No Batch mode available.
• Supports Illumina as well as Oxford nanopore data.
• Drugs included: isoniazid, rifampicin, ethambutol, pyrazi-
namide, ofloxacin, moxifloxacin, ciprofloxacin, streptomy-
cin, amikacin, capreomycin, kanamycin.
• Species identification of four MTBC species (M. tuberculosis,
M. bovis, M. africanum, and M. caprae) and 40 nontubercu-
lous mycobacteria (NTM) species (e.g., M. abscessus,
M. avium, and M. intracellulare).
• Sensitivity 82.6%; Specificity 98.5% (independent validation
set n = 1609).
2. TB Profiler (https://tbdr.lshtm.ac.uk/):
• No registration needed.
• Runs as an online service (command-line version available).
• Batch mode available.
• Supports Illumina and Oxford nanopore data.
• Tbdb database (https://github.com/jodyphelan/tbdb)
used as mutation library.

namide, fluoroquinolones, streptomycin, amikacin, capreo-
mycin, kanamycin, cycloserine, ethionamide, clofazimine,
para-aminosalicylic acid, delamanid, bedaquiline, linezolid.
• Species identification based on Coll et al. 2014.
3. PhyResSE (www.phyresse.org):
• No registration needed.
• Runs as an online service.
• Supports Illumina and Ion Torrent data.
• Self-provided masterlist used as mutation library.
namide, fluoroquinolones, streptomycin, amikacin, capreo-
mycin, kanamycin, ethionamide, clofazimine, para-
aminosalicylic acid, linezolid.
• Species identification based on Coll et al. 2014.
4. GenTB (https://gentb.hms.harvard.edu/):
• Registration required, but free of charge.
• Classification based on a random forest classifier or a Wide
and Deep neural network.
• Runs as an online service (command-line version available).
• Drugs included (random forest): rifampicin, isoniazid, eth-
ambutol, pyrazinamide, streptomycin, capreomycin, amika-
cin, ciprofloxacin, kanamycin, levofloxacin, ofloxacin, para-
aminosalicylic acid, ethionamide.
• Species identification based on Freschi et al. 2020.
5. ResFinder (https://cge.food.dtu.dk/services/ResFinder-4.1/):
• Not specifically for Mycobacterium tuberculosis (Campylo-
bacter spp., Campylobacter jejuni, Campylobacter coli,
Escherichia coli, Enterococcus faecalis, Enterococcus faecium,
Neisseria gonorrhoeae, Salmonella spp., Klebsiella, Helicobac-
ter pylori, Staphylococcus aureus).
• Runs as an online service.
• Drugs included (Mycobacterium tuberculosis specific): rifam-
picin, isoniazid, amikacin, linezolid, streptomycin, bedaqui-
line, ethambutol.
• No lineage classification included.
2.5 Tool for the Besides M. tuberculosis, there are also several other mycobacterial
Prediction of Antibiotic species that can cause serious disease in humans [22]. These
Resistances in Non- non-tuberculous mycobacteria (NTM) have gained increased
tuberculous attention in the past decades due to reports of emerging global
Mycobacteria (NTMs) infections, increasing number of deaths, and the occurrence of
several outbreaks [22, 23]. NTMs are notoriously difficult to treat
due to intrinsic resistance to a variety of antibiotics [24]. Further-
more, strains can also acquire additional resistances due to muta-
tions in resistance-related genes [22, 24]. Currently, only two
bioinformatic tools are available for whole genome sequencing-
based resistance prediction of NTMs: Mab_ariba (https://github.
com/samlipworth/Mab_ariba) and NTMprofiler (https://github.
com/jodyphelan/ntmdb/tree/main/db/Mycobacteroides_
abscessus). At present, these tools only include mutation catalogs
for resistance prediction of M. abscessus strains to macrolides and
aminoglycosides. Future efforts should therefore focus on the
establishment of additional mutation catalogues for other NTM
species as well as validation of these catalogues using both in vitro
drug susceptibility data, genotypic data and in vivo treatment
outcomes.
3 Notes
1. We recommend using datasets with >80% reads mapped to the

H37Rv reference genome (to exclude datasets with larger
DNA contaminations), >50x average genome-wide coverage
(ensuring a descent read depth), and (unambiguous) total
bases (%) > 95% (ensuring a descent genome coverage). In
addition, phylogenetic-relevant positions in Classification/
Strain_Classification.tab flagged as “bad” can indicate mixed
samples with more than one MTBC strain.
2. [SampleID] represents the identifier for a specific bacterial
sample and [LibID] is an identifier for the next-generation
sequencing library used. [Direction] is an essential field and
indicates if reads are in forward (R1) or reverse
(R2) orientation in paired-end data. Files for single-end data
must use the [Direction] R1. Other than these, file names can
be freely given, including further [*] fields.
3. By default, MTBseq uses the Mycobacterium tuberculosis
H37Rv genome (NC_000962.3) as a reference and maps
reads with the BWA mem program with default settings. After
the mapping, the resulting “SAM” files (Sequence Alignment
Map format) are converted into binary mapping files (.bam) to
reduce the file size. The mapping is sorted, indexed and puta-
tive PCR duplicates are removed, using the program
SAMTOOLS.
4. MTBseq uses the following GATK parameters:

--downsample_to_coverage 10000
--defaultBaseQualities 12
--maximum_cycle_value 600
--noOriginalAlignmentTags
5. SAMTOOLS is executed with the following parameters
-- B (Disable base alignment quality (BAQ) computation,
which is very time consuming)
-- A (Do not skip anomalous read pairs in variant calling.
Anomalous read pairs are those marked in the FLAG field
as paired in sequencing but without the properly paired
flag set.)
6. Default MTBseq variant calling thresholds for forward read
coverage, reverse read coverage, number of reads with a mini-
mum phred score of 20, and variant frequency, respectively, can
be modified as follows
--mincovf
--mincovr
--minphred20
--minfreq
7. SAMTOOLS flagstat program employs the same thresholds to
discern unambiguously covered positions as the TBvariants
module (i.e., set by the --mincovf, --mincovr, --minphred20,
--minfreq).
8. The column “ResistanceSNP” lists all mutations that are
provided with the customized resistance catalog. New muta-
tions must be specified at least with genome position, reference
base, alternative base, and the name of the antibiotic. The latest
version of the MTBseq mutation catalog has been published
recently [16].
9. To allow a confidence assessment for called mutations, we
suggest that you provide few quality control parameters such
as mapped reads (%), (any) coverage mean, (unambiguous)
total bases (%) (see also Note 1). A low mutation frequency
(e.g., <95%) can indicate heterogeneous infections by a sensi-
tive and resistant strain, or intra-patient resistance evolution,
i.e., treatment-acquired antimicrobial resistances. Reporting
the MTBC sublineage may help to identify putatively related
strains from transmission events or local outbreak. ID and
libID provide the unique identifier to document DNA ID
and library preparation.
References
1. Merker M, Kohl TA, Barilar I et al (2020) 2:e604–e616. https://doi.org/10.1016/
Phylogenetically informative mutations in S2666-5247(21)00175-0
genes implicated in antibiotic resistance in 10. Sonnenkalb L, Carter JJ, Spitaleri A et al
mycobacterium tuberculosis complex. (2023) Bedaquiline and clofazimine resistance
Genome Med 12:27. https://doi.org/10. in mycobacterium tuberculosis: an in-vitro and
1186/s13073-020-00726-5 in-silico data analysis. Lancet Microbe S2666-
2. Allix-Béguec C, Arandjelovic I, Bi L, Beckert P, 5247(23):00002–00002. https://doi.org/10.
Bonnet M, Bradley P et al (2018) Prediction of 1016/S2666-5247(23)00002-2
susceptibility to first-line tuberculosis drugs by 11. Meehan CJ, Goig GA, Kohl TA et al (2019)
DNA sequencing. N Engl J Med 379:1403– Whole genome sequencing of mycobacterium
1 4 1 5 . h t t p s : // d o i . o r g / 1 0 . 1 0 5 6 / tuberculosis: current standards and open
NEJMoa1800474 issues. Nat Rev Microbiol 17:533–545.
3. Walker TM, Kohl TA, Omar SV et al (2015) https://doi.org/10.1038/s41579-019-
Whole-genome sequencing for prediction of 0214-5
mycobacterium tuberculosis drug susceptibil- 12. Kohl TA, Utpatel C, Schleusener V et al (2018)
ity and resistance: a retrospective cohort MTBseq: a comprehensive pipeline for whole
study. Lancet Infect Dis 15:1193–1202. genome sequence analysis of mycobacterium
https://doi.org/10.1016/S1473-3099(15) tuberculosis complex isolates. PeerJ 6:e5895.
00062-6 https://doi.org/10.7717/peerj.5895
4. Miotto P, Tessema B, Tagliani E et al (2017) A 13. Consortium TCr (2022) A data compendium
standardised method for interpreting the asso- associating the genomes of 12,289 mycobacte-
ciation between mutations and phenotypic rium tuberculosis isolates with quantitative
drug resistance in mycobacterium tuberculosis. resistance phenotypes to 13 antibiotics. PLoS
Eur Respir J 50:1701354. https://doi.org/10. Biol 20:e3001721. https://doi.org/10.1371/
1183/13993003.01354-2017 journal.pbio.3001721
5. Grobbel H-P, Merker M, Köhler N et al (2021) 14. Li H, Durbin R (2009) Fast and accurate short
Design of multidrug-resistant tuberculosis read alignment with burrows-wheeler trans-
treatment regimens based on DNA sequenc- form. Bioinforma Oxf Engl 25:1754–1760.
ing. Clin Infect Dis Off Publ Infect Dis Soc https://doi.org/10.1093/bioinformatics/
Am 73:1194–1202. https://doi.org/10. btp324
1093/cid/ciab359 15. McKenna A, Hanna M, Banks E et al (2010)
6. Walker TM, Miotto P, Köser CU et al (2022) The genome analysis toolkit: a MapReduce
The 2021 WHO catalogue of mycobacterium framework for analyzing next-generation
tuberculosis complex mutations associated DNA sequencing data. Genome Res 20:
with drug resistance: a genotypic analysis. Lan- 1297–1303. https://doi.org/10.1101/gr.
cet Microbe 3:e265–e273. https://doi.org/ 107524.110
10.1016/S2666-5247(21)00301-3 16. Finci I, Albertini A, Merker M et al (2022)
7. Kadura S, King N, Nakhoul M et al (2020) Investigating resistance in clinical mycobacte-
Systematic review of mutations associated rium tuberculosis complex isolates with geno-
with resistance to the new and repurposed mic and phenotypic antimicrobial susceptibility
mycobacterium tuberculosis drugs bedaqui- testing: a multicentre observational study. Lan-
line, clofazimine, linezolid, delamanid and pre- cet Microbe 3:e672–e682. https://doi.org/
tomanid. J Antimicrob Chemother 75:2031– 10.1016/S2666-5247(22)00116-1
2043. https://doi.org/10.1093/jac/dkaa136 17. MTBseq (2022) MTBseq. In: MTBseq
8. Battaglia S, Spitaleri A, Cabibbe AM et al Github. https://github.com/ngs-fzb/
(2020) Characterization of genomic variants MTBseq_source. Accessed 20 Mar 2023
associated with resistance to bedaquiline and 18. Homolka S, Projahn M, Feuerriegel S et al
delamanid in naive mycobacterium tuberculo- (2012) High resolution discrimination of clini-
sis clinical strains. J Clin Microbiol 58:e01304– cal mycobacterium tuberculosis complex
e01320. https://doi.org/10.1128/JCM. strains based on single nucleotide polymorph-
01304-20 isms. PLoS ONE 7(7):e39855. https://doi.
9. Ismail N, Rivière E, Limberis J et al (2021) org/10.1371/journal.pone.0039855
Genetic variants and their association with phe- 19. Coll F, McNerney R, Guerra-Assunc JA et al
notypic resistance to bedaquiline in mycobac- (2014) A robust SNP barcode for typing myco-
terium tuberculosis: a systematic review and bacterium tuberculosis complex strains, Nat
individual isolate data analysis. Lancet Microbe
Commun 5(1). https://doi.org/10.1038/ mycobacterium abscessus. Nat Rev Microbiol

ncomms5812 187(18):392–407. https://doi.org/10.1038/
20. Merker M, Blin C, Mona S et al (2015) Evolu- s41579-020-0331-1
tionary history and global spread of the myco- 23. Dahl VN, Mølhave M, Fløe A et al (2022)
bacterium tuberculosis Beijing lineage. Nat Global trends of pulmonary infections with
Genet 47(3):242–249. https://doi.org/10. nontuberculous mycobacteria: a systematic
1038/ng.3195 review. Int J Infect Dis 125:120–131. https://
21. Schleusener V, Köser CU, Beckert P et al doi.org/10.1016/j.ijid.2022.10.013
(2017) Mycobacterium tuberculosis resistance 24. Van Ingen J, Boeree MJ, Van Soolingen D,
prediction and lineage classification from Mouton JW (2012) Resistance mechanisms
genome sequencing: comparison of automated and drug susceptibility testing of nontubercu-
analysis tools. Sci Rep 7:46327. https://doi. lous mycobacteria. Drug Resist Updat 15:149–
org/10.1038/srep46327 161. https://doi.org/10.1016/J.DRUP.
22. Johansen MD, Herrmann JL, Kremer L (2020) 2012.04.001
Non-tuberculous mycobacteria and the rise of
Chapter 19
Analysis of Whole Genome Sequencing Data for Detection

of Antimicrobial Resistance Determinants
Marie Anne Chattaway
Abstract
Genomic sequencing has revolutionized microbial typing methods and transformed high-throughput
methods in reference, clinical, and research laboratories. The detection of antimicrobial-resistant (AMR)
determinants using genomic methods can provide valuable information on the emergence of resistance.
Here we describe an approach to detecting AMR determinants using an open access and freely available
platform which does not require bioinformatic expertise.
Key words Antimicrobial resistance (AMR) determinants, Center for genomic epidemiology (CGE),
Sequencing, KMER, MLST, ResFinder
1 Introduction
The use of genomic sequencing as a means of detecting antimicro-

bial-resistant (AMR) determinants for surveillance is well described
[1–6] and pathogen-specific validation studies have shown high
correlation between AMR genotype and phenotype [7–9]. There
are several mechanisms for antimicrobial resistance which can be
encoded in the genome and can occur such as point mutations in
the chromosome of the pathogen or acquired via mobile genetic
elements such as plasmids [10, 11], insertion sequences or trans-
posons [12]. Despite high correlation in studies for some patho-
gens, it is important to note that one AMR determinant conferring
phenotypic resistance in one pathogen may not confer resistance in
another. That novel markers or previously unvalidated genotype to
phenotypic against specific pathogens should have phenotypic test-
ing when considering implications for clinical practices or public
health surveillance. This is also true when reviewing genomic data
to provide an evidence base for updating clinical guidelines [7, 13,
https://doi.org/10.1007/978-1-0716-3981-8_19,
211
212 Marie Anne Chattaway
14]. Despite multiple research studies and the potential for predict-
ing phenotypic resistance [15–18], further validation is required
before this concept is implemented in clinical laboratories.
The routine use of sequencing technologies and open sharing
of generated data in reference laboratories [4, 19, 20], and research
and clinical laboratories [21] has resulted in a vast amount for
analysis. High-throughput data were initially processed using
bespoke bioinformatic pipelines or specialized bioinformaticians
to process the data. However, open access platforms have now
been developed enabling non-specialized bioinformaticians to pro-
cess raw sequence data files for identification, typing, and detection
of AMR determinants. There are many platforms available for
detection of AMR determinants [1], which will have their own
manuals and publications, and some will be bespoke to a pathogen
of interest, so it is worth researching the options available to suit
your needs. In this chapter, we will describe an AMR platform that
is not pathogen-specific. The aim of the Center for Genomic Epi-
demiology (CGE) platform (http://www.genomicepidemiology.
org/) is to provide access to bioinformatics resources also for those
with limited experience and thereby allow all countries, institutions,
and individuals to take advantage of sequencing technologies. CGE
is entirely non-commercial and operates a number of free online
bioinformatics services. Ideally, you would have an identification
and basic typing of your organism of interest before assessing AMR
determinants. If not, the advantage of using sequence data is that
once you have the data, you can extrapolate different types of
information such as identification to a species level and the
Sequence Type (see Note 1) which are also available on the CGE
platform and will also be covered in this chapter. Bioinformatic
pipelines available on CGE include: the KmerFinder, k-mers are
substrings of length k contained within a biological sequence and
k-mers can be used in the context of computational genomics and
sequence analysis. There is an R package called kmer that enables
fast alignment-free clustering of biological sequences [22]. For the
context of identification of a bacterial pathogen to the species or
sub-species level, a selection of kmers within the large genome have
been selected to develop a publically available database for identifi-
cation which is called KmerFinder. This approach is computation-
ally quicker and more practicable than querying the entire genome.
Further subtyping of a bacterial pathogen can then be performed
using multilocus sequence typing (MLST) and each scheme is
pathogen specific. MLST is an unambiguous procedure for char-
acterizing isolates of bacterial species using the sequences of inter-
nal fragments of (usually) seven house-keeping genes. For each
house-keeping gene, the different sequences present within a bac-
terial species are assigned as distinct alleles and, for each isolate, the
alleles at each of the seven loci define the allelic profile or sequence
type (ST) [23–26]. Using MLST will group strains into complexes
or eBURST groups (Salmonella) and with some bacterial species
Analysis of Whole Genome Sequencing Data for Detection of Antimicrobial. . . 213
such as Salmonella, E.coli, and Shigella, the sequence type, eBURST

group or complex can be used to infer the serotype [4, 27, 28].
There are multiple services available, this chapter will focus on
identification of the strain to a species level using the KmerFinder
Platform (can be any bacterial genus) (see Subheading 3.2.1),
MLST of the strain (there are multiple schemes of different patho-
gens you can select from) (see Subheading 3.2.2), and identifica-
tion of acquired antibiotic resistance genes using ResFinder (see
Subheading 3.3).
2 Materials
A computer and access to the internet is required for these techni-

ques. The data format can vary depending on the sequencing
platform used but access to data files such as fastaq, fasta format,
or assembled data are typically required for downstream analysis.
You will need to check the output options for your sequencing
platform. Sequencing data should be quality checked before any
downstream analysis is performed, if the quality of sequencing data
is poor, then it will need to be repeated.
3 Methods
3.1 Getting Started 1. CGE runs entirely online and be accessed via the link https://
www.genomicepidemiology.org/.
2. The main page contains an overview of the platform, if you
select the service tab, it will bring you to the full list of services
(see Figs. 1 and 2).
3.2 Identification Unlike some genomic platforms, you do not need to know the
and Typing of Data presumptive identification of your strain to assess the data. The use
of the KMER ID programme so a useful tool to identify your
organism of interest (see Note 2). In order to identify your strain,
you will need to (1) select the program of interest and then
(2) upload your reads, the options given include FASTA files,
FASTQ files, or assembled files.
3.2.1 KmerFinder— 1. Select KmerFinder under the Typing section on the services
Prediction of Bacterial page (see Notes 3 and 4) (Fig. 3).
Species Using a Fast K-Mer 2. Select the host database—bacterial organism.
Algorithm
3. Other options such as scoring method, gene database, identify
threshold, and depth are set at default; these settings are
acceptable for queries but have the option to be changed if
needed.
4. Browse for your files by clicking on the Isolate File Icon.
5. Press upload once you have added your files.
Fig. 1 Welcome page for the Center for Genomic Epidemiology
6. Enter your email address, the link to your results will be sent
once processed. If there is a delay in receiving your results,
remember to check your junk mail.
7. Once the results are ready, click on the link sent to you via
email.
8. View your data and look at quality metrics to assign identifica-
tion, the first hit will be the best match with the data submitted
and will be your identification. In this example, the result is
Salmonella Typhi.
3.2.2 MLST—Multi 1. Select MLST under the Typing section on the services page (see
Locus Sequence Typing Note 5) (Fig. 4).
(MLST) from an Assembled 2. Select the MLST configuration database—your pathogen of
Genome or from a Set of interest (From the KMER identification).
Reads
3. Select min. Depth for an allele—default option is suggested.
Fig. 2 Menu of available services
4. Select the type of data input (assembled or raw fastq files).

6. Press upload once you have added your files.
once processed. Once again, remember to check your
junk mail.
8. Once the results are ready, click on the link sent to your email.
9. View your data and look at quality metrics to assign identifica-
tion, the MLST will give you the allele number and will also
state the ST if the allelic profile is recognized.
Fig. 3 (a) The front page of the KmerFinder program when the sequence data has been selected for upload. In
this example, pair-end fastq files have been selected. The progress bar will show the progress of the upload.
(b) The pop-up box once your data has been successfully uploaded. The email is entered, and a link is sent to
that email once the results are ready. (c) The output of the KmerFinder result which shows the top hits of the
submitted sequence data against the KmerFinder database. The identification of this isolate is Salmonella
enterica subspecies enterica. The KmerFinder should not be used for serovar Identification
Fig. 4 (a) The front page of the MLST program when the sequence data has been selected for upload. In this
example, pair-end fastq files have been selected. The progress bar will show the progress of the upload. (b)
The output of the MLST result which shows the MLST allelic profile of the submitted sequence data against the
Salmonella MLST database. In this example, the allelic profile is Sequence Type (ST) 36. This Salmonella ST is
associated as being the serovar Typhimurium
3.3 Identification of 1. Select ResFinder under the Phenotyping section on the ser-
Antimicrobial vices page (see Note 6) (Fig. 5).
Resistance 2. Select for mutations of interest such as “Chromosomal muta-
Determinants tions” or “acquired resistance”.
3.3.1 ResFinder— 3. Select the antimicrobial panel of interest for acquired resis-
Detection of Acquired and tance, default is all.
Point Mutations Associated 4. Select the threshold for %ID and minimum length, default is
with Antimicrobial 90% and 60% respectively.
Resistance
5. Select the appropriate species database—based on the KMER
identification.
6. Select the type of your reads—assemble or sequencing specific
output files.
8. Press upload once you have added your files (see Note 7).
Fig. 4 (continued)
once processed. Remember to check your junk mail.
10. Once the results are ready, click on the link sent to your email.
11. View your data to analyze resistance, you will get a line list of
the resistance markers found, what antibiotic they predict resis-
tance to and where the mutations occur. Unknown mutations
should be validated phenotypically (see Note 8).
12. The algorithm and appropriate reference may vary depending
on your study and the pipeline you have selected (see Note 9).
4 Notes
1. Typing platforms have been designed to process data from a

pure culture of bacteria, this is required to enable accurate
typing of the strain including detection of AMR determinants.
Refer to manufacturing instructions of your sequencing equip-
ment for the data output file type to assess if suitable for the
typing platform of interest, though most platforms are set up to
accept different sequencing output files. Although methods
and platforms for metagenomics (sequencing of mixed bacte-
rial populations, such as DNA from feces) are being developed,
many of the processes require bioinformatic expertise and are
not discussed in this chapter.
Fig. 5 (a) The front page of the ResFinder program and the selection of the parameters of the antimicrobial
resistance determinants to be included. (b) The front page of the ResFinder program when the sequence data
has been selected for upload. In this example, pair-end fastq files have been selected. The progress bar will
show the progress of the upload. (c) The output of the ResFinder program: in this example, this isolate
contains aac(6′)-Iaa aminoglycoside enzyme encoding genes which predicts resistance to amikacin and
tobramycin, the qnrS1 plasmid which predicts ciprofloxacin resistance, and the CTX-M-15 gene which
predicts resistance to beta-lactams including third-generation cephalosporins
2. It is best practice to check that your sequence data is the same

identification of your organism and that the DNA extraction
process was not from selective media which can inhibit con-
taminants which can thrive during the enrichment phase of
culturing the organism and will also be sequenced.
3. Under each service, there are multiple tabs including instruc-
tions, outputs, article abstracts, relevant citations, overview of
genes (where appropriate), and database history. This chapter
provides a basic key instruction, but it is recommended that
you read these other tabs for further details.
4. For the KMER pipeline, there are over 75 references to select
from, when publishing your work, please select the most
appropriate reference from this list.
Fig. 5 (continued)
5. For the MLST pipeline there are multiple references to select

from depending on whether you have used raw reads or assem-
bled data. When publishing your work, please select the most
appropriate reference from this list. It should also be noted the
different MLST schemes are pathogen-specific and have been
developed by other researchers, the original MLST article for
your pathogen of interest should also be cited.
6. For the ResFinder pipeline, there are multiple references to
select from depending on whether you have used raw reads or
assembled data. When publishing your work, please select the
most appropriate reference from this list.
7. If your file upload or processing fails, check that you have
correctly selected the sequence file type with the file you are
uploading.
8. It is always advised that genetic mutations are phenotypically
validated against your current laboratory antimicrobial sensi-
tivity testing (AST) protocols for interpretation. This is partic-
ularly important with new mutations or markers that are not
well defined or not well published. Mutations can also confer
resistance in one pathogen but not another so assumptions
should not be made, hence the need for phenotypic validation
and confirmation. If AST validation is not available, it is impor-
tant to define your criteria for resistance in your analysis and
note the caveat in your analysis that phenotypic confirmation
was not undertaken.
9. There are multiple references for using these tools with a
summary below. However, for the correct citation for your
specific study, please go onto the CGE website and look up
the appropriate reference depending on the tools you are using.
(a) Benchmarking of methods for genomic taxonomy. Larsen
MV, Cosentino S, Lukjancenko O, Saputra D,
Rasmussen S, Hasman H, Sicheritz-Pontén T, Aarestrup
FM, Ussery DW, Lund O. J Clin Microbiol. 2014 May;52
(5):1529–39.
(b) Multilocus Sequence Typing of Total Genome Sequenced
Bacteria.
Larsen MV, Cosentino S, Rasmussen S, Friis C,
Hasman H, Marvig RL, Jelsbak L, Sicheritz-Pontén T,
Ussery DW, Aarestrup FM and Lund O. J. Clin. Micobiol.
2012. 50(4): 1355–1361. PMID: 22238442. doi:
10.12.0/JCM.06094-11
(c) ResFinder 4.0 for predictions of phenotypes from geno-
types. Bortolaia V, Kaas RF, Ruppe E, Roberts MC,
Schwarz S, Cattoir V, Philippon A, Allesoe RL, Rebelo
AR, Florensa AR, Fagelhauer L, Chakraborty T,
Neumann B, Werner G, Bender JK, Stingl K,
Nguyen M, Coppens J, Xavier BB, Malhotra-Kumar S,

Westh H, Pinholt M, Anjum MF, Duggett NA, Kempf I,
Nyk€aenoja S, Olkkola S, Wieczorek K, Amaro A,
Clemente L, Mossong J, Losch S, Ragimbeau C,
Lund O, Aarestrup FM.Journal of Antimicrobial Chemo-
therapy. 2020 Aug 11. PMID: 32780112. doi: https://
doi.org/10.1093/jac/dkaa345
(d) PointFinder: a novel web tool for WGS-based detection of
antimicrobial resistance associated with chromosomal
point mutations in bacterial pathogens.
Zankari E, AllesÃ¸e R, Joensen KG, Cavaco LM,
Lund O, Aarestrup FM. Journal of Antimicrobial Chemo-
therapy. 2017 July 19. PMID: 29091202. doi: https://
doi.org/10.1093/jac/dkx217
(e) Camacho C, Coulouris G, Avagyan V, Ma N,
Papadopoulos J, Bealer K, Madden TL. BLAST+: archi-
tecture and applications. BMC Bioinformatics 2009; 10:
421.
(f) Clausen PTLC, Aarestrup FM, Lund O. Rapid and precise
alignment of raw reads against redundant databases with
KMA. BMC Bioinformatics 2018; 19:307.
(g) (Database) Yoon SH*, Park YK, and Kim JF. 2015.
PAIDB v2.0: exploration and analysis of pathogenicity
and resistance islands Nucleic Acids Res. 43:D624-D630
(doi: 10.1093/nar/gku985, Published in advance,
October 21, 2014).
References
1. Hendriksen RS, Bortolaia V, Tate H, Tyson 5. Davies N, Jorgensen F, Willis C, McLauchlin J,
GH, Aarestrup FM et al (2019) Using geno- Chattaway MA (2022) Whole genome
mics to track global antimicrobial resistance, sequencing reveals antimicrobial resistance
vol 7. Front Public Health Rev, p 17 determinants (AMR genes) of Salmonella
2. AMR NGHRUoGSo (2020) Whole-genome enterica recovered from raw chicken and
sequencing as part of national and international ready-to-eat leaves imported into England
surveillance programmes for antimicrobial between 2014 and 2019. J Appl Microbiol
resistance: a roadmap. BMJ Global Health 133(4):2569–2582
5(11):e002244 6. Baker KS, Dallman TJ, Field N, Childs T,
3. Neuert S, Nair S, Day MR, Doumith M, Ash- Mitchell H et al (2018) Genomic epidemiology
ton PM et al (2018) Prediction of phenotypic of Shigella in the United Kingdom shows
antimicrobial resistance profiles from whole transmission of pathogen sublineages and
genome sequences of non-typhoidal salmo- determinants of antimicrobial resistance. Sci
nella enterica. Front Microbiol 9:592 Rep 8(1):7389
4. Chattaway MA, Dallman TJ, Larkin L, Nair S, 7. Chattaway MA, Gentle A, Nair S, Tingley L,
McCormick J et al (2019) The transformation Day M et al (2021) Phylogenomics and anti-
of reference microbiology methods and surveil- microbial resistance of Salmonella Typhi and
lance for salmonella with the use of whole Paratyphi A, B and C in England,
genome sequencing in England and Wales. 2016–2019. Microbial Genomics 7(8):633
Front Public Health 7:317
8. Day MR, Doumith M, Do Nascimento V, 18. Zhuang H, Zhu F, Lan P, Ji S, Sun L et al

Nair S, Ashton PM et al (2018) Comparison (2021) A random forest model based on core
of phenotypic and WGS-derived antimicrobial genome allelic profiles of MRSA for penicillin
resistance profiles of Salmonella enterica sero- plus potassium clavulanate susceptibility pre-
vars Typhi and Paratyphi. J Antimicrob Che- diction. Microbial Genomics 7(9):1–10
mother 73(2):365–372 19. Chattaway MA, Dallman TJ, Gentle A, Wright
9. Feldgarden M, Brover V, Haft DH, Prasad AB, MJ, Long SE et al (2016) Whole genome
Slotta DJ et al (2019) Validating the AMRFin- sequencing for public health surveillance of
der tool and resistance gene database by using Shiga toxin-producing escherichia coli other
antimicrobial resistance genotype-phenotype than serogroup O157. Front Microbiol 7:258
correlations in a collection of isolates. Antimi- 20. Chattaway MA, Greig DR, Gentle A, Hartman
crob Agents Chemother 63(11):1–19 HB, Dallman TJ et al (2017) Whole-genome
10. San MA (2018) Evolution of plasmid-mediated sequencing for national surveillance of Shigella
antibiotic resistance in the clinical context. flexneri. Front Microbiol 8:1700
Trends Microbiol 26(12):978–985 21. Zhou Z, Alikhan NF, Mohamed K, Fan Y,
11. Rozwandowicz M, Brouwer MSM, Fischer J, Agama Study G et al (2020) The EnteroBase
Wagenaar JA, Gonzalez-Zorn B et al (2018) user’s guide, with case studies on Salmonella
Plasmids carrying antimicrobial resistance transmissions, Yersinia pestis phylogeny, and
genes in Enterobacteriaceae. J Antimicrobial Escherichia core genomic diversity. Genome
Chemother 73(5):1121–1137 Res 30(1):138–152
12. Partridge SR, Kwong SM, Firth N, Jensen SO 22. Wilkinson S (2018) kmer: an R package for fast
(2018) Mobile genetic elements associated alignment-free clustering of biological
with antimicrobial resistance. Clin Microbiol sequences. https://doi.org/10.5281/zenodo.
Rev 31(4):1–61 1227689
13. Nabarro LE, McCann N, Herdman MT, 23. Maiden MC, Bygraves JA, Feil E, Morelli G,
Dugan C, Ladhani S et al (2022) British infec- Russell JE et al (1998) Multilocus sequence
tion association guidelines for the diagnosis typing: a portable approach to the identifica-
and management of enteric fever in England. tion of clones within populations of pathogenic
J Infect 84(4):469–489 microorganisms. Proc Natl Acad Sci U S A
14. Herdman T, Karo B, Dave J, Katwa P, Freed- 95(6):3140–3145
man J et al (2021) SUBMITTED: increasingly 24. Maiden MC (2006) Multilocus sequence typ-
limited options for the treatment of enteric ing of bacteria. Annu Rev Microbiol 60:561–
fever in travellers returning to England, 588
2014–2019: a cross-sectional analytical study. 25. Jolley KA, Bray JE, Maiden MCJ (2018) Open-
J Med Micro 2021:JMM-D-21-00003 access bacterial population genomics: BIGSdb
15. Wang S, Zhao C, Yin Y, Chen F, Chen H et al software, the PubMLST.org website and their
(2022) A practical approach for predicting anti- applications. Wellcome Open Res 3:124
microbial phenotype resistance in staphylococ- 26. Maiden MC, Jansen van Rensburg MJ, Bray
cus aureus through machine learning analysis JE, Earle SG, Ford SA et al (2013) MLST
of genome data. Front Microbiol 13:841289 revisited: the gene-by-gene approach to bacte-
16. Nguyen M, Long SW, McDermott PF, Olsen rial genomics. Nat Rev Microbiol
RJ, Olson R et al (2019) Using machine 11(10):728–736
learning to predict antimicrobial MICs and 27. Achtman M, Wain J, Weill FX, Nair S, Zhou Z
associated genomic features for nontyphoidal et al (2012) Multilocus sequence typing as a
Salmonella. J Clin Microbiol 57(2):e01260 replacement for serotyping in Salmonella enter-
17. VanOeffelen M, Nguyen M, Aytan-Aktug D, ica. PLoS Pathog 8(6):e1002776
Brettin T, Dietrich EM et al (2021) A genomic 28. Chattaway MA, Schaefer U, Tewolde R, Dall-
data resource for predicting antimicrobial resis- man TJ, Jenkins C (2017) Identification of
tance from laboratory-derived antimicrobial Escherichia coli and shigella species from
susceptibility phenotypes. Brief Bioinform whole-genome sequences. J Clin Microbiol
22(6):bbab313 55(2):616–623
INDEX
A CD8 cells ......................................................................... 99

Cell aggregates ................................................... 13, 16, 17
Acoustic levitation......................................................... 111 Chequerboard assays....................................................... 67
Acoustic trapping ........................................ 110, 111, 115 Compartmental models ............................................84–88
Agar-based methods ....................................................... 36
Competent cells.........................................................24, 30
Agent-based models (ABMs) ......................................... 94 Coronavirus ................................................................... 164
Algorithms ...................................87, 103, 104, 133, 139, CRISPR interference (CRISPRi) ...................... 24, 27–29
164, 166, 197, 213, 218
Cytotoxic activity ............................................................ 95
Alignment free clustering ............................................. 212
Ancestral states ............................................ 122, 123, 126 D
Animal models................................................................. 12
Antibiotic combinations ...........................................60, 73 Delamanid ............................................................ 196, 206
Antibiotic monotherapy ................................................... 1 Deterministic models ...................................................... 87
Antibiotic (antimicrobial) resistance ........... 1, 23, 24, 35, Diagnoses .................................. 121, 131, 136–138, 140,
36, 43, 51, 52, 57, 59, 79–88, 93, 95, 97, 100, 146, 153–159
103, 104, 122, 138, 164, 166, 167, 186, 196, Digital time-lapse angled field microscopy......... 135, 136
203, 204, 211–223 Dimethyl sulfoxide (DMSO)..........37–40, 68, 70, 72–74
Antibiotic (antimicrobial) susceptibility test ......... 24, 47, Discrete (grid-like) space ................................................ 97
55, 60–63, 136, 222 DNA .................................... 26, 139, 154, 163, 207, 221
Antibiotic synergy .......................................................1, 73 DNA-library ...................... 164, 170, 186, 187, 191–193
Atomic force microscopy (AFM) ................................. 136 Drug distribution ............................................................ 85
Automated microbiology systems (AMS).................... 133 Drug penetration ............................................................ 86
Drug susceptibility test (DST) ....................129–140, 203
B
E
Bacterial fitness ................................................................ 24
Bacterial growth .................................3, 4, 35–41, 45, 81, Ebola virus ..................................................................... 164
100, 134, 135 eBURST ........................................................................ 213
Bacterial heterogeneity .............................................11, 80 Electrical sensors ........................................................... 110
Bacterial strain selection ................................................. 83 Electronic nose .............................................................. 136
Bayesian analysis ............................................................ 123 Electroporator ...........................................................26, 30
BD Phoenix ................................................................... 133 Escherichia coli ..................................................36, 39, 206
Bedaquiline ............................................66, 138, 196, 206 Ethionamide ......................................................... 138, 206
Behavioral rules ....................................... 94, 97, 100, 102 Evolution .............................................. 81–87, 93, 95–97,
Biofilms ............................................................... 11–20, 63 103, 104, 118, 121, 122, 132, 139, 208
Bioinformatic pipelines ............................... 123, 196, 212 Exhaust protective cabinet................................... 151, 154
Bioinformatics ................................... 164, 167, 186, 196, Extraction
197, 207, 212, 221, 223 DNA .................................26, 29, 168, 171, 177, 221
Biomass ................................................12, 16, 17, 19, 110 RNA ................ 26, 29, 147–149, 155, 156, 158, 159
Bioreactors.................................... 14, 15, 57, 59–61, 111
F
C FASTA ........................................................................... 213
Capreomycin ............................................... 138, 205, 206 FASTQ.................................................................. 197, 213
Cas9 enzyme ................................................................... 24 Fractional Inhibitory Concentration Index
CD4 cells ......................................................................... 99 (FICI).............................................................44, 45
https://doi.org/10.1007/978-1-0716-3981-8,
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Science+Business Media, LLC, part of Springer
Nature 2024
225
ANTIBIOTIC RESISTANCE PROTOCOLS
226 Index
G M
Gene essentiality...........................................24, 27, 29–31 Macrophages .................................. 2, 67, 95, 96, 99, 100
Gene expressions .................................................... 29, 110 Malaria ............................................................................. 96
Gene silencing ...........................................................24–33 MALDI-TOF ................................................................ 130
Genome ........................................ 24, 138, 163, 195, 211 Mathematical models ......................................... 79–88, 94
Genome alignment .............................................. 123, 124 Maximum likelihood analysis .............................. 122, 126
Genome assembly ......................................................... 164 Microfluidics........................................................ 112, 113,
Genome-wide association studies (GWAS) ........ 121, 122 115, 134–136
GenTB ........................................................................... 206 Microgravity .................................................................... 12
GenXpert ....................................................................... 138 Microscan ............................................................. 130, 134
Granuloma.................................................................67, 96 Microtiter plate .........................................................65–76
Middlebrook medium.......................................15, 38, 39,
H 67, 68, 74
Heat inactivation ........................................................... 147 Minimum inhibitory concentration
(MIC)........................................................... 4, 5, 8,
Helicobacter pylori .................................................. 96, 206
High containment laboratory (BSL3) ......................... 139 24, 37, 44, 45, 81, 101, 114, 133, 134, 198
HIV-1 .............................................................................. 96 Mixed infections.............................................................. 83
Mobile genetic elements
Hollow fiber .......................................................... v, 57–64
Horizontal gene transfer (HGT) ....................80, 81, 122 (MGEs) ......................................80, 163, 166, 211
Hybrid models ................................................................ 88 Model parameterization.................................................. 88
Model variables ...................................................... 97, 125
Hypervirulent ................................................................ 163
Moxifloxacin ........................................................... 66, 205
I MTBseq ...............................................196–205, 207, 208
Multi-drug resistant organisms (MDROs).................. 161
Immune cells .............................................. 80, 82–86, 94, Multi-drug resistant tuberculosis
95, 99, 100, 102, 103 (MDR-TB)....................................... 185, 187, 196
Individual-based modeling ............................................. 83 Multi-locus sequence typing (MLST) ............... 163, 167,
Infection prevention and control (IPC) ............ 161, 162, 212–215, 217, 222
167, 168 Mutant prevention concentration
Insertion deletions (indels)..........................125, 195–197 (MPC)......................................................... 81, 171
Intensive care unit (ICU) ............................................. 163 Mutation rates ........................................................ 80, 122
Isoniazid (INH) ..............................................13, 66, 112, Mutations ................................................... 44, 66, 80, 81,
114, 116, 117, 138, 185, 205, 206 87, 88, 94, 122, 138, 139, 186, 195–198, 203,
205–208, 211, 215, 218, 222, 223
K
Mycobacterium abscessus ............................. 24, 26, 29–31,
Klebsiella pneumoniae ..................................................... 45 36, 38, 205, 207
Kmer ....................................................212–214, 217, 222 Mycobacterium avium complex ...................................... 96
Mycobacterium bovis .................................. 13–18, 20, 205
L Mycobacterium smegmatis ..................... 24, 112, 114–117
Mycobacterium tuberculosis ..................11, 18, 23, 24, 27,
Laboratory information management systems
38, 65–76, 81, 83, 99, 111, 121, 122, 138, 139,
(LIMS) ............................................................... 162
145–151, 153–155, 159, 185–193, 195, 196,
Larvae .......................................................................... 2–10
204–207
Lesion specific acquired resistance ................................. 86
Mykrobe ........................................................................ 205
Library preparation ............................168, 170, 172–175,
178, 179, 186, 208
N
Ligase chain reaction (LCR) ........................................ 137
Lineage classifications .......................................... 201, 206 Non-destructive method .............................................. 110
Lipid-rich mycobacteria ................................................ 111 Non-tuberculosis mycobacteria
Liposomes........................................................................ 36 (NTM) ............................................. 196, 205, 207
Long read sequencing (LRS) ..................... 163, 164, 166 Norfloxacin (NOR)...............................67, 70, 71, 73, 74
Luria-Bertani (LB) agar/broth ................................26, 38 Nucleic acid amplification test (NAAT)....................... 137
ANTIBIOTIC RESISTANCE PROTOCOLS
Index 227
O Sequence Alignment/Map format (SAM) ......... 124, 207
Sequence type (ST) .............................212, 213, 215, 217
Ordinary differential equations Shigella ........................................................................... 213
(ODEs) .........................84, 85, 87, 88, 93, 94, 96 Short-read data (Illumina)......................... 124, 164, 167,
Organic light emitting diodes (OLEDs) .................52, 54 168, 170, 171, 176, 180, 181, 186, 187, 191,
Outbreaks ............................................161–181, 207, 208 193, 196, 205, 206
Single nucleotide polymorphisms (SNPs) ......... 163, 166,
P
195, 196, 201, 202
Paired end reads ............................................................ 199 Spatial models..................................................... 87, 95, 97
Partial differential equations (PDEs) ..................... 84, 87, SPOTi ..............................................24, 27, 30–31, 35–41
88, 93, 94 Staphylococcus aureus .............................. 1–10, 36, 39, 53,
Personal protective equipment (PPE) ......................... 161 54, 130, 137, 163, 173, 206
Phenotypes ........................................ 12, 13, 18, 99, 103, Staphylococcus epidermidis .........................................53, 54
121–124, 126, 127, 195, 203, 211, 223 Stochastic ............................................................ 87, 88, 94
Phenotypic characterization ........................................... 30 Streptococcus Group A ................................................... 163
Phenotypic knockdown .................................................. 24 Survival analysis .................................................... 121–127
PhiX ............................................................ 170, 173, 176, Survival curves...........................................................4, 6, 7
177, 181, 187, 189, 191 Synergy assessment ......................................................... 73
Photodynamic therapy (PDT)........................... 51, 52, 54
Phylogenetic inference......................................... 123–125 T
Phylogenetic survival analysis ......................123, 126–127 Tagmentation .............................173, 179, 187, 189, 192
PhyResSE....................................................................... 206 Target validation .......................................................23–33
Plasmids ............................................................. 24, 26–30, TB molecular bacterial load assay
80, 163, 164, 211, 219 (TB-MBLA).............................................. 153–159
Polymerase chain reaction (PCR) .................... 24, 26–29, TB Profiler ..................................................................... 205
125, 137, 150, 151, 158, 159, 170, 172–175, T cells ................................................ 95, 96, 99, 100, 103
186, 187, 189, 192, 207 Temporal scales ............................................................... 97
Protospacer adjacent motifs (PAM) .................. 24, 27, 32 Thermocycler ...................................................... 173, 174,
Pyrazinamdidase/nicotinamidase (PncA)...................... 66 186, 187, 189, 192
Pyrazinamide (PZA), potentiation..........................65–76, Transformation................................................... 26, 29, 80
205, 206 Transposons...........................................80, 178, 179, 211
Pyrazinoic acid (POA) .................................................... 66 Typing......................................... 163, 166, 212–215, 217
R V
Raman spectroscopy................................... 110, 111, 113, Valinomycin ..................................................................... 66
115, 134, 135 Viability testing .........................................................35–41
Random/pseudo-random migration ........................... 100 Vibrational fingerprints................................................. 111
Reference mapping .............................164, 166, 196, 197 Vitek 2 ......................................................... 130, 133, 136
Resazurin ..................................................... 68, 72, 74, 75
ResFinder.................. 167, 206, 213, 215, 219, 222, 223 W
Resistance determinants....................................... 137, 196
Resistance mutations............................................ 197, 203 Wavelength-modified Raman spectroscopy................. 134
Resistance protocols...................................................... 137 Whole genome sequencing (WGS)........... 121, 137–140,
Reverse transcriptase ................................... 150, 158, 159 163–169, 172, 173, 177, 181, 185–193,
Rifampicin (RIF) ........................................ 13, 66, 67, 70, 195–208, 211–223
71, 73, 74, 138, 154, 185, 205, 206 Within-host modeling....................................... 79, 81–88,
RNA .................... 24, 26, 29, 33, 62, 147–150, 155–159 95, 103
World Health Organisation (WHO)........... 36, 139, 145,
S 154, 167, 185, 186, 196
Salmonella...................................206, 213, 214, 216, 217 Z

SARS-CoV-2 ................................................................... 96
Scattered light integrating collector (SLIC)................ 135 Zebrafish ...................................................................... 1–10
Sensititre ........................................................................ 136 Ziehl–Neelsen (ZN)..................................................37, 39
Zika virus ....................................................................... 164

Antibiotic Resistance Protocols, Fourth Edition Springer, 2024

Uploaded by

Copyright:

Available Formats

Antibiotic Resistance Protocols, Fourth Edition Springer, 2024

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Antibiotic Resistance Protocols, Fourth Edition Springer, 2024

Uploaded by

Copyright:

Available Formats

Methods in

Molecular Biology 2833

Stephen H. Gillespie Editor

For further volumes:

ISSN 1064-3745 ISSN 1940-6029 (electronic)

If disposing of this product, please recycle the paper.

St Andrews, Fife, UK Stephen H. Gillespie

13 Rapid Drug Susceptibility Testing to Preserve Antibiotics. . . . . . . . . . . . . . . . . . . . 129

Using the Zebrafish Larval Model of Infection to Investigate

Bacterial infections are commonly treated with antibiotics, that

respond to antibiotics in an in vivo context as they do in vitro.

1. E3 10× stock: 50 mM NaCl, 1.7 mM KCl, 3.3 mM CaCl2,

10. Antibiotics of interest (see Note 3).

17. Tryptone soy broth (TSB) (30 g/L in distilled water).

These protocols are intended to be used with 33 hpf larvae infected

4. Finally, keep a group of larvae that are untreated.

Total incubation time of E3

3. Repeat step 2 with the PBS controls.

3.4 Titration of 1. Infected larvae are immersed in E3 containing the antibiotic

compared to untreated controls are exhibiting a bactericidal or

1. Preparation of working solution: 50 mL 10× E3 stock is made

6. At such low volumes shaking is not necessary if it is not

11. It is important that where possible, fish are rinsed in fresh E3 to

Three-Dimensional Rotary Culture to Model Mycobacterial

Key words Mycobacterium, RCCS, Microgravity, Low-shear suspension, Three-dimensional

Biofilms are communities of bacteria encased in an extracellular

of antimicrobial drugs. Corded biofilm-like growth of M. tb was

a controllable speed to keep the growing cell aggregations in sus-

2.1 Culture Media 1. Middlebrook 7H9 broth base (Merck).

2.2 RCCS Culture 1. RCCS-4DQ bioreactor system (Synthecon Ltd).

All work with Mycobacterium bovis BCG should be conducted in a

Once dissolved, add 60 mL glycerol followed by 100 μL 1%

3.2 RCCS Inoculum 1. Prepare the mycobacterial inoculum by thawing a glycerol

1. Middlebrook ADC (albumin, dextrose, catalase) growth sup-

6. Ensure an excess of culture inoculum, for example, 50 mL to

17. The timeframes indicated in this protocol are for culturing

S.J.W. would like to acknowledge funding from The Royal Society

cell wall carbohydrates during biofilm growth functions of three-dimensional hepatocyte

Rapid Gene Silencing Followed by Antimicrobial

Antimicrobial resistance is one of the largest global threats to health

antimicrobials and their mechanisms of action, a combination of

Fig. 1 Workflow for rapid gene silencing. (Created with BioRender.com)

7. Super Optimal broth with Catabolite repression (S.O.C)

2.2 Colony PCR 1. Mycobacterium abscessus colonies.

2.3 Agarose Gel 1. Loading dye (e.g., GelRed®).

2.5 Transformation 1. Middlebrook 7H9 Broth (M7H9) (Difco™).

2.6 Validation of 1. Middlebrook 7H9 Broth (M7H9) (Difco™).

5. Master Mix for real-time quantitative PCR.

2.7 Gene Essentiality 1. Reagent Dispenser (e.g., Multidrop Combi Reagent

3. Perform a BLASTN search of the sgRNA and the PAM against

A primer sequence for this has been provided by Rock

Component (see Note 2) Volume

Run the reaction mix in a thermal cycler at 42 °C for

12. Mix the reaction gently by tapping, or gently pipetting. Iden-

3.4 Transformation 1. Grow an M. abscessus culture to OD600 ~ 0.4 (in 100 mL

trifuged at 3000× g at 4 °C for 10 min. Every wash, the volume

3.5 Validation of 1. It is important to evidence that successful, on-target transcrip-

3.6 HT-SPOTi: 1. Make 200 μL of M7H10 agar according to the manufacturer’s

2. Prepare the reagent dispenser to dispense 200 μL into each well