Dissertation
submitted to the
Combined Faculties for the Natural Sciences and for Mathematics
of the Ruperto-Carola University of Heidelberg, Germany
for the degree of
Doctor of Natural Sciences
presented by
Kirti Prakash, MSc (Tech)
born in: Lucknow, India
Oral-examination: 01 June 2016
KIRTI PRAKASH
THE PERIODIC AND
DYNAMIC STRUCTURE OF
CHROMATIN
REFEREES:
PROF. DR. CHRISTOPH CREMER
PD DR. KARL ROHR
Copyright © 2016 Kirti Prakash
typeset with tufte-latex
tufte-latex.googlecode.com
For academic use only.
To life and light!
Summary
The organisation of chromatin is non-random and shows a broad
diversity across cell types, developmental stages, and cell cycle stages.
During G0 and G1 phase of interphase, chromatin displays a bivalent
status. The condensed chromatin (heterochromatin) at the nuclear
periphery is mostly associated with low levels of gene expression,
while the loosened chromatin (euchromatin) towards the interior of
the nucleus is associated with higher gene expression. This quiescent
picture of interphase radically changes when the cell cycle progresses
toward cell division. Firstly, during S phase, DNA is replicated,
and chromatin progressively condenses. This is followed by the G2
phase that shows a compact heterochromatin recruited towards the
centre of the nucleus. At the beginning of mitosis, the chromosomes
condense with a significant topological change in their organisation
and are segregated during the next stages of the cell division. Meiotic
chromosomes are also highly condensed as mitotic chromosomes but
show a particular functional structure, which prepares germ cells to
exchange DNA sequences between their homologous chromosomes
to generate diversity. To summarise, chromatin experiences dramatic
organisational changes during mitosis and meiosis. These changes
in chromatin organisation during the lifetime of a cell show that
chromatin is not a static entity but highly dynamic in nature.
For a variety of reasons, conventional light and electron microscopy have not been able to fully capture the finer details of
chromatin organisation and dynamics. For a long time, description
of the interphase nucleus was limited to delineate the euchromatinheterochromatin dichotomy or describe some specific nuclear elements such as the nucleolus. Advancements in molecular biology
during the last thirty years have brought an immense amount of information about how chromatin is organised and genes are regulated.
As a classical example, the globin gene has been shown to display a
highly constrained shape forced by chromatin looping that brings the
regulatory regions to the promoter of the gene. Nowadays, genomic
studies can acquire an immense amount of information regarding
chromatin organisation and gene regulation, leaving one with the
expectation that structure of individual genes could potentially be
described visually if sufficient specificity and resolution were reached.
With the advent of various super-resolution methods, in particular,
single molecule localisation microscopy (SMLM) based methods and
recently developed strategies for labelling DNA, it is now possible
to study chromatin organisation and underlying gene regulatory
mechanisms at the nanoscale.
During my PhD, I have analysed a broad range of nuclear phenotypes using SMLM. My analyses contribute to the description of
a periodic and dynamic structure of chromatin. Moreover, I have
described several elementary chromatin structures that I call chromatin domains, both in interphase and meiosis, that are potentially
associated with a local function such as gene activation or silencing.
Firstly with colleagues, I established an experimental setup to
study chromatin organisation with single molecule localisation microscopy. I investigated how UV-induced photo-conversion of conventional DNA dyes allows increasing sufficiently the labelling density
such that it is possible to study various organisational aspects of chromatin in basal interphase. An adequate imaging protocol has been
established to bring DNA minor groove binding dyes such as Hoechst
33258, Hoechst 33342 and DAPI (4’,6-diamidino-2-phenylindole) into
an efficient blinking state necessary to record single molecule locations with high precision. This method was applied to several cell
types to investigate the chromatin organisation during different stages
of the cell cycle at the highest resolution currently achievable with
light microscopy.
The results show that the method can capture several hierarchical levels of chromatin organisation. In reverse hierarchical order, I
could describe previously known chromatin territories of 1000 nm,
subchromosomal domains of 500 nm, chromatin domains of 100 to
400 nm (and further sub-categories of active or repressed domains)
and chromatin fibres below 100 nm, mostly between 30 to 60 nm.
Individual nucleosomal domains are also described, which tend to
cluster in batches of 10-15 nucleosomes, a number close to one found
in genomic studies upstream to promoter regions. Next, with colleagues, I studied the dynamics of chromatin using stress as a model
system. It was found that short-term oxygen and nutrient deprivation
provokes chromatin to shrink to a hollow, condensed ring and rod-like
configuration, which reverses back to the initial structure when the
stress conditions cease. The condensed network of rods and rings
interspersed with large, chromatin-sparse nuclear voids were 40-700
nm in dimension, capturing another level of chromatin organisation
not described before.
Finally, I explored the unique properties of chromatin during
meiosis, which has escaped analysis at the single-molecule level until
now. Single molecule analysis revealed unexpected highly recognisable periodic patterns of chromatin. Firstly, I observed that meiotic
chromatin show unique clusters of 250 nm diameter along the synaptonemal complex, extended laterally by chromatin fibres forming
loops. These clusters show a remarkable periodicity of 500 nm, a
pattern possible to spot because of the highly deterministic nature
of pachytene chromosomes and the resolution of the experimental
setup. Furthermore, guided by genomic data, I selected histone
modifications associated with different chromatin states to dissect
the morphology of meiotic chromosomes. I could examine the morphology of these chromosomes into three spatially distinct nanoscale
sub-compartments. Histone mark H3K4me3 associated with active
chromatin was found in a lateral position, potentially located at the
places of de novo double-strand breaks. Repressive histone mark
H3K27me3 was shown to display a surprising medial symmetrical
and periodic pattern, putatively associated with recombination. Finally, centromeric histone mark H3K9me3 locates at one of meiotic
chromosome ends and is potentially associated with repression of
repeated regions and pairing of homologous chromosomes at early
stages. I summarise these findings in a comprehensive final model.
Overall, I have used new information brought by super-resolution
technologies to show the dynamics of chromatin in various processes
and novel orders of chromatin compaction, which were not reported
previously. Among these new levels of chromatin compaction are the
interphase hierarchical chromatin domains, the stress pattern of cells
upon oxygen and nutrients deprivation and the novel epigenetic domains found at pachytene stage of meiosis. These architectures show
that the organisation of chromatin is more complex than thought
before, dynamic in nature and shows a high order of periodicity.
Further investigation is, therefore, necessary to understand how chromatin transits from a ’beads-on-string’ model to the intermediary
chromatin domains and finally to the commonly observed X-shaped
chromosomes.
Zusammenfassung
Chromatin ist hoch organisiert und zeigt eine große Vielfalt an Strukturen in Abhängigkeit vom Zelltyp, Entwicklungstadien und von
den Zellzyklusphasen. Während der Interphase besitzt Chromatin
beispielsweise einen bivalenten Zustand: kondensiertes Chromatin
(Heterochromatin) in der nuklearen Peripherie wird hauptsächlich mit
einem geringen Maß an Genexpression assoziiert, während aufgelockertes Chromatin (Euchromatin) im Zentrum des Nukleus mit höherer
Genexpression in Verbindung gebracht wird. Dieses inaktive Erscheinen des Chromatins in der Interphase ändert sich radikal sobald
die Zelle in die Metaphase eintritt. Zuerst, während der S-Phase,
wird DNS repliziert und Chromatin zunehmend kondensiert. Anschließend, in der G2 Phase, wird das kompakte Heterochromatin im
Zentrum des Zellkerns ausgerichtet. Im Anschluss folgt die G2-Phase,
welche zum Zentrum des Nukleus ausgerichtetes, kompaktes Heterochromatin zur Erhöhung der Transkriptionsoberfläche aufweist. Zu
Beginn der Mitose kondensieren die Chromosomen mit tiefgreifenden
topologischen Änderungen in ihrer Organisation um dann in den
weiteren Schritten der Zellteilung getrennt zu werden. Meiotisches
Chromatin ist genauso wie mitotisches Chromatin stark kondensiert,
zeigt aber zudem eine besondere deterministische Struktur, welche
die Keimzellen darauf vorbereitet, ihre DNS-Sequenz mit ihren homologen Chromosomen auszutauschen. Dies hat zur Folge, dass das
Chromatin eine dramatische organisatorische Veränderung während
der Mitose und Meiose erfährt. Diese Änderungen in der Chromatinorganisation während des Lebenszyklus einer Zelle zeigt, dass
Chromatin kein statisches, sondern ein höchst dynamisches Gebilde
ist.
Aufgrund verschiedener Einschränkungen waren bisher konventionelle Licht- und Elektronenmikroskopie nur in sehr begrenzter
Weisein der Lage, Details der Chromatinorganisation und Dynamik
zu erfassen. Für lange Zeit beschränkte sich die Beschreibung des
Nukleus in der Interphase auf den Euchromatin-Heterochromatin
Dualismus oder einige besondere nukleare Elemente wie die Nukleoli. Die Entwicklung neuer hochauflösender Methoden in der Lichtmikroskopie, insbesondere der Lokalisationsmikroskopie von einzelnen Molekülen (single molecule localisation microscopy, SMLM),
und neuer Strategien zum Markieren von DNS, erlaubt nun Untersuchungen der Chromatinorganisation und der zugrundeliegenden
Genregulationsmechanismen im Nanometer-Bereich.
Ich benutzte einen Versuchsaufbau, mit dem die Chromatinorganisation mittels Lokalisationsmikroskopie einzelner Moleküle möglich
ist. Während meiner Promotion analysierte ich eine breite Palette an
nuklearen Phänotypen mittels SMLM, welche zur genaueren Beschreibung der periodischen und dynamischen Struktur von Chromatins
beitragen. Des Weiteren untersuchte ich verschiedene Arten an
grundlegenden Chromatinstrukturen, welche ich als Chromatindomänen bezeichne, jeweils in der Interphase und der Meiose, welche potentiell mit lokaler Funktionalität wie etwa Gen-Aktivierung oder Genstillegung assoziiert sind. Ich beschreibe wie die UV-induzierte Photokonversion von gewöhnlichen DNS-Fluoreszenzfarbstoffen dazu
verwendet werden kann, die Markierungsdichte für die SMLM zu erhöhen, so dass es möglich war verschiedene organisatorische Aspekte
von Chromatin in der basalen Interphase darzustellen. Es wurde ein
experimentelles Protokoll für DNS-bindende Farbstoffe wie Hoechst
33258, Hoechst 33342 und DAPI (4’,6-diamidino-2-phenylindole),
welche an die kleine DNS Furche binden, entwickelt, um diese in
einen effizienteren "Blink"-Zustand zu versetzen und letztlich eine
höhere Lokalisationsgenauigkeit zu erhalten.
Diese Methode wurde dazu verwendet, in mehreren Zelltypen
die Chromatinorganisation während den verschiedenen Phasen des
Zellzyklus bei der zurzeit höchstmöglichen lichtmikroskopischen Auflösung zu untersuchen. Die Ergebnisse zeigen, dass man mit dieser
Methode mehrere hierarchische Ebenen der Chromatinorganisation
erfassen kann. So konnte ich bereits bekannte Chromatinabschnitte
von 1000 nm, subchromosomale Domänen von 500 nm, Chromatindomänen von 100 bis 400 nm (und weitere Unterkategorien an aktiven
oder unterdrückten Domänen) sowie Chromatinfasern unterhalb von
100 nm, üblicherweise zwischen 30 und 60 nm, abbilden. Außerdem
werden individuelle nukleosomale Domänen beschrieben, welche
dazu tendieren, sich in Gruppen von 10-15 Nukleosomen anzuhäufen,
ähnlich der Anzahl welche bei genomischen Studien oberhalb von
Promoter Regionen gefunden wurden.
Weiterhin untersuchte ich zusammen mit Kollegen die Dynamik
von Chromatin unter Einfluss von „Stress“. Wir fanden heraus, dass
kurzzeitiger Sauerstoff- und Nährstoffentzug Chromatin dazu veranlasst, sich zu einem hohlen, kondensierten Ring sowie zu stabähnlichen Konfigurationen zusammenzuziehen, wobei diese Chromatinkonfigurationen zur Ausgangsstruktur zurückkehren nachdem
die Stressbedingungen enden.
Das kondensierte Netzwerk aus Stäbchen und Ringen ist mit
großen, chromatinarmen nuklearen Hohlräumen von 40-700 nm
Durchmesser durchsetzt, was eine weitere Ebene an zuvor unbeschriebener Chromatinorganisation darstellt.
Zuletzt untersuchte ich die einzigartigen Eigenschaften von Chromatin während der Meiose, welche noch nie zuvor auf Einzelmolekülebene analysiert wurden. Die Einzelmolekülanalyse offenbarte unerwartet deutliche periodische Chromatinmuster. Zuerst beobachtete
ich, dass meitotisches Chromatin, spezifische Cluster mit 250 nm
Durchmesser entlang des synaptonemalen Komplexes aufweist, welche
lateral von schleifenformenden Chromatinfasern erweitert werden.
Diese Cluster zeigen eine bemerkenswerte Periodizität von 500
nm, ein Muster welches durch die inhärente Beschaffenheit von
pachytänen Chromosomen sowie das Auflösungsvermögen des experimentellen Aufbaus bestimmt war. Desweiteren wählte ich, ausgehend von genomischen Daten, bestimmte posttranslationale HistonModifikationen, welche mit verschiedenen Chromatinzuständen verbunden sind, um die Morphologie meiotischer Chromosome aufzugliedern. Ich konnte die Morphologie dieser Chormosome in mindestens drei räumlich abgetrennte nanoskalige Unterabschnitte zerlegen. Der Histon Marker H3K4me3, assoziiert mit aktivem Chromatin, wurde in lateraler Position gefunden, möglicherweise an Genpositionionen oder Stellen von de novo Doppelstrangbrüchen. Der
Histon Marker H3K27me3, assoziiert mit inaktivem Chromatin, zeigte
ein überraschend symmetrisches Muster auf, welches mutmaßlich mit
Rekombination im Zusammenhang steht. Der H3K27me3-Marker für
inaktives Chromatin, zeigt eine hohe Periodizität von 500 nm entlang
der Synapsis, welche möglicherweise mit der Chromatindichte in
Zusammenhang steht. Zuletzt wurde der zentrometrische Histone
Marker H3K9me3 betrachtet, welcher an einem Ende des meitotischen
Chromosoms lokalisiert und möglicherweise an der Hemmung sich
wiederholender Regionen und Paarung homologer Chromosomen in
frühen Stadien verbunden ist. Die Ergebnisse sind in einem umfangreichen, finalen Modell zusammengefasst.
Im Rahmen dieser Arbeit konnte ich zum ersten Mal mit hochauflösender Mikroskopie die Dynamik von Chromatin in verschiedenen Prozessen zeigen und neuartige Anordnungen der Chromatinverdichtung nachweisen. Unter diesen neuartigen Stufen der Chromatinverdichtung befinden sich die hierarchischen Chromatindomänen der Interphase, das Stress-Muster von Zellen aufgrund von
Sauerstoff- und Nährstoffmangel sowie die neu gefundenen epigenetischen Domänen im pachytänen Schritt der Meiose. Diese
Architekturen zeigen, dass die Organisation von Chromatin weit
komplexer ist als zuvor angenommen, dynamische Eigenschaften
sowie ein hohes Maß an Periodizität besitzt.
Weitere Untersuchungen sind daher notwendig um zu verstehen, wie Chromatin von einem Perlenschnurmodell zu intermediären
Chromatindomänen und abschließend zu den gemeinhin beobachteten
X-förmigen Chromosomen wechselt.
Contents
Summary
7
Zusammenfassung
11
Abbreviations
23
Preface
27
1
2
A Condensed History of Chromatin Research
1.1 The early research on the nucleus and chromatin . . .
1.2 Chromatin bares information: the chromosomes and
genes era (1870—1945) . . . . . . . . . . . . . . . . . . .
1.3 Chromatin as a decision center of the cellular factory:
the golden age of molecular biology and electron microscopy (1944-1980) . . . . . . . . . . . . . . . . . . . .
1.4 Chromatin as a highly structured system: genomic
data, localisation methods and modelling (1980 onwards)
1.5 The substratum of chromatin memory: epigenetic regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6 Fine-scale chromatin architecture: a new modelling area
1.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . .
1.8 Acknowledgement . . . . . . . . . . . . . . . . . . . . .
31
31
32
34
37
41
42
44
45
Investigating Chromatin Organisation using Single Molecule
Localisation Microscopy
47
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Single-molecule localization microscopy: state-of-the-art 49
2.2.1 Principle of SMLM . . . . . . . . . . . . . . . . . 49
2.2.2 The different SMLM methods: a historical perspective . . . . . . . . . . . . . . . . . . . . . . . 49
2.3 Application of SMLM to image chromatin . . . . . . . 51
2.3.1 The tao of SMLM . . . . . . . . . . . . . . . . . . 51
2.3.2 Importance of a good localization precision in
order to improve resolution . . . . . . . . . . . . 51
2.3.3 Importance of high signal density to improve
signal-to-noise ratio . . . . . . . . . . . . . . . . 53
2.3.4 Limitations of previous approaches to study
chromatin organisation . . . . . . . . . . . . . . 54
2.4 A method to reach high labelling density of chromatin
with SMLM . . . . . . . . . . . . . . . . . . . . . . . . . 54
Theory of DNA dye fluorescence . . . . . . . . .
Adapting study of DNA dyes fluorescence to
SMLM . . . . . . . . . . . . . . . . . . . . . . . .
2.4.3 Optimization of the photoconversion process .
2.4.4 Optimization of the buffer conditions . . . . . .
2.4.5 Multicolor imaging with DNA . . . . . . . . . .
2.4.6 A summary of various approaches used to study
DNA with SMLM . . . . . . . . . . . . . . . . .
SMLM microscope design and imaging pipeline . . . .
2.5.1 Sample preparation for SMLM . . . . . . . . . .
2.5.2 Imaging medium . . . . . . . . . . . . . . . . . .
Data acquisition for SMLM . . . . . . . . . . . . . . . .
Data reconstruction for SMLM . . . . . . . . . . . . . .
2.7.1 Spot finding for SMLM . . . . . . . . . . . . . .
2.7.2 Drift correction algorithms for SMLM . . . . . .
2.7.3 Data visualisation for SMLM . . . . . . . . . . .
2.7.4 Data analysis for SMLM . . . . . . . . . . . . . .
Some further considerations for localisation microscopy
2.8.1 Artefacts in localisation microscopy . . . . . . .
2.8.2 Difference between localisation precision and
accuracy . . . . . . . . . . . . . . . . . . . . . . .
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.1
2.4.2
2.5
2.6
2.7
2.8
2.9
3
56
56
57
57
57
58
61
61
61
61
63
64
66
68
69
70
70
72
73
Structure, Function and Dynamics of Chromatin
75
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2 The hierarchical organisation of chromatin . . . . . . . 77
3.2.1 Chromosome territories (scale: 1000-2000 nm) . 77
3.2.2 Sub-chromosomal domains (scale: 500-1000 nm) 80
3.2.3 Chromatin domains (scale: 100-400 nm) . . . . . 81
3.2.4 Chromatin fibres (scale: 30-100 nm) . . . . . . . 85
3.2.5 A cluster-on-a-string model to describe the fibre/domain transition . . . . . . . . . . . . . . . 87
3.2.6 Nucleosome domains (scale: 10-30 nm) . . . . . 88
3.2.7 Inference about the chromatin structures using
local density maps . . . . . . . . . . . . . . . . . 90
3.2.8 Hierarchical organisation of chromatin structure 91
3.3 The dynamics of chromatin . . . . . . . . . . . . . . . . 91
3.3.1 Contrasting arrangement of eu- and hetero-chromatin
inside the cell nucleus . . . . . . . . . . . . . . . 92
3.3.2 Classifier identifies intermediate states between
eu- and heterochromatin regions in differentiated cells . . . . . . . . . . . . . . . . . . . . . . . 93
3.3.3 Chromatin dynamics during differentiation of
mesenchymal stem cells . . . . . . . . . . . . . . 93
3.3.4 Dynamics of chromatin upon stress . . . . . . . 95
3.3.5 Reversible compaction of chromatin under stress 96
3.3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . 98
3.4 The function of chromatin . . . . . . . . . . . . . . . . . 100
High DNA synthesis in periphery of chromatin
domains . . . . . . . . . . . . . . . . . . . . . . .
3.4.2 Stress-dependent transcription at the periphery
of chromatin domains . . . . . . . . . . . . . . .
3.4.3 Histone modifications allow to further dissect
chromatin compartmentalization . . . . . . . . .
3.4.4 SMLM identifies potential sites of the transcription machineries . . . . . . . . . . . . . . . . . .
Summary and discussion . . . . . . . . . . . . . . . . .
3.4.1
3.5
101
102
102
103
103
4
Periodic and Symmetric Organisation of Meiotic Chromosomes
105
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.2 Organisation of the synaptonemal complex (SC) . . . . 108
4.2.1 Superresolution imaging of the SC substructures 108
4.2.2 Quantification of SC substructures . . . . . . . . 109
4.2.3 A model for organisation of SC . . . . . . . . . . 111
4.3 Periodic organisation of pachytene chromosomes . . . 112
4.3.1 Superresolution imaging of pachytene chromosomes reveals periodic clusters of chromatin . . 113
4.3.2 Quantification of periodic chromatin clusters . 113
4.4 Functional organisation of pachytene chromosomes . . 114
4.4.1 Rational . . . . . . . . . . . . . . . . . . . . . . . 114
4.4.2 Clustering method sorts chromatin into functional epigenetic compartments . . . . . . . . . 115
4.4.3 Centromeric histone mark (H3K9me3) labels
one end of the SC . . . . . . . . . . . . . . . . . . 116
4.4.4 Repressive histone mark (H3K27me3) shows
characteristic periodic clusters along the SC . . 116
4.4.5 Histone mark (H3K4me3) associated with active
transcription emanates radially from the axis of
the SC . . . . . . . . . . . . . . . . . . . . . . . . 118
4.5 Structure and dynamics of meiotic chromosomes . . . 119
4.5.1 Lampbrush-like structures in mammalian meiotic chromosomes . . . . . . . . . . . . . . . . . 119
4.5.2 A model for SC spiralisation during the zygotene/pachytene transition . . . . . . . . . . . 120
4.6 A model of spatial distribution of chromatin around
the SC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.6.1 A ’cluster-on-a-string’ model for spatial distribution of pachytene chromosomes . . . . . . . . 122
4.7 Summary and Conclusion . . . . . . . . . . . . . . . . . 122
5
Conclusions
127
5.1 Originality of the work presented here . . . . . . . . . . 127
5.2 A general methodology to study chromatin architecture 128
5.3 Limitations of the method and possible improvements 129
5.4 New avenues for the study of chromatin patterns during meiosis . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.5
Enlarging the spectrum of questions: chromatin organisation as a fundamental principle of nucleus formation 130
Appendices
131
Bibliography
143
Publications
165
Acknowledgements
171
List of Figures
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
Purkinje drawings of the germinal vesicle of a hen egg
Sketch of a cell . . . . . . . . . . . . . . . . . . . . . . . .
Cell nuclei from the eggs of a Salamander species . . .
The double helix . . . . . . . . . . . . . . . . . . . . . .
Sketch of a nucleus . . . . . . . . . . . . . . . . . . . . .
Comparison between superresolution microscopy and
conventional microscopy . . . . . . . . . . . . . . . . . .
The beads on a string model of nucleosomes . . . . . .
Different genomic methods to study chromatin . . . .
Comparison between predicted data and experimental
data in the genomic era . . . . . . . . . . . . . . . . . .
31
32
33
35
36
40
41
43
43
2.10
2.11
2.12
2.13
2.14
2.15
2.16
2.17
2.18
2.19
2.20
2.21
Chromatin spatial organisation . . . . . . . . . . . . . .
Schematic illustration of the underlying principle of
SMLM . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The tao of single molecule microscopy . . . . . . . . . .
Importance of good localization precision . . . . . . . .
Importance of high signal density . . . . . . . . . . . .
Photo-conversion of common DNA dyes . . . . . . . .
Transitions between excitation states during fluorescence process . . . . . . . . . . . . . . . . . . . . . . . .
Optimization of the photoconversion process . . . . . .
Comparison of various DNA dyes and other fluorescent
probes . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison of direct and indirect DNA labeling . . . .
Different kinds of background in a SMLM measurement
A schematic of the SMLM microscope . . . . . . . . . .
SMLM data reconstruction and analysis flowchart . .
Various reconstruction algorithms for SMLM . . . . .
Structure based drift correction for SMLM . . . . . . .
Quantification of sample drift . . . . . . . . . . . . . . .
Comparison of various visualisation algorithms . . . .
Single molecule autocorrelation . . . . . . . . . . . . . .
Artifacts in localisation microscopy . . . . . . . . . . . .
SMLM image of Drosophila polytene chromosome . .
Localisation precision Vs localisation accuracy . . . . .
58
60
61
62
63
66
67
68
69
70
71
72
73
3.1
3.2
Methods for chromatin research . . . . . . . . . . . . .
A simulated chromosome territory map . . . . . . . . .
76
78
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
47
49
51
52
53
55
56
57
3.28
3.29
Identification of sub-chromosomal territories . . . . . . 79
Sub-chromosomal regions corresponding to potential
transcription machineries . . . . . . . . . . . . . . . . . 80
Inter chromatin compartments observed by SMLM . . 81
SMLM reveals several levels of chromatin compaction
83
Transition between chromatin fibre and sub-chromosomal
territories . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Transition between chromatin fibres and chromatin
domains . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Differential topology of nuclear and cytoplasmic molecular set up . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Bead on a string model . . . . . . . . . . . . . . . . . . . 87
Inverse distance map of pixels based on the next 5
neighbours . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Comparison of chromatin patterns in random and experimental data . . . . . . . . . . . . . . . . . . . . . . . 88
Impact of increasing number of nearest-neighbours . . 89
Quantification of chromatin condensation . . . . . . . . 90
Inference from local chromatin density maps . . . . . . 91
Different milestones of chromatin architecture research 92
Contrasting nuclear phenotypes . . . . . . . . . . . . . 93
Pixel density classification of chromatin . . . . . . . . . 94
Compaction increases during early differentiation of
mesenchymal stem cells . . . . . . . . . . . . . . . . . . 95
Chromatin compaction upon stress . . . . . . . . . . . . 96
3D surface plot of chromatin nanostructure under ischemia conditions . . . . . . . . . . . . . . . . . . . . . . 96
Reversible compaction of chromatin . . . . . . . . . . . 97
Nearest neighbour characterization to describe the extent of chromatin reversibility . . . . . . . . . . . . . . . 97
Mean nearest neighbour distance characterisation for
different moments of the OND stress experiment. . . . 98
Localisation precision of untreated cells, subjected to 1
hour of OND or 5, 15, 60 and 240 minutes after release
from OND . . . . . . . . . . . . . . . . . . . . . . . . . . 98
DNA synthesis occurs at periphery of chromatin domains100
OND induces compaction of chromatin to mechanistically reduce transcription . . . . . . . . . . . . . . . . . 101
Distribution of histone modifications during interphase 103
Functional compartments inside the cell nucleus . . . . 103
4.1
4.2
4.3
4.4
4.5
4.6
4.7
An epigenetic model of pachytene chromosomes . .
Stages of meiosis prophase I . . . . . . . . . . . . . . .
Organisation of the synaptonemal complex proteins .
Quantification of SC substructures . . . . . . . . . . .
A model for the synaptonemal complex . . . . . . . .
Higher order periodic clusters of chromatin . . . . .
Characterization of pachytene chromosome clusters .
3.3
3.4
3.5
3.6
3.7
3.8
3.9
3.10
3.11
3.12
3.13
3.14
3.15
3.16
3.17
3.18
3.19
3.20
3.21
3.22
3.23
3.24
3.25
3.26
3.27
.
.
.
.
.
.
.
106
107
109
110
111
112
114
4.8
4.9
4.10
4.11
4.12
4.13
4.14
4.15
4.16
4.17
5.1
A.1
A.2
A.3
A.4
A.5
A.6
B.1
B.2
B.3
B.4
B.5
B.6
B.7
B.8
B.9
P.1
P.2
P.3
P.4
P.5
P.6
P.7
Chromatin compaction as a function of histone modifcations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Association of histone modifications . . . . . . . . . . . 115
Functional chromatin states . . . . . . . . . . . . . . . . 116
Centromeric position of H3K9me3 along the SC . . . . 117
H3K27me3 displays periodic paired clusters in proximity to the SC . . . . . . . . . . . . . . . . . . . . . . . . . 118
Lower orders of chromatin domains characterised by
H3K4me3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Distribution of transcriptionally active chromatin . . . 120
Helical structure of the pachytene chromosomes . . . . 121
A snake model for the formation of the synapsis . . . . 122
Symmetric and periodic organisation of pachytene chromsomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Dynamic architecture of chromatin across the various
stages of cell cycle . . . . . . . . . . . . . . . . . . . . . . 129
Chromatin density of CT/IC regions . . . . . . . . . . . 133
Linking microscopy to genomic data . . . . . . . . . . . 133
HiC data at various levels of resolution . . . . . . . . . 134
Mitotic chromosomes at high resolution . . . . . . . . . 134
OND replicates . . . . . . . . . . . . . . . . . . . . . . . 134
Fourier Ring Correlation (FRC) analysis of DNA/SMLM
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Relative localization of SYCP1 and SYCP3 within the
synaptonemal complex . . . . . . . . . . . . . . . . . . . 137
Validation of the chromatin clusters . . . . . . . . . . . 137
SEM image of chromatin organisation around SC . . . 138
Organisation of condensed chromatin . . . . . . . . . . 138
Helical structure of the pachytene chromosomes using
information from H3K27me3 clusters. . . . . . . . . . . 138
Spiralization of H3K9me3 confirmed by patterns of DNA139
Helicoidal nature of pachytene chromosomes. . . . . . 140
SYCP3 strands move apart at non-centromeric end of
the SC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Characteristics of epigenetic clusters identified to regulate pachytene chromosomes . . . . . . . . . . . . . . . 141
A simulated chromosome territory model . . . . . . . .
A model for the epigenetic landscape of meiotic chromosomes . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effect of ischemia on chromatin nanostructure . . . . .
Superresolution imaging of mitotic chromosomes . . .
3D surface plot of DNA . . . . . . . . . . . . . . . . . .
Algorithm scheme for processing single molecule localisation microscopy data . . . . . . . . . . . . . . . . . .
Drift correction strategies for superresolution imaging
modalities . . . . . . . . . . . . . . . . . . . . . . . . . .
165
165
166
167
167
168
168
P.8
P.9
Distribution of post translational modifications along
the SC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Distribution of H3K4me3 mark around the synaptonemal complex (SC) . . . . . . . . . . . . . . . . . . . . . . 169
Abbreviations
µm
3D
Ac
AFM
BALM
BaLM
bp
CCD
ChIP
ChIP-seq
CLSM
COI
CRISPR/Cas9
CTCF
Cy3
Cy5
DAPI
ddNTP
DNA
DSB
EdU
EM
emGFP
FISH
fPALM
fps
FRC
FRET
FWHM
GFP
H2B
H3
H3K9ac
H3K14ac
H3K27ac
H3K27me3
H3K4me1
Micrometer
Three dimensional
Acetylation
Atom force microscopy
Binding-activated localisation microscopy
Bleaching/blinking assisted localisation microscopy
Base pair
Charge coupled device
Chromatin immunoprecipitation
ChIP-sequencing
Confocal laser scanning microscopy
Center of intensity
Clustered regularly-interspaced short palindromic repeats/CRISPR associated protein 9
CCCTC-binding factor
Cyanine 3
Cyanine 5
4’,6-diamidino-2-phenylindole
Dideoxynucleotide
Deoxyribonucleic acid
Double strand break
5-ethynyl-2’-deoxyuridine
Electron microscopy
Emerald green fluorescent protein
Fluorescence in situ hybridization
Fluorescence photo-activated localization microscopy
Frames per second
Fourier ring correlation
Fluorescence resonance energy transfer
Full width at half maximum
Green fluorescent protein
Histone 2B
Histone 3
Histone 3 lysine 9 acetylation
Histone 3 lysine 14 acetylation
Histone 3 lysine 27 acetylation
Histone 3 lysine 27 tri-methylation
Histone 3 lysine 4 mono-methylation
H3K4me2
H3K4me3
H3K9me3
H4
HAT
HDAC
HL-1
iPALM
kb
kDa
LBC
LCR
LM
LSM
LTR
Mb
Me
MEA
MLE
mRNA
MSC
NA
nm
NND
OND
OTF
P
PALM
PBS
PDC
Pixel
Pol II
Pol II Ser2P
PRDM9
PSF
PTM
R
RNA
ROI
ROS
rRNA
S
S phase
SC
SYCP
SEM
SIM
SMLM
SNR
Histone 3 lysine 4 di-methylation
Histone 3 lysine 4 tri-methylation
Histone 3 lysine 9 tri-methylation
Histone 4
Histone acetyl-transferase
Histone deacetylase
Cardiomyocyte cell line
Interference photo-activated localization microscopy
Kilobase
Kilodalton
Lampbrush chromosome
Locus control region
Light microscopy
Least-squares method
Long terminal repeat
Megabase
Methylation
Monoethanolamine
Maximum likelihood estimation
Messenger RNA
Mesenchymal stem cell
Numerical aperture
Nanometer
Nearest neighbour distance
Oxygen and nutrient deprivation
Optical transfer function
Phosphorylation
Photo-activated localization microscopy
Phosphate buffered saline
Pixel density classifier
Picture element
Polymerase II
Polymerase II serine 2 phosphorylation
PR domain zinc finger protein 9
Point spread function
Post-translational modification
Arginine
Ribonucleic acid
Region of interest
Reactive oxygen species
Ribosomal RNA
Serine
Synthesis phase
Synaptonemal complex
Synaptonemal complex proteins
Scanning electron microscopy
Structured illumination microscopy
Single molecule localization microscopy
Signal-to-noise ratio
SPDM
STED
STORM
Su
TAD
TALE
TEM
TIRF
tRNA
TSS
Ub
UV
VH7
VV
Y
YFP
YOYO-1
Spectral precision distance microscopy
Stimulated emission depletion
Stochastic optical reconstruction microscopy
Sumoylation
Topologically associating domain
Transcription activator-like effectors
Transmission electron microscopy
Total internal reflection fluorescence
Transfer RNA
Transcription start site
Ubiquitination
Ultra-violet
A human fibroblast cell line
Vybrant DyeCycle Violet
Tyrosine
Yellow fluorescent protein
A tetracationic homodimer of Oxazole yellow
All truth passes through three stages.
First, it is ridiculed.
Second, it is violently opposed.
Third, it is accepted as being self-evident.
Arthur Schopenhauer
Preface
This thesis is a mixture of a bit of history, a bit of theory, a bit of
technical literature, accidental findings, unconventional hypothesis
and a series of re-re-writing experiences. Thesis writing is a special
experience and as my friend David Fournier would say, one should
embrace the opportunity to write a monograph in order to look
back on your work and think about its implications. Also, there is
a difference between being merely able to fit some 100 figures and
50000 words to compile a thesis without any lore, rhythm or rhyme,
and synthesizing a good design with previous facts and fresh analysis
to fit a particular purpose. One of the good things about scientific
lore is that it lives in heart of the reader for a longer time than an
accumulation of facts written within the strict bounds of modern day
scientific writing.
Outline of the thesis
In chapter 1, I present a brief historical overview of chromatin research.
While writing this chapter, I got reminded about the following quote
from G.H. Hardy
"Statesmen despise publicists, painters despise art-critics, and physiologists,
physicists, or mathematicians have usually similar feelings: there is no scorn
more profound, or on the whole more justifiable, than that of the men who
make for the men who explain. Exposition, criticism, appreciation, is work
for second-rate minds."
Indeed it was an unusual feeling to write a historical briefing and
comment on other people’s work. But contrary to what Hardy says, it
was a wonderful learning experience on how chromatin biology has
evolved over the years and how new techniques are reconfirming the
old findings. It also provided me with a good insight regarding the
problems to attack in the future.
In chapter 2, I discuss the design and development of single
molecule imaging system to study chromatin architecture with conventional DNA dyes. Furthermore, basics of single molecule localisation microscopy (SMLM) and various elements involved in processing
and analysis of single molecule data are discussed.
In chapter 3, I discuss spatial and temporal aspect of chromatin
organisation from three viewpoints: basic building blocks, function
and dynamics. In particularly, some recent data from SMLM have
shown new orders of chromatin never imaged before, the so-called
chromatin domains, an order of compaction of chromatin between the
chromatin fibre and the chromosomal territory. These domains are
most likely associated to function, either activation or repression of
genes. I connect these different features to show how the novel building blocks or patterns can be the result of differential compaction of
chromatin as function of various proteins associated with chromatin.
One major finding presented here is the ring and rod like shapes
chromosomes form when put under stress. Interestingly, it seems
that stress is the fastest way to bring two homologous chromosomes
together.
In chapter 4, I present the epigenetic make-up of meiotic chromosomes. The periodic and symmetrical organisation of pachytene
chromosomes makes me believe that simple mathematical principles
lie underneath the seemingly complex looking genome organisation.
The challenge is to figure out how various mathematical shapes and
curves like conical helices and spirals can combine into one structure
that could help chromatin avoid entanglements during assembly and
disassembly process. Nevertheless, I propose a model for pairing of
homologous chromosomes based on coupling of snakes. I hope with
major future advancements in imaging, one will be able to see the
live pairing of homologous chromosomes.
In chapter 5, I summarise the major findings of this thesis and
put forward some new theories in the spirit of the following quote
by Imre Lakatos and discuss some possible future research directions
in chromatin biology. "Where theory lags behind the facts, we are dealing
with miserable degenerating research programmes".
Contributions of the thesis
The main contributions of the thesis are summarized below.
1. SMLM of DNA dyes: Initiation of the project and design of the
experiments. Observation of the blinking property of minor groove
DNA dyes making them applicable for single molecule imaging.
Contribution to the development of the optical setup. Adapting the
existing post processing algorithms to reconstruct the data done
by this method. Development of new algorithms to analyse and
calibrate the data (Szczurek*, Prakash*, Lee* et al. 2014; ŻurekBiesiada, Szczurek, Prakash et al. 2015; Żurek-Biesiada, Szczurek,
Prakash et al. 2016; this thesis). *: equal contributions.
2. Structure of interphase chromatin: Application of SMLM of DNA
dyes to study the architecture of interphase chromatin at several
structural levels, including chromatin domains and nucleosome
domains. Development of algorithms to quantify and characterise
chromatin states (Szczurek*, Prakash*, Lee* et al. 2014; ŻurekBiesiada, Szczurek, Prakash et al. 2015; this thesis). *: equal
contributions.
3. Mechanisms of chromatin condensation and reversibility upon
stress: Development of new algorithms to quantify and analyse
chromatin contraction and relaxation (Kirmes*, Szczurek*, Prakash
et al. 2015; this thesis). *: equal contributions.
4. Epigenetic landscape of meiotic chromosomes: Initiation of the
project and design of the experiments. First super-resolution images of chromatin during meiosis. Application of single molecule
auto-correlation to describe periodic clusters of chromatin along
the pachytene chromosome. Development of a method to predict chromatin compartments based on post-translational histone
modifications. (Prakash et al. 2015; this thesis).
5. Algorithms: Development of algorithms for analysis and visualization of single molecule localisation microscopy data. (Szczurek*,
Prakash*, Lee* et al. 2014; Zurek-Biesiada, Szczurek, Prakash et al.
2015; Kirmes*, Szczurek*, Prakash et al. 2015; Best* and Prakash*
et al. 2014; Hagmann* and Prakash* et al. 2014; this thesis). *:
equal contributions.
A detailed description of publications listed here can be found at
the publication section of the thesis. All figures from previous studies
presented here are reproduced according to publisher guidelines and
permission policies.
An apology of a stargazing scientist
As a beginner and over-enthusiastic scientist, I might have come
up with unconventional hypothesis or speculated a bit too much at
certain places or missed some of the important works on chromatin
research. The topics covered are quite broad and I do not pretend to
have expertise in all areas of chromatin biology.
This thesis is my attempt to provide a brief historical account
of chromatin research, basics of localisation microscopy, systematic
characterisation of interphase and meiotic chromosomes based on my
current knowledge and level of expertise. Thus, it is likely that such a
work might open doors for some trivial mistakes. Nonetheless, I hope
that this months long intellectual pursuit to decipher the various
spatial and functional aspects of chromatin organisation helps in
advancing the field.
“The model was so pretty that we wanted to believe it no matter what the data may say.”
James D. Watson
1 A Condensed History of Chromatin
Research
The cell nucleus is a discernible cellular compartment, where the
expression and regulation of genes take place. Due to limited tools
and methods, it remained ignored for a long time (until the late
19th century), however nowadays it is a major research topic. The
structure of DNA and in-depth study of the nucleus composition
has shown that the DNA is heavily constrained by proteins either
modulating gene expression or devoted to shaping DNA into various
topological structures such as loops and globules. The DNA with
all other associated protein machinery and RNA is referred to as
chromatin.
For the most part, chromatin research has highly depended on the
advancements of microscopy, from the early days of conventional light
and electron microscopy of the cell nucleus to the recent advanced
high-resolution and live imaging methods. Much more recently,
chromatin research has gained a lot from experiments which have
tried to map chromatin in 3D inside the cell nucleus to conclude that
it is organised in distinct territories. This, essentially, is the story told
in the present chapter.
1.1
The early research on the nucleus and chromatin
The nucleus was described for the first time as early as the eighteen
century by Antonie van Leeuwenhoek1 in the nucleated blood cells
of salmon. For a long time between the late 17th century and the first
thirty years of the 18th , it was at best described as a vesicle and did
not show much complexity than the name suggests. Felice Fontana2
in 1781 describes the first sub-structure in a nucleus of a skin cell
from an eel, probably a nucleolus [Baker, 1949]. The power of a good
drawing is illustrated by Johann Evangelist Purkyně3 [Purkyně, 1830],
where he presents the germinal vesicle of the hen egg (see Figure 1.1).
He comments as follows:
”Thus the scar [germinal disk] of the ovarian egg contains a special part,
peculiar to itself, a vesicle of the shape of a somewhat compressed sphere.
This vesicle is limited by a very delicate membrane and filled with a special
fluid, perhaps connected with procreation (for which reason I might call
it the germinal vesicle); it is sunk into a white breast-shaped projection
composed of globules and perforated in the middle” according to John
Figure 1.1: Purkinje drawings of the germinal vesicle of a hen egg. Reproduced
from the main set of figures in [Purkyně,
1830].
1
Antonie
van
Leeuwenhoek
(1632—1723),
a Dutch scientist
who
discovered
micro-organisms
by developing sophisticated microscopes. Considered to be the Father of
Microbiology.
Felice Fontana (1730—1805), an Italian
scientist, considered to be the father of
modern toxicology.
2
Johann
Evangelist
Purkyně
(1787—1869),
Czech physiologist.
His broad contributions touch neurophysiology, reproduction research,
physiology of the vision, pharmacological properties of plants and study of
cell composition.
3
32 the periodic and dynamic structure of chromatin
Baker’s translation from Latin [Baker, 1949].
As Baker points out, the nuclear envelope and the nucleoplasm are
described here for the first time. This first comprehensible picture
of a nucleus also brought the first sound hypothesis regarding the
function of the nucleus, a "connection with procreation", exemplifying
the need to find quality signals in biology, and so proper animal or
cellular models with extraordinary properties. As a consequence, at
one glance, one understands what the structure looks like, and what
it can do.
Though the nucleus was observed early, its function remained
unexplored for a long time. One of the initial hypothesis (∼ 1850)
stated that nucleus could be the place where new cells emerge [Baker,
1949]. These studies used optical microscopy and were mainly morphological. Mitotic chromosomes, which are easy to observe under
a microscope, captured the focus of many researchers and were the
only nuclear component to be studied for a long time. Interphase
chromosomes, at that time, were considered only by a handful of
scientists, among them Theodor Boveri4 [Boveri, 1888]. The habit to
consider only mitosis is also exemplified by modern textbooks, which
often present the mitotic chromosomes (highest level in the hierarchy
of chromatin compaction) after the initial beads-on-the-string and 30
nm chromatin fibre for the organisation of DNA (Figure 1.2). Interphase chromatin is obviously not organized in X-shape chromosomes
and is much more complex.
In the second part of the 19th century, chromatin research was
mainly focused on the chemical quantification of nucleic acids and
the fractionation of nuclear proteins. The point of view of chemists
was slightly different from that of 18th and early 19th century biologists. Most of them wanted to describe the chemical composition
of the cell exhaustively. One of them, Friedrich Miescher5 , decided
to do so for immune cells. In 1869, while studying in Tuebingen,
he discovered a substance in pure leukocyte solution which caused
the nucleus to swallow and explode upon mild injections of an alkaline solution. He found this substance to have a high phosphate
proportion and called it nuclein, in reference to its nuclear localization;
and later called it nucleic acid [Miescher, 1871, 1874, 1897, Kossel and
Kennaway, 1911, Kossel, 1883, 1884] (for detail see [Dahm, 2008]).
Later, Albrecht Kossel6 determined the composition of DNA to be
a mixture of adenine, cytosine, guanine and thymine, an important
step to understand how information is stored in the DNA molecule
[Kossel and Kennaway, 1911, Kossel, 1883, 1884].
1.2
Chromatin bares information: the chromosomes and genes
era (1870—1945)
Ernst Haeckel7 proposed that the nucleus contains factors responsible
for heredity (1866). Miescher is the first to have postulated a nuclear
origin of a cellular functions, the motion of sperm, though without
providing hints about the mechanism [Miescher, 1874, Dahm, 2008].
Figure 1.2: Sketch of a cell: On the left,
a cell prior to mitosis. On the right, two
sister cells coming from the same mitotic event [Flemming, 1882] (citation inspired by [Cremer and Cremer, 2009b]).
Theodor Boveri (1862—1915), German
biologist, first to propose that chromosomes occupy distinct territories inside
of the cell nucleus. Also known for the
Boveri-Sutton chromosome theory and
the discovery of the centrosome.
4
5
Friedrich Miescher (1844—1895), Swiss
physician and biologist, most noted for
his work on isolation and identification
of nucleic acids.
Albrecht Kossel(1853—1927), German
biochemist, who pioneered the work
in the field of genetics.
He was
awarded the Nobel Prize in Physiology
or Medicine in 1910 for the discovery
of the five fundamental nucleic bases
that constitute nucleic acids: adenine,
thymine, guanine, cytosine and uracil.
6
Ernst Haeckel (1834—1919), German
biologist known for establishing the first
phylogenetic tree of all living forms and
for establishing a relation between development and evolution (ontogeny recapitulates phylogeny).
7
a condensed history of chromatin research
During the second half of the 19th century, probably the most famous work on nuclear functionality [Hertwig, 1875] was conducted
by Oscar8 and Richard9 Hertwig. Oscar Hertwig, using the sea urchin
as a model, showed that a male and a female germ cell merge to
form the egg [Hertwig, 1875]. Moreover, he showed that organisms
come from one cell, leaving biologists with the very complex task of
explaining the mechanism of development. Oscar Hertwig’s observations led to the first proof that a nucleus is a place that stores some
information.
An interesting point to note is that the Hertwig brothers carried
out their initial work at Jena, where Ernst Abbe10 was pioneering the
work in optics, lens design and microscopy around the same time11 .
As optical technology was still limited at the end of the nineteenth
century, knowledge about chromatin function was restricted to the
description of the metaphase chromosomes and the characterization
of chemical composition. Though chromosomes themselves were
observed by several scientists in the first half of the nineteen century,
it is only in 1882 that they were associated with mitosis. In a famous
work, Walther Flemming12 described the content of the nucleus under
the term chromatin for the first time (Figure 1.2).
After the works of Flemming and others, the mitosis begins to get
the focus of the biologists. The nucleus was rapidly hypothesised to
be one of the factors where hereditary information can potentially be
stored and transmitted to daughter cells. As such, Wilhelm Roux13
in 1883 said that there is not only a question of quantity in mitosis
(meaning that both daughter cells will inherit the same amount of
chromatin), but there is also a question of quality (not all chromatin is
equivalent), which supposes that all granules of chromatin have different functions. Around the same time, the work of Oscar Hertwig was
prolonged by August Weismann14 germ plasm theory15 . Weismann
defined chromosome heredity as the process by which information
is transmitted from one generation to the other via chromosomes of
the germ lines, based on Hertwig observations that sperm + egg =
zygote, a sufficient note to claim that one needs information from
two different cells to form a new individual, and that information
is exchanged during the merging of the cells (Figure 1.3). Friedrich
Miescher, who characterised nucleic acids using nucleus of leukocytes
and salmon sperm, hypothesised that information might be stored
on this molecule by a change of stereochemistry; during fecundation,
each gamete will contribute with a molecule of a different state to
combine in the nucleus of the fertilised egg [Miescher, 1897].
Theodor Boveri and Walter Sutton16 were the first scientists to
establish a link between chromosomes dynamics during mitosis and
Mendel’s laws of heredity17 . According to their theories, there must
be entities that are passed differently to the next generation (genes,
though the term is coined later in history; see [Boveri, 1904, 1909]).
33
Oscar Hertwig (1849—1922),German
zoologist, the first to describe fecundation as the merging of a sperm and an
egg.
8
9
Richard Hertwig (1850—1937), German zoologist known for his work on
protists and the relationship between the
nucleus and the plasma as well as for
his work on meiosis and fecundation.
Figure 1.3: Cell nuclei from the eggs
of a Salamander species (adapted from
[Hertwig, 1898]).
Ernst Abbe (1840—1905), German
physicist, a major innovator of modern
microscopy. Abbe’s innovations in optics led to great improvements in microscope design and understanding of
resolution limits of light microscopy.
10
One wonders to which extent advancements in chromatin biology research
have depended on progress in optics
and microscopy.
11
Walther Flemming (1843—1905), German biologist, who discovered mitosis.
12
Wilhelm Roux (1850—1924), German
zoologist, early embryologist, known for
studying the outcome of provoked developmental defects.
13
August Weismann (1834—1914), German evolutionary biologist, known for
the germ plasm theory.
14
According to the theory genetic inheritance takes place via germ cells such as
egg and sperm cells. Other cells such as
somatic cells do not transmit hereditary
information to the next generation.
15
Walter Sutton (1877—1916), American
geneticist, who contributed to discover
that chromosomes inheritance follow
the laws of Mendel.
17
Gregor Mendel (1822—1984), German
scientist, who discovered fundamental
laws of heredity through his experiments on pea plants.
16
34 the periodic and dynamic structure of chromatin
Later, Thomas Morgan18 was the leading player in the discovery
of genes. As a result, he is considered as the founder of the field of
genetics. Along with Hugo De Vries19 and others, he proved that
chromosome regions bare information that is variable in different
organisms and can be transmitted to next generation. To achieve this,
he provoked mutations in flies using irradiations. After proving that
the provoked mutations can be inherited, he found that the transmission of certain mutations depends on the gender of the carrier and so,
on the presence of the X chromosome. Knowing that the presence of
genes on the X chromosome could be shown by this method, Morgan
performed appropriate cross-breeding. By examining the phenotypes
of the following generations, he could determine if two alleles recombine together more often than predicted by chance, which probably
means that the distance between them on the chromosome has an
influence on the probability to recombine. The further apart, more
likely the recombination of genes will be. This rational was the base
for designing experiments leading to a first genetic map, for the X
chromosome of Drosophila [Strutevant, 1913].
The earliest visualisation of functional sub-chromosomal structures
was done on pachytene stage chromosomes by August Weismann in
1913. Images pointed at prominent bulky regions, regularly spread
along the chromosomes, which are probably equivalent to chromatin
clusters recently observed in meiosis (see chapter 4 and [Prakash et al.,
2015]). Phoebus Levene 20 in 1929 characterised the composition of
DNA and found that DNA contained nucleic bases (already identified previously by Kossel), deoxyribose, and a phosphate group. It
was much later proven by Chargaff that adenine:thymine and cytosine:guanine ratios were 1:1 [Chargaff et al., 1952].
In 1944, one of the most important experiments in molecular biology, known as Avery—MacLeod—McCarty experiment [Avery et al.,
1944], was carried out. The experiment proved that DNA and not
proteins carry the information i.e. gene. Another major step was also
to show that one gene corresponds to one enzyme (see [Olins and
Olins, 2003] for a review).
1.3
Chromatin as a decision center of the cellular factory: the
golden age of molecular biology and electron microscopy (19441980)
The The fifties and the sixties experienced the emergence of molecular
biology and the electron microscopy (EM) as methods to explore the
way the nucleus responds to the environment and uses its internal
program to maintain the survival of the cell. Before the fifties, the description of chromatin was very shallow. On one hand, chemistry had
revealed the composition of the nucleus; on another hand, the optical
description of chromatin was limited to individual sub-structures,
such as the nucleolus or Cajal bodies. Genes had been described indirectly; polytene chromosomes of Drosophila had been characterised
Thomas Morgan (1866—1945), American embryologist, who discovered that
traits transmitted to the next generation
are coded by certain regions on chromosomes, the so-called genes.
18
Hugo De Vries (1848—1935), Dutch
botanist who introduced the concept of
genes and mutation, which he incorporated in an updated theory of evolution.
19
Phoebus Levene (1863—1940), American biochemist who characterized DNA
composition as a mixture of nucleic
bases, sugars and phosphate groups.
20
a condensed history of chromatin research
by autoradiography [Ficq and Pavan, 1957, Lewis, 1945]. Nevertheless,
beyond the level of basic chemical compounds and the genetic bands
of Drosophila, not much was known about the various regulatory
processes happening inside the nucleus, especially in the interphase.
Molecular biology has emerged in the middle of the 20th century as
an attempt to describe this complexity and bridge the gap between
basic chemical assays and visual description of chromatin. This will
be achieved mostly by experimental evidence but also by modelling.
Prompted by research and thoughts from physicians Niels Bohr21 and
Erwin Schrödinger22 , who saw that the basic laws of physics have
to be at the basis of all living forms, many biologists started to use
models to address new questions.
The most famous paper of the twentieth century in the field of
biology is the 1953 model of DNA by Watson and Crick published
in journal Nature ([Watson et al., 1953] and Figure 1.4). According
to him, James Watson23 was inspired by the reading of Schrödinger
essay "What is life" [Schrödinger, 1945] in choosing his post-doctoral
topic which was on the structure of DNA. In the early fifties, many
people, including Linus Pauling24 , who had already modelled the
alpha-helix of proteins [Pauling et al., 1951], came with the idea that
the DNA molecule could be modelled. An earlier hypothesis had
claimed that the structure could be a helix, which Pauling falsely
thought to be triple [Pauling and Corey, 1953]. Based on experimental
evidence from Franklin, Watson along with his collaborator Francis
Crick, designed a double helical structure for DNA [Watson et al.,
1953]. The model explained how the main components of DNA are
placed toward each other: two strands of DNA made of alternating
deoxyribose and phosphate, interact via nucleic bases bound to their
respective deoxyribose, similarly to the zip of a jacket. The idea of
base pairing came from brand new evidence from Erwin Chargaff,
who showed that adenine: thymine and cytosine : guanine are 1:1
ratios [Chargaff et al., 1952]. The fantastic model of Watson and Crick
revealed how a genetic vocabulary is possible by a combination of
nucleic bases and how it can be inherited, by duplication of available
information. It also showed how modelling is crucial to biology and
how the vision is as important as the data themselves.
This finding prompted many subsequent central questions of biology. At that time, where DNA was freshly known to bare information
in the cell, the issue that was slowly arising was how information
bared by DNA turns into proteins. The correspondence saying that
one gene corresponds to one enzyme had just been postulated [Beadle
and Tatum, 1941]. Following Watson and Crick publication, George
Gamov25 postulated that a minimum of 3 nucleotides is necessary to
code the 20 amino acids; later experiments by Crick showed that this
prediction was true [Crick et al., 1961].
35
Niels Bohr (1885—1962), Danish physicist, famous for a model of the atom
where electron can transit between orbitals, and for an early quantum theory
describing these transitions.
21
Erwin Schrödinger (1887—1961), Austrian physicist, made fundamental contributions to the field of quantum theory.
22
Figure 1.4: The famous model of the
double helix by Watson and Crick published in 1953 [Watson et al., 1953].
23
James Watson (1928—), American
molecular biologist and visionary, codiscoverer of the structure of DNA
with Francis Crick, co-discoverer of
mRNA and early promoter of the human genome project.
Linus Pauling (1901—1994), American
chemist who made several important
contribution to biochemistry, discovering the structure of the alpha-helix and
contributing to the study of genetic diseases.
24
George Gamow ( 1904— 1968), Russian theoretical physicist and cosmologist, who made some fundamental contributions to molecular genetics.
25
36 the periodic and dynamic structure of chromatin
Another set of experiments led to the discovery of messenger RNA,
mainly characterized by Watson and colleagues [Watson, 1963], based
on ideas of different discussion groups and seminal reflections by
Francois Jacob26 and Jacques Monod27 . These investigations led to
the conclusion that the nucleus has to be an administrative decision
centre of the cell, storing information that can be used to respond
to cues. A very remarkable ability of the nucleus is to adapt to the
signals from the environment.
The pioneer studies of Salvatore Luria28 , Alfred Hershey and Max
Delbrueck29 in the 40’s, established phage lambda and bacteria models as important platforms to study genes and their regulation. Until
the mid-20th century, how genes can be converted to proteins was
not understood. Experiments from Jacob and Monod helped to describe such a phenomenon. They showed the existence of activable
gene products, proving that genes respond to external cues. Upon
presence of sugar, they demonstrated that an intermediary product,
moving to the cytoplasm, was transmittable to another bacteria which
was not exposed to the initial clue. The famous "operon paper" also
hypothesised that these products are RNA [Jacob and Monod, 1961].
RNA polymerase itself was just discovered in 1960 [Hurwitz, 2005].
The first quantification of mRNA produced by the nucleus was done
in the group of James Watson, who had also been central in thinking
about possible intermediary elements to turn genes into proteins. The
earliest sketch drawn showing the transformation of DNA into RNA
to protein is from the hand of Watson, in 1959, according to him
[Watson, 2012].
Following pioneer works on DNA copying, Frederick Sanger30
developed a set of techniques to sequence DNA molecules, first
published on tRNA [Sanger, 1981]. This led to more efficient procedures, such as the classical Sanger sequencing protocol, which
uses dideoxynucleotides (ddNTP) to interrupt copying of the DNA
molecules to sequence at different positions.
Aside from the series of experiments that have led to the ’central
dogma’ of biology (genes transfer information to messenger RNA
which is used to build protein), the development of the electron microscope has been a major advancement of the century. Conceived to
record the differential transmission of an electron canon on a sample
preparation, this microscope allows viewing cells at the nanometer
resolution. Many compartments have been discovered using EM: the
endoplasmic reticulum, the Golgi apparatus, the chloroplast [Porter
et al., 1945] by Albert Claude31 . It has also helped to describe the
ultrastructure of the interphase nucleus and distinguish euchromatin
(’active chromatin’) from heterochromatin (’inactive chromatin’, see
Figure 1.5), and describe nuclear pores along with other nuclear structures (see Chapter 1 of [Busch, 1974] for more information). Most
importantly, it was crucial to proving that a nascent RNA comes out
of pericentric chromatin and that chromosomes in interphase have an
inactive region (central) and an active region (pericentric).
In situ hybridization is another area of research that has been deci-
Francois Jacob (1920—2013), French
biologist, along with Jacques Monod,
came up with the idea that intermediate
messengers are needed to make proteins
out of genes.
26
27
Jacques Monod (1910—1976), French
biologist, worked on genetic control of
viruses and enzymes together with Francois Jacob.
28
Salvador Luria (1912—1991), Italian
microbiologist, showed that mutations
in bacteria are random, and that selection acts on random mutations according to Darwin’s theory.
Max Delbrueck (1906—1981), a
German-American biophysicist, worked
on replication and genetics of viruses.
29
Figure 1.5: Electron microscopy image showing the classical dichotomy
between euchromatin and heterochromatin in a human leucocyte. Most of heterochromatin (black arrows) is in the periphery of the nucleus, for the most part
tightly attached to the lamina, while euchromatin (space pointed by the white
arrows) is filling most of the nucleus
space, without touching lamina. Image
courtesy: Wioleta Dudka.
30
Frederick Sanger( 1918— 2013), British
biochemist, known for his work on the
structure of proteins, the sequencing of
amino acids and for the determination
of nucleotide sequences in DNA.
31
Albert Claude (1899— 1983), Belgian
cell biologist, who was the first to use
electron microscopy for the study of
cells and description of various fundamental organelles.
a condensed history of chromatin research
sive for the description and regulation of a gene. The technique was
pioneered by Joseph Gall32 [Gall, 1968a,b], who developed a method
to purify rRNA molecules. Gall and Pardue then applied hybridization to the localisation of repeated elements on the chromosomes and
could visualise for the first time nanoscale chromatin features in situ
[Gall and Pardue, 1969]. This research shows an early attempt to
localise structural and functional pattern on chromosomes, though
banding on Drosophila chromosomes had already been shown since
long.
Electron microscopy, despite limitations due to the low density of
certain cellular components and limited assortment in contrasting
compounds, has been an immense advancement in biological research
during the 20th century. Radioactive probes used in EM preparation
have been slowly replaced by confocal imaging, which allows using
fluorescence to stain several types of molecules at the same time, without long, sophisticated EM procedures. Experiments in other areas
of research have also shown that the specificity of cells phenotypes
(i.e. their function) comes from specific genetic programs executed
in the nucleus. For instance, a pioneering experiment has been the
successful transplantation of a somatic nucleus to an embryonic stem
cell by John Gurdon33 , which resulted in a healthy production of
tadpoles. This proved that the phenotype of a cell depends on the
end products of the genetic program, that are mRNA and proteins
[Gurdon et al., 1958].
1.4
Chromatin as a highly structured system: genomic data, localisation methods and modelling (1980 onwards)
The research of the 60’s and 70’s had shown many basic principles
of the life of a nucleus, the ’administration centre’ of the cell. In
the following decades, research regarding chromatin function and
dynamics has started to take off. The central questions in the field of
research on the nucleus were: how many genes does it comprise, and
what are they doing? Does the chromatin adopt a given architecture
in the nucleus? Are chromosomes, transcription machinery, genes,
regulators disposed in a certain way or are things close to stochastic?
What about the dichotomy euchromatin/heterochromatin revealed by
electronic microscopy? Are there other levels of complexity between
the nucleosome level and chromatin dichotomy? Finally, can the
nuclear centre store information and change its behaviour using past
stimuli which are no longer present? This last question, the question
of epigenetics, will be treated in the next section.
Describing the diversity of genes present in a nucleus was certainly
not a trivial task. Since Sanger first experiments, sequencing was
possible, but at a tiny scale. The first automatic sequencer appeared
in 1985, with a speed of 3000 nucleotides per day. Knowing that
the genome comprises possibly 100,000 genes, with several kilobase
pairs long in each case, researchers had to face both an experimental
and a computational challenge. Computation would require a high
37
Joseph Gall (1928— ), American cell
biologist, known for developing in situ
hybridization and work on repeated genomic regions such as telomeres.
32
John Gurdon (1933— ), a British biologist, the pioneer in the field of cellular
reprogramming.
33
38 the periodic and dynamic structure of chromatin
number of very fast processors and was not possible at a decent speed
before the 1990’s. The first competitive algorithm for high-throughput
DNA sequencing, based on graph theory and so on finding the best
path to assembling the sequences generated by enzymatic digestion
was developed in the early eighties [Kozak, 1984]. Most importantly,
the existence of a vast number of repetitive regions will have to be
tackled. A large public consortium started to emerge, on the initiative
of several prominent scientists such as James Watson. In the late
phase of the project, the public consortium was challenged by an
independent project generated by Craig Venter34 . The two projects
published the genome draft by one day of interval [Lander et al.,
2001, Venter et al., 2001]. The quest was both a biological and a
human milestone, revealing the necessity of community effort to
produce genome-wide data, a goal achievable in both academics and
entrepreneurial environment.
Many insights came from the human genome sequence: a surprisingly low amount of genes (20,000), the prominence of repeated
sequences, and the significant number of horizontal transfer events
from various organisms, including bacteria. Aside from getting a
catalogue of genes, non-coding RNA and repeated elements, information from genomes has also allowed testing many hypotheses at
genome-scale, such as identification of genetic mutations, the design
of probes to further track sequences or performing site-specific mutations. Following Joseph Gall’s pioneer method, fluorescent probes
were designed to target specific genes (Fluorescence in situ hybridization or FISH) [Langer-Safer et al., 1982, Schardin et al., 1985, Lengauer
et al., 1992]. FISH has been extensively used to study the regulation of
genes in the genome, at given loci. For instance, interaction between
genes has been imaged with this technique (’gene kissing’, [Cremer
and Cremer, 2001, 2009b,a]). Important applications of FISH include
the technique of chromosome painting, which aims at using enough
probes to be able to see most of the spatial occupancy of a given
chromosome.
Today, it is no doubt that interphase chromatin is a very complex
structure that has a relationship with the structure of chromosome
during metaphase. For instance, the genes remain in the same order
along the cell cycle. Nevertheless, during the early days of chromatin
research, it was thought that chromosomes become fragmented at
the end of the mitosis and clusters together to reform chromosomes
at the end of the interphase. Two scientists explored the alternative
hypothesis, Rabl and Boveri, and guessed that chromosomes kept
defined positions during interphase, though in a more sophisticated
manner as during mitosis [Boveri, 1888]. This representation of what
are now called "chromosome territories" remained marginal for a
long time. When the nucleus architecture was better studied using
higher levels of magnification, with the advancement of the electron
microscopy, interphase nucleus began to be described in terms of condensed and relaxed chromatin, with the idea that the chromosomes
where highly untangled in the nucleus, and that their precise indi-
John Craig Venter (1946— ), American biochemist and geneticist, known
for contributing to the first sequencing
of the human genome.
34
a condensed history of chromatin research
vidual localisation did not matter. Some isolated authors described
possible models looking closer to the modern view on chromosome
territories [Pollister, 1952], but the dichotomy hetero/euchromatin
prevailed for long. It is only during the late seventies that the idea
that interphase chromosomes occupy certain positions or territories in
3D space started to emerge. The concept was in the air but put aside
due to experimental limitations for some decades [Comings, 1980].
The first demonstration of the existence of chromosome territories
was shown by [Stack et al., 1977] using specific chemical treatments, to
make the DNA mass appear wherever it is located in the nucleus. At
the same time, other regions of the nucleus became characterised, such
as the perichromatin region, shown to be the zone experiencing most
of the transcription, splicing and eventually repair events [Cremer
et al., 1982a,b]. Recently, more complex models have been proposed to
describe the region between the chromosome territories (described in
[Fakan, 2004, Cremer et al., 2004]), some hypothesising for separation
of chromosomes, other for intermingling [Branco and Pombo, 2006].
These interchromosomal regions can potentially recruit genes from
different chromosomes to transcription machineries [Osborne et al.,
2004, Chakalova et al., 2005, Lanctôt et al., 2007, Cremer and Cremer,
2009a,b], while some have shown that the number of transcription
factories may be very restricted in adult cells (factories are reviewed
in [Sutherland and Bickmore, 2009]).
Several levels of chromatin have been modelled in textbook pictures following EM studies (see [Uhlmann, 2014, Ozer et al., 2015]
for a summary). Some regions, ’clumps of chromatin’, presented in
[Monneron and Bernhard, 1969] and visible in other studies, are
regions of loose chromatin or euchromatin, though their apparent
localisation is only related to the concept of chromosome territory
since recently [Albiez et al., 2006]. Indeed, for a long time, apart from
the euchromatin/heterochromatin dichotomy, high-order chromatin
levels have not been considered. Towards the description of potential
high-order structures during interphase, in situ hybridization of entire
interphase chromosomes is characterised in [Schardin et al., 1985].
Though chromosome territories are established, and active regions
either messy [Branco and Pombo, 2006] or confined to specific regions
[Cremer and Cremer, 2001], trans-associations between chromosomes
are now also known to happen [Spilianakis et al., 2005, Lomvardas
et al., 2006].
Increasing complexity of the nucleus has been captured recently
using techniques cataloguing interactions between DNA sequences
(chromosome conformation capture [Dekker et al., 2002]). Data have
shown interchromosomal interactions to happen preferentially in a
short range distance, though the two telomeres of a chromosome may
be interacting during interphase [Dekker et al., 2002]. Interphase
subchromosomal domains (topologically associating domains, chromatin loops, etc.) have emerged progressively as an intermediate
between the beads-on-a-string level (see next section on epigenetics)
and the chromosome territory. Ma et al. [Ma et al., 1998] attempted to
39
40 the periodic and dynamic structure of chromatin
measure these intermediary levels of complexity via optics before the
advent of genomics, hinting for chromosome domains in the range
of 1Mb. HiC data [Rao et al., 2014, Lieberman-Aiden et al., 2009]
recently helped to refine the picture and describe more complexities
in the chromosome territory regions. It only recently occurred to
scientists that intermediate levels also exist. Before HiC and other
techniques, territories of 1Mb were thought to exist, but only HiC has
allowed estimating the distribution of contacts between regions. A
more complex picture has come from data showing dynamics of contacts during different phases of the cell cycle [Naumova et al., 2013].
Having achieved wide success in generating significant amounts of
genomic data, genomics is awaiting validation from microscopy.
Benefiting from advancements in the area of fluorescence microscopy, which include the discovery of properties of the green
fluorescent protein [Prasher et al., 1992], and new theoretical frameworks (see [Betzig, 1995] and chapter 2 of this work), recent improvements in light microscopy [Hell and Wichmann, 1994, Heintzmann
and Cremer, 1999, Gustafsson, 2000, Lidke et al., 2005, Betzig et al.,
2006, Rust et al., 2006, Hess et al., 2006, Lemmer et al., 2008, Reymann
et al., 2008] have allowed studying the cell nucleus at an unprecedented resolution. These new techniques have been extensively used
by Thomas and Christoph Cremer to study chromatin patterns in high
resolution (Figure 1.6), leading to a drastic improvement of chromatin
description. Thank to them, chromosome territories are now well
described, and a picture of chromatin structure with fundamental
functional building blocks of chromatin of about 100-400 nm starts to
emerge. These so-called chromatin domains are either associated with
high or low transcription, depending on their low or high compaction
state, respectively (see section 3.2.3. on this report, [Prakash et al.,
2015, Boettiger et al., 2016]). In case of embryonic cells, these methods
have shown that the number of transcription units is much higher
than in differentiated cells (i.e. specialized tissues), but surprisingly
each of the units contained only one or two molecules (see myoblasts
image, Figure 2 in [Smeets et al., 2014]), while in non-specialized cells,
the machineries in lower number have a broader range of genes to
transcribe.
Finally, the dynamics of chromatin during interphase is emerging
as a new focus for modelling biological processes. Stimulus-induced
chromatin changes have been most of the work on dynamics of chromatin in the pre-genomics area [Jacob and Monod, 1961]. Stimulation
by UV was shown to decrease chromatin extractability, or in more
modern words, accessibility [Smith, 1962]. UV has also shown to
induce changes in chromatin localization [Cremer et al., 1982a,b].
Movements of chromatin as a possible way to regulate genes has
emerged during the seventies, with concepts such as nuclear matrix35 [Berezney and Coffey, 1976]. Understanding the dynamics of
nuclear architecture during development has progressed a lot using
band painting, to monitor the transition between interphase and mitosis and find relationship between defined chromosome positions
Figure 1.6: Comparison between recently developed superresolution microscopy techniques and conventional
light microscopy imaging. Chromatin
domains are visible on the SMLM
only; staining was performed using Vybrant DyeCycle Violet DNA dye [ŻurekBiesiada et al., 2015].
Nuclear matrix: nuclear cytoskeleton
thought to be involved in chromatin dynamics.
35
a condensed history of chromatin research
41
in mitosis and the distribution of chromosomes in 3D space during
interphase. Contraction and relaxation of chromatin has also captured
attention to study the relationship between structure of chromatin and
regulation of gene expression [Shilatifard, 2006]. Recently, application
of FRAP has helped to study chromatin dynamics happening in the
nucleus (see original method in [Axelrod et al., 1976])
1.5
The substratum of chromatin memory: epigenetic regulation
The works presented in the previous section have shown that the
nucleus contains DNA molecules baring information through a vocabulary of genes, coding for proteins and that this information can
be used to respond to external clues. This was a major focus of early
molecular biology. In the following decades, up to now, the study
of genes and genomes has continuously occupied a prominent place.
Nevertheless, researchers have become increasingly aware that genes
are not all transcribed at the same time and that the genome needs
to modulate their response in space and time. On top of this, it
has become more explicit that the spatial disposition of chromatin
within the nucleus is essential to modulate the response of the cell
to environmental cues. This modulation will be shown to happen
through modification of bases of the DNA as well as histone modifications, between other mechanisms summarised under the epithet of
epigenetics.
What remains from Jean-Baptiste Lamarck36 is probably the transmission of information between cells during development, rather
than between organisms. Nevertheless, the first statement that can be
recognized as an indirect reference to a potential epigenetic regulation of chromatin in a modern sense is the hierarchy of cells during
development: "During each cell division on the way from a fertilized
egg to an entire organism, each daughter cells obtains one-half of
the idioplasm according to mass, but not necessarily according to its
quality" [Weismann, 1892, Cremer and Cremer, 2009b]. This is close
to the definition of epigenetics. The importance of epigenetics for
transcription was taken into account very early in the development of
molecular biology. Since the discovery of chromatin and first characterisation of its composition in the nineteenth century, a community
of researchers was specifically devoted to study nuclear proteins and
their biological importance. Biochemists were nevertheless quite separate from molecular biologists, the latter more focused on high-order
processes, while biochemists and chemists by definition focused on
the lower scales, on entities such as nucleic acids or enzymes [Hurwitz, 2005]. The composition of histone became characterised in great
details by biochemists [Stedman and Stedman, 1951]. They helped
to show that histones can inhibit the work of RNA polymerases, and
hypothesised that they have a role in compaction.
Biochemical analysis had shown that histones can be acetylated or
methylated, an observation relatively easy to obtain given the small
size of the proteins. A relationship between this post-translational
Figure 1.7: The beads on a string: in
vitro visualization of nucleosomes punctating DNA. Scale bar: 30 nm. [Olins
and Olins, 2003].
36
Jean-Baptiste Lamarck (1744—1829),
French naturalist and biologist, most
known for his ideas on evolution and
inheritance of acquired characteristics.
42 the periodic and dynamic structure of chromatin
modification (both acetylation and methylation) and gene transcription was described very soon after the theorization of messenger
RNA [Allfrey et al., 1964]. Allfrey and colleagues showed that upon
mild acetylation, histone proteins lose their capacity to inhibit the
work of RNA polymerase. In the seventies, the first EM images of
nucleosomes were published and revealed the nanoscale structure of
chromatin (Figure 1.7).
Histone modifications and DNA methylation were later shown
to recruit factors. In this last case, the first proof of DNA methylation binding protein was presented by [Meehan et al., 1989], a study
showing for the first time a protein binding methylated CG dinucleotides. Another important notion of chromatin architecture and its
accessibility stems in the historical dichotomy between euchromatin
and heterochromatin. More recent works have proven that epigenetic
marks (acetylations or methylations) are not evenly distributed in
the chromatin and that one can find marks associated with regions
visually more condensed and others to regions more relaxed. Findings from recent superresolution experiments have helped to describe
epigenetic patterns in interphase for active and inactive X chromosomes, showing that clusters of activating and inactivating histone
modifications anti-correlate [Zinner et al., 2007, Cremer et al., 2015].
These findings have confirmed patterns already found previously in
mitosis or using assays such as ChIP-seq [Barski et al., 2007]. What
remain to be confirmed are the domains and loops detected with HiC
data [Rao et al., 2014]. Imaging will aim at confirming the reality of
these loops and domains and see if more patterns can be found using
pure visual methods. Work from [Chandra et al., 2012] and more
recently [Boettiger et al., 2016] going in this direction have shown
that functional chromatin domains can either display a condensed or
relaxed configuration, with different epigenetic signatures associated
in each case. An overview of various genomic methods to study
chromatin at different length scales is shown in Figure 1.8.
As the ROM (Random Access Memory) of chromatin, epigenetic
regulation is right now one of the hottest topic in biology. While the
mechanisms that guide a resting cell have been extensively studied,
the way the chromatin architecture adapts to its environment is very
little studied. For instance, how long a stress can be recorded by a
cell? Does chromatin learn from previous cues to adapt better to the
next situation? Moreover, how chromatin architecture changes from
mother to daughter cell during development, based on surrounding
environment?
1.6
Fine-scale chromatin architecture: a new modelling area
In the last years, a new way of analysing chromatin has emerged,
following data from in situ hybridization and HiC technique. Biophysicists have started to find patterns and trends in available data
to describe possible orders of chromatin that have escaped attention
until now. Chromatin is now being viewed as a self-organizing struc-
a condensed history of chromatin research
43
Figure 1.8: Different genomic methods
to study chromatin. The ever expanding range of next-generation sequencing technologies is portrayed. A number of modern sequencing approaches
have opened new possibilities to analyse epigenetics at different organisation
levels and study various gene regulatory mechanisms. Furthermore, these
methods offer numerous opportunities
for super-resolution microscopy. Image
courtesy: Wolf Gebhardt.
ture [Misteli, 2001, Karsenti, 2008, Rajapakse and Groudine, 2011].
Contrary to an object which just folds according to pure Brownian
motion and finds its structure by the random association of its separate parts, a self-organised object requires the external intervention
of compounds (for instance enzymes) that constrain its parts into
adopting a structure not formed by chance. In the case of DNA, these
elements could be either transcription factors, epigenetic modifications or their interactors. The key is then to get an overview of the
different actors, how they interact with DNA to modify its shape.
Though still in its theoretical stage, this research has brought some
insights in the recent years.
A recent study defines chromatin states as ’binders’ (such as CTCF)
and ’strings’ (DNA molecules) [Barbieri et al., 2012]. In this configuration, binders are in solution and find their targets driven by
Brownian movements. Ideally, these binders will target DNA at one
unique location. Nevertheless, if a binding site is present at different
places in relative close locations on a chromosome, DNA may bind
several of them together and form a cluster. The concept matches HiC
data, and can explain several known complexity levels of chromatin
architecture (chromatin loops, chromatin domains, chromosome territories). Tom Misteli has recently proposed a genome-wide imaging,
HIPmap; that is currently being set up to validate HiC contact maps
with microscopy [Shachar et al., 2015]. Finally, some recent models
have also attempted to study the dynamics of chromatin. For instance,
a model by [Sanborn et al., 2015] describes activation of genes as loops
forming through cohesin rings and popping out of chromosome territories to enter a compartment containing transcription machineries
(Figure 1.9).
The existence of the higher level of chromatin compaction, the chromosome territory, has long been oblivious to scientists. Modelling
Figure 1.9: Comparison between predicted data and experimental data in the
genomic era. The experimental results
(bottom panel) were generated using
HiC method and show localization of
contacts between chromosome regions,
an information important to model the
3D structure of the chromatin. The
model in the first panel predicts contacts based on dynamic modelling of
chromatin folding. In both cases, the
red dots show the putative contacts. Image modified from [Sanborn et al., 2015].
44 the periodic and dynamic structure of chromatin
helps to integrate data and see to which kinds of architecture the
distributions of contacts could fit [Barbieri et al., 2012]. Recent models
show a new creativity emerging among biologists to describe chromatin accurately, with the help of concepts from external disciplines
such as polymer physics. A lesson from this research is that morphology and structural organisation of mitotic chromosomes might be
structurally related to interphasic chromosome territories. One can
speculate that there are even more analogies and that studying either
mitotic (or meiotic, as I will discuss later) or interphase stage of the
cell cycle will help to get information about the other stages of the
cell cycle. In any case, a strong validation of proposed models, both
with genomics and microscopy, is urgently needed.
1.7
Conclusion
The history of chromatin research provides numerous lessons for
research. One of them is the necessity to put forward theories, even if
they prove later wrong. Another lesson is obviously the need to put
forward the development of new technologies. Though paradoxically,
often, new technical insights come from new questions. For instance,
how Morgan decided to study DNA to learn about evolution was a
beautiful example of perfect experimental design focused on answering a big theoretical question. Lastly, from Watson and Crick to recent
insights in the fields of genome architecture, modelling has emerged
as a necessity to describe properly chromatin structure and behaviour.
Direct observation is only useful if it is linked to genomic data, and
as such requires the help of physical models to make sense of the
complex available information.
Chromatin has been explored using various methods (light microscopy, biochemistry, histochemistry, genomics) but the usage of
these methods was blinded by the fact that scientists thought that
investigation of interphase chromatin was a dead end, that there was
nothing to learn beyond nucleosomes and euchromatin/heterochromatin
dichotomy. Later stages of research have shown that the picture was
more complex. Firstly, in situ hybridization experiments have shown
repeated regions, meaning that there are non-coding regions between
genes. Chromosome territories have then revealed a compaction
at the high order in the nucleus, which was no longer considered
as a spaghetti dish. Moreover, recently, information coming from
next-generation sequencing, chromosome capture and hybridization
methods have shown even more complexity. New concepts emerge,
such as chromatin loops, chromatin domains, which were unknown
before. This research is relatively new, and despite thinking of pioneers [Cremer and Cremer, 2001, 2009a,b], one has to understand that
the idea that chromosome functionalities are compartmentalised in
the interphase nucleus is relatively new to biological research.
The questions that challenge the researchers now are: how much
of the complexity is still to know? Are there functional domains that
one is not aware of yet? And can single-molecule imaging help us in
a condensed history of chromatin research
this task? This is the question that I have tried to answer in the work
presented here.
1.8
Acknowledgement
This chapter is an outcome of many discussions with David Fournier,
who has helped immensely in formulating and revising the text.
45
“My intention is not to replace one set of general rules by another such set: my intention
is, rather, to convince the reader that all methodologies, even the most obvious ones, have
their limits.”
Paul Karl Feyerabend
2 Investigating Chromatin Organisation using Single Molecule Localisation Microscopy
2.1
Introduction
In this chapter1 , I discuss the technical details of single molecule
localisation microscopy (SMLM) to investigate spatial and temporal
organisation of DNA. The DNA is hierarchically folded at multiple
levels to become more compacted and functionally organise itself
inside of the nucleus. This spatial arrangement in turn affects the
functionality of DNA. Thus one can characterise the organisation of
chromatin into three inter-related categories: (1) the basic building
blocks, (2) the functional organisation of chromatin and (3) the spatial
arrangement of chromatin inside the cell nucleus.
Parts of this chapter, including figures
and captions, are based on the following publications: [Szczurek et al., 2014,
Żurek-Biesiada et al., 2015, 2016, Hagmann et al., 2014, Best et al., 2014]
1
Figure 2.1: Chromatin spatial organisation: The spatial organisation of chromatin can be studied at three levels: at
the lowest orders, which include the
beads-on-a-string model and the chromatin fiber, at the level of functional
chromatin domains and at high order
chromatin patterns. The functional organisation of chromatin domains can
be modeled from perspective of various post-translational histone modifications. Depending upon the kind of
modification it is enriched with, chromatin can be either highly condensed,
in an open conformation or switch between these two extreme forms. The
spatial arrangement of chromatin can
also be viewed from three points of
views: the chromosome territory model
of DNA organisation, the bimodal classification condensed and open chromatin
in hetero- and eu-chromatin, and finally
the highly condensed configuration during metaphase.
The two classical building blocks: beads-on-a-string and 30 nm
chromatin fibre have been extensively studied by EM (Figure 2.1, first
column). Furthermore, one can view functional chromatin domains
(Figure 2.1, second column) as an emerging building block of chromatin, responsible for its basic functions. I estimate these domains to
48 the periodic and dynamic structure of chromatin
be about 100-400 nm, according to the literature and my own work
(see Chapter 3 on interphase chromatin and [Prakash et al., 2015, Boettiger et al., 2016, Rao et al., 2014] for more details). These domains
are usually the place of chromatin regulation and can display several
configurations, depending on their compaction (Figure 2.1, second
column). These compaction states are often associated with different
kinds of post-translational histone modifications and can be characterised using information from these modifications [Prakash et al.,
2015, Boettiger et al., 2016]. Domains can either be highly condensed
(usually baring H3K9me3 modification, a mark associated to the
presence of transcriptionally inactive repeated regions) or in an open
conformation (mostly at gene promoters and intergenic regions, showing enrichment for H3K4me3, a mark associated to active chromatin).
Moreover, chromatin can also be in a state where it switches between
these two extreme forms as these domains can also coexist very close
to each other (see [Chandra et al., 2012]; for an overview of the different configurations of chromatin domains using microscopy and
genomic data, see Figure 3.16). Higher orders of chromatin folding,
above the chromatin domains, include: the rather outdated bimodal
classification of chromatin into condensed and open regions (heterochromatin and euchromatin), the chromosome territory model of
DNA organisation [Cremer and Cremer, 2001] and finally the highly
condensed configuration of the metaphase chromosome (Figure 2.1,
last column).
As chromatin has a highly complex shape, theoretical models are
limited to describe its functional structure. Moreover, the organisation
of its basic building blocks, the functional and spatial domains, lies
between what can be achieved by conventional light microscopy
(LM) and electron microscopy (EM). Although EM has provided
considerable insights into the structure and organisation of chromatin,
as it is not DNA specific, it fails at capturing the underlying nature
of the patterns observed. Moreover, sample preparation for EM
includes harsh chemical cross-linking and vacuum treatment, which
also influence the sample. Light microscopy techniques can be used to
study protein and chromatin function in live cells but these methods
suffer from much lower resolution than EM.
Super-resolution light microscopy provides an interesting alternative to EM that fills the gap between the resolution achieved by a
conventional microscope [Abbe, 1873, Rayleigh, 1896] and the one
required to resolve the individual chromatin domains. The first attempts to improve the resolution used confocal [Cremer and Cremer,
1978, Sheppard and Wilson, 1981] and multiphoton microscopy [Zipfel
et al., 2003], allowing for effective background suppression. The advent of various super-resolution methods [Hell and Wichmann, 1994,
Heintzmann and Cremer, 1999, Gustafsson, 2000], in particular, Single
Molecule Localisation Microscopy (SMLM) based methods [Betzig
et al., 2006, Hess et al., 2006, Rust et al., 2006, Lemmer et al., 2008] have
allowed to study objects whose structure lies below the diffraction
limit.
investigating chromatin organisation using single molecule localisation microscopy
2.2
2.2.1
49
Single-molecule localization microscopy: state-of-the-art
Principle of SMLM
The underlying principle of most SMLM based methods is to label
proteins or DNA/RNA with fluorescent molecules that can reversibly
switch between a fluorescent state and a stable dark state. This
process of switching between states is called ’blinking’, and allows
for optical isolation of single molecules. Since only a fraction of
molecules will switch back to the fluorescent state at a given time,
their precise location can be determined. The final accuracy of the
position of a molecule depends on the number of detected photons.
With a photon count of 104 in a single glowing phase, one can achieve
a resolution down to a few nanometers [Thompson et al., 2002]. In
Figure 2.2, many such molecules are separated by a distance less
than λ/(2N A) and if all of them are ’on’ simultaneously then one
cannot resolve them. However, if only a few of them are ’on’ while
the neighbouring molecules are in dark state, they can be optically
isolated and their precise position can be determined (see Figure 2.2).
Subsequently, if one collects a series of such images where only a few
of the molecules are ’on’ in each image and localise their position,
then a final image revealing the structure of the underlying object can
be reconstructed [Small and Parthasarathy, 2014].
2.2.2
The different SMLM methods: a historical perspective
Optical isolation of individual molecules in order to circumvent the
resolution limit by acquiring signals at different times has been first
theorized in 1985 [Burns et al., 1985], and later included in a comprehensive framework by Eric Betzig of what is now known under name
single molecule localisation microscopy [Betzig, 1995]. Historically,
SMLM was first used outside of the chromatin field. Early design
by Betzig was successfully applied in 2006 [Betzig et al., 2006] to
study cell architecture in a set-up called Photo-activated localization
microscopy (PALM). PALM uses two lasers to provoke the blinking
of GFP molecules necessary for single molecule microscopy. The first
laser (561 nm ) is used as the excitation laser while the second laser
(405 nm) is used as the activation laser (see [Sengupta et al., 2014] for
a review). The structures imaged showed similar level of resolution
as electron microscopy, but helped to mark very precisely the position
Figure 2.2: Schematic illustration of the
underlying principle of SMLM using
a hypothetical object whose building
blocks lie below the diffraction limit.
This object is made of point sources indicated by black circles. Photons emitted by these point sources such as biological molecules or fluorophores are
smeared out onto the detector of the
imaging system due of the wave nature of light. This distribution of photons is commonly known as the point
spread function (PSF) of the imaging
system (indicated by the shadow in red).
PSF has a width of λ/(2N A), where
λ is the wavelength of the fluorescent
light and NA is the numerical aperture. Repeated imaging and localisation of indiviual fluorophores provides
high resolution image of the diffraction
limited object if one approximates the
position of the molecule as the center of
the spot. The photons in the airy disk,
which is the central region of the PSF,
are thought to have a repartition that
follows a Gaussian distribution if the
point source emits a sufficient number
of photons [Sengupta et al., 2014].
50 the periodic and dynamic structure of chromatin
of given proteins on top of the EM image. Set-ups such as STORM
(STochastic Optical Recovery Microscopy) use Cy5-Cy3 dye pair as
a switch to optically isolate molecules whose separation is below
Abbe’s limit. Firstly, all Cy5 molecules are pushed to dark state with
a red laser (633 nm) and then a green laser (532 nm) is used to bring
a fraction of Cy5 molecules to the florescent state. Though there is
no fluorescence from Cy3 itself, a close presence to Cy5 is required
for Cy5 to recover from the dark state to a fluorescent emitting state
[Rust et al., 2006]. dSTORM (direct STORM), a method which extends
early observations by [Rust et al., 2006, Lemmer et al., 2008] uses only
one laser to push a conventional fluorochrome, for instance Alexa647,
between dark and fluorescent state [Heilemann et al., 2008]. Below,
I briefly summarize various approaches that led to development of
SMLM in its current working form.
• First theoretical approaches: [Burns et al., 1985, Betzig, 1995]
• First practical approaches:
1. Blinking of single molecules [Hirschfeld, 1976, Moerner and
Kador, 1989, Shera et al., 1990, Xie et al., 1994]
2. Blinking of GFP [Dickson et al., 1997]
3. Blinking of quantum dots [Lidke et al., 2005]
4. (fluorescence) Photo-activated localization microscopy (PALM/fPALM,
[Betzig et al., 2006, Hess et al., 2006])
5. Stochastic Optical Reconstruction Microscopy (STORM, [Rust
et al., 2006])
• Methods using standard fluorophores:
1. Spectral Precision/Position Distance/Determination Microscopy
(SPDM, [Van Oijen et al., 1998, Lemmer et al., 2008, 2009])
2. direct Stochastic Optical Reconstruction Microscopy (dSTORM,
[Heilemann et al., 2008])
3. Bleaching/blinking assisted localization microscopy (BaLM,
[Burnette et al., 2011])
• Methods based on binding activation and kinetics: Binding-activated
localization microscopy (BALM, [Schoen et al., 2011])
Herein, I employed SPDM, the method initiated by [Lemmer et al.,
2009] and now known as dSTORM [Heilemann et al., 2008]. This
method uses a single laser to induce both photoconversion and blinking of a single fluorochrome in one excitation event. This technique
was helpful in obtaining a detailed distribution of chromatin, various protein elements and histone modifications, in interphase nuclei
(Chapter 3) and during meiosis (Chapter 4). This simple set-up not
only enables the analysis of biological samples with the fluorophores
used to image DNA molecules in conventional microscopy, but also
allows for multi-modal imaging, which can then be used to compare
investigating chromatin organisation using single molecule localisation microscopy
51
or combine with the results obtained from other microscopic techniques. In this thesis, I discuss various technical aspects of SMLM
applied to standard DNA dyes to study chromatin organisation.
2.3
Application of SMLM to image chromatin
Chromatin has been a rather recent focus in the field of SMLM. The
nano-structure of interphase chromatin was first studied by [Bohn
et al., 2010] using SMLM. Statistical methods were combined with
SMLM to study the distribution of histone H2B in HeLa and VH7
diploid human fibroblast cells. The main question posed by this study
is the possibility of the existence of a recurrent universal chromatin
nano-domain and if it can vary across the cell lines and the cell cycle.
The author reported that chromatin nano-structure is cell type specific
and is dependent on the way the chromatin is labelled. A similar
conclusion was reached by a recent study [Ricci et al., 2015]. Before I
probe further into the limitations of these two studies, I would like
to discuss two concepts that are central to localisation microscopy:
localization precision and signal density.
2.3.1
The tao of SMLM
Two key aspects of SMLM imaging are localisation precision and
signal density. Figure 2.3 shows the interplay between these factors.
While the bright fluorophores i.e. fluorophores with high photon
count are good for localization precision (∝ √1 , where N is the
N
number of photons), it is highly desired to have only a subset of
them ’on’ in each frame (amount of localized events per unit area).
To achieve low signal density for each frame, the lifetime of the dark
state needs to be significantly longer than the lifetime of the bright
state. The overall high signal density (amount of localized events per
unit area) is achieved by acquiring a large number of frame so that
the theoretical density of the underlying sample is matched.
2.3.2
Importance of a good localization precision in order to improve
resolution
The resolution of a final reconstructed image in localisation microscopy can be attributed to two separate causes. The first is the fact
that like any measurement of a physical quantity, the true value to be
measured is only approached at an infinite number of measurements.
Since the photon count of the localization of a fluorophore is limited,
the precision with which the fluorophore location can be predicted
is limited too. In this analogy a single measurement corresponds
to telling which pixel on the light sensitive array was struck by one
single photon. Only with the ensemble comprising many photons
spread over the adjacent pixels, a highly precise location of the signal can be determined. The precision can be calculated using the
Cramér-Rao Lower Bound [Neice, 2010, Small and Parthasarathy,
Figure 2.3: The tao of single molecule
microscopy: The two key aspects around
which localisation microscopy revolves
are localisation precision and signal density. The true representation of the underlying structure is reached when localisation precision and signal density are
sufficiently high. Bright fluorophores
are required for good localisation precision while for a good signal density,
it required to have only a few of them
in each frame which is compensated by
acquring a large number of frames.
52 the periodic and dynamic structure of chromatin
√
2014] and is always above λ/(2NA · N ), with λ the wavelength of
the fluorescent light, NA the numerical aperture [Hagmann et al.,
2016] and N the number of photons.
The second cause for resolution reduction results directly from the
first (the measurement uncertainty): The fact that the uncertainty of
a localized event is isotropic makes it literally impossible to tell in
which direction the measured position has to be displaced in order
to set it to the true location of the fluorophore [Hagmann et al.,
2016]. It is important to be aware that the measured position of a
fluorophore is not its true position but only the one with the highest
probability. The quality of this position estimate is reflected by the
localization precision (or uncertainty). In other words, if the same
measurement would be carried out several times under the exact
same conditions, the acquired locations would be slightly displaced
for each measurement. The scale of displacement for every single
fluorophore is reflected by its individual localization precision, given
by
s2 + a2 /12 8πs4 b2
σ2 =
+ 2 2
N
a N
where s is the width of PSF, N is the number of detected photons, b
is the background intensity and a is the size of pixels on the camera.
Localisation uncertainty 2 nm Localisation uncertainty 5 nm
Localisation uncertainty 10nm
Localisation uncertainty 20 nm Localisation uncertainty 50 nm Localisation uncertainty 100 nm
In Figure 2.4, I demonstrate that with an increasingly poor localisation precision (from 2 nm to 100 nm), the point-like-object start to
overlap resulting in a blurred image where fine features cannot be
resolved. Another point worth noting is that while the first image
has the lowest uncertainty (2 nm), the images with 10 nm and 20
nm uncertainty appear the best. This is due to the fact that higher
uncertainty covers for the missing signals by blurring with a bigger
radius (see Figure 2.5)
Figure 2.4: Importance of good localization precision: The measured position
of a fluorophore is not its true position
but only the one with the highest probability. As the photon budget of a fluorophore is limited, so is the precision
with which its position can be localised.
In this simulation, I compared the effect
of increasing the localisation uncertainty
gradually from 2 nm to 100 nm.
investigating chromatin organisation using single molecule localisation microscopy
2.3.3
53
Importance of high signal density to improve signal-to-noise ratio
One of the key to SMLM is its ability to provide structural information
at the highest possible resolution among various super-resolution
methods. The structural resolution in localisation microscopy depends
upon the localisation precision σ̄xy of the individual molecules and
the local density of the detected molecules is given by the following
formula:
σxy =
q
(2.35σ̄xy )2 + (2d¯NN )2
σ̄xy is the mean lateral localisation accuracy;
d¯NN is the mean distance to the next neighbouring molecules.
1000 points
10000 points
20000 points
25000 points
50000 points
120000 points
The final number of localised signals depends on a number of factors such as labelling efficiency, activation and re-activation frequency
and detection efficiency. For example, only a certain percentage of
molecules gets labelled and only a subset of these molecules get activated and detected. Furthermore, even a smaller subset of these
remaining molecules get reactivated again. This means that only
a small subset of the molecules forming the original structure get
activated, detected and localised, making it hard to distinguish signals from background. In Figure 2.5, with 120000 points one can
resolve fine features of Erika’s lips (yellow box), while 1000 points are
insufficient to draw any satisfactory conclusions. Thus a minimum
1
critical density of signals of at least 10
of the original structure is often
required especially if the underlying structure is unknown, a common
case for most super-resolution methods. If this is not possible then an
intelligent estimate must be made from the prior knowledge based
on other studies (for example, EM).
Figure 2.5: Importance of high signal density: In localisation microscopy,
molecules emit fluorescent bursts multiple times before they are bleached (i.e.
they stay permanently in dark state),
so I use the term ’signals’ rather than
’molecules’ here. In all 6 images, the signals were sampled with a localisation
uncertainty of 10 nm and the number of
signals were gradually increased from
1000 points to 120000 points for a 25 µm2
area. A minimum signal density (signal
density means the amount of localized
events per area) is required to make a
first assessment about the underlying
structure.
54 the periodic and dynamic structure of chromatin
2.3.4
Limitations of previous approaches to study chromatin organisation
Human genome has roughly 109 base pairs and a spherical nucleus
( 43 πr3 ) of 10 µm radius will have an overall volume of approximately
4000 µm3 . Since the cells in culture are relatively flat, one can assume
that they show an ellipsoid shape ( 43 πr2 a) with the axial thickness
(a = 2 − 3µm) smaller than the lateral radius (r = 10µm ). This
would bring down the nuclear volume to roughly 1000 µm3 . Since the
observation volume is only a fraction of the total volume, I assume
that the final volume of the nuclei imaged would be around 200
µm3 with an optical section of 500-600 nm axially). So, overall one
expects to have around 108 putative binding sites/signals in the nuclei.
As only a subset of molecules get labelled and are finally detected,
the final number of the expected signals is still quite lower. Thus a
minimum of 107 − 108 signals should be detected in order to get a
first accurate representation of chromatin distribution inside the cell
nucleus [Szczurek et al., 2014].
In one of the previous studies on chromatin organisation, emGFP
conjugated histone protein H2B was expressed [Bohn et al., 2010]
while in another similar study [Ricci et al., 2015], H2B was immunostained using Alexa Fluor 647. At present, both of these studies report
a labelling density of 100-500 signals per µm2 and total identified
signals in order of 104 − 105 . At present, both these studies lack sufficient labelling density to make good estimates about the distribution
of chromatin inside the cell nucleus.
Here, I focus on improving the method so that the overall signal
density and labelling efficiency of DNA molecules is closer to expected
binding sites. In the next section, a simple method which uses direct
DNA binding dyes and fulfils the above criteria is presented.
2.4
A method to reach high labelling density of chromatin with
SMLM
Previous studies of chromatin distribution inside of the cell nucleus
have relied on a fluorophore conjugated with antibodies targeting
chromatin via core histone proteins [Bohn et al., 2010, Ricci et al.,
2015] but none has used direct staining of DNA. Recently it was
found that DNA minor groove binding dyes, such as Hoechst and
DAPI, can undergo UV-induced photo-conversion, to be effectively
employed in single molecule localization microscopy (SMLM) with
high optical and structural resolution. A proposed mechanism is
that these minor groove DNA dyes undergo intra-molecular proton
transfer between the phenol group and the bisbenzimide nitrogen
under UV illumination [Cosa et al., 2001, Carvalho, 2010]. This UV
induced photo-conversion leads to a red shift from the blue emitting
form to a green emitting form; a second excitation at a higher wave
length will push the fluorochrome to a dark state. This situation
eventually induces multiple cycles from dark-state to fluorescent
investigating chromatin organisation using single molecule localisation microscopy
55
Figure 2.6: Figure and caption modified
from [Szczurek et al., 2014]. (A) SMLM
image of a HeLa cell nucleus stained
with Hoechst 33258. Scale bar: 1 µm.
(B1-B2) Widefield and SMLM magnifications of the heterochromatin region
highlighted in inset (B). (C) Complete
nucleus in the widefield mode. (D1-D3)
Magnified sections of widefield, SMLM
and point representation of nucleolus in
(D). (E) Histogram of localization precision with average precision around
26.6 nm. (F) Profiles of the heterochromatic region in the boxes of insets B1
and B2, comparing widefield and localisation images. (G) Hoechst 33258
excitation spectrum and the emission
spectrum shift of its photoproduct. The
green interval is the detection band of
the photoproduct.
state and back to dark state. Fluorescence of these stochastically
blinking molecules is registered by illumination with high intensity
491 nm laser, until bleached (i.e. the molecules are permanently in
the dark state). It is worth to note that green-emitting forms of both,
Hoechst 33258 and DAPI occur rather sparsely, facilitating the optical
isolation of individual dye molecules bound to DNA [Szczurek et al.,
2014]. To test if the molecules are really photoconverted, SMLM
measurements were performed without any prior illumination with
405 nm line. Subsequently, a significantly lower blinking rate in greenyellow channel was observed, resulting in a poor final reconstruction
(data not shown). In the reconstruction from these datasets, fine
structures of chromatin could not be resolved. However, with low UV
excitation (405 nm), a significant increase in number of molecules in
blinking state was observed allowing acquisition of more signals. A
630/90 band pass filter was used to record these signals.
Figure 2.6 shows an optimised adjustment of 405/491 nm laser
intensities to study the distribution of chromatin inside a HeLa cell
nucleus. Compared with previous studies of histone H2B stably
expressed with GFP [Bohn et al., 2010], 50 times more signal density
was recorded (124 signals per µm2 for H2B labelling as compared
to 5000 signals per µm2 here). The nucleus stained with Hoechst
(Figure 2.6A) shows higher density of chromatin around the nucleolus
and at the periphery of the nucleus likely associated with hetero-
56 the periodic and dynamic structure of chromatin
chromatin. Furthermore, a lower density of chromatin is observed
towards the interior of the nucleus, in line with previous studies
[Cremer et al., 2004, Albiez et al., 2006, Cremer et al., 2015].
2.4.1
Theory of DNA dye fluorescence
Molecules with multiple aromatic cycles which bind the small groove
of DNA, such as DAPI, can potentially emit fluorescence. In solution, the structure of the molecules shows a high flexibility, but while
interacting with DNA, they become stabilized and display a planar
structure, which enhances the degrees of freedom of circulating electrons across aromatic cycles or other features of the molecule. This
situation has both effects of increasing the probability of capturing
incident photons, as the surface of the molecule increases, and the
probability to observe an electron hit by a photon, as an electron
occupies more space at a given time ([Biancardi et al., 2013] and
Figure 2.7).
Figure 2.7: Transitions between excitation states during the fluorescence process. Fluorescence depends on the rapid
and lasting rotation between a ground
state (electron of low energy) and an
excited state (electron of high energy).
During relaxation process of the cycle
(red arrow), a photon is emitted that
produces light. With time, electron can
reach other intermediate states called
triplets (T1 and TN here) that can lead
to the deteriorative oxidation of the fluorochrome, resulting in bleaching.
2.4.2
Adapting study of DNA dyes fluorescence to SMLM
Application of blinking properties of DNA dyes to SMLM stems in the
photoconversion of individual molecules followed by their excitation
to the dark state, a concept pioneered by Betzig using GFP [Dickson
et al., 1997, Betzig et al., 2006]. The method has two steps:
1. Optimization of photo-conversion, to ensure that the pattern of
photo-converted molecules is optimal to generate blinking patterns.
2. Optimization of buffers, to push a higher number of fluorochrome
molecules to photo-conversion and delay bleaching of molecules.
investigating chromatin organisation using single molecule localisation microscopy
2.4.3
57
Optimization of the photoconversion process
This step optimizes the amount of molecules photo-converted to
obtain high density of signals. The signals are optimally spread in
time so that one does not acquire too many signals in a single frame,
otherwise one gets unresolved points and poor resolution. To achieve
this, the intensity and the timing of photo-conversion, via a 405 nm
wave length excitation pushing DAPI or Hoechst to a green form,
has to be calibrated properly. If photo-conversion is absent, the final
number of detected signals is low and the final reconstruction is
impaired (Figure 2.8).
2.4.4
Optimization of the buffer conditions
A strong element in the optimization of the protocol to employ conventional DNA dyes for imaging of DNA molecules is the usage
of an effective buffer. Usage of mixes with high concentrations of
glycerol seems to be the best option according to recent benchmarks
[Żurek-Biesiada et al., 2016]. Beside preserving the nuclear morphology, glycerol prevents the generation of reactive oxygen species
(ROS) from the reaction between the fluorochrome and surrounding
dioxide molecules. This reaction releases oxygen atoms with a free
radical, which are very hostile for biological molecules such as DNA
[Bernas et al., 2004]. Glycerol can compete with the fluorochrome for
interaction with the free radical, preserving the fluorochrome from
degradation [Hussels and Brecht, 2011]. This results in a better photoconversion than if the DNA dye was subjected to more Brownian
movements [Biancardi et al., 2013].
2.4.5
Multicolor imaging with DNA
To analyse the spatial arrangement of different nuclear proteins with
respect to DNA, optimization of various conventional dyes was studied along with DNA dyes and summarised in Figure 2.9. It is further
hoped that the dyes presented in Figure 2.9 will be optimised to allow
Figure 2.8: Optimization of the photoconversion process to obtain good resolution and signal density. Obtaining a
good blinking pattern is dependent on
mild photoconversion. Excessive photoconversion (i) leads to large number of
molecules photoconverted and will not
be detected along the different frames,
preventing optical isolation of single
molecules (poor resolution, average density of points). (ii) Insufficient photoconversion separates molecules properly,
but does not generate enough signals to
reconstruct the initial object (high resolution, low density of points). (iii) An
optimized set-up will generate homogenous photoconversion of molecules in
time, resulting in both good separation
of signals across time points and a high
final signal density, allowing a more accurate reconstruction of the initial object
(high resolution, high density of points).
Figure modified from [Szczurek et al.,
2014].
58 the periodic and dynamic structure of chromatin
the visualization of more fluorochromes within one experiment. This
will further enable the analysis of spatial arrangements of various
other components of the nucleus in conjunction with chromatin at the
level of individual nuclei.
Figure 2.9: Comparison of various DNA
dyes and other fluorescent probes. Quality of blinking as well as suitable wave
lengths are indicated. Asterisks indicate
that blue shift needs only a low excitation to happen. Dollars indicate that
dual color imaging can be performed
without correction of chromatic shift. Table is reprinted from [Żurek-Biesiada
et al., 2016].
2.4.6
A summary of various approaches used to study DNA with SMLM
The strategies to study chromatin inside the cell nucleus can be
divided into two broad categories: (1) Direct labelling of DNA and
(2) indirect labelling of DNA. Below I provide a brief summary on
various approaches that have been used to study chromatin/DNA
with SMLM which is pictorially summarised in Figure 2.10.
• Direct staining methods
1. YOYO-1 is a DNA intercalating dye binding between the two
strands of DNA. YOYO-1 and PicoGreen (which binds to the
major groove) were previously used to study the organisation
of bacterial chromosomes with SMLM [Schoen et al., 2011].
The authors called this method binding-activated localization
microscopy (BALM) because both YOYO-1 and PicoGreen bind
directly to the DNA and these molecules only get activated when
bound to the DNA. This is a good way to improve both labelling
efficiency and localisation accuracy as DNA binding increases
specificity and is closer to the actual target than antibodies
are. These dyes have been shown to work for lambda phage
DNA stretches but not for structures which have much higher
chromatin density, for example, mammalian cell nuclei. A report
investigating chromatin organisation using single molecule localisation microscopy
of a successful application of YOYO-1/SMLM to staining of
eukaryotic nuclei will be presented in Chapter 3.
2. Hoechst, DAPI and Vybrant Violet (VV). Conceptually this is
similar to the binding-activated localization microscopy (BALM).
DNA minor groove binding dyes such as DAPI or Hoechst do
not ’blink’ in their standard form. However, with low dose of
405 nm laser, a small proportion of molecules can be photoconverted to green form and start to blink, a property applicable
to localisation microscopy. It is proposed that upon UV illumination, these dyes, which either undergo protonation or become
photoconverted from blue to green emitting form, might also
be driven by hydrogen peroxide [Piterburg et al., 2012]. It must
be noted that only one wavelength (491 nm) is required for
activation and excitation in the case of VV.
3. Propidium Iodide or DRAQ5 are non-cyanine intercalators
that also bind directly to DNA. However, their applicability in
superresolution microscopy has yet to be demonstrated.
• Indirect staining methods
1. EdU incorporation: Zessin et. al. [Zessin et al., 2012] were
the first to describe a high resolution structure of DNA using
precursor incorporation. The authors used EdU (5-ethynyl-2’deoxyuridine) labelling using click chemistry to image nascent
DNA produced during DNA replication. Click-chemistry is
a copper-induced covalent binding between an alkyne group
of EdU and an azide present in the chemically modified fluorophore [Grammel and Hang, 2013]. This method has two main
advantages. Firstly, it is highly specific. The alkyne-azide bond
is not commonly found in any biological system and the method
is very specific and sensitive. Because of this specificity, the
method produces almost no background. Secondly, it is less
harmful compared to other popular precursors such as BrdU
[Baskin et al., 2007, Grammel and Hang, 2013].
2. Core histone proteins: Bohn et al. [Bohn et al., 2010] studied
chromatin organisation inside the cell nucleus by fusing one of
the core histone proteins (H2B) with GFP. In this study, H2B
was either stably integrated (H2B-GFP) or over-expressed (H2BemGFP) in the HeLa cells. In other studies, either immunolabelling against core histone proteins (H3, H2B) was used or
cells were transfected with H2B-mEos2 or H2B-PAmCherry for
live cell imaging [Ricci et al., 2015].
3. Histone modifications: To study functional chromatin and its
compartmentalization, histone modifications were recently used
to describe chromatin organisation [Prakash et al., 2015, Boettiger et al., 2016]. These two studies were based on antibody
labelling of specific histone modifications. The main problem
with such labelling methods is the huge size of primary and
secondary antibody (molecular weight around 150-200 kDa) and
59
60 the periodic and dynamic structure of chromatin
the linker length i.e. the distance between the true position
of the molecule and the position where the signal is detected,
which can be in the order of 7.5 nm for a primary antibody and
in the order of 15 nm for a secondary antibody.
• Sequence specific imaging of DNA: It is often desired to study
specific loci of gene or other DNA sequences instead of the entire
chromatin. This specific staining can be performed using Fluorescence In Situ Hybridization (FISH) [Weiland et al., 2011], oligonucleotide probes [Cremer and Cremer, 2001, Müller et al., 2010]
or via genome editing techniques such as transcription activatorlike effectors (TALE), [Miyanari et al., 2013, Thanisch et al., 2013]
CRISPR/Cas9 techniques [Anton et al., 2014, Chen et al., 2013,
Wood et al., 2011, Hsu et al., 2014, Doudna and Charpentier, 2014]
or SNAP-tag fusion proteins [Klein et al., 2011].
Figure 2.10: Comparison of direct and
indirect DNA labelling. Top block
shows various DNA minor groove binding dyes that can be used to study the
distribution of the chromatin inside the
cell nucleus. Bottom block shows examples of indirect methods (immunostainig and EdU incorporation) to study
the distribution of chromatin inside the
cell nucleus.
Figure 2.10 compares the overall staining of the cell nucleus with
direct and indirect staining. One can observe that direct staining
of DNA leads to a more continuous distribution of chromatin as
compared to indirect staining, where the distribution seems to be
in a speckle-like patterns (see also [Bohn et al., 2010, Ricci et al.,
2015] for recent examples of low labelling density). This happens
because in specific staining only the functional compartments, usually
associated to specific histone modifications, are revealed. For example,
H3K14ac is associated with active transcription and hence only sites
where chromatin is in a more open conformation fluoresce. Similarly,
in EdU incorporation (Figure 2.10, bottom right) only the freshly
replicated sites from S-phase are detected (see Chapter 3 for more
examples of these two cases).
investigating chromatin organisation using single molecule localisation microscopy
2.5
61
SMLM microscope design and imaging pipeline
The basic set-up for SMLM imaging is very similar to the configuration
of a wide-field microscope. In the following sections, various aspects
of the SMLM such as sample preparation, data acquisition, data
reconstruction, data visualisation and data analysis are discussed.
2.5.1
Sample preparation for SMLM
A good SMLM sample follows three criteria. (1) The nano-structure
is preserved. (2) A high number of fluorophores bind to the target.
(3) Non-specific labelling and background are minimised [Bates et al.,
2013]. For preservation of nanostructure, optimised fixation protocols
from EM can be used. Optimization of buffer is key and usage of
reactive oxygen species competitors which prevent fluorochrome from
degradation can help. Background/non-specific signal can severely
decrease the quality of a superresolution reconstruction [Betzig et al.,
2006]. A comparison of different background scenarios is shown
in Figure 2.11. Use of thin samples in addition to an appropriate
imaging buffer can help to minimize the background.
2.5.2
Imaging medium
Usage of photo-convertible DNA dyes (Hoechst, DAPI and VV) requires certain components to be present in the ’switching buffer’. In
particular, the switching buffer should contain 0.5 mg/ml Glucose oxidase, 0.04 mg/ml Catalase, 0.1 g/ml glucose in PBS. The use of MEA,
one of main components of the STORM/dSTORM imaging buffer diminishes the rate of switching in the case of DNA dyes. The switching
buffer facilitates the switching of the steady fraction of spontaneously
blinking molecule in each frame. The switching buffer is the same
as the oxygen scavenger system in STORM/dSTORM, which mainly
reduces photo-bleaching fluorescent dyes. More details on imaging
medium can be found in [Szczurek et al., 2014, Żurek-Biesiada et al.,
2015, 2016].
2.6
Data acquisition for SMLM
For imaging DNA with photoconvertible DNA dyes, several important aspects have to be considered. Firstly, low power 405 nm illumination in conjunction with a 491 nm laser should be used to increase
the number of detected molecules. Similar ideas were previously
described in [Heilemann et al., 2008] for photoswitchable fluorescent
dyes. However, these dyes have no affinity for DNA. The SMLM
measurements were performed with a vertical/upright microscope
set-up (see 2.12). The set-up has four laser sources with excitation
wavelengths at 405 nm, 491 nm, 561 nm and 647 nm and can be
dissected into 3 modules described below.
Figure 2.11: Different kinds of background in a SMLM measurement: (A)
No blinking. All signals are considered
as background. (B) No background, a
typical PALM set-up in TIRF mode with
only blinking molecules. (C) Uniform
background. (D) Highly spatial and temporal non-uniform background
62 the periodic and dynamic structure of chromatin
1. The illumination module: The laser beam enters the microscope
via mirrors (M) and dichroic mirrors (DM)[M1-M4, DM1-DM4], a
collimator arrangement (5x expansion of beam diameter) and neutral density filters (if required). This collimator is made up of two
achromates with focal lengths of 30 mm and 150 mm which expand
the beams by a factor of 5 to get an homogeneous illumination of
the sample.
2. The localisation module: To achieve the high laser intensity for
the localization mode, a lens [LL] is inserted in the optical pathway,
leading to a smaller illuminated area of higher intensity in the
object region of interest. The beam is focused by this lens into the
back focal plane of an oil immersion objective lens (63x, NA=1.4).
The sample is actuated by a piezoelectrical stage [PS] for focusing.
3. The detection module: The emitted fluorescent light goes through
the DM and is focused by a tube lens [TL] (1.0x, f=200 mm) onto
the CCD chip of a highly sensitive 12 bit black-and-white camera.
A set of appropriate blocking filters is mounted in a filter wheel
[FW] in front of the camera chip.
Before acquiring data, certain parameters have to be optimised. For
example, bleaching gradient, blinking and their relation to integration
time must be studied prior to the measurement and the final number
of acquisition frames should be chosen accordingly. One should try
to optimise the integration time of the camera as this determines
signal-to-noise ratio and eventually the background of the sample.
Acquisition of wide-field images before and after measurement helps
to compare the structure of the sample with the final reconstruction.
Illumination of the sample should be homogeneous. Furthermore, to
Figure 2.12: A schematic of the SMLM
microscope. For description of optical
parts, refer to the text.
investigating chromatin organisation using single molecule localisation microscopy
63
fasten the acquisition time and save disk space, the sample area can
be restricted to a region of interest.
2.7
Data reconstruction for SMLM
Figure 2.13 presents various steps required to reconstruct a highly
resolved image from single molecule coordinates. The operations can
be characterized at various levels which are described below.
Elements of localisation microscopy
WIDEFIELD
STACK LEVEL
FRAME LEVEL
1. Generate background map
2. Calculate difference image
ERROR CORRECTIONS
VISUALISATIONS
1.
2.
3.
4.
1. Find local maxima
2. Extract ROIs
1. Remove multi-frame points
2. Correct for sample drift
3. Remove events which don't fulfill quality
criteria (e.g. std, error thresholds)
Gaussian blurring
Triangulation
Histogram
Other methods
list of points
analysis on single molecules
ROI LEVEL
analysis on image
ANALYSIS
1. Localize fluorophore
2. Save the determined parameters
(position, photon number, std, errors, ...)
1. Cluster analysis
2. Colocalisation
• Stack level: A high number of frames with each frame containing
only a few optically isolated molecule must be generated. This data
stack can be used to analyse the position of the fluorophore across
multiple frames and characterise the fluorophore. Furthermore,
a background map for each frame must be calculated and subtracted to get the difference image. This step can be really critical
depending on the type of background in the sample (refer 2.13.)
• Frame level: Local maxima in each frame are determined and the
corresponding regions-of-interest (ROI) are extracted.
• ROI level: Each fluorophore is localised and an estimate about the
localisation precision for each molecule based on the parameters of
the model function should be calculated. An estimation about the
Figure 2.13: SMLM data reconstruction
and analysis flowchart. ROI: Region of
interest. See text for details.
64 the periodic and dynamic structure of chromatin
number of photons per molecule is then calculated, an important
parameter to study the characteristics of the fluorophores used in
the experiments.
• Error corrections: Signal from a particular fluorophore occurs
more than once during an acquisition and subsequent frames.
Removal of signals in multiple consecutive frames is important,
so that the same fluorophore is not counted twice. If the sample
moved significantly during acquisition, the drift in the sample
must be corrected, a common procedure for measurements of long
duration. Finally, the events that do not fulfil other quality criteria
are removed.
• Visualisation: The position of the single molecules should then be
visualised to get a nanoscale representation of the biological object.
Visualisation methods include for example, Gaussian blurring,
triangulation and histogram methods. A visualisation method is
selected based on the labelling density and/or the prior knowledge
about the underlying structure from other methods such as EM.
• Analysis: Finally, different kinds of analysis such as co-localisation,
cluster analysis, nearest neighbour, density analysis can be performed on either high resolution images or coordinates of single
molecules.
2.7.1
Spot finding for SMLM
Algorithms used to precisely identify the locations of fluorophores
can be broadly classified into two categories, fitting based and nonfitting based (usually Centroid) methods. While iterative fitting-based
methods can usually provide fitted parameters equal or close to the
maximum likelihood estimate, ad hoc centroid based methods are
usually very fast. However, any localisation method will struggle if
the underlying model poorly represents the observed data e.g. in case
of a high background level, presence of out of focus signals or noise,
etc. A particular challenge for the exact fluorophore determination
is posed by both spatially and temporally fluctuating background
intensities arising from out-of-focus blinking fluorophores. This can
happen if the structure is not per se 2-dimensional (e.g. PALM using
TIRF illumination, see Figure 2.11). In the following sections, I discuss
these two broad categories of localisation methods.
Fitting based methods
Photons emitted by point sources are smeared out onto the detector
of the imaging system due of the wave nature of light. This smear or
blur is described by the point spread function (PSF) of the imaging
system. The intensity distribution of a photon at the detector follows
an Airy function but for practical purposes is approximated with
a Gaussian function. The overall model function consists in a 2D
Gaussian describing the PSF, a function describing the background
investigating chromatin organisation using single molecule localisation microscopy
level and one more term to describe the noise. The model function is
represented by:
"
A
1
I ( x, y) =
exp −
2πσx σy
2
( x − x0 )2 ( y − y0 )2
+
σx2
σy2
!#
+B+N
where x, y are the coordinates of the point emitters, σx , σy are the
width of the distribution, A the peak intensity value of the distribution, 2πσx σy the normalisation factor, x0 , y0 the coordinates of the real
center of the distribution, B the background level and N the noise
parameter.
The next task is to optimise the parameter values in order to minimise the error between the signal and the model. The optimisation
can be done on all the parameters or just on the coordinates of the
emitter positions. Two common methods to optimise the parameters
are the least squares method (LSM) and the maximum likelihood
estimation (MLE). The principle behind both methods is the same.
In LSM, the difference between the signal and model on each pixel
is calculated and the error is then squared and added. Based on the
error sum, new parameter estimates are made and the sum of the
squared errors is calculated again to check if the updated parameter
values reduce the error. This process is iterated until the least summed
errors are obtained.
In the MLE, the overall process is the same except that detailed
information about the experimental conditions such as knowledge
about the PSF, noise, background is required. Similarly to LSM, the
difference between the signal and the model is calculated and statistics
of noise are then used to predict how likely the difference between
the model and the signal is. The parameters are tuned to maximise
the likelihood of the data. Cramer-Rao Lower Bound (CRLB) states
that MLE makes more precise estimates compared to LSM or other
parameter optimisation methods [Small and Stahlheber, 2014, Small
and Parthasarathy, 2014].
Non-fitting based methods
The fitting methods are often computationally intensive and difficult to handle when the sample is inhomogeneous. Another approach
is to find the coordinates of the centroid of the fluorescent spot or
in the case of isotropic emitters to take advantage of their radially
symmetric shape to estimate the coordinates. When the background
is non-uniform and so localisation density is high, due to the additive nature of background noise at regions with high localisation
density, the coordinate estimates are biased towards the center of
the ROI, often resulting in a grid like patterns (Figure 2.14 and [Best
et al., 2014]). Various filtering methods should be used to reduce the
background intensity and improve the coordinate estimation. For
more details on various algorithms estimating the emitter coordinates,
please refer to [Small and Stahlheber, 2014, Small and Parthasarathy,
2014, Chenouard et al., 2014].
65
66 the periodic and dynamic structure of chromatin
Figure 2.14: Various reconstruction algorithms for SMLM: Fitting (Gaussian)
and non-fitting (centroid) reconstruction
algorithms are compared. While nonfitting algorithms are fast, fitting algorithms are more accurate. The grid-like
patterns are frequenctly observed when
high density data are reconstructed with
centroid method (see text for more details and [Best et al., 2014]). Scale bar:
1µm.
2.7.2
Drift correction algorithms for SMLM
The correct position determination of fluorescent molecules is crucial
for the interpretation of the localisation microscopy data, e.g. to
understand the biological structure investigated. The position of
fluorophores is highly sensitive to environmental disturbances (e.g.
acoustic vibrations) and to mechanical instabilities of the microscope
hardware (e.g. thermal expansion, stage drift). These disturbances
can cause distortion in the recorded image, significantly affecting
the achieved localisation accuracy, especially when it is in the order
of tens of nano-meter or better. Previously, on line drift correction
methods have been described, which estimate and correct the drift
experienced during recording of the experiments. Predicting the drift
for future frames is a hard task, as behaviour of vibrations can not be
predicted. This can easily be done after measurement.
With colleagues, I developed a drift correction strategy based
solely on already acquired data without any fiducial markers [Hagmann et al., 2014]. It was found that in some SMLM set-ups (PALM,
STORM), most of the biological samples exhibit enough permanent
(photostable) structure that can be used at several time points to gain
information about the sample. This information was used to correct
the lateral drift based on the underlying structure visible in a raw
image sequence (averaged over 10 subsequent frames for one sample
image), to then calculate the auto-correlation between sample images.
Then two (for x and y) eighth order Fourier series were fitted to the
acquired data in order to obtain the drift vectors. Figure 2.15A-B
shows HeLa cell nucleus stained with Hoechst 33258 photo-product
before and after drift correction. The long acquisition times resulted
in considerable mechanical drift of the stage. In the Figure 2.15, the
frames of the data stack are color coded to demonstrate the drift
in the sample. The initial frames are coloured blue, while the final
frames range from yellow to white. The sample can be seen to drift
investigating chromatin organisation using single molecule localisation microscopy
67
Figure 2.15: Structed based drift correction strategies for localisation microscopy: Some variants of localisation
microscopy exhibit enough permanent
(photostable) structure that can be used
to correct the lateral drift based on the
underlying structure visible in a raw image sequence. HeLa cell nucleus stained
with Hoechst 33258 photoproduct before (A) and after drift correction (B) are
shown. (C) Widefield image of the same
nucleus. Due to long measurements, the
initial drift was found to be in the order
of 500-1000 nm. (D-E) After application
of structured based drift correction algorithm drift improved to less than 20
nm, well within the localisation precision. The magnified sections (D and
E) show significant improvement in resolution (computed using FRC method
[Nieuwenhuizen et al., 2013]). The resolution before drift correction was 125
nm and improved down to 107 nm after
[Hagmann et al., 2014].
from top right to bottom left. For the SMLM microscope used for the
measurement, the drift was in order of 100-1000 nm (Figure 2.16A)
and was determined by image phase correlation. A polynomial or
Fourier series (equation below) was fitted through the data, from
which the dislocation of every event was subtracted.
n
y = a0 + ∑ ai cos(iwx ) + bi sin(iwx )
i =1
where a0 is the intercept term in the data (associated with i = 0
cosine term), w the frequency of the signal and n the number of
harmonics (8 in our case).
In the present case of a cell nucleus with high background, the
data were corrected to a final drift under 20 nm (Figure 2.16B), which
was reasonable given the high localisation precision (average 25 nm,
Figure 2.16C). Furthermore, the resolution computed based on Fourier
ring correlation [Nieuwenhuizen et al., 2013] showed a significant
enhancement in resolution before (125 ± 12 nm ) and after (107 ± 11
nm) drift correction. The drift correction also resulted in significant
improvement of neighbour distances. The mean nearest neighbour
distances before and after drift correction were approximately 42.95
nm and 39.76 in the zoomed sections of the cell nucleus (Figure 2.15DE and Figure 2.16D).
Overall, the drift correction resulted in increased local density of
chromatin at the nuclear periphery, which is in accordance with nuclear architecture. It is believed that with appropriate drift correction
strategies, one can observe true structural features which are often accompanied with a decrease in nearest neighbour distances. Structure
68 the periodic and dynamic structure of chromatin
A
C
Figure 2.16: Quantification of sample
drift. (A) The drift along x axis was less
than 200 nm, however the drift along
y axis was in oder of 500 nm. (B) After application of structured based drift
correction algorithm, the overall drift in
both x and y directions was less than 20
nm, well within the localisation precision (C). (D) shows the nearest neighbor
distances before and after drift correction. The mean nearest neighbour distance before drift was 42 nm while it
improved to 39 nm after correction.
B
D
Mean distance before drift 42.95
Mean distance after drift 39.76
based drift correction demonstrates high background and that the
underlying structure of datasets can be used to correct drift. Using
this approach, one can successfully correct the localisation microscopy
data down to a final drift under 5 nm. The results are comparable
to fiducial markers based strategies. Moreover, the resolution of the
final reconstructions is substantially enhanced and the natural limit
of localisation precision is achieved.
2.7.3
Data visualisation for SMLM
While most imaging modalities generate an image, SMLM provides
coordinates single molecules in addition to the image common to the
other superresolution methods (STED and SIM). Previously, scattergram of point positions [Hess et al., 2006, Van Oijen et al., 1998], 2D
histogram of point positions [Heilemann et al., 2008, Lemmer et al.,
2008, Fölling et al., 2008, Egner et al., 2007], localisation precision
based Gaussian blurring [Betzig et al., 2006, Rust et al., 2006], nearest
neighbour blurring of molecule positions [Kaufmann et al., 2012] and
triangulation of localisations [Baddeley et al., 2010] have been the
most popular visualisation methods.
In Figure 2.17, localisation precision based Gaussian blurring, triangulation and nearest neighbour blurring of individual molecule are
compared. When enough number of nearest neighbour molecules
are taken into consideration (20 in the present case), the underlying
structure becomes increasingly visible. This happens because blurring
locations with a Gaussian is not the best choice when the labelling
density is low. In such a case the pointillist appearance of the emitters
investigating chromatin organisation using single molecule localisation microscopy
69
Figure 2.17: Comparison of various visualisation algorithms: Comparison of
Gaussian blurring based localisation precision, triangulation and nearest neighbour blurring of individual molecules
is shown. When enough nearest neighbour points are taken into account, the
underlying structures can be better observed. Nearest neighbour blurring
can help to bring out the underlying
structure in images lacking proper labelling density. Moreover, the points
that are not part of the structure are suppressed. Figure and caption modified
from [Prakash et al., 2015].
resurfaces in the final image. Moreover, there is a resolution loss factor
√
of 2 as stated in [Baddeley et al., 2010]. However, nearest neighbour
blurring is helpful at revealing structures especially if the sample
suffers from a low density of points. Moreover, the points that are not
part of a structure are suppressed. Finally, the visualisation method
is chosen based on the labelling density and the characteristics of the
structure to visualise (1D, 2D or 3D).
2.7.4
Data analysis for SMLM
Advent of various super-resolution microscopy methods have brought
to light various diffraction limited biological objects, revealing periodic patterns [Prakash et al., 2015, Früh et al., 2015, Xu et al., 2013]
under better resolution. The analysis of periodicity in these studies
[Früh et al., 2015, Xu et al., 2013] was based on the intensity information of the images. In addition to pixel intensity information, SMLM
provides information about the single molecule coordinates which
can also be used for autocorrelation and periodicity analysis.
Single molecule autocorrelation
For single molecule autocorrelation, each localisation was blurred
with 20 nearest neighbours and a binary mask was generated (see
Figure 2.18) using a global threshold described elsewhere [Prakash
et al., 2015, Jianzhuang et al., 1991]. The single molecules within
this binary mask were extracted for further analysis. This resulted
in keeping only the molecules which belong to the structure or the
70 the periodic and dynamic structure of chromatin
Figure 2.18: Various steps involved in
single molecule autocorrelation. (A)
Firstly, each molecule is blurred using
the value of the mean distance to 20
nearest neighbour points. (B) Secundly,
a binary mask is generated based on the
nearest neighbor image and (C) only the
signals within the binary mask are considered for further analysis. (D) Each
molecule position is then translated to
Cartesian coordinates for the ease of
analysis. (E) A histogram of the tangential and normal positions is generated.
(F) Finally the bins of the histogram are
auto-correlated to estimate the periodicity.
clusters considered. For the autocorrelation analysis, a line was fitted
manually to pass through the center of the object. The coordinates
were then translated and rotated to get a more linear orientation.
Finally, the autocorrelation values were calculated on the histogram
of the tangential distances along the central axis. The algorithm is
described in detail in [Prakash et al., 2015].
2.8
Some further considerations for localisation microscopy
There are a number of artefacts that can arise due to sample preparation mainly due to fixation conditions, permeabilization conditions,
antibody concentration and the type of buffer used. These are described in detail in [Whelan and Bell, 2015]. Here, I discuss three more
kinds of artefacts that can arise from the way microscope is set up and
the coordinates of single molecules are estimated post-acquisition.
2.8.1
Artefacts in localisation microscopy
Figure 2.19 shows negative of wide-field and localisation image of a
HL-1 cell nucleus stained with Hoechst. Due to the non-homogeneous
illumination of the sample (Figure 2.19A), the lower left of the cell
nucleus is more illuminated than the upper right part of the nucleus.
This leads to a non-uniform detection of molecules across the sample.
For example, the chromatin density at the nuclear periphery (the
upper right part of the image in (Figure 2.19A-B) is higher than the
lower left part of the image. The same pattern of density can also be
observed around the nucleolus (black arrows). Another factor that
can lead to a non-homogeneous illumination of the sample is the
Gaussian profile of the excitation beam. In such a case, the central
regions are more illuminated than the regions toward the periphery
of the nucleus.
investigating chromatin organisation using single molecule localisation microscopy
71
Figure 2.19: Artifacts in localisation microscopy: (A) Negative of a widefield
image of HL-1 cell nucleus stained with
Hoechst. (B) Negative of high resolution
reconstruction of the same cell. Dense
chromatin blob is observed around the
nucleolus (black arrows inside the nucleus) in (A) but is missing in (B). Moreover, no dense chromatin is observed
at the nuclear periphery in (A) but is
present in the high resolution reconstruction in (B). This is usually due to nonhomogenous illumination of the sample. One factor that can generate nonhomogenous illumination is the Gaussian profile of the high intensity excitation beam that causes the central region
in widefield mode to be illuminated
more than the peripheral regions. (C)
A section of the nucleus showing a gridlike pattern. Such patterns usually arise
in high density single molecule reconstructions when the Center-Of-Intensity
(COI) algorithm is used. COI biases the
localisation towards the center of the
CCD camera pixel, giving rise to such
grid-like patterns.
Another type of artefact can arise due to the data reconstruction
when there is a high number of signals in each frame (Figure 2.19C).
The high density leads to the overlap of the signal resulting in a gridlike patterns. If the centroid fitting algorithm or center-of-intensity
(COI) is used to localise the signals then the signal is biased toward
the center of the pixel resulting in grid-like patterns (see Section 2.7.1).
However, no grid-like pattern is observed with Gaussian fitting, which
is more accurate but slower when compared to centroid fitting. One
solution to solve this problem is to make sure only few signals blink
in each frame so signals do not overlap and can be optically isolated.
Then one has to take a high number of frames to get a high density
super-resolved image. This can be achieved by initial reversible photobleaching of the sample before acquisition and waiting until the
uniform blinking of the signals is achieved.
A third type of artefact is experienced when biological objects
under study are poorly labelled. The pattern which are obvious in
wide-field image are missed in the high resolution images. Such an
example can be seen in Figure 2.20. Pol II is known to form barrel-like
elongated shapes in polytene chromosomes [Zhimulëv, 1996] which
are apparent in wide-field image (Figure 2.20A, inset A1). However,
SMLM imaging does not reconstruct these apparent patterns of Pol II
72 the periodic and dynamic structure of chromatin
Figure 2.20: Dual color SMLM image
of Drosophila polytene chromosome
stained with VV (green) and Pol II Ser2PAlexa Fluor 555 (red). (A) Widefield
image of a fragment of a polytene chromosome. (B) SMLM images of the fragment as in (A). Arrows indicate regions
of loose chromatin inside the polytene
chromosome. (C) Enlarged fragment of
the images shown in A and B, embracing
dense heterochromatic and lower density eu-chromatic regions of DNA. Figure and caption modified from [ŻurekBiesiada et al., 2015].
(Figure 2.20B), following the well-known alternation patterns of dense
and light regions of polytene chromosomes [Zhimulëv, 1996]. Pol
II-rich regions are enriched for genes, and correspond to the dense
regions of electron microscopy images, while Pol II-poor regions
correspond to light regions, deprived of genes. This example shows
a major pitfall of super-resolution imaging, particularly prominent
when the underlying structure is unknown and poor or insufficient
labelling of the biological object might lead to a wrong interpretation
of the data. This also reflects the importance of a good in-depth
literature survey about the biological object and the importance of
blurring when the labelling density is limited.
2.8.2
Difference between localisation precision and accuracy
The terms localisation precision and accuracy are often used interchangeably. Localisation precision is the exactness with which the
position of a signal can be predicted, which can be slightly off the
true position of the molecule. This distance between the true position
of a molecule and the detected signal is the localisation accuracy.
Figure 2.21A, shows a detected signal slightly off from the true position of the molecule. As the signals are not spread, the localisation
precision will be good but the localisation accuracy will be poor. In
the second case (Figure 2.21B) the detected signals are spread over a
large region resulting in a poor localisation precision but an increased
localisation accuracy as the detected signals are close to the true position of the molecule. In the third case (Figure 2.21C), localisation
accuracy and precision are both high, a typical case for Hoechst and
DAPI which directly bind to DNA (high localisation accuracy) and
display a high photon count (high localisation precision).
investigating chromatin organisation using single molecule localisation microscopy
73
Figure 2.21: Localisation precision Vs
localisation accuracy: (A) Good localisation precision but poor localisation
accuracy. (B) Good localisation accuracy but poor localisation precision. (C)
Good localisation accuracy and localisation precision. The true position of a
fluorophore is indicated by a blue star
while turquoise blobs indicate the detected signal.
2.9
Summary
Single Molecule Localisation Microscopy (SMLM) is one of the major
emerging tools for the analysis of biological structures at approximately 10 nm spatial resolution. The procedure relies on sequential
detection of (a subset of) individual fluorophores. For dense regions
(fluorophores with significant overlap), a compromise between labelling density and the photoswitching behaviour of fluorophores is
needed to optically isolate molecules in each acquired frame. The
last 10 years have seen a significant progress in fluorescence imaging
with the development of new fluorescent probes and the discovery of
new properties of existing fluorochromes [Szczurek et al., 2014, ŻurekBiesiada et al., 2015], making them applicable for super-resolution
imaging.
Until recently, description of chromatin was limited to immunostaining of histone proteins, which does not provide enough labelling
efficiency and density for chromatin imaging [Bohn et al., 2010, Ricci
et al., 2015]. Nevertheless, usage of conventional dyes dramatically
increases the amount of signals recorded, opening avenues for the
description of chromatin architecture at all levels of magnification.
This is the matter of the next two chapters; Chapter 3 discusses many
structural features of chromatin found in interphase, while Chapter 4
presents recent results regarding the chromatin architecture of the
pachytene chromosome. Both these aspects of chromatin organisation
grandly benefit from the developments presented in the current chapter. In this chapter, I discussed the theory and concepts of SMLM, as
well as sample preparation, hardware implementation, data reconstruction, visualisation and quantifications. I closed the chapter with
a note about localisation accuracy, precision, resolution and some of
other technical pitfalls.
With more efficient fluorophores (i.e. with higher photon counts),
highly sensitive camera, high NA objective lenses, an even higher
spatial resolution can be expected in the future. A particular important challenge is live cell imaging, which requires both high spatial
and temporal resolution though efforts in this direction are already
74 the periodic and dynamic structure of chromatin
being made [Biteen et al., 2008, Shroff et al., 2008]. Single particle
tracking is another area where photo-switchable fluorescent probes
will facilitate tracking of high-density particle in live cells [Manley
et al., 2008]. Finally, I hope that together with SIM, SMLM can be
broadly applied to chromatin biology and other areas of cell biology
to provide insights into the nuclear architecture at nanoscale.
Stress is the shortest distance between two homologous chromosomes. Searching for a soul-mate?
3 Structure, Function and Dynamics
of Chromatin
3.1
Introduction
Chromatin is a DNA-protein polymer, which consists of DNA, structural proteins, non-structural proteins and RNA. During interphase,
chromatin encodes the information necessary to maintain the primary
functions of the cells. At the same time, the cell holds in a relatively
small volume (roughly 4000 µm3 of diameter for the nucleus) the entire genome, which is billions of base pairs long and roughly 3 meters
in length when completely unfolded. Moreover, information needs to
be retrieved from this highly condensed structure when a gene needs
to be expressed at a fast pace. To achieve this, chromatin arranges
itself in a highly ordered and compact structure. This compaction
has several hierarchical levels of folding, similar to fractals, which
show similar patterns at multiple length scales. DNA helix represents
the lowest level of the hierarchy while the metaphase chromosome
represents the highest level. Before the advent of EM studies, research
on the nucleus was mainly focused on (1) its chemical composition
and (2) the structure of highly condensed chromosomes. The status
of chromatin during interphase was oblivious and limited to the
description of punctuated elements such as nucleoli. Recently, the
overall view of interphase chromosomes has become very detailed,
and several intermediate orders have been shown to exist, though the
importance of these orders is still unknown.
The main feature of interphase chromatin that has been extensively
studied is the bead-on-string like structure called nucleosome. A
nucleosome is a structure of 11 nm diameter consisting of DNA
and histone proteins [Kornberg, 1974, Richmond et al., 1983, Luger
et al., 1997, Kornberg and Lorch, 1999]. The DNA molecule wraps
around an octamer of four histones with a linker histone (H1) at
the base leading to compaction of nucleosome clusters. Nucleosome
clusters deprived of linker histone (H1) decondense and rearrange
themselves into ’beads-on-a-string’ configurations [Thoma et al., 1979].
The DNA sequence has a highly consistent length of 147 base-pairs
across nucleosomes. Overall, there is a very good understanding
about the chromatin organisation at the DNA and nucleosomal level,
but the way chromatin is compacted at higher levels is still oblivious,
despite the recent evidence from genomic and microscopy studies.
76 the periodic and dynamic structure of chromatin
Figure 3.1: Methods from microscopy
and genomics to capture chromatin
structural patterns.
Due to its significantly higher resolution, previously EM has been
a very powerful method to study organisation of chromatin at various length scales. EM could distinguish structural elements such as
heterochromatin, nuclear pores or mRNA. Experimental destruction
of chromosomes [Cremer et al., 1982b, Cremer and Cremer, 2001]
showed that chromosomes barely intermingle inside of the nucleus
and for the most part are organized in distinct territories [Zorn et al.,
1976, 1979, Stack et al., 1977]. Nevertheless, between these two contrasting levels of chromatin organisation: the nano level (DNA double
helix, nucleosome level) and the micron level (chromosome territory
level), how chromatin is organized is not well understood. Recent
experiments from high resolution microscopy, genetics and genomics
have brought new observations to light regarding the apparent hidden complexity of interphase chromatin (Figure 1.8). Methods such
as nucleosome profiling or HiC have provided valuable information,
but at the same time convoluted the picture of nuclear processes and
structural features. As a consequence, there is still not a consensus
regarding the way interphase chromatin is organised. Genomics has
brought some strong clues, but this information is mostly populationbased and suffers from strong cell to cell heterogeneity. As a result,
it is hoped that single-cell information coming from microscopy and
also from single-cell genomics provides the missing information re-
structure, function and dynamics of chromatin
quired to understand chromatin architecture. Figure 3.1 presents
various methods from microscopy or genomics to study chromatin
architecture.
In this chapter and Chapter 4 on the organisation of meiotic chromosomes, I tried to gather some evidence regarding the organisation
of chromatin at multiple levels that I could capture during my PhD
studies. I will show how microscopy at single molecule resolution
can be a useful tool to identify the different levels of chromatin compaction, from the nucleosome level up to the level of the metaphase
chromosome. One of the challenges is the scarcity of data. Here,
using newly established SMLM applied to conventional DNA dyes,
I highlight the different levels of chromatin compaction. In reverse
hierarchical order, I will start by the concept of chromatin territory,
which states that chromosomes have defined regions within the cell
(Section 3.2.1, scale of 1 to 2 µm). I will then present subchromosomal
domains, which are bulky parts of the chromosome territory with
potential functionality (Section 3.2.2, scale of 500 nm). I will show how
chromatin domains now emerge as fundamental building blocks of
chromatin, at least from the point of view of a biologist (Section 3.2.3,
scale of 100-400 nm). These regulatory domains have a strong potential to be functional units of chromatin and their detection here
is backed up by recent data from genomics. I will finish by describing lower orders of chromatin folding, including nucleosomes and
chromatin fibres (Section 3.2.4 and Section 3.2.6). I hope that these
findings will help to define a framework for the further investigation
of chromatin structure and function.
3.2
The hierarchical organisation of chromatin
There are two classical views about how the interphase chromatin
is organised at the highest level of hierarchy. The first is the compartmentalization of nucleus into distinct and functional chromsome
territories and the other is the classical bimodal euchromatin and heterochromatin model. In this section, I will provide recent evidences
from single molecule localisation microscopy to get a more detailed
structure of interphase chromosomes and how their hierarchical nature might result from association of fundamental building blocks of
chromatin.
3.2.1
Chromosome territories (scale: 1000-2000 nm)
As mentioned in the first chapter, interphase chromatin has long been
described as an alternation of dense and loose chromatin, i.e. heterochromatin and euchromatin, with no further complexity. With time,
the idea that chromosomes follow a territorial organisation slowly
became established [Cremer and Cremer, 2001]. Later, it was shown
that chromosome territories do exist and are well conserved between
different cell types [Mayer et al., 2005]. The paradox between the
existence of chromosome territories and the lack of visible territories
77
78 the periodic and dynamic structure of chromatin
Figure 3.2: A simulated chromosome
territory map (CT) based on SMLM
data. Top panel: The nearest neighbour based clustering is used to segregate chromatin into 46 distinct clusters
representing individual chromosomes
and marking their individual territories.
White lines show the delimitations of
three chromosome territories. Yellow
line highlights a putative active interchromosomal territory (IT). Bottom panels: Magnifications of top panel. In
white are depicted three chromosome
territories and an hypothetical region of
active chromatin is displayed in yellow.
Scale bar: 2 µm.
in electron microscopy has recently been solved by the emergence of
super-resolution microscopy. By relying on the detection of single
molecules rather than on the capability of the sample to transmit an
electron flux, these methods have drawn a clearer picture of the organisation of DNA and associated proteins in the interphase nucleus
[Cremer et al., 2006]. Interphase chromatin is presently described as
chromatin networks and foci, separated by caveats, as exemplified
by EM images [Branco and Pombo, 2007, Markaki et al., 2010]. It is
thought that most of these caveats are the place of protein machineries
such as transcription factories [Iborra et al., 1996, Sexton et al., 2007]
and alternative splicing complexes known as speckles [Zirbel et al.,
1993, Puvion and Puvion-Dutilleul, 1996]. Transcription factories colocalise in DNA void regions while the speckles have been shown
to anti-localize with DNA elements [Zirbel et al., 1993, Cremer and
Cremer, 2001, Markaki et al., 2010]. In this first section, I will show
how SMLM can be used to get information about chromatin and
reveal some structural properties of CT and other nuclear features
belonging to a high-order level of structure.
structure, function and dynamics of chromatin
79
Nearest neighbour clustering identifies chromosome territories and
inter chromosomal compartments
Nearest neighbour molecules information was used to predict
territories of different chromosomes. The algorithm generates a hierarchical cluster tree of localization data using the MATLAB built-in
function linkage. This function links pair-wise fluorophores in order
to form clusters, then links the clusters again pair-wise and so on.
The process is iterated until only one cluster is found. A parameter
threshold that helps to decide when to separate territories, was set to
fit the number of expected regions (39 territories here). The resulting
image is a mosaic of clusters that are predicted to be chromosome
territories (Figure 3.2). Further validation is required to prove whether
the prediction of territories positions is correct. This could be done
using FISH or additional labelling techniques.
Interestingly, the picture also reveals regions of lower density (dark
regions) which are overlapping two to four chromosomes. This pattern is a putative inter-chromosomal territory, a region of the nucleus
thought to be the place of active transcription, while the core of the
territory displays a more condensed chromatin [Cremer and Cremer,
2001] (yellow territory in Figure 3.2, see also Appendix A.1). The alternation between light and dark regions of Figure 3.2 further validates
the CT/IT model proposed in [Cremer and Cremer, 2001].
Figure 3.3:
Identification of subchromosomal territories. Top ppanel:
The number of signals is gradually increased from 10000 to 1000000 to reconstruct the final image. Bottom panel:
corresponding values of localisation precision. Irrespective of the increase in
number of signals, the average localisation precision is found to be around 20
nm in each case. Howewer, increasing
the number of signals acquired helps
to identify sub-chromosomal territories
(top right panel).
Good labelling density is required to distinguish chromosome territories and sub-chromosomal compartments
Single molecule microscopy is highly useful to identify the precise
location of molecules and distinguish signals whose relative position
with respect to the neighbouring molecules is not distinguishable
in conventional microscopy. Moreover, by gradually increasing the
number of detected molecules, one is able to refine the description of
various nuclear structures. Figure 3.3 shows the minimum number
of signals required to distinguish various nuclear compartments. For
80 the periodic and dynamic structure of chromatin
instance, the nucleolus is only visible for 100000 signals detected.
Conversely, a much higher number of signals is required to distinguish individual chromatin domains and the dense chromatin at the
nuclear periphery commonly seen in EM images (Figure 3.3). Another
thing worth noting is that localisation precision remains around 20
nm throughout our analysis.
SMLM reveals putative sites for transcription
Our SMLM setup is useful at probing chromatin structures without
prior knowledge about the underlying DNA topology or sequence.
One such interesting structure revealed is the discrete sites of gene
expression called transcription machineries [Sutherland and Bickmore,
2009] different from the nucleolus. The nucleolus is known to be
a large nuclear compartment devoted to constant transcription of
ribosomal RNA (see the seminal study of nucleolus RNA by Gall
and Pardue in [Gall and Pardue, 1969]), found in both widefield
and confocal images. The representation is globally the same at
microscopic level using SMLM even with a low number of frames, for
instance 5000. Nevertheless, when augmenting the number of frames
to 20,000, new structures start to emerge, of round shape, similar in
nature to nucleoli structure, but 50 times smaller (Figure 3.4, right
image, see also Figure 3.29). It is hypothesized that these are the sites
of machineries transcribing a cluster of genes. One can relate these
patterns to a possible low diversity of genes transcribed at single cell
level.
Figure 3.4: Sub-chromosomal regions
corresponding to potential transcription
machineries distinct from nucleolus are
revealed by SMLM with a high frame
number. On the left, a widefield image
of a sample. On the middle panel, a picture with 5000 frames. In both widefield
and low frame number, only two giant
nucleolus foci are visible. On the right,
a high number of frames reveals emerging structures corresponding to potential transcription factories (Figure and
caption modified from [Żurek-Biesiada
et al., 2015]).
3.2.2
Sub-chromosomal domains (scale: 500-1000 nm)
Conventional light microscopy (LM) has already shown some subchromosomal structures of chromatin [Mora-Bermúdez and Ellenberg,
2007, Dietzel et al., 2004], but due to the resolution limitations of LM
(∼ 200 nm), the precise organisation of chromatin during interphase is
barely described. Though transmission electron microscopy has been
spectacular at describing the nanoscale structure of cellular objects
[Haggis, 1992, Dehghani et al., 2005] (see comparison to high resolution microscopy in [Betzig et al., 2006]), superresolution methods have
structure, function and dynamics of chromatin
81
been a real advancement in localising individual molecules inside the
nucleus.
Here, SMLM was used to image the chromosome territories of
an interphase nucleus of a primate fibroblast (Figure 3.5). Magnification and manual segmentation of chromatin structure helps at
describing sub-regions of chromosome territories, as proposed in
[Cremer and Cremer, 2001]. Between different dense regions, one
can easily distinguish corridors, which are likely to give freedom for
transcription machinery to function (Figure 3.5, top inset), though
only staining for RNA Pol II or messenger RNA would allow to
confirm this hypothesis. The pattern observed follows the model of
chromosome territories described previously [Cremer and Cremer,
2001], with chromosome territories showing the shape of grapes, each
of the grains representing a sub-chromosomal domain (Figure 3.6A).
Figure 3.5: Inter chromatin compartments observed by SMLM (Figure and
caption modified from [Żurek-Biesiada
et al., 2015]). A Vero-B4 monkey kidney fibroblast-like cell stained with 500
nM Vybrant DyeCycle Violet is shown.
The blowup (top right) shows the interchromatin compartments, which possibly provide a path for mRNA to escape
into the cytoplasm. The blue nucleus
(bottom right panel) is the non-photo
converted widefield image. The periphery of the nucleolus is associated to heterochromatin (bottom inset). The signal
density is of approximately 5000 single
molecules/µm2 .
3.2.3
Chromatin domains (scale: 100-400 nm)
Recent proof of a universal chromatin domain morphology in mammalian systems
Genomics has recently generated exciting maps of chromosome
interactions within the nucleus, via chromosome capture technologies
(3C, HiC and others, see Figure 1.8). These maps have allowed to
identify structural regions called Topologically Associating Domains
(TAD). Functionally, this kind of structure comprises genes and remote
regulatory regions, and correlate strongly with replication domains
[Ibn-Salem et al., 2014, Pope et al., 2014]. Structurally, a TAD is an
element of about 185 kb in mammals, ranging from 100 kb to 1 Mb
[Rao et al., 2014]. Low-throughput experiments have been made to
validate these structures using DNA probes, however no study has
tried to give these genomic data a visual perspective. Thus, a detailed
82 the periodic and dynamic structure of chromatin
description of fundamental building blocks of chromatin, beyond
nucleosome and chromosome territory scale linking genomic data
to microscopy data is of immediate importance. For example, how
do one-dimensional genome browser or HiC plots (from sequencing
data) correlate to microscopy data? Say one observes 100 nm clusters
containing 40 signals of activating mark H3K4me3 on microscopy
images and finds several peaks on the same region in a ChIP-seq
dataset for H3K4me3 (in a similar cell line). How can one relate the
two kinds of data? Some models have been described, that attempt
to address this problem, but these are mostly theoretical without
experimental basis (Figure A.2).
Recent studies based on FISH and super-resolution microscopy
have estimated the sizes of clusters to be between 50-400 kbps and
around 100-500 nm spatially ([Prakash et al., 2015, Boettiger et al.,
2016], and Chapter 4 of this thesis). The regions probed were selected
based on combination of histone modifications, distinguishing active,
inactive and repressed regions. Overall, the studies provide evidence
that there is a functional fundamental building block of chromatin
around 100-400 nm confirmed by microscopy data (see Figure 3.6,
the different categories of active and repressed chromatin, red and
blue boxes in panel A, respectively). From the center of a domain to
the periphery, a gradient is observed, from highly dense chromatin,
which is likely compacted, to chromatin of low density, which is loose
and probably an active place of transcription (Figure 3.6B), white box).
In a different analysis, chromatin at the periphery of a domain was
associated to intense DNA transcription and stress-dependent gene
activity (see Section 3.4 for more details).
In a recent study [Boettiger et al., 2016], active, inactive and repressed chromatin domains were studied based on ChIP-seq data of
histone modifications from [Filion et al., 2010]. The authors chose
H3K4me2 (which marks both enhancers and promoters [Barski et al.,
2007, Wang et al., 2008]) as the mark to study active chromatin domains. H3K27me3 was chosen to mark the repressed and condensed
chromatin. It must be noted that H3K9me3 is the mark of the most
condensed chromatin while H3K27me3 often localises at the periphery of H3K9me3 and is slightly smaller in size than H3K9me3 clusters
[Prakash et al., 2015]. Nonetheless, in the study from [Boettiger et al.,
2016], the active regions based on H3K4me2 could roughly be divided
into two categories: short active regions, around 50 kb and of 100-200
nm of radius, and long active regions, around 100-200 kb and of
200-400 nm. Repressed regions based on H3K27me3 designated as
regions of most condensed chromatin (corresponding probably to the
early concept of ’heterochromatin’), were restricted to a 100-200 nm
size.
structure, function and dynamics of chromatin
83
SMLM identifies chromatin domains without prior knowledge about
the function or sequence
The data from [Boettiger et al., 2016] were generated using DNA
probes coupled to fluorochromes and directed toward specific DNA
sequences. Interphase chromatin using our SMLM setup that uses conventional DNA binding dyes or EdU coupled to a fluorochrome was
also explored. The image of interphase chromatin (Figure 3.6) identifies different chromosome territories, sub-chromosomal domains
(panel A, white broken lines), chromatin domains (panel B, white
stars), categories of chromatin domains (panel A, blue and red for
repressed and active domain, respectively), and loci with different
levels of compaction (inset A1 and A2, with denser region in the
center in blue and active region in the outer region in red).
Figure 3.6: SMLM reveals several levels of chromatin compaction, among
them functional chromatin domains: (A)
View of a single chromosome territory
already presented in Figure 3.5. White
broken lines delimit the different subchromosomal domains. The boxes show
chromatin domains of about 400 nm
inside each sub-chromosomal domain.
Red box: active domain. Blue: repressed
domain. White: intermediate pattern,
zoomed in insets A1 and A2. (A1) magnification of mildly compacted domain.
(A2) dissection of domain into repressed
chromatin (delimited by blue broken
line) and loose chromatin (red broken
line). (B) Same picture as A, with positions of putative functional chromatin
domains shown with white stars. Note
that each sub-chromosomal territory has
several chromatin domains.
In order to highlight the hierarchical levels of chromatin folding,
EdU was coupled to a fluorochrome, showing the positions of freshly
replicated DNA molecules within the nucleus. As the targets are theoretically unbiased, a complete view of chromatin distribution within
the nucleus was possible (Figure 3.7). Domains are found to be in the
range of 100 to 500 nm, as predicted by other methods (see Figure 3.16
for comparison, red rows). Moreover, magnification of the domains
reveals smaller highly identifiable spots, likely to be individual loci,
possibly nucleosomes (Figure 3.7A1 and Figure 3.7A2).
84 the periodic and dynamic structure of chromatin
Figure 3.7: A549 human lung cancer
cells were incubated with EdU (DNA
base pair precursor) for 15 min and fixed
with formaldehyde. Incorporated DNA
basepairs were detected using click-it reaction with Alexa 555 leading to staining
of freshly synthetised DNA. These samples provide a higher resolution with
SMLM as there is very little influence of
signals from the background.
Universality and functionality of chromatin domains
With respect to recent literature, chromatin domains have started
to emerge as the units of genetic regulation, with two important
characteristics:
1. Universality of the patterns: In HiC experiments, chromatin domains (the term used in genomics is topologically associating domain, or TAD) were found to be about 90% cell-type independent
(2763/3000 conserved boundaries between two assayed cell types
[Dixon et al., 2012]). The 100-400 nm scale is confirmed in recent
studies [Prakash et al., 2015, Boettiger et al., 2016], indicating a very
high consistency and possibly universality of chromatin domains.
However, one should be cautious in comparing the data, because
most of HiC data come from mammals while the work of [Boettiger
et al., 2016] was conducted in Drosophila. Moreover, the relevance
of the patterns found in HiC data is challenged by the low amount
of reads (poor sequencing depth) generated by this type of analysis
(Figure A.3).
2. Functionality of the patterns: Chromatin domains facilitate function by bringing genes and their regulatory elements closer to each
other, with an optimum at 50 kb distance [Doyle et al., 2014]. The
first comprehensible equivalence between genomic and physical
distance provided by [Boettiger et al., 2016] confirms this value. A
pure threading pattern of fibres would never allow these sorts of
patterns, which hints at folding for both active and inactive regions
of the genome at TAD, following models from [Rao et al., 2014].
In terms of functionality of chromatin domains, [Prakash et al.,
2015, Boettiger et al., 2016] show that these chromatin domains
come from epigenetically characterized regions. Thus, the results
presented in Figure 3.6 and Figure 3.7 further show the potential to
find different morphologies of compaction among these patterns.
structure, function and dynamics of chromatin
3.2.4
Chromatin fibres (scale: 30-100 nm)
Measurements of DNA is key to describe chromatin accurately. Describing a heterogeneous polymer as complex as chromatin requires
to identify its most fundamental elements. Early studies of chromatin
spectrum showed several peaks at 110, 55, 37, 27 and 22 Å, shown
to be in the direction of the axis of the helix (values highlighted in
[Finch and Klug, 1976]). Later this proved to be corresponding to a
beads-on-a-string pattern with a width of 8 nm, below an expected
11 nm diameter. The value of 11 nm was confirmed later (see [Luger
et al., 1997] for instance). First measurements of DNA and DNH
(nucleohistone) led to an early incorrect model of chromatin. The
molecules of DNA were thought to be in the center of the structure,
and the histones at the periphery, though diameter was correctly predicted to be 10 nm [Luzzati and Nicolaieff, 1963]. Later experiments
proved that the real structure follows a different configuration, with
DNA molecules wrapping around an octamer of 4 histone proteins.
In-vitro evidences from Olins [Olins and Olins, 1974] have described
strings of nucleosomes more or less packed and have provided solid
evidence for the existence of fibres, but this, only in in vitro studies.
Seminal study from Finch and Klug led to the description of a
higher compacted state of chromatin, that was named ’chromatin fibre’
with a width of 300 Å. Nevertheless, a close look at pictures reveals
that there is an intrinsic variability and that the width of elementary
chromatin size spans from 30 nm to 70 nm (Finch and Klug 1976).
Though Maeshima et al. [Maeshima et al., 2014] argue that this finding
is highly depending on salt concentration, these pictures are likely
to capture some elementary features of chromatin. The study also
showed that each transversal section of chromatin comprises about
4-7 nucleosomes, leaving a first estimation of correspondence between
distances in nm and kb. As found, 12 (visible) nucleosomes of about
200 bp each (if compaction is maximum) multiplied by a factor 2
(to take into account nucleosomes that may not visible), compaction
is estimated to be 5-fold. As will be seen later, that is largely an
underestimation. An earlier finding from Joseph Gall estimates the
diameter of a chromatin fibre around 40-60 nm [Gall, 1963].
Different patterns have been seen for chromatin fibres, with more or
less nucleosomes visible across transversal plane but this is explained
by different relaxation levels of the fibre, highly depending on environmental conditions [Bassett et al., 2009]. Opponents to chromatin
fibres have proposed a solenoid model [Dubochet et al., 1986], though
possible in theory, does not threaten the concept of fibre and the
possibility to build higher order building blocks. Enlightening study
from Woodcock and collaborators [Woodcock et al., 1984] has shown
how a fibre-like (or solenoid-like) chromatin of 30 nm diameter leads
to higher orders of chromatin folding. A very revealing image from
their report is reproduced here showing the transition between three
turns of the 30 nm chromatin fibre and a more compacted pattern
of about 100 nm diameter, a putative chromatin domain (Figure 3.8).
85
86 the periodic and dynamic structure of chromatin
It is a rare case of pattern showing transition between two levels of
chromatin compaction, giving a hint about the way structure builds
itself hierarchically.
This direct proof of evidence of hierarchy between the fibre and
chromosomal domains has not been highlighted in literature, to a
point that some researchers now question the existence of fibres.
Scarcity of evidence led some authors to recently dispute the concept
of chromatin fibre [Fussner et al., 2011], but the question remains
open as the article failed at disproving the notion and based its
assumption on a spectrum, which can be very well disturbed by the
multilevel complexity of chromatin. More evidences account for the
high conservation of a level of chromatin compaction in order 40-60
nm across cell types and species [Mayer et al., 2005, Harmon and
Sedat, 2005]. Generally accepted is the idea that fibres can display
several levels of compaction and diameter ranges between 30 to 60
nm (Figure 3.16).
Various groups have experimented new optical devices in order
to get insight into fundamentals of chromatin architecture. Most of
direct images of in vivo chromatin have been low resolution, and
so far only Scanning Electron Microscopy (SEM) and Atom Force
Microscopy (AFM) have been able to give a good idea of the actual
shape of chromatin. Spectacular images of chromatin from 3T3 cells
taken with SEM by Haggis [Haggis and Pawley, 1988, Haggis, 1992]
show chromatin architecture as a combination of bulky (’balls’) and
elongated regions connecting ’balls’ together, the ’linker’ domains (see
Figure 3.9). Measurements of the two categories reveal a configuration
of fibres compatible with previous findings of Joseph Gall [Gall, 1963]:
Figure 3.8: Transition between the 30 nm
chromatin fibre and a small chromatin
domain of 100 nm from [Woodcock et al.,
1984]. White bars show several turns of
the fibre, whose width can be appreciated with the scale to be around 30
nm, while the green domain is a complex folding of the chromatin fibre with
possible regulatory function, a putative
chromatin domain. Chromatin domains
are presented in Section 3.2.3.
• ’balls’ display an apparent size of 42 nm.
• ’linkers’ which bind the ’balls’ of length 100 nm on average. Width
is close to 40 nm.
The compaction here is not a thread, but a rather cluster on a string.
Each cluster or ’ball’ is a product of compaction of many nucleosomes
(size: 7 nm), possibly several hundreds, for about 50 kb of DNA
length. Besides, AFM has been used by Ushiki and Hoshi [Ushiki
and Hoshi, 2008] to image mitotic chromosomes. Authors find similar
’ball and stick’ patterns, whose sizes are approximately 50 to 60 nm
(see the comprehensive scale in [Hoshi and Ushiki, 2001] and figure
4c in [Ushiki and Hoshi, 2008]).
To summarize, history of chromatin research in the last decades has
demonstrated the existence of at least two levels of DNA compaction,
beyond nucleosomes. One is the 30 nm chromatin fibre, which has
been demonstrated beyond shadow of a doubt in several instances.
The picture can be a little refined by saying that this element is in
most cases not elongated, but rather shrinks to form higher order
structures, questioning the epithet ’fibre’. Nevertheless, in several
instances (relaxed interphase chromatin, pachytene loops), the term
fibre is highly appropriate. One should forget the picture of chromatin
as a thread of fibres and replace it by a more complex fibre, which
Figure 3.9: Differential topology of nuclear and cytoplasmic molecular set up
revealed by SEM. Cytoplasmic compartment (cyt) is made mostly from highly
elongated structures, the cytoskeleton,
while nuclear compartment (n) is a web
of threads (linkers) connecting bulky
regions (balls). Magnification x45,000.
From [Haggis, 1992].
structure, function and dynamics of chromatin
87
gives birth after folding to a second level of DNA compaction, which
is the building block of chromatin regulation (the chromatin domain,
whose size is of 100-400 nm).
Finally, I mention the accepted idea that chromatin fibres are one
level of ordered compaction of DNA during mitosis (see [Ushiki and
Hoshi, 2008] for nice AFM pictures showing both fibres and chromosome domains; see also Figure A.4). Beautiful imaging of condensin
proteins together with DNA have helped build one of the best models
of chromatin architecture proposed until today, with a condesin core
holding chromatin around in a regular pattern ([Maeshima et al.,
2014], Figure 6).
Figure 3.10: Bead on a string model.
Monkey kidney fibroblast-like cell
stained with Vybrant Violet to reveals
positions of chromatin with several levels of magnification, A, B, C and F. In the
magnified section C, it is interesting to
note the beads-on-a-string-like patterns
that are commonly seen in EM images
of 10 nm in vitro nucleosome structures,
which fit with the scale bar (200 nm in
the inset). The average density is around
3000-4000 molecules per µm2 . The localization image was acquired with a high
intensity laser excitation 491 nm, detecting more than 1300000 signals in an optical plane through section of approx. 500600 nm. The mean localisation precision
was 15 nm with a standard deviation of
4 nm. In D random distribution of C is
presented, in G random distribution of
F. In E and H the positions of the single
molecules are displayed.
3.2.5
A cluster-on-a-string model to describe the fibre/domain transition
In order to describe transition between fibres to chromatin domains,
one could imagine chromatin domains as clusters on a string. The
88 the periodic and dynamic structure of chromatin
chromatin fibres of 30 nm would be the string that link two or more
domains together. On Figure 3.9, one can see the white peanut-shape
regions as the clusters, while fibres could be the darker grey regions
between them. This model is one level higher compared to the beadon-a-string model in the hierarchy of chromatin compaction and is
a sort of intermediate chromatin organisation between the lowest
levels of compaction and chromosome or sub-chromosome territories.
Imaging of chromosome in meiosis shows that this type of structure
is maintained during meiosis and is potentially even more visible
on SMLM preparations of meiotic chromosomes if staining against
histone modifications such as H3K4me3 is used [Prakash et al., 2015].
3.2.6
Nucleosome domains (scale: 10-30 nm)
A number of studies have dealt with chromatin fibres at a scale of
30 nm and have tried to establish fibres as the basic structures of
chromatin both in inter- and metaphase [Frenster et al., 1963, Hay
and Revel, 1963]. A few studies [Davies, 1968, Davies and Small,
1968] have pointed out that chromatin is more or less distributed
randomly towards the interior (mostly euchromatin) while periodic
lumps of chromatin are attached to the nuclear periphery (mostly
heterochromatin). In the following section, I revisit this idea from
single molecule co-ordinate information and try to explore it further.
Figure 3.11: Inverse distance map of pixels based on the next 5 neighbours. The
distance of the center of a pixel i to the
next 5 locations is measured and averaged. 1/meandistance is then assigned
as a grey value to pixels i. The procedure is iterated over all pixels in the
image (even those where no fluorophore
resides in). The image is then scaled to
8 bit. Is the worm-like pattern a real
structure? The cell imaged is a HeLa
cell stained with Hoechst 33258.
Nearest-neighbour method identifies worm-like patterns
In order to study the structure of chromatin at the lowest order, I
analysed images of fibroblasts (Figure 3.10). On the magnified image,
the distribution of bead sizes shows an average diameter of 10 nm,
in line with the historical measurements of these structures [Oudet
et al., 1975, Olins and Olins, 2003]. In order to give more weight
to these beads and strings associations and validate the patterns, I
used blurring of nearest-neighbours. Applied to HeLa cells, this
technique resulted in finding worm-like patterns of roughly 5-10 nm
of periodicity (Figure 3.11), close to the magnitude of the theoretical
nucleosome size [Oudet et al., 1975]. On the magnification presented
in this figure, one can note some highly identifiable shapes in the
decondensed chromatin, which form circles and crosses.
Comparison of the worm-like pattern to simulated random patterns
To understand if the worm-like pattern or beads on the string
pattern is a real structure, I simulated the same type of data using a
stochastic distributed process to produce the locations. Surprisingly, I
could find similar and comparable worm-like structure with overlaid
density variation (Figure 3.12).
If the building block is real, the shapes created at scale of 100 nm
on this picture could be potential artefacts and the result of a random
superimposition of basic patterns. Though one cannot conclude on
the relevance of the shape of the chromatin patterns seen around 100
Figure 3.12: Comparison of chromatin
patterns in random and experimental
data. (Left) Simulation. An image was
generated with locations produced by
a stochastic gaussian distributed process. (Right) Measurement. Comparable worm-like structure with overlaid inverse distance variation are observable.
The cell is the same as the one on Figure 3.11. For the simulated image, the
density map is based on the list of localisations or single molecule coordinates.
For every pixel in the image the local
density is measured based on the mean
distance to the next 5 neighbours in the
list of localisations. The reciprocal value
of this mean distance is then mapped to
the corresponding pixel.
structure, function and dynamics of chromatin
89
nm, one can still observe more accurate patterns that can be seen at
higher scale by using nearest-neighbour methods.
Figure 3.13: Comparison of chromatin
patterns generated with random simulated data to experimental data when
different number of nearest-neighbours
are taken into account. Top block: Random case. Bottom block: SMLM Measurement. From left to right, influence
of the number of next neighbours is considered. In each block, the second row
shows the region highlighted with a red
box in the first row. In the case of the
random configuration, worm-like structures are visible at all scales. In the case
of the real data, the worm-like structure
vanishes when enough next neighbours
are taken into account. Same cell as in
Figure 3.11.
Nearest neighbour method helps to capture further chromatin complexity using combinations of building blocks
I then increased the number of nearest neighbours taken for calculation of mean distance and I observed that in the random data
worm-like structures were visible at all length scales. According to the
theoretical definition, random data is one which is not correlated and
therefore does not change under scaling. Interestingly, the worm-like
structures vanished when enough next neighbours were taken into
account (see Figure 3.13.).
While the low scale patterns found using nearest neighbour are
likely to correspond to topologically associating domains of low scale
of around 100 kb (’chromatin domain 2’), the higher order elements
defined here correspond to the ’chromatin domain 1’ and TAD of
about 1Mb (see Figure 3.16 for nomenclature). The former are around
100 per chromosome, while the latter are 10 per chromosome and
may be equivalent to active genomic regions of high gene content.
Quantification of chromatin condensation
I next quantified chromatin compaction in SMLM images by analyzing distances to 5, 10, 20, 50, 100 nearest neighbours (Figure 3.14).
90 the periodic and dynamic structure of chromatin
Considerable differences were observed between the experimental
data and randomly simulated data. First, the nearest neighbour distances (NND) in the experimental measurements were considerably
less than the random simulated when enough number of nearest
neighbour was taken into account. Significant differences started to
emerge using 20 to 100 nearest neighbours (65 nm vs 80 nm). This
fact was indicative of local condensation in the experimental data and
also provided the expected degree of condensation. Secondly, the
distribution of NND in the experimental started to broaden up as the
number of nearest neighbour molecule increased. The broadening
of peaks can be attributed to the fact that the sparser distributions
of chromatin might be represented at the boundary of condensed
chromatin, a common feature of chromatin organisation (please refer
to Figure 3.6).
Measurement
3.2.7
Simulation
Inference of further intermediate chromatin structures using local
chromatin density maps
Images of isolated molecules appear to be similar to the beads-on-astring-like patterns that are commonly seen in EM images of folded
chromatin in vitro (Figure 3.15). To further investigate this hypothesis,
the experimental data were compared with the random simulated
data. Comparison of the local inverse distance maps with the distance
maps generated by a random process shows that at a larger scale
there is a local condensation of the fluorophores whereas at a finer
scale the structure becomes random. In the case of real data, the large
structures are different from the structures that one observes at a finer
scale - which are ’wormlike’ shape. In random data, the worm-like
pattern is visible at all scales. This result shows that one cannot
resolve the fine substructure because there is none. The molecule
signals are real, but the patterns at low scale are not meaningful. Only
at a larger scale structure becomes dominant. Hence, organisation
of chromatin is clustered at large scales but distributed randomly at
fine scales. If our result is confirmed in the future by complementary
experiments, our method will prove competence at detecting beadson-a-string patterns in situ, which is not possible with other current
Figure 3.14: Quantification of chromatin
condensation. Distances to 5, 10, 20, 50,
100 nearest neighbours are compared
between measurements and simulations.
While the distribution of 5 and 10 nearest neighbours are similar between simulation and measurement, there is a significant drop in neighbouring molecules
when 20 nearest neighbours are taken
into account. For 100 nearest neigbours,
a big drop is observable with a very
broad distribution, as an indication of
over smoothing. The line inside each
coloured box-plot in the graph indicates
the median, ’+’ denotes the mean, the
boxes delimit the 25% and 75% quartiles,
while the whiskers show bounding of
9% and 91% of the data.
structure, function and dynamics of chromatin
91
methods so far.
Figure 3.15: Inference of pattern from
local chromatin density maps. (Left)
A composite image shows the impact
of blurring and condensation enhancement when different number of nearest
neighbour (NN) are taken into account.
(Right) Same cell with 5 nearest neighbour blurring. The far left block in both
images (5 NN) are identical. Scale bar:
1µm.
3.2.8
Hierarchical organisation of chromatin structure
I finally summarize information from literature and personal findings
regarding chromatin architecture and its different levels of hierarchical
compaction in Figure 3.16. The table shows the hierarchy of chromatin
organisation from nucleosome to chromosome territory, including
the building block of a size around 50 nm. In Section 3.2.6, a 50
nm pattern is visible with superresolution microscopy performed on
interphase cells. Similar configurations are further shown in stem cells
and confirm the existence of functional chromatin clusters around
50-100 nm using the meiotic chromosome as a model, and histone
modifications H3K27me3 and H3K4me3 as targets, instead of DNA
itself (see the Chapter 4 on meiotic chromatin).
3.3
The dynamics of chromatin
Only recently, gene regulation has started to be seen as the result of
chromatin dynamics inside of the cell nucleus. A way for chromatin to
be dynamically regulated is to switch between heterochromatin and
euchromatin states. Electron microscopy, though of high resolution,
is poor at identifying regions of low density, such as euchromatin
[Cremer et al., 2006], though usage of an expended protocol for
Giemsa staining has been shown to turn euchromatin denser and so
reveal chromosome territories [Stack et al., 1977]. In any case, for a
long time, the description of chromatin was confined to distinguish
dense and clear regions as compact chromatin (heterochromatin) and
relaxed chromatin (euchromatin) respectively, the latter being the
place of transcription. Besides, other nuclear components such as the
nucleolus were also identified.
92 the periodic and dynamic structure of chromatin
Figure 3.16: Different milestones of
chromatin architecture research, with
color classification related to scale investigated. Yellow region at the top
shows milestones toward discovery of
chromatin fibre architecture and similar
scale fibre patterns (30-60nm). Red region shows main data regarding sizes
of chromatin domains (range 100-500
nm). Orange region shows a remarkable
study of the transition between the two
kinds of chromatin architecture that are
fibres and domains. Finally, blue region
at the bottom lists the studies which epitomize higher order compaction level of
chromatin, including chromosome territories.
For historical purpose, I have showed in the first chapter a photograph of a classical euchromatin/heterochromatin dichotomy, generated from a human leukocyte (Figure 1.5). A characteristic very
thin layer of heterochromatin is present close to the nuclear envelope, while most of the center of the nucleus is the place of a loose
chromatin (white arrows). The nucleolus, which is the place of active transcription of ribosomal RNA, is the large dense cluster in the
middle. Some stem cells show a contrasting phenotype with euchromatin in the exterior and hetero-chromatin toward the interior. This
is discussed in the following section.
3.3.1
Contrasting arrangement of eu- and hetero-chromatin inside the
cell nucleus
Previously, chromatin was broadly divided into classical bimodal
model i.e. euchromatin and heterochromatin regions. Euchromatin
is the open and transcriptionally active compartment of chromatin,
mostly found in the interior of the cell, and mostly associated with
histone acetylation and certain histone variants. Heterochromatin is
condensed and is mostly found at the nuclear and nucleolar periphery.
It is also usually referred to as ’silent DNA’. Repressive histone marks
and hyper-methylated DNA are characteristics of heterochromatin.
structure, function and dynamics of chromatin
93
Figure 3.17: Contrasting phenotype of
interphase chromatin. Different cell
lines can display very different chromatin phenotypes. The standard belief
(A) is that dense chromatin sits at the nuclear periphery while sparse chromatin
is located in the interior (here, a monkey
fibroblast cell). In (B) one can observe
that in Mesenchymal Stem Cells (MSC,
mouse), the dense chromatin is absent
at the nuclear periphery but is rather
found in the interior of the nucleus, in
places commonly known as chromocenters. These are rich in H3K9me3, a mark
of centromeric chromatin [Smeets et al.,
2014, Cremer et al., 2015].
A comparison of fibroblast cells and mesenchymal stem cells shows
that this textbook picture is rather the characteristic of the pattern
observed in the fibroblast. Images show a clear contrast between
compact and relaxed chromatin, as well as a predominant nucleolus (Figure 3.17A, monkey fibroblast). On the opposite side, stem
cells have a broader spectrum of gene expression and show several
foci of heterochromatin in the middle, possibly by migration of heterochromatin away from the lamina (Figure 3.17B), from a mouse
mesenchymal stem cell).
3.3.2
Classifier identifies intermediate states between eu- and heterochromatin regions in differentiated cells
I developed a new method called Pixel Density Classifier (PDC) to
classify localisation microscopy data into regions with different spatial
densities and structural features. I applied PDC to select different
levels of chromatin compaction on SMLM images (Figure 3.18). Setting the parameter to three different levels of chromatin compaction,
I could define three compartments with different densities: one is
associated to a loosened state of chromatin, present in the nucleus as
large foci (Figure 3.18C)and corresponding most likely to transcription machinery compartments (which bare chromatin speckles, place
of splicing machinery; see similar patterns stained with antibodies
against speckles on Figure 6 in [Cremer and Cremer, 2001]). The
second degree of compaction (Figure 3.18D) is the surrounding area
of these foci; this category has the largest distribution of the three
and mostly corresponds to euchromatin, that is, the part of chromatin
deployed out of the chromosome to be transcribed. Finally, dense
chromatin (heterochromatin) is found at the outer part of the nucleus
(Figure 3.18E).
3.3.3
Chromatin dynamics during differentiation of mesenchymal stem
cells
In order to study dynamics of chromatin, I present an example in
Figure 3.19 which shows two images of differentiated and undif-
94 the periodic and dynamic structure of chromatin
Figure 3.18: Pixel density classification
of chromatin. (A) Monkey fibroblast
cell in SMLM using photoconversion of
DNA dyes. (B) Decomposition of chromatin into three density states (pixel
density of 2, 5 and 9 molecules/100
nm2 for blue, green and red regions
respectively). (C-E) Image decomposition shows at least three different status
of chromatin of increasing compaction,
from C to E. Intermediate levels are actually the most abundant (D) and cannot be classified strictly into eu- or heterochromatin state. (C1-E1) Magnifications of C-E.
ferentiated mysenchymal stem cells. From this data, one can draw
scenarios regarding the timing of chromatin dynamics during differentiation, even if refined synchronization is needed to capture the
entire dynamics.
Step captured in Figure 3.19A slightly precedes Figure 3.19B. The
chromatin has almost the same level of compaction. Nevertheless,
binning the level of compaction into three levels shows that first figure
displays elongated patterns of chromatin of intermediate compaction
level, forming threads, showing the progression of chromatin towards
condensation in the middle. Figure 3.19B is a slightly later time point,
when the compaction is maximum.
Such particular patterns have been also observed previously in
[Popken et al., 2015]. The intermediate stages of the stem cells associated with central heterochromatin islands are reported. Differences
between ESC versus differentiated cells chromatin status have been
also exemplified in [Mikkelsen et al., 2007], confirming the model.
structure, function and dynamics of chromatin
95
Figure 3.19: Compaction increases during early differentiation of mesenchymal
stem cells. (A) Undifferentiated cell. (B)
Differentiated cell. (A1 and B1) show the
different categories of chromatin compaction with number of categories set
to three. Red is the most condensed
chromatin, blue the least condensed and
green is an intermediate level.
3.3.4
Dynamics of chromatin upon stress
Stress as a good model to study chromatin dynamics1
In order to capture the changes that chromatin can experience
during a dynamic process, cardiomyocytes under a cellular stress
mimicking infarctus, ischemia, were imaged using SMLM. Stress
has been shown to remodel chromatin [Johnson and Barton, 2007,
Hargreaves and Crabtree, 2011] and seems to be a good source to
study dynamics of chromatin. Extreme patterns of stress highlighting
specific ring-shape patterns, likely for mechanistic protection, were
previously reported in [Everid et al., 1970]. This study shows compaction patterns of chromatin following citrate treatment of chicken
erythrocytes.
In the present study, the effect of environment on chromatin nanostructure with SMLM and 3D structured illumination microscopy
(SIM) was evaluated. Following a short-term oxygen and nutrient
deprivation (OND) of the cardiomyocyte cell-line HL-1, chromatin
architecture experiences a dramatic changes, adapting to large ring
patterns, of 1 to 5 µm in diameter, with sparse voids inside. These
rings most likely correspond to individual chromosomes, hinting
at an intense chromatin compaction. The findings show a dramatic
adaptation of chromatin to environment, probably due to a major
change of energy status.
Distinct changes of nuclear morphology upon stress
In order to mimic ischemia, SMLM was employed on HL-1 cells
transiently exposed to OND. OND procedure was performed by
collaborators as follows: cells were put in a hypoxia chamber (Whitley
This section is partially based on
[Kirmes et al., 2015].
1
96 the periodic and dynamic structure of chromatin
Figure 3.20: Chromatin compaction
upon stress, stained with YOYO-1 (a-f)
or Edu (g-l) and imaged via SMLM (figure and caption modified from [Kirmes
et al., 2015]). YOYO-1 staining. a-c: untreated cells; d-f: cell subjected to one
hour of OND. g-l: 24 hours of 10 µM
EdU labeling. g-i: untreated cells; j-l:
after one hour of OND. After fixation,
EdU was fixed to AlexaFluor 488 via
click chemistry in situ (see Zessin et al.
2012 for method). Regions devoided of
chromatin are indicated by an asterisk,
atolls are indicated by a white arrow. c,
f, i and l: wide-field images corresponding to b, e, h and k, respectively.
Hypoxystation, see [Kirmes et al., 2015] for details) during one hour
without any nutrient (ion solution with no organic molecules), to then
recover by restoration of normal oxygen concentration and nutrients
(Claycomb media). Localization maps were generated in minimum
nine replicates for each condition, by integrating 30,000 observations
capturing photons re-emitted by the samples after 50 ms of exposure.
The images generated spots at the positions of molecules under the
diffraction limit, at a structural resolution of 103 nm, which is actually
above the theoretical resolution of 67 nm that can be expected.
HL-1 cells treated one hour with OND displayed a major change
of their nuclear architecture. Chromosomes shifted from classical
interphase territory distribution to a characteristic ring-shape. Most of
the chromosomes are attached to one another, mostly at the periphery
of the nucleus (Figure 3.20d,j). Certain chromosomes are bound to the
lamina, with elongated patterns nicely exemplified by 3D surface plot
in Figure P.8. Almost all chromosomes are bound together and form
a sort of continuous chain. No part of chromatin shows the diffuse
chromatin pattern observed in untreated condition.
3.3.5
Reversible compaction of chromatin under stress
Next, a time course of chromatin organisation upon hypoxia (Figure 3.22) was evaluated. Stress induced posture happens within the
one hour of induction, while relaxation takes up to four hours to
reach complete normal state (Figure 3.22). The typical stress induced
pattern is already vanished after 5 minutes, leaving the cell with a few
caveats that disappear totally after 4 hour of relaxation. Additional
images of the time course can be found in appendix (Figure A.5).
This pattern has been related by collaborators to H3K14ac signal
during the same hypoxia process; result from confocal imaging shows
that H3K14ac is reversibly vanishing during hypoxia (Figure 3.27
and [Kirmes et al., 2015], Supplementary Figure 4). This result hints
Figure 3.21: 3D surface plot of chromatin nanostructure under ischemia
conditions. (A) Untreated cell. (B) Cell
subjected to 1 hr of OND. The height
represents the density of signal in the
reconstructed image.
structure, function and dynamics of chromatin
97
Figure 3.22: Reversible compaction of
chromatin (Figure and caption modified
from [Kirmes et al., 2015]). Representative SMLM images of Vybrant Dyecycle
Violet-stained nuclei, either untreated,
subjected to 1 hour of OND or 5, 15, 60
and 240 minutes after release from OND
are shown to demonstrate the reversible
compaction of chromatin.
at the restoration of the normal phenotype of the cell. Overall, the
patterns observed upon stress demonstrate a protective rearrangement mechanism of chromosomes, associated with a reduction of
transcription-related histone modifications.
Nearest neighbour analysis to describe the extent of chromatin reversibility
B
A
Untreated
OND 1hr
Reperfusion 5 min
20 nearest
neighbours
50 nearest
neighbours
100 nearest
neighbours
200 nearest
neighbours
Reperfusion 15 min
300 nearest
neighbours
Reperfusion 60 min
400 nearest
neighbours
Reperfusion 240 min
500 nearest
neighbours
I then tried to describe comprehensively the dynamics of chromatin
and its reversible characteristic by computing the distance distribution
of the nearest neighbour molecules to regions of interest (ROI) and
see how molecules are forming clusters during stress. I find that
Figure 3.23: Nearest neighbour characterization to describe the extent of chromatin reversibility. (A) 500 next nearest
molecules were used to describe the extent and reversibilty of chromatin compaction upon one hour of OND. (B) The
effect of different numbers of nearest
neighbours (20, 50, 100, 200, 300, 400,
500) was evaluated on the analysed position shown in (A) for OND 1h condition.
X-axis: unit is the nanometer.
98 the periodic and dynamic structure of chromatin
molecules clustering upon 1 hr of hypoxia tend to be kept away
to each other at approximately 100 nm (Figure 3.23). The normal
distribution is broader, showing a more diffuse pattern of chromatin,
while the OND-induced condition is narrower and shows further
distance, describing the observed clusters. The pattern shifts back to
normal upon return to normal environmental conditions, with nearest
neighbour distance showing broader distribution again, describing
the diffuse pattern (Figure 3.23, 240 minutes).
To parametrise the method, I had previously tested the influence
of the number of nearest neighbours to calculate the average distance
to neighbours (Figure 3.23B). In the case of 1h OND, both median
distance and breadth of distribution increase with the number of
nearest neighbours, showing that low sampling favours neighbours
from the same cluster while high sampling captures more complexity
and more distant patterns.
I further plotted the distance to nearest neighbours for the different
conditions (Figure 3.24). Interestingly, patterns show that the distribution of distances tends to have an exaggerated normal phenotype
upon 4 hr of relaxation, with a mean distance lower for 4h relaxation
than normal condition, and a narrower peak. This result hints at an
even lower amount of caveats in the cell under relaxation, caused by
an even higher diffusion than for normal state (as the distribution is
shifted to the left in Figure 3.23A), associated potentially with small
clustering difficult to observe by sight, of slightly granular shape (as
the distribution is narrower, a characteristic of clustering behaviour
seen in Figure 3.23A).
Such a pattern may be necessary for the cell to recover from damage, or is a sign of permanent DNA damage. Finally, the recovering
pattern after 4h of relaxation may account for an anarchical distribution of chromatin, resulting in a random pattern of signal position.
Figure 3.24: Mean nearest neighbour distance characterisation for different moments of the OND stress experiment.
The relationship between the number
of nearest neighbours used in the analysis to the mean distance to the nearest
neighbour molecules is shown for untreated, 1 hour OND or 5, 15, 60 and 240
minutes relaxation. It is observed that
chromatin adapts to a more closed configuration after 240 min of relaxation.
Resolution estimation of DNA data
Monitoring the resolution for the different steps of the experimental
procedure shows that resolution is better for the normal and post
OND-induced conditions, while the OND states are more difficult
to resolve (Figure A.6), probably due to condensation and formation
of highly dense chromatin. This can come from a poor localisation
precision of the signals in the case of the OND state (Figure 3.25).
3.3.6
Conclusion
I have shown in this section that our super-resolution setup is able
to capture the dynamics of chromatin upon biological processes,
using hypoxia as model. Such observations would be very difficult
to observe using lower resolution methodologies as exemplified by
comparisons of our results with widefield images and confocal images.
Dynamics of chromatin under stress have shown exceptional patterns
rarely described before (see early study by [Everid et al., 1970] on
Figure 3.25: Localisation precision of untreated cells, subjected to 1 hour of OND
or 5, 15, 60 and 240 minutes after release
from OND.
structure, function and dynamics of chromatin
influence of citrate on chicken erythrocytes, which created foci of
stress in chromatin). Our set up can also be used for biomedicine in
order to further describe the consequence of ischemia for myocardiac
cells.
Regarding cellular differentiation, chromatin dynamics is a drastic
change in gene expression management. Differentiated cells show
clear contrast of compact and relaxed chromatin, with an active
region corresponding to the nucleolus which shows that translation
is preponderant compared to transcription, leading to betting all
nuclear energy on transcribing rRNA (Figure 3.17A from an adult
mouse fibroblast). On the opposite side, stem cells have a broader
spectrum of gene expression and show several foci of heterochromatin
in the middle, by migration of heterochromatin away from the lamina
(right picture on Figure 3.19, from a mouse mesenchymal stem cell).
From these basic morphologies, one can hypothesize that some
kinds of cells do not need to transcribe a lot of genes, but ribosomal
RNA; while cells with a stem capacity have to transcribe many specialized genes associated to differentiation and maximize transcription by
migrating their dense regions toward the center, to increase the available space for genes to deploy their transcription machinery. Similar
patterns have been also observed previously [Popken et al., 2015]. In
this article, authors report intermediate stages of cells differentiation
associated to central heterochromatin islands.
Findings presented in this chapter show both that cells have an
intrinsic capacity to condense chromatin (the basis of all developmental processes) and that a cell about to die maximizes its protection
mechanism by showing an extreme condensation phenotype. I do not
claim here that the model of compaction holds true for every kind of
cells, but that it could rather be a specific properties of certain stem
cells which have a high potential of compaction (compactibitily) as
the genome is not completely folded yet and can be assumed in a
’naive’ state.
A model for chromatin dynamics
In the analysis of the differentiation of mesenchymal stem cells
and more importantly of the reaction to hypoxia of HL-1 cells, a
similar pattern is found: reprogramming of the cell, whether it is for
enhancing gene expression at genome scale, or condensing chromatin
for protecting DNA integrity, requires condensation or reshuffling in
the center of the nucleus. DNA is still present at the lamina, though
in little amount compared to classical euchromatin/heterochromatin
dichotomy; in the case of mesenchymal stem cells, mostly all over the
lamina, while in the case of HL-1 cells, at certain foci which possibly
play the role of anchors. In a situation of massive gene regulation as
early differentiation events, a big part of the genome can be bound to
lamina and so be silenced. The situation for highly stressed cells is
different: the entire chromatin has to be packed. In order to facilitate
the process, the amount of chromatin attached to the lamina, a process
99
100
the periodic and dynamic structure of chromatin
necessary for maintaining coherence of chromatin, has to be as little
as possible, so the packing can be done on most of the chromatin.
3.4
The function of chromatin
Chromatin was initially defined as a nuclear material composed of
histone proteins [Kossel, 1884], other proteins [Mirsky and Ris, 1951]
and possibly RNA [Busch, 1974]. In order to refine the description
of chromatin structure and correlate structure to function, information from histone modification and other nuclear components were
employed. This information helped to further identify compartments
in SMLM images, something which was not possible with EM previously [Haggis, 1992]. In this section, different labelling strategies
are used in order to understand how the dynamic architecture of
chromatin relates to function. Importantly, I show that the periphery
of chromatin domains (see Section 3.2.3 for a thorough description
of these domains) is associated with active transcription and related
enhanced replication activity.
Figure 3.26: (A) DNA synthesis occuring at the periphery of chromatin domains in a monkey kidney fibroblast
cell. Purple channel shows chromatin
and green regions of newly synthesized
DNA. White represents the regions overlapping. (A1) zoomed-in section corresponding to yellow box in (A). The vast
majority of EdU signals anti-colocalize
with chromatin and is observed at the
periphery of chromatin domains (B).
Scale bar in A is 1000 nm while in A1 is
500 nm.
structure, function and dynamics of chromatin
3.4.1
101
Periphery of chromatin domains is associated with high DNA synthesis
Early replication is associated to regions of high transcription [Boulos
et al., 2015]. To further describe chromatin domains, I decided to test
where replication sites localise in the nucleus. If the outer part of
chromatin domains is the place of more active transcription, then replication sites should also localise at the periphery of the domains. To
investigate this hypothesis, sites of DNA replication were targeted by
incorporating EdU, an DNA-basepair analogue (Figure 3.26). Alexa
647 was conjugated with Cu(I)-induced ’click chemistry’. Intensity
of pixels coming from DNA or EdU signals were recorded on an
homogeneous section of the preparation. It is observed that in newly
synthesized DNA sites, the signal locates at the periphery of the chromatin domains (chromatin domain 1 or sub-chromosomal domains, of
about 1 µm, in the classification presented in Figure 3.16), confirming
that this region of the nucleus is the site of new synthesised DNA
and active transcription, while the center of chromatin domains is a
place of repressed transcription, due to condensation of chromatin.
Figure 3.27: OND induces compaction
of chromatin to mechanistically reduce
transcription (Figure and caption modified from [Kirmes et al., 2015]). HL-1
cells were subjected to both anti-H3K14
immunostaining and counter-staining
with Vybrant DyeCycle Violet. Signals
were analysed using a SMLM set up. ab: untreated cells; d-e: cells after one
hour of OND stress. b-e: magnifications
of white boxes in a and d, respectively.
c-f: wide-field images corresponding to
b and e, respectively. Regions deprived
of chromatin are highlighted with an asterisk and atolls are shown with a white
arrow.
102
3.4.2
the periodic and dynamic structure of chromatin
Stress-dependent transcription at the periphery of chromatin domains
As previously discussed chromatin can protect itself via contraction
(Section 3.3.4). HL-1 cells were transiently exposed to OND and
stained for both DNA and H3K14ac, a histone modification known
to be associated to active promoter state [Karmodiya et al., 2012].
Cells were fixed, immunostained with AlexaFluor647 conjugated antiH3K14ac antibody and counter-stained with Vybrant DyeCycle Violet
[Żurek-Biesiada et al., 2015]. The dynamics of histone modification
H3K14ac was monitored during stress. Result shows that this mark
experiences a great regression (Figure 3.27).
Resting state of the nucleus is associated with a classical diffuse
pattern of chromatin, with denser regions observable close to the
nuclear periphery (Figure 3.27). The threading network is complex
and is interspaced by small inter-nuclear compartments. Staining for
an active mark H3K14ac shows punctuate distribution throughout
the nucleus (Figure 3.27). Foci of H3K14ac are found to be present at
the edge of chromatin regions (Figure 3.27b) and in very few places
inside of the mass, a result in agreement with theory on nuclear compartmentalization, which identifies inter-chromatin regions as places
of transcription and transport of messenger RNAs [Cremer et al.,
2015]. Wide-field images are not sufficient to show the tight relationship between chromatin and histone modifications (Figure 3.27c). As
H3K14ac is related to active promoters, it is hypothesized that the cells
imaged here are subjected to a high reduction of transcription. This
clearly advocates for a physical mechanism of self-protection of chromatin by contraction, which leads to the shut-down of transcription
in order to preserve the physical integrity of the DNA.
3.4.3
Histone modifications allow to further dissect chromatin into active
and inactive domains
In order to gain insight into the epigenetic landscape of the interphase
nucleus, and study how its structure may be regulated by histone
modifications, several histone modifications were stained (Figure 3.28).
A fluorophore density per 100 nm2 of 0.0125 for H3K9me3; 0.0119 for
H3K27me3 and 0.0089 for H3K14ac was found ( Figure 3.28). Globally, this identifies centromeric mark (H3K9me3) as a mark of high
density, active chromatin mark (H3K14ac) as the least dense, and the
repressive chromatin mark (H3K27me3) as having the best balance
between strength of signal and overall repartition. Comparison between H3K14ac and H3K27me3 marks reveals that the former has a
distribution shifted to the left (Figure 3.28, bottom left image). This
shows that the H3K27me3 clusters are broader than the H3K14ac
and so are less dispersed. Overall, these findings confirm the highest
density of chromatin associated to repressive histone marks.
structure, function and dynamics of chromatin
103
Figure 3.28: Various histone modifications identify different functional compartments of the nucleus. The figure
aside describes the distribution of various post-tranlational histone modifications in interphase. Often histone modifications have distinct molecular signature suggesting different functional roles
in gene regulatory mechanisms. While
H3K9me3 is mostly enriched at chromocenters as expected, H3K27me3 is distributed more in a speckle-like patterns.
The acetylation patterns (H3K14ac) vary
in density and are more diverse in distribution. For functional histone modification patterns during meiosis, see Chapter 4. The nearest neighbour plot gives a
measure of the spread of 20 closest signals. Localisation precision plots show
that the different sizes of the clusters are
not due to the fluorophore i.e. the fluorophore behaves the same in all three
cases in terms of photon counts.
3.4.4
SMLM identifies potential sites of transcription machineries in the
mammalian nucleus
Transcription factories are self organising discrete sites where polymerases and other transcription factors are concentrated [Cook, 2010].
Similarly to Section 3.2.1, transcription machineries are identified
using SMLM. As shown in Figure 3.29, while nucleoli are visible with
a low number of signals, certain central structures start to appear
when enough number of signals are accumulated (white arrows). The
organisation of the structures is thought to be an active region of
transcription because their round shape is analogical to nucleoli and
they are likely to be a collection center of chromatin coming from
several chromosomes, similar to the territories/inter-territories model
presented in Section 3.2.2.
3.5
Summary and discussion
SMLM was successfully applied to conventional DNA dyes (Hoechst
and DAPI) to describe different levels of chromatin organisation, from
nucleosome to chromosome territories. High labelling density (more
than 107 signals per 500-600 nm optical section) of these minor groove
binding DNA dyes helped to infer structural features and density
variations in chromatin organisation at the nanoscale. It was found
that chromatin is clustered at large scales but distributed randomly at
the finest scale. At a lower order of organisation, statistical analysis
of chromatin distribution revealed distinct differences from a random
density distribution. At an intermediate order, 100-400 nm domains
emerged as the functional building blocks of chromatin architecture,
with differential condensation states. The outer side of these domains
being less condensed as compared to the more condensed inner
domains.
Figure 3.29: Possible functional compartments emerge from increasing the
number of signals on a given figure.
Left panel: 100,000 signals, right panel:
1,000,000 signals. White arrows: structures that are not visible with 100,000
signals but are with 1,000,000 signals.
104
the periodic and dynamic structure of chromatin
Furthermore, some examples of dynamic processes that chromatin
undergoes were shown, with special emphasis on stress. The compaction of chromatin observed upon stress conditions indicates that it
might be experiencing a phase separation process. At its core ’phase
separation’ means ’like dissolves like and unlike separates/phases
out’. For example, GC-rich might separate out from AT-rich regions.
From a polymer physics perspective, monomers of the same species
phase separate from monomers of a different species. At a chemical
level, a polymer composed of hydrophilic and hydrophobic components may arrange itself differentially in order to maximise or
minimise interactions with the surroundings medium. In OND conditions, one observes large and dense ring/rod like structures. It is
known that most AT-rich region are heterochromatic due to the presence of 2 hydrogen bonds (less active) and that GC-rich regions are
euchromatic due to the presence of 3 hydrogen bonds (more active).
Since Hoechst and DAPI have high specificity for AT-rich sequences,
the GC-rich block of the chromosomes phase separates to form non
overlapping micelles [Ostashevsky, 1998]. Staining of DNA using
different oligonucleotide probes specific for simple repetitive DNA
sequences might provide further clues.
Finally, using multicolour staining, the structure-to-function relationship of chromatin was studied with focus on metrics that could
hint at chromatin activity. The active chromatin regions were on
exterior and in decondensed regions (Section 3.4.1 and 3.4.2.) while
the inactive regions in the interior of the domain (Section 3.4.3), with
an increase of chromatin density. EdU and several histone modifications helped in identifying novel chromatin compartments. Histone
modifications associated with active chromatin were always on the
periphery of a chromatin while histone modification associated with
repressed chromatin were more toward the interior of a chromatin domain. Also, several new orders of chromatin domains were reported
in this chapter.
Overall, it is hoped that the data presented in this chapter will
help acknowledging that chromatin folds at various length scales and
that domains in the range 100-400 nm constitute the fundamental
functional elements of nucleus organisation. Focus on understanding
the biology of these domains will be of great help to understand the
mechanisms behind chromatin folding and organisation.
Only a structure built on profound symmetry and periodicity can possibly fit 3 meters of
DNA inside 10 micron wide nucleus.
4 Periodic and Symmetric Organisation of Meiotic Chromosomes
In this chapter1 , single molecule localization microscopy (SMLM) is
combined with analytical tools to describe the chromatin organisation
of the pachytene chromosomes. DNA is found to be non-randomly
distributed along the length of chromosome proteinaceous backbone,
the synaptonemal complex (SC), in condensed clusters. Furthermore, chromatin is organized in spatially distinct functional clusters
associated to specific epigenetic marks (Figure 4.1). For functional
characterization, various post-translational histone modifications were
selected based on information from genomic data and the three following chromatin compartments were identified in the pachytene
chromosome:
1. Radial chromatin identified by trimethylation of histone H3 at lysine
4 (H3K4me3) — indicative of actively transcribed chromatin.
2. Polar chromatin identified by trimethylation of histone H3 at lysine
9 (H3K9me3) — indicative of centromeric chromatin.
3. Tangential chromatin identified by trimethylation of histone H3 at
lysine 27 (H3K27me3) — indicative of repressed chromatin. This
compartment is remarkably associated to the backbone of the
chromosome, showing a putative implication in the regulation of
recombination.
Periodic clusters of H3K27me3 are found at 500 nm intervals along
the SC, while H3K9me3 mark is associated with a large and dense
cluster at one of the ends of SC. H3K4me3 is arranged in a radial hairlike loop pattern emerging laterally from the SC, observed for the first
time in mammals, especially in context of post-translational histone
modifications. For the first time, SMLM is combined with immunostainings and DNA stains to map the global epigenetic landscape of
pachytene chromosomes. In this study, single molecule localisation
microscopy protocol was optimised to combine direct staining of SC
proteins either with DNA molecules or modified histones. This work
can be used as a rational to explore the chromatin organization of any
cell nucleus, giving both an overview of DNA molecules and refine
the picture with meaningful functional annotations.
Finally, taking into account the arrangement and composition
of chromatin, as well as the region-specific distribution of post-
Parts of this chapter, including figures
and captions, have been published as
[Prakash et al., 2015, Prakash, 2012]
1
106
the periodic and dynamic structure of chromatin
Figure 4.1: A model for spatial distribution of chromatin at pachytene
stage of meiosis prophase I (figure and
caption modified from [Prakash et al.,
2015]): Based on localisation maps of
post-translational histone modifications,
I can dissect the meiotic chromosome
structure into at-least three distinct morphologies, highlighted by the differential, nanoscale organization: (1) Radial
chromatin identified by trimethylation
of histone H3 at lysine 4 (H3K4me3):
indicative of actively transcribed chromatin, (2) Tangential chromatin identified by trimethylation of histone H3 at
lysine 27 (H3K27me3): indicative of repressed chromatin and (3) Polar chromatin identified by trimethylation of histone H3 at lysine 9 (H3K9me3): indicative of centromeric chromatin. Image
courtesy: David Fournier.
translational histone modifications, a model of the chromatin architecture along the SC is discussed in this chapter. It is tempting
to speculate on a possible super-coiling of meiotic chromosomes,
not seen in mitosis or interphase, possibly needed to mechanically
simulate chromosome pairing and recombination.
4.1
Introduction
Meiosis is an essential event in the life cycle of sexual organisms, as it
creates new genetic combinations at each generation. The two consecutive major events that participate in generating genetic diversity
during meiosis are chromosome segregation and genetic recombination. Chromosome segregation is the sorting of each chromosome of a
pair into a different two daughter cell, an event happening in the two
cell divisions of meiosis, which results in a major shuffling of alleles.
Recombination also participates to diversity by exchanging sequences
of 1-2kb of two homologous chromosomes during an event called
crossing-over. This event is more likely to generate innovation as the
portion involved are smaller and can potentially edit gene scaffolds,
regulatory intergenic regions, or non-coding genes. The places of
recombination are not random and are heavily constrained by DNA
sequence (mainly GC content, see [Clément, 2012]) and chromosome
structure.
Figure 4.2 (inspired by [Lichten, 2001, Baudat et al., 2013] and other
studies mentioned below) details the mammalian meiotic process,
preceded by DNA duplication, as described previously in mouse. Particularity of mouse chromosomes compared to human is the location
of the centromere at one telomere, resulting in one-arm chromosomes
exclusively. During interphase, chromosomes are complex objects
consisting of one chromatid and the telomeric centromere (Figure 4.2
periodic and symmetric organisation of meiotic chromosomes
107
Figure 4.2: Stages of meiosis: During
leptotene, the duplicated chromosomes
condense. In zygotene, the SC starts
to form the synapsis of homologous
chromosomes. During pachytene, homologous chromosomes are completely
synapsed and crossing-overs occur. In
diplotene, SC starts to disappear. In diakinesis, nuclear envelope starts to fragment and the bivalent chromosomes are
ready for subsequent divisions. Blue
and red distinguish chromosome pairs.
Continuous or broken lines distinguish
the two chromosomes of a given pair.
Dots: Cohesin. Ovals: Centromeres.
Light green: SC. Black arrows: Recombination sites. Dark green broken lines:
Microtubules. Image courtesy: David
Fournier.
, Step 1). After DNA replication, the chromosomes display two chromatids, associated at certain locations via cohesin protein. During
early meiosis, cohesin spots dramatically increase, leading chromosomes to form typical chain of loops, called axial element (Figure 4.2,
Step 2, inset). The chromosomes are unpaired and elongated. Moreover, chromosomes start to attach to the nuclear envelopes via their
two extremities (Figure 4.2, Step 3). During zygotene stage, twochromatids chromosomes start to pair, most likely using extruding
DNA generated by double strand breaks (DSB) as probes to find
the partner (see pairing in Figure 4.2, Step 4 represented with light
green). The search for the partner chromosome is essentially a random process. Chromosomes move along the lamina until they find
another chromosome; if the two chromosomes match, they will start
pairing (see dynamics in [Sato et al., 2009]). Pairing shows formation
of a giant protein scaffold, the SC, which holds the two homologous
chromosomes at a distance of between 150-200 nm from each other,
which is a prerequisite for recombination to happen consistently in a
3D space. The pairing usually starts close to the centromere, as the
DSB are more likely in this region [Pratto et al., 2014]. Pachytene
stage shows the most constrained DNA, with complete pairing all
along the chromosome (Figure 4.2, Step 5). Chromosome pairs at this
time are all attached around the same spot of the nuclear envelope
that is called the bouquet [Berríos et al., 2014, Gall and Pardue, 1969,
Sato et al., 2009]. During pachytene, recombination happens via the
108
the periodic and dynamic structure of chromatin
resolution of chromatid DSB from the homologous chromosome and
creating a chimera chromosome with extensive portion of both original chromosomes (Figure 4.2, Step 6, black arrows). Diplotene is a
relaxation of chromatin, with unzipping starting from centromeres,
leaving progressively the pair attached only at the location of the
recombination site (Figure 4.2, Step 7). Meiosis I ends with the migration of chromosomes to the two asters along microtubules, an event
that entangles the recombined chromosomes and leads to migration
of the two chromosomes of each pair to different sides of the cell (Figure 4.2, Step 8). Meiosis I generates two cells with 2n chromosomes
(Figure 4.2, Step 9), while meiosis II completes gametogenesis by
generating overall four gametic cells with n chromosomes (Figure 4.2,
Step 10).
In this description of meiosis, synapsis is an essential event for
recombination of the homologous chromosomes. Though advances
of genomics have proven that recombination can be studied by DNA
sequence analysis, getting the detailed structure of the pachytene
chromosome is essential to understand the mechanistic steps that
have to take place in order to make recombination happen. Here, I
propose a detailed description of meiotic chromosomes at pachytene
stage. I start by exploring the organization of SYnaptonemal Complex
Proteins (SYCPs), which form the SC, the backbone of pachytene
chromosome, and then I describe the spatial distribution of chromatin,
including some of its functional compartments associated to certain
histone modifications, in relation to SYCP.
4.2
4.2.1
Organisation of the synaptonemal complex (SC)
Superresolution imaging of the SC substructures
The synaptonemal complex (SC) is an important structural component
of meiosis that is used to localise pachytene chromosomes in routine
experiments [Baudat et al., 2013]. Two important components of SC,
namely SYCP3 (the lateral element, Figure 4.3A) and SYCP1 (the
central element, Figure 4.3B) were first characterised in order to study
the architecture of meiotic chromosomes.
Chromosome spreads from oocytes were obtained from collaborators (for details see [Prakash et al., 2015]. SYCP3 and SYCP1
C-terminus were labelled on separate preparations using antibodies
coupled with Alexa Fluor 555. On average, 2500 photons per cycle
were detected for Alexa Fluor 555 and localization maps of the two
proteins were generated by integrating roughly 20,000 observations.
Each frame captured photons acquired during 100 ms of integration
time. Individual localisations were blurred using the mean distance
to the 20 nearest molecules (method described in detail in [Kaufmann
et al., 2012]). This particular blurring method was chosen for clarity
of the visuals and comparison to other visualisation methods is done
in Chapter 2, Section 2.7.3.
Main images of SC proteins are shown in Figure 4.3. The colours
periodic and symmetric organisation of meiotic chromosomes
A
A1
SYCP3
B
B1
SYCP1 C-terminus
show a density map, with denser regions close to yellow and regions
with low density close to red. Qualitatively, one can see the doublestrand nature of both proteins, which is not apparent in wide-field
mode of the same image (Figure 4.3, inset A1 and B1). A dual color
image showing relative organisation of SYCP1 and SYCP3 is shown
in Figure B.1 and confirms the double-stranded patterns (Figure 4.4).
4.2.2
Quantification of SC substructures
Quantifications confirmed the general structural features of the synaptonemal complex. Profiles of pixel density show the improvement of
the superresolution setup compared to a wide-field image of the same
object, the double-stranded SYCP3 and SYCP1 scaffold (Figure 4.4A):
clearly, the two strands cannot be distinguished with conventional
microscopy (for confocal microscopy images, see [Baudat et al., 2013]).
Distribution of SYCP1 and SYCP3 signals along the SC shows a clear
separation of the two lateral elements, and a more central localisation
for SYCP1 (Figure 4.4B). By computing the distance from the maxima
of each individual strand of SYCP3, I estimated the distance between
the two strands of SYCP3 to be around 181 nm and the width of an
109
Figure 4.3: Organisation of the synaptonemal complex proteins (figure and
caption modified from [Prakash et al.,
2015]): SYCP3 (A) and C-terminus of
SYCP1 (B) protein are imaged using antibodies coupled with Alexa Fluor 555.
Insets (A1 and B1) show the widefield
equivalents of the underlying localization image.
110
A
C
the periodic and dynamic structure of chromatin
B
SYCP3
D
individual strand to be 60 nm (Figure 4.4B). Moreover, analysis of
SYCP3 patterns at the ends of the chromosomes shows that the SC is
narrower at the telocentric end compared to the other end (Figure 4.11,
Figure B.8). I used the same approach for SYCP1, whose distance
between the strands was found to be 88 nm and the width of the
strands themselves 47 nm.
The inter-distance between strands was calculated by taking the
position with the maxima of molecule distribution on each strand.
Individual fluorophores were detected with a precision of 11 nm for
SYCP3 and 16 nm for SYCP1. Moreover, a Fourier Ring Correlation
(FRC) resolution of 43 nm and 56 nm was found for SYCP3 and SYCP1
C-terminus, respectively. For SYCP3, the nearest neighbour distance
was 14 +/- 5 nm and for SYCP1, this distance was 19 +/- 7 nm
(Figure 4.4C). These distances were used as the standard deviation of
the Gaussian distribution applied to the individual molecule position
for blurring, a slight modification of Gaussian blurring (see Figure 2.17
of methods chapter for comparison of visualisation methods using
the SC data as example).
Furthermore, in a two colour experiment both SYCP3 and SYCP1 Cterminus were stained. Immunostainings were performed using either
Alexa 555 coupled to an antibody against SYCP3 or Alexa 488 coupled
to an antibody against the C-terminal of SYCP1 (Figure B.1). The
merged image clearly shows that SYCP1 is located inside of SYCP3
(Figure B.1C), of similar double-strand nature (insets). I roughly
summarize the findings in a visual that shows SYCP3 proteins on
the outside of the SC and SYCP1 on the inside (Figure B.1D). These
results are in line with observations from other studies [Gustafsson
et al., 2008, Syrjänen et al., 2014, Schücker et al., 2015]. A recent
SMLM study reported the width of individual strands of SYCP3 to
Figure 4.4: Quantification of SC substructures (figure and caption modified
from [Prakash et al., 2015]): (A) Distribution of SYCP3 along the transversal
plane of the SC (i.e. perpendicular to
main axis) in SMLM (red) versus widefield configuration (blue). (B) Distribution of C-terminal end of SYCP1 and
SYCP3 along the transversal plane (red
and blue plots respectively). The distance between the positions with highest molecule occupancy is 180.7 nm for
SYCP3 and 87.9 nm for the C-terminus
of SYCP1. (C) Average distance to the
20 nearest neighbors was used to calculate the extension of blurring of the
individual molecules. For SYCP3, the
distance used was 14.31 nm +/- 5.43
nm and 18.81 +/- 7.40 nm for SYCP1.
(D) A schematic diagram illustrating the
thickness and the interstrand distance
of SYCP3 and SYCP1.
periodic and symmetric organisation of meiotic chromosomes
111
be around 56 nm while the distance between strands to be around
165 nm [Schücker et al., 2015]. The width of SYCP1 C-terminus was
reported to be 45 nm with overall width of central region (composed
of transverse filaments (TF) and the central element (CE)) to be around
148 nm, similar to the values found in my analysis (Figure B.1D).
4.2.3
A model for organisation of SC
In this section, I suggest a model for the SC incorporating further
information from [Syrjänen et al., 2014]. As the antibodies are directed toward the C-terminal of SYCP1 proteins, information obtained
from localization of these molecules confirms the inner position of
the SYCP1 C-terminal domain (Figure 4.5), with N-terminal sides
of SYCP3 molecules directed toward the center of the SC. SYCP1
dimers could interact to form a ladder-like structure, while SYCP3
molecules link neighbour DNA loops, as demonstrated in the model
of [Syrjänen et al., 2014]. The figure shows the two chromosomes
modelled with two sister chromatids shown in light and dark blue.
SYCP3 triads can act to bring distant locations on a chromosome together by self assembly [Syrjänen et al., 2014]. This would eventually
lead to the shortening of the chromosome axis longitudinally while
lengthening or looping out of chromatin radially. The chromatids for
both homologs are separated by a distance of roughly 150-200 nm
i.e. the distance between the two strands of SYCP3 (green box). The
inter-loop DNA is figured with a black line. This region is likely to
be the site of repetitive DNA sequence and may be enriched with
histone marks that condense chromatin like H3K27me3 [Pauler et al.,
2009]. SYCP1 is depicted as a blue box. Similarly to SYCP3, SYCP1
might play an important role in the final assembly of the pachytene
chromosomes.
Figure 4.5: A model for the synaptonemal complex explaining the results presented in this chapter with evidences
and models from [Syrjänen et al., 2014]
is shown. The view shows the two chromosomes modeled with two sister chromatids in light and dark blue; the interloop DNA is figured with a black line.
SYCP1 is depicted as a blue box and
SYCP3 with a green box; both show their
N and C terminal domains.
112
4.3
the periodic and dynamic structure of chromatin
Periodic organisation of pachytene chromosomes
The pachytene chromosomes have a very constrained pattern in 3D
space [Schücker et al., 2015], which is highly identifiable on mammalian chromosome spreads. This is due to the fact that recombination process requires two chromosomes of a pair to be kept apart 100
nm of each other in order exchange of chromatids to happen [Petronczki et al., 2003]. As a result, both the highest order of chromatin
complexity and density is attained at pachytene stage of meiosis,
the moment when recombination takes place. Pachytene chromosomes are visible in EM (for instance in [Moses et al., 1977]) and SEM
(Figure B.3) preparations of mouse samples and display a typical
caterpillar shape. Though EM and SEM are among the most powerful
techniques to probe nanoscale structures, they do not stain efficiently
and specifically particular proteins. High resolution light microscopy
based measurements of pachytene chromosome are therefore necessary and occupy the next section of the present chapter.
C
A
A
A1
DNA
SYCP3
D
E
Figure 4.6: Super resolution microscopy
reveals higher order clusters of chromatin patterns along the pachytene chromosome (figure and caption modified
from [Prakash et al., 2015]): (A) Samples
immunostained using Alexa 555 conjugated anti-SYCP3 and counter-stained
with Vybrant Violet, a photocovertible
DNA dye (see Chapter 2 for more details). Inset A1 shows the wide-field
image for comparison. B. DNA density map of the pachytene chromosome.
Scale bar same as in (A). (C) Magnified
section corresponding to yellow box in
B. Scale bar: 250 nm. (D) Distributions
of SYCP3 and DNA molecules around
central axis of SYCP3. (E) Autocorrelation plot of chromatin patterns along
the SC axis based on chromosome section displayed in panel C. Periodicity is
found to be 550-700nm tangentially and
the cluster diameter is estimated to be in
the range 170-225 nm. Blue lines show
the 95 percent confidence bounds (+/0.08) of the autocorrelation function.
periodic and symmetric organisation of meiotic chromosomes
4.3.1
Superresolution imaging of pachytene chromosomes reveals periodic
clusters of chromatin
Experimental procedure was close to the one described for the characterization of SC proteins. The protocol for imaging DNA has been
described in [Żurek-Biesiada et al., 2015]. Briefly, samples were immunostained using Alexa 555 conjugated anti-SYCP3 and DNA with
Vybrant DyeCyble Violet (Life Technologies), a dye with similar properties as Hoechst and DAPI [Szczurek et al., 2014, Żurek-Biesiada
et al., 2015]. Blinking efficiency was stimulated using our custom
switching buffer consisting of glycerol with 10% (wt/vol) imaging
buffer (stock comprising 0.25 mg/mL glucose oxidase, 0.02 mg/mL
catalase, 0.05 g/mL glucose in PBS). For SYCP3, localisation maps
were generated integrating roughly 20,000 observations and about
100,000 observations for DNA. A two-color image was generated after
application of nearest-neighbour blurring to the raw single molecule
data (Figure 4.6A). Density map of the DNA channel reveals periodic
clusters of chromatin, characterized by a central high density region
(Figure 4.6B). Elongated extensions of the DNA are localised laterally,
possibly involved in recombination [Gall, 2012, Callan, 2012]. General
shape of chromatin is reproducible (Figure B.6) and is validated by
SEM experiments (see Figure B.3).
4.3.2
Quantification of periodic chromatin clusters
Quantification helped to further describe the density distribution and
the periodicity of the chromatin clusters. Pixel density classification
(Figure B.4) reveals that the clusters have two major regions: a central
region which bares the highest density (3.5 signals/nm2 ) and an outer
region of lower density (1.5 signals/nm2 see Figure B.4, purple and
green regions, respectively).
In order to characterize the periodicity of the clusters and distinguish them from background, I first applied auto-correlation, an
algorithm to find similarity of signals (Figure 4.6C). Periodicity lies
in the range 550-700 nm tangentially along the SC axis (Figure 4.6E,
Figure B.6A3). The average diameter of the clusters is found to be in
range of 170-225 nm (Figure B.9). Later, a new algorithm was developed to further characterize the clusters (Figure 4.7). First, a binary
mask was generated based on distance to 20 nearest neighbour and
signals within this mask were taken for further analysis. A randomly
generated image with the same number of localisations is presented
for comparison (Figure B.2B). The random distribution served as a
control to be compared to the SMLM data. Nearest neighbour distances were used to characterise the local condensation in SMLM and
randomly simulated data (Figure B.2C). In the experimental data, the
mean distance to 500 nearest neighbours was around 86 nm while in
the simulated random data, it was 110 nm (see [Prakash et al., 2015],
supplementary data). This non-random distribution of DNA was
indicative of the fact that the clustering observed in SMLM images of
DNA is a consequence of structural packing of DNA.
113
114
the periodic and dynamic structure of chromatin
Figure 4.7:
Characterization of
pachytene chromosome clusters: (A)
Pachytene chromosomes stained with
Vybrant Violet. (B) Randomly simulated
dataset.
4.4
4.4.1
Functional organisation of pachytene chromosomes
Rational
The power of high resolution light microscopy over electron microscopy is to specifically label particular structures on a given sample
(for comparison on same sample refer to [Lippincott-Schwartz and
Patterson, 2009], Figure 5). With SMLM, one can not only observe
nano-scale structures but is also able to describe them in molecular
terms. For example, individual functional compartments of chromatin whether active, inactive, or repressed can be described and
categorised independently. Each of these categories experiences a
different epigenetic regulation and is enriched with various combinations of epigenetic modifications.
I was curious to investigate the contribution of epigenetics, more
specifically histones post-translational modifications (PTM) associated
to the structure of pachytene clusters observed in SMLM experiments
(Figure 4.7). In interphase, PTM are known to regulate gene expression by modifying DNA accessibility (see example on Figure 4.8 and
Chapter 1 , Section 1.5). In meiosis, apart from de novo H3K4me3,
which is thought to be a major regulator of meiotic recombination by
promoting double strand breaks [Mihola et al., 2009, Sommermeyer
et al., 2013], not much is known about the epigenetic regulation of
chromatin. As a result, I tried to define a rational to select meaningful
histone modifications in order to investigate and better characterise
Figure 4.8: Chromatin compaction as a
function of histone modifcations: Chromatin is not an inert structure and its
compaction and relaxation can be modulated by many external and internal
factors, among them post-translational
modifications of histones. A few of
these modifications and their impact
on chromatin structure has been characterised, however, the list of these modifications is constantly growing. Some
histone modification cause chromatin
to be more compact while others can
make chromatin to take a more relaxed
and open configuration [Bannister and
Kouzarides, 2011].
periodic and symmetric organisation of meiotic chromosomes
115
the chromatin architecture of the pachytene chromosomes.
I decided to take a naive approach to systematically select relevant
histone marks for the study. Assuming that chromatin states such as
gene activation or gene repression are often defined by combinations
of post-translational histone modifications (also known as the ’histone
code’), it is likely that positions of several histone marks, having
similar function, may co-localise throughout the genome. Finding
these clusters of marks would tell what is the minimal number of
marks to study in order to get a rough picture of chromatin functionality. Therefore, I studied the co-occurrences of histone marks
and derived meaningful functional clusters. As the data is sparse
for mouse oocytes meiotic chromosome, I used data from classical
studies of [Barski et al., 2007, Wang et al., 2008] in human immune
cells to benchmark histone mark associations. This method is described in detail in [Prakash, 2012]. The Boolean model helped to
select genomic positions which have strong ChIP-seq signals to study
the association of histone marks at these sites. This resulted in a table
where associations between histone modifications are marked with a
coloring scheme (Figure 4.9).
Figure 4.9: Co-occurences and clustering of histone post-translational modifications in CD4 T cells: Blue circles
represent positive correlations between
histone modifications while red circles
represent negative correlations. Size and
color depth indicate the measure of the
correlations. Histone modifications with
no significant correlation are left blank.
The dark blue rectangles along the diagonal of the correlation matrix represent
clusters of histone modifications and are
based on the results of hierarchical clustering. The cluster on the top left corner
shows H3K4me3 correlating with several acetylation marks. The bottom right
cluster shows positive correlation of
H3K27me3 with H3K79me1, H3K9me2
and H3K27me2. The fourth cluster
along the diagonal shows strong correlation between H3K9me3, H4K20me3 and
H3K79me3.
4.4.2
Clustering method sorts chromatin into functional epigenetic compartments
Hierarchical clustering of 39 different histone modifications led to
identification of five functional clusters. The first cluster was found
to be associated with gene activation, comprising acetylation marks
and H3K4me3, a well-known mark associated to initiation of double
116
the periodic and dynamic structure of chromatin
strand breaks during recombination (Figure 4.9, top-left cluster displayed as a blue box); a second cluster was found to be associated to
H3K9me3, a mark involved in constitutive heterochromatin of centromeres, structures which have a strong importance in synaptonemal
complex initiation (Figure 4.9, fourth cluster from the left); lastly, I
identified a cluster of repressive marks (Figure 4.9, fifth cluster from
the left), from which H3K27me3 has been previously identified as anticorrelated to meiotic DSBs [Buard et al., 2009]. Genomic data confirm
the anti-correlation of H3K9me3 and H3K27me3 clusters during germ
cell differentiation [Liu et al., 2014]. On the other hand, H3K27me3
and H3K4me3 are shown by genomics to anti-correlate genome-wide
[Hammoud et al., 2014]. As a result of this analysis, I decided to
stain the following 3 representative marks to further characterize
chromatin organisation: H3K4me3, H3K9me3 and H3K27me3. The
function and genomic regions where these marks are known to be
enriched is summarised in Figure 4.10.
4.4.3
Centromeric histone mark (H3K9me3) labels one end of the SC
H3K9me3 is associated to centromeres and to the initiation of doublestrand breaks at the early stages of SC formation [Hernández-Hernández
et al., 2012, Mikkelsen et al., 2007]. In mouse chromosomes, centromeres localise at one of the telomeres (often referred to as telocentres), resulting in chromosomes having only one arm. Chromatin
staining of the pachytene chromosome shows a huge cluster of DNA
at one end of chromosomes (Figure 4.6B). I used SMLM to probe if
these dense chromatin regions were enriched with the centomeric
mark H3K9me3. Images obtain show that one of the ends of SYCP3
is highly enriched with H3K9me3 (Figure 4.11), compared to the
other end of SYCP3. Moreover, the centromeric end is remarkably
distinguishable by a narrower SYCP3 structure compared to the other
end (Figure 4.11B). This observation is further presented in replicates
(Figure B.7 and Figure B.8A-B). Furthermore, at the telocentric end of
the chromosome (Figure 4.11C), H3K9me3 co-localizes with a high
chromatin density. The top cluster on chromatin image (Figure 4.6B)
has a similar average axial spread of 1-1.5 µm as in case of H3K9me3
(Figure 4.11A) image.
4.4.4
Repressive histone mark (H3K27me3) shows characteristic periodic
clusters along the SC
H3K27me3 is associated to repeat regions of the genome and some of
them can be deleterious for the proper transmission of information
to the next generation [Pauler et al., 2009]. Next, I wondered if
H3K27me3 might possibly contribute to the periodic clusters seen
in Figure 4.6C. In agreement with previous findings [HernándezHernández et al., 2010], H3K27me3 post-translational modification
was found to be localised close to the SC, directly at the outside of
the regions occupied by SYCP3 proteins (Figure 4.12). The important
overlap between H3K27me3 and SYCP3 was also visible in wide-field
Figure 4.10: Functional chromatin states:
Chromatin can be broadly classified
into 3 functional chromatin states and
these states are characterised by three
different histone modifications. The
constitutive heterochromatin is found
in centromeres, telomeres and repetitive sequences. Constitutive chromatin
is highly condensed, transcriptionally
silent and is marked by H3K9me3. Facultative heterochromatin is relatively
less condensed, transcriptionally reversible and marked by H3K27me3. Euchromatin, which is rich in genes and
transcriptionally active is marked by
H3K4me3.
periodic and symmetric organisation of meiotic chromosomes
A
A1
B
H3K9me3
SYCP3
C
D
D1
Nearest neighbor blurring
D2
Gaussian blurring
D3
Single molecules
Triangulation
(Figure 4.12A). However, the SMLM image revealed an unexpected
pattern. H3K27me3, instead of spanning the entire length of the
chromosome, displayed large periodic clusters, disposed in pairs
(Figure 4.12B), found every 450-650 nm along the SC according to
auto-correlation analysis (Figure 4.12D). Size of clusters was found in
the 90-130 nm range (Figure B.9) and the average lateral distance from
the strands of SYCP3 in the 40-50 nm range (Figure 4.12C), though
there is an overlap with SYCP3 in several instances, possibly due to
twisting of SYCP3. The proximity of H3K27me3 and SYCP3 on the
SMLM cliches explains why the two features seem to co-localise with
confocal images (∼ 200 nm resolution of the imaging system).
The patterns shown on Figure 4.12 were reproducible on all chromosomes (see [Prakash et al., 2015] supplementary data). The periodicity of H3K27me3 and chromatin clusters found in Figure 4.6C seem
to match and I hypothesize that the larger chromatin clusters with
lateral extensions have strong enrichment for H3K27me3 mark at their
base, an information that I incorporate in a model of the pachytene
117
Figure 4.11: H3K9me3 displays a centromeric position along the SC (figure
and caption modified from [Prakash
et al., 2015]): Two-color SMLM image
(A) immunostained with anti-SYCP3
(Alexa 555) and anti-trimethylated histone H3K9 (Alexa 488). Inset (A1)
shows the underlying wide-field image. Large dense clusters of repressive centromere mark H3K9me3 can be
seen at the presumptive telocentric end
of the pachytene chromosomes. High
H3K9me3 density at the telocentric ends
of SYCP3 is concomitant with high chromatin density at the top axial end in (Fig.
4.6B). (B) The strands of SYCP3 where
H3K9me3 co-localises (presumptive telocentric end) are found to be closer ( ∼
135 nm) than at the presumptive telocentric end (∼ 180 nm). (C) H3K9me3 distribution hints for spiralization of DNA
at the centromeric end of the SC. Point
of representation of single molecules
of SYCP3 and H3K9me3 is shown in
(C). (D) compares different visualization
strategies. Spiralization of DNA was
observed at one of the extremities of
synapsed chromosomes (D1-D3), probably working as an anchor for the two
chromosomes to start synapsis. This spiralization exemplifies the relevance of
blurring methods for visualisation, nevertheless, feature is visible in all three
visualisation methods.
118
the periodic and dynamic structure of chromatin
C
A
A1
H3K27me3
SYCP3
D
chromosome (see Section 4.6.1).
The pattern of the clusters itself indicates their high functional
nature. Symmetry and perfect pairing means that deposition of this
repressive mark is dependent on the genomic content of chromosomal
regions and is reproducible between chromosomes. What is the function of these regions? The present analysis does not provide any answer, but one speculates that these periodic clusters might correspond
to the silent repeated regions of genomes, such as long interspersed
elements (LINE) or long-terminal repeat (LTR) , as H3K27me3 is often
found to be associated to this kind of element [Pauler et al., 2009].
This repressive feature is key to prevent the mobility of jumping
elements during meiosis, whose movements can have consequence
for the progeny.
4.4.5
Histone mark (H3K4me3) associated with active transcription emanates radially from the axis of the SC
Finally, I present here the distribution of H3K4me3, a mark known to
be highly involved in generation of double strand breaks, along the
pachytene chromosome. Instead of localising at the axis, H3K4me3
showed localisation in the lateral to the central axis of SYCP3 (Figure 4.13), possibly corresponding to the extensions identified with
DNA staining (Figure 4.6). H3K4me3 extensions were found to be 500
nm wide which were in accordance with chromatin protrusions (Figure 4.6D). Similarly to H3K27me3, H3K4me3 forms clusters of 50 nm
diameter (Figure B.9). The average spread of the radial emanations
of H3K4me3 ranged from 300 to 500 nm, with the overall distribution of H3K4me3 qualitatively similar to the chromatin distribution
(Figure 4.6D, Figure 4.13A1). By applying autocorrelation to the tangential distances from the central axis of SYCP3 (Figure 4.13A3), I
estimated the average spread (approx. 500 nm) of the protrusions to
be larger than their spacing (approx. 200 nm).
Figure 4.12: H3K27me3 displays periodic paired clusters in proximity to
the SC (figure and caption modified
from [Prakash et al., 2015]). Two color
SMLM image (A) immunostained with
anti-SYCP3 (Alexa Fluor555) and antitrimethylated histone H3K27 (Alexa
Fluor488). Inset (A1) shows a widefield version of the image for comparison. Overlaps between H3K27me3
and SYCP3 channels are shown in
white. (B) The periodic and symmetric clusters of H3K27me3 mark shown
as raw data (single molecules in green
or after blurring procedure in purple).
(C) Differential distribution of SYCP3
(top) and H3K27me3 clusters (bottom).
Both distributions are symmetrical, but
H3K27me3 are found slightly outside of
the SYCP3 strands, laterally 50 nm apart
from the SC, with a 90-130 nm diameter.
(D) Autocorrelation on yellow section of
panel B reveals the tangential distance
of the clusters from the central axis of
SYCP3 to be 40 to 50 nm and reveals
a periodicity of approximately 500 nm
(see also [Prakash et al., 2015]), supplementary data). Blue lines correspond to
the 95 percent confidence bounds (+/0.1633) of the autocorrelation function.
periodic and symmetric organisation of meiotic chromosomes
A1
A
A2
A3
H3K4me3
SYCP3
4.5
4.5.1
Structure and dynamics of meiotic chromosomes
Lampbrush-like structures in mammalian meiotic chromosomes
Lampbrush chromosomes (LBC) are giant chromosomes found in
oocytes of diverse organisms, both vertebrate and invertebrate. They
are characterized by paired loops of transcriptionally active chromatin
that extend laterally from an axis of inactive chromatin. Among
vertebrates, they have been extensively studied in amphibians, reptiles,
fish, and birds, but have not been convincingly demonstrated in
mammals [Gall, 2012, Callan, 2012].
Though known to be necessary for the pachytene chromosome to
function normally, presence of loops in mammals has not been so far
optically demonstrated ([Heng et al., 1996], does not show convincing
images). Data in this thesis are a good step in this direction, as some
of the images show H3K4me3-marked processes forming circular
patterns (Figure 4.14) resembling SEM (Figure B.3) and EM data of
pachytene chromosomes observed in amphibians [Rattner et al., 1980].
Some putative looping structures are found on our chromatin SMLM
pictures (Figure 4.6) but it is possible that the pattern is the result of
contact of two adjacent extensions of the chromatids. Nevertheless,
the size of lateral chromatin extensions (500 nm) (Figure 4.6D) is
similar to the size of H3K4me3 extensions (Figure 4.14D). H3K4me3
is better at spotting possible loops because the labelling density is
lower than DNA (DNA staining: 5,000 single-molecule signals per
square micrometer [Szczurek et al., 2014] , histone staining: 100 single-
119
Figure 4.13: Lower orders of chromatin
domains characterised by H3K4me3:
Two color SMLM image (A) immunostained with anti-SYCP3 (Alexa Fluor
555) and anti-trimethylated histone
H3K4 (Alexa Fluor 488). H3K4me3 clusters reveal lower orders of chromatin
domains along the pachytene chromosomes (figure and caption modified
from [Prakash et al., 2015]). The average
spread of H3K4me3 shows periodicity
along the lateral direction to the central
axis of SYCP3, possibly indicating individual H3K4me3 clusters (A1). The average cluster diameter in (A) was found
to be 47 nm (see also Figure B.9). (A2)
shows the histogram of the tangential
distances taken along the central axis of
SYCP3 in the white boxe indicated in
(A). (A3) shows the periodicity of clusters of H3K4me3. The periodicity in A
was found to be around 216 nm. The
blue lines in the A3 plot correspond to
the 95 percent confidence bounds (A3,
+/- 0.1414).
120
the periodic and dynamic structure of chromatin
A
B
Nearest neighbor blurring
C1
C
WIDEFIELD
D
H3K4me3
SYCP3
SMLM
E
F
molecule signals per square micrometer [Bohn et al., 2010]), which
creates more opportunity to distinguish thin or subtle structures. This
point is another example of the importance of staining chromatin
with both DNA dyes and histone modifications.
4.5.2
A model for SC spiralisation during the zygotene/pachytene transition
The above data provide important information regarding the organisation of pachytene chromosomes. They also provides some hints
regarding the dynamics of pachytene chromosomes during meiosis.
Though presently there is no time series data of meiosis steps, it is
worth mentioning that the images hint at a spiralization of chromosomes in pachytene stage, possibly for functional reasons. It was
shown that H3K9me3 displays a helical wrapping around the SC
(Figure 4.11). DNA staining itself also shows a trend to form spirals,
either only at the extremity (Figure 4.6) or all along the body of the
chromosome (Figure B.6). One may object that the situation studied is
not native, as the chromosomes have been processed through a rather
harsh protocol. Nonetheless, evidences show that the helical structure of single chromosomes is preserved in pachytene chromosomes
showing bouquet-like patterns close to native state (Figure B.7, two
pair of chromosomes on left).
I further show high-order patterns of H3K4me3 on certain chromosomes, showing clear helical patterns. As H3K4me3 is probably
spanning most of the space occupied by chromatin, one can approximate the positions of the H3K4me3 for overall spatial occupancy of
Figure 4.14: Distribution of transcriptionally active chromatin (figure and
caption modified from [Prakash et al.,
2015]): Radially emanating chromatin
loops are punctuated with transcriptionally active histone mark (H3K4me3):
Two color wide-field image (A) immunostained with anti-SYCP3 (Alexa
Fluor 555) and anti-trimethylated histone H3K4 (Alexa Fluor 488) with (B)
showing the corresponding SMLM image. (C) shows radial loop like patterns
emerging laterally compared to SYCP3,
with C1-C3 showing (C) with different
visualisation methods. (D) The distribution of H3K4me3 around SYCP3 is
characteristically similar to chromatin
distribution stained with Vybrant DyeCycle Violet (Fig. 4.6D). The average
spread (∼ 300 nm ) of radially emanating chromatin loops perpendicular
to the lateral elements of SC is uneven
across the SYCP3 strands punctuated
with periodic patterns of H3K4me3. (E)
shows the histogram of the tangential
distances taken along the central axis
of SYCP3 in (C). Using an autocorrelation function (F), the estimated average
spread of the radial loops was found to
be larger than their spacing (∼ 200 nm).
periodic and symmetric organisation of meiotic chromosomes
B
A
1
2
H3K4me3
SYCP3
3
6
7
5
4
C
D
E
chromatin. To demonstrate the spiralization, I developed an algorithm
to identify alternate patterns of H3K4me3 clusters on each side of
the SC (Figure 4.15B). Most of the neighbour clusters were found
to be separated along the axis of SC, a pattern even more evident
on the first image with the SYCP3 staining. Same type of alternative configuration was found using data from H3K27me3 imaging
(Figure B.5). Such chromosomes may role around each other in the
fashion of mating snakes (see snake model, Figure 4.16). Pairing of
the two chromosomes could start by a tight enrolment of chromatin
at the centromere, H3K9me3-rich (the spiral pattern of H3K9me3 in
Figure 4.11 ascertains this fact), to rapidly induce the wrapping of
the rest of the chromosome. Recent data from 3D-SIM confirms a
complete 3D twist of SYCP3 along the axis of the chromosome, as predicted by the snake model (David Fournier, personal communication,
data not shown). Moreover, helical patterns are clearly identifiable
in whole cell preparations of Xenopus oocytes from early studies on
pachytene chromosomes [Gall and Pardue, 1969], probably working
as an anchor for the two chromosomes to start synapsis.
It is speculated that the helical patterns have a structural or a
mechanistic function. Spiralisation is probably essential for pairing
to happen, by inducing a wrapping of the two chromosomes around
each other once pairing has started. Spiralisation may also help
recombination events to form, following the initiation of synapsis to
impose certain structural constraints that help crossing-overs to form.
4.6
A model of spatial distribution of chromatin around the SC
Finally, I summarize the composition of high-order chromatin into
lower-order histone modification clusters using distance profiles (Figure 4.17, left). One can observe that while chromatin has the broadest
121
Figure 4.15: Helical structure of the
pachytene chromosomes: General pattern of H3K4me3 confirms hypothesis
of helical structure of the pachytene
chromosomes. (B) The distribution of
H3K4me3 hints for potential rotation
of chromatin along the entire length of
SC. The clusters of H3K4me3 are colorcoded based on nearest neighbor distances to indicate that the rotation follows the rotation of SYCP3 in (A). (C)
The average spread of H3K4me3 showed
periodicity along the normal direction to
the central axis of SYCP3. The average
cluster diameter in white box (A) was
found to be 45 nm. (D) shows the histogram of the tangential distances taken
along the central axis of SYCP3 in the
white boxes. (E) shows the periodicity
of the clusters of H3K4me3 loops. The
periodicity was found to occurr at intervals of 209 nm. The blue lines in the
plot correspond to 95 percent confidence
bounds ( +/- 0.1613).
122
the periodic and dynamic structure of chromatin
extension, H3K9me3 follows the same shape as the chromatin profile and is normally distributed along the central axis of SC, hinting
that the telocentric region is mostly covered by the H3K9me3 cluster, while other regions that extended laterally are deprived of this
centromeric mark. Differently, H3K27me3 cluster follows the shape
of the synaptonemal complex, hinting for its tangential localization.
H3K4me3 profile is asymmetrically distributed and comparison with
chromatin profiles shows a level of complexity that chromatin alone
cannot capture, as possibly most of the DNA is not active. Most of
the profiles are slightly lopsided, that I speculate to be mostly due to
chromosome not being in the sample plane and the angle creating the
asymmetry.
4.6.1
A ’cluster-on-a-string’ model for spatial distribution of pachytene
chromosomes
The different clusters observed (H3K4me3, H3K27me3 and H3K9me3)
are obviously of much higher order than basic bead-on-a-string structures of DNA and histone complexes. As a result, I hypothesize that
the patterns observed in the pachytene chromosomes are rather a
large chunk of those beads, possibly of a state close to chromatin fibres.
I describe them as ’clusters-on-a-string’, to point that they are likely
bulky regions epigenetically regulated, surrounded by disorganized
DNA fibres. These structures are probably to relate to chromatin
domains found in interphase (see Section 3.2.3 and Figure 3.16). More
has to be explored in order to see what is the structure of the clusters
and the regions in between. For this, more informational staining will
have to be performed, using either polymerases, splicing machinery
or other enzymes.
Finally, I summarize our findings in a model of the pachytene
chromosome, which displays roughly three epigenetic compartments,
capturing most of the epigenetic information in chromatin according
to the histone code table (Figure 4.17, right). While H3K4me3 is
covering the entire chromatin landscape, including lateral extensions,
synaptonemal complex region and telocentric end of the chromosome,
H3K27me3 is confined to the SC, a fact confirmed by a trend on
distribution plots in Figure 4.17, left. I speculate that the two marks
anti-correlate in the SC. H3K4me3-rich spots being prone to recombination, following binding of PRDM9. H3K27me3-rich spots focused
on repressing intermediate regions, either to silent undesirable elements such as repeated elements, which may strengthen its integrity
by over-twisting.
4.7
Summary and Conclusion
Different variations of DNA and histone modifications staining, at
single molecule resolution, have revealed two orders of chromatin
in the meiotic chromosomes that were unknown before. The analysis shows that histone modifications can be useful not only to add
Figure 4.16: A snake model for the formation of the synapsis. (A) Zygotene
stage. The snakes represented on the
cartoon start copulation by their enrolment close to the head. The heads of the
snakes find each other at the telocentric
ends of the zygotene chromosome by a
yet unknown mechanism. (B) Pachytene
stage. Starting from head, impulse is
transmitted to the rest of the body. Attachment of the paired chromosomes
at their telocentric end is followed by
SC formation potentially followed by a
twisting mechanism, which starts at the
telocentric end to be loosely transmitted to the rest of the chromosome. I
speculate that this process may facilitate
recombination. Compare (B) with Figure B.8.
periodic and symmetric organisation of meiotic chromosomes
123
Figure 4.17: Symmetric and periodic organisation of pachytene chromsomes:
The two strands of the synaptonemal
complex are displayed in black. In purple, chromatin covers the entire surface and is thought to be ubiquitously
marked by tangential mark H3K4me3,
shown in blue. Tangential, symmetrical and periodic H3K27me3 clusters are
displayed with yellow spots. Polar chromatin associated to H3K9me3 is shown
as an orange spiral pattern.
functional information but also to study the chromatin organisation
in more details. It also shows that the level of resolution of histone
modification is highly useful to characterize very fine regions such as
loops; H3K4me3 staining revealed loop-patterns not observable with
the DNA staining while H3K9me3 hints toward spiralization of DNA
at telocentric end, a pattern less visible on DNA images.
The above results confirm the predicted anti-correlation from the
genomic data of the three histone modifications. H3K9me3 and
H3K27me3 clusters anti-correlate, as shown by [Liu et al., 2014] using ChIP-seq data. Similarly to genomic data from ChIP-seq [Hammoud et al., 2014], H3K27me3 and H3K4me3 are shown to anticorrelate genome-wide, with lateral H3K4me3 marks being devoid of
H3K27me3 mark, while axial H3K4me3 clusters show possible partial
overlap with H3K27me3 clusters. Additional co-immunostaining of
pairs of histone modifications should further reveal dependency and
correlation with other marks.
Spreading technique can question the reliability of the patterns
observed and one can wonder if they would be observed in intact
nuclei. Though this is unlikely for two reasons; one is that in virtue
of thermodynamics, a passive system will only loose complexity with
time. Spreading chromosomes can at worst destroy clusters, but
unlikely create ones, and moreover these patterns are reproducible
across several samples. Second reason is that close to native states are
124
the periodic and dynamic structure of chromatin
sometimes observed (Fig B.7), with chromosomes bound by pairs via
their centromeres, a pattern reminiscent from the native attachment
of all chromosomes to the rRNA center [Gall and Pardue, 1969]. The
helical structure of single chromosomes is preserved in these pairs
(Fig B.7).
Study of the epigenetic regulation of the pachytene chromosome
has left one with several important results. It was shown for the first
time a high resolution map of pachytene chromosome using both a
DNA dye and histone modifications, which when combined helped
to have a comprehensible view of functional chromatin (Figure 4.17).
This can easily be pursued in different contexts, especially the one of
disease phenotype, to see the implication of epigenetics in disease.
Moreover, this is also the first study to show that mammalian
chromosomes display lateral loops, similar to one observed in yeast
or amphibians. Finally, this study was the first to show high degree periodicity in chromatin organisation at multiple length scales.
More importantly, I show here clusters of DNA and various posttranslational histone modifications in a very comprehensive way for
the first time, demonstrating the importance of the meiotic chromosome as a model to study chromatin organization. The clusters are
arranged according to their functional position, and provide the extend of chromatin compartmentalisation. As such, I at least identify a
new type of functional compartment of 50 nm size (using H3K4me3)
and another one of about 120 nm (using H3K27me3).
The three histone modifications summarising the clustering generated from the histone code table have different functionalities for
the cell activity and exhibit a higher degree of specialization during
meiosis. During development, H3K4me3 and H3K27me3 histone
marks are present at promoters and have the capacity to turn on
and off genes, respectively. As a result, these genes are in a bivalent
state. Genes which become activated loose the H3K27me3 mark and
start to get transcribed. In more differentiated stages of development, H3K4me3 guides transcription, while genes having only the
H3K27me3 mark become totally repressed, eventually being rendered
inaccessible via extreme compaction of chromatin [Pauler et al., 2009].
A histone modification associated to this kind of dense chromatin
is H3K9me3, mainly associated with telomeric, repeats and satellite
sequences [Mikkelsen et al., 2007]. The final picture of the epigenetic
landscape of histone modifications can be simplified with a division
between active genes displaying H3K4me3, inactive genes displaying
H3K27me3 and inaccessible regions (rich in repeated elements) which
bare a high density of H3K9me3 mark (Figure 4.10).
In highly differentiated cells such as non-renewable adult tissues,
the relationship between epigenetic marks and gene expression is less
pronounced, with many genes either up-regulated or down-regulated,
which do not seem to exhibit any H3K27me3 or H3K4me3 modulation
([Wu et al., 2012], Epithelial-mesenchymal transition (EMT) data from
David Fournier, not published), while many genes are kept totally
silent during the entire life of the cell, with H3K27me3 spread all
periodic and symmetric organisation of meiotic chromosomes
along the gene body [Pauler et al., 2009]. As a result, the modulation
of histone modification amplitude may play a special role in meiosis
and possibly guide the structure of the chromosome.
As meiosis is the type of cell division that provides the basis
for sexual reproduction and genetic variability, defects in synapsis
and/or recombination of meiotic chromosome can lead to aneuploid
syndromes and infertility. It is likely that the SC driven chromosomes have structurally distinct histone modifications patterns and a
unique chromatin landscape compared to those identified here in the
native state. This work also demonstrates the link between the structural hierarchy of chromatin organization in meiosis and epigenetic
regulation of transcription.
I urge the scientists to use this model as a platform to study the
epigenetic make-up of genomic regions of interest, and the basics of
gene expression and epigenetic regulation. The meiotic chromosome,
while being very specialized toward one goal, recombination, happens
to be a promising candidate to study the fundamental principles that
guide chromatin architecture.
125
“Where theory lags behind the facts, we are dealing with miserable degenerating research
programmes.”
Imre Lakatos
5 Conclusions
5.1
Originality of the work presented here
In this thesis, I have shown the potential of single-molecule imaging to study chromatin architecture. The primary challenge was to
systematically identify and characterise different building blocks of
the chromatin organisation. Some elements were known since long,
for example, the nucleosomes at the lowest level and the metaphase
X-shaped chromosomes at the highest level of DNA compaction.
In between these two extreme levels, a vast spectrum of structures
remained mostly unexplored due to resolution limitation of light microscopy and lack of specificity in electron microscopy. Furthermore,
an effort to combine and synchronise observations from different studies on chromatin organisation was lacking. This thesis is an attempt
to combine old and new evidence to understand various structurefunction conundrums of chromatin organisation and understand how
chromatin structure affects gene regulation.
For genes to function properly, some fundamental structures must
exist in a finite number. I speculate that these structures should be in
a restricted size range and diversity. For instance, I found structural
elements of about 100-400 nm probably corresponding to functional
domains involved in activation or repression of transcription. After
studying chromatin organisation at different phases of cell cycle, the
following new patterns were described to refine the overall description
of chromatin:
• patterns of chromatin in interphase: territories, subchromosomal
domains (scale: 500 to 1000 nm), fundamental chromatin domains
(scale: 100 to 500 nm) and chromatin fibres (scale: 30 to 100 nm).
• patterns of condensation during differentiation of cells, and more
interestingly during stress: deprivation of oxygen and nutrients
induces a characteristic ring/rod-like shape of chromosomes (scale:
40 to 700 nm).
• patterns of functional chromatin during meiosis (scale: 40 to 600
nm). Active chromatin localises laterally towards the exterior
of pachytene chromosomes, likely facilitating transcription while
inactive chromatin locates toward the interior showing a high level
of periodicity and symmetry.
128
5.2
the periodic and dynamic structure of chromatin
A general methodology to study chromatin architecture
Further imaging at different scales should clearly identify patterns
found to be associated with various nuclear phenotypes. Toward
this, it will be more relevant to focus first on the strongest signals
(Barr bodies, nucleolus) to generate certain rules and then go for the
data of lower signals (autosomal chromosome territories, low-scale
chromatin domains). Similarly to what has been used here, combined staining of DNA molecules and functional proteins, such as
histone modifications, will help to characterise these structures in
more depth. For instance, characterization of transcription during
interphase will be essential. How many genes are transcribed during
interphase and which structure chromatin is adopting? Model cells
such as mesenchymal stem cells, whose chromocenters are associated
to significant transcription, will be useful in these cases. For comparison, functional transcription machinery could be described in other
interphase cells of lower transcription. Furthermore, information
regarding building blocks of chromatin and the interplay between
different hierarchical levels should be summarised in models to better
understand the system under consideration. Dynamics of patterns
should be determined in the light of various cell processes (either
interphase steady state, cells in embryonic development, reaction to
a stimulus). The SMLM method can be applied to many questions
regarding chromatin organisation, for instance, it could investigate
the difference of chromatin shape between the micronucleus and the
macronucleus of the Tetrahymena.
The different findings are summarised in a model of chromatin
architecture (Figure 5). The various cartoons represent different stages
of the cell cycle. In each of them, two kinds of chromatin organisation
are distinguished, one condensed and particularly folded, usually
associated with gene silencing, and the other being loose chromatin,
often related to active transcription. Interphase stage shows most of
the condensed chromatin at the periphery of the nucleus, while active
chromatin is in the centre. Upon stress (the cartoon on the top right),
chromatin starts to condense in the shape of elongated patterns at
the periphery of the nucleus while progressively, an extreme compact
pattern of chromatin appears toward the centre, forming intriguing
ring-shaped structures (middle right cartoon). Before beginning any
cell division, whether mitosis or meiosis, cells have to double their
genetic material. As a result, to maximise transcription, late S phase
to G2 phase is associated with a pattern where the nucleus is made of
islands of condensed chromatin (the chromocenters) and shows loose
active chromatin located at the periphery of the chromocenters, maximising the surface of transcription. Prophase stage of cell division
starts by the recruitment of chromocenters to the lamina, in order
chromosomes to find more easily each other in the next steps of mitosis or meiosis (central cartoon). The chromocenters drag the active
chromatin behind them, displaying a hair like pattern. Mitosis is the
product of attachment of chromatin domains with the aid of condens-
conclusions
129
ing polymers (bottom left cartoon). Meiosis is both a condensation
process of chromatin (with the help of epigenetic regulators) and a
polymerization process of synaptonemal complex proteins, which
leads to the typical structure necessary for recombination (bottom
right cartoon).
Figure 5.1: Dynamic architecture of chromatin across the various stages of cell
cycle. Image courtesy: David Fournier.
5.3
Limitations of the method and possible improvements
Many interrogations are brought by these findings. First of all, the exact physical mechanism for the powerful combination of 405/491 nm
laser combination is unknown. The Jablonski diagram of such mechanism should be established based on appropriate physics methods.
Improvement of imaging buffer should also follow a more thorough
analysis. Usage of different redox agents, susceptible to improve the
blinking of the fluorochrome molecules, should also be considered.
Proper benchmarks between the DNA dyes and various combinations
of fluorophore are still lacking and need to be set up. Most importantly is the comparison between the images generated by the SMLM
methodology and other available high-resolution technologies on the
same kinds of cells.
The randomness of some patterns at the lower level is still to be
130
the periodic and dynamic structure of chromatin
investigated (for instance in the case of DNA imaging of interphase
nuclei). This can be achieved by combining DNA staining with EdU
for the detection of new synthesised DNA in the same cell to see if
the worm-like patterns follow the elongated EdU clusters. Finally,
different ways of analysing individual molecules could be done. For
instance, what are the signals that blink exactly at the same position
several times? How much blinking does one has per molecule for
each DNA dye? Finally, a more thorough comparison between the
different DNA dyes will be relevant for the community.
5.4
New avenues for the study of chromatin patterns during
meiosis
Complementary techniques should determine which molecules provoke the epigenetic patterns found in meiotic chromosomes. Inhibition or genetic invalidation should confirm their structural roles.
The functionality of the clusters observed during meiosis should
be explored by association to functional components. For instance,
double strand break positions could be characterised on the meiotic
chromosome to see which epigenetic mark co- or anti-localize. Complementary biochemical analyses will strengthen the functionality.
Mutation of enzymes depositing the marks will provide more information regarding the assembly of epigenetically regulated clusters.
Finally, the different stages of meiosis shall be characterised to have a
dynamic picture of the epigenetic regulation and the way different
patterns are formed.
5.5
Enlarging the spectrum of questions: chromatin organisation
as a fundamental principle of nucleus formation
How do chromatin clusters form? Do sequences follow a pre-labelling,
for instance via histone modifications, before being put together in a
single cluster? Meiotic chromosomes seem to be an example platform
to test this structure-to-function relationship. Contrary to interphase
chromatin, meiotic chromatin is so well defined that there is potentially little overlap between different epigenetic domains and so a
significant potential for isolating domains in further analyses. I also
ask the following questions: How does chromatin shape changes during development? Are there defined chromosome territories positions
in early phases of embryogenesis? How epigenetics relate to gene
expression? All these dynamics are crucial to the understanding of
fundamental genomic processes and will benefit from an unbiased,
naive and structural analysis such as the ones presented in this work.
Appendices
Appendix A
Figure A.1: Areas and single molecule
localisation of inter-chromosomal territories. Chromatin data have been arbitrarily thresholded based on the density of
SM localizations and divided into two
classes: interchromatin compartments
(IC) and chromatin occupied regions.
Then the areas and single molecule localization density were quantified. On
average, IC displayed about 4 times less
DNA-associated signals. However, some
regions within IC displayed much lower
DNA-associated signal (indicated with
squared regions) [Żurek-Biesiada et al.,
2015].
Figure A.2: Comparison between objects
found in microscopy and genomics. Left
panel: model of chromosome territories
and their sub-domain, from [Cremer
and Cremer, 2001], using information
from a conventional light microscope.
Right panel: size of functional regions of
the genome (chromatin domains), modified from [Rao et al., 2014], generated
after analysis of data from HiC.
134
the periodic and dynamic structure of chromatin
Figure A.3: HiC data at various levels of
resolution. The chromosome 14 (approx
108 ) is partitioned into bins of 10000 (A),
40000 (B), 66000 (C), 100000 (D) pixels.
The figure shows that at a higher order or scale there is a large number of
intra-chromosomal interactions on the
chromosome 14, however when we start
to bin the chromosome into pixels of
smaller size these interactions start to
vanish.
Figure A.4: A comparison between
images of mitotic chromosome DNA
recorded using wide field or highresolutioncondition. (A) chromosomes
in anaphase, imaged by wide-field (bottom) and super-resolution localization
microscopy (top). (B) an inset presenting chromosomes visualized with
wide-field. (C) chromosomes shown in
panel B imaged by super-resolution localization microscopy. Data points were
blurred with using the value of their
respective localization precisions (see
Chapter 2). (D) point representation of
single molecules blurred in (C).
Figure A.5: Replicates of the results presented in figure 1.21.
135
A
Untreated
Reperfusion 5 min
Reperfusion 60 min
B
OND 1hr
Reperfusion 15 min
Reperfusion 240 min
Figure A.6: Fourier Ring Correlation
(FRC) analysis of DNA/SMLM data
(Figure and caption modified from
[Kirmes et al., 2015]). A) Representative normalised FRC curves for untreated, OND, and in recovering cells.
The red horizontal line designates the
1/7 threshold of the radially integrated
Fourier frequencies in accordance with
[Nieuwenhuizen et al., 2013]. B) Resolution estimates across all experimental
conditions based on the treshold determined in A.
Appendix B
Figure B.1: Relative localization of
SYCP1 and SYCP3 within the synaptonemal complex. Single molecule localization images of SYCP3 labelled with
Alexa 488 (A) and SYCP1 C-terminus labelled with Alexa 555 (B). (C) Dual color
image of SYCP3 and SYCP1. Insets in
A and B highlight the double strand nature of SYCP3 and SYCP1, respectively.
(D) shows an relative outline structure
of SYCP3 and SYCP1
Figure B.2: Validation of the chromatin
clusters (figure and caption modified
from [Prakash et al., 2015]). In order to
validate the clusters found in the SMLM
image of DNA (A), I compared it to a
randomly simulated dataset (B). I generated a binary mask of the nearest neighbour blurred image and then randomly
generated a number of points identical
to the SMLM image. (C) The local condensation in SMLM and random images
was characterised by detecting the 20500 nearest neighbours. In each case the
mean nearest neighbour distance was
found to be significantly shorter than
in the random simulated data-set. The
mean distance to the 500 nearest neighbour was 86 nm for chromatin and 110
nm for the random data.
138
the periodic and dynamic structure of chromatin
Figure B.3: SEM data validate the pattern of chromatin found with SMLM.
Chromatin shows a caterpillar-shape,
with several ring-shape clusters found
successively along the SC, similar to Figure 4.6B. Courtesy Wioleta Dudka.
Figure B.4: Chromatin rearrangement
upon condensation. After G2 phase
chromatin starts to condense around a
scaffold of proteins and to achieve a
maximal compaction. This moves the
silent and inacessible regions toward the
interior while the loose and active chromatin goes toward the exterior, making chromatin accessible for transcription. Three different scenarios exemplifying this transition is shown in (A) meiotic chromatin at the pachytene stage
in mouse, (B) organisation of polytene
chromosomes in Drosophila and (C) organisation of mitotic chromosomes in
anaphase.
Figure B.5: Helical structure of the
pachytene chromosomes using information from H3K27me3 clusters. (A)
SMLM image of H3K27me3 labelled
with Alexa 488. (B) SMLM image of
SYCP3 labelled with Alexa 555. (C) Two
color SMLM image of H3K27me3 and
SYCP3. (D) Clusters of H3K27me3 are
highlighted in different colours to reveal
an helicoidal structure. Twists of SYCP3
of same order as the H3K27me3 clusters are also observed (C). The size of
clusters follows the trend found computationally. Periodicity of twists is close
to 500nm. Scale bar: 1000 nm.
139
A
A1
A2
A3
Figure B.6: Spiralization of H3K9me3
confirmed by patterns of DNA (figure
and caption modified from [Prakash
et al., 2015]). Dense DNA clusters seem
to spiral at one of the ends of the SC,
confirming the spiralizartion patterns of
H3K9me3 found in Figure 4.11. The
average spread of DNA (A1) is around
500 nm. The cluster diameter in (A) was
found to be 223 nm. (A2) shows the histogram of the tangential distances taken
along the central axis of SYCP3. Chromatin clusters were found to occur with
a periodicity of 685 nm (A3). The blue
lines in the plot correspond to the 95
percent confidence bounds (+/- 0.08) of
the autocorrelation function.
140
the periodic and dynamic structure of chromatin
Figure B.7: Helicoidal nature of the
pachytene chromosomes.
Pairs of
pachytene chromosomes reminiscent
from the in situ state of chromatin are
displayed. Note that the helicoidal aspect of the paired chromosomes (left) are
similar to the one of the solitary chromosome (right).
141
B
A
H3K9me3
SYCP3
A1
Figure B.8: SYCP3 strands move apart
at non-centromeric end of the SC. The
centromeric ends of SYCP3 seem to
be closer to each other than the noncentromeric end of SYCP3 (A and B).
The average distance at the centromeric
end in A and B is found to be 142 nm
and 130 nm respectively, while the average distance at the non-centromeric
end is 171 nm and 190 nm respectively
(A1 and B1). H3K9me3 is normally
distributed at the centromeric end (figure and caption modified from [Prakash
et al., 2015]).
B1
Figure B.9: Characteristics of the epigenetic clusters identified to regulate
pachytene chromosomes.
Bibliography
Ernst Abbe. Beiträge zur Theorie des Mikroskops und der
mikroskopischen Wahrmehmung. Archiv für Mikroskopische Anatomie,
9:413–420, 1873. doi: 10.1007/BF02956173.
Heiner Albiez, Marion Cremer, Cinzia Tiberi, Lorella Vecchio, Lothar
Schermelleh, Sandra Dittrich, Katrin Küpper, Boris Joffe, Tobias
Thormeyer, Johann von Hase, et al. Chromatin domains and the
interchromatin compartment form structurally defined and functionally interacting nuclear networks. Chromosome Research, 14(7):
707–733, 2006.
VG Allfrey, R Faulkner, and AE Mirsky. Acetylation and methylation
of histones and their possible role in the regulation of rna synthesis.
Proceedings of the National Academy of Sciences of the United States of
America, 51(5):786, 1964.
Tobias Anton, Sebastian Bultmann, Heinrich Leonhardt, and Yolanda
Markaki. Visualization of specific dna sequences in living mouse
embryonic stem cells with a programmable fluorescent crispr/cas
system. Nucleus, 5(2):163–172, 2014.
Oswald T Avery, Colin M MacLeod, and Maclyn McCarty. Studies
on the chemical nature of the substance inducing transformation of
pneumococcal types induction of transformation by a desoxyribonucleic acid fraction isolated from pneumococcus type iii. The Journal
of experimental medicine, 79(2):137–158, 1944.
Daniel Axelrod, DE Koppel, J Schlessinger, Ei Elson, and WW Webb.
Mobility measurement by analysis of fluorescence photobleaching
recovery kinetics. Biophysical journal, 16(9):1055, 1976.
David Baddeley, Mark B Cannell, and Christian Soeller. Visualization
of localization microscopy data. Microscopy and Microanalysis, 16(01):
64–72, 2010.
John R Baker. The cell-theory: a restatement, history, and critique.
Quarterly Journal of Microscopical Science, 3(9):87–108, 1949.
Andrew J Bannister and Tony Kouzarides. Regulation of chromatin
by histone modifications. Cell research, 21(3):381–395, 2011.
Mariano Barbieri, Mita Chotalia, James Fraser, Liron-Mark Lavitas,
Josée Dostie, Ana Pombo, and Mario Nicodemi. Complexity of
144
the periodic and dynamic structure of chromatin
chromatin folding is captured by the strings and binders switch
model. Proceedings of the National Academy of Sciences, 109(40):16173–
16178, 2012.
A. Barski, S. Cuddapah, K. Cui, T.Y. Roh, D.E. Schones, Z. Wang,
G. Wei, I. Chepelev, and K. Zhao. High-resolution profiling of
histone methylations in the human genome. Cell, 129(4):823–837,
2007.
Jeremy M Baskin, Jennifer A Prescher, Scott T Laughlin, Nicholas J
Agard, Pamela V Chang, Isaac A Miller, Anderson Lo, Julian A
Codelli, and Carolyn R Bertozzi. Copper-free click chemistry for
dynamic in vivo imaging. Proceedings of the National Academy of
Sciences, 104(43):16793–16797, 2007.
Andrew Bassett, Sarah Cooper, Chenyi Wu, and Andrew Travers.
The folding and unfolding of eukaryotic chromatin. Current opinion
in genetics & development, 19(2):159–165, 2009.
Mark Bates, Sara A Jones, and Xiaowei Zhuang. Stochastic optical
reconstruction microscopy (storm): a method for superresolution
fluorescence imaging. Cold Spring Harbor Protocols, 2013(6):pdb–
top075143, 2013.
Frédéric Baudat, Yukiko Imai, and Bernard de Massy. Meiotic
recombination in mammals: localization and regulation. Nature
Reviews Genetics, 14(11):794–806, 2013.
George W Beadle and Edward L Tatum. Genetic control of biochemical reactions in neurospora. Proceedings of the National Academy of
Sciences of the United States of America, 27(11):499, 1941.
Ronald Berezney and Donald S Coffey. The nuclear protein matrix:
isolation, structure, and functions. Advances in enzyme regulation, 14:
63–100, 1976.
T Bernas, M Zarebski, RR Cook, and JW Dobrucki. Minimizing
photobleaching during confocal microscopy of fluorescent probes
bound to chromatin: role of anoxia and photon flux. Journal of
microscopy, 215(3):281–296, 2004.
Soledad Berríos, Catalina Manieu, Julio López-Fenner, Eliana
Ayarza, Jesús Page, Marisel González, Marcia Manterola, and Raúl
Fernández-Donoso. Robertsonian chromosomes and the nuclear
architecture of mouse meiotic prophase spermatocytes. Biological
research, 47(1):1, 2014.
Gerrit Best, Kirti Prakash, Martin Hagmann, Christoph Cremer, and
Udo Birk. Identify and localise: Algorithms for single molecule
localisation microscopy. In 18th International Microscopy Congress,
number ISBN 978-80-260-6720-7. Edited by Pavel Hozák, 2014.
E Betzig. Proposed method for molecular optical imaging. Optics
letters, 20(3):237–239, 1995.
bibliography 145
Eric Betzig, George H Patterson, Rachid Sougrat, O Wolf Lindwasser,
Scott Olenych, Juan S Bonifacino, Michael W Davidson, Jennifer
Lippincott-Schwartz, and Harald F Hess. Imaging intracellular
fluorescent proteins at nanometer resolution. Science, 313(5793):
1642–1645, 2006.
Alessandro Biancardi, Tarita Biver, Fernando Secco, and Benedetta
Mennucci. An investigation of the photophysical properties of minor
groove bound and intercalated dapi through quantum-mechanical
and spectroscopic tools. Physical Chemistry Chemical Physics, 15(13):
4596–4603, 2013.
Julie S Biteen, Michael A Thompson, Nicole K Tselentis, Grant R
Bowman, Lucy Shapiro, and WE Moerner. Super-resolution imaging in live caulobacter crescentus cells using photoswitchable eyfp.
Nature methods, 5(11):947–949, 2008.
Alistair N Boettiger, Bogdan Bintu, Jeffrey R Moffitt, Siyuan Wang,
Brian J Beliveau, Geoffrey Fudenberg, Maxim Imakaev, Leonid A
Mirny, Chao-ting Wu, and Xiaowei Zhuang. Super-resolution imaging reveals distinct chromatin folding for different epigenetic states.
Nature, 529(7586):418–422, 2016.
Manfred Bohn, Philipp Diesinger, Rainer Kaufmann, Yanina Weiland,
Patrick Müller, Manuel Gunkel, Alexa Von Ketteler, Paul Lemmer,
Michael Hausmann, Dieter W Heermann, et al. Localization microscopy reveals expression-dependent parameters of chromatin
nanostructure. Biophysical journal, 99(5):1358–1367, 2010.
Rasha E Boulos, Guénola Drillon, Françoise Argoul, Alain Arneodo,
and Benjamin Audit. Structural organization of human replication
timing domains. FEBS letters, 589(20PartA):2944–2957, 2015.
Theodor Boveri. Zellen-Studien: Die Befruchtung und Teilung des Eies
von Ascaris megalocephala. Verlag von Gustav Fischer, 1888.
Theodor Boveri. Ergebnisse über die Konstitution der chromatischen
Substanz des Zellkerns. Verlag von Gustav Fischer in Jena, 1904.
Theodor Boveri. Die Blastomerenkerne von Ascaris megalocephala und
die Theorie der Chromosomenindividualität. Engelmann, 1909.
Miguel R Branco and Ana Pombo. Intermingling of chromosome territories in interphase suggests role in translocations and
transcription-dependent associations. PLoS biology, 4(5):e138, 2006.
Miguel R Branco and Ana Pombo. Chromosome organization: new
facts, new models. Trends in cell biology, 17(3):127–134, 2007.
Jérôme Buard, Pauline Barthès, Corinne Grey, and Bernard de Massy.
Distinct histone modifications define initiation and repair of meiotic
recombination in the mouse. The EMBO journal, 28(17):2616–2624,
2009.
146
the periodic and dynamic structure of chromatin
Dylan T Burnette, Prabuddha Sengupta, Yuhai Dai, Jennifer
Lippincott-Schwartz, and Bechara Kachar. Bleaching/blinking assisted localization microscopy for superresolution imaging using
standard fluorescent molecules. Proceedings of the National Academy
of Sciences, 108(52):21081–21086, 2011.
David H Burns, James B Callis, Gary D Christian, and Ernest R
Davidson. Strategies for attaining superresolution using spectroscopic data as constraints. Applied optics, 24(2):154–161, 1985.
H Busch. The cell nucleus. 1974.
Harold G Callan. Lampbrush chromosomes, volume 36. Springer
Science & Business Media, 2012.
Cátia Diana Parente Caldeira Carvalho. Supramolecular biomimetic
binding of the DNA-dye Hoechst 33258 by a synthetic macrocycle. PhD
thesis, 2010.
Lyubomira Chakalova, Emmanuel Debrand, Jennifer A Mitchell,
Cameron S Osborne, and Peter Fraser. Replication and transcription:
shaping the landscape of the genome. Nature Reviews Genetics, 6(9):
669–677, 2005.
Tamir Chandra, Kristina Kirschner, Jean-Yves Thuret, Benjamin D
Pope, Tyrone Ryba, Scott Newman, Kashif Ahmed, Shamith A
Samarajiwa, Rafik Salama, Thomas Carroll, et al. Independence of
repressive histone marks and chromatin compaction during senescent heterochromatic layer formation. Molecular cell, 47(2):203–214,
2012.
Erwin Chargaff, Rakoma Lipshitz, and Charlotte Green. Composition of the desoxypentose nucleic acids of four genera of sea-urchin.
Journal of Biological Chemistry, 195(1):155–160, 1952.
Baohui Chen, Luke A Gilbert, Beth A Cimini, Joerg Schnitzbauer, Wei
Zhang, Gene-Wei Li, Jason Park, Elizabeth H Blackburn, Jonathan S
Weissman, Lei S Qi, et al. Dynamic imaging of genomic loci in
living human cells by an optimized crispr/cas system. Cell, 155(7):
1479–1491, 2013.
Nicolas Chenouard, Ihor Smal, Fabrice De Chaumont, Martin Maška,
Ivo F Sbalzarini, Yuanhao Gong, Janick Cardinale, Craig Carthel,
Stefano Coraluppi, Mark Winter, et al. Objective comparison of
particle tracking methods. Nature methods, 11(3):281, 2014.
Yves Clément. The evolution of base composition in mammalian genomes.
PhD thesis, Freie Universität Berlin, 2012.
DE Comings. Arrangement of chromatin in the nucleus. Human
genetics, 53(2):131–143, 1980.
Peter R Cook. A model for all genomes: the role of transcription
factories. Journal of molecular biology, 395(1):1–10, 2010.
bibliography 147
G Cosa, KS Focsaneanu, J R N McLean, JP McNamee, and JC Scaiano. Photophysical properties of fluorescent dna-dyes bound to
single-and double-stranded dna in aqueous buffered solution. Photochemistry and Photobiology, 73(6):585–599, 2001.
Christoph Cremer and Thomas Cremer. Considerations on a laserscanning-microscope with high resolution and depth of field. Microscopica acta, pages 31–44, 1978.
T Cremer and C Cremer. Rise, fall and resurrection of chromosome
territories: a historical perspective part ii. fall and resurrection of
chromosome territories during the 1950s to 1980s. part iii. chromosome territories and the functional nuclear architecture: experiments
and m. European Journal of Histochemistry, 50(4):223–272, 2009a.
Thomas Cremer and Christoph Cremer. Chromosome territories,
nuclear architecture and gene regulation in mammalian cells. Nature
reviews genetics, 2(4):292–301, 2001.
Thomas Cremer and Christoph Cremer. Rise, fall and resurrection
of chromosome territories: a historical perspective. part i. the rise
of chromosome territories. European Journal of Histochemistry, 50(3):
161–176, 2009b.
Thomas Cremer, Christoph Cremer, H Baumann, EK Luedtke,
K Sperling, V Teuber, and Christian Zorn. Rabl’s model of the
interphase chromosome arrangement tested in chinise hamster cells
by premature chromosome condensation and laser-uv-microbeam
experiments. Human genetics, 60(1):46–56, 1982a.
Thomas Cremer, Christoph Cremer, T Schneider, Hella Baumann,
Luc Hens, and M Kirsch-Volders. Analysis of chromosome positions in the interphase nucleus of chinese hamster cells by laser-uvmicroirradiation experiments. Human genetics, 62(3):201–209, 1982b.
Thomas Cremer, Katrin Küpper, Steffen Dietzel, and Stanislav Fakan.
Higher order chromatin architecture in the cell nucleus: on the way
from structure to function. Biology of the Cell, 96(8):555–567, 2004.
Thomas Cremer, Marion Cremer, Steffen Dietzel, Stefan Müller, Irina
Solovei, and Stanislav Fakan. Chromosome territories–a functional
nuclear landscape. Current opinion in cell biology, 18(3):307–316, 2006.
Thomas Cremer, Marion Cremer, Barbara Hübner, Hilmar Strickfaden, Daniel Smeets, Jens Popken, Michael Sterr, Yolanda Markaki,
Karsten Rippe, and Christoph Cremer. The 4d nucleome: Evidence
for a dynamic nuclear landscape based on co-aligned active and
inactive nuclear compartments. FEBS Letters, 2015.
Francis Crick, Leslie Barnett, Sydney Brenner, and Richard J WattsTobin. General nature of the genetic code for proteins. Macmillan Journals
Limited, 1961.
148
the periodic and dynamic structure of chromatin
Ralf Dahm. Discovering dna: Friedrich miescher and the early years
of nucleic acid research. Human genetics, 122(6):565–581, 2008.
HG Davies. Electron-microscope observations on the organization of
hetero-chromatin in certain cells. Journal of cell science, 3(1):129–150,
1968.
HG Davies and JV Small. Structural units in chromatin and their
orientation on membranes. Nature, 217:1122–1125, 1968.
Hesam Dehghani, Graham Dellaire, and David P Bazett-Jones. Organization of chromatin in the interphase mammalian cell. Micron,
36(2):95–108, 2005.
Job Dekker, Karsten Rippe, Martijn Dekker, and Nancy Kleckner.
Capturing chromosome conformation. Science, 295(5558):1306–1311,
2002.
Robert M Dickson, Andrew B Cubitt, Roger Y Tsien, and WE Moerner. On/off blinking and switching behaviour of single molecules
of green fluorescent protein. Nature, 388(6640):355–358, 1997.
Steffen Dietzel, Kourosh Zolghadr, Claudia Hepperger, and Andrew S Belmont. Differential large-scale chromatin compaction and
intranuclear positioning of transcribed versus non-transcribed transgene arrays containing β-globin regulatory sequences. Journal of cell
science, 117(19):4603–4614, 2004.
Jesse R Dixon, Siddarth Selvaraj, Feng Yue, Audrey Kim, Yan Li, Yin
Shen, Ming Hu, Jun S Liu, and Bing Ren. Topological domains in
mammalian genomes identified by analysis of chromatin interactions.
Nature, 485(7398):376–380, 2012.
Jennifer A Doudna and Emmanuelle Charpentier. The new frontier
of genome engineering with crispr-cas9. Science, 346(6213):1258096,
2014.
Boryana Doyle, Geoffrey Fudenberg, Maxim Imakaev, and Leonid A
Mirny. Chromatin loops as allosteric modulators of enhancerpromoter interactions. PLoS Comput Biol, 10(10):e1003867, 2014.
J Dubochet, M Adrian, P Schultz, and P Oudet. Cryo-electron
microscopy of vitrified sv40 minichromosomes: the liquid drop
model. The EMBO journal, 5(3):519, 1986.
Alexander Egner, Claudia Geisler, Claas Von Middendorff, Hannes
Bock, Dirk Wenzel, Rebecca Medda, Martin Andresen, Andre C Stiel,
Stefan Jakobs, Christian Eggeling, et al. Fluorescence nanoscopy in
whole cells by asynchronous localization of photoswitching emitters.
Biophysical journal, 93(9):3285–3290, 2007.
AC Everid, JV Small, and HG Davies. Electron-microscope observations on the structure of condensed chromatin: evidence for orderly
arrays of unit threads on the surface of chicken erythrocyte nuclei.
Journal of cell science, 7(1):35–48, 1970.
bibliography 149
Stanislav Fakan. The functional architecture of the nucleus as analysed by ultrastructural cytochemistry. Histochemistry and cell biology,
122(2):83–93, 2004.
Adriane Ficq and Crodowaldo Pavan. Autoradiography of polytene
chromosomes of rhynchosciara angelae at different stages of larval
development. 1957.
Guillaume J Filion, Joke G van Bemmel, Ulrich Braunschweig,
Wendy Talhout, Jop Kind, Lucas D Ward, Wim Brugman, Inês J
de Castro, Ron M Kerkhoven, Harmen J Bussemaker, et al. Systematic protein location mapping reveals five principal chromatin types
in drosophila cells. Cell, 143(2):212–224, 2010.
JT Finch and A Klug. Solenoidal model for superstructure in chromatin. Proceedings of the National Academy of Sciences, 73(6):1897–1901,
1976.
W Flemming. 1: Beiträge zur kenntnis der zelle und ihrer lebenserscheinungen. iii. Teil. Arch. f. mikr. Anat., Bd, 20, 1882.
Jonas Fölling, Mariano Bossi, Hannes Bock, Rebecca Medda, Christian A Wurm, Birka Hein, Stefan Jakobs, Christian Eggeling, and
Stefan W Hell. Fluorescence nanoscopy by ground-state depletion
and single-molecule return. Nature methods, 5(11):943–945, 2008.
John H Frenster, Vincent G Allfrey, and Alfred E Mirsky. Repressed
and active chromatin isolated from interphase lymphocytes. Proceedings of the National Academy of Sciences, 50(6):1026–1032, 1963.
Susanna Maria Früh, Ingmar Schoen, Jonas Ries, and Viola Vogel.
Molecular architecture of native fibronectin fibrils. Nature communications, 6, 2015.
Eden Fussner, Reagan W Ching, and David P Bazett-Jones. Living
without 30nm chromatin fibers. Trends in biochemical sciences, 36(1):
1–6, 2011.
Joseph Gall. Chromosome fibers from an interphase nucleus. Science,
139(3550):120–121, 1963.
Joseph G Gall. Differential synthesis of the genes for ribosomal rna
during amphibian oögenesis. Proceedings of the National Academy of
Sciences of the United States of America, 60(2):553, 1968a.
Joseph G Gall. The genes for ribosomal rna during oogenesis. Genetics, 61(1):Suppl–121, 1968b.
Joseph G Gall. Are lampbrush chromosomes unique to meiotic cells?
Chromosome research, 20(8):905–909, 2012.
Joseph G Gall and Mary Lou Pardue. Formation and detection of
rna-dna hybrid molecules in cytological preparations. Proceedings of
the National Academy of Sciences, 63(2):378–383, 1969.
150
the periodic and dynamic structure of chromatin
Markus Grammel and Howard C Hang. Chemical reporters for
biological discovery. Nature chemical biology, 9(8):475–484, 2013.
John B Gurdon, Thomas R Elsdale, and Michel Fischberg. Sexually
mature individuals of xenopus laevis from the transplantation of
single somatic nuclei. 1958.
Mats GL Gustafsson. Surpassing the lateral resolution limit by a
factor of two using structured illumination microscopy. Journal of
microscopy, 198(2):82–87, 2000.
Mats GL Gustafsson, Lin Shao, Peter M Carlton, CJ Rachel Wang,
Inna N Golubovskaya, W Zacheus Cande, David A Agard, and
John W Sedat. Three-dimensional resolution doubling in widefield fluorescence microscopy by structured illumination. Biophysical
journal, 94(12):4957–4970, 2008.
Geoffrey H Haggis. Sample preparation for electron microscopy
of internal cell structure. Microscopy research and technique, 22(2):
151–159, 1992.
GH Haggis and JB Pawley. Freeze-fracture of 3t3 cells for highresolution scanning electron microscopy. Journal of microscopy, 150
(3):211–218, 1988.
Martin Hagmann, Kirti Prakash, Gerrit Best, Rainer Kaufmann, Udo
Birk, and Christoph Cremer. Drift correction strategies for single
molecule localisation microscopy. In 18th International Microscopy
Congress, number ISBN 978-80-260-6720-7. Edited by Pavel Hozák,
2014.
Martin Hagmann, Kirti Prakash, and Christoph Cremer. Visualisation enhancement for single molecule reconstructions using wignerseitz cells. In preparation, 2016.
Saher Sue Hammoud, Diana HP Low, Chongil Yi, Douglas T Carrell,
Ernesto Guccione, and Bradley R Cairns. Chromatin and transcription transitions of mammalian adult germline stem cells and
spermatogenesis. Cell Stem Cell, 15(2):239–253, 2014.
Diana C Hargreaves and Gerald R Crabtree. Atp-dependent chromatin remodeling: genetics, genomics and mechanisms. Cell research,
21(3):396–420, 2011.
Brian Harmon and John Sedat. Cell-by-cell dissection of gene expression and chromosomal interactions reveals consequences of nuclear
reorganization. PLoS Biol, 3(3):e67, 2005.
Elizabeth D Hay and JP Revel. The fine structure of the dnp component of the nucleus: An electron microscopic study utilizing
autoradiography to localize dna synthesis. The Journal of cell biology,
16(1):29–51, 1963.
bibliography 151
Mike Heilemann, Sebastian Van De Linde, Mark Schüttpelz, Robert
Kasper, Britta Seefeldt, Anindita Mukherjee, Philip Tinnefeld, and
Markus Sauer. Subdiffraction-resolution fluorescence imaging with
conventional fluorescent probes. Angewandte Chemie International
Edition, 47(33):6172–6176, 2008.
Rainer Heintzmann and Christoph Cremer. Laterally modulated
excitation microscopy: improvement of resolution by using a diffraction grating. In BiOS Europe’98, pages 185–196. International Society
for Optics and Photonics, 1999.
Stefan W Hell and Jan Wichmann. Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion
fluorescence microscopy. Optics letters, 19(11):780–782, 1994.
HH Heng, John W Chamberlain, Xiao-Mei Shi, Barbara Spyropoulos,
Lap-Chee Tsui, and Peter B Moens. Regulation of meiotic chromatin loop size by chromosomal position. Proceedings of the National
Academy of Sciences, 93(7):2795–2800, 1996.
Abrahan Hernández-Hernández, Rosario Ortiz, Ernestina Ubaldo,
Olga M Echeverría Martínez, Gerardo H Vázquez-Nin, and Félix
Recillas-Targa. Synaptonemal complex stability depends on repressive histone marks of the lateral element-associated repeat sequences.
Chromosoma, 119(1):41–58, 2010.
Abrahan Hernández-Hernández, Gerardo H Vázquez-Nin, and
Rosario Ortiz Hernádez. Epigenetics of the Synaptonemal Complex.
INTECH Open Access Publisher, 2012.
Oscar Hertwig. Beiträge zur kenntniss der bildung, befruchtung und
theilung des thierischen eies... W. Engelmann, 1875.
Oskar Hertwig. Lehrbuch der Entwicklungsgeschichte des Menschen und
der Wirbelthiere. G. Fischer, 1898.
Samuel T Hess, Thanu PK Girirajan, and Michael D Mason. Ultrahigh resolution imaging by fluorescence photoactivation localization
microscopy. Biophysical journal, 91(11):4258, 2006.
Tomas Hirschfeld. Optical microscopic observation of single small
molecules. Applied optics, 15(12):2965–2966, 1976.
Osamu Hoshi and Tatsuo Ushiki. Three-dimensional structure of
g-banded human metaphase chromosomes observed by atomic force
microscopy. Archives of histology and cytology, 64(5):475–482, 2001.
Patrick D Hsu, Eric S Lander, and Feng Zhang. Development and
applications of crispr-cas9 for genome engineering. Cell, 157(6):
1262–1278, 2014.
Jerard Hurwitz. The discovery of rna polymerase. Journal of Biological
Chemistry, 280(52):42477–42485, 2005.
152
the periodic and dynamic structure of chromatin
Martin Hussels and Marc Brecht. Effect of glycerol and pva on the
conformation of photosystem i. Biochemistry, 50(18):3628–3637, 2011.
Jonas Ibn-Salem, Sebastian Köhler, Michael I Love, Ho-Ryun Chung,
Ni Huang, Matthew E Hurles, Melissa Haendel, Nicole L Washington, Damian Smedley, Christopher J Mungall, et al. Deletions of
chromosomal regulatory boundaries are associated with congenital.
Genome Biology, 2014.
Francisco J Iborra, Ana Pombo, Dean A Jackson, and Peter R Cook.
Active rna polymerases are localized within discrete transcription
factories in human nuclei. Journal of cell science, 109(6):1427–1436,
1996.
François Jacob and Jacques Monod. Genetic regulatory mechanisms
in the synthesis of proteins. Journal of molecular biology, 3(3):318–356,
1961.
Liu Jianzhuang, Li Wenqing, and Tian Yupeng. Automatic thresholding of gray-level pictures using two-dimension otsu method. In
Circuits and Systems, 1991. Conference Proceedings, China., 1991 International Conference on, pages 325–327. IEEE, 1991.
Amber Buescher Johnson and Michelle Craig Barton. Hypoxiainduced and stress-specific changes in chromatin structure and
function. Mutation Research/Fundamental and Molecular Mechanisms of
Mutagenesis, 618(1):149–162, 2007.
Krishanpal Karmodiya, Arnaud R Krebs, Mustapha OuladAbdelghani, Hiroshi Kimura, and Laszlo Tora. H3k9 and h3k14
acetylation co-occur at many gene regulatory elements, while
h3k14ac marks a subset of inactive inducible promoters in mouse
embryonic stem cells. BMC genomics, 13(1):424, 2012.
Eric Karsenti. Self-organization in cell biology: a brief history. Nature
Reviews Molecular Cell Biology, 9(3):255–262, 2008.
Rainer Kaufmann, Jörg Piontek, Frederik Grüll, Manfred Kirchgessner, Jan Rossa, Hartwig Wolburg, Ingolf E Blasig, and Christoph
Cremer. Visualization and quantitative analysis of reconstituted
tight junctions using localization microscopy. PloS one, 7(2):e31128,
2012.
Ina Kirmes, Aleksander Szczurek, Kirti Prakash, Iryna Charapitsa,
Christina Heiser, Michael Musheev, Florian Schock, Karolina Fornalczyk, Dongyu Ma, Udo Birk, et al. A transient ischemic environment
induces reversible compaction of chromatin. Genome biology, 16(1):
1–19, 2015.
Teresa Klein, Anna Löschberger, Sven Proppert, Steve Wolter, Sebastian van de Linde, and Markus Sauer. Live-cell dstorm with snap-tag
fusion proteins. Nature methods, 8(1):7–9, 2011.
bibliography 153
Roger D Kornberg. Chromatin structure: a repeating unit of histones
and dna. Science, 184(4139):868–871, 1974.
Roger D Kornberg and Yahli Lorch. Twenty-five years of the nucleosome, fundamental particle of the eukaryote chromosome. Cell, 98
(3):285–294, 1999.
A Kossel and EL Kennaway. Über nitroclupein. Hoppe-Seyler´ s
Zeitschrift für physiologische Chemie, 72(5-6):486–489, 1911.
Albrecht Kossel. Zur chemie des zellkerns. Zeitschrift für physiologische Chemie, 7(1):7–22, 1883.
Albrecht Kossel. Ueber einen peptonartigen bestandtheil des zellkerns. Zeitschrift für physiologische Chemie, 8(6):511–515, 1884.
Marilyn Kozak. Compilation and analysis of sequences upstream
from the translational start site in eukaryotic mrnas. Nucleic acids
research, 12(2):857–872, 1984.
Christian Lanctôt, Thierry Cheutin, Marion Cremer, Giacomo Cavalli,
and Thomas Cremer. Dynamic genome architecture in the nuclear
space: regulation of gene expression in three dimensions. Nature
Reviews Genetics, 8(2):104–115, 2007.
Eric S Lander, Lauren M Linton, Bruce Birren, Chad Nusbaum,
Michael C Zody, Jennifer Baldwin, Keri Devon, Ken Dewar, Michael
Doyle, William FitzHugh, et al. Initial sequencing and analysis of
the human genome. Nature, 409(6822):860–921, 2001.
Pennina R Langer-Safer, Michael Levine, and David C Ward. Immunological method for mapping genes on drosophila polytene
chromosomes. Proceedings of the National Academy of Sciences, 79(14):
4381–4385, 1982.
P Lemmer, M Gunkel, D Baddeley, R Kaufmann, A Urich, Y Weiland,
J Reymann, P Müller, M Hausmann, and C Cremer. Spdm: light
microscopy with single-molecule resolution at the nanoscale. Applied
Physics B, 93(1):1–12, 2008.
Paul Lemmer, Manuel Gunkel, Yanina Weiland, Patrick Müller,
David Baddeley, Rainer Kaufmann, A Urich, H Eipel, Roman Amberger, Michael Hausmann, et al. Using conventional fluorescent
markers for far-field fluorescence localization nanoscopy allows resolution in the 10-nm range. Journal of microscopy, 235(2):163–171,
2009.
Christoph Lengauer, Eric D Green, and Thomas Cremer. Fluorescence in situ hybridization of yac clones after alu-pcr amplification.
Genomics, 13(3):826–828, 1992.
Ea Bo Lewis. The relation of repeats to position effect in drosophila
melanogaster. Genetics, 30(2):137, 1945.
154
the periodic and dynamic structure of chromatin
Michael Lichten. Meiotic recombination: breaking the genome to
save it. Current Biology, 11(7):R253–R256, 2001.
Keith Lidke, Bernd Rieger, Thomas Jovin, and Rainer Heintzmann.
Superresolution by localization of quantum dots using blinking
statistics. Optics express, 13(18):7052–7062, 2005.
Erez Lieberman-Aiden, Nynke L van Berkum, Louise Williams,
Maxim Imakaev, Tobias Ragoczy, Agnes Telling, Ido Amit, Bryan R
Lajoie, Peter J Sabo, Michael O Dorschner, et al. Comprehensive
mapping of long-range interactions reveals folding principles of the
human genome. science, 326(5950):289–293, 2009.
Jennifer Lippincott-Schwartz and George H Patterson. Photoactivatable fluorescent proteins for diffraction-limited and super-resolution
imaging. Trends in cell biology, 19(11):555–565, 2009.
Ning Liu, Michael Fromm, and Zoya Avramova. H3k27me3 and
h3k4me3 chromatin environment at super-induced dehydration
stress memory genes of arabidopsis thaliana. Molecular plant, 7
(3):502–513, 2014.
Stavros Lomvardas, Gilad Barnea, David J Pisapia, Monica Mendelsohn, Jennifer Kirkland, and Richard Axel. Interchromosomal interactions and olfactory receptor choice. Cell, 126(2):403–413, 2006.
Karolin Luger, Armin W Mäder, Robin K Richmond, David F Sargent,
and Timothy J Richmond. Crystal structure of the nucleosome core
particle at 2.8 å resolution. Nature, 389(6648):251–260, 1997.
Vittorio Luzzati and A Nicolaieff. The structure of nucleohistones
and nucleoprotamines. Journal of molecular biology, 7(2):142–163, 1963.
Hong Ma, Jagath Samarabandu, Rekandu S Devdhar, Raj Acharya,
Ping-chin Cheng, Chunling Meng, and Ronald Berezney. Spatial
and temporal dynamics of dna replication sites in mammalian cells.
The Journal of cell biology, 143(6):1415–1425, 1998.
Kazuhiro Maeshima, Ryosuke Imai, Sachiko Tamura, and Tadasu
Nozaki. Chromatin as dynamic 10-nm fibers. Chromosoma, 123(3):
225–237, 2014.
Suliana Manley, Jennifer M Gillette, George H Patterson, Hari Shroff,
Harald F Hess, Eric Betzig, and Jennifer Lippincott-Schwartz. Highdensity mapping of single-molecule trajectories with photoactivated
localization microscopy. Nature methods, 5(2):155–157, 2008.
Y Markaki, M Gunkel, L Schermelleh, S Beichmanis, J Neumann,
M Heidemann, H Leonhardt, D Eick, C Cremer, and T Cremer.
Functional nuclear organization of transcription and dna replication a topographical marriage between chromatin domains and the
interchromatin compartment. In Cold Spring Harbor symposia on
quantitative biology, volume 75, pages 475–492. Cold Spring Harbor
Laboratory Press, 2010.
bibliography 155
Robert Mayer, Alessandro Brero, Johann Von Hase, Timm Schroeder,
Thomas Cremer, and Steffen Dietzel. Common themes and cell type
specific variations of higher order chromatin arrangements in the
mouse. BMC cell biology, 6(1):44, 2005.
Richard R Meehan, Joe D Lewis, Stewart McKay, Elke L Kleiner, and
Adrian P Bird. Identification of a mammalian protein that binds
specifically to dna containing methylated cpgs. Cell, 58(3):499–507,
1989.
F Miescher. Hoppe-seyler’s medicinisch-chemische untersuchungen
1871, 4, 441-460. Ueber die chemische Zusammensetzung der Eiterzellen,
1871.
Friedrich Miescher. Die Spermatozoen einiger Wirbelthiere: ein Beitrag
zur Histochemie. Birkhäuser, 1874.
Friedrich Miescher. Die histochemischen und physiologischen Arbeiten
von Friedrich Miescher. Vogel, 1897.
Ondrej Mihola, Zdenek Trachtulec, Cestmir Vlcek, John C Schimenti,
and Jiri Forejt. A mouse speciation gene encodes a meiotic histone
h3 methyltransferase. Science, 323(5912):373–375, 2009.
Tarjei S Mikkelsen, Manching Ku, David B Jaffe, Biju Issac, Erez
Lieberman, Georgia Giannoukos, Pablo Alvarez, William Brockman,
Tae-Kyung Kim, Richard P Koche, et al. Genome-wide maps of
chromatin state in pluripotent and lineage-committed cells. Nature,
448(7153):553–560, 2007.
AE Mirsky and Hans Ris. The desoxyribonucleic acid content of
animal cells and its evolutionary significance. The Journal of general
physiology, 34(4):451, 1951.
Tom Misteli. The concept of self-organization in cellular architecture.
The Journal of cell biology, 155(2):181–186, 2001.
Yusuke Miyanari, Céline Ziegler-Birling, and Maria-Elena TorresPadilla. Live visualization of chromatin dynamics with fluorescent
tales. Nature structural & molecular biology, 20(11):1321–1324, 2013.
William Esco Moerner and Lothar Kador. Optical detection and
spectroscopy of single molecules in a solid. Physical Review Letters,
62(21):2535, 1989.
Ariane Monneron and W Bernhard. Fine structural organization
of the interphase nucleus in some mammalian cells. Journal of
ultrastructure research, 27(3):266–288, 1969.
Felipe Mora-Bermúdez and Jan Ellenberg. Measuring structural
dynamics of chromosomes in living cells by fluorescence microscopy.
Methods, 41(2):158–167, 2007.
Montrose J Moses, LB Russel, and NL Cacheiro. Mouse chromosome
translocations: visualization and analysis by electron microscopy of
the synaptonemal complex. Science, 196(4292):892–894, 1977.
156
the periodic and dynamic structure of chromatin
Patrick Müller, Eberhard Schmitt, Anette Jacob, Jörg Hoheisel, Rainer
Kaufmann, Christoph Cremer, and Michael Hausmann. Combo-fish
enables high precision localization microscopy as a prerequisite
for nanostructure analysis of genome loci. International journal of
molecular sciences, 11(10):4094–4105, 2010.
Natalia Naumova, Maxim Imakaev, Geoffrey Fudenberg, Ye Zhan,
Bryan R Lajoie, Leonid A Mirny, and Job Dekker. Organization of
the mitotic chromosome. Science, 342(6161):948–953, 2013.
Andrew Neice. Chapter 3 - methods and limitations of subwavelength imaging. In Peter W. Hawkes, editor, Advances in Imaging and
Electron Physics, volume 163, pages 117 – 140. Elsevier, 2010. doi:
http://dx.doi.org/10.1016/S1076-5670(10)63003-0.
Robert PJ Nieuwenhuizen, Keith A Lidke, Mark Bates, Daniela Leyton Puig, David Grünwald, Sjoerd Stallinga, and Bernd Rieger. Measuring image resolution in optical nanoscopy. Nature methods, 10(6):
557–562, 2013.
Ada L Olins and Donald E Olins. Spheroid chromatin units (ν
bodies). Science, 183(4122):330–332, 1974.
Donald E Olins and Ada L Olins. Chromatin history: our view from
the bridge. Nature reviews Molecular cell biology, 4(10):809–814, 2003.
Cameron S Osborne, Lyubomira Chakalova, Karen E Brown, David
Carter, Alice Horton, Emmanuel Debrand, Beatriz Goyenechea, Jennifer A Mitchell, Susana Lopes, Wolf Reik, et al. Active genes dynamically colocalize to shared sites of ongoing transcription. Nature
genetics, 36(10):1065–1071, 2004.
Joseph Ostashevsky. A polymer model for the structural organization of chromatin loops and minibands in interphase chromosomes.
Molecular biology of the cell, 9(11):3031–3040, 1998.
P Oudet, M Gross-Bellard, and P Chambon. Electron microscopic
and biochemical evidence that chromatin structure is a repeating
unit. Cell, 4(4):281–300, 1975.
Gungor Ozer, Antoni Luque, and Tamar Schlick. The chromatin fiber:
multiscale problems and approaches. Current opinion in structural
biology, 31:124–139, 2015.
Florian M Pauler, Mathew A Sloane, Ru Huang, Kakkad Regha,
Martha V Koerner, Ido Tamir, Andreas Sommer, Andras Aszodi,
Thomas Jenuwein, and Denise P Barlow. H3k27me3 forms blocs over
silent genes and intergenic regions and specifies a histone banding
pattern on a mouse autosomal chromosome. Genome research, 19(2):
221–233, 2009.
Linus Pauling and Robert B Corey. A proposed structure for the
nucleic acids. Proceedings of the National Academy of Sciences, 39(2):
84–97, 1953.
bibliography 157
Linus Pauling, Robert B Corey, and Herman R Branson. The structure of proteins: two hydrogen-bonded helical configurations of the
polypeptide chain. Proceedings of the National Academy of Sciences, 37
(4):205–211, 1951.
Mark Petronczki, Maria F Siomos, and Kim Nasmyth. Un menage a
quatre: the molecular biology of chromosome segregation in meiosis.
Cell, 112(4):423–440, 2003.
M Piterburg, H Panet, and A Weiss. Photoconversion of dapi following uv or violet excitation can cause dapi to fluoresce with blue or
cyan excitation. Journal of microscopy, 246(1):89–95, 2012.
AW Pollister. Nucleoproteins of the nucleus. Exp Cell Res, Suppl II,
pages 59–74, 1952.
Benjamin D Pope, Tyrone Ryba, Vishnu Dileep, Feng Yue, Weisheng
Wu, Olgert Denas, Daniel L Vera, Yanli Wang, R Scott Hansen,
Theresa K Canfield, et al. Topologically associating domains are
stable units of replication-timing regulation. Nature, 515(7527):402–
405, 2014.
Jens Popken, Alexander Graf, Stefan Krebs, Helmut Blum, Volker J
Schmid, Axel Strauss, Tuna Guengoer, Valeri Zakhartchenko, Eckhard Wolf, and Thomas Cremer. Remodeling of the nuclear envelope
and lamina during bovine preimplantation development and its
functional implications. PLoS One, 2015.
Keith R Porter, Albert Claude, and Ernest F Fullam. A study of
tissue culture cells by electron microscopy methods and preliminary
observations. The Journal of experimental medicine, 81(3):233–246, 1945.
Kirti Prakash. A binary combinatorial histone code. Master thesis,
Aalto University, 2012.
Kirti Prakash, David Fournier, Stefan Redl, Gerrit Best, Máté Borsos,
Vijay K Tiwari, Kikuë Tachibana-Konwalski, René F Ketting, Sapun H
Parekh, Christoph Cremer, et al. Superresolution imaging reveals
structurally distinct periodic patterns of chromatin along pachytene
chromosomes. Proceedings of the National Academy of Sciences, 112(47):
14635–14640, 2015.
Douglas C Prasher, Virginia K Eckenrode, William W Ward, Frank G
Prendergast, and Milton J Cormier. Primary structure of the aequorea victoria green-fluorescent protein. Gene, 111(2):229–233, 1992.
Florencia Pratto, Kevin Brick, Pavel Khil, Fatima Smagulova,
Galina V Petukhova, and R Daniel Camerini-Otero. Recombination initiation maps of individual human genomes. Science, 346
(6211):1256442, 2014.
Jan Evangelista Purkyně. Symbolae ad ovi avium historiam ante incubationem. 1830.
158
the periodic and dynamic structure of chromatin
Edmond Puvion and Francine Puvion-Dutilleul. Ultrastructure
of the nucleus in relation to transcription and splicing: roles of
perichromatin fibrils and interchromatin granules. Experimental cell
research, 229(2):217–225, 1996.
Indika Rajapakse and Mark Groudine. On emerging nuclear order.
The Journal of cell biology, 192(5):711–721, 2011.
Suhas SP Rao, Miriam H Huntley, Neva C Durand, Elena K Stamenova, Ivan D Bochkov, James T Robinson, Adrian L Sanborn, Ido
Machol, Arina D Omer, Eric S Lander, et al. A 3d map of the human genome at kilobase resolution reveals principles of chromatin
looping. Cell, 159(7):1665–1680, 2014.
JB Rattner, M Goldsmith, and BA Hamkalo. Chromatin organization
during meiotic prophase ofbombyx mori. Chromosoma, 79(2):215–224,
1980.
Lord Rayleigh. Xv. on the theory of optical images, with special
reference to the microscope. The London, Edinburgh, and Dublin
Philosophical Magazine and Journal of Science, 42(255):167–195, 1896.
Jürgen Reymann, David Baddeley, Manuel Gunkel, Paul Lemmer,
Werner Stadter, Thibaud Jegou, Karsten Rippe, Christoph Cremer,
and Udo Birk. High-precision structural analysis of subnuclear
complexes in fixed and live cells via spatially modulated illumination
(smi) microscopy. Chromosome Research, 16(3):367–382, 2008.
Maria Aurelia Ricci, Carlo Manzo, María Filomena García-Parajo,
Melike Lakadamyali, and Maria Pia Cosma. Chromatin fibers are
formed by heterogeneous groups of nucleosomes in vivo. Cell, 160
(6):1145–1158, 2015.
TJ Richmond, JT Finch, B Rushton, D Rhodes, and A Klug. Structure
of the nucleosome core particle at 7 a resolution. Nature, 311(5986):
532–537, 1983.
Michael J Rust, Mark Bates, and Xiaowei Zhuang. Stochastic optical
reconstruction microscopy (storm) provides sub-diffraction-limit
image resolution. Nature methods, 3(10):793, 2006.
Adrian L Sanborn, Suhas SP Rao, Su-Chen Huang, Neva C Durand,
Miriam H Huntley, Andrew I Jewett, Ivan D Bochkov, Dharmaraj
Chinnappan, Ashok Cutkosky, Jian Li, et al. Chromatin extrusion
explains key features of loop and domain formation in wild-type
and engineered genomes. Proceedings of the National Academy of
Sciences, page 201518552, 2015.
Frederick Sanger. Determination of nucleotide sequences in dna.
Bioscience reports, 1(1):3–18, 1981.
Aya Sato, Berith Isaac, Carolyn M Phillips, Regina Rillo, Peter M
Carlton, David J Wynne, Roshni A Kasad, and Abby F Dernburg.
Cytoskeletal forces span the nuclear envelope to coordinate meiotic
chromosome pairing and synapsis. Cell, 139(5):907–919, 2009.
bibliography 159
Margit Schardin, Thomas Cremer, HD Hager, and M Lang. Specific
staining of human chromosomes in chinese hamster x man hybrid
cell lines demonstrates interphase chromosome territories. Human
genetics, 71(4):281–287, 1985.
Ingmar Schoen, Jonas Ries, Enrico Klotzsch, Helge Ewers, and Viola
Vogel. Binding-activated localization microscopy of dna structures.
Nano letters, 11(9):4008–4011, 2011.
Erwin Schrödinger. What is life, 1945.
Katharina Schücker, Thorge Holm, Christian Franke, Markus Sauer,
and Ricardo Benavente. Elucidation of synaptonemal complex organization by super-resolution imaging with isotropic resolution.
Proceedings of the National Academy of Sciences, 112(7):2029–2033, 2015.
Prabuddha Sengupta, Schuyler B van Engelenburg, and Jennifer
Lippincott-Schwartz. Superresolution imaging of biological systems
using photoactivated localization microscopy. Chemical reviews, 114
(6):3189–3202, 2014.
Tom Sexton, David Umlauf, Sreenivasulu Kurukuti, and Peter Fraser.
The role of transcription factories in large-scale structure and dynamics of interphase chromatin. In Seminars in cell & developmental
biology, volume 18, pages 691–697. Elsevier, 2007.
Sigal Shachar, Ty C Voss, Gianluca Pegoraro, Nicholas Sciascia,
and Tom Misteli. Identification of gene positioning factors using
high-throughput imaging mapping. Cell, 162(4):911–923, 2015.
CJR Sheppard and T Wilson. Effects of high angles of convergence
on v (z) in the scanning acoustic microscope. Applied Physics Letters,
38(11):858–859, 1981.
E Brooks Shera, Newton K Seitzinger, Lloyd M Davis, Richard A
Keller, and Steven A Soper. Detection of single fluorescent molecules.
Chemical Physics Letters, 174(6):553–557, 1990.
Ali Shilatifard. Chromatin modifications by methylation and ubiquitination: implications in the regulation of gene expression. Annu.
Rev. Biochem., 75:243–269, 2006.
Hari Shroff, Catherine G Galbraith, James A Galbraith, and Eric
Betzig. Live-cell photoactivated localization microscopy of nanoscale
adhesion dynamics. Nature methods, 5(5):417–423, 2008.
Alex Small and Shane Stahlheber. Fluorophore localization algorithms for super-resolution microscopy. Nature methods, 11(3):267–
279, 2014.
Alexander R Small and Raghuveer Parthasarathy. Superresolution
localization methods. Annual review of physical chemistry, 65:107–125,
2014.
160
the periodic and dynamic structure of chromatin
Daniel Smeets, Yolanda Markaki, Volker J Schmid, Felix Kraus, Anna
Tattermusch, Andrea Cerase, Michael Sterr, Susanne Fiedler, Justin
Demmerle, Jens Popken, et al. Three-dimensional super-resolution
microscopy of the inactive x chromosome territory reveals a collapse
of its active nuclear compartment harboring distinct xist rna foci.
Epigenetics & chromatin, 7(8), 2014.
Kendric C Smith. Dose dependent decrease in extractability of dna
from bacteria following irradiation with ultraviolet light or with visible light plus dye. Biochemical and biophysical research communications,
8(3):157–163, 1962.
Vérane Sommermeyer, Claire Béneut, Emmanuel Chaplais,
Maria Elisabetta Serrentino, and Valérie Borde. Spp1, a member
of the set1 complex, promotes meiotic dsb formation in promoters
by tethering histone h3k4 methylation sites to chromosome axes.
Molecular cell, 49(1):43–54, 2013.
Charalampos G Spilianakis, Maria D Lalioti, Terrence Town,
Gap Ryol Lee, and Richard A Flavell. Interchromosomal associations between alternatively expressed loci. Nature, 435(7042):637–645,
2005.
Stephen M Stack, David B Brown, and WC Dewey. Visualization of
interphase chromosomes. Journal of cell science, 26(1):281–299, 1977.
Edgar Stedman and Ellen Stedman. The basic proteins of cell nuclei.
Philosophical Transactions of the Royal Society of London B: Biological
Sciences, 235(630):565–595, 1951.
A Strutevant. The linear arrangement of six sex-linked factors indrosophila as shown by their mode of association. Molecular and
General Genetics MGG, 10(1):293–294, 1913.
Heidi Sutherland and Wendy A Bickmore. Transcription factories:
gene expression in unions? Nature Reviews Genetics, 10(7):457–466,
2009.
Johanna Liinamaria Syrjänen, Luca Pellegrini, and Owen Richard
Davies. A molecular model for the role of sycp3 in meiotic chromosome organisation. Elife, 3:e02963, 2014.
Aleksander T Szczurek, Kirti Prakash, Hyun-Keun Lee, Dominika J
Żurek-Biesiada, Gerrit Best, Martin Hagmann, Jurek W Dobrucki,
Christoph Cremer, and Udo Birk. Single molecule localization microscopy of the distribution of chromatin using hoechst and dapi
fluorescent probes. Nucleus, 5(4):331–340, 2014.
Katharina Thanisch, Katrin Schneider, Robert Morbitzer, Irina
Solovei, Thomas Lahaye, Sebastian Bultmann, and Heinrich Leonhardt. Targeting and tracing of specific dna sequences with dtales in
living cells. Nucleic acids research, page gkt1348, 2013.
bibliography 161
F Thoma, Th Koller, and A Klug. Involvement of histone h1 in
the organization of the nucleosome and of the salt-dependent superstructures of chromatin. The Journal of cell biology, 83(2):403–427,
1979.
Russell E Thompson, Daniel R Larson, and Watt W Webb. Precise
nanometer localization analysis for individual fluorescent probes.
Biophysical journal, 82(5):2775–2783, 2002.
Frank Uhlmann. A silent revolution in chromosome biology. Nature
Reviews Molecular Cell Biology, 15(7):431–431, 2014.
Tatsuo Ushiki and Osamu Hoshi. Atomic force microscopy for
imaging human metaphase chromosomes. Chromosome Research, 16
(3):383–396, 2008.
AM Van Oijen, J Köhler, J Schmidt, M Müller, and GJ Brakenhoff. 3-dimensional super-resolution by spectrally selective imaging.
Chemical Physics Letters, 292(1):183–187, 1998.
J Craig Venter, Mark D Adams, Eugene W Myers, Peter W Li,
Richard J Mural, Granger G Sutton, Hamilton O Smith, Mark Yandell, Cheryl A Evans, Robert A Holt, et al. The sequence of the
human genome. science, 291(5507):1304–1351, 2001.
Z. Wang, C. Zang, J.A. Rosenfeld, D.E. Schones, A. Barski, S. Cuddapah, K. Cui, T.Y. Roh, W. Peng, M.Q. Zhang, et al. Combinatorial
patterns of histone acetylations and methylations in the human
genome. Nature genetics, 40(7):897–903, 2008.
James Watson. The double helix. Hachette UK, 2012.
James D Watson. Involvement of rna in the synthesis of proteins.
Science, 140(3562):17–26, 1963.
James D Watson, Francis HC Crick, et al. Molecular structure of
nucleic acids. Nature, 171(4356):737–738, 1953.
Yanina Weiland, Paul Lemmer, and Christoph Cremer. Combining fish with localisation microscopy: Super-resolution imaging of
nuclear genome nanostructures. Chromosome research, 19(1):5–23,
2011.
August Weismann. Das Keimplasma; eine Theorie der Vererbung. Fischer,
1892.
Donna R Whelan and Toby DM Bell. Image artifacts in single
molecule localization microscopy: why optimization of sample
preparation protocols matters. Scientific reports, 5, 2015.
Andrew J Wood, Te-Wen Lo, Bryan Zeitler, Catherine S Pickle, Edward J Ralston, Andrew H Lee, Rainier Amora, Jeffrey C Miller, Elo
Leung, Xiangdong Meng, et al. Targeted genome editing across
species using zfns and talens. Science, 333(6040):307–307, 2011.
162
the periodic and dynamic structure of chromatin
CL Woodcock, L-LY Frado, and JB Rattner. The higher-order structure of chromatin: evidence for a helical ribbon arrangement. The
Journal of cell biology, 99(1):42–52, 1984.
Chung-Yin Wu, Ya-Ping Tsai, Min-Zu Wu, Shu-Chun Teng, and
Kou-Juey Wu. Epigenetic reprogramming and post-transcriptional
regulation during the epithelial–mesenchymal transition. Trends in
Genetics, 28(9):454–463, 2012.
X Sunney Xie, Robert C Dunn, et al. Probing single molecule dynamics. Science, pages 361–361, 1994.
Ke Xu, Guisheng Zhong, and Xiaowei Zhuang. Actin, spectrin, and
associated proteins form a periodic cytoskeletal structure in axons.
Science, 339(6118):452–456, 2013.
Patrick JM Zessin, Kieran Finan, and Mike Heilemann. Superresolution fluorescence imaging of chromosomal dna. Journal of
structural biology, 177(2):344–348, 2012.
Igor Fedorovich Zhimulëv. Morphology and structure of polytene chromosomes. Academic Press, 1996.
Roman Zinner, Kathrin Teller, Rogier Versteeg, Thomas Cremer, and
Marion Cremer. Biochemistry meets nuclear architecture: multicolor
immuno-fish for co-localization analysis of chromosome segments
and differentially expressed gene loci with various histone methylations. Advances in enzyme regulation, 47(1):223–241, 2007.
Warren R Zipfel, Rebecca M Williams, and Watt W Webb. Nonlinear magic: multiphoton microscopy in the biosciences. Nature
biotechnology, 21(11):1369–1377, 2003.
RM Zirbel, UR Mathieu, A Kurz, Thomas Cremer, and Peter Lichter.
Evidence for a nuclear compartment of transcription and splicing
located at chromosome domain boundaries. Chromosome Research, 1
(2):93–106, 1993.
Christian Zorn, Thomas Cremer, Christoph Cremer, and Jürgen
Zimmer. Laser uv microirradiation of interphase nuclei and posttreatment with caffeine. Human genetics, 35(1):83–89, 1976.
Christian Zorn, Christoph Cremer, Thomas Cremer, and Jürgen Zimmer. Unscheduled dna synthesis after partial uv irradiation of the
cell nucleus: distribution in interphase and metaphase. Experimental
cell research, 124(1):111–119, 1979.
Dominika Żurek-Biesiada, Aleksander T Szczurek, Kirti Prakash,
Giriram K Mohana, Hyun-Keun Lee, Jean-Yves Roignant, Udo Birk,
Jurek W Dobrucki, and Christoph Cremer. Localization microscopy
of dna in situ using vybrant® dyecycle violet fluorescent probe: A
new approach to study nuclear nanostructure at single molecule
resolution. Experimental cell research, 2015.
bibliography 163
Dominika Żurek-Biesiada, Aleksander T Szczurek, Kirti Prakash,
Gerrit Best, Giriram K Mohana, Hyun-Keun Lee, Jean-Yves Roignant,
Jurek W Dobrucki, Christoph Cremer, and Udo Birk. Quantitative
super-resolution localization microscopy of dna in situ using vybrant® dyecycle violet fluorescent probe. Data in Brief, 7:157–171,
2016.
“We have a habit in writing articles published in scientific journals to make the work
as finished as possible, to cover all the tracks, to not worry about the blind alleys or to
describe how you had the wrong idea first, and so on. So there isn’t any place to publish,
in a dignified manner, what you actually did in order to get to do the work.”
Richard Feynman
Publications
In peer reviewed journals
1. Single molecule localization microscopy of the distribution of
chromatin using Hoechst and DAPI fluorescent probes
• Authors: Szczurek*, Aleksander T , Prakash*, Kirti , Lee*,
Hyun-Keun , Żurek-Biesiada, Dominika J , Best, Gerrit , Hagmann, Martin , Dobrucki, Jurek W , Cremer, Christoph and Birk,
Udo
* first co-authors
• Description: In this report, we demonstrate that DNA minor
groove binding dyes, such as Hoechst and DAPI, can undergo
UV-induced photoconversion, to be effectively employed in single molecule localization microscopy (SMLM) with high optical
and structural resolution.
• Journal: Nucleus 2014 (cover page)
Figure P.1: A simulated chromosome territory model using Hoechst fluorescent
probe.
• Author contributions: K.P., C.C., and U.B. initiated the project.
A.S., H-K.L., K.P., and D.Z-B. designed the experiments. A.S.
prepared the samples and performed the single-molecule measurements. H-K.L., K.P., C.C., and U.B. developed and constructed the microscopy apparatus. K.P., G.B., M.H., and U.B.
contributed to the software used for data analysis. D.Z. performed the confocal experiments. K.P., A.S., H-K.L., G.B., and
M.H. performed the data analysis. C.C., J.D., and U.B. supervised the work. All authors contributed in writing of the
manuscript.
• Web: http://www.tandfonline.com/doi/pdf/10.4161/nucl.29564
2. Superresolution imaging reveals structurally distinct periodic
patterns of chromatin along pachytene chromosomes
• Authors: Prakash, Kirti , Fournier, David , Redl, Stefan , Best,
Gerrit , Borsos, Máté , Tiwari, Vijay K , Tachibana-Konwalski,
KiKuë , Ketting, René F , Parekh, Sapun H , Cremer, Christoph ,
Birk Udo J.
• Description: In this study, we found that chromatin is nonrandomly distributed along the length of the synaptonemal
Figure P.2: A model for the epigenetic landscape of meiotic chromosomes.
Chromatin (lateral extensions) is constrained by periodic clusters of histone
modifications along the synaptonemal
complex (SC, thick axial gray lines).
Active chromatin (H3K4me3) emanates
radially in loop like structures (red
dots) while the repressive chromatin
(H3K27me3) is confined to axial regions
(green balls) of the SC. Centromeric chromatin (H3K9me3) hints at spiralization
of DNA at one of the extremities of SC
(blue lines at the top).
166
the periodic and dynamic structure of chromatin
complex (SC) and displays differential condensed clusters. Chromatin structure is further divided into different functional compartments using histone modifications. Taking into account
the arrangement and composition of chromatin, as well as the
region-specific distribution of post-translational histone modifications, we discuss a model of the chromatin architecture along
the length of the SC. This study is a successful application of
our newly developed SMLM method applied to DNA dyes.
• Journal: Proceedings of the National Academy of Sciences
(PNAS) 2015
• Author contributions: K.P. designed research; K.P. and S.R.
performed research; K.P., S.R., G.B., M.B., V.T., K.T.-K., R.K., S.P.,
C.C., and U.B. contributed new reagents/analytic tools; K.P. and
D.F. analysed data; K.P. and D.F. wrote the paper with input
from all the other authors.
• Web: http://www.pnas.org/content/112/47/14635.short
3. A transient ischemic environment induces reversible compaction
of chromatin
• Authors: Kirmes, Ina , Szczurek, Aleksander , Prakash, Kirti ,
Charapitsa, Iryna , Heiser, Christina , Musheev, Michael , Schock,
Florian , Fornalczyk, Karolina , Ma, Dongyu , Birk, Udo , Cremer,
Christoph and Reid, George
• Description: Using SMLM, we evaluated the environmental
effects of ischemia on chromatin nanostructure. We found
short-term oxygen and nutrient deprivation (OND) of the cardiomyocyte cell-line HL-1 induces a dramatic and reversible
compaction. Chromatin adapts to a previously undescribed
sub-nuclear configuration comprising of discrete, DNA dense,
hollow, atoll-like structures, which upon removal of transient
ischemic-like conditions, reverses to the open structure in untreated cells.
• Journal: Genome Biology 2015
• Author contributions: C.C. and G.R. conceived of the study
and designed the experimental strategy. I.K., A.S., C.H., I.C.,
M.M., D.M. and K.F. performed experiments. I.K., A.S., K.P.,
F.S., M.M., U.B. and G.R. analyzed data. U.B. prepared the
movie. U.B., C.C. and G.R. supervised this project. G.R. wrote
the paper and coordinated the supplemental information. All
authors discussed the results and contributed to the manuscript.
All authors read and approved the final manuscript
• Web: http://www.genomebiology.com/2015/16/1/246
4. Localization microscopy of DNA in situ using Vybrant® DyeCycle Violet fluorescent probe: A new approach to study nuclear
nanostructure at single molecule resolution
Figure P.3: Deprivation of oxygen and
nutrients provokes an entirely novel and
previously undescribed chromatin architecture consisting of a condensed network of rods and swirls interspersed between large, chromatin-sparse nuclear
voids.
publications
167
• Authors: Żurek-Biesiada, Dominika , Szczurek, Aleksander
T , Prakash, Kirti , Mohana, Giriram K , Lee, Hyun-Keun ,
Roignant, Jean-Yves , Birk, Udo , Dobrucki, Jurek W and Cremer,
Christoph
• Description: We report here that the standard DNA dye Vybrant
Violet can be used for chromatin imaging using SMLM and
helps to describe the nanoscale structure of chromatin. This
technique enabled the localisation of a large number of DNAbound molecules, usually resulting in an excess of 106 signals
in a ∼ 500 nm optical section of a cell nucleus.
• Journal: Experimental Cell Research 2015
• Author contributions: D.Z-B. and A.S. planned the experiments.
D.Z-B. and A.S. performed the experiments, drafted and revised
the manuscript, D.Z-B., K.P. and A.S. performed image data
reconstruction. G.K.M. prepared polytene chromosome samples,
J.Y. R., U.B., J.D., C.C. supervised the work and contributed to
writing the manuscript.
Figure P.4: Mitotic chromosomes stained
with Vybrant® DyeCycle Violet fluorescent probe.
• Web: http://www.sciencedirect.com/science/article/pii/
S001448271530080X
5. Quantitative super-resolution localization microscopy of DNA
in situ using Vybrant® DyeCycle Violet fluorescent probe
• Authors: Żurek-Biesiada, Dominika , Szczurek, Aleksander T
, Prakash, Kirti , Gerrit Best, Mohana, Giriram K , Lee, HyunKeun , Roignant, Jean-Yves , Birk, Udo , Dobrucki, Jurek W and
Cremer, Christoph
• Description: In this manuscript, parameters that influence the
quality of SMLM reconstruction using Vybrant DyeCycle Violet,
for instance number of frames, wave length or composition of
buffer, are investigated, using quantifications and experimental
methods.
• Journal: Data in Brief 2016
• Web: http://dx.doi.org/10.1016/j.dib.2016.01.041
Figure P.5: 3D surface plot of DNA.
168
the periodic and dynamic structure of chromatin
In conferences
1. Identify and localise: Algorithms for single molecule localisation microscopy
• Authors: Kirti Prakash*, Gerrit Best*, Martin Hagmann, Udo
Birk, and Christoph Cremer.
* first co-authors
• Description: Here, we present a comparative analysis of a range
of available localisation algorithms regarding their complexity,
applicability and performance by testing them on both synthetic
and experimental data. Experimental data come from both
sparse and dense regions, with low and high background levels,
to determine which method is suited for a given dataset.
Elements of localisation microscopy
WIDEFIELD
STACK LEVEL
FRAME LEVEL
1. Generate background map
2. Calculate di erence image
✁
1. Find local maxima
2. Extract ROIs
ERROR CORRECTIONS
VISUALISATIONS
1. Remove multi-frame points
2. Correct for sample drift
3. Remove events which don't ful ll quality
criteria (e.g. std, error thresholds)
1. Gaussian bluring
2. Triangulation
3. Histogram
4. Other methods
✄
list of points
analysis on single molecules
analysis on image
ANALYSIS
ROI LEVEL
✂
1. Localize uorophore
2. Save the determined parameters
(position, pophoton number, std, errors, …)
1. Cluster analysis
2. Colocalisation
Figure P.6: Algorithm scheme for processing single molecule localisation microscopy data.
• Conference: International Microscopy Congress (IMC) 2015
• Author contributions: K.P. designed research, performed research and analysed the data. K.P. and G.B. wrote the code.
• Web: http://www.imc2014.com
2. Drift correction strategies for superresolution imaging modalities
• Authors: Martin Hagmann*, Kirti Prakash*, Rainer Kaufmann,
Udo Birk and Christoph Cremer.
* first co-authors
• Description: We present two drift correction strategies based
solely on acquired data without any fiducial markers. Using both approaches we successfully corrected localization microscopy data down to a final drift under 5 nm. We demonstrate
that with this procedure the resolution of the final reconstructions was substantially enhanced.
Figure P.7: HeLa cell nucleus stained
with Hoechst 33258 photoproduct before
and after drift correction.
• Conference: International Microscopy Congress (IMC) 2015
• Author contributions: K.P. and M.H. designed research, performed research and analysed the data
• Web: http://www.imc2014.com
3. Superresolution imaging of meiosis prophase I chromatin in
pachytene stage.
• Authors: Kirti Prakash, Gerrit Best, Mate Borsos, Stefan Redl,
Kikue Tachibana-Konwalski, Rene Ketting, Sapun Parekh, Udo
Birk, and Christoph Cremer
• Description: We combined single molecule localisation microscopy with next generation sequencing data and computer
simulations to map and analyse the distribution of chromatin
and several of its post-translational modifications along the lateral elements of the SC.
• Conference: Focus On Microscopy (FOM) 2015
Figure P.8:
Distribution of posttranslational histone modifications
around the synaptonemal complex (SC).
publications
169
• Author contributions: K.P. conceived the project and planned
the experiments. K.P. and S.R. performed the experiments. K.P.
performed image data reconstruction and analysed the data.
K.P., S.R., G.B., R.K., S.P., U.B. and C.C. analyzed the results.
• Web: http://www.focusonmicroscopy.org/2015/home.html
4. Lampbrush-like structures in mammalian meiotic chromosomes
• Authors: Kirti Prakash
• Description: Lampbrush chromosomes (LBC) are transcriptionally active chromosomes found in meiosis prophase I of most
animals, except mammals. In LBC, chromosomes adapt special
loop-like structures that emanate radially from the axis of the
chromosomes, most likely to facilitate transcription. These loops
have never been observed previously in mammals, somatic cells
or diploid. Here, it is shown for the first time that mammalian
chromosomes with clusters of transcriptionally active chromatin
show patterns similar to the amphibian LBC.
• Conference: Nuclear Organization and Function, CSHL 2016
•
Figure P.9: Distribution of H3K4me3
mark around the synaptonemal comWeb: https://meetings.cshl.edu/meetings.aspx?meet=NUCLEUS& plex (SC): H3K4me3 images revealed
lampbrush-like shapes in mammalian
year=16
chromosomes.
Acknowledgements
I thank Christoph Cremer for providing me with the opportunity to
enter a new area of research and participate in the establishment of
a new lab. I would also like to thank him for allowing me to work
on my ideas and set up my collaborations. I am further in debt to all
Cremer group members. I thank Udo Birk, Martin Hagmann, Gerrit
Best, Margund Bach, Hyun-Keun Lee, Florian Schock, Aleksander
Szczurek, Amine Gouram, Jan Neumann for the useful discussions.
I thank my friends and colleagues at IMB. Maria Hanulova, Sandra
Ritz and Katharina Boese were my unofficial lab mates and many
thanks for proofreading my thesis. I thank Wolf Gebhardt with whom
I had many interesting scientific discussion. During my initial days
at IMB, I had the opportunity to work with Heinz Eipel from whom I
learnt a lot about science. I would also like to thank George Reid and
Jean-Yves Roignant for their support. Finally, I thank Miguel Andrade
for the encouragement and many useful pieces of advice.
A lot of algorithms that made way into this thesis were co-written
with Martin Hagmann and Gerrit Best. Frederik Fleissner, Sandra
Ritz and Udo Birk helped in formulating the German version of the
thesis summary/abstract. David Fournier, Akshay Prakash, Shashi
Bhagat and Wioleta Dudka helped with most of the cartoon figures
that made their way into the thesis. I thank my collaborators for the
samples. In particular, I thank Wolf Gebhardt, Aleksander Szczurek,
Ina Schaefer, Sabine Puetz, Sapun Parekh, Dominika Zurek-Biesiada,
Paulina Rybak, Jurek Dobrucki, Giriram Kumar, Jean-Yves Roignant,
Ina Kirmes, Dongyu Ma, Mate Borsos, Stefan Redl, Kikue TachibanaKonwalski, Rene Ketting, George Reid and Wioleta Dudka.
I learnt a lot from some people whom I met during various courses,
conferences and lab visits. In particular, I would like to thank Thomas
Cremer, Marion Cremer, Brad Amos, Jonas Ries and Rainer Heintzmann. I further thank Sapun Parekh and Joseph G. Gall for their
support and understanding during the last few months of my thesis
writing. Sapun, also thanks for proofreading my thesis.
I am further in-debt to Wioleta Dudka for proofreading of my
thesis multiple times and for providing many valuable criticisms.
I have incorporated nearly all of her suggestions which helped in
removing a lot of ambiguities and redundancies.
This thesis won’t have been through and thorough without the
immense help of David Fournier, who not only critically read and
corrected parts of this dissertation but also constantly encouraged
172
the periodic and dynamic structure of chromatin
and pushed me to write this detailed monograph. David, thank you
for being there.
My family and friends have been a constant source of support.
In particular, I thank Wiola and David for their encouragement and
for handling my mood swings especially during last 3-4 months of
writing, rewriting and re-re-writing experience.
Finally, I hope this thesis provides enough new insights to encourage a young enthusiast to pursue this fascinating field.
Kirti Prakash