Journal of Crohn's and Colitis, 2019, 626–633
doi:10.1093/ecco-jcc/jjy205
Advance Access publication December 12, 2018
Original Article
Original Article
Redefining the Practical Utility of Blood
Transcriptome Biomarkers in Inflammatory
Bowel Diseases
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
Jerzy Ostrowski,a,b Michalina Dabrowska,a Izabella Lazowska,c
Agnieszka Paziewska,b Aneta Balabas,a Anna Kluska,a Maria Kulecka,b
Jakub Karczmarski,a Filip Ambrozkiewicz,a Magdalena Piatkowska,a
Krzysztof Goryca,a Natalia Zeber-Lubecka,b Jaroslaw Kierkus,d
Piotr Socha,d Michal Lodyga,e Maria Klopocka,f Barbara Iwanczak,g
Katarzyna Bak-Drabik,h Jaroslaw Walkowiak,i Piotr Radwan,j
Urszula Grzybowska-Chlebowczyk,k Bartosz Korczowski,l
Teresa Starzynska,m Michal Mikulaa
a
Department of Genetics, Maria Sklodowska-Curie Institute – Oncology Centre, Warsaw, Poland bDepartment of
Gastroenterology and Hepatology, Medical Center for Postgraduate Education, Warsaw, Poland cDepartment
of Pediatric Gastroenterology and Nutrition, Medical University of Warsaw, Warsaw, Poland dDepartment of
Gastroenterology, Hepatology and Feeding Disorders, Children’s Memorial Health Institute, Warsaw, Poland
e
Department of Internal Medicine and Gastroenterology with IBD Subdivision, Central Clinical Hospital of the
Ministry of the Interior, Warsaw, Poland fVascular Diseases and Internal Medicine, Nicolaus Copernicus University
in Torun, Collegium Medicum, Bydgoszcz, Poland gDepartment of Pediatrics, Gastroenterology and Nutrition,
Wroclaw Medical University, Wroclaw, Poland hDepartment of Pediatrics, School of Medicine with the Division
of Dentistry in Zabrze, Medical University of Silesia, Katowice, Poland iDepartment of Pediatric Gastroenterology
& Metabolic Diseases, Poznan University of Medical Sciences, Poznan, Poland jDepartment of Gastroenterology,
Medical University of Lublin, Lublin, Poland kDepartment of Pediatrics, School of Medicine in Katowice, Medical
University of Silesia, Katowice, Poland lMedical College, University of Rzeszow, Rzeszow, Poland mDepartment of
Gastroenterology, Pomeranian Medical University, Szczecin, Poland
Corresponding author: Jerzy Ostrowski, MD, PhD; Cancer Center-Institute, Roentgena 5, 02-781 Warsaw, Poland. Tel.: +48
225462575; e-mail: jostrow@warman.com.pl
Abstract
Background and Aims: The study investigates the practical utility of whole-blood gene expression
profiling to diagnose inflammatory bowel diseases [IBDs].
Methods: The discovery cohorts included 102 and 51 paediatric IBD patients and controls, and 95
and 46 adult IBD patients and controls, respectively. The replication cohorts included 447 and 76
paediatric IBD patients and controls, and 271 and 108 adult IBD patients and controls, respectively.
In the discovery phase, RNA samples extracted from whole peripheral blood were analysed using
RNA-Seq, and the predictive values of selected biomarkers were validated using quantitative
polymerase chain reaction [qPCR].
Results: In all, 15 differentially expressed transcripts [adjusted p ≤0.05] were selected from the
discovery sequencing datasets. The receiver operating characteristic curves and area under the
curve [ROC-AUC] in replication analyses showed high discriminative power [AUC range, 0.91–0.98]
for 11 mRNAs in paediatric patients with active IBD. By contrast, the AUC-ROC values ranged from
© European Crohn’s and Colitis Organisation (ECCO) 2018.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/
licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For
commercial re-use, please contact journals.permissions@oup.com
626
627
Blood Transcriptome Biomarkers in IBD
0.63 to 0.75 in comparison among inactive paediatric IBDs and active/inactive adult IBDs, indicating
a lack of discriminative power. The best multi-mRNA diagnostic classifier showed moderate
discriminative power [AUC = 0.81] for paediatric inactive IBD, but was not able to discriminate
active or inactive adult IBD patients from controls. The AUC-ROC values did not confirm an ability
of the mRNAs abundances to discriminate between active ulcerative colitis and active Crohn’s
disease in paediatric or adult populations.
Conclusions: This study identifies and validates blood transcriptional biomarkers that could be
used in clinical settings as diagnostic predictors of IBD clinical activity in paediatric, but not adult,
IBD patients.
Key Words: RNA-Seq; whole-blood gene expression, biomarker; inflammatory bowel disease
2.2. Subjects
Peripheral blood cells share >80% of their transcriptome with
that of other tissues, and profiling of gene expression in these
cells is employed in descriptive and comparative analyses of
autoimmune and inflammatory diseases.1–3 Crohn’s disease [CD]
and ulcerative colitis [UC] are chronic disorders that result from
altered activation of intestinal immunopathological processes
in response to the intestinal microbiota.4 Whereas the onset of
both of these inflammatory bowel diseases [IBDs] can occur from
early childhood to beyond the sixth decade of life, differences
in the polygenic architecture of paediatric- and adult-onset IBD5
may indicate a relationship between genetically attributable risk
and pathogenic variances in children and adults. Compared with
adult-onset IBDs, paediatric IBDs are typically characterized by a
more extensive disease course, a change in disease location over
time, and a more frequent family history of IBD. Patients diagnosed between the ages of 20 and 30 years have a relatively less
variable phenotype, and those diagnosed after the age of 60 years
often have a mild disease severity.5
Previous studies revealed that gene expression profiles
obtained from whole-blood cells may differentiate CD and UC
from non-IBD conditions, and active from inactive CD.6 However,
the practical utilities of measuring whole-blood expression levels in IBD patients have not been resolved. To date, only small
populations of IBD patients have been examined,6–10 the results
from discovery phases of the studies have not been replicated,
and candidate biomarkers used in the studies have been selected
from microarray-based datasets. In the current study, candidate
mRNA biomarkers of IBD were screened using RNA sequencing,
and the replication study was performed using polymerase chain
reaction [PCR]-based testing. Both phases of the study used relatively large cohorts of paediatric and adult patients with defined
clinical activities of IBD.
A total of 915 patients with IBD and 280 healthy control individuals
were recruited at the gastroenterology departments of various Polish
hospitals. Criteria to participate in the study included a confirmed
CD or UC diagnosis. There were no exclusion criteria except indeterminate colitis or concomitant infections. Overall, 488 patients [303
children aged 17 years and 185 adults] and 427 patients [245 children and 182 adults] were diagnosed with CD and UC, respectively.
Clinical characteristics and biochemical parameters available from
the enrolled patients at the time of blood collection for gene expression profiling are summarized in Supplementary Table S1, available
as Supplementary data at ECCO-JCC online. IBD was diagnosed
by experienced gastroenterologists during a standard diagnostic
work-up, using the Porto criteria modified in accordance with the
recommendations of the European Crohn’s and Colitis Organisation
[ECCO] in children, and according to ECCO guidelines in adults.
Patients were recruited during a course of hospital treatment or during a scheduled visit at outpatient departments. The Crohn’s disease
activity index [CDAI], the ulcerative colitis activity index [UCAI],
and their paediatric versions [PCDAI/PUCAI] were determined to
estimate disease severity.11–13 The patients were then assigned to two
subgroups: patients in remission or with mild IBD were considered
to have inactive disease, and those with moderate or severe IBD were
considered to have active disease. Before inclusion most patients
were given mesalazine, but for majority of them the blood samples
were collected before additional medications [immunosupressants,
glucocorticoids, biological therapy] were ordered. The control group
of children consisted of surgical, orthopaedic, or ophthalmological
patients who did not suffer from inflammation or intestinal diseases.
The control group of adults were healthy individuals who were
mainly recruited at cancer screening programmes.
2. Methods
2.1. Ethics statement
All procedures involving human participants were performed
in accordance with the ethical standards of the institutional and/
or national research committees, and with the 1964 Helsinki
Declaration and its later amendments, or comparable ethical standards. The study was approved by the Ethics Committee of the
Medical Center for Postgraduate Education, Warsaw, Poland [decision 69/PW/2011 granted on 16 May 2011]. Informed consent was
obtained from all participants in the study.
2.3. RNA extraction
Peripheral blood was collected using the Tempus RNA Isolation Kit
[Thermo Fisher Scientific], and total RNA was isolated according
to the manufacturer’s instructions. RNA quality and quantity were
analysed using a NanoDrop spectrophotometer, and samples with
A260/A280 ratios of 1.8–2.1 were assessed further using an Agilent
2100 Bioanalyzer.
2.4. RNA sequencing, reads processing, and
statistical analysis
Preparation and sequencing of the RNA-Seq libraries was performed
as described previously.14 Signal processing and base calling were
conducted with Torrent Suite version 5.0.4. Reads were mapped to
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
1. Introduction
628
J. Ostrowski et al.
the hg19 AmpliSeq Transcriptome version 1 genome. Read counts
per gene were obtained with HTSeq-count version 0.615 using
default parameters. Normalization and differential gene expression
estimations were performed using DESeq2, using default parameters
and options.16 A gene was considered differentially expressed when
the adjusted p-value was less than 0.05.
2.5. Quantitative RT-PCR
2.6. Data accessibility statement
The RNA sequencing datasets generated during the study were deposited in European Nucleotide Archive under the PRJEB28822 accession number.
3. Results
3.1. The discovery step
To screen for candidate biomarkers, we used the Ion AmpliSeq
Transcriptome Human Gene Expression Kit to sequence 293 RNA
samples extracted from whole peripheral blood. Of these, 51 and 46
samples were from paediatric and adult patients with CD, 51 and 49
samples were from paediatric and adult patients with UC, and 50
and 46 samples were from control children and adults, respectively.
In total, 31 patients in each paediatric IBD subgroup and seven patients in each adult IBD subgroup had active disease [with a score
above 30] at blood collection. Table 1 shows a summary of the main
epidemiological variables for the discovery cohorts.
On average, the RNA-Seq analyses generated 11 827 252 reads
per sample, of which 88% were on target and mapped to the reference genome. The RNA-Seq data were combined according to
patient age, diagnosis, and clinical activity [remission or mid-active
versus active disease], and the pair-wise comparisons identified 148
and 111 differentially expressed genes [DEGs] (false discovery rate
[FDR] ≤0.05; FC >2) between paediatric IBD patients and controls,
and between adult IBD patients and controls, respectively [Table 2
3.2. The replication study
Next, we assessed the diagnostic potential of the 15 genes described
above using newly recruited IBD patients and controls. The replication cohorts included 391 CD patients [252 children and 139
adults], 327 UC patients [195 children and 132 adults], and 184
controls [76 children and 108 adults]. Of these, 53 and 25 of the
paediatric patients and 34 and 27 of the adult patients had active
CD and active UC, respectively, at blood collection. Table 5 shows
a summary of the main epidemiological variables for the replication
cohorts.
The transcript levels were determined by qRT-PCR, and the FCs
between the IBD subgroups and control groups were compared. In
the paediatric population, the levels of all 15 mRNAs were significantly different [adjusted p ≤0.05] between the control group and
the active UC or active CD subgroups. Furthermore, the expression
Table 1. Summary of the main epidemiological variables for the discovery cohorts.
Demographics
Paediatric CD
Paediatric UC
Paediatric controls
Adult CD
Adult UC
Adult controls
Medication
Sex; [female/male]
Age;
range [median, years]
5-ASAs
Immunosupressants
Glucocorticoids
Biological therapy
30/21
32/19
28/22
23/23
28/21
26/20
2–17 [13]
1–17 [15]
1–17 [8]
19–69 [34]
21–66 [36]
38–62 [46]
90.2%
96.1%
0
86.9%
97.9%
0
45.1%
23.5%
0
60.9%
22.4%
0
39.2%
0
0
26.1%
26.5%
0
11.8%
29.4%
0
32.6%
12.2%
0
ASA, aminosalicylic acid; UC, ulcerative colitis; CD, Crohn’s disease.
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
Nine genes that were statistically significantly differentially expressed
and had absolute fold changes [FCs] greater than 2 [as estimated by
DESeq2] between controls and patients with the active form of disease [regardless of age] were verified by quantitative RT-PCR [qRTPCR]. Six additional genes were also chosen for verification based on
the adult cohort data only. The qRT-PCR analyses were performed
as described previously,17 using TaqMan chemistry. The geometrical
mean expression level of the RPLP0 and UBC mRNAs was used as a
normalization factor. Gene expression levels were calculated using the
∆∆Ct method, and the results were analysed using the Mann-Whitney
U-test; p ≤0.05 was considered statistically significant.
and Supplementary Table S2, available as Supplementary data at
ECCO-JCC online]. None of the mRNAs was able to differentiate
between CD and UC in paediatric patients, and only 36 mRNAs
[with FCs ranging from 0.7 to 1.3] were able to differentiate between
these IBDs in adult patients.
To search for biomarkers with diagnostic utility, we selected
15 mRNAs that displayed the highest levels of differential expression between active IBD patients and healthy controls; all but one
[KLRF1] were upregulated in IBD. This list included nine mRNAs
[S100A12, OPLAH, ATP9A, ANOS1, FCGR1A, ITGB4, UTS2R,
MMP9, and COX6B2] that showed at least a 2-fold difference between the active IBD and control groups and adjusted p-values less
than 0.05 in both the paediatric and adult populations, and six
mRNAs [ANXA3, CACNA1E, GALNT14, IL18R1, KLRF1, and
PFKFB3] that showed the highest FCs between the active IBD and
control groups in the adult population. The expression levels of these
mRNAs were confirmed by qRT-PCR using the same RNA samples
as those used in the RNA-Seq analyses. The expression levels of all
15 genes were statistically significantly different [adjusted p ≤0.05]
between paediatric patients with active IBD and the corresponding
controls [Tables 3 and 4], and similar results were obtained for the
active UC and CD paediatric subgroups [data not shown]. By contrast, none of these genes was able to differentiate between control
children and paediatric patients with inactive IBD. In the adult population, the expression levels of all but one [ATP9A] from the set of
nine genes, and three [GALNT14, KLRF1, and PFKFB3] from the
set of six genes, were significantly different between the active IBD
and healthy control groups. In addition, only one gene [KLRF1] was
differentially expressed between healthy controls and adults with
inactive IBD.
629
Blood Transcriptome Biomarkers in IBD
Table 2. Summary of numbers of differentially expressed genes in
different comparisons.
Comparison
DGE
FC >2.0
% DEG
Paediatric active IBD vs control children
Paediatric inactive IBD vs control children
CD children vs UC children
Adult active IBD vs control adults
Adult inactive IBD vs control adults
CD adults vs UC adults
5249
2637
0
1627
4203
35
148
11
0
101
0
0
30.35
18.19
0
11.08
30.28
0.24
DGE, number of differentially expressed genes; FC, fold change; %DEG,
percentage of DEGs among tested genes; IBD, inflammatory bowel disease;
CD, Crohn’s disease; UC, ulcerative colitis.
that differentiated control children from those with inactive UC
[ITGB4, MMP9, UTS2R, KLRF1, and GALANT14] and inactive
CD [ANOS1, CACNA1E, MMP9, and UTS2R] had an AUC value
of 0.81, and therefore displayed only moderate discriminative
power. Furthermore, the AUC-ROC values did not confirm the
ability of gene expression profiling to discriminate between active
UC and active CD in paediatric patients, even though statistically
significant differences in mRNA levels were observed between the
two groups [Table 6].
Unexpectedly, none of the genes that were significantly differentially expressed between the adult IBD and healthy adults groups
in pair-wise comparisons showed sufficient discriminatory power
in the ROC-AUC analyses, when either single mRNAs [Table 7]
or combinations of mRNAs were analysed [not shown]. In adults,
the best multi-mRNA diagnostic classifiers for active UC [FCR1B,
KLRF1, CACNA1E, ANXA3, and GALANT14] and active CD
[KLRF1, CACNA1E, ANXA3, and GALANT14] had AUC values
of 0.78 and 0.77, respectively. As seen for the paediatric population,
the AUC-ROC values did not confirm the ability of gene expression
profiling to discriminate between active UC and active CD in adults
[Table 7].
4. Discussion
Endoscopic/radiological examination followed by pathological
evaluation, supported by laboratory testing of C-reactive protein
[CRP] levels, the erythrocyte sedimentation rate, and faecal calprotectin [among others], is the ‘gold standard’ for diagnosis of IBD. The
severity of IBD can vary from mild to severe, and its diagnosis can be
delayed, especially for early-onset IBD when the clinical symptoms
are ambiguous, non-specific, or indolent. Biomarkers have the potential to be useful for IBD diagnosis, identification of disease subtypes,
and prognosis to therapeutic adjustment, and might therefore complement the use of clinical parameters. Candidate biomarkers are
typically identified by high-throughput methods using either microarray or next-generation sequencing [NGS] technologies, and are
subsequently validated by standard molecular methods. Microarrays
represent a non-linear model of the relationship between readouts
and actual amounts of mRNAs in the samples analysed. Due to the
multiple steps involved in microarray hybridization and imaging,
the results of such analyses are strictly dependent on the statistics
included in the data processing. NGS can provide linear relationships between relative numbers of sequenced transcript fragments
and true gene expression levels18; however, this method is hampered
by its bias for identifying particular subsets of transcripts and the
unknown minimum size of the sequenced transcriptome required for
sufficient coverage of low-abundance mRNAs. Although neither of
these technologies is optimal for measuring gene expression at the
global level, they are both commonly employed for discovery of candidate biomarkers.19
Table 3. Differences in mRNAs levels from the nine gene sets selected by RNA sequencing and validated by qRT-PCR in the training cohorts.
Comparisons
ITGB4
MMP9
COX6B2
UTS2R
FCGR1B
OPLA
S100A12
ATP9A
ANOS1
Paediatric active IBD vs control children
Paediatric inactive IBD vs control children
Adult active IBD vs control adults
Adult inactive IBD vs control adults
4.80E-06
8.80E-01
1.50E-03
3.40E-01
1.40E-04
6.30E-01
3.20E-03
4.30E-01
3.00E-04
7.30E-01
6.00E-04
4.50E-01
4.10E-06
7.30E-01
6.00E-04
5.60E-03
4.10E-06
7.30E-01
1.80E-05
8.20E-01
4.70E-04
7.40E-01
2.30E-04
4.30E-01
3.30E-06
5.10E-01
1.70E-03
4.30E-01
5.90E-04
6.30E-01
3.50E-01
5.90E-01
3.30E-06
5.00E-01
5.50E-05
7.90E-01
In bold fonts are the statistically significant results with adjusted p-value <0.05.
IBD, inflammatory bowel disease.
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
levels of 13 of the 15 mRNAs examined were significantly different between the paediatric control group and the paediatric inactive
UC or inactive CD subgroup [Table 6]. In the adult population, the
expression level of UTS2R differed significantly between the control group and the group comprising patients with either active or
inactive UC, and the expression levels of 10 and nine mRNAs differed significantly between the healthy adults and those with active
or inactive CD, respectively [Table 7]. Gene expression levels did not
correlate with patients’ age, with a Kendall’s tau coefficient lower
than 0.1.
Next, the diagnostic potentials of the 15 mRNAs described
above were assessed using receiver operating characteristic [ROC]
curves and area under the curve [AUC] analyses. ROC curves enable assessment of the relationship between the sensitivity and specificity of a biomarker over various cut-offs.10 AUC-ROC values
greater than 0.8 were assumed to represent moderate [good] discriminative power, and those greater than 0.9 were assumed to represent high [excellent] discriminative power between the analysed
groups. In the paediatric population, nine genes [ANOS1, ANXA3,
CACNA1E, GALNT14, ITGB4, MMP9, OPLA, PFKFB3, and
S100A12] showed high discriminative power [AUC range, 0.91–
0.98] between the control group and each active IBD subgroup
[active UC or active CD] [Figure 1]. In addition, ATP9A showed
high discriminative power between the control and active UC
groups, and FCGR1B showed high discriminative power between
the control and active CD groups. By contrast, whereas several
genes were significantly differentially expressed between paediatric
patients with inactive IBDs and control children, the AUC-ROC
values of these genes ranged from 0.63 to 0.75, and therefore did
not confirm their discriminative properties. To determine whether
the diagnostic ability of an mRNA signature is higher than those
of the single mRNAs, we calculated AUC-ROC values for combinations of mRNAs with the best discriminatory powers. Using linear
models of normalized expression values and a stepwise inclusion
approach, we found that the best multi-mRNA diagnostic classifier
630
J. Ostrowski et al.
Table 4. Differences in mRNAs levels from the six gene sets selected by RNA sequencing and validated by qRT-PCR in the training cohorts.
Comparison
ANXA3
CACNA1E
GALNT14
IL18R1
KLRF1
PFKFB3
Paediatric active IBD vs control children
Paediatric inactive IBD vs control children
Adult active IBD vs control adults
Adult inactive IBD vs control adults
5.51E-05
0.996873
0.076794
0.529562
1.46E-05
0.996873
0.195521
0.973081
6.56E-06
0.996873
0.046505
0.529562
0.003566
0.996873
0.878352
0.529562
0.020675
0.996873
9.40E-06
1.49E-05
0.000209
0.996873
0.046505
0.529562
In bold fonts are the statistically significant results with adjusted p-value <0.05.
IBD, inflammatory bowel disease.
Table 5. Summary of the main epidemiological variables for the replication cohorts.
Demographics
Sex; [female/male]
Age;
range [median, years]
5-ASAs
Immunosuppressants
Glucocorticoids
Biological therapy
102/150
107/88
32/44
86/53
78/54
64/44
2–17 [15]
1–17 [15]
1–17 [8]
18–70 [29]
18–73 [35]
43–64 [58]
90.1%
97.4%
0
87.8%
95.4%
0
45.2%
22.6%
0
33.1%
23.5%
0
39.3%
0
0
20.9%
24.2%
0
11.9%
30.3%
0
8.6%
18.2%
0
ASA, aminosalicylic acid; UC, ulcerative colitis; CD, Crohn’s disease.
Blood cells suspended in a fluid matrix connect the entire biological system of an organism and constitute the first line of the immune
defence system; hence, blood samples are often used as a surrogate
for traditional tissue specimens in clinical diagnoses.3 Changes in the
whole-blood transcriptome are associated with injuries and a wide
range of diseases, including autoimmune, inflammatory, infectious,
psychiatric, cardiovascular, neurological, and neoplastic disorders,
as well as various environmental stresses.20
Microarray studies have led to the development of a noninvasive
test for the diagnosis of IBD which is based on expression profiles
of peripheral leukocytes.21 In 2008, Alsobrook et al. reported that
the peripheral blood expression levels of a classifier set of six genes
[BLCAP, UBE2G1, GPX1, RAP1A, CALM3, and NONO] are able
to distinguish IBD patients from healthy controls with accuracy, sensitivity, and specificity rates of 84%, 89%, and 75%, respectively.22 To
date, a few other studies have confirmed that expression profiling of
blood samples provides a noninvasive method of distinguishing clinically active from inactive IBD,6 endoscopically active UC patients from
patients in clinical remission,8 active CD from CD in remission,7 and
paediatric IBD patients in clinical remission from healthy controls.23
In sum, IBD-related blood gene expression in the previous studies has
only been examined in small populations, and analyses of the DEGs
generated results that could not unambiguously differentiate healthy
from diseased samples. The only exception was the study of Burakoff
et al.7 describing four separate gene panels which discriminated either
CD or UC patients with either mild or moderate-to-severe severity
from other categories with an ROC–AUC ranging between 0.89 and
0.99. Our RNA-Seq-based study performed on 293 RNA samples
determined a detectable expression of 19 out of 22 of those genes,
but only one, PGM1, showed a moderate [AUC = 0.81] discriminative
power between the control and active CD groups.
To re-examine the practical utility of whole-blood gene expression profiling in IBD diagnosis, we identified 15 potential biomarker
transcripts [S100A12, OPLAH, ATP9A, ANOS1, FCGR1A, ITGB4,
UTS2R, MMP9, COX6B2, ANXA3, CACNA1E, GALNT14,
IL18R1, KLRF1, and PFKFB3] by sequencing RNA samples isolated from a total of 293 IBD patients and control individuals. These
mRNAs were subsequently validated by qRT-PCR analyses of RNA
samples from 902 newly recruited IBD patients and controls. Our
findings confirmed that measurement of blood mRNA levels has
a diagnostic potential for paediatric patients with active IBD, but
not for paediatric patients with inactive IBD, or adults with active
or inactive disease. Furthermore, although we found that analysing combinations of mRNAs has a moderate power to discriminate
between paediatric inactive IBD patients and control children, we
could not confirm that the accuracy of such multi-classifier mRNA
signatures is higher than those of single mRNAs in adult patients.
Furthermore, the blood mRNA signatures were unable to distinguish UC from CD in the adult or paediatric populations. Among 15
transcripts analysed, several encode proteins functionally connected
with inflammatory processes and immune response. The S100A12,
also known as calgranulin C, has recently emerged as marker of
inflammation to predict IBD better than calprotectin from stool
samples.24–26 The FCGR1A encodes the neutrophils’ expressed CD64
receptor, and its abundance is positively correlated with immune
inflammation syndromes including IBD.27,28 MMP-9 is one of major
metalloproteases activated in intestinal tissues of patients with active
IBD,29 and serum MMP-9 levels could aid differentiation of active
UC from active CD.30 IL18R1 encodes IL18 receptor, essential for
IL18-mediated signal transduction. IL18 is instrumental for controlling the outgrowth of bowel microbiota; however during instances
of severe inflammation, exacerbated IL-18 expression causes a loss
of goblet cells, ultimately depleting mucosal barrier function leading
to colitis in a mouse model.31 The KLRF1 encodes C-type lectin-like
activating receptor NKp80, which is expressed on all NK cells in
the peripheral blood. It recognizes activation-induced C-type lectin
[AICL] that is upregulated on activated monocytes and NK cells,
and NKp80–AICL interaction promotes NK cell-mediated control
of monocytes.32 The KLRF1 was the only transcript with expression
downregulated in IBD when compared with healthy controls.
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
Paediatric CD
Paediatric UC
Paediatric controls
Adult CD
Adult UC
Adult controls
Medication
631
Blood Transcriptome Biomarkers in IBD
Table 6. Comparisons of mRNA levels analysed by qRT-PCR in the paediatric replication cohorts.
mRNA
Crohn’s disease vs control children
Active
Active
Inactive
Active ulcerative
colitis vs active
Crohn’s disease
Inactive
AUC
Adj. p-value
AUC
Adj. p-value
AUC
Adj. p-value
AUC
Adj. p-value
AUC
Adj. p-value
0.96
0.98
0.92
0.96
0.88
0.90
0.98
0.83
0.94
0.83
0.97
0.96
0.98
0.97
0.81
1.9E-11
1.13E-11
6.12E-10
1.9E-11
3.93E-08
9.3E-09
1.13E-11
1.86E-06
2.62E-10
1.05E-06
1.13E-11
2.03E-11
1.13E-11
1.69E-11
5.74E-06
0.69
0.66
0.55
0.68
0.63
0.63
0.66
0.57
0.71
0.69
0.71
0.69
0.63
0.67
0.70
1.6E-05
0.000316
0.223911
3.71E-05
0.003564
0.003013
0.000222
0.105793
5.14E-06
1.9E-05
5.14E-06
1.6E-05
0.003564
9.36E-05
1.26E-05
0.97
0.95
0.82
0.96
0.87
0.95
0.95
0.83
0.94
0.71
0.97
0.96
0.95
0.96
0.63
2.47E-18
2.21E-17
1.25E-09
3.95E-18
1.92E-12
2.2E-17
1.69E-17
4.49E-10
8.15E-17
4.77E-05
2.47E-18
5.68E-18
2.2E-17
5.56E-18
0.013751
0.75
0.69
0.58
0.74
0.62
0.72
0.68
0.50
0.69
0.68
0.73
0.68
0.66
0.70
0.70
3.49E-08
9.99E-06
0.074931
1.07E-07
0.007121
4.97E-07
2.94E-05
0.999028
1.23E-05
3.01E-05
1.61E-07
2.71E-05
0.000149
3.19E-06
8.87E-06
0.62
0.55
0.51
0.61
0.53
0.66
0.55
0.57
0.52
0.52
0.56
0.54
0.55
0.56
0.55
0.001611
0.198678
0.853082
0.00308
0.3854
6.75E-06
0.198678
0.088585
0.50386
0.558309
0.198678
0.313489
0.198678
0.171036
0.198678
In bold fonts are the statistically significant results with adjusted p-value <0.05 and AUC values >0.8.
AUC, area under the curve; adj, adjusted.
Table 7. Comparisons of mRNA levels analysed by qRT-PCR in the paediatric replication cohorts.
mRNA
ANOS1
ANXA3
ATP9A
CACNA1E
COX6B2
FCGR1B
GALNT14
IL18R1
ITGB4
KLRF1
MMP9
OPLA
PFKFB3
S100A12
UTS2R
Ulcerative colitis vs control adults
Crohn’s disease vs control adults
Active
Active
Inactive
Active ulcerative
colitis vs
active Crohn’s disease
Inactive
AUC
Adj. p-value
AUC
Adj. p-value
AUC
Adj. p-value
AUC
Adj. p-value
AUC
Adj. p-value
0.58
0.54
0.58
0.51
0.51
0.51
0.53
0.57
0.50
0.62
0.50
0.55
0.57
0.54
0.69
0.11068
0.51191
0.11068
0.89914
0.89914
0.90343
0.58935
0.15341
0.94155
0.00507
0.94155
0.27203
0.15341
0.51191
1.7E-06
0.56
0.49
0.48
0.54
0.51
0.56
0.52
0.53
0.50
0.60
0.48
0.55
0.62
0.56
0.76
0.82618
0.93717
0.93717
0.93717
0.93717
0.82618
0.93717
0.93717
1.00000
0.58882
0.93717
0.82618
0.46185
0.82618
0.00079
0.71
0.69
0.54
0.71
0.64
0.71
0.64
0.53
0.65
0.71
0.60
0.55
0.59
0.70
0.66
0.001198
0.003
0.571025
0.001198
0.024824
0.001198
0.024729
0.655729
0.017966
0.001198
0.104351
0.399874
0.140337
0.001519
0.015584
0.64
0.60
0.62
0.61
0.55
0.62
0.60
0.54
0.50
0.74
0.57
0.50
0.54
0.61
0.65
0.00072
0.00884
0.00373
0.00373
0.21516
0.00198
0.01125
0.29748
0.99496
0.00000
0.06547
0.99496
0.32333
0.00458
0.00030
0.55
0.57
0.52
0.62
0.55
0.64
0.56
0.51
0.52
0.63
0.57
0.56
0.55
0.58
0.56
0.0880
0.0287
0.6237
0.0001
0.0997
0.0001
0.0577
0.7161
0.5403
0.0001
0.0287
0.0577
0.0880
0.0206
0.0709
In bold fonts are the statistically significant results with adjusted p-value <0.05.
AUC, area under the curve; adj, adjusted.
IBDs are heterogeneous chronic disorders characterized by a succession of variable severity of relapses and remissions, which in turn
relates to variable therapeutic decisions.33 In this study, the severity
of IBD was assessed using the CDAI or UCAI and their paediatric
equivalents [PCDAI/PUCAI]. However, although these indexing systems are commonly used due to the simplicity of collecting data for
calculating the severity scores, they predominantly rely on symptomatology without consideration of other aspects of disease severity,
such as inflammatory activity and structural damage. Patients with
severe symptoms may have mild inflammation, and disease severity does not always reflect the true disease activity. Furthermore,
IBD may co-exist with other non-inflammatory chronic intestinal conditions, such as irritable bowel syndrome.33 Consequently,
our method of assigning the severity of IBD as mild, moderate, or
severe may reflect current working practice rather than a clinically
proven classification. Accordingly, the cut-off gene expression levels for distinguishing active from inactive IBD identified here should
be considered with some caution. Similar problems apply to commonly used markers of intestinal inflammation, such as C-reactive
protein [CRP] and faecal calprotectin.33 For example, up to 50%
of patients with clinically active UC, and some patients with active
CD, display normal CRP levels.6 In addition measurement of faecal
calprotectin is relevant in follow-up examinations of IBD patients,
but more studies are needed to establish accurate reference values
for different ages, disease subtypes, disease localization/extension,
and responses to therapy.25,34,35 The results presented here confirm
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
ANOS1
ANXA3
ATP9A
CACNA1E
COX6B2
FCGR1B
GALNT14
IL18R1
ITGB4
KLRF1
MMP9
OPLAH
PFKFB3
S100A12
UTS2R
Ulcerative colitis vs control children
632
J. Ostrowski et al.
Author Contributions
Ulcerative colitis
1.0
Sensitivity
0.8
PFKFB3
ANXA3
GALANT14
MMP9
S100A12
CACNA1E
ANOS
OPLAH
ITGB4
ATP9A
0.6
0.4
0.2
0
1.0
0.5
0
Specificity
Sensitivity
0.8
MMP9
ANOS
CACNA1E
S100A12
OPLAH
GALANT14
FCR1B
PFKFB3
ANXA3
ITGB4
0.6
0.4
0.2
0
1.0
0.5
0
Specificity
Figure 1. Diagnostic potential of candidate mRNA biomarkers. Receiver
operating characteristic [ROC] curves were generated and area under
the curve [AUC] values were calculated using qRT-PCR expression values
obtained for the paediatric active ulcerative colitis [UC] and active Crohn’s
disease [CD] subgroups.
that whole-blood mRNA expression profiling may be an effective
tool to monitor IBD severity in paediatric, but not adult, patients.
Although this study included large groups of patients, the amounts
of RNA obtained from the blood samples were only sufficient for
confirmatory and replication studies on a limited number of selected
transcripts. Therefore, we cannot exclude the possibility that, among
the thousands of DEGs identified via high-throughput sequencing,
we may have missed other potentially useful biomarkers due to
methodological limitations.
In summary, this study identified and validated the use of wholeblood transcriptional biomarkers as predictors of IBD clinical activity in paediatric but not adult IBD patients. These promising results
were derived from analyses of a large number of patients with a
wide range of clinical activity, and were conducted simultaneously in
paediatric and adult cohorts.
Funding
This work was supported by the National Science Centre [2011/02/A/
NZ5/00339]. The funder had no role in in the study design; in the collection,
analysis, and interpretation of the data; in the writing of the report; or in the
decision to submit the paper for publication.
Conflict of Interest
The authors do not declare any conflict of interest.
Supplementary Data
Supplementary data are available at ECCO-JCC online.
References
1. Mesko B, Poliska S, Nagy L. Gene expression profiles in peripheral
blood for the diagnosis of autoimmune diseases. Trends Mol Med
2011;17:223–33.
2. Gliddon HD, Herberg JA, Levin M, Kaforou M. Genome-wide host
RNA signatures of infectious diseases: discovery and clinical translation.
Immunology 2018;153:171–8.
3. Liew CC, Ma J, Tang HC, Zheng R, Dempsey AA. The peripheral blood
transcriptome dynamically reflects system wide biology: a potential diagnostic tool. J Lab Clin Med 2006;147:126–32.
4. Wallace KL, Zheng LB, Kanazawa Y, Shih DQ. Immunopathology of inflammatory bowel disease. World J Gastroenterol 2014;20:6–21.
5. Ostrowski J, Paziewska A, Lazowska I, et al. Genetic architecture differences between pediatric and adult-onset inflammatory bowel diseases in
the Polish population. Sci Rep 2016;6:39831.
6. Barnes EL, Liew CC, Chao S, Burakoff R. Use of blood based biomarkers
in the evaluation of Crohn’s disease and ulcerative colitis. World J
Gastrointest Endosc 2015;7:1233–7.
7. Burakoff R, Pabby V, Onyewadume L, et al. Blood-based biomarkers used
to predict disease activity in Crohn’s disease and ulcerative colitis. Inflamm
Bowel Dis 2015;21:1132–40.
8. Planell N, Masamunt MC, Leal RF, et al. Usefulness of transcriptional
blood biomarkers as a non-invasive surrogate marker of mucosal
healing and endoscopic response in ulcerative colitis. J Crohns Colitis
2017;11:1335–46.
9. Burakoff R, Hande S, Ma J, et al. Differential regulation of peripheral
leukocyte genes in patients with active Crohn’s disease and Crohn’s disease
in remission. J Clin Gastroenterol 2010;44:120–6.
10. Hsu MJ, Chang YC, Hsueh HM. Biomarker selection for medical diagnosis using the partial area under the ROC curve. BMC Res Notes
2014;7:25.
11. Best WR, Becktel JM, Singleton JW, Kern F Jr. Development of a Crohn’s
disease activity index. National Cooperative Crohn’s Disease Study.
Gastroenterology 1976;70:439–44.
12. Seo M, Okada M, Yao T, Ueki M, Arima S, Okumura M. An index of
disease activity in patients with ulcerative colitis. Am J Gastroenterol
1992;87:971–6.
13. Hyams JS, Ferry GD, Mandel FS, et al. Development and validation of
a pediatric Crohn’s disease activity index. J Pediatr Gastroenterol Nutr
1991;12:439–47.
14. Lichawska-Cieslar A, Pietrzycka R, Ligeza J, et al. RNA sequencing reveals widespread transcriptome changes in a renal carcinoma cell line.
Oncotarget 2018;9:8597–613.
15. Anders S, Pyl PT, Huber W. HTSeq – a Python framework to work with
high-throughput sequencing data. Bioinformatics 2015;31:166–9.
16. Love MI, Huber W, Anders S. Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15:550.
17. Mikula M, Rubel T, Karczmarski J, Goryca K, Dadlez M, Ostrowski J.
Integrating proteomic and transcriptomic high-throughput surveys
for search of new biomarkers of colon tumors. Funct Integr Genomics
2011;11:215–24.
18. Zhao S, Fung-Leung W-P, Bittner A, Ngo K, Liu X. Comparison of RNASeq and microarray in transcriptome profiling of activated T cells. PLoS
One 2014;9:e78644.
19. Zhang W, Yu Y, Hertwig F, et al. Comparison of RNA-seq and microarraybased models for clinical endpoint prediction. Genome Biol 2015;16:133.
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
Crohn’s disease
1.0
JO designed the study. All authors participated in generation, collection, assembly, analysis, and interpretation of data. MK and KG performed statistical
analysis. JO, MK, and MM drafted the manuscript that was reviewed by all
the authors.
633
Blood Transcriptome Biomarkers in IBD
28.
29.
30.
31.
32.
33.
34.
35.
new biomarker for gastroenterologic diagnostics. Am J Gastroenterol
2009;104:102–9.
Minar P, Haberman Y, Jurickova I, et al. Utility of neutrophil Fcγ receptor
I [CD64] index as a biomarker for mucosal inflammation in pediatric
Crohn’s disease. Inflamm Bowel Dis 2014;20:1037–48.
Bailey CJ, Hembry RM, Alexander A, Irving MH, Grant ME,
Shuttleworth CA. Distribution of the matrix metalloproteinases stromelysin, gelatinases A and B, and collagenase in Crohn’s disease and normal
intestine. J Clin Pathol 1994;47:113–6.
Matusiewicz M, Neubauer K, Mierzchala-Pasierb M, Gamian A, KrzystekKorpacka M. Matrix metalloproteinase-9: its interplay with angiogenic
factors in inflammatory bowel diseases. Dis Markers 2014;2014:643645.
Nowarski R, Jackson R, Gagliani N, et al. Epithelial IL-18 equilibrium
controls barrier function in colitis. Cell 2015;163:1444–56.
Kumar S. Natural killer cell cytotoxicity and its regulation by inhibitory
receptors. Immunology 2018;154:383–93.
Peyrin-Biroulet L, Panés J, Sandborn WJ, et al. Defining disease severity
in inflammatory bowel diseases: current and future directions. Clin
Gastroenterol Hepatol 2016;14:348–54.e17.
Ministro P, Martins D. Fecal biomarkers in inflammatory bowel disease:
how, when and why? Expert Rev Gastroenterol Hepatol 2017;11:317–28.
Di Ruscio M, Vernia F, Ciccone A, Frieri G, Latella G. Surrogate fecal biomarkers in inflammatory bowel disease: rivals or complementary tools of
fecal calprotectin? Inflamm Bowel Dis 2017;24:78–92.
Downloaded from https://academic.oup.com/ecco-jcc/article/13/5/626/5240223 by guest on 30 June 2022
20. Cabrera SM, Chen YG, Hagopian WA, Hessner MJ. Blood-based signatures in type 1 diabetes. Diabetologia 2016;59:414–25.
21. Burczynski ME, Peterson RL, Twine NC, et al. Molecular classification of Crohn’s disease and ulcerative colitis patients using transcriptional profiles in peripheral blood mononuclear cells. J Mol Diagn
2006;8:51–61.
22. Alsobrook J, Ma T, Leighton J, et al. Novel genomic biomarkers that differentiate between inflammatory bowel disease and normal patients using
peripheral blood specimens. Am J Gastroenterol 2008;103(s1):S439.
23. van Lierop PP, Swagemakers SM, de Bie CI, et al. Gene expression analysis
of peripheral cells for subclassification of pediatric inflammatory bowel
disease in remission. PLoS One 2013;8:e79549.
24. Heida A, Van de Vijver E, van Ravenzwaaij D, et al.; CACATU consortium. Predicting inflammatory bowel disease in children with abdominal
pain and diarrhoea: calgranulin-C versus calprotectin stool tests. Arch Dis
Child 2018;103:565–71.
25. Galgut BJ, Lemberg DA, Day AS, Leach ST. The value of fecal markers
in predicting relapse in inflammatory bowel diseases. Front Pediatr
2017;5:292.
26. Heida A, Kobold ACM, Wagenmakers L, van de Belt K, van Rheenen PF.
Reference values of fecal calgranulin C [S100A12] in school aged children
and adolescents. Clin Chem Lab Med 2017;56:126–31.
27. Tillinger W, Jilch R, Jilma B, et al. Expression of the high-affinity IgG
receptor FcRI [CD64] in patients with inflammatory bowel disease: a