Peripheral blood gene expression patterns discriminate among chronic inflammatory diseases and healthy controls and identify novel targets.

Mesko et al. BMC Medical Genomics 2010, 3:15
http://www.biomedcentral.com/1755-8794/3/15Open AccessR E S E A R C H A R T I C L E
Research articlePeripheral blood gene expression patterns
discriminate among chronic inflammatory diseases
and healthy controls and identify novel targets
Bertalan Mesko†1, Szilard Poliska1†1,4, Andrea Szegedi3, Zoltan Szekanecz5, Karoly Palatka6, Maria Papp6 and
Laszlo Nagy*1,2,4
Abstract
Background: Chronic inflammatory diseases including inflammatory bowel disease (IBD; Crohn's disease and
ulcerative colitis), psoriasis and rheumatoid arthritis (RA) afflict millions of people worldwide, but their pathogenesis is
still not well understood.
It is also not well known if distinct changes in gene expression characterize these diseases and if these patterns can
discriminate between diseased and control patients and/or stratify the disease. The main focus of our work was the
identification of novel markers that overlap among the 3 diseases or discriminate them from each other.
Methods: Diseased (n = 13, n = 15 and n = 12 in IBD, psoriasis and RA respectively) and healthy patients (n = 18) were
recruited based on strict inclusion and exclusion criteria; peripheral blood samples were collected by clinicians (30 ml)
in Venous Blood Vacuum Collection Tubes containing EDTA and peripheral blood mononuclear cells were separated by
Ficoll gradient centrifugation. RNA was extracted using Trizol reagent. Gene expression data was obtained using
TaqMan Low Density Array (TLDA) containing 96 genes that were selected by an algorithm and the statistical analyses
were performed in Prism by using non-parametric Mann-Whitney U test (P-values < 0.05).
Results: Here we show that using a panel of 96 disease associated genes and measuring mRNA expression levels in
peripheral blood derived mononuclear cells; we could identify disease-specific gene panels that separate each disease
from healthy controls. In addition, a panel of five genes such as ADM, AQP9, CXCL2, IL10 and NAMPT discriminates
between all samples from patients with chronic inflammation and healthy controls. We also found genes that stratify
the diseases and separate different subtypes or different states of prognosis in each condition.
Conclusions: These findings and the identification of five universal markers of chronic inflammation suggest that
these diseases have a common background in pathomechanism, but still can be separated by peripheral blood gene
expression. Importantly, the identified genes can be associated with overlapping biological processes including
changed inflammatory response. Gene panels based on such markers can play a major role in the development of
personalized medicine, in monitoring disease progression and can lead to the identification of new potential drug
targets in chronic inflammation.
Background
Chronic inflammatory diseases such as inflammatory
bowel disease (IBD; including Crohn's disease - CD and
ulcerative colitis - UC), psoriasis and rheumatoid arthritis
(RA) exist as a substantial burden in social and economic
terms worldwide. Despite the importance of these dis-
eases, it is still not clear if characteristic gene expression
signatures can discriminate this group of diseases from
healthy controls, the various diseases from each other or
whether it is possible to stratify the diseases based on
gene expression changes.
* Correspondence: nagyl@med.unideb.hu
1 Department of Biochemistry and Molecular Biology, University of Debrecen, BioMed Central
© 2010 Mesko et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
These chronic conditions have common features such
as the autoimmune origin, the frequent co-morbidity and
Debrecen, Hungary
† Contributed equally
Full list of author information is available at the end of the article

Mesko et al. BMC Medical Genomics 2010, 3:15
http://www.biomedcentral.com/1755-8794/3/15
Page 2 of 13
a few genes such as IL10, IL23R, SLC22A4 and SLC22A5
that have been identified as contributors to their genetic
background [Table 1]. However their prevalence and the
tissues affected are clearly different.
RA is a systemic autoimmune disorder, with a preva-
lence between 0.5-1.0% [1], that causes inflammation and
tissue damage in joints and tendon sheaths. Psoriasis
which is a chronic disorder of the skin and joints where
the psoriatic plaques are areas of inflammation and
excessive skin production, affects approximately 2% of
the population only in the USA [2]. 1-2% of Western pop-
ulations suffer from IBD [3] in which the common fea-
tures are the inflammation of the intestines and the
autoimmune origin.
Global and selective gene expression analyses have
already been performed in order to gather hints on the
mechanisms of these medical conditions by using human
biopsy samples such as colon tissue in IBD [4]; skin tissue
in psoriasis [5] and synovial tissue biopsy in RA [6]. How-
ever peripheral blood is a more accessible source of cells
and may be easier to use for screening processes. Further-
more as circulating peripheral blood mononuclear cells
(PBMCs) are key cells of inflammation, it may also reflect
disease mechanisms.
In addition studies of gene expression profiling of
PMBCs may provide a more cost effective and less inva-
sive alternative to biopsy or invasive measurements [7].
Examples of the clinical implications of this approach
include the analysis of human breast cancer progression
[8] and PBMC profiles in RA, systemic lupus erythemato-
sus, type I diabetes and multiple sclerosis [9].
It appears therefore that gene expression profiling from
PBMCs is a validated tool for discovery and also may be
used to explore the pathogenetic background of these
medical conditions [10-12]. However comparative studies
on the existence of distinct and overlapping gene expres-
sion patterns are lacking. We sought to fill this gap by car-
rying out a comparative analysis of peripheral gene
expression patterns of a panel containing 96 genes in var-
ious chronic inflammatory diseases and healthy controls.
Methods
Patient recruitment
The Research Ethics Committee of University of Debre-
cen Medical and Health Science Center approved the
Table 1: Known SNP - disease associations
Gene IBD Psoriasis Rheumatoid arthritis
ADAM33 NA rs512625 PMID: 18560587 NA
IL10 rs3024505 PMID: 18836448 NA rs1800896 PMID: 18615156
IL13 NA rs1800925 PMID: 19554022 NA
IL23R rs2201841 PMID: 18338763 rs11209026 PMID: 18369459 NA
IL4 rs2243250 PMID: 18064451 NA NA
IL8 NA NA PMID: 18799095
PADI4 NA NA rs2240340 PMID: 12833157
PTGS2 rs20432 PMID: 16273614 NA rs5275 PMID: 18381795
PTPN22 NA rs1217414 PMID: 18341666 rs2476601 PMID: 18466513
SLC22A4 rs3792876 PMID: 17476680 rs11568506 PMID: 18614543 rs3792876 PMID: 15107849
SLC22A5 rs3792876 PMID: 17476680 rs2631367 PMID: 16255050 rs2631367 PMID: 15107849Known associations between single nucleotide polymorphisms and chronic inflammatory diseases are shown in the table with SNP ID and
PMID number referring to the publication that documents the association. NA means there is no documented association on SNP level. Bold
cells mean we could detect the association in our study between that specific disease and healthy controls.

Mesko et al. BMC Medical Genomics 2010, 3:15
http://www.biomedcentral.com/1755-8794/3/15
Page 3 of 13
clinical protocol and the study that were in compliance
with the Helsinki Declaration. Signed informed consent
was obtained from all healthy and diseased individuals
who provided blood sample. Inclusion and exclusion cri-
teria were developed using the best evidence currently
available. Online supplement is provided for additional
information about inclusion and exclusion criteria [Addi-
tional File 1: Figure S2].
The study included 13 patients with IBD; 15 with psori-
asis and 12 with RA, all of whom had active disease and
were medication-free at the time of blood draw [Table 2].
Blood was also obtained from a group of healthy control
individuals (18 patients) that did not show significant dif-
ferences compared to the diseased groups regarding age.
After the subjects fasted overnight for 12 hours, all of the
blood samples were obtained locally between 8:00 AM
and 9:00 AM; and were processed within one hour after
sample collection.
Peripheral blood mononuclear cell collection and RNA
processing
Venous peripheral blood samples were collected by clini-
cians (30 ml) in Venous Blood Vacuum Collection Tubes
containing EDTA (BD Vacutainer K2E). PBMCs were
separated by Ficoll gradient centrifugation.
Total RNA was extracted from PBMCs using Trizol
reagent (Invitrogen), according to the manufacturer's
protocol. RNA quality and quantity were checked on
NanoDrop and Agilent Bioanalyser 2100 (Agilent Tech-
nologies).
TaqMan mRNA analysis by RT-QPCR
Gene expression data was obtained using TaqMan Low
Density Array (TLDA) (Applied Biosystems) which is a
384-well micro fluidic card that enables to perform 384
simultaneous real-time PCR runs and which has been
used for gene expression profiling in several studies
[13,14]. This low- to medium- throughput micro fluidic
card allows for 2 samples to be run in parallel against 96
TaqMan® Gene Expression Assay targets that are pre-
loaded into each of the wells on the card. cDNA was gen-
erated with High Capacity cDNA Reverse Transcription
Kit according to manufacturer's protocol. 2 micrograms
of RNA were used per sample in the RT-PCR runs. 400 ng
(4 μl) cDNA was used in each sample. 196 μl nuclease-
free water and 200 μl 2× TaqMan Universal PCR Master
Mix (Applied Biosystems) were added for the Real-Time
Quantitative PCR measurements. This mixture was then
equally divided over four sample-loading ports of the
TLDA, each connected to one set of the 96 genes of inter-
est. The arrays were centrifuged once (1', 1300 RPM on
room temperature) to equally distribute the sample over
the wells. Subsequently, the card was sealed to prevent an
exchange between wells. RT-QPCR amplification was
performed using an Applied Biosystems Prism 7900HT
sequence detection system with the following thermal
Table 2: Patient parameters
Disease Status Control IBD Psoriasis Rheumatoid
arthritis
n 18 13 15 12
Sex (male/female) 8/10 4/9 5/10 2/10
Age (years) 40.07 ± 21.3 27.92 ± 8.49 30.53 ± 9.3 45.83 ± 18.24
Clinical subtype NA CD/UC 8/4 Arthritis positive/
negative 4/11
Bone erosion
positive/negative
4/8
CDAI UCDAI PASI DAS28
Clinical severity NA 270.4 ± 67.18 8 ± 1.2 27.47 ± 9.62 6.12 ± 1.05
124-365 6-9 15-48 4.07-7.56Definition of abbreviations: IBD = Inflammatory bowel disease, CD = Crohn's Disease, CDAI = Crohn's Disease Activity Index, UC = Ulcerative
colitis, UC DAI = Ulcerative Colitis Disease Activity Index, PASI = Psoriasis Area Severity Index, DAS28 = Disease Activity Score 28, NA = non
available. Data are presented as mean ± SD; also range in Clinical Severity.

Mesko et al. BMC Medical Genomics 2010, 3:15
http://www.biomedcentral.com/1755-8794/3/15
Page 4 of 13
cycler conditions: 2 min at 50°C and 10 min at 94.5°C, fol-
lowed by 40 cycles of 30 s at 97°C and 1 min at 59.7°C.
Gene List Selection Process
A database containing 400 genes which are associated
with the 3 chronic inflammatory diseases and inflamma-
tion was constructed by using the data derived from the
Human Genome Navigator that lists all the genes related
to a specific disease and the evidence the relation is based
on; the current medical literature and international data-
bases (e.g. OMIM, Entrez Gene). Genes were also
selected from genome-wide association studies as well as
microarray analyses focusing on skin biopsy in psoriasis,
colon biopsy in IBD and synovial fluid in RA. (Figure 1)
In the second step, genes were put in order by using a
score list that was based on the number of publications
mentioning the disease-gene association, relation to
other diseases and expected expression in PBMC accord-
ing to previous studies and our pilot experiment in which
we analyzed selected genes in PBMC samples through
individual RT-QPCR assays. This selection process
resulted in 150 genes.
In the third step, genes that did not show expression
(missing signal in QPCR measurements) in the pilot
study were removed, leading to a final list of 96 genes
Additional File 1: Figure S1.
Statistical Analyses
Relative gene expression levels of each gene were calcu-
lated by comparative Ct method and results were normal-
ized to glyceraldehyde-3-phosphate dehydrogenase
(GAPDH) expression for each sample. Statistical analyses
of the normalized gene expression data were performed
in Prism (GraphPad). Due to the fact that our data did not
follow normal distributions, the gene expression in
groups with different numbers of samples was compared
separately using the non-parametric Mann-Whitney U
test. P-values < 0.05 were considered to be statistically
significant. This method that does not include correction
for multiple comparisons is widely used in the analysis of
TLDA data [15,13] and is explained in details in [16].
Gene interactions were analyzed by using the Direct
Interactions and Biological Processes functions of Gene-
Spring GX (Agilent Technologies).
Principal component analysis (PCA), a standard, non-
parametric tool that reduces a complex data set to a lower
dimension, was used in order to reveal the internal struc-
ture of the data sets and to project the differences
between diseased and healthy groups based on each set of
significantly changing genes.
Results
Identification of gene panels discriminating chronic
inflammatory disease patients from healthy controls
A panel of 96 genes has been selected by the algorithm
depicted on Figure 1. Patients were recruited and PBMC
gene expression was determined as described in Materi-
als and Methods. When we compared the gene expres-
sion results we have found 53 genes that show significant
differences between diseased and healthy samples. First,
we looked at the gene panels that differentiate each dis-
ease from the set of control samples. Altogether 25 genes
show significant differences between IBD; 16 genes
between psoriasis; 33 genes between RA and controls
(Figure 2a-c). Principal component analysis of these gene
sets also separates the different groups of samples. In
order to get hints of the complex transcriptional basis of
inflammatory diseases and to find potential targets that
might play a role in the pathogenesis of the conditions,
the interaction among multiple genes needs to be
revealed. Gene interaction analyses highlighted IFNG,
IL4, IL10, MMP9 and TIMP1 in IBD; IL10, IL13 and
PTGS2 in psoriasis; IL8, IL10 and PTGS2 in RA. These
genes had the most direct interactions that might reveal
their key role in the pathogenetic background of the dis-
eases. Gene interaction analysis on each set of genes is
provided in the supplement material Additional File 1:
Figure S3-5.
Gene sets showing overlapping or differential expression
In order to analyze the gene expression patterns of these
conditions, first we created a Venn diagram highlighting
those genes that differentiate between a disease and the
set of control samples (Figure 3a).
Figure 1 Flowchart of gene selection process. A flow chart of the
gene selection process shows the steps through which we chose the
final 96 genes out of the database containing 400 genes.

Mesko et al. BMC Medical Genomics 2010, 3:15
http://www.biomedcentral.com/1755-8794/3/15
Page 5 of 13
Performing Gene Interaction Analysis resulted in the
identification of a network that highlights the genes with
the most interactions such as CCL5, IFNG, IL4, IL8, IL10,
IL13, MMP9, PTGS2 and TIMP1 (Figure 3b) clearly
showing that not only individual genes but entire net-
works are impacted and hence can be identified.
Although there are genes that showed disease specific
signatures, all of these genes showing significant differ-
ences between diseased and healthy samples form a net-
work which might represent the common background of
the pathogenesis of this type of inflammation.
As the next step in our analysis all the significantly
changing genes were assigned to functional categories
that were created based on EASE, the Expression Analy-
sis Systematic Explorer that provides statistical methods
for discovering biological themes within gene lists [17].
genes without categories) categories with 21, 11, 7, 9 and
7 genes respectively. In order to illustrate the similarities
in the pathogenetic background of these diseases, the
functional categories, the number of genes and the dis-
ease gene panels are visualized. It shows that there are
similar number of genes in the same functional categories
in each of the diseases, though the number of unique
genes (11 in IBD, 6 in psoriasis and 21 in RA) separating
only a disease from the control group is also high (Figure
4).
Diseases subtype stratification by differentially expressed
genes
We also sought to identify potential markers that differ-
entiate between distinct subtypes or states of prognosis in
each of the three diseases to stratify the disease based on
gene expression patterns (Figure 5). Analyses of individ-
Figure 2 Fold changes of genes differentiating between diseased and control samples. Fold change values of genes, showing statistically sig-
nificant (Mann-Whitney U test) differential expression between diseased and control patients, were generated from RT-QPCR measurements and rep-
resent the difference of the means of the diseased and control groups. Principal component analysis was performed and separates the two groups of
samples. 2a represents the IBD, 2b the psoriasis and 2c the RA gene panel.Genes were grouped into inflammatory response; cell
growth and maintenance; proteolysis; metabolism and
unclassified (including genes in unique categories or
ual genes are provided in the supplement material Addi-
tional File 1: Figure S7-9.