DNA microarray allows molecular profiling of rheumatoid arthritis and identification of pathophysiological targets


This study was undertaken to evaluate the possibility to obtain a molecular signature of rheumatoid arthritis (RA) comparatively osteoarthritis (OA), and to lay the bases to develop new diagnostic tools and identify new targets. Microarray technology was used for such an analysis. The gene expression profiles of synovial tissues from patients with confirmed RA, and patients with OA were established and compared. A set of 63 genes was selected, based, more specifically, on their overexpression or underexpression in RA samples compared to OA. Results for six of these genes have been verified by quantitative PCR using both samples identical to those used in the microarray experiments and entirely separate samples. Expression profile of the 48 known genes allowed the correct classification of additional RA and OA patients. Furthermore, the distinct expression of three of the selected genes was also studied by quantitative RT-PCR in cultured synovial cells. Detailed analysis of the expression profile of the selected genes provided evidence for dysregulated biological pathways, pointed out to chromosomal location and revealed novel genes potentially involved in RA. It is proposed that such an approach allows valuable diagnosis/prognostics tools in RA to be established and potential targets for combating the disease to be identified.


Rheumatoid arthritis (RA) affects approximately 0.5% of the world adult population and represents a significant cause of disability. It is a chronic, debilitating joint disease of unknown origin, which manifests as diffuse symmetrical polyarthritis. Joint inflammation is characterized by synovial hyperplasia associated with infiltration of many different inflammatory cell types, and which progresses to cartilage and bone destruction. Key issues in this pathology are: the mechanisms of induction of inflammation, the immunological network leading to disease perpetuation/chronicity, and identification of mechanisms that promote disease resolution. Answers to these fundamental questions would lead to effective treatments for a range of inflammatory disorders.

Dysregulated immune responses within the inflamed synovial membrane probably contribute to cartilage and bone erosion. The large array of cytokines and factors produced in the affected joints, and the multiple cell interactions dictate the evolution of arthritis. The levels of proinflammatory molecules, particularly the monokines TNF-α and IL-1 β, exceed those of anti-inflammatory moieties such as IL-10 and TGF-β.1, 2 Biological therapies targeting TNF-α and now IL-1 and IL-6 have clearly defined the central importance of cytokine production and macrophage-induced inflammation in RA, but these achieve only transient clinical responses, unless given repetitively.3, 4, 5, 6 Moreover, only 40% of patients reach the American College of Rheumatology (ACR) 50% response. This highlights the clinical necessity for generating further, novel, pathogenesis-led interventions.

To reach this goal, it is crucial to identify and quantify the genes expressed by the cells that are present and interacting in the tissue. Although an approach which is focused on one gene or gene product is important, a real effort has to be made to avoid a too restrictive vision, since articular diseases such as RA are multifactorial disorders, and a variety of possibilities must be taken into account. In order to increase the understanding of the mechanisms involved in these diseases, and consequently to identify new therapeutic targets and develop new diagnosis tools, it is of prime importance to precisely analyse the genes expressed and their level of expression in the tissue. Until recently,7, 8 no comprehensive, systematic study of the genes expressed or repressed during RA had been reported. Expression profiling has already shown its usefulness in identifying genes in specific cell types under defined conditions and in establishing characteristic patterns of gene expression in a variety of diseases.9, 10 In a proof concept study, DNA array technology was used to study gene expression in RA,11 showing that this approach is feasible and can give interesting results.

Other studies have focused on culture of synovial fibroblasts (SF) by combining a DNA array and an RNA arbitrary-primed PCR.12, 13 However, these studies did not allow appreciation of the overall modification of gene expression in RA. Furthermore, although working with homogenous cell populations such as SF has the clear advantage to simplify the analysis of DNA array experiments and to study a subpopulation of cells known to be involved in the disease, it has also major drawbacks such as modification of gene expression due to cell culture, purity and synchronization of the culture, selection of particular subset of cells.

It has been recently demonstrated that the ACR classification criteria were not useful for predicting the diagnosis of RA in patients, within the first year of arthritis onset.14 This is certainly due to the fact that the 1987 ACR criteria were developed as a tool for classifying patients who had already been identified as RA patients by a rheumatologist and not as a means of actually diagnosing RA, although these criteria are increasingly used as the gold standard for the diagnosis of RA. These results collectively demonstrate the real need for new evaluation methods for the diagnosis of the disease. As a first step and to evaluate the possibility to reach this goal, we present here an analysis of genes differentially expressed in synovial tissue from RA and OA patients undergoing reconstructive surgery or synovectomy. We used microarrays containing 5760 probes derived from a cDNA library. By measuring expression of these genes in RA and OA tissue samples, we were able to define a selection of 48 known and 15 unknown genes that are differentially expressed in RA and OA tissues.

Data analysis revealed that molecules belonging to restricted biological processes, for instance transcription/transduction/cell cycle and proteases, were preferentially differentially expressed when comparing OA and RA tissues. In addition, results concerning some of these products have been validated by another techniques on greater number of samples.


Global gene expression profile

A total of 15 RA and OA tissues (five RA and 10 OA) were analysed by cDNA microarray, and expression data for 5760 genes were obtained. Only tissues from patients with a confirmed diagnosis and without any other associated disease were used (see Materials and methods). With expression data for 4254 expressed genes that passed prefiltering, a comparison of gene expression between RA and OA tissues led to the selection of 63 differentially expressed genes on the basis of their significant differential expression between OA and RA patients (Table 1). From those 63 genes, 48 were known genes and 15 were unknown ESTs. Expression of 31 of the identified genes, such as guanylate-binding protein 1 (GBP1), TAF2E, clusterin (CLU), DDX17 and CDK7 was ascertained by redundancy hits, two or more independent clones, in the cDNA arrays (Table 1 and Web Table A, see Supplementary information).

Table 1 Set of genes with increased and decreased mRNA expression ratio

Differentially expressed genes in RA compared to osteoarthritis

Among the 63 selected, differentially expressed genes, 15 showed higher expression and 48 lower expression in RA than in OA. We found that these genes participated in a limited number of functions (Table 1). In total, 36% of the selected genes had unknown functions. Of the genes with known functions, 77% were involved in four main classes of cellular functions: transcription and the cell cycle (33%), signal transduction (19%) metabolism (16%), and proteases and inhibitors (9%) (Figure 1).

Figure 1

View of biological functional family gene repartition of the known selected genes. Each gene is described in more detail in Table 1.

Clustering of samples with a selected set of differentially expressed genes

We clustered samples by using all 63 differentially expressed genes. To do so we used an unsupervised algorithm that groups samples into a predetermined number of clusters on the basis of their gene expression patterns.15, 16 The resulting hierarchical clustering is shown in Figure 2a, and as expected all OA and RA samples were correctly classified. In this setting all the RA samples were arranged side by side and all OA were also arranged side by side.

Figure 2

Expanded view of distinct gene expression signatures of selected differentially expressed genes between OA and RA samples. Expression level was normalized per gene, data were log-transformed and the relative value to the mean among the 15 samples is shown by color: red; relatively high expression, green; low expression. The clustering program arranges samples (10 OA and five RA) along the horizontal axis so that those with the most similar expression profiles are placed adjacent to each other. The organization of the results was made with the same number of genes and arrays used in the experiment and means of redundant hits were used. Rows represent genes (unique cDNA element) and columns represent experimental samples. OA=osteoarthritis, RA=rheumatoid arthritis. (a) Expanded view of distinct gene expression signatures of 63 selected differentially expressed genes between OA and RA samples. (b) Expanded view of distinct gene expression signatures of 48 known selected differentially expressed genes between OA and RA samples.

The 48 known genes that were identified as differentially expressed in RA vs OA tissues were then subjected to cluster analysis. Figure 2b shows that the selection of these 48 genes was good enough to group all the RA samples together and to group all OA together.

Validating the selected genes

Quantitative RT-PCR analyse of selected genes

Six of the selected genes were further studied by real-time, quantitative PCR in order to validate the results obtained using microarray. Three (CTSL (cathepsin L), GBP1, GLO1 (glyoxalase 1)) were chosen for their higher expression and three (CLU, DXS1357E (B-cell receptor-associated protein 31), RH70/DDX17 (RNA helicase 70 kDa)) for their lower expression in RA compared to OA. The genes for this in-depth study were also chosen because of their different cellular functions and for their potential relevance to disease. For these genes, we compared the results obtained when the same samples were analysed using both the microarray technique and real-time, quantitative PCR. Although the number of samples analyzed simultaneously by the two techniques is rather limited, the results shown in Table 2 demonstrate that the mean ratio of expression between RA and OA of these six genes is fully concordant when microarrays and real-time PCR results were compared. Increasing the number of samples analysed by real-time PCR further validated these results. In this case, more than 50% of the samples that we used had not been tested by microarray. As shown in Table 2, the ratio of expression of the six tested genes fully corroborated the results obtained with the microarray technique, which further validated the set of selected genes. In addition, Western blot analysis using commercially available polyclonal antibodies revealed an overexpression of GBP-1 protein in RA tissue compared to OA (data not shown).

Table 2 Relative quantification of transcription for six distinct genes

The selected gene cluster allowed differential diagnosis between OA and RA

To validate our findings, new samples from RA (n=2) and OA (n=3, one of which was the contralateral knee of a patient already tested) were processed for microarray procedure. These new data were added to those of the first experiment and subject to a cluster analysis focused on the 63 genes previously selected. As shown in Figure 3a, each of these five new samples was correctly classified in the OA and RA categories. Furthermore, the two samples obtained from the same patient 1 year on (OA 2 and OA 11) demonstrated a very similar expression profile. These results showed that the set of genes is probably a good disease marker. Similar cluster analysis conducted on the 48 known genes that were identified as differentially expressed in RA vs OA tissues confirmed the diagnosis capabilities of the selected genes (Figure 3b).

Figure 3

The selected set of 63 (a) or 48 (b) genes allows patients classification. Analyses were performed with five new samples added to those in Figure 2. New samples were from two RA patients (RA6 and ACJ§, a 32-year-old woman in whom the disease had begun 20 years earlier as chronic juvenile arthritis) and three OA patients (OSN1 and OSN2, #two patients who developed osteonecrosis before OA, and OA11*, a second sample of OA2 obtained 1 year later from the contralateral knee).

Expression of selected genes in cultured synovial cells

Since SF appear to play a major role in the pathogenesis of RA, we measured the expression of three of the selected genes in the synovial cells: one gene with higher (GBP-1) and two genes (CLU, RH70/DDX17) with lower expression in RA compared to OA. Third-passage RA- and OA-synovial cells from seven OA and four RA patients were used to evaluate the gene expression by real-time PCR. While the expression of GBP1 and RH70/DDX17 did not differ when comparing SF-RA and SF-OA, CLU transcripts in cultured SF-RA were significantly lower than those from SF-OA (Figure 4). Thus, the low expression of CLU noted in RA synovial tissue collected ex vivo persisted in SF expanded in vitro, suggesting that the downregulation of this gene may be an intrinsic property of RA synoviocytes.

Figure 4

Quantitative RT-PCR determination of gene expression in cultured synovial cells. Results are arbitrary units (see Materials and methods) and expressed as ratios of RA or OA cultured synovial cells to control RNA. All the experiments were performed using SFs at the third passage in tissue culture; at this time the contaminating cells were less than 2%. Data are from four RA and seven OA patients. *P<0.001, RA vs OA. Determinations were carried out at least in triplicate. Nonparametric tests were used. The Mann–Whitney test was used to evaluate the difference between groups. A P-value<0.05 was considered statistically significant.


Among the main clinical challenges RA represents, two are particularly crucial: (1) the difficulty in establishing a rapid and reliable diagnosis; (2) the need for new therapeutic targets. Herein, we report on microarray studies of OA and RA synovial tissues leading to the selection of a limited number of genes differentially expressed in the two types of diseases, which allowed correct diagnosis in a series of additional OA and RA patients. For some genes, real-time PCR has been used to ascertain their differential expression in the same samples used for microarray experiments, and in a new, independent set of samples. Furthermore, real-time PCR has also been used to study three of the differentially expressed genes in synovial cultured cells.

One-third of the genes we described as being implicated in the distinction between OA and RA are either unknown or have unknown functions. Among the 48 known genes, some were already reported to be differentially expressed between RA and OA such as CTSL, CTSD and TIMP2. Notably, more than two-third of the genes belong to three main classes: transcription and cell cycle (33%), transduction (19%) and metabolism (16%). The restricted number of class these known genes belong to was particularly interesting because reveals, at the molecular level, differences in features known or suspected as being different in OA and RA.

Indeed, we found that five of the selected genes with known functions belong to the GTP/GDP metabolism class, making these genes and the respective pathways attractive candidates for new therapeutic attention and interventions. More precisely, four of those five genes were downregulated in RA. Interestingly, Neumann et al13 reported recently the decreased expression of Rho A, a GTPase related to formation of stress fibers, in cultured RA synoviocytes compared to OA synoviocytes. This result has to be linked to the decreased expression we observed of ARHA, a Ras-related GTP binding protein of the rho subfamily, which regulates reorganization of the actin cytoskeleton. On the other hand, GBP-1 was found to be upregulated, a result that was recently reported by Van der Pouw Kraan.8

Looking at the chromosomal localization of the selected genes revealed several interesting features. Between the 63 selected genes, 49 had reliable chromosomal location. Among those genes, three were located at the chromosome 6p21 region that harbors the HLA genes,17 and which is responsible for one-third of the genetic influence on the development of RA. These results are in line with a recent study of a distinctive gene expression profile in rheumatoid synovium using microarray analysis reporting the localization of nine genes of interest in 6p21.7 We also found that four genes were located to chromosome X with three of them to Xq28, results that cannot be explained by a hazardous distribution or by a biased patient recruitment since the same proportion of male and female patients were used for either diseases. Moreover, six other genetic regions containing susceptibility loci for inflammatory diseases, namely 22q12–13; 19q13; 17q21–22, 19p12–13 (an IBD susceptibility region18), 1p36 and 18q12 (RA susceptibility regions19) contained altogether 10 genes that we found differentially expressed between RA and OA. Such preferential grouped chromosomal location could be the result of different regulation/modulation of transcription between the two diseases and could help on the identification of genetic risk factors that is a challenge for complex diseases.

Doing gene correlation in a disease as complex as RA was fully behind the scope of this work and is presently difficult to envisage out of perfectly defined in vitro experimental conditions.20

Recently, Ship et al22 and Pomeroy et al21 described prediction of outcome for large, B-cell lymphoma and central nervous system embryonal tumours, based on gene expression profiling and supervised learning. Such approaches necessitate both analysing a large number of specimens, more than 40 even for a quite homogenous disease, and access to samples very early in the course of the disease and before any treatment. These three main requirements represent all major difficulties when studying RA.

Identifying a set of genes that could be used to molecularly distinguish RA and OA was one of the ideas behind this study. Analysing expression profiling of the selected set of genes is already highly informative in this context since, by using only the set of known genes, (i) two different samples from the same patient were similarly classified, (ii) those patients with OA associated with primary osteonecrosis were correctly classified among the others OA. In this work, the classification results were obtained without weighting the expression of any gene, assuming that they have all the same weight in the classification. Presently, we are working on the improvement of the method of classification. Further investigations are currently under progress to identify the genes that have a higher prediction factor. Using first SAM to estimate a value for each missing data, PAM was then used for classification. PAM method is used to find genes in a set of DNA chips, which best classifies samples and to validate a set of genes to classify samples. Data point to a set of 14 out of 63 genes (Web Text A, and Web Figures 1–4, see Supplementary information).

Although 32 out of the 48 of the known genes that belong to the selected set of genes reported herein were not present in Van der POUW KRAAN's arrays8 (Web Table B, see Supplementary information), seven out of the 16 comparable genes between the two studies were found identically selected. This comparison shows that both approaches, in two different laboratories, could lead to identification of same genes.

In a second series of experiment, we focused our work on an in-depth study by real-time PCR of six selected genes, namely CLU, CTSL, DXS1357E, GBP1, GLO1 and RH70/DDX17, to further validate the results obtained using the microarray technique at two different levels. First, by comparing the results obtained when the same samples were analysed using both techniques and second, by testing an independent set of tissue samples. In both settings, the ratio of the expression of the six genes tested using the real-time PCR technique fully corroborated the results obtained using the microarray technique.

Since the six genes studied in more details have been selected on the basis of (1) obvious genes given the concept of the disease; (2) genes belonging to a family or metabolic pathway specifically modified in the particular disease; (3) novelty (gene with unknown function or for which few data are available); and (4) magnitude of the differential expression, we think that they are good candidates not only for diagnostic purposes but also as pathophysiological factors.

For instance, clusterin is a glycoprotein with multiple activities23, 24, 25, 26, 27 and many of these activities could be related to the physiological processes involved in RA. We found a highly significant, lower expression of CLU when comparing RA to OA (P<0.001), making the measurement of CLU gene expression, an highly interesting test for distinguishing between OA and RA tissues (Devauchelle et al, submitted). Interestingly, differences in expression of CLU mRNA between OA and RA tissues were also found, in the same range, when comparing CLU mRNA expression in cultured OA and RA synoviocytes.

DDX17/RH70 is a gene that has been very recently described as a bidirectional, RNA helicase.28 Since it is expressed at lower levels in RA compared to OA tissue, it could be responsible for lower efficiency mRNA splicing or mRNA alterations in RA. Furthermore, DDX17 associates with U1snRNP, a target autoantigen, making it a possible target for an autoimmune response and/or an active player in the emergence of autoantigen.

Although the number of samples analysed is still too low to achieve definitive conclusions, results have shown that equivalent expression of CTSL was found between RA patients and patients affected by OA associated with primary OSN and higher than detected in samples of patients affected of OA alone. Interestingly, the expression of another cathepsin (CTSB) was higher in OA compared to RA. Since CTSL has been involved in bone degradation,29, 30 these results suggest that CTSL gene expression is a potential marker for bone erosion, and that it might be a candidate for therapeutic intervention.

On the other hand, expression of GBP1 and CTSL (a cysteine proteinase involved in joint destruction and bone resorption in RA tissues) was upregulated in RA tissues confirming results reported.8 Both genes are known to be IFN-γ responsive through the STAT1 signalling31, 32 and particularly in synovial cells.33, 34 Interestingly, we provided evidence that the differential expression of GBP-1 and CTS-L observed in tissues was lost in cultured SF. These results raised the possibility that these genes were activated in RA tissue potentially through the action of cytokine. To test this hypothesis, we are currently studying whether the IFN-γ responsiveness of these genes is different in OA and RA synovial cells.


The differences in gene expression may suggest biological pathways for therapeutic intervention and facilitates identification of more homogeneous groups of patients for clinical studies and trials. Our concern in this work was to use a powerful technology to lay the bases to define novel molecular diagnostic tools and to identify potential targets for therapeutic intervention in RA. On the basis of microarray technology, our work points to a set of distinct genes whose elevation or impaired expression correlates with RA. These findings open up the possibility in the future of the clinical use of a ‘rheumatoid chip’, as a diagnostic tool and may enable scientists to better tailor therapeutic strategies. Differences in gene expression profiles will provide a unique perspective to allow us to distinguish between different, pathogenic RA and OA subsets, based on molecular criteria.

Patients and methods

Synovial tissue and cell culture

Seven patients with RA (five females, two males, mean disease duration of 14.3 years, ranging from 3 to 32 years), who fulfilled the RA criteria of the ACR35 were evaluated. The mean age was 55±9.2 years. The mean±s.d., C reactive protein level was 17±12 mg/l. Four RA patients were receiving corticosteroids, and four received disease-modifying antirheumatic drugs. One had received anti TNF therapy, which was stopped 8 weeks prior to surgery. One patient, a 32-year-old woman in whom the disease had begun 20 years earlier as polyarticular juvenile idiopathic arthritis (rheumatoid factors positive) was also tested. In total, 12 patients with OA (seven females, five males, mean disease duration of 7.6 years, ranging from 1 to 20 years) were used for comparison. The mean age was 72±9.3 years. The mean±s.d., C reactive protein level was 8.5±20 mg/l. All patients with OA received analgesic drugs, three received nonsteroidal anti-inflammatory drugs. RA and OA samples were mainly obtained from patients undergoing knee replacement surgery. None of the patients received intra-articular injection of corticoid in the month preceding surgery. All the patients included in the study were also seen by a rheumatologist (GF or VD) to confirm the diagnosis.

Tissues were obtained, with informed consent, from patients who underwent remedial synovectomy or arthroplasty of the knee. Synovial tissue was dissected from surgical specimens taking particular care that there was minimal contamination by nonsynovial tissue, and samples were immediately processed for RNA extraction and when possible, treated prior to cell culture with collagenase dispase. Synoviocytes were cultured in RPMI 1640 supplemented with 10% fetal calf serum, 100 μg/ml streptomycin and 500 U/ml penicillin. Jurkat, THP-1, U937 and HaCat36 cells were cultured in the same medium. To prepare RNA from the cell lines, the cells, apart from HaCat and activated Jurkat, were seeded at 500 000/ml, 15 h before harvesting. HaCat cells were seeded at 10 000 cells/cm2 and cultured for 10 days before harvesting for RNA extraction. Medium was changed every 2 days. Activated Jurkat cells were incubated at 1 × 106 cells/ml for 6 h in the presence of PHA (1 μg/ml) and PMA (10 ng/ml) before harvesting.

RNA preparation

After dissection, synovial tissues were promptly frozen in RLT® RNA extraction buffer (Qiagen, Rneasy kit). Total RNA was extracted and treated with DNase I to eliminate genomic DNA contamination. The integrity of the RNA was analysed by gel and RT-PCR and concentration was measured by absorbance at 260 nm.

Generation and postprocessing of cDNA microarrays

Gene expression analysis was performed using cDNA microarrays containing 5760 human probes that had been sequenced twice. The mean size of probes was around 1.500 pb. Altogether, the microarrays allowed detection of 4652 different genes (Web Table C, see Supplementary information). The PCR products were prepared in 96-well plates, purified by ethanol precipitation, washed in 70% ethanol, dried, dissolved in TE/DMSO (50/50) and stored at −80°C. The quality, size and concentration of each PCR product were assessed by agarose gel electrophoresis using a calibrated standard DNA solution (Invitrogen). PCR products were arrayed onto poly-L-lysine-coated slides (Menzel Glaser) using the 4-pins GMS 417 arrayer (Genetic Microsystems).

Before hybridization, the slides were hydrated over boiling water and dried at 80°C. Spotted DNA was crosslinked by UV irradiation (130 mJ at 254 nm). Slides were totally immersed in a freshly prepared blocking solution consisting of 2% succinic anhydride, 20 mM sodium borate pH 8 in methyl-2-pyrrolidinone. Slides were washed in water and immersed in 100% ethanol at −20°C before a final centrifugation for 7 min at 100 g at room temperature. Slides were then prehybridized at 42°C in 3 × SSC, 0.1% SDS, 0.1% BSA solution for 30 min. Following prehybridization, the slides were washed successively in water, propanol-2 and 100% ethanol. Finally, the slides were centrifuged for 5 min at 100 g at room temperature.

cDNA synthesis and hybridization

Labeled cDNAs were synthesized by reverse transcription of total RNA (20 μg) from synovial samples in the presence of oligo(dT-VN) using the Superscript II reverse transcription kit (Gibco-BRL) and amino-allyl dUTP. The reference RNA consisted of a mix of equivalent amounts of Jurkat, activated Jurkat, THP-1, U937 and HaCat cell lines) (20 μg total RNA). The use of a common control cDNA probe allows the relative expression of each gene to be compared across all samples.

The cDNAs were purified on Microcon filters (Amicon) as described on P. Brown's Web site (http://brownlab.stanford.edu/protocols.html) and on Service de Génomique Fonctionnelle's, Web site (http://www.genopole.org/html/en/connaitre/cite/outils/plateforme_sgf.htm). Purified DNAs were spotted onto poly-L-lysine-coated slides (Menzel Gläser), using an automated arrayer (Genetic Micro Systems). Prehybridization and hybridization conditions were those described by Brown and colleagues (http://brownlab.stanford.edu/protocols.html) with a few minor modifications (http://www.genopole.org/html/en/connaitre/cite/outils/plateforme_sgf.htm). Briefly, Cy3- and Cy5-labeled cDNAs were mixed with 30 μl of hybridization solution (50% formamide, 1.75 × Denhardt's, 0.35% SDS and 6 × SSPE, 10 μg of human Cot-1 DNA (Invitrogen), 10 μg poly(A) (Sigma-Aldrich) and 10 μg of yeast tRNA (Invitrogen)). The probe was heated for 2 min at 100°C, incubated at 37°C for 20–30 min and placed between a slide and coverslip. The arrays were incubated overnight at 42°C in an humidified slide chamber (Corning), and washed initially in a bath of 0.1 × SSC with 0.1% SDS for 2 min at 42°C, then in 0.1 × SSC with 0.1% SDS for 10 min at room temperature followed by two washing steps in 0.1 × SSC for 15 min each at room temperature. The arrays were dried by centrifugation at 600 rpm for 6 min.

Hybridized arrays were scanned using a GenePix 4000B scanner (Axon Instruments, Inc.). Fluorescence intensities of Cy3 and Cy5 were measured separately at 532 and 635 nm and fluorescence ratio measurements were determined with the GenePix Pro 3.0 software (Axon Instruments, Inc.).

Image acquisition and data analysis

The resulting 16-bit data files were imported into an image analysis software (Axon Genepix 3.0, Axon). Feature ratios were calculated after background subtraction using Genepix 3.0. This software flags spots as absent, based on spot characteristics. Other bad spots were manually flagged and no flagged spots were included in subsequent analysis. Data were scaled so that the average median ratio values for all spots were normalized to 1.0. Additionally, only spots with more than 70% of their pixels fluoresced with an intensity greater than 2 standard deviations above background noise in one of the two wavelengths were selected for further analysis.

The data were normalized by applying a uniform scaling factor, assuming that the arithmetic median of the ratios for every spot considered was equal to 1. Further parameters were additionally used for the selection of the differentially expressed genes: (1) only the genes selected on the base of flag filtering and present in at least 70% of the samples of each type of disease were analyzed; (2) the mean difference of expression in each group should be at least 1.6; (3) Student test used should be significant (<0.05); and (4) each of the redundant hits was taken into account separately.

The verification whether OA and RA populations have homogeneous variance has been performed via the Fisher unilateral test of equality of variance in Excel (TEST.F function). Probes rejected at a 5% P-value have been considered having statistically different variance in the two conditions. The test whether OA and RA populations have the same statistical mean has been performed via a bilateral t-test in Excel (T.TEST function); homo or heteroschedasticity of variance has been assumed from the results of the previous Fisher test. Probes rejected at 5% P-value have been considered having significantly different means (Web Table D, see Supplementary information).

Clustering were performed using Cluster and TreeView software.37 Normalization was made by gene, and data were log-transformed. To avoid an analysis bias, the organizations of the results were made with the same number of genes and arrays used in the experiment. To avoid false weighting of some genes expression, only one value per gene (mean of redundant hit when applicable) was used. Several analyses were made. The data presented resulted from a hierarchical clustering made with an uncentered correlation.

Real-time PCR analysis

Real-time PCR was performed using a LightCycler system (Roche diagnostics, Mannheim, Germany), according to the manufacturer's instructions. Reaction were performed in a 20 μl volume with 0.5 μl primers, 2 μl of LightCycler FasStart reaction mix SYBR Green I (Roche Diagnostics), and adequate concentrations of MgCl2. β-Actin was used as the housekeeping gene for relative quantification to normalize target gene expression. To confirm amplification specificity of a single specific gene product and to exclude primer dimers, the PCR products from each primer pair were subjected to a melting curve analysis. The RT-PCR products of each primer PCR were also subjected to electrophoresis on a 2% agarose gel to confirm the presence of the specific gene amplicon.

Sequence of PCR primers were as follows:

  • β-Actin forward, 5′-IndexTermGGG TCA GAA GGA TTC CTA TG-3′ (237 pb) (57°C);

  • β-Actin reverse, 5′- IndexTermGGT CTC AAA CAT GAT CTG GG-3′;

  • GBP1 forward, 5′-IndexTermAAC CAT CAA CCA GCA GGC TAT-3′ (187 pb) (62°C);

  • GBP1 reverse, 5′-IndexTermTTG TCC ATC TGC TTC CAA GTC-3′;

  • CLU forward, 5′-IndexTermGCG AAG ACC AGT ACT ATC TG-3′ (201 pb) (57°C);

  • CLU reverse, 5′-IndexTermTTT TGC GGT ATT CCT GCA GC-3′;

  • RH70/DDX17 forward, 5′-IndexTermACCGATAGAGCTGGTTAT GC-3′ (221 pb) (60°C);

  • RH70/DDX17 reverse, 5′-IndexTermACTGCTGGCTAGAGCTCTGT-3′;

  • GLO1 forward, 5′-IndexTermCACTCTACTTCTTGGCTTAT-3′ (204 pb) (60°C);

  • GLO1 reverse, 5′-IndexTermTGTATACATCAGGAACAGCA-3′;

  • DXS forward, 5′-IndexTermGAA GAG AAC AGG AGC CTG AA-3′ (190 pb) (60°C);

  • DXS reverse 5′-IndexTermCCA TGG GAC CAT CTA CTG CA-3′;

  • CTSL forward, 5′-IndexTermATGAAGGCAGTTGCAACTGT-3′ (156 pb) (62°C);

  • CTSL reverse, 5′-IndexTermCTGTGCTTTCAAATCCGTAG-3′.

Each primer efficiency was tested using corresponding annealing temperatures and MgCl2 concentrations (2–7 mM). Primer efficiency was calculated using the standard curve method (E=10−1/slope, where E represents the primer efficiency). Each standard curve was determined with multiple dilutions steps and replicates, and stored as a coefficient file used for the analysis. An efficiency-corrected calculation was performed by the LightCycler Relative Quantification software using the so-called coefficient files, according to the manufacturer's recommendations.

PCR quantification: For calculation of the final result, we used the software Rel-Quant (Roche Diagnostics). The PCR quantification was a relative quantification with a normalized calibrator with efficiency correction. The calculation was based on the crossing point of a sample and the efficiency of the PCR reaction, according to the general PCR equations. The normalized ratio is the result of the target/reference ratio of each sample divided by the target/reference ratio of our calibrator. We used actin as our housekeeping, reference gene. The calibrator was the same pool of control RNA as that for the microarray experiments.


  1. 1

    Vervoordeldonk MJ, Tak PP . Cytokines in rheumatoid arthritis. Curr Rheumatol Rep 2002; 4: 208–217.

  2. 2

    Arend WP, Dayer JM . Cytokines and cytokine inhibitors or antagonists in rheumatoid arthritis. Arthritis Rheum 1990; 33: 305–315.

  3. 3

    Arend WP, Dayer JM . Inhibition of the production and effects of interleukin-1 and tumor necrosis factor alpha in rheumatoid arthritis. Arthritis Rheum 1995; 38: 151–160.

  4. 4

    Elliott MJ, Maini RN, Feldmann M et al. Randomised double-blind comparison of chimeric monoclonal antibody to tumour necrosis factor alpha (cA2) vs placebo in rheumatoid arthritis. Lancet 1994; 344: 1105–1110.

  5. 5

    Lard LR, Visser H, Speyer I et al. Early vs delayed treatment in patients with recent-onset rheumatoid arthritis: comparison of two cohorts who received different treatment strategies. Am J Med 2001; 111: 446–451.

  6. 6

    Goekoop YP, Allaart CF, Breedveld FC, Dijkmans BA . Combination therapy in rheumatoid arthritis. Curr Opin Rheumatol 2001; 13: 177–183.

  7. 7

    Van Der Pouw Kraan TC, Van Gaalen FA, Huizinga TW et al. Discovery of distinctive gene expression profiles in rheumatoid synovium using cDNA microarray technology: evidence for the existence of multiple pathways of tissue destruction and repair. Genes Immun 2003; 4: 187–196.

  8. 8

    van der Pouw Kraan TC, van Gaalen FA, Kasperkovitz PV et al. Rheumatoid arthritis is a heterogeneous disease: evidence for differences in the activation of the STAT-1 pathway between rheumatoid tissues. Arthritis Rheum 2003; 48: 2132–2145.

  9. 9

    Perou CM, Sorlie T, Eisen MB et al. Molecular portraits of human breast tumours. Nature 2000; 406: 747–752.

  10. 10

    Alizadeh AA, Eisen MB, Davis RE et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling [see comments]. Nature 2000; 403: 503–511.

  11. 11

    Heller RA, Schena M, Chai A et al. Discovery and analysis of inflammatory disease-related genes using cDNA microarrays. Proc Natl Acad Sci USA 1997; 94: 2150–2155.

  12. 12

    Masuda K, Masuda R, Neidhart M et al. Molecular profile of synovial fibroblasts in rheumatoid arthritis depends on the stage of proliferation. Arthritis Res 2002; 4: R8.

  13. 13

    Neumann E, Kullmann F, Judex M et al. Identification of differentially expressed genes in rheumatoid arthritis by a combination of complementary DNA array and RNA arbitrarily primed-polymerase chain reaction. Arthritis Rheum 2002; 46: 52–63.

  14. 14

    Saraux A, Berthelot JM, Chales G et al. Ability of the American College of Rheumatology 1987 criteria to predict rheumatoid arthritis in patients with early arthritis and classification of these patients two years later. Arthritis Rheum 2001; 44: 2485–2491.

  15. 15

    Golub TR, Slonim DK, Tamayo P et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999; 286: 531–537.

  16. 16

    Tamayo P, Slonim D, Mesirov J et al. Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc Natl Acad Sci USA 1999; 96: 2907–2912.

  17. 17

    MacKay K, Eyre S, Myerscough A et al. Whole-genome linkage analysis of rheumatoid arthritis susceptibility loci in 252 affected sibling pairs in the United Kingdom. Arthritis Rheum 2002; 46: 632–639.

  18. 18

    Lawrance IC, Fiocchi C, Chakravarti S . Ulcerative colitis and Crohn's disease: distinctive gene expression profiles and novel susceptibility candidate genes. Hum Mol Genet 2001; 10: 445–456.

  19. 19

    Cornelis F, Faure S, Martinez M et al. New susceptibility locus for rheumatoid arthritis suggested by a genome-wide linkage study. Proc Natl Acad Sci USA 1998; 95: 10746–10750.

  20. 20

    Iyer VR, Eisen MB, Ross DT et al. The transcriptional program in the response of human fibroblasts to serum. Science 1999; 283: 83–87.

  21. 21

    Pomeroy SL, Tamayo P, Gaasenbeek M et al. Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 2002; 415: 436–442.

  22. 22

    Shipp MA, Ross KN, Tamayo P et al. Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med 2002; 8: 68–74.

  23. 23

    Hogasen K, Mollnes TE, Harboe M et al. Terminal complement pathway activation and low lysis inhibitors in rheumatoid arthritis synovial fluid. J Rheumatol 1995; 22: 24–28.

  24. 24

    Trougakos I, Gonos E . Clusterin/apolipoprotein J in human aging and cancer. Int J Biochem Cell Biol 2002; 34: 1430.

  25. 25

    Petropoulou C, Trougakos IP, Kolettas E, Toussaint O, Gonos ES . Clusterin/apolipoprotein J is a novel biomarker of cellular senescence that does not affect the proliferative capacity of human diploid fibroblasts. FEBS Lett 2001; 509: 287–297.

  26. 26

    Bettuzzi S, Scorcioni F, Astancolle S et al. Clusterin (SGP-2) transient overexpression decreases proliferation rate of SV40-immortalized human prostate epithelial cells by slowing down cell cycle progression. Oncogene 2002; 21: 4328–4334.

  27. 27

    Poon S, Rybchyn MS, Easterbrook-Smith SB et al. Mildly acidic pH activates the extracellular molecular chaperone clusterin. J Biol Chem 2002; 277: 39532–39540.

  28. 28

    Lee CG . RH70, a bidirectional RNA helicase, co-purifies with U1snRNP. J Biol Chem 2002; 277: 39679–39683.

  29. 29

    Maciewicz RA, Wardale RJ, Wotton SF, Duance VC, Etherington DJ . Mode of activation of the precursor to cathepsin L: implication for matrix degradation in arthritis. Biol Chem Hoppe Seyler 1990; 371 (Suppl): 223–228.

  30. 30

    Hill PA, Buttle DJ, Jones SJ et al. Inhibition of bone resorption by selective inactivators of cysteine proteinases. J Cell Biochem 1994; 56: 118–130.

  31. 31

    Stark GR, Kerr IM, Williams BR, Silverman RH, Schreiber RD . How cells respond to interferons. Annu Rev Biochem 1998; 67: 227–264.

  32. 32

    Darnell Jr JE . STATs and gene regulation. Science 1997; 277: 1630–1635.

  33. 33

    Lemaire R, Huet G, Zerimech F et al. Selective induction of the secretion of cathepsins B and L by cytokines in synovial fibroblast-like cells. Br J Rheumatol 1997; 36: 735–743.

  34. 34

    Hu X, Herrero C, Li WP et al. Sensitization of IFN-gamma Jak-STAT signaling during macrophage activation. Nat Immunol 2002; 3: 859–866.

  35. 35

    Arnett FC, Edworthy SM, Bloch DA et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988; 31: 315–324.

  36. 36

    Boukamp P, Petrussevska RT, Breitkreutz D et al. Normal keratinization in a spontaneously immortalized aneuploid human keratinocyte cell line. J Cell Biol 1988; 106: 761–771.

  37. 37

    Eisen MB, Spellman PT, Brown PO, Botstein D . Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 1998; 95: 14863–14868.

Download references


This work was supported in part by ‘Institut National de la Santé et de la Recherche Médicale’ (Inserm), the Commissariat à l'Energie Atomique (CEA), the ‘Association de Recherche sur la Polyarthrite Rhumatoïde’ (ARP), and by Abbott laboratories.

Author information

Correspondence to G Chiocchia.

Additional information

Conflict of interest: No conflict of interest.

Supplementary Information accompanies the paper on Genes and Immunity's website (http://www.nature.com/gene).

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Devauchelle, V., Marion, S., Cagnard, N. et al. DNA microarray allows molecular profiling of rheumatoid arthritis and identification of pathophysiological targets. Genes Immun 5, 597–608 (2004). https://doi.org/10.1038/sj.gene.6364132

Download citation


  • rheumatoid arthritis
  • transcriptomics
  • diagnosis
  • microarray

Further reading