SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Objective

To generate a molecular description of synovial tissue from rheumatoid arthritis (RA) patients that would allow us to unravel novel aspects of pathogenesis and to identify different forms of disease.

Methods

We applied complementary DNA microarray analysis to profile gene expression, with a focus on immune-related genes, in affected joint tissues from RA patients and in tissues from osteoarthritis (OA) patients as a control. To validate microarray data, real-time polymerase chain reaction was performed on genes of interest.

Results

The gene expression signatures of synovial tissues from RA patients showed considerable variability, resulting in the identification of at least two molecularly distinct forms of RA tissues. One class of tissues revealed abundant expression of clusters of genes indicative of an involvement of the adaptive immune response. Detailed analysis of the expression profile provided evidence for a prominent role of an activated signal transducer and activator of transcription 1 pathway in these tissues. The expression profiles of another group of RA tissues revealed an increased tissue remodeling activity and a low inflammatory gene expression signature. The gene expression pattern in the latter tissues was reminiscent of that observed in the majority of OA tissues.

Conclusion

The differences in the gene expression profiles provide a unique perspective for distinguishing different pathogenetic RA subsets based on molecular criteria. These data reflect important aspects of molecular variation that are relevant for understanding the biologic dysregulation underlying these subsets of RA. This approach may also help to define homogeneous groups for clinical studies and evaluation of targeted therapies.

Rheumatoid arthritis (RA) is a chronic inflammatory joint disease affecting synovial tissue in multiple joints in which immune and nonimmune cellular systems mediate pathology. Despite uncertainty about its etiology, RA is thought to be an immune-mediated disease that promotes inflammation and tissue destruction. Rheumatoid synovial tissue is characterized by intimal lining layer hyperplasia and infiltration of the sublining by macrophages, plasma cells, T and B cells, and other inflammatory cells that promote inflammation and tissue destruction (1–3). However, the pathogenesis of RA is still poorly understood, and fundamental questions remain to be answered regarding the precise molecular nature and biologic significance of the inflammatory changes.

Owing to the lack of knowledge of the etiology and pathogenesis of RA, there is an awareness that current methods for classifying this disease, based as they are on a set of clinical variables supplemented with minimal laboratory evidence in the form of molecular markers, are not optimal. For RA, the classifying diagnosis is based on the presence of 4 of 7 criteria, which include 5 clinical variables supplemented with radiographic evidence for erosions and the presence of rheumatoid factor as laboratory evidence (4, 5). Moreover, the current modes of classification fall short with regard to the challenge posed by the fact that diagnosed RA is a heterogeneous disease in itself. The clinical presentation of RA may reveal striking heterogeneity, with a spectrum ranging from mild to severe disease. In addition, marked variability in the features of synovial inflammation among RA patients has been described (6–8). The wide variation in responsiveness to different modes of antirheumatic treatment is consistent with this notion (9, 10). Hence, the data suggest that distinct pathogenetic mechanisms contribute to disease in RA.

A powerful way to gain insight into the molecular complexity and pathogenesis of arthritides has arisen from complementary DNA (cDNA) microarray technology (11–17), which provides the opportunity to determine differences in gene expression of a large portion of the genome in search of genes that are differently expressed between clinically diagnosed arthritides (18). Therefore, we used microarrays containing ∼18,000 cDNA representing predominantly genes thought to be of relevance in immunology.

In the present study, we have conducted a systematic characterization of gene expression in synovial tissues from affected joints of patients with RA and compared this expression with that in tissues from patients with osteoarthritis (OA), a degenerative joint disease characterized by progressive loss of cartilage, as a control (19–22). Both diseases are complex clinical entities that share clinical and demographic characteristics, but they also harbor key differences in tissue destruction and prognosis. Although the etiology of OA is also unknown, it is believed that, in contrast to RA, biomechanical insults and cartilage abnormalities are primary events in the development of OA, with minor subsequent activation of the immune response (23). In our interpretation, we focused particularly on RA to determine whether gene expression profiling could subdivide clinically diagnosed RA on the basis of molecular criteria. In addition, by this approach novel insights into the biologic dysregulation underlying RA might be found.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Patients and synovial tissues.

Thirty patients (21 with RA and 9 with OA) undergoing total joint replacement surgery were included in the gene expression profiling study. Table 1 summarizes the characteristics of these patients. The RA patients met the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 criteria for RA (4, 5), and the OA patients had primary OA of the joint, based on the ACR criteria (24, 25). Synovial tissue biopsy samples, obtained from an independent group of 10 OA patients and 9 RA patients with active arthritis of the knee joint, were selected from various regions by arthroscopy (2.7-mm arthroscope; Storz, Tuttlingen, Germany) under local anesthesia (26) for immunochemistry analysis. The samples were snap-frozen en bloc in Tissue-Tek OCT (Miles, Elkhart, IN). The frozen blocks were stored in liquid nitrogen. Cryostat sections (5 μm) were mounted on glass slides (Star Frost adhesive slides; Knittelgläser, Braunschweig, Germany). The glass slides were sealed and stored at −70°C until immunohistologic analysis.

Table 1. Demographic and clinical data of the patients with rheumatoid arthritis (RA) and osteoarthritis (OA) whose synovial tissue was studied for gene expression profiling*
 RAtotal (n = 21)RAlow (n = 8)RAinterm (n = 4)RAhigh (n = 9)OA (n = 9)
  • *

    Except where indicated otherwise, values are the number (%) of patients. Three RA subgroups were defined according to their gene expression profiles: a high-inflammation subgroup (RAhigh), an intermediate-inflammation subgroup (RAinterm), and a low-inflammation subgroup (RAlow). ESR = erythrocyte sedimentation rate; DMARDs = disease-modifying antirheumatic drugs.

  • P = 0.008 versus RAlow and RAinterm groups, by t-test.

Age, mean years (range)60 (22–85)69 (56–85)50 (22–77)63 (24–77)64 (52–85)
No. of women/no. of men17/45/34/08/19/0
Laboratory variables     
 ESR, mean mm/hour (range)34 (7–66)27.5 (9–57)19.5 (7–41)47.8 (20–66)25 (9–53)
 Leukocytes × 106/ml, mean (range)7.7 (5.1–16.8)9 (6.2–16.8)7.9 (5.1–10.0)6.4 (4.5–11.5)5.7 (3.4–10.1)
 Erosive disease18 (86)6 (75)4 (100)8 (89)0 (0)
 Rheumatoid factor positive17 (81)7 (88)2 (50)8 (89)0 (0)
Medication     
 DMARDs19 (90)7 (88)4 (100)8 (89)1 (11)
 Prednisone3 (14)2 (25)0 (0)1 (11)1 (11)

Messenger RNA (mRNA) isolation from tissue samples and labeling.

After surgical resection, the synovial tissue (∼1 gm) was dissected and quickly frozen in liquid nitrogen and stored at −80°C. Total cellular RNA and mRNA were isolated from the tissues by TRIzol reagent and the FAST TRACK 2.0 kit (both from Invitrogen, Carlsbad, CA), respectively, according to the manufacturer's instructions. For gene expression profiling by DNA microarray analysis, fluorescent cDNA probes were prepared from a 1-μg experimental mRNA sample by oligo(dT)-primed polymerization using Superscript II reverse transcriptase in the presence of Cy5-labeled dCTP as described (online at http://cmgm.stanford.edu/pbrown/protocols.html). A common reference mRNA sample that consists of a mixture of mRNA isolated from 11 different cell lines (15), synovial tissue, fibroblasts, and activated peripheral blood mononuclear cells was labeled with Cy3.

Microarray procedures.

DNA microarray analysis was done essentially as described by Eisen and Brown (27). The Cy5-labeled experimental cDNA and the Cy3-labeled common reference sample were pooled and hybridized to the Lymphochips containing ∼18,000 cDNA spots representing genes of relevance in immunology as described previously (14). One OA sample was hybridized twice.

Data analysis.

Analysis of microarray gene expression data was performed as described (14), except that we used genes with an expression level higher than 1.6 times background instead of 1.4 times background. The use of a common reference sample allows the comparison of the relative expression levels across the tissue samples (14). Hierarchical clustering (28) (online at http://rana.lbl.gov/EisenSoftware.htm) was applied to the gene axis as well as the tissue axis. The results were visualized with Treeview (online at http://rana.lbl.gov/EisenSoftware.htm). Full data can be viewed at the Stanford Microarray Database (29) (online at http://genome-www.stanford.edu/microarray).

Immunohistochemistry and microscopic analysis.

Stainings with monoclonal antibodies directed against signal transducer and activator of transcription 1 (STAT-1) (clone 1 cat. no. 610115; BD Transduction Laboratories, San Jose, CA) were performed as described previously (30). Coded sections stained for STAT-1 were analyzed in a random order by computer-assisted image analysis. Eighteen high-power fields (hpf) were analyzed, representing an area of 2.1 mm2. The hpf images were analyzed for total cell counts and integrated optical density using a specialized algorithm (Syndia version 1.1) written in Quips for the Qwin analysis system (Leica, Cambridge, UK), as described previously in detail (31).

Statistical analysis.

Statistical analysis on microarray data was performed using the Significance Analysis of Microarrays (SAM) method (32) (online at http://www-stat.stanford.edu/∼tibs/SAM). Pearson's correlation coefficient was used to determine the degree of correlation between clusters (expression ratios of all genes within a cluster were averaged per tissue). A t-test was used to test for differences in patient characteristics between the patient groups and for differences in STAT-1 protein expression levels between RA and OA patients.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Gene expression pattern and disease.

We focused most of our analyses on a set of 1,066 cDNA, including different cDNA representing the same genes, whose transcripts varied in abundance by at least 2-fold from their median level in at least 4 clinical samples. Hierarchical clustering revealed a remarkable ordered variation in gene expression patterns in RA and OA tissues with clusters of genes having similar expression patterns. To demonstrate that experimental noise or artifact was negligible, distinct clones representing the same genes were typically invariably clustered in adjacent rows. Without information about the identity of the samples, the tissues were also organized on the basis of overall similarity in their gene expression patterns (Figure 1). Correction of the weighting factor for distinct clones representing the same gene spotted multifold did not affect the tissue clustering. The structure of the hierarchical dendrogram indicated that the gene expression patterns in RA tissues were considerably heterogeneous. The algorithm divided the RA tissues into two groups. One group contained almost exclusively RA tissues that clustered together with one outlying OA tissue. Within this group, two subbranches, labeled green and red, were evident. The remaining RA tissues clustered together with the OA tissues in a group labeled blue.

thumbnail image

Figure 1. Cluster diagram of the expression of 1,066 complementary DNAs (cDNA) in 30 experimental samples using the Lymphochips. Log-transformed data (base 2) are represented in a matrix format wherein each row displays expression results for a single gene across the arrays and each column shows the relative expression levels for all the genes in each tissue (28). Red represents relative expression greater than the median expression level across all tissues, and green represents an expression level lower than the median expression level. The color intensity represents the magnitude of the deviation from the median. Gray indicates missing or excluded data. Left panel shows a thumbnail representation of the hierarchical clustering of the selected 1,066 cDNA. Colored bars labeled A–F to the right of this panel identify the locations of a category of clustered genes with a related expression profile. Complementary DNAs are grouped on the basis of similarity in their relative expression across the different tissues (A = T/B cell cluster; B = antigen-presenting cell cluster; C = transcription/signaling factors cluster; D–F = stromal cell–related gene clusters). The dendrogram at the top lists the samples studied and provides a measure of the relatedness of the expression profile in each sample. The branches of the dendrogram are color-coded by category of tissue sample (see Results for description of tissue groups). Right panel shows an expanded view of cluster A, with genes listed that are characteristic for the defined gene cluster. Genes without designations are new genes of unknown function derived from various lymphoid cDNA libraries. RA = rheumatoid arthritis; OA = osteoarthritis.

Download figure to PowerPoint

The position of any tissue in the dendrogram is determined by distinct biologic themes represented by clusters of coordinately expressed genes. Therefore, the molecular signature provides the basis for a biologic interpretation. The most prominent distinction between the two dominant tissue clusters was formed by genes represented in cluster A, which contained a collection of genes that are expressed by T and B cells. This cluster revealed a difference in expression of genes involved in an adaptive inflammatory response. The elevated expression of immunoglobulin genes present in this T/B cell cluster indicates a high biologic activity of B cells (Figure 1). Moreover, this cluster showed an increased expression of matrix metalloproteinases (MMPs) 1 and 3, STAT-encoding genes, and STAT-induced genes.

Increased expression of genes present in cluster A (the T/B cell cluster), together with genes in cluster B that were characteristic for antigen-presenting cells (APCs), was typical of the red-labeled subbranch, which contained exclusively RA tissues. Cluster B (Figure 2), the APC cluster, contained many genes that encode HLA class I, HLA class II, and associated molecules. In the APC cluster, specific markers were found for several immune cell types. Thus, the expression signature of the RA tissues from the red subbranch indicated an influx of inflammatory cells, which, in addition to B and T cells, provided evidence for the presence of monocytes and granulocytes.

thumbnail image

Figure 2. Scaled-down representation of distinct gene expression cluster B defined by hierarchical clustering. The diagram presents an expanded view of the gene expression cluster B shown in Figure 1. The genes on the right are known genes that are characteristic for the defined gene cluster. See Figure 1 for definitions and explanations.

Download figure to PowerPoint

A remarkable expression pattern was featured by the C cluster, which contained genes whose products are involved in intracellular signaling and transcriptional regulation (Figure 3). Genes in this cluster were highly expressed in those tissues showing high expression of genes in the APC (B) cluster, as well as in a subgroup of tissues from the blue-labeled group.

thumbnail image

Figure 3. Scaled-down representation of distinct gene expression clusters C–F defined by hierarchical clustering. The diagram presents an expanded view of the gene expression clusters C–F shown in Figure 1. The genes on the right are known genes that are characteristic for the defined gene clusters. See Figure 1 for definitions and explanations.

Download figure to PowerPoint

Another gene expression profile that distinguished between the tissue groups included genes that were grouped together in the D, E, and F gene clusters (Figure 3), which contained genes involved in stromal cell differentiation. The expression of these genes was relatively low in the red-labeled tissues. The D cluster included the genes for MMP-11, MMP-13, osteonectin (secreted protein, acidic and rich in cysteine), and cyclin D1, which has a role in cell turnover. In addition to this cluster, two other gene clusters with stromal cell–related genes were identified. One of these was the E gene cluster (involving genes for protocadherin-γ, secreted frizzled-related protein 1, tissue inhibitor of metalloproteinases 2, membrane metallo-endopeptidase, and CD36 antigen [type I collagen/thrombospondin receptor]). This cluster also contained genes that define an activated transforming growth factor β3 (TGFβ3) pathway, such as TGFβ3, TGFβ receptor type III, and TGFβ-stimulated proteins TSC-22 and Id3. Genes grouped in the F cluster form a diverse set of genes involved in miscellaneous biologic processes. The abundant expression of the gene clusters D, E, and F, indicative of extracellular matrix (ECM) remodeling and cell turnover, was inversely correlated with a low infiltrate gene expression signature, in particular, expression of the T/B cell (A) gene cluster (r = −0.74, P < 0.0001).

Significance analysis of RA subtypes.

Clearly, the gene signatures indicated the existence of subtypes of RA tissues. To confirm this, we reclustered the RA tissues without including OA tissues (Figure 4). Indeed, the hierarchical dendrogram revealed a subdivision of RA tissues almost identical to that seen in Figure 1, based on a considerable difference in their gene expression profiles, including having the T/B cell and APC gene clusters on one side and the stromal cell gene clusters on the other side. Of note, we could identify 3 subgroups according to their gene expression profiles: a high-inflammation subgroup (RAhigh, red branch; n = 9), an intermediate-inflammation subgroup (RAinterm, green branch; n = 4), and a low-inflammation subgroup (RAlow, blue branch; n = 8).

thumbnail image

Figure 4. Discovery of rheumatoid arthritis (RA) subtypes by gene expression profiling. Shown is hierarchical clustering of RA tissues only. Three RA subgroups were defined according to their gene expression profiles: a high-inflammation subgroup (RAhigh, red branch), an intermediate-inflammation subgroup (RAinterm, green branch), and a low-inflammation subgroup (RAlow, blue branch). T/B = T/B cell cluster; APC = antigen-presenting cell cluster; TFs = transcription/signaling factors cluster; stromal = stromal cell–related gene clusters. See Results for description of tissue groups. See Figure 1 for explanations.

Download figure to PowerPoint

Based on this subclassification of RA patients, we next tested which of the genes differed significantly in expression between the tissues from the RAhigh group and those from the RAlow group, by applying the SAM method (32). The comparison of the two groups identified 150 independent cDNA sequences (72 unknown cDNA) with a statistically significant difference in expression of at least 2-fold in at least one copy of the gene when multiple copies of the same gene were spotted (Table 2). Based on the same criteria, we also tested which of the genes differed significantly in expression between the tissues from the OA group and those from the RAhigh group and identified 197 independent sequences (including 85 unknowns). Remarkably, there was a large overlap in genes that differed significantly in expression between the RAhigh and RAlow tissue groups and between the RAhigh and OA tissue groups.

Table 2. Comparison of statistical differences in gene expression between the RAhigh and the RAlow subgroups and between the RAhigh subgroup and the OA group*
 RAhigh vs. RAlowRAhigh vs. OA
Fold difference, mean ± SEMq, meanFold difference, mean ± SEMq, mean
  • *

    Statistical analysis on microarray data was performed using the Significance Analysis of Microarrays method (online at http://wwwstat.stanford.edu/∼tibs/SAM). The q value (see ref. 48) is the lowest false discovery rate at which a gene is called significant. It is similar to the familiar P value, adapted to the analysis of a large number of genes. A q value of 5 for a particular gene means that at a false discovery rate of 5%, this gene will still be listed as significant. In this table, mean q values <5 (in boldface) are considered significant. See Table 1 for description of rheumatoid arthritis (RA) tissue subgroups and for definitions. OA = osteoarthritis.

  • If a gene was spotted more than once, the mean ± SEM values of the fold change of distinct complementary DNAs representing the same gene were calculated and presented, including their geometric mean q values.

B cells/T cells    
 CD79A = BCR α chain3.24 ± 0.160.333.22 ± 0.870.20
 CD752.600.332.890.15
 NG725.730.3312.250.15
 CD272.21 ± 0.160.331.86 ± 0.190.26
 SLP65 = BLNK, B cell linker2.200.331.820.36
 Immunoglobulins (n = 39)17.44 ± 1.830.3512.49 ± 2.000.18
 CD214.52 ± 3.021.363.52 ± 2.020.22
 PBEF, pre–B cell colony-enhancing factor1.83 ± 0.373.631.32 ± 0.2315.97
 CD95.17 ± 4.233.655.07 ± 4.351.62
 CD202.00 ± 0.083.723.48 ± 0.830.26
 CD691.45 ± 0.314.931.23 ± 0.256.64
 NK41.49 ± 0.094.942.81 ± 0.170.15
 CD3 epsilon1.714.952.020.36
 CD79B = BCR β chain1.93 ± 0.225.211.94 ± 0.202.01
 T cell receptor gamma chain1.39 ± 0.3711.932.080.15
 CD371.00 ± 0.1419.881.67 ± 0.191.10
 T cell receptor beta chain1.18 ± 0.1526.841.35 ± 0.313.59
 CD480.99 ± 0.1240.241.84 ± 0.602.78
Proteases    
 Caspase-15.75 ± 3.970.593.16 ± 1.241.41
 MMP15.111.0613.120.15
 MMP35.501.7917.160.15
 Cathepsin L1.84 ± 0.154.511.91 ± 0.091.18
 Disintegrin protease1.807.782.001.63
Antigen presentation    
 PSMB92.26 ± 0.080.332.23 ± 0.080.15
 CD141.900.952.330.15
 CD741.76 ± 0.181.642.65 ± 0.400.26
 TAP-1 = peptide transporter1.901.792.660.15
 HLA class II (n = 37, 51)1.75 ± 0.052.522.71 ± 0.130.21
 B2M, beta-2-microglobulin1.76 ± 0.342.831.81 ± 0.471.36
 HLA class I (n = 12, 14)1.59 ± 0.093.912.08 ± 0.150.30
 Similar to TAP2E2.234.461.964.46
Chemokines/receptors    
 BLC = BCA-13.300.336.710.15
 FLMP receptor1.741.062.210.15
 SDF11.901.571.1438.68
 CXCR41.58 ± 0.152.262.75 ± 0.080.15
 IP-101.88 ± 0.732.782.12 ± 1.242.58
 RANTES1.39 ± 0.1311.381.54 ± 0.413.74
 CCR51.4013.372.150.15
 CCR11.2713.371.481.91
Oxidative stress    
 NCF11.97 ± 0.292.662.13 ± 0.341.19
 CYBB, cytochrome b-245 beta1.72 ± 0.317.742.01 ± 0.481.30
 HSPC022 protein1.3021.862.35 ± 0.260.15
Transcription factors    
 AA8056834.60 ± 0.060.333.38 ± 0.030.15
 c-fos3.65 ± 0.610.471.68 ± 0.283.99
 RAR-alpha-11.93 ± 0.490.842.090.15
 IRF-11.80 ± 0.221.141.81 ± 0.211.05
 STAT-12.77 ± 1.131.542.80 ± 0.540.43
 JunB1.91 ± 0.321.851.49 ± 0.405.38
 Staf501.77 ± 0.232.981.83 ± 0.180.99
 ICSBP11.44 ± 0.1012.162.38 ± 0.010.15
 SREBF21.38 ± 0.1213.001.60 ± 0.363.07
 GMEB21.3018.112.010.28
Signal transduction    
 TEK tyrosine kinase14.530.3310.330.15
 PPP1R2, protein phosphatase inhibitor 22.500.331.990.15
 FYB, FYN-binding protein1.65 ± 0.432.551.84 ± 0.732.30
 GSbeta3.76 ± 2.713.824.10 ± 3.022.55
 HCLS11.66 ± 0.203.991.80 ± 0.210.84
 RKAG16.67 ± 5.676.8511.300.15
 SIP-1103.03 ± 1.907.702.01 ± 0.652.85
G-protein signaling    
 GBP11.75 ± 0.111.202.07 ± 0.110.19
 ARL71.64 ± 0.452.832.12 ± 0.190.15
 DOCK21.58 ± 0.024.701.96 ± 0.080.15
 RHO GDI 21.41 ± 0.154.761.70 ± 0.230.66
 RGS11.48 ± 0.146.451.94 ± 0.190.23
 PTPN61.43 ± 0.1212.452.11 ± 0.200.60
Adhesion    
 CD502.18 ± 0.381.102.07 ± 0.010.23
 PSCD11.38 ± 0.147.581.54 ± 0.142.59
 PTPRK1.689.342.052.71
 L-selectin1.6113.372.050.77
 CD181.01 ± 0.0825.602.12 ± 0.290.39
 CD1031.12 ± 0.1325.871.79 ± 0.251.91
Cytokines/receptors    
 IL-6Rβ1.98 ± 0.240.331.33 ± 0.225.72
 IL-151.560.531.32 ± 0.125.28
 CSF-1R2.012.121.755.78
 IL10Rα1.70 ± 0.122.802.00 ± 0.300.38
 GM-CSFRα1.58 ± 0.124.442.02 ± 0.320.54
 IL-2Rγ1.48 ± 0.147.771.76 ± 0.251.59
 IL-6Rα1.1131.9111.43 ± 10.130.83
Cell proliferation    
 GAPCenA3.43 ± 0.090.331.89 ± 0.230.71
 BTG22.46 ± 0.390.761.77 ± 0.160.55
 ISG201.63 ± 0.152.801.57 ± 0.091.25
Cell surface molecules    
 FcϵRI-alpha2.340.332.500.15
 CD452.03 ± 0.220.953.07 ± 0.480.23
 CD531.67 ± 0.213.761.83 ± 0.471.53
 CD321.46 ± 0.176.601.85 ± 0.230.85
Miscellaneous    
 BCKDHA2.220.332.100.15
 Centaurin beta 23.520.332.490.15
 Bone morphogenetic protein 62.300.331.413.87
 Similar to TBC12.280.331.980.28
 SEL1L1.99–0.150.331.75 ± 0.110.34
 FGFR2, fibroblast growth factor receptor 22.090.331.681.63
 MCL1, myeloid cell leukemia sequence 12.00 ± 0.110.681.55 ± 0.152.22
 RIZ2.88 ± 1.320.773.19 ± 1.880.88
HNRPH1, heterogeneous nuclear ribonucleoprotein H11.81 ± 0.260.971.74 ± 0.220.35
 Apolipoprotein L32.13 ± 0.071.311.94 ± 0.070.87
 HSP701.81 ± 0.251.621.81 ± 0.320.83
 AIM21.81 ± 0.261.762.19 ± 0.801.02
 WASPIP1.73 ± 0.242.221.84 ± 0.260.53
 MDS019 phorbolin-like protein1.73 ± 0.142.272.12 ± 0.220.42
 CHI3L2, chitinase 3-like 23.41 ± 0.292.751.53 ± 0.1742.74
 ALOX51.68 ± 0.193.312.42 ± 0.040.15
 LLT1 C-type lectin2.773.661.8813.95
 EVI2B1.80 ± 0.204.262.39 ± 0.160.15
 LCP11.28 ± 0.047.042.18 ± 0.200.19
 PTB-42.18 ± 1.4312.344.37 ± 3.903.06
 CORO1A1.50 ± 0.2512.701.84 ± 0.332.03
 MFNG1.34 ± 0.3014.401.73 ± 0.513.08
 Gamma-parvin1.0614.491.61 ± 0.212.77
 SRM30001.29 ± 0.0216.901.85 ± 0.210.83
 BIN21.39 ± 0.0216.931.90 ± 0.320.50

A striking feature in the list of genes with significantly increased expression in the RAhigh tissues was the presence of the STAT-1 gene. In addition, a number of genes that fit in the STAT-1 activation pathway, including potential STAT-1–inducing receptors (e.g., IL-2Rγ, BCR, CCR5) and STAT-1 target genes (e.g., STAT-1 itself, MMPs, GBP1, ICSBP, IP-10, caspase-1, TAP-1, and IRF-1), were differentially expressed, which can be viewed as one of the hallmarks of the expression pattern of highly expressed genes in the inflammatory rheumatoid synovium. The expression levels of STAT-1 in the RA patients were confirmed by real-time polymerase chain reaction (PCR). With this technique, the difference in STAT-1 mRNA expression between the RAhigh and RAlow groups was again seen (3.5-fold difference; P = 0.02). No significant differences in gene expression between the RA low and OA tissues were revealed by the SAM method using the Lymphochip.

In an independent study using synovial biopsy material from patients with OA and RA, we could confirm the differential expression of STAT-1 at the protein level (Figure 5). This analysis revealed a higher number of STAT-1–positive cells in the tissues of RA patients (n = 9) than in the tissues of OA patients (n = 10) (P = 0.018 by t-test plus Welch correction) (Figure 5B). The STAT-1 intensity per positive cell increased with increasing numbers of STAT-1–positive cells (r = 0.52, P = 0.02) (data not shown). Positive cells were predominantly found in the lymphoid aggregates and intimal lining layer (Figure 5A). In analogy to the molecular profiling data, we observed remarkable variability in STAT-1 expression in the rheumatoid synovial biopsy specimens tested (Figure 5B).

thumbnail image

Figure 5. Detection of signal transducer and activator of transcription 1 (STAT-1) in synovial tissue sections derived from patients with rheumatoid arthritis (RA) and osteoarthritis (OA). A, Immunohistochemical staining of STAT-1 in RA and OA synovial tissue sections. L = intimal lining layer; A = lymphoid aggregates (original magnification × 200). B, Increased STAT-1 expression in RA compared with OA synovial biopsy specimens. Bars show the mean percentage of STAT-1–positive cells per total number of nuclei. In A, the RA synovial tissue section was obtained from a patient with 77.6% STAT-1–positive cells, while the OA synovial tissue section was obtained from a patient with 10.9% STAT-1–positive cells.

Download figure to PowerPoint

Relationship between molecular profiles and disease parameters.

The question then arose as to whether the molecular definition of RA tissue subgroups was associated with clinical differences between RA patients. To address this possibility, the relationship with clinical and demographic parameters was determined (Table 1). No differences in clinical parameters were revealed. For erythrocyte sedimentation rates (ESRs), a statistically significant difference was observed between patients whose tissues featured a fulminant inflammation signature (the RAhigh group) and those whose tissues clustered in the RAlow and RAinterm groups (P = 0.008 by t-test). Thus, the molecular dissection of RA tissues from joints that are subject to erosive disease apparently identifies differences in systemic parameters.

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

OA and RA are related complex clinical entities resulting from the interaction of multiple gene products. The objective of this study was to generate a molecular description of these diseases focused on immune-related genes that would allow us to differentiate these arthritides, subclassify RA as a heterogeneous disease, unravel novel aspects of RA pathogenesis, and identify genes and biologic processes not hitherto associated with this disease. We identified molecularly distinct forms of RA synovium, which had expression patterns that suggested the occurrence of different pathologic mechanisms. One RA type expressed genes characteristic of ongoing inflammation, while another type expressed genes indicative of an increased tissue remodeling activity reminiscent of the profile seen in OA tissues.

Previously, Zanders et al (17) applied high-density cDNA arrays printed on nylon membranes. Their analysis revealed an overall increased expression of inflammation-related genes in rheumatoid tissue compared with normal synovium. Since those investigators performed an analysis on pooled RA synovial tissue compared with pooled tissue from healthy controls, their strategy excluded an evaluation of disease heterogeneity and might therefore have diluted differences that existed between tissues. We chose to profile gene expression in whole synovial tissue, which comprises heterogeneous cell types, with the specific purpose of gaining a global representative insight into the molecular changes associated with OA and RA (19, 33, 34).

Gene expression profiling of synovial membranes of RA and OA patients revealed a spectrum of expression that ranged from an extensive inflammatory expression signature, indicative of massive infiltration of mononuclear cells, to an ECM remodeling activity, indicative of fibrosis that is accompanied by a scarce cellular infiltrate. What is also clear from the cluster analysis is that all but one of the OA tissues formed a homogeneous group with a gene expression profile that reflected the fibrosis arm of the spectrum. These data are in accordance with data from histologic studies that suggested that the amount of fibrosis was inversely proportional to the extent of the cellular infiltrate, and that fibrosis was mainly encountered in OA tissue (22). In the present study, we have identified a class of genes grouped in the stromal cell gene clusters that are likely to be implicated in tissue fibrosis in OA.

The RA synovia showed considerable heterogeneity in gene expression, indicating the existence of molecularly distinct classes of rheumatoid synovium. The T/B cell and APC signatures were prominent features in the RAhigh tissues. These features correlate with increased expression of T and B cell genes, genes involved in inflammation, and high expression of major histocompatibility complex (MHC) class I and class II genes as hallmarks of the activation of the immune system (33, 35–37).

Although it is not a priori evident whether measurements of gene expression levels either reflect genuine gene activity or are representative of the cellular composition, it is most likely that these discrete, cell-specific gene expression patterns at least indicated the abundant presence of immune cells such as T and B cells in the RAhigh tissues. The presence of such characteristic cell-specific gene clusters correlates with reported data on infiltration of mononuclear cells into the rheumatoid synovium (6, 19, 33, 34). Overexpression of these gene clusters is consistent with increased ESRs. Another class (RAlow) showed high expression of stromal cell genes that is accompanied by a profile of a scarce cellular infiltrate. An intermediate class (RAinterm) displayed a gene expression profile that included both genetic profiles. These results indicate the existence of extreme differences in cellular infiltration between RA synovial tissues. The combined analysis of cellular complexity, together with a comprehensive overview of the concomitant gene expression profile, provides opportunities for further research.

The expression data suggest involvement of two distinct disease processes in rheumatoid pathogenesis, which is in accordance with data from several studies that indicated biologic heterogeneity in RA (6, 19, 33, 34). The existence of a spectrum of molecular variation that may be translated into distinct pathophysiologic mechanisms at the site of the lesion would fit a model proposed by Firestein and Zvaifler (3), who suggested two processes in the destruction stage of RA. One is a T cell–mediated process that might progress to a T cell–independent process that is centered on autonomous fibroblast-like synoviocyte (FLS) aggression. This model is further supported by data from several animal models in which FLS acquire a degree of independence from T cell control in late destructive disease, implying that an autonomous role of stromal cell elements is responsible for tissue destruction. Hence, both the T cell–involved and the autonomous stromal cell form of disease might drive destruction of bone and cartilage.

Clearly, the molecular dissection of the rheumatoid synovium allows for a biologic interpretation of processes that take place in the tissues. Among the genes that were significantly increased in the RAhigh tissue group were genes indicative of an activated STAT-1 pathway (i.e., STAT-1 itself and a number of genes that are dependent on STAT-1 for their transcription). The induction of these genes is known to occur via the activation of janus-activated kinases, resulting in the phosphorylation and subsequent translocation of specific STAT proteins to the nucleus (38), where it directly activates transcription of target genes including STAT-1 itself (38, 39). We confirmed the differential STAT-1 expression by real-time PCR. Moreover, heterogeneity in STAT-1 protein expression was also found upon immunohistologic analysis of specimens from RA patients with active disease. These findings are in accordance with recently reported data on increased STAT-1 protein expression in RA synovial tissues obtained during joint replacement surgery, compared with normal tissues (40).

Gene expression studies on macrophages and fibroblasts from wild-type and STAT-1–deficient mice revealed ∼66 genes that were part of the interferon-γ (IFNγ)–induced STAT-1–dependent genetic profile (41, 42). A number of the genes from our analysis that met our filtering criteria corresponded to the STAT-1–dependent genes mentioned in those reports. This number will probably increase when purified cells are used as the source for gene expression profiling.

Activation of STAT-1 is involved in MHC class I–restricted antigen presentation through up-regulation of TAP-1, and it indirectly up-regulates MHC class II genes (Table 2). Other genes induced by STAT-1, such as MMPs, caspase-1, IP-10, IRF-1, GBP1, and ICSBP, are also selectively up-regulated in RAhigh tissue (Table 2). Although not definitely proven, this suggests that STAT-1 may be responsible for the activation of these genes.

The STAT-1–dependent chemokine IP-10 attracts T cells and monocytes, which suggests that STAT-1 activation may also be partly responsible for the inflammatory cell influx. The expression of STAT-1 coincided with increased expression of genes encoding receptors activating STAT pathways (e.g., IL-6Rβ, IL-2Rγ, CCR5, and BCR). The data indicated that in addition to the increase in the number of STAT-1–positive cells, the intensity per cell was increased. Hence, conditions in the inflamed joint contribute to an increase per cell. A likely explanation would be that increased STAT-1 expression is a consequence of increased cytokine activity in the synovium. STAT-1 can be activated by a number of cytokines, including the type I and type II IFNs, interleukin-6 (IL-6), IL-9, IL-11, oncostatin M, and leukemia inhibitory factor, and by the chemokines RANTES and macrophage inflammatory protein 1α (43, 44). Although the prototype STAT-1–inducing cytokine IFNγ is barely detectable in the rheumatoid synovium, low doses of IFNγ are able to sensitize the STAT-1 activation pathway, which has been shown to yield increased STAT-1 activity upon subsequent activation (40).

Obviously, the various cytokines present in the RA synovium create a complex situation with simultaneous activation of multiple signaling pathways that may influence STAT-1 signaling (e.g., STAT-1 activation may negatively influence the tumor necrosis factor α and IFNα/β signaling pathways [for review, see ref. 45]). The importance of STAT activation in arthritis has been demonstrated in an animal model; periarticular administration of adenoviral suppressor of cytokine signaling 3 dramatically reduced the severity of collagen-induced arthritis and synovial IgG production (46). These findings justify further research on the cell-specific expression of STAT signaling components, including the activating receptors and their ligands, which is crucial for our understanding of the molecular and cellular events that take place in the effector phase.

The abundant expression of specific chemokines and their receptors revives data from studies proposing that cell-specific expression of selective chemokines and their cognate receptors is involved in the accumulation of mononuclear cells in the inflamed synovium (47). In the RAhigh tissues, we found significantly higher expression levels of the chemokine RANTES (and its receptors CCR1 and CCR5), which activates STAT-1. The higher expression of stromal cell–derived factor 1 plus its ligand CXC receptor 4 may account for the attraction and retention of T cells, B cells, and monocyte/macrophages, while the expression of B lymphocyte chemoattractant explains the increased B cell activity in the RAhigh synovial tissue with greatly enhanced production of immunoglobulins.

What does the molecular classification of RA patients into the subgroups RAhigh and RAlow mean in clinical terms? Based on the expression profiles, it suggests that different pathologic mechanisms may contribute to disease. However, we realize that the design of this study does not allow any firm conclusions to be drawn concerning the clinical parameters associated with the molecular phenotype. Further studies, which are necessary to address this issue, may provide a means to dissect and analyze the rheumatoid synovium of these patients and perform a thorough clinical association study based on molecular variation among patients. Moreover, since the molecular differences most likely reflect distinct pathophysiologic processes underlying disease, it is tempting to speculate that these differences predict individual responsiveness to treatment. Hence, the molecular stratification of patients may be helpful in assembling homogeneous populations of patients, which will improve the likelihood of observing efficacy in clinical trials.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We are grateful to Drs. Pat Brown and David Botstein, in whose laboratories most of the work described in this report was performed. We thank Dr. L. M. Staudt (National Cancer Institute, National Institutes of Health, Bethesda, MD) for providing cDNA that were printed on the Lymphochips.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES