Genome-Wide Comparison between IL-17A- and IL-17F-Induced Effects in Human Rheumatoid Arthritis Synoviocytes¹

Rheumatoid arthritis (RA)³ is a complex chronic disorder leading to joint destruction and characterized by synovium hyperplasia, neoangiogenesis, and local infiltration by immune cells. Among the inflammatory mediators expressed within RA synovium and synovial fluid, IL-17A is associated with disease severity (1). This proinflammatory cytokine is produced almost exclusively by Th17 cells. Mouse models have suggested a critical role of IL-17A in RA since IL-17A deficiency or antagonism has profound antiarthritic effects (2, 3). Data coming from human in vitro experiments demonstrated their effects on joint degradation through the induction of RANKL, metalloproteinases (4, 5, 6), and neutrophil survival, in part through GM-CSF (7).

Results on IL-17F, another member of the IL-17 family, have suggested that the genes encoding for IL-17A and IL-17F are coming from a mechanism of duplication since they are located adjacent. Moreover, both cytokines exhibit a similar expression pattern, driven by the nuclear receptor RORγt (8, 9, 10), and have 50% identity (11). They also exhibit a similar cysteine knot configuration (11), acting both as homodimer but also as IL-17A-IL-17F heterodimer (12, 13, 14), and they induce common transducing pathways through a complex composed of the IL-17RA and IL-17RC receptors (15). Thus, the common features shared by IL-17A and IL-17F suggest that they may have a similar regulatory effect. However, it is still unknown whether IL-17A and IL-17F induce similar, overlapping, or divergent gene expression profiles.

To investigate their possible implication in RA, the expression of IL-17A and IL-17F was studied in osteoarthritis (OA) and RA synovium and their regulatory effects were examined on cultured RA synoviocytes at the mRNA and protein levels. Since cooperative effects with other proinflammatory cytokines are common, the effects of IL-17A and IL-17F were compared when used alone or in the presence of TNF-α. High throughput cDNA microarrays were used to obtain an extensive evaluation of these interactions, focusing on the genes synergistically induced by IL-17A or IL-17F and TNF-α.

RA synoviocytes were obtained from synovial tissue obtained from RA patients undergoing joint surgery who fulfilled the American College of Rheumatology’s criteria for RA (16). In brief, synovial tissues were minced into small pieces and then incubated for 2h at 37°C with proteolytic enzymes (1 mg/ml collagenase and hyaluronidase; Sigma-Aldrich). Synoviocytes were cultured in DMEM (Invitrogen) supplemented with 10% FCS, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin at 37°C in a humidified 5% CO₂ incubator and used at passages three to seven, which were >99% negative for CD45, CD1, CD3, CD19, CD14, and HLA-DR and were positive for the expression of CD44 (Abs obtained from BD Pharmingen). The effect of IL-17A and IL-17F alone or in combination with TNF-α was compared both at the protein and mRNA levels. Synoviocytes seeded in 96-well plates (1 × 10⁴ cells/well) were stimulated for 48 h with IL-17A or IL-17F (0.1–100 ng/ml) alone or in combination with TNF-α (0.5 ng/ml) to examine protein secretion by ELISA. Synoviocytes seeded in 6-well plates (5 × 10⁵ cells/well) were stimulated for 1–12 h with IL-17A or IL-17F (0.5 ng/ml) alone or in combination with TNF-α (0.5 ng/ml) to examine target gene mRNA expression by real-time RT-PCR. For microarray analysis, the cells were stimulated for 12 h.

Human recombinant TNF-α was purchased from Sigma-Aldrich. Human recombinants IL-17A and IL-17F were purchased from R&D Systems. CCL20/MIP-3α and IL-6 levels were quantified in removed supernatants by ELISA (R&D Systems and eBioscience, respectively). Human anti-IL-17A mAb (clone 41809) and anti-IL-17F mAb (clone 197315) were purchased from R&D Systems. Monoclonal anti-human TNFRII/TNFRSF1B Ab (clone 22210) used for neutralizing assays and PE-conjugated mouse anti-human TNFRII mAb (clone 22235) and PE-conjugated mouse IgG2A monoclonal Ig isotype control were purchased from R&D Systems.

OA and RA synovial tissues were fixed in 10% phosphate-buffered formaldehyde. After paraffin embedding, samples were cut into 4-μm sections, mounted on glass slides, and treated for deparaffination (OTTIX Plus; DiaPath). An Ag retrieval procedure was performed by incubation in citrate buffer (pH 6) for 40 min at 99°C. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 5 min before the application of primary Ab. The sections were then incubated for 1 h with primary Abs: 10 μg/ml mouse monoclonal anti-IL-17A (IgG2b) or 10 μg/ml mouse monoclonal anti-IL-17F (IgG1). In negative control sections, irrelevant Ab (mouse IgG2b) was applied at the same concentration as the primary Ab. After washing, the sections were incubated with biotinylated anti-mouse Ab for 15 min, followed by streptavidin-peroxidase complex for 15 min and 3,3′-diaminobenzidine chromogen solution (DakoCytomation). The sections were then counterstained with Mayer’s hematoxylin.

Synoviocytes seeded in 6-well plates (5 × 10⁵ cells/well) were serum deprived for 2 h and then stimulated for 1, 3, 6, or 12 h with IL-17A or IL-17F (50 ng/ml) alone or in combination with TNF-α (0.5 ng/ml) in DMEM/10% FCS. RNA was extracted using TRIzol reagents (Life Technologies) and purified using RNeasy kits (Qiagen). The concentration of RNA was quantified by spectrophotometry at 260 nm (SmartSpec 3000; Bio-Rad).

One microgram of total RNA was reverse transcribed using a ThermoScript RT-PCR System (Invitrogen Life Technologies). Briefly, total RNA was denatured by incubating for 5 min at 65°C with 4 μM oligo(dT) primer and then reverse transcribed by using a final concentration of 0.5 mM dNTP, 40 U/μl RNase OUT, 0.01 M DTT, and 10 U/μl of ThermoScript reverse transcriptase. Reverse transcription was performed by incubation at 50°C for 60 min followed by 85°C for 5 min. The obtained cDNA was diluted 1/10 with distilled water and 10 μl was used for amplification. Specific primer sets for IL-6, CXCL8/IL-8, CXCL5/ENA-78, RANTES/CCL5, CXCR4, and GAPDH were optimized for the LightCycler instrument (Roche Molecular Biochemicals) and purchased from Search-LC. Primer-specific nucleotide sequences for TNFR type I (GenBank accession no. NM_001065), TNFR type II (GenBank accession no. NM_001066), IL-32 (GenBank accession no. NM_001012631), lipoprotein lipase (LPL; GenBank accession no. NM_000237), and CDC42EP3 (GenBank accession no. NM_006449) were synthesized by Eurogentec: TNFRI forward: 5′-ctcctgtagtaactgtaagaa-3′ and TNFRI reverse: 5′-gtctaggctctgtggctt-3′; TNFRII forward: 5′-ccccaccagatctgtaacg-3′ and TNFRII reverse: 5′-tatcggcaggcaagtgagg-3′; IL-32 forward: 5′-ggagacagtggcggcttat-3′ and IL-32 reverse: 5′-ggcaccgtaatccatctctt-3′; LPL forward: 5′-ccatgacaagtctctgaataagaa-3′, LPL reverse: 5′-ccccaaacactgggtatgtt-3; and CDC42EP3 forward: 5′-agactcggctggatctgc-3′, CDC42EP3 reverse: 5′-gaccacaaccaggacaaacc-3′. The PCR was performed using the LightCycler FastStart DNA SYBR Green I kit (Roche Molecular Biochemicals) according to the protocol provided with the parameter-specific kits (45 amplification cycles, denaturation at 96°C, primer annealing at 68 °C with touchdown to 58°C, amplicon extension at 72°C). The copy number of target mRNA was normalized by the housekeeping gene GAPDH.

Two micrograms of RNA from RA synoviocytes were analyzed using HG-U133A arrays (Affymetrix) according to the manufacturer’s instructions (in vitro transcribed labeling protocol). RNA integrity number was assessed using RNA 6000 nano chips (mean ± SD: 8.3 ± 0.1) and the Agilent 2100 Bioanalyzer (Agilent Technologies). Total RNA was used to prepare double-stranded cDNA containing the T7 promoter sequence. cRNA was synthesized and labeled with biotinylated ribonucleotide (GeneChip IVT Labeling Kit; Affymetrix). The fragmented cRNA was hybridized onto HG-U133A oligonucleotide arrays (22,283 probe sets). The arrays were washed and stained using the fluidic station FS450 (Affymetrix; protocol EukGE-WS2v4). The array was scanned with the Agilent G2500A GeneArray Scanner. Array hybridization quality controls were analyzed (mean ± SD of scaling factor: 1.7 ± 0.8; GAPDH ratio: 1.0 ± 0.1; and β-actin ratio: 2.3 ± 0.1).

Statistical analysis was performed using the Affymetrix Data Mining Tool Software (version MAS 5.0). There were 12,074, 12,912, 12,238, 12,391, and 12,779 probe set signals detected (present call) after IL-17A, IL-17F, TNF-α, IL-17A plus TNF-α, and IL-17A plus TNF-α, respectively. Gene products that showed a signal intensity of 30 or greater were considered as significantly expressed and genes that showed a change of 2-fold or greater compared with the control situation were considered as significantly regulated. A global comparison between the regulatory effects of IL-17A and IL-17F was first assessed using the comparison call from the MAS 5.0 algorithm. The signal intensity of 705 and 438 probe sets showed a change of 2-fold or greater when compared with the untreated condition upon IL-17A and IL-17F stimulation, respectively. From these subsets, a list of 680 distinct probe sets (601 genes) induced by IL-17A and/or IL-17F showed a signal intensity of 30 or greater. A hierarchical clustering was then applied to compare the effects of IL-17A or IL-17F alone or in combination with TNF-α. The signal intensity of 705, 438, 1032, 1493, and 1288 probe set signals showed a change of 2-fold or greater when compared with the untreated condition upon the stimulation with IL-17A, IL-17F, TNF-α, IL-17A plus TNF-α, and IL-17F plus TNF-α, respectively. From these subsets, a common list of 2408 probe sets (1979 genes) showed a signal intensity of 30 or greater and was analyzed by hierarchical clustering using Spotfire (Spotfire DecisionSite 8.2) on Z-transformed row data. The NETAFFX web site (www.affymetrix.com) was used to select candidate genes. Two ways were used for independent confirmation of the results: in silico analysis using published results and experimental validation by real-time RT-PCR. Three levels of analysis were investigated for in silico validation: cell type specificity, constitutive mRNA expression, and inducible effect upon IL-17A or TNF-α stimulation. The validation by real-time RT-PCR was tested on four target genes (IL-6, IL-8, MMP-3, and CXCR4). Due to the allometric relation between PCR- and microarray-derived data, the correlation was calculated after log transformation.

The following approach was used to quantify the cooperation between IL-17A or IL-17F and TNF-α from large-scale expression data sets. For each gene, H defined the value of gene induction expressed as fold change when compared with the untreated situation and I defined a variable which depends both on IL-17A and TNF-α pathways: H_{TNF + IL-17A} = H_TNF + H_IL-17A + I_(TNF,IL-17A).

To avoid a bias leading to an underestimation of poorly expressed genes, data were normalized. H′ defined the normalized values of induction and I′ defined the associated value of cooperation as follows: H′_{TNF + IL-17A} = H_{TNF + IL-17A}/(H_TNF + H_IL-17A); H′_TNF = H_TNF/(H_TNF + H_IL-17A); and H′_IL-17 = H_IL-17A/(H_TNF + H_IL-17A).

Thus, a cooperation index (I′) was defined as follows: I′_{(IL-17A, TNF)} = H′_{TNF + IL-17A} − (H′_TNF + H′_IL-17A). I′ was not calculable for 11 genes of 1979. A cutoff of 2 was arbitrarily taken to represent significant agreement beyond chance for genes to be classified in the group of inhibition (I′ < −2), additivity (−2 ≤ I′ < 2), or synergy (I′ ≥ 2) depending on the type of cooperation between IL-17A or IL-17F and TNF-α. A part of the apparent cooperation effect might be artifactual due to random variations. To verify the statistical significance of I′, a linear mixed model was used in which the variance of the cooperation index among genes was tested. The dependent covariate in this model was the signal and independent covariates were the experimental conditions (control, IL-17A, TNF-α, and IL-17A plus TNF-α). The four different signals of each gene were considered as repeated measurements and I′ was introduced as a random covariate. A non-null variance of I′ was considered as evidence of heterogeneity of I′ among genes. The three gene subsets (inhibition, addition, synergy) were compared for the presence of AU-rich elements (ARE) using the ARE-mRNA database (http://brp.kfshrc.edu.sa/ARED/) (17).

Synoviocytes (10⁵ cells/condition) were seeded in 12-mm cover slides and stimulated with cytokines. FcRs were blocked using PBS/3% BSA After 1 h of incubation with PE-conjugated anti-TNFRII, cells were fixed and permeabilized using BD Biosciences Cytofix/Cytoperm. Nuclear staining was performed by incubation with Hoechst for 1 h. Fluorescent images were captured by using a Nikon Eclipse 50i microscope connected to a Nikon digital sight DS-U1 camera and merged with the NIS-Elements F 2.20 software.

Protein levels were expressed as mean ± SEM. mRNA expression of target genes was normalized with GAPDH mRNA expression and data are expressed as the fold induction compared with untreated controls. Results are expressed as the mean ± SEM. Statistical inference was performed using Mann-Whitney U test, Pearson χ² tests, and mixed linear models. Values of p < 0.05 were considered statistically significant. All data are the result of at least three separate experiments.

We have previously demonstrated the selective expression of IL-17A protein in RA synovial tissue when compared with OA samples (18). To extend this study to IL-17F, we performed immunohistochemical analysis of both IL-17A and IL-17F on serial sections from OA and RA synovium (Fig. 1). Both cytokines were expressed in RA but not OA synovial tissues and a more sustained staining for IL-17F compared with IL-17A was observed in the synovium from three different RA patients. IL-17A and IL-17F were expressed in the same areas, particularly in lymphocytic aggregates and hyperplasic lining cells (Fig. 1, A and B). Moreover, IL-17F-secreting cells had a plasma cell-like morphology similar to that we had described for IL-17A-secreting cells (Fig. 1, inset in C and D) (18).

Although the effect of IL-17A, alone or in combination with TNF-α, has been studied in RA synoviocytes, the effect of IL-17F still remained to be clarified. First, we compared the effect of IL-17A and IL-17F (0.1–100 ng/ml) on IL-6 secretion by synoviocytes. IL-17F dose-dependently induced IL-6 secretion, but to a lesser extent compared with IL-17A (Fig. 2,A). At 100 ng/ml, IL-17A induced a 4-fold higher induction of IL-6 secretion compared with IL-17F (29.2 ± 3 ng/ml vs 7.1 ± 1.4 ng/ml, respectively, p < 0.05). We compared the effect of IL-17A and IL-17F in the presence of TNF-α and demonstrated that although IL-17F alone had a limited effect alone, it potently induced IL-6 mRNA expression in the presence of TNF-α (18-fold increase compared with the basal level; Fig. 2 B).

We then examined by real-time RT-PCR the time course expression of target genes known to be induced (IL-6, IL-8/CXCL8, and CXCL5) (19, 20, 21) or repressed (RANTES/CCL5) by IL-17A. IL-17A, IL-17F, or TNF-α induced IL-6 mRNA expression as early as 1 h, which increased over 12 h, leading to a 20-, 3-, and 16-fold induction of IL-6 mRNA levels, respectively. After 12 h of exposure, the combination of IL-17A and TNF-α induced a 2-fold induction of IL-6 mRNA expression compared with the combination of IL-17F plus TNF-α (120- and 60-fold increase compared with the basal level; Fig. 2,C). IL-17F had a limited effect on IL-8 mRNA levels compared with IL-17A. At 12 h of stimulation, IL-17A induced a 30-fold higher in IL-8 mRNA expression compared with IL-17F alone. In the presence of TNF-α, both IL-17A and IL-17F further induced IL-8 mRNA expression (Fig. 2,D). We observed a similar trend in ENA-78/CXCL5 mRNA expression: at 12 h of stimulation, IL-17A, IL-17F, or TNF-α alone induced a 69-, 2.5-, and 35-fold induction, respectively, which was further enhanced upon combined stimulation with IL-17A or IL-17F plus TNF-α (Fig. 2,E). IL-17A or IL-17F had no significant effect on RANTES mRNA expression between 1 and 12 h of stimulation, whereas TNF-α was effective, with a 147-fold increase over control at 12 h. We observed that not only IL-17A, but also IL-17F induced a massive inhibition of TNF-α-induced RANTES expression. At 12 h, IL-17A or IL-17F inhibited 89 or 75% of TNF-α-induced RANTES mRNA expression, respectively (Fig. 2 F).

To further extend these studies to a large number of genes, mRNA profiles from IL-17A- and IL-17F-stimulated cells were studied using microarrays. The expression of several cell-specific markers was analyzed to characterize the population used in the experiment. As described in Table I, the synovium-cell derived population was negative for CD19, CD3E, von Willebrand factor, markers which are specifically expressed by B cells, T cells, and endothelial cells, respectively (22). CD68, a marker of monocytes/macrophages, was expressed at very low levels, whereas high levels of fibroblastic markers CD90 (23, 24), CD44 (25), prolyl 4-hydroxylase (26), and CD55 (27) were detected. These results indicated that the cell population was mainly composed of type B synoviocytes, with a residual proportion of type A synoviocytes. These results extend the observations made at the protein level by flow cytometry with cells >99% negative for CD3, CD19, and CD14.

As a way of confirmation, we also analyzed the mRNA expression of genes based on published knowledge from synoviocyte studies. Table I shows genes known to be constitutively overexpressed in RA synoviocytes compared with OA synoviocytes (caldesmon 1 (28), biglycan (28), aggrecan 1 (28), VCAM-1 (29, 30), IGFBP-5 (29, 31), CXCL12 (29, 31), CD147/basigin (32), MMP-2 (32), and IL-6 (22, 28)), genes known to be expressed at low levels compared with OA synoviocytes (CXCL2 (28) and CXCL6 (28)) and genes known not to be expressed in RA synoviocytes (IL-10, IL-13, TNF-α, IL-15, and IL-18 (22)). The accuracy of our microarray datasets was further tested through the analysis of expression data upon stimulation (Table I). The known regulatory effects of IL-17A or TNF-α on RA synoviocytes were confirmed in our microarray data sets. This included the positive regulatory effect of IL-17A and TNF-α on CCL20 (20), CYP7B1 (33, 34), IL-6 (12, 35), LIF (35, 36), IL-8 (12), and IL-23p19 (37) expression, along with the specific regulatory effect of TNF-α on VCAM-1 mRNA expression (38, 39). Moreover, no regulatory effect on IL-12p35 mRNA expression was observed upon IL-17A or TNF-α stimulation, as previously described (37). Finally, real-time RT-PCR experiments performed with cells from four different RA donors confirmed the regulatory effects of IL-17A and TNF-α on IL-6 and IL-8 mRNA expression.

Among the 601 genes considered as significantly expressed and regulated upon IL-17A and/or IL-17F stimulation, we observed that IL-17A and IL-17F had similar regulatory effects on 424 genes (70.6%; Fig. 3 A). Only one gene, CDC42EP3, was classified as regulated in the opposite way by IL-17A and IL-17F. However, such regulatory effect was not significant (−1.9 and +2.0-fold induction after IL-17A and IL-17F stimulation, respectively; NS). The lack of significant effect was confirmed by real-time RT-PCR (+1.5 and +2.0-fold induction after IL-17A and IL-17F stimulation respectively; NS). Interestingly, 165 genes were specifically regulated by IL-17A (27.4%) but only 11 were specifically regulated by IL-17F (1.8%). Again, the level of induction and/or expression in the latter group of genes was poorly significant. Based on this analysis, IL-17A and IL-17F had a very similar regulatory effect on RA synoviocytes (McNemar’s χ²: 101; df = 1; p < 0.00001), but at the same time, IL-17A regulated the expression of more genes than IL-17F (McNemar’s χ²: 133; df = 1; p < 0.00001).

We then compared IL-17A- and IL-17F-induced signals in the presence of TNF-α. A subset of 1979 genes considered as significantly expressed and regulated upon stimulation with at least one stimulus was analyzed by hierarchical clustering (Fig. 3 B). IL-17A induced a distinct expression profile when compared with TNF-α, as demonstrated by the metric distance in the hierarchical clustering. The profile induced by IL-17F was closer to the control than to IL-17A-stimulated cells. In the presence of TNF-α, the expression patterns of IL-17A and IL-17F were rather similar.

To study the cooperative effect between IL-17A or IL-17F and TNF-α, we used the cooperation index (I′) resulting in inhibitory, additive, or synergistic interactions (see Materials and Methods for details). Fig. 4^,A represents the distribution of genes according to I′ and showed the presence of outlier genes. These outliers were either positive (I′_{(IL17A, TNF)} > 2; synergy) or negative (I′_{(IL17A, TNF)} < 2; inhibition). Fig. 4^,B shows that these cooperation effects were observed for genes independently of signal intensity. The linear mixed model showed that the observed variability of the cooperation index (I′) was statistically significant (likelihood ratio test: 34.17; 2 df; p < 0.0001). Among the 1968 genes, 80 genes showed a pattern of inhibition (I′_{(IL17A, TNF)} < 2), 1758 a pattern of additivity (−2 < I′_{(IL17A, TNF)} < 2), and 130 a pattern of synergy (I′_{(IL17A, TNF)} > 2) according to the cooperation index. Among the genes synergistically induced by IL-17A or IL-17F and TNF-α, we found again IL-6, IL-8, CXCL5, but also IL-23p19, E-selectin, Egr-1, and G-CSF. We validated by real-time RT-PCR the synergistic induction of IL-8, a chemoattractant for neutrophils, of CXCR4, a chemokine receptor implicated in the migration of hematopoietic and metastatic cells, and of LPL, which catalyzes the hydrolysis of chylomicrons and very low-density lipoproteins (Fig. 4 C). Furthermore, we validated the inhibitory effect of IL-17A and IL-17F on TNF-α-induced IL-32 expression, a proinflammatory cytokine overexpressed in RA synoviocytes when compared with OA synoviocytes (40). Additional examples of such regulation can be found in a supplemental file⁴ added to this article.

Multiple mechanisms are likely to drive cooperation between IL-17A or IL-17F and TNF-α in RA synoviocytes. Among them, posttranscriptional regulations, mainly through ARE, have been previously described (41). Thus, we looked for the presence of ARE in our gene subsets using the ARE-mRNA database (http://brp.kfshrc.edu.sa/ARED/). No ARE enrichment was found in synergistically induced genes (n = 130 genes) when compared with the subsets of additivity and inhibition (n = 1818 genes) (χ²: 0.7091; df = 1; p = 0.4). Moreover, the mRNA expression of the well-known ARE-BP (AUF1, TTP, HuR, TIAR, TIA-1, BRF1; (42)) was not regulated by IL-17A or IL-17F alone or in combination with TNF-α (data not shown). We then tested whether a feedback mechanism involving IL-17A/F or TNF receptors could be associated with the cooperative effect between IL-17A/F and TNF-α. The time course mRNA expression of IL-17RA, IL-17RC, TNFR type 1 (p55), and TNFR type 2 (p75) was studied upon IL-17A or IL-17F stimulation alone or in combination with TNF-α (data not shown). The stimulation with TNF-α had no effect on IL-17RA, IL-17RC, TNFR type 1, or TNFR type 2 in RA synoviocytes. Conversely, IL-17A or IL-17F alone or in combination with TNF-α up-regulated TNFRII mRNA expression, which peaked after 3–6 h of stimulation, whereas the expression of TNFRI, IL-17RA and IL-17RC remained stable (Fig. 5,A). We showed by immunofluorescence an up-regulation of TNFRII surface expression upon IL-17A stimulation (Fig. 5,B). Finally, we showed that TNFRII neutralization using a mAb inhibited the synergistic effects of IL-17A or IL-17F and TNF-α on CCL20/MIP-3α (Fig. 5 C). Thus, part of the cooperation effects between IL-17A/F and TNF-α is possibly mediated at the level of receptor expression.

The recent discovery of Th17 cells as additional effector T cells distinct from Th1 and Th2 cells challenged the traditional concept of a Th1-driven RA disease (43). These cells are characterized by the production of a distinct profile of effector cytokines, including IL-17A, IL-17F, IL-21, and IL-22. In RA patients, IL-17A is present in both synovium and synovial fluid (4, 44). We showed here that both IL-17A and IL-17F are expressed within RA synovium, whereas no expression was detected in OA samples. IL-17F-expressing cells are characterized by a plasma cell-like morphology, with a large nucleus and a reduced cytoplasm, as we have previously described for IL-17A (18). Moreover, IL-17F expression was strongest than IL-17A in RA synovial tissues. These observations are consistent with previous studies reporting higher levels of IL-17F secretion compared with IL-17A upon stimulation of CD4⁺ T cells (10, 14, 45).

Using high throughput microarrays, we showed that IL-17A and IL-17F have a significant similar regulatory effect on RA synoviocytes, IL-17F being less active quantitatively. Moreover, IL-17A and IL-17F induced similar expression profiles in the presence of TNF-α, as demonstrated by the hierarchical clustering analysis. These expression data, taken together with the specific expression of IL-17A and IL-17F in RA synovium, are in line with a role for both cytokines in joint degradation, neutrophil attraction, and angiogenesis in the context of chronic inflammation.

The cooperative effect between IL-17A and TNF-α was analyzed using the cooperation index. The majority of regulated genes were additively induced by IL-17A and TNF-α (89%), whereas the patterns of inhibition and synergy only represented 11% of the relevant genes. Among the synergistically induced genes, we validated by real-time RT-PCR the regulation of LPL and CXCR4. CXCR4 is a chemokine receptor mainly expressed on immune cells, except for stem cells and tumor cells in which CXCR4 play a key role for tissue homing. Its expression on synoviocytes is still controversial (46, 47, 48). Our report demonstrates for the first time the inducible expression of CXCR4 on RA synoviocytes. Indeed, IL-17A and IL-17F synergistically induced CXCR4 in the presence of TNF-α, while no regulatory effect was seen for all other chemokine receptors (data not shown). The cognate ligand of CXCR4 is CXCL12/stromal-derived factor-1, which is overexpressed in synovial tissue and synovial fluid from RA patients compared with OA patients (46, 49, 50). In RA joints, its expression is detected in synovial lining, synovial high endothelial venules, and synovial fluid (46, 49, 51). Thus, these data suggest a role for IL-17A and IL-17F in synoviocyte migration toward the CXCL12 gradient.

We showed that both IL-17A and IL-17F inhibited TNF-α-induced IL-32 mRNA expression. IL-32 is overexpressed in RA synovium when compared with OA, and its expression is mainly detected in lymphocytic infiltrates (52, 53). Furthermore, IL-32 was associated with disease severity (52). Interestingly, it was shown in epithelial cells that IFN-γ and IL-18 were potent inducers of IL-32 expression (54). Its induction by Th1-associated cytokines remains to be established in synoviocytes to clarify whether IL-17-induced IL-32 inhibition could represent a negative regulatory mechanism of the Th1 pathway.

We showed that the differences in expression patterns (inhibition, addition, synergy) were not linked to a preferential distribution of ARE in the promoter of target genes nor to a specific regulation of ARE-BP. Although ARE, which are key players in the mRNA life time, were found in the 3′ untranslated region of many inflammatory mediators (55), they may not be crucially involved in synergistic cooperation between IL-17A or IL-17F and TNF-α. Conversely, both IL-17A and IL-17F up-regulated TNFR type 2 receptor mRNA and protein expression. Such a positive regulatory loop may in part participate in the potentiating effect of IL-17A or IL-17F on TNF-α-induced target gene mRNA expression, as demonstrated by the effect of TNFRII neutralization on CCL20/MIP-3α secretion. In human bronchial epithelial cells, it has been demonstrated that blockade of both TNFRI or TNFRII down-regulated IL-17A or IL-17F plus TNF-α induced G-CSF secretion (56). Finally, the distribution of the cooperation index highlighted the complexity of the networks implicated in gene regulation. Previous studies supported the existence of multiple mechanisms implicated in the cross-talk between IL-17- and TNF-associated pathways. These could lead to either activation or inhibition of gene expression: activation of signaling pathways (MAPKs, Akt/protein kinase B) (57, 58), recruitment of specific transcription factors (C/EBP, NF-κB, AP-1, IFN regulatory factor 1) (59, 60, 61), and mRNA stabilization (41).

In conclusion, we demonstrated that IL-17A and IL-17F are specifically expressed in RA synovium when compared with OA. In RA synoviocytes, both cytokines induced a similar expression pattern in the presence of TNF-α. Accordingly, IL-17F appears as a target in Th17-mediated diseases such as RA.

The authors have no financial conflict of interest.