HIF-2α-induced chemokines stimulate motility of fibroblast-like synoviocytes and chondrocytes into the cartilage-pannus interface in experimental rheumatoid arthritis mouse models

Introduction Pannus formation and resulting cartilage destruction during rheumatoid arthritis (RA) depends on the migration of synoviocytes to cartilage tissue. Here, we focused on the role of hypoxia-inducible factor (HIF)-2α-induced chemokines by chondrocytes in the regulation of fibroblast-like synoviocyte (FLS) migration into the cartilage-pannus interface and cartilage erosion. Methods Collagen-induced arthritis (CIA), K/BxN serum transfer, and tumor necrosis factor-α transgenic mice were used as experimental RA models. Expression patterns of HIF-2α and chemokines were determined via immunostaining, Western blotting and RT-PCR. FLS motility was evaluated using transwell migration and invasion assays. The specific role of HIF-2α was determined via local deletion of HIF-2α in joint tissues or using conditional knockout (KO) mice. Cartilage destruction, synovitis and pannus formation were assessed via histological analysis. Results HIF-2α and various chemokines were markedly upregulated in degenerating cartilage and pannus of RA joints. HIF-2α induced chemokine expression by chondrocytes in both primary culture and cartilage tissue. HIF-2α -induced chemokines by chondrocytes regulated the migration and invasion of FLS. Local deletion of HIF-2α in joint tissues inhibited pannus formation adjacent to cartilage tissue and cartilage destruction caused by K/BxN serum transfer. Furthermore, conditional knockout of HIF-2α in cartilage blocked pannus formation in adjacent cartilage but not bone tissue, along with inhibition of cartilage erosion caused by K/BxN serum transfer. Conclusion Our findings suggest that chemokines induced by IL-1β or HIF-2α in chondrocytes regulate pannus expansion by stimulating FLS migration and invasion, leading to cartilage erosion during RA pathogenesis.


Introduction
Rheumatoid arthritis (RA) is a chronic autoimmune disease causing synovial inflammation, bone erosion, and cartilage destruction. The inflammatory process initially affects a single joint, but the disease usually progresses to affect nearly all joints [1]. RA is characterized by abnormal synovial hyperplasia caused by increased proliferation and activation of synoviocytes in inflamed synovia [2,3]. The pannus, an aggressive front of hyperplastic synovium, invades and destroys mineralized cartilage and bone [4][5][6]. Migration and invasion of activated fibroblast-like synoviocytes (FLS) into cartilage and bone are critical events during invasive pannus formation in RA synovium [4][5][6]. However, little is currently known on the molecular mechanisms underlying migration and invasion of activated FLS and formation of pannus.
FLS in the inflamed synovium produce numerous catabolic regulators, including chemokines, cytokines, and matrix metalloproteases (MMPs), which are directly implicated in joint destruction [7,8]. Production of these factors, in addition to cell-cell interactions, is crucial for progression of RA [8,9]. For instance, FLS produce CX 3 CL1, which attracts T cells to the joint [10]. T cells promote tumor necrosis factor alpha (TNFα) production, in turn causing FLS proliferation [10]. FLS in inflamed synovium also produce several other chemokines, including CXCL8, CXCL9, CXCL10, CXCL12, CCL2, and CCL3 [7,[11][12][13]. Chemokines are key players in inducing directed chemotaxis in nearby responsive cells and are classified into four subfamilies, specifically CXC, CC, CX 3 C, and XC [14]. These signaling proteins are essential for leukocyte trafficking during chronic inflammatory diseases, such as RA, atherosclerosis, and adipose inflammation [15], and additionally stimulate migration, proliferation, and MMP activation in FLS during RA pathogenesis [16]. Chemokines are also produced by chondrocytes in cartilage tissues of RA joints [17][18][19], but the potential role of hypoxia-inducible factor (HIF)-2α-induced chemokines in regulation of RA pathogenesis remains to be explored.
Previously, we demonstrated that HIF-2α (encoded by Epas1), a transcription factor regulated by oxygen tension [20], acts as an essential catabolic regulator of osteoarthritis (OA) by modulating the expression of various catabolic factors, including MMPs, in chondrocytes [21]. HIF-2α additionally regulates inflammatory RA by modulating angiogenesis, various functions of FLS, and interleukin (IL)-6-dependent T-helper (T H ) 17 cell differentiation [7]. We initially demonstrated that HIF-2α modulates the expression of various chemokines in both chondrocytes and FLS. Since no links have been established between HIF-2αinduced chemokines, pannus formation, and cartilage erosion, we focused on the potential role of chondrocytederived chemokines in FLS migration and invasion during RA pathogenesis. Here, we examined the hypothesis that HIF-2α-induced chemokine production regulates interactions of chondrocytes with FLS, resulting in formation and invasion of pannus into cartilage tissue and consequently cartilage erosion in RA-affected joint tissues.

Mice and experimental arthritis
Male DBA/1 J, C57BL/6, Epas1 +/− , Epas1 fl/fl , TNFα transgenic (TG), and K/BxN mice were used for experimental RA studies. Epas1 +/− C57BL/6 mice were backcrossed against the DBA/1 J strain for eight generations to generate Epas1 +/− DBA/1 J [7]. TNFα TG and K/BxN mice were obtained from Taconic Farms (Hudson, NY, USA) and Jackson Laboratory (Bar Harbor, ME, USA), respectively. Mice were housed in specific pathogen-free barrier facilities and used in accordance with protocols approved by the Animal Care and Ethics Committees of Chonnam National University and Gwangju Institute of Science and Technology, Gwangju, Republic of Korea. Collagen-induced arthritis (CIA), K/BxN serum transfer, and TNFα TG mice were employed as experimental RA models. CIA was induced in DBA/1 J mice using a previously described standard protocol [7]. For K/BxN serum transfer, K/BxN and control sera were collected from arthritic K/BxN and nonarthritic BxN mice, respectively. K/BxN or control serum (200 μl) was administered intraperitoneally on days 0 and 2, and mice sacrificed 14 days after serum transfer [22]. TNFα TG mice (20 weeks old, heterozygous) were used as a model for spontaneous inflammatory arthritis, with non-TG wild-type (WT) littermates as controls [23].
FLS and articular chondrocyte culture and conditioned medium preparation FLS were isolated from joint synovium of WT and Epas1 +/− mice according to previously described protocols [7]. FLS between passages 4 and 8 were used for further analysis. Flow cytometry using antibodies against the fibroblast marker, CD90, and the macrophage marker, CD14 (Abcam), led to the identification of pure FLS (>90 % CD90 + /<1 % CD14 + ). Chondrocytes were isolated from cartilage tissue by digestion with 0.2 % collagenase type II, as described previously [21,24], and cells on culture day 3 used for further treatments. To prepare conditioned medium (CM), chondrocytes or FLS were treated with IL-1β or infected with 800 multiplicity of infection (MOI) of empty adenovirus (Ad-C) or adenovirus expressing HIF-2α (Ad-Epas1) for 24 hours in the absence of serum.

RNA isolation, RT-PCR, quantitative RT-PCR, and western blotting
Total RNA was extracted from mouse articular chondrocytes and knee joint cartilage using TRI reagent. For isolation of RNA from knee joints, cartilage tissues were sliced using a surgical blade and RNA isolated using the Purelink RNA mini kit (Thermo Fisher Scientific). RNA was reversetranscribed and cDNA amplified using PCR. Transcript levels were quantified with quantitative real time-PCR (qRT-PCR) performed using SYBR Premix Ex Taq™ (TaKaRa Bio Inc., Shiga, Japan). All qRT-PCR reactions were performed in duplicate, and threshold cycle values from target genes normalized to that of glyceraldehyde-3phosphate dehydrogenase (GAPDH). Primer sequences for PCR were described in an earlier report [7]. Western blot analysis was performed to detect secreted CXCL1, CXCL2, and CCL5 in CM using the following antibodies: rabbit anti-CXCL1 (Abcam), goat anti-CXCL2 (R&D Systems), and goat anti-CCL5 (R&D Systems).

Cell migration and invasion assay
For the transwell migration and invasion assay, FLS from WT (Epas1 +/+ ) or Epas1 +/− mice (C57BL/6) were seeded on membranes of inserts with 8.0 μm pores (Corning Costar, Corning, NY, USA ) and the lower chambers filled with CM, which served as the chemoattractant. Serum-free CM was prepared from primary cultures of chondrocytes or FLS isolated from WT or Epas1 +/− mice (C57BL/6).
For neutralization of chemokines, 0.5 μg anti-IgG (control), anti-CXCL2, or anti-CCL5 (R&D Systems) were added to the lower chamber. For the invasion assay, inserts of the transwell were precoated with matrigel (BD Bioscience, San Jose, CA, USA). After 24 hours of incubation, transwell inserts were fixed with 4 % paraformaldehyde for 10 minutes and stained with 0.05 % crystal violet. Cells on the bottom layer were captured in five randomly selected fields at 200× magnification using a light microscope and quantified with Image J software (NIH, shareware) .

Microarray analysis
Mouse articular chondrocytes and FLS were infected with Ad-Epas1 or empty virus at an MOI of 800 for 24 hours. Total RNA was extracted from mouse articular chondrocytes or mouse FLS cells using TRI reagent (Molecular Research Center, Cincinnati, OH, USA). Three replicates for each cell type were isolated and processed. RNA of mouse articular chondrocytes was analyzed using Agilent microarrays (Agilent Mouse Whole Genome 4 × 44 K Microarray), in accordance with the Agilent protocol (Genomic Tree Inc., Daejeon, Korea ). RNA of mouse FLS was analyzed using Affymetrix GeneChip arrays (Affymatrix GeneChip Mouse Gene 2.0 ST Array) using the Affymetrix protocol (DNALINK Inc., Seoul, Korea). All microarray raw data are available through the GEO database [GEO:GSE73658, GEO:GSE73659].

Statistical analysis
The nonparametric Mann-Whitney U test was used for analysis of data based on an ordinal grading system, such as synovitis, pannus, and Mankin scores. Data obtained with qRT-PCR assays, cell migration and invasion, and HIF-2αpositive cells were initially tested for conformation to normal distribution using the Shapiro-Wilk test and subsequently analyzed with Student's t test (pair-wise comparisons) or analysis of variance (ANOVA) with post hoc tests (multicomparison), as appropriate. Significance was accepted at the 0.05 level of probability (P <0.05).

HIF-2α is upregulated in both cartilage and pannus of the experimental mouse RA joint
To elucidate the possible links and regulatory mechanisms between pannus formation and cartilage erosion, we initially examined HIF-2α expression levels in chondrocytes from cartilage and FLS in pannus of inflamed joint tissues. CIA [7], K/BxN serum transfer [22], and TNFα TG mice [23] were employed as inflammatory arthritis models. CIA mice displayed severe synovitis, pannus formation, and cartilage erosion in the ankle joint, along with a marked increase in HIF-2α protein levels in chondrocytes of damaged cartilage and FLS in pannus adjacent to cartilage tissue (Fig. 1a). Similarly, TNFα TG mice (20 weeks old) exhibited synovitis, pannus formation, and cartilage erosion with increased HIF-2α levels in chondrocytes and FLS of cartilage and pannus, respectively (Fig. 1b). Additionally, HIF-2α was markedly upregulated in cartilage and pannus in the K/BxN serum transfer model, with severe synovitis and cartilage erosion (Fig. 1c). Quantitation of HIF-2α-positive cells in cartilage and pannus of RA-affected joints revealed significantly increased numbers of chondrocytes and FLS in all the models of inflammatory arthritis examined (Fig. 1d).
Next, we examined whether HIF-2α regulates the expression of chemokines in mouse cartilage tissue. Consistent with previous findings [7], intra-articular (IA) injection of Ad-Epas1 promoted pannus formation (Fig. 3a). RT-PCR and qRT-PCR analyses of cartilage tissue revealed that HIF-2α induces a significant increase in mRNA levels of the chemokines examined (Fig. 3b). Furthermore, HIF-2α-overexpressing chondrocytes in cartilage tissue of mice IA injected with Ad-Epas1 showed increased levels of CXCL1, CXCL2, and CCL5 proteins, which co-localized with HIF-2α (Fig. 3c). Similarly, CXCL1, CXCL2, and CCL5 proteins were increased in HIF-2α-overexpressing chondrocytes in cartilage tissue of CIA mice and co-localized with HIF-2α (Fig. 3d). Our results collectively indicate that HIF-2α upregulates chemokines in both primary culture chondrocytes and cartilage tissue.

HIF-2α-induced chemokines by chondrocytes regulate FLS migration and invasion
FLS, a major cell type in pannus tissue, migrate, attach to, invade, and destroy cartilage tissue by secreting proteases within the RA joints [4]. Since a major function of chemokines is recruitment of adjacent cells as a chemoattractant, we used an in vitro modified Boyden chamber assay to further examine whether increased expression of chemokines by HIF-2α in FLS and chondrocytes plays a role in FLS migration and invasion. Our data initially confirmed that CM from FLS infected with Ad-Epas1 stimulates both migration and invasion of FLS (Fig. 5a). . e HIF-2α binding sites in promoters of indicated chemokines. ChIP assays in chondrocytes infected with 800 MOI Ad-C or Ad-Epas1. The histone H3 antibody was used as the positive control and IgG as a negative control. f The indicated HRE sites in each chemokine promoter were amplified from immunoprecipitated DNA (left) and pre-immunoprecipitated DNA (right; input). Values presented as mean ± standard error of the mean (*P <0.005). FLS fibroblast-like synoviocytes, HIF hypoxia-inducible factor, HRE hypoxia-inducible factor responsive element Next, we examined whether HIF-2α-induced chemokines by chondrocytes also regulate migration and invasion of FLS. Notably, CM from Ad-Epas1-infected chondrocytes promoted both migration and invasion of FLS (Fig. 5b). The correlation between chondrocyteoriginating HIF-2α and chemokine function in FLS migration and invasion was further validated. CM from chondrocytes treated with IL-1β stimulated the migration and invasion of WT (Epas1 +/+ ) FLS (Fig. 5c). However, Epas1 +/− FLS exhibited significantly reduced migration and invasion in the presence of IL-1β-treated chondrocyte CM (Fig. 5c). Furthermore, addition of anti-CXCL2 and anti-CCL5 antibodies to CM of IL-1βtreated chondrocytes blocked FLS migration (Fig. 5d). Based on the results, we propose that HIF-2α-induced chemokines by chondrocytes regulate FLS migration and invasion.

Cartilage-specific knockdown of HIF-2α inhibits FLS migration at the cartilage-pannus interface
Finally, we elucidated the in vivo significance of HIF-2αderived chemokine production in chondrocytes during RA cartilage destruction. For this purpose, HIF-2α was locally downregulated in joint tissues via IA injection of Ad-Cre in Epas1 fl/fl mice and experimental RA induced by K/BxN serum transfer. Immunostaining in synovium and cartilage revealed that Ad-Cre injection effectively reduced the levels of HIF-2α induced by K/BxN serum transfer (Fig. 6a). Local deletion of Epas1 in joint tissues significantly suppressed all the hallmarks of RA, including synovitis and synovial hyperplasia, pannus formation and invasion into calcified cartilage and bone, and cartilage destruction (Fig. 6b, c). To elucidate the role of HIF-2α-regulated chemokines in cartilage, we used chondrocyte-specific conditional knockout (cKO) mice (Epas1 fl/fl ;Col2a1-cre) in a K/BxN serum transfer model. In cKO mice, K/BxN serum transfer did not cause upregulation of HIF-2α in cartilage whereas HIF-2α was upregulated in synovial tissue (Fig. 6d). Consistently, Epas1 fl/fl ;Col2a1-cre mice exhibited significantly reduced cartilage erosion and pannus formation adjacent to cartilage tissue after K/BxN serum transfer, compared with Epas1 fl/fl mice (Fig. 6e, f ). However, K/BxN serum transfer caused a similar degree of synovitis and bone loss at the bone-pannus junction (Fig. 6e, f ), indicating that HIF-2α knockout in cartilage inhibits pannus expansion

Discussion
Our current results suggest that RA-triggering factors, such as IL-1β and HIF-2α, enhance the expression of several chemokines in chondrocytes, including Cxcl1, Cxcl2, Cxcl5, Cxcl10, Ccl2, Ccl5, and Ccl7. These HIF-2α-induced chemokines in chondrocytes stimulate migration and invasion of FLS, leading to pannus expansion adjacent to cartilage tissue of RA joints. Functional blockage of HIF-2α-induced chemokines, such as Cxcl2 and Ccl5, ameliorated FLS motility. Moreover, Cxcl1, Cxcl2, Cxcl5, Cxcl10, and Ccl5 were identified as direct target genes of HIF-2α in chondrocytes. Local deletion of HIF-2α in joint tissues or conditional knockout in cartilage tissue inhibited pannus formation adjacent to cartilage tissue and cartilage destruction caused by K/BxN serum transfer, demonstrating the significance of chondrocyte-originated chemokines in pannus expansion and cartilage destruction during RA pathogenesis.
Formation of invasive pannus, a thickened layer of synovial tissue that erodes cartilage and bone, is the pathologic hallmark of RA [2,3]. FLS is a main effector cell type in the progression of RA, promoting pannus formation with cartilage erosion and progressive joint destruction by secreting diverse cytokines, chemokines, and MMPs [5,8,9]. Here, we demonstrated upregulation of HIF-2α in chondrocytes from damaged cartilage and FLS in the pannus region of various RA mouse models, including CIA, TNFα TG, and K/BxN serum transfer. To elucidate the potential functions of HIF-2α, we screened the gene expression profiles in HIF-2αoverexpressing chondrocytes and FLS via microarray analysis. Functional enrichment analysis of microarray data revealed upregulation of genes associated with RA pathogenesis, such as those involved in extracellular matrix (ECM) remodeling and inflammatory responserelated, cell migration-related and invasion-related processes (data not shown). Among these, we focused on chemokines expressed in both chondrocytes and FLS. Chemokines regulate chemotaxis in nearby responsive cells and are implicated in chronic inflammatory diseases, such as RA, atherosclerosis, and adipose inflammation [14]. . c RT-PCR and western blot analyses of the indicated molecules in primary cultures of chondrocytes isolated from WT (Epas1 +/+ ) or Epas1 +/− mice treated with IL-1β for 24 hours. RT-PCR was carried out to detect transcript levels of Epas1, Cxcl1, Cxcl2, and Ccl5. Secreted chemokine proteins were examined via western blot. d Immunostaining of HIF-2α, CXCL1, CXCL2, and CCL5 in cartilage tissue of WT and Epas1 +/− DBA1/J mice with CIA. Values presented as mean ± standard error of the mean (*P <0.05, **P <0.01). Scale bar: 50 μm. CIA collagen-induced arthritis, DAPI 4′,6-diamidino-2-phenylindole, HIF hypoxia-inducible factor, IL interleukin Release of chemoattractants by FLS signals various immune cells to migrate into the RA synovium [4]. Chemokines, such as CX 3 CL1, CCL2, and CXCL10, and their receptors, CCR2, CCR5, and CXCR3, regulate FLS migration or invasion during RA pathogenesis [16,27,28]. In addition to FLS, chondrocytes produce various chemokines and related catabolic factors during cartilage destruction. For instance, human OA chondrocytes generate various chemokines (CXCL1, CCL2, and CCL5) and chemokine receptors (CCR1, CCR2, CCR3, CCR5, CXCR1, and CXCR2) by proinflammatory cytokines [17][18][19]. CXCR1/CXCR2 production by human OA chondrocytes stimulates MMP3 release [19], and constitutively expressed CXCR2 in normal chondrocytes maintains cartilage homeostasis [29]. However, the functions and regulatory mechanisms of HIF-2α-induced chemokines in RA pathogenesis are yet to be elucidated. Our current results clearly demonstrate that HIF-2α upregulates various chemokines as direct target genes in primary culture chondrocytes. Additionally, upregulated HIF-2α in cartilage tissue stimulates chemokine expression in chondrocytes. We detected three CCL (CCL2, CCL5, and CCL7) and four CXCL chemokine families (CXCL1, CXCL2, CCXCL5, and CXCL10) that were highly upregulated by IL-1β or HIF-2α. Among these, Cxcl1, Cxcl2, Cxcl5, Cxcl10, and Ccl5, but not Ccl2 and Ccl7, were identified as direct target genes of HIF-2α. Ccl2 and Ccl7 genes may be regulated indirectly or independently of HIF-2α. NF-κB is a potential transcription factor responsible, since the promoter region contains a NF-κB binding site and IL-1βinduced HIF-2α upregulation is controlled by NF-κB signaling [21]. Furthermore, our data show that HIF-2αinduced chemokines by chondrocytes stimulate in vitro migration and invasion of FLS. The observation that addition of blocking antibodies against CXCL2 and CCL5 to CM of chondrocytes abrogates FLS migration clearly demonstrates an essential role of HIF-2α-induced chemokines by chondrocytes in FLS motility. Although chemokines are produced by both chondrocytes and FLS, it remains to be established whether HIF-2α-induced chemokines by chondrocytes act in a manner complementary to or redundant with those produced by FLS in the development and pathogenesis of experimental RA. Different chemokine subtypes are upregulated in FLS and . c Migration and invasion of FLS treated (12 hours) with CM prepared from WT (Epas1 +/+ ) or Epas1 +/− chondrocytes treated with IL-1β (2 ng/ml, 24 hours) (n = 4). d Migration of FLS treated with CM prepared from chondrocytes exposed to IL-1β (2 ng/ml, 24 hours) in the presence of IgG or neutralizing antibodies against CXCL2 and CCL5. Values presented as mean ± standard error of the mean (*P <0.05, **P <0.005). Scale bar: 50 μm. Ab antibody, CON control, FLS fibroblast-like synoviocytes, IL interleukin chondrocytes. For instance, CXCL3 and CXCL2 are the most highly upregulated in Ad-Epas1-infected FLS, while CXCL2, CXCL5, CXCL1, and CCL5 are significantly upregulated by HIF-2α-overexpressing chondrocytes. Among these, CXCL2 expression was the most upregulated in Ad-Epas1-infected chondrocytes. Chondrocyte-derived CXCL2 appears to be a major chemoattractant triggering FLS migration, based on data showing that blockage of CXCL2 with specific antibodies inhibits FLS migration in IL-1βtreated CM. All chemokines exert biological effects by stimulating G protein-linked receptors [14]. Microarray analysis revealed the presence of CXCR2 (for CXCL1, CXCL2, and CXCL5), CXCR3 (for CXCL10), and CCR5 (for CCL5) [14] in both mouse FLS and articular chondrocytes, although the expression levels of these receptors were not affected by HIF-2α overexpression (data not shown).
Chondrocytes in cartilage tissue are embedded in cartilage matrix, raising the question of how chondrocytederived chemokines regulate the motility of FLS, which are localized in synovial tissues. One possibility is that chemokines diffuse from the cartilage matrix to affect synovial cells [30]. Although we did not validate this theory in the current study, we were able to demonstrate the significance of chondrocyte-derived chemokines in the expansion of invasive pannus in vivo. Local deletion of HIF-2α in joint tissues via IA injection of Ad-Cre in Epas1 fl/fl mice suppressed pannus formation and cartilage destruction. Furthermore, cKO of HIF-2α in cartilage tissue, which inhibited chemokine production, led to sufficient blockade of pannus formation adjacent to cartilage tissue, but not in bone tissue, resulting in inhibition of cartilage destruction. Cartilage is eroded by breakdown of ECM via the action of several proteases, such as MMPs and cathepsin, during RA. Our group previously showed that stimulation of chondrocytes with IL-1β increases secretion of MMPs in association with marked HIF-2α transcriptional activity [7,21]. MMPs participate in ECM remodeling and play important roles in the progressive destruction of joints in RA. FLS are considered a major effector cell type involved in cartilage erosion, based on their ability to produce massive Fig. 6 Knockdown of HIF-2α in joint tissues inhibits cartilage destruction at the cartilage-pannus interface. a-c Epas1 fl/fl mice were IA-injected with Ad-C or Ad-Cre (1 × 10 9 plaque-forming unit ) to locally delete HIF-2α in joint tissue. K/BxN serum was transferred into mice to induce experimental RA. Representative images of HIF-2α immunostaining in the synovium and cartilage of ankle joints (n = 8) a. Detection b and quantitation c of synovitis, cartilage destruction (Mankin scores), and pannus formation (n = 10). d-f K/BxN serum was transferred to Epas1 fl/fl (WT) or Epas1 fl/ fl ;Col2a1-cre mice. Representative images of HIF-2α immunostaining in the synovium and cartilage of ankle joints d. Detection e and quantitation f of synovitis, cartilage destruction (Mankin scores), and pannus formation (n = 5). Values presented as mean ± standard error of the mean (*P <0.01, **P <0.001; NS not significant). Scale bar: 50 μm. B bone, C cartilage, cKO conditional knockout, HIF hypoxia-inducible factor, P pannus, WT wild type amounts of degradative enzymes, such as MMPs. Since both FLS and cartilage are major cell types that produce ECMdegrading enzymes [7,21], cartilage erosion during RA may be attributable to a multistep and complex process including "FLS attachment to cartilage" and "synthesis of enzymes from FLS and cartilage that degrade the ECM". This theory is supported by the finding that deletion of HIF-2α in both synovium and cartilage exerts a greater ameliorating effect on cartilage destruction than local deletion in cartilage.

Conclusions
In summary, HIF-2α is overexpressed in damaged chondrocytes as well as FLS in the pannus region in several experimental RA mouse models, and directly regulates the expression of Cxcl1, Cxcl2, Cxcl5, Cxcl10, and Ccl5. In addition, we have identified a previously unrecognized role of chondrocyte-derived chemokines and provided a mechanistic explanation for the regulatory processes underlying the invasive properties of FLS. Although our findings from experimental RA mouse models remain to be verified in human RA pathogenesis, we carefully suggest new insights into the effects of chondrocytes on synovial cell motility into the cartilage-pannus interface in RA pathogenesis and highlight the relevance of HIF-2α as a potential novel therapeutic target in RA.