Abstract
Objective. To investigate the expression of Gremlin 1 (GREM1), an antagonist of bone morphogenetic protein, in rheumatoid arthritis (RA) synovia and its involvement in the hyperplasia and invasiveness of fibroblast-like synoviocytes of RA (RA-FLS).
Methods. Computational analysis was introduced to identify FLS-predominant regulators. GREM1 expression was examined by immunohistochemistry, real-time PCR, and ELISA. FLS proliferation and apoptosis were determined using tetrazolium-based colorimetric assay and APOPercentage assay, respectively. FLS migration and invasion were evaluated by wound migration and Matrigel invasion assay, respectively. Expressions of Bax, Bcl2, pErk1/2, and pAkt were detected by Western blot analysis.
Results. Through global transcriptome profiling, we identified a GREM1 gene predominantly expressed in RA-FLS. Indeed, the GREM1 expression was higher in synovia, synovial fluids, and FLS of patients with RA than in those of patients with osteoarthritis, and its levels correlated well with proinflammatory cytokine concentrations. Knockdown of GREM1 transcripts using short interfering RNA (siRNA) reduced the proliferation and survival of RA-FLS along with downregulation of pErk1/2, pAkt, and Bcl2 expressions, whereas it induced Bax expression. Conversely, the addition of recombinant GREM1 to RA-FLS showed the opposite results. Moreover, GREM1 siRNA decreased the migratory and invasive capacity of RA-FLS, whereas exogenous GREM1 increased it. The GREM1-induced FLS survival, migration, and invasion were completely blocked by neutralizing antibodies to ανβ3 integrin on RA-FLS, suggesting that ανβ3 integrin mediates the antiapoptotic and promigratory effects of GREM1.
Conclusion. GREM1 is highly expressed in RA joints, and functions as a regulator of survival, proliferation, migration, and invasion of RA-FLS.
Rheumatoid arthritis (RA) is an autoimmune disease characterized by chronic inflammation and progressive joint destruction by activation of various inflammatory cells1. Particularly, fibroblast-like synoviocytes (FLS), the major cells in the synovial lining layer, directly participate in chronic inflammation through the secretion of inflammatory cytokines and matrix-degrading enzymes2. In addition, they have highly migratory and invasive features to the bone and cartilage after detaching from synovia, which leads to further joint destruction1,2,3. Moreover, FLS of patients with RA (RA-FLS) harbor somatic mutations of several oncogenes or tumor suppressor genes, such as H-Ras and p53, and proliferate abnormally despite the presence of cell death receptors and ligands4,5,6.
Gremlin 1 (GREM1) is a 20.7 kDa protein belonging to the cerberus and dan subfamily of bone morphogenetic protein (BMP) antagonists7. This protein forms heterodimers with BMP-2, BMP-4, and BMP-7, and inhibits the interaction of BMP with its receptor on the cell surface8,9,10. Inhibition of BMP by GREM1 in mice allows the expression of the fibroblast-growth factor and sonic hedgehog, which are necessary for proper limb development11. In several studies, GREM1 was shown to be important for the proliferation, migration, and invasion of various cancer cells12,13,14. In addition, GREM1 exerts a proangiogenic activity in the endothelium of human lung tissues15. Despite the progress in the understanding of GREM1 in the development and cancer, little is known about GREM1 expression in RA and its involvement in FLS biology and RA pathogenesis.
Here, we identified a GREM1 gene predominantly expressed in RA-FLS through a global transcriptome profiling of synovium and FLS. We also validated that GREM1 was highly expressed in synoviums, synovial fluids (SF), and FLS of patients with RA as compared with those of patients with osteoarthritis (OA). GREM1 knockdown by short interfering RNA (siRNA) decreased the proliferation and survival of RA-FLS. In parallel, GREM1 siRNA decreased the expressions of pAkt, pErk1/2, and Bcl2 in RA-FLS, while it increased Bax expression. Conversely, the addition of recombinant GREM1 to RA-FLS exhibited the exact opposite results. Additionally, GREM1 siRNA reduced the migratory and invasive capacity of RA-FLS, whereas recombinant GREM1 induced it. The GREM1-induced FLS survival, migration, and invasion were completely blocked by neutralizing antibodies to ανβ3 integrin on RA-FLS. GREM1 is highly expressed in RA joints and promotes survival, proliferation, migration, and invasion of RA-FLS through its functional receptor ανβ3 integrin. These results suggest that GREM1 is crucial for abnormal proliferation and invasiveness of RA-FLS and could therefore be a potential target for anti-FLS therapy.
MATERIALS AND METHODS
Transcriptome data analysis
Gene expression data collected from FLS (GSE7669 and GSE4061) and synovial tissues (GSE12021) were obtained from the Gene Expression Omnibus16. Each of the datasets was normalized using the quantile method17. Differentially expressed genes (DEG) in RA versus OA, RA versus normal control, or RA versus common reference mRNA (CRM) samples were determined using the integrative hypothesis testing method, as previously reported18,19. DEG were selected as genes with false discovery rate < 0.05 and fold change ≥ 1.5. As a result, 17 genes were identified as commonly upregulated genes in all 3 comparisons.
Selection of potential regulators predominantly expressed in RA-FLS
Out of the 17 genes using gene ontology annotation, 13 genes were found to have regulatory activities (e.g., transcriptional regulators and/or signaling molecules). We further performed a selection of RA-FLS–predominant genes by comparison of expression levels of 13 regulators in RA-FLS and FLS of patients with OA (OA-FLS) with those of 28 normal tissues or 60 cancer cell lines. For this, the expression dataset for FLS (GSE7669) was combined with 28 normal tissue expression datasets obtained from the BioGPS database20 and the NCI60 cancer cell line (a panel of 60 diverse human cancer cell lines used by the U.S. National Cancer Institute) gene expression dataset21, separately, and then the expression levels of 13 regulators were compared with the mean expression levels of 28 normal tissues or NCI60 cell lines using the 1-sided Student t test.
Immunohistochemical analysis for GREM1
For the immunohistochemistry of GREM1, 5 μm sections of paraffin-embedded synovia of patients with RA and OA were treated with pepsin for 30 min and blocked with bovine serum albumin (BSA) for 30 min at room temperature. Tissue sections were then incubated with an anti-GREM1 antibody (1:100; Santa Cruz Biotechnology) overnight at 4°C. Nonspecific mouse immunoglobulin (Sigma-Aldrich) was used for a negative control. Each slide was washed 3 times in Tris-buffered saline (TBS) and incubated with biotinylated anti-mouse IgG (Vector Laboratories) in a humid chamber for 30 min. The GREM1-positive cells were detected by adding peroxidase-conjugated streptavidin (Vector Laboratories), followed by 3′3-Diamino-benzidine tetrahydrochloride (Vector Laboratories).
Double immunofluorescence staining of RA synovia
Seven-μm sections of cryo-embedded synovia of patients with RA were incubated in acetone for 30 min at −20°C and blocked with BSA for 30 min at room temperature. Tissue sections were then incubated with anti-CD55 antibodies (1:200; Santa Cruz Biotechnology) plus anti-GREM1 antibodies (Santa Cruz Biotechnology) or anti-CD68 antibodies (1:50; Invitrogen) plus anti-GREM1 antibodies (Santa Cruz Biotechnology) overnight at 4°C. Each slide was washed 3 times in PBS and incubated with anti-mouse Alexa 488 for CD55 or CD68 (Invitrogen) and anti-goat Cy3 for GREM1 (Abm) in a dark humid chamber for 40 min. The slides were fixed with DAPI mounting solution.
Isolation and culture of FLS
FLS were isolated from the synovial tissues of patients with RA or OA, as described previously22. Cells from passage 3 to 6 were used throughout the experiments. Our study protocol was approved by the Institutional Review Board of the Catholic Medical Center (VC14TIMI0206).
ELISA
The levels of BMP-2, interleukin (IL) 1β, IL-6, and transforming growth factor (TGF) β in the SF of patients with RA and OA were measured with an ELISA kit (R&D) according to the manufacturer’s instructions. The amount of GREM1 was also measured in the SF and culture supernatants of RA-FLS using an ELISA kit (USCN Life Science Inc.).
RNA isolation and real-time PCR
Total mRNA was isolated from the FLS of patients with RA and OA using an RNease Kit (Quiagen) and reverse transcribed to complementary DNA (cDNA). Quantitative real-time PCR was performed in the StepOnePlus Real-Time PCR System (Applied Biosystems) using SYBR Premix (Applied Biosystems) according to the manufacturer’s instructions. GAPDH was used as an internal control for PCR amplification. Transcript levels were calculated relative to controls and expressed as −ΔΔCt.
Transfection of GREM1 siRNA
For the downregulation of GREM1 transcripts, FLS were transfected with 50 nM GREM1 siRNA (Dharmacon) using DharmaFECT transfection reagent (Dharmacon) in Opti-MEM medium (Invitrogen) according to the manufacturer’s instructions. After transfection, FLS were cultured at 37°C for 4 h and then placed again in culture medium for FLS. After 48 h, GREM1 mRNA expression levels in FLS were determined by real-time PCR using GREM1 primers (Bioneer).
Western blot analysis
FLS were lysed in lysis buffer for 15 min at 4°C. The concentration of lysate was measured by the Bradford assay (Bio-Rad). The membrane was blocked by 5% skim milk in TBS containing 0.25% Tween 20 and incubated overnight at 4°C with antibodies to pAkt (1:1000; Cell Signaling Technology), Akt (1:1000; Cell Signaling Technology), pErk (1:1000; Cell Signaling Technology), total Erk (1:1000; Santa Cruz Biotechnology), Bax (1:1000; Santa Cruz Biotechnology), Bcl2 (1:1000; Santa Cruz Biotechnology), and β-actin (1:10000; Abm). Membranes were then visualized using an enhanced chemo-luminescent technique.
Cell proliferation assay
After the addition of recombinant human GREM1 (rhGREM1; R&D), the FLS proliferation rate was determined using a 5-bromo-2′-deoxyuridine (BrdU) incorporation assay kit (Roche Life Science) according to the manufacturer’s instructions. In RA-FLS transfected with GREM1 siRNA, cell proliferation was also assessed by tetrazolium (MTT) assay after stimulation of tumor necrosis factor (TNF)-α or TGF-β for 72 h as previously described23.
Detection of apoptosis
FLS apoptosis levels were determined in the presence of rhGREM1 (R&D) or GREM1 siRNA using an APOPercentage apoptosis assay kit (Biocolor Ltd) as described previously23. Photo images of APOPercentage dye-labeled cells, which stained pink under a light microscope, were used to quantify the extent of apoptosis. In some experiments, FLS apoptosis was assessed in the presence of BMP-2 (100 ng/ml; R&D), a GREM1 antagonist24.
Wound migration assay
Wound migration was assessed as previously described25. Briefly, FLS were seeded to 80% confluence on 6 well plates, scratched with pipette tips, and then treated with 200 ng/ml of rhGREM1 in the presence or absence of BMP-2 (R&D) for 12 h. In other sets of experiments, the migration of GREM1-deficient RA-FLS was also assessed 12 h after transfection with GREM1 siRNA, when cell viability was intact. The migrated FLS were quantitated by counting the cells that moved beyond the reference line.
Matrigel invasion assay
The BD BioCoat Matrigel Invasion Chamber (Becton Dickinson) was used to assess FLS invasion, according to the manufacturer’s instructions. Briefly, after the transfection of GREM1 siRNA or the addition of rhGREM1, RA-FLS were allowed to migrate in Matrigel invasion chamber for the indicated times. The noninvading cells were removed afterward by scrubbing with a cotton-tipped swab, and the cells on the lower surface of the membrane were stained with Diff-Quik stain (Baxter Diagnostics). For quantification, cells were counted in 8 random fields.
Flow cytometry analysis for ανβ3 integrin
RA-FLS were incubated with anti-ανβ3 integrin antibody as the primary antibody (Chemicon international) at 4°C for 30 min. The cells were then washed in PBS with 0.5% BSA and labeled with phycoerythrin-conjugated goat anti-mouse IgG (BD Biosciences) at 4°C for 30 min. The expression of ανβ3 integrin on FLS was detected using a FACSCanto flow cytometer (BD Biosciences).
Statistical analysis
Data are expressed as the mean ± SD. Comparisons of the numerical data between groups were performed by the paired or unpaired Mann-Whitney U test. P values < 0.05 were considered statistically significant.
RESULTS
Identification of GREM1 gene by profiling the global gene expression of RA-FLS
Disease-associated genes have been shown to be frequently upregulated in a certain disease condition, and their expression reflects tissue origin26,27. Thus, we hypothesized that key genes might be highly expressed under RA conditions and also that their expression levels should be predominant in the synovia. To assess this hypothesis, we first collected 3 gene expression datasets derived from studies for FLS (GSE7669 and GSE4061) and synovial tissues (GSE12021)28,29,30. Three sets of upregulated genes, including 227 genes (RA-FLS vs OA-FLS), 4086 genes (RA vs normal synovial tissues), and 432 genes (RA-FLS vs CRM samples), were identified by integrative hypothesis testing (Details under Materials and Methods; Figure 1A). We then focused on the 17 genes that are commonly upregulated in all 3 comparisons. Among the 17 genes, we further selected 13 genes with regulatory activities (e.g., transcriptional regulators and/or signaling molecules) using gene ontology annotation (Supplementary Table 1; all supplementary material available from www.ribjd.com/base/data/data01.php?com). Interestingly, the list of 13 genes included well-known regulators involving RA pathogenesis. For example, CDH11 is highly expressed in synovial cells at the pannus region in rheumatoid synovitis and induces the invasive property of FLS31. FAS and TNC are expressed at the areas of inflammation and tissue damage in RA synovia32,33. These results suggest that others of the 13 genes, albeit not studied in RA pathogenesis, can be essential for pathologic actions of RA-FLS. Thus, we subsequently tried to select the genes specific for RA-FLS by comparison of expression levels of 13 potential regulators with transcriptome profiles of 28 normal tissues or 60 cancer cell lines. As a result, we identified 4 genes — GREM1, SFRP4, THBS1, and SOCS5 — out of the 13 regulators predominantly expressed in RA-FLS over 28 normal tissues and 60 cancer cell lines (Figure 1A and Supplementary Figure 1). Interestingly, GREM1 gene showed the greatest statistical significance in terms of specificity (Figure 1A), and its involvement in RA pathogenesis has never been studied. Taken together, this result proposed that GREM1 could be a novel regulator with predominant expression in RA-FLS.
Elevated GREM1 expression in RA joints. (A) Transcriptome analysis of FLS and synovial tissues. Venn diagram depicts the number of overlapping genes with significant overexpression in 3 RA transcriptome datasets. (B) Immunohistochemical staining of synovia of patients with RA (a, b, c, and d) and OA (e and f) using anti-GREM1 antibody (a, b, d, e, and f) and isotype control (c). Rectangular area in (a) is magnified to (b) and (c). Bars = 200 µm. (C) Double immune-fluorescent staining of an RA synovium (n = 3) using anti-CD55 antibody plus anti-GREM1 antibody (upper panel) or anti-CD68 antibody+ and anti-GREM1 antibody (lower panel). Bars = 100 µm. (D) Concentration of GREM1 in the sera of non-RA controls (n = 20), sera of patients with RA (n = 80), and synovial fluids of patients RA (n = 19), primary (1′) OA (n = 25), and secondary (2′) OA (n = 8) as determined by ELISA. Secondary OA indicates noninflammatory joint fluids (white blood cell count/mm3 < 1000) of patients with RA. Data show the mean ± SD. * P < 0.05. *** P < 0.0001. GREM1: Gremlin 1; RA: rheumatoid arthritis; FLS: fibroblast-like synoviocytes; OA: osteoarthritis; CRM: common reference mRNA; SFRP4: secreted frizzled-related protein 4; THBS1: thrombospondin 1; SOCS5: suppressor of cytokine signaling 5.
Increased expression of GREM1 in FLS and SF of patients with RA
To validate our postulation, we first performed immunohistochemical staining for GREM1 in the synovial tissues of patients with RA and OA. The high level of GREM1 protein expression was observed in the synovial tissues of patients with RA (Figure 1B: a, b, and d) as compared with those of OA (Figure 1B: e and f), particularly in the lining layer. To assess the GREM1-expressing cells in the RA synovium, double immunofluorescence staining was performed using anti-CD55 antibody (green), which was specific to FLS in the synovia2, and anti-GREM1 antibody (red). The result showed that the superimposition of CD55+ staining on GREM1 staining (yellow) highlighted the lining layer of RA synovia. As a control, anti-CD68 antibody (green), which stained macrophage lineage2, was done, showing that the cells positive for GREM1 were distinct from CD68+ cells in the synovia and the colocalization of GREM1 with CD68 was lacking in the sublining layer. This result indicated that GREM1 was expressed in FLS in the synovial lining region (Figure 1C).
GREM1 is a secreted glycoprotein. Thus, we measured the concentration of secreted GREM1 in the SF obtained from patients with RA and OA by ELISA. While the concentration of GREM1 was not elevated in the sera of patients with RA (n = 80) as compared with those of non-RA controls (n = 20), it was significantly higher in the SF of patients with active RA (n = 19) than in those of the patients with primary (n = 25) and secondary OA (n = 8; Figure 1D). The SF of secondary OA indicated noninflammatory joint fluids (white blood cell count/mm3 < 1000) of patients with RA. These results indicated that abnormal GREM1 response was specific to RA joints, the actual sites of chronic inflammation with synovial hyperplasia. Collectively, RA-FLS in the synovial lining layer were a major cellular component expressing the GREM1 protein, and they may secrete soluble GREM1 into the SF of patients with RA.
Proinflammatory cytokines induced GREM1 expression
Figure 1 shows that FLS are a major cell type expressing GREM1 protein. Based on this finding, we examined the basal expression levels of GREM1 in FLS of patients with RA and OA. Real-time PCR analysis revealed that the GREM1 mRNA expression level was more than 3-fold higher in RA-FLS (n = 6) than in OA-FLS (n = 6; Figure 2A). Proinflammatory cytokines, including IL-1β, TNF-α, TGF-β, and hypoxic stimulus CoCl2, increased the GREM1 mRNA expression in RA-FLS, but this was not noted with OA-FLS (Figure 2B). In contrast, IL-4, an antiinflammatory cytokine, decreased GREM1 mRNA expression in RA-FLS (Figure 2B). Moreover, as determined in the culture supernatants of RA-FLS by ELISA, treatment of proinflammatory cytokines resulted in an increase in the secretion of the GREM1 protein (Figure 2C). Additionally, GREM1 concentrations in the SF of patients with RA (n = 27) correlated well with levels of IL-1β, IL-6, and TGF-β in the same SF (Figure 2D). Taken together, these results suggest that not only is GREM1 constitutively overexpressed in RA-FLS, but its levels can also be further increased by proinflammatory cytokines and hypoxia.
Association of GREM1 expression with proinflammatory cytokines. (A) Basal expression of GREM1 mRNA in RA-FLS (n = 8) versus OA-FLS (n = 8). FLS (6 × 105) were cultured in Dulbecco modified Eagle’s medium (DMEM) supplemented 10% FBS for 24 h. The mRNA level was measured by real-time PCR. GAPDH was used as an internal control for PCR amplification. ** p < 0.01 versus OA-FLS. (B) Stimulated expression of GREM1 mRNA in RA-FLS (n = 6) and OA-FLS (n = 6). FLS (6 × 105) were cultured in DMEM supplemented with 1% FBS and stimulated with IL-1β (10 ng/ml), TNF-α (10 ng/ml), TGF-β (10 ng/ml), IL-4 (10 ng/ml), IL-6 (10 ng/ml), and CoCl2 (200 µM) for 12 h. The mRNA level was determined by real-time PCR. GAPDH was used as an internal control for PCR amplification. * P < 0.05 versus unstimulated RA-FLS. † P < 0.05. (C) GREM1 protein expression in RA-FLS. RA-FLS (1 × 104) were treated with IL-1β (10 ng/ml), TNF-α (10 ng/ml), and TGF-β (10 ng/ml) for 24 h, and then GREM1 concentrations in the supernatants were measured by ELISA. Data are the relative values to basal level of GREM1 secreted by unstimulated RA-FLS. Mean level of GREM1 produced by unstimulated RA-FLS (none) was 1.9 ± 0.8 ng/ml. ** p < 0.01 versus unstimulated cells. (D) Correlation of GREM1 and proinflammatory cytokine concentrations (conc.) in the SF of patients with RA, which were measured by ELISA. Spearman rank correlation coefficient (r) and its level of significance (p) were computed by SPSS software. GREM1: Gremlin 1; RA: rheumatoid arthritis; FLS: fibroblast-like synoviocytes; OA: osteoarthritis; DMEM: Dulbecco’s modified Eagle’s medium; FBS: fetal bovine serum; IL-1β: interleukin 1β; TNF-α: tumor necrosis factor-α; TGF-β: transforming growth factor β.
GREM1 promoted FLS proliferation and survival
RA-FLS, similar to cancer cells, abnormally proliferate and are resistant to apoptotic death, resulting in the formation of a hyperplastic pannus1,2. Moreover, GREM1 controls the proliferation of various cancer cells12,13,14. Therefore, we investigated the contribution of GREM1 to FLS proliferation and survival using siRNA against GREM1 transcripts as well as using rhGREM1. As seen in Figure 3A, we first confirmed that the GREM1 mRNA expression was markedly reduced by transfection of RA-FLS with GREM1 siRNA for 48 h. We then tested whether GREM1 regulates FLS proliferation in response to proinflammatory cytokines. The result showed that GREM1 siRNA completely abrogated TNF-α–induced and TGF-β–induced RA-FLS proliferation (Figure 3B). To the contrary, proliferation of RA-FLS was increased by the treatment of rhGREM1 for 5 days, as determined by the BrdU incorporation assay (Figure 3C), suggesting that GREM1 promotes FLS proliferation. A signaling pathway critical for FLS proliferation, such as pAkt and pErk1/2, has been described in RA34. As shown in Figure 3D, transfection of GREM1 siRNA decreased the levels of pAkt and pErk1/2, whereas exogenous rhGREM1 increased it. The induction of pAkt and pErk1/2 in RA-FLS occurred as early as 5 min and remained high up to 30 min after treatment with rhGREM1, supporting the notion that GREM1 is an important mediator of FLS proliferation. However, the expression levels of Akt and total Erk were altered by neither GREM1 siRNA nor rhGREM1.
Effects of GREM1 on FLS proliferation. (A) Downregulation of GREM1 mRNA expression in RA-FLS by siRNA. Levels of GREM1 mRNA were measured by reverse transcription PCR (upper panel) or real-time PCR (lower panel). The bar plot in the lower panel represents the relative abundance of GREM1 mRNA expression normalized by GAPDH mRNA level. ** P < 0.01 versus control (scrambled) siRNA-transfected RA-FLS. (B) FLS proliferative responses to TNF-α and TGF-β. Twenty-four h after transfection with GREM1 siRNA, RA-FLS (1 × 104 cells) were treated with TNF-α (10 ng/ml) or TGF-β (10 ng/ml) for 72 h. Optical density was measured by tetrazolium assay (left). Under the same culture conditions, manual cell counts were also performed by trypan blue exclusion to identify viable cells (right). Data are the mean ± SD of 3 independent experiments in duplicate. * P < 0.05 and ** P < 0.01 versus unstimulated cells. † P < 0.05. †† P < 0.01. (C) FLS proliferation in response to human recombinant GREM1 (rhGREM1). RA-FLS were cultured in DMEM supplemented with 1% FBS, stimulated with rhGREM1 (200 ng/ml) for 5 days, and subjected to BrdU incorporation assay. Data are the mean ± SD of 3 independent experiments in duplicate and expressed as optical density (O.D.). * P < 0.05 versus unstimulated cells. (D) GREM1 control of pAkt and pErk1/2 expressions in RA-FLS. In the left panel, the cells were transfected with GREM1 siRNA for 48 h. In the right panel, the cells were cultured with rhGREM1 (200 ng/ml) in DMEM supplemented with 0.5% FBS for the indicated times. Expressions of pAkt, Akt, pErk1/2, and total Erk were determined by Western blot analysis. Data represent 3 independent experiments with similar results. GREM1: Gremlin 1; FLS: fibroblast-like synoviocytes; RA: rheumatoid arthritis; siRNA: short interfering RNA; TNF-α: tumor necrosis factor-α; TGF-β: transforming growth factor β; DMEM: Dulbecco modified Eagle’s medium; FBS: fetal bovine serum.
The activation of pErk1/2 maintains mitochondrial integrity through the upregulation of antiapoptotic Bcl2 expression or through the inhibition of proapoptotic Bax expression35,36. Thus, we tested whether GREM1 controlled FLS survival and apoptosis-related molecules, Bcl2 and Bax. As seen in Figure 4A, the number of apoptotic cells markedly increased in GREM1-deficient RA-FLS as compared with control siRNA-transfected RA-FLS, as determined by the APOPercentage assay. Moreover, GREM1 knockdown using siRNA decreased the expression of Bcl2 in RA-FLS, but increased the Bax expression (Figure 4C). On the contrary, the rhGREM1 addition significantly inhibited starvation-induced apoptosis of RA-FLS (Figure 4B). In parallel, rhGREM1 addition induced Bcl2 expression in RA-FLS, while reducing the Bax expression (Figure 4C). Additionally, GREM1 siRNA significantly decreased the viability of RA-FLS after 48 h and 72 h of transfection, as determined by the MTT assay (Figure 4D). Together, these results suggest that GREM1 confers apoptotic resistance to RA-FLS, contributing to synovial hyperplasia in RA.
GREM1 regulation of FLS survival. (A) Apoptosis of RA-FLS by GREM1 siRNA. RA-FLS (n = 3) were transfected with GREM1 siRNA or control (scrambled) siRNA and then cultured in DMEM supplemented with 2% FBS for 48 h. The extent of apoptosis was determined by APOPercentage apoptosis assay. Apoptotic cells were stained. Bar plot on the right shows the number of apoptotic cells. Data are expressed as the mean ± SD of 3 independent experiments in duplicate. * P < 0.05 versus control siRNA-transfected cells. Bars = 200 µm. (B) GREM1 protection of starvation-induced apoptosis. Degrees of apoptosis were assessed by APOPercentage apoptosis assay. RA-FLS (n = 3) were cultured in the absence (upper panel) or presence (lower panel) of rhGREM1 (200 ng/ml) for 12 h. Data show the mean ± SD of 3 independent experiments in duplicate. * P < 0.05 versus unstimulated cells. Bars = 200 µm. (C) GREM regulation of Bcl2 and Bax expressions. RA-FLS (1 × 105 cells) were transfected with GREM1 siRNA for 48 h (left) or stimulated with 200 ng/ml of GREM1 for 24 h (right) in DMEM supplemented with 0.5% FBS. The expressions of Bcl2 and Bax were determined by Western blot analysis. The results are the representative of 3 independent experiments. Con siRNA: control siRNA. (D) GREM1 siRNA induction of FLS death. The FLS (n = 3) were cultured in DMEM supplemented with 1% FBS for the indicated times. Cell viability was determined by MTT assay. Data are normalized with respect to control siRNA transfected cells. * P < 0.05. ** P < 0.005. Values are the mean ± SD of 3 independent experiments in triplicate. GREM1: Gremlin 1; FLS: fibroblast-like synoviocytes; RA: rheumatoid arthritis; siRNA: short interfering RNA; DMEM: Dulbecco modified Eagle’s medium; FBS: fetal bovine serum; MTT: tetrazolium.
GREM1 increased FLS migration and invasion
In addition to abnormal proliferation and apoptotic resistance, increased invasiveness and migration were unique features of RA-FLS and were essential to joint destruction1,31. Thus, we investigated whether GREM1 regulated migration and invasion of RA-FLS. As shown in Figure 5A, under nonlethal time conditions (after 12 h), the knockdown of GREM1 transcripts suppressed the wound migration of RA-FLS. Further, invasion of RA-FLS in a Matrigel chamber was also significantly blocked by transfection with GREM1 siRNA (Figure 5C). Conversely, rhGREM1 (200 ng/ml) treatment significantly increased FLS migration and invasion as compared with untreated cells (Figures 5B and 5D). Collectively, these data indicated that GREM1 was a positive regulator of the migration and invasion of RA-FLS.
GREM1 control of migration and invasion of RA-FLS. (A) GREM1 siRNA suppression of wound migration of RA-FLS, as indicated by images on the phase-contrast microscopy (left) and by the number of manually counted migrated cells (right). After 12 h of transfection with GREM1 siRNA or control siRNA, RA-FLS (1 × 105 cells) were wounded with the tip of a micropipette and incubated in DMEM containing 1% FBS for 12 h. Cells migrating beyond the reference line were photographed and counted. Original magnification × 100. Data show the mean ± SD of 3 independent experiments. * P < 0.05 versus control siRNA. Bars = 200 µm. (B) GREM1 induction of wound migration of FLS. The RA-FLS (n = 3) were cultured with rhGREM1 (200 ng/ml) in DMEM containing 0.5% FBS for 12 h. Data show the mean ± SD of 3 independent experiments. * P < 0.05 versus untreated cells. Bars = 200 µm. (C) GREM1 siRNA suppression of FLS invasion. After 24 h of transfection with GREM1 siRNA or control siRNA, RA-FLS (5 × 105 cells) were incubated in a Matrigel invasion chamber for 12 h. Invasive cells were stained using a Diff-Quik kit, and the number of stained cells was counted. Representative photo images are shown on the left, where the rectangular area in the upper panel is magnified to the lower panel. The mean ± SD of 3 independent experiments is presented on the right. * P < 0.05 versus control siRNA. Bar = 200 µm. (D) GREM1 induction of FLS invasion. The RA-FLS (5 × 105 cells) were incubated in a Matrigel invasion chamber for 12 hours. The upper chamber contained DMEM supplemented with 0.1% FBS plus rhGREM1 (200 ng/ml), and the lower chamber contained serum-free medium. Bar = 200 µm. Data are expressed as the mean ± SD. * P < 0.05 and ** P < 0.01 versus untreated cells. GREM1: Gremlin 1; RA: rheumatoid arthritis; FLS: fibroblast-like synoviocytes; siRNA: short interfering RNA; DMEM: Dulbecco modified Eagle’s medium; FBS: fetal bovine serum.
ανβ3 integrin mediated GREM1-induced increase in FLS survival and migration
GREM1 is a well-known antagonist of BMP7, so we investigated whether GREM1 action on FLS originates from its antagonism to BMP. We focused particularly on BMP-2 among the BMP family, since BMP-2 is dominantly expressed in the RA synovium and promotes FLS apoptosis24. As shown in Supplementary Figure 2A (www.ribjd.com/base/data/data01.php?com), cotreatment of rhGREM1 blocked the BMP-2–induced apoptosis of RA-FLS, suggesting that GREM1 antagonizes proapoptotic activity of BMP-2. However, the GREM1-induced increase in FLS migration and invasion were rarely affected by BMP-2 (data not shown). Paradoxically, the addition of BMP alone increased FLS migration and invasion (data not shown), indicating that GREM1 antagonism to BMP-2 was not a major mechanism responsible for all pathologic actions of GREM1.
We demonstrated that RA-FLS expressed 3 vascular endothelial growth factor receptors — Flt-1, KDR, and neuropilin 1, of which neuropilin 1 is the major one37. In endothelial cells, GREM1 acts as a noncanonical KDR ligand, triggering a potent angiogenic response in a BMP-independent manner38. A report has also demonstrated that GREM1 stimulated the formation of KDR/ανβ3 integrin complexes in endothelial cells, and that anti-β3 antibodies blocked GREM1-induced angiogenesis39. Thus, we tested whether the effect of GREM1 on RA-FLS was mediated by the KDR or ανβ3 integrin. We first confirmed that the ανβ3 integrin was highly expressed on RA-FLS, and its levels were further instigated by the stimulation of IL-1β and TNF-α (Supplementary Figure 3). We also found that the rhGREM1-induced increase in FLS survival was completely abrogated by treatment with neutralizing antibodies to the ανβ3 integrin (Figure 6A). Moreover, GREM1-induced FLS migration and invasion were completely blocked by antibodies to the ανβ3 integrin, but not by antibodies to KDR, Flt-1, and neuropilin 1 (Figures 6B and 6C). Notably, we found that anti-ανβ3 antibody reduced the phosphorylation of Akt, but not pErk, in RA-FLS in the presence of rhGREM1 (Figure 6D), indicating that GREM1 induced the phosphorylation of Akt through the ανβ3 integrin. These results, together with previous reports38,39, suggested that GREM1-induced FLS survival, migration, and invasion were mainly dependent on the ανβ3 integrin.
Involvement of ανβ3 integrin in GREM1-induced FLS survival, migration, and invasion. (A) Dependency of GREM1-induced FLS survival on ανβ3 integrin. RA-FLS (n = 3) were cultured with rhGREM1 (200 ng/ml) in the absence or presence of anti-ανβ3 integrin antibody (10 µg/ml) for 24 h. Degrees of apoptosis were assessed by APOPercentage apoptosis assay. Data show the mean ± SD of 3 independent experiments in duplicate. ** P < 0.01 and *** P < 0.001 versus unstimulated cells. Bars = 200 µm. (B) Effect of antibodies to VEGF receptors or anti-ανβ3 integrin antibody on the wound migration of synoviocytes. RA-FLS (n = 3, 1 × 105 cells) were incubated in DMEM containing 0.5% FBS with 200 ng/ml of rhGREM1 in the absence or presence of 10 µg/ml of anti-VEGF receptor antibodies [including anti-Flt-1 antibody, antineuropilin 1 antibody (anti-NP-1), and anti-KDR antibody] or anti-ανβ3 integrin antibody (10 µg/ml) for 12 h. Cells migrating beyond the reference line were photographed and counted. Data show the mean ± SD of 3 independent experiments and are presented as a ratio (%) relative to unstimulated conditions. * P < 0.001 versus rhGREM1 only. Bars = 200 µm. (C) Dependency of GREM1-induced FLS invasion on ανβ3 integrin. RA-FLS (5 × 105 cells) were incubated in a Matrigel invasion chamber for 12 h. The upper chamber contained DMEM supplemented with 0.1% FBS plus rhGREM1 (200 ng/ml) in the absence or presence of anti-ανβ3 integrin antibody (10 µg/ml), and the lower chamber contained serum-free medium. Data are expressed as the mean ± SD. * P < 0.05 versus untreated cells. † P < 0.05 versus rhGREM1 only. Bar = 200 µm. (D) Phosphorylation of Akt and Erk1/2 in RA-FLS treated with the anti-ανβ3 integrin antibody. The cells were precultured with the anti-ανβ3 integrin antibody (10 µg/ml) for 1 h and then incubated with rhGREM1 (200 ng/ml) in DMEM supplemented with 0.5% FBS for 15 min. Expressions of pAkt, total Akt, pErk1/2, and total Erk were determined by Western blot analysis. Data are the representative of 3 independent experiments with similar results. GREM1: Gremlin 1; FLS: fibroblast-like synoviocytes; RA: rheumatoid arthritis; siRNA: short interfering RNA; VEGF: vascular endothelial growth factor; DMEM: Dulbecco modified Eagle’s medium; FBS: fetal bovine serum.
DISCUSSION
A key issue in developing a systems approach to diseases is to extract disease-relevant information by integrating global datasets and to apply it to clinical settings40. The identification of cell-specific key molecules is especially essential in understanding a particular pathophysiology in complex tissue microenvironments or in discovering novel targets for personalized therapeutic approaches41,42. Here, we introduced a computational approach through the integration of transcriptome profiles to identify FLS-predominant regulators. Through an unbiased analysis of RA transcriptomes in FLS and synovial tissues, as well as through the integration with transcriptome profiles of various normal tissues and cancer cell lines, we successfully identified 4 genes – GREM1, SFPR4, THBS1, and SOCS5 – that were predominantly overexpressed in RA-FLS with regulatory activity. In particular, GREM1 had the greatest statistical significance, and its expression and function were validated in our patients’ samples. These data suggest that our systems approach can help clinicians unveil a novel mechanism underlying disease pathogenesis and identify new targets for RA. The involvement of the other 3 genes in FLS pathology can be similarly determined.
In our present study, we demonstrated first that GREM1 was found in the SF, FLS, and synovial tissues of patients with RA at high levels, and its expression in RA-FLS was induced by proinflammatory cytokines, including IL-1β, TNF-α, and TGF-β, but not by antiinflammatory cytokine IL-4. Moreover, GREM1 concentrations in SF correlated well with IL-1β, IL-6, and TGF-β levels, suggesting that the inflammatory milieu contributes to high GREM1 expression in RA joints. Interestingly, GREM1 expression in OA-FLS was not changed by proinflammatory stimuli, suggesting that cytokine induction of GREM1 might be specific for RA-FLS. Our data further indicated that GREM1 is involved in FLS proliferation.
It is well known that IL-1β, TNF-α, and TGF-β levels are greater in the sera, SF, and synovial tissues of patients with RA than in those of patients with OA1,2,3,43. Also, it has been reported that the expression level of the IL-1β receptor is also higher in the synovial membrane of patients with RA than in that of patients with OA44. TGF-β receptor is highly expressed in RA-FLS as compared with OA-FLS45. Therefore, RA-FLS may have a greater opportunity to be stimulated by these cytokines than OA-FLS. Moreover, the IL-1β–induced or TNF-α–induced nuclear factor-κB (NF-κB) pathway and the TGF-β–induced Smad pathway are more activated in the synovia of patients with RA than those of patients with OA2. We believe that cell responsiveness to IL-1β, TNF-α, and TGF-β is generally increased in RA-FLS, which leads to a higher expression of GREM1 in RA-FLS than in OA-FLS upon IL-1β or TGF-β stimulation. Under these circumstances, proinflammatory cytokines induce GREM1 expression in RA-FLS, and such an increased level of GREM1, in turn, could lead to synovial hyperplasia, which facilitates the further secretion of cytokines and thus establishes a vicious cycle (Supplementary Figure 4, www.ribjd.com/base/data/data01.php?com).
It remains undetermined how GREM1 expression is regulated by inflammatory stimuli. IL-1β is well known to regulate the expression of target genes through the activator protein 1 and nuclear factor-κB pathway46. Therefore, we presume that the interaction of these transcription factors with the promoter sites of GREM1 gene might be involved in the upregulation of GREM1 expression upon cytokine stimulation. However, the basal expression levels of GREM1 mRNA were also higher in RA-FLS than in OA-FLS, which suggests that elevated GREM1 expression is intrinsic to RA-FLS. Imprinted anomalies of RA-FLS have been identified, which are associated with altered microRNA expression and DNA methylation47. Therefore, it is possible that epigenetic dysregulation may contribute to the increased GREM1 expression in RA-FLS at the basal state.
We also showed that GREM1 is essential to FLS survival. GREM1 knockdown resulted in increased apoptotic death of RA-FLS with Bcl2 downregulation and Bax upregulation. Conversely, the addition of exogenous rhGREM1 to RA-FLS showed the opposite result, suggesting that secreted GREM1 by RA-FLS confers apoptotic resistance to FLS themselves in an autocrine/paracrine manner. Additionally, GREM1 siRNA decreased pErk1/2 and pAkt activities, while rhGREM1 increased such activities. Since pErk1/2 is an upstream regulator of Bcl2 and Bax35,36, GREM1 promotes FLS survival by regulating Bcl2/Bax expression, possibly through the activation of pErk1/2. Together, our data suggest that overproduced GREM1 within RA joints protects RA-FLS from apoptosis, thereby maintaining a unique hyperplastic phenotype.
From the hyperplastic synovia, RA-FLS can be detached, and then they migrate to and invade adjacent bone and cartilage1,2,44. At the reattached sites, RA-FLS destroy cartilage and bone by releasing matrix degrading enzymes. In our study, similar to results in cancer cells14, rhGREM1 increased the migration and invasion of RA-FLS. Conversely, siRNA-mediated elimination of GREM1 transcripts impaired these processes, suggesting that GREM1 is a key regulator of FLS migration and invasion. Considering that FLS migration and invasion are critical for joint destruction1,2, these results suggest that a specific blockade of GREM1 activity may delay bone and cartilage destruction by intervening in these processes. Interestingly, it has been reported that skeletal overexpression of GREM1 impairs bone formation and causes osteopenia. Thus, we expect that GREM1 inhibitors can induce bone remodeling in a way that increases bone formation by osteoblasts as well as decrease bone destruction by blocking the promigratory and invasive activity of FLS.
BMP are currently considered pleiotropic cytokines that influence the proliferation, growth, differentiation, and apoptosis of different cell types45,46,48,49. BMP-2 promotes apoptosis of RA-FLS and osteoblasts24,48,50,51, and its receptors are expressed on RA-FLS52. Here, we found that RA-FLS were exposed to the skewed production of GREM1 over BMP-2, as evidenced by the ratio of GREM1 over BMP-2 in the SF, which was significantly higher in patients with RA than in patients with OA (Supplementary Figure 2B). However, in contrast to FLS survival, cotreatment of BMP-2 did not block GREM1-induced FLS migration and invasion (data not shown), suggesting that the effects of GREM1 on FLS biology are only partially because of the GREM1 inhibition of BMP-2 (Supplementary Figure 4).
It has been reported that GREM1 directly binds to KDR and ανβ3 integrin heterodimer in endothelial cells39. However, antibodies to KDR did not affect GREM1-induced RA-FLS migration in our study, which may be because of the very low expression of KDR in RA-FLS22. Instead, blocking experiments using anti-ανβ3 antibody demonstrated that ανβ3 integrin, like endothelial cells, predominantly mediates the GREM1-induced increase in FLS survival, migration, and invasion. The ανβ3 integrin can induce activation of pAkt and pErk in various cell types53,54. Interestingly, we found that the anti-ανβ3 antibody nearly completely abrogated the recombinant GREM1-induced increase in pAkt expression but not pErk expression in RA-FLS. These data indicate that secreted GREM1 increases the phosphorylation of Akt through ανβ3 integrin and also suggest that GREM1-induced pErk activation is mediated possibly through different receptors from ανβ3 integrin. Taken together, although we failed to identify the direct binding partner(s) of GREM1 on RA-FLS, GREM1 may play proliferative, prosurvival, and promigratory roles through its functional receptor ανβ3 integrin (Supplementary Figure 4).
Through global transcriptome profiling, we identified a GREM1 gene that was predominantly overexpressed in RA-FLS. In fact, GREM1 expression was elevated in synovia, SF, and FLS of patients with RA, and correlated with proinflammatory cytokine concentration. We also demonstrated the functional involvement of GREM1 in promoting the hyperplasia, migration, and invasiveness of RA-FLS, the key pathologic features of invasive pannus in RA. In this respect, GREM1 may be an effective therapeutic target for RA capable of demoting hyperplastic synovitis.
Acknowledgment
The authors thank all members of the Institute of Bone and Joint Diseases at the Catholic University of Korea.
Footnotes
Supported by grants from the Korea Healthcare Technology R&D Project, Ministry for Health, Welfare and Family Affairs (No. HI14C3417) and the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (2014R1A2A1A11049812, 2014R1A6A3A04054066, and 2015R1A3A2032927).
- Accepted for publication October 13, 2015.
