Transglutaminase 2 cross-linking activity is linked to invadopodia formation and cartilage breakdown in arthritis

All experimental procedures involving animals were conducted under protocols approved by the Ethics Committee on Animal Research of the University of Sherbrooke, in accordance with regulations of the Canadian Council on Animal Care. Female Lewis/SsNHsd rats (100 to 124 g) were purchased from Charles River Laboratories (St-Constant, QC, Canada). Arthritis was induced by intradermal injection of type II collagen at the base of the tail, as described [11]. In the case of in vivo TG2 inhibition, rats received one intraarticular injection of 3 × 10⁹ units of a lentivirus harboring TG2-shRNA or control (scrambled) shRNA, 10 days after CIA induction.

Tissue sections from the left hind-knee joints were processed immediately after excision by following a standardized paraffin-embedding protocol. Tissue sections were rehydrated, treated with trypsin 1%, and immunohistochemical staining was performed according to the standard avidin-biotin immunoperoxidase complex technique by using a mouse monoclonal ε-(γ-glutamyl)-lysine bond antibody (Abcam (Cambridge, MA, USA) ab424; 1:100), a rabbit polyclonal collagenase cleaved type I and II collagen (Col 2 3/4 short) antibody (Ibex Pharmaceutical (Montreal, QC, Canada) Ab 50-1035; 1:75), or a mouse monoclonal TG2 (Thermo Scientific (Rockford, IL, USA) TG100; 1:50). DAB, SG, or VIP (Vector Laboratories Inc., Burlington, ON, Canada) was used for the detection of labeled proteins. Some sections were counterstained with Harris hematoxylin. Tissue sections from rats in each group were treated simultaneously to palliate interexperimental staining variations.

Sections were stained by using hematoxylin/eosin and the safranin O/Fast green staining protocols to allow visualization of cartilage structure. Pathology scores were evaluated for each tissue section by three blinded observers using a modified Mankin scoring system, as described [27]. In the cartilage structure subcategory of the scoring system, the extent of cartilage invasive zones was evaluated instead of cartilage clefts, the better to account for the cartilage changes observed in CIA.

Analysis of joint section (synovial membrane and cartilage) was performed by using an Axioskop 2 phase-contrast/epifluorescence microscope (Carl Zeiss, Inc., Toronto, ON, Canada) equipped with a 10× or 40× objective at the same level of incident illumination for each slide in the experiment. The condenser was adjusted to bright-field transmission light microscopy. Photomicrographs of six (for 40× magnification) or three (for 10× magnification) random areas within the synovial membrane or cartilage were captured with Retiga SRV cooled color digital camera (Qimaging, Surrey, BC, Canada). The intensity of labeling in the synovial membrane or a representative portion of cartilage in each image was analyzed by using the IHC quantification technique validated by Pham et al. [28]. Images were converted to CMYK in the FIJI software (Open Source). Next, the gray image of the yellow (Y) channel was extracted, and chromogen intensities were analyzed by using the Image Pro software (Media Cybernetics Inc., Bethesda, MD, USA). Results are expressed as the sum of labeling intensity (density) relative to the total area, from two different tissue sections for each rat.

Synovial membranes were removed from the right hind-knee joints of experimental rats, and cells were isolated by collagenase type IV digestion, as described [29]. Four human cell lines isolated from patients diagnosed with RA and undergoing arthroplasty (RA-FLS) or three cell lines isolated from control joints with no evidence of disease (C-FLS) were obtained from Asterand (Detroit, MI, USA). All synovial tissues were obtained with Institutional Review Board approval and appropriate informed consent. Cells were cultured in DMEM-F12 supplemented with 10% FBS and 40 μg/ml gentamycin. The cells were used between passages 3 and 8.

pcDNA3.1-wtTG2 and pcDNA3.1-W241A-TG2 were generously provided by Dr. Gail V.W. Johnson (University of Rochester, Rochester, NY, USA). GFP-lentiviral shRNAs targeting rat TG2 and control (scrambled) shRNA plasmids were from ezBiolab (Carmel, IN, USA). Control and mouse/rat FXIIIA targeting pLKO.1-shRNA plasmids were from Thermo Scientific. Viral particles were generated by transient transfection of 293T cells with the ViraPower lentiviral expression system (Invitrogen, Burlington, ON, Canada). Synoviocytes were tested in the different assays 24 hours after transient transfection by using Fugene 6 (Roche Diagnostics, Laval, QC, Canada) or 48 hours after lentivirus infection with Polybrene (5 μg/ml; Santa Cruz Biotechnology, Santa Cruz, CA, USA). With this procedure, a high efficiency (about 60%) of lentivirus infection was achieved in both rat and human synoviocytes. Cells expressing FXIIIa-shRNAs were selected with a 72-hour puromycin treatment.

Coverslips were prepared by using Oregon Green⁴⁸⁸-conjugated gelatin (Invitrogen) at a final concentration of 1%, as described [30]. Thirty thousand cells were seeded on each coverslip and incubated for 40 hours. Cells were fixed with 1% paraformaldehyde and stained with DAPI and Alexa⁶⁴⁷-conjugated phalloidin (Invitrogen). Cells were visualized with fluorescence microscopy. Invadopodia were identified by actin-rich areas of matrix degradation characterized by loss of green fluorescence. Three hundred cells were counted per coverslip. In the case of TG2-overexpressing cells, TG2 was labeled by using anti-TG2 antibodies (Thermo Scientific CUB7402), and invadopodia formation was counted for TG2-transfected cells only. For inhibition and activation assays, dithiothreitol (DTT; Sigma, Oakville, ON, Canada), Cystamine (Sigma), KCC-009 (provided by Dr K.M. Rich, Washington University, St. Louis, MO, USA), Z-DON (Zedira, Darmstadt, Germany), TGF-β (Peprotech, Rocky Hill, NJ, USA), LY364947 (Tocris Bioscience, Bristol, UK), or TGF-β-neutralizing antibody (R&D Systems (Minneapolis, MN, USA) Ab100-NA; 15 μg/ml) was added 30 minutes after cell plating. In the case of invadopodia assays of GFP-shRNA-expressing cells, Texas-Red-conjugated 1% gelatin slides were prepared as described [31], and the number of invadopodia (loss of red fluorescence) was counted for GFP-positive cells. To quantify the areas of degradation, pictures of fluorescent matrix were analyzed by using ImagePro software, and degradation areas associated with positive cells were calculated in pixels, for a minimum of 25 cells per coverslip.

Synoviocytes cultured on coverslips for 4 to 6 hours were prepared as described [11] and stained by using antibodies against p-cortactin (Cell Signaling (Danvers, MA, USA) Ab 45569; 1:75), TG2 (ThermoScientific CUB 7402; 1:50), and stained for actin by using Texas Red- or Alexa⁶⁴⁷-conjugated phalloidin (Invitrogen; 1:200 or 1:50). Negative control slides were treated with isotype-matched primary antibodies, followed by secondary antibodies. Confocal images were acquired by using a Fluoview 1000 scanning confocal microscope (Olympus, Richmond Hill, ON, Canada) in-line with an inverted microscope equipped with a 60× oil-immersion objective. Color channels were scanned sequentially to avoid overlapping signals.

To evaluate the activity level of TG in vivo and in vitro, EZ-link pentylamine-biotin (PAB) (Pierce/Thermo Scientific) was used as an exogenous TGase substrate. For the live cell activity assay, 30,000 cells were allowed to adhere to 0.2% gelatin-coated coverslips for 30 minutes and incubated for 3 hours with 0.75 mM PAB after addition of treatments, as specified in the figure legends. Coverslips were fixed with 1% PFA, permeabilized with 0.05% saponin, and blocked with 2% BSA. Cross-linked biotin was detected with Alexa⁵⁴⁶-conjugated streptavidin (Invitrogen). Mean fluorescence intensity per cell was calculated from photographs of 20 randomly selected cells for each treatment. To assess TGase activity in articular homogenates, tissue lysates were prepared from freshly excised control and CIA rats (21 days after induction) synovial membranes, homogenized by using a bead mill (Retsch, Newtown, PA, USA) in TGase reaction buffer (50 mM Tris, pH 7.5; 150 mM NaCl; 5 mM CaCl₂; Complete (Roche)). Lysates were centrifuged, and the protein content of the supernatants was determined by using the Bradford method (Bio-Rad Laboratories, Mississauga, ON, Canada). Equal amounts of protein were supplemented with 1 mM DTT and inhibitor (Z-DON, 100 μM; KCC-009, 250 μM) or vehicle. TGase enzymatic activity was started by addition of 5 mM PAB and allowed to proceed at 37°C for 4 or 20 hours. Reaction was stopped by addition of SDS loading buffer and sample boiling. Samples were separated with SDS-PAGE and transferred onto nitrocellulose membranes. To assess TGase activity in articular joints, rat hind knees were injected with 0.5 mM PAB in PBS, 1.2 mM CaCl₂, 0.5 mM MgCl₂, with or without 10 mM Z-DON or 250 M KCC-009, and incubated at 37°C for 4 hours. Tissues were processed as described earlier. In both experiments, biotin-labeled proteins were detected with streptavidin-conjugated alkaline phosphatase.

Whole-cell extracts were prepared by lysis of overnight serum-starved cells in RIPA buffer. ECM protein extracts were prepared from cells grown for 72 hours in complete medium, as described [22]. In brief, cells were detached with 2 mM EDTA/PBS. The remaining cell-assembled matrix was washed with 2 mM EDTA/PBS/0.1% deoxycholate and scraped in 2× strength Laemmli gel-loading buffer to obtain the ECM fraction. Proteins were immunoblotted as described [29] by using anti-TG2 (ThermoScientific TG100; 1:500), anti-LAP (R&D Systems Ab-246-NA; 1:1,000) or anti-actin (Sigma Ab A5060; 1:5,000) antibodies. Band intensities were analyzed by using the Quantity One software (Bio-Rad Laboratories).

Quantikine ELISA for Human TGF-β 1 (R&D Systems) was used, following the manufacturer's instructions, with acid-activated A-FLS supernatants. For luciferase assays, cells were transiently transfected with p3TP-Lux [32] by using 1 mg/ml PEI (Polysciences, Warrington, PA, USA) in DMEM-F12 supplemented with KnockOut Serum Replacement (Invitrogen). Twenty-four hours after transfection, cells were starved and treated overnight, as indicated in the figure legends. Cell lysates were assayed for luciferase activity, as described [29].

The paired or unpaired Student t test or one-way ANOVA test was used to assess statistical significance. A P value smaller than 0.05 was considered significant.

Knee joints of control and CIA rats were prepared for immunohistochemistry (IHC) and labeled with an ε-(γ-glutamyl)-lysine bond-specific monoclonal antibody to obtain an estimation of the cross-linking activity of TGase or TG2-specific antibodies. As revealed by the intensity of labeling, both TG2 and cross-linking activity were detectable at day 18 after the induction of arthritis and increased progressively in the synovial membrane and adjacent cartilage of CIA rats, except at day 32, when the intensity of TG2 labeling declined (Figure 1A and Additional file 1, Figure S1). At days 18 to 32, TG2 and cross-linking labeling were predominantly expressed in the hyperplastic synovial lining and at the synovial lining/cartilage junction, with less activity found within the chondrocyte layer of the cartilage (Figure 1A and Additional file 1, Figure S1). To define whether TG2 is linked to the cross-linking activity found in synovial membranes, we assessed in vivo TGase activity by measuring biotin-pentylamide (BAP) incorporation in homogenates of normal and arthritic synovial membrane or in whole articular joints in the presence or absence of the inhibitors KCC-009, an irreversible TGase inhibitor [33], or Z-DON, a selective TG2 inhibitor [34]. The results indicated that BAP incorporation into proteins from arthritic synovium homogenates was increased (as revealed by the intensity of the bands) compared with normal rat synovium homogenates. In addition, the bands from both normal and arthritic homogenates incubated in the presence of the TG2-specific inhibitor Z-DON showed reduced intensity compared with vehicle, whereas a stronger reduction was obtained by using the TGs inhibitor KCC-009 (see Additional file 2, Figure S2A). Furthermore, direct assessment of TGase activity into articular joints (through intraarticular (IA) BAP injection) indicated that the TGase activity inhibited by Z-DON was located mainly in the synovial lining of arthritic joints (see Additional file 2, Figure S2B). Similar results were obtained with the KCC-009 inhibitor (data not shown).

To evaluate the extent of cartilage damage, the articular tissues were immunostained by using an antibody that detects a newly exposed epitope within collagenase-cleaved type II collagen [35]. Results showed that the intensity of degraded collagen staining in both synovial membrane and underlying cartilage increased with time during the development of arthritis (Figure 1B). In sections stained with hematoxylin/eosin and safranin O, evaluation of histopathologic changes, according to a modified Mankin scoring system [27], revealed a loss of structural integrity of the cartilage, a reduction in proteoglycan content, and synovial membrane hyperplasia, which were more pronounced as arthritis progressed (see Additional file 3, Figure S3). The obtained histologic scores closely reflected changes observed with TG2, TGase, and collagen-degradation assessments. In addition, double labeling of joint tissues revealed robust TGase activity that colocalized with collagen degradation at sites of synovial membrane invasion (Figure 1C). These results clearly suggested an association between cartilage degradation and TGase activity in CIA joints.

We next examined whether silencing the expression of TG2, the main active transglutaminase in articular cartilage [36], affected cartilage degradation. For this, TG2-shRNA- or control-shRNA-expressing lentivirus was injected IA into CIA rats. Injections were performed at day 10 to avoid interference with the initiation of the immune response. Tissue sections from rat hind paws were immunostained with an antibody directed against TG2 or ε-(γ-glutamyl)-lysine epitope to assess efficiency of the treatment. Results indicated that the high levels of TG2 labeling and TGase activity observed in arthritic synovial membranes from control-shRNA-treated rats were markedly reduced in TG2-shRNA-treated animals, reaching levels comparable to control (nonarthritic) joints (Figure 2A, B). In parallel, immunostaining by using an antibody directed against fragments of type II collagen showed that the extensive zones of collagen degradation observed in the joints of CIA rats were similarly reduced in tissues of CIA rats treated with TG2-shRNA (Figure 2C). Collectively, these data suggested that TG2 is the main enzyme involved in cross-linking activity and cartilage degradation in CIA rats.

Because it is well established that FLSs play a major role in cartilage degradation, we next assessed the levels of TG2 expression and TGase activity in rat synovial cell cultures. Immunoblot analysis of intracellular, secreted, and matrix-bound TG2 revealed an absence of consistent differences with respect to intracellular and secreted TG2 between experimental and control rats (data not shown). In marked contrast, a reproducible increase in the deposition of TG2 in the laid-down ECM was found in arthritic FLS cultures compared with control FLSs (Figure 3A). In addition, A-FLSs cultured on a gelatin matrix displayed consistently higher TGase activity (averaging 2.5-fold) than did C-FLSs (Figure 3B and Additional file 4, Figure S4A).

Recent evidence indicated that the ECM degradation activity of rat A-FLS in vitro and in arthritic joints is associated with the formation of ECM-degrading cell structures, which are reminiscent of invadopodia produced by cancer cells or podosomes in osteoclasts [10, 11]. We therefore investigated whether the modulation of TG2 in A-FLSs affected their capacity to form ECM-degrading invadopodia. Confocal microscopy analysis revealed that 81% ± 4.3% of invadopodia structures, identified by clusters of actin and p-cortactin, two bona fide markers of invadopodia [37], colocalized with TG2 (Figure 3C). Of significance, knockdown of TG2 caused a drastic decrease in the percentage of cells forming invadopodia, as well as in the extent of ECM degradation induced by invadopodia (Figure 3D, E). In contrast, knockdown of FXIIIa, a parental TG, which is also expressed in synovial cells (Figure 3F), failed to reduce invadopodia formation or ECM degradation (Figure 3D, E). Immunoblotting experiments confirmed the efficiency of TG2 and FXIIIa knockdown (Figure 3F). Furthermore, TG2 knockdown resulted in a strong decrease in total TGase activity, whereas knockdown of FXIIIa had a milder effect (Figure 3G and Additional file 4, Figure S4B), confirming that TG2 was responsible for the majority of the TGase activity in A-FLSs. Together, these results indicated that TG2 was involved in the formation of ECM-degrading invadopodia in arthritic FLSs.

We next examined whether modulation of the TGase activity of TG2 affected the capacity of FLSs to form invadopodia. For this, TGF-β, a potent inducer of TG2 and its TGase activity [21, 38], and DTT, a disulfide-bond reducer that unfolds TG2, inducing its TGase activity [39], were added to C-FLS at the onset of invadopodia assays. Results showed that each treatment induced a strong increase in cross-linking activity (2.2- and 2.5-fold increases, respectively) as compared with untreated C-FLS, whereas the relative levels of TG2 protein were only mildly affected (Figure 4A and Additional file 4, Figure S4C). In contrast, cystamine, a competitive inhibitor of TGase [40], strongly reduced the cross-linking activity of C-FLS (Figure 4A and Additional file 4, Figure S4C). These TGase inhibitors/activators were also tested in invadopodia assays. TGF-β and DTT treatment induced 4.3- and 3.1-fold increases, respectively, in the percentage of control synovial cells forming invadopodia (Figure 4B). The stimulatory effect of TGF-β was blocked by shRNA-dependent TG2 depletion (Figure 4B). In addition, preincubation of A-FLSs with cystamine, the TGase inhibitor KCC-009, or the TG2-specific inhibitor Z-DON, induced a significant dose-dependent decrease in the formation of invadopodia without affecting cell viability (Figure 4C and data not shown). These results indicated that modulation of the TGase activity of TG2 influenced invadopodia formation in FLSs. Further to evaluate the role of the TGase activity of TG2, C-FLSs were transfected with a TGase-defective mutant (TG2-W214A), which is efficiently secreted (data not shown), or wild-type TG2 (TG2 WT), and assayed for invadopodia formation. Overexpression of TG2-W214A resulted in weak invadopodia formation. Furthermore, this mutant TG2 failed to promote ECM degradation, in contrast to control cells harboring WT TG2 (Figure 4D, E). These results strongly suggest that TG2 acted through TGase activity for the regulation of invadopodia formation in rat FLSs.

To determine whether our findings could be applied to human arthritis, studies were performed by using synovial cells derived from control (arthritis-free; C-FLSs) and rheumatoid arthritis patients (RA-FLSs). Treatment of C-FLSs with either TGF-β or DTT caused 2.7- and 6.5-fold increases, respectively, in invadopodia formation (Figure 4F). In contrast, silencing TG2 expression by using TG2-shRNA or inhibition of TGase activity with cystamine in synovial cells from RA patients resulted in marked decreases in invadopodia formation, reaching levels similar to those found in control FLS (Figure 4G). These data led us to conclude that TG2 acted, at least in part, through its TGase activity to regulate invadopodia formation in rat and human synovial cells.

To gain insight into the mechanism by which TG2 influences invadopodia formation and ECM degradation, we investigated the ability of TG2 to affect the bioavailability of TGF-β. TG2 has been shown to enhance TGF-β production and to contribute to TGF-β bioactivation by favoring the binding of latent TGF-β-binding protein-1 to ECM, which is a prerequisite to TGF-β release [22, 41]. Overexpression of TG2 in C-FLSs resulted in a 4.3-fold increase in the production of TGF-β (Figure 5A), whereas a 4.6-fold increase in invadopodia formation was observed (Figure 5B) as compared with control cells transfected with an empty vector. The invadopodia-promoting effect of TG2 was completely inhibited by TGF-β1-3 neutralizing antibodies (Figure 5B). A similar inhibition was observed by treating TG2-overexpressing cells with the TGF-β R1 inhibitor, LY364947 (Figure 5C). In addition, silencing TG2 expression in arthritic FLSs inhibited TGF-β production (Figure 5D) and reduced TGF-β activity, as measured by using the p3TP-LuX reporter construct (Figure 5E). Taken together, these results indicate that endogenous TGF-β was part of the mechanism by which TG2 influenced invadopodia formation in arthritic synoviocytes. They further suggest that TGF-β expression and activity were regulated by TG2 in these cells.