Genetic deficiency of Wnt5a diminishes disease severity in a murine model of rheumatoid arthritis

All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC), Boston University School of Medicine (BUSM). Wnt5a whole body inducible cKO mice (Wnt5a cKO) were generated as previously described [10] by selective breeding of the Wnt5a floxed mice [21] with the UBC cre promoter mice (Jackson Labs, Bar Harbor, ME, USA, stock number 007001). Female littermate Wnt5a floxed cre-negative mice (control) and cre-positive mice (Wnt5a cKO) were used for this study. Wnt5a recombination was induced in the mice at 6 weeks of age by five daily intraperitoneal injections of tamoxifen delivered in corn oil (80ug tamoxifen per g body weight each day). Both control and Wnt5a cKO mice were treated with tamoxifen. Baseline mice (at least four per genotype) were euthanized following 2 weeks of tamoxifen wash out. Tissue was collected and processed as described in the following section. Mice were used in the arthritis model following 2 weeks of tamoxifen wash out (n = 13 per genotype for RA studies). Female C57BL6J (Jackson Laboratories) mice 10 − 12 weeks of age were used for the in vitro fusion assays (n = 3).

Rheumatoid arthritis development was induced in control and Wnt5a cKO mice using the K/BxN serum transfer model of arthritis as previously described [22]. Briefly, serum from K/BxN mice (the offspring of mice expressing the KRN T cell receptor transgene and the major histocompatibility complex (MHC) class II molecule A(g7), which develop spontaneous arthritis) was isolated at 8 weeks of age. Passive transfer of arthritis was induced in control and Wnt5a cKO mice by intraperitoneal injection of 150 μl of K/BxN serum at day 0 (d0) and d2. Severity of arthritis was monitored by two measures prior to serum injection and every other day for 28 days. The clinical score of each joint was determined using a well-established scoring metric [22‐24]. The reported clinical score is the summation of the scores for all four appendages. Ankle thickness was measured at the thickest point of the ankle using digital Vernier calipers (Fine Science Tools). Significance for the clinical score and ankle thickness measurements was determined using the statistical package in Prism GraphPad using two-way analysis of variance (ANOVA). Significance was determined using the Sidak post-hoc test. Tissues were collected at d7 (n = 6 per genotype) or d28 (n = 7 per genotype) following induction of arthritis. Serum was collected by cardiac puncture after euthanasia by carbon dioxide inhalation. Paw joints were snap frozen for mRNA isolation and ankle joints were fixed in 10% formalin and decalcified in 0.375 M EDTA.

Paw joints were pooled for each individual animal prior to mRNA extraction. The joints were cleaned of skin, homogenized with Qialyzer (Qiagen) and mRNA was isolated with the Qiagen RNeasy Fibrous tissue kit according to the manufacturer’s directions. cDNA was transcribed from 1 μg of RNA using High Capacity cDNA Kit (Applied Biosystems) and quantitative PCR (qPCR) was performed using SYBR-based detection (primer sequences are shown in Table 1). Fold changes were calculated after normalization of the average hypoxanthine guanine phosphoribosyl transferase (HPRT) and beta-actin relative to baseline mice. P values ≤0.05 from the t test calculated using GraphPad Prism were considered significant.

Paraffin-embedded tissue sections were stained with hematoxylin and eosin (H&E). Disease severity was analyzed by a pathologist (KGH) blinded to the genotypes as previously described [25] (n = 6 per genotype). Clinical score, cartilage destruction, bone erosion, inflammation, polymononuclear (PMN) infiltration, extra-articular inflammation, lymphocyte infiltration, pannus formation and synovial lining average were evaluated as previously described [23, 25‐28]. Tartrate acid resistant phosphatase (TRAP) staining was performed as specified by the manufacturer (Sigma Kit) (n = 6 per genotype at d7 and at least 4 per genotype for baseline analysis). Stained sections were visualized on a Keyence BZ-X700 and analyzed with the on-board image processing software to identify total TRAP+ area in at least four fields per mouse in both ankle joints.

L control cells (ATCC, CRL-2648) and L Wnt5a cells (ATCC, CRL-2814) were used to generate controls and Wnt5a conditioned media as previously described [29]. L cells were maintained in DMEM with 10% FBS and pen/strep/Lglut. Conditioned medium for the fusion assays were generated in complete alpha-MEM with 10% FBS and pen/strep/Lglut. This conditioned medium was diluted 1:4 with fresh complete alpha-MEM medium for the fusion assays.

BMDM were isolated from female C57BL6J mice (Jackson Laboratories) and cultured in alpha-MEM with 10% FBS and pen/strep/Lglut. Following differential plating to purify monocyte cells in the presence of macrophage colony-stimulating factor (MCSF) (100 ng/mL), 10^5 cells per well were plated in 24 well plates in the presence of diluted L cell control or diluted Lcell Wnt5a-treated medium. BMDM were induced to fuse with the stimulation cocktail (MCSF (50 ng/mL) and RANKL (50 ng/mL)), which was replenished at d2 and d4. RNA was isolated at d2 and d5 using the miRNeasy Micro Kit (Qiagen) and 1 μg of RNA was transcribed using the Quantitect kit (Qiagen). SYBR chemistry was used to detect transcripts and these were analyzed on a ViiA7 using the primers listed in Table 1. Significance was determined in GraphPad Prism using two-way ANOVA with the Tukey post-hoc test, considering p values ≤0.05 as significant. To enumerate osteoclasts, d5 cultures were stained for TRAP (Sigma) and co-stained with 4',6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific). Cultures were visualized using a Keyence microscope. Osteoclasts were identified as TRAP+ multinucleated cells (3 nuclei or more). Total nuclei, total TRAP+ cells, total osteoclasts and total nuclei in osteoclasts were enumerated in at least six random high powered fields per condition per experiment. The osteoclast frequency per 1000 cells was calculated as the ratio of total osteoclasts per high-powered field (HPF) divided by total nuclei in the same HPF. The percentage of cells in osteoclasts is the total nuclei in osteoclasts per HPF divided by the total nuclei in the same image. Statistical significance for enumeration tests was performed using Prism Software and the unpaired t test and considering significance at p ≤ 0.05. All in vitro studies were performed three times. Enumeration studies were performed in duplicate three times.

Due to the association of Wnt5a with clinical arthritis, we sought to determine its role in the murine STIA model of RA. In apparent contrast to clinical arthritis [18], Wnt5a transcript levels were reduced at d7 after RA induction (Additional file 1: Figure S1). However, active RA induced the levels of the Wnt5a co-receptor ROR2 mRNA at this time point (Additional file 1: Figure S1). To further evaluate the role of Wnt5a in the development of RA, a murine model of Wnt5a deficiency was subjected to the STIA model. Using primers that hybridize within the floxed Exon 2 of Wnt5a (Additional file 1: Figure S2A), the Wnt5a cKO mice exhibited >90% decrease in Wnt5a in the paws of Wnt5a cKO mice after 2 weeks of tamoxifen-induced ablation (Additional file 1: Figure S2A) compared with littermate controls. Baseline measures in healthy Wnt5a cKO mice did not demonstrate alterations in inflammatory markers (Additional file 1: Figure S2B), in ankle morphology (Additional file 1: Figure S3), or in osteoclast activation (Additional file 1: Figure S4). Following STIA induction of RA, there was attenuated development of arthritis in the Wnt5a cKO mice in terms of ankle thickness, which plateaued below the thickness developed in the littermate controls and resolved in a similar timeframe (Fig. 1a). In a separate cohort of animals, the arthritic process was examined in control and Wnt5a cKO mice at the earliest peak of inflammation (d7). Measurement of the disease severity was performed by a pathologist blinded to genotype using previously established histopathologic scoring metrics [23, 24], using H&E stained sections from control and Wnt5a cKO mice at d7 (Fig. 2a). The Wnt5a cKO mice displayed statistically significant reduction in overall disease severity (overall H&E score) and this cohort exhibited a significant reduction in clinical score (Fig. 2b). Histologic analysis of the arthritic joints also revealed that Wnt5a cKO mice exhibit significant reductions in some parameters of inflammation, including measures of reduced extra-articular inflammation and PMN infiltration (Fig. 2b). Although not statistically different, trends of reduced pannus formation, synovial lining average and bone erosion were also observed in the Wnt5a cKO mice compared to control mice (Fig. 2b and Additional file 1: Figure S5). Numerous studies have reported that Wnt5a promotes cytokine expression [13‐16]. However, only trends of decreased TNFα, IL6, and IL1β were observed in Wnt5a cKO at day 7 following RA induction (Additional file 1: Figure S6A), which were not statistically significant.

We next evaluated the osteoclast activity in the arthritic ankles from control and Wnt5a cKO mice. TRAP staining (Fig. 3a) demonstrated robust activity in the control mice, with obvious staining at multiple destruction pits, whereas Wnt5a cKO mice displayed a markedly reduced TRAP-positive area as quantified in Fig. 3b. Consistent with these findings, decreased levels of TRAP and cathepsin K (CTSK) mRNA were detected in the arthritic paws of Wnt5a cKO mice (Fig. 3c), but there were no changes in the calcitonin receptor (CAL) or carbonic anhydrase 2 (CAR2) mRNA in Wnt5a cKO mice (Additional file 1: Figure S6B). There was a trend toward lower expression of the osteoclast differentiation signal RANKL and reduction in the expression of MCSF observed in the arthritic paws of Wnt5a cKO mice (Fig. 3c). Consistent with their roles both in inflammation and in osteoclast activity, levels of MMP2 and MMP9 mRNA were significantly decreased in the arthritic paws of Wnt5a cKO mice (Fig. 3c).

To attain a deeper understanding of the mechanism of Wnt5a action in the process of osteoclastogenesis, in vitro analyses were performed to determine whether Wnt5a has an intrinsic ability to promote osteoclast fusion. MCSF-dependent BMDM were stimulated to fuse in the presence of L cell medium conditioned by cells that were engineered to overexpress Wnt5a (Wnt5a) or control media (conditioned by L cells without Wnt5a overexpression). TRAP-positive (dark color) multinucleated osteoclasts formed after 5 days of RANKL stimulation (Fig. 4a). Osteoclast fusion was stimulated by Wnt5a, both in terms of total incidence (osteoclast frequency per 1000 cells) and in the percent of nuclei fused into osteoclasts (Fig. 4b). In contrast, Wnt5a did not significantly change the nuclei per osteoclast total or total nuclei in osteoclasts (Additional file 1: Figure S7).

Transcripts were analyzed in fusing BMDM cultures after 2 and 5 days of RANKL stimulation to further evaluate the impact of Wnt5a on osteoclast fusion. Wnt5a mRNA levels were not altered during the fusion process (Additional file 1: Figure S8). Treatment with Wnt5a did not significantly alter the expression of osteoclast genes (CTSK or TRAP) or an inhibitor of osteoclastogenesis (OPG) (Additional file 1: Figure S8). However, Wnt5a promoted the expression of DCSTAMP (Fig. 4c), a cell surface molecule that is essential for osteoclast fusion [30]. Consistent with observations in the STIA model, fusing macrophages treated with Wnt5a expressed increased levels of MMP9 but not MMP2 (Fig. 4d). In contrast, Wnt5a did not affect the expression of other nuclear factor activating T cells 1c (NFATc1) downstream genes, including MMP14, CLCN7, CAR2, RCAN2 and OSCAR, or NFATc1 itself (Additional file 1: Figure S9).