Wnt antagonism without TGFβ induces rapid MSC chondrogenesis via increasing AJ interactions and restricting lineage commitment

Summary Human mesenchymal stem cells (MSCs) remain one of the best cell sources for cartilage, a tissue without regenerative capacity. However, MSC chondrogenesis is commonly induced through TGFβ, a pleomorphic growth factor without specificity for this lineage. Using tissue- and induced pluripotent stem cell-derived MSCs, we demonstrate an efficient and precise approach to induce chondrogenesis through Wnt/β-catenin antagonism alone without TGFβ. Compared to TGFβ, Wnt/β-catenin antagonism more rapidly induced MSC chondrogenesis without eliciting off-target lineage specification toward smooth muscle or hypertrophy; this was mediated through increasing N-cadherin levels and β-catenin interactions—key components of the adherens junctions (AJ)—and increasing cytoskeleton-mediated condensation. Validation with transcriptomic analysis of human chondrocytes compared to MSCs and osteoblasts showed significant downregulation of Wnt/β-catenin and TGFβ signaling along with upregulation of α-catenin as an upstream regulator. Our findings underscore the importance of understanding developmental pathways and structural modifications in achieving efficient MSC chondrogenesis for translational application.


INTRODUCTION
Cartilage and joint diseases are one of the most common clinical conditions affecting quality of life worldwide because of increases in the two major risk factors of age and obesity. Current treatments for arthritic diseases are largely limited to symptom palliation using analgesics for mild cases, and immunosuppressants and/or surgery for more serious cases. 1 However, these treatments are not disease-modifying or curative; moreover, cartilage is unable to regenerate. Although more novel treatment using autologous cartilage tissue or chondrocytes are being tested, the rarity of the cell/ tissue source, ex vivo culture difficulties, and low cell viability after transplantation have resulted in limited success with these methods. 2 Thus, continued investigation into more efficacious methods of regenerating cartilage is ongoing to find curative treatments for these common and debilitating diseases.
Human multipotent mesenchymal stem cells (MSCs) are versatile somatic stem cells with immunomodulatory properties. First isolated in adult bone marrow (BM), MSCs have subsequently been found in numerous post-natal organ/tissues 3 as well as directly differentiated from pluripotent stem cells such as human embryonic stem cells (ESCs) 4,5 and induced pluripotent stem cells (iPSCs). 6,7 MSCs readily differentiate into chondrocytes, osteoblasts, adipocytes, and fibroblasts, and their easy accessibility compared to many other cell types including chondrocytes render them ideal for use in cartilage-related diseases. However, chondrogenic differentiation efficiency is low because of the requisite cumbersome 3-dimensional (3D) pellet culture and the relative lack of knowledge on molecular mechanisms involved in chondrogenesis, compared to osteogenesis and adipogenesis. The most common factors used to induce MSC chondrogenesis are TGFb1 and TGFb3, two pleomorphic growth factors, but information on mechanisms involved for either of these factors are surprisingly scarce despite decades of use. 8 Moreover, TGFb is well known to induce fibrosis, as well as ossification in cartilage. 9,10 Thus, there clearly is a need for more precision and efficiency in achieving MSC chondrogenesis for therapeutic application.

Wnt/b-catenin antagonism significantly enhanced MSC chondrogenesis whereas agonism resulted in the opposite and upregulation of the master osteogenic transcription factor Runx2
To investigate the effects of Wnt modulation on MSC chondrogenesis, we treated 3D pellet-cultured human MSCs-including iPSC-MSCs, ESC-MSCs and BM-MSCs-in a standard complete chondrogenic induction medium (ChM) containing TGFb3 along with a small molecule Wnt/b-catenin agonist CHIR which inhibit GSK3b to disrupt the b-catenin destruction complex, or antagonist XAV, a tankyrase inhibitor which stabilizes the b-catenin destruction complex. TGFb3 was utilized in ChM because it has been found to possess higher chondrogenic potential than TGFb1. 18 We found that in ChM conditions, Wnt inhibition by XAV increased condensation of the pellets whereas agonism by CHIR decreased pellet integrity in all 3 sources of MSCs at Day 20. Quantification of glycosaminoglycans (GAGs) showed a significant increase in the expression of this structural extracellular matrix of cartilage with Wnt/b-catenin antagonism, whereas the opposite was seen with Wnt/b-catenin agonism ( Figure 1A). Because iPSC-MSCs can be patient-specific as well as continually derived, we focused on using this source in further studies. By observing morphological characteristics, including pellet size and integrity, and quantification of GAGs ( Figure 1B), we found a dose-dependent effect of Wnt/b-catenin modulation on MSC chondrogenesis. Analysis of chondrogenic gene expression levels after Wnt/b-catenin antagonism at Day 3 showed significant upregulation of two key chondrogenic genes collagen 2A1 (COL2A1) and aggrecan (ACAN) but not SOX9; Wnt/b-catenin agonism, on the other hand, resulted in significant downregulation of all 3 genes ( Figures 1C-1E). Of interest, Wnt/b-catenin antagonism decreased expression of the master osteogenic transcription factor RUNX2 whereas agonism showed the opposite effect ( Figure 1F). These findings collectively demonstrate that Wnt antagonism significantly enhances MSC chondrogenic differentiation whereas agonism results in the opposite effect and upregulates the master osteogenic transcription factor RUNX2.
TGFb rapidly increased alpha smooth muscle actin (aSMA) and RUNX2 expression in MSCs during chondrogenic induction Although TGFb1 and TGFb3 have long been used for MSC chondrogenesis, it is also well established that TGFb can induce smooth muscle differentiation 9 and upregulate an osteogenic program. 19 By analyzing the transcriptomic data of human primary BM-MSCs and smooth muscle cells from public database (GSE128949 for human primary BM-MSCs; GSE109859 for human primary smooth muscle cells) with Ranked Gene Set Enrichment Analysis (GSEA), gene set of Signaling by TGFb Family Members was positively enriched in human BM-MSCs compared to smooth muscle cells with normalized enrichment score (NES) equals to 2.17 and the nominal p value less than 0.001 ( Figure 2A). Relative changes in mean expression levels also showed that the core components in TGFb signaling (TGFB1, TGFB3 TGFBR1, TGFBR2, SAD2 and SMAD3) as well as aSMA (ACTA2), a marker of smooth muscle differentiation and also fibrosis, 20 were all significantly upregulated in human primary smooth muscle cells (SMC) compared to MSC ( Figure 2B). These results indicated that TGFb signaling is positively correlated with the smooth muscle lineage. In addition to smooth muscle lineage, TGFb/BMP signaling is also known to induce MSC osteogenic differentiation. 21 Figure 2C), with quantitative fluorescent intensity demonstrating significantly higher aSMA expression (up to 3-fold) in TGFb-treated groups compared to ChBM ( Figure 2D). RUNX2 mRNA expression was also significantly upregulated in TGFb1and TGFb3-treated groups on Day 3 compared to ChBM ( Figure 2E). These results indicate that TGFb rapidly induce non-chondrogenic lineages including smooth muscle and ossification during MSC chondrogenesis.

Wnt/b-catenin antagonism alone induced more rapid MSC chondrogenesis than TGFb
Because Wnt/b-catenin antagonism strongly enhances MSC chondrogenesis and TGFb agonism significantly induce other non-chondrogenic lineage markers during this process, we examined the possibility of replacing TGFb completely with Wnt/b-catenin antagonism to achieve more specific MSC chondrogenic differentiation. After 20 days of chondrogenic differentiation, MSC pellets treated with either TGFb3 or XAV formed well-condensed spheres and strong Alcian blue staining compared to ChBM, whereas pellets treated with CHIR did not undergo condensation but became disintegrated at the end of the culture time period ( Figure 3A, top panel). Strikingly, we found that GAG production as evidenced by Alcian blue staining was more rapidly induced in pellets cultured with XAV than TGFb3 at Day 10, approximately halfway through the full differentiation time period ( Figure 3A, bottom panel). This was verified with quantification of GAG production at Day 10 in which significant increases in MSC pellets cultured with XAV was seen, not only when compared to ChBM culture but also to TGFb3 treatment, whereas CHIR treatment resulted in minimal GAG production ( Figure 3A, right panel). Moreover, induction of chondrogenesis by Wnt antagonism at this earlier time point was not further enhanced with TGFb3 supplementation, implicating that iScience Article there is no synergistic effect ( Figure S1). Histological cross-sections of Day 10 pellets also revealed stronger Alcian blue staining representing GAG production, as well as a more solid and homogeneous pellet structure in XAV-treated conditions, compared to TGFb-treated or ChBM cultured conditions ( Figure S2). Immunohistochemical staining for chondrogenic markers type II collagen (COL2) and aggrecan (ACAN) also showed stronger signals in XAV-treated conditions compared to TGFb-or ChBM-conditions (Figure S3). We further validated these findings using micromass culture, which is a more convenient method to induce chondrogenic differentiation in standard 2D culture condition and useful for in vitro highthroughput drug screening. 22 Similar to 3D pellet culture, micromass culture of MSCs with XAV treatment at Day 10 demonstrated better structural condensation and stronger Alcian blue intensity compared to ChBM culture and even TGFb3-treated culture, whereas CHIR treatment resulted in minimal evidence of any micromass structure or Alcian blue staining. Quantitation of GAG production was in line with the morphologic findings, with XAV treatment resulting in the highest production of GAG ( Figure 3B). XAV treatment also most significantly upregulated expression of both chondrogenic genes COL2A1 and ACAN, whereas TGFb3 treatment did not increase expression of either genes and CHIR treatment resulted in minimal expression of either genes at Day 3 ( Figures 3C and 3D). To further ascertain that the outcome is not specific to the utilized antagonist and/or agonist, iCRT-3, a non-tankyrase Wnt antagonist 23 and lithium chloride (LiCl) a well-studied Wnt/b-catenin pathway agonist 24 were additionally used ( Figure S4). At Day 3 and Day 5, pellets cultured with either XAV or iCRT-3 resulted in condensed and spherical morphology while progressive disintegration was seen of pellets cultured with either CHIR or LiCl ( Figure S4A). Moreover, Alcian blue staining and quantification at the early time point of Day 6 showed that both XAV and iCRT-3 already induced significant increases in GAG production not only over baseline/control levels but also at a level similar to or better than with TGFb3 ( Figure S4B). These results strongly indicate that iScience Article antagonism of the Wnt/b-catenin pathway rather than antagonist-specific effects is responsible for the rapid induction of MSC chondrogenesis.
To validate our in vitro data, we performed subcutaneous transplantation into wild-type mice of mouse BM-MSCs cultured in ChBM medium with injections of various modulators alone (TGFb3, XAV, or CHIR) every three days for 20 days ( Figure 3E). Histological analyses of in vivo differentiated samples demonstrated that the highest level of GAG production was in the XAV-treated group; moreover, Wnt/b-catenin antagonism significantly increased GAG production compared to TGFb3 agonism ( Figure 3F). These results indicated that Wnt inhibition significantly enhances MSC chondrogenesis in vivo. All these results demonstrate that replacement of TGFb with Wnt/b-catenin antagonism achieves more rapid and specific MSC chondrogenesis. We further performed gene expression array of two iPSC-MSCs and one bone marrow-derived MSC with XAV-or TGFb-based chondrogenic induction. GSEA results revealed that chondrogenic pathways are positively enriched in both conditions, whereas gene sets with regards to bone development and smooth iScience Article muscle differentiation are negatively enriched in XAV-compared to TGFb-induced chondrogenesis; moreover, Wnt antagonism does not enrich for adipogenesis, demonstrating the strong lineage-specificity of inhibiting this pathway for chondrogenesis. As for adipo-related lineage, we found no significant difference in the enrichment of Adipose tissue development gene set, indicating that Wnt antagonism is indeed highly specific for MSC chondrogenesis ( Figure S5). Collectively, these results demonstrate that compared to TGFb, MSC chondrogenesis using Wnt antagonism is highly specific and not result in off-target lineage commitment toward smooth muscle lineage or hypertrophy.
Wnt/b-catenin antagonism but not TGFb agonism during MSC chondrogenesis decreased canonical Wnt/b-catenin transcriptional activity including RUNX2 expression The significant enhancement of MSC chondrogenesis with Wnt/b-catenin antagonism alone led us to investigate the role of the canonical Wnt/b-catenin pathway during the differentiation process. As a transcription factor, b-catenin undergoes translocation from the cytoplasm to the nucleus when activated. Using immunofluorescent staining and quantification, we found that pellet-cultured MSCs treated with XAV had the lowest nuclear b-catenin levels compared to control ChBM culture or TGFb3 treatment, whereas CHIR treatment dramatically increased nuclear b-catenin levels ( Figures 4A and 4B), which was expected because CHIR is known to strongly induce b-catenin transcriptional activity. 25 Assessment of b-catenin pathway downstream gene expression levels demonstrated that after XAV treatment, expression of AXIN2 and iScience Article TCF7, two well-established b-catenin-activated downstream genes, and the osteogenic master transcription factor RUNX2 were all significantly decreased, whereas CHIR treatment strongly increased expression of these genes as expected (Figures 4C and 4D). Surprisingly, TGFb3 treatment resulted in increased expression of not only RUNX2 but AXIN2 as well, implicating a weak agonistic effect of this factor on canonical Wnt/b-catenin pathway. These findings demonstrate that inhibition of Wnt/b-catenin transcriptional activity is critical for MSC chondrogenesis, and that TGFb treatment may not be optimal during this process because of its weak agonism for the pathway.
Wnt/b-catenin antagonism but not TGFb agonism increased N-cadherin expression and interactions with b-catenin at AJs as well as enhancing actin cytoskeleton-mediated condensation In addition to its role as a transcription factor, b-catenin is also a component of the AJ, a key structure of cell-cell adhesion which is a critical aspect during cartilage condensation. 26 To investigate how Wnt/b-catenin modulation during MSC chondrogenesis affect AJs, we performed immunofluorescent staining for N-cadherin, a major AJ component in mesenchymal cell types like MSCs which is also essential during chondrogenesis. 27 We found N-cadherin expression to be strongly and most significantly upregulated in MSC micromass culture with XAV treatment after 24 h, compared to all other conditions ( Figures 5A and  5B). To further investigate whether there were increased interactions between b-catenin and the increased N-cadherin levels induced by Wnt/b-catenin antagonism, we performed proximity ligation assay (PLA) between these two molecules. We found that only XAV treatment during MSC chondrogenesis could significantly increase N-cadherin/b-catenin interactions as evidenced by increased PLA signals (Figures 5C and  5D). Of interest, whereas CHIR treatment resulted in the strongest expression of b-catenin ( Figure S6), there was minimal PLA signal detected, indicating little interaction/colocalization between N-cadherin and b-catenin ( Figures 5C and 5D). To validate the role of N-cadherin in Wnt antagonism-induced MSC chondrogenesis, N-cadherin blocking antibody was applied. Alcian blue staining at Day 6 showed that blocking of N-cadherin interactions significantly decreased the enhancement of chondrogenesis induced by XAV treatment ( Figure S7). These findings demonstrate that Wnt inhibition enhances MSC chondrogenic differentiation through increasing N-cadherin levels and interaction/colocalization between N-cadherin and b-catenin at AJs.
To validate interactions between N-cadherin and b-catenin in MSC chondrogenic differentiation, we first examined the N-cadherin pathway using the Pathway Interaction Database (PID_NCADHERIN_PATHWAY) and found that with N-cadherin (CDH2), b-catenin (CTNNB1) is a core participant within the signaling pathway, with numerous components of the AJ: a-catenin (CTNNA1), p120/d-catenin (CTNND1), and plakoglobin (JUP). Moreover, the three major GTPases responsible for actin cytoskeleton organization-RhoA, CDC42, and Rac1-are all found as downstream effectors in the signaling pathway ( Figure 5E), demonstrating that actin cytoskeleton organization is highly regulated by the N-cadherin signaling pathway. To validate the bioinformatics analyses, during MSC chondrogenesis we treated with cytochalasin D (CytoD), an inhibitor of actin polymerization, which significantly disrupted the condensation process with the pellets completely disintegrated by Day 10. In contrast, cells treated with XAV+CytoD partially maintained the pellet structure and chondrogenic differentiation ( Figures 5F and 5G). All these results suggest that Wnt/b-catenin antagonism promotes MSC chondrogenic differentiation by increasing N-cadherin expression and its interactions with b-catenin at AJs as well as enhancing actin cytoskeleton-mediated condensation.

Significant downregulation of Wnt/b-catenin and TGFb-related pathways in transcriptomes of human primary chondrocytes compared to osteoblasts
To assess the physiological relevance of our findings, we performed bioinformatics analyses using human primary BM-MSC, chondrocyte, and osteoblast transcriptome data from public database (GSE108186 for human primary BM-MSCs; GSE68038 for human primary chondrocytes; and GSE121892 for human primary osteoblasts). Initial analyses using principal component analysis (PCA) demonstrated that chondrocytes, osteoblasts, and MSCs are three highly distinct populations ( Figure 6A). We then evaluated the transcriptomic changes of MSC lineage specification toward chondrogenic and osteogenic lineages by comparing mRNA expression profiles of chondrocytes and osteoblasts to MSCs (Chondro versus MSC and Ostb versus MSC, respectively). To uncover genes and pathways relevant to cartilage maintenance but not hypertrophy and ossification, we compared the transcriptomic profiles of chondrocytes to osteoblasts (Chondro versus Ostb) ( Figure 6B). GSEA based on major developmental pathways 28 Figure 6D, right two panels). We then analyzed by cell type the mRNA expression levels of specific genes in the canonical Wnt/b-catenin pathway, during chondrogenesis, as well as during osteogenesis for further validation of the GSEA results ( Figure 6E). To better assess the directionality of expression for each gene, we compared the changes of Robust Multi-array Average (RMA) levels of specific genes in human chondrocytes to that of MSCs (Chondro versus MSC) and osteoblasts (Chondro versus Ostb) ( Figure 6F). The results showed that the osteogenic/hypertrophy genes RUNX2, COL1A1 and ALPL were significantly downregulated in primary chondrocytes compared to MSCs, and RUNX2, ALPL, SPP1, COL1A1 and COL10A1 were downregulated in chondrocytes compared to osteoblasts. In contrast, chondrogenic genes including SOX5, SOX6, COL2A1 and ACAN were more highly expressed in chondrocytes than MSCs or osteoblasts, with the surprising exception of SOX9 which was more highly expressed in MSCs and osteoblasts than in chondrocytes. GSK3B, an inhibitory gene of the Wnt/b-catenin pathways, were more highly expressed in primary chondrocytes, whereas Wnt/b-catenin downstream genes AXIN1, AXIN2 were less expressed in chondrocytes compared to MSCs or osteoblasts.
Using Upstream Regulator Analysis in Ingenuity Pathway Analysis (IPA) which can provide a more causal relationship between transcription factors/master regulators and downstream pathways, 29 we found that TGFb1, Smad2/3, WNT3A, TCF4, BMP2, and TGFBR2 were all predicted to be downregulated transcription factors/master regulators in human primary chondrocytes compared to MSCs. Furthermore, TGFb1, CTNNB1, and WNT3A were predicted to be downregulated regulators in human primary chondrocytes compared to osteoblasts. Specifically, a-catenin-a core components of the AJ-was predicted as an upregulated regulator in chondrocytes compared to both MSCs and osteoblasts ( Figure 6G). All these bioinformatic results were in line with our data demonstrating participation of AJs in WNT/b-catenin antagonism-induced chondrogenesis ( Figure 5). Additional validation using another recent published dataset of human primary chondrocyte transcriptomes from various developmental stages 30 also demonstrated that both TGFb and Wnt signaling pathways are negatively enriched in adult articular chondrocytes compared to limb bud pre-chondrocytes ( Figure S8). Overall, these results underscore that the Wnt/b-catenin pathway as well as the TGFb pathway is downregulated in primary human chondrocytes compared to osteoblasts and MSCs.

DISCUSSION
MSC therapy likely offers the best possibility of a curative treatment for cartilage and joint diseases, which currently lacks such significant disease-modifying treatments. However, the relatively more difficult 3D pellet differentiation protocol and the need for using a protein-based growth factor, TGFb, continue to be obstacles for robust clinical applications. TGFb is a cytokine with complex functions, and has pleomorphic roles in specification of multiple mesenchymal lineages, not only for chondrogenesis but also osteogenesis. 10,19,21 In many transgenic mouse studies, Wnt inhibition has been clearly shown to promote cartilage development 13,31 whereas Wnt activation induces osteogenesis. 12,32 Although some early in vitro stem cell differentiation studies did not consistently find Wnt inhibition to promote chondrogenesis, these reports utilized biologically derived Wnt inhibitory ligands which are known to have off-target effects as well as potency issues. 33,34 In addition, some studies utilized non-MSCs which may not be the appropriate iScience Article system to investigate chondrogenic lineage specification through Wnt/b-catenin. 33 Using highly potent small molecule agonists and antagonists of the Wnt/b-catenin pathway, we demonstrated that Wnt/b-catenin antagonism efficiently induce in vitro chondrogenesis in human MSCs from multiple sources, and in vivo using murine BM-MSCs. Moreover, in comparison with TGFb3, the most potent member of the family for induction of chondrogenesis, Wnt/b-catenin antagonism more rapidly induced chondrogenesis without inducing other non-chondrogenic, off-target, lineages. Our findings are further supported by transcriptome analysis of primary human MSCs, chondrocytes, and osteoblasts, which demonstrated that the Wnt/b-catenin and TGFb family signaling pathways were downregulated in chondrocytes relative to the other 2 cell types ( Figures 6C-6G). Our findings therefore strongly implicated that Wnt/b-catenin antagonism using potent small molecules can efficiently and more precisely promotes MSC chondrogenesis than TGFb. Our report of replacing TGFb agonism with Wnt antagonism increases induction efficiency and allows for the use of more stable and less costly small molecules rather than protein-based factors, and limits off-target lineage commitment which could possibly better preserve the chondrogenic phenotype. Because cost and maintaining lineage in the transplanted cell continue to be major obstacles in cartilage cell therapy, 35 our findings here could have important translational implications.
We found that the efficient chondrogenic commitment mediated by Wnt/b-catenin antagonism involved strong upregulation of N-cadherin expression and N-cadherin/b-catenin interaction at the AJ which enhanced pellet condensation ( Figure 5), a critical step in chondrogenesis and cartilage formation. 36 Because initiation of chondrogenic induction of MSCs requires better cell-cell interaction rather than cell-matrix adhesion, non-adhesive 3D pellet culture or micromass culture are considered as standard methods to induce MSC chondrogenic differentiation. Previous reports showed that neutralization of N-cadherin results in the inability of mesenchymal cells to condense and therefore inhibits subsequent chondrogenesis 37 ; conversely, addition of N-cadherin biomimetic peptides can enhance neocartilage formation by human MSCs. 38 These findings demonstrated the crucial role of N-cadherin-mediated cell condensation in promoting chondrogenic differentiation. We found that expression of N-cadherin, and interactions between N-cadherin and b-catenin were significantly upregulated with Wnt inhibition during MSC chondrogenic differentiation ( Figure 5), implying that Wnt/b-catenin antagonism can increase N-cadherin levels and cell condensation, both critical processes for chondrogenesis. Moreover, disruption of actin polymerization-also a critical process during in vivo condensation 39 -by CytoD was partially rescued with Wnt/b-catenin antagonism ( Figures 5F and 5G), demonstrating the key role of this pathway on multiple aspects of chondrogenesis. Even more striking, using Upstream Regulator Analysis which predict involvement of transcription factors/master regulators, a-catenin was predicted as an upregulated factor, whereas TGFb as well as Wnt/b-catenin members were predicted to be downregulated in primary chondrocytes compared to MSCs as well as osteoblasts ( Figure 6G). Our findings therefore strongly implicate the critical role of b-catenin as a structural protein as well as further support the importance of N-cadherin in the AJ and the cytoskeleton-mediated condensation during MSC chondrogenesis.
Our study revealed the many off-target lineages specification by TGFb during MSC chondrogenesis. One of the most important molecules/pathways in developmental biology, TGFb is known to be involved in specification of multiple mesodermal lineages, as well as mediate pathological fibrotic processes. In the development of the skeletal system, TGFb not only induces chondrogenic differentiation and modulates the process of hypertrophy 40 but also osteogenic differentiation as well. 19 In addition, TGFb promotes differentiation of MSCs into smooth muscle cells 9 ; this was clearly reflected in our transcriptomic analysis which showed that signaling by TGFb family was highly enriched in SMCs compared to MSCs (Figures 2A and 2B). Of interest, the evidence for involvement of TGFb signaling in chondrogenesis appears to be largely in vitro, because no defect in cartilage formation seen in either TGFb1-or TGFb3-deficient mice. 8 Despite the well documented evidence on the pleomorphic effects of TGFb on multiple mesenchymal lineages, no study has concurrently evaluated commitment into these lineages. We found that even under chondrogenic 3D pellet culture conditions, TGFb increased the expression of both the smooth muscle-lineage marker aSMA and the osteogenic/hypertrophic marker RUNX2 (Figure 2). Such significant off-target lineage specifications may contribute to a lower chondrogenic differentiation efficiency, and be responsible for ossification and/or fibrosis observed in transplanted chondrocytes derived from TGFb-differentiated MSCs. 41,42 Although two recent studies reported that removal of TGFb and/or applying Wnt antagonist in the late stage of MSC chondrogenic differentiation reduced hypertrophic cartilage formation, 43,44 we found that addition of TGFb rapidly upregulated canonical Wnt/b-catenin downstream genes AXIN2 and TCF7 in addition to RUNX2-the upstream transcription factor controlling the  Figures 4C and 4D), which is in line with previous reports that TGFb is a weak agonist for Wnt/b-catenin signaling. 47 Moreover, the use of MSCs rather than pluripotent stem cells may lead to less non-chondrogenic specification and a more uniformed outcome. 44 Our findings with previous transgenic mice data collectively support that Wnt/b-catenin antagonism rather than TGFb agonism may be the most critical and appropriate pathway to efficiently and specifically induce MSC chondrogenic lineage commitment.
The master transcription factor for chondrogenesis is SOX9, 48 however, we did not find further upregulation of this transcription factor with Wnt inhibition during MSC chondrogenic induction using 3D pellet culture ( Figure 1C), unlike the significant upregulation of more downstream chondrogenic genes COL2A1 and ACAN (Figures 1D and 1E). Our in vitro data was surprisingly corroborated in transcriptome analysis of human samples, in which SOX9 was downregulated in chondrocytes compared to osteoblasts and undifferentiated MSCs (Figures 6E and 6F). Similar to our data, a discrepant trend in expression levels between SOX9 and mature chondrogenic markers including COL2A1 and ACAN has been reported previously. [49][50][51] These findings collectively strongly suggest that high levels of Sox9 expression are a necessary event in development and early stages of chondrogenesis, but less evident at later stages including in vitro differentiation in somatic progenitors/stem cells such as MSCs.
In summary, using multiple sources of human MSCs including ESC-MSCs and iPSC-MSCs we found that replacement of TGFb agonism with Wnt/b-catenin antagonism resulted in robust and specific in vitro and in vivo MSC chondrogenesis by eliminating off-target lineage specification into osteogenesis/hypertrophic cartilage and smooth muscle. Wnt/b-catenin antagonism also more efficiently induced MSC chondrogenesis by increasing N-cadherin levels as well as N-cadherin-b-catenin interactions at the AJ to enhance condensation (Figure 7). These findings are also corroborated by bioinformatic analysis of human primary MSC, chondrocyte, and osteoblast transcriptomes, in which downregulation of both Wnt/b-catenin and TGFb pathways, with upregulation of a-catenin-related processes was seen in primary chondrocytes compared to MSCs and osteoblasts. Our study underscores the importance of structural

OPEN ACCESS
modification in MSC chondrogenesis, as well as having a thorough understanding of key developmental pathways in lineage specification. The capacity to use small molecules rather than a protein growth factor for stem cell differentiation is also more cost-effective, and therefore highly relevant for translational application.

Limitations of the study
In this study, we found that Wnt antagonism alone can induce MSC chondrogenic differentiation through increasing AJ interactions and condensation, whereas restricting off-target lineage commitment which can occur with the classical differentiation protocol using TGFb. But the detailed molecular mechanisms involved are still unclear and should be clarified in future studies. Moreover, the aim of our animal experimental results is to provide in vivo proof-of-concept. The therapeutic efficacy of Wnt antagonism for MSC chondrogenesis should be further investigated in disease models with cartilage defects.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
This work was partially funded by the NHRI (11A1-CSPP06-014 & 11A1-CSGP08-048 to B.L.Y.). We also thank the Optical Biology Core Facility of the NHRI for their support on confocal microscopy. iScience Article