A PINCH-1–Smurf1 signaling axis mediates mechano-regulation of BMPR2 and stem cell differentiation

Proper control of stem cell fate decision is crucial for embryonic development, tissue homeostasis, repair, and regeneration. Stem cell differentiation is regulated by multiple signaling pathways, including those of TGFβ/bone morphogenetic protein (BMP), integrin, Hippo, Wnt, and FGFs (Blank et al., 2008; Chen et al., 2016). Furthermore, there are extensive cross talks between these signaling pathways, which, collectively, determine the final outcome of stem cell fate decision. Importantly, stem cell differentiation is controlled by not only biochemical, but also mechanical signals from extracellular environment or niche (Vogel and Sheetz, 2009; Dupont et al., 2011; MacQueen et al., 2013; Chen et al., 2016; Vining and Mooney, 2017). Pioneering studies by McBeath et al. (2004) have shown that mesenchymal stem cell (MSC) fate decision is regulated by cell shape and cytoskeletal tension. Furthermore, changes in ECM stiffness have been found to exert profound effects on stem cell differentiation (Mammoto and Ingber, 2009; Wozniak and Chen, 2009; Dingal and Discher, 2014). Because of their importance, the signaling mechanisms through which mechano-environment regulates stem cell differentiation are an important area of current biological and medical research.

It has been well documented that BMP signaling pathways are critical for control of stem cell differentiation (Zhang and Li, 2005; Beederman et al., 2013; Wang et al., 2014; Garg et al., 2017). Several BMPs, including BMP2, BMP6, BMP7, and BMP9, have been shown to promote MSC osteoblastic differentiation (Cheng et al., 2003; Noël et al., 2004; Beederman et al., 2013). BMPs exert their effects on cells through interacting with cell surface heterotetrameric complexes consisting of two dimers of type I and II serine/threonine kinase receptors, in which the constitutively active type II receptor transphosphorylates the type I receptor, leading to activation of the type I receptor, phosphorylation of Smad1/5/8, and downstream signaling (Shi and Massagué, 2003; Sieber et al., 2009; Miyazono et al., 2010; Gomez-Puerto et al., 2019). BMPR2 is a BMP-specific type 2 receptor that is crucial for embryonic development, vasculogenesis, and osteogenesis (Onishi et al., 1998; Garimella et al., 2007; Lehnerdt et al., 2007; Kim et al., 2017; Andruska and Spiekerkoetter, 2018; Gomez-Puerto et al., 2019). Lack of BMPR2 in mice is lethal in the early embryonic stage (Beppu et al., 2000), while mice expressing a BMPR2 mutant with reduced signaling capability die at midgestation with cardiovascular and skeletal defects (Délot et al., 2003). BMPR2 is critically involved in promoting MSC differentiation toward osteoblastic lineage (Wu et al., 2010; Yang et al., 2010; Zeng et al., 2012; Cao et al., 2015; Kim et al., 2017). Interestingly, overexpression of Smurf1, a C2-WW-HECT domain E3 ubiquitin ligase (Zhu et al., 1999), in HEK239T cells reduced the level of BMPR2 (Murakami et al., 2010). It remains to be determined, however, whether Smurf1 mediates BMPR2 degradation in MSCs and, if so, whether it mediates the upstream signals and the molecular mechanism that controls this process.

Another signaling pathway that is critical for control of stem cell differentiation is that of integrins, transmembrane receptors mediating cell-ECM adhesion and signaling (Schwartz, 2010; Yim and Sheetz, 2012; Humphrey et al., 2014; Horton et al., 2016). PINCH-1 is a widely expressed and evolutionally conserved cytoplasmic component of the integrin signaling pathway (Tu et al., 1999; Zhang et al., 2002; Wu, 2004, 2005; Legate et al., 2006; Kovalevich et al., 2011). In this study, we show that Smurf1 binds BMPR2 and controls its degradation in MSCs in response to mechanical signals from ECM. Furthermore, we identify PINCH-1 as a key regulator of Smurf1-mediated binding and degradation of BMPR2 in MSC differentiation, suggesting a cross talk between the integrin and BMP signaling pathways in this process. Mechanistically, PINCH-1 binds directly to the Smurf1 C2 domain, to which BMPR2 binds. Overexpression of PINCH-1 inhibited the Smurf1–BMPR2 interaction and consequently Smurf1-mediated degradation of BMPR2. Conversely, depletion of PINCH-1 increased Smurf1-mediated binding and degradation of BMPR2. Notably, ECM stiffening increased PINCH-1 level and its complex formation with Smurf1, resulting in inhibition of Smurf1 interaction and degradation of BMPR2 and consequently augmented BMP signaling and osteogenic differentiation (OD). Depletion of PINCH-1 was sufficient to inhibit the ECM stiffening–induced increase of BMPR2 signaling and OD, whereas overexpression of PINCH-1 blocked ECM softening–induced degradation of BMPR2 and adipogenic differentiation (AD). These results identify a novel PINCH-1–Smurf1 signaling axis that links mechano-environment to BMPR2 degradation and MSC fate decision.

In our investigation of MSC differentiation, we noticed that PINCH-1 protein level was increased during OD (Fig. 1 A). PINCH-1 mRNA level, however, was not increased under the same condition (Fig. 1 B). PINCH-1 level was more than doubled soon after (within 1 d) induction of OD and was tripled by day 4 (Fig. 1 A). The levels of BMPR2 and phosphor-Smad1/5 were also increased, albeit the increases initially lagged behind those of PINCH-1 level (Fig. 1 A). No increases of BMPR2 mRNA (Fig. 1 B) or BMPR1A and Smad1 levels (Fig. 1 A) were observed under the same condition.

To test whether the increase of PINCH-1 level is functionally important, we depleted PINCH-1 from MSCs (Fig. 1 C, lane 3). PINCH-1–deficient MSCs, unlike control MSCs expressing a normal level of PINCH-1, were unable to undergo OD when cultured under OD condition (Fig. 1, D–F). Furthermore, depletion of PINCH-1 was sufficient to induce spontaneous AD in MSCs cultured under normal growth condition (Fig. 1, I and J) and enhance AD in MSCs cultured under AD condition (Fig. 1, D, G, and H). Reexpression of 3xFLAG-tagged PINCH-1 (3f-P1) in PINCH-1–deficient MSCs (Fig. 1 C, lane 5) effectively reversed the differentiation defects induced by depletion of PINCH-1 (Fig. 1, D–H). To test the function of PINCH-1 in MSC differentiation in vivo, we transplanted PINCH-1–deficient MSCs and control MSCs into mice and analyzed their abilities to differentiate into osteoblasts or adipocytes. The tissues were dissected from the mouse recipients and analyzed by hematoxylin and eosin (H&E) staining (Fig. 2 A). Western blotting (WB) analyses confirmed that PINCH-1 level in the tissues derived from PINCH-1 knockdown MSCs was significantly lower than that in control MSCs (Fig. 2 B). Consistent with the results in cell culture, depletion of PINCH-1 markedly inhibited OD (Fig. 2 C) and concomitantly increased AD (Fig. 2 D) in vivo. Collectively, these results suggest that PINCH-1 level is critical for control of MSC fate decision (i.e., higher levels of PINCH-1 favor OD, whereas lower levels of PINCH-1 favor AD).

We next sought to determine the mechanism by which PINCH-1 regulates MSC differentiation. Because the increase of BMPR2 followed that of PINCH-1 during OD (Fig. 1 A), we tested whether the increase of BMPR2 is caused by that of PINCH-1. To do this, we depleted PINCH-1 from the MSCs and found that it significantly reduced BMPR2 protein (Fig. 3 A, compare lane 3 with lanes 1 and 2), but not mRNA level (Fig. 3 B). Knockdown of integrin-linked kinase (ILK) or kindlin-2 did not significantly reduce BMPR2 level (Fig. S1). Consistent with our previous studies (Guo et al., 2018), knockdown of kindlin-2 did reduce the level of YAP1 (Fig. S1). In contrast to a marked reduction of BMPR2 level, depletion of PINCH-1 did not significantly change the levels of BMPR1A and Smad1 (Fig. 3 A) or Rac1 activity (Fig. 3 C). Consistent with the reduction of BMPR2 level, the level of Smad1/5 phosphorylation was also reduced in response to depletion of PINCH-1 (Fig. 3 A). To further test this, we reexpressed 3f-P1 in PINCH-1 knockdown MSCs and found that it substantially increased BMPR2 and phosphor-Smad1/5 levels (Fig. 3 D). Additionally, knockdown of PINCH-1 diminished nuclear localization of phosphor-Smad1/5, which was restored by reexpression of 3f-P1 (Fig. 3 F). Similar results were obtained with a BMP reporter containing BMP-responsive elements (BREs) fused to a luciferase reporter gene (Korchynskyi and ten Dijke, 2002; Yadav et al., 2012; Fig. 3 E). These results demonstrate that PINCH-1 is critical for regulation of BMPR2 level and its downstream signaling.

We next tested whether the down-regulation of BMPR2 expression is responsible for PINCH-1 deficiency–induced inhibition of OD. To do this, we overexpressed 3xFLAG-tagged BMPR2 (3f-BMPR2) in PINCH-1–deficient MSCs (Fig. 4 A, lane 4) as well as normal MSCs (Fig. S2 A, lane 3) and assessed the effects on Smad1/5 phosphorylation and MSC differentiation. Consistent with previous studies (Wu et al., 2010; Yang et al., 2010; Zeng et al., 2012; Cao et al., 2015; Kim et al., 2017), overexpression of 3f-BMPR2 enhanced Smad1/5 phosphorylation (Fig. S2 A), resulting in augmented OD and reduced AD (Fig. S2, B–F). Overexpression of 3f-BMPR2 in PINCH-1–deficient MSCs effectively restored the defects on Smad1/5 phosphorylation (Fig. 4 A), nuclear localization (Fig. 4 C), BMP reporter activity (Fig. 4 B), and OD (Fig. 4, D–F) and suppressed PINCH-1 deficiency–induced AD (Fig. 4, D, G, and H). Collectively, these results suggest that PINCH-1 regulates MSC differentiation through, at least in part, control of BMPR2 level.

We next sought to determine the mechanism by which PINCH-1 regulates BMPR2 level. Because PINCH-1 regulates BMPR2 protein but not mRNA level (Fig. 3, A and B) and previous studies in HEK293 cells have implicated a role of Smurf1 in BMPR2 degradation (Murakami et al., 2010), which we have confirmed (Fig. S3), we tested the possibility that Smurf1 is involved in PINCH-1 regulation of BMPR2 level in MSCs. To do this, we overexpressed Smurf1, and used a different E3 ubiquitin ligase NEDD4 as a control, in MSCs. Overexpression of Smurf1 (Fig. 5 A, compare lane 3 with lanes 1 and 2), but not that of NEDD4 (Fig. 5 B, compare lane 3 with lanes 1 and 2), markedly reduced the level of BMPR2, which was reversed by the presence of proteasome inhibitor MG132 or lysosomal inhibitor leupeptin (Fig. S4 A). Conversely, depletion of Smurf1 from MSCs increased BMPR2 level (Fig. 5 C, compare lane 3 with lanes 1 and 2). Neither overexpression nor knockdown of Smurf1 significantly altered PINCH-1 level (Fig. 5, A and C).

Next, we sought to test whether Smurf1 binds BMPR2 and, if so, the Smurf1 domain responsible for the binding. Because overexpression of WT Smurf1 diminished BMPR2 level (Fig. 5 A, lane 3), we expressed a 3xFLAG-tagged catalytically inactive Smurf1 mutant CA (3f-CA; Wei et al., 2017), in which Cys⁶⁹⁹ within the catalytic HECT domain is substituted with Ala, in MSCs and tested its complex formation with BMPR2 by coimmunoprecipitation (coIP). The results showed that BMPR2 was indeed coimmunoprecipitated with 3f-CA (Fig. 5 D, lane 4). In control experiments, Smurf1-binding protein kindlin-2 (Wei et al., 2017), but not irrelevant protein GAPDH, was also coimmunoprecipitated with 3f-CA (Fig. 5 D, lane 4), confirming the specificity of the experiments. To further test this, we analyzed the interaction between 3f-CA and BMPR2 using proximity ligation assay (PLA). Kindlin-2 and irrelevant protein Ki-67 were used as positive and negative controls, respectively. The results showed that BMPR2 and 3f-CA (Fig. 5 E), like kindlin-2 and Smurf1 (Fig. S5 A) but not Ki-67 and Smurf1 (Fig. S5 B), formed a complex in MSCs. To test whether Smurf1 binds BMPR2 directly, we expressed recombinant MBP-BMPR2 cytoplasmic domain (CD; residues 203–1,038) and GST-Smurf1, respectively. GST-tagged dynein light chain tctex-1 (DYNLT), which is known to bind BMPR2 CD (Machado et al., 2003), and GST-tagged NEDD4, the overexpression of which does not reduce BMPR2 level (Fig. 5 B), were used as positive and negative controls, respectively. Consistent with the coIP and PLA experiments, GST-Smurf1 (Fig. 5 F, lane 4), but not GST (Fig. 5 F, lane 2), readily pulled down MBP-BMPR2-CD, suggesting that Smurf1 directly interacts with BMPR2 CD. In control experiments, MBP-BMPR2-CD was pulled down by GST-DYNLT (Fig. 5 F, lane 8) but not GST-NEDD4 (Fig. 5 F, lane 6), confirming the specificity of the assay. To further test this, we generated MBP-BMPR2 kinase domain (KD; residues 203–504) and tested its Smurf1 binding. The results showed that GST-Smurf1 (Fig. 5 G, lane 4), but neither GST (Fig. 5 G, lane 2) nor GST-NEDD4 (Fig. 5 G, lane 6), readily pulled down MBP-BMPR2-KD, suggesting that BMPR2 KD located within the CD is sufficient to bind Smurf1. To map the Smurf1 domain that mediates BMPR2 binding, we generated GST fusion proteins containing the various domains (C2, WW, or HECT) of Smurf1. GST-C2 (Fig. 5 H, lane 4), but neither GST-WW (Fig. 5 H, lane 6) nor GST-HECT (Fig. 5 H, lane 8), pulled down MBP-BMPR2-CD. Similarly, GST-C2 (Fig. 5 I, lane 4), but neither GST-WW (Fig. 5 I, lane 6) nor GST-HECT (Fig. 5 I, lane 8), pulled down MBP-BMPR2-KD. Next, we expressed FLAG-tagged Smurf1 mutant in which the C2 domain is deleted (3f-ΔC2) in MSCs (Fig. 5 D, lane 6) and confirmed that 3f-ΔC2 (Fig. 5 D, lane 8), unlike 3f-CA (Fig. 5 D, lane 4), failed to coimmunoprecipitate BMPR2. Collectively, these results suggest that the Smurf1 C2 domain and BMPR2 KD are involved in mediating the Smurf1–BMPR2 interaction.

To test whether BMPR2 binding is required for Smurf1-mediated degradation of BMPR2, we expressed FLAG-tagged WT Smurf1, catalytically inactive CA mutant, BMPR2 binding–defective ΔC2 mutant, and ΔWW mutant in HEK293 cells. As expected, overexpression of WT Smurf1 reduced BMPR2 level (Fig. 5 J, compare lanes 1 and 2). Importantly, overexpression of BMPR2 binding–defective ΔC2 (Fig. 5 J, lane 3), like that of catalytically inactive CA (Fig. 5 J, lane 5), failed to reduce BMPR2 level. By marked contrast, overexpression of ΔWW, which contains C2 and HECT domains but lacks WW domain, effectively reduced BMPR2 level (Fig. 5 J, lane 4). These results suggest that the BMPR2-binding C2 domain and the catalytic HECT domain are necessary for Smurf1-mediated degradation of BMPR2.

We next tested whether Smurf1 is involved in PINCH-1 deficiency–induced reduction of BMPR2 level. To do this, we depleted Smurf1 and found that it reversed PINCH-1 deficiency–induced reduction of BMPR2 level (Fig. 6 A). Consistent with this, treatment of cells with MG132 or leupeptin reversed PINCH-1 deficiency–induced reduction of BMPR2 level (Fig. S4 B), suggesting that proteasome- and lysosome-mediated degradation are involved in PINCH-1 deficiency–induced, Smurf1-mediated reduction of BMPR2 level. As expected, depletion of Smurf1 from PINCH-1–deficient MSCs restored Smad1/5 phosphorylation (Fig. 6 A), nuclear localization (Fig. 6 C), BMP reporter activity (Fig. 6 B), and OD (Fig. 6, D–F). Concomitantly, it suppressed PINCH-1 deficiency–induced AD (Fig. 6, D, G, and H). Collectively, these results suggest that Smurf1 is a key mediator of PINCH-1 in its regulation of BMPR2 expression, signaling, and MSC differentiation.

We next sought to determine how PINCH-1 regulates Smurf1-mediated degradation of BMPR2. To this end, we first tested whether PINCH-1 interacts with Smurf1. PINCH-1 was readily coimmunoprecipitated with 3f-CA (Fig. 7 A, lane 4). Consistent with this, endogenous Smurf1 and PINCH-1 formed a complex in cells (Fig. 7 B). To test whether they directly interact, we expressed MBP–PINCH-1 and GST-Smurf1, respectively, and tested their binding in a pull-down assay. GST-ILK, which is known to directly bind PINCH-1 (Tu et al., 1999), and α-parvin, an ILK-binding protein that does not directly bind PINCH-1 (Tu et al., 2001; Qin and Wu, 2012), were used as positive and negative controls (Fig. 7 C), respectively. The results showed that MBP–PINCH-1 was pulled down by GST-Smurf1 (Fig. 7 C, lane 4) and GST-ILK (Fig. 7 C, lane 8), but neither GST (Fig. 7 C, lane 2) nor GST-α-parvin (Fig. 7 C, lane 6). To identify the Smurf1 domain mediating PINCH-1 binding, we analyzed the PINCH-1 binding activity of individual Smurf1 domains. GST-C2 (Fig. 7 D, lane 4) and, to a much lesser extent, GST-WW (Fig. 7 D, lane 6), but not GST-HECT (Fig. 7 D, lane 8), bound MBP–PINCH-1. Consistent with this, 3f-ΔC2 (Fig. 7 A, lane 8), unlike 3f-CA (Fig. 7 A, lane 4), failed to coimmunoprecipitate PINCH-1. The findings that PINCH-1 binds to C2, to which BMPR2 also binds (Fig. 5, D, H, and I), raised the possibility that PINCH-1 may inhibit Smurf1 interaction and consequently degradation of BMPR2. To test this, we analyzed the Smurf1–BMPR2 interaction either in the presence or absence of PINCH-1 and found that the presence of MBP–PINCH-1 (Fig. 7 E, lane 8) but not MBP (Fig. 7 E, lane 7) indeed reduced the amount of BMPR2 bound to Smurf1. Furthermore, the binding of BMPR2 to GST-C2 was inhibited by PINCH-1 but not MBP in a dose-responsive manner (Fig. 7 F). Similarly, overexpression of PINCH-1 in cells also reduced the amount of BMPR2 coimmunoprecipitated with Smurf1 (Fig. 7 G, compare lane 6 with lane 5), confirming that PINCH-1 inhibits the binding of BMPR2 to Smurf1. To test the effect of PINCH-1 on Smurf1-mediated degradation of BMPR2, we overexpressed either 3f-Smurf1 alone or 3f-Smurf1 together with 3f-P1 in MSCs. As expected, overexpression of 3f-Smurf1 reduced the level of BMPR2 (Fig. 7 H, lane 3). Overexpression of 3f-P1, however, completely reversed the reduction of BMPR2 level induced by overexpression of 3f-Smurf1 (Fig. 7 H, lane 4). Thus, the binding of PINCH-1 to Smurf1 effectively inhibits Smurf1-mediated binding (Fig. 7 K) and degradation of BMPR2.

MSC fate decision is controlled to a great extent by mechano-environment. To test whether the PINCH-1–Smurf1–BMPR2 signaling axis identified in the foregoing experiments is involved in this process, we plated MSCs on ECM with different stiffness and analyzed the effects. PLA experiments showed that the level of the PINCH-1–Smurf1 complex in MSCs on soft ECM was markedly reduced compared with that in MSCs on stiff ECM (Fig. 7, I and J). Furthermore, PINCH-1 protein (Fig. 8 A, compare lane 1 with lane 3; and Fig. 8 B), but not mRNA (Fig. 8 C), level was significantly reduced in response to ECM softening, whereas the level of Smurf1 was not altered (Fig. 8 A). Consistent with a critical role of PINCH-1 in regulation of BMPR2 level, the protein (Fig. 8 A, compare lanes 1 and 3; and Fig. 8 B), but not mRNA (Fig. 8 C), level of BMPR2 was also reduced in response to ECM softening. Concomitant to the reduction of BMPR2 level, Smad1/5 phosphorylation was dramatically reduced (Fig. 8 A, compare lanes 1 and 3). The levels of Smad1 and kindlin-2 were not reduced in response to ECM softening (Fig. 8 A). To further investigate this, we tested the effects of ECM stiffness on the Smurf1–BMPR2 complex in MSCs. Human MSCs (hMSCs) expressing 3f-CA were used (Fig. 8 E), as a complex consisting of BMPR2 and catalytically active Smurf1 was barely detectable, presumably because BMPR2 was quickly degraded upon interaction with catalytically active Smurf1. As expected, the level of the 3f-CA–PINCH-1 complex was reduced in response to ECM softening (Fig. 8, D and F). Importantly, the level of the 3f-CA–BMPR2 complex was markedly increased in response to ECM softening (Fig. 8, D and G). Furthermore, overexpression of PINCH-1 in MSCs on soft ECM (Fig. 8 E, lane 4), which increased the level of the 3f-CA–PINCH-1 complex (Fig. 8, D and F), significantly reduced the level of the 3f-CA–BMPR2 complex (Fig. 8, D and G). Consistent with this, overexpression of PINCH-1 in MSCs significantly increased BMPR2 level and downstream signaling despite being in contact with soft ECM (Fig. 8 A, compare lanes 3 and 4). Conversely, depletion of PINCH-1 in MSCs plated on stiff ECM, which reduced the level of the 3f-CA–PINCH-1 complex (Fig. 8, D and F), significantly increased the level of the 3f-CA–BMPR2 complex (Fig. 8, D and G). Consistent with a role of Smurf1 interaction in BMPR2 degradation, depletion of PINCH-1 significantly reduced BMPR2 level and downstream signaling despite the presence of stiff ECM (Fig. 8 A, compare lanes 1 and 2). Functionally, knockdown of PINCH-1 was sufficient to block OD and induce AD in MSCs, despite the fact that they were in contact with stiff ECM (Fig. 9, A–E). Conversely, overexpression of PINCH-1 (Fig. 9, A–E) or BMPR2 (Fig. 9, F–K) in MSCs on soft ECM was sufficient to induce OD and inhibit AD.

We next sought to investigate the mechanism by which ECM stiffness regulates PINCH-1 level. Previous studies have shown that soft ECM enhances integrin internalization through caveolae/raft-dependent endocytosis (Du et al., 2011). Thus, we hypothesize that ECM stiffness may influence PINCH-1 level through alteration of caveolae/raft-dependent integrin internalization. To test this, we analyzed the levels of PINCH-1 in MSCs on ECM with different stiffness either in the presence or absence of methyl-β-cyclodextrin (MBCD), a caveolae/raft inhibitor that is known to suppress integrin internalization (Du et al., 2011). We confirmed that treatment with MBCD suppressed integrin internalization (Fig. 10 A). Importantly, suppression of integrin internalization with MBCD completely reversed the reduction of PINCH-1 level induced by soft ECM (Fig. 10 B). To further test this, we knocked down caveolin-1 and found that it also reversed soft ECM–induced reduction of PINCH-1 level (Fig. 10 C). Finally, soft ECM–induced reduction of PINCH-1 level was completely reversed by proteasome inhibitor MG132 or lysosomal inhibitor leupeptin (Fig. 10 D), suggesting that proteasome and lysosome are involved in soft ECM–induced down-regulation of PINCH-1 level.

MSC fate decision is controlled to a great extent by mechano-environment (Vogel and Sheetz, 2009; Dupont et al., 2011; MacQueen et al., 2013; Chen et al., 2016; Vining and Mooney, 2017). Our studies presented in this report identify a novel signaling pathway consisting of PINCH-1, Smurf1, and BMPR2 that links mechano-environment to MSC fate decision. Specifically, we have found that the level of BMPR2, a key component of the BMP signaling pathway, is regulated by a mechano-responsive PINCH-1–Smurf1 complex. ECM stiffening increased the level of PINCH-1 and its complex formation with Smurf1, which, in turn, inhibited Smurf1 interaction with BMPR2 and thus blunted Smurf1-mediated BMPR2 degradation, resulting in elevation of BMPR2 level and consequently augmenting BMP signaling and MSC OD. Depletion of PINCH-1 was sufficient to inhibit BMP signaling and reprogram MSC fate from OD to AD, even in the presence of mechanical cues that favor OD (i.e., stiff ECM), suggesting that the PINCH-1–mediated signaling pathway is crucial for mechano-regulation of MSC fate decision. Furthermore, PINCH-1 deficiency–induced reprogramming of MSC fate (i.e., from OD to AD) can be effectively reversed by either depletion of Smurf1 or overexpression of BMPR2. Thus, PINCH-1 regulates MSC fate decision through, at least in part, control of Smurf1-mediated degradation of BMPR2. How does ECM stiffness influence the level of PINCH-1? Because ECM stiffening does not increase PINCH-1 mRNA level (Fig. 8 C), ECM stiffness likely regulates PINCH-1 at the protein level. Indeed, inhibition of proteasome with MG132 or lysosome with leupeptin markedly reversed soft ECM–induced reduction of PINCH-1 level (Fig. 10 D), suggesting that proteasome and lysosome are involved in this process. Previous studies have shown that ECM stiffness exerts a strong effect on integrin internalization through caveolae/raft-dependent endocytosis (Du et al., 2011). Consistent with the previous studies, treatment with caveolae/raft inhibitor MBCD inhibited integrin internalization (Fig. 10 A). Importantly, inhibition of integrin internalization completely blocked the reduction of PINCH-1 level induced by soft ECM (Fig. 10 B). Depletion of caveolin-1 also reversed soft ECM–induced down-regulation of PINCH-1 level (Fig. 10 C). Based on this and the previous studies (Du et al., 2011), we propose a model in which soft ECM promotes caveolae/raft-dependent integrin internalization, resulting in increased proteasome- and lysosome-mediated degradation of PINCH-1, which reduces the amount of PINCH-1 bound to Smurf1 and consequently increases Smurf1-mediated interaction/degradation of BMPR2 and suppresses BMP signaling and OD.

In addition to identifying the PINCH-1–Smurf1–BMPR2 axis as a key signaling pathway that mediates mechano-control of MSC differentiation, our studies have revealed important insights into the molecular mechanism through which PINCH-1 regulates Smurf1 binding and degradation of BMPR2. Smurf1, which was originally identified based on its interaction with Smad1 and targeting it for degradation (Zhu et al., 1999), contains a C2 domain in its N-terminus, two WW domains in the middle, and a catalytic HECT domain in its C-terminus. Smurf1 can interact not only with R-Smads (e.g., Smad1 and 5), but also with other proteins, including Runx2, RhoA, ARHGEF9/hPEM2, BMPR1A, MEKK2, Talin head, Prickle1, kindlin-2, STAT1, KLF2, WFS1, and ING2 (Zhu et al., 1999; Kavsak et al., 2000; Podos et al., 2001; Murakami et al., 2003; Wang et al., 2003; Ying et al., 2003; Zhao et al., 2003; Yamashita et al., 2005; Yamaguchi et al., 2008; Huang et al., 2009; Narimatsu et al., 2009; Nie et al., 2010; Guo et al., 2011; Xie et al., 2011; Yuan et al., 2012; Cao and Zhang, 2013; Wei et al., 2017). Through interacting and targeting different substrates for degradation, Smurf1 regulates a variety of cellular processes including cell differentiation, proliferation, polarity, adhesion, migration, DNA damage response, and immune responses (Cao and Zhang, 2013). Therefore, determining the molecular basis underlying specific Smurf1–substrate interactions, as well as their regulatory mechanisms, is important for understanding the mechanisms whereby Smurf1 functions in these diverse cellular processes. In this study, we have found that the interaction with BMPR2 is mediated by Smurf1 C2 domain (Fig. 5). Importantly, PINCH-1 also interacts with the C2 domain, and PINCH-1 interaction inhibits Smurf1 binding and degradation of BMPR2 (Fig. 7). These findings provide an explanation as to why overexpression of PINCH-1 inhibited the Smurf1–BMPR2 complex and increased BMPR2 level in MSCs (Figs. 7 and 8). Furthermore, they also explain why knockdown of PINCH-1 increased the Smurf1–BMPR2 complex and reduced BMPR2 level in MSCs (Figs. 3 and 8). It is interesting to note that the levels of Smad1 and BMPR1A were not altered by knockdown or overexpression of PINCH-1 (Fig. 3). This is consistent with the fact that Smad1 and BMPR1A bind to Smurf1 WW or HECT, but not C2 domain (Aragón et al., 2011; Chaikuad and Bullock, 2016), which is the primary binding target of PINCH-1 (Fig. 7).

While the findings presented in this paper clearly demonstrate that PINCH-1 acts as a key regulator of BMP signaling and MSC differentiation in response to mechanical signals, other proteins and signaling pathways also contribute to mechano-regulation of MSC differentiation. For example, we recently found that kindlin-2 is also involved in mechano-regulation of MSC differentiation (Guo et al., 2018). However, depletion of kindlin-2, unlike that of PINCH-1, did not significantly reduce the level of BMPR2 (Fig. S1 B), suggesting that the mechanism by which kindlin-2 regulates MSC differentiation differs from that of PINCH-1, which is consistent with our previous studies showing that kindlin-2 mediates mechano-regulation of MSC differentiation through control of YAP1/TAZ expression and signaling (Guo et al., 2018). Thus, it is likely that multiple signaling pathways, including those consisting of PINCH-1–Smurf1–BMPR2 (the current study) or kindlin-2–YAP1/TAZ (Guo et al., 2018), are involved in mechano-regulation of MSC differentiation. Interestingly, knockdown of ILK, a PINCH-1–binding protein (Tu et al., 1999; Zhang et al., 2002), did not significantly reduce the level of BMPR2 (Fig. S1 A), suggesting that ILK probably does not play a direct role in PINCH-1–mediated interaction with Smurf1 and inhibition of BMPR2 degradation. This is consistent with the fact that PINCH-1 is capable of interacting with Smurf1 or its C2 domain and inhibiting Smurf1 interaction with BMPR2 in the absence of ILK (Fig. 7, C–F). Thus, while complex formation with ILK is critical for many PINCH-1–mediated functions, as previously shown by us and others (Tu et al., 1999; Wu and Dedhar, 2001; Guo and Wu, 2002; Zhang et al., 2002; Fukuda et al., 2003; Wu, 2004, 2005; Yang et al., 2005; Li et al., 2007; Wickström et al., 2010; Rooney and Streuli, 2011; Qin and Wu, 2012), it is likely that not all functions of PINCH-1 are dependent on ILK. Furthermore, while ILK does not seem to play a major role in PINCH-1–mediated interaction with Smurf1 and inhibition of BMPR2 degradation, the studies presented in this paper did not test the role of ILK in mechano-regulation of MSC differentiation and therefore do not rule out the possibility that ILK may contribute to this process through other mechanisms.

Although the studies presented in this report focus on the mechanism of the PINCH-1–Smurf1–BMPR2 signaling pathway and its function in mechano-regulation of MSC differentiation, the PINCH-1–Smurf1–BMPR2 signaling pathway identified in the current study may also function in other physiological and pathological processes involving Smurf1 and BMPR2. In this regard, it is particularly interesting to note that mutations in the BMPR2 gene are linked to the development of human pulmonary arterial hypertension (PAH; Deng et al., 2000; Lane et al., 2000). It was estimated that ∼80% of patients with familial PAH and 20% of patients with idiopathic PAH bear heterozygous BMPR2 mutations (Deng et al., 2000; Aldred et al., 2006; Cogan et al., 2006). Numerous studies have shown that BMPR2 deficiency is a major causal factor in the pathogenesis of PAH (Atkinson et al., 2002; Orriols et al., 2017; Andruska and Spiekerkoetter, 2018). However, because the penetrance of BMPR2 mutations is only ∼20% (Newman et al., 2004), mechanisms other than BMPR2 mutations must exist that can also cause BMPR2 deficiency in PAH. Recent studies have suggested an important role of Smurf1 in BMPR2 degradation and pathogenesis of PAH (Murakami et al., 2010; Rothman et al., 2016). Thus, because PINCH-1 plays a key role in regulation of Smurf1-mediated binding and degradation of BMPR2 (the current study), it will be interesting to investigate in future studies whether PINCH-1 is involved in the pathogenesis of at least a subpopulation of PAH patients (e.g., those with reduced BMPR2 level but lacking BMPR2 mutations).

The pLKO.1-TRC, psPAX2, and pMD2.G vectors were obtained from Addgene (plasmid #10878, plasmid #12260, and plasmid #12259). The pLKO.1 vectors expressing shRNAs targeting human PINCH-1 (Sh-PINCH-1) or scrambled shRNA (Sh-con) sequence were generated using the following sequences: Sh-PINCH-1, 5′-AAG​GTG​ATG​TGG​TCT​CTG​CTC-3′; and Sh-con, 5′-ACG​CAT​GCA​TGC​TTG​CTT​T-3′. To generate lentiviruses encoding the above shRNAs, HEK293T cells were co-transfected with pLKO.1 encoding the various shRNAs and lentiviral constructs (psPAX2 and pMD2.G). The expression vectors encoding human PINCH-1, BMPR2, Smurf1-CA, Smurf1-ΔC2, or BRE-Luc (pLVX-PINCH-1, pLVX-BMPR2, pLVX-Smurf1-CA, pLVX-Smurf1-ΔC2, or pLVX-BRE-Luc) were generated by cloning the corresponding cDNA sequences into the pLVX-IRES-hyg vector (Clotech). The sequence corresponding to the shRNA targeting region in the PINCH-1 expression vectors was changed to 5′-AAG​GCG​ACG​TCG​TGT​CTG​CTC-3′ to confer resistance to the PINCH-1 shRNA (Sh-P1). The sequences of all DNA inserts were verified by DNA sequencing (Invitrogen). To generate lentiviruses, vectors encoding human pLVX-PINCH-1, pLVX-BMPR2, pLVX-Smurf1-CA, or pLVX-BRE-Luc were co-transfected with psPAX2 and pMD2.G into HEK293T cells. After the cells were incubated at 37°C and 5% CO₂ for 24–48 h, the media containing lentiviral particles were harvested. For lentiviral infection, hMSCs were cultured in basal growth medium until 70% confluence and then replaced with fresh medium containing lentivirus at an MOI of 100 for 16 h. Lentiviral infections were performed in the presence of 8 µg ml⁻¹ polybrene.

For preparation of GST fusion proteins containing WT or mutant forms of Smurf1, the corresponding cDNA sequences were cloned into pGEX-4T-1 vector, which were transfected into Escherichia coli Bl21. GST and GST fusion proteins were purified with Glutathione–Sepharose 4B matrix (GE Healthcare) following the manufacturer’s instructions. MBP-his–tagged PINCH-1 and WT or mutant forms of BMPR2 were expressed in E. coli and purified by affinity chromatography with amylose resin (Biolabs, #E8021S). Purified proteins were resolved by SDS-PAGE to verify their sizes and purity. For GST pull-down assays, GST or GST fusion proteins bound to Glutathione-Sepharose beads were incubated with MBP-his–tagged PINCH-1 and WT or mutant forms of BMPR2 for 2 h at 4°C. The beads were washed three times with buffer (20 mM Tris, pH 7.5, 150 mM NaCl, and 0.1% Triton X-100) and then analyzed by SDS-PAGE and WB with antibodies (Abs) as specified in each experiment.

siRNA directed against human Smurf1 or caveolin-1 was synthesized by Invitrogen. The sense sequences of siRNAs were as follows: Smurf1 siRNA, 5′-GCA​GCU​GCU​AGA​UAU​CUU​ATT-3′; caveolin-1 siRNA, 5′-GCA​UUU​GGA​AGG​CCA​GCU​UTT-3′; and control siRNA (Si-NC), 5′-ACG​CAT​GCA​TGC​TTG​CTT​T-3′. Transfection of siRNAs to hMSCs was performed using Lipofectamine RNAiMAX Reagent (Life Technologies). Cells were transfected with 25 pmol per 6-well culture dish (1 × 10⁵ cells per milliliter in 6-well plates).

hMSCs were isolated from human placenta as previously described (Guo et al., 2014; Wang et al., 2015). hMSCs were seeded on cell culture dishes with a growth medium consisting of DMEM (Gibco-Invitrogen), 10% FBS (Gibco-Invitrogen), and antibiotics. In some experiments, hMSCs (as specified in each experiment) were cultured in the presence of proteasome inhibitor MG-132 (10 µM), lysosomal inhibitor leupeptin (10 µM), or caveolae/raft inhibitor MBCD (10 mM) for 24 h before further analyses.

Preparation of collagen-coated polyacrylamide hydrogels with different stiffness was performed as previously described (Wang and Pelham, 1998; Cretu et al., 2010). Briefly, glass coverslips were activated successively with 100 mM NaOH for 5 min, 3-aminopropyltri-methoxysilane (Sigma-Aldrich, #281778) for 5 min, and 0.5% glutaraldehyde (Sigma-Aldrich, #G7776) for 30 min. Hydrogels with different stiffness (soft vs. stiff) were prepared with two prepolymer solutions containing different acrylamide/bis-acrylamide ratios. The elastic moduli of stiff and soft hydrogels were ∼40 kPa and 0.35 kPa, respectively. The polyacrylamide sheets were assembled on the activated coverslips upon the addition of 0.5% ammonium persulfate (Sigma-Aldrich, #09830) and 0.005% tetramethylethylenediamine and subsequently activated by incubation with 1 mM Sulfo-SANPAH (Thermo Scientific, #22589) under 452-nm UV light for 5 min. The activated polyacrylamide sheets were then coated with 0.3 mg ml^-1 collagen-I at 4°C overnight. hMSCs were cultured on collagen-I–coated gels (1.5 × 10⁵ cells per 3.0-cm-diameter coverslip) and analyzed as specified in each experiment.

hMSCs were cultured in various media as specified in each experiment. For induction of AD, hMSCs were cultured in DMEM supplemented with 10% FBS, 1% penicillin-streptomycin, 125 µM 3-isobutyl-1-methylxanthine (Sigma-Aldrich), 0.25 µM dexamethasone (Sigma-Aldrich), 2.5 µM insulin (Sigma-Aldrich), and 50 µM indomethacin (Sigma-Aldrich). 14 d later, cells were fixed in 4% paraformaldehyde and stained with oil red O (Sigma-Aldrich) until obvious signals were detected in one or more of the cell groups (e.g., sh-P1 cell group; Guo et al., 2014). The cells were then rinsed three times with PBS to stop the staining. For induction of OD, hMSCs were cultured in DMEM–high glucose supplemented with 10% FBS, 1% penicillin-streptomycin, 10 nM dexamethasone, 50 µM ascorbic acid 2-phosphate (Sigma-Aldrich), and 10 mM β-glycerophosphate (Sigma-Aldrich). 14 d later, cells were fixed in 4% paraformaldehyde and stained with Alizarin red S (Sigma-Aldrich) to observe mineralized matrix deposition (Wu et al., 2007; Li et al., 2011; Guo et al., 2014). Images were acquired at room temperature using a microscope (Eclipse TS100; Nikon) with long working distance 10×/0.25 and long working distance 20×/0.4 objectives (Nikon) equipped with a digital camera (TS100-F; Nikon) and NIS-Elements F version 4.00.00 (Nikon) image software. Alizarin red S and oil red O staining were quantified from four to five microscopic fields using ImageJ.

GFP-labeled hMSCs (2 × 10⁶ per milliliter) that were infected with Sh-P1 or Sh-con lentivirus were embedded into matrix gels as previously described (Guo et al., 2018). hMSCs in matrix gels were induced with OD medium for 3 d. The matrix plugs were then implanted subcutaneously into Nu/Nu mice. The specimens were dissected from the mouse recipients and analyzed 10 d later by H&E staining, WB, and anti-ALP and anti-perilipin immunofluorescent (IF) staining as described in Tang et al. (2013).

cDNA was synthesized from 10 ng total RNA using the ReverTra Ace qPCR RT Master Mix (Toyobo Life Science) according to the manufacturer’s protocol. Real-time PCR was performed in a 20-µl reaction volume using SYBR Green Realtime PCR Master Mix (Toyobo Life Science) on an ABI 7500 QPCR System. As an internal control, levels of GAPDH mRNA were quantified in parallel with mRNAs of the target genes. Normalization and fold changes were calculated using the ΔΔCt method. Primer sets are listed in Table S1.

WB was performed as previously described (Wu et al., 2015). For preparation of whole cell protein lysates, cells were lysed in 1% SDS lysis buffer (25 mM Tris-HCl, pH 6.8, 50 mM DTT, 10% glycerin, and 2.5% sucrose). Equal amounts (10–60 µg/lane) of total protein were separated on 10% polyacrylamide gel and transferred onto a nitrocellulose membrane. Membranes were blocked for 1 h at room temperature in TBS (50 mM Tris-HCl and 150 mM NaCl, pH 7.4) containing 0.1% Tween 20 and 5% nonfat powdered milk, followed by overnight incubation at 4°C with HRP-conjugated mouse anti-GAPDH (Santa Cruz, sc-47724 HRP), HRP-conjugated mouse anti–FLAG-M2 HRP (Sigma-Aldrich, A8592 HRP), rabbit anti–PINCH-1 (Proteintech, 20772–1-AP), rabbit anti-BMPR2 (Cell Signaling, 6979S), rabbit anti-BMPR1A (Proteintech, 12702–1-AP), rabbit anti–p-Smad1/5 (Cell Signaling, 9516S), rabbit anti-Smad1 (Cell Signaling, 9743S), rabbit anti–caveolin-1 (Cell Signaling, 3238S), mouse anti-Smurf1 (Santa Cruz, sc-100616), or mouse anti-His (Tiangen, AB102-02) Abs. After washing and incubation with appropriate HRP-conjugated secondary anti-rabbit or mouse IgG Abs (Jackson ImmunoResearch, #711-005-152 or #715-005-151), the blots were developed using an ECL kit (Bio-Rad) or the Ultra ECL Western Blotting Detection Reagent (4A Biotech, 4AW011) and then exposed using an automatic digital gel image analysis system (Tanon, 6100B) or x-ray film (Fujifilm, # super RX-N-C). The images were scanned using an imaging scanning system (EPSON Scan L365) and quantified with ImageJ. The levels of BMPR2, BMPR1A, Smad1, and PINCH-1 relative to GAPDH, and that of p-Smad1/5 relative to total Smad1, were calculated by quantification of the data from at least three independent experiments.

hMSCs (2 × 10⁴ per well) were seeded on coverslips in 24-well plates and cultured overnight. The cells were then immersed in 0.1% Triton X-100 in PBS for 10 min at room temperature, followed by washing three times with PBS. The cells were incubated with rabbit anti-ALP (Genetex, #GTX62596), rabbit anti-perilipin (Cell Signaling, 9391S), or mouse anti–p-Smad1/5 (Cell Signaling, 9516S) Abs at 4°C overnight. The cells were then washed with PBS and incubated with Alexa Fluor 594/488–conjugated anti-mouse or rabbit (Invitrogen) secondary Abs (1:300) for 1 h at room temperature. Images were acquired at 21°C using an SP8 confocal fluorescence microscope (63× oil objective, 1.4 NA; Leica) with Leica X version 1.1.0.12420 image software. Anti-ALP and anti-perilipin IF staining were quantified from five microscopic fields using ImageJ. The percentages of cells with positive nuclear p-Smad1/5 staining among total cells were calculated. At least 50 cells were analyzed in each experiment.

hMSCs infected with FLAG-Smurf1-CA were harvested and homogenized in WB and IP lysis buffer (Beyotime Biotechnology, P0013) supplemented with 1 mM PMSF (Sigma-Aldrich, #329–98-6) for 30 min at 4°C. IP was performed by mixing an equal amount of cell lysates (2–3 mg protein) with an equal amount of anti-FLAG M2-conjugated agarose beads (10–30 µl of 50% suspension; Sigma-Aldrich, A2220) and incubated for 2 h. The IP samples were washed once with the lysis buffer and twice with PBS. The samples were then analyzed by WB with rabbit anti-BMPR2 (Cell Signaling, 6979S), rabbit anti–PINCH-1 (Proteintech, 20772–1-AP), or mouse anti-FLAG (Sigma-Aldrich, F1804) Abs.

PLA was performed on fixed hMSCs with Duolink PLA technology probes and reagents (Sigma-Aldrich) following the manufacturer’s instructions. Briefly, cells (as specified in each experiment) were permeabilized with PBS containing 0.5% Triton X-100 for 15 min. After washing with PBS twice, the cells were incubated with blocking solution for 30 min at 37°C and then incubated with pairs of rabbit and mouse Abs as specified in each experiment (e.g., pairs of rabbit anti-BMPR2 [Proteintech, 19087–1-AP] and mouse anti-FLAG [Sigma-Aldrich, F1804], mouse anti–PINCH-1 [BD Biosciences, 612711] and rabbit anti-Smurf1 [Proteintech, 55175–1-AP], or mouse anti–PINCH-1 [BD Biosciences, 612711] and rabbit anti-FLAG [Proteintech, 20543–1-AP]). The cells were incubated with the Abs at 4°C overnight. The coverslips were washed three times with buffer A provided by the manufacturer for 5 min, followed by incubation with the PLA probes (secondary Abs against the primary Abs from different species bound to two oligonucleotides; e.g., anti-mouse MINUS and anti-rabbit PLUS) in Ab diluent for 60 min at 37°C. After washing three times with buffer A for 5 min, the ligation step was performed with ligase diluted in the ligation buffer for 30 min at 37°C. In the ligation step, the two oligonucleotides in the PLA probes were hybridized to the circularization oligonucleotides. The cells were washed twice with buffer A for 5 min, followed by incubation with amplification solution at 37°C for 100 min. The amplification solution contains polymerase for the rolling circle amplification step and oligonucleotides labeled with fluorophores, which binds to the product of the rolling circle amplification and thus allows detection. After washing twice with buffer B provided by the manufacturer for 10 min and once with 0.01× buffer B for 1 min, the coverslips were mounted with Duolink in situ mounting medium containing DAPI. For every experiment, a negative control was included in which one of the two Abs was replaced with control IgG from the same species. In addition, a pair of rabbit anti–Ki-67 (CST, #12202) and mouse anti-Smurf1 (Abcam, ab57573) Abs was used as an additional negative control, and a pair of mouse anti–kindlin-2 (3A3.5) and rabbit anti-Smurf1 (Proteintech, 55175–1-AP) Abs was used as a positive control. At least 20 cells were analyzed for each experiment. All experiments were repeated at least three times.

Integrin internalization assay was performed based on the method described by Du et al. (2011) with minor modifications. Briefly, hMSCs grown on soft substrate were treated with or without 10 mM MBCD for 24 h. The cells were then incubated with an Ab for β1 integrin (1:300; Millipore, #MAB1987) at 4°C for 45 min. After washing three times, cells were incubated at 37°C for 30 min to allow integrin internalization. At the end of the incubation, cell surface Abs were removed with acidic medium (pH 4.0). Internalized β1 integrin bound to the Ab was detected with FITC-conjugated anti-mouse IgG Abs (1:1,000; Abcam).

Stripwell plates (Corning) were coated with 1 µM Trx-His-BMPR2 KD in 50 mM carbonate buffer (pH 9.6) overnight at 4°C. 1.5 µM BSA (VWR Life Science) and 1.5 µM GST-Smurf1-C2 diluted in the same buffer were used as negative and positive controls, respectively. Plates were washed three times with PBS and then blocked with 1% BSA. After washing three more times with PBS, the samples of 1.5 µM GST-Smurf1-C2 containing 0.1% BSA and increasing concentrations of MBP or MBP–PINCH-1 (as specified in each experiment) were applied to wells and incubated at 37°C for 1 h. The wells were washed three times with PBS. HRP-conjugated anti-GST-Tag Ab (Proteintech, HRP-66001) and HRP detection reagent (Cytoskeleton) were used to detect GST-Smurf1-C2 bound to immobilized Trx-His-BMPR2 KD. An optical density of 490 was determined using an Epoch2 (BioTek) microplate reader.

hMSCs were infected with BRE-Luc reporter lentivirus containing BREs from the Id1 promoter fused to a luciferase reporter gene (Korchynskyi and ten Dijke, 2002; Yadav et al., 2012). 2 d later, the reporter cells were infected with various lentiviral vectors as specified in each experiment. To measure luciferase activities, the cells were harvested, suspended in culture medium, and added to wells (100 µl containing 10⁴ cells/well) of 96-well white wall plates. The plates were incubated at 37°C in a humidified cell culture incubator with 5% CO₂ for 2–3 d. Luciferase activities were analyzed with the Dual-Luciferase Reporter System (Beyotime Biotechnology). The Renilla luciferase activities were used as internal controls, and values of firefly luciferase activities were normalized to the corresponding Renilla luciferase activities. All experiments were repeated at least three times.

hMSCs or hMSCs infected with Sh-P1 or control lentivirus were cultured in normal medium for 5 d. Rac1 activities in equal amounts of protein from the cells were analyzed using a Rac1 G-LISA Rac1 activation assay kit (Cytoskeleton) following the manufacturer’s protocol.

Statistical analysis was performed using Student’s t test (two tailed) or one-way ANOVA with Tukey’s post-hoc test using GraphPad Prism 7.0. Data are presented as mean ± SEM. A P value <0.05 was considered significant. Data distribution was assumed to be normal, but this was not formally tested.

Fig. S1 shows that knockdown of ILK or kindin-2 does not significantly reduce BMPR2 level. Fig. S2 demonstrates that overexpression of BMPR2 promotes OD and inhibits AD. Fig. S3 shows that overexpression of Smurf1 reduces BMPR2 level in HEK293 cells. Fig. S4 shows that inhibition of proteasome or lysosome reverses Smurf-1 overexpression– or PINCH-1 deficiency–induced down-regulation of BMPR2 level. Fig. S5 shows PLA analyses of kindlin-2–Smurf1 and Ki-67–Smurf1 interactions. Table S1 lists primers used for quantitative RT-PCR.

This work was supported by the National Natural Science Foundation of China grant 81430068, the Ministry of Science and Technology of China grant 2016YFC1302100, the Natural Science Foundation of Guangdong Province grant 2017B030301018, and the Shenzhen Innovation Committee of Science and Technology grants JCYJ20180302174228311 and ZDSYS20140509142721429.

The authors declare no competing financial interests.

Author contributions: L. Guo, R. Wang, K. Zhang, J. Yuan, J. Wang, X. Wang, and J. Ma conducted the experiments. L. Guo and C. Wu designed the experiments and wrote the paper.