Vinculin promotes nuclear localization of TAZ to inhibit ECM stiffness-dependent differentiation into adipocytes

Mesenchymal stem cells (MSCs) are multipotent stem cells with the capability to differentiate into various types of cells, including neurons, adipocytes, myoblasts, chondrocytes and osteoblasts (Kopen et al., 1999; Pittenger et al., 1999). For instance, MSCs can be differentiated into adipocytes by stimulation with bone morphogenetic protein (BMP) 4 (Tang et al., 2004), and into osteoblasts or chondrocytes by treatment with BMP2 (Noël et al., 2004). In addition to soluble factors, MSC lineage specifications can be directed by physical changes induced by the extracellular environment, including cell area and strain, and by extracellular matrix (ECM) stiffness (Dupont et al., 2011; Engler et al., 2006; McBeath et al., 2004; Nobusue et al., 2014; Sen et al., 2009; Swift et al., 2013). MSCs cultured on soft substrates differentiate into adipocytes efficiently, whereas cells on rigid substrates preferentially differentiate into osteoblasts (Dupont et al., 2011; Engler et al., 2006).

Cell–ECM adhesion sites, called focal adhesions (FAs), play an essential role in sensing ECM stiffness and transducing the information into downstream signaling molecules. Integrins, which are transmembrane ECM receptor proteins, bind to the ECM at FAs, and anchor the actin cytoskeleton through association with cytosolic proteins such as talin and vinculin. Among the numerous cytoplasmic FA proteins, vinculin is one of the most plausible candidate sensors for ECM stiffness (Atherton et al., 2016). Vinculin consists of an N-terminal head domain (Vh), C-terminal tail domain (Vt), and a linker region connecting these two domains. Vh binds to talin and integrin, while Vt binds to F-actin (Jockusch and Isenberg, 1981; Menkel et al., 1994). The linker region binds to Arp2/3, vinexin (also known as SORBS3) and p130CAS (also known as BCAR1) (DeMali et al., 2002; Janoštiak et al., 2014; Kioka et al., 1999). Vh interacts with Vt intramolecularly, resulting in an inactive form that has low affinity to actin and talin (Johnson and Craig, 1995). Without the intramolecular interaction, vinculin becomes an activated form that can interact with F-actin at high affinity (Johnson and Craig, 1995). We previously revealed that vinculin adopts the activated form on rigid ECM and the inactive form on soft ECM. This regulation requires the binding to one of the vinexin splicing variants, vinexin α. Furthermore, this interaction is necessary for stiffness-dependent cell migration of fibroblasts (Yamashita et al., 2014). In addition, p130CAS also affects vinculin stability at FAs through direct interaction (Janoštiak et al., 2014). Although vinculin, as well as vinexin α and p130CAS, seem to function as mechanosensors in mouse embryonic fibroblasts, whether vinculin regulates the stiffness-dependent differentiation of MSCs to adipocytes is still incompletely understood.

Yes-associated protein (YAP, also known as YAP1) and its paralog transcriptional coactivator with a PDZ-binding motif (TAZ, also known as WWTR1) (hereafter YAP/TAZ) are core components of the mammalian Hippo pathway. YAP/TAZ plays a crucial role in regulating cell proliferation and differentiation through transcriptional activation (Varelas, 2014). Moderate expression of YAP induces adipogenesis through cooperating with Wnt/β-catenin signaling (Seo et al., 2013). TAZ modulates the balance of differentiation between adipocytes and osteoblasts in MSCs (Hong et al., 2005). In addition, physical cues, including cell morphology and ECM stiffness, controls YAP/TAZ subcellular localization and function (Aragona et al., 2013; Cui et al., 2015; Dupont et al., 2011; Wada et al., 2011). YAP/TAZ is preferentially accumulated within the nucleus and becomes transcriptionally active on rigid ECM to promote osteogenesis, whereas it accumulates within the cytosol on soft ECM (Hwang et al., 2015). Although intracellular tension and actin organization are suggested to be involved in this regulation, the mechanism by which ECM stiffness regulates YAP/TAZ remains poorly clarified.

In this study, we show that the FA protein vinculin functions as a mechanosensor in MSCs and promotes the nuclear localization of the transcription factor TAZ to inhibit the adipocyte differentiation on rigid ECM.

ST2, a mouse bone-marrow-derived stromal cell line, was used as a model of MSCs because immortalized cell lines are suitable for transfection and long-term culture in order to evaluate their ability to differentiate. We first investigated the differentiation of ST2 cells into adipocytes. Cells were seeded on plastic dishes and induced to form adipocytes by the addition of isobutyl-methyl xanthine (IBMX), dexamethasone (DEX) and insulin for 2 days. Medium was then replaced with fresh medium containing insulin for 2 days, followed by the addition of maintenance medium for 2 days. The mRNA expression of adipogenic markers, including peroxisome proliferator-activated receptor γ2 (PPARγ2) and adipocyte protein 2 (aP2, also known as Fabp4), were significantly increased 6 days after induction (Fig. S1A). Western blotting showed that protein expression of aP2 was increased during the differentiation (Fig. S1B), demonstrating the differentiation of ST2 cells into adipocytes under this condition. We next examined the effect of ECM stiffness on the adipocyte differentiation of ST2 cells. Cells were seeded on collagen-coated polyacrylamide (PAA) gel substrates with different levels of stiffness [1.5 or 8.7 kPa, which corresponds well to the stiffness of adipose tissues (Swift et al., 2013)], and induced to differentiate into adipocytes. Cells on soft substrates (1.5 kPa) expressed higher levels of adipogenic markers, including PPARγ2 (1.5-fold ±0.16) and aP2 (1.5-fold ±0.14) (mean±s.e.m.), than those on rigid substrates (8.7 kPa) (Fig. 1A). Furthermore, soft substrates increased the expression of aP2 protein (Fig. 1B). These data indicate that ECM stiffness regulates the differentiation of ST2 cells into adipocytes, and rigid ECM suppresses adipocyte differentiation.

ST2 cells showed an osteogenic phenotype, as indicated by increased mRNA expression of osteogenic markers, after culture on rigid (plastic) substrates without adipogenic induction (data not shown). To examine whether ECM stiffness affects the differentiation of ST2 cells into other cell types, expression of markers for osteoblast, chondrocyte and myoblast was also analyzed under this condition. As shown in Fig. 1C, rigid substrates (23 kPa) increased the expression of osteogenic markers, e.g. runt-related transcription factor 2 (Runx2), alkaline phosphatase (ALP, also known as Alpl), and osteopontin (Fig. 1C). The substrates with 8.7 kPa stiffness gave similar results. Likewise, expression of Sox9, a chondrogenic marker, was higher on rigid substrates (Fig. S1C). Expression of MyoD (also known as MyoD1), a myogenic marker, was not affected by the ECM stiffness (Fig. S1D), despite previous reports showing that primary MSCs expressed the highest levels of MyoD on 11 kPa substrates (Holle et al., 2013). Taken together, these results suggest that ECM stiffness can regulate the commitment of ST2 cells to adipocytes or osteoblasts.

Cytoskeleton stabilization buffer (CSK) treatment removes cytoplasmic proteins as well as proteins that are loosely attached to the cytoskeleton from cells, but does not remove proteins that bind tightly to the cytoskeleton. Thus, the amount of activated and cytoskeleton-associated vinculin can be visualized by CSK treatment. We recently demonstrated that the amount of CSK-resistant vinculin is increased in cells cultured on rigid substrates and that vinculin activation is involved in sensing ECM stiffness in mouse embryonic fibroblasts (Yamashita et al., 2014). Thus, we next examined the effects of ECM stiffness on the resistance of vinculin to CSK treatment in ST2 cells. First, we determined the localization of vinculin on PAA substrates with different levels of stiffness (1.5 kPa, 3.2 kPa, and 8.7 kPa) by normal immunostaining, that is, without CSK treatment (Fig. 2A). Quantified data show that the number and the area of vinculin-containing FAs per cell were comparable on all of the levels of stiffness that we tested (Fig. 2B). The integrated intensity of CSK-resistant vinculin was also comparable for each level of stiffness. In addition, the plasma membrane was stained with Cell Mask Orange (Fig. S1E) to quantify the cell-spreading area and aspect ratio. No significant difference in aspect ratio among cells grown on substrates with different levels of stiffness was observed, although the cell area was slightly smaller on soft substrates (Fig. S1F). Taken together, these data indicate that ECM stiffness does not affect the subcellular localization of vinculin under our experimental conditions. On the other hand, after treatment of cells with CSK buffer, the number and area of vinculin-containing FAs on rigid substrates were larger than those on soft substrates (Fig. 2C,D). Furthermore, the integrated intensity of CSK-resistant vinculin was also increased on rigid substrates. Collectively, these results suggest that ECM stiffness does not alter the subcellular localization of vinculin, but that rigid ECM increases the amount of activated and cytoskeleton-associated vinculin in ST2 cells, similar to what is observed in fibroblasts, and supports the idea that vinculin functions as a sensor for ECM stiffness in ST2 cells.

The transcriptional coactivators YAP and TAZ regulate the differentiation of MSCs into adipocytes and osteoblasts (Hong et al., 2005). Furthermore, ECM stiffness and contractile forces generated by the actomyosin cytoskeleton regulate the subcellular localization and activity of YAP/TAZ (Dupont et al., 2011; Wada et al., 2011). It is unknown, however, how YAP/TAZ is regulated by ECM stiffness. Therefore, we tested whether vinculin is involved in YAP/TAZ regulation and adipocyte differentiation. We first checked whether ECM stiffness regulates the subcellular localization of YAP/TAZ in wild-type (WT) ST2 cells. Cells cultured on substrates with stiffness ranges of 1.5 kPa to 8.7 kPa in sparse conditions were immunostained with anti-YAP/TAZ antibody (Fig. 3A). The subcellular localization of YAP/TAZ was evaluated by the ratio of the intensity of YAP/TAZ signal inside the nucleus to that just outside the nucleus (nuc/cyt intensity ratio) (Fig. 3B). A larger intensity ratio indicates a stronger nuclear localization of YAP/TAZ. Consistent with previous reports (Dupont et al., 2011), rigid substrates (8.7 kPa) significantly promoted the nuclear localization of YAP/TAZ compared to that seen in cells on soft substrates (Fig. 3C). These data indicate that YAP/TAZ localizes to the nucleus in an ECM stiffness-dependent manner in ST2 cells.

To examine the function of vinculin in the nuclear localization of YAP/TAZ, vinculin was stably knocked down using shRNA lentivirus. Expression of vinculin was reduced to 20% in vinculin-depleted cells (Fig. S2A). Immunostaining analysis showed that the amount of vinculin localized in both cytosol and at FAs was decreased, although cytosolic vinculin was preferentially depleted (Fig. S2B). We first examined the effect of vinculin on the subcellular localization of YAP/TAZ on a rigid substrate. Vinculin-depleted cells and vinculin-depleted cells with re-expression of vinculin were seeded onto rigid (glass) substrates and immunostained using anti-YAP/TAZ antibody (Fig. 3D). Vinculin depletion decreased the nuc/cyt ratio of YAP/TAZ by 15% (±4.9, mean±s.e.m.) compared to control cells (Fig. 3E). The re-expression of vinculin rescued the decrease (Fig. 3F). As an alternative quantification, the integrated intensity of YAP/TAZ staining in the nucleus was quantified (Fig. S2C,D). Consistent with the nuc/cyt ratio, amounts of nuclear YAP/TAZ were decreased in vinculin-depleted cells and vinculin re-expression rescued this effect. We also performed co-immunostaining of vinculin and YAP/TAZ to investigate the correlation of the nuclear YAP/TAZ and vinculin expression at a single-cell level (Fig. S2E). The integrated intensity of vinculin at FAs, and that of YAP/TAZ in nuclei of individual cells were quantified (Fig. S2F). In control cells, the Spearman's correlation coefficient between vinculin and YAP/TAZ intensity was 0.66 (P<0.001), indicating that YAP/TAZ nuclear localization is highly correlated with vinculin localization at FAs. Vinculin depletion decreased the correlation coefficient to 0.37 (P<0.05). These results indicate the importance of vinculin in promoting YAP/TAZ nuclear localization.

To confirm that nuclear YAP/TAZ is transcriptionally active and promotes the expression of target genes, the mRNA expression of Ctgf and Ankrd1, major target genes, were examined by qRT-PCR. As expected, vinculin depletion decreased the relative expression of Ctgf and Ankrd1and re-expression of vinculin rescued the decrease in the expression (Fig. 3G,H). Taken together, these data indicate that vinculin promotes the nuclear localization and activity of YAP/TAZ on rigid substrates.

Finally, we investigated the effects of vinculin depletion on ECM stiffness-regulated YAP/TAZ nuclear translocation. As shown in Fig. 4A,B, rigid substrates increased the nuclear localization of YAP/TAZ in control knockdown cells by 40% (±5.6, mean±s.e.m.) compared to that seen on soft substrates. On the other hand, nuclear localization of YAP/TAZ was increased only moderately on rigid substrate in vinculin-depleted cells. Furthermore, the nuc/cyt intensity ratio in vinculin-depleted cells was significantly lower than in control cells on rigid substrates (8.7 kPa) (Fig. 4B). On the other hand, the distribution of YAP/TAZ in each cell on soft substrates was comparable. Hence, these data suggest that vinculin regulates the ECM stiffness-dependent subcellular localization of YAP/TAZ and promotes its nuclear localization on rigid substrates.

Nuclear localization of YAP/TAZ is suppressed by the phosphorylation of YAP/TAZ, which occurs as a result of signaling through the Hippo pathway or other pathways (Kanai et al., 2000). Thus, the phosphorylation of YAP/TAZ in vinculin-depleted cells on rigid substrates was analyzed. Phosphorylation was slightly upregulated in vinculin-depleted cells (Fig. S2G). However, the activated (phosphorylated) form of LATS1, a YAP kinase in the Hippo pathway, was not changed between control and vinculin-depleted cells (Fig. S2G). This result suggests that vinculin does not alter the Hippo signaling pathway.

In addition to the Hippo pathway, the actin cytoskeleton is also involved in the nuclear localization and transcriptional activity of YAP/TAZ (Wada et al., 2011). Vinculin can bind to F-actin and change actin organization in other cell types (Bourdet-Sicard et al., 1999; Wen et al., 2009). Therefore, we examined whether the actin cytoskeleton affects the vinculin-mediated enhancement of YAP/TAZ nuclear localization and activity. First, control or vinculin-depleted cells were cultured in the presence of cytochalasin D (cytoD), a pharmacological inhibitor of actin polymerization, and YAP/TAZ nuclear localization was investigated. Phalloidin staining indicated that the cytoD treatment disrupted the actin stress fibers in both control and vinculin-depleted cells, although there were no apparent differences between control and vinculin-depleted cells in our experimental conditions. The cytoD treatment suppressed YAP/TAZ nuclear localization in control cells, but hardly affected the nuclear localization of YAP/TAZ in vinculin-depleted cells (Fig. 5A,B). To evaluate the effect of cytoD on transcriptional activity of YAP/TAZ, the relative expression of Ctgf and Ankrd1 mRNA was examined (Fig. 5C). As expected, cytoD treatment decreased the expression of Ctgf and Ankrd1 mRNA by up to 30% compared to vehicle in control cells. Vinculin depletion abrogated the decrease in the Ctgf and Ankrd1 expression induced by the cytoD treatment. Taken together, these results suggest that the vinculin-mediated enhancement of YAP/TAZ nuclear localization and activity involves actin organization.

To further gain insight into the relation between actin organization and vinculin, we used the vinculin mutants I997A (denoted IA, deficient in actin binding) and T12 (constitutively active) (Thompson et al., 2014; Cohen et al., 2005). GFP-fused vinculin mutants were expressed in ST2 vinculin-depleted cells (Fig. S2H). Then, the proportion of YAP/TAZ that showed nuclear localization in these cells on PAA substrates was evaluated. In vinculin-depleted cells re-expressing WT vinculin, YAP/TAZ nuclear localization was increased by 30% (±5.9, mean±s.e.m.) on rigid substrates (8.7 kPa) compared to on soft substrates (1.5 kPa), and showed ECM stiffness-dependent YAP/TAZ nuclear localization (Fig. 5D,E). On the other hand, in vinculin-depleted cells re-expressing IA vinculin, the YAP/TAZ nuclear localization was lower than WT on rigid substrates (8.7 kPa) and no significant difference was observed between soft and rigid substrates. In contrast, the YAP/TAZ nuclear localization in vinculin-depleted cells re-expressing T12 was high even on soft substrate (1.5 kPa). Much higher nuclear localization was observed on rigid substrates (8.7 kPa) in these T12-expressing cells compared to both that on soft substrates for T12-expressing cells and that on rigid substrates for WT vinculin-expressing cells. These results strongly suggest that actin binding of vinculin is involved in the regulation of YAP/TAZ.

We next investigated the effects of vinculin on the differentiation into adipocytes. First, we tested the differentiation of vinculin-depleted cells into adipocytes on rigid (plastic) substrates. Quantitative real-time PCR (qRT-PCR) analysis showed that vinculin depletion markedly increased the expression of mRNA encoding PPARγ2 and aP2 (Fig. 6A). Western blotting showed an increased expression of aP2 protein in vinculin-depleted cells (Fig. 6B). The differentiation was further assessed by measuring the accumulation of lipid droplets using Oil Red O staining. Vinculin-depleted cells contained more lipid droplets than control cells (Fig. 6C,D). These data suggest that vinculin depletion promotes the adipocyte differentiation on rigid substrates. To confirm the effects of vinculin depletion on the differentiation into adipocytes, vinculin-depleted cells re-expressing vinculin were induced to differentiate into adipocytes on rigid (plastic) substrates. Expression of mRNA encoding PPARγ2 and aP2 was lower in vinculin re-expressing cells than in vinculin-depleted cells (Fig. 6E), and the amount of lipid droplets was smaller than in vinculin-depleted cells (Fig. 6F,G). These results indicate that vinculin suppresses adipocyte differentiation on rigid substrates.

To check whether vinculin also affects the differentiation into osteoblasts on rigid substrates, ALP mRNA expression was quantified. Vinculin depletion significantly decreased the mRNA expression of ALP (Fig. S3A), and vinculin re-expression rescued this decrease (Fig. S3B). Taken together, these observations indicate that vinculin regulates the balance between adipogenic and osteogenic lineage commitment.

We next investigated the effect of vinculin depletion on ECM stiffness-dependent adipocyte differentiation. Vinculin-depleted cells were cultured on PAA substrates and induced to differentiate into adipocytes. Expression of mRNA encoding PPARγ2 and aP2 expression was 2-fold lower on rigid substrates (8.7 kPa) than those on soft substrates (1.5 kPa) in control cells, as with the WT ST2 cells, indicating an ECM stiffness-dependent differentiation (Fig. 7A). By contrast, although vinculin depletion led to a 2–4-fold increase in adipocyte differentiation compared with control cells on soft substrates, there was no significant difference in the expression of adipogenic markers between vinculin-depleted cells cultured on soft and rigid substrates. Stiffness-dependent differentiation was further investigated by Oil Red O staining (Fig. 7B). In control knockdown cells, the area stained by Oil Red O on soft substrates (1.5 kPa) was 3.0-fold (±0.85, mean±s.e.m.) larger than that on rigid substrates (8.7 kPa). On the other hand, the stained area was not changed by ECM stiffness in vinculin-depleted cells (Fig. 7C). Taken together, these results suggest that vinculin depletion attenuates the ECM-stiffness dependency of adipocyte differentiation.

Because we found that association of vinculin with actin is involved in regulation of YAP/TAZ nuclear localization, we examined the role of this association in adipocyte differentiation. Vinculin-depleted cells re-expressing WT or IA vinculin were cultured on rigid (plastic) substrates. Whereas re-expression of WT vinculin reduced expression of mRNA encoding PPARγ2 and aP2 to 25% (±1.8) (PPARγ2) or 28% (±1.3) (aP2), re-expression of IA vinculin only reduced these to 82% (±8.2) (PPARγ2) or 82% (±7.5) (aP2) (mean±s.e.m.; Fig. 7D). This result suggests that IA vinculin inhibits adipocyte differentiation less efficiently than WT vinculin. Oil Red O staining also confirmed the weak effect of IA vinculin on adipocyte differentiation (Fig. 7E,F). Collectively, these observations suggest that vinculin regulates ECM stiffness-dependent adipocyte differentiation through its association with actin.

Interaction of vinculin with vinexin α is involved in ECM stiffness-dependent cell migration in mouse embryonic fibroblasts (Yamashita et al., 2014). Thus, we also examined the effect of vinexin depletion on differentiation into adipocytes, but did not detect any increase in differentiation (data not shown), possibly due to other functions of vinexin such as regulating signal molecules. To assess this possibility further, we used P2 mutant vinculin, which has a reduced affinity to vinexin (Yamashita et al., 2014). To check whether the P2 mutation affects vinculin activation in ST2 cells, we first examined the resistance of vinculin to CSK treatment on rigid substrates (8.7 kPa) (Fig. S3D,E). Consistent with previous work (Yamashita et al., 2014), the P2 mutation decreased the amount of CSK-resistant vinculin (the vinculin-positive FA number, area and intensity) compared to that seen with WT vinculin (Fig. S3E). Next, to test whether the P2 mutation affects ECM stiffness-dependent YAP/TAZ nuclear localization, WT and P2 vinculin-expressing cells cultured on PAA substrates were analyzed by immunostaining (Fig. S3F,G). Rigid substrates increased YAP/TAZ nuclear localization by 30% (±6.3 mean±s.e.m.) compared to that seen on soft substrates for WT vinculin-expressing cells. On the other hand, rigid substrates increased the localization only moderately (20±6.3%) in P2-vinculin expressing cells. Furthermore, YAP/TAZ nuclear localization on rigid substrate in P2 vinculin-expressing cells was significantly lower than that on rigid substrate in WT vinculin-expressing cells. Although we also investigated the effect of P2 mutation on adipocyte differentiation, this mutation did not affect the mRNA expression of adipogenic markers either on substrates with different levels of stiffness or on extremely rigid plastic dishes (data not shown), suggesting that the moderate effect of the P2 mutation on YAP/TAZ localization is not enough for the regulation of adipocyte differentiation. Taken together, these results suggest that vinculin–vinexin interaction is involved in regulating YAP/TAZ nuclear localization.

In addition to ECM stiffness, cell spreading and morphology can affect the adipocyte differentiation of MSCs as well as YAP/TAZ nuclear localization (McBeath et al., 2004). Thus, it is possible that vinculin depletion regulates the adipocyte differentiation through cell spreading and morphology. To investigate that possibility, the effects of vinculin depletion on cell spreading and morphology were investigated. Cells were cultured on rigid substrates (glass), and then the cell-spreading area and aspect ratio were quantified (Fig. S4A,B). Quantitative data revealed no significant differences in cell area or in aspect ratio between control and vinculin-depleted cells. Furthermore, vinculin depletion did not affect the cell-spreading area and aspect ratio on PAA substrates with different levels of stiffness (1.5 kPa, 3.2 kPa and 8.7 kPa) (data not shown). These data indicate that cell spreading and morphology are not involved in vinculin-mediated regulation of adipocyte differentiation.

To examine whether the role of vinculin in YAP/TAZ regulation and adipocyte differentiation is shared among MSCs, we also tested the effect of vinculin depletion using UE7T-13 cells, a human MSC cell line (Mori et al., 2005). Vinculin-depleted cells were established using shRNA. First, we checked the effect of vinculin depletion on the YAP/TAZ transcriptional activity. Vinculin-depleted human MSCs showed lower mRNA expression of YAP/TAZ target genes, e.g. CTGF and ANKRD1 (Fig. S4C). Next, we examined the effect on the adipocyte differentiation. The vinculin-depleted human MSCs were induced to differentiate into adipocytes. The mRNA expression of adipogenic markers was investigated. We found that the expression of mRNA encoding PPARγ2 and aP2 in vinculin-depleted adipocytes was significantly higher than in control cells (Fig. S4D). Furthermore, depletion of vinculin through transient transfection of small interfering (si)RNAs directed against human vinculin into UE7T-13 cells upregulated both mRNA and protein expression of aP2 in siRNA-transfected cells (data not shown). These results suggest that vinculin-mediated regulation of YAP/TAZ activity and suppression of adipocyte differentiation is not limited to ST2 cells but is a shared property of MSCs.

To examine whether YAP or TAZ was responsible for suppression of differentiation into adipocytes by vinculin, siRNAs targeted against YAP or TAZ were transfected into control and vinculin-depleted cells and the effect of YAP or TAZ knockdown on the adipocyte differentiation was examined. Because combinatorial treatment of YAP or TAZ knockdown with adipogenic stimulation (IBMX, insulin and dexamethasone) severely damaged ST2 cells under our conditions, the effect of YAP or TAZ knockdown on the differentiation without adipogenic stimulation was examined. Under this condition, cells still differentiate into adipocytes although the efficiency is moderate. Although YAP siRNA transfection efficiently reduced the expression of YAP, YAP knockdown only slightly affected the expression of both PPARγ2 and aP2 in WT ST2 cells (Fig. S4E). Transfection of siRNAs against TAZ efficiently reduced the TAZ expression (Fig. 8A). Knockdown of TAZ increased the expression of aP2-encoding mRNAs by 2.6±0.15-fold (siRNA#1) or 1.7±0.15-fold (siRNA#2) in control cells, whereas knockdown of TAZ slightly increased the expression [1.7±0.08-fold (siRNA#1) or 1.1±0.05-fold (siRNA#2)] in vinculin-depleted cells (mean±s.e.m.; Fig. 8B). PPARγ2 expression was also increased efficiently by TAZ knockdown in control cells [2.2±0.05-fold (siRNA#1) or 1.6±0.05-fold (siRNA#2)] and moderately in vinculin-depleted cells [1.7±0.03-fold (siRNA#1) or 1.3±0.002-fold (siRNA#2)]. These results indicate that TAZ knockdown is less effective in vinculin-depleted cells because vinculin depletion already suppresses TAZ nuclear localization and activity. To further analyze this effect, we also transfected TAZ siRNA in combination with YAP siRNA. Similar to the results with TAZ knockdown alone, simultaneous knockdown of YAP and TAZ increased the level of mRNAs encoding PPARγ2 and aP2 in control cells, whereas knockdown of YAP and TAZ only slightly affected this in vinculin-depleted cells (Fig. S4F,G). Taken together, these results indicated that TAZ, at least partially, inhibits adipocyte differentiation downstream of vinculin.

Physical properties of the extracellular microenvironment have emerged as an important factor controlling the differentiation of stem cells. ECM stiffness can direct the lineage commitment of MSCs. They differentiate into neuroblasts or adipocytes on soft ECM and into osteoblasts on rigid ECM (Engler et al., 2006). The transcriptional coactivators YAP and TAZ are known as mediators of mechanical signals; however, how ECM stiffness regulates lineage specification and which molecules are involved in the YAP/TAZ regulation remain unclear. Here, we show that rigid ECM suppressed the differentiation of ST2 mesenchymal cells into adipocytes and promoted the differentiation into osteoblasts. Vinculin status, estimated by its resistance to CSK treatment, was regulated by ECM stiffness. Vinculin promoted the nuclear localization and activity of YAP/TAZ on rigid substrates, and suppressed the adipocyte differentiation and promoted osteoblast differentiation on rigid ECM. Furthermore, vinculin depletion attenuated the ECM stiffness-dependent nuclear localization of YAP/TAZ and adipocyte differentiation. Collectively, these observations indicate that vinculin functions as a sensor for ECM stiffness and regulates the differentiation of MSCs via promoting the nuclear localization of YAP/TAZ.

During adipocyte differentiation both in vitro and in vivo, the type and density of ECM components dramatically change. These changes lead to the alteration of cell morphology and cytoskeletal organization, and affect the differentiation through FAs. Although, some studies have focused on the involvement of signaling molecules localized at FAs, including FAK (also known as PTK2), in adipocyte differentiation (Lee et al., 2013; Thompson et al., 2013; Uzer et al., 2015), the functions of cytoskeletal FA proteins such as talin and vinculin in adipogenesis have not been studied. Here, our observations showed that vinculin suppressed the differentiation into adipocytes and promoted differentiation into osteoblasts on rigid ECM. To our knowledge, the present study is the first study that sheds light on the FA scaffold protein vinculin as a regulator of adipocyte differentiation.

We identified here a novel role for vinculin as a regulator of YAP/TAZ in MSCs. Vinculin promoted the nuclear localization and activity of YAP/TAZ on rigid substrates. Although the Hippo signaling pathway is known to regulate the localization and activity of YAP/TAZ to mediate the contact inhibition of cell proliferation and the control of organ size (Dong et al., 2007; Zhao et al., 2007), the phosphorylation of LATS1, a YAP kinase in the Hippo pathway, was not affected by vinculin in the present study. This is consistent with a previous report showing that ECM stiffness-dependent YAP/TAZ regulation is independent of LATS1 phosphorylation (Dupont et al., 2011). Importantly, vinculin depletion did not affect the localization and activity of YAP/TAZ under F-actin-disrupted conditions, indicating that actin organization is essential for the vinculin-mediated regulation of YAP/TAZ. The requirement of proper actin organization for mechanical regulation of YAP/TAZ has also been reported by others. Stress fiber formation is an important factor in cell morphology- and ECM stiffness-directed YAP/TAZ regulation (Dupont et al., 2011; Wada et al., 2011). Furthermore, Aragona et al. showed that knockdown of the F-actin capping or severing proteins CapZ, cofilin and gelsolin, increases F-actin bundling and promotes the nuclear localization of YAP/TAZ on soft ECM (Aragona et al., 2013). Interestingly, vinculin directly binds to F-actin (Jockusch and Isenberg, 1981; Menkel et al., 1994) and has F-actin-bundling and -capping activities (Le Clainche et al., 2010; Wen et al., 2009). Here, we demonstrated that actin-binding-deficient vinculin I997A (Thompson et al., 2014) lacks the ability to promote the change in the YAP/TAZ nuclear localization that is dependent on ECM stiffness. Taken together, these observations indicate that vinculin regulates YAP/TAZ at least partly through organization of actin cytoskeleton.

Here, we showed that vinculin status, as estimated by the resistance to CSK treatment, was regulated by ECM stiffness in ST2 MSCs, and that ECM stiffness-dependent differentiation required vinculin expression. We also demonstrated that mutation of the vinexin-binding site in vinculin reduced the ability of vinculin to promote YAP/TAZ nuclear localization. Our previous research revealed that the interaction of vinculin with vinexin α functions as a mechanosensor for ECM stiffness, and this interaction is a prerequisite for the stiffness-dependent regulation of vinculin status and cell migration in mouse embryonic fibroblasts (Yamashita et al., 2014). In agreement with our findings, Holle et al. reported that vinculin and vinexin have an essential role in myogenic differentiation (Holle et al., 2016) during the process of revising our manuscript. Thus, the present study extends our understanding of the vinculin function in mechanosensing and mechanotransduction to the regulation of cell differentiation and to MSCs. In addition to the function of vinculin as a sensor of ECM stiffness, vinculin also can sense other physical microenvironments. Janoštiak et al. have also reported that vinculin binding to p130CAS serves as a sensor for stretch and regulates the stretch-induced activation of p130CAS (Janoštiak et al., 2014). Grashoff et al. reported the requirement of tension to activate vinculin (Grashoff et al., 2010). Vinculin binding to talin at FAs occurs under force-generating conditions (Hirata et al., 2014). Taken together, these observations suggest that vinculin is used as a ‘mechanosensor’ for various physical cues.

Another important finding in this study is that vinculin determines the lineage commitment of ST2 mesenchymal stem cells. Rigid ECM promotes the nuclear localization of YAP/TAZ and osteogenesis but suppresses adipogenesis (Dupont et al., 2011; Hwang et al., 2015). TAZ regulates the balance between adipogenesis and osteogenesis (Hong et al., 2005). Here, we show that vinculin promotes expression of the osteogenic gene, ALP, while it inhibits the differentiation into adipocytes. Furthermore, knockdown of TAZ by siRNA dampened the vinculin-mediated suppression of differentiation into adipocytes on rigid ECM. Recent reports have also shown that vinculin can regulate myogenesis (Holle et al., 2013) and chondrogenesis (Koshimizu et al., 2012) in a stiffness-dependent or -independent manner, respectively. Taken together, these results show that vinculin has a significant role in mesenchymal lineage commitment.

We demonstrated that vinculin promotes the activity of YAP/TAZ. YAP/TAZ is notable for its role in cancer progression. YAP has oncogenic potential, leading to epithelial-to-mesenchymal transition and cell proliferation in breast cancer (Overholtzer et al., 2006; Sorrentino et al., 2014; Wang et al., 2015). While cancer tissues have more rigid ECM than with normal tissues, YAP is activated and contributes to ECM stiffening in cancer-associated fibroblasts (Calvo et al., 2013). Vinculin is also involved in metastasis and malignancy of various cancers, including pancreatic, prostate and breast cancers (Rubashkin et al., 2014; Ruiz et al., 2011; Wang et al., 2012). Our results indicate that vinculin could regulate tumor progression through YAP/TAZ, in addition to through signals already known to be involved in tumorigenesis.

While we were preparing the manuscript, Elosegui-Artola et al. reported that talin regulates YAP/TAZ nuclear localization on substrates above 11 kPa, in accordance with the talin-unfolding state in mouse embryonic fibroblasts (Elosegui-Artola et al., 2016). They also showed that expression of the vinculin head domain, which could work as a dominant-negative mutant, suppressed the effect of talin. Our present study demonstrated that vinculin modulated the ECM stiffness-dependent YAP/TAZ nuclear localization on substrates between 1.5 kPa and 8.7 kPa in ST2 mesenchymal stem cells. Under this condition, ECM stiffness did not affect the subcellular localization of vinculin at FAs but increased the amount of activated and cytoskeleton-associated vinculin. These observations suggest that vinculin could work as a mechanosensor in at least two fashions: one through detecting the talin unfolding on substrates above 11 kPa and the other through modulating vinculin activation on substrates between 1.5–8.7 kPa independently of talin unfolding. Alternatively, vinculin could work as a mechanosensor at different ranges of stiffness in different cellular contexts. Taken together, these observations indicate that ECM stiffness-dependent YAP/TAZ nuclear localization involves FA mechanosensitive proteins, including talin and vinculin.

In summary, we demonstrated that vinculin promotes the ECM stiffness-dependent nuclear localization of YAP/TAZ and suppresses adipocyte differentiation in a manner that depends on ECM stiffness. The effect of TAZ knockdown on adipocyte differentiation was less effective in vinculin-depleted cells than in control cells. Thus, we propose a model, in which vinculin senses ECM stiffness and regulates differentiation through promoting the nuclear localization of TAZ (Fig. 8C).

Mouse anti-vinculin [V9131; 1:20,000 western blotting (WB); 1:500 immunofluorescence (IF)] and anti-β tubulin (T4026; 1:2000) antibodies were purchased from Sigma (Saint Louis, MO). Mouse anti-β-actin (ab6276; 1:10,000) and rabbit anti-vinculin (ab73412; 1:100 IF) antibodies were purchased from Abcam (Cambridge, UK). Mouse anti-YAP (sc-101199; 1:100 IF, 1:1000 WB), rabbit anti-ERK2 (sc-154; 1:10,000) and goat anti-lamin B (sc-6216; 1:100) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit anti-aP2 (#3544; 1:2000), rabbit anti-phospho-YAP (#4911; 1:1,000), rabbit anti-LATS1 (#3477; 1:1000) and rabbit anti-phospho LATS1 (#9157; 1:1000) antibodies were purchased from Cell Signaling Technology (Boston, MA). Rabbit anti-vinexin polyclonal antibody was described previously (Kioka et al., 1999). Alexa-Fluor-568–phalloidin was purchased from Thermo (Rockford, IL). Type I collagen was purchased from Nitta Gelatin (Osaka, Japan). Insulin, IBMX and cytochalasin D were purchased from Sigma-Aldrich.

ST2, a mouse bone-marrow-derived mesenchymal stem cell line, and UE7T-13, a human bone-marrow-derived mesenchymal stem cell line, were obtained from RIKEN BRC (Tsukuba, Japan) and JCRB Cell Bank (Osaka, Japan), respectively.

ST2 cells were maintained in RPMI1640 (Sigma) supplemented with 10% fetal bovine serum (FBS; Sigma, or Biosera, Boussens, France) at 37°C under a humidified atmosphere containing 5% CO₂. For adipocyte differentiation, 7.5×10⁴ (or 1.5×10⁵ in some cases) cells were seeded into six-well plates. Adipocyte differentiation was induced by the addition of 0.5 mM IBMX, 10 µg/ml insulin and 10 µM DEX. After incubation for 2 days, the medium were changed to that containing 5 µg/ml insulin, and then to maintenance medium every 2 days thereafter.

UE7T-13 cells were maintained in DMEM (Nacalai Tesque, Kyoto, Japan) supplemented with 10% FBS. For adipocyte differentiation, medium was changed to differentiation medium (0.5 mM IBMX, 10 µg/ml insulin and 1 µM DEX) every 2 days.

Vinculin-depleted cells and vinculin re-expressing cells were established using lentivirus as described previously (Yamashita et al., 2014), with slight modifications. Briefly, lentiviruses for knockdown were generated by transfecting pMD2.G, psPAX2 (Addgene, Cambridge, MA) and pLKO.1 vectors, and lentiviruses for expression were generated by transfecting pMD2G, pRSV-Rev, pMDLg/pRRE (Addgene) and pCDH vectors. pLKO.1 harboring vinculin small hairpin RNA (mouse 5′-CCCTGTACTTTCAGTTACTAT-3′, or human 5′-GCCTAACAGGGAAGAGGTATT-3′) was purchased from Openbiosystems (Waltham, MA). pCDH-EF1-IRES-hygro vector were purchased from System Biosciences (Mountain View, CA) and vinculin was subcloned into the vector in a similar manner to that previously described (Yamashita et al., 2014). ST2 cells were infected with lentiviruses, followed by incubating with medium containing 3 µg/ml puromycin or 100 µg/ml hygromycin B. UE7T-13 cells were infected with lentiviruses, followed by incubating with medium containing 0.5 µg/ml puromycin.

ST2 cells were transfected with 2 nM Stealth™ RNAi siRNAs against TAZ (siTAZ#1, 5′-GGAAGGUGAUGAAUCAGCCUCUUGG-3′; siTAZ#2, 5′-GGAGUCCUUCUUUAAGGAGCCCGAU-3′; Invitrogen, Carlsbad, CA) or control siRNA (Stealth^TM RNAi Negative Control Medium GC Duplex#3, Cat. No. 12935-113) using Lipofectamine RNAiMAX Reagent (Invitrogen) as previously described (Tomiyama et al., 2015). UE7T-13 cells were transfected for 24 h with 20 nM Stealth RNAi™ siRNAs against human vinculin (siVinculin#1, VCL-HSS111259; siVinculin#2, VCL-HSS187662; Invitrogen) using Lipofectamine RNAiMAX Reagent (Invitrogen).

PAA gel substrate was prepared as previously described with slight modification (Damljanović et al., 2005; Yamashita et al., 2014). Briefly, 35 µl of the acrylamide mix reagents (final proportions were 5–8% acrylamide monomer, 0.03–0.3% bisacrylamide, 0.05% ammonium persulfate and 0.1% TEMED) was dropped onto glass coverslips or glass-bottom dishes. The polyacrylamide gels were coated with 200 µg/ml type I collagen (Nitta Gelatin) using sulfo-SANPAH (ProteoChem, IL). The elastic moduli of the gels were from 1.5 kPa to 23 kPa.

Whole-cell lysates were prepared in lysis buffer [1% SDS and 1% protease inhibitor cocktail (Roche) in PBS]. Amounts of lysates obtained from cells cultured on PAA gels were adjusted according to the DNA concentration. DNA concentration was measured using the Qubit dsDNA HS Assay Kit (Invitrogen). Protein concentrations of lysates obtained from cells on plastic dishes were determined with BCA reagents (Thermo) and equal amounts of lysates were analyzed by SDS-PAGE and western blotting using specific antibodies. Secondary antibodies conjugated to horseradish peroxidase (HRP) were used, and bound antibodies were detected using Immunostar LD/Zeta (Wako, Osaka, Japan) and EZ-Capture (ATTO, Tokyo, Japan).

Cells were washed twice with PBS and fixed with 3.7% formaldehyde in PBS for 15 min at room temperature. After washing with 60% isopropanol in distilled water, the cells were stained with 0.18% (w/v) Oil Red O in 60% isopropanol for 30 min, followed by washing with 60% isopropanol once and PBS twice. Images of stained cells were obtained by microscopy (Nikon ECLIPSE TE300-2) equipped with a Leica MC120HD camera. For cells cultured on PAA gels, an additional wash with 40% isopropanol and 20% isopropanol was performed. Quantification of Oil Red O-stained area was performed using ImageJ software (National Institutes of Health, Bethesda, MD).

qRT-PCR was performed as described previously (Sezaki et al., 2012). Briefly, total RNA was extracted using an RNeasy mini kit (QIAGEN, Hilden, Germany), and cDNA was synthesized using Super Script reverse transcriptase III (Invitrogen). qRT-PCR analysis was carried out with Step One^TM Real-Time PCR Systems (Applied Biosystems) using Fast SYBR^® Green Master Mix (Applied Biosystems). Relative expression levels to internal control GAPDH (or 36B4, RPS18 in some cases) were given. Sequences of specific primers used in this paper are summarized in Table S1.

Immunostaining was performed to analyze the total amount of vinculin or the amount of vinculin that was resistant to CSK buffer treatment as described previously (Yamashita et al., 2014), with slight modifications. Briefly, to analyze total vinculin, cells cultured on PAA gel substrates were fixed with 1.5% paraformaldehyde at 4°C for 45 min, followed by permeabilization with PBS containing 0.2% Triton X-100 for 5 min at room temperature. To analyze CSK-resistant vinculin, cells were first treated twice with CSK buffer (0.1% Triton X-100, 10 mM PIPES, pH 6.8, 50 mM NaCl, 3 mM MgCl₂, 300 mM sucrose) at 4°C for 30 s, followed by fixation with 4% paraformaldehyde for 15 min at room temperature. Cells were treated with 10% goat serum (or donkey serum in some cases) in PBS for blocking, and were incubated with primary antibodies at 4°C overnight and with secondary antibodies for 1 h. Images were obtained with a LSM700 confocal microscope (Carl Zeiss, Oberkochen, Germany) with a ×40 oil-immersion Plan-APOCHROM objective lens, or with a Nikon C2 confocal microscope (Nikon, Tokyo, Japan) with a ×40 Plan-APO objective lens. Quantification of the total number and average area of vinculin at FAs per cell was performed using ImageJ software as described previously (Yamashita et al., 2014).

Cells were seeded on PAA gels or collagen-coated glass coverslips at a density of 7.5×10⁴ or 1.5×10⁵ cells per well. At 24 h after seeding, cells were fixed with 4% paraformaldehyde for 15 min, followed by permeabilization with PBS containing 0.2% Triton X-100 for 5 min. Immunostaining was performed as described above. YAP/TAZ nuclear localization was evaluated by assessing the nucleus-to-cytosol intensity ratio. Intensity inside the nucleus and just outside the nucleus was quantified using ImageJ.

Statistical analysis was performed in Origin 8.5.1 software. Values that displayed a normal distribution were analyzed with a Tukey's honest significant difference test after one-way ANOVA or Student's t-test for comparison among three or more groups or between two groups, respectively. Values that did not display a normal distribution were analyzed with a Mann–Whitney U-test after Kruskal–Wallis's ANOVA for comparison among three or more groups.

We thank T. Ichikawa and T. Omachi (Kyoto University, Japan) for providing materials.