An amphipathic helix of vinexin α is necessary for a substrate stiffness-dependent conformational change in vinculin

Extracellular matrix (ECM) stiffness regulates various cellular behaviors, including cell proliferation (Klein et al., 2009; Wang et al., 2000), differentiation (Engler et al., 2006) and migration (Lo et al., 2000; Nagasato et al., 2017; Yamashita et al., 2014). In addition, some types of tumor cells generate stiffened ECM, which promotes cell proliferation and metastasis (Levental et al., 2009; Provenzano et al., 2009). Therefore, understanding how cells sense ECM stiffness is important for tissue engineering and cancer therapy.

Focal adhesions (FAs), integrin-based protein complexes formed at cell–ECM contact sites, play critical roles in sensing ECM stiffness. FAs function as mechanical linkages between ECM and force-generating actomyosin, and, thus, forces are exerted on FAs. Because stiff ECM is less deformable than soft ECM, greater forces are thought to be loaded on FAs on rigid ECM (Sun et al., 2016). ECM stiffness-dependent force seems to cause a conformational change in mechanosensors localized at FAs, which translate information from physical cues into biochemical signals and gene expression.

Vinculin, an FA scaffold protein, plays a crucial role in sensing ECM stiffness due to its direct binding to the actin cytoskeleton (Hirata et al., 2014; Omachi et al., 2017; Thievessen et al., 2013; Yamashita et al., 2014). Vinculin is composed of an N-terminal head region and a C-terminal tail region connected by a proline-rich flexible linker region (Bakolitsa et al., 2004; Borgon et al., 2004) and interacts with various proteins, including talin (TLN1 and TLN2 in mammals), through the head region, and F-actin through the tail region (Gilmore et al., 1993; Johnson and Craig, 1995a). The head–tail intramolecular interaction keeps vinculin ‘closed’ and inhibits its interaction with F-actin. Once one of the binding partners, such as talin, α-catenin and α-actinin, binds to the vinculin head domain through its amphipathic α-helix, the conformation of vinculin changes into an ‘open’ form that has a higher affinity for F-actin (Bois et al., 2005; Izard et al., 2004; Peng et al., 2012). In addition, rigid ECM as well as cytoskeletal forces are considered to be required for the vinculin conformational change (Carisey et al., 2013; Humphries et al., 2007; Liu et al., 2016; Omachi et al., 2017; Yamashita et al., 2014). A molecular dynamics simulation has also indicated that external forces promote the conformational change in vinculin (Golji and Mofrad, 2013; Sun et al., 2017).

Vinexin α (an isoform encoded by the SORBS3 gene), another FA scaffold protein, cooperates with vinculin to work as a mechanosensor of ECM stiffness (Ichikawa et al., 2017; Kuroda et al., 2017; Yamashita et al., 2014). On rigid ECM, vinexin α interacts with vinculin in an actomyosin contractility-dependent manner and induces the conformational change in vinculin to an open form, promoting cell migration and the nuclear localization of the transcriptional co-activators, Yes-associated protein (YAP, also known as YAP1) and transcriptional coactivator with a PDZ-binding motif (TAZ, also known as WWTR1) (hereafter YAP/TAZ). However, the detailed mechanism by which vinexin α induces the conformational change in vinculin on rigid ECM is still unclear. Interestingly, vinexin β, a splice variant that lacks the N-terminal half of vinexin α, interacts with vinculin through the same Src homology 3 (SH3) domains as vinexin α (Kioka et al., 1999) but does not induce the ECM stiffness-dependent behaviors of vinculin (Yamashita et al., 2014). This indicates that the SH3 domains of vinexin α are not sufficient to induce the conformational change in vinculin and suggests that additional molecular mechanisms underlie the vinculin–vinexin α interaction.

Talin, an FA protein, has also been shown to play a crucial role in mechanosensing (Austen et al., 2015; Elosegui-Artola et al., 2016). Talin interacts with the cytoplasmic tail of integrins through its N-terminal head domain (Calderwood et al., 1999) and with F-actin through its C-terminal rod domain (Hemmings et al., 1996) and therefore is subject to mechanical forces. Mechanical force loading on talin induces a conformational change in helix bundles in the rod domains, resulting in exposure of cryptic vinculin-binding sites (VBSs) and recruitment of vinculin to FAs (Ciobanasu et al., 2014; del Rio et al., 2009; Hirata et al., 2014). Therefore, the talin–vinculin interaction also functions as a sensor for ECM stiffness. However, the relationship between talin–vinculin and vinculin–vinexin α mechanosensing systems is poorly understood.

Here, we identify an amphipathic helix named H2 in the N-terminal region of vinexin α as a novel vinculin-binding region. H2 helix interacts with the vinculin D1b subdomain and promotes the formation of a talin–vinculin–vinexin α ternary complex. Mutations in the H2 region not only impair the ability of vinexin α to induce the ECM stiffness-dependent conformational change in vinculin but also diminish nuclear localization of YAP/TAZ on rigid ECM. However, the H2 region alone does not induce the conformational change in vinculin. Taken together, these results indicate that vinexin α interacts with vinculin through both its SH3 domains and H2 helix and induces the vinculin conformational change, which plays a pivotal role in sensing ECM stiffness.

Vinexin α, but not vinexin β, is necessary for the ECM-stiffness dependent behavior of vinculin (Yamashita et al., 2014). This observation indicates that an N-terminal region of vinexin α, which is not included in vinexin β, plays a critical role in regulation of vinculin behavior (Fig. 1A). However, the molecular structure and the function of the N-terminal region is poorly elucidated. To further analyze the importance of this region, secondary structure prediction was performed using the program PORTER (Pollastri and McLysaght, 2005). Although most of the region was predicted to be random coils or disordered regions, a few regions were predicted to form α-helices (Fig. S1A). Since vinculin binding partners, such as talin, α-actinin and IpaA, interact with vinculin through their amphipathic helices to induce the conformational change in vinculin (Bois et al., 2005; Izard et al., 2004, 2006), we focused on two regions named predicted Helix1 (H1) (residues 193–214) and predicted Helix2 (H2) (residues 329–355), which were predicted to form amphipathic helices (Fig. 1A,B). To reveal a role for these amphipathic helices in the interaction with vinculin, we constructed five vinexin α mutants (H1Pa, H1Pb, H2Pa, H2Pb, and H2Pc) in which hydrophobic residues in the H1 region (H1Pa and H1Pb) or the H2 region (H2Pa, H2Pb, and H2Pc) were replaced with proline residues to disrupt the secondary structure (Fig. 1C). Because GST-tagged full-length vinexin α is highly degraded in E. coli, as described previously (Yamashita et al., 2014), truncated mutant vinexin αΔN, which lacks residues 1–156 of vinexin α, and αΔN mutants having mutations in H1 or H2 helices were used to perform pulldown assays. Vinexin αΔN wild type (WT) co-precipitated with vinculin more efficiently than vinexin β (Fig. 1D,E), suggesting a possible binding region in the N-terminal region of vinexin α. Interestingly, the H2Pa, H2Pb and H2Pc mutations dramatically decreased the amount of co-precipitated vinculin to a level comparable with that obtained with vinexin β. By contrast, the H1Pa and H1Pb mutations did not affect the co-precipitation. These results clearly indicate that the H2 region is required for the high-affinity interaction between vinexin α and vinculin.

To examine whether the H2 region can adopt an α-helical conformation, the secondary structure of a chemically synthesized H2 peptide (residues 329–355) was analyzed by circular dichroism (CD) spectroscopy. As shown in Fig. S1B, the H2 peptide exhibited a spectrum pattern similar to that of synthesized talin VBS3 peptide (residues 1945–1970), which has already been revealed to form an amphipathic helix (Izard et al., 2004). These peptides showed particular negative peaks near 200 and 220 nm, which can be attributed to the superposition of negative peaks derived from a typical α-helix at 210 and 220 nm and from a random coil at 200 nm (Greenfield and Fasman, 1969). Therefore, this result suggests that the H2 region can adopt an α-helical conformation that is similar to the talin VBS3 region.

Alignment of vinculin-binding amphipathic helices among talin, α-actinin and α-catenin showed the conserved hydrophobic residues in these helices, which are involved in binding to vinculin (Bois et al., 2005; Izard et al., 2004; Peng et al., 2012), and H2 helix includes similar hydrophobic residues (Fig. 2A). Therefore, to investigate the role of the hydrophobic residues in H2 helix in the interaction with vinculin, we constructed additional vinexin α mutants (H2Q1, H2Q2, H2Q3 and H2Q4) whose hydrophobic residues were replaced with glutamine residues to impair the hydrophobic properties (Fig. 2B). GFP-tagged vinexin mutants expressed in 293T cells were pulled down using GST–vinculin. As shown in Fig. 2C, GFP–vinexin α H2Q2 and H2Q3 mutants co-precipitated with GST–vinculin much less than GFP–vinexin α WT. The H2Q1 and H2Q4 mutant showed a modest effect on the co-precipitation. To further confirm the effect of these mutations on vinculin binding, the depletion of GFP fluorescence in the supernatant was determined to analyze the fraction of the GFP-tagged vinexin mutants that bound to vinculin (Fig. 2D). Consistent with the pulldown assay, the H2Q2 and H2Q3 mutations significantly decreased the bound fraction to a level that was comparable with that found for vinexin β. These results suggest that the hydrophobicity of H2 helix is necessary for the high-affinity interaction between vinexin α and vinculin.

To investigate whether the H2 helix directly interacts with vinculin or allosterically modulates the binding of vinexin α to vinculin, we tested whether the H2 helix itself has a vinculin-binding ability. The H2 helix (residues 329–355) was purified as a GST-tagged protein, and vinculin binding was tested with a pulldown assay. GFP–vinculin was co-precipitated with the GST–vinexin α H2 helix, while introduction of the H2Q3 mutation into GST–vinexin α H2 helix clearly impaired the co-precipitation (Fig. 2E). This result indicates that H2 helix directly interacts with vinculin.

Previous studies have shown that the SH3 domains of vinexin α interact with the vinculin linker region (Kioka et al., 1999). To identify the H2 helix-binding site in vinculin, we first split vinculin into two parts, the N-terminal half containing a head region (residues 1–850) (head) and the C-terminal half containing a proline-rich linker and a tail region (836–1066) (linker-tail), and then performed pulldown assays using these constructs (Fig. 3A,B). GST–vinexin α H2 helix pulled down the head region (1–850) but not the linker-tail region (836–1066). Introducing the H2Q3 mutation into GST–vinexin α H2 helix specifically disrupted the co-precipitation with the head region (1–850). In contrast, GST–vinexin αΔN WT, which includes the SH3 domains, was co-precipitated with both constructs (Fig. 3B). These results indicate that the head region contains an H2 helix-binding site.

The vinculin head region is composed of four helix bundles, called D1 to D4 domains (Fig. 3A) (Bakolitsa et al., 2004). Thus, we next investigated which domain in the head region interacts with H2 helix. Both GST–vinexin αΔN WT and GST–vinexin α H2 helix were co-precipitated with the vinculin D1 domain (1–258) more efficiently than with the other domains (Fig. S1C). H2Q3 mutation of vinexin αΔN markedly decreased the co-precipitation with the D1 domain (1–258). Taken together, these findings show that vinexin α interacts with the vinculin D1 domain through the H2 helix, in addition to the interaction with the vinculin linker region through the SH3 domains.

The D1 domain consists of two helix bundles, the D1a and D1b subdomains, each of which contains four helices (Bakolitsa et al., 2004). Talin VBS3 and most other amphipathic helices that interact with and activate vinculin bind to the D1a subdomain (Izard et al., 2004). Replacement of the alanine at residue 50 in the D1a subdomain with an isoleucine residue (A50I) disturbs the interaction with talin VBS3 (Bakolitsa et al., 2004). To examine whether H2 helix interacts with the D1 domain in a manner similar to talin VBS3, we performed pulldown assays using a GFP–vinculin A50I mutant (Fig. 3C,D). The A50I mutation decreased the co-precipitation with talin VBS3, as expected. In contrast, the A50I mutation did not affect the co-precipitation with H2 helix. GST–vinexin α H2 helix having H2Q3 mutation did not co-precipitate with either wild-type vinculin or the A50I mutant. These results suggest that the H2-binding site of vinculin is different from the talin VBS3-binding site. To verify this observation, GFP–D1a (1–130) and GFP–D1b (127–252) helix bundles were constructed. However, GFP–D1a was co-precipitated with GST alone non-specifically, possibly due to improper folding of D1a (data not shown). Therefore, we focused on the GFP–D1 and GFP–D1b constructs. GST–talin VBS3 co-precipitated with GFP–D1 (1–258) but not GFP–D1b (127–252), while GST–vinexin α H2 helix, as well as GST–vinexin αΔN, co-precipitated with both GFP–D1 (1–258) and GFP–D1b (127–252) (Fig. 3E). These results suggest that the H2 helix of vinexin α interacts with the D1b subdomain, while talin VBS3 binds to the D1a subdomain.

The above observation that vinculin binds to H2 helix of vinexin α and talin VBS3 through different subdomains raises the possibility that vinexin α and talin simultaneously interact with vinculin. To investigate this possibility, the effect of the talin VBS on the interaction of H2 helix with vinculin was examined. Interestingly, the addition of a chemically synthesized talin VBS3 peptide increased the amount of full-length vinculin and the vinculin head region (1–850) co-precipitated with GST–vinexin α H2 helix in a dose-dependent manner (Fig. 4A). Triplicate experiments using various concentrations of GST–vinexin α H2 helix further confirmed the effect of talin VBS3 on the co-precipitation of vinculin with the vinexin α H2 helix (Fig. S1D,E). These results suggest that the interaction of talin with vinculin promotes the vinculin–vinexin α interaction.

We next examined whether vinexin α, talin and vinculin form a ternary complex. Vinexin α WT, its H2Q3 mutant or vinexin β was expressed in vinexin-knockout (KO) mouse embryonic fibroblasts (MEFs) together with GFP–vinculin. Although the expression level of vinexin α WT was similar to that of the vinexin α H2Q3 mutant, the expression level of vinexin β was much higher than that of vinexin α (Fig. 4B). Co-immunoprecipitation assays using anti-vinexin antibody showed that talin, as well as GFP–vinculin, co-precipitated with the vinexin α H2Q3 mutant significantly less than with vinexin α WT (Fig. 4C,E,F). Talin also co-precipitated significantly less with vinexin β. To determine whether vinexin α interacts with talin directly or indirectly through the interaction with vinculin, we performed co-immunoprecipitation assays using GFP- or GFP–vinculin-expressing vinculin-KO MEFs. Talin co-precipitated with vinexin α in GFP–vinculin-expressing cells but not in GFP-expressing cells (Fig. 4D). Taken together, these results suggest that vinexin α indirectly forms a complex with talin through interaction with vinculin, resulting in the formation of a ternary complex.

Vinexin α interacts with vinculin and induces a conformational change in vinculin to an open form in vitro (Yamashita et al., 2014). To determine whether the H2 helix is involved in the conformational change, we assessed the effect of H2 helix on vinculin conformation using a purified vinculin fluorescence resonance energy transfer (FRET) conformation probe that shows a high FRET ratio in a closed form compared with an open form (Chen et al., 2005). Consistent with a previous report (Yamashita et al., 2014), the addition of GST–vinexin αΔN WT decreased the FRET ratio in a concentration-dependent manner (Fig. 5A). The vinexin αΔN H2Q3 mutant did not decrease the FRET ratio, which is comparable to what was seen with vinexin β, which lacks the N-terminal region of vinexin α, including H2 helix (Fig. 5A). These results suggest that the H2 helix of vinexin α is necessary for inducing the conformational change in vinculin.

In the open form, vinculin has a higher affinity for F-actin than in the closed form, and thus, we also performed F-actin co-sedimentation assays to examine the effect of the H2 helix on vinculin affinity for F-actin. Consistent with the result from the in vitro FRET assay, GST–vinexin α WT increased the amount of vinculin that co-sedimented with F-actin in a concentration-dependent manner (Fig. 5B,C). The H2Q3 mutant, as well as vinexin β, co-sedimented significantly less vinculin (Fig. 5B,C). These results support the notion that H2 helix of vinexin α plays a critical role in inducing the conformational change in vinculin.

Next, we investigated whether the H2 helix is sufficient to induce the conformational change in vinculin. Vinculin FRET analysis showed that GST–vinexin α H2 helix did not change the FRET ratio even at high concentrations, while GST–talin VBS3 decreased the FRET ratio in a concentration-dependent manner (Fig. 5D). The F-actin co-sedimentation assay showed results consistent with this observation (Fig. 5E,F). GST–talin VBS3 increased the amount of vinculin that co-sedimented with F-actin in a concentration-dependent manner, whereas GST–vinexin α H2 helix did not affect the amount of co-sedimented vinculin (Fig. 5E,F). Taken together, these results indicate that H2 helix of vinexin α is necessary but not sufficient for induction of the conformational change in vinculin.

To examine the function of H2 helix in vivo, we first investigated the subcellular localization of vinexin proteins and their effect on vinculin recruitment to FAs on glass cover slips. The vinexin α H2Q3 mutant was predominantly localized to FAs, similar to vinexin α WT (Fig. 6A). Re-expression of vinexin α WT led to 1.8-fold and 1.4-fold increases in the intensity of GFP–vinculin at FAs and the number of GFP-vinculin-positive FAs, respectively, compared with mock cells under this experimental condition (Fig. 6B,C). Re-expression of vinexin β showed a moderate effect. In contrast, expression of the vinexin α H2Q3 mutant did not increase either GFP–vinculin intensity in FAs or the number of FAs, suggesting that the H2 helix promotes the recruitment of GFP–vinculin to FAs (Fig. 6B,C). The difference between vinexin α H2Q3 and vinexin β is possibly due to much higher expression of vinexin β than that of vinexin α WT or vinexin α H2Q3 (see Fig. 4B,C).

Next, to investigate the effect of the H2 helix on the ECM-stiffness dependent localization of vinculin to FAs, vinexin-KO MEFs re-expressing different forms of vinexins were seeded on soft (3.8 kPa) and rigid (25 kPa) polyacrylamide (PAA) gel substrates and then subjected to confocal microscopy (Fig. 7A,B; Fig. S2A). In vinexin α WT-re-expressing cells, both the intensity and the number of GFP–vinculin-positive FAs were increased on the rigid substrates compared with on the soft substrates under this experimental condition. Although mock and vinexin α H2Q3-re-expressing cells somewhat retained the ECM stiffness dependency, both the intensity and the number of GFP–vinculin-positive FAs on the rigid substrates were significantly decreased compared with vinexin α WT-re-expressing cells. Vinexin β-re-expressing cells did not show any ECM stiffness-dependency. Therefore, these results show that the H2 helix promotes accumulation of GFP–vinculin to FAs on rigid ECM.

A closed form of vinculin localized to FAs is readily washed out by treatment with CSK buffer containing Triton X-100, whereas an open form of vinculin is resistant to CSK treatment, mainly because of its tight interaction with F-actin (Omachi et al., 2017; Yamashita et al., 2014). Our previous study showed that vinexin α, but not vinexin β, increases the amount of CSK-resistant vinculin in cells on rigid ECM. To investigate whether H2 helix is involved in this increase, vinexin-KO MEFs re-expressing different forms of vinexins were seeded on PAA gel substrates with different levels of stiffness, and subjected to CSK treatment. As expected, in vinexin α WT-re-expressing cells, the GFP–vinculin intensity at FAs and the number of CSK-resistant vinculin-positive FAs were 8.4-fold and 4.7-fold higher, respectively, on the rigid substrates (25 kPa) compared with on the soft substrates (3.8 kPa) (Fig. 7C,D; Fig. S2B). By contrast, CSK-resistant vinculin-positive FAs in mock, vinexin α H2Q3 and vinexin β-re-expressing cells cultured on either the soft or rigid substrate were barely visible, and there was no significant difference between cells cultured on soft and rigid substrates (Fig. 7C,D; Fig. S2B). Combined with our previous report showing that rigid ECM promotes the interaction of vinexin α with vinculin compared to soft ECM, these results suggest that H2 helix of vinexin α is necessary for the ECM stiffness-dependent conformational change in vinculin in MEFs.

Nuclear localization of YAP/TAZ is promoted on rigid ECM in mesenchymal stem cells in a manner dependent on vinexin and vinculin expression (Kuroda et al., 2018, 2017). To explore the role of H2 helix in the nuclear localization of YAP/TAZ, vinexin-KO MEFs re-expressing vinexin α WT or H2Q3 cultured on soft or rigid substrates were immunostained with anti-YAP/TAZ antibody (Fig. 8A,B). On soft substrates, there was no significant difference in nuclear localization of YAP/TAZ between mock cells and vinexin-KO MEFs re-expressing vinexin α WT, H2Q3 mutant or vinexin β. In contrast, on rigid substrates, vinexin α WT-re-expressing cells showed more prominent nuclear localization of YAP/TAZ than mock, vinexin α H2Q3- or vinexin β-re-expressing cells. These results indicate that vinexin α H2 helix promotes YAP/TAZ nuclear localization on rigid substrates.

In this study, we identified a novel vinculin-binding site in vinexin α, the H2 helix, which forms an amphipathic helix structure. The H2 helix interacts with the vinculin D1b subdomain, which is included in the vinculin head region, and the disruption of the H2 helix results in impairment of the conformational change in vinculin and the nuclear localization of YAP/TAZ on rigid ECM. However, the H2 helix alone does not induce the conformational change in vinculin. Similarly, the SH3 domains of vinexin α are required but not sufficient to induce the conformational change in vinculin (Yamashita et al., 2014). Taken together, these observations suggest that, on rigid ECM, vinexin α interacts with vinculin through both the H2 helix and SH3 domains to induce the conformational change of vinculin and nuclear translocation of YAP/TAZ (Fig. 8C), and that the interaction regulates behaviors of vinculin that act as a sensor for ECM stiffness.

Amphipathic helices are known to associate with vinculin in two different ways, via a ‘helical bundle conversion’ and ‘a helix addition mechanism’. Amphipathic helices present in talin, α-actinin, or IpaA bind the vinculin D1a subdomain and induce dynamic remodeling of the structure of the vinculin D1 domain, leading to the conformational change in vinculin to an open form (Bois et al., 2005; Izard et al., 2004, 2006), a process called helical bundle conversion. A few amphipathic helices, including IpaA VBS2, are known to interact with the vinculin D1b subdomain (Nhieu and Izard, 2007). These interactions induce neither the dynamic remodeling of the structure of the vinculin D1 domain nor the conformational change in vinculin, a process called a helix addition mechanism. In this study, we showed that H2 helix of vinexin α interacts with the vinculin D1b subdomain and does not induce the conformational change in vinculin by itself. Interestingly, the H2 helix shows significant similarity to IpaA VBS2 (563–587) (Fig. S1F). Thus, it can be speculated that H2 helix interacts with vinculin through a helix addition mechanism. It will be of interest to examine this possibility in the future.

Because both the H2 helix and the SH3 domains of vinexin α are necessary for induction of the conformational change in vinculin, the simultaneous interaction of H2 helix and the SH3 domains with vinculin is required for induction of the conformational change. Owing to the flexibility of the proline-rich linker region to which the first and second SH3 domains of vinexin bind, the distance between the linker region and D1b subdomain is roughly estimated to be 50 Å, as determined from the crystal structures of vinculin (Bakolitsa et al., 2004; Borgon et al., 2004; Nhieu and Izard, 2007). The H2 helix and the first SH3 are 95 amino acids apart, most of which are predicted to form disordered or random coil regions. The maximum length of one amino acid in an extended peptide is 3.8 Å long (Erickson, 1994); thus 95 amino acids are enough length for connecting a 50 Å distance, suggesting that vinexin α could bind vinculin through the H2 helix and SH3 domains simultaneously. Previous studies have indicated that molecules that induce the conformational change in vinculin interact with a single domain of vinculin; talin, α-actinin, and α-catenin interact with the vinculin head region, and phosphatidylinositol 4,5-bisphosphate (PIP₂) interacts with the vinculin tail region (Bois et al., 2005; Izard et al., 2004; Johnson and Craig., 1995b; Peng et al., 2012). Thus, it is plausible that vinexin α induces a conformational change in vinculin through a novel mechanism, in which interactions at two different sites of vinculin are required.

Mechanical forces induce stretching of the talin rod, leading to exposure of cryptic VBSs and resulting in force-dependent recruitment of vinculin to FAs (Ciobanasu et al., 2014; del Rio et al., 2009; Hirata et al., 2014). This talin–vinculin interaction seems to play a crucial role in ECM stiffness sensing (Austen et al., 2015; Elosegui-Artola et al., 2016). Indeed, tension applied to talin is regulated by ECM stiffness (Austen et al., 2015; Kumar et al., 2016), especially in soft ranges (less than 4 kPa) in fibroblasts (Austen et al., 2015). The tensions detected in the Austen et al. study are enough for unfolding the most unstable talin rod domain (R3) and exposing VBSs (Yao et al., 2014). However, cell responses to ECM stiffness have a wide dynamic range, and traction force generation increases as the ECM becomes stiffer from 3 kPa to 52 kPa (Austen et al., 2015); thus, cells seem to have other systems for sensing ECM stiffness. In addition to the possibility that other talin rod domains or the number of talin molecules subjected to force contribute to sensing this range of ECM stiffness, our previous observation indicates that amount of open-form vinculin is increased depending on the interaction with vinexin α and the ECM stiffness in a stiffness range of 3.8–43 kPa (Yamashita et al., 2014), implicating that the vinculin–vinexin α machinery is also involved in sensing of ECM stiffness. However, the interplay between the talin–vinculin and vinculin–vinexin α mechanosensory systems has been never addressed. In the present study, we showed that talin VBS3 promotes the interaction between vinculin and H2 helix of vinexin α in a concentration-dependent manner, and talin and vinexin α can bind to vinculin simultaneously to form a ternary complex. Therefore, our observations suggest a collaboration between the talin–vinculin and vinculin–vinexin α machinery and a multi-tiered architecture of mechanosensory systems that allows cells to respond to a wide range of ECM stiffnesses.

Vinexin, c-Cbl associated protein (CAP) and Arg-binding protein2 (ArgBP2) constitute an adaptor protein family called the vinexin (SORBS) family and contain one sorbin homology (SoHo) domain and three SH3 domains in common (Kioka et al., 2002). Since our recent study indicated that CAP, as well as vinexin α, interacts with vinculin and induces the ECM stiffness-dependent conformational change in vinculin, it apparently seems that vinexin α and CAP are functionally redundant (Ichikawa et al., 2017). However, an alignment of amino acid sequences and secondary structure predictions implies that an amphipathic helix, such as H2 helix in vinexin α, is not conserved in the other vinexin family proteins (data not shown). Therefore, it is likely that vinexin α and CAP interact with vinculin through different mechanisms and show different responses to ECM stiffness. Indeed, vinexin α is involved in differentiation of mesenchymal stem cells into myoblasts on 15 kPa substrates, while CAP is involved in the differentiation into osteoblasts on 42 kPa substrates (Holle et al., 2016). Thus, a comparative study of the molecular mechanism of the vinexin α–vinculin and CAP–vinculin interactions is needed.

Collectively, we conclude that the H2 helix in the N-terminal region of vinexin α plays a crucial role in the ECM stiffness-dependent regulation of vinculin and cell behaviors. Given that the expression levels of vinexin in various tissues are correlated with those of vinculin (Kioka et al., 1999), regulation of vinculin function by vinexin α would be likely to play critical roles. Because it has been revealed that vinculin is involved in a variety of cell functions that are dependent on ECM stiffness, such as stem cell differentiation and tumor progression (Kuroda et al., 2017; Rubashkin et al., 2014), further studies concerning the role of vinexin in cancer progression and cell differentiation might provide novel insights into cancer therapy and tissue engineering.

The lentiviral transfer vector pCDH-EF1-IRES-Puro from System Biosciences (Mountain View, CA) was modified into pCDH-EF1-IRES-Blast as described previously (Ichikawa et al., 2017). The full-length vinexin α and β, as well as monomeric GFP-tagged full-length vinexin α, vinexin αΔN (157–733), and vinexin β were subcloned into pCDH-EF1-IRES-Blast. Monomeric GFP-tagged full-length chicken vinculin was inserted into pCDH-EF1-IRES-Puro. Vinculin head (1–850), linker-tail (836–1066), D1–D4 (1–836), D1 (1–258), D2 (253–487), D3 (491–717), D4 (718–836), D1a (1–130) and D1b (127–252) were generated by PCR and then subcloned into pCDH-EF1-IRES-Puro vector as a monomeric GFP-tagged form. The vinculin FRET conformation probe (Chen et al., 2005) was subcloned into pColdI vector (Takara Bio, Ohtsu, Japan). Vinexin αΔN (157–733), H2 (328–355) and talin1 VBS3 (1945–1970) were generated by PCR and then subcloned into pColdI-ΔHis-GST vector. Vinculin was also subcloned into pColdI-ΔHis-GST vector. Vinexin α H1Pa (M199P; F200P; I203P), H1Pb (L210P; L212P), H2Pa (I329P; L332P; L333P), H2Pb (L337P; L340P; S341P; L344P), H2Pc (L348P; I351P; L355P), H2Q1 (I329Q; L332Q), H2Q2 (L333Q; L337Q), H2Q3 (L340Q; L344Q) and H2Q4 (L348Q; I351Q; L355Q) were generated by site-directed mutagenesis using an In-Fusion® HD cloning kit (Clontech, Mountain View, CA).

Mouse monoclonal anti-vinculin (hVIN-1, cat. no. V9131, 1:20,000 dilution for immunoblotting, 1:400 dilution for immunofluorescence) and anti-talin (8d4, cat. no. T3287, 1:10,000 dilution for immunoblotting) antibodies were purchased from Sigma-Aldrich (St Louis, MO). Mouse monoclonal anti-GST antibody (5A7, cat. no. 013-21851, 1:5000 dilution for immunoblotting) was purchased from Wako (Osaka, Japan). Mouse monoclonal anti-GFP (B-2, cat. no. sc-9996, 1:2000 dilution for immunoblotting) and anti-YAP (63.7, cat. no. sc-101199, 1:100 dilution for immunofluorescence) antibodies were purchased from Santa Cruz Biotechnology (Dallas, TX). Rabbit polyclonal anti-vinexin antibody (1:10,000 dilution for immunoblotting, 1:400 dilution for immunofluorescence) was as described previously (Kioka et al., 1999). Goat anti-mouse IgG HRP conjugate (cat. no. 172-1011, 1:10,000 dilution for immunoblotting) and goat anti-rabbit IgG-HRP conjugate (cat. no. 172-1019, 1:10,000 dilution for immunoblotting) antibodies were from Bio-Rad (Hercules, CA). Alexa Fluor 546-conjugated goat anti-rabbit IgG antibody (1:1000 dilution for immunofluorescence) was from Thermo Fisher Scientific (Waltham, MA). Hoechst 33342 was from Invitrogen (Carlsbad, CA, 1:2000 dilution for immunofluorescence). Type I collagen was from Nitta Gelatin (Osaka, Japan). Puromycin and Blasticidin S were from Sigma-Aldrich and Kaken Pharmaceutical (Tokyo, Japan), respectively. Actin protein (>99% pure) purified from rabbit skeletal muscle was purchased from Cytoskeleton, Inc. (Denver, CO). Peptides of vinexin α H2 (329–355) and talin1 VBS3 (1945–1970) were synthesized and purified by Toray Research Center (Otsu, Japan). Full size blots for appropriate figures are presented in Fig. S3.

Spontaneously immortalized WT and vinexin-KO MEFs were as described previously (Kioka et al., 2010). Vinculin-KO MEFs were gifted from Dr Eileen Adamson (Burnham Institute, La Jolla, CA). Lentiviruses for protein expression were generated as described using lentiviral third-generation vectors [pCDH, pMD2.G (#12259), pRSV-Rev (#12253) and pMDLg/pRRE (#12251)] (Addgene, Cambridge, MA). The target cells were incubated with lentivirus, followed by the incubation with medium containing 1 µg/ml Puromycin or 3 µg/ml Blasticidin S for more than 7 days. All cells were grown in Dulbecco's modified Eagle's medium (DMEM; Nacalai tesque, Kyoto, Japan) supplemented with 10% fetal bovine serum (FBS; Gibco/Thermo Fisher Scientific) at 37°C in a humidified atmosphere containing 5% CO₂. PAA gel substrates were prepared as previously described (Yamashita et al., 2014), with slight modifications; PAA gels were coated with 100 µg/ml type I collagen. 2.2 and 3.8 kPa PAA gels were used as soft substrates and 25 and 43 kPa were used as rigid substrates.

Immunofluorescence of FA proteins was performed as previously described (Ichikawa et al., 2017; Yamashita et al., 2014), with slight modifications. Briefly, vinexin-KO cells re-expressing vinexin together with GFP–vinculin were cultured on 10 µg/ml type I collagen-coated No. 1 coverslips or 100 µg/ml type I collagen-coated PAA gels, and subjected to immunostaining. To visualize the total amount of vinculin or vinexin in cells on coverslips or PAA gels, cells were fixed with 2% (w/v) paraformaldehyde in PBS for 30 min at room temperature and then permeabilized with 0.2% (v/v) Triton X-100 in PBS for 5 min at room temperature. To visualize CSK-resistant GFP–vinculin in cells on PAA gels, cells were treated with CSK buffer (0.5% Triton X-100, 10 mM PIPES, pH 6.8, 50 mM NaCl, 3 mM MgCl₂, 300 mM sucrose) on ice for 2 min, followed by fixation with 4% (w/v) paraformaldehyde in PBS for 30 min at room temperature. For immunostaining, cells were incubated with 10% goat serum (Sigma-Aldrich) and 0.01% (v/v) Triton X-100 in PBS for 30 min at room temperature, followed by sequential incubation with primary and secondary antibodies diluted with PBS containing 10% goat serum and 0.01% (v/v) Triton X-100 for 1 h at room temperature or overnight at 4°C. Images were acquired using an LSM700 laser scanning confocal microscope with a Plan-Apochromat 40x/1.3 NA oil immersion objective lens (Carl Zeiss, Oberkochen, Germany). Image quantifications were performed using ImageJ software. Focal adhesions were classed as structures of 1–20 µm² in the ‘Analyze Particles’ command. Gray images were treated with ‘Enhance Contrast’ command and then inverted to increase their visibility. Immunostaining of YAP/TAZ was performed as described previously (Kuroda et al., 2018), with slight modifications. Briefly, vinexin-KO cells re-expressing vinexin together with GFP–vinculin were cultured on 100 µg/ml type I collagen-coated PAA gels, and subjected to immunostaining. The cells were fixed with 4% (w/v) paraformaldehyde in PBS for 15 min at room temperature and then permeabilized with 0.2% (v/v) Triton X-100 in PBS for 5 min. The cells were further incubated with 10% goat serum in PBS for 30 min, followed by sequential incubation with the anti-YAP antibody that crossreacts with TAZ (Kuroda et al., 2017) and second antibody diluted with PBS containing 10% goat serum for 1 h at room temperature or overnight at 4°C. Images were acquired using an FLUOVIEW FV1000 laser scanning confocal microscope (Olympus, Tokyo, Japan) equipped with an UPlan SApo 20×/0.75 NA dry objective lens (Olympus) with 3× zoom. To evaluate YAP/TAZ nuclear localization, the mean intensity of YAP/TAZ in the regions of the nucleus and cytosol (just outside the nucleus) with the same area was analyzed, followed by calculation of nucleus:cytosol intensity ratio.

Secondary structures of vinexin α H2 (329–355) and talin VBS3 (1945–1970) peptides were analyzed by circular dichroism spectrometry (J-805; JASCO, Tokyo, Japan) using a 0.1-mm quartz cell (121.027-QS, Q 10 mm; JASCO) as described previously (Sato et al., 2013; Yamashita et al., 2014). Talin VBS3 or vinexin α H2 peptides (80 µM) dissolved in PBS containing 8% ethanol were subjected to circular dichroism analysis. The obtained values were corrected by subtracting the spectrum of PBS containing 8% ethanol alone.

Purifications of GST-tagged proteins were performed as described previously (Takahashi et al., 2005; Yamashita et al., 2014). The His-tagged vinculin FRET conformation probe was expressed in E. coli Rosetta™ (DE3) strain (Novagen/Merck Millipore, Darmstadt, Germany), and then E. coli were lysed in PBS with a XL2020 Sonicator (MISONIX, Farmingdale, NY). The protein was then purified with Profinity™ IMAC Ni-Charged Resin (Bio-rad), followed by anion exchange chromatography with MonoQ 5/50 GL (GE Healthcare).

The in vitro FRET and actin co-sedimentation assays were performed as described previously (Yamashita et al., 2014) with slight modifications. Briefly, the purified vinculin FRET conformation probe (final 0.1 µM) and GST-tagged protein were incubated with pre-polymerized F-actin (final 4.6 µM) for 30 min at room temperature. The fluorescent intensity of CFP and FRET were acquired with a SpectraMax M2e (Molecular Devices, Inc., Sunnyvale, CA) for the experiment using GST-tagged vinexin, and with a Cytation 5™ imaging plate reader (BioTek Japan, Tokyo, Japan) for the experiment using GST-tagged amphipathic helices (H2 and talin VBS3). CFP emission was detected at 478 nm with excitation at 436 nm, and FRET emission was detected at 526 nm with excitation at 436 nm. Spectral crosstalk was corrected as described previously (Chen et al., 2005). In this study, the values R_Y and R_C were 0.048 and 0.402 in the measurement with SpectraMax M2e, and 0.059 and 0.494 with Cytation 5™ imaging plate reader. The corrected FRET ratio (cFRET:CFP) was obtained as the corrected FRET emission divided by the CFP emission. After the measurement of fluorescence, the samples were subjected to ultracentrifugation at 150,000 g for 1 h at 4°C for an actin co-sedimentation assay. The pellet and half of the supernatant were subjected to SDS-PAGE, followed by Coomassie Brilliant Blue (CBB) staining.

The immunoprecipitation assay was performed as described previously (Yamashita et al., 2014). Briefly, vinexin-KO cells re-expressing vinexin together with GFP–vinculin, or vinculin-KO MEFs expressing GFP–vinculin together with vinexin α were lysed with 1% (v/v) Triton X-100 in PBS. The cell lysates (300 µg) were incubated with 1.5 µg anti-vinexin antibody at 4°C for 1 h, followed by further incubation with Protein G–Sepharose 4B beads (Sigma-Aldrich). The beads were collected and washed with 0.5% (v/v) Triton X-100 in PBS three times. Co-precipitated proteins were subjected to SDS-PAGE, followed by immunoblotting with specific antibodies.

The pulldown and GFP depletion assays were performed as described previously (Ichikawa et al., 2017) with slight modification. 293T cells were transiently transfected with pCDH plasmids. At 24 h after transfection, the cells were lysed with 1% (v/v) Triton X-100 in PBS. Cell lysates were incubated with purified GST-tagged proteins at 4°C for 1 h followed by further incubation with glutathione–Sepharose 4B beads (GE Healthcare) for 1 h. The beads and the supernatant were separated by centrifugation (380 g for 2 min). For the GFP depletion assay, the GFP fluorescence of the supernatants was measured with a Cytation 5™ imaging plate reader. The fraction of bound GFP-tagged proteins were determined by calculating the reduction in GFP fluorescence of the supernatant incubated with GST–vinculin compared with that of the supernatant incubated with GST. For pulldown assay, the collected beads were washed with 0.5% (v/v) Triton X-100 in PBS three times. Co-precipitated proteins were subjected to SDS-PAGE, followed by immunoblotting with specific antibodies.

Data are presented as mean±s.e.m. Differences among compared groups were assessed by one-way ANOVA followed by multiple Tukey's post hoc tests. P<0.05 was considered statistically significant. Statistical analysis was performed using Origin 8.6J or GraphPad Prism 7 software.

We thank Dr K. Murakami and Dr K. Irie (Kyoto University) for advising us on CD spectroscopic analysis.