TRPV4 mediates the Ca²⁺ influx required for the interaction between flightless-1 and non-muscle myosin, and collagen remodeling

Fibroblasts synthesize and degrade extracellular matrix (ECM) proteins such as collagen in a tightly coordinated fashion to preserve connective tissue structure. Matrix homeostasis is maintained in part by collagen phagocytosis, a critical, fibroblast-mediated, multi-step degradative process (Everts et al., 1996) that requires the initial binding of collagen fibrils by the α2β1 integrin (Arora et al., 2005) and which is followed by fibril engulfment involving plasma membrane-derived cell extensions (Melcher and Chan, 1981). For efficient engulfment of collagen fibrils, actin filaments at sub-cortical sites actively modify to facilitate generation of plasma membrane extensions. Arising from their ability to sever actin filaments and promote actin assembly at free barbed ends, certain Ca²⁺-dependent actin-binding proteins are likely to be involved in the formation of cell extensions that is required for collagen phagocytosis. One of these proteins is flightless-1 (FliI), which interacts with non-muscle myosin IIA (NMMIIA; heavy chain encoded by MYH9) to generate contractile forces that enable reorganization and compaction of the collagen matrix (Arora et al., 2015). However, the signals that regulate the initiation of the FliI–NMMIIA interaction that is required for cell extension formation in spreading cells are unknown.

Ca²⁺ signaling controls a large array of cellular functions, which range from short-term responses (e.g. cell contraction and secretion) to long-term processes, such as the regulation of cell growth and proliferation (Berridge et al., 1998). The generation of receptor-induced cytosolic Ca²⁺ signals involves two closely related and interdependent elements: a fast and transient release of Ca²⁺ from the endoplasmic reticulum (ER), and a slower entry of extracellular Ca²⁺ through plasma membrane channels (Putney and McKay, 1999).

Stromal interaction molecule (STIM) proteins coordinate Ca²⁺ release and entry signals to maintain Ca²⁺ homeostasis (Roos et al., 2005). Small changes in Ca²⁺ concentration in the endoplasmic reticulum (ER) trigger STIM proteins to translocate and activate highly Ca²⁺-selective Orai channels localized at the plasma membrane (Vig et al., 2006), but there is currently only limited evidence to suggest that these channels are involved in coordinating matrix remodeling (Ross et al., 2017). In contrast, the presence of stretch-activated Ca²⁺-permeable channels in the plasma membrane has been demonstrated in migrating fibroblasts, and these respond to deformation with a sharp increase of cytoplasmic Ca²⁺ (Lee et al., 1999). Plasma membrane-associated Ca²⁺-permeable channels may therefore be well-positioned to participate in the signaling processes that are involved in matrix remodeling. In this context, certain transient receptor potential (TRP) family members mediate Ca²⁺ influx, which is important for fibroblast differentiation and wound healing (Davis et al., 2012). Although Ca²⁺-initiated signals can regulate actomyosin assembly and disassembly, the mechanisms that generate Ca²⁺ signals in response to matrix adhesion are not well defined.

TRP channels exhibit diverse activation mechanisms, which are critical for sensory physiology (Clapham, 2003), and these proteins are expressed and function in a variety of organisms, including fruit flies, worms, mice and humans (Montell, 2005). The TRP family of cation-permeable channels in mammals comprises 28 members that are divided among TRPA, TRPC, TRPM, TRPN, TRPP and TRPV subfamilies (Venkatachalam and Montell, 2007). Among these proteins, TRPM7 and TRPC6 are important for atrial and dermal fibroblast differentiation (Davis et al., 2012; Du et al., 2010). TRPV1 is a vanilloid receptor activated by capsaicin (Szallasi et al., 2007), while TRPV4 is expressed in neurons, endothelial cells and epithelial cells, and in tissues such as lung, heart and kidney (Hamanaka et al., 2007). TRPV4 is similar to TRPV1 in terms of its activation by warm temperatures, hypotonicity, mechanical stretch and phosphorylation (Nilius and Voets, 2004). Mechanical stretching of capillary endothelial cells through ECM adhesions stimulates Ca²⁺ influx through TRPV4 channels, which results in phosphoinositide 3-kinase (PI3K) activation to promote cell orientation (Thodeti et al., 2009).

FliI is an evolutionarily conserved member of the gelsolin family of actin-binding proteins, which have two functional domains: an N-terminal leucine-rich repeat (LRR) domain and a C-terminal domain that mediates discrete actin regulatory activities. FliI regulates actin assembly by capping filament ends, a process that ultimately promotes the formation of actin-rich cell extensions required for cell migration and matrix remodeling (Cowin et al., 2007; Kopecki and Cowin, 2008). FliI is enriched at cell–matrix adhesions (Mohammad et al., 2012) and colocalizes with molecules involved in regulating cytoskeletal reorganization including the GTPases Ras and Cdc42 (Davy et al., 2001), and with NMMIIA, which, upon activation, generates the force required for collagen remodeling through contraction (Arora et al., 2015).

Myosins comprise a superfamily of motor proteins that, together with actin filaments, make up the major contractile apparatus of cardiac, skeletal and smooth muscle (Clark et al., 2007). Similar to muscle myosins, NMMII molecules are comprised of three pairs of polypeptides: two heavy chains (HC), two regulatory light chains (RLC) and two essential light chains (ELC) that are bound to the myosin heavy chain, where they stabilize the heavy chain structure and regulate myosin structure and activity. Myosin light chains (MLC) are substrates for several kinases including the Ca²⁺-calmodulin-dependent MLC kinase (MLCK; also known as MYLK), which phosphorylates MLC (the MYL6B form) on Ser19, thereby increasing Mg²⁺-ATPase activity, which enhances myosin filament formation and contractile activity. Myosin contractility and actin cross-linking functions are essential for the determination of cell shape, migration and the maturation of cell adhesions (Choi et al., 2008; Scholey et al., 1980). At integrin adhesions, NMMII is activated by phosphorylation of the regulatory light chain, which leads to increased force generation and, through integrins, matrix remodeling (Beningo et al., 2001). NMMIIA activation is dependent on Ca²⁺, an important determinant of light chain phosphorylation, which in turn is the net result of the activities of MLCK and MLC phosphatase (Kamm and Stull, 1985).

We have previously shown that the actin-capping function of FliI contributes to the formation of actin-enriched cell extensions while, at collagen adhesion sites, FliI interacts with NMMIIA to generate contractile forces that enable compaction of the collagen matrix (Arora et al., 2015). Since Ca²⁺ regulates a large number of signaling pathways that orchestrate cytoskeletal structure and function, we examined how Ca²⁺ was involved in the association of FliI with NMMIIA at cell adhesions and how it contributes to the growth of cell extensions that are required for collagen compaction and remodeling (Friedl and Wolf, 2003). We also examined the role of TRPV4 channels in mediating the Ca²⁺ influx that is required for interaction of NMMIIA with FliI at cell adhesions. Our data show that Ca²⁺ fluxes through TRPV4 enable FliI to interact with NMMIIA filaments, thereby regulating formation of cell extensions and collagen remodeling.

Localized Ca²⁺ increases are detected at the growing filopodia of cells and in migrating cells that express stretch-activated Ca²⁺ channels (Gomez et al., 2001; Lee et al., 1999). Being aware of the regional sequestration of Ca²⁺ fluxes at localized sites, we selected a plasma membrane-targeted GCaMPG-caax probe, a fluorescent protein (GCaMPG) fused with a CAAX domain (Tsai et al., 2014) fused to GFP to measure sub-plasma membrane Ca²⁺ in cultured wild-type (WT) mouse fibroblasts (hereafter FliI WT cells). As anticipated, probe fluorescence was distributed throughout the cytoplasm but fluorescence was maximal at growing cell extension sites, where presumptive Ca²⁺-permeable channels are enriched (Wei et al., 2009) (Fig. 1A). Intracellular Ca²⁺ ([Ca²⁺]_i) in fura2-AM-loaded cells was analyzed in regions of interest (ROIs) at growing cell extensions in cells plated on collagen substrates, and compared with what was seen in ROIs selected in the centers of cell bodies (Fig. 1Bi). These data showed that [Ca²⁺]_i flux was higher near growing cell extensions (Fig. 1Bii,iii).

We examined expression levels of TRP channels in FliI WT cells by semi-quantitative PCR (primers described in Table S1). Of those TRP family channels examined, mRNA expression for TRPM7, TRPV1, TRPV2, TRPV4 and TRPP2 (also known as PKD2) was abundant in FliI WT cells (Fig. 1C). Western blot analysis did not show TRPM7 protein expression in lysates of FliI WT cells, although this protein was abundantly expressed in HeLa cells (Fig. 1Di,ii). Expression of TRPM7 mRNA in HeLa cells was confirmed by performing semi-quantitative PCR (Fig. 1Dii). Protein expression of TRPP2 was also undetectable (Fig. 1Dvi). TRPV1, TRPV2 and TRPV4 protein expression was detected in cell lysates in which equivalent amounts of cell proteins were examined (Fig. 1Diii–v). There was no difference in expression levels of TRPV in FliI WT and or in mouse fibroblasts where FliI was knocked down through use of shRNA (hereafter FliI KND cells) (Fig. 1E). We also found that knockdown of TRPV4 expression by means of two different siRNAs, and observed (70% and 65% knockdown) attenuated Ca²⁺ flux at growing cell extensions (Fig. 1Fi,ii). Knockdown of TRPV4 reduced collagen compaction by >1.5- and 2-fold for the two TRPV4 siRNAs (Fig. 1Gi,ii). In contrast, siRNA-mediated knockdown of TRPV2 (by 80%) and of TRPV1 (by 75%) did not affect Ca²⁺ fluxes, indicating that these channels are not involved in Ca²⁺ permeation (Fig. 1H,J). In assessing the role of these channels in matrix remodeling, we found no difference in collagen compaction between cells with TRPV2 or TRPV1 knockdown compared with that seen in cells transfected with a control siRNA (Fig. 1I,K).

We determined whether TRPV4 channels are functionally active in FliI WT cells loaded with fura2-AM. In cells treated with the specific TRPV4 agonist GSK1016790A, there were substantial concentration-dependent increases in [Ca²⁺]_i (10–500 nM), indicating that TRPV4 channels were functionally active (Fig. 2A). In cells treated with the TRPV4 antagonist RN1734 (AB1), there was a minimal increase in [Ca²⁺]_i during spreading (Fig. 2B). We examined the relative specificity of the TRPV4 antagonist and whether it affected other types of Ca²⁺-permeable channels. In these experiments, cells that were treated or not with TRPV4 channel antagonist were stimulated with FPL 64176 (a L-type channel agonist), BAY K8644 (a voltage-gated channel agonist), GV-58 (a N-type channel agonist) or MSP-3 (a TRPV1 Ca²⁺ channel agonist). These experiments showed minimal reduction of Ca²⁺ influx after treatment with the TRPV4 antagonist (Fig. S1). We also stimulated cells with the Ca²⁺ ionophore ionomycin in the presence and absence of the TRPV4 inhibitor but found no change of Ca²⁺ flux (Fig. S1).

We determined whether mobilization of internal Ca²⁺ was required for TRPV4-dependent Ca²⁺ influx. Blocking intracellular Ca²⁺ release through inositol 1,4,5-triphosphate receptors by treatment with Xestospongin C or by inhibition of the ER Ca²⁺ transport ATPase (by thapsigargin) had no effect on Ca²⁺ influx in spreading cells at the cell extensions in selected ROIs (Fig. 2Ci,ii). In another experiment, cells plated in low Ca²⁺ medium ([Ca²⁺]_e=0.7) and treated with thapsigargin showed rapid release of Ca²⁺ from the ER and a return to basal [Ca²⁺]_i levels, which was followed by minimal [Ca²⁺]_i changes over 10 min. Replacement of medium with normal Ca²⁺ ([Ca²⁺]_e=1.8 mM) and then treatment with the TRPV4 agonist GSK1016790A, stimulated Ca²⁺ influx (Fig. 2D), indicating that mobilization of intracellular Ca²⁺ stores was not necessary.

In addition to the role of ECM proteins in stimulating Ca²⁺ in cells spreading on matrix-coated substrates (Chan and Odde, 2008), clustering of integrins is thought to be an important component of the mechanosensing machinery that mediates local Ca²⁺ fluxes (Gomez et al., 2001). Immunostaining showed 60±8% and 40±6% (mean±s.e.m.) colocalization of TRPV4 with β1 integrin and paxillin respectively (Pearson correlation co-efficient; Fig. 2E). Consistent with these data, a co-immunoprecipitation assay showed that TRPV4 channels associate with the β1 integrin (Fig. 2F). Analysis of proteins isolated from collagen-coated magnetite beads also showed an association of TRPV4 with β1 integrin in adhesion complexes (Fig. 2G). The data showing TRPV4 enrichment in collagen adhesions and co-immunoprecipitation with the β1 integrin indicates that TRPV4 is likely located in the vicinity of collagen receptors.

We examined whether Ca²⁺ flux through TRPV4 channels required activation of the α2β1 integrin (Thodeti et al., 2009). Cells that were incubated with an α2β-blocking antibody did not spread and did not develop cell extensions. Cells loaded with fura2-AM and incubated with α2β1-blocking antibody showed reduced Ca²⁺ influx compared with that seen in cells treated with control antibody (Fig. 2H). In a separate experiment of similar design, cells treated with the TRPV4 agonist GSK1016790A exhibited normal amplitude Ca²⁺ fluxes in the presence of the α2β1 blocking antibody (Fig. 2I). In another experiment using β1 integrin-deficient cells (GD25 cells), the TRPV4 agonist GSK 1016790A induced substantial Ca²⁺ fluxes (Fig. 2J). Collectively these experiments suggest that in spreading cells that are attaching to matrix and forming cell extensions, collagen receptors are involved in TRPV4 channel activation while the β1 integrin is not required for chemical agonist-induced channel activation.

We determined whether Ca²⁺ fluxes mediated by TRPV4 channels contribute to the regulation of cell shape and the formation of cell extensions. We used FliI WT cells, GD25 cells (which lack the β1-integrin; Fassler and Meyer, 1995) and GD25-β1 cells (GD25 cells transfected with a β1 integrin expression plasmid). GD25 cells plated on collagen appeared circular with minimal cell extensions. The TRPV4 antagonist RN1734 (AB1) did not affect the shape, number or length of extensions in GD25 cells, which do not express the β1 integrin. Untreated GD25 cells had a circularity index that was 4-fold higher (P<0.01) than FliI WT or GD25-β1 cells showing that they are more rounded. The number and length of extensions were reduced by the TRPV4 antagonist (3.5-fold reduction, P<0.01, in WT cells; 2-fold reduction, P<0.01, in GD25-β1 rescued cells). Untreated FliI WT cells and GD25-β1 cells showed 5-fold (P<0.01) and 4-fold (P<0.01) more cell extensions than did GD25 cells. The length of extensions in untreated GD25 cells was 3.8-fold (P<0.01) and 2.6-fold (P<0.02) less than in FliI WT and GD25-β1 integrin rescued cells (Fig. 2Ki–iii). These data indicate that Ca²⁺ fluxes mediated by TRPV4 channels affect cell shape and cell extension formation (Fig. 2Ki–iii). In β1-integrin-deficient GD25 cells treated with the TRPV4 agonist, there was no effect on the number and length of cell extensions (Fig. S1B), reflecting the apparent requirement for β1 integrins in mediating this process.

Ca²⁺ influx through TRPV4 channels is thought to be involved in focal adhesion assembly, cell orientation and migration (Matthews et al., 2010). We determined whether Ca²⁺ influx through TRPV4 channels also plays a role in collagen compaction. FliI WT cells incubated with the TRPV4 antagonist RN1734 (AB1) showed a 1.5-fold reduction of collagen compaction (Fig. 2Li,ii).

We determined whether Ca²⁺ influx through TRPV4 channels was required for interaction of FliI with NMMIIA. Inactivation of TRPV4 channels with the specific antagonist RN1734 (AB1) completely inhibited the association of FliI with NMMIIA in spreading cells. Binding to collagen was evidently an important determinant since in cells plated on poly-lysine there was no interaction of FliI with NMMIIA (Fig. 3A). Similarly, knockdown of TRPV4 expression with siRNA inhibited the association of FliI with NMMIIA (Fig. 3Bi). In contrast, knockdown of TRPV1 or TRPV2 channels (with specific siRNAs) showed no effect on FliI–NMMIIA interactions (Fig. 3Bii,iii). These experiments show that inhibition of TRPV4 function and expression reduces the association of FliI with NMMIIA. Furthermore, in cells treated with the TRPV4 antagonist RN1734 (AB1) and plated on collagen or poly-lysine, immunoprecipitation experiments showed that collagen receptor ligation and TRPV4 channel activation were required for NMMIIA and TRPV4 association (Fig. 3C). In a separate experiment, we confirmed by immunoprecipitation that FliI associated with phosphorylated MLC in response to collagen ligation (Fig. S1C).

Cell lysates from FliI WT type fibroblasts that had been loaded with BAPTA-AM (4 μM; an intracellular Ca²⁺ chelator) and plated on collagen in low Ca²⁺ medium ([Ca²⁺]_e=0.7 mM) were immunoprecipitated with antibody against NMMIIA and immunoblots were probed with FliI antibody. There was a 1.5-fold reduction (P<0.05) in the association of FliI with NMMIIA in cells loaded with BAPTA. This association required cell attachment to collagen, as cells in suspension showed 3-fold reduced association of FliI with NMMIIA (P<0.05). Cells plated on poly-lysine showed a 3-fold reduction in the association of FliI with NMMIIA (P<0.05; Fig. 4A). Consistent with these data, BAPTA-AM treatment prevented colocalization of GFP–FliI and NMMIIA in cells plated on collagen-coated bead substrate (a substrate chosen to facilitate colocalization studies) (Fig. S1A). We found that this association was specific for cell attachment to collagen because GFP–FliI and NMMIIA did not colocalize in cells plated on bovine serum albumin (BSA)-coated beads (Fig. S1A).

FliI WT cells plated on collagen showed abundant cell extensions. Incubation with BAPTA/AM caused 2.5-fold reduction in the number of cell extensions (P<0.05), a 1.4-fold reduction in the length of cell extensions (P<0.05) and an absence of well-formed actin stress fibers (Fig. 4Bi,ii). Cells plated in low Ca²⁺ medium ([Ca²⁺]e=0.7 mM) showed 6-fold (P<0.01) reduction in the number of cell extensions and a 7.5-fold (P<0.01) reduction in the length of cell extensions. FliI KND cells did not form cell extensions (Fig. 4Bi,ii).

In view of the requirement for Ca²⁺ in the growth of cell extensions, we examined the requirement of Ca²⁺ for collagen remodeling. After loading with BAPTA/AM, FliI WT or knockdown cells showed reduced collagen compaction around the cell periphery (Fig. 4Ci). Quantification of compacted collagen fibrils (Arora et al., 2015) showed a 2-fold reduction in WT cells loaded with BAPTA-AM in regions of interest delineated by concentric circles radiating from the cell centroid. FliI KND cells treated with BAPTA showed a 4.5-fold reduction of collagen compaction compared with FliI WT cells (Fig. 4Cii).

Ca²⁺ is important for phosphorylation of the MLC, which in turn promotes myosin filament formation (Vicente-Manzanares et al., 2007). There was a substantial increase of MLC phosphorylation in cells plated on collagen-coated surfaces compared with that seen in cells in suspension and in cells treated with the MLCK inhibitor ML-7. After treatment with the Ca²⁺ ionophore ionomycin (1 μM), cells exhibited substantial MLC phosphorylation (Fig. 5A). ML-7-treated (25 μM) cells or cells maintained in suspension showed loss of the FliI–NMMIIA association whereas treatment with ionomycin enhanced the association of these proteins compared with that seen in untreated control cells plated on collagen (Fig. 5B).

ML-7 affects myosin assembly and myosin activity through its inhibition of MLCK activity (Kimura et al., 1996; Watanabe et al., 2007) while the ROCK inhibitor Y-27632 affects myosin function by inhibiting MLC phosphatase activity (Riento and Ridley, 2003; Totsukawa et al., 2000). Blebbistatin inhibits NMMIIA motor activity. Accordingly, we assessed whether inhibition of myosin activity (but without affecting myosin assembly) would impact the association of FliI with NMMIIA. In immunoprecipitation experiments, treatment with Y-27632 or blebbistatin did not affect the association between NMMIIA and FliI (Fig. 5Ci,ii). Because FliI can bind to other Ca²⁺-binding proteins such as CaM kinase II, we treated cells with a specific inhibitor of CaM kinase (KN-93) but did not observe inhibition in the association of FliI with NMMIIA (Fig. 5Ciii).

We examined the role of Ca²⁺ in the interaction of FliI with NMMII. Pulldown experiments were performed to determine the binding of purified NMMIIA rods (amino acids 1339–1960) to glutathione S-transferase (GST)-tagged fusion protein comprising all of its six gelsolin-like domains (GLDs) (FliI-GLD1–6) in the presence or absence of EGTA (1.1 mM). Pulldown assays with Sepharose–FliI-GLD1–6 beads and NMMIIA rods provided evidence for a specific interaction (K_d=0.199±0.093 μM), consistent with previous data (Arora et al., 2015). Depletion of Ca²⁺ inhibited the interaction by 7-fold (K_d=1.37±0.406 μM; Fig. 5Di,Ei,ii). FliI-GLD1–5 also showed strong binding to NMMIIA (K_d=0.243±0.090 μM), which was reduced by EGTA (K_d=1.45±0.957 μM; Fig. 5Diii,Eiii,iv). There was also evidence of weaker binding of FliI-GLD1–3 with NMMIIA rods (K_d=0.600±0.049 μM), which was only slightly reduced by EGTA (K_d=0.727±0.029 μM; Fig. 5Dii,Ev,vi). The K_d of the interaction between FliI-GLD4–6 and NMMIIA was similar to that for GLD1–3 (0.622±0.082 μM) but was reduced nearly 3-fold in the presence of EGTA (K_d=1.57±0.89 μM). Collectively, these data suggest that the Ca²⁺-sensitive interaction of NMMIIA with FliI may involve the FliI-GLD4–6 (in the C-terminus).

The role of Ca²⁺ in FliI–NMMIIA interactions was studied in intact cells with cerulean-tagged NMMIIA (donor) and Venus-tagged FliI (acceptor) probes. Photobleaching was reduced by using a low power setting of the exciting laser, which was well below the intensity required for image acquisition. De-quenched signals from the donor were seen after photo-bleaching the acceptor, demonstrating fluorescence resonance energy transfer (FRET) and an interaction between FliI and NMMIIA (Fig. 5F). The specificity of the acceptor photobleach (AP)-FRET assay was established by using cerulean-labeled NMMIIB (also known as MYH10) as a negative control, as NMMIIB does not interact with FliI (Arora et al., 2015). Pre-bleach and post-bleach images were used to calculate mean the percentage efficiency (E, ±s.d.), which was 27±9.1% for the FliI–NMMIIA interaction and 5.1±2.5% for the FliI–NMMIIB interaction. The donor-alone interaction measured in cells transfected with an empty vector construct with FliI was reduced (E=9.7±3.1%; P<0.05). FliI–NMMIIA interactions were also reduced in cells plated on a poly-L-lysine substrate that prevents activation of α2β1 integrins (E=15.3±4.2%). Treatment with the Ca²⁺ ionophore ionomycin (1 μM) increased the tightness of the interactions of FliI and NMMIIA (E=27.2±8.2%). In samples loaded with BAPTA-AM in intact cells, the percentage efficiency of the FliI–NMMIIA was E=14.1±6.5%, indicating that Ca²⁺ was an important determinant of FliI and NMMIIA interactions in intact cells (Fig. 5F).

The signaling systems that regulate the formation of cell extensions required for collagen remodeling by the phagocytic pathway are not defined. Here, we identified TRPV4 as a Ca²⁺-permeable channel in the plasma membrane that regulates FliI–NMMIIA interactions and cell extension formation to control of collagen remodeling. Consequently, TRPV4 may be a critical regulator of matrix remodeling, which highlights the central role of this membrane protein in cell signaling and matrix biology (Fig. 6).

Previous data have shown that FliI localizes to cell adhesions and contributes to appropriate control of wound healing and cell migration (Cowin et al., 2007; Kopecki and Cowin, 2008; Mohammad et al., 2012), but the mechanisms by which FliI mediates ECM remodeling and cell function were elusive. In this context, the actin-capping function of FliI contributes to the formation of actin-enriched cell extensions while, at collagen adhesion sites, FliI interacts with NMMIIA to generate contractile forces that enable compaction of the collagen matrix (Arora et al., 2015). However, the signals that regulate the initiation of the FliI–NMMIIA interaction, which is required for cell extension formation in spreading cells, are unknown. As Ca²⁺ plays a central role in controlling actomyosin contractile forces during cell motility (Wei et al., 2009), we used FliI WT and KND cells to examine the involvement of Ca²⁺ in FliI–NMMIIA interactions. We found that in intact cells, localized Ca²⁺ fluxes were concentrated at growing cell extensions, which were enriched with TRPV4 channels (Adapala et al., 2013; Scheraga et al., 2016; Thodeti et al., 2009).

We found, by performing ratio imaging, that Ca²⁺ fluxes were concentrated at the leading edge of extensions in cells spreading on collagen. The presence of stretch-activated Ca²⁺-permeable channels has been demonstrated in migrating fibroblasts, which respond to stretching forces with an increase in the cytoplasmic Ca²⁺ concentration (Lee et al., 1999). Recent studies show that TRPM7 and TRPC6-mediated Ca²⁺ influxes play an essential role in atrial and dermal fibroblast differentiation (Davis et al., 2012; Du et al., 2010). We examined the expression of several TRP family channel members by performing qPCR and immunoblotting. We found that TRPV4, TRPV2 and TRPV1 were expressed but that only TRPV4 was required for collagen compaction, as assessed through siRNA knockdown experiments or treatment with the TRPV4-specific antagonist RN1734 (Vincent et al., 2009). Furthermore, use of RN1734 showed that TRPV4-mediated Ca²⁺ fluxes were restricted to growing cell extensions. In this context, localized transient elevations of [Ca²⁺]_i initiated at the filopodia of neural growth cones generate signals that are transmitted back to the growth cone to induce cell motility (Gomez et al., 2001). Notably, sequestration of intracellular Ca²⁺ with BAPTA strongly inhibited the formation of cell extensions, but exerted a much smaller effect on the length of cell extensions, possibly indicating that those mechanisms that drive cell extension formation once they are formed, may not involve Ca²⁺-dependent processes.

We found that TRPV4 localized predominantly at paxillin and β1 integrin-immunostained focal adhesions proximal to cell extensions. Our immunoprecipitation and immunostaining experiments showed an association of TRPV4 channels with FliI and NMMIIA, suggesting that these molecules are spatially coordinated to regulate cell extension formation and collagen compaction. Conceivably, FliI and NMMIIA provide the continuity between integrin-mediated focal adhesions and matrix-gated ion channels to enable local signal transduction at cell extensions. Notably, TRPV4 activation requires application of mechanical strain at focal adhesions, which results in Ca²⁺ influx (Matthews et al., 2010). In professional phagocytic cells such as macrophages, lipopolysaccharide-stimulated phagocytosis is also regulated by TRPV4 (Scheraga et al., 2016), indicating a broader role for this channel protein in determining the interactions of cells with external particles and matrix molecules.

The structure and function of the actin cytoskeleton are very sensitive to localized changes of Ca²⁺ concentration (Tsai et al., 2015), which affects actin severing, capping and cross-linking. These processes in turn influence cell shape, endocytosis, morphogenesis, locomotion, growth and division (Furukawa et al., 2003). For example, in Dictyostelium, Ca²⁺ fluxes derived from extracellular Ca²⁺ or intracellular Ca²⁺ stores are required for cell spreading in locomotion and chemotaxis (Furukawa et al., 2003; Unterweger and Schlatterer, 1995). Consistent with these data and the notion that FliI is a Ca²⁺-dependent actin-capping protein, we found that when FliI WT cells were loaded with BAPTA-AM to clamp [Ca²⁺]_i, the association of FliI with NMMIIA, the formation of cell extensions and collagen compaction were all inhibited. Ca²⁺ is also required for phosphorylation of MLC, which is important for the unfolding of myosin into extended monomers that can assemble into filaments (Stull et al., 1993; Watanabe et al., 2007). Notably, immunoprecipitation, in vitro protein interaction and FRET experiments showed that the FliI–NMMIIA interaction was regulated by Ca²⁺, which in turn was required for the activation of MLC and the promotion of NMMIIA filament formation and binding to FliI.

Cells bind, sequester and extrude Ca²⁺ across membrane-enclosed compartments to effect changes of [Ca²⁺] over a wide range of Ca²⁺ concentrations (100 nM to 1 mM). These signals control a broad range of cellular processes (Berridge et al., 2003; Clapham, 2007). For example, gelsolin, an actin-binding protein that regulates actin assembly by severing and capping actin filaments, is activated by Ca²⁺, which triggers conformational re-arrangements to promote the binding of gelsolin to actin (Wang et al., 2009). Like gelsolin, FliI is an actin-capping protein that has six GLDs (Liu and Yin, 1998). While GLD 1 of FliI contributes to Ca²⁺-independent binding of FliI to NMMIIA filaments (Arora et al., 2015), our new data indicate that there are also important, Ca²⁺-sensitive interactions of FliI and NMMIIA, which may involve FliI-GLD4–6. In the context of ECM remodeling, these Ca²⁺-dependent interactions of FliI with NMMIIA may be central determinants of cell extension formation, which in turn affect collagen remodeling by compaction and phagocytosis (Everts et al., 1996).

We conclude that Ca²⁺ influx through TRPV4 channels regulates the interaction of FliI with NMMIIA, which in turn enables generation of cell extensions that are essential for collagen remodeling.

BAPTA-AM, fura 2-AM and Rhodamine–phalloidin were purchased from Molecular Probes. Blebbistatin (20 µM), Y-27632 (10 µM), RN-1734 (AB1) (100 µM) and Xestospongin (10 µM) were from Calbiochem (Cedarlane, Burlington, ON). Acrylamide, bis-acrylamide, ammonium persulfate (APS), GSK 1016790A (a TRPV4 agonist) (10 µM), thapsigargin (10 µM), EGTA, antibodies to paxillin, nebulin and α2β1 integrin, ML-7 (20 µM), ionomycin (1 µM), tetramethylethylenediamine (TEMED), CaM kinase inhibitor (KN-93) (10 µM), FPL 64176 (a L-type channel agonist) (300 nM), BAY K8644 (a voltage-gated channel agonist) (10 nM), GV-58 (a N-type channel agonist) (10 µM) and MSP-3 (a TRPV1 Ca²⁺ channel agonist) (1 µM) were purchased from Sigma-Aldrich (Oakville, ON). siRNAs for TRPV4, TRPV2 and TRPV1 and irrelevant control siRNAs were purchased from GE Dharmacon (Mississauga, ON). Type I bovine collagen (5.8 mg/ml solution) was purchased from Advanced BioMatrix (Carlsbad, CA). Purified rat antibodies to mouse β1 integrin (clone 9EG7) were purchased from BD Pharmingen (Mississauga, ON). Antibodies to TRPV4, TRPV2, TRPV1, TRPP2 and TRPM7 were from Alomone Labs (Jerusalem, Israel). The anti-FliI antibody was from Abcam (Cambridge, MA), and the antibody to NMMIIA was from Biomedical Technologies (Stoughton, MA). The plasmid, pGP-CMV-GCaMP6s-CAAX, was purchased from Addgene (Cambridge, MA). Live-cell imaging medium was purchased from Molecular Probes, Life Technologies. Details of the antibodies, including the dilution, can be found in Table S2.

WT and KND FliI mouse fibroblasts were prepared as described previously (Arora et al., 2015) and were maintained in DMEM (10% fetal bovine serum and 10% antibiotics). GD25 cells, originally produced by Reinhard Fassler, were obtained from the late Wolfgang Vogel (University of Toronto, Canada). Immortalized HeLa cells originally derived from cancerous cervical tumours were obtained from Andrew Wilde (University of Toronto, Mars Centre, Canada).

To construct stable FliI stable KND fibroblast cells, oligonucleotides (top strand, 5′-gatccGAAGATACACACTATGTTATTCAAGAGATAACATAGTGTGTATCTTCTTTTTTACGCGTg-3′ and bottom strand, 5′-aattcACGCGTAAAAAAGAAGATACACACTATGTTATCTCTTGAATAACATAGTGTGTATCTTCg-3′, corresponding to the sense 5′-GAAGATACACACTATGTTA-3′ and antisense 5′-TAACATAGTGTGTATCTTC-3′; lowercase letters represent restriction sites for BamH1 and EcoR1) for mouse FliI were annealed and inserted into an RNAi-Ready pSIREN-RetroQ-DsRed-Express vector (Clontech) at BamHI/EcoRI sites. Insert sequences were confirmed by sequencing. The plasmid was co-transfected with pVSV-G into GP-293 cells for retrovirus production. NIH-3T3 cells were infected with the virus; 2 weeks later, the transfected cells were sorted in PBS with 0.5% fetal bovine serum (FBS) in a Beckman-Coulter Altra flow cytometer/sorter. Cells with strong red fluorescence were cloned by limiting dilution. There was 90% knockdown of the FliI expression (Arora et al., 2015).

Cells were loaded with the Ca²⁺ indicator, fura-2-AM (3 μM) and plated on collagen substrates coated on to glass bottom Microwell dishes (MatTek corporation) and incubated at 37°C for 30 min. The medium was replaced with live cell imaging solution, a sterile solution that consisted of 140 mM NaCl, 2.5 mM KCl, 1.8 mM CaCl₂, 1.0 mM MgCl₂, 5 mM D-glucose and 20 mM HEPES (pH 7.4) with osmolality of 300 mOsm. For experiments in which low Ca²⁺ was required, EGTA was added to reduce the Ca²⁺ concentration (0.7 mM). [Ca²⁺]_i in single, attached cells was measured by selecting regions of interest at the periphery of the growing cell extensions for 15 min periods using a microscope-based ratio fluorimeter (Nikon Eclipse TE2000-U with 40× oil immersion lens). Intracellular Ca²⁺ was recorded by line scan in xt mode and data were acquired as the ratio of emission when excited at 340 nm to that when excited at 380 nm. To determine whether STIM1 or Orai1 were involved in the Ca²⁺ influx mechanism, before addition of thapsigargin, the cell culture medium was replaced with low Ca²⁺ medium. After addition of thapsigargin, [Ca²⁺]_i levels were allowed to return to basal levels and the medium was replaced with normal Ca²⁺ medium. [Ca²⁺]_i levels were monitored for 10 min to determine the involvement of STIM1 and Orai1. Cells were stimulated with TRPV4 agonist in the normal Ca²⁺ medium.

To examine the specificity of the TRPV4 channel antagonist, the antagonist was pre-incubated with cells that were then stimulated with an L-type agonist (FPL 64176) or an N-type channel agonist (GV-58), a voltage-gated channel agonist (BAY K8644) and a TRPV1 channel agonist (MSP-3) in normal Ca²⁺ medium. Ionomycin was used as a positive control.

Cellular RNA was purified by using the RNeasy Mini Kit (Qiagen, Mississauga, ON) according to the manufacturer's protocol. The concentration and integrity of the extracted RNA was determined with a Nanodrop 1000 (ThermoScientific) machine. PCR analysis was performed by using the iScript™ cDNA synthesis kit (Bio-Rad, Hercules, CA). According to the manufacturer's instructions, total RNA (1 μg) was reverse-transcribed and qRT-PCR was performed with a Bio-Rad CFX96 Real Time qPCR system using SsoFast™ Eva-Green^® Supermix (BioRad) with validated mouse primers for GADPH and specific TRP channels (Table S1). The TRPM7 primers used for HeLa cells were as follows: forward, 5′-TGCAGCAGAGCCCGATATTAT-3′ and reverse, 5′-CTCTATCCCATGCCAATGTAAGG-3′; β-actin primers were as follows: forward, 5′-CATGTACGTTGCTATCCAGGC-3′ and reverse, 5′-CTCCTTAATGTCACGCACGAT-3′.

Cells plated on collagen-coated coverslips, fixed with 4% paraformaldehyde for 10 min and permeabilized with 0.3% Triton X-100 for 10 min, were blocked with BSA (1%) in PBS for 25 min and immunostained for TRPV4, FliI, paxillin and activated β1 integrin using antibodies diluted in 0.1% BSA in PBS and applied for 1 h at 37°C. Secondary antibodies diluted in 0.1% BSAin PBS were incubated for 1 h at room temperature. For imaging, coverslips were inverted on to glass bottom dishes with mounting medium. After confocal imaging of immunostained proteins, the Pearson co-efficient of double-stained samples was computed (using Image J) to quantify the degree of colocalization (Bolte and Cordelières, 2006).

Coverslips (25 mm; VWR) were sterilized with UV light (10 min), submerged in 2% 3-aminopropyltrimethoxysilane (Sigma-Aldrich) for 15 min, washed three times with distilled water, immersed in 0.1% glutaraldehyde (Caledon Laboratories, Canada) for 15 min, washed three times with distilled water, air-dried for 10 min and coated with collagen, which was left to polymerize for 1 h at 37°C. Cells were plated on collagen-coated substrates for 5 h before fixing with 4% paraformaldehyde for 10 min and being permeabilized with 0.3% Triton X-100 for 10 min. Coverslips were inverted on to glass bottom dishes with mounting medium. Confocal reflectance microscopy was used to visualize and quantify collagen fiber orientation as described previously (Mohammadi et al., 2015). Cell morphometry was performed with Image J software (version 1.44). The number of cell extensions per cell, the mean and sum of the length of cell extensions, and circularity (a dimensionless shape index) were measured. The length of cell extensions was measured from the cell centroid to the tip of each cell extension. The sum of the lengths of cell extensions was computed by adding the length of all cell extensions for each cell. Computation of the circularity index was performed using Image J. A perfect circle has a circularity of 1 while values of increasingly non-circular shapes (i.e. cells with long extensions) approach 0.

Protein concentrations of FliI bound to GST–Sepharose beads were determined by analysis of standards separated on SDS polyacrylamide gels. Various concentrations of assembled NMMIIA rods (0.012–5.0 μM) were incubated for 30 min at room temperature in reaction buffer containing 20 mM Tris-HCl pH 7.5, 20 mM NaCl, 2 mM MgCl₂, 1 mM DTT and 0.3 mM CaCl₂ or 1.1 mM EGTA. Samples were centrifuged at 10,000 g for 5 min. Pellets consisting of Sepharose beads were washed 3× with PBS. Supernatants and pellets were analyzed on SDS-PAGE gels, which were stained with Coomassie Blue for observation.

AP-FRET was performed on a Leica SP8 confocal microscope with the Leica AP-FRET wizard. Images were acquired with a 40× NA 1.3 oil objective at 512×512 pixel resolution. Photo-bleaching was minimized during acquisition by using low laser power (10% for 448 nm and 552 nm) with 448 nm, 514 nm and 552 nm beam splitters. A region of interest (ROI; 2×2 μm) was chosen in selected areas and 100% laser power was used with eight iterations for photo-bleaching the acceptor Venus–FliI. Pre-bleach and post-bleach images were acquired using the same settings. The efficiency of energy transfer (E%) was calculated using the equation: E%=(Donor Post-bleach−Donor Pre-bleach)×100/(Donor Post-bleach).

For continuous variable data, means and standard deviations were computed. When appropriate, comparisons between two samples were made by Student's t-test with statistical significance set at P<0.05. For multiple comparisons, ANOVA was used followed by Tukey's test for assessment of individual differences. All experiments were performed at least three times in triplicate.

We thank Dhaarmini Rajshankar for help with semi-quantitative PCR, Yong Wang for designing the primers for TRPs and Nuno Coelho for help with preparation of collagen substrates. GD25 cells were created by R. Fassler, were obtained from the late W. Vogel (University of Toronto).