Supervillin slows cell spreading by facilitating myosin II activation at the cell periphery

Cell migration is a multi-step process involving surface protrusion, substrate attachment at integrin- and actomyosin-associated focal adhesions, myosin II-based retraction of the cell body, and substrate detachment (Bretscher, 1996; Lauffenburger and Horwitz, 1996; Lin and Forscher, 1995). Cell spreading, the initial step in this process, can be modeled by plating cells onto fibronectin-coated coverslips (Betapudi et al., 2006; Cai et al., 2006; Duxbury et al., 2004; Wakatsuki et al., 2003) and can be approximated as the sum of two forces (Wylie and Chantler, 2003). The outward force, driven by actin polymerization, is opposed by a myosin II-driven inward contractile force associated with F-actin bundling and remodeling in a transitional zone behind the extending cell surface (Brown and Bridgman, 2003; Cai et al., 2006; Gupton and Waterman-Storer, 2006; Medeiros et al., 2006; Wakatsuki et al., 2003).

Recent work with cells deficient in individual myosin II isoforms implicates myosin IIB as being primarily responsible for focal adhesion stabilization during rapid cell surface extension, whereas myosin IIA is implicated in the contractile forces that reduce the rate of cell spreading (Betapudi et al., 2006; Bridgman et al., 2001; Cai et al., 2006; Lo et al., 2004; Vincente-Manzanares, 2007; Wylie and Chantler, 2003). This generalization is tempered by cell type-specific differences both in myosin II isoforms and in myosin II-binding regulatory proteins (Bao et al., 2005; Betapudi et al., 2006; Cai et al., 2006; Simerly et al., 1998; Togo and Steinhardt, 2004). The fact that some cells contain only one nonmuscle myosin isoform [e.g. COS7 cells contain only myosin IIB (Betapudi et al., 2006)] indicates a large degree of functional redundancy among myosin II isoforms.

Myosin II activation is regulated by monophosphorylation of its regulatory 20 kDa light chain (myosin light chain, MLC) on serine-19 and diphosphorylation at a second residue, threonine-18 (Ikebe, 1989). Diphosphorylation also differentially stabilizes the extended, filamentous conformation of myosin (Ikebe et al., 1988; Matsumura, 2005; Watanabe et al., 2007). Many kinases can phosphorylate and activate MLC, with Rho kinase and MLC kinase (MLCK) among the chief regulators (Matsumura, 2005; Murthy, 2006). These enzymes mediate both mono- and di-phosphorylation of MLC (Choi et al., 1994; Ikebe, 1989; Itoh et al., 1992; Ueda et al., 2002). Rho kinase regulates MLC phosphorylation in the cell interior (Totsukawa et al., 2004) and controls the development of traction forces in adhesive fibroblasts (Beningo et al., 2006), perhaps via selective modulation of myosin IIB (Sandquist et al., 2006), but is not a major contributor during cell spreading (Betapudi et al., 2006; Totsukawa et al., 2004). MLCK promotes MLC phosphorylation in the cell periphery and along stress fibers (Chew et al., 2002; Totsukawa et al., 2004). MLCK also preferentially co-localizes with myosin IIA (Blue et al., 2002) and regulates integrin-dependent cell survival (Connell and Helfman, 2006; Fazal et al., 2005), perhaps via a signaling cascade downstream of Ras and MEK (Nguyen et al., 1999). This model is consistent with the requirement for the kinase activities of MLCK and extracellular signal-regulated kinase (ERK, also called MAPK), both of which are downstream of MEK during focal adhesion disassembly (Webb et al., 2004).

Negative regulation of MLC phosphorylation in nonmuscle cells is mostly mediated by a tripartite myosin phosphatase comprised of a ∼130 kDa targeting subunit called MYPT1, a ∼38 kDa catalytic subunit and a ∼20 kDa subunit (reviewed in Matsumura, 2005; Murthy, 2006). MYPT1 binds to the myosin S2 subfragment, bringing the catalytic phosphatase into proximity with MLC (Johnson et al., 1997). Several of the kinases that promote MLC activation simultaneously phosphorylate and downregulate myosin phosphatase activity (Ito et al., 2004; Matsumura, 2005).

We have identified a plasma membrane-associated cytoskeletal fraction that contains myosin II, F-actin and signaling proteins in association with a tightly bound peripheral membrane protein called supervillin (SV, p205) (Nebl et al., 2002; Pestonjamasp et al., 1997). Over 95% of SV in differentiated cells localizes and co-fractionates with plasma membranes (Nebl et al., 2002; Oh et al., 2003; Pestonjamasp et al., 1997; Wulfkuhle et al., 1999). SV also co-isolates as part of a cytoskeletal complex with β2-integrin, receptor protein tyrosine phosphatase β and the P2X₇ ATP-gated ion channel, which regulates cell growth, apoptosis and membrane blebbing (Kim et al., 2001). SV negatively regulates cell-substrate adhesion through interactions with the LIM domains of thyroid receptor-interacting protein 6 (TRIP6) and lipoma-preferred partner (LPP) (Takizawa et al., 2006). Alternatively spliced forms of SV found in striated and smooth muscle are involved in the organization of muscle costameres, stimulus-mediated contractility of smooth muscle and myogenic differentiation (Gangopadhyay et al., 2004; Oh et al., 2003). Thus, SV apparently helps transmit signals across the plasma membrane to the actin- and myosin II-associated cortex.

A role for SV in regulating membrane-associated cytoskeletal assembly is supported by direct binding to both myosin II and F-actin (Chen et al., 2003). The first 171 amino acids in SV (SV1-171) bind to the S2 regulatory domains of nonmuscle and smooth muscle myosin II heavy chains. SV sequences containing combinations of two of the three F-actin-binding sites within the SV N-terminus (SV amino acids ∼280-342, ∼344-422 and ∼700-830) also bundle or crosslink actin filaments in vitro (Chen et al., 2003). The TRIP6- and LPP-binding site lies within SV amino acids 342-571 (Takizawa et al., 2006), and the SV C-terminus (SV1542-1792) binds to calponin (Gangopadhyay et al., 2004), which also binds both actin and myosin II and regulates their interaction (El-Mezgueldi and Marston, 1996; Lin et al., 1993). Thus, direct or indirect sites of interaction with actin and myosin II are distributed throughout the SV sequence.

We therefore hypothesize that SV is part of a signaling scaffold that organizes and regulates actin and myosin II filaments at the membrane. In addition to mediating focal adhesion turnover via interactions involving TRIP6 in the center of the cell (Takizawa et al., 2006), we hypothesize that SV modulation of myosin II activity regulates the rate of cell spreading. We show here that SV does modulate cell spreading, apparently by promoting or stabilizing contractile assemblages of actin and myosin II with the long form of MLCK (L-MLCK).

SV negatively regulates cell spreading on fibronectin (Fig. 1, supplementary material Fig. S1). In populations of COS7 cells spreading onto fibronectin, we noticed that cells expressing EGFP-SV were less well spread than nontransfected cells at 30-45 minutes after plating; no such difference was apparent for cells expressing EGFP only. Initial spreading was essentially complete by 60 minutes for all cells. Defining a `spread' cell as one exhibiting a radius greater than twice that exhibited by the average initially rounded cell (Fig. 1A, panel a), we quantified the number of spread cells as a function of time after plating (Fig. 1B). COS7 cells expressing an average ∼5-fold excess of EGFP-SV (Takizawa et al., 2006) were significantly less likely to be spread at 30 minutes, as compared with control cells or cells expressing EGFP alone (Fig. 1C). This result was reminiscent of our previous observations of SV-mediated inhibition of cell-substrate adhesion (Takizawa et al., 2006).

During early cell spreading, EGFP-SV and myosin heavy chain (MHC) IIB prominently co-localized in bundles at the cell periphery, in addition to their abundant signals in the cell interior (Fig. 2A, panels a-c). EGFP by itself did not colocalize with myosin bundles after 30 minutes of spreading (Fig. 2A, panels d-f). Indeed, the circumferential myosin bundles were far less prominent in untransfected cells (not shown) or cells expressing only EGFP at this time than in cells with EGFP-SV (Fig. 2A, panel e versus panel b). By 60 minutes of spreading, peripheral and interior bundles of myosin were observed in all three types of cells (Fig. 2B and not shown).

Peripheral bundles of F-actin, activated myosin II and endogenous SV also formed during spreading of A549 human lung carcinoma cells (Fig. 3). These cells spread onto fibronectin with a time course similar to that of COS7 cells (see below). Consistent with previous reports (Cai et al., 2006; LoRusso et al., 1997; Sumoza-Toledo et al., 2006), we found that both MHC IIA and MHC IIB isoforms localize with F-actin in peripheral bundles during cell spreading, although MHC IIA staining was more prominent in A549 cells (Fig. 3a-f). These bundles were associated with vinculin-stained focal adhesions (Fig. 3f) and with singly and doubly phosphorylated MLC (Fig. 3g,h,j), indicating local myosin activation and the potential for contractility. Immunostained endogenous SV at the cell periphery overlapped with activated myosin II, although fainter overlaps of SV with F-actin in myosin-depleted lamellipodia also were seen (Fig. 3j-r). Thus, although it contains at least three binding sites for F-actin and only one for MHC II (Chen et al., 2003), SV might preferentially colocalize with myosin II during cell spreading.

To determine whether SV binds preferentially to nonmuscle MHC IIA versus MHC IIB, we analyzed the binding of 6×His-tagged recombinant S2 domains from these myosins to columns containing GST-tagged SV amino acids 1-174 (GST-SV1-174, Fig. 4). These residues contain the binding site of SV for the smooth muscle myosin S2 domain (Chen et al., 2003). The S2 domains from both nonmuscle myosin isoforms bind to GST-SV1-174 (Fig. 4, left column), but not to GST lacking SV sequences (Fig. 4, right column). Although much less MHC IIB than MHC IIA S2 domain is eluted by physiological levels of salt (Fig. 4, left, lanes 2-4), significant amounts of both myosin isoforms resist extraction with even 0.3 M NaCl, eluting only when GST-SV1-174 is removed from the column (Fig. 4, left, lane 7). As reported previously (Chen et al., 2003), no detectable binding of SV to skeletal muscle myosin II could be observed. Thus, SV can interact with both nonmuscle myosin IIA and nonmuscle myosin IIB isoforms, although associations with myosin IIB might be somewhat stronger.

As expected if endogenous SV plays a role during cell spreading, we found that reduction of SV levels in A549 cells using two different dsRNAs increased the rate of spreading (Fig. 5). Targeting of human SV coding nucleotides 1680-1704 or 2026-2050 with two rounds of dsRNA transfection over 4 days reduced the levels of SV to ⩽10% of normal amounts without affecting the levels of other proteins analyzed in this study (Fig. 5A). In both live-cell imaging (Fig. 5B; supplementary material Movies 1, 2) and counts of spread cells (Fig. 5C), A549 cells with reduced amounts of SV spread faster than either mock-transfected cells or cells treated with a control dsRNA. As with COS7 cells, the largest difference in cell spreading was observed at 30 minutes, when cells treated with either dsRNA against SV were significantly more likely to be spread than cells with normal levels of endogenous protein (Fig. 5D).

During spreading, reduction of endogenous SV also reduces the extent of peripheral bundles (Fig. 6), defined as arrays of concentrated actin and myosin II at the cell edge or at the base of a lamellipodium (Small et al., 1999). After 30 minutes of spreading onto fibronectin, only ∼18% (n=200) of A549 cells treated with a dsRNA against SV contained well-defined peripheral bundles of myosin and actin, whereas 43% (n=200) of cells treated with a control or no dsRNA contained these structures (Fig. 6A). Activated myosin II filaments were still detectable with antibodies against mono- or doubly-phosphorylated MLC, but these structures were more widely distributed throughout the cell cortex (Fig. 6B). Thus, reduced numbers of peripheral bundles of myosin II and F-actin accompany SV-mediated reductions in the rate of cell spreading.

To explore the mechanism of SV action during cell spreading, we first treated A549 cells, which contain endogenous SV, with inhibitors of myosin II function (Fig. 7). We found that the rate of cell spreading was increased by blebbistatin, which directly inhibits myosin II ATPase activity (Kovacs et al., 2004; Limouze et al., 2004; Straight et al., 2003); by ML-7 at a concentration (1 μM) that specifically inhibits MLCK (Bain et al., 2003); and by the drug U0126, an indirect inhibitor of MLCK via the upstream kinase MEK (Fincham et al., 2000; Nguyen et al., 1999). By contrast, Y-27632, which inhibits direct and indirect activation of myosin by Rho kinase (Amano et al., 1996; Kimura et al., 1996; Uehata et al., 1997), showed no significant effect (Fig. 7).

The formation of circumferential bundles of myosin II and F-actin during spreading of either A549 cells (Fig. 8) or COS7 cells (supplementary material Fig. S2) was inhibited by blebbistatin (20 μM, not shown) or by low concentrations (1 μM) of ML-7. As with SV knockdowns, these drugs reduced the percentage of A549 cells with peripheral bundles after 30 minutes of spreading, from ∼40% to ∼10% (n=200). The effect appeared to be selective for peripheral bundles because cytoplasmic myosin II punctae and filaments remained in each cell line treated with 1 μM ML-7. Reduced numbers of peripheral bundles of myosin II and F-actin were also observed after indirect inhibition of myosin II activity with U0126 (2.5 μM); inhibition of Rho kinase with Y-27632 was without effect (supplementary material Fig. S2). ML-7 and U0126 also virtually eliminated the peripheral localization of phosphorylated MLC (i.e. activated myosin II) (supplementary material Fig. S3). Taken together, these results suggest that MLCK might mediate the effects of SV on the rate of cell spreading and on the formation of peripheral myosin and actin bundles.

The effects of the myosin II inhibitors on SV-mediated changes in cell spreading varied with concentration of blebbistatin and ML-7 (Fig. 9, supplementary material Fig. S4). Low amounts of blebbistatin (20 μM) and ML-7 (1 μM) abolished SV inhibition of COS7 cell spreading, whereas higher drug concentrations could themselves slow the rate of spreading (Fig. 9A,B), as described previously (Betapudi et al., 2006; Duxbury et al., 2004; Jurado et al., 2005). Slowed spreading at 10 μM ML-7 was associated with the disruption of myosin II filaments, of F-actin bundles and of focal adhesions throughout the cell (supplementary material Fig. S4), consistent with the inhibition of all cytoplasmic MLCK and/or inhibition of another myosin II kinase at higher ML-7 concentrations (Komatsu and Ikebe, 2004). By contrast, the inhibitory effects of the MEK inhibitor U0126 (Fig. 9C) and the absence of an effect of the Rho kinase inhibitor Y-27632 (Fig. 9D) were observed at all concentrations tested. These results suggest that SV might be slowing the rate of cell spreading by facilitating localized myosin II activation by MLCK.

To test the possibility of crosstalk between SV and MLCK, we co-transfected COS7 cells with plasmids encoding EGFP-SV and Flag-tagged long and short isoforms of MLCK (L-MLCK, S-MLCK; Fig. 10) (Blue et al., 2002; Herring et al., 2000). Murine L-MLCK contains the entire S-MLCK coding sequence. In addition, L-MLCK contains an extra N-terminal sequence of 954 amino acids with multiple actin-binding sites within its immunoglobulin (Ig) domains and two DXR repeats (repeats of a motif made up of the sequence DXRXXL) (Kudryashov et al., 2004; Poperechnaya et al., 2000; Smith et al., 2002; Stull et al., 1998; Yang et al., 2006). The shared S-MLCK C-terminus contains another three DXR repeats, the kinase domain and a myosin II-binding site (Fig. 11A). A lysine-to-alanine mutation in the kinase domain generates kinase-dead (KD) MLCK mutants that inhibit the catalytic activity of MLCK (Wadgaonkar et al., 2003).

KD mutants of L-MLCK and S-MLCK and the wild-type forms all co-localized with EGFP-SV in peripheral myosin II and F-actin bundles during cell spreading (Fig. 10A). Colocalization was more pronounced for the L-MLCK proteins (Fig. 10A, panels a-h) than for the S-MLCK proteins (Fig. 10A, panels i-p) because of the high proportion of cytoplasmic S-MLCK (Blue et al., 2002; Poperechnaya et al., 2000).

To determine whether the SV-mediated effects on cell spreading require the unique L-MLCK N-terminus, we determined the rates of spreading for COS7 cells co-transfected with each MLCK construct and either EGFP-SV or EGFP (Fig. 10B). Either wild-type L-MLCK or S-MLCK, when coexpressed with EGFP (Fig. 10B, columns 3, 7), mimicked EGFP-SV (Fig. 10B, column 2) in reducing the initial rates of cell spreading (column 1). Co-expression of either wild-type MLCK isoform with EGFP-SV accentuated SV-mediated reductions in the rate of cell spreading (Fig. 10B, columns 4 and 8 versus column 2). As shown in Fig. 10B, KD L-MLCK (columns 5, 6) was less effective than wild-type L-MLCK (columns 3, 4) at reducing spreading rates for EGFP-expressing cells (column 1) and did not accentuate the effects of EGFP-SV (column 2). By contrast, KD S-MLCK (columns 9, 10) exhibited an effect similar to that of wild-type S-MLCK (columns 7, 8). Neither effect was significantly altered by the presence of EGFP-SV (columns 7 versus 8; 9 versus 10), implying a kinase- and SV-independent mechanism. The differences between the effects of KD L-MLCK and KD S-MLCK (columns 9 versus 5; 10 versus 6) suggest an additional kinase-independent effect associated with the unique L-MLCK N-terminus. Thus, these results are consistent with involvement of SV in an L-MLCK-mediated kinase-dependent mechanism and with one or more kinase-independent effects in the rate of cell spreading induced by S-MLCK (the L-MLCK C-terminus) and by L-MLCK N-terminal sequences.

Multiple mechanisms are supported by domain-mapping studies of L-MLCK and EGFP-tagged SV sequences (Figs 11, 12). Consistent with a kinase-independent mechanism, L-MLCK-specific N-terminal sequences (Fig. 11A), which lack the myosin-binding and kinase domains (Stull et al., 1998) but bundle actin filaments (Smith et al., 2002; Yang et al., 2006), slowed the rate of cell spreading by 30-53% (Fig. 11B). A ∼26% reduction was also observed with the SV fragments that can bundle F-actin (Chen et al., 2003), SV343-830 and SV171-1792 (Fig. 12A).

However, F-actin bundling cannot be the sole mechanism by which SV reduces the rate of cell spreading. The N-terminal 171 residues in SV (SV1-171), which do not bind F-actin but do bind to myosin II (Chen et al., 2003) and to the L-MLCK N-terminal Ig domains (Fig. 11C), also slowed cell spreading (Fig. 12A). In addition, EGFP-tagged SV1-171 induces the formation of myosin punctae (Chen et al., 2003) that co-localize with mono- and di-phosphorylated MLC (Fig. 12B) and with L-MLCK (Fig. 12C, panels a-c). S-MLCK was far less efficiently associated (Fig. 12C, panels g-i), and these structures appeared to actually exclude the myosin phosphatase subunit MYPT1 (Fig. 12C, panels j-l). Time-lapse imaging of EGFP–SV1-171 showed that the elongated punctae arise from apparently hypercontractile segments of stress fibers (supplementary material Movie 3). Taken together, these results suggest that F-actin bundling is responsible for at least some of the kinase-independent reductions in cell spreading and implicate SV1-171 in a kinase-dependent mechanism involving myosin II activation by L-MLCK.

We demonstrate here that SV decreases the rate of cell spreading via interactions with myosin II, F-actin and L-MLCK. SV co-localized with F-actin and myosin II filament bundles at the plasma membrane, and SV levels correlated with the extent of bundle formation during initial cell spreading. SV binds directly to myosin IIA and myosin IIB and to the N-terminus of L-MLCK, as well as to actin filaments (Chen et al., 2003). SV-associated filament bundles are enriched in L-MLCK and, to a lesser extent, in S-MLCK. Pharmacological inhibition of MLCK or myosin motor activity reverses the SV-mediated decreases in the rate of cell spreading, and MLCK overexpression phenocopies the SV effect. Cell spreading is slowed by L-MLCK and SV sequences that bundle F-actin. Cell spreading is also slowed by SV1-171, which binds myosin II and L-MLCK in vitro and co-localizes with aggregates of L-MLCK and activated myosin II in vivo. Thus, SV apparently modulates cell spreading by promoting or stabilizing actomyosin bundles and/or upregulating myosin II activation at the plasma membrane.

The major new finding reported here is that SV, a membrane protein, enhances L-MLCK-mediated myosin II function during cell spreading. Others have used myosin II inhibitors, including blebbistatin, to show that myosin II plays an active role in promoting cell spreading (Betapudi et al., 2006; Cramer and Mitchison, 1995; Duxbury et al., 2004; Sumoza-Toledo et al., 2006; Wakatsuki et al., 2003). Inhibition of MLCK with high concentrations (⩾5 μM) of ML-7 slows cell spreading (Betapudi et al., 2006; Giannone et al., 2004). We show here that SV introduces a biphasic nature to the inhibitor dose-response curves (Figs 7, 9) consistent with SV-mediated enhancement of myosin II-mediated contractility that antagonizes cell spreading. Low concentrations of either blebbistatin (20 μM) or ML-7 (1 μM) alleviate the SV-mediated effect, returning the rate of cell spreading to that observed with lower SV levels. These results suggest that subpopulations of myosin II and MLCK, distinguished by myosin isoform and/or associated proteins, might be differentially sensitive to low concentrations of these drugs and/or to regulation by SV.

Our results also suggest that SV is part of an interconnected membrane-associated scaffold that helps position myosin II and F-actin with respect to MLCK. Myosin II and actin filaments have long been known to assemble at the membrane surface (DeBiasio et al., 1988; McKenna et al., 1989; Rhee et al., 1994) and to interact with both L-MLCK and S-MLCK (Blue et al., 2002; Dulyaninova et al., 2004; Poperechnaya et al., 2000; Stull et al., 1998). The additional F-actin-binding and -bundling sequences in the unique L-MLCK N-terminus apparently increase the affinity of L-MLCK, relative to S-MLCK, for actin filaments (Smith et al., 2002; Yang et al., 2006). SV also binds to both myosin and actin filaments, with the physical separation of the F-actin- and myosin II-binding sites within SV consistent with its suggested potential for regulated recruitment of actin and myosin filaments during myofibril and stress fiber assembly at the membrane (Oh et al., 2003).

An interaction of SV with the N-terminus of L-MLCK is suggested by direct binding in vitro and by the preferential colocalization of L-MLCK with SV1-171 in vivo. Simultaneous binding of SV1-171 to myosin and to L-MLCK could provide the basis for SV-mediated activation of myosin II by helping to retain or position L-MLCK on the myosin II heavy chain.

An SV-associated scaffold could mediate local MLCK activation via phosphorylation by ERK during cell spreading (Huang et al., 2004; Klemke et al., 1997). This possibility is consistent with the ability of the MEK inhibitor U0126 to reverse SV-mediated effects on cell spreading (Fig. 9) and with the co-localization of SV, myosin II and ERK in smooth muscle cells (Gangopadhyay et al., 2004). The suggestion that ERK signaling might be regulated by SV is also consistent with the observation that ERK activation during agonist-mediated contraction is inhibited by knockdown of the smooth muscle isoform of SV in ferret aorta (Gangopadhyay et al., 2004). The interaction between SV and calponin (Gangopadhyay et al., 2004) – a protein that binds directly to F-actin, myosin II and ERK (Leinweber et al., 1999; Morgan and Gangopadhyay, 2001) – suggests a further level of interconnection.

A second possible mechanism is that the binding of SV1-171 to subfragment 2 of myosin regulates its conformation and thus increases the affinity of L-MLCK for the myosin head-rod junction (Masato et al., 1997). For example, SV1-171 binding to myosin might promote myosin monomer unfolding and thus facilitate myosin activation by MLC phosphorylation and subsequent filament assembly (Kolega, 1998; Kolega and Taylor, 1993). Conversely, SV might stabilize the myosin filament or block myosin filament disassembly. For instance, by binding to subfragment 2 of myosin, SV might block binding of myosin phosphatase to this site (Johnson et al., 1997; Mitsui et al., 1992), shifting the dynamic equilibrium of dephosphorylated versus phosphorylated MLC and leading to the observed exclusion of MYPTI and the accumulation of mono- and di-phosphorylated MLC with SV and SV1-171 (Fig. 12).

SV might also stabilize contractile assemblies by bundling and thus stabilizing actin filaments against disassembly. This hypothesis is consistent with the observations that the actin-bundling fragments of SV and L-MLCK decrease the rate of cell spreading. This possibility is also consistent with recent work showing that myosin II slows cell protrusion by accelerating actin filament fragmentation in a contractile zone away from the leading edge (Medeiros et al., 2006), a process that is antagonized by actin filament stabilization, such as by tropomyosin (Huckaba et al., 2006). Given the sequence similarity between the SV C-terminus and the actin filament-severing proteins villin and gelsolin (Pestonjamasp et al., 1997), SV-mediated filament stabilization might seem unlikely. However, sequences required for F-actin severing in gelsolin and villin (Matsudaira et al., 1985) are absent from SV (Pestonjamasp et al., 1997; Wulfkuhle et al., 1999).

Another possibility is that the combined F-actin and myosin II binding capabilities of SV regulate the actomyosin cross-bridge cycle in the way proposed for calponin and caldesmon (Szymanski, 2004). Smooth muscle isoforms of these proteins are thought to act as `latch bridges' that inhibit myosin cross-bridge cycling while binding to both F-actin and activated myosin II, stabilizing the contractile state. In this model, SV1-171 competes for binding on myosin subfragment 2 with endogenous full-length SV and/or nonmuscle caldesmon, releasing the latch state and inducing the localized contractions that apparently destabilize stress fiber integrity (supplementary material Movie 3).

These possible modes of action are not mutually exclusive because SV regulation of actin-myosin interactions at the membrane might differ with intracellular location. For instance, SV might promote myosin activation in the posited signaling scaffold with L-MLCK at the cell periphery (Medeiros et al., 2006) and inhibit myosin II function elsewhere. SV1-171 and SV343-571 both reduce the number of large focal adhesions in the center of the cell (Takizawa et al., 2006). The effect of SV1-171 is consistent with a role for SV in potentiating L-MLCK-induced focal adhesion disassembly at various intracellular sites (Webb et al., 2004), but the lack of effect of SV343-571 on cell spreading (Fig. 12) argues strongly for a second regulatory mechanism. In fact, SV343-571 decreases cell-substrate adhesion via an interaction with the focal adhesion protein TRIP6, which leads to loss of stress fibers and of mature focal adhesions in the cell interior (Takizawa et al., 2006). Thus, SV apparently participates in at least two regulated interactions involving myosin II, F-actin and focal adhesion proteins at different intracellular locations during different stages of cell migration.

Glutathione-Sepharose was purchased from Amersham Biosciences (Piscataway, NJ). ML-7 was from Calbiochem-Novabiochem (La Jolla, CA). Blebbistatin with a ± stereochemical ratio of 1:1 was from Tocris (Bristol, UK). U0126 was from Sigma (St Louis, MO). Y-27632 was from Yoshitomi Pharmaceutical Industries (Osaka, Japan). Other chemicals were from Sigma, Calbiochem-Novabiochem, Fisher (Pittsburgh, PA) or VWR International (Bristol, CT).

Rabbit polyclonal anti-nonmuscle MHC IIA, rabbit polyclonal anti-nonmuscle MHC IIB, sheep polyclonal anti-MLCK and rabbit polyclonal anti-MYPT1 were from BAbco-Covance (Richmond, CA). Rabbit polyclonal anti-ERK1/2 was from Upstate (Lake Placid, NY). Mouse monoclonal anti-β-tubulin (TUB2.1), mouse monoclonal vinculin (hVIN1) and rabbit polyclonal anti-Flag were from Sigma. Mouse monoclonal anti-phospho-Ser19 MLC, rabbit polyclonal anti-phospho-Thr18/Ser19 MLC and mouse monoclonal anti-6×His (27E8) were from Cell Signaling Technology (Beverly, MA). Goat polyclonal anti-GST antibodies were from Amersham. Horseradish peroxidase (HRP)-conjugated secondary antibodies against rabbit, mouse and goat/sheep IgGs were from Jackson ImmunoResearch (West Grove, PA). Rabbit polyclonal anti-SV (H340) has been described (Nebl et al., 2002; Oh et al., 2003).

EGFP-SV and EGFP-SV fragment vectors were described previously (Chen et al., 2003; Wulfkuhle et al., 1999). EGFP-SV171-1792 was made by ligating the fragment released from EGFP-SV171-571 with AgeI into AgeI-digested EGFP-SV (Chen et al., 2003). Plasmids encoding Flag-tagged wild-type and KD murine 220 kDa L-MLCK and Flag-tagged rabbit 130 kDa S-MLCK were gifts from Patricia Gallagher (Indiana University School of Medicine, IN) (Blue et al., 2002). Constructs encoding EGFP-tagged N-terminal fragments of human L-MLCK (amino acids 2-867 and 2-1024) and EGFP-tagged rabbit five DXR actin-binding domains (Smith et al., 2002) were provided by Anne R. Bresnick (Albert Einstein College of Medicine, Bronx, NY) (Dulyaninova et al., 2004).

GST tagged SV1-174 was described previously (Chen et al., 2003). The S2 domains of human nonmuscle myosin IIA (NP002464, coding amino acid sequences 837-1188, LLQVSRQ-THEAQIQ), human nonmuscle myosin IIB (NP005955, coding amino acid sequences 844-1192, LLQVTRQ-ETKNHEA) and human skeletal muscle myosin II (NP005954, coding amino acid sequences 843-1246, LLKSAET-ETVSKAK) were subcloned into the pFastBac1 vector (Invitrogen, Carlsbad, CA) with a C-terminal hexa-histidine (6×His) tag. The Ig and five DXR domains from L-MLCK were expressed with N-terminal 6×His tags using the pet30a vector (Novagen). 6×His-tagged myosin II S2 domains and MLCK proteins were expressed in Sf9 cells and BL21(DE3)LysS, respectively, and purified with Ni²⁺NTA agarose columns (Qiagen, Valencia, CA), according to the manufacturer's protocol.

COS7-2 cells, a highly transfectable clone of COS7 (Kowalczyk et al., 1997), were a generous gift from Kathleen J. Green (Northwestern University, IL). COS7-2 cells were grown in Dulbecco's Modified Eagle high glucose media (DME) supplemented with 10% fetal calf serum (FCS) for ⩽20 passages. A549 human non-small cell lung carcinoma cells [American Type Culture Collection (ATCC), Rockville, MD] were maintained in Ham's F12K medium, 2 mM L-glutamine, 10% FCS for up to 12 passages.

Transfection methods have been described (Takizawa et al., 2006). Briefly, COS7 cells were transfected using Effectene Transfection Reagent (Qiagen) for 24 hours, except in experiments with L-MLCK N-terminal sequences, which were lethal to cells transfected for longer than 16 hours (Dulyaninova et al., 2004). A549 cells at ∼60% confluence were transfected twice with 10 nM dsRNA and Lipofectamine 2000, as recommended (Invitrogen), and grown for 4 days. Two dsRNAs targeting human SV were made as Stealth duplexes (Invitrogen): human (h)SV1, 5′-CAGCCAUAAGGAAUCUAAAUAUGCU-3′ (coding nucleotides 1680-1704) and hSV2, 5′-GCGAUGUUUGCUGCUGGAGAGAUCA-3′ (coding nucleotides 2026-2050). The Stealth duplex sequence, 5′-GAACUAUGAAGGACCACCAGAGAUA-3′, was used as a control dsRNA. Cell growth was similar for mock and dsRNAi-treated cells (data not shown).

Cells were cultured in T25 tissue culture flasks (25 cm²) and detached by incubating at 37°C for 2 minutes with 0.5 ml trypsin/EDTA solution (0.1%/0.40 mg/ml for COS7 cells, or 0.25%/1.0 mg/ml for A549 cells). Trypsin was immediately neutralized with 5 ml culture medium containing 5 μg/ml aprotinin, and the cells were washed once with 10 ml PBS, pH 7.2-7.4. Cells were adjusted to 3.0×10⁴ cells/ml of serum- and antibiotic-free media (2.5 ml/well for COS7 cells, or 2.0 ml/well for A549 cells), and were transferred into wells (9.60 cm²) of six-well plates containing 18×18 mm coverslips coated with 10 μg/ml bovine plasma fibronectin (Sigma). For experiments with chemical inhibitors, 2× concentrations of cells were added to wells containing equal amounts of media with 2× concentrations of inhibitors. At each time point, cells were fixed with 4% paraformaldehyde and counted using a 40× Pan-NeoFluor oil immersion objective (NA 1.3) with phase-contrast optics on a Zeiss Axioskop microscope (Carl Zeiss MicroImaging, Thornwood, NY).

Methods for indirect immunofluorescence microscopy have been described previously (Chen et al., 2003; Takizawa et al., 2006). Briefly, cells were fixed for 20 minutes with 4% paraformaldehyde in PBS, 1 mM MgCl₂, 1 mM EGTA and permeabilized for 5 minutes with 0.1% Triton X-100, PBS before immunostaining. Depending on the experiment, cells were stained for nonmuscle MHC IIA (1:400 dilution), MHC IIB (1:400), β-tubulin (1:400), vinculin (1:200), phospho-Ser19 MLC (1:200), phospho-Thr18/Ser19 MLC (1:200), MYPT1 (1:100), Flag (1:100), and/or SV (1:100). Cross-adsorbed secondary antibodies, conjugated with Alexa-Fluor-350, -488, or -594 were from Invitrogen Molecular Probes (Eugene, OR). F-actin was visualized with phalloidin conjugated to Texas-Red or Alexa-Fluor-350 (Molecular Probes). Slides were analyzed with a 100× Pan-NeoFluar oil immersion objective (NA 1.3) on a Zeiss Axioskop fluorescence microscope with a RETIGA 1300 CCD camera (QImaging Corp., Burnsby BC, Canada) and OpenLab software (Improvision, Lexington, MA). Each channel of the epifluorescence micrographs was scaled automatically using the Auto Level submenu in Adobe Photoshop 7.0 (San Jose, CA). Due to the high background in Alexa-Fluor-350–phalloidin images (Figs 3, 6, 8, 10, 12; supplementary material Figs S2, S3, S4), background was subtracted after Auto Level manipulation. This operation might also remove weak Alexa-Fluor-350 signals. Image intensity profiles were obtained using the Plot Profile subprogram in ImageJ 1.37v (NIH, Bethesda, MD).

dsRNA-transfected A549 cells were allowed to settle onto fibronectin (10 μg/ml)-coated 35 mm chambers (FluoroDish, World Precision Instruments, Sarasota, FL) containing 3 ml DME without phenol red on a DMIRE 2 inverted microscope (Leica Microsystems, Allendale, NJ) in a Plexiglass environmental chamber heated to 37°C. Phase contrast images of the chamber surface were acquired with an HC PL Ph2, 20×, 0.50 NA, air objective lens and a Retiga Exi cooled CCD camera (QImaging), using Open Lab 3.5.2 software (Movies 1, 2). EGFP-SV1-171-transfected COS7 cells on square coverslips (22 mm) in a chamber containing DME + 10% FCS without phenol red were visualized similarly using an HCX PL Apo, 63×, 1.32 NA, oil immersion objective lens (Movie 3). Images were processed with OpenLab to generate movie files.

Binding assays with myosin II S2 domains have been described (Chen et al., 2003). Briefly, 6×His-tagged myosin S2 domains (100 μg) were incubated for 1 hour at 4°C with glutathione-Sepharose (50 μl) containing pre-bound GST-SV1-174 or GST (100 μg) proteins in binding buffer (10 mM MOPS, pH 7.5, 1 mM ATP, 5 mM MgCl₂, 50 mM NaCl, 0.05% β-mercaptoethanol). Beads were transferred to a 2 ml column, and the columns were washed with 0.5 ml of binding buffer and then with 100-μl portions of binding buffer containing increasing concentrations (0.1, 0.15, 0.2 and 0.3 M) of NaCl. Binding assays with L-MLCK fragments (100 μg) were performed similarly, except that the binding buffer was 10 mM MOPS, 100 mM imidazole, pH 7.5, 100 mM KCl, 50 μM EGTA, 0.05% β-mercaptoethanol, 3.25% glycerol, and the column was washed with 0.25 ml portions of a buffer containing 20 mM MOPS, 20 mM imidazole, pH 7.5, 0.1 mM EGTA, 0.1% β-mercaptoethanol with increasing concentrations (0.1, 0.1, 0.15, 0.2 and 0.3 M) of NaCl. GST proteins and bound myosin or L-MLCK fragments were eluted from the washed columns with 250 μl of 5 mM glutathione in washing buffer with 0.3 M NaCl. Fractionated proteins were analyzed by immunoblot analysis using anti-GST antibodies (goat polyclonal; 1:1000; Amersham) and anti-6×His antibodies (mouse clone 27E8; 1:1000).

SDS-PAGE and immunoblotting have been described (Takizawa et al., 2006). Primary antibodies and dilutions: anti-SV (H340), 1:10,000; anti-MHC IIA, 1:10,000; anti-MHC IIB, 1:10,000; anti-MLCK, 1:3000; anti-ERK1/2, 1:3000; anti-GST, 1:1000; and 6×His, 1:1000. HRP-conjugated secondary antibodies and dilutions: anti-rabbit IgG, 1:10,000; anti-mouse IgG, 1:10,000; anti-goat/sheep IgG; 1:10,000.

Research was supported by NIH grants GM33048 to E.J.L., and AR41653 and HL073050 to M.I. We thank Patricia Gallagher and Anne Bresnick for the gifts of MLCK plasmids, and Fumio Matsumura for antibody against phosphorylated MLC used in preliminary experiments. We also thank Anne Bresnick, Fumio Matsumura, Tara Smith and Jessica Crowley for critical reading of the manuscript, and Louise Ohrn and Donna Castellanos for technical assistance.