Smooth muscle archvillin: a novel regulator of signaling and contractility in vascular smooth muscle

The mechanisms by which protein kinase C (PKC) and extracellular-signal-regulated kinases (ERK1/2) govern smooth-muscle contractility remain unclear. Calponin (CaP), an actin-binding protein and PKC substrate, mediates signaling through ERK1/2. We report here that CaP sequences containing the CaP homology (CH) domain bind to the C-terminal 251 amino acids of smooth-muscle archvillin (SmAV), a new splice variant of supervillin, which is a known actin- and myosin-II-binding protein. The CaP-SmAV interaction is demonstrated by reciprocal yeast two-hybrid and blot-overlay assays and by colocalization in COS-7 cells. In differentiated smooth muscle, endogenous SmAV and CaP co-fractionate and co-translocate to the cell cortex after stimulation by agonist. Antisense knockdown of SmAV in tissue inhibits both the activation of ERK1/2 and contractions stimulated by either agonist or PKC activation. This ERK1/2 signaling and contractile defect is similar to that observed in CaP knockdown experiments. In A7r5 smooth-muscle cells, PKC activation by phorbol esters induces the reorganization of endogenous, membrane-localized SmAV and microfilament-associated CaP into podosome-like structures that also contain F-actin, nonmuscle myosin IIB and ERK1/2. These results indicate that SmAV contributes to the regulation of contractility through a CaP-mediated signaling pathway, involving PKC activation and phosphorylation of ERK1/2.

Archvillin is a ~250-kDa membrane skeleton protein that localizes to cortical punctae in differentiating C2C12 cells and, with dystrophin, to costameres, which are sites of force transduction at the plasma membrane in striated muscle . Archvillin is a larger isoform of nonmuscle supervillin, which binds tightly to membranes, F-actin and smooth-muscle and nonmuscle myosin II (Chen et al., 2003;Pestonjamasp et al., 1995;Pestonjamasp et al., 1997). Supervillin is isolated with signaling and lipid-raft-organizing proteins from neutrophil plasma membranes, and has been implicated in the regulation of cell adhesion (Nebl et al., 2002;Pestonjamasp et al., 1997). Supervillin and archvillin both contain sequences that activate the transcriptional activity of the androgen receptor (Ting et al., 2002), and nuclear localization has been observed Pestonjamasp et al., 1997;Ting et al., 2002). Overexpression of the musclespecific archvillin sequences in differentiating C2C12 cells inhibits myotube formation, consistent with a dominantnegative effect during early myogenesis .
In this paper, we identify smooth-muscle archvillin (SmAV) as a unique isoform of supervillin (SV) that interacts with CaP. SmAV is most similar to, but is distinct from, skeletal-muscle archvillin. We discovered the CaP-SmAV association by yeast two-hybrid analyses with CaP as bait. The purposes of the present study were: (1) to confirm the CaP-SmAV interaction; and (2) to test whether SmAV contributes to the regulation of contractility through the CaP-mediated signaling pathway, which requires PKC activation and phosphorylation of ERK1/2.
The interaction was confirmed both by reciprocal blot overlays and by colocalization of CaP and C-terminal sequences of supervillin (c-SV) that lack the actin-and myosin-II-binding sites. CaP and c-SV colocalize to or near the plasma membrane after overexpression in COS-7 cells and before and after permeabilization with detergent. CaP and SmAV colocalize at endogenous levels to the cortices of freshly isolated smooth-muscle cells, after agonist stimulation. Antisense knockdown of SmAV expression decreases PKCactivated contractility and phosphorylation of ERK1/2, signaling effects similar to those observed previously for a CaP knockdown in this system (Je et al., 2001). SmAV, CaP and ERK1/2 localize with F-actin and nonmuscle myosin IIB to PKC-induced 'podosome-like' structures in A7r5 smoothmuscle cells, suggesting that these structures represent sites of CaP-and SmAV-mediated signaling. Taken together, our results show that SmAV is an essential regulator of contractility in differentiated smooth muscle and suggest that SmAV acts in concert with CaP, PKC and ERK1/2 in the plasma membrane cortex subsequent to agonist stimulation.

Materials and Methods
Yeast two-hybrid screening CaP cDNA (Je et al., 2001) was cloned into the bait vector pBD-GAL4 Cam (Stratagene, La Jolla, CA). A HybriZAP ® 2.1 two-hybrid ferret aorta cDNA library was constructed as described (Je et al., 2001), excised from phage and transformed into yeast strain YRG-2 containing the bait vector. Transformants with target plasmids were screened for histidine prototrophy and for β-galactosidase activity according to the manufacturer's instructions (Stratagene). Interacting plasmid DNA (pAD-GAL4-2.1-c-SmAV) was isolated from a positively interacting colony and identified by DNA sequencing and BLAST search. To confirm the interaction of CaP and the C-terminus of SmAV, the bait and prey were exchanged. CaP cDNA was cloned into pAD-GAL4-2.1 vector (pAD-GAL4-2.1-CaP) and the cDNA of c-SmAV was cloned into pBD-GAL4 Cam vector (pBD-GAL4 Camc-SmAV). The deletion constructs of CaP were made by PCR amplification of the cDNA residues corresponding to amino acids 1-136 and 1-163. They were cloned into the pAD-GAL4-2.1 vector (pAD-GAL4-2.1-Cap136 and pAD-GAL4-2.1-CaP163, respectively).

Expression constructs
A glutathione-S-transferase (GST) expression construct encoding the C-terminal 251 amino acids of SmAV (GST-c-SmAV) was generated by directionally ligating the two-hybrid target vector into the EcoRI/XhoI sites of pGEX-6P-1 (Amersham Pharmacia Biotech, Arlington Heights, IL). Similarly, a chimera of GST and the Cterminal 510 amino acids of bovine supervillin (accession number AF025996; GST-c-BSV) was generated by ligating a partial cDNA into the SmaI/NotI sites of pGEX-6P-3. Recombinant His-tagged ferret h1 CaP was constructed by PCR amplification of the open reading frame (Je et al., 2001) with oligonucleotide primers containing BamHI (sense) and HindIII (antisense) restriction sites. Amplified products were inserted in frame into BamHI/HindIIIdigested pET 30b vector (Novagen, Milwaukee, WI) for bacterial expression and into the pCDNA4/HisMax TOPO vector (Invitrogen Life Technologies, Baltimore, MD) for expression in COS-7 cells. Construction of EGFP-c-SV (EGFP-SV1010-1792) has been described (Wulfkuhle et al., 1999).
COS-7 cells COS-7 cells were grown to 90-95% confluence in six-well plates with Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. Each well was transfected with 10 µg plasmid DNA (10 µg each for double transfections) in Lipofectamine 2000 reagent (Invitrogen). After a 48-hour incubation at 37°C, cells were trypsinized, plated onto coverslips and incubated for another 4 hours at 37°C. Coverslips were washed with PBS and fixed with 2% paraformaldehyde, permeabilized with 0.1% Triton X-100, blocked with 10% goat serum and reacted with primary and secondary antibodies. In some experiments, cells were pre-permeabilized with 0.1% or 0.5% Triton X-100 before fixation.
Ferret aorta cell isolation Ferrets (Marshall Farms, North Rose, NY) were euthanized by an overdose of chloroform and abdominal aortas were excised quickly to a dish containing oxygenated physiological saline solution (PSS; 120 mM NaCl, 5.9 mM KCl, 25 mM NaHCO3, 11.5 mM dextrose, 2.5 mM CaCl2, 1.2 mM MgCl2 and 1.2 mM NaH2PO4, pH 7.4). All procedures were approved by the Boston Biomedical Research Institute Animal Care and Use Committee. Single aorta cells were isolated using a variation of a previously described method (Menice et al., 1997). Each 100 mg of aorta (wet weight) were digested with medium A, which is composed of 1.2 mg CLS 2 collagenase (type II, 390 U mg -1 , lot number MIE4817; Worthington Biochemical, Lakewood, NJ) and 1.6 mg elastase (4.0 U mg -1 ; Boehringer Mannheim, Indianapolis, IN) in a final total volume of 7.5 ml Ca-and Mg-free Hanks' balanced salt solution with 0.2% bovine serum albumin (BSA). Tissue pieces were incubated in a shaking water bath at 34°C for 90 minutes, filtered on a nylon mesh, rinsed and incubated for 20 minutes in digestion medium B (medium A with only 0.74 mg elastase). Tissue pieces were filtered, rinsed and reincubated in digestion medium C [medium B plus 0.5 mg of 5000 U soybean trypsin inhibitor (type II-S; Sigma)]. For all experiments, cells on one coverslip were tested to confirm that they shortened in response to phenylephrine. Cells were fixed and stained for immunofluorescence as described for COS-7 cells. To prevent cell shortening during phenylephrine-induced stimulation, both stimulated and unstimulated cells were placed in a Hanks' solution to which 300 mM sucrose was added to increase tonicity, as previously described (Khalil et al., 1995;Parker et al., 1998). This procedure has been shown not to interfere with signal transduction in smooth-muscle cells, as measured by agonist-induced increases in intracellular calcium concentration and by translocation of CaP and PKC, and is thought to prevent contraction directly at the cross-bridge level (for review, see Parker et al., 1998).

Imaging
COS-7 and aorta cells Images were obtained with a Kr/Ar laser (Radiance 2000, BioRad Laboratories) scanning confocal microscope equipped with Nikon 60ϫ (NA 1.4) or 40ϫ (NA 1.4) oil immersion objective. Images were recorded and quantified with Laser Sharp 2000 (BioRad Laboratories) for Windows NT. Ratio analyses of fluorescent intensity distributions were performed as described (Khalil et al., 1994). The 'surface' fluorescence value was defined as the peak pixel intensity over the outer cell surface, and the 'cytoplasmic' value was taken as the peak value in the core of the cell (see Fig. 8, inset). Each ratio was the mean of values from three line scans from non-nuclear areas of each cell.

Colocalization analysis
The degree of colocalization between pairs of stained proteins in COS-7 cells and aorta smooth-muscle cells was quantified by a modification of a previously published method (Parker et al., 1998) and by using LaserPix image software (BioRad Laboratories). Images were thresholded to eliminate background signal outside cell boundaries. A colocalization coefficient between red and green fluorophores (CR-G) was computed as, where the denominator is the sum of red pixels intensities greater than the threshold and the numerator is the sum of the intensities of red pixels that are greater than threshold and that also have a green component. The amount of reverse colocalization [i.e. (CG-R)] was computed in a similar manner by replacing the denominator of Eqn 1 with the sum of intensities of green pixels that are greater than the threshold. The colocalization coefficients are, by definition, between 0 and 1. A value of 0 indicates no colocalization and a value of 1 indicates complete colocalization.
The pixel intensities in each image were randomized by reading the pixel information within each image into a mathematical matrix. The elements of each matrix were then assigned to new locations within the matrix using a random number generator. After converting the randomized matrices back into images, the colocalization coefficients between the pair of random images were computed. A comparison of the colocalization coefficients for the random images to that for the original images allows an assessment of the degree to which colocalization between a pair of fluorophores was due to chance.

Tissue fractionation
Aortic ferret tissue was equilibrated for 1 hour at room temperature in oxygenated (95% O2, 5% CO2) PSS. Viability was tested with a depolarizing solution consisting of PSS in which 51 mM NaCl was stoichiometrically replaced by KCl. The tissue was further incubated for 1 hour in PSS and quickly frozen in a dry-ice/acetone/ dithiothreitol slurry in either the presence or the absence of 10 µM phenylephrine for 10 minutes. Tissue was pulverized and homogenized at 4°C with 3 ml buffer I (20 mM Tris HCl, pH 7.5, 50 mM NaCl, 250 mM sucrose, 10 mM dithiothreitol (DTT), 3 mM EGTA, 5 mM MgCl2, 1 mM ATP and protease inhibitors) (Je et al., 2001). Homogenized tissue was centrifuged at 165,000 g for 1 hour and the supernatant collected as the cytosolic fraction. The pellet was shaken for 1 hour at 4°C in 3 ml buffer II (20 mM Tris HCl, pH 7.5, 250 mM sucrose, 0.5% Triton X-100, 10 mM DTT, 3 mM EGTA, 5 mM MgCl2, 1 mM ATP and protease inhibitors) and centrifuged at 165,000 g for 1 hour. This supernatant was collected as the Tritonextractable ('membrane') fraction. The pellet was resuspended in 3 ml buffer III [20 mM Tris HCl, pH 7.5, 250 mM sucrose, 0.5% Triton X-100, 1.2% sodium dodecyl sulfate (SDS), 10 mM DTT, 3 mM EGTA, 5 mM MgCl2, 1 mM ATP and protease inhibitors], extracted at 4°C for 1 hour and centrifuged at 165,000 g for 1 hour. This final supernatant was collected as the Triton-resistant ('cytoskeletal') fraction.
Overlay assay Equal amounts of proteins were resolved on SDS-polyacrylamide gels and transferred to PVDF membranes. Relative protein content was confirmed by naphthol blue black (NBB) staining of the membrane before incubation with the overlay solution. Blots were blocked with 5% milk in TBS and incubated for 1 hour with 1.25 µg ml -1 recombinant ferret CaP, GST-c-SmAV or GST in 154 mM NaCl, 10 mM Tris HCl, pH 7.4, 10 mg ml -1 BSA and protease inhibitors (Leinweber et al., 1999). Blots were fixed with 0.5% paraformaldehyde in PBS, neutralized with 2% glycine in PBS, washed six times with PBS and immunostained as for western blots.
Antisense knockdown of SmAV expression in smooth-muscle tissue Experiments were performed as previously described (Kim et al., 2000). The antisense oligonucleotide used was 5′-CCTTGC-CAGAGGTACACCTCAT-3′, corresponding to amino acids HEVYLWQG (residues 1912-1919) of SmAV. A randomized version of this sequence (5′-CACTAGAGTCCAGCTACGCTCT-3′) served as the negative control. BLAST searches confirmed a lack of homology to any other relevant coding sequences. Nucleotides were phosphorothioated to minimize degradation and FITC tagged at their 5′ ends. Oligonucleotides were loaded once each day until day 4, as described previously (Je et al., 2001). Fluorescence microscopy at the end of each experiment confirmed internalization of the oligonucleotide into all cells of the tissue (Je et al., 2001).

Contractility of the muscle strips was measured daily in response to
Journal of Cell Science 117 (21) 51 mM KCl. No significant decreases in the amplitude of contraction to KCl were observed over the 4 day course of the experiment. The SmAV expression level and ERK1/2 phosphorylation levels were measured by densitometry of western blots from day 4 tissue extracts.

Statistics
Unless otherwise noted, significance of difference between means was taken at P>0.05 by a two-tailed Student's t test.

Identification of SmAV as a CaP-interacting protein
We used reciprocal yeast two-hybrid screens to identify a new interaction partner for CaP. We initially used the cDNA encoding ferret (h1) basic CaP as bait in an untargeted screen of a ferret aorta cDNA library. Out of thousands of transformants, one specific interacting clone was identified and verified ( Fig. 1A databases (Altschul et al., 1997) showed 94% identity with human supervillin. In the reciprocal experiment, the supervillin homolog in the bait vector also interacted with CaP as prey (Fig.  1A,C6). A deletion mutant (CaP 1-163) lacking the CLIK motif  of CaP (Fig. 1B) interacted with the supervillin homolog (Fig. 1A,C8), as did a further deletion mutant (CaP 1-136) lacking both the TnI-Like domain and the CLIK motif but containing the single type 3 calponin homology (CH) domain  of calponin (Fig. 1A,C7). The CH domain of CaP is of interest because it has also been identified as an ERK-binding domain (Leinweber et al., 1999). The complete cDNA sequence for the supervillin-related protein in ferret aorta revealed a new splice variant that might be characteristic of smooth muscle. The encoded protein (Fig. 2) is predicted to exhibit a relative molecular mass of 231,457.96 and an isoelectric point of 6.44 (http://us.expasy.org/tools/pi_tool.html). The CaP-binding sequence initially identified by the yeast two-hybrid assay is the C-terminal 251 residues (1823-2073). Alignment with previously described supervillin family members revealed 82% similarity (78% identity) with human archvillin and 78% similarity (74% identity) with human supervillin. The Cterminal 1416 residues are 88% identical to the corresponding sequence of both supervillin and archvillin.
Archvillin, a skeletal-muscle-specific isoform of supervillin, contains two muscle-specific inserts encoded by four differently spliced exons in both human and murine tissues . The first muscle-specific archvillin insert is encoded by exons 3, 4 and 5 in skeletal muscle archvillin ( Fig. 2A, yellow). The large exon 4 is conserved in the smooth-muscle archvillin protein sequence (residues 264-567), except for a short hypervariable region ( Fig. 2A, region II). However, muscle insert sequences encoded by exons 3 and 5 are absent ( Fig. 2A, regions I,III), as is the second muscle-specific archvillin insert ( Fig. 2A, green), which is encoded by exon 9 in murine and human archvillins . Because of the greater similarity to skeletal-muscle archvillin, we have named the protein found in aortic smooth muscle SmAV for smoothmuscle archvillin. Thus, SmAV is intermediate in amino acid sequence between skeletal-muscle archvillin and non-muscle supervillin, presumably owing to alternative exon splicing. The binding sites for myosin II and F-actin that have been recently mapped within the N-terminus of bovine supervillin (Chen et al., 2003) appear to be present and highly conserved within SmAV. SmAV contains a site from residues 1 to 175 (Fig. 2B, pink) with 80% identity to the binding site for nonmuscle and smooth-muscle myosin II in supervillin. Similarly, residues 565-627, 629-707 and 984-1114 (Fig. 2B, brown) contain residues with 97%, 93% and 91% identity, respectively, to the three actin-binding domains of supervillin.
Journal of Cell Science 117 (21) Analysis of the SmAV sequence by motif searching programs at high stringency (http://scansite.mit.edu/ and CD Search, NCBI, http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi) revealed several potentially significant regulatory sites. One of the two nuclear targeting sequences ( Fig. 2A, blue) within the first insert in skeletal-muscle archvillin  is conserved in SmAV. C-Terminal sequences homologous to six domains in gelsolin and villin, characteristic of supervillins and archvillins (Pestonjamasp et al., 1997), are also present in SmAV (Fig. 2B, red ovals and short rectangles). Additionally, several kinase-binding sites and phosphorylation sites are predicted (Fig. 2B). Because CaP has been linked with PKCand ERK-dependent signaling pathways, it is noteworthy that the SmAV sequence predicts the presence of two ERK-binding sites and a motif consistent with phosphorylation by PKC and a proline-directed kinase.

Direct SmAV-CaP binding in vitro
Binding between CaP and SmAV was confirmed by blot overlay assays (Fig. 3). As is shown in Fig. 3A, recombinant CaP from the overlay buffer binds to blotted GST fusion proteins containing the C-terminus of either bovine supervillin (GST-c-BSV, amino acids 1282-1792) or SmAV (GST-c-SmAV, amino acids 1822-2073). Neither protein binds appreciably to equal amounts of GST (Fig. 3A). Quantitative densitometry analysis (Fig. 3B) of the ratio of the CaP immunostained band to that of the total protein band obtained by NBB staining confirms that there is significant binding to both GST-c-BSV and GST-c-SmAV but little binding to GST alone. Specific binding was also observed in the reverse experiment [i.e. overlays with GST-c-SmAV or GST in the overlay buffer and CaP and calmodulin (for comparison) on the blot] (Fig. 3C). Even though as much or more calmodulin was present on the blot, more GST-c-SmAV (as visualized with anti-GST antibody) bound to CaP than to calmodulin (Fig. 3C). By contrast, with an equal amount of GST alone in the overlay buffer, no binding was detected to either CaP or calmodulin on blots, even at long exposures (Fig. 3C). Quantitative analysis (Fig. 3D) of the ratio of the GST immunostained band to that of the total protein band obtained by NBB staining confirmed these observations.
To confirm the domain of CaP responsible for binding to SmAV, purified turkey-gizzard CaP was chymotrypsin digested and fragments were identified by N-terminal sequencing and mass spectrometry (Leinweber et al., 2000). These fragments were used in overlay assays with GST-c-SmAV in the overlay buffer. The results indicate that GST-c-SmAV specifically binds to the 20 kDa fragment and to the undigested full-length CaP, but that the overlay signal with the 14 kDa fragment is almost undetectable (Fig. 3E). The quantitative analysis of the anti-GST-immunostained bands normalized to those for total protein confirms these results (Fig. 3F). The 20 kDa fragment to which GST-c-SmAV bound contains residues 7-152 (Fig. 1B), which encompass the single type-3 CaP-homology (CH) domain of CaP .

Colocalization in COS-7 cells
To verify the interaction between CaP and supervillin family proteins in vivo, we tested for protein colocalization in COS-7 cells, which are transformed epithelial cells that are easily transfectable with multiple plasmids. We transfected these cells with plasmids encoding His-tagged ferret h1CaP and a previously described C-terminal fragment of bovine supervillin tagged with enhanced green fluorescent protein (EGFP-c-SV, amino acids 1010-1792) (Wulfkuhle et al., 1999). This C- terminal bovine supervillin fragment is 94% identical (96% homologous) to ferret SmAV in the region containing the putative CaP-binding site. We used this construct, which lacks the N-terminal sequences of supervillin that interact with Factin and myosin II (Chen et al., 2003;Wulfkuhle et al., 1999), because CaP also binds actin . Thus, colocalization of EGFP-c-SV with CaP associated with actin filaments can be informative in a way that colocalization with full-length SmAV would not be. CaP and EGFP-c-SV did colocalize in COS-7 cells (Fig. 4). In cells expressing EGFP-c-SV only, this protein was concentrated primarily in cytoplasmic punctae and at the cell periphery (Fig. 4A, left, green). All nine randomly acquired images displayed this qualitative pattern. By contrast, cells expressing exogenous CaP by itself displayed both peripheral (12 out of 13 cells) and linear (13 out of 13 cells) staining (Fig.  4A, right, red), consistent with CaP's association with actincontaining microfilaments (Parker et al., 1998). In cells transfected with both constructs, overlap in staining was observed at punctae reminiscent of cell-substrate adhesion plaques (12 out of 12 cells) (Fig. 4B, inset, arrows) and along short filaments in the cell interior (11 out of 12 cells) (Fig. 4B, inset, arrowheads) as well as at cell peripheries (12 out of 12 cells). Calculation of colocalization coefficients confirms that there is a statistically very significant overlap of CaP with SmAV and of SmAV with CaP, compared with that for the same images when randomized (Table 1).
As a negative control, we co-transfected CaP and the empty EGFP vector (Fig. 4C). As previously reported (Wulfkuhle et al., 1999), EGFP alone labeled the nucleus and little colocalization with CaP was observed. CaP again showed a prominent filamentous staining but EGFP does not (Fig. 4C).
To test more rigorously for colocalization of CaP and EGFPc-SV, we pre-extracted doubly transfected COS-7 cells with Triton X-100 before fixation and staining. This procedure releases proteins that are unbound or only loosely bound to cytoskeletal structures. As previously reported (Wulfkuhle et al., 1999), essentially all EGFP-c-SV expressed by itself in COS-7 cells was released by pre-extraction with Triton X-100 before cell fixation (Fig. 5A, right, green). By contrast, exogenously expressed CaP continued to exhibit a linear appearance in Triton-extracted cells (Fig. 5B, red). In cells expressing both CaP and EGFP-c-SV, the extractability of EGFP-c-SV with 0.5% Triton X-100 (Fig. 5C, right, green) and with 0.1% Triton (not shown) was greatly decreased. In these pre-extracted cells, CaP and EGFP-c-SV continued to colocalize extensively in cytoplasmic punctae and along linear structures (presumably microfilaments) even after Triton preextraction (Fig. 5C, right, overlap in yellow). These results suggest that CaP stabilizes the association of the supervillin Cterminus with the cytoskeleton and that EGFP-c-SV might potentiate the localization of CaP to cytoplasmic punctae.

Co-fractionation of CaP and SmAV from smooth muscle
The colocalization of overexpressed CaP and EGFP-c-SV in COS-7 cells suggests an interaction between these two proteins but does not necessarily indicate that the endogenous proteins interact in differentiated smooth muscle. As one test of coassociation of endogenous SmAV and CaP in differentiated smooth muscle tissue, we fractionated at high speed (165,000 g) both unstimulated smooth-muscle tissue and tissue stimulated for 10 minutes with the agonist, phenylephrine (Fig.  6). Three fractions were obtained: a cytosolic fraction (proteins released by homogenization in the absence of detergent); a Triton-extractable ('membrane') fraction (proteins released with 0.5% Triton X-100); and a Triton-insoluble ('cytoskeletal') fraction (proteins released with 1.2% SDS and 0.5% Triton X-100). In both unstimulated and stimulated tissues, SmAV and CaP co-fractionated in the cytoskeletal fraction (Fig. 6), consistent with an affinity for cytoskeletal structures and/or Triton-insoluble membrane domains.

Agonist-dependent colocalization
To define more precisely the localizations of endogenous SmAV and CaP in differentiated smooth-muscle cells, we performed quantitative confocal imaging studies on freshly isolated cells from the aorta of the ferret (Fig. 7). Using an antibody raised against the N-terminal 340 amino acids of human supervillin (anti-H340) (Nebl et al., 2002), we detected a single band at ~231 kDa in western blots of whole-cell homogenates of ferret aorta (Fig. 7A). The size of this protein is consistent with the size of SmAV predicted from the deduced Journal of Cell Science 117 (21) amino acid sequence and suggests that SmAV is the dominant member of the supervillin/archvillin family in aorta smooth muscle. The specificity of this antibody further suggests that it can be used to image SmAV in this tissue.

Table 1. Calponin and SmAV colocalize significantly in COS-7 cells, and agonist stimulation increases the colocalization coefficient in aorta cells
In images of unstimulated smooth-muscle cells that had been freshly isolated from aortic tissue, we found, as has been previously reported (Parker et al., 1998), that CaP is excluded from the nucleus, localizing primarily to linear structures (Fig.  7B, left, green), presumably microfilament bundles (Parker et al., 1998). These previous studies determined that CaP colocalizes with actin in these bundles (Parker et al., 1998). SmAV also is excluded from the nucleus but, in contrast to CaP, is concentrated in a non-filamentous, more globular, pattern throughout the cell interior (Fig. 7B, middle, red). Some punctate staining at the cell periphery is also detected (arrowheads). A merged image (Fig. 7B, right) of the CaP and SmAV signals shows little convincing overlap. Quantification of colocalization coefficients confirmed that the degree of colocalization is lower than that in the COS-7 cells (Table 1). Furthermore, the proportion of SmAV that overlaps with CaP was not statistically different from that due to chance, as measured in randomized versions of the same images. The reverse comparison (i.e. the proportion of CaP that overlaps with SmAV), although low, was significantly greater than that for the randomized images. The difference from the random images is presumably due to the fact that the CaP is organized into discrete bundles overlying the more diffuse pattern of SmAV.
Interestingly, a very different pattern emerges when the cells are stimulated with the α-adrenergic agonist phenylephrine (Fig. 7C). As previously reported (Menice et al., 1997), the addition of phenylephrine to ferret aortic cells causes CaP to translocate to the cell periphery (Fig. 7C, left, green). Under these conditions, the SmAV distribution also shifts towards the cell periphery (Fig. 7C, middle, red), where it distributes with CaP (Fig. 7C, right, yellow). This results in an increase in the colocalization coefficients (Table 1). Furthermore, in the presence of phenylephrine, both the overlap of SmAV on CaP and the overlap of CaP on SmAV are significantly greater than that seen in the corresponding randomized images. Ratio analysis also confirms that the individual translocations of each protein to the cell periphery are significant (Fig. 8).

CaP and SmAV co-translocate with similar time courses
To confirm the statistical significance of the SmAV redistribution, we performed digital image analysis with an algorithm in which the surface fluorescence is divided by the fluorescence in the cell core (Khalil et al., 1994) (Fig. 8, inset). The phenylephrine-induced translocations of both SmAV and CaP were highly statistically significant (Fig. 8A,B). However, although the average ratios for SmAV and CaP were 1.35 and 1.42, respectively, in unstimulated cells, stimulation caused the ratio for SmAV to increase to 2.22, whereas the ratio for CaP increased to 3.83. These results suggest that a larger proportion of the SmAV remains in the center of the cell.
To test whether SmAV and CaP co-translocate with a similar time course, ratios of SmAV distributions were determined as a function of time after phenylephrine addition (Fig. 8C). SmAV targeting to the cell periphery reaches a maximum within 4 minutes of stimulation, and persists through the 10 minutes of observation. This time course is indistinguishable from that of CaP, as previously determined under similar experimental conditions in the same cell type (Menice et al., 1997).

SmAV knockdown decreases PKC-dependent contractility in vascular tissue
To investigate whether SmAV plays a regulatory role in the contractile function of differentiated smooth muscle similar to that documented previously for CaP (Je et al., 2001), we used an antisense knockdown approach that preserves contractility in organ culture (Fig. 9). We achieved ~50% knockdown of SmAV expression in ferret aorta smoothmuscle tissue (Fig. 9A). By contrast, neither actin nor ERK1/2 total protein values changed with antisense treatment (data not shown). We have previously shown that this method transfects antisense oligonucleotides equally into all cells across the thickness of the tissue preparation (Je et al., 2001). Although no significant change was observed in the magnitude of contractility mediated by depolarization (Fig.  9B), SmAV knockdown led to a significant decrease in the magnitude of contractions triggered by activation of PKC with a phorbol ester (Fig. 9C). Treatment with a randomized version of the nucleotides used in the antisense sequence had no significant effect on the magnitude of either type of contraction, as compared with sham-treated muscles (Fig.  9B-D). Consistent with our previous observations that the proportion of phenylephrine-stimulated contraction that persists in a Ca-free buffer (about one-third of the total contraction) is specifically PKC dependent (Khalil et al., 1994;Menice et al., 1997), we found that calciumindependent, phenylephrine-induced contractions were also significantly reduced by SmAV knockdown (Fig. 9D). This phenotype, with respect to both phorbol-ester-and Journal of Cell Science 117 (21) Ca-independent phenylephrine-stimulated contractions, is essentially identical to that observed for CaP knockdown in this tissue (Je et al., 2001). Finally, even though (as mentioned above) total ERK1/2 protein levels do not change with antisense treatment, SmAV knockdown inhibits PKCmediated activation of both ERK1 and ERK2 (Fig. 9E), as detected with an anti-phospho-ERK1/2 antibody. A similar inhibition of PKC-induced ERK1/2 activation has also been observed after CaP knockdown (Je et al., 2001). These results implicate SmAV in both PKC-dependent contractility and signaling to ERK1/2.

Localization to the cell cortex
To define further the intracellular site(s) at or near the plasma membrane to which SmAV and CaP relocate after activation of PKC, we immunolocalized endogenous levels of these proteins in cultured A7r5 vascular smooth muscle cells before (Fig.  10A) and after (Fig. 10B) treatment with a phorbol ester, PDBu. The spread morphology of these cells facilitates resolution of subcellular localizations especially between diverse membrane domains. Phorbol-ester treatment of these cells induces a disassembly of actin filament bundles and the concomitant appearance of podosome-like, tubular structures containing F-actin, myosin II and other cytoskeletal proteins that arise from the ventral cell surface and protrude several micrometers into the cell cytoplasm (Brandt et al., 2002;Fultz and Wright, 2003;Hai et al., 2002;Kaverina et al., 2003). In untreated cells, SmAV is localized to membrane ruffles and non-filamentous structures in the cell interior (Fig. 10Aa). CaP is localized primarily to F-actin-containing filaments (Fig.  10Ag,k). In PDBu-treated A7r5 cells (Fig. 10B), we found that endogenous SmAV colocalizes with F-actin ( Fig. 10Ba-c) and endogenous h1 CaP (Fig. 10Bg-i) at podosomes (arrowheads) and membrane ruffles (arrows). Podosomes also contain nonmuscle myosin IIB (Fig. 10Bd-f) and ERK1/2 ( Fig. 10Bjl). Because the antibodies used in this study are specific on immunoblots (not shown), these results suggest that PKC induces the relocation of SmAV, CaP, nonmuscle myosin IIB and ERK1/2 to podosome-like structures at or near the plasma membranes of vascular smooth-muscle cells. The phenylephrine-induced shifts in protein distributions in A7r5 cells (Fig. 10) are qualitatively similar to those observed in the differentiated aortic smooth-muscle cells (Figs 7, 8), suggesting that the peripheral distribution of SmAV and CaP in the aortic cells might arise from an abundance of similar podosome-like structures (i.e. complexes containing both membrane and cytoskeletal components).

Discussion
This paper describes for the first time a smooth-muscle isoform of supervillin. This isoform (SmAV) was identified as an interacting protein with CaP in a yeast two-hybrid assay. We have verified this interaction using both in vitro and in vivo techniques, including functional assays. SmAV appears to be the predominant supervillin/archvillin homolog in smooth muscle, because western blots of smooth-muscle homogenates with an antibody against supervillin show only a single band at the expected molecular mass of SmAV. Furthermore, only a single RT-PCR product was found for each primer pair during the cloning of the cDNA.
It is noteworthy that the SmAV sequence contains several predicted nuclear targeting sequences and yet imaging of differentiated smooth muscle cells clearly indicates an absence of SmAV from the nucleus in both stimulated and unstimulated cells. By contrast, supervillin (Pestonjamasp et al., 1997) and skeletal-muscle archvillin  can exhibit strong nuclear localization. This apparent difference in the isoforms might be explained by a retention of SmAV in the cytoplasm by virtue of its interaction with CaP and/or with other smoothmuscle cytoskeletal elements.
SmAV associates with CaP biochemically and functionally. We observe SmAV-CaP interactions in reciprocal yeast two- hybrid assays, direct binding on reciprocal blot overlays and colocalization in doubly transfected COS-7 cells. Endogenous SmAV and CaP co-translocate with similar kinetics in response to PKC-stimulating agonists in freshly isolated cells from smooth muscle, and are co-extracted in the detergent-resistant cytoskeletal fraction from this tissue. Endogenous SmAV and CaP also are co-targeted with ERK1/2 to PKC-induced 'podosome-like' structures in A7r5 cells. Finally, antisense knockdown of SmAV inhibits stimulus-mediated contraction through a signaling pathway that is indistinguishable from that previously described for CaP (Je et al., 2001).
Because CaP and supervillin/archvillin both contain binding sites for F-actin and/or myosin II (Chen et al., 2003;Gimona et al., 2002;Gusev, 2001), it is important to realize that the interaction described here does not require the presence of either actin or myosin. Interactions were observed both in vitro with purified recombinant proteins and in vivo, when CaP was co-transfected into COS-7 cells with an EGFP-tagged chimeric protein containing the CaP-binding domain of supervillin but lacking interaction sites for F-actin or myosin II.
By contrast, the co-fractionation of SmAV and CaP in a cytoskeletal/detergent-resistant fraction in both resting and stimulated tissues is consistent with multiple cytoskeletal associations and might reflect a regulatory role for these proteins during actin-and myosin-mediated contractility. Detergent-resistant cell fractions contain many components, including cholesterol-rich membrane domains such as those containing caveolin and other lipid-raft-organizing components, as well as cytoskeletal proteins. In fact, supervillin and skeletal-muscle archvillin associate closely with both lipid-raft-organizing proteins and cytoskeletal elements (Nebl et al., 2002).
The agonist-stimulated translocation of endogenous SmAV and CaP to the cell periphery in both differentiated smoothmuscle cells and A7r5 cells further suggests a role for these proteins in mediating interactions between the cortical membrane skeleton and the contractile apparatus. CaP has previously been reported to translocate from contractile filaments containing α-actin to the β-actin containing network of the submembranous cortex (Parker et al., 1998). Thus, we suggest that CaP and SmAV translocate from different sites within the cell to a common cytoskeletal structure that is in contact with a detergent-resistant component of the cell membrane.
The peripheral localization of CaP and SmAV in differentiated smooth-muscle cells might reflect a recruitment into structures analogous to the 'podosome-like structures' in immortalized A7r5 cultured cells, to which CaP and SmAV also colocalize after PKC activation by PDBu. Podosomes are actin-and myosin-II-associated tubular invaginations of the basal surface of the plasma membrane that have been implicated in remodeling of the cortical actin cytoskeleton (Brandt et al., 2002;Fultz et al., 2000;Gimona et al., 2003;Fultz and Wright, 2003;Li et al., 2001). They also contain αactinin, vinculin and transfected h2 CaP Hai et al., 2002). However, in the same study , h1 CaP transfected into these cells was not detected in podosomes even though displacement of CaP from actin Journal of Cell Science 117 (21) Forces are normalized to the amplitude of contraction of each muscle in response to 51 mM KCl on day 1. **Pр0.01 for antisense compared with sham; ++ Pр0.001 for antisense compared with random. (D) Contraction of muscle strips in response to 10 µM phenylephrine in the absence of extracellular calcium. Forces are normalized to the amplitude of contraction of each muscle in response to 51 mM KCl physiological saline on day 1. *Pр0.05 for antisense compared with sham; + Pр0.05 for antisense compared with random. (E) Densitometric analysis of phospho-ERK1/2 on immunoblots for antisense-or shamtreated muscles exposed to DPBA, normalized to that for muscles treated with a random sequence. (Inset) A typical western blot for phospho-ERK1/2. **Pр0.05. All values are from between three and five separate experiments. filaments is thought to be a prerequisite for podosome formation (Burgstaller and Gimona, 2004). Interestingly, we found that endogenous h1 CaP is indeed localized to the podosomes. Perhaps the dynamics of h1 CaP distribution change when h1 CaP levels increase after transfection. This is the first report of which we are aware for the localization of supervillin, archvillin, endogenous h1 CaP or ERK1/2 to these podosome-like structures. Hai et al. have suggested that podosomes represent molecular scaffolds for PKC-mediated effects leading to Ca 2+ sensitization in smooth muscle (Hai et al., 2002). Because the targeting of ERK to the cell periphery is PKC dependent and associated with Ca 2+ sensitization in smooth muscle (Collins et al., 1992;Khalil et al., 1995), our demonstration of PDBu-mediated recruitment of ERK1/2 to podosomes supports this hypothesis.
The specific interaction between SmAV and the CH domain of CaP is interesting because this domain is known to bind ERK (Leinweber et al., 1999). However, the N-terminal end of SmAV is predicted to contain ERK binding and phosphorylation sites. This raises the possibility that the binding of the CH domain of CaP to SmAV might displace ERK1/2 from CaP and free ERK to interact with SmAV or other proteins in the podosome.
The question arises of the mechanism by which CaP/SmAV downregulation leads to the observed inhibition of contractility. CaP has been previously suggested to function as an adapter protein to facilitate the targeting of PKC and ERK1/2 to a signaling complex in the cell cortex (Leinweber et al., 2000;Menice et al., 1997). Once CaP, ERK1/2 and SmAV arrive at the cell cortex, SmAV might displace ERK1/2 from CaP, allowing ERK to be activated by the upstream kinase, MEK. Subsequent ERK1/2-mediated phosphorylation of caldesmon at Ser-789 might then relieve the caldesmon-mediated inhibition of actomyosin interactions (Sobue et al., 1982). In this model, the absence of either SmAV or CaP inhibits ERKdependent effects but has little or no effect on contractility triggered by calcium-dependent signaling events. It is worth noting that the peak of myosin light chain phosphorylation in this tissue occurs by 1 minute after phenylephrine stimulation and then declines (Menice et al., 1997), preceding the maximal translocation of SmAV (at 4 minutes). Thus myosin phosphorylation pathways might predominate during the Fig. 10. Localization of endogenous SmAV, F-actin, nonmuscle myosin II, CaP and ERK1/2 before and after PDBu treatment of A7r5 smoothmuscle cells. (A) In untreated A7r5 cells, endogenous SmAV (a) appears to be membranous, whereas F-actin (b,e,h), nonmuscle myosin II (d) and CaP (g,k) are primarily associated with microfilaments, and ERK1/2 appears to be cytosolic (j). (B) After PKC activation with PDBu, endogenous SmAV (a,g), F-actin (b,e), nonmuscle myosin II (d), CaP (h,k) and ERK1/2 (j) all localize to podosome-like structures (arrowheads). SmAV, F-actin, myosin IIB and CaP also concentrate at or near membrane ruffles (arrows). Merged images (c,f,i,l) show the overlap of signal from panels on the left (green) with the signal from the middle panels (red); the overlap appears yellow. Scale bar, 25 µm. onset of smooth muscle contractions, whereas caldesmon phosphorylation might play a greater role in the sustained increase in contractility. The presence of predicted ERKbinding motifs and proline-directed phosphorylation sites in SmAV further suggests that interactions with membrane and/or cytoskeletal proteins might be under feedback regulatory control from an ERK pathway.
Finally, the phorbol-ester-induced colocalization of CaP and SmAV with ERK to podosomes might imply a role for these or analogous cortical signaling structures in the regulation of MAPKs in general. Thus, this paper might provide a 'missing link' between previous work documenting the importance of the MAPK cascade to early myogenesis of C2C12 cells (Bennett and Tonks, 1997;Cuenda and Cohen, 1999;Weston et al., 2003) and results showing a dominant-negative effect of archvillin sequences on C2C12 myogenesis . Thus, the stimulus-triggered targeting of CaP and SmAV to the cell cortex might participate in both the regulation of contractility and the coordination of cortical signaling pathways in differentiated and proliferative muscle cells.