|Home | About | Journals | Submit | Contact Us | Français|
Phorbol ester induces actin cytoskeleton rearrangements in cultured vascular smooth muscle cells. Calponin and SM22 α are major components of differentiated smooth muscle and potential regulators of actin cytoskeleton interactions. Here we show that actin fibers decorated with h1 CaP remain stable, whereas SM22 α-decorated actin bundles undergo rapid reorganization into podosomes within 30 min of PDBu exposure. Ectopic expression of GFP α-actinin had no effect on the stability of the actin cytoskeleton and α-actinin was transported rapidly into PDBu-induced podosomes. Our results demonstrate the involvement of CaP and SM22 α in coordinating the balance between stabilization and dynamics of the actin cytoskeleton in mammalian smooth muscle. We provide evidence for the existence of two functionally distinct actin filament populations and introduce a molecular mechanism for the stabilization of the actin cytoskeleton by the unique actin-binding interface formed by calponin family-specific CLIK23 repeats.
Stimulation of cultured rat aortic smooth muscle A7r5 cells with phorbol 12,13-dibutyrate (PDBu) has been used previously to study protein kinase C (PKC)-dependent remodeling of the actin cytoskeleton (Fultz et al., 2000 ; Li et al., 2001a , 2001b , 2002 ; Brandt et al., 2002 ). PDBu causes the induction of force development and contraction in A7r5 cells and leads to the gradual disassembly of actin stress fibers. An additional feature of cultured A7r5 cells is the formation of columnar, peripheral bodies in response to treatment with phorbol-12-myristate-13-acetate (TPA) or PDBu (Fultz et al., 2000 ). The observed tubular organization of the columns, arising from the bottom of the PDBu-treated cells and the presence of actin-associated and focal adhesion proteins are defining features of podosomes (Fultz et al., 2000 ; Li et al., 2001b ; Hai et al., 2002 ) and it was demonstrated recently that, like the podosomes of osteoclasts or macrophages, the A7r5 podosome bodies contain both actin, α-actinin, and vinculin (Hai et al., 2002 ). Conventional PKC (α or βI) activity is necessary for the dramatic cytoskeletal rearrangement and we have proposed that podosomes in A7r5 cells may represent molecular scaffolds where PKC phosphorylates regulatory proteins necessary for Ca2+-sensitization in smooth muscle cells (Hai et al., 2002 ).
A7r5 cells exhibit a phenotype similar to adult smooth muscle cells (Firulli et al., 1998 ), display contractile responses to vasopressin, phenylephrin, and elevated K+-levels, and express a variety of smooth muscle marker genes, including smooth muscle myosin heavy chain and alpha smooth muscle actin. Expression levels for two other markers, namely h1 calponin (CaP) and SM22α are, however, reduced compared with cells derived from primary smooth muscle cultures. CaP and SM22α are each present in almost equimolar amounts to the major actin-associated protein, tropomyosin (TM), in smooth muscle tissue and their expression is a hallmark of differentiated smooth muscle. We have shown previously that CaP stabilizes actin stress fiber bundles in REF 52 rat embryo fibroblasts and A7r5 cells against the actin-depolymerizing agents cytochalasin and latrunculin, but also delays the cytoskeletal disrupting effects of the specific Rho kinase inhibitor Y-27632 (Danninger and Gimona, 2000 ). In vitro, CaP enhances the bundling of actin filaments (Tang et al., 1997 ) and reinforces the strain of α-actinin cross-linked actin solutions (Leinweber et al., 1999 ). By contrast, the closely related calponin family member SM22α (also known as transgelin) induces the formation of actin networks, promoting the rapid conversion of loose actin networks into viscous actin gels, but fails to induce or stabilize actin bundle formation (Shapland et al., 1993 ; Lawson et al., 1997 ). Thus, these two components appear to perform specialized, yet unidentified functions within the actin cytoskeleton, pivotal for smooth muscle differentiation and function.
Calponin family molecules feature two unique sequence motifs, an N-terminal calponin homology (CH) domain (Gimona et al., 2002 ) and one or more copies of a unique 23–29-residue C-terminal tandem repeat (Leinweber et al., 2000 ; Kranewitter et al., 2001 ). In calponin, the three repeats are situated in the C-terminal third of the molecule and form an independent actin-binding domain (Danninger and Gimona, 2000 ; Kranewitter et al., 2001 ). Likewise, the single calponin-like (CLIK23) module in SM22α is necessary for actin binding both in vitro and in vivo (Fu et al., 2000 ). Proteins consisting only of CLIK23 repeats have been identified exclusively in worms (UNC-87 from Caenorhabditis elegans; OV9M from Onchocerca volvulus), and these molecules associate with the myofibrillar apparatus of the bodywall muscle (Goetinck and Waterston, 1994a , 1994b ). UNC-87 binds to actin directly via the CLIK23 modules and forms compact bundles of actin filaments both in vitro and in vivo (Kranewitter et al., 2001 ). The binding of CaP (Leinweber et al., 1999 ) and UNC-87 (Kranewitter et al., 2001 ) is unaffected by the presence of saturating amounts of other actin-binding proteins like α-actinin, filamin, or tropomyosin, suggesting that these molecules do not compete for the same binding site along the actin filament.
In this present study we have investigated the role of calponin family proteins in the regulation of cytoskeletal rearrangements in response to phorbol ester-induced PKC activation employing immunofluorescence microscopy, dual live videomicroscopy and electron microscopy on extracted cytoskeletons. Our results reveal that calponin and SM22α regulate the sensitivity of at least two different actin populations in living cells and demonstrate that the calponin repeats are sufficient for stabilization of the actin cytoskeleton.
GFP α-actinin was a kind gift from Markus Geese (Gesellschaft Fuer Biotechnologische Forschung, Braunschweig, Germany). GFP α-actinin ABD (encompassing residues 34–246) was a gift from Dr. Wolfgang Kranewitter (Lambda Diagnostics, Austria) and was generated by the PCR technique using the appropriate primers and the full-length nonmuscle α-actinin cDNA as a template. GFP-calponin (h1, h2, h1Δt, h2Δt, h1 ABS 1 k.o.) and GFP UNC-87 constructs were described previously (Kranewitter et al., 2001 ; Burgstaller et al., 2002 ). All constructs were sequenced using a LI-Cor model 4000 automated sequencer (MWG Biotech, Ebersberg, Germany). The DsRed-SM22 construct was obtained by subcloning a full-length mouse SM22 cDNA (Gimona and Mital, 1998 ) into the h-dsRed vector (Clontech, Heidelberg, Germany) and was a kind gift from Dominique Brandt (University of Hannover, Germany).
A7r5 rat smooth muscle cells (ATCC, Manassas, VA) were grown in low glucose (1000 mg/l) DMEM without phenol red, supplemented with 10% FBS (PAA, Linz, Austria) and penicillin/streptomycin (Life Technologies, Austria), at 37°C and 5% CO2. For transient expression, cells were grown in 60-mm plastic culture dishes and transfected using Superfect (Qiagen, Hilden, Germany) at 70% confluence, essentially as described elsewhere (Kranewitter et al., 2001 ). Expression and stability of the constructs was assessed by Western blotting using an mAb against GFP (Clontech). Cells were replated onto 15-mm cover slips 16 h posttransfection and prepared for immunofluorescence microscopy after an additional 48 h on glass coverslips. Cells were washed three times in PBS (138 mM NaCl, 26 mM KCl, 84 mM Na2HPO4, 14 mM KH2PO4, pH 7.4), extracted in 3.7% formaldehyde/0.3% Triton X-100 in PBS for 5 min, and fixed in 3.7% formaldehyde (Merck, Darmstadt, Germany) in PBS for 30 min. Alexa 568 phalloidin was from Molecular Probes (Leiden, NL). Fluorescent images were recorded on a Zeiss Axioscope equipped with an Axiocam driven by the manufacturer's software package (all Zeiss, Vienna, Austria) using a 63× oil immersion lens.
Cells were observed in an open, heated chamber (Warner Instruments, Reading, UK) at 37°C on a Zeiss Axiovert TV-135 inverted microscope equipped with epifluorescence, phase-contrast and DIC optics. The objectives 40 ×/NA 1.3 Plan-Neofluar and 100×/NA 1.4 Plan-Apochromat were used with or without 1.6 optovar intermediate magnification. Tungsten lamps, 100 W, were used for fluorescence and phase contrast illumination. Data were acquired using a back-illuminated, cooled charge-coupled-device camera (Princeton Research Instruments, Princeton, NJ) driven by a 16-bit controller. The camera controller was driven by IPLab software (Visitron Systems, Eichenau, Germany), and shutters were used on the illumination ports to minimize photodamage (Anderson et al., 1996 ). The digital images were analyzed on an Apple Power Macintosh G3, using IPLab and Adobe Photoshop 2.5 and 5.5 software.
Stacks of optical sections (z step = 0.1 μm) were captured (63× objective NA 1.4, exposure time 800 ms, 488 nm LASER excitation) using a confocal spinning disk system (QLC100 confocal head from Visitech, England) mounted on a Zeiss Axiovert 100M microscope (Zeiss, Oberkochen, Germany). Images were acquired using a Micromax camera (Princeton Instruments) driven by IPLab version 3.5.5 software (Scanalytics, Fairfax, VA) running on a Macintosh G4 computer. Collected confocal stacks were processed using the demo version of Huygens software (Scientific Volume Imaging, The Netherlands) and further processed with ImageJ 1.28 software (http://rsb.info.nih.gov/ij/) on a PC computer.
Analytical SDS gel electrophoresis on 8–22% gradient polyacrylamide mini-slab gels and Western blotting onto nitrocellulose (Amersham, Austria) was performed as described elsewhere (Gimona et al., 1990 ). Transferred proteins were visualized using horseradish peroxidase–coupled secondary antibodies and the ECL chemiluminescence detection system (Amersham, Vienna, Austria).
Cytoskeletons were prepared for electron microscopy according to the critical point drying protocol by Svitkina and Borisy (1998 ), using modifications from the negative staining method (Small and Sechi, 1998 ). A7r5 cells were replated on 15-mm glass coverslips, treated with 1 μM PDBu for 30 min and extracted for 1 min with 0.25% Triton X-100/0.5% EM grade glutaraldehyde (GA) in cytoskeleton buffer (CB; 150 mM NaCl, 5 mM EGTA, 5 mM MgCl2, 5 mM glucose, 10 mM MES, pH 6.1). After fixation with 1.0% GA in CB overnight, cytoskeletons were postfixed with 0.1% freshly prepared tannic acid (TA; low molecular weight) in water for 20 min, rinsed with H2O three times for 2 min, and incubated subsequently in 0.2% aqueous uranyl acetate for 20 min. For dehydration and critical point drying, the coverslips were transferred to a wire basket, separated by layers of lint-free lens paper. Dehydration was done in a graded series of ethanol dilutions (10, 20, 40, 60, 80, and 100% EtOH 5 min each, 0.15% uronic acid (UA) in 100% EtOH for 20 min, 100% EtOH twice, anhydrous EtOH twice for 5 min each) under continuous stirring, and one more UA postfixation step. The samples were transferred to the critical point device filled with dry EtOH and processed according to the instructions (10 exchanges of EtOH against water-free CO2 with 10-min intervals plus stirring).
Immediately after drying, the samples were transferred to an Edwards E306 high vacuum coater and rotary shadowed at a 45° angle with a 2.5–3.0-nm layer of platinum as measured with a FTM6 quartz crystal monitor. Subsequently, the samples were coated with a continuous layer of carbon from a pointed source. Cut replicas were floated off on 8% hydrofluoric acid, transferred to detergent-containing water, and mounted onto formvar-coated 50-mesh hexagonal EM grids. Electron microscopy was performed on a Zeiss EM-10A with a 50-μm objective aperture. Images were acquired on Kodak Electron SO-163 plate film and digitized with a Umax Astra 2400S scanner.
Monoclonal antibodies to vinculin (clone hVin1), calponin (clone hCaP), tropomyosin (clone TM 311), α-actinin (clone 72.5), α-smooth muscle actin (clone 14A), and β-cytoplasmic actin (clone AC-74) were purchased from Sigma (St. Louis, MO). A polyclonal antibody specific for smooth muscle myosin heavy chain was from Biomedical Technologies Inc. (Stoughton, MA). The affinity-purified rabbit polyclonal antibody against SM22α was prepared in the laboratory in Salzburg using recombinant mouse SM22α as the antigen and was used as described previously (Hirschi et al., 1998 ). Secondary antibodies and phalloidin labeled with Alexa 350 (blue), Alexa 488 (green), or Alexa 568 (red) were from Molecular Probes (Leiden, The Netherlands).
Even after prolonged cultivation, A7r5 cells expressed a set of smooth muscle marker proteins such as smooth muscle α-actin, smooth muscle myosin, and smooth muscle tropomyosins, but also the differentiation markers h1 CaP and SM22α (Figure 1A). In addition these cells expressed significant amounts of β-cytoplasmic actin as well as α-actinin, filamin, vinculin, FAK125, and paxillin (unpublished data).
A7r5 cells display a specific reactivity to micromolar amounts of PDBu and form small, podosome-like structures in the cell periphery (Figure 1B; see also Fultz et al., 2000 ; Hai et al., 2002 ). The outer limit of the podosomes is enriched in α-actinin, leading to a ring-like appearance. F-actin is present also in the center of the podosomes and the column-shaped structures transverse the cell body from the ventral surface all the way up to the dorsal side of the plasma membrane (Figure 1C). Transmission electron miscroscopic images of similar preparations further demonstrated that podosomes formed preferentially at the ends of actin stress fiber bundles (Figure 2A). The core of the podosomes contained a number of short, loosely arranged actin fibers with no detectable inner structural organization (Figure 2, B and C).
The homodimeric actin cross-linking molecule α-actinin was reorganized into podosomes in GFP α-actinin–transfected cells (Figure 3A), identical to the endogenous protein (Hai et al., 2002 ). As evident from time lapse videomicroscopy analyses, rearrangement of the actin cytoskeleton and translocation of α-actinin were initiated at the cell periphery, whereas the central stress fibers still displayed the characteristic striated pattern of α-actinin along the F-actin bundles (Figure 3A). Live cell imaging also revealed that in parallel to peripheral actin cytoskeleton remodeling, PDBu induced contractile movements in the central region of the cell. The cross-linking function of α-actinin, however, was not required for the translocation into podosomes since a GFP fusion construct containing exclusively the actin-binding domain (ABD) of α-actinin was reorganized in a manner identical to the full-length molecule (Figure 3B).
Surprisingly, h1 CaP was not recruited to podosomes. Transfection of A7r5 cells with a GFP h1 CaP construct resulted in the almost complete incorporation of the molecule into stress fibers showing the characteristic association of CaP only with central stress fibers. Cells expressing GFP h1 CaP also displayed a reduced number of peripheral podosomes in response to PDBu and retained an intact and organized actin stress fiber network (Figure 4B). By contrast, GFP- (unpublished data) or DsRed-tagged SM22α were readily incorporated into numerous podosomes (Figure 4A), pointing toward different sensitivities of the two molecules to PKC- and Src-dependent cytoskeleton remodeling.
Imaging of cells transiently transfected with GFP h1 CaP or DsRed-SM22α under live conditions showed the identical behavior as seen above with the fixed and stained cells (unpublished data). Because transient transfection resulted in both cases in a significant overexpression of the respective protein, the effects could be artifactual, although identical amounts of DNA were used for transfection. Thus, we next investigated the behavior of the two molecules simultaneously using double-transfected cells. In resting cells, h1 CaP and SM22α were found colocalized along actin stress fibers in the central part of the cell (yellow color), whereas SM22α was also present in the cell periphery (compare also with Figure 4). On the addition of 1 μM PDBu, DsRed-SM22α underwent a rapid translocation into the transiently forming podosomes, whereas the h1 CaP-decorated thin filaments remained intact throughout the observation period of 60 min (Figure 4C). These results indicate the presence of at least two different populations of actin filaments with markedly different sensitivities to cytoskeletal remodeling and actin dynamics in response to PDBu-induced PKC activation.
The results described above raised the question whether the differences in dynamic remodeling reflects actin isoform sorting in existing actin stress fibers consisting of mixed actin filaments. To address this question, we repeated the above experiments under live conditions using cells transiently transfected with a GFP-tagged β-cytoplasmic actin construct. In unstimulated cells, the localization of GFP-β-actin overlapped almost completely with that of F-actin both along stress fibers and at the cell periphery (unpublished data). However, as shown in Figure 5A, GFP-β-actin was remodeled and translocated to podosomes in response to PDBu. Moreover, in cells double-transfected with GFP β-actin and DsRed-SM22α, both proteins translocated to podosomes at the identical rate, and β-actin filaments were almost completely reorganized into podosomes and peripheral membrane ruffles (Figure 5B). In addition we stained methanol-fixed, PDBu-treated cells with specific monoclonal antibodies to α-smooth and β-cytoplasmic actin, respectively. Both actin isoforms became incorporated into the newly forming podosomes (Figure 5C), suggesting that there is no explicit preference for an actin isotype in the formation of podosomes. Notably, under these conditions we consistently observed a ring of β-cytoplasmic actin surrounding a more densely packed core composed of α-smooth muscle actin.
Together with previous observations, the results obtained this far led us to conclude that h1 CaP is capable of, at least partially, inhibiting the PKC-dependent, Src-mediated remodeling of actin filaments, but the underlying mechanism remained elusive. Smooth muscle h1 CaP uses an unconventional actin-binding interface composed of two independently functioning actin-binding sites, termed ABS1 and ABS2 (Mino et al., 1998 ; Danninger and Gimona, 2000 ). In particular ABS2, formed by multiple copies of the unique calponin repeats (termed CLIK23 repeats) has been shown to form a noncompeting actin-binding site also in other members of the CaP family (see Figure 6A), which otherwise lack a functional ABS1 or other potential actin-binding domains (Kranewitter et al., 2001 ; Burgstaller et al., 2002 ). To further explore the potential molecular mechanism mediating the stabilizing effect of h1 CaP, we used a set of CaP mutants to study the relative contributions of the two actin-binding regions.
The nonmuscle calponin variant h2 CaP binds to actin filaments in vitro and in vivo but its cellular localization is not restricted to the central stress fiber bundles as it is for h1 CaP. Stimulation of GFP h2 CaP-expressing cells with PDBu resulted in the translocation of h2 CaP into podosomes, indicating that the initial, specific localization of the CaP molecule influences the subsequent stabilization of actin filaments (Figure 6B).
The h1 and h2 isoforms of CaP differ markedly in two positions, namely their acidic C-terminal tails and a short sequence, termed ABS1, situated just N-terminal of the triple CLIK23 repeats. We first investigated the role of the ABS1 sequence in h1 CaP, which is mutated and inactive in h2 CaP. In agreement with our previous studies, mutation of this site had no influence on the actin-binding ability of the mutant, and GFP h1CaP ABS1 k.o. stabilized actin filaments identical to the full-length molecule (Figure 6B). Actin binding in the h2 CaP isoform is regulated by the inhibitory tail sequence at the C termini and deletion of this sequence has been shown to enhance actin binding by releasing the inhibition of the ABS2 (Burgstaller et al., 2002 ). Mutants of both h1 and h2 CaP lacking the C-terminal tail localized to the central stress fibers. Deletion of the regulatory tail now also enabled the h2 CaP variant to remain attached to actin bundles in the presence of PDBu and caused a similar cytoskeletal stabilization as seen before with the h1 CaP isoform (Figure 6C). Thus, the ABS2, formed by multiple copies of the CLIK23 module, appeared necessary and sufficient for the stabilizing effect of CaP.
To further test this hypothesis, we finally transfected A7r5 cells with a GFP-tagged version of the C. elegans muscle protein UNC-87, which contains seven copies of the CLIK23 repeat (see Figure 6A for the molecular domain structure of UNC-87) and which binds to actin with high affinity (Kranewitter et al., 2001 ). Like h1 CaP, UNC-87 associated preferentially with central actin stress fibers, also in the presence of PDBu, and UNC-87-bound filaments were stabilized against depolymerization and/or reorganization into podosomes (Figure 6C). A truncated version encompassing only the five C-terminal CLIK23 repeats showed an identical behavior, supporting the essential function of the CLIK23 repeats in actin filament stabilization. We also observed the segregation of two different actin filament populations in cells double transfected with GFP UNC-87 and DsRed-SM22α (unpublished data).
Remodeling of the actin cytoskeleton has been hypothesized as a pivotal event in the regulation of smooth muscle contraction. Calponin and SM22α are highly specific markers for differentiated smooth muscle. They represent almost 2% of the total protein mass in adult smooth muscle and are thus expected to perform smooth muscle-specific functions related to actin cytoskeleton-dependent contractile processes. Our previous studies (Danninger and Gimona, 2000 ) have underscored the ability of CaP to stabilize actin filaments in living cells and to reduce actin remodeling, thus pointing toward a structural role for this molecule. This hypothesis is supported by biochemical data, which showed that α-actinin and CaP cooperate in the formation of mechanically resilient actin gels in vitro (Leinweber et al., 1999 ). Remarkably, this latter study also embarks already on the concept of noncompetitive actin-binding interfaces of the two molecules along the filament, despite the presence of potentially interfering CH domain-based actin-binding domains (Gimona et al., 2002 ). Together these data support our contention that the unconventional actin-binding sites of CaP are the primary interfaces for contacting the actin filament (Danninger and Gimona, 2000 ; Burgstaller et al., 2002 ).
In contrast to CaP, the close relative SM22α has been shown to stabilize loose actin gels and to cause actin filament gelation (Shapland et al., 1993 ), but was ineffective in in vitro studies to enhance filament stabilization. Although the single C-terminal CLIK23 module is necessary and sufficient for actin binding (Fu et al., 2000 ), the affinity for F-actin appears significantly lower than that of CaP, and binding of SM22/transgelin to F-actin is sensitive to even low ionic strength. Nevertheless, both native and GFP- or DsRed-tagged SM22 constructs (Zhang et al., 2002 ) decorate actin filaments in smooth muscle, indicating a physiological role for SM22 in thin filament regulation, despite the low binding affinity in vitro.
Within the highly organized stress fibers, actin filaments undergo a constant dynamic rearrangement. The rate of treadmilling depends essentially on the balance between polymerization (as determined by the incorporation of free monomers into the polymer chain) and the severing and subsequent depolymerization of existing actin polymers mediated by proteins like gelsolin or members of the ADF/cofilin family. Thus, cross-linking of actin filaments appears to be an appropriate method to stabilize actin filaments against depolymerization and to reduce actin dynamics in living systems. One of the striking findings of this study, however, is that CaP, but not the strong cross linker α-actinin can prevent stress fibers from reorganization in response to phorbol ester-induced PKC activation. This result may be interpreted in the following way: α-actinin cross-linking contributes to the net “mechanical” stability of two cross-linked F-actin strands (Hanein et al., 1998 ; Tang et al., 2001 ; Volkmann et al., 2001 ) and aids in the parallel alignment of adjacent actin polymer bundles. CaP, by contrast, appears to contribute a “conformational” stability preventing the binding or activation of e.g., severing proteins, the activity of which is required for the continuous, controlled disassembly of actin filaments (see Figure 7). Because CaP does not compete for actin binding with proteins like gelsolin or ADF/cofilin, this protection activity may be achieved by subtle alterations of the three-dimensional structure of the CaP-bound actin filaments. Indeed, CaP has been demonstrated to alter the structure of the actin filament upon binding (Bartegi et al., 1999 ).
We have delineated the CLIK23 repeat module as the molecular domain responsible for the stabilization of actin filaments. Previous work for our group (Danninger and Gimona., 2000 ; Kranewitter et al., 2001 ; Burgstaller et al., 2002 ) and others (Bartegi et al., 1999 ) has indicated that the actin-binding interface formed by the multiple CLIK23 repeats does not compete with the site occupied by actin-binding molecules using a tandem CH domain ABD, like α-actinin (Hodgkinson et al., 1997 ; Gimona et al., 2002 ). Thus, our findings that both α-actinin and the isolated ABD are strongly enriched in podosomes and that overexpression of α-actinin fails to prevent PDBu-induced actin remodeling are consistent with the hypothesis that the unique actin-binding interface of CaP and UNC-87 is necessary and sufficient for actin filament stabilization. This view is further supported by our experiments using the h2 variant of CaP (Burgstaller et al., 2002 ). Whereas full-length h2 CaP is partially translocated into podosomes upon PDBu treatment, a mutant lacking the regulatory tail sequence and rendering the ABS2 constitutively accessible, is not.
Forced expression of h1 CaP prevents a subset of actin filaments from undergoing remodeling into podosomes, whereas SM22α overexpression had no apparent effect on the general morphology of unstimulated A7r5 cells. Thus, the specific stabilization of actin filaments is independent of the actin isotype but is driven rather by the type of actin-binding protein associated with the filament. However, this latter explanation does not exclude the possibility that β- and α-actin recruit different sets of actin-binding proteins, and we cannot rule out the intriguing possibility that SM22 and CaP associate with and distinguish different actin isoforms and prepare them for remodeling and stabilization, respectively. The localization of SM22α, but not CaP, to the sites of membrane ruffling/actin polymerization argues for a role of SM22α in promoting actin polymerization. Thus, our observation of a specific interaction of CaP with the central stress fibers is consistent with the hypothesis of structural differences within actin filaments (Egelman and Orlova, 2001 ), likely driven by the binding of the actin-interacting component(s).
A striking feature of PDBu-induced cytoskeleton remodeling is the fact that actin reorganization starts at the cell periphery, in close proximity to the end of actin stress fibers and the border of an adjacent focal adhesion. We have documented previously that CaP binding to F-actin terminates shortly before the filaments overlap the focal adhesion site (Burgstaller et al., 2002 ; and Figure 5) leaving a “bare zone” of undecorated actin filaments. Inspection of the turnover of various focal adhesion components in response to PDBu-induced cytoskeleton remodeling reveals that these zones indeed mark the origins of podosome formation and Arp2/3-dependent de novo actin polymerization activity (Kaverina, Stradal, and Gimona, unpublished results). Together, these findings lend further support to the contention that CaP serves to stabilize mature actin bundles in the cell center.
Thin filament stability and dynamics are likely to be regulated by competition between molecules increasing the stability of actin filaments (like tropomyosin), and the severing and depolymerizing activities of proteins like ADF/cofilin. In addition to cytoskeletal rearrangements required for motility and cell protrusion, these effects likely contribute also to the contractile machinery in muscle cells (see also Cooper, 2002 ). Ono and Ono (2002 ) have recently demonstrated a role for TM in regulating actin dynamics in C. elegans bodywall muscle by competition with ADF/cofilin but acknowledged that other factors, like UNC-87, may exist in addition to TMs, which may perform similar roles. From our results we strongly suggest that calponin-family proteins serve to play such a regulatory role in smooth muscle and that this process involves structural alterations of the actin filament rather than direct competition for actin-binding sites (Bartegi et al., 1999 ). Finally, the actin filament–stabilizing effect partially explains CaP's function as a potential tumor suppressor (Horiuchi et al., 1999 ), similar to the demonstrated activities of high-molecular-weight tropomyosins TM1 and TM2 (Gimona et al., 1996 ; Shah et al., 1998 ; Prasad et al., 1999 ). Considering that PDBu-induced formation of podosomes is also observed in human intestinal smooth muscle cells (HISM; unpublished data), this particular cytoskeletal remodeling may be a general feature of smooth muscle cells, which is tightly regulated in normal tissue by balanced levels of h1 CaP and SM22α. Reducing the endogenous levels of CaP would thus favor the formation of podosomes and amplify the potential for cell migration, causing an increased invasive behavior of dedifferentiated smooth muscle cells in malignant vascular anomalies.
We are grateful to Ulrike Tischler for expert technical assistance, J.V. Small for critical comments and helpful discussions, and Dominique Brandt for supplying the DsRed SM22 cDNA. This work was supported in part by grants from the Austrian Science Foundation (P-14285 and P-15120 to M.G. and P-15710 to G.P.R.).
Abbreviations used: CaP, calponin; F-actin, filamentous actin; ABD, actin binding domain; PDBu, phorbol 12,13-dibutyrate; TPA, phorbol-12-myristate-13-acetate; GFP, green fluorescent protein.
Online version of this article contains video material. Online version is available at www.molbiolcell.org.