Caldesmon mutant defective in Ca²⁺-calmodulin binding interferes with assembly of stress fibers and affects cell morphology, growth and motility

Calmodulin (CaM) plays a crucial role in the regulation of cell motility (Chin and Means, 2000). An increase in intracellular Ca²⁺ concentration leads to the formation of a Ca²⁺-CaM complex that subsequently signals cell locomotion. Inhibition of the Ca²⁺-CaM complex significantly slows migration of cultured rabbit gastric cells (Ranta-Knuuttila et al., 2002). Likewise, inhibition of CaM activity by caged peptide leads to the rapid cessation of locomotion and membrane extension in eosinophils (Walker et al., 1998). Consistent with its role in regulating cell motility, the spatial distribution of Ca²⁺-CaM has been shown to be crucial for the extension of filopodia in growth cones (Cheng et al., 2002). Furthermore, CaM activation is greatest in lamellae at the leading edge of fibroblasts in early stages of the wound-healing process (1-4 hours after wounding), followed by a relocation to the posterior end (Hahn et al., 1992). Its involvement in the regulation of cell motility indicates that the Ca²⁺-CaM complex is able to control the dynamics of the cytoskeleton via downstream effectors. The regulation of myosin light chain kinase (MLCK) by Ca²⁺-CaM and its role on cell contractility have been intensively studied (Gough and Taylor, 1993; Walker et al., 1998). However, the role of the Ca²⁺-CaM complex in actin localization and polymerization is poorly understood, although in vitro studies have demonstrated that actin filaments are regulated by Ca²⁺-CaM (Pritchard et al., 1987) and have identified several actin-binding proteins whose activities are regulated by the Ca²⁺-CaM complex (Shirinsky et al., 1992; Sobue et al., 1983).

One of these actin-binding proteins, caldesmon (CaD) (Smith et al., 1987) has been shown to stabilize microfilaments (Ishikawa et al., 1989a; Ishikawa et al., 1989b; Warren et al., 1994), to enhance actin binding of tropomyosin (Yamashiro-Matsumura and Matsumura, 1988) and to inhibit myosin ATPase activity in vitro (Smith et al., 1987). Most of the functional domains of CaD are located in its C-terminal half (Marston and Redwood, 1991). Previously, we subcloned the C-terminal fragment of human fibroblast caldesmon (CaD39) and demonstrated that the bacterially synthesized C-terminal fragment CaD39 retained the abilities to bind to actin, to potentiate tropomyosin binding and to inhibit myosin ATPase activity (Novy et al., 1993). When stably expressed in Chinese hamster ovary (CHO) cells, CaD39 stabilized microfilaments and protected endogenous tropomyosin from turnover (Warren et al., 1994). Furthermore, we demonstrated that the activity of force-expressed CaD39 was regulated similarly to endogenous CHO CaD (Warren et al., 1996).

The effects of CaD on the actomyosin system can be reversed in vitro by Ca²⁺-CaM (Smith et al., 1987). Two well-conserved domains have been identified within the C-terminal half of CaD, sites A and B, that are responsible for the interaction with Ca²⁺-CaM (Marston et al., 1994; Mezgueldi et al., 1994; Zhan et al., 1991). In human fibroblast CaD, the sequences for sites A and B are: ⁴⁵⁹MWEKGNVF and ⁴⁸⁹SRINEWLTK (underlined tryptophan (W) residues are the crucial residues mutated below). In human smooth muscle CaD, site A and site B localized between amino acids 715-722 and 744-752, respectively (Hayashi et al., 1992; Humphrey et al., 1992). It has been shown that the tryptophan residues within the two domains are crucial for Ca²⁺-CaM binding (Graether et al., 1997). Despite a wealth of in vitro evidence, there is very little evidence suggesting that CaD activity is indeed regulated by Ca²⁺-CaM in living cells or that it plays a role in controlling actin filament dynamics. To study the regulation of CaD by Ca²⁺-CaM in vivo and to investigate further the role of CaD in Ca²⁺/CaM-regulated cell motility, we created a CaD39 mutant, CaD39-AB, in which both of the tryptophan residues essential for Ca²⁺-CaM binding were replaced with alanine. Here, we report that, with the exception of Ca²⁺-CaM binding, CaD39-AB retains actin-binding properties similar to that of wild-type CaD39 and is still able to potentiate tropomyosin binding to actin and to inhibit actomyosin ATPase. Expression of CaD39-AB in CHO cells leads to disruptions of stress fibers and focal adhesions, a defect in cell morphology, and a delay in cell-cycle progression. Moreover, CaD39-AB impairs cell motility during wound healing. Cells expressing CaD39-AB migrate much more slowly than control cells and change their direction of movement more frequently than normal cells.

Point mutations were introduced into CaD39 fragment by splicing through overlapping extension (SOE) (Horton et al., 1989; Novy et al., 1993), a two-step polymerase chain reaction that replaced the tryptophan residues essential for Ca²⁺-CaM binding with alanine residues. Briefly, CaD39 subcloned in a pCB6hx vector was amplified in two separate reactions using a pair of 5′ vector primer (5′-GGTAGGCGTGTACGGTGGGAGGTCTAT-3′) and a mutagenic primer (see below), or the complementary mutagenic primer and a 3′ vector primer (5′-TGGAGTGGCAACTTCCAGGGCCAGGA-3′). The two resulting fragments were then gel purified and used in a second round of PCR reactions with the flanking primers (5′ and 3′ vector primers) to generate either the 39CaMA or 39CaMB mutant. The double mutant CaD39-AB was generated by mixing the 39CaMA mutant with the mutagenic primers for 39CaMB in sequential PCR. After gel purification, the PCR product was digested with XbaI and subcloned into either the eukaryotic expression vector pCB6hx or the prokaryotic expression vector pET8c. The mutagenic primers are listed below with the mutated codon underlined.

39CaMA: 5′-CATTCCCTTTCTCCGCCATACTCTTGA-3′ and 5′-CATCAAGAGTATGGCGGAGAAAGGGAATG-3′;

39CaMB: 5′-GGTTTTAGTTAGCGCTTCATTGATGCGGC-3′ and 5′-GCATCAATGAAGCGCTAACTAAAACCCCAG-3′

Plasmids pETCaD39 and pETCaD39-AB were transformed individually into BL21(DE3)pLysS. Recombinant proteins were expressed and purified as previously described (Novy et al., 1993). Briefly, the soluble fraction of heat-treated bacteria lysate was subjected to ammonium sulfate fractionation, followed by separation of the protein mixture through two rounds of fast performance liquid column chromatography (Mono S column and Superose 12 column). Purified CaD39 and CaD39-AB proteins were identified by 12.5% sodium-dodecyl-sulfate polyacrylamide-gel electrophoresis (SDS-PAGE) and western-blot analysis with the monoclonal antibody C21 against CaD.

Binding of CaD39-AB or CaD39 to Ca²⁺-CaM was performed by a modified enzyme-linked immunosorbent assay (ELISA) as previously described (Lin et al., 1991). 100 μl of a 5 μg ml^–1 solution of bovine brain calmodulin (Sigma-Aldrich, St Louis, MO) was added to wells of Immulon II microtiter plate and incubated overnight at 4°C. After removing excess Ca²⁺-CaM, an increasing amount of either CaD39 or CaD39-AB was added into the wells and incubated at 37°C for 30 minutes. The wells were washed three times with PBS-Tween-Ca²⁺ (10 mM sodium phosphate buffer, pH 8.2, 0.15 M NaCl, 0.05% Tween-20, 0.2 mM CaCl₂). Bound CaD39 or CaD39-AB was then incubated with the monoclonal antibody C21 against CaD (used at 1000-fold dilution) at 37°C for 30 minutes. After washing three times with PBS-Tween-Ca²⁺, the wells were incubated for an additional 30 minutes at 37°C with horseradish-peroxidase-conjugated anti-mouse IgG at 1000-fold dilution. The enzyme activity in each well was detected by adding substrate ABTS [0.03% H₂O₂ and 1 mM 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonate acid) in 0.1 M citrate buffer, pH 4.2] and assaying the color intensities at 405 nm using an ELISA reader (SpectraMax Plus, Molecular Devices, Sunnyvale, CA).

Actin binding abilities of the recombinant proteins were measured in a cosedimentation assay (Novy et al., 1993). 9.3 μM actin filaments were incubated with various amounts of CaD39 or CaD39-AB in binding buffer containing 10 mM imidazole buffer (pH 7.0), 30 mM KCl, 0.1 mM EGTA and 1 mM DTT. After incubation at room temperature for 40 minutes, the reaction mixtures were centrifuged at 130,000 g for 30 minutes in an airfuge. The pellet from each reaction was resuspended in binding buffer. Both the supernatants and the corresponding pellets were analysed by 12.5% SDS-PAGE after they were dissolved in SDS sample buffer. The proteins were stained with Coomassie Blue, scanned and analysed using Scion Image software (Scion, Frederick, MD). Additional actin-binding assays were performed by incubating 3.6 μM recombinant human fibroblast tropomyosin isoform 3 (hTM3) (Novy et al., 1993), 9.3 μM F-actin and 0.5 μM CaD39 or CaD39-AB under similar conditions.

ATPase activities were measured at 29°C as described previously (Novy et al., 1993) in 20 mM imidazole buffer, pH 7.0, 50 mM KCl, 2 mM MgCl₂, 1 mM dithiothreitol, 0.1 mM EGTA, 0.04 μM rabbit skeletal heavy meromyosin (HMM; Cytoskeleton, Denver, CO), 12 μM skeletal muscle actin, 2.5 μM recombinant human tropomyosin isoform 5 (hTM5) and 2.8 μM CaD39 or CaD39-AB. The reactions were started by the addition of 1 mM ATP containing 5 μCi [γ-³²P]-ATP. In the reversal of CaD inhibition, 2.5 μM CaM and 2 mM Ca²⁺ were included in the reaction.

CHO cells were cultured in Dulbecco's modified Eagle's medium (DMEM) plus 10% fetal bovine serum (FBS) at 37°C in 5% CO₂. Plasmid DNA was introduced into CHO cells using Lipofectamine reagent (Invitrogen, Carlsbad, CA). 24 hours after transfection, cells were selected in DMEM plus FBS medium containing 500 μg ml^–1 G418 (Gibco, Gaithersburg, MD). For establishing stable clones expressing CaD39-AB, single cells were cloned after 2 weeks of drug selection. Each cell clone was expanded and tested for expression of CaD39-AB. Stable lines m4-65 and m5-8 were selected for this study. The nonexpressing G418-resistant cell line C3 was used as an experimental control. A previously isolated stable line 39C15 that expresses wild-type CaD39 was also included for comparison in this study (Fig. 1). To detect the expression of force-expressed protein in CHO cells, the monoclonal antibody C21 (Lin et al., 1991) was used in immunofluorescent microscopy. As described before (Li et al., 2003), C21 antibody only recognized overexpressed exogenous proteins when used at 1000-fold dilution. To detect focal adhesions, cells were stained with anti-vinculin monoclonal antibody (Sigma-Aldrich, St Louis, MO) at a 300-fold dilution. To examine the organization of stress fibers, fixed and permeabilized cells were incubated with rhodamine-conjugated phalloidin (Molecular Probes, Eugene, OR). The fluorescent staining was visualized using confocal microscopy. z-Axis series of xy optical planes were obtained at 0.2 μm intervals using a BioRad Radiance 2100MP (BioRad, UK). Projections and merged images were constructed using the co-localization module Laser Sharp 2000 software (BioRad, UK).

Confluent cells were wounded by scraping the coverslip with a sterile pipette. After 1 hour of recovery in a 37°C incubator, the coverslip was transferred to a Dvorak-Stotler perfusion chamber (Lucas-Highland, Chantilly, VA) that positioned on the stage of a Zeiss inverted microscope prewarmed to 35°C. Cell behavior was monitored for 5 hours through a 40× objective and a high performance CCD camera (Optronics, Galito, CA). Images were digitized directly onto the hard drive of a Macintosh computer equipped with a Data Translation frame grabber board (Data Translation, Marlboro, MA) and 2D-DIAS software. Cells at the leading edge were manually traced and the perimeters converted to β-spline replacement images that were used to compute the positions of the cell centroids (Soll, 1995). Methods for computing parameters derived from 2D-DIAS have been presented elsewhere (Soll, 1995; Soll and Voss, 1998). Parameters for cell motility, including velocity, net path length, directional change and persistence, were computed from the position of the cell centroid. Area, a parameter of cell shape, is defined as the area of the final shape at frame n minus the area of any holes. Roundness is a percentage measure of how efficiently a given amount of perimeter encloses the area. A circle has the largest area and has a roundness of 100%. `Positive flow', a measure of the area change per unit, was calculated by overlapping cell outlines between frames n and n–1. The regions in frame n not contained in frame n–1 were considered to be `expansion zones'. The summed area of expansion zones divided by the total cell area in frame n multiplied by 100 represents the positive flow. Likewise, the summed area of `contraction zones' divided by the total cell area in frame n multiplied by 100 represents `negative flow'.

Tropomyosin-enriched actin filaments were isolated from 15 100-mm dishes of 39C15 and 16 150-mm dishes each of m5-8 and m4-65 as previously described (Lin et al., 1984). Briefly, the cytoskeleton of interphase cells was extracted in KTG solution (0.1 M PIPES, pH 6.9, 5 mM MgCl₂, 0.2 mM EGTA, 0.05% Triton X-100, 4 M glycerol) and homogenized in the presence of 5 mM ATP and 5 mM Mg²⁺. After centrifugation at 10,000 g, the supernatant was incubated with anti-tropomyosin monoclonal antibody. Tropomyosin-enriched filaments were pelleted by low-speed centrifugation (10,000 g) and analysed by western blots using C21 antibody. The loading volumes were adjusted among different expressing cell lines so that equivalent amounts of the tropomyosin-enriched microfilaments were later analyzed by western blot. As demonstrated previously, the yield of tropomyosin-enriched microfilaments isolated by this method is roughly proportional to the amount of stress fibers in the cells (Lin et al., 1984; Matsumura et al., 1983), indicating that this method preferentially isolates actin bundles. Growth rates for various cell lines were determined by inoculating cells that are exponentially growing into 60-mm dishes. The initial cell number was 4×10⁴ cells per dish. One dish was harvested every 24 hours for determination of cell numbers.

It was shown previously that two Trp residues at the C-terminal fragment of human fibroblast CaD (CaD39), W460 and W494, were crucial for Ca²⁺-CaM binding (Graether et al., 1997). Using site-directed mutagenesis, a mutant of the CaD39 fragment was generated in which both W460 and W494 were substituted with alanine, and named CaD39-AB (Fig. 1). To characterize its in vitro binding properties, CaD39-AB and the wild-type CaD39 fragment were bacterially expressed and purified as described previously (Li et al., 2003). The purified proteins were subjected to a Ca²⁺-CaM binding assay using a modified ELISA method (Lin et al., 1991). In the presence of Ca²⁺, wild-type CaD39 was able to bind to CaM (Fig. 2). The binding was EGTA sensitive (data not shown) (Lin et al., 1991). Replacing the tryptophan residues with alanine almost entirely abolishes the binding of CaD39-AB to Ca²⁺-CaM (Fig. 2).

Actin-binding ability of CaD39-AB was examined in a cosedimentation assay in which the purified protein CaD39-AB or CaD39 was incubated with skeletal muscle F-actin. After high-speed centrifugation, comparable amounts of CaD39-AB and CaD39 were detected in the pellet fractions (Fig. 3A,B), whereas neither CaD39 nor CaD39-AB was detected in pellet fractions in the absence of actin filaments (data not shown). Further quantitative analyses in three additional experiments revealed no significant differences between CaD39-AB and the wild-type CaD39 in either the affinity or the kinetics of actin binding (Fig. 3C), suggesting that the actin binding ability was not affected by the mutation at the Ca²⁺-CaM binding sites. We previously found that CaD39 potentiated actin binding of tropomyosin (Novy et al., 1993). Therefore, we further examined the actin-binding affinities of recombinant hTM3 in the presence of CaD39 or CaD39-AB. As shown in Fig. 4, in a low salt concentration (30 mM KCl and no Mg²⁺), where hTM3 alone barely binds to actin (Fig. 4, 1S and 1P), the addition of either CaD39 or CaD39-AB greatly increased the amount of hTM3 bound to actin filaments (Fig. 4, 2P and 3P), suggesting that the ability of the mutant to enhance actin binding of tropomyosin is similar to that of the wild-type CaD39 fragment and intact CaD protein (Novy et al., 1993).

The ability of CaD39-AB to inhibit actin/tropomyosin-activated HMM ATPase activity was compared with that of wild-type CaD39 using a myosin HMM ATPase assay as described previously (Novy et al., 1993). As shown in Fig. 5, the actin/tropomyosin-activated HMM ATPase activity decreased by 49.2% in the presence of 2.8 μM wild-type CaD39. The inhibition of ATPase by CaD39 could be reversed by 2.5 μM Ca²⁺-CaM. Similarly, CaD39-AB inhibited the actin/tropomyosin-activated HMM ATPase to about 42.6% of the control activity in the absence of CaD (Fig. 5). This inhibition could not be significantly reversed by the addition of Ca²⁺-CaM. These results suggest that the ability of CaD39-AB to inhibit myosin ATPase is neither impaired by the mutations at the Ca²⁺-CaM binding sites nor effectively reversed by the addition of Ca²⁺-CaM.

To investigate the consequences of expressing CaD39-AB in CHO cells, we generated stable clones expressing the mutant fragment. After transfection and cell cloning, cell lines C35, C41, m4-2, m5-8 and m4-65 were obtained, expressing low to high amounts of exogenous protein (Fig. 6A). Two overexpressing clones, m5-8 and m4-65, were used for further characterization of the cells expressing CaD39-AB, and the nonexpressing drug-resistant line C3 served as a negative control. To verify that mutant phenotypes observed in the overexpressing cells resulted from expression of CaD39-AB rather than that of wild-type CaD39, we also included a CaD39-overexpressing cell line, 39C15 (Warren et al., 1994) for comparison. The stable clones used in this study are summarized in Fig. 1.

To detect the localization of CaD39-AB in the stably expressed cell lines, we used immunofluorescent microscopy with monoclonal antibody C21 against the C-terminal fragment of CaD (Lin et al., 1991). When used at a 1000-fold dilution, C21 antibody could only detect the presence of overexpressed exogenous proteins in the stable clones (Fig. 7E,H), but barely any cross-reactivity could be detected with endogenous CHO CaD in nonexpressing C3 cells (Fig. 7B). Consistent with previous findings (Warren et al., 1994), the overexpressed CaD39 protein was found on stress fibers and ruffle regions of 39C15 cells (Fig. 7E,F). By contrast, CaD39-AB in m4-65 cells displayed diffuse staining throughout the cytoplasm but seemed to concentrate at ruffle regions (Fig. 7H). Surprisingly, the expressed CaD39-AB was not found on stress fibers counter-stained with rhodamine-labeled phalloidin (Fig. 7I). This staining pattern of CaD39-AB was observed in all the isolated stable clones irrespective of the expression level (data not shown). Furthermore, the transiently expressed CaD39-AB mutant (36 hours after transfection) also showed a diffuse staining pattern and did not incorporate into the stress fibers (data not shown). Compared with the control cells, phalloidin staining is more prominent in 39C15 cells (Fig. 7F), suggesting that CaD39 stabilizes the structure of stress fibers (Warren et al., 1996). However, the stress fiber structure in CaD39-AB-expressing cells appears to be disintegrated (Fig. 7I).

The observation that CaD39-AB was not associated with stress fibers was further supported by the isolation of tropomyosin-enriched microfilaments. When tropomyosin-enriched actin bundles were isolated and analysed on Coomassie-Blue-stained gels (Fig. 8B) and western blots (Fig. 8C-E), the amounts of associated tropomyosin 5 (TM5) in both 39C15 and m4-65 cells was very similar (Fig. 8D). However, after calculating the differences in the initial cell numbers used for the actin bundle isolation and in the volumes loaded in the gel, we found that the yield of actin bundles isolated from m4-65 cells was 2.2-times less than that isolated from 39C15 cells. Furthermore, the amount of associated CaD39-AB in either the m5-8 or m4-65 cells (Fig. 8C, lane 2, and Fig. 8C,E, lane 3) was significantly lower than that of CaD39 in the 39C15 clone (Fig. 8C,E, lane 1). The total amount of CaD39-AB expressed in the clone m5-8 was comparable to that of CaD39 expressed in the isolated CaD39-expressing clone 39C15, whereas the amount of CaD39-AB in m4-65 was significantly higher (Fig. 8A). We have previously demonstrated that the molar ratio of the exogenous CaD39 to the endogenous CaD protein is about 80 in the 39C15 cells (Warren et al., 1994). Therefore, the expression level of CaD39-AB in m5-8 and m4-65 were no less than 80 times higher than that of the endogenous CHO CaD. Overall, these data demonstrate that force-expressed CaD39-AB is unable to bind stress fibers, whereas CaD39 is readily incorporated into stress fibers. Because the in vitro binding assay revealed that actin-binding affinity of CaD39-AB was not impaired, it is possible that CaD39-AB is still able to bind actin filaments in vivo but that those CaD39-AB-containing microfilaments are somehow prohibited from forming stress fibers. In support of this suggestion, when total cytoskeletal extracts were analysed by western blot, the amount of CaD39-AB associated with total actin filaments was comparable to that of CaD39 in overexpressing cells (data not shown).

To investigate whether CaD39-AB is able to disrupt the assembly of stress fibers and focal adhesions, we double-labeled the cells of the overexpressing clone m4-65 with phalloidin and an anti-vinculin monoclonal antibody. As shown in Fig. 9, control C3 cells displayed straight, well-developed stress fibers across the cell bodies (Fig. 9A). Similar phalloidin staining patterns were also observed in CaD39-overexpressing 39C15 cells (Fig. 9E). By contrast, stress fibers in CaD39-AB-overexpressing cells were short, fragmented and abnormally localized to cell peripheries (Fig. 9I). The tropomyosin localization revealed by anti-tropomyosin antibodies (CG3 and LC24) in these CaD39-AB-expressing cells was very similar to stress-fiber staining pattern (data not shown). In agreement with the actin organization, staining of vinculin revealed that focal adhesions in both the control cells and the cells expressing wild-type CaD39 exhibited linear organization along stress fibers (Fig. 9B,F). Prominent vinculin staining could be observed at both ends of the stress-fiber arrays (Fig. 9C,D,G-H). In m4-65 cells, vinculin-containing focal adhesions appeared more prominently at cell peripheries, whereas those inside the cell bodies were mostly small and punctate (Fig. 9J,K). Moreover, most of the focal adhesions at cell peripheries did not overlap with stress fibers (Fig. 9K), although these two structures were usually in close proximity (Fig. 9L). Excluding the cortical region, CaD39-AB-overexpressing cells often exhibited no or weak focal-adhesion structures at the tips of stress fibers (Fig. 9L), suggesting that formation/maturation of focal adhesions was impaired by the disruption of stress fibers in overexpressing cells.

Under phase-contrast microscopy, we observed significant changes in cell shape in CaD39-AB-overexpressing cells compared with either wild-type CHO or CaD39-overexpressing cells (Fig. 7). To characterize better the morphological defects, cells from clones m5-8 and m4-65, as well as cells from clone 39C15, were seeded in culture dishes so that similar cell densities were obtained after 2 days of culture. After fixation, the area and roundness of each individual cell were quantified using the 2D-DIAS software program (Soll and Voss, 1998). As shown in Table 1, cells from both m5-8 and m4-65 clones were significantly larger and rounder than those of clone 39C15. However, 39C15 cells showed no difference from parent CHO cells in cell morphology (Warren et al., 1994). The severity of the morphological defects roughly paralleled the expression levels of CaD39-AB (Table 1), demonstrating that changes in cell shape are caused by the expression of CaD39-AB.

While culturing, we noticed that the CaD39-AB-expressing clones grew slowly. To determine whether there was a cell-cycle defect in CaD39-AB-expressing clones, we measured growth rate. We found that the number of 39C15 cells increased at a rate similar to that of CHO cells (Warren et al., 1994) (R. D. Eppinga and J.J.-C.L., unpublished). By contrast, the number of cells in both the m4-65 and m5-8 lines increased more slowly (Fig. 10). The cell doubling times of the 39C15, m5-8 and m4-65 clones were 16 hours, 21 hours and 24 hours, respectively, indicating that the expression of CaD39-AB mutant suppresses cell growth.

To investigate whether overexpression of CaD39-AB affects cell movement, cells from the non-expressing cell line (C3) and the overexpressing cell lines (m4-65 and m5-8) were analysed for cell motility in a wound-healing assay. Using the 2D-DIAS software program, motility parameters were quantified. The average of each parameter is presented in Table 2. CaD39-AB-overexpressing cells were significantly slower than control cells. Furthermore, in overexpressing cells the persistence (velocity/directional change) was significantly reduced and directional change was significantly higher when translocating, suggesting that they turn more frequently than control cells. These motility defects were apparent in a comparison of centroid tracks (Fig. 11). Over a 4-hour period after wounding, the centroid tracks of translocating m4-65 cells were short and jagged. Compared with m4-65 cells, the centroid paths of control cells were longer and had more persistent directions. Similar defects were observed in another CaD39-AB-expressing cell line, m5-8 (Table 2). In contrast to CaD39-AB-expressing cells, no motility defects were detected in 39C15 cells overexpressing the wild-type CaD39 fragment (Warren et al., 1996). Taken together, these results demonstrated that expression of a CaD39 defective in Ca²⁺-CaM binding impairs speed and directionality during wound healing.

We further examined the dynamics of lamellipodia extension during wound healing. Figs 12 and 13 show time-lapse images of migrating cells from C3 and m5-8 clones, respectively. Consistent with the characterization of lamellipodia dynamics previously demonstrated in normal fibroblasts (Fisher et al., 1988), control C3 cells extended a thin lamellipodium along the substrate, typically along the entire free edge of the wound (Fig. 12A). Difference pictures of representative control cell (Fig. 12A, arrow) with color-coded lamellipodia are shown in Fig. 12B. In contrast to control cells, cells overexpressing CaD39-AB failed to form continuous lamellipodia in the direction of the wound (Fig. 13A). Instead, multiple small, discrete protrusions were observed at the edges of these cells, each extending and retracting independently (Fig. 13B). Consistent with the observed defects in expansion zones in difference pictures, the parameters of positive flow and negative flow, measures of expansion areas and contraction areas, were significantly lower in mutant cells expressing CaD39-AB compared with that in control C3 cells (Table 2).

In early studies, two opposing theories were proposed regarding the mechanism for the regulation of CaD by Ca²⁺-CaM (Sobue et al., 1983; Smith et al., 1987). In the `flip-flop' model (Sobue et al., 1983), Ca<sup>2+</sup>-CaM competes with actin for the binding sites on CaD and therefore reverses the inhibitory effects of CaD on actomyosin ATPase by dissociating it from actin filaments. In favor of this model, the Ca<sup>2+</sup>-CaM binding domain has been shown to be juxtaposed to the actin-binding domain (Fraser et al., 1997; Huber et al., 1998a; Wang et al., 1997). By contrast, the `four state' model (Smith et al., 1987) proposes that Ca²⁺-CaM can reverse the inhibitory effects of CaD by two means: either by dissociating CaD from the actin filaments or by forming an uninhibited ternary complex of Ca²⁺-CaM-CaD-actin-tropomyosin. The fact that mutant CaD39-AB protein defective in Ca²⁺-CaM binding retains the actin-binding ability militates against the `flip-flop' model. Instead, it suggests that the binding sites for actin and Ca²⁺-CaM are independent of each other and, therefore, that Ca²⁺-CaM might be able to reverse the inhibitory effect of CaD on microfilaments without completely abolishing its binding to actin-tropomyosin. Consistent with our result, Helfman et al. (Helfman et al., 1999) reported that transiently transfected non-muscle CaD remained associated with stress fibers even when CaD activity was suppressed by Ca²⁺-CaM. Recent studies of the thin filaments have indicated a more complex mechanism for CaM regulation of CaD activity in the smooth muscle cells. Notarianni et al. (Notarianni et al., 2000) have isolated, from smooth-muscle thin filaments, a structurally novel CaM that can bind to CaD in the absence of Ca²⁺. This might indicate the presence of another inhibitory state of CaM-CaD-actin-tropomyosin complex that appears to be more sensitive to Ca²⁺ concentration change than the CaD-actin-tropomyosin complex. It was proposed that, when associated with small amount of Ca²⁺, the complex of the novel CaM and CaD could rearrange itself, which would in turn allow the actin-tropomyosin quickly to alter its conformation and to activate myosin ATPase activity (Marston et al., 1998). Whether this novel isoform of CaM also exists in non-muscle cells remains to be determined.

The CaD domains for myosin ATPase inhibition have been mapped to the C-terminal half of full-length CaD (Wang and Chacko, 1996), three of which overlap with the actin-binding sites. The mutations at Ca²⁺-CaM-binding sites used in this study neither disrupt actin binding nor interfere with the ability to inhibit actomyosin ATPase. Interestingly, Huber et al. (Huber et al., 1998a) have generated a mutant of the 10 kDa C-terminal fragment of chicken gizzard CaD Cg1 whose activity could not be regulated by Ca²⁺-CaM. Although the actin binding was not impaired by the mutation, the ability of Cg1 to inhibit myosin ATPase activity was significantly reduced. The difference in the ability to inhibit myosin ATPase by Cg1 and CaD39-AB is probably due to the different natures of mutants. In addition to the replacement of the tryptophans at site B, Cg1 also bore mutations of another five amino acids that flanked the tryptophan residue (Huber et al., 1998b). It is likely that mutation of these adjacent residues, rather than the mutations of the tryptophan itself, caused the decrease in the inhibition of myosin ATPase activity.

CaD39-AB displays a similar binding ability to actin filaments in vitro, just as the wild-type CaD39 does, yet it is not incorporated into stress fibers when force-expressed in CHO cells. Moreover, it appears to disrupt the organization of stress fibers and focal adhesions in overexpressing cells. It is likely that the CaD39-AB still binds to microfilaments in vivo but prevents them from forming actin bundles by bundling proteins. In support of this notion, Ishikawa et al. (Ishikawa et al., 1998) have shown in vitro that, together with tropomyosin, CaD completely prevents the bundling protein fascin from binding to actin filaments and thus inhibits its actin-bundling activity. Inhibition of actin bundling by CaD and tropomyosin can be reversed by the addition of Ca²⁺-CaM, suggesting that regulation of CaD by Ca²⁺-CaM plays an important role in controlling the actin-bundling activity of fascin.

We have previously reported that stress fibers in CaD39-overexpressing cells become more resistant to disruption by cytochalasin (Warren et al., 1994), suggesting a role for CaD in stabilizing stress fibers. A similar result was obtained in the transient transfection of full-length CaD, in which expression of exogenous CaD led to the formation of thicker stress fibers (Yamashiro et al., 2001). We propose that CaD serves dual functions in the regulation of stress fiber assembly. First, CaDand tropomyosin together prevent actin filaments from bundling. The presence of Ca²⁺-CaM is required in the ternary structure of actin, tropomyosin and CaD to regulate CaD activity so that tropomyosin and CaD are in the appropriate position to allow the binding of fascin to actin filaments (Ishikawa et al., 1998), leading to assembly of stress fibers. Once stress fibers are formed, associated CaD can stabilize the structures and prevent it from disassembly, possibly by inhibiting the activities of actin-severing proteins such as gelosin (Ishikawa et al., 1989b) and cofilin (Yamashiro et al., 2001). It is likely that the requirement of Ca²⁺-CaM for actin bundling is transient, in which case Ca²⁺-CaM would be released after assembly of stress fibers, allowing it to regulate CaD activity in a temporally and spatially specific way. Taken together, CaD plays dual roles in actin dynamics: first, in the regulation of bundling protein activity and, second, in the stabilization of stress fiber structures in the presence of Ca²⁺-CaM.

When overexpressed in CHO cells, CaD39-AB might integrate randomly into actin filaments and prohibit bundling in the presence of tropomyosin, whereas actin filaments associated with endogenous CaD can bundle together to form few stress fibers through Ca²⁺-CaM regulation. Because the expression level of the endogenous CaD is less than that of CaD39-AB (less than 1/80th of the level), force-expressed CaD39-AB will be functionally competing with the endogenous CaD in the overexpressing cells, which locks actin filaments in an inhibitory conformation for actin bundling and leads to the disintegration of stress fibers in overexpressing cells.

It is noteworthy that the disruption of stress fiber assembly in overexpressing cells accompanies disruption of focal-adhesion structures. The association of focal adhesions with CaD39-AB-containing microfilaments that are not bundling might result in poor colocalization between vinculin and stress fibers. Moreover, it has been shown that formation of focal adhesions is dependent on the development of tension generated in the actomyosin system (Kaverina et al., 2002). Inhibition of actomyosin ATPase by CaD39-AB, along with the disruption of actin integrity in overexpressing cells, might result in insufficient tension within the actomyosin system for the formation/stabilization of focal adhesion structures. The diffused staining of CaD39-AB and the disorganized structures of stress fibers and focal adhesions in the overexpressing clones resemble the phenotypes that Helfman et al. (Helfman et al., 1999) observed in their transient expression of wild-type CaD. They showed that the expression of full-length CaD in fibroblasts inhibits the assembly of stress fibers, disrupts focal adhesions and causes a defect in cell contractility, all of which can be reversed by elevated cytoplasmic Ca²⁺ concentrations.

Coincident with disruption of stress fibers and focal adhesions, expression of CaD39-AB delays cell-cycle progression and alters cell morphology. The CaD39-AB-overexpressing cells grow much more slowly and exhibit a larger and less polarized cell shape than the parent CHO cells or CaD39-overexpressing cells. The defects in cell shape and cell cycle probably result from the disruption of stress fibers and focal adhesions by CaD39-AB mutant protein. It has been shown that disruption of stress fibers by cytochalasin D inhibits the synthesis of DNA, RNA and proteins, as well as S-phase entry (Iwig et al., 1995). Furthermore, cytochalasin D also blocks growth-factor-induced expression of cyclin D1 and phosphorylation of pRb (Bohmer et al., 1996). It is thus believed that an intact cytoskeleton system is important in the cyclin/cyclin-dependent-kinase pathway, possibly by modulating the proper localization of G1-phase regulators (Lohez et al., 2003). Disruption of stress fibers in overexpressing cells might result in a delay in the G1/M-phase transition, thus slowing down cell-cycle progression. Meanwhile, disintegrated stress fibers and focal adhesions might impair the generation of proper tension in the cortical cytoskeleton required for maintaining the polygonal shape of CHO cells. Because Ca²⁺-CaM cannot reverse the inhibitory effect of CaD39-AB on myosin ATPase activity, cell contractility would be greatly impaired by the expression of CaD39-AB mutant, which in turn leads to a rounder, larger cell.

Disorganized microfilaments and weakened contractility also contribute to the motility defects of CaD39-AB-expressing cells in the wound-healing assay. It has been shown that the actin network is highly organized during migration. Within the cell body, actin is highly organized into fibers either perpendicular or parallel to the direction of cell migration. Transverse fibers perpendicular to the direction of migration produce contractions that narrow the anterior portion of a migrating cell and induce detachment from neighboring cells (Hahn et al., 1992), whereas stress fibers aligned with direction of translocation provide contraction along the axis of movement that contributes to forward migration (Wang, 1984). Lack of integrated stress fibers, together with a deficiency in myosin ATPase activity, might impair the generation of coordinated contractility across the cell body and therefore slow down cell movement. Moreover, it has been shown that changes in actin organization and cytoskeleton-based contraction forces are crucial to the orientation of lamella extension and the direction of cell migration (Parker et al., 2002).

During wound healing, membrane extension in CaD39-AB-expressing cells is not as active as in control cells. Instead of producing a broad lamellipodium over the front edge, a CaD39-AB-expressing cell extends small multiple protrusions. Immunofluorescence microscopy has showed concentrated CaD39-AB staining at the leading edge during wound healing (data not shown). In addition, Ca²⁺-CaM is known to concentrate in the lamellipodia during the period that we monitored wound healing (Hahn et al., 1992). Therefore, defects in membrane extension might be directly caused by the presence of CaD39-AB within the lamellipodia, which affects actin bundling and/or actin polymerization. Portions of lamellipodia contain bundles of actin filaments called microspikes (Ballestrem et al., 1998), and actin-bundling proteins such as fascin are believed to play an important role in their assembly (Small et al., 2002). It is possible that CaD39-AB at the leading edge inhibits bundling activity, suppresses the formation of microspikes and thus destabilizes the structure of membrane protrusion (Fisher et al., 1988). Alternatively, but not exclusively, in vitro studies have demonstrated that CaD inhibits Arp2/3-induced actin polymerization and that this inhibition can be reversed by Ca²⁺-CaM (Yamakita et al., 2003). Loss of regulation by Ca²⁺-CaM might allow CaD39-AB to inhibit Arp2/3 activity at the leading edge, which might in turn reduce membrane protrusion (Pollard et al., 2000).

Taken together, our results have identified Ca²⁺-CaM as a key regulator of CaD activity. Regulation of CaD by Ca²⁺-CaM plays an important role in the assembly of stress fibers, in the generation of cell contractility and probably also in Arp2/3-induced polymerization of actin. Furthermore, motility defects shown in CaD39-AB-expressing cells resemble those observed in CaM-deficient cells (Walker et al., 1998), suggesting that CaD is involved in Ca²⁺-CaM control of cell locomotion.

We thank D. Wessels for constructive discussion and J.-Y. Choi, W. Pierce and J. Sholley for their great technical assistance. This work is supported by grant HD18577 from the US National Institutes of Health and a grant from the W. M. Keck Foundation.