Regulation of Actin Binding and Actin Bundling Activities of Fascin by Caldesmon Coupled with Tropomyosin*

Human fascin is an actin-bundling protein and is thought to play a role in the formation of microfilament bundles of microspikes and stress fibers in cultured cells. To explore the regulation of fascin-actin interaction, we have examined the effects of culture cell caldesmon and tropomyosin (TM) on actin binding activity of human fascin. Caldesmon alone or TM alone has little or no effect on the actin binding of fascin. However, caldesmon together with TM completely inhibits actin binding of human fascin. When calmodulin is added, the inhibition of fascin-actin interaction by caldesmon and TM becomes Ca2+ dependent because Ca2+/calmodulin blocks actin binding of caldesmon. Furthermore, as phosphorylation of caldesmon by cdc2 kinase inhibits actin binding of caldesmon, phosphorylation can also control actin binding of fascin in the presence of TM. As expected by the inhibition of fascin-actin binding, caldesmon coupled with TM also inhibits actin bundling activity of fascin. Whereas smooth muscle caldesmon alone or TM alone shows no effect, caldesmon together with TM completely inhibits actin bundling activity of fascin. This inhibition is again Ca2+ dependent when calmodulin is added to the system. These results suggest important roles for caldesmon and TM in the regulation of the function of human fascin.

Human fascin is an actin-bundling protein and is thought to play a role in the formation of microfilament bundles of microspikes and stress fibers in cultured cells. To explore the regulation of fascin-actin interaction, we have examined the effects of culture cell caldesmon and tropomyosin (TM) on actin binding activity of human fascin. Caldesmon alone or TM alone has little or no effect on the actin binding of fascin. However, caldesmon together with TM completely inhibits actin binding of human fascin. When calmodulin is added, the inhibition of fascin-actin interaction by caldesmon and TM becomes Ca 2؉ dependent because Ca 2؉ /calmodulin blocks actin binding of caldesmon. Furthermore, as phosphorylation of caldesmon by cdc2 kinase inhibits actin binding of caldesmon, phosphorylation can also control actin binding of fascin in the presence of TM. As expected by the inhibition of fascin-actin binding, caldesmon coupled with TM also inhibits actin bundling activity of fascin. Whereas smooth muscle caldesmon alone or TM alone shows no effect, caldesmon together with TM completely inhibits actin bundling activity of fascin. This inhibition is again Ca 2؉ dependent when calmodulin is added to the system. These results suggest important roles for caldesmon and TM in the regulation of the function of human fascin.
Fascins belong to a unique family of actin-bundling proteins (1,2), which include sea urchin fascin (3)(4)(5)(6)(7)(8), HeLa 55-kDa actin-bundling protein (9,10), and the gene products of Drosophila singed (11)(12)(13). All of these proteins make F-actin aggregate side-by-side into bundles (4, 8 -10) and are localized in the structures containing actin bundles including filopodia and stress fibers of cultured cells (1,9,10,14), bristles of Drosophila, actin bundles of Drosophila nurse cells (11,13), and microspikes and microvilli of sea urchin eggs and coelomocytes (6,7). Some of these structures such as filopodia and microspikes are known to be dynamic structures responding to various biological signals. For example, the fertilization of sea urchin eggs induces the formation of fascin-containing microvilli on their surfaces (7). Filopodia contain fascin-actin bundles and are actively extending and retracting during cell movement of fibroblasts. Fascin should thus be involved in the assembly and disassembly of actin bundles in such structures. However, the mechanisms for the regulation of actin binding of fascin are not well understood.
One way to regulate fascin-actin interaction is phosphorylation. We have shown that human fascin is phosphorylated at Ser-39 in vivo in human neuroblastoma cells upon treatment with 12-O-tetradecanoylphorbal-13-acetate, a tumor promoter (16). The same site is phosphorylated by protein kinase C, which results in the inhibition of actin binding of fascin (16,17). However, the stoichiometry of fascin phosphorylation by protein kinase C is low, suggesting that kinases other than protein kinase C may be involved. Furthermore, the stoichiometry of in vivo phosphorylation is also low in the absence of TPA, which suggests the presence of other mechanisms to control fascin-actin interactions under normal conditions.
Another possibility for the control mechanism of fascin-actin interaction is that other actin binding proteins modulate actinfascin association. Our previous study indeed showed that skeletal muscle tropomyosin (TM 1 ) inhibits actin-fascin binding (18). However, we found that TM isolated from cultured rat cells has only a slight effect on actin binding of human fascin (18). On the other hand, human fascin inhibits actin binding of cultured rat cell TMs, depending on TM isoforms. Rat cultured cells contain at least six isoforms (19 -23), which are divided into two groups based on the molecular weights and the strength of their actin binding. The isoforms with high M r (high M r TMs) bind to actin more strongly than the isoforms with low M r (low M r TMs) (21). Fascin completely inhibits actin binding of low M r TMs but shows little effect on high M r TMs. These observations suggest that TM could control the fascin-actin interactions if actin binding of TM is high enough. It is thus possible that a protein that can increase actin affinities of TMs may be able to modulate fascin-actin interactions. Caldesmon is a candidate as caldesmon and TM increase each other's actin binding affinities (24 -27). Caldesmon is an actin-and calmodulin binding protein (28,29), the actin binding of which is regulated by Ca 2ϩ /calmodulin. Caldesmon is also reported to bind to other proteins including myosin and TM. In in vitro reconstituted system, caldesmon is known to inhibit actomyosin ATPase. We showed that caldesmon together with TM inhibits actin-severing activity of gelsolin (30). These results suggest the regulatory roles of caldesmon in actomyosin-based motility and the organization of the actin cytoskeleton.
We have thus examined whether caldesmon has any effects on the interactions between human fascin and TMs. We have found that caldesmon, when coupled with TMs, not only blocks the binding of human fascin to actin but also dissociates fascin from actin. Our results suggest that caldesmon coupled with TM controls the formation of fascin-actin bundles.

MATERIALS AND METHODS
Cell Culture-SV40-transformed rat embryo cells (REF-WT4A) and HeLa cells were cultured in Dulbecco's modified Eagle's medium con-taining 10% newborn calf serum and 10% calf serum, respectively, in an atmosphere of 5% CO 2 and 95% air at 37°C . For large scale culture of  HeLa cells or SV40-transformed rat embryo cells (REF-WT4A), cells were grown in large plates (245 ϫ 245 ϫ 20 mm, Nunc), and serum concentration was reduced to 6%.
Proteins-Cultured cell TMs were purified from REF-WT4A cells and separated into two types, high M r TMs (a mixture of TM 1 and TM 2) and low M r TMs (a mixture of TM 4 and TM 5) as previously reported (21). Rat nonmuscle caldesmon was purified from REF-WT4A cells as described in Ref. 15. Smooth muscle caldesmon and TM (a mixture of alpha and beta forms) were from chicken gizzard (31), human fascin was from HeLa cells (9), recombinant human fascin was purified as described (17), actin was from rabbit striated muscle (19), and a cyclin-cdc2 kinase complex was from HeLa mitotic cells as described previously (32). Calmodulin was purchased from Sigma.
Actin Binding Assay-Actin binding of cultured cell TMs, caldesmon, and human fascin was assayed in the following four conditions. (iv) F-actin (final concentration, 12 M) was first incubated for 1 h at room temperature with human fascin (2.2 M) in the same imidazole buffer. Caldesmon, which had been phosphorylated by cdc2 kinase as described previously (32), was then added together with either low M r TMs (5.2 M) or high M r TMs (2.1 M). As a control, caldesmon without phosphorylation was added instead of phosphorylated caldesmon. In all actin binding experiments described above, the samples were further incubated for 1 h and centrifuged at 140,000 ϫ g (Beckman Airfuge, 28 p.s.i.) for 20 min. Both supernatants and pellets were dissolved in an equivalent volume of SDS sample buffer and analyzed by 12.5% SDSpolyacrylamide gel electrophoresis. Gels were stained with Coomassie Brilliant Blue R-250 and scanned with a Joyce Loebl Chromoscan 3 densitometer (Vickers Instrument Inc., Malden, MA) to quantitate the amounts of actin, human fascin, caldesmon, and TMs.
Actin Bundling Assay-Actin bundling activity was measured by low speed cetrifugation assay and fluorescence microscopy. In low speed centrifugation assay, recombinant human fascin (0 -0.89 M) was mixed with gizzard TM (2.3 M), gizzard caldesmon (1.7 M), calmodulin (12 M), and F-actin (7.1 M) in 100 mM KCl, 20 mM dithiothreitol, 20 mM imidazole-HCl buffer (pH 7.0) either in the presence of 1 mM CaCl 2 or 1 mM EGTA. After incubation for 30 min at room temperature, the mixtures were centrifuged at 8,000 ϫ g for 20 min. Both supernatants and pellets were dissolved in an equivalent volume of SDS sample buffer, and the amount of actin was determined by SDS-polyacrylamide gel electorphoresis followed by densitometry. In fluorescence microscopy, F-actin filaments (7.1 M) containing 10% rhodamine-phalloidinlabeled F-actin (33) were mixed with 0.89 M recombinant fascin, 2.3 M gizzard TM, 1.7 M gizzard caldesmon, and 12 M calmodulin in 100 mM KCl, 20 mM dithiothreitol, 20 mM imidazole-HCl buffer (pH 7.0) either in the presence of 1 mM CaCl 2 or 1 mM EGTA. After incubation for 30 min at room temperature, the mixture was sandwiched between a coverslip and slide glass, sealed with petroleum jelly, and observed with a fluorescence microscope (Axioplan; Zeiss, Oberkochen, Germany) equipped with a 100ϫ oil lens (Plan-NEOFLUAR; Zeiss).
Other Procedures-SDS-polyacrylamide gel electrophoresis was performed as described by Blatter et al. (34) using 12.5% polyacrylamide except that the buffer system of Laemmli et al. (35) was used. Protein concentrations were determined by the Bradford method (36). We have chosen these concentrations of TMs and caldesmon because they saturate F-actin (12 M) at these concentrations (21,26). Our preparation of low M r TMs contains higher amounts of TM-4 than TM-5 (21), which explains the low actin binding ability of our low M r TM preparation; Helfman's group has revealed, using recombinant TM isoforms, that TM-4 shows much weaker actin binding than TM-5 (37).

Caldesmon, Together with TMs, Can Dissociate Human Fas
As Fig. 1 shows, caldesmon alone (open squares) or TM alone (open circles, high M r TMs; open triangles, low M r TMs) has little or no effect on actin-fascin binding (compare with a control, ϫ). On the contrary, when both caldesmon and TMs (closed circles, high M r TMs; closed triangles, low M r TMs) were added, fascin-actin binding was greatly abolished. Although 2.2 M human fascin shows saturated binding to actin in the absence of either TMs or caldesmon, the addition of both TMs and caldesmon causes the dissociation of more than 90% of fascin from actin. It should be noted that caldesmon and TMs were added after the formation of fascin-actin complexes, indicating that caldesmon together with TMs can cause the dissociation of fascin from actin.
In the same experimental conditions as described above, we have determined how actin binding of high M r or low M r TMs is affected by fascin. Fig. 2A shows actin binding of high M r TMs in the presence of varying amounts of fascin. In the absence of caldesmon, human fascin slightly inhibits actin binding of high M r TMs by about 20% (open circles), which is consistent with our previous result (18). In the presence of caldesmon, fascin does not inhibit actin binding of high M r TMs at a concentration below 2 M fascin (closed circles). However, above 2 M fascin, actin binding of high M r TMs is decreased to a similar extent as observed in the absence of caldesmon. triangles), which is consistent with our previous result (18). In the presence of caldesmon, however, the inhibition of TM-actin binding by fascin is greatly reduced; at 2.4 M low M r TMs (closed triangles), actin binding of low M r TMs is inhibited by about 30%, and at 5.2 M low M r TMs (closed squares), virtually no inhibition is observed.
We have then examined dependence on caldesmon concentrations of the dissociation of fascin-actin complexes (Fig. 3). F-actin was first saturated with 3.6 M human fascin, and then varying concentrations of caldesmon with a constant, but saturating amount of TMs (open squares, 2.1 M high M r TMs; open triangles, 5.2 M low M r TMs) were added. As Fig. 3 shows, the dissociation of fascin from actin depends on the concentrations of caldesmon added. As the caldesmon concentration is increased from 0 to 3.3 M, the dissociation increases linearly from 0 to 80 -90%. It should be noted that high M r TMs are more effective than low M r TMs, reflecting the fact that high M r TMs bind actin more strongly than low M r TMs. As a control, we examined effects of caldesmon alone on fascin-actin binding and found that caldesmon inhibits actin binding of fascin by less than 20% (open circles). These results suggest that an increase in actin-TM interactions by caldesmon allows TMs to compete with fascin in actin binding.
We have also examined whether fascin affects the actin binding of caldesmon in the same conditions described for Fig.  3. As Fig. 4  Regulation of Fascin-Actin Binding by Ca 2ϩ ⅐Caldesmon⅐ Calmodulin-The binding of caldesmon is regulated by calmodulin in a calcium-dependent way. In the absence of Ca 2ϩ , calmodulin does not bind to caldesmon, and caldesmon binds to actin. On the other hand, in the presence of Ca 2ϩ , calmodulin binds to caldesmon, and a Ca 2ϩ ⅐calmodulin⅐caldesmon complex does not bind to actin (28,29). We have thus examined whether actin binding of fascin can be regulated in a calcium-dependent way by controlling actin binding of caldesmon with Ca 2ϩ /calmodulin. As Fig. 5A shows, when calcium is absent, calmodulin has no effect on actin binding of caldesmon, and thus, caldesmon coupled with high M r TMs inhibits actin-fascin binding almost completely. In the presence of calcium, caldesmon binding to actin is inhibited by calmodulin; the molar ratio of caldesmon to actin is decreased from 0.19 to 0.07. At the same time, the molar ratio of fascin to actin is increased from 0.007 to 0.07. Fascin does not bind to calmodulin, and actin binding of fascin is not regulated by Ca 2ϩ /calmodulin (data not shown). Thus, the regulation of fascin-actin binding is through the regulation of caldesmon-actin binding. Similar calciumdependent regulation of fascin-actin binding is observed when low M r TMs were used instead of high M r TMs (Fig. 5B). The molar ratio of fascin to actin is increased from 0.04 to 0.12 when calcium is added.
Regulation of Fascin-Actin Binding through Phosphorylation of Caldesmon-Phosphorylation of caldesmon by cdc2 kinase causes the dissociation of caldesmon from actin (32,38). We have thus examined effects of caldesmon phosphorylation on actin binding of fascin and found that phosphorylation of caldesmon by cdc2 kinase also controls fascin-actin binding. As Fig. 6 1 and 5).
Regulation of Actin Bundling Activity of Fascin by Caldesmon and TM-The inhibition of fascin-actin binding by caldes-mon coupled with TM indicates that actin bundling activity of fascin is also blocked by caldesmon together with TM. We have confirmed this by two independent assays, i.e. low speed centrifugation, in which only bundles of actin filaments are precipitated, and fluorescent microscopy to directly observe actin bundles. Instead of cultured cell caldesmon and TM, we used smooth muscle caldesmon and TM for this purpose because the properties of smooth muscle caldesmon and those of TM are very similar to those of nonmuscle caldesmon and TM. Fig. 7A and Table I show the effect of TM and caldesmon on actin bundling activity of fascin examined by the low speed centrifugation assay. When 7.1 M actin is incubated with 0.86 M fascin, about 75% of actin is precipitated (lanes 3 and 4). The addition of either caldesmon alone (lanes 5 and 6) or TM alone (lanes 7 and 8) to the fascin-actin mixture does not alter the amount of actin pelleted by low speed centrifugation. On the other hand, when both caldesmon and TM are added (lanes 9 and 10), the amount of pelletable actin is reduced to 20%, a level of which is similar to that found in the control without fascin (lanes 1 and 2). To further confirm the inhibition, we have directly observed the fluorescently labeled actin-fascin bundles (Fig. 8). In the absence of fascin, no actin bundles are observed (Fig. 8A). When fascin is mixed with actin filaments, very thick F-actin bundles are formed (Fig. 8C). Similar bundles are observed when caldesmon alone (Fig. 8D) or TM alone (Fig. 8E) is added to fascin-actin mixtures. On the contrary, when fascin is mixed with F-actin in the presence of both caldesmon and TM, no such bundles but very thin filaments are observed (Fig. 8F). These thin filaments seem to represent bundles consisting of a few actin filaments. These observations are consistent with the results obtained by the low speed centrifugation assay described above.
The fluorescent microscopy method has allowed us to examine how quickly caldesmon coupled with TM can disassemble preformed fascin-actin bundles. The disassembly of fascin-actin bundles occurs rather quickly. Within 3 min, roughly 80% of thick fascin-actin bundles disappear; in 20 min, no such bundles are observed.
Ca 2ϩ /calmodulin can regulate the actin binding activity of fascin when both TM and caldesmon are present. It is thus expected that Ca 2ϩ /calmodulin also controls the actin bundling activity of fascin in the presence of caldesmon and TM. As Fig.  9A and Table I show, fascin precipitates only 16% of actin filaments when caldesmon, TM, and calmodulin are present in the presence of 1 mM EGTA (lanes 1 and 2). This level of actin precipitation is comparable with that in the absence of fascin, indicating that caldesmon coupled with TM and calmodulin inhibits the actin bundling activity in the absence of Ca 2ϩ . Again, very thin filaments were observed by fluorescent microscopy (Fig. 9B), which is similar to those shown in the absence of calmodulin (see Fig. 8F). When EGTA is replaced with CaCl 2 , the amount of precipitated actin increases to 54% (lanes 3 and 4). Fluorescent microscopy has revealed the appearance of thick actin bundles (Fig. 9C). These results confirm that calmodulin together with caldesmon and TM can confer Ca 2ϩ sensitivity to the actin bundling activity of fascin.

DISCUSSION
Regulation of the Actin Bundling of Fascin by Caldesmon and TM-We have revealed that caldesmon coupled with cultured cell TM inhibits actin binding and bundling activities of fascin. Our results suggest a possible mechanism that caldesmon together with TM regulates the assembly and disassembly of actin bundles present in filopodia, membrane ruffles, and stress fibers. In fact, all three proteins are, at least to some  extent, colocalized in cultured cells. Caldesmon is known to be present in stress fibers, as well as membrane ruffles (26,39,40). Fascin is localized in filopodia, membrane ruffles, and stress fibers (1,9,10,14). Although high M r TMs are found only in stress fibers, low M r TMs are reported to be localized in the membrane periphery including membrane ruffles (37,41).
It is worthy of note that the disassembly of fascin-actin bundles by caldesmon and TM occurs very rapidly, within 3 min. Because actin bundling by fascin is also a rapid reaction, these results suggest that caldesmon together with TM could control the formation of filopodia, which is known to be a very dynamic structure. We have also shown that calmodulin together with caldesmon and TM confers Ca 2ϩ sensitivity to the inhibition of fascin's actin bundling activity. This regulation may work in some of Ca 2ϩ -mediated alterations in the actin cytoskeleton. For example, sea urchin eggs show the elongation of fascin-containing microvilli on the surface upon fertilization (7,42). Because fertilization is accompanied by an increase in Ca 2ϩ concentration, it is possible that a calmodulin-caldesmon-TM system is involved in the formation of fascin-actin bundles of microvilli, although caldesmon has not yet been identified in sea urchin eggs.
Competition in Actin Binding between Fascin and TM-Our results indicate that fascin competes with TM and caldesmon for actin binding and that the binding affinity of each protein for actin seems to be a major factor to determine which protein is dissociated from actin. For example, fascin (K a ϭ 5-7 ϫ  (18). These results correlate well with the fact that low M r TMs show the lowest affinity, whereas skeletal muscle TM has the highest among these TM isoforms. Furthermore, the result that caldesmon coupled with TM can dissociate fascin-actin complexes is explained by the fact that caldesmon and TM increases each other's actin binding. The apparent binding constants become one order higher (24 -27).
Recent structural study by Lehman et al. (44) has demonstrated that the binding site of caldesmon is shared with many actin binding proteins. Indeed, two of the major binding sites of gelsolin are shared with the binding sites of caldesmon and TM, which explains our previous results that caldesmon together with TM inhibits severing action of gelsolin (30,45). Perhaps, fascin also shares the binding sites on actin with caldesmon and TM. However, it is worthy of note that high M r TMs or caldesmon can still bind to actin, which is almost saturated with fascin, suggesting that the binding site of fascin on actin is not entirely the same as those of caldesmon and TM. The partial overlap of actin binding sites may explain why the dissociation of fascin-actin bundles is rapid. If the actin binding sites of these proteins completely overlap, then the binding of the second protein to actin will only occur after the dissociation of the first protein from actin. It is also possible that the simultaneous binding of TM and caldesmon may shift the binding sites of TM and caldesmon in such a way that TM and caldesmon can compete better with fascin for actin binding. Indeed, the three-dimensional reconstruction study (44) has suggested that caldesmon seems to induce movement of TM.
Possible Roles of Fascin, Caldesmon, and TMs in Microfilament Organization during Cell Transformation-The functions of fascin seem to contrast with those of caldesmon. There is an increasing body of evidence showing that caldesmon stabilizes microfilaments in nonmuscle cells (28), probably by increasing actin binding of TM (24 -27). In vitro, caldesmon and TM combined inhibit both the actin-severing and actin-capping activities of gelsolin, even in the presence of Ca 2ϩ (30,45). Increased expression of caldesmon in vivo leads to the stabilization of microfilaments (46 -48). Conversely, inhibition of caldesmon via microinjection of specific caldesmon antibody results in the disruption of microfilaments (49). On the other hand, fascin inhibits actin binding of cultured cell TM, when caldesmon is absent. Fascin also inhibits actin binding of caldesmon when TM is absent. Thus, fascin is likely to contribute to destabilization of microfilaments by counteracting the actions of caldesmon and TM.
It should be noted that the expression of caldesmon is downregulated in cell transformation (39,50,51) whereas that of fascin is up-regulated (52,53). In addition, the levels of high M r TMs are decreased in many transformed cells (54 -58). These changes may contribute to microfilament reorganization upon cell transformation. For example, more fascin coupled with less caldesmon and less TM would result in less binding of TM to actin, thereby causing the destabilization of microfilaments. More fascin expression is also likely to induce more bundling of actin filaments, leading to uncoordinated formation of filopodia and membrane ruffles. In fact, we have recently shown that overexpression of fascin induces membrane protrusions including microspikes and lammelipodia (52). Furthermore, many transformed cells show increased levels of calmodulin (59). Thus, it is also possible that an increased level of calmodulin could decrease the binding of caldesmon, again resulting in the decreased stability of microfilaments and increased formation of actin bundles in filopodia and membrane ruffles.