Microtubule-associated Protein 2c Reorganizes Both Microtubules and Microfilaments into Distinct Cytological Structures in an Actin-binding Protein-280–deficient Melanoma Cell Line

All reagents were obtained from Sigma Chemical Co. (St. Louis, MO) unless otherwise specified. Colchicine was prepared as a 10 mM stock solution in ethanol and diluted to 50 μM final concentration in the culture media. Cytochalasin D was prepared as a 1 mM stock solution in ethanol and diluted to 2 μM in the media. Taxol was prepared as a 10 mM stock in ethanol, diluted to 1 μM in the media, and incubated with the cells for 20–30 min.

The melanoma cell lines were cultured in minimal essential media (MEM) (GIBCO BRL, Gaithersburg, MD) supplemented with 10 mM Hepes, 8% newborn calf serum, and 2% FCS. Sf9 cells were obtained from American Type Culture Collection (ATCC No. CRL 1711; Rockville, MD) and were used to propagate wild-type and recombinant viruses. Sf9 cells were grown as a monolayer at 27°C in Grace's medium (GIBCO BRL) supplemented with 10% (vol/vol) FCS.

Cells were plated on a 25-mm glass coverslip and placed in a Leiden incubation chamber (Medical Systems Corp., Greenvale, NY) kept at 37°C and in a 5% CO₂ atmosphere on the microscope stage by means of a temperature controller (model TC102; Medical Systems Corp.). Differential interference contrast (DIC) and phase contrast microscopy were performed on an inverted microscope (model Axiovert 405M; Carl Zeiss, Inc., Thornwood, NY) with variously a 40×, 0.9 NA Plan-neofluor, a 63×, 1.4 NA Planapochromatic, or a 100×, 1.4 NA Planapochromatic lens.

MAP2c, mature MAP2, and tau were purified from baculoviral expression constructs in Sf9 cells. The expression of these MAPs in Sf9 cells was described previously (Leclerc et al., 1996). Briefly, cDNAs corresponding to MAP2c, mature MAP2, or the three-repeat juvenile tau isoform were cloned into the transfer vector pVL1392 and then introduced into the genome of the baculovirus, Autographa californica multiple polyhedrosis virus (AcMNPV) by homologous recombination. The transfer vector was cotransfected with a modified form of AcMNPV, designated AcRP6-SC (Kitts et al., 1990), into Sf9 cells using the Lipofectin procedure (Gibco Laboratories, Grand Island, NY). After 48 h at 27°C, the infection supernatant was purified by a standard plaque-assay for occlusion body negative recombinant viral plaques by limiting dilution (Fung et al., 1988). The recombinant virus was amplified to yield titers of 10⁸ to 10⁹ plaque-forming U/ml. MAP2c, mature MAP2, and tau were purified from lysed infected cells by adjusting the low-speed supernatant to 0.75 M NaCl, boiling for 5 min, and centrifuging at 200,000 g for 30 min. After precipitation with ammonium sulfate, the proteins were collected by low-speed centrifugation, resuspended in 1.2 M KI buffer (1.2 M KI, 1 mM EGTA, 10 mM imidazole, 0.5 mM DTT, pH 7.4), and then collected again by a low-speed centrifugation. After purification, the proteins were dialyzed extensively against injection buffer (120 mM KCl, 10 mM KH₂PO₄, pH 7.2), concentrated to between 1 and 3.5 mg/ml, aliquotted and frozen at −80°C before use. Some mature MAP2 preparations included the protease inhibitor diisopropyl fluorophosphate, and some was obtained commercially from bovine brain (Cytoskeleton, Denver, CO). None of these modifications altered any of the results.

Microinjection was performed on the inverted microscope equipped with a Narishige micromanipulator and pneumatic injector (Narishige USA, Inc., Greenvale, NY). Needles were pulled on a vertical needle puller and beveled as needed on a rotating grinder (both from Narishige USA, Inc.). All injections were performed with the cells kept at 37°C and in a 5% CO₂ atmosphere. Injection volumes were estimated at <20% of total cell volume both by comparing cell images during injection and based on subsequent cell viability, as previous studies suggest that cells do not survive with injection volumes >15% of total cell volume (Cooper et al., 1987). A series of control injections was performed with injection buffer alone in varying volumes for all experiments. Since, with time, M2 cells slowly establish a peripheral gel from actin filament interpenetration alone (Cunningham, 1995), all injections were performed in cells less than 36 h after plating. Each result reported is based on injections of at least 50 cells, unless otherwise noted.

For immunofluorescence, cells were fixed for 15 min at room temperature in 3.7% paraformaldehyde in PBS and 0.12 M sucrose. They were then washed in PBS and permeabilized in either 0.3 or 0.5% Triton X-100 for 5 min. For some experiments, cells were simultaneously fixed and extracted for 10 min in fixative plus 0.1% saponin and 250 nM FITC-phalloidin (Sigma Chemical Co.), and 10 mM EGTA to stabilize actin filaments. After blocking of nonspecific binding sites with 5% BSA in PBS, the cells were incubated first with the specific primary antibody overnight at 4°C, then with secondary antibody at 50 μg/ml of either rhodamine or fluorescein-conjugated goat anti–mouse IgG (Cappel Labs, Durham, NC) or donkey anti–rat IgG (Jackson Laboratories, Bar Harbor, ME) for 2 h at room temperature, and then washed in PBS with 0.2% gelatin in between. For actin staining, fluorescein or rhodamine-labeled phalloidin (Sigma Chemical Co.) was added at 250 ng/ml to the secondary antibody incubation. Antibodies used were AP18 against MAP2c (Binder et al., 1986), and YL1/2 rat monoclonal against tubulin (Accurate Chemical and Scientific, Westbury, NY). The cells were viewed on a microscope (model Axioplan 405M; Carl Zeiss, Inc.) equipped for epifluorescence or an inverted microscope (model Diaphot-TMD; Nikon, Inc., Melville, NY). Images were analyzed with the Image I software program (Universal Imaging Corp., West Chester, PA) and processed in Adobe Photoshop (Adobe Systems Inc., Mountain View, CA) to maximize signal to noise ratios.

The effects of each MAP on the viscosity of a solution of actin filaments were compared by falling ball viscometry (Pollard, 1982), and the viscoelastic properties of actin gels were measured by dynamic oscillatory methods (Janmey, 1991). In the falling ball assay, 0.5 mg/ml actin was mixed with 1:1,500 gelsolin and varying concentrations of the candidate cross-linking protein. After addition of salts to initiate polymerization, the mixture was drawn into a 100 μl capacity glass capillary tube and allowed to polymerize for 2 h. The tubes were then slanted to a 45° angle, and the time required for a 0.025-inch stainless steel ball to travel the length of the tube was measured. The measurements are presented as relative viscosity compared to actin filaments polymerized without cross-linker.

The dynamic shear viscoelastic properties were measured using a torsion pendulum. Measurements were performed in a buffer at pH 7.4 consisting of 2 mM MgCl₂, 150 mM KCl, 0.2 mM CaCl₂, 0.2 mM DTT, 0.5 mM ATP, and 0.3 mM Pi. The storage modulus was calculated from the frequency of free oscillations ranging between 0.1 and 10 Hz for 1.6 mg/ml actin polymerized in the presence of gelsolin (1:200) and ABP-280 (1:133) or MAP2c (1:100) as previously described (Janmey, 1991). Viscoelasticity was also measured in collaboration with Soren Hvidt (Roskilde, Denmark) using a Rheometrics (Piscataway, NJ) RFS instrument operating in a forced oscillation mode, as previously described (Janmey et al., 1988).

Bundle formation was measured by dynamic light scattering using a Brookhaven Instruments BI30AT apparatus (Holtsville, NY) with a 128channel autocorrelator and a 10 mW He-Ne laser emitting 633-nm light. Actin (3 μM) was polymerized for 4 h with various concentrations of MAP2c in siliconized glass tubes (6-mm inner diameter). The total scattering intensity and the autocorrelation function of quasielastically scattered light were determined by standard methods (Hartwig et al., 1992). An increase in polymer mass/length ratio would retard the decay of the autocorrelation function and increase the total scattering intensity if MAP2c caused lateral association of actin filaments. This method has been used previously to document the bundling activity of tensin (Lo et al., 1994) and myristoylated alanine-rich C kinase substrate (Hartwig et al., 1992).

M2 cells lack ABP, and when freshly plated, they appear rounded or polygonal in shape, do not form lamellae, are unable to constrain intracellular organelles from their periphery, and display prolonged, extensive membrane blebbing from the entire cell perimeter (Fig. 1,A). Transfection of the ABP cDNA into this line completely corrects these defects (Fig. 1,B). Within 5 min after injection of 50 μM MAP2c (maximum estimated final intracellular concentration is 5 μM), the M2 cells stopped membrane blebbing and formed a flattened peripheral lamella (Fig. 2). In some cells, several separate lamellae developed, while in other cells the peripheral flattening was circumferential. Consistent with the identity of these peripheral structures as lamellae, intracellular organelles (which average ∼1 μm in diameter in these cells) were excluded.

In most of the cells injected (75%), multiple, long processes (average 40 ± 12 per cell) emerged from the lamellar margin 10–20 min after MAP2c microinjection. By phase contrast microscopy long, rigid, spikelike structures that probably represented microtubule bundles traversed the processes and originated well within the cell interior. In some cells, these processes formed a radial array around the entire cell circumference. Time-lapse videomicroscopy revealed that in all cases process formation began after the elaboration of the lamella. Ruffling of the cell edge was frequently observed, and often preceded process formation (Fig. 3). However, once process formation occurred, no further ruffling was observed in those areas, although some cells displayed separate areas of continued ruffling.

The morphologic effects of MAP2c were concentration dependent. Microinjection of 25 μM MAP2c within the micropipette gave similar results to those above, except that only 50% of the cells developed processes. At 12 μM MAP2c within the micropipette, there was less peripheral clearing, only 30% of the cells had processes, and the number of processes per cell decreased to an average of 15. 6 μM MAP2c did not induce lamellae or processes. Every concentration, including 6 μM, did cause the cessation of blebbing. Therefore, although some of the morphologic effects may require MAP2c in excess of physiological concentrations, exceedingly low concentrations rescue the blebbing phenotype.

Control injections with either injection buffer alone or with purified supernatant from Sf9 cells infected with the wild-type virus did not induce any morphologic changes in the cells. Taxol (1 μM), which stabilizes microtubules and thus promotes their polymerization, also did not lead to process formation or lamellar formation in these cells. Additional controls used the A4, A6, and A7 clonal sublines created from the transfection of the cDNA for ABP-280 into cells from the ABP-deficient line M2. The three lines differ in their expression of ABP-280 with A6 expressing much less than the other two (Cunningham et al., 1992). In contrast to the morphology of the M2 cells, cells from all three transfected lines have a peripheral lamella and increased cortical viscosity by rheological measurements. Consistent with the low ABP-280 in the A6 cells, these cells also have a lower cortical viscosity (Cunningham et al., 1992). Injection of 50 μM MAP2c into A6 cells resulted in process formation in 50% of the injected cells (24 out of 50) and an average of 11 processes per cell. Injection of cells from A4 or A7 cell lines induced processes in only 2 of 102 (<2%) cells. None of the injections caused an enlargement of the lamella. These observations suggested a negative correlation between the ability of MAP2c to form processes and the levels of ABP-280 in cells.

Having observed two distinct cytological structures—lamellae and processes—we sought to determine whether MAP2c was present in both of them. Indirect immunofluorescence with antibodies to MAP2c revealed diffuse labeling that ranged from faint to moderate intensity within the cell body and the lamellae. Focal regions of the cell, such as ruffles (Fig. 4,B) and processes (Fig. 4 E), often contained more intense MAP2c immunoreactivity. It was clear that microinjected MAP2c was not uniformly distributed throughout the cytoplasm.

Some regions of the cell where MAP2c label was observed are known to be highly enriched in actin filaments and have a relative paucity of microtubules. We searched for cellular regions where MAP2c was likely to be present in the absence of microtubules. The membrane ruffle is such a region that is known to lack microtubules. The identification of the ruffle was confirmed by staining with FITC-phalloidin, and colocalization of MAP2c to the same site was observed by double labeling with a MAP2c antibody (Fig. 4, A–C). An association of MAP2c with the cytoskeleton was indicated by the retention of staining after detergent extraction at the time of fixation. The presence of MAP2c in a membrane ruffle strongly points to the presence of this molecule in an actin-rich region known to lack microtubules. Tubulin antibodies demonstrated that microtubules often course through lamellae, but there are sites within lamellae that lack microtubules. Although all lamellae reacted with MAP2c antibodies above background levels, regions of the lamellae that lacked staining with tubulin antibodies remained reactive with MAP2c antibodies (Fig. 4, D–F). This observation suggested that MAP2c was present in regions of the lamellae that lacked tubulin, but contained actin filaments.

In contrast to MAP2c, the microinjection of the mature isoform of MAP2 did not usually induce either a lamella or processes. All concentrations of mature MAP2 up to 50 μM within the micropipette induced the cessation of blebbing in M2 cells. In 30 of 35 cells injected (85%), no processes formed after MAP2 injections; in the remaining cells, 1–3 processes formed and rarely was there a small lamella (Fig. 5 A). These processes were wider and more tapered than those observed in the MAP2c-injected cells, and by phase microscopy, the thickened bundles within their processes did not extend very far into the cell body. A similar inhibition of blebbing was seen with injections of decreasing MAP2 concentrations down to 6 μM, at which point there was no detectable effect on 36 out of 36 injected cells.

Likewise, injections of three repeat tau (50 μM in the micropipette) did not lead to cortical gelation with lamella or process formation in 47 of 60 cells. However, in contrast to vehicle injections and like mature MAP2, tau did result in the complete cessation of blebbing (Fig. 5 B). Two cells elaborated three to five short uniform processes that contained thickened bundles of microtubules that originated within the cell body by phase contrast microscopy. Injections of more dilute tau did not change the cellular morphology.

Evidence that the MAP2c-induced processes were microtubule based was derived from treatment of the cells with colchicine (Fig. 6,B). M2 cells microinjected with MAP2c and then incubated with colchicine at 50 μM final concentration showed a decrease or disappearance of processes in 27 of 30 process-bearing cells (90%), within 20 min. Thus, most of the microtubules within the processes were sensitive to colchicine. However, after colchicine treatment the MAP2c-induced lamella remained nearly intact. In 17 cells, there was a slight decrease (at most 25%) in the width of the peripheral lamellae. To determine whether microtubule polymerization was initially required to form the peripheral lamellae, cells were preincubated with 50 μM colchicine and then injected with MAP2c (Fig. 6 C). Of 26 viable cells remaining after injection, lamellae developed in 18 (70%), but processes did not form in any cell. This result suggested that after MAP2c microinjection, lamellae could form independently of the microtubules.

To assess the cytoskeletal systems involved in supporting MAP2c-induced lamellae, the M2 cells were treated with 3–5 μM cytochalasin D (CD), which inhibits barbedend actin polymerization (Fig. 6 A). 25 μM MAP2c was loaded into the micropipette and injected; 20 min later, when the cells had ceased blebbing and developed a cortical veil, we added 3–5 μM CD to the medium. Of 83 cells so treated, 72 (86%) had a marked decrease in their lamellae (average reduction in width of cleared zone: 80 ± 20%) after 10–20 min of treatment with cytochalasin. Although there was a nearly complete collapse of the lamellae in many cells, the growth of the processes was actually enhanced. In 58 cells (70%), the average number of processes per cell doubled in number from 40 (± 12) to 85 (± 14). Furthermore, the average length of the processes increased from 27 μm (± 5) to 61 μm (± 15) in these cells. A similar experiment in A7 ABP-280–expressing cells also resulted in extensive process formation in nearly all injected cells (data not shown).

We next questioned whether intact microfilaments were required to initiate process formation by treating M2 with 3–5 μM CD for 10 min before they were microinjected with MAP2c. Under these conditions neither lamellae nor processes formed. These results raised the possibility that a focal site of “weakness” or loss of microfilament integrity in the cortical actin skeleton is required to initiate directed microtubule growth, rather than a complete depolymerization of the F-actin.

Microinjection of MAP2c into M2 cells induces a reorganization of F-actin. In control noninjected or vehicle-injected cells, F-actin as detected with rhodamine-labeled phalloidin forms a submembranous rim (Fig. 7,A). Indeed, M2 cells do have a subcortical F-actin skeletal system; but in contrast to A7 cells, which contain ABP-280, M2 cells have a decrease in highly cross-linked actin (Cunningham, 1995). Injection of the M2 cells with MAP2c results in the loss of the cortical F-actin rim and a more diffuse labeling of the cells with rhodamine-phalloidin (Fig. 7 B). Much of the diffuse label corresponds to the broad lamellae induced by MAP2c.

Not surprisingly, MAP2c injection also induces a reorganization of the microtubules in a manner that has been reported previously (Umeyama et al., 1993). For 24 h after plating, M2 cells have few discernible microtubules, and as they spread further they develop a dispersed microtubule network. However, the injection of MAP2c induces large bundles of microtubules in the cell body and processes. Microtubule bundles within processes often appear to originate deep within the cell body. Microtubules sometimes extend in a more dispersed pattern into the lamella. Particularly, in cells that do not form processes, bundled microtubules loop around the perimeter of the cell. When MAP2c was injected into A7 cells, similar microtubule bundles formed, but as noted above, the cells did not form processes.

Because the morphologic changes induced by MAP2c in the M2 cells imply effects on the cortical actin network, we measured the ability of the protein to increase the viscosity of a dilute solution of actin filaments in vitro. Such solutions of short polymers are usually fluid, but if some of the polymers are either increased in length or tied together by a cross-linking protein, the decrease in overall free rotational motion of the filaments results in an increased viscosity of the total solution. This increase is most dramatic in solutions where most of the filaments are effectively immobilized, resulting in a phase transition to a gel-like state (Flory, 1953).

Increases in actin viscosity due to the effects of a candidate cross-linking protein can be quantitated in several ways. One method is to measure the differences in the impedance presented to a small ball moving through the solution, as measured by the amount of time required for the ball to move a preset distance, in a falling ball viscometer (Pollard, 1982). If the length of the actin filaments in a solution of standard actin concentration is kept constant, various candidate cross-linking proteins can then be compared by the amount of each required to achieve a certain viscosity, or to cause gelation. If the affinity of the candidate protein for actin filaments is uniformly high, then the efficiency of each protein in producing gelation depends upon the geometry of the filaments allowed by the crosslinker (Flory, 1953; Almdal et al., 1993). In mammalian cells, the most efficient gelation protein is ABP-280 because of its ability to form near right-angle cross-links between filaments (Janmey et al., 1990).

We measured the ability of each of the MAPs to increase the viscosity of, or to gel, a solution of short actin filaments in a falling ball viscometer. Actin filaments polymerized in the presence of MAP2c reached a gel state at concentrations similar to those at which ABP-280 gels actin (Fig. 8). In contrast, tau and mature MAP2 were unable to gel the same concentration of actin filaments at these or any other molar ratios tested, although they did increase viscosity. Mature MAP2 was superior to tau in increasing the viscosity of actin filaments.

Similar results were obtained by measuring the elastic shear moduli of solutions of actin filaments polymerized in the presence of each protein in a torsion pendulum (Janmey, 1991). The elastic shear modulus is a measure of a solution's resistance to shear stress, and actin gels formed by MAP2c/actin mixtures had a greatly increased elastic shear modulus (Fig. 9). In contrast, tau at equivalent molar ratios to actin resulted in a much lower increase in the elastic shear modulus of an F-actin solution. The increase seen with MAP2c was dose dependent with MAP2c concentration (Fig. 9 A) and was not due to greater filament length, as MAP2c did not affect the kinetics or extent of actin polymerization as judged by pyrene-labeled actin assembly (data not shown).

To determine if the initial increases in elastic shear moduli seen with MAP2c could be due solely to bundling of actin filaments, we then measured the relative light scatter of solutions of actin filaments with and without varying concentrations of MAP2c. At dilute molar ratios of MAP2c to actin, where an increase in the elastic shear modulus was already apparent, there was no increase in relative light scatter. Higher molar ratios of 1:86 increased the light scattering (Fig. 9,A, inset). These results suggest that dilute concentrations of MAP2c caused actin gelation by organizing the actin filaments into an isotropic network; only at higher molar ratios or lower ionic strength (data not shown) do actin bundles form. Because this behavior is similar to ABP-280, we directly compared their effects on actin. The shear moduli of networks composed of short actin filaments cross-linked by either MAP2c or ABP-280 (Fig. 9 B) showed that similar molar ratios of MAP2c and ABP280 (corresponding to a smaller weight ratio for MAP2c) produced actin networks of very similar shear moduli, and both were much greater than those of the corresponding actin solutions without cross-links.

Process emergence involves an exquisitely controlled coordination between two cellular structures: an actin-rich lamella and a microtubule-rich shaft. The consolidation of the lamella into a shaft requires interactions between microtubules and microfilaments. The studies described here support the hypothesis that MAP2c can serve as a mediator between the actin filaments and the microtubules because it can engage in functional interactions with both systems. Other studies have described in vitro interactions between MAPs and actin, and some of these have identified the microtubule-binding domain within the MAP as the approximate binding site (Griffith and Pollard, 1978, 1982; Nishida et al., 1981; Satillaro et al., 1981; Pollard et al., 1984; Asai et al., 1985; Satillaro, 1986; Correas et al., 1990; Dentler and Adams, 1992; Cross et al., 1993; Moraga et al., 1993). All of these studies predate the discovery of MAP2c. A few studies have noted other proteins that may link microtubules and actin filaments, including ezrin (Birgbauer et al., 1991), calcineurin (Ferreira et al., 1993), and p150^glued (Waterman-Storer et al., 1995).

The in vitro data here show that the interaction of MAP2c with actin filaments differs from tau and mature MAP2 in that binding at low concentration results in the formation of an isotropic actin network rather than actin bundles (Fig. 9). The lack of a significant increase in light scatter in the presence of low molar amounts of MAP2c indicates that the large increase in elastic modulus at these concentrations cannot be attributed to lateral associations between actin filaments resulting in bundles, as was previously proposed for MAP2 (Pollard et al., 1984; Satillaro et al., 1981). More likely, the gel formed by MAP2c is due to an isotropic network of filaments branching at high angles (Flory, 1953). The smaller increases in viscosity and elastic shear moduli seen with tau and mature MAP2 are more likely due to their actin-bundling ability. In fact, MAP2c and tau show comparable increases in light scattering (data not shown), which also suggests that at concentrations of MAP2c above those required to induce actin filament isotropy, MAP2c, like tau, bundles actin filaments. MAP2c is unique in its ability to induce an isotropic network of actin filaments at concentrations below those required for mature MAP2 or tau, or even MAP2c itself, to bundle actin filaments.

Studies of chymotryptically digested fragments indicate that the actin-binding domain of MAP2c and MAP2 approximates the tubulin-binding domain (Satillaro, 1986), and a synthetic peptide corresponding to one of the repeated tubulin-binding sites in both MAP2 and tau also binds actin (Correas et al., 1990). Therefore, for actin filament cross-linking to occur, either multiple actin-binding domains exist (Selden and Pollard, 1986) or dimerization of the MAP2c molecule must occur. That MAPs can set microtubule–microtubule spacing (Chen et al., 1992; Leclerc et al., 1996) raises the possibility of dimer formation. Dynamic light scattering assays on pure preparations of MAP2c suggest the existence of oligomerized forms (Janmey, P.A., unpublished observations). Dimerization of tau and MAP2 isoforms has been reported (Wille et al., 1992a, b), and both form antiparallel dimers arranged so that the binding domains would be separated by the overlap length of the dimer (Wille et al., 1992a,b). These overlap lengths would clearly differ between MAP2c and mature MAP2.

In addition to different overlap lengths, other conformational properties of MAP2 versus MAP2c may account for their differential organizational properties in both in vitro and in vivo assays. In neurons as well as in other eukaryotic cells, MAP2 becomes highly phosphorylated. All of the MAPs in these studies were synthesized and purified from Sf9 cells, which appear to phosphorylate MAP substrates similarly to the in vivo pattern. Nevertheless, significant phosphorylation differences may occur between mature MAP2 and MAP2c. A comparison of the primary structures of MAP2 and MAP2c with other known actinbinding proteins reveals little homology. However, a similar lack of homology is found among actin-binding domains in other proteins such as gelsolin and α-actinin (Way et al., 1992). The binding of these proteins to F-actin relies on the formation of specific folds that interact with grooves in the actin filament (McLaughlin and Weeds, 1995). MAP2 and MAP2c may bind via similar mechanisms with differences in the spacing of their charge distributions accounting for the differences in actin gelation ability between the two isoforms.

Like MAP2c, tau and mature MAP2 rescue the blebbing phenotype and induce microtubule bundles in M2 cells; but unlike MAP2c, tau and mature MAP2 fail to induce either lamellae or processes. The M2 cell phenotype arises from a defect in actin cross-linking because of the absence of ABP-280. This defect in the cortical actin cytoskeleton gives rise to membrane blebs because there is decreased impedance to internal solvent flow, resulting in rapid expansions of the plasma membrane that outpace the usually stabilizing local actin polymerization (Cunningham, 1995). Proteins that increase the viscosity of the peripheral actin network decrease this solvent flow and reduce blebbing because of an accelerated formation of a peripheral actin gel. It therefore seems likely that all of the MAPs studied here are competent to increase the viscosity of the peripheral actin network. In contrast to MAP2c, the interaction of tau and mature MAP2 with actin in vitro induces bundle formation rather than the high angle crosslinks of an isotropic network. This difference may account for the failure of tau and mature MAP2 to induce lamellae in M2 cells.

All of the MAPs studied here bundle microtubules when they are expressed in cells, including M2 cells. Why then do tau and mature MAP2 fail to induce processes in the M2 cells? Likewise, the treatment of M2 cells with taxol, which bundles microtubules, does not lead to process formation. Clearly, polymerization and bundling of microtubules alone is insufficient for process formation. To some extent process formation depends on the cell type—all of the tau and MAP2 isoforms are competent to form processes in Sf9 cells—whereas in fibroblasts and M2 cells they induce microtubule bundles, but not processes. What then are the critical components that lead from microtubule polymerization to the elaboration of processes? And why should MAP2c, which has a microtubule-binding domain that is identical to mature MAP2 and highly homologous to tau, be especially competent to initiate process outgrowth in M2 cells? The organization of the subcortical actin cytoskeleton and the specific molecules competent to interact with this filamentous system may represent the critical component for successful process outgrowth and account for the specificity of interacting molecules.

In the M2 cells, MAP2c appears to have the expected property of cross-linking microtubules in the induced processes and the unexpected property of cross-linking actin filaments in ruffles and lamellae. These data suggest that MAP2c can form microfilament–microfilament cross-links and microtubule–microtubule cross-links; our data cannot address the question of whether microtubule–microfilament cross-links also occur. How might this dual cross-linking property serve in process elaboration? Penetration of microtubules into an organized actin framework, such as that of the growth cone, is required for process initiation and growth. Process emergence appears to involve the capture of several microtubules at a discrete site near the membrane, the bundling of these microtubules proximally where they funnel into the process shaft, and the deformation of the cell boundary as an incipient or elongating process.

One of the early requirements for process formation is the establishment of a microtubule orientation which directs elongation orthogonal to the cell perimeter. Actin filaments may contribute to establishing the correct microtubule orientation. When the margin of a cell consists of actin bundles, rather than an isotropic actin organization characteristic of a lamella, elongating microtubules often have a tendency to loop around the cell perimeter without penetrating the bundles or forming a process. This pattern is observed in the marginal band of dogfish red cells (Sanchez and Cohen, 1994) and sometimes in cells expressing MAP2c (Edson et al., 1993; Leclerc et al., 1996). In contrast, microtubules involved in process formation do not show this radial organization; their direction of elongation is orthogonal to the cell perimeter. Some degree of actin organization may be required to engage sufficient numbers of microtubule plus ends at a single site near the membrane to initiate a process. Support for this step in process formation is the observation here that M2 cells treated with cytochalasin before MAP2c injection failed to elaborate processes. F-actin may capture microtubules at a discrete and possibly specific site on the subcortical membrane, and once assuming a radial direction of elongation, efficient microtubule growth proceeds more efficiently with less resistance from an organized actin barrier. Evidence for the concomitant focal organization of actin at the initiation site of a process comes from observations of processes induced by tau expression in Sf9 cells (Knowles et al., 1994).

Once microtubule growth is directed approximately orthogonal to the membrane, the actin polymerization state appears capable of controlling microtubule elongation (Lin and Forscher, 1993). In the A4 and A7 cells there is sufficient ABP-280 to form actin cross-links and prevent MAP2c-induced processes. The lack of ABP-280 in the M2 cells and concomitant decrease in actin cross-linking may allow access for MAP2c–actin interactions. Once initiated, microtubule elongation is most efficiently translated into process elongation when there is minimal actin resistance. In this manner the cortical actin cytoskeleton serves not only as a barrier over most of the cell perimeter but also as a gate at sites of process emergence. Biochemical and mutagenesis studies with particular attention to the phosphorylation states of MAP2c will be required to determine whether this molecule can serve as a switch to mediate functionally distinct interactions between microfilament populations and microtubule populations leading to coordinate regulation of these two cytoskeletal systems.

Extracellular signals that affect local actin organization may also determine the specific site and orientation for microtubule growth. Regulated interactions between actin filaments and MAP2c may determine whether MAP2c will engage the cortical actin cytoskeleton in a manner that allows lamellar formation and penetration of microtubules through the actin meshwork. As observed in Aplysia growth cone, contacts induce a focal F-actin assembly that, in turn, guides microtubule redistribution to a target site (Lin and Forscher, 1993). Actin organization is responsive to extracellular signals via such intracellular messengers as phosphatidylinositol 4,5-bisphosphate and the small G proteins, rac and rho; however, the mechanism of signal transduction to microtubules is unknown. Possible intermediates are molecules found at the ends of microtubules such as the adenomatous polyposis coli gene product (Nathke et al., 1996), on microtubules such as mitogen-activated protein kinase (Morishima and Kosik, 1996), or calmodulin, which can affect the interaction of MAP2 and tau with actin (Kotani et al., 1985). The very complex array of MAP phosphorylation states, known to have an effect on microtubule dynamics (Drechsel et al., 1992) may also serve to regulate interactions with actin.

We thank Elizabeth Ackerman, Maura Schutt, and Cailin Monahan for technical assistance, Dr. Thomas Stossel for a critical reading of the manuscript, and Dr. Craig Garner for the generous gift of rat MAP2c cDNA.

This work was supported by National Institutes of Health grants CA57479, DK38452, and HL48743 (C.C. Cunningham), AR38910 (P.A. Janmey), and NS29031 and AG06601 (Kenneth S. Kosik). N. Leclerc acknowledges a Canadian Medical Research Council Fellowship.

ABP

actin-binding protein

CD

cytochalasin D

DIC

differential interference contrast

MAP

microtubuleassociated protein

Sf9

Spodoptera frugiperda