Localization of myosin II A and B isoforms in cultured neurons

The neuronal growth cone is a highly dynamic structure that governs axon assembly and pathway selection during development and following nervous system trauma. Although the interactions between the growth cone and the extracellular environment are complex, a crucial and instructive consequence of this interaction has been discerned: tension develops in axons resulting from growth cone exploration, and this tension regulates both axon formation rate and the direction of outgrowth (pathway selection) (Heidemann et al., 1991; Lamoureux et al., 1989). We are studying the molecular mechanisms underlying tension production by growth cones.

Actin and myosins are abundant in growth cones (Lewis and Bridgman, 1992; Miller et al., 1992; Espreafico et al., 1992; Ruppert et al., 1993) and their interaction is capable of producing tension in vitro. There are at least eleven classes within the myosin superfamily (Sellers and Goodson, 1995) and three are known to be represented in vertebrate brain, myosins I, II, and V. Myosin I has been demonstrated to move actin filaments along a phospholipid substratum (Zot et al., 1992), has been localized to Golgi vesicles in intestinal epithelial cells (Fath and Burgess, 1993), and tends to be localized near the plasma membrane in growth cones (Lewis and P. C. Bridgman, unpublished observation). These data suggest involvement of myosins I in vesicle transport and linkage of actin to the plasma membrane. Myosin V has a lipid binding domain (Cheney et al., 1993) and a primitive myosin V, myo2, is essential for vesicle transport in Saccharomyces (Johnston et al., 1991). Furthermore, in the dilute (myosin V) mutants that have the most severe phenotype i.e. death following seizures beginning 3 weeks after birth, brain development appears to be normal (Strobel et al., 1990), arguing against a critical role for this myosin in gross aspects of growth cone navigation. Thus, myosin V, like myosin I, is a good candidate for a vesicle transporter. Whether these two myosins also contribute to tension production remains to be determined.

Myosin II is the conventional, two-headed myosin that forms filaments via interactions between the coiledcoil tail domains of different molecules (Warrick and Spudich, 1987). Structural differences between brain myosin II and other types of nonmuscle myosin were initially proposed by Burridge and Bray (1975). Recently, three isotypes of myosin II that differ in the primary sequences of their heavy chains have been found in vertebrate brain (Murakami and Elzinga, 1992; Sun and Chantler, 1991). Two isoforms that are prominently expressed in mammalian brain are designated myosin heavy chain A and B (MHC-A and MHC-B) (Simons et al., 1991). Alternative splicing of the MHC-B message gives rise to three additional isotypes (Itoh and Adelstein, 1995). Prior to the discovery of these isoforms, Bridgman and Dailey (1989) examined the distribution of myosin II in rat superior cervical ganglion (SCG) neurons. However, this work was carried out with an antibody prepared against platelet myosin, the heavy chain of which is exclusively MHC-A (Murakami et al., 1991). We subsequently learned that this antibody will recognize MHC-B in homogenates of neural tissue and also on western blots, albeit with a lower apparent affinity than that for MHC-A (see Miller et al., 1992). We undertook the current study to determine the distribution of each myosin II isoform individually. During the course of our study two groups reported the localization of myosin II isoforms in growth cones (Cheng et al., 1992; Miller et al., 1992). Both groups emphasized the localization of myosin II to the leading edge of the growth cone periphery, and Cheng et al. proposed a role for myosin IIB, which appears to be the predominant neuronal isoform, in leading edge extension. The distribution we had observed differed from theirs and did not suggest a role in protrusion. To investigate this further, we fixed growth cones during live observation, permitting us to identify sites that were undergoing protrusion or retraction at the time of fixation, and localized myosin IIB. Our observations support a role for myosin IIB in retraction of the growth cone periphery, but not in protrusion. The distributions of both myosin II A and B in growth cones are consistent with roles for these myosins in tension production by growth cones.

Initially two peptides, GKADGAEAKPAE and SDVNETQPPQSE, were synthesized based on the derived amino acid sequences at the carboxylterminal end of human macrophage MHC-A (Saez et al., 1990) and T-cell MHC-B (Phillips et al., 1995), respectively. The peptides were conjugated to keyhole limpet hemocyanin (Calbiochem) with glutaraldehyde (Sigma grade II) and rabbits were immunized and bled by Berkeley Antibody Company (CA). Both antibodies were purified on peptide antigen columns prepared by coupling 20 mg of the appropriate peptide to 25 ml of Affigel 15 (Bio-Rad). Antibodies were then further purified on a Protein A column (Pierce). The final concentration of both antibodies was between 0.7 and 1.0 mg/ml. During the course of this study, we learned that the corresponding fragment for the rat MHC-B is identical to that in the human, but the rat MHC-A differs from the human by 3 amino acids (K.I., unpublished observation). We therefore prepared an antibody against the carboxy terminus of rat MHC-A, GKADGADAKATE. The species specific anti-MHC-A yielded immunoblots and immunofluorescence that was indistinguishable from that of the anti-human MHC-A. The anti-MHC-A used is indicated in each figure legend.

The dye Cy3 was directly conjugated to affinity purified anti-MHC-B using a kit available from Biological Detection Systems. A molar ratio of 5 (Cy3:anti-MHC-B) was obtained.

SCG cells were plated onto laminin coated plastic culture dishes as previously described (Bridgman and Dailey, 1989). After overnight growth, the dishes were washed with 37°C L15 medium, and then transferred to a cold surface, whereupon the L15 was immediately exchanged for icecooled 11% TCA. The precipitated cellular materials were scraped off of the dish, centrifuged, washed twice with 1 ml of cold acetone, dried and prepared for SDS-PAGE using conventional methods. We used a low percentage of acrylamide (5.5%) so as to increase the completeness of the electric transfer to nitrocellulose. Following transfer, the membrane was stained with Ponceau reagent (Sigma) and the borders of the lanes were marked. Lanes were cut into strips and labelled for MHC immunoreactivity with the following dilutions of affinity purified antibodies: MHC-A, 1:1,000; MHC-B, 1:2,000. Antibodies were detected using the Enhanced Chemiluminescent Method (Amersham).

SCG explants were cultured on coverglasses overnight as described previously (Lewis and Bridgman, 1992), and mounted onto a chamber designed to permit perfusion during DIC observation (Berg and Block, 1984). Advancing growth cones were imaged using a Newvicon camera and time lapse video enhancement (averaging 4 background subtracted frames every 5 seconds). Glutaraldehyde fixative (37°C, see Immunofluorescence, below) was perfused through the chamber, usually during protrusion of a portion of the leading edge. During perfusion, fine focus became difficult, but it was nonetheless possible to observe the fixation process by continuous imaging. Within seconds of initiating perfusion, the growth cones developed a flattened, coarser appearance and regions near the base of growth cone that had appeared smooth developed circumscribed regions with a reverse shadow cast morphology. No further changes in their appearance occurred, from which we inferred that fixation had been largely completed in the initial seconds. These cultures were processed as described below for immunofluorescence observation of anti-MHC-B and β-actin staining. To enable overlapping of the immunofluorescence images (which were obtained with a CCD camera) and the DIC images, the image distortion due to the Newvicon camera was digitally compensated. To ascertain whether a given region of the growth cone perimeter was undergoing protrusion or retraction, changes in the position of the edge were determined over the 2-frame (10 second) period preceding the fixation.

Cultured SCG cells were fixed and permeabilized by one of two methods.

Cultures were perfused with 37°C fixative (0.25% glutaraldehyde in 100 mM cacodylate (pH 7.4) with 5 mM CaCl₂ and 10 mM MgCl₂). After 10 minutes, cultures were washed 3× with fixative buffer without aldehyde. For phalloidin labelling, cultures were then simultaneously permeabilized and labelled with 100 μl of PBS containing 0.02% saponin and 133 nM rhodaminephalloidin (Sigma). (When labelling of actin was carried out with anti-β-actin, phalloidin was omitted from the permeabilization buffer, and coverslips were prepared as described in the next paragraph.) After 10 minutes, cultures were washed 3× in PBS and mounted for fluorescence photomicroscopy in Vectashield mounting medium (Vector Laboratories). One corner of the coverglass was marked with a permanent marker so that the orientation could be determined following the antibody staining. After photographing about 30 growth cones per coverglass, the coverglasses were prepared for antibody staining as described below.

To quench unreacted glutaraldehyde moieties, decrease autofluorescence, and eliminate the phalloidin staining, cultures were treated with 1% OsO₄ in PBS for 5 minutes at 4°C, washed exhaustively (7×) in PBS at room temperature, then treated for 30 minutes with 5% β-mercaptoethanol (Pierce) in PBS (prepared within 20 seconds of use), and washed as before. Cultures were then blocked with PBS containing 8 mg/ml BSA, 0.5% fish gelatin (v/v), and 5% normal goat serum (v/v). Antibodies were diluted into PBS containing 20% blocking buffer. The following primary antibody concentrations were used: anti-rat MHC-A, 1:500; anti-human MHC-A, 1:7,200; anti-MHC-B, 1:3,150; anti-Cy3-coupled MHC-B, 1:500; monoclonal anti-β-actin, 1:1,800 (Sigma), fluorescein goat anti-mouse 1:400; Cy3 goat anti-rabbit, 1:800; fluorescein goat anti-rabbit monovalent Fab, 1:140; Cy5 goat anti-mouse 1:200. All secondary antibodies were obtained from Jackson Research except fluorescein goat anti-mouse (Molecular Probes). For phalloidin stained coverslips, growth cones previously photographed were relocated and rephotographed.

Cultures were rapid frozen by immersion in liquid nitrogen-cooled 50% ethane, 50% propane, and then stored in liquid nitrogen. Substitution and fixation were performed by transferring coverslips into a container of 0.5% paraformaldehyde in methanol cooled to −80°C, followed by warming to −20°C. Coverslips were then transferred to phosphate buffered saline (PBS) at 22°C, blocked and labelled as described above, except that higher concentrations of anti-MHC could be used (anti-rat MHC-A, 1:100; anti-human MHC-A, 1:400; anti-MHC-B, 1:175).

For labelling of both MHC isoforms in the same growth cones, coverslips were labelled first with rabbit anti-MHC-A (and in some cases mouse anti-β-actin), next with a monovalent fluorescein anti-rabbit (and Cy5 goat anti-mouse), and finally with Cy3-conjugated rabbit anti-MHC-B.

Fluorescence images were obtained with a cooled charge-coupled device (CCD). Digitized images were particularly useful for aligning the triple-stained images and brightness/contrast adjustment. To align triple-stained images of growth cones, the actin staining was used as the reference, and the myosin images were individually positioned so as to align edge foci with the F-actin bundles at the edge of the growth cone. For the triple-stained non-neuronal cell images, precise colocalization of the MHC isoforms was evident, and used to position the MHC images, and the actin staining was positioned next. In addition, these images revealed that the fluorescence associated with the rhodamine-phalloidin obtained from Sigma was not completely eliminated by the oxidation/reduction treatment. The Cy3 anti-MHC-B labelling appeared to be elevated in the periphery of phalloidin stained growth cones, but not in the periphery of β-actin stained growth cones. To subtract this staining, the elevation in peripheral Cy3 anti-MHC-B staining was estimated and the grayscale values of the original phalloidin image were factored down so as to be approx-imately equal to the estimated elevation in Cy3 MHC-B staining. The factored down grayscale values of the phalloidin image were then subtracted from the Cy3 anti-MHC-B image.

Antibodies prepared against isoform specific regions in the carboxy terminus of MHC-A and MHC-B recognize bands on western blots that comigrate with skeletal muscle myosin (Fig. 1). The band identified as MHC-A migrates slightly more rapidly than MHC-B, and these bands separate if the duration of SDS-PAGE is extended (not shown; see Kawamoto and Adelstein, 1991). The antibodies do not cross-react at all on western blots (Maupin et al., 1994; Phillips et al., 1995). The distributions of the two major types of myosin II were examined by immunofluorescence light microscopy. MHC-B labelling was bright in both neurons and non-neuronal rat superior cervical ganglion (SCG) cells, whereas MHC-A staining was bright in nonneuronal cells, but much dimmer in neurons.

By directly conjugating Cy3 to anti-MHC-B, we were able to colocalize both myosin II isoforms in the same growth cones and non-neuronal cells (see Materials and Methods). In growth cones, MHC-B staining, though less intense than when a secondary antibody was used, could easily be detected through the eyepieces of our fluorescence microscope, but MHC-A was barely visible. The staining pattern of MHC-A could not be determined without the use of a sensitive CCD camera and image processing software to enhance brightness and contrast. The brightest MHC-A foci were approximately of the same intensity as the dim MHC-B foci detected in CCD images of the growth cone periphery. To compare the distribution of MHC-A and MHC-B staining (Fig. 2) the peak brightnesses of the two MHC images were normalized, i.e. the MHC-A staining intensity was amplified relative to that of MHC-B. Two methods of fixation were used for the immunostaining. Glutaraldehyde fixation enabled excellent labelling of actin, but the MHC staining usually appeared to have a higher background than growth cones fixed by freeze substitution. Although all but the most intense actin staining is compromised by freeze substitution, the myosin staining appears brighter and areas of increased fluorescence path length are not sites of increased diffuse staining. We will refer to four regions of wellspread growth cones: the peripheral domain (P-domain) is composed of the F-actin-rich lamellipodia, the central domain (C-domain) refers to the more proximal, thickened, organellerich region (see Bridgman and Dailey, 1989), the marginal zone refers to a transition region between the P-andCdomains, and the base of the growth cone refers to the proximalmost portion of the growth cone that has not completed differentiation into a segment of neurite.

The pattern of staining of MHC-A depended more strongly on the fixation conditions employed than did that of MHC-B. In glutaraldehyde fixed growth cones MHC-A staining (Fig. 2A) is brightest in the C-domain. In contrast, the brightness of the staining in freeze substituted growth cones was even throughout the C-domain and marginal zone (Fig. 2E). It is possible that the elevation of C-domain labelling in the glutaraldehyde fixed cells (which was also evident for MHC-B staining, compare Fig. 2C with Fig. 2F) represents non-specific labelling. Staining foci of the two isoforms overlapped frequently in growth cones prepared by both methods (Fig. 2D,G, orange), but not exclusively (Fig. 2A,C,D, arrow; E-G, open arrowhead). Perhaps because of the higher background associated with the glutaraldehyde fixed growth cones, colocalization of MHC-A and -B foci in the C-domain is more easily observed in the freezesubstituted growth cones (Fig. 2E-G, arrow). The brightest staining of the B isoform was associated with F-actin bundles that also had MHC-A staining foci (Fig. 2, filled arrowheads). The MHC-B staining appeared to have a slightly more peripheral localization than the MHC-A staining, owing to the elevation of the MHC-B staining in the marginal zone. MHC-B was typically elevated at sites that appeared to have undergone substantial retraction (Fig. 2C,D,F, asterisks). MHC-A labelling, while present at these sites, was not always elevated (Fig. 2E, asterisks). In neurites, MHC-A staining was often uniform, lacking the rimlike quality often observed for MHC-B staining (see below). In summary, while MHC-A and-B were both concentrated in the C-domain, MHC-A had a more uniform distribution and lacked the marginal zone elevation characteristic of MHC-B.

Confocal microscopy was used to examine the distribution of MHC staining in neuronal cell bodies fixed by freeze substitution. MHC-A staining (Fig. 3A) was again dimmer than that of MHC-B (Fig. 3B). The lower level of diffuse staining within the cell body (vs MHC-B), made it easier to discern the pattern of cortical MHC-A staining (Fig. 3A, ventral surface). MHC-A staining on the dorsal surface was difficult to detect (not shown). Lack of dorsal cortical staining with anti-MHC-A was also observed by Maupin et al. (1994) in fibroblasts, but the dorsal surface of non-neuronal cells in our SCG cultures was intensely stained for MHC-A. The dorsal (not shown) and ventral (Fig. 3B) surfaces of the cell bodies had linear arrays of MHC-B staining (arrowhead), suggesting that cortical myosin II was organized into filaments. The arrays were thicker and longer on the ventral surface than on the dorsal surface. It was difficult to determine whether MHC-B staining was organized into filaments within the cytoplasm due to the high level of diffuse staining. The distribution of actin matched that of the MHC-A and -B in the cell bodies (not shown).

Unlike growth cones, non-neuronal SCG cells stained brightly for both MHC-A and MHC-B (Fig. 4A,C, respectively). Following freeze substitution fixation, the large bundles of F-actin (e.g. stress fibers) present in non-neuronal cells can be discerned with anti-β-actin staining (B). The non-neuronal cell shown in Fig. 6 was at the front of the advancing halo of neurites emanating from an SCG explant. The MHC-A staining pattern of the non-neuronal cells (A) could easily be discerned through the eyepieces, but not that of the two small growth cones that had grown over the non-neuronal cell (open arrowheads). The staining patterns of MHC-A and -B within the non-neuronal cells were similar but not identical. In some non-neuronal cells, MHC-A appeared to be distributed closer to the leading edge than MHC-B (A,B,D, small triple arrow), and MHC-B appeared more concentrated than MHC-A toward the nucleus (A,B,D, large triple arrow). This was observed previously for human melanoma cells (Maupin et al., 1994). Given the striking difference in the ratio of MHC-A:MHC-B staining between growth cones and non-neuronal cells, we investigated the relationship between MHC-B staining, F-actin, and growth cone dynamics more closely.

MHC-B colocalized extensively with F-actin, in particular F-actin bundles. The actin staining looked similar regardless of whether we used rhodamine-phalloidin (Fig. 5A,B, rendered in green) or an antibody that recognized β-actin (Fig. 6I). Long, radially disposed F-actin bundles typical of forming lamel-lipodia within the P-domain are often dotted with dim MHC-B foci (Figs 5A, 6J, open arrowheads). MHC-B was almost always found along the subset of these bundles that extend into the marginal zone. Bundles that terminate at widely separated areas of the leading edge appear to converge and intermingle in the marginal zone, consistent with the possibility that myosin II crosslinks these bundles (Figs 5A,B, 6J). The C-domain often had bright MHC-B staining that colocalized with F-actin bundles, but at a lower density than in the marginal zone (Figs 5A, 6G,I,J). The C-domain was also associated with an increase in the diffuse MHC-B staining that could not be correlated with F-actin staining. The edges of the C-domain (the sides and trailing edges of the growth cone), were sites of intense MHC-B staining that colocalized with thick F-actin bundles (Figs 5A, 6G,I,J, short filled arrows).

Although neurites were brightly labelled for MHC-B (Figs 5A, 6G-J, small open arrows), colocalization of MHC-B and F-actin bundles was difficult to establish in neurites. The labelling often appeared to be brightest along one or both edges of the neurite, suggesting a concentration of MHC-B in the cortex. However, in neurites that had varicosities along their length, the staining was brightest at the varicosities, suggesting that MHC-B is also present in the cytosol.

To determine the relationship between growth cone dynamics and the distribution of MHC-B, we used DIC microscopy to observe advancing growth cones (Fig. 6A-E), fixed them during their advance (Fig. 6F), and stained them for MHC-B and β-actin (Fig. 6G,I). The following associations emerged from our live observations (Table 1). During growth cone advance, lamellipodia and filopodia undergoing protrusion have very little MHC-B (Fig. 6G,H, long arrow). If a leading edge lamellipodium was retracted to the marginal zone, MHC-B was elevated at this leading edge (n=4), but the leading edges of partially retracted lamellipodia were not sites of intense MHC-B staining (n=3, including Fig. 6G,H, o).

The sides of the growth cone are sites at which protrusion and retraction occur in adjacent areas, often perpendicular to the direction of advance. If retraction had occurred just prior to fixation, MHC-B staining was found close to the side edge (e.g. Fig. 6G,H, asterisk) but if the growth cone was spreading laterally, the associated protrusion was only weakly stained, as observed for leading edge lamellipodia.

As neurite formation occurs, the trailing edge of the growth cone moves forward and inward, and trailing edge filopodia often appear to be dragged along prior to being retracted. The trailing edges of the growth cone were typically a site of intense MHC-B staining and trailing filopodia were labelled more intensely than leading edge filopodia (Fig. 6H, compare the staining of filopodia between the asterisk and the short, filled arrow with that of leading edge filopodia). Due to the V-shape of the rear edge of the growth cone in Fig. 6, it is difficult to categorize it as the trailing edge vs the side. In Fig. 5A, a wider growth cone illustrates the trailing edge staining (double arrow). Occasionally filopodia on the side or base of the growth cone appeared to be anchored to other neurites or debris on the substratum (Fig. 6G,H, convex arrowhead). These structures appeared to be under tension because they were straight along their length, and, in some cases, deflected neurites. Such tensile elements were more brightly stained for MHC-B than average filopodia.

In almost all cases, at least one edge of the base of the growth cone had elevated MHC-B staining, and in 40% of the cases elevated MHC-B staining was observed on both sides (Table 1; Fig. 6G,H, short arrows). We re-examined the DIC videotapes to determine if there were any motile or morphological correlates to the variability in the edge staining observed at the base. At the time of fixation two phenomena were observed at the base of the growth cone: retraction of flattened lamellipodia towards the forming neurite, and recruitment of already thickened cytoplasm to form the cylindrical shape characteristic of the neurite. In six of eight cases in which onesided staining was observed, that staining was located on a side that had just undergone lamellipodial retraction, and the non-stained side had undergone further centripetal recruitment of already thickened cytoplasm. However, bases that had staining on both sides were undergoing recruitment of thickened cytoplasm. Together, these data on the relationship between MHC-B staining and changes in the perimeter of the growth cone establish a strong correlation between myosin II B and sites of lamellar retraction, and a negative correlation with sites of lamellar extension.

As described above, the most intense MHC-B labelling occurred away from the growth cone perimeter, in the marginal zone (Fig. 6A, dotted boundary). Although the advance of the marginal zone is difficult to monitor, our DIC observations indicated that this is the site at which the constitutive retro-grade flow (Forscher and Smith, 1988) appears to cease. The concentration of actomyosin IIB in this area (Fig. 6G,I,J, filled arrowhead) suggests that it is the focal point of myosin IIB activity.

This paper reports that myosin II, in particular MHC-B, is abundant in neuronal cell bodies and neurites, and has a striking distribution in growth cones. The distributions of the major isoforms of myosin II have been published previously by other workers, but our results are different from theirs, provide a more extensive analysis of the association with actin (Cheng et al., 1992; Miller et al., 1992), show the staining patterns of the two isoforms in the same growth cone, and show the relationship of this distribution to growth cone motility. We report that MHC-B is concentrated along F-actin bundles throughout the organellerich central domain (C-domain) in growth cones, and is especially bright in the margin between this domain and the actinrich peripheral domain (P-domain). Cheng et al. (1992) reported that MHC-B is most concentrated in the periphery of the growth cone. We also found MHC-B staining foci in this region, but they were present at a much lower concentration than in the C-domain and marginal region. The myosin distribution that we observed also differed from that reported by Miller et al. (1992), who found that myosin II was concentrated throughout the P- and C-domains, excluding filopodia. Colocalization of myosin II with F-actin bundles could not be discerned in the above studies. This colocalization is clear in our high resolution two-andthree-color images. In addition, staining for the two myosin isoforms in the same growth cone established that they partially colocalize along F-actin bundles in growth cones. Finally, our observations of growth cone behavior just prior to fixation revealed a positive correlation between sites of lamellipodial retraction and the concentration of myosin.

The differences between our findings and the previously reported findings could result from the use of different populations of rat peripheral neurons (the other groups used DRG cells, whereas we used SCG cells) or from differences in the fixation methods. The other investigators prefixed their cells in −20°C methanol whereas we used either a rapid freeze method (−190°C) followed by freeze substitution, or perfusion with warm glutaraldehyde-cacodylate buffer. In our experience with SCG neurons, the fine structure of the growth cone is better preserved by our methods.

What do our data on the distribution of myosin II in advancing growth cones suggest about the dynamics of myosin II organization during growth cone advance? The motile events occurring in the marginal zone, where myosin IIB is most concentrated, are difficult to monitor. The presence of a low level of staining distal to this site, and a higher level proximal to this site, is consistent with rapid assembly of actomyosin bundles at the leading edge of the marginal zone, perhaps in concert with retrograde flow (or retraction) of F-actin rich structures, followed by disassembly/disruption as the C-domain advances (see McKenna et al., 1989). Recent work has shown that myosin filaments contain fewer myosin molecules in the periphery than in the central domain (Verkhovsky et al., 1995) consistent with the possibility that myosin filaments grow during centripetal transport. The sharp increase in the size and intensity of MHC foci at the marginal zone suggests that myosin filament assembly is differentially regulated at this site. Since the advancing trailing edge is also more intensely labelled than the C-domain, actomyosin dynamics at this site may undergo a similar assembly/disassembly process. In contrast to the marginal zone, both assembly and disassembly of actomyosin would occur in front of (central to) the direction of the retraction of the trailing edge.

The localization of the major myosin isoform in neurons afforded by our methods suggests several hypotheses on the role of conventional myosin in effecting growth cone shape changes. The DIC and fluorescence observations indicate a strong correlation between intense myosin II B staining foci and sites of lamellar retraction/retrograde flow (Table 1). This is consistent with the possibility that myosin II filament assembly assists in crosslinking actin filaments as they are transported centrally from the periphery. In the marginal zone, such crosslinking may result in connecting widely separated regions of the growth cone, perhaps facilitating the integration of actinbased force production throughout the growth cone. Cross-linking may also be an important step in the process that leads to the apparent cessation of the retrograde flow at the marginal zone. In addition to this passive (i.e. non-motor-dependent) assembly of actomyosin II bundles, myosin II may contract, and thereby oppose protrusion or destabilize existing protrusions. This possibility is suggested by experiments in which myosin function was disrupted (De Lozanne and Spudich, 1987; Honer et al., 1988; Knecht and Loomis, 1987). The abundance of myosin II in the marginal zone and the C-domain may help restrict protrusion away from these regions, confining it to the forward leading edge (see Lee et al., 1994). Marginal zone actomyosin bundles may also be involved in changing the direction of growth cone advance by contracting against, and thereby destabilizing, lateral lamellipodia. Indirect evidence consistent with actomyosin bundles exerting tension along their axes was provided by the observation that filopodia that appear to be deflecting an encountered neurite have more actomyosin staining than average filopodia. Alternatively, if the bundles do not contract, they could serve a purely structural role, enabling an alternative motor mechanism to pull on a large strip of the growth cone cytoskeleton by pulling on a portion of the bundle.

How might myosin II, being concentrated in the C-domain, influence growth cone advance? The C-domain undergoes two processes in which myosin II is likely to be involved: advancement and differentiation into an axon. These processes are probably not identical, since elastic tension develops in axons, apparently as a result of axon formation lagging behind advance of the growth cone C-domain (Heidemann et al., 1991). If the cytoskeleton is more stably substratumlinked toward the marginal zone than in the C-domain, which is consistent with the higher apparent density of actomyosin bundles in the marginal zone, myosin II-mediated contraction would pull the C-domain forward (see Lee et al., 1994). A role for conventional myosin in overcoming substratum adhesion is implied by the work of Jay et al. (1995), and a role for conventional myosin in trailing edge retraction has been suggested for fibroblasts (Small, 1989). Myosin II is at present the best candidate for this retraction-based advance of the trailing edge of the growth cone. Myosin II is also often elevated along the rear edge of the growth cone, and the cortex of the base of the growth cone, consistent with a role in constricting the base of the growth cone into a cylinder during axon formation. Such a role is predicted by the dependence of cleavage furrow constriction on conventional myosin function (Mabuchi and Okuno, 1977).

MHC-A colocalized with F-actin bundles and appeared to be present at a lower level than MHC-B. In chick brain (Kawamoto and Adelstein, 1991) and human brain (Itoh and Adelstein, 1995), both protein and mRNA levels are higher for MHC-B than MHC-A (approximately 4:1). The distribution of MHC-A was slightly different from that of MHC-B. Whereas MHC-B was clearly concentrated in the marginal zone, MHC-A was evenly distributed throughout the C-domain and marginal zone. In contrast, non-neuronal cells derived from the SCG occasionally had MHC-A staining that was closer to the peripheral edge than MHC-B. The apparently low levels of MHC-A suggest that myosin II-B generates most of the myosin II-mediated force in SCG neurons, but we cannot presently eliminate a crucial role for myosin II-A in this process (see Waterston, 1989). An antibody (Biomedical Technologies, Inc.) prepared against platelet myosin II (MHC-A) was used in our previous work (Bridgman and Dailey, 1989), and also that of Cheng et al. (1992) and Miller et al. (1992). Cheng et al. (1992) found no cross-reactivity of the anti-platelet antibody with MHC-B, but this claim conflicts with both our unpublished data, and the immunoprecipitation data published by Miller et al. (1992). Thus, this paper is the first to report the distribution of myosin II A in neurons using an isoform specific antibody.

Myosin II staining in the neurite was typically as bright as in the growth cone C-domain, although it often had a periodic distribution. Myosin II may serve a purely structural role, perhaps in maintaining the cylindrical shape of the neurite. Such a role is consistent with the finding that no changes in neurite tension are observed in the absence of growth cone advance in vitro (Lamoureux et al., 1989).

Cell bodies stained brightly for conventional myosin. Confocal microscopy revealed that cortical myosin II staining was organized into linear arrays. The long, thick arrays of actomyosin II on the ventral surface may participate in adhesion of the soma to the substratum, in analogy to stress fibers in non-neuronal cells. Cortical actomyosin II may have functions in addition to stabilizing the linkage of somata to the substratum. Maintaining the spherical shape of neurons presumably requires considerable surface tension. In addition, somas exhibit very little protrusive activity following completion of process production. Both of these functions may be subserved by cortical tension that involves myosin II activity.

Our data on the distribution of conventional myosins in neurons differ from those obtained from previous investigations and provide higher resolution information on the subcellular localization of each of the myosin isoforms, their association with F-actin, and their relationship to protrusion and retraction of the growth cone perimeter. The localization that we report suggests important roles for conventional myosin in growth cone navigation, axon assembly, and soma cortex homeostasis. We are presently testing these hypotheses by perturbing myosin II function in living neurons.

This work was aided by a grant from Paralyzed Veterans of America Spinal Cord Research Foundation (to M.W.R.) and by a grant from the NIH (to P.C.B.).