Myosin Va binding to neurofilaments is essential for correct myosin Va distribution and transport and neurofilament density

The myosin Va family of nonmuscle myosin genes includes the dilute gene from mouse, the myr6 gene from rat, the MYO2 and MYO4 genes from Saccharomyces cerevisiae, the brain myosin Va gene from chicken, the Drosophila myosin V gene, and the myosin Va gene from human (Berg et al., 2001). The protein products of these genes consist of a globular head domain sharing ∼40% homology with the head domain of myosin II, followed by a myosin I–like neck domain containing six IQ motifs that bind four calmodulin light chains and two light chains (Espreafico et al., 1992). The central stalk-like region contains a myosin II–like α helical coiled-coil domain and a second globular domain of unique amino acid sequence. At least three isoforms of myosin V exist: Va (Mercer et al., 1991), Vb (Espreafico et al., 1992), and Vc (Rodriguez and Cheney, 2002). The distinctive functions of these differentially expressed isoforms are unknown. Myosin Va is the specific focus of our studies.

Proteins of the myosin V family serve as molecular motors to transport diverse molecules. Analyses of the MYO2 gene implicate yeast myosin V in the assembly and transport of actin, the translocation of specific proteins to sites of polarized growth, and the transport of vesicles (Santos and Snyder, 1997). The yeast MYO4 gene product transports components necessary for regulating gene expression and targets mRNA molecules to sites of polarized growth (Bobola et al., 1996). In rat and chick neuronal tissue, myosin Va mediates actin-dependent movement of synaptic vesicles and ER (Tabb et al., 1998), and supports filopodial extension in growth cones (Wang and Jay, 1997).

Myosin Va protein in brain is highly expressed (Cheney et al., 1993) and broadly distributed regionally (Mercer et al., 1991; Espreafico et al., 1992), with prominent localizations in growth cones and neurites of cultured neurons, astrocyte processes (Espindola et al., 1992), and axons of several sensory organs (Hasson et al., 1997). Additionally, myosin Va has been reported to immunolocalize to the intermediate filament (IF)^* compartment in a variety of cultured cell types (Engle and Kennett, 1994). Although these findings document that myosin Va is a significant component of CNS neurons and sensory organs, its neural functions are still poorly defined.

Mutations in the mouse myosin Va gene, dilute, cause not only defective melanosome movement but also profound neurological symptoms, including a convulsive disorder that results in death by 3–4 wk of age (Searle, 1952). Mutations in the human myosin Va gene cause Griscelli's disease, a childhood disease characterized by pronounced hypopigmentation and a convulsive disorder that leads to premature death (Pastural et al., 1997). A similar phenotype is associated with mutations of the ashen locus harboring the Rab27a gene (Menasche et al., 2000), which regulates myosin Va recruitment to melanosomes (Wu et al., 2001). Some features of the dilute phenotype also result from the flailer mutation, which originates from germ line exon shuffling between the guanine nucleotide binding protein β 5 and myosin Va genes, leading to the synthesis of a mutant hybrid protein that competes with wild-type myosin Va (Jones et al., 2000). These phenotypes clearly underscore the importance of myosin Va for normal neurological function.

In view of the importance of myosin Va in brain and its possible association with IFs (Engle and Kennett, 1994), we investigated whether myosin Va interacts with neurofilaments and how it influences the properties of neurofilaments in neurons. Neurofilaments predominate in axons and, to a lesser extent, in dendrites and perikarya of the nervous system where they modulate the polar shape and size of the neuron (Nixon and Shea, 1992). Unlike other classes of IFs, neurofilament proteins are heteropolymers formed from three distinct subunits with apparent molecular masses of 200 kD (NF-H), 150 kD (NF-M), and 70 kD (NF-L) on SDS–polyacrylamide gels. In some axons, neurofilaments form a three-dimensional network with microtubules and actin filaments, mediated by one or more members of a family of cross-linking proteins (Svitkina et al., 1996; Yang et al., 1996). How the organization of this network is achieved in axons is not known.

Neurofilaments, microtubules, and microfilaments, as well as their associated proteins, are transported from the perikaryon at speeds of 0.1–2.0 mm/d, collectively referred to as the slow phase of axonal transport. This range of rates is distinct from that of fast axonal transport, which carries mainly vesicular constituents by a microtubule-dependent mechanism powered by kinesin in the anterograde direction (Hirokawa et al., 1991) and by dynein in the retrograde direction (Schnapp and Reese, 1989). It remains controversial as to whether different filamentous cytoskeletal structures are transported exclusively as polymers, subunits/oligomers, or in both forms under different conditions (Hirokawa et al., 1997; Wang et al., 2000; Yabe et al., 2001). Short-range transport mechanisms also mediate restricted longitudinal or lateral movement of organelles within a specific compartment of the neuron. Roles for myosin Va in this form of transport have been suggested (Huang et al., 1999); however, the molecular mechanisms responsible for the movements of cytoskeleton components within axons are poorly understood.

In the study presented here, we establish by multiple criteria that the neurofilament is a major ligand for myosin Va in mouse central and peripheral nervous tissue. We demonstrate that myosin Va binds selectively to the NF-L subunit of the filament, and that myosin Va content, transport, and distribution in axons are regulated in vivo, in part, by levels of NF-L. Our further observations show that myosin Va moves in the slow phase of axonal transport, prominently associates with neurofilaments, and alters neurofilament density selectively when deleted in dilute-lethal (dl) mice, suggesting that myosin Va plays a role in the behavior of neuronal cytoskeletal elements in addition to its well-recognized role in translocating membranous organelles.

An affinity-purified polyclonal antisera against the carboxy-terminal end of myosin Va (Evans et al., 1997) detected a single band of the expected apparent molecular mass (190 kD) in immunoblots of protein from spinal cord, optic, and sciatic nerves (Fig. 1 A). The same protein was identified with a second affinity-purified rabbit polyclonal antiserum raised to the amino-terminal domain of myosin Va (unpublished data). Quantitative immunoblot analysis of total protein extracts from the nervous tissues of the mouse revealed that myosin Va levels are more than twofold higher in the spinal cord than in optic or sciatic nerves (Fig. 1, A and B; Table I). The major proportion of myosin Va (55–67%) immunoreactivity in each of these tissues was associated with the Triton-insoluble cytoskeleton (Fig. 1 B; Table II). By comparison, kinesin heavy chain, another motor protein, was detected only in Triton-soluble fractions (Fig. 1 B). The absence of myosin Va immunoreactive protein in spinal cord fractions from homozygous dl mice carrying a mutation in the myosin Va gene (Fig. 1 C) established the specificity of the affinity-purified polyclonal myosin Va antibody and confirmed the identity of the cytoskeleton-associated 190-kD protein from normal mice as myosin Va. For comparison with myosin Va, a known myosin Va ligand (actin) and NF-L were also analyzed by quantitative Western blot on the same tissue fractions. The proportions of actin were roughly comparable in Triton-soluble and Triton-insoluble fractions of the brain and spinal cord. However, in the optic and sciatic nerves, cytoskeletal fractions contained only a third of the total actin, but 55–61% of the total myosin Va (Table II).

Myosin Va immunoreactivity was broadly distributed in neurons within the mouse central and peripheral nervous systems (Fig. 2), as previously noted. Purkinje cells and their elaborate dendritic arborizations were well immunolabeled, as previously shown (Bridgman, 1999) (Fig. 2 A), and other neuronal populations were strongly immunoreactive (Fig. 2 A, inset). Axons in spinal cord (Fig. 2 B) and sciatic nerve fibers (Fig. 2 C) were also well labeled. By contrast, immunoreactive myosin Va was absent in the corresponding nervous tissue regions of dl mice (unpublished data).

Ultrastructural inspection of immunogold-labeled sections of mouse optic axons revealed that myosin Va immunolabeling in axons is most abundant in association with neurofilaments (Fig. 2, D, F, and G). In some cases, trains of anti–myosin Va gold particles traced the outline of a single neurofilament (Fig. 2, F and G). Sections from optic nerves of 21-d-old dl mice (Fig. 2 E) showed <6% of the total gold labeling seen in age-matched control mice (Fig. 2 D). Immunoreactivity was notably absent from microtubules and myelin (Fig. 2, D–G, arrowheads). To assess the relative distribution of immunoreactive myosin Va, we counted gold particles on, or immediately adjacent to, specific cellular structures. In 25 micrographs of immunolabeled optic axons, the number of gold particles associated with neurofilaments was approximately fourfold greater than that associated with vesicular organelles and the subaxolemmal compartment. Myosin Va immunoreactivity was also associated with structures previously identified to be sites of myosin Va localization (Bridgman, 1999), including the subaxolemmal membrane (unpublished data), smooth ER (Fig. 2 H), mitochondria (Fig. 2 J), and small, membrane-bound vesicles within nerve terminals or preterminal axon swellings (Fig. 2 I).

Because neurofilaments are heteropolymers consisting of three subunits, NF-H, NF-M, and NF-L, we used a blot overlay analysis to investigate which of the neurofilament subunits bind to myosin Va. Purified neurofilament subunits were separated on SDS–polyacrylamide gels and blotted onto nitrocellulose membranes, which were then incubated in solutions containing decreasing concentrations of myosin Va fused with β-galactosidase (Fig. 3 A, lanes 2–5). As a control, purified β-galactosidase protein was incubated with immobilized neurofilament proteins in a concentration equal to the highest concentration of myosin Va–β-gal fusion protein used (Fig. 3 A, lane 6). Bound myosin Va–β-gal or β-galactosidase alone was detected by Western blot analysis using an anti–β-galactosidase antibody. Fig. 3 A demonstrates that myosin Va bound to the NF-L subunit in a concentration-dependent manner, but did not bind to NF-M or NF-H. β-Galactosidase alone did not bind (Fig. 4 A, lane 6), indicating that the association of myosin Va–β-gal fusion protein with NF-L is due to the myosin Va portion of the fusion protein.

To confirm the specificity of the association between myosin Va and the NF-L subunit, we expressed a cloned myosin Va–β-galactosidase fusion protein (myosin Va–FP) or a cloned NF-L–c-Myc fusion protein (NF-L–FP) in bacteria. Crude extracts prepared from each expression system were mixed together to allow binding between myosin Va and NF-L, which were then immunoprecipitated with an antibody to either myosin Va–FP (anti–β-gal monoclonal antibody) or NF-L–FP (anti-Myc polyclonal antibody). The immunoprecipitates were separated on SDS-PAGE and blotted onto nitrocellulose for Western blot analysis. Myosin Va–FP and NF-L–FP were detected with a mixture of anti–β-gal and anti–c-myc antibodies. Anti-Myc Ab (Fig. 3 B, lane 1) or anti–β-gal Ab (Fig. 3 B, lane 4) immunoprecipitated both myosin Va–FP and NF-L–FP. As expected, in unmixed bacterial extracts, Myc antibody immunoprecipitated only NF-L–FP (Fig. 3 B, lane 2) but not myosin Va–FP (Fig. 3 B, lane 3), whereas β-gal antibody specifically precipitated myosin Va–FP (Fig. 3 B, lane 5) but not NF-L–FP (Fig. 3 B, lane 6). In contrast to NF-L, bacterially expressed NF-M or NF-H subunits alone did not coimmunoprecipitate with myosin Va–FP (unpublished data).

Because the NF-L subunit of the neurofilament triplet is the prime ligand for myosin Va, it was of interest to determine whether myosin Va would coimmunoprecipitate with NF-L from neuronal tissue. Using the NF-L mAb NR-4 (Sigma-Aldrich), we immunoprecipitated NF-L from Triton-insoluble cytoskeletal preparations of brain and spinal cord and observed that the immunoprecipitates were enriched not only for NF-L (Fig. 3 C, bottom) but also for myosin Va (top, lanes 1, 2, 4, and 5). In the absence of NF-L antibody, no myosin Va or NF-L sedimented after centrifugation of the Sepharose beads (lane 3), indicating that precipitation of myosin Va and NF-L was antibody dependent.

The specificity of myosin Va binding to cytoskeletal proteins was further investigated by blot overlay analysis of Triton-insoluble cytoskeletal fractions from mouse spinal cord (Fig. 4 A, lanes 1–5) and sciatic nerve (Fig. 4 A, lanes 6–10). In both tissues studied, the major binding partner of myosin Va was NF-L (Fig. 4 B); however, we also observed binding of myosin Va to a 45-kD protein in spinal cord (Fig. 4 A, lanes 2–5) and to a 55-kD protein in sciatic nerve (Fig. 4 A, lanes 6–9). Based on these apparent molecular masses on SDS gels, we probed the same blots with antibodies to glial fibrillary acidic protein (GFAP), a 45-kD protein (Fig. 4 D), and peripherin (Fig. 4 C), a 55-kD protein, and observed that the immunoreactivity of these proteins overlapped precisely with the positions of the protein bands detected by myosin Va binding. Thus, from the dozens of different proteins present in the cytoskeleton preparations, myosin Va bound specifically to three structurally related IF proteins. It is not surprising that actin was not detected in this overlay assay because this myosin Va clone (Fb8; Engle and Kennett, 1994) does not have a complete actin binding site.

Based on the foregoing evidence of a myosin Va–NF-L association, we next examined whether the levels of NF-L in axons influence the axonal content of myosin Va by analyzing mice in which neurofilament levels were altered. Targeted mice lacking the NF-L gene are devoid of NF-L and virtually all neurofilaments (Zhu et al., 1997). NF-L is absent in the sciatic nerves of these mice (Fig. 5, A and C, lane 2), and we observed a 55% reduction in the levels of myosin Va (Fig. 5, A and B, lane 2). By contrast, myosin Va levels were not altered in sciatic nerves of NF-H–deleted mice (Fig. 5, A and C, lane 3), in which NF-L and neurofilament levels in axons are normal (Rao et al., 1998). The effects on myosin Va of increasing NF-L levels and filament number in axons were also examined in mice overexpressing the NF-L protein (Xu et al., 1993). NF-L levels and neurofilament counts in sciatic nerves from NF-L–overexpressing mice have been previously shown to be ∼1.5-fold higher than the corresponding control mice (Xu et al., 1993). We confirmed the reported increase in NF-L levels (Fig. 5, D and F, lane 2) and observed by quantitative immunoblot analyses that myosin Va levels were comparably increased (50%) in the sciatic nerves of NF-L transgenic mice (Fig. 5 D, lane 2). These results indicate that axonal myosin Va content is influenced by levels of NF-L, neurofilament number, or both.

In earlier studies, we tentatively identified a Triton-soluble radiolabeled protein advancing 1–2 mm/d in the slow phase of axonal transport as a form of myosin (Lewis and Nixon, 1988). This protein, however, comigrated on SDS gels with multiple phosphovariants of NF-H, which precluded its full characterization. To eliminate NF-H as a confounding variable, we conducted the present myosin Va transport study in NF-H–deleted mice (Rao et al., 1998). Retinal ganglion cells of NF-H–deleted mice were labeled by intravitreous injection of [³⁵S]methionine, and after 3 d, we prepared Triton-soluble and Triton-insoluble fractions from optic nerves or optic tracts pooled from three mice and immunoprecipitated myosin Va. Myosin Va coimmunoprecipitated with only a limited set of labeled proteins, which included NF-L, NF-M, spectrin (a known neurofilament binding protein; Frappier et al., 1991), actin, as well as several unidentified proteins (Fig. 6 A). The identities of immunoprecipitated myosin Va, NF-L, and actin were confirmed by Western blot analysis on the same blots after autoradiography (Fig. 6 C). No radiolabeled proteins were precipitated by preimmune IgG used under the same conditions before myosin Va immunoprecipitation (unpublished data).

Most of the radiolabeled myosin Va was still present in the optic nerve 3 d after isotope injection, indicating that myosin Va principally moves in the slow phase of axonal transport. To establish the transport pattern more precisely, additional NF-H–null mice were injected intravitreously with [³⁵S]methionine. After 3 or 7 d, the optic pathways from three mice were then cut into eight consecutive 1.1-mm segments, which were fractionated into Triton-soluble and -insoluble fractions and subjected to electrophoresis, electroblotting, and autoradiography. The myosin Va in Triton-soluble (Fig. 7, A and B) and cytoskeleton-associated fractions (Fig. 7, C and D) displayed a similar transport pattern, moving as a wave averaging 0.73–0.94 mm/d (peak) to 2.2 mm/d (front), which correspond to a typical slow component b rate of transport (0.8–2 mm/d) in optic axons (Nixon, 1991).

Triton-soluble actin displayed a transport profile (Fig. 7 E) similar to that of myosin Va. Cytoskeleton-associated actin also moved into optic axons at a rate of 1.1 mm/d (Fig. 7 G) (Yuan et al., 2000), but a significant proportion of this insoluble actin slowed to a rate similar to that of NF-L (Fig. 7 I, 0.31 mm/day). The slower movement of neurofilaments than much of its associated protein myosin Va suggests a dynamic association during axonal transport, which is not unexpected in light of observations that microtubule-associated proteins, the family of proteins that associate with microtubules in axons, are also transported more rapidly than tubulins in vivo (Nixon et al., 1990; Mercken et al., 1995).

Direct evidence for an in vivo interaction of myosin Va with NF-L was sought by investigating whether the elimination of neurofilaments altered the movement or principal location of myosin Va in axons. To visualize radiolabeled myosin Va on gels in pulse-labeling studies, we analyzed axonal transport patterns in optic nerves of mice deleted of both NF-L and NF-H. In these neurofilament-deficient axons, the average rate of myosin Va translocation was significantly faster than that in normal axons at 3 or 7 d after [³⁵S]methionine injection (Fig. 8, A and B). Transport of tubulin and major slow component b proteins was unaltered in neurofilament-deficient mice (Fig. 8, E and F). Immunogold labeling NF-L-null mice showed that, in the absence of neurofilaments, myosin Va distribution is restricted to the subaxolemmal compartment and to the surfaces of membranous vesicles (Fig. 8, G and H). An increased incidence of subaxolemmal labeling implied that a proportion of the neurofilament-associated myosin Va was redistributed in these mice. These observations demonstrate unequivocally that a significant proportion of myosin Va normally interacts with neurofilaments in vivo.

To further establish the physiological significance of the myosin Va–neurofilament interaction, we analyzed the ultrastructure and cytoskeletal protein content of peripheral nerves from dilute mice and their normal littermates. Neurofilament number was increased nearly twofold (P < 0.001) in axons of dl mice compared with littermate controls (mean ± SEM, 1177 ± 105 vs. 666 ± 51 NFs/axon, respectively [n = 75–85 axons]). Regression analyses indicate that the density of neurofilaments was also significantly higher (P < 0.0001) in axons of all caliber sizes in the dilute mice (251 ± 13 vs. 113 ± 7 NF/μm²) (Fig. 9, A and C, inset). Western blot analyses confirmed an increased content of NF-L of ∼60% in dl sciatic nerves compared with controls (Fig. 9 E). By contrast, levels of β_III-tubulin and actin in the same nerve extracts were not significantly altered by deletion of myosin Va. The distribution of axon cross-sectional areas in dilute mice was similar to control mice, indicating that the higher neurofilament densities did not alter axonal calibers (Fig. 9 D). Coupled with the analyses of NF-L–null and transgenic mice (Fig. 5), these studies provide strong evidence for a physiological interaction between myosin Va and neurofilaments, which modulates neurofilament organization in axons.

Our results identify the molecular motor, myosin Va, as a neurofilament-associated protein and establish NF-L as a major ligand of myosin Va in nervous tissue. Multiple lines of evidence support these conclusions. First, more than half of the total myosin Va in CNS tissues coisolates with the neurofilament-rich Triton-insoluble cytoskeleton. Second, immunoreactive myosin Va is abundant in nerve fiber tracts of brain, spinal cord, and peripheral nerves. Immunoelectron microscopic analyses show that the majority of the myosin Va in axons is associated with neurofilaments. Third, electroblot overlay analyses and in vitro and in vivo coimmunoprecipitation studies with either myosin Va or NF-L antibodies establish that the binding of myosin Va to neurofilaments is mediated through a specific association of myosin Va with the NF-L subunit. Fourth, we demonstrate in vivo that myosin Va levels, distribution, and transport in axons are influenced by the level of NF-L subunits. Finally, deletion of the myosin Va gene in dilute mice selectively alters neurofilament content and organization in axons.

In addition to establishing a novel myosin Va–neurofilament association, our results confirm previously observed associations of myosin Va with ER membranes, synaptic vesicles (Tabb et al., 1998), and actin. Extending observations that myosin Va is an actin binding protein (Espreafico et al., 1992; Cheney et al., 1993), we demonstrated that in vivo–labeled actin and myosin Va coimmunoprecipitate and cotransport and that myosin Va antibodies decorate the actin-rich subaxolemmal compartment (Fig. 2; Fig. 8, G and H; Kobayashi et al., 1986). It is not surprising that the myosin Va–neurofilament interaction has been less well appreciated than the associations of myosin Va with actin or membranous organelles because previous myosin Va localization studies have focused on organelle-rich and neurofilament-poor cellular compartments in cell bodies and dendrites in brain, rather than in axons.

The NF-L subunit is one of only a limited number of proteins that interacts directly with myosin Va. The further observation that the two other major ligands in the spinal cord and sciatic nerve are also IF proteins, GFAP and peripherin, suggests that myosin Va plays a more general role in IF behavior or that IFs mediate important functions of myosin Va. By performing multiple protein sequence alignment with hierarchical clustering (Corpet, 1988; Multalin version 5.4.1) on NF-L, peripherin, and GFAP, we found 132 amino acid homologies in a sequence of 350 residues corresponding to the head and rod domains. The observation that as many as 82 of these amino acid positions differ in NF-H and NF-M may partly explain the relative selectivity of the myosin Va binding to the three “core” subunits of IFs. The observation that other high abundance proteins, such as tubulin, bound negligible amounts of myosin Va underscores the specificity of the myosin Va–IF interaction. Moreover, that another molecular motor, kinesin, is found in negligible amounts in cytoskeletal fractions indicates that tight association with cytoskeletal structures is not a general feature of motor molecules.

The association of myosin Va with neurofilaments raises new possibilities regarding its function in the nervous system where levels are exceptionally high (Cheney et al., 1993). One of these possibilities is modulating the organization of the axoplasm. We observed that myosin Va deletion in dilute mice creates a denser packing of neurofilaments in axons, suggesting a role in neurofilament spacing. Because NF-L and actin bind to myosin Va at separate sites, myosin Va may be capable of dynamically cross-linking IFs and microfilaments. Recent studies have emphasized the role of molecules other than neurofilaments themselves in regulating lateral spacing of filaments in axons. Molecules, including BPAG and plectin, have recently been shown to cross-link IFs, microfilaments, and microtubules (Svitkina et al., 1996; Yang et al., 1996), although, unlike myosin Va, these proteins have no known ATPase or motor activity. If myosin Va does in fact link filament systems, it is likely to be in the service of dynamically rearranging these structures within the cytoskeletal network. Neurofilaments could either act as an anchor from which myosin Va could move other proteins or vesicular organelles (e.g., actin) or as a cargo of myosin Va. In regard to the first possibility, neurofilaments provide a three-dimensional lattice interconnecting the microtubule system with the subaxolemmal compartment (Yang et al., 1996). This stationary network of filaments conceivably could represent a system of tracks well suited for myosin Va to guide microfilaments or membranous organelles laterally within axons to achieve the proper radial organization of these structures. Neurofilaments have also been shown to be possible ligands of membrane-associated enzymes and receptors (Terry-Lorenzo et al., 2000; Kim et al., 2002), raising the possibility that movements of molecules of this type along neurofilaments may be mediated by myosin Va. Myosin Va has been implicated in moving actin short distances within growth cones (Evans et al., 1997; Bridgman, 1999; Huang et al., 1999). Finally, short-range rearrangements of neurofilaments within the axon, such as those that occur during early postnatal development (Sanchez et al., 1996), might also require motor activity.

Actin and neurofilaments also move long distances by slow axonal transport. The average rate of myosin Va transport that we observed along optic axons is similar to that of actin and close to the rate at which purified myosin Va moves along actin cables from dissected Nitella cells in vitro (2.5–3.8 mm/d) (Cheney et al., 1993). Interestingly, Willard (1977) identified two polypeptides (195 kD and 200 kD) that cosediment with actin in an ATP-reversible way and were transported along optic axons at slow transport rates. Transport of at least some neurofilaments in growing axons of cultured neurons involves a series of rapid movements punctuated by long periods of immobility (Wang et al., 2000). This pattern suggests that motors, such as kinesin, which are capable of mediating fast transport rates, may be attractive candidates for powering neurofilament movement. Because myosin Va can bind directly to kinesin (Huang et al., 1999), the interaction of myosin Va with neurofilaments could represent one mechanism to facilitate neurofilament movement along microtubules. Although the results in dilute mice imply that myosin Va is not essential for movement of neurofilaments into axons, the considerably increased neurofilament number in axons could reflect impaired slow transport. However, a particular pattern of neurofilament distribution along axons, by itself, is not predictive of transport kinetics. An increased rate of incorporation of transported neurofilaments into the stationary cytoskeletal network along axons (Nixon, 1998) would yield a similar picture. Definitive tests of these possibilities require long-term labeling studies that are precluded in dilute mice by the frailty and early death of these mice after 3–4 wk.

In conclusion, these novel interactions of myosin Va with IFs indicate previously unrecognized roles for myosin Va in regulating cytoskeleton dynamics in the nervous system. Although long or short range transport and/or rearrangements of neurofilaments represent several of the possible roles, interactions between neurofilaments and myosin Va might instead, or in addition, be important in modulating myosin Va–mediated movements of other cytoskeletal proteins, membranous organelles, or membrane-associated proteins. A variety of experimental approaches will be required to investigate the range of intriguing possibilities.

Control mice were the normal mouse strain C57BL/6J. Brain, spinal cord, sciatic, and optic nerves from 4-mo-old mice were dissected as previously described (Nixon and Logvinenko, 1986). NF-L–null mice were screened according to Zhu et al. (1997). NF-L Tg mice were screened according to Xu et al. (1993). NF-H–deleted mice were screened according to Rao et al. (1998). NF-H/NF-L double knockout mice (HL-DKO) were generated by crossbreeding and screening according to Rao et al. (1998) and Zhu et al. (1997). Dilute mice were bought from Jackson ImmunoResearch Laboratories. All the tissues were frozen on dry ice and stored at −80°C. To extend survival of dl mutants to 4 wk, pups were fed manually every 8 h from postnatal day 10 with kitten formula (KMR; PetAg Inc.) dissolved in sterile water and mixed with maize syrup.

Total protein homogenates from brain, spinal cord, sciatic, and optic nerves were made according to Rao et al. (1998). Protein concentration was determined using bicinchoninic acid (BCA) assay kit (Pierce Chemical Co.). Protein extracts of known amounts from different neuronal tissues were fractionated on SDS-PAGE gels containing 6 or 7.5% polyacrylamide and transferred to nitrocellulose membranes. Triton-insoluble and -soluble fractions were made according to Nixon et al. (1990). After protein estimation, equal amounts of Triton-insoluble and -soluble fractions were boiled, run on a 7% SDS–polyacrylamide gel, and immunoblotted with affinity-purified polyclonal antibody directed against the COOH terminus of the myosin Va fusion protein (Evans et al., 1997), NF-L (NR-4), α-tubulin (DMIA), and kinesin heavy chain mAbs as specified by the manufacturer. The blots were processed with the ECL system (Amersham Biosciences) or the alkaline phosphatase system (Promega). Band images were quantified with a BioImage whole band analyzer from Kodak.

C57BL/6J mice were anesthetized and fixed by cardiac perfusion using 10% neutral buffered formalin in TBS, pH 7.4. The brain, cervical spinal cord, and sciatic nerves were dissected and 40-μm-thick vibratome sections were processed for immunocytochemistry (Cataldo et al., 1990) using a polyclonal antibody directed against myosin Va. Several sections were processed in tandem in the absence of primary antibody and served as controls.

3–4-wk-old dl and their control mice and NF-L–null mice were anesthetized and perfused fixed with 4% paraformaldehyde–0.2% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4). The tissues were fixed for an additional 2 h at room temperature in 10% formalin. The brain and optic pathway were removed and 40-μm-thick vibratome sections of the cerebral peduncle and the optic pathway at a distance of 700 μm from the retina were dehydrated in series of alcohols and embedded in Epon as previously described (DeMey, 1983). Ultrathin sections were placed on nickel-coated grids and processed for postembedding immunostaining (DeMey, 1983) using a polyclonal antibody directed against myosin Va. Grids incubated without primary antibody served as controls. All grids were poststained in uranyl acetate and lead citrate and inspected using a JEOL 100EX electron microscope.

Sciatic nerves were dissected and analyzed morphometrically to determine axonal cross-sectional areas for the entire fiber population and neurofilament numbers for a representative subpopulation of axons, as previously described (Rao et al., 1998).

Neurofilament triplet subunits were purified as previously reported (Balin et al., 1991). The myosin Va fusion protein (∼120 kD) (Engle and Kennett, 1994) consists of a small 3.5-kD piece of β-galactosidase fused to that portion of myosin Va representing 57% of the full-length protein containing half of the actin binding site, calmodulin binding domains, α helical coiled-coil domains, and a small portion of the COOH-terminal tail domain. The fusion protein was expressed in XL-1 Blue Escherichia coli (Stratagene) cells and purified using an anti–β-galactosidase affinity column (Promega) per the manufacturer's protocol. 10 μg of purified neurofilament triplet proteins were run per lane on a 7% SDS–polyacrylamide gel, blotted onto nitrocellulose, and cut into strips. The strips were blocked and incubated in 10 μg/ml, 1 μg/ml, 0.1 μg/ml, 0.01 μg/ml, or, as a control, 0 μg/ml of myosin Va fusion protein in PBS. Bound myosin Va–β-gal was detected with anti–β-galactosidase mAb (Boehringer) according to the manufacturer's instructions and blots were processed with the ECL system (Amersham Biosciences) and by exposure to Kodak XAR film. The purified neurofilament triplet proteins immobilized on the nitrocellulose were visualized by Coomassie blue staining. Cytoskeletal blot overlay assays were performed same as indicated above for purified NF proteins.

Myosin Va–β-gal fusion protein was expressed as described above and the crude extract was used for immunoprecipitation. NF-L myc-tagged plasmid (Heins et al., 1993) was transfected into BL21-DE3 host cells, induced, NF-L–containing lysates were made, and protein concentrations determined. 2 μg of NF-L–FP crude extract and myosin V–FP crude extract (containing ∼3 μg of myosin Va–FP) were mixed together in 1× RIPA buffer. Protein A/G–agarose (Boehringer) was also added to preclear. The mixture was incubated overnight at 4°C to allow binding between NF-L–FP and myosin Va–FP. As controls, 2 μg of NF-L–FP or 3 μg of myosin Va–FP, in 1× RIPA buffer were also precleared with protein A/G–agarose overnight at 4°C. Protein A/G–agarose was removed with centrifugation, and NF-L–FP/myosin Va–FP complex, NF-L–FP alone, and myosin Va–FP alone were immunoprecipitated for 8 h at 4°C with either 5 μl of a polyclonal Myc antisera (to precipitate NF-L–FP) or with 2.5 μl of 1:300 dilution of anti–β-galactosidase mAb (Promega). Protein G–agarose (50 μl) was added and the mixtures were nutated overnight at 4°C. Protein G–agarose was pelleted, washed, boiled in Laemmli buffer, fractionated on 7.5% SDS–polyacrylamide gels, blotted onto nitrocellulose membranes, and probed with a mixture of anti–β-gal and anti-Myc antisera.

Triton X-100–insoluble fractions were suspended in TBS, and SDS was added to a final concentration of 1%. The samples were diluted 1:4 in neurofilament extraction buffer (60 mM Tris-HCl, pH 7.4, 190 mM NaCl, 6 mM EDTA, 1.25% Triton X-100, 1 mM PMSF). The samples were sonicated for 20 s and protein was estimated by bicinchoninic acid (BCA) method. A monoclonal antibody to NF-L (NR-4) was added to 0.6–2.4 mg of cytoskeletal fractions at a dilution of 1:10 and the samples were incubated overnight at 4°C. The antigen antibody complex was precipitated using protein A/G–Sepharose (Santa Cruz Biotechnology, Inc.), washed, boiled in Laemmli buffer, fractionated on 7% SDS gels, and immunoblotted with myosin Va and NF-L antibodies.

NF-H–null mice were injected intravitreally with 100 μCi of [³⁵S]methionine into each eye and killed after 3 d, as previously described (Nixon and Logvinenko, 1986). Cytosolic and cytoskeletal fractions of optic nerves and tracts obtained were used to immunoprecipitate labeled myosin Va using an affinity-purified polyclonal antibody to myosin Va (Evans et al., 1997). The samples were incubated overnight at 4°C. The antigen–antibody complex was precipitated using protein A/G–Sepharose, electrophoresed, electroblotted, and exposed to X-ray film followed by Western blotting for myosin Va, NF-L, and actin.

The retinal ganglion cells of 3–4-mo-old NF-H and NF-H/NF-L double null mice were radiolabeled in vivo with 100 μCi of [³⁵S]methionine by intravitreal injection. 3 and 7 d after injection, optic pathways from groups of three animals were cut into eight consecutive 1.1-mm segments. Triton-soluble and -insoluble NF-rich cytoskeleton preparations from each segment were subjected to SDS-PAGE, electrotransfer of proteins, phosphor-imaging, and autoradiography.

The authors wish to express deep appreciation to Janet Rosdil, Lucy Morales, and Hayley McAuliff (Ludwig Institute for Cancer Research, San Diego, CA) for help with manuscript preparation and Dr. Asok Kumar (Ludwig Institute for Cancer Research) for his help with electron microscopy. We are grateful to Dr. Zuoshang Xu (University of Massachusetts Medical School, Worcester, MA) for NF-L–overexpressing mice and Dr. Don Cleveland (Ludwig Institute for Cancer Research) for the NF-L–Myc construct.

This work was supported by United States Public Health Service grant P30-DK19525 and by Merit grant AG05604 and National Institutes of Health National Research Service Award training grant in neurodegeneration, No. 2-T32-AG00222 (L.J. Engle).