ABSTRACT
The control of radial growth of axons is of functional importance because caliber is a principal determinant of conduction velocity in myelinated nerve fibers. Neurofilaments, the major cytoskeletal protein in myelinated nerves, appear to be intrinsic determinants of caliber. Evidence supporting this derives first from the linear relationship between neurofilament content and axonal diameter. Further, following distal axonal injury in a peripheral nerve, caliber is reduced in the proximal axonal stumps. This reduction in caliber is itself due to selective suppression of neurofilament gene expression, thereby leading to lower levels of newly synthesized neurofilament subunits transported into the axon and a consequent decrease in axonal neurofilament content. To demonstrate directly the physiological consequence of altering normal neurofilament accumulation, we have elevated neurofilament expression by introducing additional genes into transgenic mice. The clear result is that increases in NF-L content alone are not sufficient to increase axonal caliber. To test the consequence of disruption of normal filament accumulation, we have identified dominant assembly-disrupting mutants in NF-L and NF-M and have used these to produce transgenic animals in which neurofilament assembly should be disrupted.
INTRODUCTION
Axonal growth from an initial neuronal precursor cell is achieved in two general phases. The first of these is the extension of an asymmetric process that must find its way to its ultimate target. For motor and sensory neurons of the human peripheral nervous system, the resulting axonal process may be up to one meter in length, exceeding the diameter of the cell body by a factor of 104-105. Once appropriately targeted, a second phase of growth is initiated in which the axons expand radially. Initially, multiple thin, immature axons are ensheathed by single Schwann cells in the peripheral nervous system. As caliber increases, some Schwann cells make exclusive association with individual axons and begin to ensheath them with compact myelin.
To be sure, a major question, explored now on many fronts, is how initial neurite extension is achieved and how specific neurons are guided to appropriate targets. But the subsequent radial growth too is an important phase, since axonal caliber is the principle determinant of conduction velocity in myelinated nerve fibers (Hursh, 1939; Gillespi and Stein, 1983). Indeed from the work of Voyvodic (1989), a plausible view has emerged that caliber may be the crucial determinant of whether an axon becomes myelinated. Evidence in favor of this arose following experimental elimination of 50-95% of the sympathetic postganglionic axons innervating the rat submandibular salivary gland. Although not normally myelinated, more than 60 % of the remaining axons become ensheathed in myelin, with essentially all axons of diameter greater than a threshold of 1.6 pm becoming myelinated.
Underlying the extension, maintenance and radial growth of axons is a cytoskeletal framework of structural elements. Most prominent in the early neurite extension phase are microtubules and actin filaments, and their combined actions provide tensile and compressive forces whose net sum supports extension of the initial axonal asymmetry (see Heidemann et al. this volume). A third class of filaments, the intermediate filaments of axons (called neurofilaments, NF), are also present during outgrowth, but they are not abundant elements and their role, if any, is obscure in this phase. During radial growth, however, NF synthesis is elevated about 5-10 fold, and for myelinated axons NFs generally become the most abundant component of the axonal cytoskeleton, outnumbering microtubules by up to an order of magnitude.
Precisely what the functional role is of NF in either phase of axonal elaboration is unproven, an uncertainty echoed for the corresponding intermediate filament (IF) arrays in other animal cells. Indeed, essentially all vertebrate cells contain within their cytoplasms an array of filaments assembled from one or more classes of subunits that belong to the family of IF proteins. Each of these subunits contains within it a conserved 310 amino acid a-helical domain that through coil-coiled interactions assembles into structurally similar 8-10 nm filaments (Steinert and Roop, 1988). On the basis of sequence differences and the conservation of intron positions, the known IF subunits have been divided into five generally accepted classes. Classes I and II contain the acidic and basic keratins, respectively, both of which are expressed in epithelial cells. Class III consists of vimentin found in mesochymal cells, desmin expressed in muscle cells, glial fibrillary acidic protein expressed in glial cells, and peripherin expressed during the outgrowth phase of a subset of neurons in the peripheral nervous system. Class IV are the major neurofilament proteins, which are probably expressed as an abundant intermediate filament component in all mature neuronal cells. Class V comprises the nuclear lamins that form the filamentous network underlying the nuclear membrane in all animal cell nuclei. In addition, a newly discovered IF subunit, nestin, is accumulated in the early neurite outgrowth phase of neuroepithelial cells. Originally it was proposed that nestin represents a sixth IF class (Lendahl et al. 1990), but the structure of the gene and sequence similarity suggest that it is more properly placed within the class IV subunits.
In mammals, NF are comprised primarily of three subunits, NF-L, NF-M, and NF-H (Hoffman and Lasek, 1975), whose relative molar abundance, respectively, in mature axons is approximately 5:2:1. Following an ≈100 residue head domain, each subunit carries an ≈310 amino acid u-helical rod domain characteristic of all IF proteins. The 68 kd NF-L, 95 kd NF-M and 115 kd NF-H subunits differ in size primarily due to the variable length of the carboxy-terminal tail domains (Geisler et al. 1983; Geisler and Weber, 1986; Lewis and Cowan, 1986; Levy et al. 1987; Myers et al. 1987; Julian et al. 1988; Lees et al. 1988), which in mouse contain 142, 438, and 676 amino acids respectively, in NF-L, NF-M and NF-H. Although the arrangement of the subunits in the final polymer is not established, in vitro reconstitution experiments have shown that NF-L and NF-M can individually assemble into filaments (Geisler and Weber, 1981; Liem and Hutchinson, 1982; Gardner et al. 1984; Tokutake et al. 1984), whereas NF-H is incompetent for self-assembly (Geisler and Weber, 1981). Still, since all three subunits contain the rod domain, almost certainly each is an integral filament subunit in vivo. Precise roles for the head and tail regions are not yet proven, but immunocytochemical and in vitro reassembly experiments suggest that the long tail domains of NF-M and NF-H extend from the core of the filament and seem likely to be involved in establishing proper filament spacing to other axonal structures (Hirokawa et al. 1984; Hisanaga and Hirokawa, 1988).
With the exception of the keratin network that has recently been demonstrated as essential for the structural integrity of the epidermis (Vassar et al. 1991), a convincing functional role for the remaining IF proteins has yet to be documented. For NF, a functional role was first suggested on the basis of a close correlation between the number of axonal NF and the cross-sectional areas of large myelinated nerve fibers (Friede and Samorajski, 1970; Weiss and Mayr, 1971; Berthold, 1978). An example of the linear correlation between cross-sectional area and neurofilament number in mature myelinated motor axons of the sciatic nerve (taken from the work of Hoffman et al. 1984) is shown in Fig. 1A. What is clear is that in each of these axons there is a nearly constant NF density (number of NF/cross-sectional area). Beyond this correlation, stronger evidence that NF content determines caliber is provided by the observation of the events that follow nerve injury induced by either crushing or severing (known as axotomy). Following distal axotomy, a reduction in axonal diameter begins adjacent to the cell body and spreads anterogradely along the nerve fiber at a rate equal to the velocity of NF transport within the axon (as sketched in Fig. IB). This process is referred to as somatofugal axonal atrophy. By following the appearance of newly made proteins in the proximal axons, it was shown that the axotomy-induced diminition in NF content was ac-companied by a selective reduction in the amount of newly made NF protein transported into and assembled within the proximal axon (Hoffman et al. 1984, 1985).
In order to test more directly the potential role of NF as a determinant of axonal caliber (or in other neurite functions), we have now begun to use molecular genetics to investigate the in vivo consequences of altering NF levels. We first document that following axonal injury the reduction in NF synthesis (which is mirrored by a reduction in caliber) is itself a direct result of reprogramming the synthesis of neurofilament subunits. From this, and with cloned genes encoding each of the three subunits in hand, we have used two different approaches to test the consequence on axonal structure and function of altering NF levels in neurons in vivo. One approach has been to generate transgenic mice expressing elevated levels of wild-type NF subunits in neurons in vivo. The resulting increased expression and assembly has allowed a first test of the hypothesis that NF expression is an intrinsic determinant of axonal caliber. In a second approach, we have sought to identify assembly-disrupting dominant mutations within NF subunits and to express such mutants in transgenic animals so as to disrupt the normal assembly process.
RESULTS
Axotomy-induced changes in NF gene expression correlate with changes in caliber
As a first test of the hypothesis that NF gene expression, at least in part, specifies axonal caliber in large myelinated fibers, we have further exploited axotomy to induce reduction in caliber of axons of the sciatic nerve. Although the consequent axonal shrinkage is well correlated with reductions both in NF content and in the amount of newly synthesized NF subunits transported into the axon (see Fig. 1C), a key remaining question is whether the measured changes reflect a true reduction in NF gene expression or alternatively result from changes in the gating of newly synthesized NF subunits into the axon. To examine this question in the sensory axons of the rat sciatic nerve, we measured the effects of axotomy on the levels of NF-L mRNAs accumulated within the corresponding nerve cell bodies, located in the lumbar dorsal root ganglia (DRG). mRNA accumulation and protein synthesis are restricted to the perikarya and proximal dendrites (see, for example, Lasek et al. 1973). At various times after axotomy, levels of mRNAs encoding NF-L were examined in total RNA isolated from DRG. Equal amounts of RNA from each sample were electrophoreti-cally separated by size, blotted to nitrocellulose and NF-L RNAs detected specifically by hybridization with a 32P-labeled probe prepared from a mouse NF-L cDNA clone. As shown in Fig. 2 (bottom panel), NF-L mRNA levels decline markedly in abundance beginning four days after axotomy, reaching a 5—7 fold reduction within two weeks post-axotomy. In situ hybridization not only confirmed this quantitative loss in NF-L mRNAs, but also demonstrated that, as expected, NF-L mRNAs were restricted to the neuronal cell bodies (see Hoffman et al. 1987). Levels of NF-L mRNAs do not recover to normal until about 70 days post-axotomy, by which time nerve regeneration is complete and caliber has recovered within the proximal axon (see Fig. 1C). In contrast to the reduction in NF-L mRNAs, the mRNAs for the growth-associated protein GAP43 (Fig. 2, top panel), as well as for tubulin (Fig. 2, middle panel) and actin (not shown), two cytoskeletal components required for neurite re-extension, are elevated 3-5 fold over a time course very similar to that of the reduction in NF-L RNAs (Hoffman, 1989). Hence, the loss of NF-L RNA is specific and not the consequence of a general reduction in mRNA content following axonal injury. Further, since the magnitude and time course of the diminution in NF-L mRNA are comparable to that seen in both the amount of NF transported into the axon (Fig. 1C) and to the NF content within the proximal axon, we conclude that axotomy induces a re-programming of NF expression, resulting in reduction of proximal NF content concomitant with shrinkage in caliber. These findings support the view that expression of a single set of neuron specific genes (the NF genes) is a direct determinant of axonal caliber.
Over-expression of NF-L in the peripheral nervous system of transgenic mice: increased NF density without affecting caliber
To test directly the consequences of accumulation of excess NF on axonal morphology, mice were constructed carrying a transgene encoding wild-type mouse NF-L under the control of the strong transcriptional promoter from Murine Sarcoma Virus (MSV, see Fig. 3A). While we were concerned that this promoter would probably yield expression in a variety of cell types and that this might be detrimental to the survival of the transgenic animals, we reasoned that such non-neuronal expression could also provide useful information concerning NF assembly and organization in the absence of neuronal influences.
In any event, we obtained 13 transgenic founders that carried between 1 and 300 integrated copies of the transgene (Monteiro et al. 1990). Immunoblotting of whole cell extracts from a series of tissues demonstrated that in several lines substantial levels of transgene products could be detected. For example, NF accumulated to 2.5 % of total cell protein in kidney and 0.5 % of total cell protein in skeletal muscle. A curious finding was that despite expression of transgene RNA at a level 20 fold higher than that of the endogenous NF-L in whole brain, NF-L protein did not accumulate to a level in excess of that in non-transgenic littermates (data not shown). The failure of abundant transgene RNA to elevate NF-L protein in brain is surprising, but since NF-L can accumulate (and assemble) in non-neuronal tissues, this cannot be the result of an inadvertent mutation within the transgene. Rather, it must indicate the presence of a post-transcriptional regulatory mechanism that acts to limit NF-L accumulation in some (and perhaps most) brain cells.
Although we did not detect any increase in NF-L within the neurons of the central nervous system, examination of NF-L accumulation in peripheral neurons demonstrated accumulation of substantial transgene RNA and a corresponding increase in protein product. For example, in one transgenic line, blot analysis of RNAs isolated from dorsal root ganglia revealed a >5 fold increase in NF-L mRNAs in transgenic ganglia, while NF-M mRNA levels were unchanged (compare normal and transgenic in Fig. 3B). Immunoblot analysis of protein extracts from sciatic nerves (which contain axons of both the sensory and motor neurons) revealed that in this transgenic line (and two others) NF-L accumulated to a level 3-4 fold higher than that in non-transgenic littermates (Fig. 3C: compare control nerve extracts, lanes 1 and 2, with transgenic extracts, lanes 3 and 4). That transgene expression was absent in the surrounding Schwann cells and largely confined to the large axons was confirmed by localization of NF-L proteins using immunocytochemistry in paraffin sections of sciatic nerves (Monteiro et al. 1990). In addition, in situ hybridization in the DRG, which contain the cell bodies of the sensory neurons, revealed that all NF-L encoding mRNAs (transgene and endogenous) were primarily, and probably exclusively, localized within the neuronal cell bodies. No evidence was generated either by immunological or by in situ hybridization techniques for the presence of NF-L RNAs within the surrounding Schwann cells.
The morphologic consequence of increased NF-L accumulation in the sciatic nerves of transgenic animals was analyzed by electron microscopy. Sciatic nerves isolated from adult (greater than 6 months old) transgenic animals and from control, non-transgenic littermates were fixed in 5% glutaraldehyde, embedded in epon, and transverse sections were obtained and analyzed by electron microscopy. Comparison of the number of neurofilaments within both large and small caliber myelinated axons revealed an increase in the density of neurofilaments in the transgenic axons (Fig. 3D,E). By carefully counting NF number, we determined that the NF density in transgenic axons was approximately 2.2 fold higher (88±25 NFμm−2 in transgenic axons and 40±20 NF,ííHI−2 in controls). In transgenic axons a number of the NF were not uniformly aligned parallel to the longitudinal axis of the axon (see Fig. 3E). Since we did not count these nonaligned neurofilaments, the differences in NF content between control and transgenic axons is actually greater. This increase in NF density is seen in essentially all myelinated axons of transgenic animals. Since the sciatic nerve contains both sensory and motor axons, we conclude that NF-L content is increased in both sensory and motor fibers.
In view of the preceding evidence suggesting an integral role played by NF in establishing axonal caliber, we compared the mean axonal diameters in the sciatic nerve of control animals and in transgenic animals expressing 3-4 fold elevated amounts of NF-L. Axonal cross-sectional areas were measured for 5500 normal and transgenic axons and mean diameters were calculated. Histograms of the frequency of axonal diameters revealed no significant changes in size between normal and transgenic axons (Fig. 4). These findings demonstrate that an increase in filaments assembled primarily from NF-L has little effect on axonal caliber. Of course, we cannot exclude the possibility that a greater increase in NF-L accumulation would ultimately have resulted in an increase in axonal caliber, or other axonal properties. This question may now be tested directly by mating different lines of the existing transgenic animals in an effort to elevate NF-L expression further.
Identification of dominant assembly disrupting mutants of NF-L and NF-M
As an alternative method for altering the normal level of assembled NF, we sought to identify mutants in NF-L or NF-M that could disrupt normal assembly of wild-type proteins. Such mutants, if they can be identified, could then be introduced into transgenic animals to disrupt the assembly of wild-type proteins and any effect on radial growth, myelination, or other axonal property could then be followed.
In an initial search for such mutants, we generated a set of amino- and carboxy-terminal truncation mutants of both NF-L and NF-M. The assembly products of the corresponding mutants were analyzed following DNA transfection into cells expressing either wild type NF-L or wild type vimentin, an intermediate filament subunit with which both NF-L and NF-M co-assemble during normal neurite outgrowth (Bignami et al. 1982). In a first series of such mutants the influence of the 142 amino acid tail of NF-L was examined by constructing a set of chimeric genes containing a strong promoter element from murine sarcoma virus, a cDNA sequence containing the coding domain of mouse NF-L but truncated to varying extents within the carboxy terminal coding domain. To facilitate detection of the product of each mutant gene, we tagged the carboxy terminus of each polypeptide with a 12 amino acid segment (from the human c-myc polypeptide) to which a specific monoclonal antibody is available (Evan et al. 1985). The mutant constructs, schematically drawn in Fig. 5, are denoted by NFL-CAx, where x refers to the number of amino acids deleted from the carboxy-terminus of NF-L.
To confirm that the expected protein product did accumulate in cells after transient transfection of each mutant gene, immunoblotting was used to analyze total cellular proteins 40 h post-transfection. Recipient cells were either mouse fibroblasts (L cells) that endogenously express vimentin, or mouse fibroblasts (NF9 cells) that stably produce wild-type mouse NF-L as the most abundant cell protein as a consequence of permanent DNA transfection with an MSV promoted wild-type NF-L gene (Monteiro and Cleveland, 1989). The blots revealed that appropriately sized mutant polypeptides accumulated in each case, with an average molar proportion of mutant to wild type vimentin or wild type NF-L of between 1:10 and 1:40. To determine whether a mutant subunit was competent for co-assembly with the wild type subunits and whether it affected the assembly products of the wild type, we used double immunofluorescence to localize both wildtype and mutant NF-L polypeptides simultaneously. An example of this is shown in Fig. 6. In cells transfected with NFL-CA11, which is missing only the carboxy-terminal 11 residues of the tail, we consistently found co-assembly of mutant with wild-type NF-L (Fig. 6A,B). However, deletion of about half of the tail domain yielded a mutant subunit that failed to assemble in most transfected cells and which partially disrupted the array of wild-type NF-L filaments (Fig. 6C,D). This was an unexpected finding since the structurally conserved rod domain (rather than the tail) was thought to be the major element required for filament assembly. Deletion of additional tail sequences invariably yielded assembly-incompetent subunits that disrupted the network of wild type NF-L filaments (Fig. 6E,F, showing mutant NFL-CÁ164 that is deleted in the entire tail domain as well as 22 amino acids of the rod domain). We conclude that efficient co-assembly of mutant NF-L subunits requires at least 80 amino acids of the proximal tail domain in addition to an intact rod segment. Larger truncations of the tail and all truncations into the rod yield assembly-incompetent subunits that disrupt wild-type subunit assembly even when present at only a few percent of wild type.
To investigate the assembly properties of the NF-M polypeptide, we generated a similar set of carboxyterminal and amino-terminal mutants, starting from a mouse NF-M gene in which the presumptive NF-M transcriptional promoter was replaced with that from MSV (Monteiro and Cleveland, 1989). The resulting truncation mutations are presented schematically in Fig. 7, in which the amino-terminal mutations are indicated by NFM-NAy, and the carboxy-terminal truncations are denoted by NFM-CAx, where x and y represent the number of amino acids truncated from the tail and head, respectively. As before, all mutant polypeptides were tagged with the 12-amino acid myc epitope to provide an immunologic tag for uniquely following the corresponding mutant. To analyze the assembly characteristics of the corresponding polypeptides, transient DNA transfection was again used. NF-M truncated in 75 of the 103-amino acid head domain yielded a polypeptide that remained assembly-competent with vimentin (Fig. 8A and B) or wild-type NF-L (not shown). Similarly, truncation of 391 amino acids of the 438-residue tail domain of NF-M yielded a polypeptide that remained assembly-competent (Fig. 8C,D). However, although subunits truncated to leave 28 amino acids of the head (NFM-NA75) or 47 amino acids of the tail (NFM-CA391) retained full co-assembly competence, the influence of both head and tail sequences on assembly properties was clearly demonstrated by combining both amino and carboxy-terminal deletions. The resulting construct, NFM-NA75/CA391, produced a dominant phenotype that was reminiscent of more severe carboxy-terminal mutants. As shown by the cell at the right in Fig. 8E,F, the mutant polypeptide was co-localized into a disrupted array that contained punctate cytoplasmic aggregates and a collapsed perinuclear mass. Similar results were obtained when the same mutant was transfected into cells expressing high levels of wild-type NF-L (data not shown). Quantitative immunoblotting analysis demonstrated that disruption was obtained even when the mutant protein was present at about 20 % of the wild type subunits.
These findings demonstrate that non-helical domains do play important roles in NF-M assembly. More importantly, along with the companion mutations in NF-L, they identify all carboxy-terminal truncation mutations near to or within the rod to be potent assembly disruptors. Given the success we have already had in obtaining high levels of expression of NF transgenes in the peripheral nervous system, we believe that if mice expressing one of these mutants can be obtained this will allow an unambiguous test of the consequences of disruption of normal NF arrays.
DISCUSSION
Within the wide spectrum of axonal diameters occurring in mammalian nerve fibers, each class of neurons has a relatively restricted range of axonal calibers. This is of functional significance not only because diameter is the principle determinant of conduction velocity in myelinated fibers, but because diameter per se may be a primary stimulus that directs which axon is to become myelinated. Several present lines of evidence have suggested that NF accumulation is itself a primary determinant of caliber in large myelinated fibers. In normal nerve fibers NF are the most numerous cytoskeletal elements and NF density remains constant over a wide range of calibers, thus yielding a close correlation between axonal cross-sectional area and the axonal NF content. Although modest differences in axonal NF density (—25%) have been reported among different classes of neurons (Price et al. 1988) and in different regions of smaller caliber nerve fibers (Nixon and Longvinenko, 1986; Parhad et al. 1987), greater variations of NF density are not normally observed. Even in those pathologic cases where substantial increases in NF density are found, such as induced by chronic intoxication with iminodipropionitrile (IDPN) (Parhad et al. 1987), an agent that through an unknown mechanism selectively impairs NF transport (Griffin et al.1978), elevation of NF content is accompanied by large increases in axonal caliber. In this instance, it seems quite plausible to propose that initial accumulation of excess filaments is accompanied by radial growth, but that ultimately other factors (such as the inability of the associated myelin sheath to accommodate further axonal distension) may disrupt the normal linkage of NF content with axonal caliber.
A final, and perhaps most compelling piece of evidence supporting the role of NF in controlling NF caliber has come from our demonstration that reductions in NF gene expression after axotomy (see Fig. 2) are associated with decreased axonal caliber in large myelinated fibers. This is also supported by evidence that radial growth of myelinated fibers, which occurs during postnatal development, coincides temporally with postnatal increases in NF gene expression in large sensory neurons. This increase is absent in the small neurons of unmyelinated fibers, that do not undergo significant radial growth (Muma et al. 1991). Although the exact mechanisms by which NF content influences axonal caliber are unknown, these observations collectively support the hypothesis that NF gene expression is, at the least, a prerequisite for the radial growth of axons.
This apparent role of NF in the control of axonal caliber is not shared by peripherin, the class III IF protein expressed in some peripheral neurons (Parysek and Goldman, 1988). In contrast to NF, peripherin is expressed at relatively high levels (in comparison to mature neurons) during both early development (Seymour and Oblinger, 1988) and axon regeneration (Oblinger et al. 1989). These findings indicate that the radial growth of axons during postnatal development correlates with a decline in the abundance of peripherin in neurons, and the somatofugal atrophy of regenerating axons correlates with increased expression of peripherin. Furthermore, unlike NF, which are expressed at highest levels in neurons with large-caliber myelinated axons (e.g. large sensory neurons), peripherin is preferentially expressed in neurons with small-caliber unmyelinated axons (e.g. small sensory neurons) (Escurat et al. 1990). These comparisons lend further support to the hypothesis that NF play a distinct role in the control of axonal caliber.
We have focused on two methods to test directly the role of NF in axonal caliber, or indeed in other axonal properties. In the first of these, we have introduced wildtype NF-L genes into transgenic mice to elevate the level of NF protein. This has proven successful but the outcome seems at odds with the initial prediction: NF-L content in peripheral axons was elevated fourfold, with a corresponding increase in filament number, but no effect on axonal caliber was been detected. Since we have also shown that there is no corresponding change in NF-M and NF-H expression (for example, see Fig. 3), our findings demonstrate that an increase in filament assembled primarily from NF-L has little effect on axonal caliber. While we cannot exclude the possibility that a greater increase in NF-L accumulation would ultimately have resulted in altered axonal caliber (as happens in animals intoxicated with iminodipropionitrile), a simpler view is that filaments constructed primarily of NF-L alone do not have the space-determining properties of authentic filaments assembled from normal levels of all three subunits.
Indeed, in this regard, of particular interest are the extended carboxy-terminal tail domains shared by NF-M and NF-H, a feature not shared by other IF proteins (with the exception of nestin). For example, the lack of a similar carboxy-terminal domain for GFAP, the IF protein of fibrous astrocytes, correlates with the absence of interfilament cross-bridges and the close spacing of adjacent filaments (Hirokawa et al. 1984). In the case of NF-H, it has been argued that the tail is associated with NF crossbridges (Hirokawa et al. 1984; Hisanaga and Hirok-awa, 1988). However, since the tail is highly enriched in both charged amino acids and carries 50 potential phosphorylation sites in a repeated domain, rather than a role in such potential cross-bridging, it is as likely that these extended, highly charged arms establish minimum inter-filament distances. Certainly such repulsive interactions could serve to establish spatial properties, such as the nearly constant filament density seen in myelinated axons. In any case, whether through cross-bridging or repulsion or a combination of such interactions, an integral role of the tail domains in the control of axonal caliber seems highly likely. This can now be explored directly by producing additional transgenic mice expressing transgenes encoding NF-M and NF-H subunits. We have recently succeeded in producing such mice and by mating those animals with our existing NF-L overexpressing lines, a definitive answer as to the morphologic and phenotypic consequence of over-expressing wild-type NFs should be forthcoming shortly.
Our second approach to investigating NF function in vivo has focused on the identification and utilization of dominant assembly-disrupting mutants of NF subunits. We have found that such mutants can be easily produced for both NF-L and NF-M by deletion into the conserved, carboxy-terminal region of the œ-helical rod domain. Introduction of such mutant genes into transgenic animals should now allow a second direct test of the consequence on axonal caliber, or on other axonal functions, of disrupting the normal assembly properties of NF filaments in vivo. Animals expressing such transgenes have recently been obtained using transgene constructs bearing either a neuron-specific promoter (from the NF genes themselves) or the stronger promoter from murine sarcoma virus. Analysis of the phenotypes of these animals should yield substantial insight as to the effect of disrupting normal NF assembly during axonal extension and/or radial growth.
ACKNOWLEDGEMENTS
This work has been supported by grants from the National Institutes of Health to D.W.C. and P.N.H. and by a grant from the March of Dimes to D.W.C.