Disruption of the NF-H Gene Increases Axonal Microtubule Content and Velocity of Neurofilament Transport: Relief of Axonopathy Resulting from the Toxin β,β′-Iminodipropionitrile

The mouse genomic NF-H DNA was isolated from a mouse phage genomic library as described previously (Julien et al., 1988). The targeting vector was constructed by insertion of a blunt-ended 1.1-kb XhoI and BamHI fragment of pMC1neoPoly A (Stratagene, La Jolla, CA) into the second XmaI site of the NF-H gene as shown in the Fig. 1 A. The final targeting fragment was excised out by BamHI digestion and then purified by b-Agarase (GIBCO BRL, Gaithersburg, MD).

The culture of R1 embryonic stem (ES) cells (a gift from J. Roder, Mount Sinai Research Institute, Toronto, Canada) and transfection were carried out as previously described (Mansour et al., 1988; Ramirez-Solis et al., 1993). ES cells (10⁷) were transfected with 25 mg of purified targeting DNA using the BRL Cell Porator at 330 mF 230 V in DME at room temperature. The cells were treated with 180 mg/ml of G418 at 24 h after electroporation for 3 d, and then the concentration of G418 was increased to 360 mg/ml. After 7–8 d under G418 selection, colonies were picked up, trypsinized, and then plated onto 96-well plates with SNL feeder layer (feeder layer and LIF 1,000 units/ml were used all time in the ES cell culture). Cells were trypsinized 4 d later and 30% of the cells were plated for DNA extraction onto gelatin-coated 96-well plates while 70% of them were frozen in duplicate plates. DNA was extracted after 5 d of culture and digested with EcoRI in the plates (Ramirez-Solis et al., 1993). The ES cell clones that yielded an expected 2.8-kb recombinant band on Southern blot (see Fig. 1) were expanded and reprobed by Neo probe to exclude the clones with random insertions. Three ES cell clones were injected into C57BL/6 blastocysts as described before (Mansour et al., 1988). Germline transmission was obtained by mating male chimeras with C57BL/6 females. The use of animals and all surgical procedures described in this article were carried out according to The Guide to the Care and Use of Experimental Animals of the Canadian Council on Animal Care (see http://www.pharma.mcgill.ca/grant.htm).

Mice were killed by pentobarbital overdose and tissue samples were immediately frozen after dissection in liquid nitrogen and stored at −80°C. Total RNA was isolated by homogenization in guanidinium thiocyanate according to the method of Chomczynski and Sacchi (1987). The RNA samples (20 mg) were fractionated on a 1% agarose–formaldehyde gel before blotting on a Hybond N⁺ membrane (Amersham Corp., Arlington Heights, IL). The DNA probes were labeled with [α-³²P]dATP by random priming. Hybridizations were performed for 16 h at 42°C in 50% formamide, 1 M NaCl, 50 mM Tris, pH 7.5, 0.2% Ficoll, 0.2% polyvinyl pyrrolidone, 0.1% sodium pyrophosphate, 1% SDS, and 100 mg/ml sonicated and denatured salmon sperm DNA. Blots were washed in 2× SSC–0.1% SDS at room temperature for 10 min, in 2× SSC–0.1% SDS at 65°C twice for 10 min, and in 0.5× SSC–0.1% SDS at room temperature once for 15 min. Blots were exposed to film (X-OMat AR; Eastman Kodak, Rochester, NY) with intensifying screens for 18 h at −80°C.

The mice were killed by intra-peritoneal injection of overdose pentobarbital and the nervous tissue samples were frozen in liquid nitrogen and stored at −80°C. Total protein extracts were obtained by homogenization of samples in SDS–urea buffer (0.5% SDS, 8 M urea in 7.4 phosphate buffer). The supernatant was collected after centrifugation at 10,000 g for 20 min. The protein concentration was estimated by the Bradford procedure (Bio-Rad Laboratories, Hercules, CA). The proteins (10 mg) were fractionated on a 7.5% SDS-PAGE and either visualized by Coomassie blue staining or blotted onto a nitrocellulose membrane for Western blot analysis. The mAbs against NF-L and NF-H were purchased from Amersham Corp. and Sternberger Inc. (Lutherville, MD), respectively. The mAbs against tau, tubulin, and actin were purchased from Boehringer Mannheim Corp. (Indianapolis, IN). The mAb against acetylated tubulin was purchased from Sigma Chemical Co. (St. Louis, MO). The Western blot results were revealed by Renaissance^® (a Western blot chemiluminescence kit from Dupont-NEN™, Boston, MA).

Five sets of 2–3-mo-old NF-H +/+, NF-H +/−, and NF-H −/− littermates were intoxicated with two doses (first day and fifth day) of IDPN intraperitoneally (1.5 g/kg body weight in 50% saline solution; Aldrich Chemical Co., Milwaukee, WI) and maintained on 0.02% IDPN in drinking water. IDPN-treated mice were analyzed at 0, 7, and 15 d after intoxication.

Mice were killed by overdose of pentobarbital, perfused with 0.9% NaCl, and then with fixative (2.5% glutaraldehyde, 0.5% PFA in 0.1 M sodium phosphate buffer, pH 7.4). Tissue samples were immersed in fixative for 2 h, rinsed in phosphate buffer, and then postfixed in 1% phosphate buffered osmium tetroxide. After three washes with phosphate buffer, each sample was dehydrated in a graded series of ethanol and embedded in Epon. The thin sections were stained with Toluidine blue and examined under a Polyvar microscope. The counting of axons and evaluation of myelin thickness in the L5 ventral root were carried out with the Image-1 software from Universal Imaging Corporation (West Chester, PA). The ultrathin sections were stained with lead citrate and examined with a Philips CM10 electron microscope (Philips Electron Optics, Mahwah, NJ).

Ultrathin cross sections (100 nm) of L5 ventral root axons (>5 mm in diam) at the dorsal root ganglia (DRG) region were photographed at a magnification of 11,500× and enlarged to a final magnification of 81,650× by printing. Positions of 10-nm filaments were marked by puncturing the print with a needle, and then captured by a CCD camera and digitized using Imagine I. Digitized x-y coordinates were then computed by an IBM program written in C++ programming language. Nearest neighbor distance was calculated for each filament. Neurofilament packing density was calculated using the tile-counting procedure as previously reported (Hsieh et al., 1994) with the following modifications. Square tiles of 100 nm × 100 nm were used, and tiles that did not contain any end-on neurofilaments were eliminated by the program.

The density and distribution of microtubules vary substantially according to the type of neurons, axonal caliber size, and myelination status of the axon (Hsieh et al., 1994). Analysis of microtubules was carried out as follows. Ultrathin cross sections (100 nm) of L5 ventral root axons of 5.0– 5.5 mm in diameter at the level of DRG region having a clear myelin sheath were photographed at a magnification of 3,100× and enlarged another 8 times by printing. The position of 25-nm microtubules were taken in the same way as described above for neurofilament calculation. Axon area for the density calculation did not include myelin sheath. All calculations were done by a computer program with a graphic display written specifically for this application.

Sciatic nerves between DRG and obturator tendon were dissected out from newly killed mice and homogenized in a 2-ml glass-to-glass tissue grinder in 1 ml of taxol/Triton X-100 PHEM extraction medium (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, 2 mM MgCl, 0.1% DMSO 10 μg/ml leupeptin, and 0.6 mM PMSF) (Black et al., 1986). The homogenates were incubated at room temperature for 30 min to solubilize unassembled tubulin and then centrifuged for 45 min in a Beckman TLA 45 rotor at 30,000 rpm. The pellets were dissolved in 100 μl of 1% SDS-Tris buffer (50 mM Tris, pH 7.4) and the protein content was estimated by the Bio-Rad Dc Protein Assay kit (Bio-Rad Modified Lowry method). The proteins (10 μg) were fractionated on a 7.5% SDS-PAGE and either visualized by Coomassie blue staining or blotted on a nitrocellulose membrane for Western blot analysis. Densitometric analysis was carried out by the Gel-Pro™ Analyzer (version 2.0) from Meia cybernetics (Silver Spring, MD).

Proteins carried by slow axonal transport in sciatic motor axons were metabolically labeled and fractionated mainly as previously reported (Kirkpatrick and Brady, 1994). Aliquots of 0.5 mCi of [³⁵S]methionine (DuPont-NEN) in 1-ml vol were injected into the anterior horn area of L2-L4 spinal cord twice on each side, 1-mm deep from the dorsal surface and 1 mm from the middle groove at a rate of 0.05 ml/min. 30 d after labeling, the sciatic nerves with ventral roots (L3 to L5) were dissected out from both sides and cut into 3-mm consecutive segments. Proteins were fractionated by SDS-PAGE and radiolabeled proteins detected by fluorography.

A genomic fragment of NF-H was isolated from a mouse library using a mouse cDNA probe (Julien et al., 1986). The mouse NF-H gene contains four exons separated by three introns (Julien et al., 1988). A targeting vector of ∼8 kb was constructed by inserting a TK/Neo cassette into the XmaI site of exon I as shown in Fig. 1 A. The DNA fragment was transfected by electroporation into R1 ES cells and neomycin resistance colonies were selected (Nagy et al., 1993). A total of 288 G418-resistant clones were picked up. Three ES cell clones out of eleven selected for homologous recombination event were microinjected into mouse blastocysts to generate chimeric mice. The majority of chimeras derived from this ES cell line were males and they were able to transmit agouti coat color to their progeny when mated to C57BL/6 females. About half of agouti progeny (F1) were heterozygous for the NF-H null mutation. The genotype of the mice was determined by Southern blot analysis of tail DNA. A Mendelian transmission of the NF-H disrupted gene was obtained by the breeding of heterozygous NF-H +/− F1 males and females. Thus, the NF-H null mutation in mice does not cause lethality during embryonic development or after birth. The homozygous NF-H knock-out mice were fertile and appeared phenotypically normal.

Northern blot analysis was carried out with 10 μg of total brain mRNA of adult NF-L +/+, +/−, and −/− mice (Fig. 1,B). No NF-H mRNA was detected in NF-H knockout mice while ∼50% decrease in NF-H mRNA was detected in NF-H heterozygous +/− mice. The depletion of NF-H mRNA did not affect the level of expression of either NF-M or NF-L mRNAs. Coomassie blue–stained gels of Triton X-100–insoluble extracts at 4°C from the spinal cord and sciatic nerve showed absence of NF-H protein in NF-H −/− mice (Fig. 1,C). This was further confirmed by immunoblotting (Fig. 1,D). No band corresponding to NF-H was detected with SMI-31 and SMI-32 antibodies (Sternberger Inc.) that recognizes the hyperphosphorylated and hypophosphorylated forms of NF-H, respectively. The lack of NF-H protein did not affect the levels of NF-L protein but it resulted in a modest increase of ∼20% in NF-M levels (Fig. 1,C). Thus, unlike the situation with NF-L knockout mice where dramatic decreases in the levels of both NF-M and NF-H proteins were observed (Zhu et al., 1997), the absence of NF-H did not cause dramatic changes on the stability of the other two neurofilament subunits, NF-M and NF-L. However, it is noteworthy that the loss of NF-H protein increased dramatically the signal of NF-M cross-reactivity with the SMI-31 antibody (Fig. 1 D). This is likely due to a compensatory increase in the phosphorylation state of NF-M that contains, like NF-H, multiple Lys-Ser-Pro phosphorylation sites (Levy et al., 1987; Myers et al., 1987).

Since there are some lines of evidence suggesting a role for NF-H protein in neurofilament spacing and axonal caliber (deWaegh et al., 1992; Hsieh et al., 1994; Nixon et al., 1994), light and electron microscopy was carried out on L5 ventral roots of adult NF-H +/+, NF-H +/−, and NF-H −/− mice. Surprisingly, the NF-H depletion caused only modest changes in the caliber of myelinated axons (Figs. 2 and 3,A). Like NF-H +/+ mice, the NF-H −/− mice yielded a bimodal distribution of axonal calibers representing the small and large myelinated axons (Fig. 3,A). In the NF-H −/− mice, the caliber of large myelinated axons was only slightly smaller (∼10%) than those of normal mice. To examine whether interfilament spacing was affected in large ventral root axons lacking NF-H, we recorded the positions of each filament in individual axons and calculated nearest neighbor distances for each neurofilament. No significant changes in the number of assembled neurofilaments were observed in NF-H +/− and NF-H −/− mice (Fig. 3,B). The NF-H subunit was clearly dispensable for the formation of 10-nm filaments and its absence caused no substantial changes in the average nearest neighbor distance, neurofilament packing density and axon caliber (Figs. 3 and 4).

The most remarkable change in the axonal cytoskeleton of ventral root axons lacking NF-H was a ∼2.4-fold increase in microtubule content (Fig. 3 D). Considering that the density and distribution of microtubules may vary according to the type of neurons, axonal caliber size, and myelination status of the axon (Hsieh et al., 1994), our analysis was carried out on large myelinated axons (∼20 mm² ± 2.5 axoplasmic area) of the L5 ventral root, a well-defined population of axons. We counted the number of microtubules on electron micrographs of individual transverse sections of axons magnified 25,000×. Our study involved four pairs of NF-H +/+ and NF-H −/− littermates with two sets derived from three distinct ES cell clones and, for each animal, 7–24 axons were analyzed. The number of microtubules per axonal area (20 mm²) was of 193 ± 86 (n = 50) in myelinated axons from control littermates and of 463 ± 94 (n = 58) in the NF-H knockout mice.

Taxol-stabilized cytoskeletal preparations from the sciatic nerve were analyzed to further confirm whether the levels of assembled tubulin were increased. Sciatic nerves between DRG and obturator tendon were dissected and homogenized in taxol/Triton X-100 extraction buffer as described previously (Black et al., 1986). After 30 min at room temperature, the homogenates were centrifuged at 45,000 g for 45 min. The pellets were dissolved in 1% SDS–Tris buffer (50 mM Tris, pH 7.4) and the proteins fractionated on SDS-PAGE. As shown on the Coomassie blue–stained gel of these cytoskeletal preparations and corresponding histograms in Fig. 5, the levels of assembled tubulin was increased of ∼2.5-fold in samples from the NF-H knockout mice as compared with normal mice. The rise in assembled microtubules in axons lacking NF-H was not due to an increase of total tubulin levels in the sciatic nerve (Fig. 1,D) implying that the absence of NF-H contributed to increase the stability of microtubules. We examined by immunoblotting using the TAU-1 antibody whether the levels of tau, a microtubule-associated protein (MAP) that suppresses microtubule dynamics, were altered in the sciatic nerve of NF-H −/− mice. No changes in levels of tau were evident in samples from total nerve extracts (data not shown). However, the TAU-1 antibody detected elevated levels of low molecular weight tau in cytoskeletal preparations from the sciatic nerve of NF-H −/− mice (Fig. 6). From these results, it is possible that NF-H could affect microtubule stability through competition with tau. Thus, a depletion of NF-H levels might yield more binding sites for tau on the surface of microtubules resulting into extra stabilization of axonal microtubules.

Since overexpression of mouse or human NF-H genes in transgenic mice slowed axonal transport of neurofilaments (Collard et al., 1995; Marszalek et al., 1996), we examined whether the velocity of axonal transport of NF-L and NF-M proteins would be increased in NF-H–deficient mice. This was done by the injection of [³⁵S]methionine into the L4-L5 spinal cord with subsequent analysis after 30 d of incorporated radioactivity for NF-L and NF-M proteins in 3-mm segments of the sciatic nerve. In normal mice, at 30 d after injection, the peak of incorporated radioactivity for the NF-L and NF-M proteins occurred in the 9-mm segment of the sciatic nerve while the leading edge of radiolabeled neurofilament proteins was found at the 15-mm segment corresponding to a velocity of ∼0.5 mm/d (Fig. 7). In contrast, for the NF-H −/− mice, the leading edge of radiolabeled NF-L and NF-M proteins was found at the 21-mm segment implying a velocity of ∼0.7 mm/d (Fig. 7,C). To facilitate comparison, the signals were quantified by image analysis and converted to percentage of total radioactivity (Fig. 7, D–F). Clearly, the peaks of radiolabeled NF-L and NF-M proteins are broadened in the NF-H −/− axons as compared with NF-H +/+ axons. Whereas <5% of radiolabeled NF-L and NF-M proteins were detected in the 15-mm sciatic segment of normal mice, ∼25% of newly synthesized NF-L and NF-M migrated to the 15-mm segment in NF-H −/− mice. In contrast, no dramatic changes in the velocity of NF-M and NF-L transport occurred in the NF-H +/− mice (Fig. 7, B and E).

IDPN is a toxic agent known to segregate microtubules from neurofilaments resulting in the excessive accumulation of neurofilaments in the proximal part of the axon and with microtubules forming a central grouping in axons (Griffin et al., 1978; Parhad et al., 1988). The availability of NF-H knockout mice provided us a unique opportunity to assess the role of NF-H in axonopathy induced by this neurotoxin. Intoxication of normal mice by IDPN led to the formation of prominent neurofilamentous swellings in proximal axons originating from spinal motor neurons (Fig. 8, B and C). In contrast, after IDPN treatment, littermates heterozygous (Fig. 8, E and F) and homozygous (Fig. 8, H and I) for the NF-H null mutation did not develop such axonal swellings in motor axons.

Electron microscopic studies of transverse sections of the L5 ventral root axons within the DRG, 7 d after IDPN intoxication, revealed the absence of IDPN-induced segregation of microtubules from neurofilaments in NF-H −/− mice (Fig. 9). We conclude from these results that NF-H is a key mediator of motor axonopathy induced by IDPN exposure.

Our analysis of NF-H knockout mice provides direct evidence for roles of NF-H protein as a determinant of microtubule density in axons, as a modulator of velocity of neurofilament protein transport in axons and as a key mediator of axonopathy induced by IDPN intoxication. A number of previous reports suggested that neurofilaments can interact between each other and with microtubules via the COOH-terminal domains of NF-H or NF-M but the significance of such interactions has remained elusive (Hisanaga et al., 1990, 1993). The presence of highly charged amino acids in the long tail domains of NF-H also led to the suggestion that this subunit would be implicated as regulator of neurofilament packing density and perhaps control axonal caliber (deWaegh et al., 1992; Hsieh et al., 1994; Nixon et al., 1994). Evidence for this notion has come from studies on the dysmyelinating mutant Trembler mouse and on hypomyelinating transgenic mice expressing in Schwann cells either a diphtheria toxin A or SV-40 large T antigen in which reduced levels of NF-H phosphorylation resulted in decreases in axon caliber (deWaegh et al., 1992; Cole et al., 1994). Therefore, as the NF-H subunit is dispensable for the formation of a neurofilament network, it is rather surprising that the disruption of the NF-H gene had so little effects on the spacing density of neurofilaments and caliber of axons. These results can reflect in part the existence of compensatory mechanisms. For instance, it is plausible that an hyperphosphorylation of the NF-M subunit, as revealed by stronger immunodetection signals with the SMI-31 antibody (Fig. 1 D), can compensate in part for NF-H function. Unlike NF-H −/− mice, the NF-M −/− mice have decreased levels of NF-L and dramatic reduction of axonal calibers (Elder et al., 1998; our unpublished observations). Therefore, it is clear that the NF-H and NF-M proteins are not playing equivalent roles in neurofilament function. The combined studies with NF-M −/− and NF-H −/− mice indicate that NF-M rather than NF-H protein is a key modulator of axonal caliber.

Another unexpected consequence of NF-H gene disruption was the dramatic increase of ∼2.4-fold in the density of microtubules in ventral root axons (Fig. 3,D). This was further corroborated by similar increases in levels of assembled acetylated tubulin detected in cytoskeletal preparations from the sciatic nerve (Fig. 5). How could microtubule density in axons be regulated by the NF-H subunit? One possibility, is that the absence of NF-H could somehow increase the rate of axonal transport of tubulin protein or of preassembled microtubules from the cell body. However, the rise in assembled microtubules in NF-H −/− axons was not accompanied by a corresponding increase in the levels of total axonal tubulin (Fig. 1,D). This suggests that the lack of NF-H contributed to increase the stability of microtubules. One plausible mechanism by which NF-H could affect microtubule stability is through competition with tau, a microtubule-associated protein (MAP) that suppresses microtubule dynamics (Miyasaka et al., 1993). In vitro binding affinity assays demonstrated that both hypophosphorylated NF-H and tau, compete for similar binding sites on the COOH-terminal region of tubulin (Miyasaka et al., 1993). Therefore, it is conceivable that a depletion of NF-H levels might yield more binding sites for tau on the surface of microtubules resulting into extra stabilization of axonal microtubules. This view is supported by the increased immunodetection of low molecular weight tau with TAU-1 mAb on immunoblots of cytoskeletal preparations from the sciatic nerve of NF-H −/− mice (Fig. 6).

Recent studies with transgenic mice overexpressing murine NF-H led to the suggestion that NF-H can influence the rate of neurofilament transport in axons (Marszalek et al., 1996). The axonal transport studies shown here are consistent with this view. The absence of NF-H increased the velocity of axonal transport of NF-M and NF-L proteins by ∼0.2 mm/d (Fig. 7). While this is not a trivial change, this is much less than an eightfold change of velocity suggested by studies of slow axonal transport during development (Willard and Simon, 1983). The exact molecular mechanism by which NF-H can affect neurofilament transport remains unknown. It has been proposed that the NF-H could shift an equilibrium toward assembly of stationary neurofilaments. The effect of NF-H on the mobility of neurofilament proteins could be indirect. The NF-M and NF-H proteins can compete for heterodimer formation with NF-L (Athlan and Mushynski, 1997). Overexpressing NF-H provoked decreases in NF-M levels (Collard et al., 1995; Marszalek et al., 1996) while the absence of NF-H levels resulted in a ∼20% increase in NF-M levels. Therefore, one possibility is that changes in levels NF-H affect the levels of NF-M which in turn could be a key regulator of NF-L transport. Future axonal transport studies with NF-M knockout mice and double NF-H; NF-M knockout mice should help to verify this hypothesis.

The IDPN intoxication studies described here further support the view that neurofilaments and microtubules are highly interdependent structures and that the NF-H subunit plays a key role in mediating these interactions. It is remarkable that motor axonopathy due to IDPN intoxication was mitigated not only in NF-H −/− mice but also in NF-H +/− mice. Thus, a ∼50% decrease in the levels of NF-H protein was sufficient to alter the properties of axonal cytoskeleton and to confer resistance to IDPN intoxication even though little effect was observed in axonal transport. A key role for NF-H as a mediator of IDPN toxicity is consistent with the report of a transient changes in the phosphorylation state of this protein preceding the onset of segregation of neurofilaments from microtubules and neurofilament transport impairment (Tashiro et al., 1994). While NF-H is clearly acting as a key mediator of IDPN-induced axonopathy, other molecules involved in the phenomenon remain to be defined. Some of these factors could be cell-type specific as neurofilamentous swellings were still detectable in DRG sensory axons of both NF-H +/− and NF-H −/− after IDPN treatment while motor axons were exempt of swellings (data not shown).

Our finding that NF-H is a modulator of microtubule density and a key mediator in IDPN-induced axonopathy may be of relevance to neurodegenerative diseases involving cytoskeletal abnormalities. The abnormal neurofilament depositions in perikaya and neuronal processes are frequently observed in neurodegenerative diseases such as ALS (Carpenter, 1968; Chou, 1992; Hirano et al., 1984), Parkinson's disease (Schmidt et al., 1991) and toxic neuronopathies induced by neurotoxin exposure (Asbury and Johnson, 1978; Graham et al., 1984; Griffin et al., 1978; Troncoso, 1992). Transgenic mouse studies provided compelling evidence that a disorganization of the neurofilament network can play a causative role in the selective degeneration of motor neurons (Côté et al., 1993; Xu et al., 1993; Lee et al., 1994). Studies with transgenic mice overexpressing human NF-H led to the suggestion that the toxicity of disorganized neurofilaments is due to a general disruption of axonal transport (Collard et al., 1995). While transport impairment has been interpreted as reflecting a physical block by neurofilament accumulations, the data presented here suggest that changes in NF-H levels can also perturb the dynamics and function of microtubules, which are key organelles of intracellular transport.

It is noteworthy that the lack of NF-H also resulted in a decreased levels of assembled actin in extracts form the sciatic nerve (Fig. 5 A). The NF-H protein was recently found to be connected to actin filaments in sensory neurons by a linker protein encoded by a neuronal splice form of the BPAG1 gene (Yang et al., 1996), which is responsible for the autosomal recessive dystonia musculorum. The large primary sensory axons in BPAG1 knockout mice degenerate within 4 wk after birth, clearly demonstrating that perturbation of the neurofilament-actin connections may have profound effects on axonal integrity. Future breeding studies with the NF-H −/− mice described here will provide a unique approach to assess the in vivo contribution of NF-H protein to BPAG1-mediated degeneration and to other neurodegenerative diseases such as motor neuron disease.

This work was supported by the American Health Assistance Foundation and the Amyotrophic Lateral Sclerosis (ALS) Association (USA). Q. Zhu was a recipient of a fellowship from the NeuroScience Network, and J.-P. Julien has a MRC senior scholarship.

ALS

amyotrophic lateral sclerosis

DRG

dorsal root ganglia

ES

embryonic stem

IDPN

β,β′-iminodipropionitrile

NF-H/NF-L/NF-M

neurofilament heavy/neurofilament light/ neurofilament medium