Selective loss of neurofilament expression in Cu/Zn superoxide dismutase (SOD1) linked amyotrophic lateral sclerosis


Address correspondence and reprint requests to Pamela J. Shaw, Academic Neurology Unit, The Medical School, University of Sheffield, Beech Hill Road, Sheffield, South Yorkshire S10 2RX, UK. E-mail:


Neurofilament pathology is a hallmark of sporadic and familial amyotrophic lateral sclerosis (SALS and FALS). The disease mechanisms underlying this pathology are presently unclear, but recent evidence in SALS patients suggest that reductions in neurofilament light subunit (NFL) mRNA may contribute to the death of motor neurones. Mutations in the gene encoding Cu-Zn superoxide dismutase (SOD1) represent the best-studied cause of FALS, and a number of laboratory models of SOD1-mediated disease exist. Here we have used microdissected lumbar spinal cord motor neurones from human SOD1 FALS patients as well as G93A SOD1 transgenic mice and demonstrated that reduced NFL mRNA levels are seen in both. To probe the molecular mechanisms underpinning these observations, we generated NSC34 motor neurone-like cell lines expressing wild-type and mutant SOD1. NSC34 cells expressing G37R or G93A SOD1 showed selective reductions in NFL and NFM mRNA and protein. These data suggest that NFL mRNA reductions are common to SALS and FALS patients, and that cells and mice expressing mutant SOD1 may enable us to characterize the molecular mechanism(s) responsible for the loss of neurofilament mRNA.

Abbreviations used

amyotrophic lateral sclerosis




diethyl pyrocarbonate


Dulbecco's modified Eagle's medium


enhanced chemiluminesence


familial amyotrophic lateral sclerosis


glyceraldehydephosphate dehydrogenase mRNA messenger RNA




neurofilament heavy subunit


neurofilament light subunit


neurofilament medium subunit


normal Cu-Zn superoxide dismutase


phosphate-buffered saline


polyvinylidene difluoride


room temperature


sporadic amyotrophic lateral sclerosis


sodium dodecyl sulfate


Cu/Zn superoxide dismutase


Tris-buffered saline.

Amyotrophic lateral sclerosis (ALS) is a late onset neurodegenerative disease characterized by the degeneration and cell death of motor neurones of the spinal cord, brain stem and motor cortex. Although 90% of cases are sporadic, the remainder are familial and, of these, 20% are due to mutations in the gene encoding Cu-Zn superoxide dismutase (SOD1; Rosen et al. 1993). A common feature of the pathology of both sporadic and familial forms of the disease, including those with SOD1 mutations, is the presence of axonal and perikaryal neurofilament accumulations, suggesting a role for neurofilaments in the pathogenesis of ALS (Carpenter 1968; Hirano et al. 1984; Rouleau et al. 1996; Ince et al. 1998). Neurofilament proteins are members of the predominant intermediate filaments expressed in large calibre neurones. The three subunits, neurofilament light (NFL), medium (NFM) and heavy (NFH) are 68 kDa, 115 kDa and 160 kDa, respectively. All three proteins are highly expressed in motor neurones, which is consistent with their potential to contribute to the development of ALS.

Several laboratories have reported NFH mutations in sporadic ALS patients (Figlewicz et al. 1994; Tomkins et al. 1998; Al-Chalabi et al. 1999). These mutations are small in-frame insertions or deletions and are located in the large NFH side-arm domain. Combining data from several studies demonstrates that these mutations occur in 1% of the approx. 1300 cases examined. It is unlikely that NFH mutations are of sufficient penetrance to cause ALS; however, they may represent a risk factor for the disease. It is noteworthy that the mutations are located within the region of NFH containing multiple Lys-Ser-Pro repeats. These sites are a consensus for proline-directed serine/threonine protein kinases (Veeranna et al. 1998; Li et al. 1999; Ackerley et al. 2000; Brownlees et al. 2000). Interestingly, phosphorylated and non-phosphorylated neurofilaments are apparently partitioned within motor neurones in the axonal and perikaryal compartments, respectively (Peng et al. 1986). In ALS, increased immunoreactivity with antibodies recognizing phosphorylated forms of these proteins has been reported in the perikaryon of surviving motor neurones (Itoh et al. 1992). This suggests that defects in neurofilament phosphorylation or localization are involved in the degeneration of motor neurones seen in ALS.

The mechanisms responsible for the neurofilament-related changes observed in ALS are unknown. However, one possible mechanism is related to the relative expression levels of the different neurofilament proteins (Strong 1999; Julien and Beaulieu 2000). Studies in transgenic mice have confirmed that overexpressing or knocking out specific neurofilament proteins can affect the axonal calibre of motor neurones, and lead to the accumulation of neurofilaments in the perikaryon (reviewed in Julien and Beaulieu 2000). Therefore, changes apparently similar to those observed in ALS can arise as a result of aberrant neurofilament gene expression. In support of this as a disease mechanism, two groups have reported altered neurofilament mRNA levels in the spinal motor neurones of ALS patients. RNA in situ hybridization was used to demonstrate significant reduction in the expression of NFL in motor neurones, as well as α-internexin and peripherin, two other intermediate filament proteins (Bergeron et al. 1994; Wong et al. 2000). These changes were all reported in sporadic ALS cases without SOD1 mutations.

The expression of mutant SOD1 is a well-studied cause of ALS. We hypothesized that selective loss of neurofilament mRNA might also occur in mutant SOD1-mediated disease. Therefore we utilized models of SOD1-mediated ALS, as well as human post-mortem spinal cord tissue, to investigate this possibility. Here we now show: (i) NFL mRNA is depleted in microdissected human spinal motor neurones from ALS cases with the I113T SOD1 mutation; (ii) G93A mutant SOD1 transgenic mice (Gurney et al. 1994) show a significant reduction in NFL mRNA levels at later stages of disease (140 days) as well as a smaller reduction prior to onset of clinical signs (60 days); (iii) motor neurone-like NSC34 cell lines stably expressing mutant but not wild-type SOD1 also show reduced neurofilament mRNA and protein levels. These findings lend support to the hypothesis that loss of NFL is important in the pathogenesis of familial as well as sporadic ALS.

Materials and methods

Generation and maintenance of NSC34 cells

NSC34 cells were grown in Dulbecco's modified Eagle's medium (DMEM; high glucose, with pyruvate, with l-glutamine) containing 10% (v/v) heat-inactivated fetal bovine serum and 1% penicillin/streptomycin solution (Cookson et al. 1998). Generation of NSC34 cells stably expressing normal or mutant human SOD1 has been described elsewhere (Menzies et al. 2002).

SOD1 transgenic mice

Mouse lines transgenic for G93A mutant human SOD1 first described by Gurney et al. (1994) were purchased from Jackson Laboratories (West Grove, PA, USA) and maintained as described previously (Fray et al. 2001). Tissue was taken from mice prior to the onset of disease (60 days old) and at end stage (140 days old).

Laser capture microdissection

For investigation of mRNA expression levels, individual human and mouse motor neurones or dorsal horn neurones from the lumbar spinal cord were microdissected using a laser capture microdissector (Arcturus, Herts, UK). Prior to microdissection, 10-µm sections of snap-frozen blocks from the lumbar spinal cord were fixed (75% ethanol, 30 s), washed [diethyl pyrocarbonate (DEPC)-treated water], then stained with toludine blue (1 min) and destained and dehydrated through graded ethanol concentrations and xylene (2 × 5 min).


RNA was extracted from microdissected cells or from NSC34 cells using Trizol reagent (Gibco Life Technologies, Paisley, U.K) following the manufacturer's instructions. RNA was treated with RNAse-free DNAse then reverse transcribed (RT) with Moloney murine leukaemia virus reverse transcriptase using random hexamers in the presence of RNAse inhibitor (42°C, 1 h). One microlitre of cDNA was amplified using Reddy-mix polymerase chain reaction (PCR) master mix (Abgene). Primers used are given in Table 1. For microdissected human motor neurones and dorsal horn neurones, the cycling parameters used were 95°C for 5 min and 50 cycles of 95°C for 15 s and 60°C for 1 min. The same parameters were used for mouse motor neurones except that the number of cycles was reduced to 35. The cycling parameters used for cultured cells were, 94°C for 30 min, 30 cycles of 94°C for 30 s, 60°C for 30 s (for NFL) and 54°C for other primers, 72°C for 45 s and, finally, 72°C for 10 min. PCR was carried out using a DNA engine peltier thermal cycler (GRI, Nottingham, UK). The product was then separated by agarose gel electrophoresis (2%) and analysed using an AlphaImager system and AlphaInnotech software (Flowgen, Aldermaston, UK).

Table 1.  PCR primers for investigation of neurofilament expression
 Forward primer (5′ to 3′)Reverse primer (5′ to 3′)
Human microdissected cells
Murine microdissected cells
Cultured cells


Paraffin-embedded sections were dewaxed (2 × 10 min xylene) and rehydrated through graded ethanol. Slides were microwaved for 20 min in sodium citrate, pH 6.0 and washed in TBS (50 mm Tris pH 7.6, 137 mm NaCl). Endogenous peroxidase was blocked with 3% H2O2 for 15 min and non-specific binding was blocked by incubation with a 1% casein solution. Primary antibodies were incubated overnight at 4°C followed by incubation with a biotinylated secondary antibody and avidin-conjugated peroxidase (Dako, Kyoto, Japan). Antibody binding was visualized with DAB/H2O2.

Protein extraction and immunoblotting

Protein was extracted from 100 mg of snap-frozen human thoracic spinal cord. Tissue was homogenized in 25 mm Tris, 0.1 m sucrose, 0.5 mm EDTA plus protease inhibitor cocktail (Sigma, St Louis, MO, USA) using an ultra-turrax homogenizer and centrifuged (5000 g, 10 min). The supernatant was collected and protein concentration determined using Coomassie blue reagent (Pierce, Tattenhall, UK). Protein extractions were carried out as described previously for NSC34 cells (Cookson et al. 1998).

Ten micrograms of protein extract was separated by acrylamide gel electrophoresis and transferred to polyvinylidene difluoride (PVDF) membrane (Immobilon; Millipore, Watford, UK). Membranes were blocked in 5% w/v dried skimmed milk in TBS/0.1% Tween (20 mm Tris–HCl, pH 7.6; 137 mm NaCl; 0.1% Tween) for 1 h at room temperature. Primary antibody was bound by incubation for 1 h at room temperature or overnight at 4°C (see Table 2). Membranes were washed in TBS/0.1% Tween, 3 × 5 min before binding of the horseradish peroxidase (HRP)-conjugated secondary antibody (1 h at room temperature; see Table 3) and then washed 3 × 10 min in TBS/0.1% Tween. Secondary antibodies were detected using ECL (Amersham, Little Chalfont, UK) following the manufacturer's directions. Following ECL detection, the membrane was washed extensively and boundantibodies removed in stripping buffer [62.4 mm Tris–HCl, pH 6.7; 2% sodium dodecyl sulphate (SDS); 0.5% v/v β-mercaptoethanol] at 50°C for 30 min. The membrane was reprobed with an antiactin antibody. For analysis immunoblots were photographed and densitometry carried out using AlphaInnotech software (Flowgen).

Table 2.  Primary antibodies used and conditions for use
AntigenAntibody typeRaised inDilution factorConditionsSupplier
SOD1MonoclonalSheep25001 h, RTBinding site
NFL (NR4)MonoclonalMouse500 WB 200 ICCOvernight, 4°CSigma
NFL (NA1214)PolyclonalRabbitWB 10 000 ICC 2000Overnight, 4°CAffiniti
NFM (NN 18)MonoclonalMouse500Overnight, 4°CSigma
NFM (NA 1216)PolyclonalRabbit500Overnight, 4°CAffiniti
NFH (N52)MonoclonalMouseWB 10 000 ICC 2000Overnight, 4°CSigma
NFH (NA 1211)PolyclonalRabbit500Overnight, 4°CAffiniti
Tubulin (TUB 2.1)MonoclonalMouse500Overnight, 4°CSigma
Actin (AC-40)MonoclonalMouse20001 h, RTSigma
Table 3.  Secondary antibodies used for Western blotting
Anti-Species raised inDilution factorSupplier information
SheepDonkey 500Sigma


NSC34 cells were seeded onto 16-mm diameter coverslips at a density of 2000 cells per coverslip and cultured for 5 days. Cells were washed with phosphate-buffered saline (PBS) before fixing in 5% acetic acid in ethanol for 10 min at − 20°C. Primary antibodies were diluted in block buffer [2% bovine serum albumin (BSA); 0.1% Tween in PBS] and incubated overnight at 4°C (see Table 2 for concentrations). After washing in PBS, fluorescent secondary antibodies were bound, either tetramethylrhodamine-conjugated goat anti-rabbit IgG or Oregon green 488-conjugated goat anti-mouse IgG (Molecular Probes, Leiden, the Netherlands). Both antibodies were diluted at 1 : 200 in block buffer and incubated for 1 h at room temperature. Cells were counterstained with Hoechst 33258. Staining was viewed on a Zeiss Axiovert 200 microscope and photographed using an Openlab system (Improvision, Coventry, UK).


Selective loss of neurofilament light subunit mRNA expression in motor neurones from familial ALS patients

The expression of mRNA encoding neurofilament subunits (NFL, NFM and NFH) was investigated in motor neurones from the lumbar spinal cord of ALS cases. Motor neurones were microdissected from post-mortem lumbar spinal cord from two cases of familial ALS with an I113T SOD1 mutation and two normal control cases. RT-PCR for each of the neurofilament subunit genes revealed a selective down-regulation in the levels of NFL mRNA in familial ALS cases compared with controls (Fig. 1a). To investigate the robustness of this finding, the results obtained from four independent experiments were normalized for GAPDH expression, combined and statistically analysed. The reduction in NFL expression was shown to be significant (p < 0.05 by anova). This decrease was seen only in NFL, no alterations were seen in NFM and NFH, or in GAPDH mRNA levels.

Figure 1.

Neurofilament subunit mRNA expression was investigated by RT-PCR in motor neurones (a) and dorsal horn neurones (b), in controls (control) and patients with an I113T SOD1 mutation (FALS). RT-PCR was carried out for each of the neurofilament subunits: NFL, NFM and NFH. GAPDH was used as a control for quantity of starting material. (c) Graph showing the results obtained from four independent experiments after normalizing for GAPDH expression. NFL mRNA is significantly reduced (*p < 0.05 anova).

The decrease seen in neurofilament light subunit mRNA levels in motor neurones did not occur in dorsal horn neurones (Fig. 1b). Dissected dorsal horn neurones from I113T SOD1 familial ALS cases and control cases (two cases in each instance) demonstrated similar levels of NFL mRNA expression.

Expression of neurofilament light subunit in hyaline conglomerate inclusions

Hyaline conglomerate inclusions have been previously been shown to contain both phosphorylated and non-phosphorylated neurofilament heavy subunits (Schmidt et al. 1987; Itoh et al. 1992; Ince et al. 1998). In order to investigate the presence of neurofilament light and medium subunits in these inclusions, it was necessary to use antibodies that would not cross-react between neurofilament subunits. Specificity of antibodies in human tissue was therefore investigated by western blotting (Fig. 2). The antibodies against NFL and NFM employed in this study can be seen not to cross-react with other neurofilament subunits. In immunohistochemical studies, both antibodies demonstrated clear staining in the cell bodies and axonal processes of neurones of both control and familial ALS cases (Fig. 3). Hyaline conglomerate inclusions were seen within the cell bodies in two cases with I113T SOD1 mutations and were visualized using both antibodies specific for NFL and NFM.

Figure 2.

Western blotting was used to establish the specificity of antibodies to neurofilament light (NFL, antibody NA1214), neurofilament medium (NFM, antibody NN18) and neurofilament heavy (NFH, antibody N52), using protein homogenates from normal human thoracic level spinal cord.

Figure 3.

(a and b) Immunocytochemistry using anti-NFH antibodies in L3 level human spinal cord demonstrated the presence of hyaline conglomerate inclusions, denoted by arrows, in motor neurones of ALS cases with an I113T SOD1 mutation (b) which were not seen in a control case (a). These inclusions also stained positively using antibodies specific for NFL (c) and NFM (d).

Expression of neurofilament mRNA is reduced in a cell culture model of ALS

Expression of NF mRNA was investigated in a motor neurone like cell line (NSC34) which stably expressed wild-type or mutant human SOD1. Expression of human SOD1 was monitored throughout all experiments by western blotting (Fig. 4). The upper band represents transfected human SOD1 and the lower band endogenous mouse SOD1. In the presence of mutant SOD1 (both G37R and G93A mutations), mRNA transcripts for NFL and NFM are greatly reduced, below the level detectable by RT-PCR (Fig. 5). mRNA could be detected for each of the neurofilaments in control cells expressing normal human SOD1 or transfected with empty vector. There were no differences observed in the level of transcripts for either neurofilament heavy or actin. These observations were replicated in two clonal cell lines for each cell type, and in three separate experiments.

Figure 4.

Expression of human SOD1 in transfected NSC34 cells demonstrated by western blotting. Human SOD1 and murine SOD1 can be resolved into two bands by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS–PAGE) gels due to the slower migration of human SOD1, which therefore appears as the upper band. Human SOD1 expression is seen in two clonal cell lines each for cells transfected with normal human SOD1 (3 and 4), G37R mutant SOD1 (5and6) and G93A mutant SOD1 (7 and 8). No human SOD1 expression is seen in cells transfected with vector only (1 and 2).Human SOD1 expression was monitored at each passage of NSC34 cells.

Figure 5.

Neurofilament mRNA expression was investigated in NSC34 clonal cell lines expressing mutant human SOD1 (G93A and G37R). As controls expression was investigated in cells transfected with vector only (vector) or expressing normal human SOD1 (nSOD1). Two separate clonal cell lines were investigated for each cell type. RT-PCR results are shown for NFL, NFM and NFH, as well as an actin control. Levels of NFL and NFM but not NFH, were dramatically reduced in clonal cell lines expressing mutant but not wild-type human SOD1. The results shown are representative of three independent experiments.

Levels of neurofilament protein expression in cells expressing mutant human SOD1

Western blotting on cell homogenates was carried out using antibodies to NFL, NFM and NFH to investigate the levels of neurofilament protein expression in cells stably expressing mutant human SOD1. As a control, another cytoskeletal protein, tubulin was also investigated. Figure 6(a) shows examples of the results obtained for each protein investigated. Densitometry was used to compare the levels of expression of each of the neurofilament subunits in the presence or absence of mutant SOD1. These measurements are expressed as a ratio to actin band density after stripping membrane and reprobing with an actin-specific antibody as a control for the amount of protein loaded. The results shown in Fig. 6(b) are combined from three independent experiments. The levels of expression of NFL and NFM were significantly reduced in the presence of both G37R and G93A mutant human SOD. anova with Newman–Keuls post-test gave p < 0.001 for each cell line expressing mutant SOD1 against both cell lines transfected with vector only, for both NFL and NFM. Protein levels were also significantly different between cells expressing mutant SOD1 and cells expressing normal SOD1 in all cases (NFM; p < 0.01 NFL; p < 0.05). Cells expressing normal human SOD1 also demonstrated a decrease in expression of neurofilament light and medium subunits compared to vector-transfected cells (p < 0.001 in all cases), but the decrease seen in the presence of mutant SOD1 was much greater. NFH appeared to be present at lower levels in all cell types expressing human SOD1 than in cells transfected with vector alone (p < 0.01 by anova). The presence of mutant SOD1 in the cells did not appear to further affect the level of NFH. There was no difference in expression levels of tubulin in the cell lines investigated. Neurofilament expression levels were monitored over a wide range of passages (6–25) to ensure that differences were not seen as a result of ageing of the cells. The same pattern of neurofilament expression was seen in cells at all passages.

Figure 6.

(a) Western blotting was used to investigate neurofilament subunit expression in cells expressing G93A or G37R mutant human SOD1 or normal SOD1 and transfected with vector only as controls. Two clonal cell lines are shown for each cell type. NFL, NFM and NFH expression was investigated, using tubulin as a control. Arrows indicate the band of interest on each western blot. These results are representative of three independent experiments. (b) Densitometry was used to quantify the relative expression levels of the neurofilament subunit proteins. The same filters shown in (a) were stripped and re-probed with an actin-specific antibody. The integrated density value (IDV) as a ratio of the level of each subunit to actin expression is shown and results are combined from three independent experiments. Black bars represent cells transfected with vector only, dark grey bars represent cells expressing normal human SOD1, and cells expressing G37R or G93A mutant human SOD1 are shown by light grey and white bars, respectively. Statistical analysis was carried out by anova with Newman–Kuels post-test. Statistical significance is denoted by +versus vector and *versus normal SOD1. *p < 0.05, **p < 0.01 and +++p < 0.001.

Consistent with these findings, the decrease in neurofilament protein levels could also be observed by immunocytochemistry (see Fig. 7). The changes in neurofilament expression levels were consistent across each cell population. No focal accumulations of neurofilament proteins were observed in any cell lines.

Figure 7.

Immunocytochemistry showed a decrease in the levels of NFL and NFM in the presence of both G93A and G37R mutant SOD1 (results shown only from G93A cells). No such decrease was measured in cells transfected with vector only or expressing normal human SOD1 (results shown only from vector transfected cells). No decrease was seen in the level of NFH staining.

Neurofilament subunit expression in non-clonal cell lines

To ensure that the changes in neurofilament mRNAs and proteins observed were not a result of the clonal selection procedure we also investigated neurofilament expression in non-clonal (parental) transfected NSC34 cell lines. As for clonal cell lines cells expressing G37R mutant SOD1 showed an almost total loss of NFL and NFM proteins (Fig. 8). A modest decrease in expression levels was also present in non-clonal cell lines expressing G93A mutant SOD1. Therefore it is unlikely that loss of neurofilament mRNA occurred as a direct result of the clonal selection procedure.

Figure 8.

Neurofilament protein expression was investigated in non-clonal cell lines expressing either G37R or G93A mutant human SOD1. The level of NFL and NFM is decreased in cells expressing mutant SOD1 (G37R, G93A), but not in control cell lines expressing normal human SOD1 (nSOD1) or transfected with vector only (vector). The expression of human SOD1 can be seen in the lower blot (SOD1). Here the upper band represents human SOD1 and the lower band represents endogenous mouse SOD1.

Loss of neurofilament light subunit mRNA expression in transgenic mice expressing G93A mutant human SOD1

The expression of neurofilament light subunit mRNA was investigated in spinal motor neurones of mice expressing G93A mutant human SOD1. As for human post-mortem tissue, motor neurones were micro-dissected from lumbar spinal cord of transgenic mice prior to disease onset and at end stage. NFL expression was investigated by RT-PCR in each of these groups of cells compared to non-transgenic littermates. A decrease in the level of NFL expression was seen in both 60-day-old and 140-day-old mice (see Fig. 9), and this difference reached significance in the 140-day-old mice (p < 0.01 by t-test). Similar results were observed in two pairs of G93A mutant and non-transgenic littermate controls.

Figure 9.

NFL expression was investigated in motor neurones of mice transgenic for G93A mutant human SOD1 (G93A) at 60 days old and 140 days old. Levels are also shown in non-transgenic littermates (ntg) of the same age. Expression levels were quantified by densitometry and normalized for the expression levels of actin mRNA. The results shown are values obtained from three independent RT-PCR reactions from one pair of G93A transgenic and littermate control mice. They are representative of data obtained in another pair of mice. There was a significant decrease in the level of expression of NFL in transgenic G93A mice at the 140-day time point (**p < 0.01). At 60 days, prior to clinical disease onset, a decrease in NFL was also observed which did not reach statistical significance.


Here we show a decrease in expression of NFL mRNA, but not NFM or NFH, in the motor neurones of familial ALS cases with I113T SOD1 mutations. This demonstrates that a selective loss of NFL mRNA is a common feature of familial and sporadic ALS (FALS and SALS), as previous in situ hybridization studies of sporadic ALS cases have demonstrated a decrease in expression of NFL (Bergeron et al. 1994; Wong et al. 2000).

To further investigate the role of mutant SOD1 in the down-regulation of NFL mRNA, we quantified neurofilament expression in a motor neurone-like cell line stably expressing wild-type and mutant SOD1. Under basal, unstressed culture conditions these cells are viable, thus allowing investigation of the cellular and biochemical effects of the mutation. In these cells, expression of NFL and NFM was found to be decreased at both mRNA and protein levels, with no alteration in NFH. This was a robust finding that was replicated at a range of cell passage numbers (passage 7–25) in pooled transfectants and clonal cell populations, and is supported by expression profiling studies in G93A mutant SOD1 mice (Olsen et al. 2001).

Transgenic mice expressing mutant SOD1 are well-characterized models of human FALS (Gurney et al. 1994; Wong et al. 1995; Bruijn et al. 1997). One of the advantages of these models is that early cellular and biochemical changes can be identified, such as decreases in the rate of axonal neurofilament transport (Williamson and Cleveland 1999). We therefore extended our studies on neurofilament expression to the G93A line of transgenic mice (Gurney et al. 1994). We showed that a significant reduction in NFL mRNA occurred at end stage in these mice (140 days; p < 0.01). Prior to the onset of clinical signs (60 days), we detected a small but insignificant reduction in the level of NFL mRNA in G93A mice, suggesting that NFL mRNA levels are not only reduced in the advanced stages of motor neurone degeneration. This is consistent with the possibility that there is a constitutive reduction in NFL mRNA in G93A mice, or that a reduction in levels of NFL mRNA accompanies the disease pathogenesis.

If the selective reduction in NFL mRNA leads to a reduced level of NFL protein, it is tempting to speculate that this may be sufficient to induce motor neurone-specific disruption of the intermediate filament skeleton in ALS. A feature of the pathology of familial ALS cases, including those with I113T SOD1 mutations, is the presence of hyaline conglomerate inclusions (Ince et al. 1998). These inclusions have previously been shown to include NFM and NFH, but the presence of NFL has not been widely reported. Neurofilament accumulations are a well-characterized phenomenon in mice which over- or underexpress specific intermediate filament proteins (reviewed in Julien and Beaulieu 2000). Therefore the reduction in NFL mRNA and protein in FALS and SALS cases may induce NFM and NFH to accumulate in the perikaryon and proximal axon of motor neurones. However, here we show that all three neurofilament proteins can be identified in hyaline conglomerate inclusions in I113T SOD1 patients, though accurate quantification of the relative amounts of each neurofilament subunit was not possible using immunohistochemistry. This suggests that the inclusions either arise prior to any significant loss of NFL protein in the perikaryon, or that the inclusions include pre-assembled axonal neurofilaments that are retrogradely transported into the cell body (Cleveland 1999; Julien and Beaulieu 2000). Although we have shown that reduced NFL mRNA in motor neurones from I113T SOD1 patients frequently contain neurofilament inclusions, further investigation of a larger number of patients is necessary before we can make any general conclusions on this mechanism in ALS in general. However, there is a precedent in the literature for decreased NFL expression as well as neurofilament-positive inclusions in sporadic ALS cases (Strong 1999; Wong et al. 2000).

This is the first investigation to report reduction of neurofilament mRNA and protein in a cellular model of SOD1-mediated ALS. Perhaps this is due to the suitability ofNSC34 cells as a model motor neurone cell line, since they express many of the cellular features of motor neurones, including neurofilament proteins (Cashman et al. 1992). However, it is important to consider the possible effects of generating single-cell clones expressing potentially toxic proteins. One possibility is that the decreases in intermediate filament mRNAs observed in cell lines have arisen simply by the selection and expansion of clones that lacked NFL and NFM expression, independent of exogenous mutant SOD1 expression. This seems unlikely for two reasons: Firstly, there was no effect noted in NFH expression, whereas all cell lines expressing mutant SOD1 protein demonstrated a decrease in NFL and NFM levels that was not seen in any of the control cell lines. Secondly, we observed decreases in NFL and NFM in pooled transfectants that had not undergone single-cell cloning. Although these decreases were more variable, the likely explanation for this variability is the existence of different numbers of cells with varying levels of SOD1 expression within the pooled transfectants compared to the clonal cell lines. This implies that mutant SOD1 expressing clones lacking NFL have arisen as a result of mutant SOD1 expression, and not by selecting pre-existing NSC cells with reduced NFL/NFM expression. Consistent with this is the observation that mutant SOD1 toxicity is enhanced in fibroblasts from G93A transgenic mice that over express NFL (Ricart et al., Society for Neuroscience poster 2001, unpublished data). On the other hand, expression of mutant SOD1 may lead to a selective down-regulation of NFL/NFM mRNA levels. The precise mechanism responsible for this is unclear; however, NFL and NFM are in adjacent regions of the genome, and display the same patterns of expression during development (Julien et al. 1986). Therefore the same transcriptional regulation might down-regulate the expression of both genes, and the induction/loss of these regulators may lead to down-regulation of NFL/NFM in mutant SOD1-mediated ALS. The identification of functional transcription factors and regulatory proteins in motor neurones will represent an important step towards understanding the role of transcriptional regulation in ALS. There is mounting evidence to suggest that factors affecting NFL mRNA stability including p190RhoGEF may also play a role in regulating neurofilament expression at the post-transcriptional level (Schwartz 1995; Canete-Soler et al. 1998a, 1998b, 1999, 2001; Canete-Soler and Schlaepfer 2000). The reason for the reduction in NFM levels observed in the NSC-34 cell model and not in the human or mouse ALS tissue is unclear, but is likely to reflect a difference in the cell biology of the NSC34 cell model, compared to motor neurones in situ in the spinal cord. Further work is needed to investigate these mechanisms in healthy and diseased motor neurones. Neurotrophic factors represent potential candidate signalling molecules that could contribute to the down-regulation of neurofilament expression in mice and cell lines expressing mutant SOD1. Nerve growth factor (NGF) has the effect of increasing the half-life of NFL mRNA (Lindenbaum et al. 1988), and neurotrophic factors used to induce differentiation of cultured rat dorsal root ganglion cells induce neurofilament expression (Sawin and Mitchison 1995; Gaspar et al. 1997). Additionally, ageing rats show decreased levels of neurofilament proteins, with a concomitant decrease in NGF p75 and trkA receptor expression (Parhad et al. 1995). There is evidence that this may be relevant to ALS, as many studies have suggested alterations in the expression of growth factor receptors in disease (Duberley et al. 1995, 1997; Dore et al. 1996; Nishio et al. 1998; Mutoh et al. 2000). Furthermore, axonal strangulation has been proposed as a model of motor neurone toxicity in ALS (Cleveland 1999; Julien and Beaulieu 2000). Blockage of the axonal process could inhibit retrograde signalling from receptors located at the distal axon terminal, leading to a breakdown in multiple signalling pathways that might regulate NFL transcription.

Another possibility concerns the hypothesis that neurofilament accumulation in the perikaryon is protective in ALS. It has been suggested that perikaryal neurofilaments may act to sequester phosphate groups from dysregulated protein kinases such as CDK5, and that this may represent a survival mechanism in large neuronal cells (Nguyen et al. 2001). Down-regulation of NFL could conceivably protect neurones by inhibiting neurofilament assembly, and thus increasing NFM/NFH levels in the perikaryon.

To investigate the effects of neurofilaments in the pathogenesis of ALS, a series of transgenic mice that either overexpress or are knockouts of specific neurofilaments have been crossed with lines of transgenic mice expressing mutant forms of SOD1. Crossing NFL knockout mice with G85R mutant SOD1 mice gave a protective effect (Williamson et al. 1998). Indeed, Julien and Beaulieu (2000) and Nguyen et al. (2001) have made a case for a general protective effect of the increased perikaryal neurofilaments in SOD1-mediated ALS, based on crosses with mice overexpressing NFH and G37R as well as G93A SOD1 mutant mice. Here we show that NFL mRNA is significantly reduced at end-stage disease in G93A mice, and that young mice (60 days) without clinical signs of disease have a small reduction in NFL mRNA. These new findings lend support to a role for SOD1-mediated deregulation of neurofilament expression inALS.

In conclusion, we have presented new data to support a role for altered neurofilament expression in mutant SOD1-mediated ALS. Our experiments raise two new possibilities. First, that reduced NFL mRNA contributes to motor neurone injury induced by mutant SOD1 and, second, that reduced NFL mRNA might represent an adaptive and potentially cytoprotective response within stressed motor neurones. Further studies are required to determine the underlying mechanisms that link mutant SOD1 to reductions of neurofilament mRNA and protein, including studies on the regulation of neurofilament transcription and mRNA stability in neurones. Our data are consistent with disease models involving increased toxicity of SOD1 in the presence of NFL (Ricart et al., Society for Neuroscience poster 2001, unpublished data), or the protective effects of perikayal neurofilament inclusions (Julien and Beaulieu 2000; Nguyen et al. 2001) induced by the stochiometric effects of reduced NFL mRNA expression.


This work was funded by grants awarded by the Wellcome Trust and the UK Motor Neurone Disease Association. We also thank Neil Cashman for generously providing us with NSC34 cells.