Toxic neurofilamentous axonopathies– accumulation of neurofilaments and axonal degeneration

Authors

  • J. Llorens

    Corresponding author
    • Departament de Ciències Fisiològiques II, Universitat de Barcelona and Institut d'Investigació Biomèdica de Bellvitge (IDIBELL), L'Hospitalet de Llobregat, Catalunya, Spain
    Search for more papers by this author

Correspondence: Jordi Llorens, Departament de Ciències Fisiològiques II, Universitat de Barcelona, Feixa Llarga s/n, 08907 L'Hospitalet de Llobregat, Catalunya, Spain.

(fax: +34 93 402 4268; e-mail: jllorens@ub.edu).

Abstract

A number of neurotoxic chemicals induce accumulation of neurofilaments in axonal swellings that appear at varying distances from the cell body. This pathology is associated with axonal degeneration of different degrees. The clinical manifestation is most commonly that of a mixed motor–sensory peripheral axonopathy with a disto-proximal pattern of progression, as in cases of chronic exposure to n-hexane and carbon disulphide. It has been demonstrated that protein adduct formation is a primary molecular mechanism of toxicity in these axonopathies, but how this mechanism leads to neurofilament accumulation and axonal degeneration remains unclear. Furthermore, little is known regarding the mechanisms of neurofilamentous axonopathy caused by 3,3′-iminodipropionitrile, an experimental toxin that induces proximal axon swelling that is strikingly similar to that found in early amyotrophic lateral sclerosis. Here, we review the available data and main hypotheses regarding the toxic axonopathies and compare them with the current knowledge of the biological basis of neurofilament transport. We also review recent studies addressing the question of how these axonopathies may cause axonal degeneration. Understanding the mechanisms underlying the toxic axonopathies may provide insight into the relationship between neurofilament behaviour and axonal degeneration, hopefully enabling the identification of new targets for therapeutic intervention. Because neurofilament abnormalities are a common feature of many neurodegenerative diseases, advances in this area may have a wider impact beyond toxicological significance.

Introduction

Neurofilamentous axonopathies are a major issue in experimental and clinical neurotoxicology [1]. In these pathological conditions, abnormal accumulation of neurofilaments is found in axonal swellings that occur at varying distances from the cell body [2-6]. In some instances, neurofilament inclusions are also observed in the cell perikarya [7-9]. Abnormal neurofilament accumulation also frequently occurs in several different neurodegenerative diseases [10-13], and the pathological features of toxic and neurodegenerative conditions may be remarkably similar, suggesting the existence of similar or coincident mechanisms of pathogenesis. Thus, neurofilamentous axonopathies are also of interest in the area of neurodegenerative pathogenesis.

The initial focus of this review will be the current understanding of neurofilament biology. Then the neurofilamentous axonopathies and the known mechanisms of toxicity of several major causative agents will be discussed. The final aim will be to provide an overview of recent studies that help to illuminate two major unresolved issues: the identity of the key mechanisms responsible for neurofilament accumulation, and the relationship between neurofilament accumulation and axonal degeneration.

Neurofilament composition, transport and regulation

Neurofilaments are the intermediate filaments found in neurons, with a typical diameter of 10 nm (i.e. diameter between that of the wider microtubules and the narrower actin filaments; for review see [10, 12]). They are heteropolymer chains containing diverse combinations of intermediate filament proteins, including the three major subunits, which are termed neurofilament-high (NF–H), -medium (NF–M) and -low (NF–L), according to their apparent molecular weights determined by gel electrophoresis. These three proteins are type IV intermediate filament proteins, a family of cytoskeletal proteins [14] that include, amongst others, the epithelial keratins (types I and II), the widely expressed vimentin (type III) and the nuclear lamins (type V). Another type IV intermediate filament protein, alpha-internexin [15], and a type III protein, peripherin [16], also assemble into neurofilaments and are currently regarded as additional neurofilament subunits [17]. The subunit composition of the neurofilament varies according to the neuron cell type, species and developmental stage [12]. The neurofilament cytoskeleton controls axonal calibre, which directly determines the electrical conduction velocity of the axon [12].

Neurofilaments are transported towards the axon terminal within the ‘slow component-a’ of axonal transport, with rates in the range 0.3–3.0 mm per day [18]. This range was determined from classical radioisotopic pulse-labelling experiments in which labelled neurofilament proteins were identified at several different time-points and distances from the cell bodies after injection of a radioactive amino acid into the region of the neuronal soma [19, 20]. Such experiments provided evidence of a wave of labelled neurofilament proteins moving along the axon. However, the nature of the transport and the form of the transported neurofilaments were subjects of intense debate until a clearer picture started to emerge at the beginning of this century.

The currently accepted theory of neurofilament transport (Fig. 1) is the stop-and-go theory [18], according to which the slow overall rate of transport measured by the classic experimental approach in mature axons is the result of rapid, infrequent and asynchronous movement of neurofilament polymers in both directions, anterograde (from the cell body to the axon terminal) and retrograde (towards the cell body), as supported by experimental evidence [21-23]. Each individual filament may move in either direction, but has a slight preference for anterograde movement, and spends most of the time at rest in a stationary position. This behaviour determines the slow anterograde characteristics of the overall transport even though each discrete anterograde and retrograde movement occurs at the rate of fast axonal transport (2–5 μm s−1, equivalent to 200–400 mm per day). One remaining point of debate is whether all neurofilaments may move at any time-point, or whether a highly stable stationary cytoskeleton is formed in mature axons [24, 25]. Neurofilament movements proceed by sliding on microtubule tracts using the common kinesin (plus-end directed, anterograde in axons) and dynein (minus-end directed, retrograde in axons) motor proteins [26-29]. The interaction of neurofilaments with motor proteins and the factors controlling these interactions are not clearly understood [11]. The stop-and-go theory is supported by mathematical modelling studies which assume six possible states for neurofilament polymers (Fig. 1B) and successfully predict cytoskeletal behaviour as a function of the transition rates between each state [25, 30, 31]. Nevertheless, it has been argued that other mechanisms, including diffusion and axonal growth, need to be taken into account to fully explain axon cytoskeletal dynamics given the diversity of individual or animal anatomy, cell types and developmental stages [32].

Figure 1.

The stop-and-go theory of neurofilament transport. (a). According to this theory, neurofilaments (orange) move bidirectionally along microtubules (yellow). Kinesin motor proteins (blue) move the neurofilaments towards the axon tip (anterograde movement), whereas dynein motor proteins (red) are responsible for retrograde movement towards the cell body. The anterograde or retrograde sense of the movements is dictated by the orientation of axonal microtubules, all of which have their plus ends at the distal side. At any one time-point, only a small fraction of neurofilaments move, as they spend most of the time resting. (Adapted from [18]). (b). Schematic representation of the six-state model of Li et al. [25] describing the behaviour of neurofilaments according to this theory. Each discrete anterograde and retrograde movement occurs at the rate of fast axonal transport, which is determined by the velocity of the motor proteins. The rates of transition from one state to another determine the behaviour of the neurofilament population, typically described as the slow component-a of axonal transport.

Neurofilament protein expression is highly regulated at the gene expression and posttranscriptional levels [13]. In addition, neurofilament proteins are highly modified at the posttranslational level [12]. The most extensively studied modification is phosphorylation. Neurofilaments are extensively phosphorylated, and their phosphorylation status varies as a function of cell type, location within the neuron and developmental stage. In addition, abnormal distribution patterns of neurofilament phosphorylation are a common finding in many human neurodegenerative diseases and experimental animal models [33, 34]. Changes in neurofilament phosphorylation have long been considered to determine axonal growth, neurofilament transport and neurofilament stability; however, conclusive proof of this has remained elusive and conflicting data have emerged. Thus, the roles of neurofilament phosphorylation remain unclear [11, 12, 34].

Extraneuronal factors may regulate neurofilament behaviour. It has been demonstrated that myelinating Schwann cells locally modulate neurofilament phosphorylation, axonal calibre and slow axonal transport [35-37]. It has been suggested that this regulation is mediated by the myelin-associated glycoprotein [38], but the mechanisms involved are not understood [12, 39]. Glutamate receptors have also been highlighted as potential determinants of neurofilament transport [40, 41].

Neurofilament pathology and neurodegenerative diseases

Abnormalities in neurofilament expression, transport and regulation have been identified in several neurodegenerative diseases. A pathogenic role of these abnormalities has been supported by the identification of a number of mutations in the genes encoding neurofilament and neurofilament-associated proteins as causes of neurological disease. As shown in Table 1, pathogenic mutations in these genes have been identified for amyotrophic lateral sclerosis [42, 43], other lower motor neuron diseases [44], Charcot-Marie-Tooth disease [45, 46] and spastic paraplegias [47]. The relationship between neurofilaments and neurodegenerative pathogenesis has been assessed (for reviews see [10-13, 34, 48-51]).

Table 1. Neurofilament and motor protein mutations in neurodegenerative diseases
DiseaseGeneGene productReference
  1. CMT2A, Charcot-Marie-Tooth disease type 2A; CMT2, Charcot-Marie-Tooth disease type 2; SPG10, spastic paraplegia type 10; LMND, lower motor neuron disease; ALS, amyotrophic lateral sclerosis.

CMT2A KIF1B Kinesin family member 1BZhao et al. [45]
CMT2 NEFL Neurofilament lightMersiyanova et al. [46]
SPG10 Kif5A Kinesin family member 1A (KIF5A)Wang and Brown [47]
LMND DCTN1 Dynactin subunit p150GluedPuls et al. [44]
ALS PRPH PeripherinCorrado et al. [43]
ALS NFH Neurofilament heavy chainAl-Chalabi et al. [42]

Toxic neurofilamentous axonopathies

About 50 single chemical entities or families of similar chemicals have been unequivocally or strongly associated with human peripheral neuropathy [52]. Of these, at least a dozen involve significant changes in neurofilament distribution (Table 2). Other agents that cause similar effects have only been identified in experimental animal studies. Some clinically irrelevant agents are nevertheless actively studied because their mechanisms of action may be important for understanding human pathologies. For several axonopathy-inducing agents, the neurofilament effect predominates as the main pathological feature. These agents include, amongst others, n-hexane and related compounds, carbon disulphide and related compounds and 3,3′-iminodipropionitrile (IDPN). The compounds related to n-hexane include its toxic metabolite (2,5-hexanedione), aliphatic analogues (methyl-n-butyl ketone, 3,4-dimethyl-2,5-hexanedione and 3-methyl-2,5-hexanedione) and aromatic analogues (1,2-diethylbenzene, 1,2-diacetylbenzene and 1,2,4-triethylbenzene) [53, 54]. N,N-diethyldithiocarbamate and bis(diethylthiocarbamoyl)disulphide (disulphiram) are related to carbon disulphide [55]. A derivative of IDPN that causes the same effects, N-hydroxy-IDPN, has also been identified [56].

Table 2. Agents associated with peripheral neuropathy that cause significant changes in neurofilament distribution
In humans
Acrylamide
Allyl chloride
Arsenic
Carbon disulphide
Dimethylaminopropionitrile
Diphtheria toxin
Ethylene oxide
Hexachlorophene
n-Hexane and its analogues
Misonidazole
Organophosphorus compounds
Piridoxine
In experimental animals
Aluminium chloride
1,2-Diacetylbenzene
3,3’-Iminodipropionitrile

The neurotoxicity of n-hexane was discovered in industrial settings, in which the chemical was used as a solvent in poorly ventilated rooms, including shoe and furniture factories. In many countries, current workplace regulations have succeeded in reducing exposure [57], thus diminishing the incidence of the disease. However, recent data suggest that subclinical neurotoxic effects may occur with exposure to sub-Threshold Level Values [58]. In the overt cases, the clinical picture is that of a symmetrical sensory–motor neuropathy progressing in a disto-proximal sequence. Peripheral nerve biopsies show paranodal axonal swellings, filled with neurofilaments and distal axonal degeneration. Long tracts in the central nervous system may also be affected. An aggressive progression of the n-hexane neuropathy is observed in cases of high exposure to the solvent resulting from substance abuse (e.g. glue sniffing). Workers exposed to the related solvent methyl-n-butyl ketone develop similar clinical and pathological signs. A very similar neuropathy is also caused by a chemically unrelated solvent, carbon disulphide, used in the vulcanized rubber, rayon fibre and other industries. Carbon disulphide exposure also results from its release from disulphiram (Antabuse), used for alcohol aversion therapy, and from N,N′-diethyldithiocarbamate, which has a wide range of agricultural and industrial uses [55]. The n-hexane and carbon disulphide neuropathies have been extensively studied in animal models that accurately reproduce the clinical and pathological findings in humans. In addition, animal studies have characterized the similar axonopathies caused by several chemical analogues for which no cases of human exposure have been described. These human and experimental data have been reviewed [53, 59, 60]. Another similar axonopathy is that caused by acrylamide; it has been observed in humans and is successfully reproduced in experimental animals. The acrylamide axonopathy resembles that caused by n-hexane, but with more prominent axonal degeneration and less neurofilament accumulation [61]. Acrylamide has received considerable attention since the discovery of its generation during cooking [62].

Humans are unlikely to be exposed to significant amounts of IDPN, but the axonopathy caused by this chemical in experimental animals is of great theoretical interest. IDPN was initially tested in animals as an analogue of beta-aminopropionitrile, the first proposed causative agent of lathyrism. Rodents exposed to this compound developed a prominent hyperactivity syndrome [63], later found to be caused by its toxic action on the vestibular sensory hair cells, leading to loss of balance and sense of gravity [64-66]. However, the initial pathological studies revealed that it also causes proximal neurofilamentous axonopathy [2]; that is, neurofilament-filled swellings of the proximal axon segments. As stated by Carpenter who first described the axonal pathology of amyotrophic lateral sclerosis in 1968 [67], ‘the spheroids produced by IDPN have a striking morphological similarity, both at light and electron microscopy levels, to the spheroids of motor neuron disease’.

Mechanisms underlying toxic neurofilamentous axonopathies

Findings by several groups during the 1980s and 1990s (for reviews see [53, 59, 60, 68, 69]) led to well-supported hypotheses to explain the initial mechanisms of neurotoxic action of n-hexane and its aliphatic analogues. On the basis of these hypotheses, all these neurotoxic conditions were categorized as γ-diketone axonopathies (Fig 2A). A similar mechanism of neurotoxic action is now well accepted for carbon disulphide (Fig 2B). More recently, aromatic compounds have also been shown to cause similar effects to n-hexane, and their study had added support to the γ-diketone hypothesis [54, 70, 71]. n-Hexane and methyl-n-butyl ketone, which are relevant in occupational settings, are substrates of microsomal activities that transform them into the γ-diketone, 2,5-hexanedione. Considerable data support the conclusion that the gamma spacing of the two carbonyl compounds in 2,5-hexanedione is critical to the generation of neurofilamentous axonopathy. Analogues such as 2,4-hexanedione which do not keep the γ-diketone structure are not neurotoxic (Fig 2C). The same is true for the aromatic analogues: the γ-diketone 1,2-diacetylbenzene is active and the δ-diketone 1,3-diacetylbenzene is not [70, 71]. It has been demonstrated that these γ-diketones covalently bind to amine groups in proteins, with a particularly strong affinity for ε-amino groups of lysines, generating 2,5-dimethyl-pyrrole adducts. It is widely accepted that pyrrole adduct formation is a necessary step for the axonal effects of γ-diketones [69]. The next proposed step is oxidation of the pyrrolyl adduct; this generates a protein-bound electrophile which reacts with protein nucleophiles to result in covalent cross-linking of proteins. Available data demonstrate the ability of the γ-diketones to cause protein cross-linking, including neurofilament–neurofilament cross-links. Nevertheless, the relevance of this step to neurofilament accumulation and axonal degeneration is still debatable [72]. Like γ-diketones, the neurotoxic properties of carbon disulphide seem to be due to the ability to form protein adducts and cross-links (Fig 2B).

Figure 2.

Axonopathy-inducing agents, their mechanisms of bioactivation, active metabolites and inactive analogues [53, 54, 59, 60, 68]. (a). The γ-diketone 2,5-hexanedione is the common neurotoxic metabolite of the solvents n-hexane and methyl-n-butyl ketone. The γ-diketones react with the ε-amino groups of lysines to generate pyrrole adducts in many proteins. Oxidation of these pyrrole adducts has been shown to cause protein–protein cross-links. (b). Carbon disulphide generates dithiocarbamate and isothiocyanate adducts on proteins, which are able to generate dithiocarbamate ester and thiourea cross-links between proteins. (c). The aromatic analogues of n-hexane, 1,2-diethylbenzene and 1,2,4-triethylbenzene are neurotoxic because they generate the aromatic γ-diketones 1,2-diacetylbenzene and 1,2,4-triacetylbenzene. Isomers that do not have a γ-diketone structure are not neurotoxic.

One problem with the cross-linking hypothesis is that this mechanism may operate on many proteins, but the γ-diketones produce quite selective neurotoxicity. However, two explanations have been proposed. First, not all proteins are equally susceptible to cross-linking, as the abundance of lysines in the protein and the chemical environment of each lysine ε-amino group in the protein determine the likelihood of cross-linking [69, 71]. Secondly, neurofilaments are amongst the proteins with the longest half-lives in the body; that is, they have a lower turnover rate and are thus more likely to accumulate in an abnormally modified form. This explanation has received support from experimental and human studies demonstrating cross-linking of erythrocyte spectrin due to γ-diketone and carbon disulphide exposure. This finding was predicted because erytrocyte spectrin has also a low turnover rate as it has no turnover during the average erythrocyte circulation of 120 days [68].

The conclusion that protein adduction is a key step in the γ-diketone and carbon disulphide axonopathies led to the unified theory of pathogenesis that included protein adduction as the common cause of chemically induced axonal neuropathies [73]. The theory included the proposal that IDPN is metabolized to reactive metabolites able to generate protein adducts [74]; this theory has received some support [56, 75-77], but has not been proven. In fact, in contrast to the n-hexane analogues and carbon disulphide, the mechanism of toxicity of IDPN remains unclear. In general, the hypothesis that protein adduction is a necessary step in all cases of chemically induced neurofilament accumulation remains unproven. Nevertheless, protein adduct formation has been proposed as a major molecular mechanism in neurotoxicity including, but not restricted to, neurofilamentous axonopathies [69].

Two major differences were initially noted between the n-hexane/carbon disulphide and the IDPN axonopathies which were thought to reveal key aspects of the mechanisms that generate lesions and their final consequences. The first difference was the location of the neurofilament accumulations, which were found in the proximal segments of axons in IDPN, but distally in n-hexane and carbon disulphide axonopathies. It was initially thought that proximal and distal axonopathies were two different disease conditions with different mechanisms of action. However, n-hexane analogues (3,4-dimethyl-2,5-hexanedione and 1,2-diacetylbenzene) that cause swelling in the proximal axon segments have been identified [4, 70, 71, 78]. Another analogue, 3-methyl-2,5-hexanedione, was found to cause accumulation of neurofilaments throughout the nerve, but more frequently in intermediate axonal regions [5]. Other studies also demonstrated changes in the precise location of the neurofilament accumulations depending on the exposure rate or the experimental animal species [79]. As a result it is now widely accepted that the different locations of the neurofilament accumulations are not determined by different mechanisms of neurotoxic action.

The second major difference noted between the n-hexane/carbon disulphide and the IDPN axonopathies was that the accumulation of neurofilaments was associated with axonal degeneration and muscular weakness in the former (see above), whereas no degeneration or weakness was evident with IDPN [80]. This raised the issues of whether the neurofilament effect is the cause of the degeneration effects, or whether they are different phenomena caused by common molecular mechanisms or whether they are caused by different mechanisms. Several studies have shown that the γ-diketones that more rapidly cause protein cross-linking tend to cause more proximal swellings, and are more neurotoxic in terms of functional deficits [4, 78, 81, 82], suggesting a link between them. However, the features of IDPN axonopathy are not consistent with the conclusion that agents that cause more proximal axonopathies are functionally more neurotoxic. According to LoPachin and Lehning [83], chemicals that lead to axonal swelling also cause axonal atrophy, and it is this atrophy that is directly responsible for the loss of function resulting in clinical muscular weakness. However, others have concluded that swelling and atrophy are two aspects of the same phenomenon affecting the axon, neurofilament redistribution and that it is axonal degeneration rather than either of these morphological features that leads to the neurological deficits [84]. This conclusion is supported by several experiments of the effects of chronic IDPN exposure which show prominent axonal swelling and atrophy without apparent functional deficits [65, 80]. Thus, how axonal degeneration is induced and how this relates to the alterations in neurofilament transport or distribution are key issues that need to be addressed.

Neurofilament transport and the toxic neurofilamentous axonopathies

The ultimate mechanisms that cause accumulation of neurofilaments remain to be identified. In the first description of the IDPN proximal axonopathy, Chou and Hartmann [2] introduced the concept of ‘axostasis’ to explain neurofilamentous axonopathies. This implied the movement of the axoplasm from the proximal to the distal axon segments, and the appearance of axonal swelling following alterations in axonal transport. Identification of the different protein constituents of the cytoskeleton and of the different rates of transport led to the proposal that agents causing neurofilamentous axonopathies do so by specifically altering neurofilament transport. Consistent with this hypothesis, IDPN was found to cause a decrease in this transport [20], and unaltered synthesis and delivery into the axon from the cell body in combination with decreased axonal movement was assumed to cause the proximal axonopathy. The effect of the γ-diketones on neurofilament axonal transport was more difficult to determine, as it was reported to be accelerated by 2,5-hexanedione [85] and decreased by 3,4-dimethyl-2,5-hexanedione [86]. In any case, the cross-linking hypothesis postulated that cross-linked neurofilaments would accumulate because they would not be transported across the axonal constriction at the nodes of Ranvier, thus explaining the localization of the swellings proximal to the nodes (see [87]). According to this view, the swellings would consist of cross-linked neurofilaments unsuitable for transport. Although this prediction had not been experimentally demonstrated, it gained wide acceptance because it was consistent with the theory that neurofilament transport includes only anterograde movements.

However, this view is much more difficult to accommodate within the currently accepted stop-and-go theory of neurofilament transport, in which, as explained above, the behaviour of the neurofilaments is much more dynamic. If the slow axonal transport is the overall result of rapid forward and backward movements and if neurofilament cross-linking causes removal of the altered subunits from the transportable pool, the generation of focal swellings is not easy to explain. Modelling studies indicate that small changes in the rates governing the state transitions shown in Fig. 1B are sufficient to cause overt changes in neurofilament behaviour, and reveal the rate at which neurofilaments reverse their direction of movement as the most likely parameter governing the overall rate of their transport and distribution along axons [31]. A first possible conclusion from these results is that similar neurofilament swellings may result from similar changes in neurofilament dynamics irrespective of the initial trigger. Rather than a passive accumulation of cross-linked neurofilaments, the axonal swellings in the γ-diketone and carbon disulphide axonopathies may be the consequence of altered state transition rates. In these cases, protein adduction and perhaps cross-linking would lead to the altered neurofilament behaviour. However, causes other than adduction and cross-linking can be hypothesized in the case of IDPN or neurodegenerative-associated axonopathies. A second possible conclusion is that any factor involved in the regulation of neurofilament transport, not only the neurofilaments, could generate the pathology.

Experimental animal studies have begun to address toxic neurofilamentous axonopathies in the light of recent understanding of neurofilament biology. Sabri et al. [88] examined the impact of exposure to the axonopathy-inducing agent 1,2-diacetylbenzene on the neuronal content of kinesin, dynein, NF-M and tau, and observed a number of alterations that were not found after exposure to the nonneurotoxic isomer 1,3-diacetylbenzene. In a more detailed analysis, Zhang et al. [89] examined protein–protein interactions in cosedimentation assays using microtubules and neurofilaments prepared from rats exposed to 2,5-hexanedione. These authors observed no relevant changes in the distribution of neurofilaments, alpha- or beta-tubulins, kinesins (KIF1A, KIF3 and KIF5), dynein or the dynein-associated protein dynactin. By contrast, the microtubule preparations showed a substantial reduction in the cosedimentation of the microtubule-associated proteins (MAPs) MAP1A, MAP1B heavy chain, MAP2 and tau, suggesting that the γ-diketones may cause the axonopathy by selectively impairing the binding of MAPs to microtubules. Furthermore, the candidate target Spna2, also known as α-II spectrin, was identified by proteomic analysis of 2,5-hexanedione-exposed animals [90]. This led to the hypothesis that calpain- and/or caspase-mediated proteolysis of this structural protein is central to the development of axonal swellings in the γ-diketone axonopathy. However, a study of the toxicity of 1,2-diacetylbenzene in Spna2 mutant mice lacking a calpain- and/or caspase-sensitive domain did not support the hypothesis [90]. Soler-Martín et al. [91] hypothesized that the IDPN axonopathy could be explained by a shift in the dominant direction of transport, from anterograde to retrograde (a rightward shift in the equilibrium state in Fig 1B). If this were the case, the proximal neurofilament-filled swellings would contain not only neurofilaments from the cell body but also neurofilaments recruited from more distal axon segments. Also, the proximal accumulation of neurofilaments would be associated with a wave of loss of neurofilaments progressing in a disto-proximal direction. To test this prediction, the authors examined distal motor axons from IDPN-exposed rats to determine neurofilament content. The results showed a significant loss of neurofilaments at the motor endplate but not at the more distal axon segments. Of note, these results did not confirm the hypothesis, as endplate loss can be attributed not only to retrograde transport but also to lack of delivery and local proteolysis; synaptic terminals are sites of neurofilament degradation by calcium-activated proteases [92, 93]. However, the results do not disprove the hypothesis either because the possibility that the mature axon contains a highly stable stationary neurofilament cytoskeleton, with only a small fraction of neurofilaments undergoing transport [24], would mean that the methods used were not sensitive enough to detect a distal to proximal movement of the small potentially mobile fraction. In addition, lack of sensitivity could be an explanation even if all neurofilaments are available for movement [25], but do so at a very low rate in the distal axons evaluated.

Axonal degeneration and the toxic neurofilamentous axonopathies

As noted above, the relationship between neurofilamentous pathology and the axonal degeneration that occurs in toxic axonopathies has not been satisfactorily defined. Because accumulation of neurofilaments was found to be prominent and precede axonal degeneration, initially they were attributed a causative role. However, IDPN may cause neurofilament accumulation with little evidence of neuronal degeneration [65, 80]. To investigate the relationship between neurofilaments and axonal degeneration, the effects of the γ-diketones and acrylamide were tested in animals lacking axonal neurofilaments. Different conclusions were first obtained using the crayfish [94] and the neurofilament-deficient quail [95], but more convincing data from a transgenic mouse model lacking axonal neurofilaments indicated that these cytoskeletal elements are not necessary for the development of degenerative axonal changes in these axonopathies [96]. Thus, axonal degeneration is likely to occur in parallel to the effects on the neurofilament and not subsequently.

Some studies have now begun to search for proteins other than neurofilaments that may be the target of axonopathy-inducing chemicals and mediate axonal degeneration. By proteomic comparison of the differential effects of the neurotoxic 1,2-diacetylbenzene and the nonneurotoxic 1,3-diacetylbenzene, Tshala-Katumbay and colleagues [97] identified 22 candidate proteins. Amongst these proteins, they highlighted two as candidates for involvement in nerve fibre degeneration, protein disulphide isomerase and gelsolin, which were found to be downregulated by 1,2-diacetylbenzene. Protein disulphide isomerase is a chaperone that regulates protein folding; it associates with superoxide dismutase, and may play a protective role against the protein aggregation that commonly occurs in motor neurons in amyotrophic lateral sclerosis [98]. Gelsolin, which has an important role in actin dynamics, is protective against excitotoxic and apoptotic neuronal death [99] and is mutated in the familial amyloid polyneuropathy type IV [100]. It is interesting that these two proteins were also identified by proteomic analysis of the spinal cord of rodents exposed to 2,5-hexanedione, thus supporting the conclusion that their loss may be of pathogenic significance [101]. The mechanism responsible for this loss remains to be elucidated; an initial hypothesis is that these proteins are good targets for γ-diketone-induced lysine ε-amino adduction, as predicted by their amino acid sequence [97, 101].

Another important question is whether the neurofilament effects may or may not be related to later axonal degeneration, even though other more direct mechanisms cause the degeneration in the γ-diketone axonopathy. This would be especially relevant to understand the significance of accumulation of neurofilaments in diseases such as amyotrophic lateral sclerosis. Although there is usually little or no axonal degeneration in IDPN-induced proximal axonopathy [65, 80], degeneration in this model has been reported [102]. It has been speculated that the achievement of a critical mass of accumulated neurofilaments or an effect caused by indirect impairment of fast axonal transport provides possible links between neurofilamentous axonopathies and axonal degeneration. However, the finding that the proximal neurofilament accumulations are associated with early loss of neurofilaments at the neuromuscular junction [91] highlights an additional link that should be investigated. Accumulating evidence supports the notion that amyotrophic lateral sclerosis is likely to be a distal axonopathy in which retraction of the terminals is a major factor in motor neuron degeneration [103-106] (for a review see [107]). Findings of studies in a homologous mouse model of spinal muscular atrophy indicate that altered maturation of the cytoskeleton of the presynaptic element probably leads to its degeneration [108]. Thus, the possibility that axonal accumulation of neurofilaments is associated with extensive loss at the synaptic level should be evaluated in other toxic axonopathies and neurodegenerative conditions as a potential mechanism contributing to synaptic detachment and subsequent axonal degeneration.

Conclusion

Toxic neurofilamentous axonopathies are an important and interesting subset of neurotoxic diseases that may be also useful as models for understanding the pathogenic mechanisms of neurodegenerative disease. Several decades of research have resulted in an extensive description of these disease conditions, as well as insight into the mechanisms of toxic action. However, the mechanisms underlying neurofilament accumulation and axonal degeneration, and the relation between these two phenomena, remain unclear. As understanding of the basic biological processes involved in axonal growth, survival and transport increases, toxic axonopathies offer new opportunities for understanding pathogenic mechanisms of widespread medical significance.

Conflict of interest statement

The author has no conflicts of interest to declare.

Acknowledgements

The author was supported in part by grants from the Ministerio de Economia y Competitividad (Spain) (BFU2006-00343/BFI and BFU2009-06945) and from the Generalitat de Catalunya (2009 SGR 1059).

Ancillary