The neurobiology of amyotrophic lateral sclerosis

Authors

  • André Bento-Abreu,

    1. Laboratory for Neurobiology, Experimental Neurology, K.U.Leuven, Herestraat, 49 PO Box 1022, 3000 Leuven, Belgium
    2. Vesalius Research Center, VIB, Leuven, Belgium
    Search for more papers by this author
  • Philip Van Damme,

    1. Laboratory for Neurobiology, Experimental Neurology, K.U.Leuven, Herestraat, 49 PO Box 1022, 3000 Leuven, Belgium
    2. Vesalius Research Center, VIB, Leuven, Belgium
    Search for more papers by this author
  • Ludo Van Den Bosch,

    1. Laboratory for Neurobiology, Experimental Neurology, K.U.Leuven, Herestraat, 49 PO Box 1022, 3000 Leuven, Belgium
    2. Vesalius Research Center, VIB, Leuven, Belgium
    Search for more papers by this author
  • Wim Robberecht

    1. Laboratory for Neurobiology, Experimental Neurology, K.U.Leuven, Herestraat, 49 PO Box 1022, 3000 Leuven, Belgium
    2. Vesalius Research Center, VIB, Leuven, Belgium
    Search for more papers by this author

Dr A. Bento-Abreu, as above.
E-mail: andre.bentoabreu@vib-kuleuven.be

Abstract

Amyotrophic lateral sclerosis is a degenerative disease affecting the motor neurons. In spite of our growing insights into its biology, it remains a lethal condition. The identification of the cause of several of the familial forms of ALS allowed generation of models to study this disease both in vitro and in vivo. Here, we summarize what is known about the pathogenic mechanisms of ALS induced by hereditary mutations, and attempt to identify the relevance of these findings for understanding the pathogenic mechanisms of the sporadic form of this disease.

Introduction

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease with enormous impact on the quality of life of patients and their carers. Although its incidence is only 1–2 per 100 000, ALS is not rare: the life time risk of developing ALS is estimated to approach 1/400–1/700 (Johnston et al., 2006). Men are somewhat more frequently affected than women (male to female ratio is ∼1.5). Onset usually is in the sixth to seventh decade of life.

ALS mainly but not exclusively affects the lower motor neurons in the brainstem and ventral horn of the spinal cord (hence the name amyotrophic) and the upper motor neurons in the cortex that give rise to the corticospinal tract which descends through the lateral spinal cord (hence the name lateral sclerosis). This results in muscle atrophy and weakness, fasciculations, and spasticity (Rowland & Shneider, 2001). Although evidence of both upper and lower motor neuron involvement needs to be present to make the diagnosis, lower motor neuron involvement predominates at presentation in some patients, while upper motor neuron involvement can be most prominent in others. The spinal region (limb onset) is affected first in most patients while, in about one out of four or five, the onset is bulbar. ALS is a progressive disease and, although survival is variable, it is fatal in most patients after 3–5 years of evolution, most often due to bulbar dysfunction and respiratory insufficiency.

Biologically, ALS is more than a motor neuron disorder. It affects many other neuronal systems, but mostly to a degree below clinical detection threshold. The neuronal circuitries in the frontal region are, however, prominently affected. Many patients have (sometimes subclinical) evidence of frontal dysfunction and ∼15% develop a frontal dementia (Phukan et al., 2007).

ALS is familial in ∼10% of patients. It is usually inherited in an autosomal dominant way, but recessive and even X-linked forms exist (Valdmanis et al., 2009; Van Damme & Robberecht, 2009). In some familial forms, motor neuron and frontal involvement coincide; in other families, ALS is seen in some members while frontal dementia is seen in others. Most patients, 90% of them, have no family history and are considered to have the sporadic form of ALS.

Currently, there is no cure for ALS. One drug, riluzole, has been demonstrated to significantly increase survival and is well tolerated, but the magnitude of its effect is limited (Bensimon et al., 1994; Lacomblez et al., 1996; Tripathi & Al-Chalabi, 2008). Symptomatic therapy remains the mainstay of treatment, and has made a clear difference in survival (Van Damme & Robberecht, 2009).

Most of what we know about the pathogenesis of ALS comes from studies on the genetic forms. The significance of these monogenic forms in understanding the far more prevalent sporadic ALS is uncertain. Here, we will review some aspects of the current thinking on the etiology and pathogenesis of familial ALS and critically review its significance for the sporadic form of this dramatic motor neuron degeneration.

Lessons from familial ALS

Over the last two decades several genes, mutations in which cause ALS, have been identified (see Table 1). We here summarize what is known about the most common one of them. It should be noted that most of the mutations known to underly hereditary ALS have also been found in (a small set of) apparently sporadic ALS patients. Often (most of the time), it is uncertain whether these are really new mutants or whether they have been misclassified as ‘sporadic’. This can happen in cases of non-paternity, in the presence of unknown, unreliable or incomplete family history, or if the parents of a patient are too young to draw conclusions, or have deceased before the age of penetrance of the phenotype. Incomplete penetrance, known to occur for some mutations, is another variable to take into account.

Table 1.   Monogenic forms of amyotrophic lateral sclerosis and frontotemporal lobar degeneration
 GeneChrInheritanceOnsetAggregates ofOccurrence of
TDP-43FUS/TLFLTDMND
  1. ANG, angiogenin; CHMP2B, charged multivesicular body protein 2B; FIG 4, factor-induced gene 4 protein; FTLD, frontotemporal lobar degeneration; FUS/TLS, fused in sarcoma/translocated in liposarcoma; GNR, progranulin; MATP, microtubulus-associated protein tau; SETX, senataxin; SOD1, superoxide dismutase 1; TARDPB, transactive response DNA-binding protein 43; VAPB, vesicle-associated membrane protein B; VCP, valosin-containing protein. +, (+), – and ? denote present, rarely present, absent and unknown, respectively.

Familial ALS
 ALS1SOD121qAD (AR)Adult??(+)+
 ALS2Alsin2qARChildhood??+
 ALS3?18qADAdult??+
 ALS4SETX9qADChildhood/adolescent??+
 ALS5?15qARChildhood??+
 ALS6FUS/TLS16pAD (AR)Adult+++
 ALS7?20pADAdult??+
 ALS8VAPB20qADAdult??+
 ALS9ANG14qADAdult+?++
 ALS10TARDBP1pADAdult+?++
 ALS11FIG 46qADAdult??+
 ALS-FTD1?9qADAdult??++
 ALS-FTD2?9pADAdult+?++
Familial FTLD
 FTLD-MAPTMAPT17qADAdult+(+)
 FTLD-GRNGRN17qADAdult+++
 FTLD-CHMP2BCHMP2B3pADAdult++
 FTLD-VCPVCP9pADAdult++(+)

Mutations in superoxide dismutase 1 (SOD1): the first and hitherto most common cause of familial ALS

Mutations in the SOD1 gene (chromosome 21) remain the most common cause of familial ALS (Rosen et al., 1993). They are found in ∼20% of the families and thus account for ∼2% of all ALS.

SOD1 is an enzyme of 153 amino acid residues, ubiquitously expressed and active as a homodimer. It catalyses the conversion of superoxide free radicals to hydrogen peroxide, which can be further detoxified to water and oxygen by glutathione peroxidase or catalase. It should be distinguished from SOD2 (a mitochondrial SOD) and SOD3 (an extracellular SOD).

Missense mutations, affecting almost every amino acid residue of the protein (and a few small deletions and insertions, in addition to rare C-terminal truncating non-sense mutations) are known to give rise to familial ALS, irrespective of their effect on dismutase activity (Borchelt et al., 1994; Robberecht et al., 1994; Rosen et al., 1994). Transgenic mice or rats overexpressing mutant SOD1 develop motor neuron degeneration with progressive muscle weakness, muscle wasting and reduced life span (Gurney et al., 1994; Ripps et al., 1995; Wong et al., 1995; Bruijn et al., 1997; Howland et al., 2002). In addition, spontaneous mutations in the SOD1 gene give rise to a recessively inherited age-dependent motor neuron disease in dogs (Awano et al., 2009). More recently, mutant SOD1 models have been generated in zebrafish (Lemmens et al., 2007) and Caenorhabditis elegans (Witan et al., 2008; Wang et al., 2009a), suitable for genetic and small compound screening (Fig. 1).

Figure 1.

 Animal models of mutant SOD1-induced motor neuron degeneration. (A) Mutant SOD1G93A mouse with weakness and muscle atrophy of the lower limbs, which is accompanied by (B and C) loss of large motor neuron cell bodies in the ventral horn of the spinal cord (C, compared to a control B). (D and E) Zebrafish embryos injected with mutantSOD1A4V RNA have reduced motor axon outgrowth with abnormal branching (F, compared to wild-type injected embryos G). An SV2-staining (in red) was used to visualize the motor axons.

Almost all SOD1 mutations behave as autosomal dominant traits, and phenotype–genotype correlations have been described (Cudkowicz et al., 1998; Regal et al., 2006; Siddique & Siddique, 2008). One mutation, D90A, is recessive in populations of Scandinavian origin but dominant in others (Andersen et al., 1995; Robberecht et al., 1996). The mechanism underlying the resistance of certain populations to monoallelic expression of this mutation (or the susceptibility of others) is of high interest but hitherto unknown.

Not surprisingly, dysfunction of the axon, containing 99% of the motor neuron cytoplasm, is among the earliest manifestations of the mutant SOD1-induced degenerative process. This dysfunction appears as retraction of motor axons from neuromuscular junctions resulting in denervation and muscle weakness (Fischer et al., 2004). The pivotal significance of the axonal compartment explains the finding that preserving the cell body by interfering with the later stages of the degenerative process is insufficient to affect the clinical disease (Gould et al., 2006; Dewil et al., 2007a).

A toxic gain-of-function of the mutant protein underlies motor neuron toxicity, as these rodent models retain their endogenous SOD1 activity and SOD1-deficient mice have no overt phenotype of motor neuron degeneration (Reaume et al., 1996). The expression level of the SOD1 mutant protein for a given mutation determines disease severity, higher levels yielding a more aggressive phenotype. This has been well documented for the G93A-mutant SOD1 mouse model (Alexander et al., 2004; Fig. 2).

Figure 2.

 Dose effect of mutant SOD1 in the mouse model. Survival in mutant SOD1G93A mice as function of transgene copy number (adapted from Alexander et al., 2004).

The mechanism through which mutant SOD1 induces motor neuron degeneration remains incompletely understood, even nearly two decades after their discovery, but most probably involves several (interacting) pathways rather than a single pathogenic mechanism.

A gain-of-function for mutant SOD1

SOD1 is an important enzyme in the defence against superoxide anions, most of which are inadvertent reaction products in the mitochondria due to incomplete efficiency (‘leakiness’) of oxidative phosphorylation. Many studies have reported the presence of oxidative damage to proteins, lipids or DNA in patients with familial or sporadic ALS as well as in several mutant SOD1 mice (Barber & Shaw, 2010). It remains uncertain whether these changes are primary or secondary in nature. Two oxidation-modified proteins are particularly worth mentioning. SOD1 itself was found to be heavily oxidized (Andrus et al., 1998); this may at least contribute to the newly acquired toxic property of the protein (Ezzi et al., 2007). Another interesting protein damaged by oxidation is the glial glutamate transporter EAAT2, which provides a direct link between mutant SOD1-induced toxicity and excitotoxic motor neuron death (see below; also Trotti et al., 1999).

It has been suggested that oxidative damage is caused by aberrant oxidation reactions catalysed by mutant SOD1. However, expression of a mutant SOD1 without any oxidoreductive activity (obtained by mutating the histidine residues that are necessary for copper loading of the protein) still results in motor neuron degeneration in the mouse (Wang et al., 2003). This suggests that its enzymatic activity is not needed for the protein to be pathogenic. Alternative mechanisms have been suggested. Mutant SOD1 may bind with greater affinity to Rac1 than wildtype SOD1 does. Rac1 is a protein that regulates Nox2, an active subunit of the NADPH oxidase complex (Harraz et al., 2007). Inappropriate activation of Nox2 results in hazardous production of superoxide anions. Of notice, deletion of Nox2 slowed disease progression and improved survival of mutant SOD1 mice (Marden et al., 2007).

Alternatively, oxidative stress may be induced by mitochondrial dysfunction caused by abnormal recruitment of mutant SOD1 to the mitochondrial compartment (Shi et al., 2010). In the mutant SOD1 mouse, mitochondria undergo vacuolar degeneration in motor neurons (Jaarsma et al., 2001; Liu et al., 2004; Pasinelli et al., 2004). Misfolded mutant SOD1 has been found to bind to the outer mitochondrial membrane in a cell- and tissue-specific manner (Liu et al., 2004; Vande Velde et al., 2008). This may result in increased leakiness of the mitochondria (with reduced energy production and increased free radical generation), interfere with their calcium-buffering capacity (important in excitotoxicity; see below) or initiate apoptosis.

Evidence for an unexpected and newly discovered function for mutant SOD1 came from the finding that this protein is aberrantly secreted by motor neurons. Mutant SOD1 interacts with chromogranin (CHB)A and B, and is shuttled into the secretory pathway (Urushitani et al., 2006). The extracellular mutant protein was found to be toxic for motor neurons (Zhao et al., 2010). Because of this finding, a possible association between (non-hereditary) ALS and the CHBA and -B genes has been investigated. One study has shown the P413L CHGB variant to be associated with sporadic ALS and to determine age at onset (Gros-Louis et al., 2009).

Mutant SOD1-induced motor neuron degeneration: a foldopathy?

The most generally accepted hypothesis on the pathobiology of mutant SOD1 relates to its propensity to aggregate (Shaw & Valentine, 2007). ALS-causing mutations in SOD1 often result in decreased protein stability or net repulsive charge, which affect the folding and assembly of SOD1 dimers (Nordlund & Oliveberg, 2008). When synthesized, a protein has to be folded properly, a complex process in which several chaperone systems aid. Failure of this process results in protein misfolding and accumulation. The cell attempts to correct this by activating the so-called unfolded protein response (UPR), which includes the upregulation of a variety of chaperone proteins. If no refolding of the protein can be obtained, the protein is exported from the endoplasmic reticulum (ER) to the cytosol, where the ubiquitinated protein is degraded by the proteasome system. This process is thought to be at play in ALS (Kanekura et al., 2009). Mutant SOD1 has been found to accumulate in the ER and to inhibit derlin-1, the protein that transports proteins destined to be degraded from the ER to the cytosol (Nishitoh et al., 2008). Furthermore, a decrease in proteasome activity has been found in mutant SOD1-overexpressing cells and tissue (Urushitani et al., 2002; Kabashi et al., 2004, 2008a; Cheroni et al., 2009). Mutant SOD1 thus induces ER stress, and may overload the UPR response and the proteasome system. Upregulation of ER stress molecules has been correlated with the vulnerability of motor neurons in mutant SOD1 mice (Saxena et al., 2009). Overexpression of heat-shock proteins (HSPs) in vitro rescues the cell from mutant SOD1-induced toxicity (Patel et al., 2005). Disappointingly, neither HSP27 nor HSP70 overexpression in vivo affected motor neuron degeneration in mutant SOD1 mice (Liu et al., 2005; Krishnan et al., 2008). It is obvious that overexpressing one component of this sophisticated system may be insufficient.

The misfolded mutant SOD1 that escapes the cellular degradation system may interact with aberrant binding partners (such as mitochondrial membranes or chromogranins; see above) or form oligomers which then may proceed to the formation of higher molecular species and finally aggregate into intracellular inclusions (Johnston et al., 2000; Rakhit et al., 2002; Ezzi et al., 2007; Teilum et al., 2009). It is thought but not certain that this process is toxic for the neuron. Which stage of formation of inclusions is responsible for toxicity is uncertain, as is the question whether wildtype SOD1, which can form heterodimers with mutant SOD1 or can be recruited to coaggregate with it, contributes to this toxicity (Bruijn et al., 1998; Jaarsma et al., 2000; Fukada et al., 2001; Lemmens et al., 2007; Wang et al., 2009b). Aggregates may deplete the cell of essential constituents by coaggregation, or may physically disturb cellular processes such axonal transport (axonal strangulation; De Vos et al., 2007). However, just like for many other neurodegenerative diseases, it remains unknown whether aggregation of mutant protein is a hazardous or a protective phenomenon. It may well be that the first stages of the process (oligomerisation) are toxic (Johnston et al., 2000; Wang et al., 2002) while the aggregates themselves are essentially inert.

Non-cell autonomous motor neuron death

The convergence of damage in non-neuronal cells surrounding motor neurons has a larger impact on motor neuron survival then initially anticipated (Ilieva et al., 2009). Several types of non-neuronal cells, such as microglia and astrocytes, are activated in the course of the neurodegenerative process in ALS (Hall et al., 1998). This neuroinflammatory response, probably provoked by signals from diseased motor neurons, has neurotrophic and neurotoxic properties.

The contribution of non-neuronal cells to the pathogenesis of motor neuron degeneration has been studied in mutant SOD1 mice, in which the transgene was excised in specific cell types. It was found that deleting mutant SOD1 from microglia slowed motor neuron degeneration but did not affect disease onset (Boillee et al., 2006). Interestingly, the activation of microglial cells was not affected, showing that this reaction itself is not harmful, a finding that is consistent with the observation that preventing T-lymphocyte activation in the ALS spinal cord reduces microglial activation but accelerates disease (Beers et al., 2008). Replacement of mutant SOD1 microglia with transplanted wildtype microglia had a beneficial effect (Beers et al., 2006), but inhibition of microglial proliferation had no effect on disease progression (Gowing et al., 2008). This shows that, at least in this mouse model, microglial cells containing the mutant protein have a detrimental effect. On the other hand, wildtype microglia appear to be protective (Chiu et al., 2008). The role of microglia as pathogenic and/or protective cells is complicated by the question whether hematogenic macrophages populate the adult spinal cord. Several of the conclusions drawn from earlier experiments are indeed questioned by recent experiments using parabiosis (Ajami et al., 2007; Mildner et al., 2007).

Deletion of mutant SOD1 from astrocytes also slowed disease progression in the mutant SOD1 mouse model (Yamanaka et al., 2008). Interestingly, microglial activation was reduced in this experiment, suggesting an interaction between the two cell types. The nature of the interaction between motor neurons and astrocytes is likely to be multifactorial (Van Den Bosch & Robberecht, 2008). Astrocytes may release toxic factors (Nagai et al., 2007) or provide surrounding cells with less trophic support. Few of the astrocytic factors or motor neuron targets have been identified to date. Reduced expression of the glutamate transporter EAAT2 in astrocytes (Rothstein et al., 1992, 1995) and a reduced astrocyte-induced upregulation of the α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor subunit GluR2 (Van Damme et al., 2007) may enhance excitotoxic motor neuron death (see below). The transcription factor Nrf2, which regulates the expression of antioxidant enzymes containing an ARE element (antioxidant response element) was able to counteract the toxicity of mutant SOD1-containing astrocytes and prolong survival of mutant SOD1 mice (Vargas et al., 2008). Of major interest is the finding that transplanting wildtype astrocytes into the mutant SOD1 spinal cord delayed disease (Lepore et al., 2008).

Counterintuitively, deletion of mutant SOD1 from Schwann cells aggravated disease (Lobsiger et al., 2009), possibly via the dismutase effect of SOD1 in this cell type. Deletion from endothelial cells had no effect (Zhong et al., 2008) (see below). Influencing mutant SOD1 synthesis in muscle cells did not affect motor neuron degeneration in the mutant SOD1 mouse (Miller et al., 2006; Towne et al., 2008). However, overexpression of insulin-like growth factor isoforms exclusively in muscle did slow down progression (Dobrowolny et al., 2005). Therefore, the exact role of muscle in ALS remains an interesting topic of research.

Lowering mutant SOD1 as a therapeutic option

The removal of mutant SOD1, the primary cause of motor neuron toxicity, is an obvious therapeutic strategy. This has been achieved by the viral delivery of RNAi against SOD1 (Ralph et al., 2005; Raoul et al., 2005), by intracerebroventricular administration of antisense oligonucleotides (Smith et al., 2006) and by crossbreeding mutant SOD1 mice with mice that express an shRNA against mutant SOD1 (Xia et al., 2006). Hence, gene silencing holds great promise as a therapy for ALS (and in fact for many neurodegenerative diseases; Maxwell, 2009). The first clinical studies investigating the feasibility of these approaches in humans are under way.

As toxicity from aberrant secretion of mutant SOD1 is likely to play a role, targeting this pool of mutant SOD1 may be of interest. The burden of extracellular SOD1 could be reduced using an active or a passive immunization strategy, and this led to a slower disease progression in mutant SOD1 mice (Urushitani et al., 2007).

The SOD1 mouse as a model for studying ALS therapy

The mutant SOD1 mouse (and rat) has been used extensively to study compounds or approaches with possible therapeutic value (Turner & Talbot, 2008). The validity of this model has been questioned because some of the compounds with a positive effect in the mouse were negative in human studies. There may be other explanations. The effects observed in the mouse were often small, and may be easily missed in a clinically and genetically heterogenous human ALS population. Furthermore, the differences in pharmacokinetics between mice and humans were often largely neglected. In addition, the ‘positive’ results obtained in mice often came from (inadequately powered) studies in which administration of the compound began before disease onset, while in humans therapeutic trials are done in patients who have had ALS for at least one, sometimes even several, years.

The question is whether the mutant SOD1 mouse is a good model in which to study sporadic ALS. Obviously it is not ideal: sporadic ALS is definitely etiologically different from monogenic mutant SOD1-related familial ALS. Recent studies on transactivation response DNA-binding protein with molecular weight 43 kDa (TDP-43) suggest that there may also be a pathogenic difference, which will be discussed below.

TDP-43: a new player shifts attention to RNA and links familial with sporadic ALS

The role of TDP-43 was first suspected when it was identified as one of the major constituents of the intraneuronal inclusions characteristically observed in ALS and in frontotemporal lobar degeneration (FTLD)–ubiquitin (FTLD-U; Neumann et al., 2006). Subsequently, mutations in the TARDBP gene encoding TDP-43 were identified in some pedigrees with familial ALS (Gitcho et al., 2008; Kabashi et al., 2008b; Sreedharan et al., 2008; Yokoseki et al., 2008).

TDP-43 is a widely-expressed 414-amino-acid protein encoded by the TARDBP gene on chromosome 1 (Pesiridis et al., 2009; Geser et al., 2010). It has two RNA-binding domains and a glycine-rich domain in the C-terminal part, with which it binds to various heterogenous nuclear nucleoproteins (hnRNPs). It is more abundantly present in the nucleus than in the cytoplasm. The exact role of TDP-43 is incompletely understood, but it is thought to play a role in a variety of processes such as processing, stabilisation and transport of RNA (Buratti & Baralle, 2009; Geser et al., 2010). A well known example is its role in the splicing of cystic fibrosis transmembranous conductance regulator mRNA (Buratti et al., 2001). Of interest is the finding that another target for the action of TDP-43 in mRNA processing is the protein SMN, deficiency of which results in spinomuscular atrophy, an infantile or juvenile onset motor neuron disorder (Burghes & Beattie, 2009). Overexpression of TDP-43 enhances exon 7 inclusion during SMN splicing, a crucial event in yielding fully active SMN protein (Bose et al., 2008). SMN deficiency in its turn is thought to cause spinomuscular atrophy through defective RNA processing or transport (Burghes & Beattie, 2009). The possible link between SMN and TDP-43 is of major interest when thinking of a common pathway for motor neuron degeneration.

The more than 25 mutations found in the TARDBP gene are, primarily, missense mutations and are almost exclusively located in the C-terminal (glycine-rich) part of the protein (Lagier-Tourenne & Cleveland, 2009). There is also a truncating mutation in this gene (Daoud et al., 2009). TARDBP mutations are rare: they probably account for < 5% of familial ALS, i.e. < 1% of all ALS (Ticozzi et al., 2009a). The major interest in them comes from the finding mentioned above, that wildtype TDP-43 containing inclusions are found in the majority of sporadic ALS patients (Neumann et al., 2006; Fig. 3). Here, we will refer to this abnormal form of TDP-43 as TDP-43SALS/FTLD in contrast to ‘normal’ TDP-43, reminiscent of the naming in prion disease, where PrPC refers to the normal PrP and PrPSc refers to the pathogenic form of PrP in sporadic and infectious Creutzfeldt–Jakob disease; it does not differ from normal PrPC in its amino acid sequence. Mutant TDP-43 refers to the mutant proteins causing the hereditary forms of ALS, just as with mutant PrP and Creutzfeldt–Jakob disease, and will be referred to as TDP-43mutant.

Figure 3.

 TDP-43 containing inclusions in a motor neuron in the spinal cord of a sporadic ALS patient (courtesy of T Tousseyn and R Sciot, Pathology, University Hospital Leuven).

An overwhelming number of papers on the role of TDP-43 in neurodegeneration have been published over the last 2 years. A common finding seems to be that TDP-43mutant and TDP-43SALS/FTLD are mislocated, hyperphosphorylated, abnormally processed and ubiquitinated. The distribution of TDP-43 pathology spreads far beyond the motor system, suggesting the biological multisystem nature of the condition we clinically know as ALS (Geser et al., 2009).

TDP-43mutant and TDP-43SALS/FTLD are mainly present in the cytoplasm and appear to be depleted in the nucleus (Neumann et al., 2006; Winton et al., 2008; Sumi et al., 2009; Barmada et al., 2010). It therefore has been suggested that depletion of TDP-43 in the nucleus results in failure of RNA metabolism in this compartment, possibly resulting in the generation of abnormal splice variants. Alternatively, mRNA species in the cytoplasm that require the action of TDP-43 may be mistargeted or even degraded. Of interest in this regard is the finding that TDP-43 interacts with NF-L (neurofilament-light) mRNA, which may play a pathogenic role in ALS (Strong et al., 2007; Strong, 2010). Ongoing studies aim to identify RNA abnormities in TDP-43SALS/FTLD and TDP-43mutant cells and to establish their pathogenic role. This is obviously not easy given the large number of RNA species and the need to use unbiased approaches. In addition, it should be noted that these studies should not be limited to mRNAs, as recent studies have identified a role for microRNAs in neurodegeneration in general and in ALS in particular (Williams et al., 2009).

Mislocation may also result in pathogenicity due to a cytoplasmic gain-of-function rather than nuclear depletion (loss-of-function). There appears to be a correlation between cytoplasmic expression of TDP-43 or its C-terminal fragments and toxicity in vitro (Johnson et al., 2009; Nonaka et al., 2009; Zhang et al., 2009; Barmada et al., 2010), but it remains to be demonstrated that this is a causal correlation.

TDP-43mutant and TDP-43SALS/FTLD also appear to be abnormally processed, as C-terminal small molecular weight species, and in particular a fragment with a molecular weight of 25 kDa, are found in disease conditions (Neumann et al., 2006; Hasegawa et al., 2008). It has been suggested that caspase-3 is a TDP-43-processing enzyme (Zhang et al., 2007, 2009; Dormann et al., 2009). Expression of C-terminal fragments results in aggregate formation in vitro (Igaz et al., 2009), but the specificity of this processing and its significance for the pathogenesis remains to be shown (Dormann et al., 2009; Nishimoto et al., 2010). Of interest, the cleavage appears to be region-specific. In spinal cord, most of the TDP-43 recovered is full length (Igaz et al., 2008). TDP-43mutant and TDP-43SALS/FTLD are also hyperphosphorylated (the S409/410 sites are best characterized; Hasegawa et al., 2008; Inukai et al., 2008; Kametani et al., 2009; Neumann et al., 2009). Again, it is unclear whether these are primary or secondary modifications (Dormann et al., 2009).

Overexpression of TDP-43mutant in zebrafish results in a phenotype resembling that seen with overexpression of mutant SOD1 (Lemmens et al., 2007; Kabashi et al., 2010). Knockdown of TDP-43 results in a similar motor neuron phenotype (Kabashi et al., 2010), which suggests an essential role of endogenous TDP-43 for motor axon development, as has also been suggested in Drosophila (Feiguin et al., 2009; Li et al., 2010), but which is unlikely to be a model for nuclear depletion through cytoplasmic sequestration. The essential role of TDP-43 for early embryonic development in mammals has recently been shown using an elegant gene-trap approach, demonstrating early lethality of TARDBP-knockout mice (Sephton et al., 2010). TDP-43 is a developmentally regulated protein essential for early embryonic development. Loss of murine TDP-43 disrupts motor function and plays an essential role in embryogenesis (Kraemer et al., 2010). Interestingly, the heterozygous knockout mice (TARDBP+/−) showed signs of motor dysfunction, although no abnormalities in their motor neurons were apparent.

Overexpression of mutant TDP-43A315T driven by the prion promoter in mouse yielded expression of the transgene in neurons and glial cells throughout the nervous system and resulted in degeneration of motor neurons and of layer V cortical neurons (Wegorzewska et al., 2009). Expression of the TDP-43A315T was about three-fold that of endogenous TDP-43. These mice developed a paralyzing disease characterised by loss of upper and lower motor neurons. Interestingly, degenerating neurons contained ubiquitinated aggregates which, in spite of thorough investigation, did not contain the mutant TDP-43A315T. Loss of TDP-43 immunoreactivity from the nucleus was seen occasionally but did not seem to be a prominent finding. On the other hand, 25-kDa fragments appeared early in the disease. Unfortunately, this study did not report the findings in wildtype TDP-43-overexpressing mice.

Not surprisingly, based on the effects seen in other models such as Drosophila (Feiguin et al., 2009; Li et al., 2010), overexpression of human wildtype TDP-43 driven by the Thy1 promotor in mice gave rise to a phenotype as well. This promoter results in postnatal neuronal expression of the transgene. Expression of wildtype TDP-43 to a degree similar to that of TDP-43A315T in the previous study resulted in no phenotype. When increasing the level of wildtype TDP-43 expression, animals developed gait abnormalities and showed evidence for degeneration of motor neurons and neurons in layer V of the frontal cortex (Wils et al., 2010). The severity of the phenotype was parallel to the degree of TDP-43wt expression. In the diseased neurons, nuclear and cytoplasmic aggregates of ubiquitinated and phosphorylated TDP-43 were found. In this study, C-terminal 25-kDa fragments were found in the nuclear fraction. In this report, no TDP-43mutant was overexpressed.

It is difficult to compare these two models. Overexpression of TDP-43 seems to be toxic and may switch TDP-43 into TDP-43SALS/FTLD. The presence of a mutation favours this switch, although it needs to be taken into account that, in the TDP-43mutant study, glial cells also expressed the transgene and this may contribute to the process of neuronal degeneration, as we have learnt from the SOD1 model. The wildtype TDP-43 study, however, suggests that expression limited to neurons appears sufficient to induce degeneration.

The question is how to interpret the many findings in terms of pathogenic mechanism at play in vivo, and thus in non-overexpression conditions. It remains uncertain whether the cleavage, phosphorylation and ubiquination of TDP-43 are important for pathogenicity or not. Propensity of TDP-43 to aggregate, further enhanced by the presence of mutations, is an almost universal finding (Johnson et al., 2009; Nonaka et al., 2009; Zhang et al., 2009), although the most relevant model generated hitherto did not contain TDP-43-containing aggregates (Wegorzewska et al., 2009). Furthermore, the significance of the depletion of TDP-43 from the nucleus (found in many but not in all studies) as an underlying ‘compartmental’ loss-of-function mechanism needs to be established. Alternatively, the sequestering of TDP-43 in the cytoplasm may be the underlying gain-of-function mechanism. Does cytoplasmic TDP-43 gain a toxic biochemical function? Is the formation of aggregates, or one of the (oligomeric) species that are a step in the dynamics of this process, the mechanism of disease? Are essential cellular constituents trapped into these aggregates, resulting in an ‘unrelated’ loss of function?

In summary, the finding of TDP-43 in ALS and FTLD neurons and the identification of TDP-43 mutations in familial ALS was a second leap forward in ALS research. It has drawn attention to the possible role of RNA processing in the pathogenic mechanism of these diseases, even though the involvement of RNA in the mechanism itself remains to be demonstrated (Lemmens et al., 2010). Of major importance is of course the possible involvement of TDP-43 in sporadic ALS. It looks as if TDP-43 may play a role similar to α-synuclein in Parkinson’s disease (PD) and amyloid precursor protein (APP) in Alzheimer’s disease (AD). α-Synuclein mutations are a rare cause of familial PD, and α–synuclein-containing inclusions are seen in the sporadic form of PD. APP mutations are a rare cause of AD, but abnormally processed APP under the form of Aβ is a hallmark of sporadic AD. APP and α-synclein overexpression give rise to AD and PD in humans. This has not been observed for TDP-43 in ALS yet.

Finally, it needs to be pointed out that, while TDP-43-containing aggregates are seen in the large majority of sporadic ALS patients, they were noted to be absent in many (Mackenzie et al., 2007; Robertson et al., 2007; Tan et al., 2007), but not all (Shan et al., 2009) studies on mutant SOD1 ALS. This may suggest that the mechanisms underlying mutant SOD1-induced motor neuron degeneration and that of sporadic ALS may be different. This still needs to be studied in depth but it has further fuelled the doubts about whether mutant SOD1 models are of use in studying sporadic ALS.

FUS/ TLS: further evidence for a central role of proteins involved in RNA processing

Recently, two studies identified mutations in FUS ( fused in sarcoma)/TLS (translocated in liposarcoma) as a cause of ALS in Cape Verde, the US, Australia and the UK (Kwiatkowski et al., 2009; Vance et al., 2009). FUS/TLS mutations were also found in other populations in Europe, Japan and the US and it is estimated that FUS/TLS mutations cause familial ALS in 4–5% of cases (Belzil et al., 2009; Blair et al., 2010; Chio et al., 2009b; Damme et al., 2009; Drepper et al., 2010; Ticozzi et al., 2009b; Corrado et al., 2010; Groen et al., 2010; Suzuki et al., 2010). In addition, one de novo truncation mutation was reported (Dejesus-Hernandez et al., 2010).

The FUS/TLS gene is located on chromosome 16. Also known as hnRNPP2, it belongs to the FET family of RNA-binding proteins and it is an hnRNP. The protein consists of an N-terminal region rich in glutamine, glycine, serine and tyrosine residues (QGSY region) immediately followed by a glycine-rich domain. It contains an RNA-recognition motif (RRM) and multiple arginine, glycine, glycine (RGG) repeats implicated in RNA binding, a zinc finger and a C-terminal region that is highly conserved (Lagier-Tourenne & Cleveland, 2009). FUS/TLS is involved in pre-mRNA splicing as well as in the export of fully processed mRNA to the cytoplasm and thus shuttles between the nucleus and the cytoplasm (Zinszner et al., 1997). It may also play an important role in transport of mRNA (Yoshimura et al., 2006). In addition, it is important in gene regulation and it was recently shown that FUS/TLS can serve as a transcriptional regulatory sensor of DNA damage signals leading to gene-specific repression of gene transcription (Wang et al., 2008). FUS/TLS is ubiquitously expressed and under normal conditions it is mainly localized in the nucleus (Hackl & Luhrmann, 1996). In cultured hippocampal pyramidal neurons, FUS/TLS was localized not only in the nucleus but also in the dendrites (Fujii et al., 2005). This punctuate dendritic localization was dependent on an intact microtubule and actin network, and activation of mGluR5 metabotropic glutamate receptors stimulated FUS/TLS accumulation at the spines of excitatory synapses (Fujii et al., 2005). FUS/TLS-knockout mice die immediately after birth (Hicks et al., 2000) or are rarely alive at weaning (Kuroda et al., 2000). In an outbred strain, FUS/TLS-knockout mice survived but showed male sterility and reduced fertility of females (Kuroda et al., 2000). It was reported that heterozygous FUS/TLS mice were indistinguishable from wildtype littermates (Kuroda et al., 2000). Neurons deficient in FUS/TLS showed abnormal spine morphology and lower spine density (Fujii et al., 2005).

It is estimated that FUS/TLS mutations account for ∼5% of familial ALS and thus again for < 1% of total ALS (Lagier-Tourenne & Cleveland, 2009). FUS/TLS-linked ALS is a dominant disease, except in the original Cape Verdian family in which the FUS/TLS mutation is recessive (Kwiatkowski et al., 2009). Although in-frame deletions and insertions have been reported, most mutations are missense and the majority are located in the last exon. Mutant FUS/TLS accumulates in the cytoplasm of neurons (Kwiatkowski et al., 2009; Vance et al., 2009). Interestingly, FUS/TLS is also a component of nuclear polyQ aggregates in a cellular model of Huntington’s disease, as well as in patients with polyQ diseases, indicating that changing FUS/TLS to an insoluble form may be a common process in polyQ diseases and ALS (Doi et al., 2008, 2010).

Our knowledge on the role of FUS/TLS in the pathogenesis of ALS is still limited. Whether the RNA processing function of the protein is relevant or whether the mutant protein acquires an unrelated toxic function is not yet known and is an area of intensive research.

Other causes of familial ALS

Several other genes have been identified, mutations in which cause ALS, but these mutations occur in a very limited number of patients (Van Damme & Robberecht, 2009) (Table 1).

Mutations in vesicle-associated membrane protein-associated protein B (VAPB) are mainly found in Brazil (Nishimura et al., 2004). VAPB is involved in the unfolded protein ER response mentioned above (Kanekura et al., 2009). Mutant protein (P56S is the most studied mutation) looses this function and makes motor neurons vulnerable to ER stress induced by unfolded proteins (Suzuki et al., 2009). Studies in Drosophila showed that VAPB fragments interact with the ephrin system and that mutants are not correctly processed, resulting in a loss of function (Tsuda et al., 2008). However, VAPB-mutant protein is also prone to misfolding and aggregation (Teuling et al., 2007; Tsuda et al., 2008), again suggesting that aggregation is involved in the gain-of-function mechanism of these dominant mutations.

A surprising and exciting observation is the identification of variants in factor-induced gene 4 (FIG 4), a phosphoinositide 5-phosphatase in ALS patients (Chow et al., 2009). This enzyme regulates PI(3,5)P2 levels, which are involved in autophagy (Ferguson et al., 2009). FIG 4 is known to cause CMT4J if the two alleles are mutated (Chow et al., 2009). Heterozygous loss-of-function mutations in FIG 4 are found in 2% of sporadic and familial ALS patients (Chow et al., 2009).

Figure 4.

 Schematic overview of the mechanism of excitotoxic neuronal death. Glutamate (Glt) released from the presynaptic neuron stimulates AMPA receptors present on the postsynaptic neuron. If the GluR2 subunit is absent in the AMPA receptor complex, excessive intracellular calcium can induce neuronal death. Under normal conditions, glutamate is cleared from the synaptic cleft by glutamate transporters (EAAT2) present in the astrocytes, a process that is disturbed in ALS. See text for more details.

Angiogenin (ANG) mutations are found in both familial and sporadic ALS patients and will be discussed later.

Finally, we mention alsin, mutations in which cause recessive motor neuron disease, probably more resembling an infantile ascending paraparesis, and senataxin (SETX), mutations in which cause ALS4, which actually is more similar to a distal hereditary motor neuropathy with some pyramidal findings (Valdmanis et al., 2009). Dynactin (DCTN1) variants have been found in sporadic ALS patients (Munch et al., 2004, 2005) after the identification of the G59S mutation in the p150Glued subunit (encoded by DCTN1) of the dynactin complex in a family with a lower motor neuron syndrome with vocal cord involvement (Puls et al., 2003). The latter mutation has been modeled in mice (Laird et al., 2008). The contribution of DCTN1 variants to ALS may be limited (Vilarino-Guell et al., 2009). Studies investigating the possible pathogenic role of the DCTN1 variants are underway.

and from frontotemporal lobar degeneration

In many patients with ALS a pure motor phenotype is encountered, with no apparent cognitive impairments or behavioral problems. Although the actual combination of ALS and FTLD has been recognized for a very long time, a stronger link between the two diseases has been discovered in recent years. FTLD is the second most common type of dementia under the age of 65 (Ratnavalli et al., 2002; Neary et al., 2005). Degeneration in prefrontal and anterior temporal areas leads to variable clinical presentations of changes in personality and social conduct, and/or disturbances in language with impaired word retrieval and/or comprehension. The combination of ALS with FTLD is estimated to occur in ∼5–10% of patients in cohorts of FTLD (Neary et al., 2000) and may be as high as 15% in ALS cohorts (Lomen-Hoerth, 2004). More subtle cortical abnormalities in ALS patients and signs of motor neuron degeneration in FTLD patients (Lipton et al., 2004; Lomen-Hoerth, 2004; Mackenzie & Feldman, 2005) occur in a much larger proportion of patients.

FTLD is classified based upon the protein content of the neuronal inclusions found in the brain (Mackenzie et al., 2010). FTLD-tau is characterized by tau-positive inclusions and FTLD-U by tau-negative, ubiquitin-positive inclusions. Of the latter the majority are TDP-43-positive (FTLD-TDP) and a minority are FUS/TLS-positive (FTLD-FUS). About 40% of FTLD is familial, and then autosomal dominant in nature. Microtubulus-associated protein tau (MAPT) mutations account for ∼20–40% of these and give rise to FTLD-tau. Progranulin (GRN) and valosin-containing protein (VCP) mutations result in FTLD-TDP. Charged multivesicular body protein 2B (CHMP2B) mutations are rare and do not result in TDP-43-positive inclusions.

The K317M mutation in the MAPT gene has been observed in families with FTLD, Parkinsonism and ALS (Zarranz et al., 2005). Mutations in GRN and CHMP2B have also been observed in FTLD patients with symptoms and signs of motor neuron degeneration (Parkinson et al., 2006; Schymick et al., 2007a; Spina et al., 2007) and in rare sporadic ALS patients (Parkinson et al., 2006; Schymick et al., 2007a; Sleegers et al., 2008). Mutations in TARDBP and FUS are also encountered in some patients with the combination of FTLD and ALS (Benajiba et al., 2009; Blair et al., 2010). FUS/TLS mutations (Van Langenhove et al., 2010) and TARDBP mutations (Borroni et al., 2009) have been described in patients with pure FTLD.

Thus clinical, genetic and neuropathological data support the notion that ALS and FTLD are closely related and may represent two extremes of a spectrum of neurodegenerative disorders. The overlap is evident not only for some of the genetic forms but also in the majority of sporadic patients, in whom accumulation of the same disease protein is found. Therefore, observations made in ALS or FTLD also have implications for the entire group of neurodegenerative disorders and similar therapeutic strategies may be valuable in the two conditions.

for the understanding of sporadic ALS

Much of what we know about ALS comes from the study of the genetic forms of this disease. Essentially the pathogenesis of sporadic ALS remains enigmatic. It is generally accepted but certainly far from proven that it is the result of an interaction between an environmental factor and a genetic susceptibility. The latter has been investigated in genome-wide association studies, some of which we review below. In addition, we will mention the data from animal models which suggest that hypoxic stress may be involved in the mechanism of sporadically occurring motor neuron degeneration. Finally, we briefly review the evidence that glutamate-induced cell death (excitotoxicity) may contribute to the motor neuron degeneration seen in ALS.

Environment in sporadic ALS: the big unknown

In spite of its obvious relevance, very little is known about possible contribution from the environment. Many studies have been published but few results have been found to be reliable. The review of these studies is beyond the scope of this paper. We only mention a few intriguing findings. The incidence of ALS is quite uniform over Western populations overall. Increased incidences have been found in the Western Pacific island of Guam and the Kii peninsula of Japan. This has been related to excitotoxicity in the form of exposure to environmental toxins such as β-N-methylamino-l-alanine (BMAA), which can induce a similar disease phenotype in primates (Banack & Cox, 2003; Cox et al., 2003; Rao et al., 2006). BMAA is present in cycad seeds, which constituted a dietary item in these populations. In addition, BMAA is produced by cyanobacteria in diverse ecosystems and is present in brain and spinal cord tissues from sporadic ALS and AD patients as well as from brains of ALS patients, although the exact contribution of BMAA to human disease is still unclear (Vyas & Weiss, 2009).

Gulf war veterans may also have an increased risk of developing ALS (Horner et al., 2008) but, again, this phenomenon is poorly explored. Soccer players may equally have an increased risk, but lots of uncertainty remains (Wicks et al., 2007; Chio et al., 2009a).

The idea that an environmental toxin may play a role has also been approached genetically. Paraoxonases are enzymes encoded by the PON genes that are involved in detoxification of various exogenous compounds. Although initial association studies were contradictory, and a large genome-wide association study did not find an association (Wills et al., 2009), it is clear that further work is needed before PON polymorphisms are considered noncontributory.

Genome-wide association studies in sporadic ALS: the importance of axonal architecture and function

The basis for accepting a genetic factor in sporadic ALS is narrow. It is mainly based upon one twin study involving 77 twins in whom the inheritability was estimated to be between 0.38 and 0.85 (Graham et al., 1997). Several genome-wide association studies have been performed in order to identify this heritability at the molecular level, but their replication potential has been disappointingly limited (Dunckley et al., 2007; van Es et al., 2007, 2008, 2009; Schymick et al., 2007b; Cronin et al., 2008; Chio et al., 2009c; Landers et al., 2009; Simpson et al., 2009). Interestingly, three of them have identified factors related to the axonal compartment or vesicle release.

One study on 1821 sporadic ALS patients and 2258 controls from the US and Europe found no association in itself, but identified an SNP in the gene encoding the kinesin-associated protein 3 (KIFAP3) to be associated with disease duration (Landers et al., 2009). The variant associated with increased survival was associated with decreased KIFAP3 expression.

In another study involving 781 patients and 702 controls, a polymorphic marker in the elongation protein 3 homolog (ELP3) gene was found to protect against the occurrence of ALS (Simpson et al., 2009). This finding were shown to have biological relevance as, within the same study, an independent genetic screen in Drosophila identified two different loss-of-function mutations in the fly homologue of Elp3 that induced aberrant axonal outgrowth and synaptic defects. Furthermore, the knockdown of Elp3 in the zebrafish induced motor axonal abnormalities, and lower expression levels of Elp3 were found in the brains of individuals with the ALS at-risk genotype. Taken together, these results suggest that low Elp3 expression renders the motor neuron vulnerable to neurodegeneration (Simpson et al., 2009).

Interestingly, Elp3 is mainly localized in the cytosol in neuronal cells (Pokholok et al., 2002; Simpson et al., 2009), suggesting the existence of additional cytosolic targets for acetylation in these cells. Given the fact that α-tubulin acetylation is a key regulator of axonal transport (Westermann & Weber, 2003; Hammond et al., 2008) and that impairment of this process leads to neurodegeneration in general and to motor neuron degeneration in particular (De Vos et al., 2008), α-tubulin emerged as an obvious candidate for acetylation (Gardiner et al., 2007). In fact, an elegant study by Creppe et al. (2009) demonstrated that Elp3 acetylates α-tubulin and regulates migration and differentiation of cortical neurons. Furthermore, the role of Elongator on α-tubulin acetylation was recently corroborated in C. elegans, in which Elongator mutants also exhibited decreased neurotransmitter levels (Solinger et al., 2010), perhaps due to defects in vesicle transport and release. Of interest, mutations in Elp1, the scaffolding subunit for the enzymatically active Elp3, cause familial dysautonomia, a recessive degenerative disease of the autonomic nervous system (Anderson et al., 2001; Slaugenhaupt et al., 2001).

Recently, another genome-wide association study of 2323 individuals with sporadic ALS and 9013 control subjects identified unc-13 homolog A (UNC13A) as susceptibility gene for sporadic ALS (van Es et al., 2009). UNC13A is member of the UNC13 family of presynaptic proteins that regulates synaptic vesicle exocytosis and thus synaptic transmission. In C. elegans and Drosophila, elimination of the UNC13 homologue (unc-13 and dunc13, respectively) resulted in accumulation of docked vesicles at neuromuscular presynaptic release sites, thus suppressing neurotransmitter release (Aravamudan et al., 1999; Richmond et al., 1999). In C. elegans, unc-13 controls both cholinergic and GABAergic synapses (Richmond et al., 1999) whereas in mouse hippocampus, UNC13 homologue, Munc13, regulates both glutamatergic and GABAergic synapses (Varoqueaux et al., 2002, 2005). Moreover, Munc-13-deficient mice show only residual acetylcholine release at the neuromuscular junction and present morphological abnormalities in the muscle, neuromuscular synapses and spinal motor neurons (Varoqueaux et al., 2005).

UNC13 regulates neurotransmission by controlling both the docking (Siksou et al., 2009) and priming of synaptic vesicles into a fusion-competent state (Rosenmund et al., 2002). Considering the central role that UNC13 proteins play in neurotransmitter, including glutamate, release and the identification of the UNC13A gene as a susceptible gene for sporadic ALS, it is reasonable to postulate that UNC13A is contributing to the glutamate excitotoxicity seen in ALS. A better characterization of UNC13A in ALS mice models as well as in ALS patients is needed to establish a function for UNC13A in ALS.

Possible role of the hypoxia-induced response: the neurovascular link

Vascular endothelial growth factor (VEGF) is a well characterized angiogenic factor with a possible role in neurodegeneration (Bogaert et al., 2006). Its role in motor neuron degeneration was established when it was found that lowering VEGF levels in the mouse through a deletion in its hypoxia-sensitive regulatory sequence resulted in an adult-onset and progressive motor neuron disorder (Oosthuyse et al., 2001). The motor neurons showed vacuolar changes and the disease was denervating in nature. Subsequently, it was demonstrated that low VEGF levels were also found in the cerebrospinal fluid and spinal cord of ALS patients (Devos et al., 2004; Brockington et al., 2006), and that polymorphisms in the VEGF gene that are associated with low expression were overrepresented in at least a subset of ALS patients (Lambrechts et al., 2009). Intracerebroventricular administration of VEGF (Storkebaum et al., 2005), and virally mediated (Azzouz et al., 2004) or transgenic motor neuron-specific overexpression (Wang et al., 2007), increased the life-span of mutant SOD1 rodents, while decreasing VEGF expression worsened the motor neuron degeneration of mutant SOD1 mice (Lambrechts et al., 2009). Induction of VEGF in a zebrafish model of ALS rescued the axonal abnormalities (Lemmens et al., 2007). It was therefore thought that a vascular component contributed to the pathogenesis of ALS. This concept is supported by the finding of microhemorrhages in the spinal cord of ALS mice (Zhong et al., 2008). Deletion of mutant SOD1 from endothelial cells, however, did not affect disease in these mice (Zhong et al., 2009). Unexpectedly, it was found that VEGF is a trophic factor for motor neurons in vitro (Van Den Bosch et al., 2004), suggesting that this factor acts directly on neural cells. Moreover, VEGF-B, which is another member of the VEGF family but which has no angiogenic activity, has similar effects in mutant SOD1 models (Poesen et al., 2008). Interestingly, VEGF protects motor neurons from excitotoxic motor neuron death by upregulating the GluR2 subunit (see below) both in vivo and in vitro (Bogaert et al., 2010), possibly through the Akt pathway (Dewil et al., 2007b). This links the activity of this neurovascular factor to excitotoxicity. In conclusion, a vascular mechanism is not necessary but may contribute to the mode of action of VEGF. Whether administration of VEGF to human ALS patients is of therapeutic interest is currently under investigation.

Of interest is that missense mutations in the hypoxia-sensitive factor angiogenin have been identified in familial and sporadic ALS, albeit in a handful of patients (Greenway et al., 2006). Angiogenin is a member of the ribonuclease A (RNase) superfamily. It protects motor neurons from hypoxic death in vitro (Subramanian et al., 2008; Sebastia et al., 2009) and administration of angiogenin to mutant SOD1 mice increased their life span (Kieran et al., 2008). The mutations identified affect the protective effect of angiogenin but it is unknown whether this loss-of-function is of relevance to the in vivo effect in motor neuron degeneration.

These findings suggest that a (genetic) susceptibility of motor neurons to hypoxia may be a contributing factor in sporadic ALS, even independent of a vascular context. The question then arises whether hypoxia is a hazard the normal nervous system has to deal with, or whether an environmental factor contributing to sporadic ALS has a hypoxic element to it.

Excitotoxicity in ALS

Glutamate released from the presynaptic neuron is the main excitatory neurotransmitter in the central nervous system and plays a very important role in normal brain function. Glutamate stimulates ionotropic glutamate receptors on the postsynaptic neuron, a process resulting in the influx of sodium and calcium. Under pathological conditions, an increase in the synaptic glutamate levels and/or an increased sensitivity of the postsynaptic neuron to this glutamatergic stimulation can result in neuronal death, a phenomenon called excitotoxicity. Although overstimulation of N-methyl-d-aspartic acid (NMDA) receptors is classically involved in this process, motor neurons seem to be more sensitive to the overstimulation of the α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) type of glutamate receptors (Van Den Bosch et al., 2006).

There is overwhelming evidence for a role of glutamate-induced excitotoxicity in ALS mediated by the overstimulation of the AMPA-type of glutamate receptors (Van Den Bosch et al., 2006). The effect of riluzole in ALS patients and of AMPA antagonists in mutant SOD1 mice are two of these arguments (Bensimon et al., 1994; Canton et al., 2001; Van Damme et al., 2003; Tortarolo et al., 2006). In addition, intrathecal or intraspinal administration of AMPA receptor agonists induced motor neuron degeneration (Hugon et al., 1989; Ikonomidou et al., 1996; Corona & Tapia, 2007), and inhibition of glutamate uptake resulted in motor neuron death in organotypic spinal cord cultures by overstimulation of AMPA receptors (Rothstein et al., 1993; Saroff et al., 2000).

Motor neurons appear to be very sensitive to excitotoxicity for several reasons (Fig. 4). They combine the presence of a high number of calcium-permeable AMPA receptors (Carriedo et al., 1996; Van Den Bosch et al., 2000) with a low calcium-buffering capacity due to the low expression level of calcium-binding proteins (Alexianu et al., 1994). An immediate consequence of the lower amount of calcium-buffering proteins is that their mitochondria play a prominent role in calcium metabolism (Grosskreutz et al., 2010).

AMPA receptors are tetramers composed of a variable association of four subunits (GluR1–4) and the calcium permeability of the receptor is determined by the GluR2 subunit. Receptors with GluR2 have a very low calcium permeability compared to GluR2-lacking receptors. The calcium impermeability of GluR2-containing AMPA receptors is explained by the presence of a positively charged arginine instead of the genetically encoded neutral glutamine. This arginine residue at the Q/R site is introduced by the editing of GluR2 pre-mRNA, a process that is virtually complete under normal conditions. Motor neurons express low levels of the GluR2 subunit, leading to a higher calcium permeability of the AMPA receptor and an increased sensitivity to excitotoxicity (Greig et al., 2000; Heath et al., 2002; Van Damme et al., 2002; Kawahara et al., 2003). The role of GluR2 in motor neuron degeneration appears quite important. Editing of the GluR2 mRNA has been reported to be disturbed in sporadic ALS patients (Kawahara et al., 2004), suggesting an increased calcium permeability of their AMPA receptors and thus increased vulnerability to excitotoxicity. Overexpression of an ‘uneditable’ GluR2 subunit resulted in late-onset motor neuron degeneration in the mouse (Feldmeyer et al., 1999). Deleting the GluR2-encoding gene in mutant SOD1 mice accelerated motor neuron degeneration (Van Damme et al., 2005), while providing motor neurons with extra GluR2 increased significantly the life span of the mutant SOD1 mouse model (Tateno et al., 2004). Astrocytes from the ventral spinal cord determine the expression level of the GluR2 subunit in motor neurons and thus protect the motor neuron from excitotoxicity (Van Damme et al., 2007). The presence of mutant SOD1 in astrocytes abolished this protective effect, which may contribute to the non-cell autonomous nature of mutant SOD1-induced motor neuron degeneration.

Spinal motor neurons do not express calcium-binding proteins such as parvalbumin and calbindin, while motor neurons that are spared during ALS clearly express these proteins (Alexianu et al., 1994). This results in a much higher calcium-buffering capacity in these resistant motor neurons (Vanselow & Keller, 2000). Providing motor neurons with extra calcium buffering proteins resulted in a higher resistance of cultured motor neurons to excitotoxicity and a longer life span of the mutant SOD1 mice (Beers et al., 2001; Van Den Bosch et al., 2002). Given the absence of calcium-buffering proteins, mitochondria play a more important role in the calcium metabolism in motor neurons. Calcium overload of mitochondria resulted in depolarization of mitochondria and the generation of oxygen species (Carriedo et al., 2000), which may inhibit glutamate uptake in the neighboring astrocytes (Rao et al., 2003), thus establishing a vicious circle that can be interrupted by inhibiting the calcium-permeable AMPA receptor (Yin et al., 2007).

Increased extracellular levels of glutamate were found in the mutant SOD1 mouse model as well as in patients (Pioro et al., 1999; Alexander et al., 2000). Clearance of glutamate from the synaptic cleft is mainly taken care of by the glial transporter EAAT2 (also called GLT-1). In synaptosomes isolated from affected brain areas and spinal cord of ALS patients a diminished glutamate transport has been found, due to the loss of this protein (Rothstein et al., 1992, 1995). This was also found in mice and rats overexpressing mutant SOD1 (Bruijn et al., 1997; Howland et al., 2002). Mutant SOD1 damaged the intracellular carboxyl-terminal part of EAAT2 by triggering caspase-3 cleavage at a single defined locus, linking excitotoxicity and activation of caspase-3 as converging mechanisms in the pathogenesis of ALS (Trotti et al., 1999; Boston-Howes et al., 2006). In addition to mutant SOD1, axonal loss also resulted in the loss of EAAT2 expression in the astroglia (Yang et al., 2009). This is an immediate consequence of the loss of signal transmission from neurons to astroglia that is necessary for neuron-dependent astroglial EAAT2 transcriptional activation through the recruitment of the nuclear factor kappa B-motif binding phosphoprotein (KBBP), the mouse homolog of human heterogeneous nuclear ribonucleoprotein K (hnRNP K) and implicated in RNA splicing as well as in transcription (Bomsztyk et al., 2004). The recruitment of KBBP to the EAAT2 promoter is required for neuron-dependent EAAT2 transcriptional activation (Yang et al., 2009). The loss of EAAT2 can be a feedforward mechanism that propagates neuronal injury through the elevation of extracellular glutamate.

Crossbreeding EAAT2-overexpressing mice with mutant SOD1 mice delayed disease onset but had no effect on survival (Guo et al., 2003), while upregulation of the EAAT2 transporter by treatment with the β-lactam antibiotic ceftriaxone increased the life-span of the mutant SOD1 mice (Rothstein et al., 2005). This indicates that an induction of the EAAT2 expression and a higher clearance of glutamate from the synaptic cleft can protect motor neurons during ALS, at least in the mutant SOD1 mouse model.

Concluding remarks

Common themes in the pathogenesis of ALS

Although still hypothetical, it looks as if themes arise that may be common pathways leading to or contributing to motor neuron degeneration (see Fig. 5). Intracellular (axonal) transport (motors and cytoskeleton) is one of them (De Vos et al., 2008). KIFAP3 (kinesin), Elp3 (tubulin), UNC13A (vesicle release) and dynactin (dynein) are examples. Interestingly, mutations in other transport-related proteins have been identified in related motor neuropathies such as Charcot–Marie–Tooth disease (e.g. NEFL; Mersiyanova et al., 2000) and hereditary spastic paraparesis (e.g. KIF5A; Reid et al., 2002). Another emerging theme has to do with RNA processing (TDP-43, FUS/TLS, Elp3), a theme also encountered in spinomuscular atrophy, senataxin-related motor neuron disease and others (Lemmens et al., 2010). It can be predicted that more RNA-interacting proteins that play an etiologic or mediating role in ALS will be identified. Neurovascular molecules seem to establish another mechanism in ALS (VEGF, angiogenin) and related diseases (e.g. progranulin in FTLD; Lambrechts et al., 2006). The involvement of ER stress is yet another one (SOD1, VAPB and others; Kanekura et al., 2009). In addition, there is the mechanism of excitotoxicity that comes up in many models generated so far and that could explain the selective vulnerability of motor neurons (Van Den Bosch et al., 2006). Finally, there is the contribution of glial cells to motor neuron death (Ilieva et al., 2009). It remains to be seen how these themes will fit together. Most importantly, however, it is uncertain whether they are also at play in human motor neuron degeneration. This is difficult to investigate, as the human material we have is usually from patients in the terminal stages of disease, often poor in quality and, for many researchers, difficult to get hold of.

Figure 5.

 Schematic overview of the possible pathogenic factors involded in motor neuron degeneration. Increased production of reactive oxygen species (ROS) which lead to oxidative damage may occur due to defective activation of Nox2 induced by mutant SOD1 (mtSOD1) or mitochondrial dysfunction caused by mtSOD1. mtSOD1 can also be secreted to the extracellular space and be toxic for motor neurons. mtSOD present in astrocytes and microglia may also contribute to motor neuron degeneration. Mutant forms of FUS/TLS, TDP-43 (mtFUS/TLS and mtTDP-43, respectively) and SOD1 induce ER stress and heat-shock response (HSR), and may lead to proteasome dysfunction and subsecuent aggregation of these misfolded proteins. Intracellular inclusions of these proteins may result in toxicity to the cell by impairing axonal transport (axonal strangulation), and they also induce mitochondrial defects (energy failure), affecting axonal transport as well. FUS/TLS is involved in pre-RNA splicing and transport of mRNA, and TDP-43may regulate processing, stabilization and transport of RNA, highlighting the possible role of these proteins on RNA metabolism in ALS. Diminished levels of VEGF, a trophic factor for motor neurons, and defective release of synaptic vesicles may also contribute to motor neuron degeneration. Finaly, excitotoxicity plays an important role in motor neuron death (as explained in Fig. 4). See text for more details.

For ∼15 years, ALS research has been limited to mutant SOD1-induced motor neuron degeneration, as it was the only known cause of this disease. The discovery of other disease-causing mutations and the generation of animal models for them will allow a much broader approach and enable investigators to study compounds with a potential therapeutic effect in several different models. Hopefully these new opportunities will soon yield novel treatment strategies and make a difference for the many patients with ALS, their families and caregivers.

Acknowledgements

A.B. is supported by the Laevers Foundation for ALS research and Fundacao para a Ciencia e a Tecnologia of the Portuguese Government (Postdoctoral grant BPD/SFRH/2009/66777). P.V.D., L.V.D.B. and W.R. are supported by grants from the Fund for Scientific Research Flanders (F.W.O. Vlaanderen), from the University of Leuven (Methusalem) and the Interuniversity Attraction Poles (IUAP) program P6/43 of the Belgian Federal Science Policy Office. W.R. is supported by the E von Behring Chair for Neuromuscular and Neurodegenerative Disorders. The authors report no financial or other conflict of interest.

Abbreviations
AD

Alzheimer’s disease

ALS

amyotrophic lateral sclerosis

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid

ER

endoplasmic reticulum

FLTD

frontotemporal lobar degeneration

FLTD-U

FLTD–ubiquitin

FUS

fused in sarcoma

hnRNP

heterogenous nuclear nucleoprotein

PD

Parkinson’s disease

SOD1

superoxide dismutase 1

TDP-43

transactivation response DNA-binding protein with molecular weight 43 kDa

TLS

translocated in liposarcoma

VAPB

vesicle-associated membrane protein-associated protein B

VEGF

vascular endothelial growth factor

Ancillary