Amyotrophic lateral sclerosis (ALS) is a complex neurodegenerative disease with clinical, pathological and genetic overlap with frontotemporal dementia (FTD). No longer viewed as one disease with a single unified cause, ALS is now considered to be a clinicopathological syndrome resulting from a complex convergence of genetic susceptibility, age-related loss of cellular homeostasis, and possible environmental influences. The rapid increase in recent years of the number of genes in which mutations have been associated with ALS has led to in vitro and in vivo models that have generated a wealth of data indicating disruption of specific biochemical pathways and sub-cellular compartments. Data implicating pathways including protein misfolding, mRNA splicing, oxidative stress, proteosome and mitochondrial dysfunction in the pathogenesis of ALS reinforce a disease model based on selective age-dependent vulnerability of a specific population of cells. To the clinical neurologist, however, ALS presents as a disease of focal onset and contiguous spread. Characteristic regional patterns of involvement and progression suggest that the disease does not proceed randomly but via a restricted number of anatomical pathways. These clinical observations combined with electrophysiological and brain-imaging studies underpin the concept of ALS at the macroscopic level as a ‘system degeneration’. This dichotomy between cellular and systems neurobiology raises the fundamental questions of what initiates the disease process in a specific anatomical site and how the disease is propagated. Is the essence of ALS a cell-to-cell transmission of pathology with, for example, a ‘prion-like’ mechanism, or does the cellular pathology follow degeneration of specific synaptic networks? Elucidating the interaction between cellular degeneration and system level degeneration will aid modeling of the disease in the earliest phases, improve the development of sensitive markers of disease progression and response to therapy, and expand our understanding of the biological basis of clinical and pathological heterogeneity.
Introduction: Clinical and pathological features indicate that ALS is a system failure
The first clinicopathological descriptions of ALS by Jean-Martin Charcot in Paris in the mid-19th century are contemporaneous with the evolution of the concept of upper and lower motor neurons and the recognition of the anatomical role both of the anterior horn cells of the spinal cord in the actualisation of voluntary movement and of the corticospinal tracts in conveying descending modulation and control (Charcot & Joffroy, 1869). Amyotrophic lateral sclerosis is the commonest form of motor neuron disease, and the third most frequent neurodegenerative disease after Alzheimer's disease and Parkinson's disease. The annual population incidence is 2/100 000 per year, and there appears to be a uniform geographical incidence of ‘sporadic’ forms of ALS, although monogenic forms, which account for 5–10% of cases, show variation in incidence between populations, presumably due to founder effects (Byrne et al. 2011; Chio et al. 2013). The malignant nature of the disease progression is indicated by the fact that the median survival is approximately 2 years from diagnosis and 3 years from symptom onset (Chio et al. 2011). However, there is considerable clinical variation, with a long ‘tail’ in the survival curve; approximately 10% of patients are still alive at 10 years after onset. A better understanding of the biological underpinning of this variation in outcome could shed considerable light on the nature of ALS.
There is a general consensus among neurologists who specialise in ALS that the condition has a focal clinical onset. The overwhelming majority of patients are clearly able to describe alteration in motor function in one body region at a specific time, typically within a timescale of weeks to months. Approximately 30% of cases begin in one upper limb, 35% in one lower limb and 30% in the muscles of speech and swallowing, with the remainder accounted for by initial respiratory failure or various forms of axial weakness (head-drop or truncal weakness). Where the onset is characterised by weakness in more than one body region at presentation, this is generally a manifestation of rapidly progressive disease with a poor prognosis. Furthermore, the disease does not progress randomly, but through a discrete number of recognisable patterns of clinical evolution (Chio et al. 2011). The fact that such patterns do not vary between populations with different genetic backgrounds implies that progression is an inherent feature or an emergent property of the way in which the corticomotorneuronal system promoting voluntary movement fails.
Although classical ALS is conventionally divided into spinal (limb) or bulbar (speech and swallowing) onset, this simple dichotomy fails to capture accurately the clinical heterogeneity of the condition. A more useful approach to classification, allowing more accurate prognostication and also proactive management of nutrition and respiration, is to consider the site of symptom onset, the intrinsic rate of progression (e.g. the slope of functional decline using a validated clinical measure such as the ALS Functional Rating Scale), whether symptoms remain regionally localised or rapidly spread to other anatomical segments, the timing of involvement of respiratory and bulbar muscles, and the age of the patient. Older age is an independent variable predicting short survival, which may reflect a higher burden of co-morbidities, an age-related lack of functional reserve in the motor system, or that an age-related neurodegenerative process is likely to be more aggressive in the aged brain. The most robust predictor of survival is the latency from symptom onset to confirmed diagnosis in a specialist clinic. A firm diagnosis of ALS requires involvement of more than one anatomical region (bulbar, upper limb, lower limb); weakness and wasting initially confined to one limb is the commonest presentation of ALS, but has a broad differential diagnosis. Slowly progressive forms of ALS are often atypical in presentation, increasing diagnostic uncertainty and requiring more ancillary diagnostic tests.
Pathologically, ALS is characterised in the majority of cases by ubiquitinated inclusions, which stain for the protein TDP-43 (Neumann et al. 2006). Careful studies correlating regional distribution of pathology with the clinical burden of disease in life suggests that there is a radial gradient of involvement from an initial clinical and pathological focus, and that this proceeds in a non-random fashion across contiguous anatomical segments (Ravits et al. 2007). Furthermore, there is evidence of simultaneous involvement of the corticospinal tract and spinal motor neuron circuits. Although there are many patients with clinical features showing upper or lower motor neuron predominance, it is the combination of the two that is so characteristic of ALS, particularly in its most aggressive forms.
In addition to the undoubted contribution of aging as a critical co-factor in the development of ALS, attention has also focused on the possible role of developmental factors in neurodegenerative diseases like ALS. It is clear that Alzheimer's disease and Parkinson's disease are the clinical endpoints of degenerative system failures that begin decades before (Bateman et al. 2012). It is possible that the way in which the corticomotorneuronal system is established in development varies significantly between individuals and that this variation is governed partly by genetic influences. The observation that patients with ALS are more likely to have a high level of physical fitness, a lower body-mass index and to have fewer other co-morbidities suggests that the disease might be more likely to affect a population of individuals with a motor system with specific characteristics (Veldink et al. 2005; Turner et al. 2011).
Direct corticomotorneuronal connections to anterior horn cells appears to be a recent evolutionary development, unique to higher primates. Human neocortical evolution over the last 200 000 years has been rapid and is presumably under strong selection pressure. The number of fibres in the corticospinal tract has increased by about 30–40% in the 5 million years since a common human-chimpanzee ancestor (Lemon, 2008). It has been suggested that this rapid and recent evolution of the motor system, in large part driven by the development of language and manual dexterity, while advantageous at a species level, may have come with an inherent penalty of increasing the possibility of system level failure in individuals with the appropriate genetic susceptibility profile (Eisen, 2009). For example, the clinical phenomenon of the ‘split hand syndrome’ in which there is preferential wasting of the thenar eminence, has been proposed to be a reflection of the corticomotorneuronal representation and input for fractionated thumb movement (Weber et al. 2000).
Although the clinical association between ALS and frontotemporal dementia has been known for decades, it has only been the subject of serious research activity in the last 10 years. It is now clear that up to 50% of patients with ALS show impairments on psychological tests of executive function (planning, attention and reasoning) if tested in sufficient detail, and in approximately 5% of cases FTD is the dominant clinical feature (Phukan et al. 2012). The neuropathology of ALS and FTD cases characterised by ubiquitin stained inclusions is identical at the cellular level, demonstrating TDP-43 inclusions and loss of nuclear staining, and varies only by distribution. With some notable exceptions (e.g. SOD1, see below) mutations in the genes associated with ALS can also cause FTD. Remarkably, in families with an identified genetic mutation, individuals can manifest ALS, FTD or combined ALS-FTD, suggesting that other events, or multiple ‘hits’, interact with a genetic mutation to determine the manifestation of the disease within the confines of a specific neuroanatomical pathway. An understanding of why two apparently different and unconnected neuroanatomical networks (corticomotorneuronal and frontotemporal) are involved in the same disease, and often combined in the same patient, is a priority for clinical and pathological research.
The cellular basis of motor neuron vulnerability in ALS
In addition to neuropathology, the most important biological insights in recent decades have come from the identification of genes associated with monogenic forms of ALS. Approximately 5% of patients have other affected first degree or second degree relatives and are considered to have familial ALS. There are now more than 10 genes for which there is general consensus of mutations having a direct causal relationship with ALS and others in which there is incomplete evidence (Al-Chalabi et al. 2012). Depending on the population studied, 35–45% of the familial cases are explained by a dynamic hexanucleotide expansion mutation in the C9orf72 gene on chromosome 9 (DeJesus-Hernandez et al. 2011; Renton et al. 2011), 20% of fALS by mutations in SOD1 (Rosen et al. 1993) and 4–5% each by mutations in TDP-43 or FUS/TLS (Sreedharan et al. 2008; Vance et al. 2009). Therefore, approximately 25% of genetic cases are yet to be explained. Importantly, although there are some observable trends and clinical clues such as the co-occurrence of FTD in families with C9orf72 mutations, it is not possible to distinguish reliably familial from sporadic ALS on clinical features alone (Byrne et al. 2012). This suggests that research into genetically determined cases of ALS would yield information of general relevance for the much commoner sporadic ALS. In fact, for each of the genes identified as causing fALS, mutations have been identified in apparently sporadic cases, at a low level (< 1%) for SOD1, TDP-43 and FUS, but remarkably in 5–7% of sporadic cases in populations of European descent for C9orf72 (Majounie et al. 2012). This indicates that that the strict distinction between familial and sporadic cases is an artificial dichotomy with genetic factors more likely to operate as a continuum, encompassing familial cases with high penetrance at one extreme, and true sporadic cases at the other (Talbot, 2011).
The number of individual upstream genetic factors contributing to ALS is therefore great and attempts to use the known functions of these genes to form functional groups that might give insights into motor neuron vulnerability are only partially successful. These are reviewed in detail elsewhere (Robberecht & Philips, 2013) but can be summarised under a number of broad themes:
Both TDP-43 and FUS are RNA-binding proteins. Evidence suggests that each protein regulates the processing of non-overlapping subsets of transcripts by binding to conserved intronic motifs, which are possibly commoner in neurons. In each case the number of identified gene targets is greater that 5000 (Lagier-Tourenne et al. 2012). In keeping with this apparently generalised splicing function, inactivation of TDP-43 or FUS by gene targeting or RNAi knockdown is lethal to cells. It is therefore difficult to envisage that loss of function mutations can be the sole mechanism of motor neuron degeneration in ALS, though aggregation and mislocalisation may be the final trigger to cell death. In keeping with this, loss of TDP-43 from its normal location in the nucleus is a pathological hallmark of ALS in 95% of cases (Neumann et al. 2006) (Fig. 1). This suggests that the cellular ‘lesion’ in TDP-43-associated ALS is a shift in the cellular compartment from nuclear to cytoplasmic, presumably induced by age-related cellular stress. When this has reached a critical level, the nucleus is depleted of TDP-43, and this presumably triggers cell death (Igaz et al. 2011). A number of other RNA-binding proteins are mutated in ALS and also non-ALS motor neuron diseases such as spinal muscular atrophy (Baumer et al. 2010a) but given the diversity of function of such proteins no overarching mechanism has yet emerged to explain this relatively cell-selective degeneration. Mutations in C9orf72, the commonest cause of ALS and also frontotemporal dementia, may lead to disruption of RNA-binding proteins through aberrant interaction with the expanded hexanucleotide repeat, but other mechanisms have also been proposed.
Proteins involved in the maintenance of cellular protein homeostasis form a second group. These include ubiquilin-2, sequestome-1, optineurin and VCP, mutations in which have been described at a low frequency in ALS (1–2% of familial cases). As protein aggregation is a common feature of neurodegenerative diseases, this pathway is a mechanistically plausible point of vulnerability in ALS. However, as with RNA-binding proteins, there is currently no clear explanation for the selective vulnerability of motor neurons over other CNS cells. SOD-1, which until the discovery of mutations in C9orf72 was the commonest and most studied cause of familial ALS, appears to be a protein with an unusual propensity to misfold and can be tentatively considered to be in this pathway.
A range of proteins functioning in axonal transport (including neurofilament subunits, peripherin, VAPB and others) have been implicated in rare families with motor neuron diseases. Although these can be classified within the ALS spectrum, there are some atypical features such as lower motor neuron predominance and slow progression (Robberecht & Philips, 2013).
Even in apparently simple monogenic forms of ALS it is necessary to invoke a multiple hit mechanism in which aging is a critical co-factor. Although the median age of onset of fALS cases is approximately 10 years earlier than sALS cases, the majority of familial cases, especially those due to C9orf72 mutations, still occur in the same age group (55–75 years) where most sALS cases arise. The putative age-related changes in cells which after many years lead to loss of tolerance of genetic mutations, many of which are highly toxic in cell culture models, include damage from oxidative stress exceeding the buffering capacity of the cell, damage to the genome and to sub-cellular organelles and alterations in the functional capacity of the proteasome.
How and when does ALS begin and how does ALS spread?
The latency between first symptoms and a formal diagnosis of ALS in a neurology specialist clinic varies, but averages 1 year (Chio et al. 2011). What is happening in the nervous system prior to the point of symptom onset is currently a matter of speculation. Although clinical observation suggests that such patients have been functioning normally right up until symptom onset, and a limited longitudinal neurophysiological study in SOD1 mutation carriers indicates that the number of motor units is normal up to few months prior to first symptoms (Aggarwal & Nicholson, 2001), the idea that ALS might arise abruptly from a normal nervous system is only one possibility. If ALS, in common with AD and PD, has a prolonged period of system dysfunction without obvious clinical manifestation, this has important implications, as a pre-clinical phase is likely to be the best target for therapeutic trials.
Most neurodegenerative diseases appear to have three phases (Fig. 2). During normal development of the nervous system, genetic variation between individuals may be an important influence in determining later life susceptibility to degeneration, but there is no evidence of dysfunction. In the pre-clinical phase, susceptible individuals are likely to have abnormal neuronal function, at a network or individual synapse level, but this does not lead to clinical symptoms either because the system in question has intrinsic reserve capacity or because there are other compensatory mechanisms in operation. It is therefore only when the reserve is exceeded or the compensatory mechanisms break down, that patients develop symptoms and the clinical phase begins.
The apparent focal onset of ALS in clinical terms might suggest that a precise ‘event’ located in time and in anatomical space triggers the disease. However, it is just as likely that the whole corticomotorneuronal system is in a state of partially compensated decline and the ‘site of onset’ simply reflects stochastic events that determine where a generalised disease rises above the ‘clinical horizon’. In this case, studying the clinical heterogeneity of ALS and stratifying patients according to extremes of phenotype, even if it is important for clinical management, might not provide much meaningful insight into the biological basis of the motor neuron degeneration. An intriguing observation in support of this possibility comes from families in which a single genetic mutation can give rise to almost every possible clinical sub-type of ALS (Fig. 3). In this family a single genetic mutation manifests as typical spinal or bulbar onset ALS, a slowly progressive lower limb variant that is lower motor neuron predominant or as an upper motor neuron predominant form described as ‘stiff limbs’. This degree of heterogeneity is not uncommon in genetically determined ALS, although there are some general trends with different genetic subtypes. For example, SOD1-related ALS almost never causes significant cognitive impairment (Wicks et al. 2009), FUS-associated ALS can be an aggressive form with onset in young adulthood (Baumer et al. 2013) and there are some broad pointers to a C9orf72 mutation such as a strong family history of dementia (Byrne et al. 2012).
While understanding the initiating steps of ALS is likely to be important for the development of therapies, of equal relevance could be understanding the mechanism of disease propagation and finding ways to halt this. A hypothesis that has received much attention recently is that neurodegeneration may spread from cell to cell by a mechanism akin to that seen in prion diseases (Nonaka et al. 2013). A number of ALS-associated proteins (principally TDP-43 and FUS) contain ‘prion-like’ domains, which in principle could promote protein aggregation. However, as well as the fact that even in classical prion diseases such as Creutzfeldt–Jakob disease (CJD) the method of spread has not been well characterised, there is only indirect evidence that TDP-43 actually undergoes prion-like aggregation, a process which is defined as permissive templating of native protein by a conformationally altered, mutant form of protein. Furthermore, other classical features of prion diseases such as evidence of infectivity are currently lacking.
There are a range of other possible mechanisms with equal explanatory power, such as transynaptic spread of mutant protein, spread through adjacent glia or even through circulating factors. Finally, the higher order network organisation of the corticomotorneuronal system, which is incompletely understood, could provide an explanation for how the disease appears to spread through anatomically contiguous segments if the equilibrium of inter-related sub-networks of fibres is altered by cell death in an adjacent region. The elucidation of such mechanisms will require new and more sophisticated approaches to the analysis of animal models and to human biomarker studies.
Conclusions: new opportunities to integrate cellular and systems based research in ALS
Since its initial description in the late 19th century, ALS has been recognised by clinical neurologists as a relatively stereotyped diagnostic entity and treated, for the purposes of clinical enquiry and clinical trials, as a single disease. In the modern era of molecular and cellular biology this concept is no longer tenable and has been replaced by a view of ALS as a clinicopathological syndrome with multiple upstream causes. However, the clinically stereotyped nature of the disease progression, and emerging evidence from multimodal brain imaging studies, suggests that a system level vulnerability is an important aspect of the biological underpinning of ALS. The recent discovery that approximately 10% of all ALS cases, and a similar number of FTD cases, are due to an expanded hexanucleotide repeat in the C9orf72 gene provides an important new departure for ALS research. Much has been learned about the pre-clinical phase of disorders such as Huntington's disease by longitudinal multimodal studies of gene carriers (Tabrizi et al. 2013). The large number of potential subjects who could be recruited from ALS clinics, both those with early disease and pre-symptomatic carriers of the mutation, will for the first time allow studies of the pre-symptomatic phase of ALS that may contribute to a complete understanding of the relationship between a genetic mutation and the corticomotorneuronal system in which the disease reveals itself.
Work in Professor Talbot's laboratory is supported with funding from the MND Association.