Controversies and priorities in amyotrophic lateral sclerosis

Has the concept of ALS as one disease become untenable?

From a clinical viewpoint, ALS has been identified by most neurologists as a single entity because it characteristically involves a recognised pattern of progressive motor neuronopathy, with respiratory failure as the mode of death for most patients. This core phenotype is confidently diagnosed by an experienced clinician when progression is rapid, although the mean diagnostic delay has stubbornly remained around 1 year from symptom onset,¹⁰ reflecting the reality of more slowly progressing cases. Although ALS is a clinically recognisable condition, there are now strong reasons to consider the classic phenotype as a manifestation of a wider aetiopathogenic spectrum.

Although median survival is 3 years from symptom onset, the survival curve for the remaining cases tapers out to 20 years or more. Many of the long-term survivors appear to have either pure upper motor neuron (termed primary lateral sclerosis) or lower motor neuron (termed progressive muscular atrophy) involvement.¹¹ and ¹² There is an urgent need for reliable subclassification that goes beyond the number of body territories with combined clinical upper motor neuron and lower motor neuron signs that underpin the current diagnostic criteria.¹³ There is increasing frustration with this diagnostic framework (panel 1), which was developed principally as a research aid, since a substantial proportion of patients die with a diagnostic label of “possible ALS”¹⁴ even though the clinical progression is the same as that of “definite ALS”, but with simply fewer demonstrable upper motor neuron signs.

At least 16 genes of major (typically dominant) effect are associated with ALS.¹⁵ Recognisable phenotypes have been reported for some known mutations (table 1), such as the SOD1 A4V variant, which produces a rapidly progressive, primarily lower motor neuron syndrome; ¹⁶ and FUS mutations linked to a juvenile onset, primarily lower motor neuron syndrome. ¹⁷ However, perhaps the most striking development has been the recent recognition of a characteristic phenotype associated with the relatively common C9orf72 hexanucleotide repeat expansion. Patients with the expansion have a younger age of onset, shortened survival, increased rates of cognitive and behavioural impairment, and a strong family history of neurodegenerative disease. ¹⁸ and ¹⁹ Postmortem studies also suggest that C9orf72-associated ALS has a different pathological signature to other forms of ALS.

Heterogeneity in neuropathology can also be discerned. Cytoplasmic inclusions of TDP-43 are found in nearly all autopsied cases of sporadic ALS, and suggest a consistent final common pathway. However, such inclusions are notably absent in ALS associated with pathogenic mutations of SOD1, ²⁰ and the pathophysiological processes underlying this and other abnormal protein aggregation have not been fully elucidated. Expansions of the C9orf72 hexanucleotide repeat have a characteristic neuropathological signature, with TDP-43 inclusions but in a different distribution including hippocampal and cerebellar regions, and with prominent staining for p62, ²¹ supporting the view that different primary pathophysiological processes have a final common, clinically indistinguishable, endpoint. As such, the clinical, pathological, and genetic frameworks traditionally used to explain disease heterogeneity remain incomplete, with the corollary that therapeutic interventions might need to be tailored across different subgroups of patients, on the basis of the prevailing upstream pathological events. Understanding of the physiological roles of TDP-43 and C9orf72 will be an essential first step.

Although cognitive and behavioural impairment are now recognised features of ALS, not all patients become impaired. Population based studies suggest that 14% of incident cases have frontotemporal dementia, and a further 30% have evidence of cognitive impairment, mostly but not exclusively in executive dysfunction.²² One of the strongest predictors of cognitive or behavioural impairment in ALS is the presence of the C9orf72 repeat expansion. Taken together, these findings point perhaps to a distinct phenotype in ALS that includes roughly 10% of all cases. Patients with early features of executive impairment progress more rapidly, whereas those who are cognitively normal at the time of diagnosis have a better prognosis and largely do not exhibit features of cognitive decline on serial testing. ²³ Furthermore, cognitive impairments common to classic ALS have been notably less prevalent in progressive muscular atrophy ²⁴ and primary lateral sclerosis. ²⁵ Such observations might reflect an increased resistance in some individuals to the penetration of pathology beyond its initial clinical focus, with obvious therapeutic value if the biological substrate for this resistance could be identified.

Going forward, clinical and genetic studies will require stratification with respect to the presence or absence of the C9orf72 expansion and deeper analysis of clinical phenotyping and genotyping will clearly be needed, including development of a greater understanding of the extreme phenotypes. This process has recently been commenced in several population cohorts with C9orf72-related repeats, ¹⁸ and has shown distinct clinical and family signatures. Separately, studies of the typically more slowly progressive clinical phenotypes such as primary lateral sclerosis ²⁶ and progressive muscular atrophy ²⁷ have also been undertaken, although major gaps persist in our understanding of molecular changes associated with these disorders, in addition to delineation of the extent and timing of upper motor neuron versus lower motor neuron involvement. ²⁸ Moreover, there is a particular lack of detailed clinicopathological studies in primary lateral sclerosis, ²⁹ regarded by some as a separate disease altogether. The development and application of novel in-vivo neurophysiological and neuroimaging methods to specific clinical subtypes ³⁰ and ³¹ have the potential to circumvent the end-stage pathology perspective of traditional histopathology.

Should ALS even be regarded as a neuromuscular disease?

Many ALS physicians are also specialists in neuromuscular disease, and the characteristic presence of progressive muscle wasting, weakness, and fasciculations clearly reflects degeneration of anterior horn cells and lower motor neurons. The electromyogram is an adjunct to clinical evaluation, and the Awaji criteria, based on features of denervation and reinnervation, have recently been established as a useful diagnostic aid.³² However, as originally recognised by Charcot, corticospinal tract involvement is nearly always seen pathologically, even in cases without any clinical evidence of upper motor neuron involvement. The emergence of a close biological relation between ALS and frontotemporal dementia argues strongly in favour of a realignment of specialist expertise towards neurodegenerative and behavioural clinics, rather than concentration exclusively within the neuromuscular domain. And although the transgenic SOD1 model might have some value for understanding of the process of anterior horn cell degeneration, the extramotor manifestations of the disease, including significant cognitive and behavioural impairments, command attention that is not provided by existing animal models.

Further to such discussion, a causal primacy of lower motor neuron over upper motor neuron degeneration³³ remains an issue of debate. Many of the initial pathological changes in models of ALS occur in the peripheral motor system, supporting a dying-back view of pathogenesis.³⁴ However, ALS can also be viewed primarily as a disease of the upper motor neurons, or corticomotoneurons, which connect monosynaptically with anterior horn cells,³⁵ supported by the clinical observation that the oculomotor, abducens, and Onuf's motor nuclei, which all lack direct corticomotorneuron connections, are strikingly resistant to degeneration.

At the cortical level, premotor and motor regions might seem dominant, although frontal and anterior temporal regions, in addition to subcortical pathways, are key structures. Because motor planning, motor abstraction, emotional expression, and language, which are important aspects of human neocortical evolution, are all affected in ALS, these might hold clues to selective vulnerability and disease progression.³⁶ As such, future studies need to consider ALS as a systems degeneration, rather than focusing on individual neuronal populations. Advanced MRI is uniquely able to study the brain at such a network level. Application of graph theory to structural data concerning white matter tract integrity,³⁷ and independent component-type analysis to cerebral functional data,³⁸ in parallel with detailed clinical and neuropsychological characterisation, offers a comprehensive in-vivo approach to probing disease activity in ALS.³⁹ and ⁴⁰

Can apparently sporadic ALS be inherited?

In populations of European origin, only 5–10% of patients with ALS report a family history of the disease (figure 2). It seems likely that there is a complex set of genetic determinants, each with potentially small but significant effect.¹⁵ Standard practice has been to reassure patients without an apparent family history of ALS. However, early studies of the C9orf72 hexanucleotide repeat expansion suggest that the repeat expansion is also present in a sizeable minority of apparently sporadic cases, ¹⁹ and ⁴¹ and careful interrogation of kindreds of these patients, supported by death certification, shows increased rates of neurodegenerative conditions, most notably frontotemporal dementia. Such findings suggest that a family history of frontotemporal dementia, which might have been misclassified as Alzheimer's disease, must now be considered in a revised definition of familial ALS.

Furthermore, the definition of familial ALS is increasingly being recognised as inconsistent.⁴² The definition is based on the presence of known disease in other members of a kindred. In this context the size of the kindred also becomes important in determining heritability. Specifically, the presence of a small number of first-degree and second-degree relatives might erroneously lead to a diagnosis of sporadic ALS, whereas large extended kindreds might serve to increase the risk of a second case of ALS occurring by chance alone. Perhaps more reassuringly for patients with sporadic disease is the limited evidence obtained from clinic-based populations to date, suggesting that although there is approximately an eight-fold increase to first-degree relatives of patients with ALS, the absolute increase in risk seemed to be very small in a follow-up study over 16 years.⁴³

As with previous familial linkage to ALS, there appears to be incomplete penetrance of the C9orf72-related repeat expansion. Incomplete family histories and small family sizes might blur the distinction between sporadic and familial ALS further, suggesting that the reassuring figures for recurrence within offspring cannot be applied across all patients. ⁴⁴ A clearer definition of familial ALS, followed by more extensive analyses of disease penetrance, might ultimately lead to a change in the routine counselling of patients. Whether the routine screening of all apparently sporadic cases of ALS for the known familial genetic abnormalities is the optimum approach at present, with the potential to register psychological harm, needs urgent debate.

The explosion in ALS genetics has challenged other long held doctrines relating to known genes of major effect, and evidence of an oligogenic cause in familial and sporadic ALS further compounds the problem of ALS heritability. Until recently, there had been an assumption that single genes of major effect with incomplete penetrance accounted for most cases of familial ALS. As each new ALS gene has been identified, ALS DNA banks have been screened to identify pathogenic variants in the new gene. DNA samples containing known variants are then excluded from further analyses. However, emerging evidence suggests a more complex genetic cause, in which the phenotypic variability might be due to a small number of genetic variants that co-occur with greater frequency than would be expected by chance. Recently described examples include FUS and TARDBP mutations in combination with ANG mutations and C9orf72 repeat expansions combined with TARDBP, SOD1, and FUS mutations. ⁴⁵ If replicated, these findings will have important implications for the design and interpretation of ALS genetic studies. Whether this explanation accounts for the substantial phenotypic variations within C9orf72 kindreds remains to be established.

That the most important gene in ALS was identified with conventional genetic approaches, rather than the newer technologies of genome-wide association studies (GWAS) and next-generation sequencing perhaps seems ironic. Whereas the chromosome 9 locus was identified a decade before its characterisation in sporadic ALS,⁴⁶ the intronic repeat expansion could never have been identified with more modern sequencing methods because of the insensitivity of these technologies to detect very large intronic repeat expansions.

The history of GWAS in ALS started with fairly small studies (table 2) that initially had negative results or results that were not replicated in independent studies.⁴⁷ and ⁵⁰ Through international collaborations and data sharing, ongoing combined analysis of available data has now provided consistent results. Nevertheless, by contrast with other neurodegenerative diseases such as Parkinson's or Alzheimer's disease, the results of GWAS looking for common variability have been disappointing in ALS, again suggesting that ALS is not one clinical entity, but is heterogeneous by nature. The best example of consistent GWAS results to date was the identification of the chromosome 9 locus, which appears aberrant in 5–8% of sporadic ALS. The larger GWAS effectively fine mapped the 9p locus from several megabases to a 110 kb region containing three genes, allowing the C9orf72 discovery. Nevertheless, continuing international collaboration will be needed to generate sample sizes that are comparable with other complex diseases, such as diabetes and inflammatory bowel disease. Also, the combined analysis of GWAS data of ALS with other neurodegenerative diseases (eg, Alzheimer's disease, Parkinson's disease, and frontotemporal dementia) will probably be of great value.

In view of the recent findings of intronic variation and repeated sequences in ALS, and the notion that rare genetic variants as opposed to the common genetic variants underlie pathogenesis, high-throughput exome sequencing and analysis of non-coding DNA by whole-genome sequencing in large samples are now needed to fully elucidate the genetic contribution to ALS pathogenesis.⁶¹ As has been the case for other diseases such as breast cancer, the discovery of one gene that affects a large minority of patients will probably help to identify other genes, since all datasets will now need to be reanalysed after removal of people with the C9orf72 expansion.

Is there a premorbid phenotype that reflects vulnerability to ALS?

Despite conflicting results regarding premorbid athleticism from case-control studies,⁶² and ⁶³ a clinical impression remains that patients who develop ALS frequently have a higher than average level of physical activity or fitness, perhaps best exemplified by the American baseball player Lou Gehrig. Whether the findings of reduced coronary artery disease premorbidly in patients with ALS,⁶⁴ and a cardioprotective vascular profile of patients⁶⁵ and their first degree relatives,⁶⁶ are significant in terms of premorbid fitness or athleticism remains uncertain. Several occupations likely to be linked with high physical activity and fitness have been associated with increased incidence of ALS, including professional footballers in Italy, the military, and manual workers.⁶⁷ and ⁶⁸ Separately, a lower than average body-mass index appears more commonly associated with ALS, linked to the observation of hypermetabolism that often develops in ALS. The biological processes accounting for these changes remain unclear. Mitochondrial impairment has been long considered as a possible pathogenic mechanism in neurodegeneration, and increasing the efficiency of oxidative phosphorylation remains one of the proposed mechanisms of the drug dexpramipexole, recently the subject of a phase 3 trial for ALS.⁶⁹

It has been argued that physical activity might be harmful to a minority of the population who carry a complex genetic profile that adversely affects the metabolic response to vigorous muscle activity, including oxidative stress.⁷⁰ Many patients with ALS specifically ask whether they can safely continue to exercise regularly without fear of accelerating their disease.⁷⁰ Although low-grade exercise programmes might even have some benefit,⁷¹ there is no firm evidence that exercise exerts a harmful effect, although avoidance of very strenuous activity seems reasonable.

An alternative formulation, which is not mutually exclusive, is that athleticism or fitness in patients with ALS is a reflection of a shared set of genetic determinants that predispose to neurodegeneration, without there being a direct causal link between ALS and exercise.⁶² In a cortically focused model of pathogenesis there might even be preferred loci for neurodegeneration,⁷² and ⁷³ perhaps reflected in regional involvement,⁷⁴ and an athletic profile might then also reflect a distinct motor system architecture. There would be an evolutionary advantage for a configuration associated with fitness, that is perhaps only rendered disadvantageous for a minority as a result of the relatively recent increases in life expectancy. Potential evolutionary trade-offs have been noted in other neurodegenerative disorders.⁷⁵

When does ALS begin?

Longitudinal studies in presymptomatic individuals carrying pathogenic SOD1 mutations have shown apparently abrupt changes in both lower motor neuron loss ⁷⁶ and the development of cortical hyperexcitability ⁷⁷ within a year of symptom onset. Whether the nervous system of people who later develop ALS was normal immediately before disease onset or whether, like in Alzheimer's and Parkinson's diseases, a more prolonged presymptomatic phase of disease might have been detectable, seems uncertain. In support of a prolonged presymptomatic phase, changes in parenchymal metabolites (N-acetylaspartate in relation to choline and myoinositol) were detected in the spinal cord of presymptomatic individuals carrying pathological SOD1 mutations long before clinical disease onset. ⁷⁸ The intriguing suggestion that the risk of ALS develops from factors operating at the time of conception or perhaps during early development warrants further consideration. ⁷⁹

Detailed study of presymptomatic individuals carrying pathogenic mutations could provide unique insight into the nature of the earliest events before the tipping point of symptom onset. This pursuit is likely to be relevant to the long-term strategy of primary prevention, although these studies will be necessarily long term and costly. Although there might be ethical concerns associated with testing and studying of asymptomatic family members, a robust framework for this process has recently been established.⁸⁰

How does ALS spread?

Typically, ALS has a striking focal clinical presentation, although this view must be interpreted with the understanding that most biological systems have a degree of redundancy and also in view of emerging data covering presymptomatic changes. For example, at least a third of muscle fibres have already been lost before clinical wasting occurs in ALS.⁸¹ Nonetheless, post-mortem studies suggest that the pathological changes in ALS radiate from an area in which maximum upper motor neuron and lower motor neuron dysfunction aggregate together in the spinal region of first symptom onset.⁸² Spread of symptoms then appears to occur via an anatomically contiguous space within both upper motor neuron and lower motor neuron systems, although the rate might appear to differ within these two anatomical axes. Separately, a complete model of the ALS system failure requires incorporation of a third extramotor axis; namely, the substrate for cognitive and behavioural impairments.

Irrespective of the basis for the regional pattern of involvement in ALS, important and unexplained differences remain in sex and age, particularly in relation to the site of disease onset. For example, respiratory-onset disease tends to be more prevalent in men,⁸³ and a predilection for isolated corticobulbar disease has been noted in elderly women.⁸⁴ Bulbar-onset ALS is strikingly under-represented in young-onset patients,⁸⁵ and the biological reasons for these phenomena require dedicated investigation.⁸⁶ By contrast, regional onset could be purely stochastic⁸² or perhaps related to patterns of use or development.⁷⁴ The convergence of upper motor neuron and lower motor neuron features might then reflect the topography of muscle innervation. Separately, there are also physiological differences between short and long axons that must be considered, including axonal membrane conductances and the Na+/K+ pump, the dysfunction of which might be relevant to pathogenesis.⁸⁷

The development of advanced neuroimaging techniques has allowed novel assessment of structural and functional connections within the brain of patients with ALS that has further guided concepts about disease spread, individual susceptibility, and prognosis. Cortical correlates for regional symptoms can be shown with volumetric MRI.⁸⁸ Neuronal connections are organised in networks and appear to have their own characteristics. Properties include the overall connectivity of the network, clustering of network nodes, the efficiency of the network, and the degree to which there are identifiable hubs.⁸⁹ Recent investigations concerning functional connectivity in ALS have identified that patients with a strong connectedness of the motor network have a fast progression of disease, suggesting a potential role for assessment of the functional organisation of the motor network as a marker of disease progression.³⁹ and ⁹⁰ Positive correlation might also suggest patterns of disease spread along functional connections of the motor network in ALS, by contrast with the theory that ALS solely affects a fixed set of primary motor connections.⁴⁰ These more widespread effects on the brain network suggest that focal damage in primary motor regions in ALS might ultimately manifest in connectivity disturbances elsewhere in the brain, with patterns of change potentially serving as biomarkers for disease progression. At the synaptic level, parallels have also been drawn for a prion-like spread of abnormal protein conformation and propagation of cellular toxicity in ALS,⁹¹ and ⁹² perhaps involving exosomes or autophagy.⁹³

How can the course of ALS be predicted and the design of therapeutic clinical trials improved?

Prediction of the rate of progression in a heterogeneous disorder such as ALS has clear value in care-planning, but also in the assessment of potential therapeutic agents. Typical inclusion criteria for therapeutic trials have tended to be based on time from symptom onset (typically with a cutoff less than 3 years) and a minimum level of respiratory function as measured by vital capacity. Within only these criteria there will inevitably be a wide range of prognoses, thereby greatly reducing statistical power.

However, broad phenotypes with consistent prognostic patterns can be identified in a population-based context. These patterns include rapid decline in patients with respiratory-onset disease,⁸³ reduced survival in those with executive impairment,²³ and slower progression in those with upper motor neuron-predominant disease (consistent with the extreme example of primary lateral sclerosis).¹² and ⁹⁴ Consistent regional (generally contiguous) patterns of symptom spread are also recognisable,^{95, 96} and ⁹⁷ and simple clinical parameters at diagnosis can allow robust stratification of prognosis for clinic-based patients.⁹⁸ and ⁹⁹ In view of these developments, a formal staging system for ALS has been proposed that attempts to better incorporate clinical phenotype.¹⁰⁰ Stratification by rate of disease progression (time to spread to a new region) and degree of cognitive impairment, rather than the traditional divisions by site of onset or upper motor neuron- dominance versus lower motor neuron dominance, might yet prove more useful for future clinical trial enrolment.

At the individual patient level, prognostication soon after symptom onset remains difficult; for instance, it is difficult to distinguish typical bulbar-onset ALS (with progression to limb weakness) from a more atypical, isolated bulbar onset, which might remain isolated sometimes for several years before the involvement of other body regions (sometimes termed progressive bulbar palsy).¹⁰¹ Intrinsic heterogeneity among trial participants can obscure treatment effects that might be more obvious in specific clinical (or genetic) subgroups, and various strategies have been proposed to improve efficiency of clinical trials in ALS.¹⁰² Subgroup analysis of patients involved in previous clinical trials might have value; for example, some have suggested that riluzole could have an effect in delaying cognitive involvement.

Pending validation of the numerous and varied candidate biomarkers (summarised in panel 2),^{103, 104} and ¹⁰⁵ the revised ALS functional rating score (ALSFRS-R; panel 3) is a frequently used outcome measure in therapeutic trials.¹⁰⁶ This score has good intrarater and inter-rater reliability, and prognostic value as a single measurement at the point of diagnosis as well as an indicator of the rate of change.¹⁰⁷ However the slope of its decline might not be consistently linear, with deviations noted in early and late stages of the disease,¹⁰⁸ so that statistical assumptions based on a linear model in trials involving patients early in their disease might be incorrect. Patient dropouts due to disease progression further complicate the design of ALS clinical trials, in addition to inherent and progressive problems for patients to attend the protocol centres, leading to potentially critical loss of statistical power. A method to combine survival and ALSFRS-R into one primary outcome measure was developed as the primary endpoint for the failed phase 3 trial of dexpramipexole.¹⁰⁹ There is also increasing momentum to consider a further adjustment of the ALSFRS-R to include assessment of executive function, although further data are needed to validate available cognitive tools for ALS.

Motor unit number estimation, axonal excitability, electrical impedence myography, the neurophysiological index (NI), and cortical excitability all have potential as alternative outcome measures.¹⁰⁵ The neurophysiological index has been shown to have good intrarater reliability and sensitivity to change in ALS, including the differentiation of fast and slow progressing patients.¹¹⁰ Longitudinal analysis supports incorporation of the NI into future therapeutic trials.¹¹¹ More generally, that future trials must routinely incorporate candidate biomarkers if the historical reliance on survival as the primary outcome measure is to be overcome is widely agreed.

Has the SOD1 mouse model outlived its usefulness?

The transgenic mouse model overexpressing SOD1 contains multiple copies of mutant human DNA. The model has been used extensively to dissect the likely pathogenic mechanisms of mutant SOD1 and to evaluate the possible benefits of new therapeutic agents. There has been a disappointing lack of drugs that meaningfully affect disease progression in the model and also in the translation of any modest disease-delaying benefits from mouse to human ALS. The repeated failures of phase 2 and 3 trials in human beings based on SOD1 mouse studies have led to calls for a reappraisal of the usefulness of the model.

Although the SOD1 mouse recapitulates several core clinical features of ALS—namely, a predictable and progressive loss of motor neurons, associated with muscle weakness commencing at approximately 90 days, and premature death—the pathological cascade plays out over a short period of time. ⁸ Therein might lie an inherent problem in the attempt to translate the results of therapeutic challenges to the human disease, in which symptoms do not emerge for several decades and where the disease develops through a combination of factors that are beyond genetic mutations and not captured by any model to date. Separately, therapeutic agents that have been tested and judged beneficial in the mouse model have been administered well before symptom onset.

A further failure of translation might partly relate to errors in early mouse studies, which were often inadequately powered and confounded by factors including differential effects of sex, failure to control for variations in copy number, differences in animal husbandry, and variability in mouse strain genetic backgrounds, leading to potentially erroneous conclusions. However, to address these issues, two recent workshops have generated guidelines that could largely resolve the confounders of early mouse studies, thereby rendering comparative analysis more robust.^{112, 113, 114} and ¹¹⁵

Proponents of the mouse model point out that translation from mouse to human being will always be imperfect because of cross-species differences in neuroanatomy, pharmacokinetics, and dynamics, as well as longevity. They also point to the reproducible murine phenotype and the similarities in neuropathology, including the early and selective degeneration of anterior horn cells. Conversely, some issues need to be further considered including the requirement for large copy numbers of mutant DNA to generate a phenotype, the absence of phenotype in the mouse with highly penetrant human mutations that lead to a severe phenotype in human beings (eg, the A4V mutation), and the observation that SOD1-related ALS is not associated with TDP-43 inclusions—the pathological signature common to nearly all other forms of human ALS, including those associated with hexanucleotide repeat expansions in C9orf72. Although recent genetic advances have led to the development of new animal models based on transgenic TDP-43 and FUS mutants, no other model identified to date has recapitulated the clinical disease phenotype to the extent of the SOD1 mouse, despite the aforementioned caveats.

Novel approaches to understanding of ALS pathogenesis

The cost of whole-genome sequencing is falling exponentially to a level that will soon make this approach feasible for a large number of patient samples. An internationally collaborative approach to sequence whole genomes from large cohorts of well characterised patients with ALS seems the next logical step. As a result, the characterisation of the genetic architecture of ALS might soon be complete, but is likely to be only the first step towards identification of key pathways amenable to therapeutic intervention. A limitation of the downstream transcriptomic approach is the difficulty in distinguishing primary (initiating) and secondary (responsive) changes in gene expression from the mass of data generated, although this approach holds promise.¹¹⁶ High-throughput proteome-wide studies are still beyond the horizon.

The bioinformatics task presented by the enormous volume and complexity of both genomic and RNA expression data has become the immediate challenge. It requires the neuroscience research community to engage those with the mathematical expertise needed to meaningfully interrogate and interpret such datasets, in conjunction with more sophisticated clinical phenotyping of patients.

Although animal modelling for ALS remains challenging, and rodent models have limitations, new models that engage technological developments hold promise for rapid assessment of therapeutic potential. Such models include the Caenorhabditis elegans worms, established for high-throughput, large-scale functional validation of candidate protective genes, including lines expressing forms of TDP-43 or FUS. ^{117, 118} and ¹¹⁹ Selection of nematode promoters can now also be used to express these proteins in neurons or more specifically only in motor neurons. In addition to the development of phenotypic assays, such studies could also unlock a holy grail in ALS research—the mechanism of disease spread at a molecular level.

The development of induced pluripotent stem cell technology offers an additional and potentially cost-effective approach as a platform for early modelling pathology and testing therapeutic candidates (so-called disease-in-a-dish). It marks a significant divergence from the traditional view of stem cells as a potential therapy for ALS.¹²⁰ One of the major challenges will be to overcome the great heterogeneity across cell colonies, and to model the in-vivo network environment of such cells, which might be as important as any individual cellular dysfunction observed in isolation.¹²¹ Emerging themes might be common to other neurodegenerative disorders, in particular protein degradation pathways,⁶ and ¹²² making overlap in treatment strategy between ALS and other disorders increasingly likely.¹²³

Search strategy and selection criteria

We searched PubMed (1966, to Dec 31, 2012), Embase (1980, to Dec 31, 2012), and the Cochrane Library using the search terms “amyotrophic lateral sclerosis” or “motor neuron disease” in combination with “diagnosis”, “epidemiology”, “fronto-temporal dementia”, “imaging”, “neurophysiology”, “management”, and “neuroprotection”. Further articles were included from reference lists, review articles, and major textbook chapters. Abstracts and reports from relevant meetings were also included. The final reference list was generated on the basis of originality and relevance to the topics covered in this report. Emphasis was placed on publications from the past 5 years, but did not exclude commonly referenced and highly regarded older publications.