State of play in amyotrophic lateral sclerosis genetics

Considerable progress has been made in unraveling the genetic etiology of amyotrophic lateral sclerosis (ALS), the most common form of adult-onset motor neuron disease and the third most common neurodegenerative disease overall. Here we review genes implicated in the pathogenesis of motor neuron degeneration and how this new information is changing the way we think about this fatal disorder. Specifically, we summarize current literature of the major genes underlying ALS, SOD1, TARDBP, FUS, OPTN, VCP, UBQLN2, C9ORF72 and PFN1, and evaluate the information being gleaned from genome-wide association studies. We also outline emerging themes in ALS research, such as next-generation sequencing approaches to identify de novo mutations, the genetic convergence of familial and sporadic ALS, the proposed oligogenic basis for the disease, and how each new genetic discovery is broadening the phenotype associated with the clinical entity we know as ALS.

ALS is an adult-onset neurodegenerative disorder characterized by rapidly progressive paralysis and death from respiratory failure, typically within 2 to 3 years of symptom onset¹. There are no effective cures for ALS, though the benzothiazole riluzole slows the rate of progression and prolongs survival by three months¹. As in other neurodegenerative diseases, ~10% of ALS is classified as familial, whereas the remaining 90% of cases are considered sporadic, as they appear to occur randomly throughout the community¹.

Unraveling the genetic etiology of ALS has provided fundamental insights into the cellular mechanisms underlying neuron degeneration, as well as facilitating disease modeling and the design and testing of targeted therapeutics; thus, it is not surprising that considerable resources have been devoted to finding pathogenic mutations. SOD1 mutations were found to cause familial ALS in 1993 (ref. ²), but there was a long hiatus until the discovery of the next ALS gene, TARDBP, in 2008 (ref. ³). Today, the pace of gene discovery has greatly accelerated, fuelled in large part by advances in sequencing and genotyping technology. In the last 3 years alone, six new ALS genes have been discovered4–10, and we now know the genetic etiology of two-thirds of familial cases and about 11% of sporadic ALS cases.

This Review will focus on recent genetic discoveries in ALS and how genetics is changing our understanding of this fatal, mysterious disorder. In the first section, we examine the main genes implicated in the pathogenesis of ALS. We will then discuss attempts to find genetic variants underlying sporadic ALS using genome-wide association studies (GWAS). Finally, we will describe emerging concepts in ALS genetics and where we ultimately think the road ahead leads.

Familial ALS genes

To provide a historical context, we present the main ALS genes in the order in which they were discovered (Fig. 1 and Table 1).

Figure 1.

Timeline of gene discoveries in familial and sporadic ALS. Values represent the proportion of ALS explained by each gene in populations of European ancestry. References are provided in the main text.

Table 1. Genes causing ALS.

Values represent the percentage of ALS explained by each gene in Caucasian populations. Relevant references are provided in the main text.

Gene	Location	Inheritance	Percentage explained		Putative protein function
			Familial ALS	Sporadic ALS
TARDBP	1p36	AD	4	1	RNA metabolism
SQSTM1	5q35	AD	1	<1	Ubiquitination; autophagy
C9ORF72	9p21	AD	40	7	DENN protein
VCP	9p13	AD	1	1	Proteasome; vesicle trafficking
OPTN	10p13	AR & AD	<1	<1	Vesicle trafficking
FUS	16p11	AD & AR	4	1	RNA metabolism
PFN1	17p13	AD	<1	<1	Cytoskeletal dynamics
SOD1	21q22	AD & AR	12	1–2	Superoxide metabolism
UBQLN2	Xp11	XD	<1	<1	Proteasome

AD = autosomal dominant; AR = autosomal recessive; XD = X-linked dominant; ALS = amyotrophic lateral sclerosis, and DENN = Differentially Expressed in Normal and Neoplasia.

Superoxide dismutase 1 (SOD1)

The identification of dominant missense mutations in SOD1 20 years ago represented the first demonstration that linkage analysis could successfully pinpoint the underlying genetic cause of a rare neurodegenerative disease². Although over a hundred SOD1 mutations have been reported, reliable genetic evidence of pathogenicity exists for only a portion of these¹¹. Nevertheless, population-based studies show that mutations in this gene account for ~12% of familial cases and ~1% of sporadic cases¹².

Considerable phenotypic heterogeneity occurs across the various SOD1 mutations. For example, the A4V mutation, which is the most frequent variant in North America, gives rise to an aggressive form of ALS that typically leads to death within a year after symptom onset¹³. In contrast, the homozygous D90A mutation in the same gene is associated with an indolent course, with patients developing respiratory failure only after 10 years of illness¹⁴. Cognitive impairment is not a prominent feature of SOD1 disease, though patients with D90A manifest cognitive issues at the later stages of the disease, perhaps reflecting their protracted survival and the corresponding longer time to manifest disease spreading to non-motor prefrontal areas¹⁵.

The discovery that SOD1 mutations cause ALS led directly to the development of the SOD1 transgenic mouse. Though important in elucidating the cellular mechanisms by which disruption of this gene predispose to motor neuron degeneration¹⁶, the use of this model to select agents for human trials has been increasingly called into question¹⁷. Indeed, the pathology of human SOD1 ALS is now thought to be distinct from that of all other types of ALS, in that it lacks the TDP-43 and/or FUS pathology present in nearly every other instance¹⁸. In recognition of this fact, there have been recent efforts to selectively treat patients carrying mutated SOD1 using antisense oligonucleotide therapy designed to knock down expression of the gene¹⁹.

TAR DNA-binding protein (TARDBP)

A landmark event in our understanding of ALS pathogenesis was the discovery that the TDP43 protein is a major component of the ubiquitin-positive neuronal inclusions that are the pathological hallmark of both ALS and frontotemporal dementia (FTD)²⁰. This seminal observation provided vital evidence that these distinct conditions constitute a disease spectrum, rather than being discrete clinical entities. It also directly led to the discovery of mutations in TARDBP (which encodes TDP-43) in autosomal dominant ALS and FTD families^3,21. These pathogenic variants were mostly located in the C terminus, which is involved in ribonucleoprotein binding and splicing.

Mutations in TARDBP account for ~4% of familial ALS cases and a smaller percentage of sporadic ALS cases²². Although TARDBP mutations are found in ALS families across the globe, some regional variability does exist. For example, the A382T mutation of the protein is particularly frequent in Sardinia, reflecting the conserved nature of that island population combined with a historical founder effect²³. Nevertheless, the overall mutational frequency of the gene remains much lower than the occurrence of TDP-43 neuropathological changes observed in autopsies.

The discovery of the central role of TDP-43 in ALS pathogenesis has highlighted the importance of RNA processing. Additional support for this hypothesis comes from the discovery of FUS, another RNA-binding protein (see below). Considerable efforts have been made to understand exactly how mutations in these genes disrupt RNA transcription and to identify RNA species modified by both proteins²⁴.

Fused in sarcoma (FUS)

Shortly after the cloning of TARDBP, missense mutations of FUS were identified as the cause of chromosome 16p–linked familial ALS^25,26. Although this gene accounts for a small percentage of familial cases (~4%), its discovery caused considerable excitement in the field because the protein shares functional homology with TDP-43. Furthermore, FUS mutations cluster in the RNA-binding domain at the C terminus of the protein, as seen with TARDBP. These observations reinforced the importance of abnormal RNA metabolism in motor neuron degeneration.

Neuropathological analyses of patients carrying FUS mutations have yielded important insights into the disease. Though some debate remains, it would appear that FUS cases are characterized by FUS-immunoreactive cytoplasmic inclusions with a striking absence of the ubiquitin-positive and TDP-43-positive aggregates that distinguish most ALS cases^25,26. One possible explanation for this finding is that FUS acts downstream of TDP-43 in the same pathway²⁷.

Optineurin (OPTN)

OPTN mutations were initially described as a cause of autosomal recessive ALS in Japanese families in 2010 (ref. ⁹). Since then, it has become clear that heterozygous mutations in this gene rarely cause familial ALS among people of European ancestry²². This may reflect different genetic etiologies underlying motor neuron degeneration across major ethnic groups. Indeed, autosomal recessive ALS may be more common in Japan compared to Europe and North America.

Nevertheless, the phenotypic pleiotropy associated with OPTN is intriguing. Mutations in this gene have long been described as a cause of primary open angle glaucoma²⁸, and more recently the OPTN locus was implicated in a GWAS of Paget’s disease of bone²⁹. Along with VCP and SQSTM1 (see below), this provides information supporting a clinical overlap between Paget’s disease and ALS. Although OPTN regulates diverse cellular processes, including membrane trafficking, protein secretion, cell division and host defense against pathogens³⁰, it remains to be determined exactly how mutations in this gene may give rise to such a wide range of clinical phenotypes.

Valosin-containing protein (VCP)

Also in 2010, we reported that mutations in VCP are responsible for 1–2% of familial ALS cases⁵, a finding that was subsequently confirmed by other studies^31,32. This was the first time that whole-exome sequencing had been successfully applied to a major neurodegenerative disorder. The power of this new technology is based on its use of the massively parallel sequencing capabilities of next-generation sequencing platforms to rapidly identify rare variants in the ~2% of the genome that encodes proteins. Moreover, whole-exome sequencing is a realistic strategy for detecting pathogenic variants in small families where linkage analysis would not be possible due to a shortage of DNA samples from affected individuals. This is particularly true in ALS, where the rapidly fatal nature of the syndrome makes it challenging to collect families of sufficient size to perform linkage analysis.

Mutations in VCP were already known to underlie an unusual clinical syndrome characterized by FTD, inclusion body myopathy and Paget’s disease of the bone (IBMPFD)³³. Thus, the discovery of mutations in the same gene resulting in an ALS phenotype represented another step in our unraveling of the genetic links between motor neuron degeneration and FTD. The coexistence of inclusion body myopathy in these patients was similarly interesting because it demonstrated that mutations in a single gene could result in pathology on both sides of the neuromuscular junction. This has given rise to the concept of multisystem proteinopathy, in which multiple tissues are affected³⁴. Recent work has convincingly demonstrated that mutations in VCP cause mitochondrial uncoupling leading to a significant reduction of cellular ATP production, perhaps explaining disparate effects across multiple tissues³⁵.

Ubiquilin 2 (UBQLN2)

Missense mutations in UBQLN2, located on the short arm of chromosome X, were initially identified in apparently autosomal dominant ALS pedigrees lacking male-to-male transmission⁴. Although isolated ALS was the predominant phenotype associated with mutations in this gene, occasional patients had concomitant symptoms of FTD. It is already evident that UBQLN2 mutations are not a common cause of familial ALS³⁶. Nevertheless, ubiquilin 2 pathology has been observed in ALS patients who do not carry mutations in the gene, suggesting that this protein, which regulates proteasome degradation of ubiquitinated proteins, may be an important component of the final common pathway mediating motor neuron degeneration⁴.

Hexanucleotide repeat expansion in C9ORF72

We and others recently reported that a massive hexanucleotide repeat expansion in C9ORF72 is the cause of chromosome 9p21–linked ALS and FTD^7,8. Long sought after, cloning of this locus has reinvigorated the ALS and FTD research field for a variety of reasons. First, the pathogenic expansion accounts for a remarkable percentage of both familial ALS (~40%) and familial FTD (~25%) and genetically explains the majority of the overlap of these two disparate clinical syndromes³⁷. Second, the repeat expansion has been found to account ~7% of apparently sporadic ALS cases in people of European ancestry, marking the first time that a genetic etiology has been identified for more than just the occasional sporadic case. Third, this is the first time that a large intronic repeat expansion has been implicated in ALS. Given that repeat expansions are known to disrupt RNA metabolism in other neurodegenerative diseases, it again points toward the importance of this pathway in the maintenance of motor and frontal cortex neurons. Finally, there is a real chance that gene therapy based on antisense oligonucleotides will be effective in slowing progression in ALS patients who carry the expansion (approximately one in ten ALS cases in European-ancestry populations). There are already concerted efforts to bring such a therapy forward into human trials.

Several important questions concerning C9ORF72 remain unanswered. For example, what is the full range of phenotypes associated with this mutation? The expansion may underlie a small portion of clinical Alzheimer’s disease and Parkinson’s disease cases^38,39, but what is its role in neuropsychiatric disorders? A corollary question is why some patients present with motor dysfunction and others with frontal lobe impairment. Although there are anecdotal reports of anticipation (the onset of symptoms at an earlier age in subsequent generations) in C9ORF72 families⁴⁰, is this a robust finding? If so, is the repeat expansion itself unstable, and what role does such instability play in determining disease phenotype and severity? Early data point to considerable instability of the expansion both across families and within individuals (somatic mutation)⁴¹. Is the GGGGCC repeat pure or interrupted, and how might that influence phenotype? What genomic and cellular substrates underlie the variable penetrance observed among individuals carrying the expansion?³⁷ Methylation has been shown to be an important determinant of age of symptom onset in other repeat-expansion diseases, and there are data to suggest that this may be relevant in C9ORF72 (ref. ⁴²).

Finally, and perhaps most pressing for the development of gene therapies targeting the locus, how does the repeat expansion give rise to disease? RNA foci are observed in fibroblasts derived from C9ORF72 patients, implicating disruption of RNA metabolism through sequestration of RNA-binding proteins and other RNA species⁸. However, a more direct role for C9ORF72 protein in the nucleus has not been excluded, suggesting that haploinsufficiency may also be relevant. C9ORF72 hexanucleotide repeats form highly stable RNA G-quadruplexes, which may influence telomere stability and RNA transcription, splicing, translation and transport⁴³. More recently, repeat-associated non-ATG (RAN) protein translation has been observed⁴⁴. Although it is not known whether the inclusions generated from such aberrant translation are involved in neuronal cell death, this process has been directly implicated in cellular toxicity associated with fragile X tremor ataxia, a different noncoding repeat expansion disease⁴⁵.

Sequestosome 1 (SQSTM1)

SQSTM1 encodes p62, a major pathologic protein deposited in neurodegeneration. P62 regulates ubiquitin binding and activation of nuclear factor kappa-B signaling. Mutations in this gene are known to cause Paget’s disease of bone⁴⁶. A candidate gene screening approach identified missense and deletion variants in ~1% of ALS cases¹⁰. Follow-up studies have shown a similar mutational frequency rate, though, as of yet, segregation of variants with disease within a large family has not been reported^47,48.

Profilin 1 (PFN1)

In late 2012 Wu et al. reported that missense mutations in PFN1 segregated with disease in several large ALS kindreds⁶. Although this gene has only been recently described as pathogenic mutational screening of sizeable ALS and FTD cohorts from several populations already demonstrate that PFN1 mutations are not a prominent cause of neurodegeneration⁴⁹. Nevertheless, this discovery implicates a new cellular mechanism in the pathogenesis of ALS, namely disruption of the cytoskeletal architecture of the neuron.

Other genes

Mutations in several other genes have been reported as rare causes of ALS or ALS-like syndromes (Table 2). In the case of ALS2, the associated phenotype is more consistent with slowly progressive, juvenile-onset, hereditary spastic paraplegia, rather than adult-onset ALS^50,51. Similarly, mutations in SETX are more typically associated with an ataxia phenotype, rather than ALS^52,53.

Table 2. Other genes implicated in the pathogenesis of ALS.

Gene	Location	Inheritance	Predominant clinical syndromes	Putative protein function
DCTN1	2p13	AD	PMA; Perry syndrome	Axonal transport
ALS2	2q33	AR	Juvenile PLS; infantile HSP	Vesicle trafficking
CHMP2B	3p11	AD	Familial ALS; sporadic ALS; FTD	Vesicle trafficking
FIG4	6q21	AD & AR	CMT; familial ALS	Vesicle trafficking
HNRNPA2B1	7p15	AD	Multisystem proteinopathy; ALS	Dysregulated protein polymerization
ELP3	8p21	Undefined	Sporadic ALS	RNA metabolism
SETX	9q34	AD	Juvenile ALS; ataxia with oculomotor apraxia	RNA metabolism
HNRNPA1	12q13	AD	Multisystem proteinopathy; ALS	Dysregulated protein polymerization
ATXN2	12q24	Undefined	Sporadic ALS; ataxia	Endocytosis; RNA translation
ANG	14q11	AD	Familial ALS; sporadic ALS	Angiogenesis
SPG11	15q14	AR	Juvenile ALS; HSP	DNA damage repair
VAPB	20q13	AD	PMA; FALS	Vesicle trafficking
NEFH	22q12	AD	Familial ALS; sporadic ALS	Axonal transport

AD = autosomal dominant; AR = autosomal recessive; ALS = amyotrophic lateral sclerosis; CMT = Charcot-Marie-Tooth disease; HSP = hereditary spastic paraplegia; PLS = primary lateral sclerosis; PMA = progressive muscular atrophy.

In other cases, the genetic evidence supporting a role as an ALS gene is not fully convincing. For example, there have been no reports of FIG4 or DAO mutations as a cause of ALS since their initial publications^54,55. By the same token, only a single pedigree showing convincing segregation of ALS with a DCTN1 mutation has been described, and mutations in this gene are now more commonly linked to Parkinson’s disease with hypoventilation and depression (Perry syndrome)^56,57. Apart from the Brazilian ALS families in which the original VAPB mutations were reported, there have been no additional publications of segregating mutations in this gene^58,59. Though there is good evidence for a private mutation in CHMP2B in a large Danish family with FTD, there is only minimal data to support its involvement in ALS pathogenesis^60–62. The role of the candidate gene ANG, nominated on the basis of its angiogenic properties, remains ambiguous for both familial and sporadic ALS^22,63–65.

Genome-wide association studies of ALS

Susceptibility loci. There have been 14 GWAS published in ALS, and they have already made significant contributions to our understanding of ALS genetics^66–79. Our own GWAS of ALS in Finland was the first to identify a significant association peak on the short arm of chromosome 9 in this founder population⁷⁵. This remains the most consistent signal yet observed^73,76,78 and was instrumental in the cloning of the C9ORF72 locus, as the dense nature of the single nucleotide polymorphisms (SNPs) on the genotyping platform narrowed the scope of the search.

A variety of other risk loci have been nominated on the basis of ALS GWAS. Many of them, such as FGGY, ITPR2 and DPP6 have not stood up to replication in large cohorts^71,77,80. Others, such as UNC13A, appear more robust, although true replication of this locus in an independent cohort is still pending^78,81. More recently, a GWAS of ALS in people of non-European ancestry identified new loci at 1q32 and 22p11 as potentially relevant to ALS pathogenesis, though the discovery cohort was small (n = 506) for an outbred population such as the Han Chinese⁷⁹. The field eagerly anticipates independent replication of these findings in larger case-control cohorts.

ALS genetic research has been largely focused on populations of European ancestry, and attention has only recently shifted to other ethnicities. Early indications are that the genetic architecture of ALS is distinct in other regions of the world. For example, the frequency of the C9ORF72 repeat expansion is far lower among Japanese than among Europeans³⁷, whereas the inverse pattern is observed for OPTN mutations. Future genetic discoveries in non-European populations will likely unravel the diverse means by which motor neuron degeneration can occur.

Age at symptom onset and survival phenotypes

In addition to searching for susceptibility loci, there have been genome-wide efforts to identify genetic variants influencing ALS phenotype. For example, genetic variation in the KIFAP3 and EPHA4 loci has been reported to significantly influence survival among ALS patients^72,82, and a large meta-analysis of ALS GWAS recently reported that age of onset is modulated by a locus on the short arm of chromosome 1 (ref. ⁷⁸). The motivation for these efforts is based on the notion that such loci represent final common pathways involved in the death of motor neurons and that the cellular mechanisms that drive neurodegeneration forward may be distinct from the initiating events. Final common pathways may also be more attractive targets for therapeutic intervention, as such agents are more likely to be effective across the gamut of ALS cases. Despite this, caution against over-interpretation of such secondary phenotypes is required. Indeed, attempts to replicate the effect of the KIFAP3 locus on survival have met with varied success⁷⁴.

Lessons learned from ALS GWAS

The current state of play in ALS illustrates a recurring theme with GWAS methodology, namely that the sheer number of association tests performed on the same data set often leads to false-positive findings⁸³. Although a variety of statistical tools can be employed to counter this, independent replication remains the gold standard. Such replication studies are greatly facilitated by the public availability of raw genotype data generated in earlier GWAS. The release of such data allows other researchers around the world to access, reanalyze and combine it with their own GWAS, thereby increasing the power of their studies at no added cost.

By their very nature, GWAS of outbred populations require several thousand case and control samples to have sufficient power to identify risk loci. The Coriell ALS DNA Repository (http://www.coriell.org/), funded by the ALS Association, the Muscular Dystrophy Association and the National Institute of Neurological Disorders and Stroke, distributes well-phenotyped biological samples from ~2,000 case and ~6,000 population control subjects⁸⁴. The availability of this resource has invigorated the field by lowering the barriers for laboratories to engaging in genetic research.

The discovery of more and more causative genes confirms that ALS is not a monolithic clinical entity but rather consists of a group of diseases unified by the common theme of progressive motor neuron degeneration. Such phenotypic and genetic heterogeneity confounds the ability of GWAS to identify associated regions. A standard approach to overcoming this obstacle has been to increase the size of case-control cohorts genotyped as part of GWAS. This approach has been successful in other neurodegenerative diseases, such as Parkinson’s disease and Alzheimer’s disease^85,86, and lends credence to the view that similarly sized studies are likely to reveal more loci important to the pathogenesis of ALS.

A complementary technique has been to study conserved populations that display elevated rates of ALS. The reduced genetic heterogeneity found in such communities dramatically increases power to identify new genes responsible for disease. In Finland, our GWAS identified association signals on chromosome 9 (corresponding to the pathogenic C9ORF72 repeat expansion) and on chromosome 21 (corresponding to the D90A allele of SOD1)⁷⁵. These loci account for nearly all familial ALS cases in Finland and explain the high disease incidence observed in that population isolate. Our studies of Sardinian ALS patients have revealed two causative genes operating in that island population, namely the A382T-encoding allele of TARDBP and the C9ORF72 repeat expansion^23,37. Similar efforts to exploit the Irish population have been less successful, but this may be due to the small size of the cohort or because this country does not have a higher rate of ALS than other European populations^69,87.

Future directions for ALS GWAS

Older versions of genome-wide arrays selected common SNPs (minor allele frequency >5% in the general population) to capture the maximum variation across the genome. This allowed the common disease/common variant hypothesis to be tested for a particular disease. In contrast, newer genotyping platforms (for example, the HumanExomeBeadchip from Illumina) focus on rare variants (minor allele frequency <5%) located in the coding portion of the genome. GWAS performed with these chips will test the hypothesis that a disease is caused by multiple rare variants. Given the relative rarity of ALS in the general population (incidence ~2 per 100,000 in Europe), these new types of GWAS hold great promise.

Emerging themes in ALS genetics

De novo mutations

Spontaneously occurring mutations are a well known cause of neurological and non-neurological conditions, such as neurofibromatosis type 1 (ref. ⁸⁸) and Hirschsprung’s disease⁸⁹. Indeed, de novo mutations of FUS and SOD1 have been described in sporadic ALS cases^90–92. Exome sequencing of parent–case offspring trios offers a new method to systematically search the genome for such variants. A recent publication from Gitler and colleagues highlights the potential power of this method and led to the identification and functional characterization of mutations in the chromatin regulator SS18L1 (also known as CREST) as possibly underlying ALS⁹³. This publication represents an excellent first step, though it should be recognized that not all de novo variants are necessarily pathogenic. Validation of these ‘hits’ in larger cohorts of familial and sporadic ALS cases is necessary to ensure that this new technique really flourishes.

Breakdown in classification of familial and sporadic ALS

The discovery that C9ORF72 is responsible for a sizeable proportion of sporadic ALS cases in populations of European ancestry has led to a fundamental shift in our understanding of the disease. Although categorization of ALS cases as familial or sporadic on the basis of family history retains some utility, it is clear that the underlying biology is more nuanced and that such clinical classification should not be viewed in absolute terms⁹⁴. This has profound implications for the clinical care and genetic testing in ALS⁹⁵. It also lends credence to the notion that a genetic component underlies all of ALS and suggests that environmental and stochastic factors act as precipitating factors in genetically susceptible individuals, rather than being directly causative in their own right.

Oligogenic basis for ALS

An oligogenic basis has recently been proposed in ALS families exhibiting Mendelian inheritance⁹⁶. Under this model, mutations in two or more genes are required for an individual to develop disease. In some ways, this is reminiscent of the Knudson two-hit hypothesis implicated in carcinogenesis. While such events can occur in isolated populations, these are likely to be chance events driven by the high frequency of founder mutations in those regions⁹⁷. A more parsimonious explanation for finding two ‘mutations’ in an ALS patient from an outbred population is that one of these variants is not truly pathogenic. In that regard, it is telling that many of the patients reported to carry two mutations in the Dutch population consisted of a mutation in an ALS gene and what may be a benign polymorphism in ANG⁹⁶.

Nevertheless, one cannot a priori discount the possibility that ALS patients carry a primary genetic variant that drives susceptibility to disease and a secondary locus that influences age of symptom onset, symptom presentation and/or disease progression. Genome-wide data from several thousand patients across multiple countries will be required to resolve whether such genetic oligogenicity is relevant to ALS. The public availability of such large-scale data sets will allow future researchers to test for epistatic interactions among genes.

Extension of ALS phenotype and overlap between ALS and FTD

The more we learn about the genetics of ALS, the broader the associated phenotype becomes. For example, recognizing that ALS and FTD represent overlapping clinical syndromes has been a major step forward in our understanding, and this convergence has been strengthened by the discovery of C9ORF72 and other genes. In addition, VCP and SQSTM1 show that the condition, at least in some cases, is not limited to the CNS and that diverse tissues such as muscle and bone may be affected. The discovery of mutations in HNRNPA2B1 and HNRNPA1 in families with ALS and in families with muscle-, bone and neurodegeneration reinforces the notion that at least some forms of ALS are part of a more widespread disease process³⁴. The term “multisystem proteinopathy” has been suggested to describe these diverse findings³⁴.

Genes involved in other forms of motor neuron disease, such as hereditary spastic paraplegia and Charcot-Marie-Tooth disease, and even forms of neurodegeneration that are now regarded as quite distinct, may ultimately be linked to the ALS phenotype rubric. Expansions of the polyglutamine repeat in the coding region of ATXN2 cause a form of spinocerebellar ataxia (SCA2), and there appears to be clinical overlap between ALS and SCA2, with intermediate-length expansions reported to increase the risk of ALS^98,99. Intriguingly, ANG has been implicated in the pathogenesis of Parkinson’s disease¹⁰⁰, although it is unclear why this locus was not detected in a previous GWAS of Parkinson’s disease involving 12,000 cases and 20,000 controls⁸⁶.

Conclusions and future directions

We have come a long way since the discovery of the first ALS gene 20 years ago, and there is no doubt that genomics will continue to drive the research field forward. Analysis of increasingly large genetic data sets will improve our understanding of neurodegeneration. Although exome sequencing is in fashion, whole-genome sequencing will rise to prominence as costs continue to drop. Genome sequencing will yield even greater insight into the genetic architecture of ALS by providing a complete catalog of rare variants present in cases and allowing us to explore the role of noncoding and intergenic genetic variation in the pathogenesis of motor neuron degeneration. Combined with the promise of effective gene therapies, it is plausible that the next 20 years will see the mysteries of ALS simply melt away.