Amyotrophic lateral sclerosis: translating genetic discoveries into therapies

Abstract

Recent advances in sequencing technologies and collaborative efforts have led to substantial progress in identifying the genetic causes of amyotrophic lateral sclerosis (ALS). This momentum has, in turn, fostered the development of putative molecular therapies. In this Review, we outline the current genetic knowledge, emphasizing recent discoveries and emerging concepts such as the implication of distinct types of mutation, variability in mutated genes in diverse genetic ancestries and gene–environment interactions. We also propose a high-level model to synthesize the interdependent effects of genetics, environmental and lifestyle factors, and ageing into a unified theory of ALS. Furthermore, we summarize the current status of therapies developed on the basis of genetic knowledge established for ALS over the past 30 years, and we discuss how developing treatments for ALS will advance our understanding of targeting other neurological diseases.

Introduction

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease that primarily affects motor neurons, leading to progressive muscle weakness and death from respiratory failure within 2–5 years of symptom onset¹. The word ‘amyotrophic’ is derived from Greek, referring to muscle loss due to lack of nourishment, whereas ‘lateral sclerosis’ refers to scarring or hardening of the lateral column of the spinal cord. The term ‘amyotrophic lateral sclerosis’ was introduced by Jean-Martin Charcot, who described the clinical syndrome 150 years ago². Consequently, ALS is also referred to as ‘Charcot’s disease’ in France, whereas it is known as ‘motor neuron disease’ in the UK. In China, it is called ‘jian dong ren’, meaning ‘gradually frozen men’. In the USA, the disease is recognized as Lou Gehrig’s disease after a famous baseball player who died of ALS in 1941. ALS and frontotemporal dementia (FTD) are part of a spectrum, and cognitive dysfunction is a part of the ALS disease course that was previously under-recognized.

ALS is relatively rare, with a worldwide incidence of ~2 per 100,000 person-years^3,4. Despite this rarity, more than half a million individuals have died of ALS in the 80 years since Lou Gehrig gave his famous ‘luckiest man’ speech in Yankee stadium⁵. The number of affected individuals is soaring, mainly because of population ageing, and ~400,000 people will live with ALS globally by 2040 (ref. 6). The high fatality rate and the severe physical burden associated with ALS have further enhanced worldwide awareness of the disease, especially after the public attention attracted by the Ice Bucket Challenge^7–10.

Approximately 10% of patients with ALS have a family history of the disease, and more than 30 ALS-associated genes have been identified in these familial ALS cases so far^1,11. Although the remaining 90% of cases are apparently sporadic, pathogenic mutations identified in familial ALS were also described in many sporadic cases¹². Familial aggregation studies, twin studies and heritability studies based on genome-wide association studies (GWAS) confirmed a substantial genetic component (61% by twin studies and 21% by GWAS) in ALS^13–15.

Pathogenic variants in the most common ALS-associated genes, superoxide dismutase 1 (SOD1), TAR DNA-binding protein (TARDBP), fused in sarcoma (FUS) and chromosome 9 open reading frame 72 (C9orf72), account for approximately 60% of familial cases and about 10% of sporadic ALS. This equates to one in six ALS cases explained by these four genes¹² (Table 1). More recently, whole-exome sequencing and whole-genome sequencing have identified several other ALS-causing genes to the point that we now know the underlying genetic mutation in about three-quarters of familial and one-fifth of sporadic ALS cases in populations of European ancestries¹⁶ (Table 1). In this Review, we summarize what is known about the genetic basis of ALS and describe how new genomic methodologies and collaborative efforts aim to unravel the remaining unsolved portion of the disease. We also discuss the genetic overlap between ALS and other conditions, ALS studies in populations of diverse genetic ancestry, gene–environment interactions and how genomic discoveries are the starting point for targeted drug development and personalized medicine¹⁷.

Table 1 |.

Summary of ALS-associated genes

Year	Locus	Gene	Inheritance	Familial (%)a	Sporadic (%)a	Disease-associated mechanism	Other associated phenotypesb	Refs.
1993	21q22.11	SOD1	Autosomal dominant, autosomal recessive, de novo	12	1–2	Oxidative stress, excitotoxicity, mitochondrial dysfunction, axonal transport disruption	Frontotemporal dementia, spastic tetraplegia and axial hypotonia	19
1994	22q12.2	NEFH	Autosomal dominant	Unknown	Unknown	Axonal transport disruption	Axonal Charcot-Marie-Tooth disease type 2CC	212
2001	2q33.1	ALS2	Autosomal recessive	Unknown	Unknown	Vesicular trafficking defects	Juvenile primary lateral sclerosis, infantile hereditary spastic paraplegia	213
2003	2p13.1	DCTN1	Autosomal dominant	Unknown	Unknown	Axonal transport disruption	Distal hereditary motor neuropathy type VIIB, Perry syndrome	214
2004	20q13.32	VAPB	Autosomal dominant	Unknown	Unknown	Proteostasis defects	Finkel-type spinal muscular atrophy	215
2004	9q34.13	SETX	Autosomal dominant	Unknown	Unknown	Altered ribostasis	Autosomal recessive spinocerebellar ataxia type 1	216
2006	3p11.2	CHMP2B	Autosomal dominant	Unknown	Unknown	Proteostasis defects, vesicular trafficking defects	Frontotemporal dementia	217,218
2008	1p36.22	TARDBP	Autosomal dominant, autosomal recessive, de novo	4	1	Altered ribostasis, nucleocytoplasmic transport defects	Frontotemporal dementia	38,39
2009	16p11.2	FUS	Autosomal dominant, autosomal recessive, de novo	4	1	Altered ribostasis, nucleocytoplasmic transport defects	Frontotemporal dementia, essential tremor	46,47
2010	9p13.3	VCP	Autosomal dominant, de novo	1	1	Proteostasis defects	Frontotemporal dementia, Charcot-Marie-Tooth disease type 2Y, inclusion body myopathy with early-onset Paget disease	144
2010	15q21.1	SPG11	Autosomal recessive	Unknown	Unknown	DNA damage	Hereditary spastic paraplegia, Charcot-Marie-Tooth disease type 2X	219
2010	10p13	OPTN	Autosomal dominant, autosomal recessive	<1	<1	Autophagy, inflammation	Adult-onset primary open-angle glaucoma	220
2011	Xp11.21	UBQLN2	X-linked dominant	<1	<1	Proteostasis defects	None	221
2011	5q35.3	SQSTM1	Autosomal dominant	1	<1	Autophagy, inflammation	Frontotemporal dementia, distal myopathy, childhood-onset neurodegeneration with ataxia, dystonia and gaze palsy, Paget disease of bone-3	222
2011	9p21.2	C9orf72	Autosomal dominant	40	7	Autophagy, global RNA alterations, intracellular trafficking defects, nucleocytoplasmic transport defects, proteostasis defects	Frontotemporal dementia	55,56
2012	17p13.2	PFN1	Autosomal dominant	<1	<1	Impaired axonal growth and cytoskeletal organization	None	223
2013	7p15.2	HNRNPA2B1	Autosomal dominant	Unknown	Unknown	Altered ribostasis	Inclusion body myositis with early-onset Paget disease with or without frontotemporal dementia 2, multisystem proteinopathy	224
2013	12q13.13	HNRNPA1	Autosomal dominant, de novo	Unknown	Unknown	Altered ribostasis	Inclusion body myositis with early-onset Paget disease with or without frontotemporal dementia 3, multisystem proteinopathy	224
2014	2q35	TUBA4A	Autosomal dominant	<1	<1	Impaired axonal growth and cytoskeletal organization	Frontotemporal dementia	225
2014	5q31.2	MATR3	Autosomal dominant	<1	<1	Altered ribostasis	Distal myopathy with vocal cord and pharyngeal weakness	226
2014	22q11.23	CHCHD10	Autosomal dominant	<1	<1	Mitochondrial dysfunction	Frontotemporal dementia, spinal muscular atrophy (Jokela type), isolated mitochondrial myopathy	227
2015	12q14.2	TBK1	Autosomal dominant	<1	<1	Autophagy, inflammation	Frontotemporal dementia	112
2016	4q33	NEK1	Not established	2	2	DNA damage, impaired cytoskeletal organization and cell cycle	Short-rib thoracic dysplasia 6 with or without polydactylism	112,114
2016	16p13.3	CCNF	Autosomal dominant	4	2	Proteostasis defects	Frontotemporal dementia	228
2016	21q22.3	CFAP410	Not established	Unknown	Unknown	Impaired cytoskeletal organization	Axial spondylometaphyseal dysplasia, retinal dystrophy with macular staphyloma	107
2017	10q22.3	ANXA11	Autosomal dominant	Unknown	Unknown	Dysregulation of calcium homeostasis and stress granule dynamics	Inclusion body myopathy and brain white matter abnormalities	229
2018	12q13.3	KIF5A	Autosomal dominant	<1	<1	Impaired cytoskeletal organization and axonal transport	Charcot-Marie-Tooth type 2, hereditary spastic paraplegia	72,73
2018	10q24.31	ERLIN1	Autosomal recessive	Unknown	Unknown	Dysregulation of inositol 1,4,5-trisphosphate intracellular ion channels	Hereditary spastic paraplegia	78
2019	3p21.1	GLT8D1	Autosomal dominant	Unknown	Unknown	Impaired ganglioside synthesis	None	230
2019	17q21.2	DNAJC7	Not established	Unknown	Unknown	Not established	None	84
2021	4p16.3	HTT	Autosomal dominant	Unknown	Unknown	Not established/nucleocytoplasmic transport defects	Huntington disease, Lopes-Maciel-Rodan syndrome	91
2022	9q22.31	SPTLC1	De novo	Unknown	Unknown	Disruption of sphingolipid metabolism	Hereditary sensory and autonomic neuropathy type 1A	97,98

Genes are listed chronologically based on their year of discovery.

^aPercentage of amyotrophic lateral sclerosis (ALS) cases explained by mutations in the corresponding disease-causing genes.

^bPhenotypes associated with the genes extracted from the Online Mendelian Inheritance in Man database.

Mendelian genes causing ALS

The four most common ALS genes

SOD1.

Thirty years ago, one of the first successful applications of linkage analysis in neurological disease identified a chromosome 21q locus segregating with the disease in 23 families with ALS cases¹⁸. Mutation screening and segregation analysis subsequently identified pathogenic variants in SOD1 as the causative gene¹⁹. This ground-breaking discovery showed for the first time that a genetic mutation could underlie ALS and that modern genetic approaches could reliably detect causative mutations. More than 200 SOD1 mutations have been reported, accounting for approximately 12% of familial and 1–2% of sporadic ALS in populations of European ancestries^20,21. Most disease-causing SOD1 variants are missense mutations²⁰. Although most of these mutations are inherited in an autosomal dominant manner, Asp91Ala, which is frequent in the Scandinavian population, is associated with a recessive form of ALS^22–24.

The clinical phenotype of ALS is variable across the various SOD1 mutations but can also be remarkably stereotypical for a particular disease-causing variant. For example, the Ala5Val substitution, which is one of the most common SOD1 mutations among North American patients with ALS, leads to a rapidly progressive, dominant form of the disease²⁵. Similarly, the Gly73Ser and Ser106Leu mutations present with a more severe phenotype than many other SOD1 mutations (for example, Gly38Arg and Gly94Cys)^26–31. By contrast, patients homozygous for the Asp91Ala allele progress much more slowly, often developing urinary incontinence and sensory abnormalities in the late stages of their illness²².

Mutant SOD1 protein has a toxic gain of function. It seems to cause neuronal cell death via several mechanisms, including excitotoxicity, oxidative stress, non-cell-autonomous toxicity of neuroglia, mitochondrial dysfunction and axonal transport disruption³². Misfolded wild-type SOD1 was also suggested to be involved in sporadic ALS³³. These mechanisms provided essential insights into the pathogenesis of ALS, showing how wild-type and mutant SOD1 could drive the disease process independently. Building on these observations, SOD1-related ALS was among the first neurological diseases for which molecular therapy development was attempted to block the production of the misfolded protein^34–36.

TARDBP.

Another milestone in ALS research was the 2006 discovery by V. Lee, J. Trojanowski and colleagues of the nuclear depletion and cytoplasmic accumulation of TAR DNA-binding protein (TDP43) in motor neurons of patients with ALS³⁷. Two subsequent studies identified mutations in the C terminus-encoding region of the gene TARDBP as a cause of autosomal dominant ALS^38,39. The prevalence of TARDBP mutations in ALS is lower than that of mutations in SOD1, accounting for only 4% of familial and 1% of sporadic cases¹². Regardless of the underlying pathogenic variants, abnormal cytoplasmic TDP43 inclusions are found in the brain and spinal cord in 97% of ALS cases⁴⁰. This widespread pathological feature has become a hallmark of ALS and FTD, a rare and progressive disease resulting from neurodegeneration in the frontal and temporal lobes of the brain. Subsequent research has shown that TDP43 can be deposited in other neurological conditions (for example, FTD and limbic-predominant age-related TDP43 encephalopathy (LATE))⁴¹. The C-terminal domain of TDP43, also known as the glycine-rich region, has prion sequence similarity⁴². As for FUS and SOD1, prion-like motifs in TDP43 have been implicated in cell-to-cell transmission that may induce misfolding and aggregation of normal counterparts^43–45. The role of TDP43 as an RNA-binding protein also implicated RNA processing as a central process in the motor neuron degeneration observed in ALS^40,42.

FUS.

A year after the discovery of TARDBP and its involvement in ALS as a dysregulated RNA-binding protein, pathogenic variants were detected in a similar RNA-binding protein encoded by FUS^46,47. Disease-causing variants are clustered in the region of the gene that encodes the C terminus of the protein, which functions as a nuclear localization signal, and they account for 4% of familial and 1% of sporadic ALS^12,44. Although mutations in FUS are predominantly associated with autosomal dominant ALS, de novo mutations have also been reported^46,48. In contrast to patients carrying mutated forms of other ALS-associated genes, some patients with FUS mutations may present at a young age and with an aggressive disease course^49–51.

The identification of FUS provided additional evidence, along with TARDBP, that points to RNA metabolism disruption as a putative pathogenic mechanism in ALS. Reminiscent of TARDBP, FUS mutations are associated with a similar nuclear to cytoplasmic mislocalization and the formation of cytoplasmic inclusions of the mutated protein, highlighting the role of these cellular processes in ALS pathogenesis^46,47,50,52. Recent research has focused on the role of stress granules in the pathogenesis of TARDBP- and FUS-related ALS^53,54, although the pathogenic role of these phenomena remains unclear.

C9orf72.

The next breakthrough in ALS research was the discovery of a hexanucleotide (GGGGCC) repeat expansion in the C9orf72 gene in 2011 (refs. 55,56). Since then, it has become clear that C9orf72 repeat expansion is the most common genetic cause of ALS and FTD, accounting for ~40% of familial and ~7% of sporadic ALS in populations of European ancestries and an equivalent proportion of FTD cases⁵⁷. Wild-type alleles carry fewer than 30 repeats, whereas pathogenic alleles have been reported to range between 700 and 1,600 repeats⁵⁵. Expanded repeats are detected using a repeat-primed PCR; accurate sizing of the expansion relies on Southern blotting^55,56,58, although initial studies suggest that long-read sequencing may also be helpful⁵⁹. Studies examining repeat size and clinical heterogeneity among patients with C9orf72 expansion have yielded contradictory results⁶⁰, with some reporting a negative correlation between expansion size and age at onset⁶¹, whereas others found a positive correlation^62,63.

Multiple mechanisms have been proposed to explain C9orf72-related neurodegeneration. Expanded repeats may cause a toxic gain of function that involves disruption of RNA metabolism similar to TARDBP- and FUS-mediated pathogenesis. For example, the expansion-containing RNAs sequester important cellular factors into nuclear RNA foci⁶⁴. In addition, a non-canonical mechanism of translation initiation, known as repeat-associated non-AUG-initiated (RAN) translation, was shown to produce dipeptide repeats from multiple reading frames in C9orf72-positive cases^12,65,66. These dipeptides interfere with cytoskeletal dynamics and axonal transport by interacting with kinesin motor proteins, such as KIF5A, which has also been implicated in ALS⁶⁷ (see below). A loss-of-function mechanism because of C9orf72 haploinsufficiency may likewise trigger neurodegeneration in ALS⁶⁸. Although considerable in vitro and in vivo evidence supports each of these mechanisms, additional studies are required to learn precisely how the C9orf72 repeat expansion causes neuronal death. An intriguing possibility is that multiple pathological processes operate in tandem to trigger the neurodegenerative process. Indeed, the C9orf72 expansion has also been reported in other neurological disorders, including Alzheimer disease, Parkinson disease, progressive muscular atrophy and Huntington disease-like syndrome^69,70.

Recent gene discoveries in ALS

Mutations in several other genes have been described as less frequent causes of ALS, including VCP, PFN1, MATR3, CHCHD10, TBK1 and NEK1 (ref. 16) (Fig. 1). Here, we discuss five ALS genes identified after 2018 using whole-exome and whole-genome sequencing, each providing novel insights into disease pathogenesis and a novel starting point for therapeutic development.

Fig. 1 |. Proportion of familial and sporadic ALS cases attributed to mutations in the corresponding disease-causing genes.

Mutations in the known amyotrophic lateral sclerosis (ALS) genes explain approximately 76% of familial and 25% of sporadic ALS²³¹. The four most common ALS-associated genes, C9orf72, SOD1, TARDBP and FUS comprise 60% of familial and 11% of sporadic ALS. The proportion attributed to recently identified or rarely implicated genes has not been established.

Kinesin family member 5A (KIF5A).

Although the KIF5A locus had been suggested as a candidate genetic risk factor⁷¹, it was identified as a definitive ALS gene based on a genome-wide study of common and rare variants in 2018 (refs. 72,73). The ALS-related mutations result in the skipping of exon 27, resulting in a protein with a novel 39 amino acid residue C-terminal sequence. The same mutations have also been shown to cause FTD⁷⁴. These mutations lead to a toxic gain of function involving abolished auto-inhibition of the kinesin activity, resulting in hyperactive axonal transport⁷⁵. Discovering mutations in KIF5A and several other motor protein and cytoskeleton-related genes, such as DCTN1, PFN1, SPAST and TUBA4A, highlighted the pivotal role of the cytoskeleton in ALS pathogenesis⁷⁶. It also suggests that therapeutic approaches that enhance cytoskeletal integrity or maintain axonal transport machinery may benefit patients with ALS⁷⁷.

Endoplasmic reticulum lipid raft associated protein 1 (ERLIN1).

ERLIN1 was identified as a causative gene for an early-onset, autosomal recessive form of ALS based on a whole-exome sequencing analysis of a large, consanguineous Turkish family⁷⁸. Many affected family members manifested symptoms in early life (before 25 years) and displayed a slowly progressive form of ALS. ERLIN1 mutations were first reported in three families with hereditary spastic paraplegia, and since then, only one additional mutation has been described in a Chinese hereditary spastic paraplegia pedigree, indicating a rare role of ERLIN1 in neuromuscular diseases^79,80.

ERLIN1 encodes endoplasmic reticulum lipid raft associated protein (ERLIN1), which is part of a protein complex implicated in degrading inositol 1,4,5-trisphosphate intracellular receptor (IP₃R) ion channels. Dysregulation of IP₃Rs disrupts calcium release from the neuronal endoplasmic reticulum, leading to synaptic loss and dysfunction⁸¹. Aberrant activity of IP₃Rs was noted in Huntington disease⁸² and ITPR1-related spinocerebellar ataxia⁸³, again showing that mutations in a single gene may be associated with diverse neurological conditions (that is, pleiotropy).

DnaJ heat shock protein (Hsp40) family member C7 (DNAJC7).

Rare protein-truncating and damaging missense variants in DNAJC7 were enriched in patients with ALS in an exome-wide association analysis⁸⁴. Further exploration in Asian cohorts identified potentially deleterious DNAJC7 variants in sporadic cases of ALS^85–88. However, pathogenic DNAJC7 mutations co-segregating with the disease in an ALS pedigree have not been reported to date, which would confirm mutations in this gene as disease-causing. Nonetheless, functional evidence supports involvement of the gene in ALS. DNAJC7 encodes a member of the Hsp40 heat shock protein family, and other members of the same family act as molecular chaperones involved in the folding and clearance of misfolded proteins⁸⁹. For example, Hsp90 contributes to the elimination of misfolded TDP43, and DNAJC7 in mice may regulate the degradation of ALS-associated proteins, such as SOD1, TDP43 and FUS⁹⁰. Protein misfolding and aggregation are a central theme in ALS, and mitigating misfolding and promoting clearance by modulating the DNAJC7 function presents an attractive therapeutic target. Such an intervention would be broadly useful for other ALS-related proteinopathies.

Huntingtin (HTT).

Pathogenic CAG repeat expansions in the first exon of the HTT gene were recently identified as a rare cause of pure ALS/FTD phenotypes based on large whole-genome sequence datasets⁹¹. Although infrequent, the repeat expansion was present in multiple patient cohorts⁹¹, and recent work has confirmed the original findings^92,93. None of the patients displayed chorea, abnormal involuntary movements that are a core symptom of Huntington disease, during their illness, implying a pure ALS/FTD phenotype in these cases. The post-mortem findings of two patients with ALS with HTT repeat expansions revealed the classic motor neuron loss and TDP43 pathology of ALS/FTD, ruling out mimic syndromes as an explanation. The neostriatal atrophy that pathologically defines Huntington disease was also absent in these cases⁹¹. Interestingly, mutant huntingtin protein disrupts the nucleocytoplasmic transport⁹⁴, a process also implicated in the pathogenic C9orf72 repeat expansions⁹⁵. Although HTT expansions are rare in ALS/FTD, this finding again emphasizes the molecular overlap across distinct neurodegenerative clinical entities. It highlights the power of genomics to define and classify them molecularly. Clinically, it opens opportunities to bridge to disease-modifying interventions, such as antisense oligonucleotide (ASO) therapies designed to lower the amount of mutant proteins in the brain (see below)⁹⁶.

Serine palmitoyltransferase long chain base subunit 1 (SPTLC1).

Recently, de novo mutations in the SPTLC1 gene were identified as a cause of juvenile ALS, a rare form of ALS that occurs before the age of 25 years^97,98. Mutations in this gene cause autosomal dominant hereditary sensory autonomic neuropathy, type 1A (HSAN1A), by disrupting an essential enzyme complex in the sphingolipid synthesis pathway⁹⁹. These data broaden the phenotypes associated with SPTLC1 mutations to include juvenile ALS and implicate disruption of sphingolipid metabolism in motor neuron disease. Cell-based assays of serine palmitoyltransferase activity confirmed that the recurrent de novo Ala20Ser mutation altered the function of the encoded enzyme, leading to increased aberrant utilization of alanine and glycine as substrates^97,100. Notably, nutritional supplementation with serine is a current therapeutic strategy for HSAN1A¹⁰¹ and may be a viable option among juvenile patients with ALS, although further work is needed to confirm this⁹⁷.

Other aspects of ALS genetics

Genome-wide association studies in ALS

As for other neurological diseases, GWAS have been a major contributor to our knowledge of the genetic basis of ALS¹⁰². The first ALS GWAS was published in 2007 and did not identify any significant genetic loci¹⁰³. Since then, much larger GWAS projects have revealed additional risk loci for ALS. In 2009, an association analysis for sporadic ALS first implicated the UNC13A locus¹⁰⁴. The association signal in chromosome 9p21 and SOD1 reached genome-wide significance in a Finnish cohort of familial ALS^105,106. These efforts paved the way for identifying the C9orf72 repeat expansion in 2011 (refs. 12,55,56). Further GWAS identified and replicated variants in C9orf72, CFAP410, KIF5A, MOBP, SARM1, SOD1, TBK1, TNIP1 and UNC13A^72,107–110 (Fig. 2). Initial studies comprised individuals with European genetic ancestry, whereas recent multi-ethnic meta-analyses contained Asian cohorts^{105,107,108,111}. Notably, GWAS and familial studies worked synergistically to unravel the genetic architecture of ALS with many Mendelian genes involved in ALS, such as SOD1, NEK1, C9orf72 and TBK1, also showing association in GWAS^{19,55,71,105,106,112–114}.

Fig. 2 |. Thirty years of gene discovery in ALS and FTD.

The first Mendelian gene for amyotrophic lateral sclerosis (ALS) was identified in 1993 (ref. 19), and the first genome-wide association study (GWAS) for ALS was published in 2007 (ref. 103). Further GWAS and gene hunting studies identified several significant loci (P < 5 × 10⁻⁸) and genes associated with ALS, frontotemporal dementia (FTD) or ALS/FTD. Summary-level associations were extracted from the EBI-GWAS website¹⁰².

Pathway and multi-omic analyses in ALS

Genetic data have implicated various pathways and cellular mechanisms in the pathogenesis of familial ALS, such as protein homeostasis, cytoskeleton alterations and RNA metabolism¹⁶. More recently, our knowledge of the biological processes involved in the sporadic form of the disease has also improved; enrichment analysis based on large-scale datasets identified genes and pathways that can be targeted for therapeutic intervention. These include neuronal projection, membrane trafficking and signal transduction mediated by ribonucleotides¹¹⁵. Detailed dissection of these pathways identified six differentially expressed genes associated with increased ALS risk (ATG16L2, ACSL5, MAP1LC3A, MAPKAPK3, PLXNB2 and SCFD1)¹¹⁵. Autophagy was also central to C9orf72-related neurodegeneration¹¹⁵, an observation consistent with the encoded protein being an integral part of the autophagy initiation complex^116,117.

Genetic data can also identify the cell types involved in ALS, often yielding surprising insights and new avenues for research. For instance, an approach combining GWAS data and single-nucleus RNA sequencing data identified cortical GABAergic interneurons and oligodendrocytes as associated with ALS risk¹¹⁵. Nearly 20 years ago, autopsy studies implicated the same class of interneurons¹¹⁸. The confluence of genetic and pathological data suggests that damage to GABAergic interneurons is an early event in ALS. Similarly, this data-driven approach demonstrates that oligodendrocytes contribute to the disease pathogenesis and are worthy of additional research.

Various omics-based approaches have provided insights into mechanisms of ALS pathogenesis. Spatial transcriptomics is an emerging technique that maps gene expression in individual cell types by generating spatially resolved RNA sequencing profiles from tissues of interest. Using this cutting-edge approach, Maniatis and colleagues¹¹⁹ identified TYROBP–TREM2-mediated signalling as an early mechanism associated with microglial gene expression in ALS mice. More recently, large-scale transcriptomics analysis of human post-mortem ALS spinal cords identified new candidate genes with cell-type-specific functions; microglia were assigned as a putative cell type for ACSL5 and oligodendrocytes for FNBP1, SH3RF1 and NFASC¹²⁰. Transcriptome-wide analysis incorporating clinical data revealed that expression of microglial gene modules correlated with disease duration, highlighting the role of glial activation in ALS¹²⁰. A multi-omic analysis of human induced pluripotent stem cell-derived spinal motor neurons identified altered lipid metabolism pathways in mutant SOD1 and C9orf72 spinal motor neuron populations¹²¹. Additional metabolomics profiling confirmed the upregulation of lipid metabolism and identified elevated arachidonic acid levels in spinal motor neurons, pinpointing a potential therapeutic target¹²¹.

Although these studies are relatively new, they highlight the application of genetic data to predict functional outcomes. Future large-scale studies will shape future drug development as personalized medicine takes centre stage.

Whole-genome sequencing and other mutation types in ALS

Various other mutation types have been implicated in ALS, highlighting the diverse genetic architecture of the disease. Repeat expansions, structural alterations, non-coding mutations, de novo variation and activation of human endogenous retroviruses (HERVs) fill the contours of the ALS genetics puzzle. Our ability to detect these distinct types of mutation has been driven by advances in next-generation sequencing technologies and the evolution of more powerful bioinformatics tools.

The discovery of the C9orf72 repeat expansions implicated this type of structural variation in ALS⁵⁵. This realization, and the availability of reliable in silico repeat detection tools, paved the way for the subsequent identification of HTT repeat expansion as a cause of ALS⁹¹. In addition, repeat expansions in ATXN1, ATXN2 and NIPA1 have now been associated with an increased risk of developing ALS^122–126. Other structural variants have been identified occasionally in patients with ALS, including an insertion in ERRB4 and an inversion in VCP¹²⁷.

Non-coding variants within the introns and untranslated regions of genes may account for a considerable proportion of the missing heritability of ALS¹²⁸. Previously, a linkage analysis combined with targeted sequencing in a large family identified an intronic variant that introduces a pseudoexon into the SOD1 transcript, leading to truncation of the protein¹²⁹. More recently, next-generation sequencing analysis showed an increased burden of untranslated region variants in SOD1, TARDBP, FUS, OPTN, VCP and UBQLN2 (ref. 130). Variants in the 3′ untranslated region of the IL18RAP gene may protect against developing ALS¹³¹. Intronic mutations leading to aberrant splicing of SOD1 and KIF5A have also been described in large ALS families, confirming the pathogenic potential of this often-overlooked part of the genome^{74,129,132,133}. Additionally, aberrant transcript splicing has long been recognized as a core feature of ALS due to the loss of the normal function of TDP43 of repressing non-conserved cryptic exons, leading to nonsense-mediated decay of the associated transcripts^134,135. The UNC13A risk factor can act in a similar manner by potentiating cryptic exon inclusion¹³⁶.

Although similar molecular mechanisms are involved in familial and sporadic ALS, the disease-causing mutations identified in familial cases to date explain only a small proportion of sporadic cases. This may be due to incomplete penetrance in the families, lack of DNA from other relatives, small family structure, or lack of the proband knowing clinical information about another. In addition, spontaneous de novo mutations have been described in the FUS, SOD1 and VCP genes, explaining at least some sporadic cases^48,137,138. Parent–offspring trio sequencing identified spontaneous variants in SPTLC1 as the underlying cause in sporadic juvenile ALS cases^97,98. This research suggests that de novo mutations may be more crucial in the pathogenesis of ALS than previously recognized.

A limitation of the current genetic studies is that they are constrained by the number of nucleotides sequenced in each read (that is, ~200 bp on Illumina platforms). However, nearly 40% of the genome is made of highly repetitive retroviral elements, and short reads derived from these regions are challenging to align to reference genomes and are thus often discarded from analyses. Long-read sequencers, such as the Pacific Biosciences (PacBio) or Oxford Nanopore technologies, are needed for this purpose. These retrovirus-derived genomic regions may have a critical role in the pathophysiology of ALS, and it is crucial to determine whether mutations in these regions contribute to ALS. For example, autopsy studies show that reactivation of HERVs, particularly the HERV-K subtype HML-2, is associated with ALS and may influence the clinical course of the disease¹³⁹. Reverse transcriptase, a marker of endogenous retroviral activity, has been detected in the blood and cerebrospinal fluid (CSF) of patients with ALS, suggesting retroviral involvement in ALS¹⁴⁰. Moreover, the HML-2 envelope triggers an ALS-like syndrome in transgenic mice¹⁴¹. Antiretroviral therapies that inhibit HERV-K in patients with ALS are currently in clinical trials¹⁴². Despite these compelling data, the information provided by long-read sequencing will be required to confirm the role of HERVs in the pathogenesis of ALS because of their repetitive nature and typical length (~7 kb).

As the size and diversity of cohorts increase, so will the ability to resolve the genetic architecture of ALS, providing further insight into the cellular processes that underlie motor neuron death. Making these data readily available to researchers is equally essential and will maintain the momentum of gene discovery in ALS by lowering the costs associated with genomic research. Several online platforms, such as Answer ALS, ALS Compute, New York Genome Center ALS Consortium, Project MinE and the database of genotypes and phenotypes (dbGaP), are already sharing genetic data and enabling future research.

Overlap between ALS and other diseases

A central theme emerging from genetic studies is the overlap of ALS with other diseases and metabolic traits. An early hint of this came with the discovery of VCP mutations in families with ALS that also manifested FTD^143,144. Mutations in this gene had been implicated as a cause of an unusual syndrome comprising inclusion body myositis, Paget disease of the bone and FTD. The rarity of VCP mutations led to the belief that the co-occurrence of ALS and FTD in the same patient was an outlier phenomenon. It was not until the revelation of C9orf72 as a common cause of both ALS and FTD that the importance of the phenotypic overlap was appreciated⁵⁵. Since then, several other genes causing both ALS and FTD, such as CHCHD10, KIF5A, FUS, HTT, SPTLC1, SQSTM1, TARDBP and TBK1 (refs. 12,69), have been identified, strengthening their convergence into a single spectrum. It is now recognized that up to 60% of patients with ALS develop cognitive impairment during their illness¹⁴⁵. Conversely, up to one-third of patients with FTD develop motor dysfunction¹²⁷.

Genetic evidence also suggests that energy metabolism is dysregulated in ALS. Older epidemiological studies had reported high levels of cholesterol and other lipids in the central nervous system (CNS) and circulation in ALS^97,146–156. However, these results were based on small numbers of patients, and several other studies observed no difference in blood lipid levels between patients with ALS and controls¹⁵⁶ or reported a negative association^157–159. A possible explanation for excess cholesterol in the CNS of patients with ALS is that the cholesterol is released after neuronal death and is not removed properly^147,159. More recently, Mendelian randomization studies based on genetic data from sizeable patient cohorts confirmed a causal link between blood LDL-cholesterol levels and ALS^109,160,161. Furthermore, a metabolomics analysis identified recurrent dysregulation in pathways related to lipid metabolism in independent ALS cohorts¹⁶².

Mendelian randomization studies also suggested that type 2 diabetes is protective against ALS^163–166. These data are consistent with population-based observational studies showing that a high body mass index and type 2 diabetes are associated with a lower risk of ALS^167,168. These observations suggest that metabolic disturbance may be a fundamental component of the ALS syndrome, similar to what has been observed in more common neurodegenerative diseases¹⁶⁹.

Aside from the phenotypic overlap, there is also considerable evidence of pleiotropy, with mutations in the same gene causing two distinct neurological diseases (but not in the same patient). For example, mutations in the N-terminal domain-encoding region of KIF5A are a known cause of Charcot–Marie–Tooth type 2 and hereditary spastic paraplegia^170,171. Another example of genetic intersection is the SPTLC1 gene: the disease-causing mutations result in diverse pathologies leading to either ALS or a form of peripheral neuropathy known as hereditary sensory autonomic neuropathy type 1 (refs. 97,99,101). Expanded repeats in the genes associated with hereditary cerebellar ataxia (ATXN1 and ATXN2) and hereditary spastic paraplegia (NIPA1) are also risk factors for ALS^122–126. Genetic data show that ALS is a complex and nuanced disease with phenotypic and aetiological overlap with other neurological and systemic conditions.

Diverse populations and rare ALS subtypes

The incidence of ALS may be lower among populations of African, Asian and Hispanic ancestries than among populations of European ancestries^3,172. It is unclear whether this rate disparity is due to differences in epidemiological study design or in population demographics, environmental factors or genetic predisposition³. There is compelling evidence that the genetic architecture of ALS varies with geography¹⁷³. The pathogenic repeat expansion of C9orf72 is a widespread cause of the disease in populations of European ancestries because of a founder effect but is diminishingly rare in populations of Asian ancestries¹⁷⁴ (Fig. 3). By contrast, mutations in SOD1 and FUS are more common in Japanese patients with ALS¹⁷⁵.

Fig. 3 |. Frequency of causal mutations in the four most common ALS genes in diverse populations.

The incidence of amyotrophic lateral sclerosis (ALS) varies in distinct populations. Although the C9orf72 repeat expansion is rare in Asia, it is the most frequent cause of ALS in Scandinavian populations. Similarly, a founder mutation in TARDBP explains approximately 35% of the disease in Sardinia.The frequency of familial cases explained by the mutated genes in different geographical regions is obtained from references^173,232,233; the regions are North America (Canada and USA); Europe (Italy, Germany, Spain, Switzerland, France, Belgium, UK, Netherlands and Slovenia); Sardinia; Scandinavia (Sweden and Finland); Asia (Japan, Korea, China, India and Iran); and Australia.

Isolated populations with a high incidence of ALS have been closely studied to understand the disease^176,177. For example, nearly one-third of ALS cases in the island population of Sardinia are due to a founder mutation in TARDBP (Ala383Thr)¹⁷⁸, and the Asp91Ala mutation in SOD1 is more frequent in the Scandinavian population^23,24 (Fig. 3). Likewise, a VAPB mutation (Phe56Ser) accounts for 44% of familial ALS in Portuguese–Brazilian populations¹⁷⁹. Investigating unique ALS variants also offers meaningful opportunities to understand ALS. For example, exome sequencing of families with Brown–Vialetto–Van Laere syndrome (ALS plus deafness and blindness) identified mutations in a series of genes encoding riboflavin receptors^180,181. This finding led to high-dose vitamin B2 supplementation as an effective treatment that halts progression in this rare form of ALS¹⁸⁰.

Gene–environment interactions

Epidemiological studies showed that several environmental and lifestyle factors, such as agricultural and industrial chemicals, certain occupations, cigarette smoking and heavy exercise, may be involved in ALS¹⁶⁰. However, these studies may have been hampered by myriad confounding variables and the difficulty of establishing causal relationships based on observational data. Mendelian randomization overcomes this limitation by using genetic elements to explore the causal relationship between exposure and outcome. This analytical technique relies on genotypes that remain constant over a lifetime and are randomized during game-togenesis. Because of this, it provides a reliable identification of causal associations by reducing confounding and excluding reverse causality^161,182. To date, Mendelian randomization studies point to alcohol consumption, educational attainment, physical exercise, dyslipidaemia and smoking as ALS risk factors^{109,160,161,166,183}. Epigenetic mechanisms, such as DNA methylation, microRNA (miRNA) dysregulation, chromatin remodelling and histone modifications, have been shown to operate in ALS^184–189 and are suspected of mediating the effects of these environmental factors on disease pathogenesis. How these exposures alter the cell’s epigenetic profile to bring about ALS is unknown.

These environmental and lifestyle exposures are common in the population, yet ALS is rare. Gene–environment interactions may explain this incongruity; extrinsic or genetic factors alone would not lead to motor neuron degeneration. Instead, only those exposed individuals with an underlying genetic predisposition would be at risk of disease¹⁹⁰. Gene–environment interactions may also be especially relevant in late-onset, rare diseases whose sporadic occurrence appears stochastic. In reality, it hints at a much more complicated theory of the disease¹⁶⁰.

Similarly, infections can also unmask or precipitate ALS. In particular, patients with HIV or HTLV-I infection can occasionally develop ALS. When ALS occurs in patients with HIV, it is often rapidly progressive; however, if treated with antiretroviral drugs early in the syndrome, the symptoms can be halted or reversed in some cases¹⁹¹.

A model of ALS that integrates genetics, environment and ageing

Various models have been proposed to explain the occurrence of ALS. Although each schema has merit, they typically postulate that disease-causing events occur sequentially or do not account for the interactions between the various factors that drive cell death^11,192,193. For example, the multistep model of ALS portrays the path to the disease as a series of steps that occur over time from genetic and environmental insults¹⁹³. Aside from the obvious difficulty in calculating the number of steps involved in the disease process, it implies that these events occur separately. Instead, numerous factors impinge on motor neurons in a complex and interdependent manner to cause neurodegeneration.

Here, we propose an updated model whereby disease-causing mutations, genetic modifiers, ageing and environmental factors come together synergistically to cause disease. Under this paradigm, ALS risk can be visualized as a series of overlapping concentric rings (Fig. 4). At the centre are ALS and the long-term health of the motor neurons. The first ring represents the principal disease-causing mutation, such as in C9orf72 or SOD1, or polygenic risk in the case of sporadic ALS. The second ring represents genetic variation influencing phenotypic manifestations of the disease, such as onset age, site of symptom onset and progression rate. These ancillary genetic effects can be monogenic or polygenic. The third ring symbolizes environmental and lifestyle factors that affect a patient’s risk of developing ALS. These include cigarette smoking, hyperlipidaemia, excessive exercise and military service¹⁶⁰. The outer ring denotes the ageing process, which is still one of the most substantial risk factors for ALS.

Fig. 4 |. A model of ALS that integrates genetics, environment and ageing.

Amyotrophic lateral sclerosis (ALS) may occur owing to multiple overlapping factors, including a primary disease-causing mutation, a genetic predisposition that influences disease manifestation, environmental and lifestyle factors, and ageing. Together, they comprise a multifaceted, complex system, in which the integrity of each element is vital to maintain the integrity of the motor neurons.

Our model conveys the essential concept of the underlying factors operating collectively. This captures the interdependent forces that lead to neurodegeneration, as well as the view that each has a role in disease occurrence. Such multiform interactions between genetic, environmental and lifestyle factors are now recognized as central drivers of neurodegeneration¹⁹⁴. For instance, several studies have shown that an individual could be genetically predisposed to cigarette smoking, which then increases the risk of developing ALS^161,166. Gene–environment interactions may be particularly relevant in the pathogenesis of rare, sporadic diseases that occur in later life, such as ALS¹⁶⁰.

The model is reminiscent of Pritchard’s omnigenic archetype proposed for complex diseases, while also incorporating salient clinical aspects of ALS¹⁹⁵. Indeed, this multidimensional model applies to almost all neurodegenerative diseases, such as Alzheimer disease and Parkinson disease. Instead of offering a mechanistic view of cellular pathophysiology, a perspective that shifts with each discovery, it provides a holistic and enduring compendium into how geneticists think about diseases. In this way, our prototype is a helpful guide for designing future genetic and environmental studies by emphasizing the need to incorporate diverse factors into the analysis.

Our ring model also has implications for ALS treatment. The complex ecosystem may explain treatment failures in molecular therapies targeting monogenic forms of ALS; addressing ancillary factors may be necessary to slow disease progression once symptoms emerge. Applying molecular therapies to patients with ALS will test this hypothesis in the next few years. These first-in-human clinical trials may also establish a better understanding of how the components interact to cause motor neuron degeneration.

Molecular therapies in ALS

Current treatment options for ALS are limited, with only three drugs currently on the market. Riluzole, a small molecule that targets voltage-dependent Na⁺ channels to reduce motor neuron excitotoxicity, was approved by the FDA in 1995. Edaravone, a small-molecule antioxidant, gained FDA approval in 2017. More recently, the FDA approved a combination of sodium phenylbutyrate and ursodoxicoltaurine (Albrioza) that is aimed at ameliorating endoplasmic reticulum stress and mitochondrial dysfunction.

Although these drugs represent progress in the ALS field, they have only a modest effect on slowing disease progression^196,197. Unsurprisingly, efforts continue towards developing drugs with greater efficacy and longer-lasting impact. To this end, molecular therapy approaches have gained traction in ALS. The advancements in our understanding of ALS genomics described above coincide with the first oligonucleotide and gene therapies hitting the markets, most notably embodied by the success of nusinersen and onasemnogene abeparvovec-xioi for treating spinal muscular atrophy¹⁹⁸.

Molecular therapies come in various modalities that offer the ability to selectively silence, upregulate, edit or introduce a gene or transcript to correct the underlying cause of disease either temporarily or permanently (Box 1). Various molecular therapy modalities can potentially treat ALS depending on the nature of the causal mutation. Diseases caused by a gain-of-function mutation may be mitigated by silencing the mutant gene or transcript. Alternatively, upregulation or introduction of a gene or transcript may correct disease caused by a loss-of-function mutation. An early example of gene therapy attempting to treat ALS by replenishing a supporting protein is Engensis (also known as VM202), a non-viral gene therapy that uses a plasmid to deliver the hepatocyte growth factor gene directly to nerves and muscle¹⁹⁹. A phase I/II clinical trial (Clinicaltrials.gov identifier NCT02039401) showed that intramuscular injections of the therapy were generally well-tolerated and may slow disease progression.

Box 1. Therapeutic modalities to treat ALS.

Molecular therapies include DNA or RNA moieties that are introduced into cells for therapeutic benefit. Molecular therapy that silence the expression of the causal gene can potentially treat diseases that result from gain-of-function mutations. Various modalities can accomplish gene silencing, the most promising of which are oligonucleotide therapeutics, with more than a dozen of this class having gained approval from the FDA. Antisense oligonucleotides (ASOs) are single-stranded oligonucleotides, typically 14–25 nucleotides in length. ASOs are used to alter splicing patterns of genes (for example, nusinersen) or directly decrease the expression level of a transcript. In that instance, the binding of the ASO to a complementary strand sequence in the target mRNA leads to recruitment of RNase H and subsequent degradation of the target mRNA, thereby preventing the translation of the aberrant mRNA. At least five ASOs are in clinical trials for amyotrophic lateral sclerosis (ALS) and more are in development (Table 2).

Small interfering RNAs (siRNAs) and short hairpin RNAs (shRNAs), which function as precursors to siRNAs, are oligonucleotide therapies that use RNA interference (RNAi). RNAi involves recruitment of the RNA-induced silencing complex (RISC), leading to cleavage of the target mRNA, silencing the expression of the target gene. At least five RNAi drugs have been approved, predominantly for rare diseases. Ongoing research shows promise for bringing RNAi-based ALS treatments to the clinic in the future²³⁴.

Gene therapy can treat diseases arising from loss-of-function mutations by introducing or upregulating the wild-type gene or transcript. Gene delivery using viral vectors such as adeno-associated virus and lentivirus, or non-viral vectors, to deliver DNA to the patient’s cells to produce a functional protein, is being extensively tested in the clinic²³⁵. Engensis, a therapy that uses a plasmid to deliver the hepatocyte growth factor gene (HGF), is in a clinical trial for ALS. A newer approach that may lead to successful therapies aims to increase the translation of mRNA from the wild-type allele in disease that results from a mutation in one allele. Small activating RNAs (saRNAs) are also an emerging class of RNA therapeutics designed for RNA activation²³⁶. Other novel RNA-based approaches to upregulate gene expression are under investigation, such as targeted augmentation of nuclear gene output (TANGO) from Stoke Therapeutics²³⁷.

Regardless of the nature or outcome of the mutation, there is also great potential for genome editing to correct a mutation directly at the DNA level. Gene editing with CRISPR–Cas9, zinc finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs) correct mutations at the DNA level. CRISPR–Cas-based gene editing first reached the clinical stage for sickle cell disease, and various CRISPR-based therapies are in clinical trials for other blood disorders, cancers, genetic blindness and more^238–240. ZFN- and TALEN-based therapies have not yet reached this stage for ALS, although the potential for their use is exciting.

Here, we focus on therapies directed at RNA, namely ASOs, as most agents in clinical trials for ALS fall into this class. In that regard, ALS has become an archetypal neurological disorder to study these new methodologies. In comparison with Alzheimer disease and Parkinson disease, the end point in ALS clinical trials is easier to measure and is reached more quickly because of the rapidly progressive nature of the disease. To date, therapeutic development has focused on dampening the toxic gain-of-function effects of the common mutant ALS genes SOD1, FUS and C9orf72. The rationale of these efforts is that removing the aberrant RNA transcripts, for example, with ASOs, will stop the disease process. Although no disease-modifying ALS treatment has yet been approved, several potential therapies are in clinical trials (Table 2).

Table 2 |.

Current clinical trials of molecular therapies for ALS

Treatment	Modality	Target	Phase	Sponsor	clinicaltrials.gov identifier
ION363 (jacifusen)	ASO	FUS	III	Ionis Pharmaceuticals, Inc.	NCT04768972
BIIB067 (tofersen)	ASO	SOD1	III	Biogen	NCT04856982
WVE-004	ASO	C9orf72	I (ALS), II (FTD)	Wave Life Sciences Ltd	NCT04931862
BIIB105 (ION541)	ASO	ATXN2	I	Biogen	NCT04494256
BIIB078 (IONIS-C9Rx)	ASO	C9orf72	I	Biogen	NCT04288856
Engensis (VM202)	Non-viral plasmid gene delivery	HGF addition	II	Helixmith Co. Ltd	NCT05176093, NCT04632225, NCT02039401

ALS, amyotrophic lateral sclerosis; ASO, antisense oligonucleotide.

ION363, informally known as jacifusen after the first patient to receive the drug, is an ASO that targets FUS mRNA to reduce FUS expression²⁰⁰. ION363 was first administered through compassionate-use protocols to a patient with life-threatening ALS linked to the Pro525Leu mutation, and it reduced total and mutant FUS in brainstem tissue from the patient. However, the ASO has potentially broader applicability, and, as of 2021, ION363 has entered a phase III clinical trial to evaluate its efficacy in a larger number of patients with FUS-related ALS. Early indications are that the ASO silences FUS expression, although the effect on clinical outcome remains unclear²⁰⁰.

BIIB067, known as tofersen, is an ASO that inhibits the production of the SOD1 protein. Although tofersen reduced SOD1 levels in CSF, it failed to meet its primary end point in a phase III clinical trial but did show encouraging results in some secondary end points^201,202. Maximizing the beneficial effect of tofersen may entail starting the treatment as early as possible¹⁷. Future efforts will build on this lesson, not just with tofersen, but across the gamut of molecular therapies for neurological diseases. This trend is already emerging. Post-hoc analysis of the phase III study of tominersen showed a beneficial effect among patients with Huntington disease who received treatment early in their illness²⁰³. Recruitment is underway for a phase III trial of tofersen in pre-symptomatic carriers of SOD1 mutation with elevated neurofilament, a CSF biomarker of neuronal death and a harbinger of symptom onset²⁰⁴.

Early reports suggested that the ASO afinersen, developed to suppress the expression of C9orf72 transcripts containing the repeat expansion, may be tolerated at low doses, and it reduced levels of toxic dipeptides within the CSF, a marker of the repeat expansion²⁰⁵. However, in a recent upset in the field, another C9orf72-targeting ASO BIIB078 (formerly IONIS-C9Rx) failed to show clinical benefit in a phase I trial, even trending toward greater decline at a high dose. Nonetheless, efforts to treat C9orf72-ALS continue with the C9orf72-targeting ASO WVE-004 currently in phase I trials for ALS and a phase II trial for FTD.

BIIB105 (formerly ION541) is an ASO designed to degrade ATXN2 mRNA, and it has entered clinical trials for patients with ALS carrying pathogenic CAG repeat expansion in this gene. This therapy is also being tested in patients with ALS without CAG expansions based on the premise that ATXN2 interacts with TDP43, and modulating the gene will prevent TDP43 cytoplasmic accumulation²⁰⁶. Even partial success in this early-stage clinical trial would represent proof of principle in neurodegenerative disorders by demonstrating the benefit of modifying an entire pathway involved in a disease.

Next steps in developing molecular therapies for ALS

Several issues must be addressed to realize the full potential of ASOs and other DNA- and RNA-based therapies as treatments for ALS, and care should be taken to incorporate lessons from previous clinical trials in the ALS space over the past 30 years. These include determining the best time to start treatment, optimizing drug delivery, ensuring adequate target engagement and discerning how to conduct clinical trials involving only a handful of patients. First-in-human dosing of new molecular therapies in ALS is typically done in clinical trials in symptomatic patients or under compassionate-use protocols in individuals in whom disease has progressed to a debilitating and life-threatening stage by the time they receive the first dose. On a molecular level, the regular pool of ~500,000 motor neurons is depleted in the later stages of the disease, decreasing the targets on which an experimental treatment can exert an effect. As the efficacy and safety profile of oligonucleotide therapeutics becomes established, we will see more clinical trials testing these therapies in patients at earlier stages of the disease or even pre-symptomatically.

Direct injection into the CSF results in a favourable distribution of oligonucleotides in the CNS²⁰⁷. However, repeat injections may obstruct CSF pathways leading to hydrocephalus²⁰⁸. Future improvements in delivery may improve the biodistribution and pharmacokinetics of ASOs for neurological diseases²⁰⁹. For example, lipid nanoparticles and other nanomedicines are increasingly being used to enhance the delivery of drugs to target tissue. Newer work has focused on conjugating ligands such as cholesterol moieties to therapeutic molecules, enabling them to cross the blood–brain barrier and potentially opening the intravenous route of administration for CNS diseases²¹⁰. There are even early indications that the oral administration of ASOs is feasible, which would enormously reduce the burden of long-term treatment in chronic diseases²¹¹.

Molecular therapy development to date has focused mainly on the more-common genes responsible for ALS. As oligonucleotide therapeutics become established, we will see more clinical trials testing these therapies in rarer genetic forms of the disease. Establishing the efficacy of ASOs and other individualized therapies will rely on our ability to robustly quantify clinical outcomes in a few or even a single patient. Each human study is an opportunity to learn about the effectiveness and safety of this relatively new class of therapeutics and to understand the cellular mechanisms that drive ALS. Conducting small clinical trials applies across the rare disease spectrum, not just to ALS. Because of this, collaboration and data sharing to aggregate experiences, such as the N=1 Collaborative network, will support this paradigm shift in rare and ultra-rare clinical trial design.

The future of ALS genetics and ASO therapy

Our understanding of ALS has considerably improved over the past three decades, driven mainly by identifying the genetic causes of this fatal neurodegenerative disease. At the same time, we now have several elegant examples of clinical proof of concept for oligonucleotide therapies and other DNA- and RNA-based therapeutic modalities, leading to the emergence of personalized drug development platforms. ASOs are well suited to such platform development as they are rapidly customizable, relatively simple and inexpensive to manufacture and demonstrate a growing safety and efficacy record. As we learn more about the genetics of ALS, and whole-genome and long-read sequencing is applied more broadly in the clinic, there will be more and more opportunities to develop personalized treatments for familial ALS. Over time, these therapeutic approaches may also be extended to the sporadic form of the disease, where targeting multiple genes in a single patient may be necessary.

The current high cost of developing personalized therapies is preventing widespread access. However, the future optimization of therapeutic development platforms will lower the cost and shorten their development time, thereby improving access to new therapies that go straight from the gene to the patient. We envisage a future time when pharmacies in all hospitals are equipped with nucleic acid synthesizers that can be instantly activated to produce safe and effective therapies personalized to the patients admitted to the neurology wards. ALS remains at the forefront of this bold new campaign, offering real hope to patients with neurodegenerative diseases and their families.

Glossary

De novo mutations

Mutations not inherited from either parent and present for the first time in a family.

End point

The targeted outcome of a clinical trial to determine the efficacy and safety of the therapy. The amyotrophic lateral sclerosis functional rating — revised (ALSFRS-R) score is the commonly used primary end point for ALS clinical trials.

Enrichment analysis

A statistical approach to identify over-represented or differentially expressed genes or biological pathways contributing to a phenotype.

Familial aggregation studies

Methods to assess the relative risk of a trait in affected family members compared with unaffected family members. Familial aggregation is the tendency for a trait to occur or cluster in multiple family members beyond what would be expected by chance

Founder effect

The reduction in genetic diversity when a newly established population is descendent from a small number of ancestors.

Genome-wide association Studies

(GWAS). Studies in which genetic variants across the genome are genotyped and tested for association with a trait.

Haploinsufficiency

A situation in which a diploid organism has only one copy of a gene when both copies are required for proper function.

Hydrocephalus

Abnormal cerebrospinal fluid in the ventricles of the brain

Linkage analysis

An approach that examines the cosegregation of the chromosomal region (or regions) with traits of interest, usually designed to localize disease-causing genes in co-segregating segments.

Mendelian randomization

The use of genetic variation to address the causal effect of one trait on another, minimizing the issues of confounding and reverse causality.

Mimic syndromes

A diverse group of conditions, the clinical features of which may resemble amyotrophic lateral sclerosis

Missing heritability

The discrepancy between the heritability explained by trait-associated genetic variants from genome-wide association studies and the amount of total heritability from familial aggregation or twin studies

Prion-like motifs

Unstructured protein domains that can undergo conformational switches in response to environmental stress. Their phase transition activity renders RNA-binding proteins prone to misfolding in neurodegenerative diseases.

Segregation analysis

A study that examines the carrier status of a genetic variant in the affected and unaffected members of a family. It is used to validate whether a specific genetic variant is causal for the disease in a family and the established mode of inheritance based on the pedigreestructure.

Twin studies

Studies that compare the resemblance of monozygotic (identical) and dizygotic (non-identical) twins. An increased concordance rate in monozygotic twins compared with dizygotic twins provides evidence of a genetic component.

Whole-exome sequencing

Targeted sequencing method that selectively determines exome sequences, which are highly enriched for protein-coding regions.

Whole-genome sequencing

Sequencing method that examines the entire genome of an individual