ALS Genetics: Gains, Losses, and Implications for Future Therapies

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder caused by loss of motor neurons from the brain and spinal cord. The ALS community has made remarkable strides over three decades by identifying novel familial mutations, generating animal models, elucidating molecular mechanisms, and ultimately developing promising new therapeutic approaches. Some of these approaches which reduce expression of mutant genes are in human clinical trials, highlighting the need to carefully consider the normal functions of these genes and potential contribution of gene loss-of-function to ALS. Here we highlight known loss-of-function mechanisms underlying ALS, potential consequences of lowering levels of gene products, and the need to consider both gain- and loss- of function to develop safe and effective therapeutic strategies.

In this review, Kim et al. discuss evidence for loss-of-function mutations and mechanisms in amyotrophic lateral sclerosis (ALS), and propose the need to consider both gain- and loss-of function in designing effective and safe therapeutic strategies.

INTRODUCTION

Nearly three decades following the discovery of the first amyotrophic lateral sclerosis (ALS) gene, the field has now entered a therapeutic phase. The promise of human genetics is coming to fruition, with several clinical trials underway to test therapies directly targeting mutant ALS-causing genes and with more in the process of development (Figure 1). Likewise, there is great hope that the remarkable pace of ALS gene discovery will empower the development of even more therapeutic approaches, including ones for sporadic forms of ALS.

Figure 1: ASOs Targeting Mutant ALS Genes.

A. ASOs targeting mRNA from three ALS genes—SOD1, C9ORF72, and FUS—for degradation have been administered to human ALS patients. These ASOs are at various stages of development, as indicated. B. ASOs are delivered broadly throughout the central nervous system after injection into the cerebrospinal fluid at the base of the spinal cord (intrathecal administration).

ALS is an adult onset neurodegenerative disorder that is characterized by degeneration of upper and lower motor neurons, clinically manifesting in a progressive loss of motor function and ultimately death often due to respiratory failure (Taylor et al., 2016). While the most salient clinical feature of ALS is motor involvement, ALS shares many neuropathological and genetic features with another neurodegenerative disorder—frontotemporal dementia (FTD)—the second leading cause of dementia after Alzheimer Disease. The two disorders share common genetic mutations and are often pathologically characterized by ubiquitinated inclusions of the 43-kD TAR DNA-binding protein (TDP-43) and abnormal accumulations of other RNA-binding proteins (Ferrari et al., 2011; Liscic et al., 2008; Morris et al., 2012; Neumann et al., 2006; Orr, 2011).

In the past decade, the development and clinical application of antisense oligonucleotides (ASOs) has taken the field by storm. The approach is elegantly simple: if a mutated gene causes disease because of a toxic function, then target that toxic gene product for degradation – stopping the deleterious RNAs or proteins in their tracks (Figure 1). There are variations to this theme in which other types of oligonucleotides are used to boost production of a gene product by blocking undesirable splicing events, like in treatment of the childhood neuromuscular disorder spinal muscular atrophy (Finkel et al., 2016; Hua et al., 2011; Wan and Dreyfuss, 2017; Wood et al., 2017). ASOs have shown striking efficacy in preclinical studies using animal models or human cells (Becker et al., 2017; McCampbell et al., 2018; Meng et al., 2015; Scoles et al., 2017; Smith et al., 2006; Sztainberg et al., 2015), and clinical trials are underway to test ASOs targeting at least two mutant genes in ALS patients (Miller et al., 2020; ClinicalTrials.gov identifier: NCT02623699; ClinicalTrials.gov identifier: NCT03626012) (Figure 1). There are even efforts to tailor ASOs to target patients on an individual basis (Daskalakis, Stella, 2020). Gene therapy using viral vectors to deliver RNA interference is also emerging as a powerful therapeutic approach to complement the ASO method (Bravo-Hernandez et al., 2020). There is great optimism in the field that these will provide the first disease-modifying therapies for this otherwise fatal disease.

Approximately 10% of ALS cases are considered “familial” (fALS), or transmitted within families, while the remaining are considered “sporadic” (sALS), or presenting without a clear familial history. Although less common than sALS, fALS has played an outsize role in our understanding of disease mechanisms through the discovery of ALS-causing mutations within families and subsequent experimental perturbations of these mutant genes. Since fALS occurs within the same family across multiple generations, genetic approaches can be used to pinpoint the mutated gene that tracks with people who developed ALS and away from those who did not.

This type of analysis has led to the discovery of almost all of the major ALS genes (Chia et al., 2018). After an ALS gene is discovered and validated, the next step is to define how that mutation affects the protein encoded by the gene. In other words, does the mutation cause a loss of function or some sort of gain of function (including gain of a toxic function)? If a mutation causes disease by conferring a loss of function, then a gene therapy approach to restore the gene’s function would be warranted. However, if the mutation causes disease because of a gain of function, then a therapeutic approach to inhibit the mutant gene’s functionality or lower the levels of its gene products would be more appropriate. As evident, sorting this out will be critical for guiding therapy development for ALS.

So far, most of the mutations that have been identified in fALS are thought to cause disease via gain-of-function (GOF) mechanisms, meaning the mutation endows the protein with increased wild type (WT) functionality or altered function that is deleterious. Diseases caused by toxic GOF mutations often exhibit autosomal dominant or semi-dominant inheritance patterns, such that one mutant copy of the gene is sufficient to cause disease despite the presence of another WT gene and presumably half the amount of WT protein. Many observations have implied GOF mechanisms as key drivers in ALS, such as dominant inheritance patterns, generation of toxic protein aggregates, and altered properties of molecules that impair biological processes (Taylor et al., 2016). For instance, the first ALS gene that was described in 1993—superoxide dismutase 1 (SOD1)—is widely considered to cause ALS through a GOF mechanism, not only because of compelling genetic studies in animal models but also because of its dominant inheritance patterns and ability to cause protein aggregation (Bruijn et al., 1997; Deng et al., 1993; Gurney et al., 1994; Rosen et al., 1993; Wong et al., 1995).

Other known causes of ALS, such as mutations in RNA-binding protein encoding genes like TARDBP, FUS, and HNRNPA1, might work through GOF mechanisms as alluded to by aggregation or other altered physical properties. These mutant proteins can disrupt RNA function and metabolism, proteostasis, or axonal cytoskeletal dynamics through the formation of toxic protein aggregates (Taylor et al., 2016). However, the normal functions of these RNA-binding proteins are also disturbed by mutations. They are redistributed from their normal localization in the nucleus to the cytoplasm in disease, and nuclear depletion of these proteins results in widespread alterations in the regulation of their RNA-binding targets, consistent with a loss of function (Klim et al., 2019; Lagier-Tourenne et al., 2012; Ling et al., 2015; Melamed et al., 2019; Polymenidou et al., 2011).

Sometimes the ALS-causing mutations are not located in the coding portion of the gene. For example, the most common genetic cause of ALS is a mutation in the C9ORF72 gene (DeJesus-Hernandez et al., 2011; Renton et al., 2011). This mutation is a hexanucleotide repeat expansion that is located in C9ORF72’s intron or promoter. The expanded GGGGCC repeat gets transcribed into repeat-containing RNAs and then translated into aberrant proteins, both of which could cause disease via GOF toxicity (Gitler and Tsuiji, 2016) (Figure 2A). These two mechanisms seem unrelated to the normal function of the C9ORF72 gene and have motivated therapeutic strategies to target these mutant RNAs and/or proteins (Donnelly et al., 2013; Jiang et al., 2016; Lagier-Tourenne et al., 2013).

Figure 2: C9ORF72.

A. G₄C₂ repeat expansions in a non-coding region of C9ORF72 are hypothesized to cause ALS by at least one of three main mechanisms: 1) RNA foci-mediated toxicity, 2) dipeptide repeat protein (DPR)-mediated toxicity, and/or 3) reduced levels of C9ORF72 protein. B. C9ORF72 forms a complex with SMCR8 and WDR41 and plays a role in autophagy initiation via Rab1a-dependent recruitment of the ULK1 complex to phagophore assembly sites. C. The C9ORF72/SMCR8/WDR41 complex acts as a GDP/GTP exchange factor for Rab8a and Rab39b GTPases, which are involved in autophagy. D. TBK1 interacts with the C9ORF72/SMCR8/WDR41 complex and regulates autophagy by phosphorylating SMCR8. E. C9orf72 knockout mice and Tbk1 dendritic cell knockout mice have similar phenotypes affecting the immune system.

Loss-of-function (LOF) mutations, also called inactivating mutations, result in a gene product that is partially or wholly inactivated. LOF diseases are often caused by recessive mutations, meaning both alleles must carry the mutation in order for the mutant phenotype to occur. Consequently, effects from a mutant gene copy are normally masked in the presence of a normal copy in individuals heterozygous for the mutation. Dominantly inherited LOF diseases can result from haploinsufficiency, which means that having half as much of a functional gene product is not enough to prevent disease. Pathogenic mutations considered to be primarily LOF have been identified in ALS, such as mutations in OPTN and TBK1. As alluded to above, however, even genetic mutations that are considered to work predominantly through GOF mechanisms—such as SOD1 and C9ORF72—have also demonstrated LOF properties. Moreover, genetic variants can contribute to disease even if LOF is not itself sufficient to cause ALS, as exemplified by the immune phenotypes present in patients with mutations in C9ORF72 (Fredi et al., 2019; Ismail et al., 2013; Miller et al., 2016; Turner et al., 2013), or in mice harboring mutations in Tbk1 (Xu et al., 2018) or Optn (Ito et al., 2016).

It is the ultimate tautology that all genes have functions. But given that several therapies under development aim to reduce toxicity by knocking down expression of mutant ALS genes (sometimes by indiscriminately targeting both WT and mutant alleles), it is important to also consider the normal functions of these genes and the potential consequences of losing or reducing their levels. In this review, we critically discuss evidence for LOF mechanisms in ALS via known genetic mutations or loss of normal protein function. We highlight that considering both GOF and LOF mechanisms will be crucial in understanding the molecular underpinnings of ALS, as the field works to develop novel therapeutic strategies for this currently untreatable disorder.

1. SOD1

The first ALS gene was identified in 1993, when mutations causing fALS were linked to the gene encoding the Cu/Zn superoxide dismutase, SOD1 (Deng et al., 1993; Rosen et al., 1993). To date, 187 SOD1 variants have been described in patients with ALS (http://alsod.iop.kcl.ac.uk/), mutations which present in approximately 18.9% and 1.2% of fALS and sALS cases (SOD1-ALS), respectively (Zou et al., 2017), and typically display an autosomal dominant inheritance pattern (Andersen et al., 1998, 1996, 1995). SOD1 is ubiquitously expressed (Fridovich, 1995) and localizes to several cellular compartments including the cytoplasm, nucleus, mitochondrial intermembrane space, lysosomes, and peroxisomes (Chang et al., 1988; Crapo et al., 1992; Keller et al., 1991; Sturtz et al., 2001). The enzyme normally protects cells from toxic reactive oxygen species by catalyzing the dismutation of superoxide anion to oxygen and hydrogen peroxide (Fridovich, 1986).

While some ALS-causing SOD1 variants result in decreased SOD1 enzymatic activity in transformed human lymphocytes, others retain activity comparable to that of WT SOD1 (Borchelt et al., 1994; Hayward et al., 2002). A systematic analysis of many different ALS-causing SOD1 mutants found no correlation between SOD1 enzymatic activity and ALS onset or disease progression in humans (Cleveland et al., 1995), arguing against a causal role of SOD1 LOF in SOD1-ALS. There is now also ample evidence supporting GOF in SOD1-ALS. For example, mice expressing the human ALS-associated variants SOD1^G93A, SOD1^G37R, and SOD1^G85R exhibit overt ALS-like phenotypes including mitochondrial dysfunction, SOD1 aggregation, motor neuron death, and paralysis (Bruijn et al., 1997; Gurney et al., 1994; Wong et al., 1995). Furthermore, the presence or absence of endogenous mouse Sod1 has no effect on the decreased survival times of mice expressing the SOD1^G85R transgene (Bruijn et al., 1998), demonstrating that this phenotype is due to a toxic GOF, as opposed to a LOF, mechanism.

With the overwhelmingly compelling role of GOF mechanisms in SOD1-ALS, ASOs targeting SOD1 are being developed and tested as potential therapeutics for humans (Figure 1A) (McCampbell et al., 2018; Miller et al., 2013). But because this therapy will reduce SOD1 levels (the ASOs do not selectively target the mutant gene), we should consider the resulting LOF effects as they relate to ALS and other disorders. For instance, even though Sod1 knockout mice develop normally and do not exhibit overt ALS-like phenotypes such as motor neuron death and paralysis (Reaume et al., 1996), they do show noteworthy deficits of the motor system including decreased mitochondrial density in motor neurons (Fischer et al., 2011; Hayes et al., 2019), denervation and dysfunction at neuromuscular junctions (NMJs) (Fischer et al., 2012, 2011; Flood et al., 1999; Ivannikov and Van Remmen, 2015; Larkin et al., 2011; Shi et al., 2014), decreased peripheral nerve conduction velocity and reduced myelin thickness (Flood et al., 1999; Hamilton et al., 2013), increased motor neuron death after axotomy (Reaume et al., 1996), and secondary muscle pathology (Flood et al., 1999; Sakellariou et al., 2014). These animals also exhibit accelerated loss of skeletal muscle mass with age accompanied by decreased stride length, rotarod performance, and voluntary wheel running, with no change in spontaneous activity (Flood et al., 1999; Muller et al., 2006). Some of these animals also develop abnormal gait and tremors (Muller et al., 2006) that are indicative of neuromuscular impairments. Sod1 knockout mice also exhibit wide-ranging neuronal and non-neuronal abnormalities outside of the motor system (Saccon et al., 2013). Sod1 heterozygous mice—which leave one WT allele intact—do not exhibit death of motor neurons. But they display a wide range of phenotypes that may directly relate to ALS, such as increased susceptibility to glutamate toxicity (Schwartz et al., 1998) and axonal damage (Reaume et al., 1996), denervation (Fischer et al., 2012, 2011), and loss of select neuronal subtypes with aging (Keithley et al., 2005). Together these observations suggest that SOD1 LOF may not be causal but could modify aspects of ALS pathogenesis or instigate other unwanted effects (Saccon et al., 2013).

Additionally, two recent human genetics studies suggest that SOD1 homozygous LOF can cause debilitating phenotypes, even though it alone does not cause ALS (Andersen et al., 2019; Park et al., 2019). Human patients carrying homozygous truncating mutations in SOD1 (c.335dupG) did not produce any detectable SOD1 activity in erythrocytes (Andersen et al., 2019) or leukocytes (Park et al., 2019). Heterozygous carriers had approximately half the SOD1 activity of WT subjects. Both patients with homozygous SOD1 c.335dupG variants had similar symptoms including progressive loss of motor abilities, tetraspacticity, and hyperreflexia with symptom onset beginning in early (Miller et al., 2020; Mueller et al., 2020)childhood, while heterozygous carriers were unaffected (Andersen et al., 2019; Park et al., 2019). Additionally, a sibling of the patient described by Andersen et al. had similar symptoms to the patient and died at the age of six (Andersen et al., 2019).

Importantly, SOD1 LOF in humans is distinct from ALS; for ALS the key observation is that heterozygous carriers of the LOF mutation are unaffected, indicating that chronic partial LOF does not cause ALS or other diseases. On the other hand, complete SOD1 LOF does result in significant, debilitating abnormalities of the motor system, such that we must consider potential LOF effects when silencing SOD1 as a potential treatment for ALS, and perhaps target the mutant allele specifically. Current clinical efforts to target SOD1 result in incomplete suppression of SOD1 and preliminary clinical evidence indicates that this level of gene knockdown is safe for SOD1-ALS patients (Miller et al., 2020; Mueller et al., 2020, ClinicalTrials.gov identifier: NCT02623699).

2. C9ORF72

The most common known genetic cause underlying both ALS and FTD is a GGGGCC (G₄C₂) hexanucleotide repeat expansion in the chromosome 9 open reading frame 72 (C9ORF72) gene. The mutation accounts for approximately 34% of familial ALS and nearly a quarter of familial FTD (C9ALS/FTD) in populations of European ancestry, where it is most prevalent (van Blitterswijk et al., 2013; Zou et al., 2017). While healthy individuals can have from ~2 to 25 G₄C₂ repeats, C9ALS/FTD patients harbor hundreds to even thousands of these repeats (DeJesus-Hernandez et al., 2011; Renton et al., 2011; Rohrer et al., 2015). Depending on the transcript variant, the expansion is located in intron 1 or the promoter region of C9ORF72, bidirectionally transcribed (Kumar et al., 2017; Mizielinska et al., 2013; Niblock et al., 2016), and then translated via a non-canonical form of translation (repeat-associated non-AUG (RAN) translation) (Ash et al., 2013; Mori et al., 2013c; Zu et al., 2013) (Figure 2A). The C9ORF72 gene is expressed not only in the central nervous system, but also in the periphery, such as in lymphocytes, bone marrow, and spleen (McCauley and Baloh, 2019; Suzuki et al., 2013), and its function is still not fully understood.

On one hand, pathogenic mechanisms underlying the C9ORF72 repeat expansion have been attributed to GOF mechanisms such as (1) repeat-RNA-mediated toxicity in the form of RNA foci and (2) production of toxic homopolymeric dipeptide repeat proteins (DPRs) through RAN translation (Freibaum and Taylor, 2017; Kwon et al., 2014; Mizielinska et al., 2013; Mori et al., 2013c; Swinnen et al., 2018; Taylor et al., 2016; Wen et al., 2014) (Figure 2A). Briefly, sense and antisense RNA foci have been shown to sequester RNA-binding proteins (RBPs) in induced pluripotent stem cell (iPSC)-differentiated neurons (Conlon et al., 2016; Donnelly et al., 2013; Mori et al., 2013b; Sareen et al., 2013; Xu et al., 2013), which may lead to LOF in one or more of those RBPs. However, there has not been conclusive evidence directly proving that phenotypes caused by C9orf72 mutations are due to altered function in these RBPs. Additionally, DPRs can cause toxicity in neurons (Kwon et al., 2014; Lee et al., 2016; May et al., 2014), with arginine-rich GR and PR repeats being considered particularly toxic. Expression of these causes significant cell death in vitro (Tao et al., 2015; Wen et al., 2014), severe degeneration in Drosophila eyes (Freibaum et al., 2015; Lee et al., 2016; Mizielinska et al., 2014), and neurodegeneration and behavioral abnormalities in mouse models (Choi et al., 2019; Hao et al., 2019; Zhang et al., 2018; Y. J. Zhang et al., 2019). The suggestion of a GOF mechanism contributing to disease has led to the current development of ASOs against C9ORF72 for potential human therapy (Figure 1B) (Donnelly et al., 2013; Jiang et al., 2016; Lagier-Tourenne et al., 2013). Given that this would still result in reduced levels of C9ORF72 (even in the case of a mutant allele-specific targeting approach), we must also consider the potentially deleterious effects of C9ORF72 LOF. This brings us to another prominent—but not mutually exclusive—hypothesis for repeat expansion toxicity: (3) haploinsufficiency owing to the partial loss of normal C9ORF72 protein expression in repeat expansion carriers (Renton et al., 2011; DeJesus-Hernandez et al., 2011) (Figure 2A).

The possibility of a LOF mechanism is exemplified by immune phenotypes in heterozygous or homozygous C9orf72 knockout mice. For instance, mice from multiple studies in which C9orf72 was ablated from all tissues have shown signs of immune system dysregulation in the form of enlarged lymph nodes, splenomegaly, autoimmunity, increased cytokine release, and massive infiltration of histiocytes/macrophages and lymphocytes in multiple organs post-mortem (Figure 2E) (Atanasio et al., 2016; Burberry et al., 2020, 2016; Jiang et al., 2016; O’Rourke et al., 2016; Shao et al., 2019; Sudria-Lopez et al., 2016; Sullivan et al., 2016). While some studies have shown no effects of C9orf72 loss on survival (O’Rourke et al., 2016), several have demonstrated higher premature mortality upon C9orf72 loss (Burberry et al., 2020, 2016; Jiang et al., 2016), with further exacerbation of premature death when DPRs are expressed (Zhu et al., 2020). These phenotypes are perhaps not surprising given the high levels of expression of C9orf72 in myeloid cells (McCauley and Baloh, 2019). In fact, most recently, mice with myeloid cell-specific knockout of C9orf72 were shown to exhibit immune defects similar to global C9orf72-deficient mice via a STING (stimulator of interferon genes)-dependent mechanism, implying that the immune defects are in large part due to a cell-type specific loss of C9orf72’s normal ability to suppress STING-induced inflammation (McCauley et al., 2020). Interestingly, mice with a dendritic cell knockout of Tbk1—a gene in which LOF mutations can cause ALS (Freischmidt et al., 2015)—have similar phenotypes to C9orf72 knockout mice (Xiao et al., 2017) (Figure 2E), suggesting a potential connection between C9ORF72 and TBK1 in immune system regulation (McCauley and Baloh, 2019).

In addition, autoimmunity and premature mortality phenotypes of C9orf72 deficient mice can be mitigated upon transplantation of WT mouse bone marrow into mutant mice, and vice versa: transplanting mutant bone marrow into WT was sufficient to cause autoimmunity (Burberry et al., 2016). A similar effect was seen with fecal transplants, such that transplanting gut microflora from C9orf72 mutant mice grown in a protective environment (i.e. one that causes less inflammation), or reducing gut microbial burden with broad spectrum antibiotics, attenuated inflammatory phenotypes in C9orf72 knockout mice grown in a proinflammatory environment (Burberry et al., 2020). Despite accruing evidence for LOF, it is worth noting that these C9orf72-deficient mouse lines have failed to recapitulate ALS pathology or significant changes in motor neuron degeneration/function (e.g., using grip strength and accelerating rotarod tests) (Jiang et al., 2016; Sudria-Lopez et al., 2016), unless crossed with gain-of-toxic-function mouse models (Shao et al., 2019; Zhu et al., 2020). When C9orf72 ablation is neuron- and glia-specific, the LOF is insufficient to cause behavioral or pathologic phenotypes, although these mice show decreased body weight compared to WT mice (Koppers et al., 2015).

Immune deficiency associated with reduced levels of C9ORF72 has also been observed in humans, including precedence and/or concurrence of C9ALS/FTD with other autoimmune or immune-mediated conditions such as systemic lupus erythematosus (commonly referred to as lupus), myasthenia gravis, and multiple sclerosis (MS) (Fredi et al., 2019; Ismail et al., 2013; Miller et al., 2016; Turner et al., 2013). For instance, a study of 650 ALS patients from North England reported that the C9ORF72 expansion was present in 4 of 5 individuals with MS—an immune-mediated demyelinating disorder that results in progressive motor and other disabilities—that preceded ALS (MS-ALS), but not in those with MS alone (Ismail et al., 2013). Moreover, if MS preceded ALS, the C9-ALS patients were more rapidly progressing (Ismail et al., 2013). However, given the small number of MS-ALS patients described and other population-based studies that report MS-ALS cases presenting at a rate expected by chance, the possibility of misdiagnoses cannot be ruled out (van Doormaal et al., 2013).

On a subcellular level, decreased autophagy and enlarged and/or defective lysosomes have been described in C9orf72-null cell lines and mice (Ji et al., 2017; O’Rourke et al., 2016; Zhu et al., 2020). In human and mouse knockdown/knockout cell models and primary neurons from mice, C9ORF72 has also been shown to play a mechanistic role in early stages of autophagy signaling pathways by participating in heteromeric complexes (e.g. with SMCR8, WDR41, ATG13, ULK1) involved in autophagy initiation and/or regulation (Amick et al., 2016; Ho et al., 2019; Ji et al., 2017; Lan et al., 2019; Sellier et al., 2016; Webster et al., 2016; Yang et al., 2016) (Figure 2B, 2C). Moreover, TBK1 has been shown to interact with SMCR8, suggesting that they may belong to a common autophagy regulatory pathway (Sellier et al., 2016) (Figure 2D). In both human cells and C. elegans, C9ORF72 also regulates TFEB, a transcription factor and master regulator of autophagic and lysosomal genes (Ji et al., 2020). Defects in autophagosome formation have also been demonstrated in iPSC-derived motor neuron models from C9ALS/FTD patients, along with abnormal accumulation of glutamate receptors and altered axonal trafficking (Abo-Rady et al., 2020; Selvaraj et al., 2018; Shi et al., 2019, 2018). Given that formation of cytoplasmic protein aggregates due to defective protein quality control is a key pathologic hallmark in ALS and numerous other neurodegenerative disorders, these impairments in protein degradation and clearance seem particularly relevant in driving disease (Ramesh and Pandey, 2017).

Despite supportive evidence for LOF mechanisms contributing to C9ALS, human ALS patients with LOF mutations in C9ORF72 other than the G₄C₂ hexanucleotide repeats have not yet been identified. Moreover, homozygous carriers of C9ORF72 mutations seem to exhibit clinical phenotypes that are no worse than heterozygous individuals (Fratta et al., 2013). Together, the above observations suggest that multiple mechanisms are at play: one could imagine that the reduced levels of normal C9ORF72 and subsequent immune deficiency/subcellular defects increase a neuron’s susceptibility to GOF effects of DPRs or RNA foci. We suspect that a multi-targeted therapeutic approach that not only targets and mitigates G₄C₂ repeat-containing RNAs but also increases C9ORF72 levels will likely be necessary to treat C9ALS/FTD patients (Figure 5A). Alternatively (or concomitantly), a STING inhibitor to suppress the increased interferon response seen in whole blood and brain tissue from C9ALS/FTD patients could perhaps be used to mitigate immune effects caused by lowered C9ORF72 levels (McCauley et al., 2020). We must also consider environmental contributions, as C9orf72 knockout mice with identical genetic backgrounds have exhibited strikingly different survival depending on their environment (Burberry et al., 2020), potentially raising the importance of boosting reduced C9ORF72 levels. Given that much of our current understanding of C9ALS/FTD pathogenesis relies on cultured cells and animal models, it will be critical for future studies to further investigate disease-relevant models and tissues derived from patients to more fully understand the underlying pathogenic mechanisms. This will allow the field to not only gain a more relevant perspective on the mechanistic contributions of C9ORF72 deficiency to disease, but also to better address the gap in knowledge pertaining to selective regional and cell-type-specific vulnerabilities in C9ALS/FTD.

Figure 5: Next-Generation Therapies: Targeting LOF and GOF Effects.

A. G₄C₂ repeat expansions in C9ORF72 cause production of sense and antisense RNA foci and DPRs, as well as reduced levels of C9ORF72, all of which may contribute to disease. A combination therapy that targets G₄C₂ repeat-containing RNAs and increases levels of C9ORF72 could mitigate both the LOF and toxic GOF effects resulting from this mutation. B. TDP-43 is depleted from the nucleus and forms cytoplasmic inclusions in most ALS cases. Nuclear depletion of TDP-43 alters mRNA metabolism and results in decreased levels of STMN2, which may contribute to ALS pathogenesis. A combination therapy that targets toxic cytoplasmic TDP-43 and restores STMN2 levels through gene therapy or an ASO that blocks cryptic exon inclusion could be more beneficial to patients than either treatment alone.

3. TDP-43

Transactive response (TAR) DNA-binding protein 43 (TDP-43), encoded by the TARDBP gene, is a DNA/RNA-binding protein normally localized in the nucleus (Polymenidou et al., 2011; Tollervey et al., 2011) (Figure 3A). TDP-43 regulates various steps of RNA metabolism, including mRNA splicing, RNA transport, translation, and microRNA biogenesis (Alami et al., 2014; Buratti et al., 2001; Deshaies et al., 2018; Kawahara and Mieda-Sato, 2012; McDonald et al., 2011; Neelagandan et al., 2019). It is also the most widespread and characterized pathologic protein in the ALS/FTD spectrum; in disease states, TDP-43 is depleted from the nucleus and found as hyperphosphorylated, aggregated cytoplasmic inclusions in ~97% of ALS and ~50% of FTD patients (Dormann and Haass, 2011; Giordana et al., 2010) (Figure 3B). Although causal mutations in TARDBP are rare and account for only ~4% of fALS and <1% of sALS (Zou et al., 2017), the nearly ubiquitous presence of pathologic TDP-43 inclusions and nuclear TDP-43 depletion in ALS indicates that ALS-causing mechanisms eventually converge on this central pathological axis.

Figure 3: TDP-43.

A. Under physiological conditions, TDP-43 predominantly localizes to the nucleus and plays a role in RNA metabolism. B. In most cases of ALS, TDP-43 is depleted from the nucleus and forms cytoplasmic aggregates. TDP-43 pathology may contribute to ALS through 1) a loss of normal function in the nucleus and/or 2) a toxic GOF via formation of cytoplasmic aggregates. C. Nuclear TDP-43 binds to STMN2 pre-mRNA transcripts and suppresses inclusion of the cryptic exon 2a. This mRNA isoform encodes functional STMN2, a microtubule-associated protein involved in neurite outgrowth and repair. D. Reduced levels of nuclear TDP-43 cause inclusion of cryptic exon 2a and premature polyadenylation of STMN2 mRNA, resulting in decreased levels of STMN2 and impaired axonal regeneration in vitro.

On one hand, the presence of pathologic inclusions and their correlation with affected brain regions suggests a toxic GOF mechanism (Arai et al., 2006; Neumann et al., 2006). On the other hand, the clearance of TDP-43 from its usual nuclear localization suggests a loss of TDP-43’s canonical nuclear function. Given the ubiquity of both phenotypes, naturally many have attempted to directly alter TDP-43 mRNA or protein levels to investigate the resulting effects—yet, evidence suggests that directly targeting its protein or mRNA levels for therapy is probably not favorable.

First, TDP-43 is not only essential for early embryonic development in mice, but is also required for normal neuronal function—such as the ability to control muscle movement—in mice, Drosophila, and zebrafish (Feiguin et al., 2009; Kraemer et al., 2010; Yang et al., 2014). Conditional TDP-43 knockout mice crossed to various tissue-specific Cre lines similarly display diverse cellular and behavioral phenotypes ranging from electrophysiological abnormalities to motor movement deficits (Wu et al., 2019), and knocking down TDP-43 in cultured mammalian cells leads to reduced dendritic complexity (Herzog et al., 2017). Moreover, TDP-43 autoregulates its own synthesis by binding to an alternatively spliced intron in the 3’ UTR of its pre-mRNA, rendering modification of its levels difficult (Ayala et al., 2010; Fukushima et al., 2019; Polymenidou et al., 2011). Such perplexities have led many to investigate its interactors, regulators, and binding targets.

As a DNA/RNA-binding protein, TDP-43 has been shown to interact with RNA transcripts of more than 6,000 genes in mice, where it binds preferentially to UG-rich sequences that are located in introns, non-coding RNAs, and intergenic regions (Polymenidou et al., 2011; Tollervey et al., 2011). Thus, it is not surprising that knocking down TDP-43 levels in adult mouse brains using ASOs leads to down- and up-regulation of some ~240 and ~360 genes, respectively—many of which encode proteins involved in synaptic activity and neuronal development—and alteration of nearly 1,000 splicing events (Polymenidou et al., 2011). One method by which TDP-43 affects RNA splicing events is by suppressing the inclusion of cryptic exons (Ling et al., 2015), which are non-conserved exons included in some but not all transcripts (Zavolan et al., 2003). Cryptic exons often cause frameshift mutations or introduce premature stop codons, alternative transcriptional start sites, or alternative polyadenylation sites, which can lead to nonsense-mediated decay (NMD) or disrupted translation and, consequently, the downregulation of critical proteins (Tan et al., 2016). In Drosophila and mouse motor neuron models of TDP-43 LOF, expressing an engineered chimeric protein that mimics TDP-43’s role as a splicing regulator not only decreases aberrant splicing, but also rescues other ALS-related phenotypes such as motor deficits, axon degeneration, and motor neuron loss caused by TDP-43 reduction (Donde et al., 2019). Therefore, exploring proteins downregulated by TDP-43 due to splicing defects in neurons may reveal critical pathways downstream of TDP-43.

One particular splicing target of TDP-43, Stathmin-2 (STMN2), seems likely to play a key role in ALS. Studies in human SH-SY5Y neuroblastoma cells revealed that knocking down TDP-43 levels leads to changes in expression of hundreds of genes, among which STMN2, a regulator of microtubule stability, is one of the most downregulated genes (Klim et al., 2019; Melamed et al., 2019). ALS patient fibroblast-derived neurons expressing an ALS-causing TDP-43 mutation display reduced STMN2 expression (Klim et al., 2019; Melamed et al., 2019). The STMN2 gene harbors a cryptic exon (exon 2a) that is normally not included in the mature STMN2 mRNA (Figure 3C) (Klim et al., 2019; Melamed et al., 2019). TDP-43 functions to repress inclusion of this cryptic exon but when TDP-43 is lost or its function is impaired, this exon gets incorporated into the mature mRNA (Figure 3D). Unfortunately, this exon harbors a stop codon and a poly-adenylation signal, which results in truncated STMN2 mRNA and protein (Melamed et al., 2019). Aberrant splicing and reduced STMN2 protein levels seem to be a major feature of ALS (except those with SOD1 mutations) (Klim et al., 2019; Melamed et al., 2019). Importantly, upregulation of STMN2 is able to rescue axonal regeneration defects caused by TDP-43 depletion in human iPSC-derived neurons, supporting the idea that TDP-43-dependent neurodegeneration acts, at least in part, via a LOF mechanism (Klim et al., 2019; Melamed et al., 2019). While STMN2 is a potentially attractive therapeutic target (e.g., by preventing its mis-splicing or by upregulating STMN2 levels) (Figure 5B), further studies investigating alterations in STMN2 levels in larger patient populations—whether that be through transcriptomic data or post-mortem brain specimens—will be crucial. Notably, the cryptic exon in STMN2 is not conserved to the mouse gene (Ling et al., 2015), requiring studies in human cell lines and/or the generation of humanized mouse models.

In addition to acting as a regulator of alternative splicing, TDP-43 can also regulate transposable elements, which are mobile DNA sequences that can hop around to different positions within the genome (Krug et al., 2017; Li et al., 2012). These include retrotransposons, transposable elements that copy themselves via an RNA intermediate. Several compelling studies have provided evidence that TDP-43 binds to and represses retrotransposon transcripts (Krug et al., 2017; Li et al., 2012); suppression of the Drosophila TDP-43 homolog leads to activation of certain transposable elements, whereas genetic rescue with TDP-43 expression suppresses this activation, indicating that the endogenous function of TDP-43 is required for transposable element repression (Romano et al., 2020). Moreover, nuclear loss of TDP-43 has been associated with increased chromatin accessibility to transposable elements in neurons derived from FTD-ALS patients (Liu et al., 2019). Induction of retrotransposon expression in normal human cells promotes a senescence-like phenotype (Belancio et al., 2010), and re-insertion of active retrotransposons can alter the genome of individual cells and affect normal cellular functions/lead to cell death (Savage et al., 2019). Thus, the expression of retrotransposons is tightly repressed in most somatic tissues to prevent DNA damage and ensure genomic integrity (Goodier, 2016). Taken together, these results suggest that dysregulation of transposable elements due to TDP-43 LOF may contribute to ALS pathogenesis.

In sum, despite evidence for TDP-43 GOF mechanisms in ALS, accruing evidence suggests an important, perhaps primary role of TDP-43 LOF in ALS. Further investigation of genes dysregulated in the absence of TDP-43 are required to shed light on the role of TDP-43 in regulating biological processes, which could lead to therapies that compensate for the loss of nuclear TDP-43 (Figure 5B). Given that TDP-43 is not itself a favorable target, it will be important that future studies aim to more deeply characterize pivotal RNAs and proteins that are regulated by TDP-43 as well as the potential role of TDP-43 as a transposable element regulator. These and other underexplored nuclear functions of TDP-43 may open up critical new avenues for elucidating disease pathogenesis and generating new therapeutic opportunities.

4. FUS/TLS

Mutations in FUS/TLS (fused in sarcoma/translocated in liposarcoma) can cause ALS and rare forms of FTD (Broustal et al., 2010; Kwiatkowski et al., 2009; Van Langenhove et al., 2010; Vance et al., 2009). To date, more than 50 different FUS mutations have been described in patients with ALS (Deng et al., 2014), many of which disrupt the nuclear localization signal and result in mislocalization of FUS to the cytoplasm (Deng et al., 2014; Dormann et al., 2010; Ito et al., 2011; Lagier-Tourenne et al., 2010). FUS mutations are present in ~4% of fALS patients and <1% of sALS patients (Deng et al., 2014; Zou et al., 2017) and typically display an autosomal dominant inheritance pattern. In post-mortem brains of FUS-ALS patients, FUS-positive inclusions are present in the cytoplasm and, less frequently, in the nucleus of neurons and glia (Hewitt et al., 2010; Kwiatkowski et al., 2009; Neumann et al., 2009; Vance et al., 2009).

The ubiquitously expressed DNA/RNA-binding protein FUS localizes predominantly to the nucleus under physiological conditions (Zinszner et al., 1997), and is normally involved in DNA repair (Wang et al., 2008) and several aspects of RNA metabolism including transcription (Masuda et al., 2015; Tan and Manley, 2010), alternative splicing (Hoell et al., 2011; Ishigaki et al., 2012; Lagier-Tourenne et al., 2012; Rogelj et al., 2012), mRNA transport (Fujii et al., 2005), mRNA stability (Kapeli et al., 2016; Udagawa et al., 2015; Yokoi et al., 2017), and microRNA biogenesis (Gregory et al., 2004; Morlando et al., 2012). The protein also localizes to the neuromuscular junction (So et al., 2018) as well as presynaptic terminals (Schoen et al., 2015) and postsynaptic dendrites in neurons (Aoki et al., 2012; Belly et al., 2005; Fujii et al., 2005; Yasuda et al., 2013), where it plays a role in local translation (López-Erauskin et al., 2018; Yasuda et al., 2013) and synaptic structure/function (Fujii et al., 2005; Fujii and Takumi, 2005).

FUS pathogenesis in ALS, similar to TDP-43 pathogenesis, is thought to arise from 1) a toxic GOF due to FUS aggregation and/or 2) a LOF resulting from cytoplasmic mislocalization of FUS and subsequent loss of nuclear function. The GOF hypothesis has been supported by studies demonstrating that overexpression of WT human FUS causes progressive motor neuron degeneration in mice (Ling et al., 2019; López-Erauskin et al., 2018; Mitchell et al., 2013), and that mice expressing exogenous FUS with ALS-associated mutations (Sharma et al., 2016) or without a nuclear localization signal (ΔNLS-FUS) exhibit ALS-like phenotypes such as motor neuron degeneration (Shiihashi et al., 2016). However, LOF of the FUS homolog caz in Drosophila causes reduced lifespan (Wang et al., 2011), neuromuscular junction disruption (Sasayama et al., 2012; Wang et al., 2011; Xia et al., 2012), and impaired locomotive ability (Sasayama et al., 2012; Xia et al., 2012). In zebrafish, fus knockdown similarly results in impaired motor activity (Armstrong and Drapeau, 2013; Kabashi et al., 2011), shortened axonal projections from motor neurons (Kabashi et al., 2011), and neuromuscular junction dysfunction (Armstrong and Drapeau, 2013).

Yet it seems that LOF alone is insufficient to cause FUS-ALS. In mice, loss of one Fus allele (Fus^+/−) does not cause an overt phenotype (Hicks et al., 2000; Kuroda et al., 2000), whereas global Fus knockout causes perinatal death in inbred C57BL/6 mice (Hicks et al., 2000). In outbred mice, Fus knockout causes male sterility and reduced female fertility (Kuroda et al., 2000) as well as hyperactivity, reduced anxiety-like behavior, and hippocampal vacuolation, but no overt ALS-like phenotype (Kino et al., 2015). Additionally, conditional knockout of Fus in mice postnatally does not lead to motor neuron degeneration (Sharma et al., 2016). Together, these results suggest that FUS LOF alone does not cause FUS-ALS but may cause other deleterious phenotypes.

In brains of mice and humans, FUS interacts with the RNA transcripts of over 5,500 genes (Lagier-Tourenne et al., 2012), where it binds primarily through a GUGGU-binding sequence or AU-rich stem loops in a saw-tooth pattern (Hoell et al., 2011; Ishigaki et al., 2012; Lagier-Tourenne et al., 2012; Rogelj et al., 2012). Knockdown of Fus with ASOs in adult mouse brains leads to altered expression and splicing of 610 and 374 mRNAs, respectively (Lagier-Tourenne et al., 2012), including Fus itself, resulting in autoregulation through NMD (Zhou et al., 2013). The subsets of RNAs with altered expression after FUS or TDP-43 knockdown are largely non-overlapping; however, both FUS and TDP-43 knockdown result in decreased expression of mRNAs that contain long introns, many of which encode proteins important for the function and activity of synapses (Lagier-Tourenne et al., 2012).

As suggested by the mRNA targets found in the studies described above, Fus knockdown results in synaptic dysfunction in vitro and in vivo. For example, Fus knockdown causes enlarged growth cones (Orozco et al., 2012), decreased neurite length (Ishigaki et al., 2017; Orozco et al., 2012), abnormal dendritic spines (Fujii et al., 2005; Yokoi et al., 2017), and decreased synaptic transmission (Udagawa et al., 2015) in rodent primary neurons. These phenotypes are mediated in part by downregulation of the AMPA receptor subunit GluA1 (Udagawa et al., 2015) and the dendritic spine maturation protein SynGAP (Yokoi et al., 2017) as well as upregulation of tauopathy-associated 4-repeat tau (4R-T) due to inclusion of Mapt exon 10 (Ishigaki et al., 2017; Orozco et al., 2012).

In vivo, knockdown of Fus in mouse hippocampus results in accumulation of phosphorylated tau, hippocampal neuron loss, and decreased adult neurogenesis (Ishigaki et al., 2017). Moreover, a novel FUS variant (p.Q140H) was identified in two sisters who exhibited a familial form of FTD characterized by astrocyte-predominant tauopathy and altered tau splicing (Ferrer et al., 2015), suggesting a potential link between FUS mutations and tau pathology in human FTD (Ishigaki et al., 2017). Knockdown of Fus in mouse hippocampus also causes FTD-like behavioral phenotypes such as social interaction defects, hyperactivity, and disinhibition (Ishigaki et al., 2017; Udagawa et al., 2015), which can be partially rescued by 4R-T suppression (Ishigaki et al., 2017) or increased GluA1 expression (Udagawa et al., 2015). Similarly, forebrain-specific knockout of Fus causes FTD-like behaviors in mice, which are partially rescued by increased SynGAP α2 expression (Yokoi et al., 2017).

In humans, the FUS nonsense mutation (p.Q290X) is associated with essential tremor, a neurodegenerative disorder characterized by postural or motion tremor (Merner et al., 2012). In lymphoblastoid cells derived from essential tremor-affected individuals, FUS mRNA is largely degraded by NMD, suggesting that complete FUS LOF may cause the clinical phenotype in humans (Merner et al., 2012). Importantly, a 26-year-old ALS patient was recently given a personalized ASO treatment targeting her FUS P525L mutation, which causes an aggressive form of ALS with juvenile onset (Figure 1C) (Arnold, 2019). The patient began treatment after the disease had already begun to cause respiratory problems and passed away less than a year after being given her first injection (Carrie Arnold, 2019). Thus, it is crucial to examine the effects of decreased FUS expression, as potential therapeutics targeting FUS are being further explored and used in the clinic.

5. OPTN and TBK1

Three ALS-causing mutations in the optineurin gene (OPTN), also known as FIP-2, were first discovered by homozygosity mapping in six Japanese individuals from consanguineous marriages (Maruyama et al., 2010). These mutations were 1) a homozygous exon 5 deletion causing a frameshift and premature stop codon, 2) a homozygous nonsense mutation (p.Q398X) that generates a truncated protein, and 3) a heterozygous missense mutation (p.E478G) that causes reduced expression likely due to NMD, all of which suggest LOF (Maruyama et al., 2010). Other heterozygous nonsense and missense mutations in OPTN have subsequently been identified in sALS and fALS patients (Tümer et al., 2012; van Blitterswijk et al., 2012), suggesting that OPTN haploinsufficiency contributes to ALS (Del Bo et al., 2011; Li et al., 2015; van Blitterswijk et al., 2012). Mutations in OPTN seem to be rare in sALS, corresponding to 0.4% of cases in Japanese/Chinese and 0.3% of European populations, respectively, while their incidence in fALS is about 4% and 1.5% in the same populations (Fifita et al., 2017).

Optineurin (OPTN) is ubiquitously expressed in most tissues and harbors a ubiquitin-binding domain that is required for recognition of ubiquitinated proteins and for its role as an autophagy adaptor (Markovinovic et al., 2017; Slowicka et al., 2016; Toth and Atkin, 2018). Consistent with a LOF mechanism in ALS, knockout of optn leads to alteration in axonal trafficking dynamics and increased cell death in zebrafish embryos (Paulus and Link, 2014), and knockdown of Optn in N2a and retinal ganglion cells (RGCs) from mice results in enhanced cell death (Akizuki et al., 2013; Sippl et al., 2011). Additionally, studies of the E478G mutant in human cancer cell lines and mouse embryonic fibroblasts have reported impairments in pathogen clearance (xenophagy) (Ivannikov and Van Remmen, 2015; Noad et al., 2017; van Wijk et al., 2017), protein aggregate clearance (aggrephagy) (Korac et al., 2013; Shen et al., 2015) and degradation of damaged mitochondria (mitophagy) (Heo et al., 2015; Lazarou et al., 2015; Moore and Holzbaur, 2016; Richter et al., 2016; Wong and Holzbaur, 2014; Zachari et al., 2019), further supporting a loss of normal OPTN function in autophagy. Its role in autophagy is also regulated by another known ALS gene, TANK-binding kinase 1 (TBK1), which via phosphorylation can enhance clearance of autophagic cargos (Moore and Holzbaur, 2016; Padman et al., 2019; Richter et al., 2016; Wild et al., 2011) by increasing the affinity of OPTN for ubiquitinated substrates and its translocation to damaged mitochondria (Heo et al., 2015; Li et al., 2018; Richter et al., 2016) (Figure 4A). In addition to overlaps in autophagy-related pathways, OPTN and TBK1 share roles in the innate immune system (Figure 4B).

Figure 4: TBK1 and OPTN.

A. TBK1 phosphorylates several autophagy receptors including OPTN. Phosphorylation of OPTN by TBK1 increases its interactions with ATG8 proteins/LC3 and ubiquitinated cargo. B. TBK1 plays a role in inflammatory pathways by acting downstream of proteins that sense bacterial lipopolysaccharides and viral RNA/DNA. OPTN promotes TBK1 activation in response to viral RNA. These pathways result in expression of proinflammatory cytokines and type I interferons. C. Several proteins, including TBK1 and OPTN, have been shown to inhibit RIPK1-dependent inflammation and cell death. D. Tbk1^+/− mice with reduced myeloid Tak1 expression and mice with oligodendrocyte or myeloid-specific knockout of Optn exhibit neuroinflammatory and ALS/FTD-like phenotypes.

TBK1, also known as T2K or NAK, is a serine/threonine kinase that plays a crucial role in the innate immune system by phosphorylating IRF3 and inducing the production of type-I interferon (IFN) and proinflammatory cytokines after viral infection (Figure 4B) (Cai et al., 2014; Fitzgerald et al., 2003; Ishii et al., 2006; Liu et al., 2015; Tanaka and Chen, 2012; C. Zhang et al., 2019). It also plays an important role in regulating autophagy by interacting with and phosphorylating autophagy adaptors, such as NDP52, p62, and OPTN (Figure 4A) as well as the C9ORF72/SMCR8/WDR41 autophagy complex (Ahmad et al., 2016; Heo et al., 2015; Korac et al., 2013; Moore and Holzbaur, 2016; Noad et al., 2017; Richter et al., 2016; Thurston et al., 2016; van Wijk et al., 2017; Vargas et al., 2019; Wild et al., 2011; Yang et al., 2016) (Figure 2D). TBK1 was identified as an ALS gene in two whole-exome sequencing (WES) studies from North American and European populations (Cirulli et al., 2015; Freischmidt et al., 2015), revealing eight mutations. Introducing these mutant alleles in human cell lines confirmed loss of TBK1 expression for most mutants, supporting a LOF mechanism in ALS (Brenner et al., 2019; Freischmidt et al., 2015). Over 80 ALS-associated mutations in TBK1 have been discovered to date (de Majo et al., 2018; Freischmidt et al., 2015).

In mice, Tbk1 LOF alone does not seem to cause ALS-like phenotypes, as heterozygous Tbk1^+/− animals do not exhibit signs of motor neuron degeneration or impaired autophagy at 200 days of age (Brenner et al., 2019). However, Tbk1^+/− mice with reduced expression of Tak1 in myeloid cells exhibit ALS/FTD-like phenotypes including TDP-43 aggregation, neuroinflammation, axonal degeneration, and neuron loss (Xu et al., 2018). Additionally, Tbk1^+/− mice overexpressing the ALS-causing SOD1 G93A mutation show accelerated clinical onset, muscular denervation, and impaired autophagy (Brenner et al., 2019). Point mutations that decrease Tbk1’s kinase activity also accelerate disease onset in SOD1^G39A mice but surprisingly extend lifespan (Gerbino et al., 2020). Together, these findings indicate that TBK1 LOF mutations can contribute to and modify ALS pathogenesis (Brenner et al., 2019; Gerbino et al., 2020; Xu et al., 2018).

The relationship between the two independent LOF ALS genes TBK1 and OPTN is one that warrants closer scrutiny. For instance, expression of the ALS-causing p.690–713del TBK1 mutant in human HEK293T cells abolishes its interaction with OPTN (Freischmidt et al., 2015). Moreover, this mutant TBK1 is still able to phosphorylate and interact with IRF3, suggesting that impairment of the interaction between TBK1 and OPTN may be sufficient to cause ALS (Freischmidt et al., 2015). This possibility is further strengthened by the finding that the p.E696K missense mutation in TBK1 abolishes TBK1’s interaction with OPTN in vitro and in vivo despite still being able to interact with other protein partners (Freischmidt et al., 2015; Li et al., 2018; Moore and Holzbaur, 2016; Richter et al., 2016), implying that the disrupted interaction with OPTN is specific. Since these initial studies, other missense mutations in TBK1 have been identified in ALS patients that also reduce levels of activated TBK1 and abolish its interactions with both IRF3 and OPTN (de Majo et al., 2018).

The overlapping functions of TBK1 and OPTN in autophagy and the innate immune system also suggests an impairment of common functions in the OPTN/TBK1 axis, such as RIPK1-dependent cell death, in the development of ALS. For instance, conditional knockout of Optn in microglia and oligodendrocytes from mice leads to axonal pathology, elevated production of proinflammatory cytokines, and accumulation of apoptosis/necroptosis markers that include phosphorylated RIPK1, which can be rescued by inhibition of RIPK1 or a kinase-dead mutant (Figure 4D) (Ito et al., 2016). Similarly, homozygous knockout of Tbk1 in mice is embryonic lethal owing to liver failure, which can be rescued by inactivating Ripk1 (Xu et al., 2018). Tbk1 has also been shown to inhibit Ripk1 together with Tak1 in mice; this is noteworthy given that Tak1 levels are upregulated in Tbk1 knockout cells and decline with aging in humans, more so in ALS patients compared to healthy individuals of the same age (Figure 4C) (Xu et al., 2018). Additionally, RIPK1 levels are elevated in the spinal cords of symptomatic SOD1^G93A mice compared to age-matched controls (Dermentzaki et al., 2019). Together, this evidence suggests that perhaps OPTN and TBK1 LOF mechanisms in ALS may center on their role in the innate immune system, and that RIPK1-dependent signaling may play a key role in ALS pathogenesis.

In direct contrast to dysregulation of the OPTN/TBK1 axis via LOF mechanisms in ALS, GOF missense mutations in OPTN (Rezaie et al., 2002) and copy number variations in TBK1 (Fingert et al., 2011; Kawase et al., 2012; Ritch et al., 2014) have been identified in patients with some forms of glaucoma (Minegishi et al., 2016). In particular, the p.M98K missense mutation in OPTN induces RGC death and enhanced autophagosome formation (Sirohi et al., 2013). Moreover, the p.E50K mutation in OPTN leads to enhanced interaction of OPTN with TBK1 (Li et al., 2016; Minegishi et al., 2013), and its overexpression leads to RGC death (Chalasani et al., 2007) and accumulation of insoluble OPTN. These phenotypes of RGC death and OPTN accumulation can be reversed by treatment with the TBK1 inhibitor BX795 (Minegishi et al., 2013), while amlexanox, another TBK1 inhibitor, corrects the abnormal intracellular localization of OPTN E50K (Minegishi et al., 2016), suggesting potential therapeutic routes through inhibition of TBK1 in glaucoma. However, it is important to emphasize that complete or partial LOF of TBK1 affects cell survival and function differently depending on the cell type, indicating that TBK1-targeting therapeutic strategies should not be indiscriminately applied to all glaucoma patients.

6. GRN

Progranulin (PGRN), encoded by the GRN gene, is a secreted glycoprotein that regulates inflammatory and neuroprotective responses (Kao et al., 2017; Karamysheva et al., 2019). Since heterozygous null mutations in GRN were first described as causative for tau-negative FTD with TDP-43 pathology (Baker et al., 2006; Cruts et al., 2006), missense mutations have also been found in ALS patients with or without FTD across different cohorts (Schymick et al., 2006; Sleegers et al., 2008). Some of these mutations predict a LOF mechanism, although pathogenicity of these variants has not been investigated in much detail. Grn knockout mice show exacerbated neuronal damage after traumatic brain injury and lysosomal dysfunction (Tanaka et al., 2013), while knockdown of grn in zebrafish leads to impaired motor function (Chitramuthu et al., 2017). Knockdown of tardbp and fus in zebrafish causes similar motor neuron degeneration phenotypes which are rescued by the expression of human GRN (Chitramuthu et al., 2017). Overexpression of human GRN also seems to extend the survival and slow disease progression in a mouse model of TDP-43 proteinopathy (Beel et al., 2018), suggesting that boosting levels of PGRN could be beneficial for ameliorating ALS/FTD pathogenesis.

7. NEK1 and C21ORF2

NEK1 was first identified as an ALS gene through whole-exome sequencing of more than 2,000 ALS patients (Brenner et al., 2016; Cirulli et al., 2015; Gratten et al., 2017; Kenna et al., 2016; Nguyen et al., 2018). Rare variants in NEK1 are observed in 3–4% of ALS patients in European populations, among which 1% are confirmed heterozygous LOF variants (Brenner et al., 2016; Cirulli et al., 2015; Kenna et al., 2016; Nguyen et al., 2018).

As a member of the NIMA-related serine/threonine kinase protein family (NEK), NEK1 participates in controlling cell cycle, DNA damage repair, and ciliogenesis (Chen et al., 2011; Higelin et al., 2018; Pelegrini et al., 2010; Shalom et al., 2008). In fact, the first mechanistic characterization of an ALS-linked NEK1 mutation (p.R812X; heterozygous) in patient-derived motor neurons showed that upon DNA damage by irradiation, there was increased DNA damage as well as a compromised DNA damage response (Higelin et al., 2018). Moreover, knocking down Nek1 expression in 661W photoreceptor cells increases cell death by RIPK1-dependent apoptosis (Amin et al., 2018), supporting a LOF mechanism.

NEK1 has also been described to interact with the ALS-associated proteins ALS2, VAPB and C21orf2, sharing cellular functions with the latter in DNA damage response and ciliogenesis (Cirulli et al., 2015; Fang et al., 2015; Wheway et al., 2015). While the biological role of C21orf2 remains poorly characterized, it is known to participate in formation and maintenance of primary cilia (Khan et al., 2015; L. Wang et al., 2017; Wheway et al., 2015) and to directly interact with NEK1. Interestingly, mutations in both NEK1 and C21ORF2 are associated with Axial Spondylometaphyseal Dysplasia, an autosomal recessive disease characterized by dysplasia of axial skeleton and retinal dystrophy (Wang et al., 2016; Z. Wang et al., 2017). Given that C21ORF2, NEK1, and other novel ALS genes share common cellular functions/disease pathways, therapies targeting these pathways could be efficacious for a broad group of patients despite variations in genotype. However, it will be necessary to first better characterize the functions and mechanisms associated with these genes, their effects in animal models, and their roles in disease pathogenesis in ALS/FTD patients.

CURRENT CLINICAL EFFORTS

The success of targeting SOD1 mRNA for degradation in animal models of SOD1-ALS (McCampbell et al., 2018) led to the development and recent conclusion of a phase 1–2 clinical trial of the ASO tofersen (BIIB067), which binds to both mutant and WT SOD1 mRNA (Miller et al., 2020, 2013). 48 of the 50 enrolled SOD1-ALS patients received all five planned doses, of which those that received 100 mg of tofersen (the highest dose given in the trial) showed a ~36% decrease of SOD1 concentrations in the cerebrospinal fluid (CSF) compared to a ~3% decrease in the placebo group (Miller et al., 2020). Due to intrathecal injection of ASOs for wide distribution throughout the central nervous system (Schoch and Miller, 2017), most patients exhibited lumbar puncture-related adverse events (Miller et al., 2020). Although this trial was not powered to test clinical or biological efficacy beyond SOD1 levels in CSF, some exploratory outcomes showed that disease progression was slowed in the 100 mg dose group, with a particularly dramatic effect for the four fast-progressing patients (Miller et al., 2020). In this subgroup, the ALSFRS-R (ALS functional rating scale-revised) score at day 85 changed from baseline by 0.84 points in the treated group compared to −16.73 points in the placebo group. The treated group also showed better performance in two other clinical measures—slow vital capacity and hand-held dynamometry—compared to the placebo group (Miller et al., 2020). While this study was small and requires continued follow-up on the treated patients, the results so far are inspiring and augur well for future ASO-based therapeutic strategies for ALS. Further assessments of the safety and efficacy of tofersen are now being investigated in a phase 3 clinical trial in SOD1-ALS patients (ClinicalTrials.gov identifier: NCT02623699).

Non-ASO-based therapies are also in the works (Hester et al., 2009) . For example, an adeno-associated virus (AAV)-based gene therapy has been intrathecally administered to two SOD1-ALS patients to lower SOD1 levels using microRNAs (Mueller et al., 2020). One patient had lower SOD1 levels in spinal cord at autopsy and transiently lower SOD1 levels in CSF during life, while the other patient has shown no changes in SOD1 levels (Mueller et al., 2020). Moreover, the first patient developed meningoradiculitis (extreme nerve pain) after infusion, while the second patient did not have this complication perhaps due to pre-treatment with immuno-suppressive drugs (Mueller et al., 2020). Further investigations of AAV-based therapies and novel delivery routes (Bravo-Hernandez et al., 2020) will be necessary to assess their efficacy and side effects, given the small sample size to date. Regardless of whether the therapy is ASO- or AAV-based, achieving clinical benefit with only partial SOD1 suppression will help to mitigate potential problems associated with complete SOD1 LOF.

ASOs targeting C9orf72 sense strand hexanucleotide repeat-containing RNAs for degradation have also shown promising results in preclinical studies (Jiang et al., 2016; Lagier-Tourenne et al., 2013), leading to a current study of ASOs in human C9ALS patients. The ASO BIIB078 targets sense strand C9ORF72 transcripts that contain the hexanucleotide G₄C₂ repeat (Figure 1A), and the safety and tolerability of this ASO are currently being assessed in a phase 1 clinical trial for patients with C9ALS (ClinicalTrials.gov identifier: NCT03626012). If shown to be safe and efficacious, perhaps targeting both sense and antisense strand repeat-containing RNAs will be more efficacious than targeting the sense strand alone, as both strands contribute to RNA foci and DPR production (Gendron et al., 2013; Lagier-Tourenne et al., 2013; Mizielinska et al., 2013; Mori et al., 2013a).

Lastly, jacifusen—an experimental ASO (not tested extensively in preclinical studies) that targets mRNA encoding mutant FUS^P525L (Figure 1A)—has been developed, as these mutations in FUS can cause juvenile onset and rapidly progressing forms of ALS. To date, the late Jaci Hermstad and two other FUS-ALS patients have received jacifusen, which is scheduled to be given to eight additional patients with FUS mutations (Figueiredo, Marta, 2020). With advancements in ASO-based therapies, we may expect personalized ASOs to become more common in treating ALS patients—even those with rare mutations.

A CALL FOR MULTI-TARGETED THERAPEUTIC APPROACHES

While mutations in many ALS genes have demonstrated evidence for toxic GOF mechanisms, such effects are often intrinsically linked to LOF effects that may also contribute to disease onset, progression, and/or severity. As such, we suspect that a multi-targeted therapeutic approach targeting both GOF and LOF mechanisms will be necessary to treat the majority of ALS patients.

For example, G₄C₂ repeat expansions in C9ORF72 result in RNA foci, DPRs, and reduced levels of C9ORF72, all of which may contribute to ALS pathogenesis. While G₄C₂ repeat-targeting ASOs would aim to mitigate RNA foci- and DPR-mediated toxicity, they would not affect the reduced levels of C9ORF72. Thus, a combination therapy that targets not only sense and antisense G₄C₂ repeat-containing RNAs but also increases levels of C9ORF72 whether that is via gene therapy, a small molecule, or CRISPR activation (CRISPRa) (Gilbert et al., 2014) may be more effective in ameliorating both the LOF and toxic GOF effects caused by this mutation (Figure 5A). For ALS genes that are thought to primarily be LOF, we should also consider the possibility of boosting their levels, such as for individuals harboring mutations in GRN, TBK1, OPTN, NEK1 or C21ORF2.

A multi-targeted therapeutic approach may also be beneficial in general when TDP-43 is depleted from the nucleus and forms cytoplasmic inclusions, which occurs in nearly all (~97%) ALS cases (Taylor et al., 2016). Cytoplasmic TDP-43 aggregates have probable toxic GOF properties, thus targeting them may be beneficial for ALS patients. Directly reducing levels of TDP-43 is not likely to be advantageous as TDP-43 plays a role in vital cellular processes, but targeting modifiers of TDP-43-mediated toxicity, such as Ataxin-2, may be a viable therapeutic strategy; reducing levels of Ataxin-2 with ASOs significantly extended lifespan and reduced pathology/ behavioral deficits in a mouse model of TDP-43 proteinopathy (Becker et al., 2017), and mutations in the Ataxin-2 gene (ATXN2) increase stability of the Ataxin-2 protein and risk for ALS in humans (Elden et al., 2010).

Yet even if we target toxic GOF properties of cytoplasmic TDP-43, such a strategy may fall short in addressing negative consequences resulting from loss of TDP-43 function in the nucleus. One potential way to address this shortcoming is to target known processes closely affected by nuclear depletion of TDP-43, such as the decreased levels of STMN2 that may be contributing to ALS pathogenesis (Klim et al., 2019; Melamed et al., 2019). Perhaps combined therapies can aim to compensate for the negative effects of TDP-43 LOF by restoring STMN2 levels through gene therapy or an ASO that blocks cryptic exon inclusion (Figure 5B), to enhance the efficacy of approaches that target toxic cytoplasmic TDP-43.

CONCLUDING REMARKS

In just over 25 years, the field has shifted from identifying the first fALS gene to entering an exciting era of therapeutic development, ranging from the delivery of personalized ASOs for a single patient to finding druggable targets for common disease pathways in sALS (Becker et al., 2017; Klim et al., 2019; Melamed et al., 2019). As detailed above, the need to consider both loss and gain of function in developing new therapies is clear. While some LOF effects may not alone be enough to cause ALS phenotypes—such as reduced levels of C9ORF72 or TBK1—there is substantial evidence that they can significantly modify the toxic effects of the disease. Even for familial mutations that almost certainly work through GOF mechanisms like SOD1 mutations, complete inhibition must be carefully considered, as detrimental phenotypes have been observed in both humans and mice with loss of SOD1 activity.

Some potential ASO therapies, such as tofersen for SOD1-ALS, indiscriminately target both WT and mutant alleles—yet it is not known whether the incomplete silencing of the normal allele will produce unwanted side effects. To circumvent this caveat, other ASOs that target specifically the mutant allele are under consideration for C9ALS and ALS caused by FUS^P525L mutations. Regardless, no current therapies for humans with ALS are aimed at mitigating deleterious LOF effects that may contribute to or modify disease pathogenesis, and these LOF effects should be considered.

This review showcases the importance and sometimes prevalence of the role of LOF in ALS, and the immense opportunity to target such functions. Whether LOF is the primary mechanism or a modifier of severity/susceptibility to ALS, it is clear that both GOF and LOF mechanisms deserve attention and careful consideration as the field moves forward to combat and prevent this devastating disease.