Short structural variants as informative genetic markers for ALS disease risk and progression.

There is considerable variability in disease progression for patients with amyotrophic lateral sclerosis (ALS) including the age of disease onset, site of disease onset, and survival time. There is growing evidence that short structural variations (SSVs) residing in frequently overlooked genomic regions can contribute to complex disease mechanisms and can explain, in part, the phenotypic variability in ALS patients. Here, we discuss SSVs recently characterized by our laboratory and how these discoveries integrate into the current literature on ALS, particularly in the context of application to future clinical trials. These markers may help to identify and differentiate patients for clinical trials that have a similar ALS disease mechanism(s), thereby reducing the impact of participant heterogeneity. As evidence accumulates for the genetic markers discovered in SQSTM1, SCAF4, and STMN2, we hope to improve the outcomes of future ALS clinical trials.

Background

Many key cellular processes are known to be disrupted in amyotrophic lateral sclerosis (ALS) such as RNA metabolism, mitochondrial dysfunction, oxidative stress, protein aggregation, impaired axonal transport, and cytoskeletal dysfunction, reviewed in [1]. Variation in the expression of genes involved in these processes may increase disease risk and/or influence the rate of disease progression [2-4]. At present, there is a lack of genetic markers for the different ALS disease subtypes, as well as a lack of genetic indicators of disease risk and/or trajectory. The heterogeneous clinical presentation and diverse rates of progression makes identifying ALS patients with similar disease mechanisms extremely challenging, and undoubtedly contributes to clinical trial failures [5].

Establishing molecular targets and genetic markers for ALS can lead to improved patient stratification for clinical trials and might enable positive treatments to be identified for specific patient subgroups. Van Eijk and Eijkemans [6] recently demonstrated that genotypic data for unc-13 homolog a (UNC13A), myelin-associated oligodendrocyte basic protein (MOBP), and the repeat expansion in c9orf720-SMCR8 complex subunit (C9orf72) could influence both primary and secondary outcomes including survival, ALS functional rating scale (ALFSRS) and forced vital capacity (FVC) measures. Additionally, a retrospective meta-analysis of three lithium carbonate clinical trials revealed that contrary to the reported negative outcomes, patients with the UNC13A (C/C) genotype had actually responded to the lithium carbonate [7]. This study provides evidence that genetic markers can inform clinical trial outcomes and should be incorporated into clinical trial design. Intuitively, genetic regions that are highly variable, known as structural variants (SVs), will likely be more informative as genetic markers than single nucleotide polymorphisms (SNP), due to a larger number of potential genotypes [8]. These variable regions of the genome have been predominantly under-characterized [9, 10]; however, the scientific community is beginning to appreciate the need to investigate polymorphic loci as potential disease modifying regions.

Structural variants (SVs) have typically been defined as insertions, deletions, inversions, and microsatellites that can be repeated hundreds of times within the genome that are greater than 1 kb in length [11]. Short structural variants (SSVs) have been predominately overlooked in the context of ALS and encompass that same class of variants as the classical definition (e.g., short tandem repeat, microsatellite, insertion/deletion, inversion, polynucleotide repeat) but are much shorter in size, typically < 50 bp in length [12]. Approximately 4 million SSVs exist within the human genome and have been previously described [13, 14]. It is possible that some of the “missing heritability” in ALS could be explained by more common SSVs with small effect sizes that have not yet been identified [13, 15]. Importantly, changes in the size and composition of both SV/SSVs can have a significant impact on the binding of regulatory elements that modulate RNA processing and gene expression [16]. SSVs have been implicated in many complex diseases, including ALS and other neurodegenerative diseases [17], such as Parkinson’s and Alzheimer’s disease [18-20]. The ability of SSVs to alter gene expression is dependent on their location within and around the gene or intergenic region, with their effects occurring via several mechanisms including the following: influencing the binding of regulatory elements, mRNA splicing and processing, genome folding and higher order structure, and translation. This may differentiate mechanisms of disease pathogenesis, including risk of disease, risk for a specific phenotype, symptom presentation, disease course, and response to treatment, between individuals [21].

Recent in silico mapping of known ALS-linked genes has predicted a number of as yet unresolved short tandem repeats within each of these genes, that are likely polymorphic, and could influence gene expression and contribute to disease risk for ALS [15, 22]. Importantly, in the most recent ALS genome wide association study (GWAS) conducted by Van Rheenen and Van der Spek [23], 15 risk loci were identified, with 8 loci previously reported in GWAS studies [24-26]. This GWAS study was unique; in addition to screening pathogenic rare burden variants, it also incorporated short tandem repeats, RNA-sequencing, and methylation datasets to prioritize causal genes within identified ALS risk loci. Of note, a polymorphic tetra-nucleotide repeat downstream of ALS-linked gene NIMA-related kinase 1 (NEK1) was reported to be associated with increased ALS disease risk. The NEK1 SSV was in linkage disequilibrium with the top hit NEK1 SNP reported in this study and may help to further explain phenotypic variability and disease penetrance between patients carrying different NEK1 mutations, particularly since the SNP alone could not reliably determine its contribution to ALS risk [23]. Furthermore, another recent study has investigated SVs in 6580 whole genome sequences (4315 ALS and 1880 controls) from the Project MinE cohort to determine genotype-phenotype correlations in known ALS genes. Al Khleifat, Iacoangeli [27] reported that structural variants in C9orf72 (repeat expansion), Valosin containing protein VCP (inversion) and erb-b2 receptor tyrosine kinase 4, ERBB4 (deletion) are variously associated with ALS disease risk and phenotype. The variant caller used in this study detected multiple classes of SVs, however, small inversions and insertion/deletions < 200 bp could not be analyzed by the Manta platform [27]. Further investigation into variants outside of the know ALS genes in addition to wet lab validation is required to gain a holistic understanding of the contribution of genetic variation to ALS risk and phenotype [27].

In addition to contributing to disease risk, SSVs can also be an informative tool for clinical trial participant selection, as demonstrated in Alzheimer’s disease, in the case of translocase of outer mitochondrial membrane 40 (TOMM40) [28]. Although the age-of-onset distributions for Alzheimer’s disease have been well-established since 1993 [29], the basis of this distribution was only partially explained by the apolipoprotein E (APOE) genotype, suggesting that other genetic factors must contribute to age-of-onset for the disease [30]. In fact, the combination of the APOE genotype alongside the SSV genotype, a poly-T repeat in TOMM40, was subsequently shown to account for > 98% of the clinical age-of-onset distributions in the Caucasian population [20, 28]. Combining these genotypes allowed the generation of the clinical age-of-onset risk algorithm [31] to be developed to inform participant selection for the TOMMORROW clinical trial [32], with selection based on the genotype of individuals and their corresponding risk of disease/age-of-onset prediction. Evidently, this finding demonstrates that informative SSVs can be used as an enrichment tool for clinical trials, thus informing participant selection.

Innovative approach for SSV discovery

In ALS, the clinical presentation can manifest differently in family members sharing the same ALS-linked variant (incomplete penetrance), suggesting multiple factors, including both genetic and non-genetic factors, can contribute to the progression of the disease [15, 21, 33]. With this in mind, we have used an established structural variant evaluation system (dbSSV) [22] to identify candidate SSVs within and around confirmed ALS loci [15]. Using a systematic approach to candidate GWAS loci, analysis of SSVs in these genomic regions will likely identify common variants that subtly influence gene function, in some cases leading to ALS. The dbSSV software [22] focuses specifically on identifying SSVs and scores them against 24 different properties, including data describing the location and definition of the variation, variability indicators, repeat context, gene context, transcription factor and microRNA binding sites, other regulatory markers, conservation, position within a linkage disequilibrium block, GWAS signals, and tissue-specific regulatory signals. Based on these scores, a short list of SSVs is generated with each total score suggesting the likelihood of the variant having significant biological effects and contributing to disease risk. Using this method, we have now identified and published several novel genetic markers, discussed below. Investigating SSVs in ALS-linked genes will help to better understand differences between individual patient phenotypes and disease progression.

Insertion/deletion in SQSTM1 is associated with disease in familial SOD1 patients

Sequestosome 1 (SQSTM1) encodes P62, an adaptor protein that binds ubiquitylated protein aggregates and delivers them to the autophagosome for degradation. With an essential role in protein clearance, it is not surprising that P62 dysfunction is implicated in neurodegenerative diseases that are governed by abnormal protein inclusions. Thus, SSVs within SQSTM1 may contribute to the diverse presentations observed between ALS patients by influencing P62 expression and autophagic clearance. In our recently published association study of a North American cohort of familial ALS (fALS) and sporadic (sALS) patients (n = 403) and age matched controls (n = 562), a small cytosine/adenine (CAAA) insertion deletion (I/D) was associated with fALS, particularly in familial superoxide dismutase 1 (SOD1) mutation positive patients, but not with sALS patients [34]. Furthermore, the presence of the insertion variant appears to translate to a stepwise decrease in SQSTM1 expression in healthy olfactory neurosphere-derived cells, with the I/I genotype resulting in a 2.5-fold reduction in transcript levels [34]. The observations of the insertion/deletion influencing SQSTM1 expression has previously been reported by an independent study in a screen of 17 different healthy tissue types [35]. By this weight of evidence, the CAAA variant needs to be further examined as a contributor to ALS disease mechanisms, particularly since SQSTM1 plays a critical role in autophagy, and mutations in this gene can be a direct cause of ALS and other neurodegenerative diseases [36, 37]. Future therapeutic studies to modulate autophagy should take into consideration the potential impact of this SSV on P62 expression.

Poly-T repeat in SCAF4 is a genetic marker for disease risk and survival in familial ALS

Studies of common SOD1-ALS mutations have suggested that a disease modifying factor located nearby SOD1 may influence the penetrance of SOD1 p.D91A (D90A) and p.A5V (AV4) mutations (historical names noted in brackets are not reflective of the amino acid position), thus contributing broadly to ALS risk [38, 39]. Our laboratory reported an 11-18 thymine repeat, located within the 3′UTR of the neighboring gene, SR-related CTD associated factor 4 (SCAF4) [40]. The function of SCAF4 has recently been elucidated, with an essential role in RNA processing through regulating transcript elongation [41]. Variants in SCAF4 have been reported to cause impaired RNA processing and can result in neurodevelopmental disorders [42]. Interestingly, the poly-T variant in SCAF4 is flanked by two binding sites for transcription factor RNA polymerase II subunit A (POLR2A); therefore, changes in the length of this variant may influence the binding of POLR2A, thus affecting transcription of nearby genes (SOD1) and may therefore play a role in ALS related neurodegeneration. A case control association study in fALS patients (n = 190) and healthy age matched controls (n = 560) revealed that the 18 T repeat is associated with ALS risk for the entire cohort, including those without SOD1-linked mutations (n = 27). The 18 T allele was also associated with a 26 month reduction in survival time but was not associated with age at disease onset [40]. Future studies should investigate the functional impact of this variant and determine whether thymine repeat length influences transcript elongation of genes that are regulated by SCAF4. Additionally, this SSV may also help to identify patients that could benefit from a SOD1 targeted therapy, such as the SOD1 suppressing antisense therapy Tofersen (currently in phase III extension study) [43].

Short tandem repeat in STMN2 is associated with sporadic ALS disease risk and clinical phenotype

The microtubule regulator stathmin-2 (STMN2) was recently identified as a gene with strong therapeutic potential for ALS [44, 45]. STMN2 is involved in axonal grown and repair and is directly regulated by the ALS-linked gene TAR DNA-binding protein 43 (TARDBP). When TARDBP protein TDP-43 is depleted or mislocalized, as occurs in ALS, STMN2 protein is also depleted [44, 45]. To understand the implications of this TDP- 43:STMN2 relationship and ALS phenotype, it is essential to characterize the natural variation in STMN2 expression.

We recently examined the STMN2 gene for candidate SSVs that may affect the binding of regulatory elements and, therefore, influence gene expression. We identified a variable length cytosine/adenine (CA) short tandem repeat within STMN2 in a cohort of Caucasian sALS patients (n = 321) and healthy age matched controls (n = 332). We have identified the first genetic association with STMN2 and both sALS disease risk and age-at-disease onset [46]. In 143 patients where end-point survival data was available, when categorized according to site of disease onset (bulbar vs spinal), the bulbar cases carrying the risk genotype had significantly shorter survival times than other bulbar cases. Moreover, this effect on survival was not abrogated when controlling for sex or age. Furthermore, in an Australian longitudinal sALS cohort (n = 67), ALSFRS scores were significantly lower in patients carrying the risk genotype. Following these clinical associations, stathmin-2 mRNA expression was shown to be reduced in sALS patient olfactory neurosphere-derived cells. When accounting for CA genotype, a trend for reduced expression of stathmin-2 mRNA was also observed in sALS patients and control laser-captured spinal motor neurons [46].

This work points to a novel mechanism by which this SSV may regulate STMN2 gene expression and could further explain the recently elucidated STMN2 cryptic exon mechanism [44, 45] and its influence on neurodegenerative disease phenotype [47, 48]. The genetic validation of STMN2 significantly adds to the weight of evidence that this gene is important in ALS [46]. The STMN2 genetic marker may therefore be a useful tool for cohort selection in clinical trials or to stratify patient response. Further, this discovery has broad implications for clinical assessment and therapeutic development and should be incorporated in future clinical trials targeting this gene.

Application of SSVs to clinical trials

Recently, the Treatment Research Initiative to Cure ALS (TRICALS) has brought the urgent need to reform ALS clinical trial design to the forefront of the literature [5, 49]. In particular, the major concerns raised are related to the stringent patient selection criteria and analytical strategy of phase 3 clinical trials. Van Eijk and Nikolakopoulos [49] highlight the use of patient risk profiles as a strategy to provide more informative selection criteria that will help to improved randomization, enable risk-based subgroup analyses, and increase the statistical power of clinical trials. Patient risk profiles are based on a multivariate analysis of several patient characteristics (i.e., age of onset, site of onset, vital capacity, diagnostic delay, ALSFRS etc.), creating a “prognostic summary” for each patient [49]. This is likely to help increase the number of eligible patients for trials, while reducing patient drop out and exclusion rates, thus increasing the generalizability of the trial [49]. The authors noted that they did not include C9orf72 repeat expansion as a factor in their predictive model, as it did not significantly improve the predictive power of the model and may pose additional logistical constraints (i.e., genotyping and counseling patients) prior to clinical trial randomization [49]. However, one must consider that although the C9orf72 repeat expansion is applicable to 40% of fALS patients and 6% sALS patients [50], incomplete penetrance is likely to be the complicating factor that influences the predictive capacity of this variant. Incorporating genetic variants into the prognostic summary may not improve prognostic predictions, since the clinical characteristics that the variant may be associated with are already individually incorporated into the model. The true value of patient genotype information is in balancing the clinical trial treatment arms, allowing sub-group analyses for patients that may have similar underlying disease mechanisms or ALS risk factors. It is likely that the eligibility window for patient risk profiles will need evaluation on a study-by-study basis, particularly since this will be influenced by the type of trial, i.e., broad vs genotype targeted treatment [49]. However, the use of patient genotype information will be pertinent to maximizing the information gained from a prospective clinical trial on potential responders. Across the most frequently studied indications, it has been shown that genetically validated targets are twice as likely to succeed in clinical development, transitioning from phase 1 to approval [51]. Therefore, the use of association studies to identify ALS genetic markers will help to guide selection of appropriate drug targets for sub-populations and help lower the rate of failure in clinical development programs.

Tofersen is currently the most promising disease modifying therapeutic for ALS and recently completed phase III (NCT02623699) with its long-term extension study currently postponed (NCT03070119) [43]. Tofersen was initially designed as a genetically targeted therapeutic for SOD1 mutation positive patients [43]; however, evidence is accumulating that this therapy may also benefit other forms of sALS, where SOD1 misfolding and aggregation is also a pathological feature [52, 53]. The use of genetic information in identifying the target population (mutation positive patients) in its initial early trial stage was integral to the phase I/II trial success [43]. Incorporation of genetic information in early-stage clinical development allowed initial efficacy to be shown and can be built upon in the hope of treating a broader patient population. Genetic variants that are associated with SOD1 aggregation within sALS cohorts will be necessary to identify those sALS patients who are likely to respond to a SOD1 targeting therapy. With the evolution of personalized treatment approaches using antisense oligonucleotides for neurodegenerative disorders [54], there is an absolute need to incorporate informative SSV genetic markers into early stage drug development. This will help to identify patients that have similar ALS disease mechanisms that would benefit from a targeted antisense therapy.

Conclusion

Currently, there is no effective treatment for ALS, and over the past 20 years, more than 60 controlled trials of putative ALS therapeutics have failed to demonstrate clinical efficacy [55, 56]. Current treatment options are non-specific and only extend survival by ~3 months in some patients, and furthermore, at present, there is no method to determine which patients are more likely to respond to a particular ALS therapeutic.

The continued discovery and evaluation of novel SSVs will undoubtedly shed light on the pathogenic mechanisms of ALS neurodegeneration. At present, there are few biomarkers/genetic markers that allow patient stratification according to disease mechanism [57-60] and treatment efficacy can only be evaluated by clinical measures during current clinical trials [34, 38]. With increasing evidence from our laboratory that SSVs do contribute to ALS risk and have disease-modifying effects [15, 34, 40, 46], investigations need to incorporate SSVs into genetic studies and clinical trial design [5]. There is an urgent need to establish well-characterized genetic markers that can be used to inform on the validity of certain treatment approaches. As ALS is a complex and heterogeneous disorder, with a varied clinical phenotype and disease trajectory, personalized medicine approaches will be more likely to result in successful treatments because they can directly target the underlying disease mechanism [54]. Therefore, it is crucial to be able to identify patient subgroups and develop compounds that are more likely to be effective in genetically defined subgroups of patients, thus reducing the impact of participant heterogeneity.

The authors recognize the challenges we face in current ALS clinical trials. Moving forward, we must examine the potential of these SSVs as a tool for patient stratification in retrospective clinical trials cohorts. This will result in the accelerated development of these genetic markers, fast tracking them into current and future clinical trials. Undoubtedly, as the data accumulates for these genetic markers, we are hopeful this will translate into identifying responder populations of ALS patients, allowing drug development to continue for specific subsets of patients. This is likely to significantly change the way clinical trials are conducted in ALS moving forward.