Rare variant analyses validate known ALS genes in a multi-ethnic population and identifies ANTXR2 as a candidate in PLS

Abstract

Background: Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease affecting over 30,000 people in the United States. It is characterized by the progressive decline of the nervous system that leads to the weakening of muscles which impacts physical function. Approximately, 15% of individuals diagnosed with ALS have a known genetic variant that contributes to their disease. As therapies that slow or prevent symptoms, such as antisense oligonucleotides, continue to develop, it is important to discover novel genes that could be targets for treatment. Additionally, as cohorts continue to grow, performing analyses in ALS subtypes, such as primary lateral sclerosis (PLS), becomes possible due to an increase in power. These analyses could highlight novel pathways in disease manifestation.

Methods: Building on our previous discoveries using rare variant association analyses, we conducted rare variant burden testing on a substantially larger cohort of 6,970 ALS patients from a large multi-ethnic cohort as well as 166 PLS patients, and 22,524 controls. We used intolerant domain percentiles based on sub-region Residual Variation Intolerance Score (subRVIS) that have been described previously in conjunction with gene based collapsing approaches to conduct burden testing to identify genes that associate with ALS and PLS.

Results: A gene based collapsing model showed significant associations with SOD1, TARDBP, and TBK1 (OR=19.18, p = 3.67 × 10^-39; OR=4.73, p = 2 × 10^-10; OR=2.3, p = 7.49 × 10^-9, respectively). These genes have been previously associated with ALS. Additionally, a significant novel control enriched gene, ALKBH3 (p = 4.88 × 10^-7), was protective for ALS in this model. An intolerant domain based collapsing model showed a significant improvement in identifying regions in TARDBP that associated with ALS (OR=10.08, p = 3.62 × 10^-16). Our PLS protein truncating variant collapsing analysis demonstrated significant case enrichment in ANTXR2 (p=8.38 × 10^-6).

Conclusions: In a large multi-ethnic cohort of 6,970 ALS patients, rare variant burden testing validated known ALS genes and identified a novel potentially protective gene, ALKBH3. A first-ever analysis in 166 patients with PLS found a candidate association with loss-of-function mutations in ANTXR2.

Keywords: ALS; Amyotrophic lateral sclerosis; Burden testing; PLS; Rare-variant analyses; peripheral lateral sclerosis.

Figure 1.. Q-Q plots of gene- and domain-level collapsing of ALL functional coding variants in ALS cohort.

(A) The results for a standard gene-level collapsing of 6,970 ALS cases and 22,524 controls. P-values were generated using an exact two-sided Cochran-Mantel-Haenszel (CMH) by gene by cluster. The genes with the top associations that achieved study-wide significance of p<4.9×10⁻⁷ (SOD1 (OR=19.18), TARDBP (OR=4.73), TBK1 (OR=2.3), and ALKBH3 (OR=0.26)) are labeled. SOD1, TARDBP, TBK1 have been previously implicated in rare variant association studies of ALS. Yellow and green lines indicate the 2.5^th and 97.5^th percentile of expected p-values, respectively. (B) The results for the domain-based collapsing restricting qualifying variants to those with subRVIS domain percentage score < 25 of 6,970 cases and 22,524 controls. P-values were generated using an exact two-sided Cochran-Mantel-Haenszel (CMH) by gene by cluster. The genes with the top associations (SOD1 (OR=20.63), TARDBP (OR=10.08), and TBK1 (OR=3.15)) are labeled. (C) Standard gene-level collapsing model showed 44 qualifying variants in cases (red circles) and 31 in controls (blue circles) for TARDBP (D) subRVIS domain collapsing improved association by removing control variants (cases = 43; controls = 15). Regions with subRVIS domain percentage below 25 are highlighted in orange while those above this threshold are highlighted in blue. A one tailed z-score showed that there were significantly less controls in the intolerant domain as indicated by subRVIS domain percentage score < 25 (p=0.031).

Figure 2.. Plot of gene- and domain-level collapsing of ALL SOD1 functional coding variants.

Standard gene-level collapsing model showed 93 qualifying variants in cases (red circles) and 18 in controls (blue circles) for SOD1. subRVIS domain collapsing improved association by removing control variants (cases = 90; controls = 16). Regions with subRVIS domain percentage below 25 are highlighted in orange while those above this threshold are highlighted in blue. However, a one tailed z-score showed that the differences in the number of controls in the intolerant domain was not significantly lower than those in the entire gene as indicated by subRVIS domain percentage score < 25 (p=0.4).

Figure 3.. Plot of gene- and domain-level collapsing of ALL TBK1 functional coding variants.

Standard gene-level collapsing model showed 73 qualifying variants in cases (red circles) and 143 in controls (blue circles) for TBK1. subRVIS domain collapsing improved association by removing control variants (cases = 47; controls = 72). Regions with subRVIS domain percentage below 25 are highlighted in orange while those above this threshold are highlighted in blue. However, a one tailed z-score showed that the differences in the number of controls in the intolerant domain was not significantly lower than those in the entire gene as indicated by subRVIS domain percentage score < 25 (p=0.3).

Figure 4.. Q-Q plot of gene-level collapsing of protein truncating variants (PTV) in PLS cohort.

The results for a standard gene-level collapsing of 166 PLS cases and 17,695 controls. P-values were generated using an exact two-sided Cochran-Mantel-Haenszel (CMH) by gene by cluster. The gene with the top associations that achieved genome-wide significance of p<8.38×10⁻⁶ (ANTXR2 (OR=174.57)) is labeled. ANTXR2 has not been previously implicated in rare variant association studies of PLS. Yellow and green lines indicate the 2.5^th and 97.5^th percentile of expected p-values, respectively.

Figure 5.. Forest plot of ALS genes by model.

Rare variants in “ALS Confirmed” genes were significantly associated with ALS in all gene-based collapsing models except the control synonymous model. Rare variants in “ALS Plus” genes were associated with ALS in “Missense & PTV” gene-based collapsing model. There was no association with ALS of rare variants in “ALS Replication Needed” and “ALS Weak Evidence” genes. Pooled odds ratio, 95% confidence intervals, and p-values were generated from exact two-sided Cochran-Mantel-Haenszel (CMH) tests.