Genomic characterization of Huntington's disease genetic modifiers informs drug target tractability

Abstract

Huntington's disease is caused by a CAG repeat in the HTT gene. Repeat length correlates inversely with the age of onset but only explains part of the observed clinical variability. Genome-wide association studies highlight DNA repair genes in modifying disease onset, but further research is required to identify causal genes and evaluate their tractability as drug targets. To address these gaps and learn important preclinical information, we analysed genome-wide association study data from a large Huntington's disease age-of-onset study (n = 9064), prioritizing robust candidate Huntington's disease modifier genes using bioinformatic approaches and analysing related information for these genes from large-scale human genetic repositories. We supplemented this information with other Huntington's disease-related screens, including exome studies of Huntington's disease onset and high-throughput assessments of mHTT toxicity. To confirm whether Huntington's disease modifiers are shared across repeat expansion disorders, we also analysed age-of-onset genome-wide association study data from X-linked dystonia-parkinsonism caused by a (CCCTCT)_n expansion. We also studied modifier-related associations with rare diseases to inform potential off-target therapeutic effects and conducted comprehensive phenome-wide studies to identify other traits linked to these genes. Finally, we evaluated the aggregated human genetic evidence and theoretical druggability of the prioritized Huntington's disease modifier genes, including characteristics recently associated with clinical trial stoppage due to safety concerns (i.e. human genetic constraint, number of interacting partners and RNA tissue expression specificity). In total, we annotated and assessed nine robust candidate Huntington's disease modifier genes. Notably, we detected a high correlation (R ² = 0.78) in top age-of-onset genome-wide association study hits across repeat expansion disorders, emphasizing cross-disorder relevance. Clinical genetic repositories analysis showed DNA repair genes, such as MLH1, PMS2 and MSH3, are associated with cancer phenotypes, suggesting potential limitations as drug targets. LIG1 and RRM2B were both associated with neurofibrillary tangles, which may provide a link to a potential role in mHTT aggregates, while MSH3 was associated with several cortical morphology-related traits relevant to Huntington's disease. Finally, human genetic evidence and theoretical druggability analyses prioritized and ranked modifier genes, with PMS1 exhibiting the most favourable profile. Notably, HTT itself ranked poorly as a theoretical drug target, emphasizing the importance of exploring modifier-based alternative targets. In conclusion, our study highlights the importance of human genomic information to prioritize Huntington's disease modifier genes as drug targets, providing a basis for future therapeutic development in Huntington's disease and other repeat expansion disorders.

Keywords: human genetics; monogenic disease; repeat expansion.

Graphical Abstract

Figure 1

Huntington’s disease AOO-associated variants display a broad spectrum of effect sizes and have potential cross-repeat expansion disorder relevance. (A) The variant effect size (i.e. years earlier/later) of Huntington’s disease AOO GWAS hits is inversely correlated to allele frequency. A rare tag variant for the HTT loss of CAA interruption (LOI) displays the greatest impact on the age of onset, while common variants with allele frequencies of >25% have a more subtle impact on this trait (i.e. <2 years, on average) but are carried by more affected individuals. Prioritized candidate genes for each respective variant have been labelled and annotated by pathway and effect direction for that variant. The dashed line represents the linear regression of log-transformed AOO effect size on MAF of the 21 genome-wide significant variants assessed. (B) Combined allele frequency of loss of function and predicted deleterious variants in an exome sequencing data set of AOO in Huntington’s disease stratified by very early onset (n = 250) and very late onset (n = 250). Only Huntington’s disease modifier genes with evidence for association with onset time (FAN1, LIG1, MSH3 and PMS1) are plotted. (C) Overlapping Huntington’s disease AOO GWAS hits show similar effect sizes in a GWAS of age-associated penetrance of XDP, a repeat expansion disorder caused by a different motif (i.e. CCCTCT) in the TAF1 gene. These results are highly correlated (R² = 0.78, linear regression P = 7.3 × 10⁻⁴), suggesting a shared effect between these two repeat expansion disorders. Bars extending from each point represent standard error. The dashed line represents the linear regression of XDP effect size on Huntington’s disease AOO effect size, with the yellow-shaded area representing the 95% confidence interval for the regression line. Points are filled based on whether there was a significant association (linear regression P < 0.05) with XDP AOO in the related GWAS (n = 353).

Figure 2

Analysis of clinical genetic repositories and human GWASs can provide insights into potential unintended consequences of therapeutically targeting Huntington’s disease modifier genes and clarify their role in pathogenic processes. (A) DNA repair genes, especially mismatch repair pathway members, MLH1, PMS2 and MSH3, account for the majority of ClinVar pathogenic/likely pathogenic alleles due to their involvement in cancer. Therapeutically targeting these genes should be approached with caution since some mismatch repair deficiency syndromes are associated with the brain/central nervous system. (B) Prioritized gene–trait associations for Huntington’s disease modifiers that were identified through the analysis of the large repository of unbiased GWASs in the OTG database. Prioritized gene–trait pairs with Locus2Gene (L2G) >0.5 are plotted (i.e. indicating that the related GWAS trait signal is likely to arise from genetic variation linked to the Huntington’s disease modifier gene). The pairs identified include important neurobiological associations, including cortical thickness (MSH3) and neurofibrillary tangles (LIG1 and RRM2B).

Figure 3

HTT and Huntington’s disease modifier genes display different metrics for genomic features associated with past clinical trial success. Relevant thresholds for classification as favourable targets are indicated with dashed lines. (A) Genetic constraint metrics obtained from gnomAD [probability of being pLI and the negative observed/expected ratio for loss of function (LoF) variants] illustrate that both HTT and TCERG1 are highly constrained. The pLI is known to be a dichotomous-like metric (https://gnomad.broadinstitute.org). Most genes can be classified as non-constrained (pLI < 0.1) or constrained (pLI > 0.9), as can be seen in the figure. We include another measure of constraint, which is more quantitative (−observed/expected ratio, negative transformation used to make comparisons with pLI on the plot more intuitive) to illustrate the most constrained genes further. (B) HTT displays significantly more interacting partners (n = 868) than Huntington’s disease modifier genes based on IntAct Molecular Interaction (MI) scores >0.42. (C) The majority of Huntington’s disease modifier genes, as well as HTT, display low tissue specificity for RNA expression, except for PMS1 and LIG1, which demonstrate tissue-enhanced RNA specificity.

Figure 4

Huntington’s disease genetic modifier drug target profiles can be used to assess theoretical druggability by providing diverse evidence related to clinical trial success. Huntington’s disease modifier genes have been ordered based on the number of favourable scores obtained in our analyses. (A) Annotation of Huntington’s disease AOO GWAS hits by candidate modifier gene by number of independent GWAS hits and whether they are supported by only a rare variant association. Genes with multiple hits at a locus have more robust evidence for involvement with Huntington’s disease AOO. (B) Aggregated human genetic evidence relating to Huntington’s disease modifiers and HTT provides insights into the robustness of findings relating to these genes. Colours represent evidence in support (blue) or refutation (orange) for their involvement in relevant phenotypes/features, with lack of information displayed in grey. While LIG1 has a large amount of human genetic support, crucially, it has not been shown to influence somatic repeat instability in human stem cell models of Huntington’s disease. (C) Ranking Huntington’s disease modifier genes by favourable (teal green) and unfavourable (deep purple) metrics based on criteria associated with clinical trial stoppage due to safety concerns provides an unbiased way to prioritize candidate Huntington’s disease modifier genes. Genes that are theoretically better targets to prioritize for future study are on the left. PMS1 was ranked the most favourably for these factors. Notably, HTT ranked unfavourably for several criteria, which is in line with the recent failure of HTT-lowering trials. A, B and C share the same x-axis, representing the Huntington’s disease modifier gene name. hPSC, human pluripotent stem cell; MSN, medium spiny neuron; SRI, somatic repeat instability.