Effects of brain microRNAs in cognitive trajectory and Alzheimer's disease

Abstract

microRNAs (miRNAs) have a broad influence on gene expression; however, we have limited insights into their contribution to rate of cognitive decline over time or Alzheimer's disease (AD). Given this, we tested associations of 528 miRNAs with cognitive trajectory, AD hallmark pathologies, and AD clinical diagnosis using small RNA sequencing from the dorsolateral prefrontal cortex of 641 community-based donors. We found 311 miRNAs differentially expressed in AD or its endophenotypes after adjusting for technical and sociodemographic variables. Among these, 137 miRNAs remained differentially expressed after additionally adjusting for several co-occurring age-related cerebral pathologies, suggesting that some miRNAs are associated with the traits through co-occurring pathologies while others through mechanisms independent from pathologies. Pathway enrichment analysis of downstream targets of these differentially expressed miRNAs found enrichment in transcription, postsynaptic signalling, cellular senescence, and lipoproteins. In sex-stratified analyses, five miRNAs showed sex-biased differential expression for one or more AD endophenotypes, highlighting the role that sex has in AD. Lastly, we used Mendelian randomization to test whether the identified differentially expressed miRNAs contribute to the cause or are the consequence of the traits. Remarkably, 15 differentially expressed miRNAs had evidence consistent with a causal role, laying the groundwork for future mechanistic studies of miRNAs in AD and its endophenotypes.

Keywords: Alzheimer’s disease; Beta-amyloid; Brain microRNA; Cognitive decline; Cognitive trajectory; Neurofibrillary tangles.

Fig. 1

Summary of differentially expressed miRNAs from the minimally adjusted and fully adjusted models. See Table 2 for variables adjusted for in each model. a number of differentially expressed miRNAs per trait. b Upset plots for the number of DE miRNAs unique to or shared among the traits beta-amyloid, neurofibrillary tangles, AD clinical diagnosis, and trajectory of global cognition. The vertical bars show the number of DE miRNAs for the combination of traits indicated by the black circles in the matrix. Beta-amyloid had no DE miRNAs with the fully adjusted model and therefore is not included in the bottom panel. c Volcano plots for each trait. Dashed horizontal line indicates p-value corresponding to FDR < 5%. Points with gray fill indicate non-significant associations. Points with crosses indicate associations with miRNAs miR-132/212-3p/5p

Fig. 2

miRNA-trait pairs with significant sex-biased differential expression. The limma moderated t-statistic is the ratio of the beta estimate to the miRNA expression standard error after the standard errors have been moderated across miRNAs. The plot shows the moderated t-statistic from the male-only (orange) and female-only (green) analyses for each miRNA-trait pair (involving five unique miRNAs) with a significant sex interaction term, which allows comparison of the direction and magnitude of miRNA-trait association in each sex. Full results are in Supplementary Table 13

Fig. 3

Identifying candidate causal miRNAs for AD and AD endophenotypes. A. Overview of workflow. The 122 DE miRNAs that had at least one miR-QTL in ref. [38] were tested using SMR and HEIDI. The SMR test identifies miRNAs associated with AD or AD endophenotypes through shared genetic association, which may occur through three models. The HEIDI test distinguishes the linkage model, in which the miRNA is associated with AD or AD endophenotype due to linkage of separate causal variants, from the causality and pleiotropy models, in which the miRNA is associated with AD or AD endophenotype due to a shared causal variant. The statistical tests do not distinguish causality and pleiotropy, so miRNAs consistent with either of these models are candidate causal miRNAs. B. Forest plots for 15 miRNAs consistent with a causal role in AD endophenotypes. Fifteen miRNAs passed the SMR/HEIDI thresholds and are candidate causal miRNAs. The SMR beta point estimates (filled circles) and 95% confidence intervals (solid lines) are plotted for candidate causal miRNAs for cognitive trajectories (top), beta-amyloid (bottom left), and neurofibrillary tangles (bottom right). None of the DE miRNAs for AD clinical diagnosis passed the SMR/HEIDI thresholds. Full results are presented in Supplementary Table 14