Multiomic profiling reveals timing of menopause predicts prefrontal cortex aging and cognitive function

Abstract

A new case of dementia is diagnosed every 3 s. Beyond age, risk prediction of dementia is challenging. There is growing evidence of underlying processes that connect aging across organ systems and may provide insight for early detection, and there is a need to identify early biomarkers at an age when action can be taken to mitigate cognitive decline. We hypothesized that timing of menopause, a marker of ovarian aging, predicts brain age decades later. We used 2086 subjects with multiple "omics" measurements from post-mortem brain samples. Age at menopause (AAM) is positively correlated with cognitive function and negatively correlated with pre-frontal cortex aging acceleration (calculated as estimated biological age from DNA methylation minus chronological age). Genetic correlations showed that at least part of these associations is derived from shared heritability. To dissect the mechanism linking AAM to cognitive decline, we turned to transcriptomic data which confirmed that later AAM was associated with gene expression in pre-frontal cortex consistent with better cognition, and among those who reached menopause naturally, decreased gene expression of pathways implicated in aging. Those with surgical menopause displayed different molecular changes, including perturbed nicotinamide adenine dinucleotide (NAD+) activity, validated by metabolomics. Bile acid metabolism was perturbed in both groups, although different bile acid ratios were associated with AAM in each. Together, our data suggest that AAM is predictive of brain aging and cognition, with potential mediation by the gut, although through different mechanisms depending on the type of menopause.

Keywords: aging; dementia; menopause; multiomics.

FIGURE 1

Age at menopause and cognitive function in the elderly. (a–c) Age at menopause (AAM) and cognition scores at ages 70, 80, and 90 (d) E‐value analysis shows the effect size needed to completely explain the relationship between AAM and cognition at ages 70, 80, and 90. Effect sizes are shown for a 1‐year difference in AAM and a five‐year difference in AAM. (e) Cognitive function subdomain analysis shows all five subdomains are associated with AAM. See Table S5 for full results.

FIGURE 2

Genetic and epigenetic associations with AAM. (a) Biological age in the brain is associated with chronological age (at death) and (b) inversely correlated with cognition at last visit (c) Brain age acceleration (biological age—chronological age) is correlated with age at menopause. Full results from regression with covariates in Table S7. (d) Genetic correlations of AAM from two GWAS cohorts with 7 cognition‐ and aging‐related traits. Full results in Table S8.

FIGURE 3

Genes associated with cognition are enriched for genes associated with AAM in directions concurrent with phenotypic findings. (a, b) There is significant enrichment of genes positively associated with cognition (COG‐POS) in Broadman Area (BA) 10 among genes positively associated with AAM in both natural and surgical cohorts (c, d) There is significant enrichment of genes negatively associated with cognition (COG‐NEG) in BA 10 among genes negatively associated with AAM in both natural and surgical cohorts. (e, f) The same analysis as in (a, b) but performed using COG‐POS gene set derived from BA 44, another area within frontal cortex. The same analysis as in (c, d) but performed using COG‐NEG gene set derived from BA 44.

FIGURE 4

Pathways associated with AAM in both natural and surgical menopause, in the same direction. (a, b) Enrichment of Hallmark pathways that are associated with AAM in both natural and surgical groups. (c) The significance of overlap of leading edge gene sets between surgical and natural menopause in each pathway, based on the hypergeometric test. The response to estrogen has a higher overlap in leading edge gene sets by type of menopause than expected by chance. The rest of the pathways have an overlap no more than that expected by chance. Row names (pathways) are preserved across (a)–(c). (d–f) Ratios of bile acid concentrations that are significantly associated with age at natural menopause (d, e) and age at surgical menopause (f). (g) Comparison of effect size of AAM and each bile acid ratio from the linear regression results, by menopause type. Full results in Table S9.

FIGURE 5

Molecular mechanisms that connect AAM to aging. (a) Hallmark pathways that are significantly associated with AAM in opposite directions in natural and surgical groups. (b) miRNA that have significant expression changes between matched pre‐ and post‐oophorectomy samples from the same women (c) Pathway enrichment of experimentally validated targets of miRNA significantly perturbed after oophorectomy. (d) The concentration in the brain of nicotinamide riboside, one of the key NAD+ precursors, is associated with age at surgical menopause.