Genome-wide study links cardiometabolic factors to cognition via APOA4-APOA5-ZPR1-BUD13 and other loci in rural Indians

Abstract

Introduction: Cardiometabolic risks affect cognition during aging, yet genetic basis for both remain understudied in Indians.

Methods: This study constructs an ancestry-matched Indian haplotype reference panel for genotype imputation of 5111 rural Indians. Single-locus, gene-based, conditional genome-wide association analyses are performed on 20 cognitive and 10 cardiometabolic traits, with subsequent follow-up of identified associations through multimodal functional annotation. Furthermore, causal interrelationships between cardiometabolic and cognitive phenotypes by Mendelian randomization are investigated.

Results: One novel memory-associated and 17 novel cardiometabolic phenotypes-associated (high-density lipoprotein cholesterol [HDL-C], low-density lipoprotein cholesterol [LDL-C], triglycerides [TG], total cholesterol [TC], TG:HDL, and visceral adiposity index [VAI]) genome-wide significant loci, and multiple genes are identified. AMIGO1 (delayed-recall) and ZPR1-APOA5 (metabolic syndrome) exhibit distinct haplotype structure compared to other populations. Causal roles of cardiometabolic traits on various cognitive domains are identified via genetic instruments in APOC3-APOA4-APOA5-ZPR1-BUD13 among others.

Discussion: These findings illustrate the impact of cardiometabolic factors on cognition in a rural socioeconomically disadvantaged population, advancing efforts to address health disparities.

Highlights: Our newly constructed ancestry-matched haplotype reference panel gives better genotype imputation accuracy for the Indian population. One and 17 novel genome-wide significant single-loci were identified to be associated with cognitive and cardiometabolic traits, respectively. Several subgenome-wide hits for all phenotypes were identified. Collapsing protein truncating variants (PTVs), there were two genes identified to be associated with cardiometabolic traits at a genome-wide level of significance, correcting for multiple phenotypes tested. Haplotypic differences were identified compared to 1000 Genomes superpopulations for genes influencing delayed recall and metabolic syndrome. Adverse causal roles of cardiometabolic traits on cognition were uncovered via genetic instruments in APOC3-APOA4-APOA5-ZPR1-BUD13, among others, through Mendelian randomization.

Keywords: India; ancestry‐matched imputation panel; cardiometabolic, cognition; causal association; genome‐wide association study (GWAS).

FIGURE 1

Imputation panel performance of our ancestry‐matched panel. A, Imputation accuracy (R ²) comparison across MAF bins, or in other words, rare (MAF < 1%), low frequency (MAF between 1% and 5%), and common variants (MAF > 5%). B, Coverage plots showing proportion of high‐quality variants against their mean imputation accuracy values across gradients of low and common variants (depicted in different colors). C, MAF concordance calculated with ≈ 14.81 million overlapping variants between the imputed variants and the 1000G SAS dataset. 1000G, 1000 Genomes project; HRC, Haplotype Reference Consortium; MAF, minor allele frequency; SAS, South Asian ancestry; TLSA, Tata Longitudinal Study of Aging; TOPMed, Trans‐Omics for Precision Medicine.

FIGURE 2

Manhattan plots for genome‐wide independent signals, signals with P < 5 × 10⁻⁸, and genetic correlations. A, Manhattan plots for recognition memory, HDL‐C, LDL‐C, total cholesterol, triglycerides, TG:HDL, VAI and metabolic syndrome. Genome‐wide (P < 5 × 10⁻⁸) high confidence independent hits with minor allele frequency > 0.05, as reported in Table 1 are highlighted in red and some of the interesting hits are annotated with rsid and gene name. B, Genetic correlations between 29/30 cardiometabolic and cognitive phenotypes assessed. Significant positive correlations within cognition domains, but low or negative correlation between cardiometabolic and cognition related phenotypes can be observed. Gray boxes indicate that the correlation is non‐significant.  HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; TC, total cholesterol; TG, triglycerides; VAI, visceral adiposity index.

FIGURE 3

Adverse causal effect of high TG:HDL indicative of high insulin resistance on mean reaction time (attention domain). A, SNP exposure–SNP outcome association. Slope is indicative of positive causal effect where increase in TG:HDL increases mean reaction time, or in other words, adversely affects attention domain. B, Funnel plot showing no heterogeneity of effect estimates between the instrumental variables. HDL, high‐density lipoprotein; MR, Mendelian randomization; SNP, single nucleotide polymorphism; TG, triglyceride.

FIGURE 4

Distinct haplotype associations. A, Haplotype blocks constituting variants in AMIGO1. B, Haplotype blocks constituting variants in TM6SF2, APOE, CETP, ZPR1, and APOA5. The black outlined triangle depicts a haplotype block. The constituent variants in each block are also provided. The red triangles are shaded with lighter shades indicating lower LD and vice versa. The numbers in each triangle represents the pairwise LD between each of the variants. LD, linkage disequilibrium.