Obesity affects brain cortex gene expression in an APOE genotype and sex dependent manner

Abstract

Objective: Obesity is the top modifiable risk factor for Alzheimer's disease. We hypothesized that high fat diet (HFD)-induced obesity alters brain transcriptomics in APOE-genotype and sex dependent manners. Here, we investigated interactions between HFD, APOE, and sex, using a knock-in mouse model of the human APOE3 and APOE4 alleles.

Methods: Six-month-old APOE3-TR and APOE4-TR mice were treated with either HFD or control chow. After 4 months, total RNA was extracted from the cerebral cortices and analyzed by poly-A enriched RNA sequencing on the Illumina platform.

Results: Female mice demonstrated profound HFD-induced transcriptomic changes while there was little to no effect in males. In females, APOE3 brains demonstrated about five times more HFD-induced transcriptomic changes (399 up-regulated and 107 down-regulated genes) compared to APOE4 brains (30 up-regulated and 60 down-regulated). Unsupervised clustering analysis revealed two gene sets that responded to HFD in APOE3 mice but not in APOE4 mice. Pathway analysis demonstrated that HFD in APOE3 mice affected cortical pathways related to feeding behavior, blood circulation, circadian rhythms, extracellular matrix, and cell adhesion.

Conclusions: Female mice and APOE3 mice have the strongest cortical transcriptomic responses to HFD related to feeding behavior and extracellular matrix remodeling. The relative lack of response of the APOE4 brain to stress associated with obesity may leave it more susceptible to additional stresses that occur with aging and in AD.

Figure 1.. APOE3/4 TR mouse model.

(A) Early obesity model timeline showing APOE-TR mice in both APOE3 and APOE4 groups switched to High Fat Diet (HFD) at 6 months of age. After 4 months of diet treatment, mice were euthanized and brains were collected. (B) After 4 months of HFD treatment (n=4 per group), females showed significant weight gain in both APOE3 (p=0.016) and APOE4 (p=0.020) genotypes. In males, APOE4 mice trended toward a significant weight gain (p=0.058) while APOE3 mice showed no significant weight gain (p=0.21). Two-tailed student t-test with equal variance was performed. Control: normal chow diet; HFD: 45% fat with high fructose, p value <0.05 (*) considered significant, na = no significant difference.

Figure 2.. RNA sequencing workflow and PCA analysis.

After euthanasia, brains were first perfused with PBS; cortex, hippocampus, and cerebellum were dissected, and snap-frozen. (A) Total cortical RNA was extracted in Trizol, purified using column-based RNA purification, followed by poly-A library preparation, and 150 bp paired sequencing using the Illumina platform. (B) After pre-processing and normalization of raw data, principal component analysis (PCA) was performed. PCA analysis showed distinct male (blue) and female (red) clusters. Additional PCA analysis showed females exhibiting distinct clustering of both E3 and E4 groups with or without HFD treatment (C); this distinct clustering was not present in males (D). CON: normal chow diet; HFD: 45% fat with high fructose.

Figure 3.. Female sex and APOE3 genotype exhibit more differentially expressed genes (DEGs) in HFD group.

DEseq analysis of the transcriptome of both APOE genotypes in female and male mice. (A) MA plots of female group of mice showing significant up-regulated (red) and down-regulated (green) DEGs in APOE3 (506 DEGs, upper panel) and APOE4 (90 DEGs, lower panel). Each MA plot is accompanied with a representative heat map on the right with the numbers of DEGs (blue: down-regulated genes; yellow: up-regulated genes). (B) MA Plots of male group of mice showing up-regulated and down-regulated DEGs in APOE3 (6 DEGs) and APOE4 (1 DEG). Cutoff FC=1.5 and FDR=0.05 were applied for the analysis, and DEGs passing an adjusted q-value of <0.05 were considered as significant DEGs. For MA plots and heat maps, up-regulated DEGs are represented in red color and down-regulated genes are represented in green color.

Figure 4.. Clustering analysis revealed APOE4 genotype specific impact on transcriptome.

K-mean unsupervised clustering analysis was performed on the top 2000 most variable genes ranked by standard deviation, and grouped into 4 clusters based on expression across experimental groups. (A) Heatmap showing 4 clusters, cluster A (588 genes), cluster B (302 genes), cluster C (109 genes), and cluster D (1001 genes). Heatmap expression annotation: Red= increased, Green= decreased. (B) The top 15 biological processes (GO:BP) are presented in network annotation for cluster B, which showed those genes which induced in APOE3 upon HFD treatment but are constitutively high in APOE4 genotype. (C) Top 15 biological processes (GO:BP) are presented in network annotation for cluster D, which showed those genes which induced in APOE3 upon HFD treatment, but constitutively low in APOE4 genotype (and unresponsive to HFD in APOE4). Biological processes (GO:BP) were ranked based on q-value after FDR = 0.1. Two nodes are connected in network if 30% or more genes are being shared between two biological processes. Red: up-regulated; Green: down-regulated (none).

Figure 5.. Shared APOE DEGs, including HFD interactions.

APOE differentially expressed genes (DEGs) were identified by comparing APOE4 and APOE3 genotypes including both sexes (male and female). Analysis was performed for control diet (left, blue) and HFD (right, red) separately. We identified 26 constitutive APOE DEGs present in both the control and HFD environments (center). We further identified DEGs which are unaffected by HFD (n=25) and DEGs interacting with HFD only (n=10). In each list, down-regulated APOE4 DEGs (compared to APOE3) are blue, and up-regulated DEGs are in orange. All identified DEGs are considered significant after adjusting FDR to 0.1 with fold change ≥ 1.5.