Open chromatin profiling in mice livers reveals unique chromatin variations induced by high fat diet

Abstract

Metabolic diseases result from multiple genetic and environmental factors. We report here that one manner in which environmental factors can contribute to metabolic disease progression is through modification to chromatin. We demonstrate that high fat diet leads to chromatin remodeling in the livers of C57BL/6J mice, as compared with mice fed a control diet, and that these chromatin changes are associated with changes in gene expression. We further show that the regions of greatest variation in chromatin accessibility are targeted by liver transcription factors, including HNF4α, CCAAT/enhancer-binding protein α (CEBP/α), and FOXA1. Repeating the chromatin and gene expression profiling in another mouse strain, DBA/2J, revealed that the regions of greatest chromatin change are largely strain-specific and that integration of chromatin, gene expression, and genetic data can be used to characterize regulatory regions. Our data indicate dramatic changes in the epigenome due to diet and demonstrate strain-specific dynamics in chromatin remodeling.

Keywords: Chromatin; Diet; Gene Regulation; Gene-environment Interaction; Liver Metabolism; Metabolic Disease.

FIGURE 1.

HF diet chromatin remodeling. A, number of enriched regions of FAIRE for HF and control diets and genomic distribution by region. B, scatter plot of read counts at open chromatin sites. Shown are read counts (log 2) for each open chromatin site for control (x axis) and HF (y axis). The top 1000 open chromatin sites displaying the greatest -fold change in target counts are plotted in red. Con Tag, control tag. C, enriched Kegg Pathway terms for genes proximal to most variable FAIRE sites. D, scatter plot of distances between the 2000 most variable FAIRE sites and differentially expressed genes by their relative expression changes. 121 genes are within 200 kb (p < 0.0001). E, genome browser tracks displaying the Lpin1 locus with FAIRE-seq and RNA-seq tracks from livers of control and HF-fed mice, as well as Lpin1 eQTL SNPs (red). HF-specific FAIRE sites (highlighted yellow) overlap with eQTL SNPs associated with Lpin1. chr12, chromosome 12.

FIGURE 2.

Sites displaying increase in accessibility in B6 livers with HF diet are regulatory regions targeted by HNF4α, FOXA1, and CEBP/α. A, the top 10 motifs enriched in top 1000 variable sites in B6 genome. B, number of sites in the top 1000 variable sites in the B6 genome bound by HNF4α and CTCF. C, FAIRE-seq and ChIP-seq tracks for control and HF diets at the Lpin1 locus showing overlap of enrichment between FOXA1, CEBP/α, HNF4α, and H3K4me1 at variable chromatin sites. D, -fold change (HF read counts/control read counts) of B6 FAIRE sites classified by TF binding. Box plots show the median value, and whiskers show distribution of first and third quartile (*indicates p value < 0.0001).

FIGURE 3.

Sites with increased chromatin accessibility also display local increase in H3K4me1 enrichment. A, gene expression (FPKM) for down-regulated and up-regulated genes in control (Con) and HF fed livers. B, enrichment of H3K4me1 fragments at promoters of down-regulated (199 genes), up-regulated (145 genes), and random (200 genes) genes in control and HF diet livers. (* indicates p value < 0.05). Box plots show the median value, and whiskers show distribution of first and third quartile. C, left, heat map of FAIRE and H3K4me1 densities at the union FAIRE sites for control and HF fed livers ranked by -fold change. Center, aggregate plots for FAIRE read counts and H3K4me1 read counts from four sets of 1000 sites: Random, 1000 randomly chosen site; Top1k up, most accessible in HF as compared with control; Top1k down, most accessible in control as compared with HF. Right, box plots of -fold change (HF/control) H3K4me1 read density from each of the set of sites. Box plots show the median value, and whiskers show distribution of first and third quartile (* indicates p value <0.01).

FIGURE 4.

B6 and D2 display strain-specific diet-induced chromatin variations. A, Oil Red O (left) and H&E (right) staining from livers of B6 control, B6 HF, D2 control, and D2 HF livers. B, Venn diagram displaying overlap of B6 control-specific (not in HF) FAIRE sites and D2 control-specific sites (top), and between B6 HF-specific (not in control) FAIRE sites and D2 HF-specific sites (bottom). C, FAIRE-seq tracks displaying variations in chromatin accessibility between strains and diets. D, the top 10 motifs enriched in top 1000 variable sites in the D2 genome. E, left, heat map of normalized read counts at FAIRE sites for each group (see “Results” for details). Center, a schematic of sites in each group. Right, box plots of SNVs from each group as compared with randomly chosen sites of the same size and coverage. Box plots show the median value, and whiskers show distribution of first and third quartile (* indicates p value < 0.001).

FIGURE 5.

Integration of chromatin accessibility, eQTL, and gene expression data identifies regions of regulatory variation. A, overlapping eQTL, chromatin accessibility, and gene expression changes identify 44 genes with regulatory variation in the liver between B6 and D2. OC sites, open chromatin sites. B, Rgs16 is differentially expressed between livers of B6 and D2 mice and has 281 eQTL SNPs associated with expression in the HMDP. Of these, 3 overlap a region of variable chromatin between B6 and D2 (highlighted). C, genome browser tracks displaying the Lpin1 locus with RNA-seq and FAIRE-seq tracks from control and HF-fed mice, with Lpin1 eQTL SNPs in red. The B6-specific, HF-specific FAIRE site is highlighted in yellow. D, schematic of luciferase (Luc) constructs to examine enhancer activity of B6 Lpin1 enhancer (Enh.) variant and D2 Lpin1 enhancer variant (top). Bar graph shows relative luciferase activity, normalized to control construct. Error bars represent S.E. (*, p value <0.05, Student's t test, n = 3).