Page 37 - Read Online

Qu et al. J Transl Genet Genom 2023;7:3-16 https://dx.doi.org/10.20517/jtgg.2022.16 Page 7

sample. Comparison between biological duplicates showed mostly overlapping H3K27ac peaks [Figure 1].

A representative browser view of the H3K27ac ChIP-Seq data signal is shown in Figure 2. Gained VELs
were defined as VELs with increased H3K27ac enrichment when comparing HFD profiles to LFD profiles.
Lost VELs were defined as VELs with a decrease in H3K27ac enrichment when comparing HFD profiles to
LFD profiles.

Principal component analysis showed that the H3K27ac profiles of HFD samples were highly distinct from
those of LFD samples [Figure 3A]. Using BEDTools, the H3K27ac peak profiles of biological replicates were
merged: 93,309 peaks were identified for WT LFD, 103,430 for WT HFD, 98,706 forApc Min/+ LFD, and 98,732
forApc Min/+ HFD. When directly comparing the merged H3K27ac profile of HFD samples to that of LFD
samples for each genotype using DESeq2, a total of 45,910 VELs were identified for WT mice, 1536 of which
were statistically significant (3.35%), and 48,646 VELs were identified forApc Min/+ mice, 1427 of which were
statistically significant (2.93%) [Figure 3B]. For WT mice, the log (fold-change) of 967 gained VELs was > 1,
2
and 173 lost VELs were < -1. ForApc Min/+ mice, the log (fold-change) of 824 gained VELs were > 1, and of 270
2
lost VELs were < -1 [Figure 3B]. In contrast to the impact of diet type, the epigenetic impact of genotype
was not as significant. According to principal component analysis, only a small variance was demonstrated
between WT HFD samples andApc Min/+ HFD samples [Figure 3A]. Additionally, DESeq2 did not identify
any significant VELs when comparing WT HFD profiles toApc Min/+ HFD profiles or when comparing WT
LFD profiles toApc Min/+ LFD profiles. Overall, HFD, in only three days, was able to cause epigenetic changes
within the intestinal epithelia.

GREAT found multiple genes predicted to be associated with the identified VELs. In WT mice, 1660 genes
were predicted to be associated with gained VELs, and 345 genes were predicted to be associated with lost
VELs; InApc Min/+ mice, 1434 genes were predicted to be associated with gained VELs, and 590 genes were
predicted to be associated with lost VELs. The distributions of the VELs relative to the transcription start
site (TSS) of annotated genes were relatively similar regardless of genotype. More than 90% of the VELs
were localized more than 5 kb away from the nearest TSS, suggesting that they predominantly function as
enhancers.

RNA samples of adequate quality were isolated from the same epithelial samples used to obtain ChIP-Seq
epigenomic profiles to generate RNA-Seq transcriptomic profiles [Supplementary Figure 1]. Similar to the
epigenomic profiles, principal component analysis of the transcriptomic profiles showed a significant
distinction between HFD samples and LFD samples and a less obvious distinction between similar WT
samples andApc Min/+ samples when controlling for diet type [Figure 4A]. The transcriptomic profiles of the
different sample types were directly compared to identify all DEGs. Many DEGs were identified when
directly comparing HFD samples to LFD samples [Figure 4B]. Relatively few DEGs were identified when
directly comparingApc Min/+ samples to WT samples while controlling for diet [Supplementary Figure 2].
Overall, the RNA-Seq results were similar to the ChIP-Seq results in that diet type had a much more
significant overall impact than differences in genotype.

Impact of HFD on lipid metabolism supported by epigenomic and transcriptomic changes
Multiple KEGG pathway enrichments were identified from HFD-induced gained VELs revealed by
ChIP-Seq and HFD-induced upregulated genes revealed by RNA-Seq. Most of these pathways and
processes were involved in lipid metabolism, including the “PPAR signaling pathway”, “AMPK signaling
pathway”, “biosynthesis of unsaturated fatty acids”, “fatty acid elongation”, and “fatty acid degradation”
[Figure 5].