Male C57BL6/N mice were fed one of four diets for a period of 24 weeks and gene expression was examined in different portions of the intestines (Fig. 1A). The diets included a low-fat Vivarium chow (VIV) and three high-fat diets (HFDs) with 40% of calories derived from fat: a coconut oil diet (CO) composed of primarily saturated fats, specifically lauric acid (C14:0) and myristic acid (C12:0), with a small amount of soybean oil to provide the essential fatty acid linoleic acid (LA) (2% kcal); a soybean oil-enriched diet (SO + CO) with a high LA content (10% kcal); and a diet enriched in a genetically modified soybean oil (PL + CO) known as Plenish, which has a low LA (1.4% kcal) and high oleic acid content (~ 14% kcal) 28 (Supplementary Table S1). Previous analysis of these mice revealed that the soybean oil diet (SO + CO), and to a lesser extent the Plenish diet (PL + CO), induced obesity, diabetes, insulin resistance, and fatty liver, while the isocaloric CO diet had minimal adverse metabolic effects despite similar caloric intake as the other HFDs 3. RNA-seq was performed on a segment of each of the four tissues: duodenum (DUO), jejunum (JEJ), terminal ileum (TI), and proximal colon (PC). Differentially expressed genes (DEGs) were identified using DeSeq2, with statistical significance determined by a p-adjusted value of less than 0.05 and an absolute fold change greater than 2 (p-adj < 0.05 & Log2FC > 1.0). The DEGs (p-adj < 0.05) were further analyzed using network analysis in Cytoscape, incorporating the KEGG and Reactome databases (Fig. 1A).
HFDs alter gene expression in a differential fashion across the intestinal tract, including drug metabolism genes
Principal Components Analysis (PCA) of the 60 RNA-seq datasets revealed that the transcriptomes were primarily grouped based on tissue, with smaller variations observed between the dietary groups (Fig. 1B). Nonetheless, a considerable number of DEGs were identified when any of the three HFDs were compared to the VIV chow within a specific tissue (Fig. 1C). The duodenum (DUO) exhibited the greatest number of DEGs in all three HFD vs. VIV chow comparisons (CO: 513; SO + CO: 345; PL + CO: 483). The jejunum (JEJ) also had a substantial number of DEGs, albeit fewer than the duodenum (CO: 258; SO + CO: 179; PL + CO: 328), while the terminal ileum (TI) had a lower number of DEGs, except for the SO + CO vs. VIV chow (CO: 42; SO + CO: 189; PL + CO: 113). In contrast, the proximal colon (PC) displayed the largest number of DEGs in the CO vs. VIV comparison (CO: 293; SO + CO: 105; PL + CO: 68) (Fig. 1C). A Venn analysis revealed a moderate to minimal overlap in DEGs between the different HFDs and the VIV chow, ranging from 188 genes in the duodenum to 29 genes in the terminal ileum (Fig. 1D). These findings indicate that diets composed of different fats have distinct impacts on specific segments of the intestines.
Comparison between each of the three HFDs showed that CO vs. SO + CO consistently yielded the greatest number of DEGs (DUO: 198; JEJ: 118; TI: 22; PC: 75) (Fig. 1E). In contrast, CO vs. PL + CO exhibited a surprisingly low number of DEGs (ranging from 2 to 28) in all four tissues, except for the duodenum, which had 43 DEGs. Venn analysis of the pairwise comparisons between the HFDs revealed no overlap in DEGs among all three comparisons and relatively limited overlap between any two comparisons (Supplementary Figure S1).
Volcano plot analysis identified individual genes with significant fold change in various HFD vs. VIV chow comparisons, including several cytochrome P450 (Cyp) genes (Supplementary Figure S2). For example, Cyp2d26 was expressed at higher levels in the small intestines than the proximal colon and significantly upregulated by all three HFDs (Fig. 1F). In contrast, Cyp2c55 was expressed at much higher levels in the proximal colon than the small intestines and the HFDs tended to decrease expression, although it did not reach significance (Fig. 1G). Several other Cyp genes (Cyp4a10, Cyp4a31, Cyp4a32, Cyp4f15, Cyp2j6, Cyp2j9) were upregulated primarily in the duodenum by all three HFDs while a few genes were dysregulated in the jejunum by one or more HFDs (Cyp2u1, Cyp2c29, Cyp4f16) (Supplementary Figure S3). Expression of other Cyp genes as well as Phase 2 Ugt and Gst genes also varied across the intestines on the VIV chow and in response to the different HFDs, with a very modest impact on relatively few Phase 2 genes (e.g., Gstm1, Gsta4, Ugt1a9, Ugt1a7, Ugt2b36) and a greater impact on a number of Cyp genes (Supplementary Figure S4).
Differential expression of nuclear receptors across the intestinal tract and in response to HFD
Several members of the nuclear receptor (NRs) superfamily of ligand-dependent transcription factors are known to regulate CYP genes and play important roles in the development and function of the intestinal tract, as well as pathologies such as IBD and colon cancer 29,30. To determine their relative expression in different parts of the intestines we compared all 48 NRs across the four intestinal tissues in the mice fed VIV chow in a non row-normalized heatmap and included several non-NR transcription factors (TFs) known to play a role in intestinal physiology (Ctnnb1, Hnf1a, Hnf1b, Polr2a, Prox1, Tcf7l2). The most highly expressed NR gene throughout the intestines is hepatocyte nuclear receptor 4 alpha (Hnf4a) – its expression was greater than that of RNA polymerase 2 (Polr2a) and nearly as high as beta-catenin (Ctnnb1) – followed by the vitamin D3 receptor (Vdr), Hnf4g, and Rxra (Fig. 2A). This relative order was maintained across the three HFDs as well (Supplementary Figure S5). Some NR genes (e.g., Hnf4a, Nr1h4, Pparg) are expressed at lower levels in duodenum or jejunum, and at higher levels further along the intestinal tract while others (e.g., Hnf4g, Vdr, Nr0b2, and Ppara) have a relatively high level of expression in the beginning of the intestines and then decrease in the latter portions (Fig. 2BC). Others, such as Rxra, which is a heterodimeric partner for many other NRs, have a fairly consistent level of expression across the four tissues, decreasing only in the proximal colon (Fig. 2B).
Among the top four most highly expressed NRs, the only one that showed differential expression among the different diets was Hnf4a. Its expression in the duodenum was decreased in the intestines of mice fed any of the three HFDs compared to VIV chow (Fig. 2B). Nr0b2 (short heterodimeric partner, SHP) which acts as a transcriptional repressor, the bile acid receptor (FXR, Nr1h4) and the glucocorticoid receptor (GR, Nr3c1), which plays a critical role in the stress response, all showed a significant difference from VIV chow in one or more HFD in at least one portion of the intestines (Fig. 2B). In contrast, there was no significant difference in Ctnnb1 expression among the various diets, which is noteworthy as both HFD and mutations in the Wnt-Beta-catenin pathway are risk factors for colon cancer in humans (Fig. 2C) 31.
Finally, we examined the PPARs, which are known to play a role in the regulation of nutrient transport from the lumen into the body and have fatty acids as their ligands. While Ppard and Pparg did not show any significant difference in expression between diets within a given tissue, Ppara expression was significantly increased in the duodenum and jejunum in all three HFDs. It was also increased in CO vs PL + CO in the duodenum and in SO + CO or PL + CO vs VIV chow in the terminal ileum (Fig. 2B).
HFD impacts the expression of intestinal epithelial barrier function genes
Formation and maintenance of a healthy epithelial barrier is an important physiological function of the intestines. To analyze the effect of diet on intestinal barrier function, we used a list of 444 genes from NCBI (Supplementary Table S6) and identified 123 genes that are significantly dysregulated (p-adj < 0.05) between any two dietary groups (Fig. 3A-D). The duodenum had the greatest number of dysregulated genes (mostly downregulated) across the different diets (68 genes). Several genes exhibited lower levels of expression in one or more HFDs compared to the VIV chow in the duodenum – e.g., Ptk6 (Protein tyrosine kinase 6), Cldn10 (Claudin 10), Egf (epidermal growth factor). In contrast, Cd36 (cluster of differentiation 36, a long chain fatty acid transporter) showed increased expression in PL + CO vs VIV chow in the duodenum while NR co-activator Ppargc1a (PPARG Coactivator 1 Alpha) showed elevated expression in one of more HFD in all parts of the intestines except the jejunum (Fig. 3E). Considering that PGC1A is a co-activator of HNF4A and the PPARs 32, these diet-induced changes in Ppargc1 expression could amplify the effects of the HFDs on the NRs.
The jejunum, responsible for lipid digestion and absorption in the intestines, displayed a pattern where most of the 24 DEGs were between VIV chow and the three HFDs, with little difference between the HFDs (Fig. 3B). The exception was Scd1, which had much higher expression in the CO diet compared to the other HFDs and the VIV chow (Fig. 3F), consistent with the function of SCD1, a desaturase enzyme that introduces double bonds into saturated fatty acids.
The terminal ileum has the least HFD-dysregulated genes (18 DEGs) related to barrier function (Fig. 3C). The most dysregulated gene was Resistin-like molecule (RELM) β (Retnlb), a cysteine-rich cytokine that plays a role in insulin resistance, gastrointestinal nematode resistance, barrier integrity and susceptibility to inflammation 33. Retnlb expression was decreased by all three HFDs in the terminal ileum (as well as the duodenum) (see Fig. 6B). Since the terminal ileum is the region of the intestines that harbors many bacteria, viruses, and other pathogens, a downregulation in Retnlb caused by a HFD could weaken the body’s defenses. Another gene showing differential expression with HFDs in the proximal colon is the IBD susceptibility gene Ptpn11 (down in CO vs VIV and SO + CO), which encodes a tyrosine phosphatase involved in the homeostasis of epithelial barrier cells 34 (Fig. 3G).
HFD impacts the expression of genes associated with IBD and colon cancer
The expression of genes involved in IBD (141 genes) and colon cancer (192 genes) was also impacted by the HFDs (Fig. 4AB and Supplementary Figure S6). Interestingly, in terms of IBD-related genes, the terminal ileum was impacted the most by the HFDs, consistent with this portion of the gut being frequently inflamed in Crohn’s Disease, a form of IBD (Fig. 4A). Tlr2 (Toll-like receptor 2), Ripk3 (receptor interacting serine/threonine kinase 3), and Nox1 (NADPH oxidase 1) all decreased expression in the SO + CO and PL + CO diets compared to the VIV chow and CO diet. In contrast, Slc22a4 (a member of the solute carrier family), Vnn1 (vanin 1), Faah (fatty acid amide hydrolase), Ndfip1 (Nedd4 Family Interacting Protein 2), Maf (bZIP transcription factor) showed increased expression in the two soybean oil diets (Fig. 4A,C,E). Noteworthy IBD-related genes in the duodenum and/or jejunum that were affected by the HFDs include Duox2 (dual oxidase 2), a member of the NADPH oxidase family which was downregulated by the HFDs, and Ephx2 (epoxide hydrolase 2), which converts fatty acid epoxides to bioactive dihydrodiols, was upregulated by the HFDs (Fig. 4D).
The HFDs also affected the expression of cancer-related genes in the proximal colon (and other parts of the intestines) including Vnn1 (vanin 1), a pantetheinase with roles in oxidative stress and inflammation 35, and Tnfsf10 (tumor necrosis factor ligand superfamily, member 10) (Fig. 4E). Genes specific to colon cancer and altered only in the proximal colon include DNA repair enzymes Mgmt (O-6-Methylguanine-DNA Methyltransferase) and Parp1 (Poly(ADP-Ribose) Polymerase 1), Mtor, a mediator of response to cellular stress including DNA damage – all were downregulated by one or more HFDs. In contrast, Ly6a (Lymphocyte Antigen 6A), which regulates T cell proliferation, and Lgr5, a prominent marker for mitotically active crypt intestinal stem cells involved in the Wnt signaling pathway, were upregulated (Fig. 4F). Finally, there were several genes related to colon cancer that were altered by the HFDs but only in the small intestines. For example, Ido1 (indoleamine 2,3-dioxygenase 1) is the first and rate-limiting step in tryptophan catabolism and plays a role in antimicrobial and anti-tumor defense, neuropathology and immunoregulation, Casp3 (caspase 3) is a key executor of apoptosis and Lgals3 (galectin 3) plays a role in innate immunity and T-cell regulation and exhibits antimicrobial activity against bacteria and fungi. All three were downregulated by the HFDs (Fig. 4G).
Network analysis reveals an impact of HFDs on the immune system as well as metabolism
To obtain a more detailed understanding of the pathways impacted by HFDs in the different parts of the intestines, we conducted a Venn analysis of the DEGs in different diet comparisons in each tissue followed by Stringapp in Cytoscape to identify networks of genes, utilizing either the Reactome or the KEGG pathway databases (Fig. 5, Supplementary Figure S7). In the duodenum, genes upregulated in the CO vs. VIV comparison but not in SO + CO (C1) were involved in the metabolism of amino acids and lipids, as well in the transport of small molecule pathways (Fig. 5B). Additional metabolic categories, especially involving fatty acids, were identified in the PL + CO vs VIV comparison (Fig. 5C). In contrast, downregulated genes in the duodenum in the CO vs VIV comparison were associated with T cell receptor (TCR) signaling and the innate immune system, while genes down in the SO + CO vs VIV comparison were found in pathways related to pancreatic secretion, chemical carcinogenesis, linoleic acid metabolism and fat digestion and absorption (Supplementary Figure S7BC). Similarly, in the jejunum, there were many upregulated genes in HFD vs VIV, including fatty acid elongation, arachidonic acid metabolism and PPAR signaling and peroxisome (Fig. 5D) and fatty acid metabolism and Phase I genes (Supplementary Figure S7E). In contrast, as in the duodenum, the down regulated genes in the jejunum were related to the immune system, second messenger molecules, cytokine signaling and herpes simplex infection (Fig. 5E, Supplementary Figure S7FG). In contrast, the SO + CO vs CO comparison in the jejunum revealed upregulated genes associated with the immune system (B cell receptor signaling, hematopoietic cell lineage, cytokine-cytokine receptor interactions) as well as PPAR signaling and cell adhesion molecules (Fig. 5F). The same comparison in the proximal colon (SO + CO vs CO) also showed upregulated genes related to the immune system, ISG15 antiviral mechanism and scavenging of heme from plasma (Fig. 5G). In the SO + CO vs CO comparison, the duodenum yielded a completely different mix of upregulated metabolic pathways (including glycine, serine and threonine metabolism), fat digestion and absorption, pancreatic secretion, the renin-angiotensin system (RAS) and, intriguingly, GABAergic synapse and neuroactive ligand-receptor (Fig. 5H) as well as oxidative phosphorylation (Supplementary Figure S7D). Lastly, there was a network of genes up in the proximal colon in the SO + CO vs CO comparison involved in herpes simplex infection, RIG-I-like receptor signaling and cytosolic DNA sensing (Supplementary Figure S7H). There were no significant networks among the genes in the terminal ileum.
Impact of HFDs on the gut microbiome
Since HFDs are known to impact the microbiome 36, we generated a heatmap of microbiome-related genes that showed differential expression between any two diets (Fig. 6A). For example, Retnlb (resistin-like beta), which has antimicrobial properties, showed consistently high expression in the proximal colon compared to other tissues; it also showed decreased expression by one or more HFD in the duodenum and terminal ileum (Fig. 6B). Tlr2 (toll like receptor 2), a pattern recognition gene, and Nos2 (nitric oxide synthase 2), which plays a role in immunity against bacteria, fungi and viruses, were also decreased in one or more HFD in the terminal ileum and duodenum, respectively (Fig. 6B).
Microbiome analysis of the small intestine and colon for the HFDs and VIV chow revealed the presence of many species of bacteria, with their relative abundance influenced by the diet (Fig. 6C). Importantly, there was an increase in populations of various pathogenic and opportunistically pathogenic bacteria in both the small intestines and the colon in the HFDs compared to VIV chow – Ureaplasma cati, Turicibacter sp. and Erysipelatoclostridium sp. in the small intestines and Enterobacteriaceae in the colon 37–40. There was also a notable decrease in bacteria with the HFDs that are typically considered to be beneficial (although their impact on host health is not fully understood yet) – Segmented filamentous bacteria (SFB) in the small intestines and Prevotella oris in the colon 41,42.
Impact of HFDs on the expression of genes involved in COVID-19
Although COVID-19 primarily affects the respiratory system, it can also impact the intestinal tract, leading to diarrhea, inflammation and septic shock 43. Furthermore, patients with COVID-19-related diarrhea are more likely to require hospitalization and experience a more severe infection 43. Heatmaps revealed several COVID-19-related genes that were dysregulated by one or more of the HFDs (Fig. 7A-D), including Ace2 (angiotensin-converting enzyme 2) and Enpep (glutamyl aminopeptidase) (Fig. 7E). In the proximal colon, both genes exhibited a significant increase in expression in HFDs compared to VIV chow. Slc6a19 (solute carrier family 6 member 19) showed increased expression in the terminal ileum in PL + CO vs. VIV chow (Fig. 7F). In contrast, Tmprss2 (transmembrane Serine protease 2), Gzma (granzyme A), Irf1 (interferon regulatory factor 1), Stat1 and Stat3 (signal transducer and activator of transcription 1/3) displayed decreased expression in one or more HFDs compared to VIV chow in various sections of the intestines (Fig. 7G). Moreover, two COVID-19-related genes, Klk1 and Klk1b5, identified in the Renin-angiotensin system (RAS) in the network analysis (Fig. 5H), were upregulated by the CO diet in the duodenum (Fig. 7H). Kallikreins are serum serine proteases that play an important role in the vascular system and have been proposed as therapeutic targets for COVID-19 44,45.
To further investigate the impact of HFDs on intestinal health during COVID-19, we utilized the BioGRID database 46 to identify interactions between host proteins/genes dysregulated by the HFDs and viral proteins of SARS-CoV-2, the causative agent of COVID-19. These interactions involved ACE2, TMPRSS2, SREBPF1 with the viral S protein; ACCA2, STAT3, and SREBPF1 with the viral M protein; and STAT1, FASN, and SREBPF1 with the viral NSP proteins (Fig. 7I). The expression of Srebf1 (sterol regulatory element binding transcription factor 1), was significantly increased in the duodenum and jejunum in response to HFDs (Fig. 7J).