Impact of Various High Fat Diets on Gene Expression and the Microbiome Across the Mouse Intestines

Abstract High fat diets (HFDs) have been linked to several diseases including obesity, diabetes, fatty liver, inflammatory bowel disease (IBD) and colon cancer. In this study, we examined the impact on intestinal gene expression of three isocaloric HFDs that differed only in their fatty acid composition – coconut oil (saturated fats), conventional soybean oil (polyunsaturated fats) and a genetically modified soybean oil (monounsaturated fats). Four functionally distinct segments of the mouse intestinal tract were analyzed using RNA-seq – duodenum, jejunum, terminal ileum and proximal colon. We found considerable dysregulation of genes in multiple tissues with the different diets, including those encoding nuclear receptors and genes involved in xenobiotic and drug metabolism, epithelial barrier function, IBD and colon cancer as well as genes associated with the microbiome and COVID-19. Network analysis shows that genes involved in metabolism tend to be upregulated by the HFDs while genes related to the immune system are downregulated; neurotransmitter signaling was also dysregulated by the HFDs. Genomic sequencing also revealed a microbiome altered by the HFDs. This study highlights the potential impact of different HFDs on gut health with implications for the organism as a whole and will serve as a reference for gene expression along the length of the intestines.


Introduction
Over the last century, the dietary pattern in the U.S. has gradually shifted from a healthy diet to one with increased fat and decreased ber.Along with an increase in the amount of fat being ingested, there has also been a change in the type of fat being consumed by Americans, with seed oils, high in polyunsaturated fatty acids (PUFAs), becoming the predominant source of dietary fat.In fact, the component of the U.S. diet that has increased the most over the last century is soybean oil 1 .The major fatty acid component of soybean oil is the PUFA linoleic acid (LA, C18:2, omega-6).Our lab and many others have shown that high-fat diets (HFDs) can be linked to several diseases, such as obesity, diabetes, insulin resistance, fatty liver and susceptibility to in ammatory bowel disease (IBD) in both mice and humans [2][3][4][5] .There are also many studies describing the impact of HFDs on the gut microbiota 6,7 , physiological changes in the small intestine 8 , intestinal permeability and gastrointestinal diseases 9 .However, most gene expression studies analyze only one portion of the intestines or one type of HFD at a time [10][11][12] .
Here, we used RNA-seq to examine the impact of three HFDs on gene expression in four functionally distinct segments of the mouse intestinal tract: the duodenum, jejunum, terminal ileum and proximal colon.The duodenum is responsible for breaking down the stomach acid and food mixture, while the jejunum absorbs sugars, amino acids, and fatty acids.The terminal ileum absorbs remaining nutrients, such as vitamin B12 and bile acids, and the proximal colon is the primary site for absorption of water and salts and microbial production of short chain fatty acids (SCFAs).All four parts of the intestine are also involved in xenobiotic and drug metabolism 13 .
The HFDs used in this study are comparable to the current American diet in that they consist of 40% of calories from fat and are low in ber.The rst diet was formulated with coconut oil (saturated fat), the second with soybean oil (polyunsaturated fat, PUFA) and the third with a genetically modi ed soybean oil with a fatty acid composition similar to olive oil (monounsaturated fat).Each diet was compared to a low fat (13 kcal% fat), high-ber vivarium chow as well as to each other.RNA-seq analysis revealed dysregulation of several nuclear receptor genes and other transcriptional regulators as well as xenobiotic/drug metabolism genes throughout the small and large intestines.There was also signi cant dysregulation of genes involved in epithelial barrier function, IBD and colon cancer.Network analysis showed an upregulation in metabolism genes and, interestingly, a downregulation in numerous genes involved in the immune system, particularly those related to bacterial and viral infections, including SARS-CoV-2, the pathogen responsible for the global COVID-19 pandemic.The expression of several genes related to signaling by neurotransmitters and the microbiome was dysregulated, and genome sequencing revealed alterations in the gut bacteria by the HFDs.Taken together, our results reveal a signi cant impact of dietary fat on the intestinal transcriptome and microbiome which could potentially contribute to disease as well as brain health.

Animals
Care and treatment of animals was in accordance with guidelines from and approved by the University of California, Riverside Institutional Animal Care and Use Committee (AUP #20140014).All the methods reported are in accordance with ARRIVE guidelines.The animals were treated as previously described 3 .Brie y, male C57BL/6N mice were weaned at three weeks of age and assigned randomly to one of the four diets for 24 weeks -low fat Vivarium (VIV) chow; 40% kcal fat coconut oil (CO, 36 kcal% from coconut oil and 4 kcal% from soybean oil to provide the essential fatty acids LA and ALA), 40% kcal CO plus soybean oil (SO + CO, 21 kcal% fat calories from coconut oil and 19 kcal% from soybean oil, resulting in 10 kcal% from LA, comparable to the amount of LA in the current American diet 14 ); 40% kcal CO plus genetically modi ed soybean oil (Plenish) which has low LA (PL + CO, conventional soybean oil was replaced on a per gram basis with the genetically modi ed (GM) High Oleic Soybean Oil Plenish (DuPont Pioneer, Johnston, IA).See Supplementary Table S1 for a comparison of the diets and Deol et al., 2017 3 for the complete composition of the diets.Metabolic parameters of the mice, including body weight, glucose tolerance, insulin resistance and fatty liver were reported previously 3 .At the end of the study, animals were euthanized by CO 2 inhalation followed by cervical dislocation.Intestinal tissue was excised immediately and put in RNALater for 24 hours at room temperature and then stored at -80°C.

RNA-seq
The tissues for RNA sequencing (RNA-seq) were duodenum (DUO, 1 cm immediately downstream of the gastroduodenal junction), jejunum (JEJ, 1 cm at the approximate middle of the remainder of the small intestine), terminal ileum (TI, 1 cm immediately upstream of the ileo-cecal junction), and proximal colon (PC, 1 cm immediately downstream of the ileo-cecal junction).RNA extraction from each of the four portions of the intestines for three VIV chow-fed mice and four mice on each of the three HFDs (60 total samples) was carried out as previously described 2 .Total RNA was isolated from samples of each tissue (DUO, JEJ, TI and PC) using a miRNeasy kit (Qiagen, Inc., Valencia, CA) and evaluated for purity and concentration by NanoDrop (Wilmington, DE) and Agilent 2100 Bioanalyzer (Santa Clara, CA).Poly(A) + RNA (4 µg) with an RNA Integrity Number (RIN) of 7.8 or higher was used to construct sequencing libraries with the TruSeq Long RNA Sample Prep Kit (Illumina, San Diego, CA).RNA libraries were validated for RNA integrity by Bioanalyzer, pooled in equimolar amounts, and sequenced on an Illumina HiSeq 2000 at the UCR Genomics Core to generate 50 bp paired-end reads.Three biological replicates were sequenced for the Vivarium Chow diet (VIV) and four for all the HFDs (CO, SO + CO, and PL + CO).On average ~ 16 million reads were acquired for each biological replicate.The raw data are publicly available in Gene Expression Omnibus (GEO), accession number GSE220302.

Differential gene expression analysis of RNA-seq data
Reads were aligned to the mouse reference genome (mm10) with STAR v2.5.0a using default parameters 15 .Raw read counts were calculated with STAR using the GeneCounts option of the quantMode parameter since the libraries were unstranded.Library normalization was performed with EDASeq 16 ; within-lane normalization on GC content was performed with the LOESS method and between-lane normalization was performed with the non-linear full quantile method.Normalization factors from EDASeq were used for differential expression analysis with DESeq2 17 .Normalized read counts, FPKM (fragments per kilobase per million), and r-log (regularized log transformation) results were generated for downstream analysis.
The list of genes used in the heatmaps for nuclear receptors, epithelial barrier, IBD, colon cancer, microbiome and COVID-19 were obtained from the NCBI website (Supplementary Table S6).Differentially expressed genes (DEGs) between any two diets (p-adj ≤ 0.05) were identi ed in the RNA-seq data and displayed in the respective heatmaps, generated using the Pheatmap package in R 18 and row-normalized before plotting, unless noted otherwise.Python library "Plotly" was used to generate scatter plots for each individual gene 19 .PCA analysis, bar plots and Venn diagrams were created using the Python library 'matplotlib'.Volcano plots were generated using the ggplot2 package from R 20 .Colored spots are DEGs with (p-adj ≤ 0.05 & abs(Log2FC) ≥ 0.05).Genes symbols for outliers are highlighted and have a value in the top 95% in -Log10(p-adj) and abs(Log2FC) ≥ 1.5 in the given comparison.StringApp 21 from Cytoscape (Version 3.8.2) 22 was used to analyze and visualize potential interactions between DEGs among the different diets and tissues in the KEGG 23 and Reactome pathways 24 (FDR ≤ 0.05); a medium interaction score of 0.4 (out of 0 to 1) in the StringApp was required.Mouse Genome Informatics (MGI) and GeneCards: The Human Gene Database were used to identify the full name of a gene, as well as its functions and associated diseases 25,26 .

Microbiome analysis
The bacterial collection protocol, DNA extraction and bacterial rRNA internal transcribed spacer (ITS) analysis was performed as previously described 27 except that bacteria were collected from the small intestine or colon of male mice fed the different diets (VIV, SO, SO + CO, PL, PL + CO) for 24 weeks -the same ones used for the RNA-seq.Only the top 12 genus-level of operational taxonomic units (OTU) were plotted as mean percentage compositions for each treatment group; the remaining OTUs were combined under "Other".DNA sequencing data of the microbiome is publicly available at SRA BioProject, Accession #PRJNA615924.

Results
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, speci cally 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 oilenriched diet (SO + CO) with a high LA content (10% kcal); and a diet enriched in a genetically modi ed 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 identi ed using DeSeq2, with statistical signi cance 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 identi ed when any of the three HFDs were compared to the VIV chow within a speci c 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 ndings indicate that diets composed of different fats have distinct impacts on speci c 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 identi ed individual genes with signi cant 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 signi cantly 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 signi cance (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 signi cant difference from VIV chow in one or more HFD in at least one portion of the intestines (Fig. 2B).In contrast, there was no signi cant 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 signi cant difference in expression between diets within a given tissue, Ppara expression was signi cantly 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 identi ed 123 genes that are signi cantly 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 in ammation 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 in amed 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), Nd p1 (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 in ammation 35 , and Tnfsf10 (tumor necrosis factor ligand superfamily, member 10) (Fig. 4E).Genes speci c 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 rst 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 identi ed 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 signi cant 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 (resistinlike 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 in uenced 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][38][39][40] .There was also a notable decrease in bacteria with the HFDs that are typically considered to be bene cial (although their impact on host health is not fully understood yet) -Segmented lamentous 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, in ammation 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 signi cant 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, identi ed in the Reninangiotensin 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 signi cantly increased in the duodenum and jejunum in response to HFDs (Fig. 7J).

Discussion
To our knowledge, this is the rst comprehensive RNA-seq analysis conducted in four different sections of the intestines (duodenum, jejunum, terminal ileum and proximal colon) and comparing three distinct HFDs to a standard low-fat diet.The HFDs are unique in that they are not the standard diet made from excessive amounts of lard (50-60 kcal% fat) as is typically used in rodent studies.Rather, they are formulated with an amount of fat closer to that consumed by Americans (40 kcal%) 47 and using the most prevalent cooking oil used in the United States, soybean oil (SO) which is high in the polyunsaturated fat LA (C18:2), a genetically modi ed soybean oil Plenish (PL) low in LA and high in the monounsaturated fat oleic acid (C18:1) that is found in olive oil, and coconut oil (CO) consisting of saturated fat.Signi cant differences between diets within tissues were observed, with different diets impacting the expression of different genes in different parts of the intestines.Interestingly, the SO + CO diet resulted in a greater number of dysregulated genes compared to the CO diet in a variety of different pathways in different parts of the intestines than did the PL + CO diet, suggesting that excess LA has a greater impact than oleic acid (Figs. 1, 5), consistent with differential effects of SO and PL we have observed previously in terms of obesity, diabetes and colitis 3,4 The majority of dysregulated genes can be grouped into one of two categories -metabolism (generally increased) and the immune system (typically decreased) -and are associated with various pathological conditions and diseases ranging from colon cancer, in ammation and IBD to leaky gut and infectious diseases including COVID-19.There were also several genes involved in the metabolism or transport of neurotransmitters -including endocannabinoids, dopamine and serotonin, gamma-aminobutyric acid (GABA), glutamate and glycine -that are dysregulated by the HFDs and could impact brain health.Lastly, we observed changes in a number of transcriptional regulators -including NRs, IRFs, STATs and SREBP1 -that could play a role in regulating the expression of the genes in the other categories (Fig. 8A).

HFDs impact expression of intestinal genes involved in fatty acid and drug metabolism
Perhaps the best example of a gene involved in fatty acid metabolism that is impacted by diet is Scd1 which converts saturated fatty acids to monounsaturated fatty acids; it is upregulated by CO more than 10-fold in the jejunum (Fig. 3).Other genes include those that impact linoleic acid and its downstream metabolite arachidonic acid which is associated with pro-in ammatory processes (e.g., Cyp2c, Cyp2j, Cyp4a, Ephx2) (Fig. 5).Ephx2 converts linoleic and arachidonic acid epoxides into bioactive oxylipins; we recently showed that a diet high in SO leads to increased levels of these oxylipins in the intestines and correlates with barrier dysfunction and susceptibility to colitis in mice 4 .Changes in genes involved in amino acid metabolism (Fig. 5) were less anticipated given that the diets all contained the same amount of protein but an intriguing nding nonetheless as they could play a role in select signaling pathways as noted below.
Dysregulation of numerous genes involved in xenobiotic and drug metabolism -Cyp, Gst, Ugt -is consistent with the notion that diet impacts Phase I and Phase II reactions in the liver 48 (Supplementary Figures S2,3,4).While we previously reported effects of the CO and SO diets on Cyp gene expression in the liver 2 and others have reported varying ndings in terms of which Cyp genes are expressed where in the intestines [49][50][51][52] , to our knowledge this is the rst report of different cooking oils impacting Cyp, Gst and Ugt gene expression in different parts of the intestines.Cyp2d26, for example, is upregulated by more than one HFD: its ortholog in humans, CYP2D6, is known to metabolize numerous drugs including antidepressants, antipsychotics, analgesics, antitussives, beta-adrenergic blocking agents, antiarrhythmics, and antiemetics 25,53 (Fig. 1).These results suggest that the intestines may play a more signi cant role in drug metabolism than previously recognized and that there could be important health consequences if a basic component of one's diet -such as cooking oil -changes.HFDs impact expression of intestinal genes involved in the immune system, the microbiome and neurological signaling Given that the intestine is the rst line of defense against many foreign invaders and plays a critical role in immune function, it is notable that we identi ed many genes linked to the immune system that were downregulated by one or more HFD.For example, we observed dysregulation of genes involved in innate immunity (e.g., Retnlb, Reg3b), cytokine signaling (e.g., Ccl8, Ccl20, Ccl22, Tnfsf10), and pattern recognition (e.g., Tlr2, Tlr3) in response to the different HFDs, even without exposure to an external pathogenic agent.Retnlb and Reg3b both have antibacterial properties and Reg3b is regulated by Retnlb 33,54-56 (Figs.4-6).
We also found that the HFDs altered the gut microbiome (Fig. 6).While the effects of diet on the microbiome are well established, especially in terms of ber and polyphenols 57,58 , less well studied are the effects of different dietary fatty acids.For example, we observed an increase in Enterobacteriaceae in the small intestine, a group of organisms known to enhance the in ammatory response 59 .Further investigation is required to determine whether changes in the microbiome are a direct result of the diets or, alternatively, are a result of changes in the host immune system by the diet which subsequently contributes to dysbiosis.We reported recently that fatty acids such as linoleic acid can contribute to the growth of certain pathogenic bacteria in vitro, analogous to changes observed in vivo on an SO diet high in linoleic acid 4 .
Host genes that are implicated in the tryptophan-serotonin pathway, which is known to be impacted by the gut microbiota, were also dysregulated by the HFDs.For example, Ido1 encodes a key tryptophanmetabolizing enzyme that generates the neurotransmitter serotonin and was downregulated by the HFDs (Fig. 4).In contrast, several neurotransmitter transporters were upregulated by one or more HFD compared to the VIV chow -glutamate transporter Slc1a3, dopamine transporter Slc6a3, serotonin transporter Slc6a4 (Fig. 8B).Faah, which was greatly upregulated by all three HFDs (Fig. 4), is a hydrolase for endocannabinoids and N-acylethanolamines such as 2-arachidonoylglycerol (2-AG), Narachidonoylethanolamine (AEA) 60 .This suggests that the HFDs might result in decreased levels of endocannabinoids in the gut which is what we observed with a soybean oil diet 4 .Although only a total of 55 genes were dysregulated between the SO + CO and PL + CO diets (DUO: 47 genes; JEJ: 2 genes; TI: 5 genes; PC: 1 gene), several of the high LA soybean oil-speci c genes were involved in neurotransmitter signaling -glycine transporter Slc6a9, GABA receptor Gabra4, and Gatm, an amidinotransferase involved in creatine biosynthesis critical for cognition, language and behavior.All were signi cantly downregulated in SO + CO compared to low LA/high oleic acid diet (PL + CO) (Supplementary Tables S2-S5, Fig. 8C).
Taken together, the ndings from the gene expression and microbiome analyses are consistent with the notion that the gut-microbiome-brain axis may be in uenced by what we eat and affect our brain health 57 .Indeed, we have previously reported that the same diets as used in this study impact the transcriptome of the hypothalamus and many of those genes are related to mental health 61 .

HFDs impact expression of genes involved in transcription regulation
One potential mechanism by which different dietary fats could alter the expression of so many genes in the intestines is via nuclear receptors (NRs) which respond to hydrophobic ligands, including fatty acids.While assessing the impact of the dietary fats on the transcriptional activity of NRs in the gut is beyond the scope of this study, we did observe changes in expression of two NRs that bind fatty acids -PPARa and HNF4a -as well as NR co-regulators such as SHP (Nr0b2) and PGC1A (Ppargc1a) (Fig. 2).HNF4a, down regulated by PL + CO in the duodenum, binds LA and plays a critical role in maintaining intestinal health, intestinal epithelial differentiation and barrier function [62][63][64] ; it is also dysregulated in colon cancer as well as colitis and is an IBD susceptibility gene 4,65,66 .Interestingly, Retnlb, a known HNF4a target gene 62 , is also downregulated by the HFDs in the duodenum.PPARa binds a variety of fatty acids 67,68 and is involved in lipid metabolism as well as nutrient transport and energy; it also plays a protective role against colon cancer 69,70 Ppara was upregulated by all three HFDs in the small intestines, as was Cd36, a fatty acid transporter and target of PPARa 71 The other most prominent transcription factor family that was dysregulated by the HFDs were the STAT/IRF factors involved in interferon signaling (Stat1, Stat3, Irf1, Irf5, Irf8) and hence play a critical role in the immune system.Lastly, SREBPF1, which regulates the expression of fatty acid and cholesterol metabolism genes, including Scd1, is upregulated by the HFDs and is potentially linked to COVID-19 (Fig. 7) 72,73 .

Impact of HFDs on genes involved in barrier function, IBD and colon cancer
Several diseases, including IBD, colon cancer, and a leaky gut (barrier dysfunction) have been linked to consumption of HFDs and rates of these diseases have been increasing along with increasing fat intake 9,74 .We observed changes in expression of many genes by one or more HFDs which could contribute to intestinal disease (Figs. 3, 4).For example, there was a decrease in expression of a number of anti-cancer genes including pro-apoptotic gene Casp3, DNA repair genes Mgmt and Parp1 and tyrosine phosphatase Ptpn11.There was also an increase in expression of several cancer-promoting genes such as intestinal stem cell marker Lgr5 and Vnn1 (vanin1, a biotinidase).In other cases, the HFDs seemed to be protective.Decreased expression of Duox2 in HFDs suggests a lower in ammatory response compared to the control diet 75 , which could be bene cial in halting the progression of colorectal cancer 76 , and reduced expression of Ripk3 (receptor-interacting protein (RIP) family of serine/threonine protein kinases) in the terminal ileum may help alleviate in ammation in IBD 77 .Some genes showed differential effects depending on the HFD.For example, both Cldn10 and EGF have lower expression in SO + CO vs PL + CO: reductions in both of these genes can impair barrier function.This is consistent with previous ndings from our lab and others that a high LA diet can contribute to barrier dysfunction while olive oil, a key feature of the Mediterranean diet, is considered to be anti-in ammatory 4,78,79 .

Effects of HFDs on COVID-19-related genes
HFDs, including the ones analyzed in this study, often contribute to obesity which is a signi cant risk factor for COVID-19 80 .COVID-19 patients can experience gastrointestinal symptoms, including damage to the intestinal epithelial barrier 81 .Therefore, it is perhaps not surprising that the lower gastrointestinal tract has a large number of ACE2 receptors and that its expression, along with the genes that encode accessory proteins ENPEP and SLC6A19 (BOAT1) which facilitate viral entry via ACE2 82,83 , is increased in one or more of the HFDs (Fig. 7).Like ACE2, Klk1 and Klk1b5 are part of the RAS pathway and are thought to be required for viral processing 84 ; their expression was also increased in the CO diet.
Furthermore, several host genes involved in the immune response to SARS-CoV-2 are downregulated by one or more HFD -Gzma, Irf1, Stat1, Stat3.In addition to dysregulated gene expression, several of these COVID-19-related proteins have also been found to interact with one or more SARS-CoV-2 viral proteins.
Taken together, our results suggest that these HFDs, or the metabolic dysfunction and/or the dysbiosis caused by them, might be detrimental to COVID-19 patients 85,86 .While additional studies are required, it is nonetheless intriguing to speculate that effects of a high fat diet on the intestinal tract could potentially account for some of the demographics of the pandemic across different populations 87,88

Limitations and caveats
Limitations and caveats to this study include the length of time on the diets (24 weeks) -the observed changes in gene expression could be due directly to the diets and/or to their long-term effects such as obesity, diabetes and susceptibility to colitis [2][3][4] .Others have shown changes in gene expression and the microbiome after just a few days on a HFD, which could be a re ection of the body adjusting to a new nutrient environment 12 ; the effects we observe after 24 weeks may represent more persistent effects on gene expression.In addition to increased fat content, the HFDs were also different from the low-fat Viv chow control in that they did not contain ber.That being said, we have shown that an SO-enriched diet containing ber leads to similar increases in susceptibility to colitis in mice and causes a similar amount of weight gain as one without the added ber 3,4 .Additionally, even though all three HFDs lacked ber, they often displayed different effects on gene expression suggesting that not all of the effects observed are due to a lack of ber.This study examines RNA levels only, which may not always relate to protein levels -e.g., Cyp2c55 has very high levels of RNA in the proximal colon but its protein levels are reported to be very low, similar to the small intestines 49,50 .Whole tissue was used, so in addition to intestinal epithelial cells, other cell types including immune cells would have been sampled.Single-cell RNA-seq by others show that a HFD does indeed impact different cell types in a differential fashion, and differences can be observed within days 12 .Finally, the relevance to humans must be established.Since most of the DEGs highlighted in the study are highly conserved between mouse and human, including several of the transcriptional regulators -HNF4a, PPARa, STAT1/3, IRF1, SREBPF1 are all over 80% identical between human and mouse on the protein level-we anticipate that many of the effects reported here will also be found in humans.

Figure 3 HFDs
Figure 3 for a complete list of genes.E-G.Line graphs showing normalized read counts with standard deviation (SD) of select genes on the indicated diets.Genes with signi cantly different levels of expression between the diets within a given tissue (p-adj < 0.05) are indicated as follows: a (VIV vs CO); b (VIV vs SO+CO); c (VIV vs PL+CO); d (CO vs SO+CO); e (CO vs PL+CO); f (SO+CO vs PL+CO).

Figure 5 Network
Figure 5