Dietary intake alters gene expression in colon tissue: possible underlying mechanism for diet’s influence on disease

Abstract

Background

While the association of diet and disease is well documented, the biologic mechanisms involved have not been entirely elucidated. In this study we evaluate how dietary intake influences gene expression to better understand the underlying mechanisms through which diet operates.

Methods

We used data from 144 individuals who had comprehensive dietary intake and gene expression data from RNAseq using normal colonic mucosa. Using the DESeq2 statistical package, we identified genes that displayed statistically significant differences in expression between individuals in high intake and low intake categories for several dietary variables of interest adjusting for age and sex. We examined total calories, total fats, vegetable protein, animal protein, carbohydrates, trans-fatty acids, mutagen index, red meat, processed meat, whole grains, vegetables, fruits, fiber, folate, dairy products, calcium, and Prudent and Western Dietary Patterns.

Results

Using a false discovery rate (FDR) of <0.1, meat-related foods were statistically associated with 68 dysregulated genes, calcium with three dysregulated genes, folate with four dysregulated genes, and non-meat related foods with 65 dysregulated genes. With a more stringent FDR of <0.05, there were nine meat-related dysregulated genes and 23 non-meat related genes. Ingenuity pathway analysis (IPA) identified three major networks among genes identified as dysregulated with respect to meat-related dietary variables and three networks among genes identified as dysregulated with respect to non-meat related variables. The top networks (IPA Network Score >30) associated with meat-related genes were 1) cancer, organismal injury and abnormalities, tumor morphology and 2) cellular function and maintenance, cellular movement, cell death and survival. Among genes related to non-meat consumption variables the top networks were 1) hematological system development and function, nervous system development and function, tissue morphology and 2) connective tissue disorders, organismal injury and abnormalities.

Conclusions

Several dietary factors were associated with gene expression in our data. These findings provide insight into possible mechanisms whereby diet may influence disease processes.

Introduction

Diet has been associated with several diseases including atherosclerosis [2], inflammatory bowel disease [3], steatosis and non-alcoholic fatty liver disease [4, 5], seizures [6], and various cancers including colon and breast carcinoma [7–9]. Moreover, diet has also been identified as a risk factor for Type 2 diabetes mellitus, obesity, and metabolic syndrome [10]. Beyond its role in disease status, diet has been shown to play an important role in a number of physiologic and metabolic processes including bone remodeling [11] and aspects of the immune response [12]. The biologic mechanisms involved in these processes is less clear.

Studies have evaluated the role of diet in the regulation of specific physiologic and pathologic processes, showing that single dietary metabolites and nutrients play a role in the regulation of certain signaling pathways [11, 13–15] or cytokine profiles [16]. Other studies have focused on the dietary effects on oxidative stress [17] or the role of potentially chemoprotective nutrients [18]. Some of these studies also have looked at localized gene expression within specific hypothesized pathways and processes, suggesting that diet has a role in the regulation of gene expression [7, 8, 16, 19]. Recently the effects of dietary patterns on the entire gene expression profile were investigated, suggesting a differential expression of genes in individuals with a Prudent Dietary Pattern verses a Western Dietary Pattern and identified specific canonical pathways associated with both dietary patterns [20].

In this study we hypothesize that levels of gene expression in normal colonic mucosa will be associated with dietary factors. Dietary factors we consider include sources of calories, including animal protein and vegetable protein, carbohydrates, and total fat. We also consider trans-fatty acids, total red meat, processed meat, mutagen index, dairy products, fruit, vegetables, refined grains, whole grains, sucrose, dietary fiber, calcium, and folate with gene expression profiles. Lastly, we examined the association between Prudent Dietary Pattern and Western Dietary Pattern with gene expression.

Methods

Total RNA was available from normal colonic mucosa for colon cancer cases who were part of the Diet, Activity, and Lifestyle study, an incident, population-based, case-control study of colon cancer from Utah and the Kaiser Permanente Medical Research Program (KPMRP) [21]. Cases were identified using a rapid-reporting-system and had tumor registry verification of a first primary adenocarcinoma of the colon and were diagnosed between October 1991 and September 1994 and tumor tissue blocks (used for RNA) were obtained for 97% of all Utah cases and for 85% of all KPMRP cases [22]. Individuals with known adenomatous polyposis coli (APC), Crohn’s disease, or inflammatory bowel disease were not eligible for the study. The study was approved by the Institutional Review Board of the University of Utah and at KPMRP; all study participants signed informed consent prior to participation.

Dietary Data

Data were collected by trained and certified interviewers using laptop computers shortly after diagnosis. All interviews were audio-taped as previously described and reviewed for quality control purposes [23]. Any interview deemed questionable by the interviewer was reviewed centrally; this allowed us to evaluate the data by distinguishing between an interview that was difficult for the interviewer versus one that was difficult for the participant. Additionally, we reviewed audio tapes for all individuals whose nutrient levels were considered outliers (over the 95%tile or under the 5 percentile of intake); this enabled us to correct any coding errors and determine the quality and face validity of the interview. Dietary information was obtained for the year two years prior to diagnosis using an extensive diet history questionnaire adapted from the validated CARDIA diet history; most individuals were asked to recall usual dietary intake from two to three years ago in order to obtain pre-diagnosis diet [24]. Foods were converted to nutrients using the Nutrition Coding Center Nutrient Data System Version 19 as well as being grouped into categories of similar foods. We assessed both foods and nutrients with gene expression. Foods units were standard servings per day, which was ½ cup of fruit, vegetable, or dairy product; meat servings were 2 to 3 oz of meat; grain products were ½ cup of rice-type grains or one slice of bread. Prudent and Western Dietary Patterns were developed based on principal component program [25]. Our Prudent Dietary Pattern was heavily loaded towards diets high in fruits, vegetables, whole grains, fish, and chicken, while the Western Dietary Pattern was highly loaded towards red meat, processed meats, and refined grains and high-sugar-high-fat foods. Additional questions were asked about meat consumption, doneness, and preparation methods that were combined and used to create a mutagen index score [26].

RNA processing

RNA was extracted from formalin-fixed paraffin-embedded colonic tissues. FFPE tissue has been shown to be a reliable source of RNA for use in conjunction with RNA-seq [27]. The study pathologist reviewed slides to delineate carcinoma and normal colonic mucosa. Total RNA was extracted, isolated, and purified using the RecoverAll Total Nucleic Acid isolation kit (Ambion), RNA yields were determined using a NanoDrop spectrophotometer.

Sequencing Library Preparation

Library construction was performed using the Illumina TruSeq Stranded Total RNA Sample Preparation Kit with Ribo-Zero as previously described [28]. Briefly, ribosomal RNA was removed from 100 ng total RNA using biotinylated Ribo-Zero oligos attached to magnetic beads. Following purification, the rRNA-depleted sample was fragmented and primed with random hexamers. First strand reverse transcription was accomplished using Superscript II Reverse Transcriptase (Invitrogen). Second strand cDNA synthesis was accomplished using DNA polymerase I and Rnase H under conditions in which dUTP is substituted for dTTP, yielding blunt-ended cDNA fragments in which the second strand contains dUTP. An A-base is added to the blunt ends as a means to prepare the cDNA fragments for adapter ligation and block concatamer formation during the ligation step. Adapters containing a T-base overhang were ligated to the A-tailed DNA fragments. Ligated fragments were PCR-amplified (13 cycles) under conditions in which the PCR reaction enables amplification of the first strand cDNA product. The PCR-amplified library was purified using Agencourt AMPure XP beads (Beckman Coulter Genomics).

Sequencing and Data Processing

Sequencing libraries (18 pM) were chemically denatured and applied to an Illumina TruSeq v3 single read flow cell using an Illumina cBot. Hybridized molecules were clonally amplified and annealed to sequencing primers with reagents from an Illumina TruSeq SR Cluster Kit v3-cBot-HS. Following transfer of the flowcell to an Illumina HiSeq instrument, a 50 cycle single-read sequence run was performed using TruSeq SBS v3 sequencing reagents. The single-end 50-base reads from the Illumina HiSeq2500 were aligned to a sequence database containing the human genome (build GRCh37/hg19, February 2009, from genome.ucsc.edu) plus all splice junctions generated using the USeq MakeTranscriptome application (version 8.8.1, available here: http://useq.sourceforge.net/). Alignment was performed using NovoAlign version 2.08.01 available from novocraft.com, which also trimmed any adapter sequence. Genome alignments to splice junctions were translated back to genomic coordinates using the USeq SamTranscriptomeParser application. Alignments were then sorted and indexed using the Picard SortSam application (version 1.100, available here: http://broadinstitute.github.io/picard/). Aligned read counts for each gene were calculated using pysam (https://code.google.com/p/pysam/) and samtools (http://samtools.sourceforge.net/). A python script using the pysam library was given a list of the genome coordinates for each gene, and counts to the exons and UTRs of those genes were calculated. Gene coordinates were downloaded from http://genome.ucsc.edu. Over 17000 genes that were expressed in colon mucosa were analyzed with dietary factors. Rigid quality control procedures were implemented to assure high quality data; participants failing QC were dropped from the analysis.

Statistical Methods

Of the 197 initial tumor/non-tumor tissue pairs, 22 subjects failed quality control (QC) based on low number of sequence leaving 175 subjects with high quality expression data. Of these, 144 had dietary questionnaire data. For each dietary factor, our analysis centered on contrasting individuals with lower intake to those of individuals with higher intake. Dietary data were evaluated using nutrients per 1000 calories and standard servings of food; sample tertiles were computed for the consumption levels of each dietary factor. For each such dietary variable, we determined which genes displayed statistically significant differential expression between the lowest and highest intake tertile categories after adjusting for age and sex using the Bioconductor package DESeq2 written for the R statistical programming environment; these genes were considered dysregulated. DESeq2 assumes the RNA-Seq counts are distributed according to negative binomial distributions. It utilizes generalized linear modeling and variance reduction techniques for estimated coefficients to test individual null hypotheses of zero log2 fold changes between high and low categories (i.e. no differential expression) for each gene. It employs both an independent-filtering method and the Benjamini and Hochberg (BH) [29] procedure to improve power and control the false discovery rate (FDR). The default DESeq2 options were used, including replacement of outliers, as defined by Cook’s distance, and the Wald test. For further details regarding DESeq2, see Love et al. [30]. In identifying genes with differential expression, an FDR of 0.10 was used. The fold change calculations for differentially expressed genes was determined by DESeq2 and represents the log2 change in expression level (i.e. counts) for the high versus low dietary categories.

Bioinformatics analysis was performed on the list of Ensemble IDs associated with genes identified as differentially expressed between high and low consumption categories for the dietary variables of interest using QIAGEN’s Ingenuity Pathway Analysis (IPA) [31]. We used genes from Ingenuity Knowledge Base and considered both indirect and direct relationships. Causal and interaction networks were generated. Interaction networks were limited to 35 molecules per network and 25 networks per analysis, and excluded endogenous chemicals. We focused on algorithmically derived interaction networks, which are assigned a score based on their relevance to the genes in the input dataset, the number of focus genes (i.e. dysregulated genes in our data that are in that network), and their connectivity [32]. The score is calculated as −log₁₀P, where P is generated using a Fisher’s exact test [33]. Studies have found scores >3 to be significant, with a score of 3 indicating a 1/1000 chance that the focus genes are in a network due to random chance [34–36]. Other studies have opted to utilize more stringent criteria and higher scores to ensure that their discovered networks are highly significant [37, 38]; we utilized highly stringent criteria, only including networks with scores over 20. We applied the BH multiple testing correction to assess pathways in IPA.

Availability of data are restricted to that authorized in patient consent form and in accordance with data transfer agreements and IRB requirements.

Results

The majority of our study population was male and the median age was 65, 13.2% were current smokers, and 37.3% currently took aspirin or non-steroidal anti-inflammatory drugs on a regular basis (Table 1). The median BMI was 29.6. Within this population few people consumed processed meat, with the highest level of intake being less than one serving per day. The highest red meat consumption was less than two servings per day. Over half of study participants had over two servings of vegetables per day and one serving of whole grains per day.

Table 1.

Description of Study population

		N	%
Sex	Male	80	55.6%
	Female	64	44.4%
Smoking Status	Never	60	41.7%
	Former	65	45.1%
	Current	19	13.2%
Recent NSAID Use	No	89	62.7%
	Yes	53	37.3%
Alcohol Long Term and Current (gm/day)	Low	53	36.8%
	Mod	68	47.2%
	High	23	16.0%
Race/Ethnicity	Non-Hispanic White	124	86.1%
	Black	8	5.6%
	Hispanic	9	6.3%
	Other/NA	3	2.1%
Education (years)	<High School	24	16.7%
	High School	42	29.2%
	Some College	50	34.7%
	College Grad/Higher	28	19.4%
		Median	Interquartile Range (IQR)
Age (years)		65.0	(56, 71)
BMI (kg/m2)		29.6	(26.6, 33.2)
Hours Moderate Physical Activity/week		1.5	(0.0, 3.9)
		Meat Consumption
Processed meat (2–3 oz servings/day)	Low1	0.05	(0.01, 0.07)
	Intermediate	0.19	(0.14, 0.23)
	High	0.45	(0.35, 0.59)
Red Meat (2–3 oz servings/day)	Low	0.43	(0.29, 0.61)
	Intermediate	0.95	(0.84, 1.07)
	High	1.58	(1.41, 1.93)
		Non-Meat Consumption
Calcium (mg/1000 Kcal)	Low	300.71	(261.72, 326.90)
	Intermediate	406.47	(380.31, 445.51)
	High	585.69	(541.01, 697.95)
Fiber (g/1000 Kcal)	Low	7.32	(6.75, 8.09)
	Intermediate	10.03	(9.52, 10.43)
	High	13.06	(12.15, 15.62)
Folate (mcg/1000 Kcal)	Low	118.35	(103.41, 126.46)
	Intermediate	151.61	(142.85, 161.84)
	High	203.63	(189.98, 247.94)
Vegetables (1/2 cup servings/day)	Low	1.41	(1.03, 1.90)
	Intermediate	2.85	(2.55, 3.25)
	High	4.59	(4.13, 5.74)
Whole Grains (servings/day)	Low	0.43	(0.33, 0.63)
	Intermediate	1.12	(1.02, 1.29)
	High	2.14	(1.82, 2.52)

Seven genes were differentially expressed for dietary calcium and folate (Table 2). Sixty-five genes were differentially expressed between high and low intake categories among non-meat variables (i.e. Prudent Dietary Pattern, fruits, vegetables, and whole grains). This can be further broken down to one gene with Prudent Dietary Pattern, one gene with fruit intake, 26 genes with vegetable intake, and 37 genes with whole grain intake, (Table 3, FDR <0.1). Several genes were identified as differentially expressed between consumption categories for multiple dietary variables. TXNDC17 was upregulated for both Prudent Dietary Pattern and vegetable intake; MUC5AC was down regulated for vegetable intake, whole grain intake, and dietary folate. FOXJ2, NECAP1, and C3AR1 were upregulated for both calcium intake and vegetable intake Using a more stringent FDR <0.05, we identify one gene with differential expression between calcium intake categories, one with Prudent Dietary Pattern, 13 with vegetable intake, and eight with whole grain intake (Table 2 and Table 3). Among the genes with an FDR <0.05, FOXJ2 was upregulated with high calcium intake and TXNDC17 was upregualted with high Prudent Dietary Pattern. In contrast, five of the eight genes associated with whole grains were downregulated and four of the 13 genes associated with vegetables were downregulated.

Table 2.

Associations between nutrients and gene expression (FDR<0.1)¹

	Gene	Average Adjusted Count		DESeq2 Estimated Log2 Fold Change	P Value	Multiple. Comparison Adjusted P Value
		Low Tertile	High Tertile
Calcium	FOXJ2	32.84	135.64	0.87	4.05E-10	7.07E-06
	NECAP1	23.38	68.92	0.71	3.87E-07	2.38E-03
	C3AR1	3.51	21.28	0.63	4.09E-07	2.38E-03
Folate	A1BG	3.84	8.45	0.63	2.12E-05	9.22E-02
	SLC37A2	23.04	79.47	0.61	1.83E-05	9.22E-02
	CLDN4	129.49	87.70	−0.47	1.44E-05	9.22E-02
	MUC5AC	58.73	5.49	−0.50	1.52E-05	9.22E-02

¹Associations are adjusted for age and sex

Table 3.

Associations between non-meat foods and gene expression (FDR <0.1)¹

	Gene	Average Adjusted Count		DESeq2 Estimated Log2 Fold Change	P Value	Multiple. Comparison Adjusted P Value
		Low Tertile	High Tertile
Prudent Diet	TXNDC17	6.18	16.74	0.68	4.00E-07	6.99E-03
Fruit	HMGN2	8.45	16.66	0.57	4.31E-06	7.50E-02
Vegetables	ANP32A	64.67	622.05	0.86	3.22E-10	2.90E-06
	NECAP1	30.12	160.72	0.76	5.98E-08	1.94E-04
	HELZ	639.22	10029.11	0.66	6.49E-08	1.94E-04
	SLC2A3	50.41	263.66	0.75	9.46E-08	1.95E-04
	FOXJ2	53.51	337.01	0.72	1.09E-07	1.95E-04
	ADAMTS6	9.00	40.07	0.63	5.57E-06	6.26E-03
	HSPG2	285.80	190.14	−0.48	4.47E-06	6.26E-03
	C3AR1	7.36	55.37	0.57	4.98E-06	6.26E-03
	TXNDC17	6.23	14.92	0.62	9.03E-06	9.03E-03
	PXN	124.57	91.47	−0.39	1.54E-05	1.26E-02
	MUC5AC	49.01	3.56	−0.48	1.51E-05	1.26E-02
	FURIN	55.18	36.93	−0.45	2.09E-05	1.56E-02
	DNAH14	21.82	62.62	0.58	6.69E-05	4.63E-02
	RNF123	45.86	30.57	−0.40	7.22E-05	4.64E-02
	MICALL2	81.80	57.71	−0.41	9.91E-05	5.94E-02
	COL1A1	478.40	278.35	−0.53	1.09E-04	6.15E-02
	COL12A1	94.94	58.32	−0.49	1.22E-04	6.27E-02
	STK36	23.04	14.94	−0.43	1.25E-04	6.27E-02
	REG4	75.69	30.88	−0.54	1.68E-04	7.96E-02
	AMN	65.79	41.60	−0.47	1.92E-04	8.62E-02
	PTER	10.51	16.89	0.44	2.02E-04	8.64E-02
	ACAP3	56.17	37.98	−0.41	2.24E-04	9.17E-02
	TAF1C	45.25	31.54	−0.42	2.70E-04	9.83E-02
	TTYH3	60.34	44.08	−0.39	2.58E-04	9.83E-02
	EPM2AIP1	44.29	60.95	0.34	2.73E-04	9.83E-02
	ATP11C	22.62	41.25	0.49	2.87E-04	9.94E-02
Whole Grains	MUC5AC	55.79	3.17	−0.62	2.52E-06	2.47E-02
	BCKDHA	16.28	7.89	−0.69	3.59E-06	2.47E-02
	BTG1	74.46	100.65	0.42	6.99E-06	3.20E-02
	ANKRD30BL	118.66	53.70	−0.70	1.05E-05	3.61E-02
	ABCC3	341.13	253.51	−0.41	1.54E-05	3.90E-02
	MPEG1	33.40	49.38	0.50	2.25E-05	3.90E-02
	EPS8L3	143.09	103.38	−0.42	2.27E-05	3.90E-02
	JMJD7	11.64	6.22	−0.63	1.97E-05	3.90E-02
	DVL1	39.48	25.55	−0.53	5.97E-05	5.46E-02
	HELZ2	91.24	58.30	−0.52	5.32E-05	5.46E-02
	GSTM3	4.19	9.13	0.64	5.94E-05	5.46E-02
	NR1I2	56.26	39.67	−0.46	4.25E-05	5.46E-02
	GMPPB	54.06	40.48	−0.40	4.89E-05	5.46E-02
	AMIGO3	55.35	41.00	−0.42	4.93E-05	5.46E-02
	C9orf172	10.88	6.23	−0.60	5.86E-05	5.46E-02
	TPRN	54.68	35.53	−0.53	6.45E-05	5.54E-02
	EMB	14.29	24.80	0.58	6.99E-05	5.65E-02
	RASSF7	37.58	23.88	−0.51	8.00E-05	5.98E-02
	KCTD13	10.46	5.71	−0.58	8.28E-05	5.98E-02
	RTEL1-TNFRSF6B	89.14	62.54	−0.43	9.18E-05	6.30E-02
	C9orf50	11.32	2.44	−0.61	9.77E-05	6.39E-02
	PFKL	144.14	108.76	−0.38	1.09E-04	6.76E-02
	MRPS28	3.04	7.82	0.62	1.13E-04	6.76E-02
	LTBR	69.00	55.07	−0.33	1.27E-04	6.96E-02
	MST1R	97.85	71.06	−0.41	1.26E-04	6.96E-02
	TERC	27.56	14.83	−0.58	1.37E-04	7.21E-02
	FAM129B	113.83	79.80	−0.44	1.51E-04	7.68E-02
	CTC-435M10.3	11.71	6.39	−0.57	1.57E-04	7.70E-02
	FAM83H	73.50	53.33	−0.41	1.68E-04	7.86E-02
	CHKB-CPT1B	70.38	48.67	−0.45	1.72E-04	7.86E-02
	SEMA3B	54.89	35.79	−0.50	1.95E-04	8.37E-02
	DPM2	18.77	11.53	−0.51	1.97E-04	8.37E-02
	CD151	79.03	57.71	−0.42	2.01E-04	8.37E-02
	RNF146	13.31	19.88	0.46	2.45E-04	9.39E-02
	SEMA4B	52.73	35.60	−0.45	2.39E-04	9.39E-02
	NOC2L	35.57	24.81	−0.44	2.46E-04	9.39E-02
	EML2	21.68	13.83	−0.46	2.68E-04	9.96E-02

¹Associations are adjusted for age and sex

Our IPA analysis found three interaction networks significantly associated with non-meat consumption (IPA Network Score ≥20). These IPA networks were identified as functioning with: cancer, organismal injury and abnormalities, and tumor morphology (Score=31, Focus Molecules=11) (Figure 1A); cellular function, maintenance, cellular movement, cell death and survival (Score=31, Focus Molecules=15) (Figure 1B), and drug metabolism, molecular transport, and small molecule biochemistry (Score=26, Focus Molecules=13) (Figure 1C).

Figure 1. IPA networks associated with dysregulated genes based on level of non-meat-related dietary intake.

Figure 1A. Hematological system development and function, nervous system development and function, tissue morphology (Score 31)

Figure 1B. Connective tissue disorders, organismal injury and abnormalities, cardiovascular system development and function (Score 29)

Figure 1C. Cellular compromise, cellular function and maintenance, protein degradation (Score 21)

Red indicates up-regulation and green indicates down regulation. Node shapes denote enzymes ; phosphatases Δ; kinases ; peptidases , G-protein coupled receptor ; transmembrane receptor ; cytokines ; ion channel ; transporter ; translation regulator ; transcription regulator ;; and complex/group

Among genes identified as significantly differentially expressed between high and low meat-related intake categories, 64 genes were differentially expressed with red meat, two genes were differentially expressed with processed meat, and three genes were differentially expressed with a Western Dietary Pattern (Table 4). Of these, eight genes were differentially expressed by level of red meat and LSAMP was differentially expressed by level of Western Dietary Pattern when using a more stringent selection criterion of a maximal FDR of <0.05. The majority of these genes were downregulated by red meat consumption; of all of the genes identified only IL22RA1, SLC17A4, SLC25A25, MUC17, ZFY, FRK, PEX11G, NEDD9, NR3C2, and RPS4Y1 were upregulated by red meat consumption. The two genes identified as being differentially expressed for Western Dietary Pattern also were differentially expressed for red meat; C3 was downregulated by red meat and processed meat.

Table 4.

Association between meat-related foods and gene expression (FDR<0.1)¹

	Gene	Average Adjusted Count		DESeq2 Estimated Log2 Fold Change	P Value	Multiple. Comparison Adjusted P Value
		Low Tertile	High Tertile
Western Diet	LSAMP	7.21	1.51	−0.70	2.20E-06	3.83E-02
	TDRD15	2.21	0.21	−0.57	6.21E-06	5.42E-02
Meat	ONECUT2	12.45	2.07	−0.92	2.40E-08	4.18E-04
	TIMP1	26.41	11.51	−0.76	1.39E-06	1.22E-02
	CLDN1	12.53	2.57	−0.77	3.78E-06	1.32E-02
	CMSS1	11.46	4.35	−0.77	3.34E-06	1.32E-02
	LSAMP	7.47	1.56	−0.79	3.58E-06	1.32E-02
	SPARC	118.63	74.35	−0.55	8.19E-06	2.38E-02
	TDRD15	1.98	0.14	−0.64	1.54E-05	3.84E-02
	COL1A2	314.64	214.85	−0.54	1.97E-05	4.29E-02
	CEMIP	27.24	16.30	−0.56	3.66E-05	6.73E-02
	NKD1	16.63	6.86	−0.69	3.86E-05	6.73E-02
	ZFY	4.11	9.63	0.65	5.84E-05	7.01E-02
	FZD3	17.37	9.06	−0.63	6.32E-05	7.01E-02
	COL4A2	155.42	108.31	−0.50	6.44E-05	7.01E-02
	TP53	33.67	23.40	−0.44	4.87E-05	7.01E-02
	SLC17A4	46.78	78.07	0.58	6.44E-05	7.01E-02
	CCT2	27.99	18.04	−0.54	5.51E-05	7.01E-02
	HSPA5	50.38	34.01	−0.50	8.23E-05	8.24E-02
	TPX2	18.50	8.63	−0.61	1.47E-04	8.24E-02
	FRK	15.94	23.66	0.50	1.44E-04	8.24E-02
	PTK7	18.21	10.62	−0.57	8.63E-05	8.24E-02
	NCL	109.20	80.61	−0.40	1.03E-04	8.24E-02
	IL22RA1	11.53	22.37	0.61	1.17E-04	8.24E-02
	GPR26	5.57	0.71	−0.55	1.46E-04	8.24E-02
	EGFLAM	4.92	1.58	−0.65	1.30E-04	8.24E-02
	SLC6A5	2.59	0.18	−0.51	1.10E-04	8.24E-02
	RAB31	19.12	10.49	−0.60	1.06E-04	8.24E-02
	TMEM107	37.26	23.81	−0.56	1.24E-04	8.24E-02
	C1S	63.12	47.50	−0.43	1.37E-04	8.24E-02
	C2orf27A	2.37	0.50	−0.57	1.23E-04	8.24E-02
	IGHG1	171.81	74.48	−0.66	1.00E-04	8.24E-02
	MUC5AC	50.46	5.63	−0.54	1.30E-04	8.24E-02
	HSPA9	71.14	53.20	−0.40	1.52E-04	8.26E-02
	TMEM132B	6.82	2.55	−0.65	1.56E-04	8.26E-02
	DCN	39.88	23.71	−0.55	1.98E-04	8.40E-02
	SNX10	7.08	2.27	−0.64	1.99E-04	8.40E-02
	AURKA	10.06	3.66	−0.63	2.06E-04	8.40E-02
	PEX11G	1.32	3.29	0.64	1.76E-04	8.40E-02
	COL1A1	445.08	285.52	−0.55	1.87E-04	8.40E-02
	NEDD9	84.09	112.98	0.38	1.96E-04	8.40E-02
	WDR89	11.37	5.46	−0.59	1.77E-04	8.40E-02
	SLC25A25	35.34	54.48	0.46	1.73E-04	8.40E-02
	NR3C2	113.10	154.48	0.41	1.95E-04	8.40E-02
	C1R	44.16	25.70	−0.53	2.09E-04	8.40E-02
	MUC17	107.73	269.60	0.63	2.12E-04	8.40E-02
	RPS4Y1	8.70	19.26	0.59	2.28E-04	8.82E-02
	DOCK3	6.29	2.66	−0.62	2.86E-04	9.55E-02
	CHRAC1	14.62	8.35	−0.51	2.90E-04	9.55E-02
	SPATA17	2.21	0.55	−0.58	2.82E-04	9.55E-02
	NETO1	3.40	0.46	−0.52	2.86E-04	9.55E-02
	TRIL	7.68	4.13	−0.60	2.83E-04	9.55E-02
	DDX53	1.70	0.06	−0.47	2.71E-04	9.55E-02
	HIST1H2BN	5.55	1.23	−0.61	2.74E-04	9.55E-02
	PEG10	6.38	2.44	−0.62	2.74E-04	9.55E-02
	THY1	19.83	10.36	−0.58	3.03E-04	9.65E-02
	GRM2	3.19	1.31	−0.59	3.07E-04	9.65E-02
	TTC40	4.30	0.77	−0.56	3.10E-04	9.65E-02
	TIMP2	94.56	68.12	−0.42	3.30E-04	9.71E-02
	NOS1	6.65	0.87	−0.51	3.48E-04	9.71E-02
	CLSPN	12.16	4.89	−0.61	3.54E-04	9.71E-02
	C3	102.46	57.73	−0.56	3.62E-04	9.71E-02
	CCDC150	7.05	3.32	−0.61	3.22E-04	9.71E-02
	VWDE	4.91	1.77	−0.61	3.50E-04	9.71E-02
	CCT3	36.04	24.11	−0.44	3.37E-04	9.71E-02
	HSP90B1	142.80	105.92	−0.38	3.61E-04	9.71E-02
Processed Meat	C3	101.05	50.36	−0.70	5.56E-06	6.44E-02
	KRT1	2.97	0.08	−0.60	7.40E-06	6.44E-02

¹Associations are adjusted for age and sex

In IPA analysis, three interaction networks were significantly associated (Score ≥20) with meat-related variables (i.e. red meat, processed meat, and Western Dietary Pattern). These IPA networks were identified as Hematological system developmental function, nervous system development and function, Tissue Morphology (IPA Network Score = 31, Focus Molecules 22) (Figure 2A); Connective Tissue Disorders, Organismal Injury and Abnormalities, Cardiovascular System Development and Function (Score=29, Focus Molecules=14) (Figure 2B); and cellular compromise, cellular function and maintenance, protein degradation (Score=21, Focus Molecules 11) (Figure 2C). Online Supplement 1 shows genes up and down regulated in each of these networks. Some genes were differentially expressed between high and low consumption categories for both meat and non-meat dietary variables. COL1A1 was down-regulated for vegetables and meat intake; MUC5AC was down-regulated for folate, vegetables, whole grains, and meat. Foods and nutrients analyzed that were not associated with any differential gene expression (FDR <0.1) were mutagen index, total dairy products, carbohydrates, trans-fatty acids, total dietary fat, dietary fiber, refined grain products, and sucrose. Additionally, we observed no differences in gene expression by level of vigorous physical activity.

Figure 2. IPA networks associated with dysregulated genes based on level of meat related dietary intake.

Figure 2A. Cancer, organismal injury and abnormalities, and tumor morphology (Score 31)

Figure 2B. Cellular function and maintenance, cellular movement, cell death and survival (Score 31)

Figure 2C. Drug metabolism, molecular transport, small molecule biochemistry (Score 26)

Red indicates up-regulation and green indicates down regulation. Node shapes denote enzymes ; phosphatases Δ; kinases ; peptidases , G-protein coupled receptor ; transmembrane receptor ; cytokines ; ion channel ; transporter ; translation regulator ; transcription regulator ;; and complex/group

Discussion

Our study provides support for associations between level of dietary intake and gene expression in normal colonic mucosa utilizing RNA-Seq which allowed assessment of over 17,000 genes. Some genes were differentially expressed by intake level across multiple dietary variables, generally in the same direction, although there was limited overlap between meat and non-meat variables. We observed three IPA interaction networks highly associated with meat consumption and three IPA interaction networks associated with non-meat consumption. These data suggest that dietary influence on gene expression plays an important role in a number of physiologic and pathologic processes.

The majority of genes that were differentially expressed with high red meat consumption were downregulated. These genes predominately play a role in extracellular matrix (ECM) assembly and structure, calcium binding, and the regulation of signal transduction via the Wnt signaling pathway. The canonical Wnt signaling pathway triggers β-catenin-dependent gene expression, allowing cell cycle progression at the G1/S transition [39]. In particular, we found that high intake of meat correlated with downregulation of FZD3, TMEM132B, CCDC150, EGFLAM, and NKD1. Both TMEM132B and NKD1 are known to downregulate Wnt signaling [40]. Thus meat consumption can contribute to the dysregulation of a proto-oncogenic pathway. At the same time, we found that high red meat consumption was associated with a downregulation of TP53, encoding for the tumor suppressor p53. Both of these findings are consistent with previous findings that red meat intake can induce pro-carcinogenic gene expression changes through the p53- and Wnt signaling pathways [19]. This is possibly due to the fact that red meat is high in dietary heme, a luminal irritant associated with hyperproliferation and hyperplasia through the downregulation of feedback inhibitors, including Wnt inhibitors [17]. These findings add to the literature that suggest red meat increases risk of several chronic diseases including cancer.

High red meat consumption also downregulated the expression of genes associated with ECM. Dysregulation of genes involved in the regulation of the ECM and cell matrix adhesion, such as matrix metalloproteinases (MMPs), collagen-related genes such as COL4A2, COL1A2, and COL1A1, tissue inhibitor of metalloproteinases (TIMPs), and one cut homeobox 2 (ONECUT2) have been implicated in a variety of diseases [41]. Moreover, red meat is associated with both acute oxidative stress and delayed cytotoxic stress [17]. Several genes dysregulated by meat, including THY1, NCL, COL4A2, and KRT1, contribute to metastasis and angiogenesis. We found that diet downregulated a variety of genes associated with ECM assembly and maintenance, including TIMP1, providing an additional mechanistic link between red meat and disease processes.

It is also worth noting that red meat upregulated some genes, namely genes with products involved in phosphate transport (SLC17A4 and SLC25A25) and cytokine signaling (MUC17 and IL22RA1). Diet is known to play an important role in bone health, with meat protein increasing acid load and regulating calcium balance [42]. The upregulation of genes involved in phosphate transport, as well as our observed downregulation of calcium binding genes (SPARC, CEMIP, HSPA5, HSP901B, VWDE, SLC25A25, CIR, CLDN1, and EGFLAM), may help explain the mechanism by which meat consumption modulates bone metabolism, beyond the effects of dietary protein on acid load. Moreover, the upregulation of MUC17 and IL22RA1 may play a role in meat-induced inflammation and inflammation-related diseases. The exact function of MUC17 is unknown; however, other mucins are reportedly involved in the regulation of inflammation [43] and MUC17 expression is related to certain cancers, including pancreatic cancer [44]. IL22 is produced by T lymphocytes and mediates cellular inflammatory responses via the activation of intracellular kinases and has been implicated in a variety of inflammatory diseases, including colon cancer and ulcerative colitis [45], both of which have been associated with red meat consumption [17, 46].

Several genes associated with immune response were down-regulated by high meat consumption, including TRIL and IGFGH1. We found that high red meat consumption decreased expression of C1R and C1S. C1S encodes for a component of C1 in the classic complement pathway; CIR is the R subcomponent of C1. The downregulation of the components of the complement cascade by meat consumptions seems paradoxical; however, evidence suggests that the proximal classical complement components may provide a protective effect against atherosclerosis [47, 48]. While atherosclerosis is an inflammatory disease, Hovland et al. suggest that the classic complement pathway may be important for tissue homeostasis by the clearance of cell debris and immune complexes [48]. Thus it is possible that a downregulation of proximal components of the classic pathway, such as components of C1, could paradoxically contribute to the pro-inflammatory effects of red meat. On the other hand, upregulation of MUC17 by meat could help explain its association with a number of factors, including inflammation [20, 46].

Further evaluation of the genes associated with non-meat related variables showed that genes associated with calcium were upregulated, two of the four genes associated with folate were upregulated, and the TXNDC17 gene associated with Prudent Dietary Pattern was upregulated. The genes associated with calcium consumption were involved in transmembrane transport, regulation of transcription and translation, cell survival and the complement system. Calcium plays a key role in muscle contraction and relaxation, blood coagulation, nerve transmission and keratinocyte differentiation [49]. The genes upregulated by high calcium consumption include FOXJ2, encoding for a transcription activator; NECAP1, whose product is involved in receptor mediated endocytosis; and C3AR1, encoding for a C3A receptor. A Prudent Dietary Pattern was associated with roughly three servings of milk per day in one study [20] and was associated with decreased risk of cancer and the B cell receptor pathways [20]. Our observed calcium-associated upregulation of tumor suppressors and a C3A receptor is consistent with these earlier findings.

Folate-rich foods include green vegetables and citrus fruits, and folate plays a key role as a coenzyme involved in DNA synthesis, amino acid metabolism, and methylations [50]. High versus low folate intake was associated with the upregulation of SLC37A2 and CLDN4 and the downregulation of MUC5AC. The gene products for SLC37A2 and CLDN4 are involved in carbohydrate transport and tight junction organization, respectively. Tight junction proteins, such as claudins, are related to NF-κB, mTOR and Nrf2 [51, 52], while the expression of tight junction proteins is associated with amino acids, such as arginine and tryptophan [51, 52]. As folic acid regulates amino acid metabolism, this could explain its role in modulating the expression of tight junction proteins. Dietary folate also downregulates the expression of MUC5AC, as some mucins have been associated with inflammation [43], this could help explain the reported association between folate and certain cancers and cardiovascular diseases [50]. We also found that a Prudent Dietary Pattern upregulated TXNDC17, which modulates NF-κB. NF-κB signaling plays a role in inflammation, apoptosis and a variety of cellular stress responses [15]. TXNDC17 overexpression has been shown to inhibit TNF-α induced activation of NF-κB [53], thus its upregulation could contribute to the decrease in inflammation and cancer previously noted with a Prudent Dietary Pattern [20].

We found that the majority of genes differentially expressed between intake levels of whole grains and vegetables were down-regulated, especially for vegetable intake. The genes that are differentially expressed with level of whole grain and vegetable consumption are associated with NF-κB signaling, regulation of apoptosis, cytoskeleton dynamics, and carbohydrate metabolism. Carbohydrates are important modulators of insulin action on glucose metabolism [54], which is consistent with our findings regarding PFKL, ABCC3 and GMPPB down-regulated expression with whole grain consumption and HSPG2 and REG4 which were down-regulated by high vegetable consumption. Both fiber and polyphenols attenuate inflammation [46]. Phenolic compounds are major bioactive compounds in whole grains and polyphenols are known to interfere with the NF-κB signaling pathway [55], which is consistent with our data showing a downregulation of genes involved in the NF-κB signaling. Our data suggests that the previously described polyphenolic perturbation of nuclear signaling, and anti-inflammatory effects [56], may be explained by the differential gene expression seen with both whole grain and vegetable levels of intake.

Some genes were up-regulated by high vegetable intake, including ANP32A, NECAP1, and HELZ. ANP32A, encodes for a tumor suppressor [57]; HELZ, encodes an RNA helicase, which has been shown in conjunction with other genes to regulate downstream genes as a transcription factor containing histone methyltransferase activity [58]. In association with SMYD3, HELZ plays a role in the proliferation of cancer cells [58]. It has also been suggested that HELZ plays a wider role in global translation activation and the phosphorylation (activation) of Ribosomal protein S6, which in turn activates eIF4b [59]. This indicates that the upregulation of HELZ, could help explain the association of vegetables with several factors, including inflammation [20, 46] through the translational activation of numerous genes within the inflammatory process.

IPA found three interaction networks with meat consumption and three interaction networks associated with non-meat consumption. This correlates with earlier IPA studies showing that Western Dietary Pattern is associated with five canonical pathways related to cancer, six pathways related to immune and/or inflammatory responses and three pathways related to cardiovascular signaling and that Prudent Dietary Pattern is associated with nine IPA canonical pathways associated with immune/inflammatory response, and six pathways associated with cancer [20]. It has previously been noted that Western Dietary Patterns are associated with a pro-inflammatory gene expression profile and that Prudent Dietary patterns are associated with an anti-inflammatory gene expression profile [60], which would be consistent with our described networks. Moreover our networks suggest that while both meat and non-meat related dietary variables are associated with the cellular functions of intracellular transport and ECM regulation, the signal transduction pathways differ between meat and non-meat networks. We found meat consumption to be associated with TP53, NFκB1, and MAPK signaling through the ERK1/2 signaling pathway. The Ras-MAPK system is involved in cell proliferation [61]; moreover, our findings in this network are consistent with earlier findings that high fat diets, the cooked meat carcinogen 2-Amino-1-Methyl-6-phenylimidazo[4,5-b]-pyridine (PhiP), and sodium nitrate can activate the MAPK pathways [9, 16, 62]. Non-meat consumption appears to play a greater role in the regulation of the PI3K complex, TNF-signaling, and TGFB1-signaling pathways. All of these pathways have been identified as major signaling pathways for colon cancer [63–66].

To the best of our knowledge this is one of the few studies to examine the role of diet on gene expression using data from a population-based study and comprehensive gene expression data. Our study has a number of strengths including using RNA-Seq which produces global gene expression data for each RNA sample and is an ideal method to undertake a discovery study such as ours [67, 68]. However, it is important to keep in mind that the gene expression profile is most relevant to current diet exposure. The time between tissue ascertainment and referent period for diet in our study could be from several months to three or possibly four years. While lack of findings could indicate a disparity in time between exposure and tissue sample acquirement, finding associations would imply that the exposure is recent enough, in that dietary patterns are consistent over time, to alter the expression. Additionally, we have utilized normal colonic mucosa, so genes would have to be expressed in colon tissue for detection. It should be recognized that different platforms carry different technical strengths and weaknesses that can influence results. Thus, it is essential to validate these findings in other populations using the methods we used here as well as other platforms to better understand associations between diet and gene expression. We utilized DESeq2 to assess gene expression data adjusting for age and sex. We previously have shown associations between cigarette smoking and alcohol and gene expression in these samples [69]. Likewise, we report statistically significant differential expression as an indicator of dysregulated genes, however, it is not clear the level of dysregulation that is necessary to result in functional significance.

Although this study was conducted in a rigorous manner, there are limitations. Our dietary data were collected shortly after diagnosis. Although our dietary questionnaire allowed for reporting of over 800 food items and we implemented extremely rigid quality control procedures, it is possible that recall could have been influenced by the disease status. While this is a possibility, it should be noted that our risk estimates for diet and colon cancer are almost identical to those reported by large cohort studies [70]. Additionally, these data reflect foods consumed in the early 1990s. While we believe that people still eat the foods that we report here, different dietary patterns may be more important today than the Western and Prudent diet that we report.

Our data support the hypothesis that diet is associated with gene expression. Our data showed that diet was involved in de-regulation of genes involved in numerous pathways and functions such as Wnt-signaling, MAPK, NF-κB, TP53, ECM maintenance, cytoskeletal structure, and cell cycle regulation. These data suggest that differential gene expression associated with dietary intake is a possible mechanism by which diet can influence a variety of biological processes and functions. Focused functionality studies evaluating dietary influence on gene expression are needed.

Supplementary Material

Supplemental Digital Content

Supplemental Table 1. Top genes associated with meat and non-meat networks from IPA

Abbreviations

KPMRP

Kaiser Permanente Medical Research Program

APC

adenomatous polyposis coli

BH

Benjamini and Hochberg procedure

FDR

false discovery rate

IPA

Ingenuity Pathway Analysis

Wnt

Wingless Type

ECM

Extracellular Matrix

MMP

matrix metalloproteinase

TIMPs

tissue inhibitor of metalloproteinases

FGF1

acidic fibroblast growth factor

IL

interleukin

NF-κB

nuclear factor kappa b

mTOR

mechanistic target of rapamycin

Nrf2

NF-E2-related factor 2

PAK

p21-activated kinase

MAPK

mitogen activated protein kinase

PhiP

2-Amino-1-Methyl-6-phenylimidazo[4,5-b]-pyridine