Secondary Logo

Journal Logo

ARTICLE: COLON

The Sulfur Microbial Diet and Risk of Colorectal Cancer by Molecular Subtypes and Intratumoral Microbial Species in Adult Men

Sikavi, Daniel R. MD1,2; Nguyen, Long H. MD, MS2,3,4; Haruki, Koichiro MD, PhD5; Ugai, Tomotaka MD, PhD5,6; Ma, Wenjie ScD2,3,4; Wang, Dong D. ScD4,7; Thompson, Kelsey N. PhD4,8; Yan, Yan PhD4; Branck, Tobyn MS4; Wilkinson, Jeremy E. PhD4; Akimoto, Naohiko MD, PhD5; Zhong, Rong PhD5; Lau, Mai Chan PhD5; Mima, Kosuke MD, PhD5; Kosumi, Keisuke MD, PhD5; Morikawa, Teppei MD, PhD5; Rimm, Eric B. ScD7; Garrett, Wendy S. MD, PhD8,9; Izard, Jacques PhD10,11; Cao, Yin ScD, MPH12,13; Song, Mingyang MBBS, ScD2,3,6,7; Huttenhower, Curtis PhD4,8,14; Ogino, Shuji MS, MD, PhD5,6,8,15; Chan, Andrew T. MD, MPH2,3,8,14

Author Information
Clinical and Translational Gastroenterology: August 2021 - Volume 12 - Issue 8 - p e00338
doi: 10.14309/ctg.0000000000000338
  • Open

Abstract

INTRODUCTION

Colorectal cancer (CRC) is the third most common and second most lethal cancer globally (1). A modifiable lifestyle is a major risk factor in the development of sporadic CRC (2–5). The link between dietary intake and CRC risk has been well-documented (6–8). However, the degree to which gut microbial communities mediate the relationship between diet and carcinogenesis is less well-understood (9,10). Specifically, sulfur-metabolizing bacteria have been implicated in CRC etiopathogenesis (11–16). This phylogenetically diverse group of microbes can convert dietary sulfur into hydrogen sulfide (H2S) gas (17). Data from in vitro and animal studies have linked H2S in the colorectum to several carcinogenic processes, including direct DNA damage (18–20), promotion of a proinflammatory phenotype (17,21), disruption of the protective colonic mucus bilayer (22), and inappropriate cell cycle progression (11). We recently identified a gut microbiome–derived dietary pattern—the sulfur microbial diet—associated with a greater abundance of sulfur-metabolizing bacteria in healthy adults. This dietary pattern was characterized by a higher intake of processed meats and low-calorie drinks and decreased intake of vegetables and legumes, foods previously linked to risk of CRC (23). Long-term adherence to this pattern of intake was subsequently associated with an increased incidence of distal colon and rectal cancers (23).

Separately, several studies have previously described molecular and intratumoral microbial heterogeneity of CRC by both anatomic subsite (24–26) and dietary intake (27–33)—i.e., different dietary risk factors may give rise to heterogenous tumors with respect to molecular subtypes, the intratumoral microbes they harbor, and even where in the colorectum these neoplastic lesions tend to occur. In particular, several studies have shown that the association between dietary risk factors and CRC may differ by Kirsten rat sarcoma (KRAS), v-raf murine sarcoma viral oncogene homolog B1 (BRAF), phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), long-interspersed nucleotide element-1 (LINE-1) methylation, microsatellite instability (MSI), CpG island methylator phenotype (CIMP), prostaglandin synthase 2 (PTGS2) (cyclooxygenase-2), and CTNNB1 (beta-catenin) status (27,28,31,34–40). Similarly, previous work suggests that diet can alter the abundance of 2 important microbes implicated in CRC pathogenesis, Fusobacterium nucleatum and Bifidobacterium spp. (41–43).

We hypothesized that the association between a diet associated with a greater abundance of sulfur-metabolizing bacteria and distal CRC is driven by particular molecular CRC subtypes or relative enrichment or depletion of intratumoral CRC-associated microbes (44,45). The study of these molecular and microbial CRC subtypes may offer a more complete and precise mechanistic understanding of the relationship between dietary intake, the gut microbiome, and colorectal tumorigenesis. Thus, in a large US prospective cohort, we examined the relationship between the sulfur microbial diet and incidence of CRC according to several molecular tumor markers and intratumoral microbes previously linked to CRC.

METHODS

Study population

We enrolled participants from the Health Professionals Follow-up Study (HPFS). The HPFS is an ongoing prospective cohort study of 51,529 US male podiatrists, dentists, osteopathic physicians, veterinarians, pharmacists, and optometrists. Participants were aged 40–75 years at enrollment in 1986 and followed with biennial questionnaires on medical, lifestyle, and other health-related information. Dietary intake was assessed every 4 years through validated semiquantitative food frequency questionnaires. Follow-up among eligible subjects exceeds 90% (46). The cohort has previously been described in detail (46). The study protocol was approved by the Institutional Review Boards of the Brigham and Women's Hospital, the Harvard T.H. Chan School of Public Health, and those of participating registries as required.

Assessment of the sulfur microbial diet

Self-reported dietary intake was assessed through semiquantitative food frequency questionnaires administered every 4 years from 1986 to 2010. These questionnaires have been validated and described in detail (47). The food frequency questionnaire includes 131 food items with specified serving sizes, of which participants indicated their average frequency of consumption over the past year. Intake frequency ranged from never or less than once per month to 6+ times per day and was converted to servings/d. Total caloric intake was calculated by summing energy intake across all food groups.

Using a previously described method, we linked dietary intake of food and the log-transformed relative abundance of 43 putative sulfur-metabolizing species (see Table 1, Supplementary Digital Content 1, http://links.lww.com/CTG/A652) (23). Briefly, the sulfur microbial diet was derived using long-term dietary intake and longitudinal stool metagenomes from 307 men in the Men's Lifestyle Validation Study, a nested developmental cohort within the HPFS (∼0.6% of the original HPFS population). Reduced rank regression and stepwise linear regression analyses were used to identify food groups associated with increased or decreased relative abundance of 43 sulfur-metabolizing bacterial species by summing the intake of foods retained from the final stepwise linear regression analyses weighted by their regression coefficients. The component food groups were processed meat, liquor, and low-calorie drinks (each positively associated with the abundance of sulfur-metabolizing bacteria), as well as beer, fruit juice, legumes, mixed (other) vegetables, and sweets/desserts (each negatively associated). We found that the sulfur microbial diet explained 2% of variation in Bray-Curtis distances (R2), comparable in magnitude with recent antibiotic use (1.5%).

To represent long-term usual dietary habits (48), sulfur microbial diet scores were updated at each follow-up cycle using the cumulative average method with each score averaged across all assessments before the current questionnaire. Dietary scores were then categorized into tertiles. The food-based sulfur microbial diet score represents a data-driven prediction for how much sulfur-metabolizing bacteria an individual may harbor over the long term. Notably, although the sulfur microbial diet shared some foods with dietary patterns previously linked to CRC (e.g., the Western diet) (27,49–51), sulfur microbial diet scores were not associated with Western dietary scores, which suggests that the sulfur microbial diet may capture a novel signal in the established diet-CRC relationship.

Assessment of CRC cases and subtype

Cases of incident CRC were reported by participants on biennial questionnaires or were identified by next of kin, postal authorities, the National Death Index, or death certificates. Study physicians blinded to risk factor status reviewed relevant records to confirm cases and extract data on anatomic site, histology, and stage of the tumor. Distal colorectal tumors were defined as tumors located from the splenic flexure to the rectum. For CRC cases with available tumor tissue, we retrieved formalin-fixed paraffin-embedded tissue blocks from hospitals throughout the United States, as previously described (52). Adjacent normal tissue and tumor sections from all CRC cases were reviewed by a pathologist (S.O).

The presence of intratumoral F. nucleatum (53) and Bifidobacterium spp. (54) was assessed by real-time polymerase chain reaction (PCR). Positivity was defined as a detectable level of bacterial DNA, and negativity was defined as an undetectable level as previously described. PCR and pyrosequencing were performed to detect mutations in KRAS (codons 12, 13, 61, and 146) (55), BRAF (codon 600) (56), and PIK3CA (exons 9 and 20) (57). LINE-1 methylation status was measured by PCR on bisulfite-treated DNA and pyrosequencing and categorized as high if ≥ 60% of sites were methylated and low if <60% were methylated (58). MSI status was determined using 10 microsatellite markers (D17S250, D18S55, D18S56, D18S67, D18S487, D2S123, D5S346, BAT25, BAT26, and BAT40) (59). Tumors were classified as MSI-high if 30% or more of the markers demonstrated instability. We quantified DNA methylation using polymerase chain reaction in 8 CIMP-specific promoters (MLH1, NEUROG1, RUNX3, CACNA1G, CDKN2A [p16], CRABP1, IGF2, and SOCS1) (60). We classified tumors as CIMP-high if 6 or more promoters were methylated and as CIMP-low/negative if 0–5 promoters were methylated (61). PTGS2 (cyclooxygenase-2) (62) and nuclear CTNNB1 (beta-catenin) (52) expression was measured by immunohistochemistry as described previously and graded as negative/low, intermediate, and high.

Assessments of covariates

Height and weight were reported at study inception, and weight was updated biennially. Body mass index (BMI) was calculated as weight in kilograms/height in meters squared. Physical activity was self-reported using validated questionnaires every 2–4 years (63). We also assessed and updated the age they started or stopped smoking, number of cigarettes smoked daily, family history of CRC among first-degree relatives, regular use of aspirin, previous health care engagement (assessed as a visit to a care provider in the past 2 years), and history of lower gastrointestinal endoscopy.

Statistical analyses

We excluded participants with a history of CRC or inflammatory bowel disease before study baseline or during follow-up. We also excluded individuals with missing information on dietary intake and those who reported implausible energy intake (<800 or >4,200 kcal/d) at baseline. Follow-up time accrued from study enrollment until the date of CRC diagnosis, death from any cause, or the end of follow-up (January 31, 2012), whichever occurred first.

To examine whether the association between the sulfur microbial diet and CRC differed by subtype, we used Cox proportional hazards models with a duplication method for competing risks and computed age and multivariable-adjusted hazard ratios (HRs) and their 95% confidence intervals (CIs). We tested for heterogeneity by using a meta-regression method with a subtype-specific random effect term (64). We separately evaluated heterogeneity by subtype among distal colon and rectal tumors to evaluate subsite-specific effects, given the previously observed association between the sulfur microbial diet and tumors arising from the distal colorectum (23).

To test for trend, participants were assigned the median value of their dietary pattern tertile. Multivariate analyses were adjusted for family history of CRC in a first-degree relative (yes/no), BMI (categories, <21, 21–24.9, 25–29.9, 30–31.9, and ≥32 kg/m2), physical activity (metabolic equivalent task hr/wk, quintiles), smoking (categories: never, past, and current: 1–14, 15–24, and ≥25 cigarettes/d), regular aspirin use (yes/no), total caloric intake (continuous), previous lower endoscopy within the past 2 years (yes/no), and physical examination in the past 2 years (yes/no). All analyses were additionally stratified by age (continuous) and calendar year. Covariates were chosen a priori and prospectively updated. For missing data, nonmissing data from 1 previous data cycle were carried forward. Two-sided P values <0.005 were considered statistically significant as recommended (65). SAS version 9.4 (Cary, NC) was used for all statistical analysis.

RESULTS

Among 48,246 eligible participants with 1,077,517 person-years of follow-up, the mean age of participants at study baseline was 54.2 ± 9.8 years. Subjects in the highest tertile of the sulfur microbial diet generally had poorer health-related indices. They tended to have higher BMIs, consume more alcohol, and were more likely to smoke and take aspirin regularly (Table 1).

Table 1. - Baseline age-standardized characteristics by the sulfur microbial diet score, HPFS (1986)
Sulfur microbial diet score
Tertile 1 (n = 16,123) Tertile 2 (n = 16,195) Tertile 3 (n = 15,928)
Age, yr 54.2 (10.0) 54.2 (9.8) 54.1 (9.6)
BMI, kg/m2 25.0 (3.1) 25.4 (3.2) 26.2 (3.4)
Physical activity, METS-hr/wk 21.0 (27.3) 18.3 (26.3) 16.9 (24.4)
Former smokers, % 39 41 46
Current smokers, % 7.8 8.4 13
White race, % 96 95 96
Family history of CRC, % 15 15 15
Regular aspirin use, % 28 29 31
Screening lower endoscopy within the past 2 yr, % 27 27 27
Physical examination within the past 2 yr, % 52 51 50
Alcohol intake (g/d) 11.9 (16.1) 8.1 (10.8) 14.3 (18.0)
Total energy intake, kcal/d 2,268 (623) 1863 (552) 1829 (582)
Dietary intake (servings/wk)
 Processed meats 1.9 (2.1) 2.2 (2.2) 3.7 (3.9)
 Liquor 1.1 (2.6) 1.5 (3.0) 4.8 (7.6)
 Low-calorie drinks 1.4 (2.8) 2.0 (3.4) 7.1 (9.6)
 Beer 3.6 (6.7) 1.2 (2.3) 1.0 (1.9)
 Fruit juice 8.0 (8.0) 5.0 (4.3) 3.7 (3.9)
 Legumes 4.5 (3.2) 2.7 (1.7) 2.2 (1.6)
 Other vegetables 5.1 (4.0) 3.3 (2.3) 2.7 (2.1)
 Sweets and desserts 10.9 (11.2) 6.1 (5.5) 4.8 (5.1)
All values other than age have been directly standardized to the age distribution (in 5-year age groups) of the entire study population. Mean (SD) is presented for continuous variables.
BMI, body mass index; CRC, colorectal cancer; HPFS, Health Professionals Follow-Up Study; kcal, kilocalories; METS, metabolic equivalent of tasks; m, meters.

We documented 1,264 incident cases of CRC. Among CRC cases with information on anatomic site, we identified 637 distal colon and rectal cancers. Tumor tissue data were available for approximately 40% of total CRCs and 50% of distal colon and rectal cancers. The baseline characteristics of participants with CRC whose tumors we analyzed were overall similar to those of participants whose tumors we did not analyze (data not shown).

We examined whether the association between the sulfur microbial diet and total CRC differed by relevant molecular tumor subtypes (Table 2). After accounting for multiple hypothesis testing, there was no conclusive evidence of heterogeneity. However, the association of the sulfur microbial diet with CRC incidence seemed to differ by PTGS2 status (Pheterogeneity = 0.04) when accounting for multiple testing with an α level of 0.005. Multivariable HRs for the highest tertile of sulfur microbial diet scores (vs the lowest) were 1.31 (95% CI: 0.99–1.74; Ptrend = 0.07) for PTGS2-high subtype, 0.99 (95% CI: 0.61–1.59; Ptrend = 0.95) for PTGS2-intermediate subtype, and 0.65 (95% CI, 0.36–1.15; Ptrend = 0.17) for PTGS2-negative/low subtype. There was no evidence for heterogeneity by KRAS, BRAF, PIK3CA, LINE-1, CIMP, MSI, or CTNNB1 status. Similarly, the relationship between the sulfur microbial diet and total CRC did not differ by the presence of intratumoral microbes F. nucleatum or Bifidobacterium spp. (Table 3).

Table 2. - Association between sulfur microbial diet scores and incidence of CRC by molecular subtype
Molecular subtype T1 T2 T3 P trend P heterogeneitya
KRAS
KRAS-wildtype CRC
  No. of cases (n = 339), No. 101 123 115
  Age-adjusted HR (95% CI)b 1 (referent) 1.23 (0.94–1.61) 1.23 (0.93–1.62) 0.14
  Multivariable HR (95% CI)c 1 (referent) 1.19 (0.91–1.56) 1.15 (0.87–1.52) 0.34
KRAS-mutant CRC 0.24
  No. of cases (n = 223), No. 79 76 68
  Age-adjusted HR (95% CI)b 1 (referent) 0.97 (0.70–1.33) 0.95 (0.68–1.32) 0.73
  Multivariable HR (95% CI)c 1 (referent) 0.94 (0.68–1.29) 0.89 (0.63–1.24) 0.47
BRAF
BRAF-wildtype CRC
  No. of cases (n = 517), No. 161 181 175
  Age-adjusted HR (95% CI)b 1 (referent) 1.13 (0.91–1.41) 1.18 (0.94–1.48) 0.15
  Multivariable HR (95% CI)c 1 (referent) 1.10 (0.88–1.37) 1.11 (0.88–1.39) 0.40
BRAF-mutant CRC 0.22
  No. of cases (n = 45), No. 17 18 10
  Age-adjusted HR (95% CI)b 1 (referent) 1.07 (0.55–2.07) 0.64 (0.29–1.41) 0.29
  Multivariable HR (95% CI)c 1 (referent) 1.04 (0.53–2.02) 0.60 (0.27–1.32) 0.22
PIK3CA
PIK3CA-wildtype CRC
  No. of cases (n = 434), No. 135 157 142
  Age-adjusted HR (95% CI)b 1 (referent) 1.17 (0.92–1.48) 1.13 (0.89–1.44) 0.33
  Multivariable HR (95% CI)c 1 (referent) 1.14 (0.9–1.45) 1.07 (0.84–1.38) 0.59
PIK3CA-mutant CRC 0.68
  No. of cases (n = 94), No. 33 31 30
  Age-adjusted HR (95% CI)b 1 (referent) 0.93 (0.57–1.53) 1.01 (0.61–1.67) 0.98
  Multivariable HR (95% CI)c 1 (referent) 0.90 (0.55–1.48) 0.96 (0.58–1.58) 0.85
LINE-1
 LINE-1 methylation high CRC
  No. of cases (n = 364), No. 115 131 118
  Age-adjusted HR (95% CI)b 1 (referent) 1.15 (0.89–1.49) 1.13 (0.86–1.47) 0.39
  Multivariable HR (95% CI)c 1 (referent) 1.12 (0.86–1.44) 1.05 (0.80–1.37) 0.74
 LINE-1 methylation low CRC 0.89
  No. of cases (n = 200), No. 64 71 65
  Age-adjusted HR (95% CI)b 1 (referent) 1.12 (0.79–1.57) 1.10 (0.77–1.56) 0.61
  Multivariable HR (95% CI)c 1 (referent) 1.08 (0.77–1.52) 1.02 (0.71–1.45) 0.93
CIMP
 CIMP-low/negative CRC
  No. of cases (n = 454), No. 144 154 156
  Age-adjusted HR (95% CI)b 1 (referent) 1.08 (0.85–1.36) 1.17 (0.93–1.49) 0.19
  Multivariable HR (95% CI)c 1 (referent) 1.05 (0.83–1.33) 1.10 (0.87–1.40) 0.44
 CIMP-high CRC 0.15
  No. of cases (n = 63), No. 24 26 13
  Age-adjusted HR (95% CI)b 1 (referent) 1.10 (0.63–1.92) 0.60 (0.31–1.19) 0.18
  Multivariable HR (95% CI)c 1 (referent) 1.07 (0.61–1.87) 0.56 (0.29–1.12) 0.12
MSI
 Non–MSI-high CRC
  No. of cases (n = 489), No. 157 166 166
  Age-adjusted HR (95% CI)b 1 (referent) 1.07 (0.85–1.34) 1.16 (0.92–1.45) 0.22
  Multivariable HR (95% CI)c 1 (referent) 1.04 (0.83–1.30) 1.08 (0.86–1.37) 0.51
 MSI-high CRC 0.97
  No. of cases (n = 64), No. 17 29 18
  Age-adjusted HR (95% CI)b 1 (referent) 1.74 (0.95–3.17) 1.16 (0.59–2.25) 0.67
  Multivariable HR (95% CI)c 1 (referent) 1.68 (0.92–3.07) 1.08 (0.55–2.11) 0.84
PTGS2
 PTGS2-negative/low CRC
  No. of cases (n = 89), No. 31 39 19
  Age-adjusted HR (95% CI)b 1 (referent) 1.29 (0.80–2.08) 0.69 (0.39–1.23) 0.26
  Multivariable HR (95% CI)c 1 (referent) 1.25 (0.78–2.02) 0.65 (0.36–1.15) 0.17
 PTGS2-intermediate CRC 0.04
  No. of cases (n = 106), No. 36 36 34
  Age-adjusted HR (95% CI)b 1 (referent) 1.03 (0.65–1.64) 1.05 (0.65–1.69) 0.86
  Multivariable HR (95% CI)c 1 (referent) 1.01 (0.63–1.60) 0.99 (0.61–1.59) 0.95
 PTGS2-high CRC
  No. of cases (n = 328), No. 93 118 117
  Age-adjusted HR (95% CI)b 1 (referent) 1.30 (0.98–1.71) 1.39 (1.05–1.84) 0.03
  Multivariable HR (95% CI)c 1 (referent) 1.26 (0.95–1.67) 1.31 (0.99–1.74) 0.07
CTNNB1 (Nuclear)
Nuclear CTNNB1-negative/low CRC
  No. of cases (n = 239), No. 73 96 70
  Age-adjusted HR (95% CI)b 1 (referent) 1.36 (1.00–1.86) 1.08 (0.77–1.51) 0.64
  Multivariable HR (95% CI)c 1 (referent) 1.33 (0.98–1.81) 1.03 (0.73–1.44) 0.88
Nuclear CTNNB1-intermediate CRC 0.12
  No. of cases (n = 170), No. 54 58 58
  Age-adjusted HR (95% CI)b 1 (referent) 1.09 (0.75–1.59) 1.18 (0.81–1.72) 0.40
  Multivariable HR (95% CI)c 1 (referent) 1.07 (0.73–1.55) 1.12 (0.77–1.64) 0.56
 Nuclear CTNNB1-high CRC
  No. of cases (n = 87), No. 24 28 35
  Age-adjusted HR (95% CI)b 1 (referent) 1.21 (0.70–2.10) 1.65 (0.98–2.79) 0.06
  Multivariable HR (95% CI)c 1 (referent) 1.18 (0.68–2.05) 1.56 (0.92–2.65) 0.09
BRAF, v-raf murine sarcoma viral oncogene homolog B1; CI, confidence interval; CIMP, CpG island methylator phenotype; CRC, colorectal cancer; CTNNB1, catenin beta-1; HR, hazard ratio; KRAS, Kirsten rat sarcoma; LINE-1, long-interspersed nucleotide element-1; MSI, microsatellite instable; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha; PTGS2, prostaglandin synthase 2; T, tertile.
aHeterogeneity was tested using a meta-regression method with a subtype-specific random effect term among multivariable models.
bModels stratified by age and calendar year and adjusted for total caloric intake (kcals/d).
cModels stratified as above and additionally adjusted for family history of CRC in any first-degree relative, BMI (categories), physical activity (METS-hr/wk), smoking (never vs past vs current), recent aspirin use, history of previous lower gastrointestinal endoscopy, and history of physical examination.

Table 3. - Association between sulfur microbial diet scores and incidence of CRC by intratumoral microbes of interest
Molecular subtype T1 T2 T3 P trend P heterogeneitya
Fusobacterium nucleatum
F. nucleatum–negative CRC
  No. of cases (n = 462), No. 145 170 147
  Age-adjusted HR (95% CI)b 1 (referent) 1.19 (0.94–1.49) 1.11 (0.88–1.41) 0.38
  Multivariable HR (95% CI)c 1 (referent) 1.15 (0.92–1.45) 1.05 (0.82–1.33) 0.72
F. nucleatum–positive CRC 0.35
  No. of cases (n = 60), No. 21 12 27
  Age-adjusted HR (95% CI)b 1 (referent) 0.57 (0.28–1.16) 1.37 (0.77–2.44) 0.25
  Multivariable HR (95% CI)c 1 (referent) 0.56 (0.27–1.14) 1.30 (0.73–2.32) 0.33
Bifidobacterium spp.
Bifidobacterium spp.–negative CRC
  No. of cases (n = 378), No. 115 132 131
  Age-adjusted HR (95% CI)b 1 (referent) 1.15 (0.89–1.48) 1.23 (0.95–1.60) 0.11
  Multivariable HR (95% CI)c 1 (referent) 1.12 (0.87–1.45) 1.16 (0.89–1.51) 0.27
Bifidobacterium spp.–positive CRC 0.33
  No. of cases (n = 148), No. 49 55 44
  Age-adjusted HR (95% CI)b 1 (referent) 1.14 (0.77–1.68) 0.98 (0.65–1.48) 0.93
  Multivariable HR (95% CI)c 1 (referent) 1.10 (0.74–1.62) 0.91 (0.60–1.39) 0.68
CI, confidence interval; CRC, colorectal cancer; HR, hazard ratio; spp, species; T, tertile.
aHeterogeneity was tested using a meta-regression method with a subtype-specific random effect term among multivariable models.
bModels stratified by age and calendar year and adjusted for total caloric intake (kcals/d).
cModels stratified as above and additionally adjusted for family history of CRC in any first-degree relative, BMI (categories), physical activity (METS-hr/wk), smoking (never vs past vs current), recent aspirin use, history of previous lower gastrointestinal endoscopy, and history of physical examination.

Given our previous findings linking the sulfur microbial diet and cancers of the distal colorectum, we subsequently focused on CRC that had arisen in the distal colon or rectum. We found trends for PTGS2-high (HR 1.58 [95% CI: 1.09–2.28], Ptrend = 0.02), CTNNB1-high (HR 2.10 [95% CI: 1.04–4.21], Ptrend = 0.04), and KRAS-wildtype tumors (HR 1.52 [95% CI: 1.06–2.17], Ptrend = 0.02), but we did not observe statistically significant heterogeneity (Pheterogeneity = 0.08, Pheterogeneity = 0.25, and Pheterogeneity = 0.06, respectively) (Table 4). There was no apparent heterogeneity by BRAF, PIK3CA, LINE-1, CIMP, or MSI status among distal colon and rectal cancers (Table 4). We observed that the association of the sulfur microbial diet with distal colon and rectal cancers seemed to differ by the presence of intratumoral Bifidobacterium spp., although the difference was not statistically significant (Pheterogeneity = 0.03) when accounting for adjustment for multiple testing with an α level of 0.005 (Table 5). The multivariable HR for the highest tertile of sulfur microbial diet scores (vs the lowest) was 1.65 (95% CI: 1.14–2.39; Ptrend = 0.01) for Bifidobacterium-negative distal CRC, but these associations were not observed in Bifidobacterium-negative total CRC (Ptrend = 0.27). Similar to the result with overall CRC, there was no observed heterogeneity in the association of diet and incidence of distal colon and rectal cancers by F. nucleatum status (Table 5).

Table 4. - Association between sulfur microbial diet scores and incidence of distal colon and rectal cancers by molecular subtype
Molecular subtype T1 T2 T3 P trend P heterogeneitya
KRAS
 KRAS-wildtype distal CRC
  No. of cases (n = 208), No. 53 72 83
  Age-adjusted HR (95% CI)b 1 (referent) 1.35 (0.94–1.94) 1.66 (1.16–2.36) 0.005
  Multivariable HR (95% CI)c 1 (referent) 1.30 (0.91–1.88) 1.52 (1.06–2.17) 0.02
 KRAS-mutant distal CRC 0.06
  No. of cases (n = 109), No. 36 39 34
  Age-adjusted HR (95% CI)b 1 (referent) 1.09 (0.69–1.72) 1.02 (0.63–1.64) 0.95
  Multivariable HR (95% CI)c 1 (referent) 1.03 (0.65–1.63) 0.93 (0.57–1.50) 0.76
BRAF
 BRAF-wildtype distal CRC
  No. of cases (n = 310), No. 85 108 117
  Age-adjusted HR (95% CI)b 1 (referent) 1.27 (0.94–1.70) 1.46 (1.09–1.96) 0.01
  Multivariable HR (95% CI)c 1 (referent) 1.22 (0.91–1.63) 1.34 (1.00–1.81) 0.05
 BRAF-mutant distal CRC 0.43
  No. of cases (n = 8), No. 3 3 2
  Age-adjusted HR (95% CI)b 1 (referent) 1.01 (0.20–5.00) 0.68 (0.11–4.07) 0.68
  Multivariable HR (95% CI)c 1 (referent) 0.98 (0.20–4.85) 0.62 (0.10–3.75) 0.61
PIK3CA
 PIK3CA-wildtype distal CRC
  No. of cases (n = 257), No. 69 91 97
  Age-adjusted HR (95% CI)b 1 (referent) 1.32 (0.96–1.82) 1.49 (1.08–2.05) 0.02
  Multivariable HR (95% CI)c 1 (referent) 1.27 (0.92–1.76) 1.37 (0.99–1.90) 0.06
PIK3CA-mutant distal CRC 0.53
  No. of cases (n = 39), No. 12 14 13
  Age-adjusted HR (95% CI)b 1 (referent) 1.15 (0.53–2.49) 1.17 (0.53–2.59) 0.70
  Multivariable HR (95% CI)c 1 (referent) 1.10 (0.51–2.38) 1.08 (0.49–2.39) 0.86
LINE-1
 LINE-1 methylation high distal CRC
  No. of cases (n = 186), No. 55 62 69
  Age-adjusted HR (95% CI)b 1 (referent) 1.14 (0.79–1.65) 1.36 (0.94–1.95) 0.10
  Multivariable HR (95% CI)c 1 (referent) 1.09 (0.75–1.59) 1.24 (0.86–1.79) 0.26
 LINE-1 methylation low distal CRC 0.61
  No. of cases (n = 134), No. 34 50 50
  Age-adjusted HR (95% CI)b 1 (referent) 1.48 (0.95–2.30) 1.57 (1.01–2.45) 0.05
  Multivariable HR (95% CI)c 1 (referent) 1.42 (0.91–2.20) 1.42 (0.91–2.23) 0.13
CIMP
 CIMP-low/negative distal CRC
  No. of cases (n = 285), No. 81 98 106
  Age-adjusted HR (95% CI)b 1 (referent) 1.21 (0.89–1.64) 1.39 (1.03–1.88) 0.03
  Multivariable HR (95% CI)c 1 (referent) 1.17 (0.86–1.59) 1.28 (0.94–1.74) 0.12
 CIMP-high distal CRC 0.30
  No. of cases (n = 10), No. 4 5 1
  Age-adjusted HR (95% CI)b 1 (referent) 1.26 (0.34–4.72) 0.29 (0.03–2.63) 0.34
  Multivariable HR (95% CI)c 1 (referent) 1.22 (0.32–4.54) 0.27 (0.03–2.42) 0.29
MSI
 Non–MSI-high distal CRC
  No. of cases (n = 299), No. 83 104 112
  Age-adjusted HR (95% CI)b 1 (referent) 1.27 (0.94–1.70) 1.45 (1.08–1.95) 0.01
  Multivariable HR (95% CI)c 1 (referent) 1.22 (0.90–1.64) 1.34 (0.99–1.81) 0.06
 MSI-high distal CRC 0.06
  No. of cases (n = 12), No. 2 4 6
  Age-adjusted HR (95% CI)b 1 (referent) 2.04 (0.37–11.14) 3.30 (0.66–16.44) 0.13
  Multivariable HR (95% CI)c 1 (referent) 1.96 (0.36–10.72) 3.04 (0.61–15.15) 0.16
PTGS2
 PTGS2-negative/low distal CRC
  No. of cases (n = 40), No. 13 16 11
  Age-adjusted HR (95% CI)b 1 (referent) 1.28 (0.61–2.66) 0.95 (0.42–2.13) 0.91
  Multivariable HR (95% CI)c 1 (referent) 1.22 (0.58–2.55) 0.87 (0.38–1.95) 0.74
 PTGS2-intermediate distal CRC 0.08
  No. of cases (n = 49), No. 16 16 17
  Age-adjusted HR (95% CI)b 1 (referent) 1.03 (0.51–2.07) 1.15 (0.58–2.29) 0.71
  Multivariable HR (95% CI)c 1 (referent) 1.00 (0.50–2.00) 1.05 (0.52–2.10) 0.90
 PTGS2-high distal CRC
  No. of cases (n = 206), No. 51 75 80
  Age-adjusted HR (95% CI)b 1 (referent) 1.49 (1.04–2.15) 1.71 (1.19–2.45) 0.004
  Multivariable HR (95% CI)c 1 (referent) 1.44 (1.00–2.08) 1.58 (1.09–2.28) 0.02
CTNNB1 (Nuclear)
 Nuclear CTNNB1-negative/low distal CRC
  No. of cases (n = 110), No. 27 45 38
  Age-adjusted HR (95% CI)b 1 (referent) 1.70 (1.05–2.76) 1.53 (0.93–2.53) 0.11
  Multivariable HR (95% CI)c 1 (referent) 1.64 (1.01–2.67) 1.42 (0.85–2.35) 0.20
 Nuclear CTNNB1-intermediate distal CRC 0.25
  No. of cases (n = 114), No. 37 32 45
  Age-adjusted HR (95% CI)b 1 (referent) 0.87 (0.54–1.40) 1.31 (0.84–2.04) 0.23
  Multivariable HR (95% CI)c 1 (referent) 0.84 (0.52–1.36) 1.22 (0.78–1.91) 0.37
 Nuclear CTNNB1-high distal CRC
  No. of cases (n = 60), No. 12 23 25
  Age-adjusted HR (95% CI)b 1 (referent) 1.94 (0.96–3.92) 2.27 (1.13–4.55) 0.02
  Multivariable HR (95% CI)c 1 (referent) 1.88 (0.93–3.79) 2.10 (1.04–4.21) 0.04
BRAF, v-raf murine sarcoma viral oncogene homolog B1; CI, confidence interval; CIMP, CpG island methylator phenotype; CRC, colorectal cancer; CTNNB1, catenin beta-1; HR, hazard ratio; KRAS, Kirsten rat sarcoma; MSI, microsatellite instable; LINE-1, long-interspersed nucleotide element-1; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha; PTGS2, prostaglandin synthase 2; T, tertile.
aHeterogeneity was tested using a meta-regression method with a subtype-specific random effect term among multivariable models.
bModels stratified by age and calendar year and adjusted for total caloric intake (kcals/d).
cModels stratified as above and additionally adjusted for family history of CRC in any first-degree relative, BMI (categories), physical activity (METS-hr/wk), smoking (never vs past vs current), recent aspirin use, history of previous lower gastrointestinal endoscopy, and history of physical examination.

Table 5. - Association between sulfur microbial diet scores and incidence of distal colon and rectal cancers by intratumoral microbes of interest
Molecular subtype T1 T2 T3 P trend P heterogeneitya
F. nucleatum
F. nucleatum–negative distal CRC
  No. of cases (n = 256), No. 72 92 92
  Age-adjusted HR (95% CI)b 1 (referent) 1.28 (0.93–1.75) 1.37 (0.99–1.89) 0.05
  Multivariable HR (95% CI)c 1 (referent) 1.23 (0.9–1.69) 1.27 (0.91–1.76) 0.16
F. nucleatum–positive distal CRC 0.17
  No. of cases (n = 32), No. 10 5 17
  Age-adjusted HR (95% CI)b 1 (referent) 0.5 (0.17–1.46) 1.83 (0.83–4.02) 0.10
  Multivariable HR (95% CI)c 1 (referent) 0.48 (0.16–1.41) 1.70 (0.77–3.74) 0.14
Bifidobacterium spp.
Bifidobacterium spp.–negative distal CRC
  No. of cases (n = 212), No. 49 80 83
  Age-adjusted HR (95% CI)b 1 (referent) 1.61 (1.12–2.32) 1.79 (1.24–2.57) 0.002
  Multivariable HR (95% CI)c 1 (referent) 1.55 (1.08–2.24) 1.65 (1.14–2.39) 0.01
Bifidobacterium spp.–positive distal CRC 0.03
  No. of cases (n = 84), No. 31 25 28
  Age-adjusted HR (95% CI)b 1 (referent) 0.81 (0.47–1.37) 0.96 (0.57–1.62) 0.86
  Multivariable HR (95% CI)c 1 (referent) 0.77 (0.45–1.32) 0.88 (0.52–1.49) 0.62
CI, confidence interval; CRC, colorectal cancer; HR, hazard ratio; spp, species; T, tertile.
aHeterogeneity was tested using a meta-regression method with a subtype-specific random effect term among multivariable models.
bModels stratified by age and calendar year and adjusted for total caloric intake (kcals/d).
cModels stratified as above and additionally adjusted for family history of CRC in any first-degree relative, BMI (categories), physical activity (METS-hr/wk), smoking (never vs past vs current), recent aspirin use, history of previous lower gastrointestinal endoscopy, and history of physical examination.

DISCUSSION

In a large US prospective cohort of men, we found that the association between the sulfur microbial diet and CRC incidence may differ by molecular and microbial subtypes. We found that a previously described sulfur microbial dietary pattern associated with an increased abundance of cancer-associated sulfur-metabolizing bacteria may be more strongly associated with PTGS2-high tumors. Finally, we observed a relative depletion of intratumoral Bifidobacterium spp., a taxon associated with beneficial health and anticancer effects, associated with higher sulfur microbial diet scores. These results provide early evidence that may link lifestyle and the development of specific CRC subtypes through dietary modulation of gut microbial communities.

Diet has played a long-recognized role in altering CRC risk (8,50,66). Modulation of the gut microbiome may be one mechanism through which certain dietary exposures contribute to or protect against carcinogenesis along the colorectum (10,67,68). Among intestinal microbial communities, sulfur-metabolizing bacteria are an important class of organisms implicated in CRC tumorigenesis. Through the metabolism of dietary sulfur, these organisms may release H2S, a compound whose gut luminal concentration is critically determined by the abundance of sulfur-metabolizing bacteria, as opposed to the sulfur content of foods alone (69,70). Interestingly, beer intake contributed negatively to sulfur microbial diet scores. Although some data have shown an increased risk of CRC associated with heavy beer consumption (71), our results may suggest that modest beer consumption, in combination with other food groups, could be protective against CRC through modulation of the sulfur-metabolizing bacteria. Both H2S and sulfur-metabolizing bacteria have been implicated in carcinogenesis in the colorectum, possibly through mechanisms related to free radical formation (18–20), immunomodulation (21), disruption of the protective colonic mucus bilayer (22), and impaired cell cycle arrest and apoptosis (11).

Our finding of possible diet-induced heterogeneity of risk builds on a well-established body of research on the molecular heterogeneity of CRC. Several previous studies have linked certain dietary exposures with an increased risk of specific CRC molecular subtypes (27–32). In addition, there are data supporting distinct gut microbial communities with different CRC subtypes (72,73). Our results extend these findings and provide new insights in carcinogenic processes induced by the complex interaction between dietary intake and microbial ecology.

We expand on the link between dietary exposures, gut microbial ecology, and carcinogenesis. The genus Bifidobacterium seems to inhibit colorectal carcinogenesis (45,74), modulate gut mucosal barrier function (75), and enhance antitumor immunity (76). Our finding that higher sulfur microbial diet scores may be associated with a higher incidence of Bifidobacterium-negative distal CRC may suggest that this relationship could be mediated by alterations in a distinct class of microbes capable of metabolizing sulfur. Although it is unclear why this association was observed only among distal tumors, it may be a result of heterogeneity in microbial biogeography (77,78). By contrast, we did not identify heterogeneity by F. nucleatum status. A species capable of dietary sulfur metabolism, F. nucleatum is relatively rare in healthy populations and was not included in the derivation of the sulfur microbial diet. The observed absence of an effect by F. nucleatum status may reflect distinct regional ecology in the distal colorectum, where F. nucleatum–positive tumors are less common (79,80).

Although the precise mechanism by which diet and microbial communities could increase the risk of PTGS2-high cancers is unknown, our findings may implicate carcinogenesis by proinflammatory mechanisms and abnormalities in bile acid metabolism (81,82). PTGS2 is involved in the conversion of free arachidonic acid into prostaglandins, which are potent proinflammatory mediators in colorectal carcinogenesis (83,84). PTGS2-mediated inflammation secondary to cell membrane damage (82), activation of tumor-associated fibroblasts (85), and epidermal growth factor receptor signaling (86) can be stimulated in response to deoxycholic acid, a secondary bile acid. Deoxycholic acid is a potential carcinogen that is produced in high levels in individuals who consume diets high in animal protein, a primary component of the sulfur microbial diet. Animal protein contains high levels of sulfur-containing amino acids such as taurine and cysteine, the former of which can be transformed to deoxycholic acid in the presence of taurine-respiring bacteria such as Bilophila wadsworthia (82), a microbe whose relative enrichment was associated with higher sulfur microbial diet scores (23). Taken together, these studies offer possible mechanistic explanations for the elevated risk associated with a key component of the sulfur microbial diet, directly implicating aberrant signaling along PTGS2 pathways.

There are several strengths to this study. We leveraged a large, prospective cohort with over 20 years of follow-up with validated dietary instruments. Consequently, we were able to assess the link between long-term dietary trends and development of cancer. At the same time, the prospective nature of data collection limited the potential for recall and ascertainment bias. The detailed data across a wide range of lifetime exposures allowed us to adjust for multiple potential confounders. Finally, the sulfur microbial diet was previously identified through incorporating metagenomic taxonomic assignment and metatranscriptomic functional activity.

We also acknowledge several limitations. Despite the large size of the cohort, there was limited power to detect heterogeneity in the association of dietary exposure and subtypes of CRC. The sulfur microbial diet was derived among bacterial species with sufficient prevalence and abundance in the healthy human gut. As a result, our dietary exposure may not necessarily capture specific pathogenic bacterial strains or less prevalent but highly active taxa in the CRC microbiome (e.g., F. nucleatum) (44,87,88). Our study was conducted among male, mostly white health professionals, potentially limiting the generalizability of our results. It is possible that stronger associations between diet and CRC incidence would have been observed in different cohorts with a wider range of dietary quality.

In conclusion, we demonstrate that long-term adherence to a dietary pattern associated with a greater abundance of sulfur-metabolizing bacteria may be associated with a greater incidence of PTGS2-high CRC and Bifidobacterium-negative distal CRC among adult men. Our data suggest that dietary modulation of the gut microbiome may be an attractive population-level preventive strategy in high-risk individuals. Additional studies are needed to further characterize the relationship between dietary sulfur metabolism, gut microbial ecology, and CRC among large and diverse study populations.

CONFLICTS OF INTEREST

Guarantors of the article: Daniel R. Sikavi, MD, Long H. Nguyen, MD, MS, Koichiro Haruki, MD, PhD, Tomotaka Ugai, MD, PhD, Shuji Ogino, MS, MD, PhD, Curtis Huttenhower, PhD, and Andrew T. Chan, MD, MPH.

Specific author contributions: All authors contributed to analysis and interpretation of data and critical revision of the manuscript. Daniel R. Sikavi, MD, Long H. Nguyen, MD, MS, Koichiro Haruki, MD, PhD, Tomotaka Ugai, MD, PhD share co-first authorship. Curtis Huttenhower, PhD, Shuji Ogino, MS, MD, PhD, Andrew T. Chan, MD, MPH share co-PI. D.R.S., L.H.N., W.S.G., E.B.R., J.I., C.H., S.O., and A.T.C.: contributed to study concept and design. D.R.S., L.H.N., C.H., S.O., and A.T.C.: acquired data. D.R.S., L.H.N., C.H., S.O., and A.T.C.: drafted the manuscript. D.R.S., L.H.N., K.H., T.U., C.H., S.O., and A.T.C.: performed statistical analyses. W.S.G., E.B.R., J.I., C.H., S.O., and A.T.C.: obtained funding. C.H., S.O., and A.T.C.: provided administrative, technical, or material support. C.H., S.O., and A.T.C.: provided study supervision.

Financial support: National Institutes of Health (NIH) grants U54DE023798, U01CA167552, P01CA055075, U01CA152904, and K23 DK125838 (to L.H.N.), Loan Repayment Program (to L.H.N.), and K99DK119412 (to D.W.), R21AA027608 (to Y.C.), R00CA215314 (to M.S.), and R01CA202704 (to W.S.G., J.I., C.H., and A.T.C.), R35 CA253185 (to A.T.C.), R35CA197735 (to S.O.), R01CA151993 (to S.O.), and R01 CA248857 (to S.O.); American Gastroenterological Association Research Scholars Award (to L.H.N.), Crohn's and Colitis Foundation Research Fellowship Award and Career Development Award (to L.H.N.), and Senior Investigator Award (to A.T.C.); American Cancer Society Mentored Research Scholar Grant (to M.S.); the USDA National Institute of Food and Agriculture Hatch Multistate Research Capacity Funding Program grant W4122 (to J.I.); Cancer Research UK Grand Challenge Award (C10674/A27140; to W.S.G., S.O., and C.H.); Dana-Farber Harvard Cancer Center Nodal Award (to S.O.); STARR Cancer Consortium; the Massachusetts General Hospital Stuart and Suzanne Steel Research Scholars Award (to A.T.C.); Overseas Research Fellowship (201960541) from the Japan Society for the Promotion of Science (to T.U.); Uehara Memorial Foundation fellowship grants (to K.H. and T.U.); Yasuda Medical Foundation fellowship grant (to T.U.); and Mitsukoshi Health and Welfare Foundation grant to K.H.

Potential competing interests: None to report.

Use of standardized official symbols: We use Human Genome Organization (HUGO)-approved official symbols (or root symbols) for genes and gene products, including BRAF, CACNA1G, CDKN2A, CRABP1, CTNNB1, IGF2, KRAS, MLH1, NEUROG1, PIK3CA, PTGS2, RUNX3, SOCS1, and WNT, all of which are described at www.genenames.org. Gene symbols are italicized, whereas symbols for gene products are not italicized.

Study Highlights

WHAT IS KNOWN

  • ✓ Sulfur-metabolizing bacteria have been implicated in colorectal cancer (CRC).
  • ✓ We previously showed that a diet that enriches for sulfur-metabolizing bacteria was associated with an increased risk of distal CRC in adult men.
  • ✓ It is unclear whether this relationship is driven by associations with particular molecular or microbially-enriched tumor subtypes.

WHAT IS NEW HERE

  • ✓ The sulfur microbial diet may be associated with PTGS2-high CRC in adult men.
  • ✓ The sulfur microbial diet may be associated with distal colon and rectal cancers depleted of intratumoral Bifidobacteria in adult men.

TRANSLATIONAL IMPACT

  • ✓ Dietary modulation of the gut microbiome may contribute to colorectal carcinogenesis.
  • ✓ Dietary manipulation of the gut microbiome may reduce CRC risk.

ACKNOWLEDGEMENTS

We thank the participants who graciously participated in this research, as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, and WY. We wish to acknowledge the efforts of the Channing Division of Network Medicine at the Brigham and Women's Hospital and Harvard Medical School. The authors assume full responsibility for analyses and interpretation of these data.

REFERENCES

1. Ferlay J, Colombet M, Soerjomataram I, et al. Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods. Int J Cancer 2019;144(8):1941–53.
2. Erdrich J, Zhang X, Giovannucci E, et al. Proportion of colon cancer attributable to lifestyle in a cohort of US women. Cancer Causes Control 2015;26(9):1271–9.
3. Carr PR, Weigl K, Jansen L, et al. Healthy lifestyle factors associated with lower risk of CRC irrespective of genetic risk. Gastroenterology 2018;155(6):1805–15.e1805.
4. Wang X, O'Connell K, Jeon J, et al. Combined effect of modifiable and non-modifiable risk factors for CRC risk in a pooled analysis of 11 population-based studies. BMJ Open Gastroenterol 2019;6(1):e000339.
5. Kirkegaard H, Johnsen NF, Christensen J, et al. Association of adherence to lifestyle recommendations and risk of CRC: A prospective Danish cohort study. BMJ 2010;341:c5504.
6. Mehta M, Shike M. Diet and physical activity in the prevention of CRC. J Natl Compr Cancer Netw 2014;12(12):1721–6.
7. O'Keefe SJ. Diet, microorganisms and their metabolites, and colon cancer. Nat Rev Gastroenterol Hepatol 2016;13(12):691–706.
8. Song M, Garrett WS, Chan AT. Nutrients, foods, and CRC prevention. Gastroenterology 2015;148(6):1244–60.e1216.
9. Brennan CA, Garrett WS. Gut microbiota, inflammation, and CRC. Annu Rev Microbiol 2016;70:395–411.
10. Song M, Chan AT, Sun J. Influence of the gut microbiome, diet, and environment on risk of CRC. Gastroenterology 2020;158(2):322–40.
11. Cai WJ, Wang MJ, Ju LH, et al. Hydrogen sulfide induces human colon cancer cell proliferation: Role of akt, ERK and p21. Cell Biol Int 2010;34(6):565–72.
12. Louis P, Hold GL, Flint HJ. The gut microbiota, bacterial metabolites and CRC. Nat Rev Microbiol 2014;12(10):661–72.
13. Scanlan PD, Shanahan F, Marchesi JR. Culture-independent analysis of desulfovibrios in the human distal colon of healthy, CRC and polypectomized individuals. FEMS Microbiol Ecol 2009;69(2):213–21.
14. Kanazawa K, Konishi F, Mitsuoka T, et al. Factors influencing the development of sigmoid colon cancer. Bacteriologic and biochemical studies. Cancer 1996;77(8 Suppl):1701–6.
15. Yachida S, Mizutani S, Shiroma H, et al. Metagenomic and metabolomic analyses reveal distinct stage-specific phenotypes of the gut microbiota in CRC. Nat Med 2019;25(6):968–76.
16. Song M, Chan AT. Diet, gut microbiota, and CRC prevention: A review of potential mechanisms and promising targets for future research. Curr CRC Rep 2017;13(6):429–39.
17. Carbonero F, Benefiel AC, Gaskins HR. Contributions of the microbial hydrogen economy to colonic homeostasis. Nat Rev Gastroenterol Hepatol 2012;9(9):504–18.
18. Attene-Ramos MS, Wagner ED, Gaskins HR, et al. Hydrogen sulfide induces direct radical-associated DNA damage. Mol Cancer Res 2007;5(5):455–9.
19. Attene-Ramos MS, Nava GM, Muellner MG, et al. DNA damage and toxicogenomic analyses of hydrogen sulfide in human intestinal epithelial FHs 74 Int cells. Environ Mol Mutagen 2010;51(4):304–14.
20. Attene-Ramos MS, Wagner ED, Plewa MJ, et al. Evidence that hydrogen sulfide is a genotoxic agent. Mol Cancer Res 2006;4(1):9–14.
21. Miller TW, Wang EA, Gould S, et al. Hydrogen sulfide is an endogenous potentiator of T cell activation. J Biol Chem 2012;287(6):4211–21.
22. Johansson ME, Phillipson M, Petersson J, et al. The inner of the 2 Muc2 mucin-dependent mucus layers in colon is devoid of bacteria. Proc Natl Acad Sci U S A 2008;105(39):15064–9.
23. Nguyen LH, Ma W, Wang DD, et al. Association between sulfur-metabolizing bacterial communities in stool and risk of distal CRC in men. Gastroenterology 2020;158(5):1313–25.
24. Yamauchi M, Morikawa T, Kuchiba A, et al. Assessment of CRC molecular features along bowel subsites challenges the conception of distinct dichotomy of proximal versus distal colorectum. Gut 2012;61(6):847–54.
25. Lee MS, Menter DG, Kopetz S. Right versus left colon cancer biology: Integrating the consensus molecular subtypes. J Natl Compr Canc Netw 2017;15(3):411–9.
26. Loree JM, Pereira AAL, Lam M, et al. Classifying CRC by tumor location rather than sidedness highlights a continuum in mutation profiles and consensus molecular subtypes. Clin Cancer Res 2018;24(5):1062–72.
27. Mehta RS, Song M, Nishihara R, et al. Dietary patterns and risk of CRC: Analysis by tumor location and molecular subtypes. Gastroenterology 2017;152(8):1944–53.e1941.
28. Schernhammer ES, Giovannucci E, Kawasaki T, et al. Dietary folate, alcohol and B vitamins in relation to LINE-1 hypomethylation in colon cancer. Gut 2010;59(6):794–9.
29. Slattery ML, Curtin K, Anderson K, et al. Associations between dietary intake and ki-ras mutations in colon tumors: A population-based study. Cancer Res 2000;60(24):6935–41.
30. Diergaarde B, Braam H, van Muijen GN, et al. Dietary factors and microsatellite instability in sporadic colon carcinomas. Cancer Epidemiol Biomarkers Prev 2003;12(11 Pt 1):1130–6.
31. Satia JA, Keku T, Galanko JA, et al. Diet, lifestyle, and genomic instability in the North Carolina colon cancer study. Cancer Epidemiol Biomarkers Prev 2005;14(2):429–36.
32. Slattery ML, Curtin K, Sweeney C, et al. Diet and lifestyle factor associations with CpG island methylator phenotype and BRAF mutations in colon cancer. Int J Cancer 2007;120(3):656–63.
33. Hughes LAE, Simons C, van den Brandt PA, et al. Lifestyle, diet, and CRC risk according to (Epi)genetic instability: Current evidence and future directions of molecular pathological epidemiology. Curr CRC Rep 2017;13(6):455–69.
34. Gilsing AM, Fransen F, de Kok TM, et al. Dietary heme iron and the risk of CRC with specific mutations in KRAS and APC. Carcinogenesis 2013;34(12):2757–66.
35. Kure S, Nosho K, Baba Y, et al. Vitamin D receptor expression is associated with PIK3CA and KRAS mutations in CRC. Cancer Epidemiol Biomarkers Prev 2009;18(10):2765–72.
36. Schernhammer ES, Giovannucci E, Baba Y, et al. B vitamins, methionine and alcohol intake and risk of colon cancer in relation to BRAF mutation and CpG island methylator phenotype (CIMP). PLoS One 2011;6(6):e21102.
37. Hughes LA, Simons CC, van den Brandt PA, et al. Body size, physical activity and risk of CRC with or without the CpG island methylator phenotype (CIMP). PLoS One 2011;6(4):e18571.
38. Mrkonjic M, Chappell E, Pethe VV, et al. Association of apolipoprotein E polymorphisms and dietary factors in CRC. Br J Cancer 2009;100(12):1966–74.
39. Habermann N, Ulrich CM, Lundgreen A, et al. PTGS1, PTGS2, ALOX5, ALOX12, ALOX15, and FLAP SNPs: Interaction with fatty acids in colon cancer and rectal cancer. Genes Nutr 2013;8(1):115–26.
40. Weijenberg MP, Lüchtenborg M, de Goeij AF, et al. Dietary fat and risk of colon and rectal cancer with aberrant MLH1 expression, APC or KRAS genes. Cancer Causes Control 2007;18(8):865–79.
41. Liu L, Tabung FK, Zhang X, et al. Diets that promote colon inflammation associate with risk of colorectal carcinomas that contain Fusobacterium nucleatum. Clin Gastroenterol Hepatol 2018;16(10):1622–e3.e1623.
42. Mehta RS, Nishihara R, Cao Y, et al. Association of dietary patterns with risk of CRC subtypes classified by Fusobacterium nucleatum in tumor tissue. JAMA Oncol 2017;3(7):921–7.
43. Le Leu RK, Hu Y, Brown IL, et al. Synbiotic intervention of Bifidobacterium lactis and resistant starch protects against CRC development in rats. Carcinogenesis 2010;31(2):246–51.
44. Kostic AD, Chun E, Robertson L, et al. Fusobacterium nucleatum potentiates intestinal tumorigenesis and modulates the tumor-immune microenvironment. Cell Host Microbe 2013;14(2):207–15.
45. Singh J, Rivenson A, Tomita M, et al. Bifidobacterium longum, a lactic acid-producing intestinal bacterium inhibits colon cancer and modulates the intermediate biomarkers of colon carcinogenesis. Carcinogenesis 1997;18(4):833–41.
46. Giovannucci E, Rimm EB, Stampfer MJ, et al. Aspirin use and the risk for CRC and adenoma in male health professionals. Ann Intern Med 1994;121(4):241–6.
47. Rimm EB, Giovannucci EL, Stampfer MJ, et al. Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135(10):1114–36. discussion 1127–36.
48. Willett W. Nutritional Epidemiology, Vol 40. New York: Oxford University Press, 2012.
49. Fung T, Hu FB, Fuchs C, et al. Major dietary patterns and the risk of CRC in women. Arch Intern Med 2003;163(3):309–14.
50. Magalhaes B, Peleteiro B, Lunet N. Dietary patterns and CRC: Systematic review and meta-analysis. Eur J Cancer Prev 2012;21(1):15–23.
51. Terry P, Hu FB, Hansen H, et al. Prospective study of major dietary patterns and CRC risk in women. Am J Epidemiol 2001;154(12):1143–9.
52. Morikawa T, Kuchiba A, Yamauchi M, et al. Association of CTNNB1 (beta-catenin) alterations, body mass index, and physical activity with survival in patients with CRC. JAMA 2011;305(16):1685–94.
53. Mima K, Sukawa Y, Nishihara R, et al. Fusobacterium nucleatum and T cells in colorectal carcinoma. JAMA Oncol 2015;1(5):653–61.
54. Kosumi K, Hamada T, Koh H, et al. The amount of Bifidobacterium genus in colorectal carcinoma tissue in relation to tumor characteristics and clinical outcome. Am J Pathol 2018;188(12):2839–52.
55. Imamura Y, Lochhead P, Yamauchi M, et al. Analyses of clinicopathological, molecular, and prognostic associations of KRAS codon 61 and codon 146 mutations in CRC: Cohort study and literature review. Mol Cancer 2014;13:135.
56. Ogino S, Nosho K, Kirkner GJ, et al. CpG island methylator phenotype, microsatellite instability, BRAF mutation and clinical outcome in colon cancer. Gut 2009;58(1):90–6.
57. Liao X, Lochhead P, Nishihara R, et al. Aspirin use, tumor PIK3CA mutation, and colorectal-cancer survival. N Engl J Med 2012;367(17):1596–606.
58. Ogino S, Kawasaki T, Nosho K, et al. LINE-1 hypomethylation is inversely associated with microsatellite instability and CpG island methylator phenotype in CRC. Int J Cancer 2008;122(12):2767–73.
59. Ogino S, Brahmandam M, Kawasaki T, et al. Combined analysis of COX-2 and p53 expressions reveals synergistic inverse correlations with microsatellite instability and CpG island methylator phenotype in CRC. Neoplasia 2006;8(6):458–64.
60. Ogino S, Kawasaki T, Kirkner GJ, et al. Evaluation of markers for CpG island methylator phenotype (CIMP) in CRC by a large population-based sample. J Mol Diagn 2007;9(3):305–14.
61. Nosho K, Irahara N, Shima K, et al. Comprehensive biostatistical analysis of CpG island methylator phenotype in CRC using a large population-based sample. PLoS One 2008;3(11):e3698.
62. Chan AT, Ogino S, Fuchs CS. Aspirin and the risk of CRC in relation to the expression of COX-2. N Engl J Med 2007;356(21):2131–42.
63. Wolf AM, Hunter DJ, Colditz GA, et al. Reproducibility and validity of a self-administered physical activity questionnaire. Int J Epidemiol 1994;23(5):991–9.
64. Wang M, Spiegelman D, Kuchiba A, et al. Statistical methods for studying disease subtype heterogeneity. Stat Med 2016;35(5):782–800.
65. Benjamin DJ, Berger JO, Johannesson M, et al. Redefine statistical significance. Nat Hum Behav 2018;2(1):6–10.
66. Chan DS, Lau R, Aune D, et al. Red and processed meat and CRC incidence: meta-analysis of prospective studies. PLoS One 2011;6(6):e20456.
67. Niederreiter L, Adolph TE, Tilg H. Food, microbiome and CRC. Dig Liver Dis 2018;50(7):647–52.
68. Hullar MA, Fu BC. Diet, the gut microbiome, and epigenetics. Cancer J 2014;20(3):170–5.
69. Florin T, Neale G, Gibson GR, et al. Metabolism of dietary sulphate: Absorption and excretion in humans. Gut 1991;32(7):766–73.
70. Magee EA, Richardson CJ, Hughes R, et al. Contribution of dietary protein to sulfide production in the large intestine: An in vitro and a controlled feeding study in humans. Am J Clin Nutr 2000;72(6):1488–94.
71. Zhang C, Zhong M. Consumption of beer and CRC incidence: A meta-analysis of observational studies. Cancer Causes Control 2015;26(4):549–60.
72. Purcell RV, Visnovska M, Biggs PJ, et al. Distinct gut microbiome patterns associate with consensus molecular subtypes of CRC. Sci Rep 2017;7(1):11590.
73. Hale VL, Jeraldo P, Chen J, et al. Distinct microbes, metabolites, and ecologies define the microbiome in deficient and proficient mismatch repair CRCs. Genome Med 2018;10(1):78.
74. Kim SW, Kim HM, Yang KM, et al. Bifidobacterium lactis inhibits NF-kappaB in intestinal epithelial cells and prevents acute colitis and colitis-associated colon cancer in mice. Inflamm Bowel Dis 2010;16(9):1514–25.
75. Laforest-Lapointe I, Arrieta MC. Patterns of early-life gut microbial colonization during human immune development: An ecological perspective. Front Immunol 2017;8:788.
76. Sivan A, Corrales L, Hubert N, et al. Commensal Bifidobacterium promotes antitumor immunity and facilitates anti-PD-L1 efficacy. Science 2015;350(6264):1084–9.
77. Turroni F, Foroni E, Pizzetti P, et al. Exploring the diversity of the bifidobacterial population in the human intestinal tract. Appl Environ Microbiol 2009;75(6):1534–45.
78. Cho I, Blaser MJ. The human microbiome: At the interface of health and disease. Nat Rev Genet 2012;13(4):260–70.
79. Yu J, Chen Y, Fu X, et al. Invasive Fusobacterium nucleatum may play a role in the carcinogenesis of proximal colon cancer through the serrated neoplasia pathway. Int J Cancer 2016;139(6):1318–26.
80. Mima K, Cao Y, Chan AT, et al. Fusobacterium nucleatum in colorectal carcinoma tissue according to tumor location. Clin Transl Gastroenterol 2016;7(11):e200.
81. Nguyen LH, Goel A, Chung DC. Pathways of colorectal carcinogenesis. Gastroenterology 2020;158(2):291–302.
82. Ridlon JM, Wolf PG, Gaskins HR. Taurocholic acid metabolism by gut microbes and colon cancer. Gut Microbes 2016;7(3):201–15.
83. Greenhough A, Smartt HJ, Moore AE, et al. The COX-2/PGE2 pathway: Key roles in the hallmarks of cancer and adaptation to the tumour microenvironment. Carcinogenesis 2009;30(3):377–86.
84. Wang D, Dubois RN. The role of COX-2 in intestinal inflammation and CRC. Oncogene 2010;29(6):781–8.
85. Zhu Y, Zhu M, Lance P. Stromal COX-2 signaling activated by deoxycholic acid mediates proliferation and invasiveness of colorectal epithelial cancer cells. Biochem Biophys Res Commun 2012;425(3):607–12.
86. Cheng K, Raufman JP. Bile acid-induced proliferation of a human colon cancer cell line is mediated by transactivation of epidermal growth factor receptors. Biochem Pharmacol 2005;70(7):1035–47.
87. Mima K, Nishihara R, Qian ZR, et al. Fusobacterium nucleatum in colorectal carcinoma tissue and patient prognosis. Gut 2016;65(12):1973–80.
88. Lloyd-Price J, Mahurkar A, Rahnavard G, et al. Strains, functions and dynamics in the expanded human microbiome project. Nature 2017;550(7674):61–6.

Supplemental Digital Content

© 2021 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of The American College of Gastroenterology