Gastrointestinal Malignancy and the Microbiome

Abstract

Microbial species participate in the genesis of a substantial number of malignancies—in conservative estimates, at least 15% of all cancer cases are attributable to infectious agents. Little is known about the contribution of the gastrointestinal (GI) microbiome to the development of malignancies. Resident microbes can promote carcinogenesis by inducing inflammation, increasing cell proliferation, altering stem cell dynamics, and producing metabolites such as butyrate, which affect DNA integrity and immune regulation. Studies in humans and rodent models of cancer have identified effector species and relationships among members of the microbial community in the stomach and colon that increase the risk for malignancy. Strategies to manipulate the microbiome, or the immune response to such bacteria, could be developed to prevent or treat certain GI cancers.

Introduction

Humans are colonized by a myriad of microbes including bacteria, archaea, eukaryotes, and viruses, although bacteria are the most abundant and well-studied component. The composition of the gastrointestinal (GI) microbiome is shaped by a variety of factors including diet, additional environmental elements, and the genetic background of the host. The GI microbiota is a complex ecosystem that contains >3 million genes, encoding enzymes that generate metabolites that can influence health as well as disease^{1, 2}.

Cancer of the GI tract is a leading cause of morbidity and mortality in the United States. Genetic factors have been identified that clearly increase cancer risk, including Apc mutations that cause familial adenomatous polyposis and E-cadherin mutations that lead to hereditary diffuse-type gastric cancer. However, non-genetic factors more broadly affect most GI cancers. Microbial species participate in the genesis of a substantial number of malignancies worldwide; in conservative estimates, more than 15% of all cancer cases can be attributed to infectious agents, for a neoplastic burden of 1.2 million cases/year³. Residential microbes in the GI tract can alter cancer risk by inducing oxidative and nitrosative DNA damage in response to chronic inflammation, increasing cell proliferation, altering stem cell dynamics, and producing mutagenic metabolites such as butyrate. The GI microbiota can also constrain or facilitate tumor growth by altering immune surveillance mechanisms and affecting the metabolism of chemotherapeutic agents^{1, 2}. In this review, we discuss emerging concepts and provide specific examples for the role of the GI microbiome in the development of malignancies that arise within this niche. We will focus on gastric and colonic malignancies, since these have been most thoroughly investigated, and will also briefly discuss esophageal cancers.

Gastric Cancer

Gastric adenocarcinoma is the second-leading cause of cancer-related death in the world^4,5. Helicobacter pylori is a Gram-negative bacterial species that selectively colonizes gastric epithelium; chronic infection with this organism is the strongest identified risk factor for gastric adenocarcinoma, prompting the World Health Organization to designate H pylori as a class I carcinogen. Approximately 660,000 new cases of gastric cancer/year are attributable to H pylori, making this pathogen the most common infectious agent linked to any malignancy³. However, only a percentage of colonized persons develop neoplasia. Risk is associated with H pylori strain, variations in host responses governed by genetic diversity, and/or specific interactions between host, microbial, and environmental determinants.

One H pylori determinant that influences cancer risk is the cag pathogenicity island (PAI). Genes within the cag PAI encode an antigenic effector protein (CagA) as well as proteins that form a type IV bacterial secretion system (T4SS) that exports CagA from adherent H pylori into host cells^6–9. H pylori strains that harbor the cag PAI (cag⁺ strains) are associated with a significantly increased risk of distal gastric cancer compared to cag⁻ strains¹⁰. Following translocation, CagA undergoes tyrosine phosphorylation by Src and Abl kinases; phospho-CagA subsequently interacts with and activates several host cell proteins, including a phosphatase (SHP2), leading to morphological alterations such as cell scattering and elongation^{6, 11, 12}. Non-phosphorylated CagA also exerts effects within host cells that affect oncogenesis. Unmodified CagA directly binds PAR1b, which regulates cell polarity, and inhibits its kinase activity—an interaction that promotes loss of polarity¹³. Non-phosphorylated CagA associates with the epithelial tight-junction scaffolding protein ZO1 and the transmembrane protein junction adhesion molecule-A (JAMA) to cause ineffective assembly of tight junctions at sites of bacterial attachment¹⁴. Unmodified CagA also activates β-catenin, leading to transcriptional upregulation of genes implicated in cancer^15–17. The CagA protein of certain H pylori strains can stimulate expression of IL8 by activating the transcription factor NFκB¹⁸, thereby contributing to neutrophil infiltration within the gastric mucosa. CagA also induces DNA damage in vitro and in rodent models of infection—results that have been validated in human subjects colonized with H pylori cag⁺ strains¹⁹. Contact between cag⁺ strains and host cells therefore activates multiple signaling pathways that may increase the risk for malignant transformation during prolonged colonization.

Another H pylori constituent linked to the development of gastric cancer is the secreted VacA toxin²⁰. VacA causes a wide assortment of alterations in gastric epithelial cells, including vacuolation, altered plasma and mitochondrial membrane permeability, autophagy, and apoptotic cell death²⁰. All H pylori strains possess vacA, but there is marked variation in vacA sequences among strains. The regions of greatest diversity are localized to the 5' region of the gene, which encodes the signal sequence and amino-terminus of the secreted toxin (allele types s1a, s1b, s1c, or s2), an intermediate region (allele types i1 or i2), and a mid-region (allele types m1 or m2)^{21, 22}. Strains that contain type s1, i1 and m1 forms of vacA are associated with a higher risk of gastric cancer than strains that contain type s2, i2 and m2 forms²².

VacA and CagA can also counter-regulate each other to affect host cell responses. Specifically, CagA antagonizes VacA-induced apoptosis and activates a cell survival pathway mediated by the MAPK ERK and the anti-apoptotic protein MCL1²³. CagA also activates NFAT and EGFR signaling—processes that are inhibited by VacA²³.

Recently, exciting data have shown that the opposing effects of CagA and VacA may be cell-lineage specific. The Wnt target gene Lgr5 encodes an orphan G protein-coupled receptor and marks a self-renewing, multipotent stem cell population responsible for long-term renewal of gastric epithelium²⁴. Lgr5⁺ epithelial cells have higher levels of oxidative DNA damage than Lgr5-negative cells in H pylori-infected persons with gastric cancer²⁵, indicating that H pylori specifically target Lgr5⁺ epithelial cells. In transgenic mice that over-express Le^b, H pylori adhere directly to gastric epithelial cells²⁶. Genetic ablation of parietal cells in Le^b-expressing transgenic mice permits the gastric epithelial stem cell population to expand, which is accompanied by increased H pylori colonization and inflammation within glandular epithelium^27,28. Delineation of this stem cell transcriptome has identified several pathways that are over-represented and of particular importance for carcinogenesis, including Wnt activation of β-catenin²⁹ which is, in turn, activated by CagA-mediated signaling^{15, 16}. In differentiated gastric epithelial cells, binding of VacA to LRP1, a specific receptor on the epithelial cell surface, leads to autophagic elimination of CagA³⁰. Importantly, gastric epithelial cells that express a stem cell marker, CD44 variant9, fail to degrade CagA; these findings have been verified in vivo³⁰. A subpopulation of host cells with progenitor-like features therefore appear to be uniquely susceptible to the effects of a microbial oncoprotein, which may lower the threshold for carcinogenesis.

Host genetic factors also influence the risk of gastric cancer among H pylori-infected persons. IL1β is an inflammatory cytokine that inhibits gastric acid secretion, and production of IL1β is increased in the gastric mucosa of infected vs uninfected persons³¹. Polymorphisms within the IL1β gene cluster that increase IL1β production are associated with significant increases in risk for hypochlorhydria, gastric atrophy, and distal gastric adenocarcinoma compared with low-expression IL1b genotypes^32–34. The presence of a virulent strain of H pylori in a genetically susceptible person further augments the risk for gastric cancer. Persons harboring high-expression IL1β alleles who are infected with H pyloricagA⁺ or vacA s1-type strains have 25-fold or 87-fold increases in risk, respectively, for gastric cancer compared to uninfected persons³³. A recent genome-wide association study has identified new targets for studies of gastric carcinogenesis, demonstrating that TLR1 and FCGR2A loci are associated with H pylori sero-prevalence. However, the relationship between these genetic loci and disease was not determined³⁵. Dietary factors have recently been shown to accelerate gastric carcinogenesis in rodent models of H pylori infection; specifically iron depletion accelerates the development of gastric dysplasia and cancer by promoting assembly and function of the H pylori cag PAI³⁶.

Although H pylori infection is the strongest identified risk factor for gastric cancer, recent human trials have indicated that other residents of the gastric microbiota may influence malignant transformation. A clinical study examining the effects of anti-microbial therapy against H pylori reported that treatment was associated with a significant reduction in gastric cancer incidence rates 15 years after therapy. However, only 47% of treated persons remained free of H pylori at this time point³⁷, leading to the tantalizing possibility that antibiotic therapy alters other microbial species that affect the development of gastric cancer. Another study evaluated the presence and frequency of bacteria-to-human somatic cell lateral gene transfer in human cancer specimens³⁸. Of all malignancies examined, gastric adenocarcinomas harbored the second highest number of bacterial DNA integrations. However, the most frequent integrations were derived from Pseudomonas-like DNA, and not Helicobacter DNA³⁸, supporting the premise that the non-H pylori gastric microbial ecosystem influences the development of cancer in the stomach.

The Human Gastric Microbiota

The human stomach is a relatively austere microbial niche, with colonization densities from 10¹ to 10³ CFU/g³⁹. Cutting-edge molecular techniques have provided insights into potential pathogenic roles of the gastric microbial community. H pylori-negative individuals have highly diverse gastric microbiota; their most abundant phyla are Firmicutes, Bacteroidetes, and Actinobacteria^{1, 40}. Common phylotypes present in H pylori-uninfected persons include Streptococcus, Prevotella, and Gemella⁴¹. Several studies have demonstrated that, among H pylori-colonized persons, H pylori accounts for >90% of all sequence reads⁴¹, so it greatly reduces the overall diversity of the gastric microbiota. The most abundant phyla in H pylori-colonized human stomachs are Proteobacteria, Firmicutes, and Actinobacteria^{42, 43}. One study reported that H pylori infection of the stomach increased the relative abundance of Proteobacteria, Spirochaetes, and Acidobacteria while decreasing the relative abundance of Actinobacteria, Bacteroidetes and Firmicutes, compared to uninfected stomach⁴².

In contrast to studies comparing the composition of the gastric microbiota between H pylori-infected and -uninfected subjects, few studies have examined differences in microbial composition and outcomes from diseases such as gastric cancer³⁹. One of the key steps in the histologic progression to intestinal-type gastric cancer is atrophy, which is accompanied by an increase in gastric pH due to loss of parietal cells⁴. Within the context of atrophic gastritis, non-molecular analyses have shown that, in the absence of H pylori, urease-producing members of the gastric microbiota, such as Proteus mirabilis, Klebsiella pneumonia, Staphylococcus aureus, Staphylococcus capitis, and Micrococcus species can bloom to levels that cause false positive results in H pylori urea breath tests⁴³. A terminal restriction fragment length polymorphism study found no significant differences between the microbiota of patients with and without gastric cancer⁴⁴. Specifically, the microbiota of patients with gastric cancer was dominated by species of Streptococcus, Lactobacillus, Veillonella, and Prevotella; among Streptococcus, S mitis and S parasanguinis predominated⁴⁴. However, detailed molecular studies defining the composition of the gastric microbiota in well-characterized human populations, with and without gastric cancer, are clearly lacking³⁹. Studies of rodent model systems will help identify important drivers and modifiers of diseases related to the microbiome.

H pylori infection of Mongolian gerbils can lead to gastric adenocarcinoma, without the co-administration of carcinogens^{15, 45–47}; gastric cancer in this model develops in the distal stomach, as in humans. Various H pylori wild-type and mutant strains colonize gerbils well^48–50, allowing an examination of the role of virulence determinants on parameters of gastric injury. However, Mongolian gerbils are outbred, which increases the variability of responses to any stimulus. Studies of inbred mice with defined genotypes, as well as transgenic lines, would allow for more detailed analyses of host susceptibility. However, standard models of inbred mice are frequently limited by their uncontrolled microbial diversity. Gnotobiotic mice obviate this obstacle, but they are expensive and specialized facilities and expertise are required, limiting their widespread use. There are opportunities to conventionalize gnotobiotic mice available from commercial sources. Several studies have used molecular techniques to define the composition of gastric microbial communities in rodents with H pylori-induced gastric cancer (Figure 1).

Figure 1. Differences in the composition of human and rodent gastric microbiome, based on H pylori status.

There are substantial variations in the gastric microbiota of humans, mice, and gerbils. Nonetheless, the presence of H pylori alters the constituents of these ecosystems. This diagram shows the relative abundance of phyla, determined by high-throughput sequencing, in the stomach of humans, gerbils, and mice^1,2,39,43. The risk of cancer increases with the presence of H pylori.

Rodent Models of H pylori-induced Gastric Cancer

Mongolian gerbils develop gastric cancer when colonized with certain strains of H pylori. Little is known about the composition of the gerbil gastric microbiota, but most commonly represented phyla include Firmicutes, Actinobacteria, Proteobacteria, and Bacteroidetes^{39, 43}. Similar to mice, the genus Lactobacillus dominates the gastric microbiome of uninfected gerbils^{39, 43}.

Only one study used molecular techniques to comprehensively compare differences in the gerbil gastric microbiome before and after H pylori infection⁵¹. Using temporal temperature gradient gel electrophoresis and next-generation sequencing, Sun et al. observed reduced diversity of Lactobacillus following H pylori infection⁵¹. Another group utilized quantitative PCR to track the relative abundance of 15 species of microbes in gerbil stomach⁵². In uninfected gerbils, the most abundant genera were Lactobacillus and Enterococcus, followed by equivalent levels of Atopobium and Clostridium. In gerbils that were challenged and successfully infected with H pylori, the relative abundance of Clostridium coccoides increased, compared to gerbils not infected with H pylori. In gerbils that were challenged with H pylori but were not successfully colonized, the proportion of C coccoides, C leptum, and Bifidobacterium was reduced. Of interest, gerbils that were challenged with H pylori but unsuccessfully colonized represented the only group that harbored members of the Eubacterium cylindroides and Prevotella species⁵². The importance of differences in microbial composition and the development of gastric cancer has not been determined in this model.

Mice are another commonly used model of gastric carcinogenesis. Bacteria other than Helicobacter species have been found to induce gastritis in mice—Acinetobacter lowffi can induce gastric inflammation in the absence of H pylori⁵³. Lactobacillus is the predominant genus in mouse stomach^{39, 43}. However, despite identical genetic backgrounds of mice, their commercial source can affect the composition of the gastric microbiome. Rolig et al. reported that C57BL/6 mice from 2 independent vendors had different levels of Lactobacillus species and developed disparate inflammation and injury responses when challenged with H pylori⁵⁴.

Infection of mice with H pylori can alter their gastric microbiota, which appears to depend on strain of mouse and duration of infection. In SPF Balb/c mice, H pylori infection reduced the number of Lactobacillus species and increased bacterial diversity⁵⁵. However, studies of C57BL/6 mice have produced conflicting results. In one study, acute infection of C57BL/6 mice with H pylori did not cause significant shifts in the bacterial composition of the gastric microbiota⁵⁶. However, administration of antibiotics to mice before H pylori infection altered the composition of the gastric microbiota and reduced the severity of gastric inflammation; these changes were reversed when the gastric microbiota from antibiotic-naïve mice was transferred to mice given antibiotics⁵⁴.

Other studies with mice that are genetically susceptible to gastric cancer, such as hypergastrinemic INS-GAS mice, have shown that chronic interactions between H pylori and the gastric microbiota influence malignancy. Lofgren et al. demonstrated that germ-free INS-GAS mice mono-colonized with H pylori progress more slowly to gastric intraepithelial neoplasia than H pylori-infected specific pathogen-free (SPF) INS-GAS mice⁵⁷. When the composition of the gastric microbiota was characterized in SPF INS-GAS mice, the phyla Firmicutes and Bacteroidetes accounted for >80% of the bacterial population. Infection with H pylori led to a bloom in the proportion of Firmicutes and decreased numbers of Bacteroidetes⁵⁷. Of interest, there were no significant shifts in the colonic microbiota following infection with H pylori. Another study demonstrated that progression to gastric neoplasia was similar in SPF INS-GAS mice infected with H pylori and germ-free INS-GAS mice first infected with a restricted Altered Schaedler's Flora (ASF): [ASF 356 (contains Clostridium species), ASF 361 (contains Lactobacillus murinus), and ASF 519 (contains Bacteroides species)], and then challenged with H pylori⁵⁸. These results indicate that these 3 bacterial constituents of ASF could cooperate with H pylori to increase the risk of gastric cancer.

Studies in mice have also suggested that extra-gastric constituents of the microbiota can influence gastric carcinogenesis. Colonization of C57BL/6 mice with the enterohepatic Helicobacter species H bilis or H muridarum before challenge with H pylori significantly reduced H pylori-induced injury of the stomach^{59, 60}. In contrast, pre-colonization with H hepaticus increased H pylori-induced gastric damage⁶⁰. The mechanism may involve reductions in the intensity of T helper 1-type cell responses against H pylori, which are mediated by T regulatory cells that have been sensitized to shared Helicobacter antigens⁵⁹.

Despite the mechanistic information that has emerged from these studies, there are important caveats to consider when comparing the effects of the rodent microbiome on malignancy with cancers that develop in humans. The phyla present in H pylori-infected human stomachs are not the same as those that predominate in an H pylori-infected rodent gastric niche. The density and topography of H pylori colonization in rodent stomachs does not precisely reflect that of humans. For example, acute H pylori infection of mice tends to primarily affect the gastric corpus, compared to the antral-predominant colonization that is typical of human infections. Finally, the presence of a squamous forestomach in rodents can alter the composition of the rodent gastric microbiome, in contrast to what is present in the human stomach³⁹. An exciting recent study in rhesus monkeys reported that, although Helicobacter species dominate the gastric community when present, infection with H pylori does not significantly alter the relative abundance of other taxa, except for H suis, which is reduced in proportion. These findings indicate that rhesus macaques represent another model for studying interactions between H pylori and constituents of the gastric microbiota⁶¹.

Esophageal Adenocarcinoma

Gastroesophageal reflux disease (GERD) is the strongest known risk factor for Barrett's esophagus, a metaplasia associated with an increased risk for esophageal adenocarcinoma. Among white males, the incidence of esophageal adenocarcinoma is increasing rapidly⁶² and the decreasing incidences of H pylori carriage and gastric cancer in developed countries over the past century have been diametrically opposed by this rapidly increasing incidence of GERD and its sequelae⁴. Carriage of H pylori is associated with a significantly reduced risk of developing esophageal adenocarcinoma⁶².

How can the ability of H pylori strains to increase risk for distal gastric cancer be reconciled with reciprocal effects against GERD, Barrett's esophagus, and esophageal adenocarcinoma? The location of inflammation within the gastric niche likely contributes to this dichotomy. By inhibiting parietal cell function and/or inducing the development of atrophic gastritis, inflammation in the acid-secreting gastric body induced by H pylori (especially in persons with polymorphisms that increase expression of IL1b) blunts the high-level acid secretion necessary for the development of GERD and its sequelae. Other potential mechanisms include changes in gastric hormone interactions and in T-cell populations induced by the loss of H pylori. Another possibility is that alterations in the gastric microbiome resulting from the loss of H pylori contribute to an increase in reflux-mediated esophageal adenocarcinoma. As noted above, there are distinct differences in the gastric microbial ecosystem among H pylori-infected and -uninfected subjects; it is plausible that the microbiota in one niche alter cancer risk at an adjacent site.

However, a recent single study demonstrated that the human distal esophagus has its own microbiota, which is perturbed during inflammation and metaplasia⁶³. This study defined esophageal bacterial communities using a 16S rRNA gene survey and identified 2 distinct microbiome clusters, termed type I and type II⁶³. The type I microbiome was dominated by Gram-positive bacteria, primarily Firmicutes. The type II microbiome was composed of a higher percentage of Gram-negative bacteria (53% vs 15% in type I), with the phyla Bacteroidetes, Proteobacteria, Fusobacteria, and Spirochaetes being the most highly represented⁶³. The prevalence of Streptococcus was higher in type I vs type II microbiomes, while the prevalence of Gram-negative anaerobes or microaerophiles was increased in type II vs type I populations. Importantly, a type I microbiome was significantly associated with the presence of an endoscopically intact esophagus whereas a type II microbiome was strongly linked to esophagitis and Barrett's esophagus⁶³.

Although cause and effect are difficult to establish conclusively in these types of studies, there are putative mechanisms by which different microbiome populations can affect esophageal disease (Figure 2). Gram-negative organisms, which predominate in reflux-associated type II esophageal microbiomes, produce specific immune-activating constituents such as lipopolysaccharide (LPS), which can activate innate immune responses that lead to disease⁶⁴. LPS engages the innate immune system Toll-like receptor (TLR)4, leading to NFκB activation; levels of activated NFκB are increased in esophageal specimens from patients with reflux esophagitis, Barrett's esophagus, and esophageal adenocarcinoma⁶⁴. Increased activation of NFκB is associated with increased levels of inflammatory cytokines such as IL1β, IL6, IL8, and TNFα⁶⁴. Another effect of LPS-induced signaling is activation of inducible nitric oxide synthase (iNOS). Activation of iNOS decreases the basal tone of the lower esophageal sphincter, which, over prolonged periods of time, increases the risk for reflux and its sequelae⁶⁵. Further, LPS has been shown in rodents to delay gastric emptying, which increases the amount of gastric content refluxed into the esophagus⁶⁶. Although further studies are needed, these results are exciting—the esophageal microbiome might be manipulated with antibiotics, probiotics, or inhibitors of specific host cell pathways (e.g. NFκB) to prevent disorders at this site.

Figure 2. Potential effects of the gastroesophageal microbiome on complications of gastroesophageal reflux disease.

The healthy esophagus has a type I microbiome, dominated by Gram-positive bacteria. In contrast, patients with GERD or Barrett's esophagus have a different population of esophageal microbes (a type II microbiome), comprised primarily of Gram-negative bacteria. There are several mechanisms through which these different microbial populations might affect gastroesophageal reflux. Gram-negative organisms produce immune-activating molecules such as LPS, which induces innate immune responses that lead to disease⁶⁴. LPS binds to and activates TLR4, leading to activation of NFκB; levels of activated NFκB are increased in esophageal samples from patients with reflux esophagitis, Barrett's esophagus, and esophageal adenocarcinoma⁶⁴. Increased activation of NFκB leads to increased production of inflammatory cytokines such as IL1β, IL6, IL8, and TNFα⁶⁴. In this manner, the type II esophageal microbiome contributes to esophagitis. LPS signaling also increases levels of iNOS, which reduces the basal tone of the lower esophageal sphincter. Over prolonged periods, this increases risk for reflux and its sequelae⁶⁵. LPS has also been shown to delay gastric emptying in rodents, which increases the amount of gastric contents refluxed into the esophagus⁶⁶. Production of iNOS can therefore lower the threshold for reflux of gastric contents into the esophagus.

Gastric and Esophageal Cancer Summary

The explosion of data enumerating and cataloguing non-H pylori members of the gastric microbial ecosystem, combined with animal studies highlighting the importance of collaboration between H pylori and the gastric microbiome, has challenged dogma about the pathogenesis of gastric cancer. In contrast to a model that focused on interactions between different strains of H pylori and genetic features of the host, researchers must now consider an alternative model, in which early interactions between H pylori and human tissues alter the structure of the gastric microbiome, by promoting gastric atrophy. Microbial blooms that develop in response to less-acidic conditions lead to new microbiome populations that might promote carcinogenesis.

Although less information is available about the effects of the microbiome on esophageal cancer, it is plausible that alterations of the gastric microbiome contribute to the increased incidence of esophageal adenocarcinomas—particularly those that arise within the gastroesophageal junction. However, there is an endogenous esophageal microbiome that differs between persons with healthy esophageal mucosa and those with Barrett's esophagus. Delineation of the fundamental relationships between members of the gastric and esophageal microbial communities is an important step in developing strategies to deliver beneficial effector species to a desired location, or engineering microbial species that can colonize these sites.

Colon Cancer

Colon cancer is an age-related disease; its incidence is predicted to increase by 52% in the next 17 years.⁶⁷ A culmination of multiple genetic and epigenetic events contributes to pathogenesis of colon cancer. Colorectal cancer (CRC) is initiated by mutations in tumor suppressor genes (APC, CTNNB1, p53) and oncogenes (KRAS), which transform healthy mucosa to adenoma and cancer.^{68, 69} Although mutations in these genes have clear roles in development of CRC, the events that lead to acquisition of these mutations and epigenetic changes are less clear. Arguably, understanding these earlier events offers a greater possibility for preventing CRC. Recent studies have identified microbial and environmental factors, such as diet and lifestyle, that can promote colon cancer.^{70, 71} Advances in sequencing and computational technology have facilitated determination of the role of the intestinal microbiota in CRC. We discuss human studies that have linked bacteria with colon cancer and findings from animal models.

The Microbiome in Patients with Polyps or Colon Cancer

In the colon, mammalian cells coexist with microorganisms, comprising bacteria, Archaea, viruses, and unicellular eukaryotes. Bacteria alone constitute about 90% of the total cell population in the colon.⁷² The colonic microbiota is primarily composed of Gram-negative Bacteroidetes and Gram-positive Firmicutes whereas Proteobacteria, Actinobacteria, and Fusobacteria form minor components.⁷³ Intestinal microbes help maintain homeostasis, contributing to immune development, preventing pathogen colonization, processing drug metabolites, and releasing key nutrients and energy from the diet. This complex community develops within the first 2 years after birth and remains stable throughout the lifetime, but varies among individuals.^{74, 75} The intestinal microbiota also has spatial variation, with differences between lumen- and mucosa-associated bacteria.⁷⁶

In spite of the beneficial effects of the microbiome, a small percentage of people develop colon cancer. The incidence of colon cancer is approximately 12-fold higher than cancer in the small intestine⁷⁷, which harbors approximately 10¹⁰ fewer bacteria than the colon. Chronic colon inflammation, which may be caused by alterations in the intestinal microbiota, promotes colon carcinogenesis; cancers associated with inflammatory bowel diseases are prime examples of this concept.⁷⁸ Initially colon cancer was associated with single pathogenic bacterial species, such as Streptococcus gallolyticus, H pylori, or adherent invasive E coli.^{37, 79, 80} However, studies from the last decade support the concept that multiple bacterial species contribute to CRC (Table 1). Disturbances in the composition, distribution, or metabolism of the colon microbiota might shift the homeostatic environment of the colon towards inflammation, dysplasia, and cancer.

Table 1.

Microbiota Changes Associated with Colon Cancer

Patients	Changes in the Microbiota	Reference
29 patients with colon adenomas, 31 with CRC, 34 patients with symptoms but normal colonoscopy results, 31 asymptomatic patients (controls)	Increased presence of intracellular E coli in patients with adenoma and CRC	141
Biopsies from 21 patients with adenomas and 23 without (controls)	Patients with adenomas had increased proportions of Proteobacteria, Faecalibacterium, and Dorea and decreased levels of Coprococcus and Bacteroides	81
Fecal samples from 60 patients with CRC and 119 healthy subjects (controls)	Pyrosequencing on a small subset of samples indicated microbiota differences in CRC patients vs controls. Higher Bacteroides/Prevetolla in patients with CRC, determined by quantitative PCR.	82
6 patients with colon adenocarcinomas; matched tumor and healthy tissues (5–10 cm from tumor) biopsies	Microbial communities differed between tumor and healthy tissues. Tumors had more Bacteroidetes and Coriobacteridae, and less Firmicutes and Enterobacteriaceae	90
Fecal samples from 46 healthy individuals and 46 patients with CRC	Patients with CRC had higher proportions of pathogenic bacteria including Enterococcus, Escherichia/Shigella, B fragilis, and Klebsiella and reduced proportions of butyrate-producing bacteria, including Roseburia and members of family Lachnospiraceae	123
46 patients with CRC: 21 stool samples, 32 rectal swab samples, 27 tumors tissues and matched normal surrounding tissues (2–5 cm and 10–20 cm away from tumor) 22 stool and 34 rectal swab samples from healthy volunteers (controls)	Microbiota were similar between matched samples from cancerous and non-cancerous tissues; cancer tissues had lower bacterial diversity. Microbiota differed between cancerous tissue and stool with over-representation of 6 and under-representation of 2 bacterial families in cancerous tissue compared to stool. Stool and mucosal-associated microbiota differed between CRC patients and controls.	76
88 CRC patients and matched normal tissues	Increased Fusobacterium and Streptococcaceae in tumor tissues	86
CRC and non-tumor colon tissues from 95 patients	Increased F nucleatum in tumor tissues	85
Stool samples from 10 patients with colon cancer and 11 healthy individuals (controls)	No significant differences in microbial composition and diversity between patients and controls; relatively low levels of Bacteroides and Prevotella and higher percentages of Akkermansia muciniphila in patients	84
Fecal samples 344 patients with advanced CRC and 344 healthy individuals (controls)	Increased abundance of Enterococcus and Streptococcus spp. in the patients and increased prevalence of butyrate producing bacterial groups (Roseburia, Clostridium) in controls.	89

In efforts to understand how the microbiome participates in CRC development, studies have been performed on either mucosa-associated bacteria or the stool microbiome in patients with CRC. Bacteria that adhere to the mucosa in patients with adenomas have increased diversity, compared to those of patients without adenomas.⁸¹ Studies of the stool microbiome have found an increase in Bacteroides species^{72, 82, 83}, a decrease in butyrate-producing bacteria^{72, 83, 84}, and an increase in potentially pathogenic bacteria such as Fusobacterium species.^{83, 85, 86, 87} The finding of increased Bacteroides is interesting, because toxin production could contribute to inflammation and cancer.⁸⁸ In studies of mucosa-associated bacteria that compared cancer and surrounding tissues, cancer tissues had overrepresentation of Coriobacteridae.^{89, 90} It is likely that local changes in the microbiome of cancer tissues are related to changes in the availability of nutrients and other conditions created by the cancer cells themselves.^{84, 91} Studies of the composition of the stool may someday be helpful, especially if paired with PCR-based studies of host DNA mutations in oncogenes, to screen patients for CRC or adenomas. Ideally, prospective studies would investigate whether changes in the microbiome precede the development of adenomas or cancer.

Diet, antibiotics, and stress are extrinsic and intrinsic factors that contribute to microbiome dysbiosis.^{73, 74} High-throughput 16S sequencing studies have led to models of mechanisms by which bacteria could contribute to CRC development. In one model, called the driver-passenger model, certain populations of bacteria (drivers) initiate CRC by damaging DNA in the intestinal epithelium.⁹² Initiation of tumorigenesis alters the intestinal environment, resulting in overgrowth of opportunistic bacteria (passengers), and possible depletion of driver bacteria.⁹² This model contrasts the α bug model, wherein bacteria with virulent properties (α bugs) initiate tumorigenesis and create helper bugs by modulating the gut microbiota. Together, these bugs take over commensal bacterial populations⁹³ (see models in Figure 3).

Figure 3. Bacteria associated colon cancer.

In a homeostatic colonic environment, bacteria are separated from the colonic epithelial layer by mucus, anti-microbial peptides, and secreted immunoglobulin A (IgA). Certain environmental factors induce microbial dysbiosis, leading to interactions between microbes and intestinal epithelial cells. Dysbiotic microbiota induces colon carcinogenesis by (A) releasing reactive oxygen species or reactive nitrogen species (ROS/RNS) and causing DNA damage^{133, 134, 135, 136} (B) releasing toxins that directly damage DNA or induce cell proliferation genes⁸⁸ (C) generating MAMPs, which activate the innate immune response, recruit immune cells, and induce inflammation⁸⁷.

Viruses

Various studies, over the past decades, have demonstrated roles for viruses in CRC. John Cunningham (JC) virus, human papilloma viruses (HPV), and BK virus are some common viruses associated with CRC. However not all strains of every virus have equal oncogenic potential. Although studies have correlated viral DNA with CRC, it is not clear if viruses are sufficient for carcinogenesis. Several studies have reported the presence of JC virus in CRC patients.^{94, 95} T-antigen, an oncogenic early protein encoded by JC virus, interacts with β-catenin to activate genes that regulate proliferation and transformation.⁹⁶ In vitro and patient-based studies have correlated the presence of human cytomegalovirus (HCMV) with cancer progression and CRC.^{97, 98} In one study, HCMV was detected in 42% of cancer tissues and 5.6% of matched non-cancerous tissue. However, increased local inflammation might activate CMV. Also, in patients with CRC, disease severity correlated with HPV infection. HPV DNA was more commonly detected in early-stage than late-stage cancers. However, no association between viruses and CRC has been reported in other studies.⁹⁹

It is not clear if particular viruses are not detected in cancer patients because the viruses have similar functions to driver bacteria—they are involved in only early stages of cancer development, and are not required for tumor progression. Compared to the quantity of studies on the microbiota, few studies have focused on the virome as a factor in onset or progression of CRC. It is likely that a longer list of potentially pathogenic enteric viruses will eventually be associated with CRC.

Diet and Microbe Metabolites

Diet is an important environmental factor associated with CRC. Epidemiology studies have indicated that the Western diet, rich in meat and fat, is a risk factor for CRC, whereas diets rich in fiber protect against CRC.^{100, 101} Colonic bacteria have many genes that regulate metabolism; they metabolize dietary components that pass through the small intestine undigested. A fiber-rich diet results in saccharolytic fermentation of carbohydrates that leads to production of short-chain fatty acids (SCFAs), including acetate, butyrate, and propionate.¹⁰² Butyrate has anti-tumorigenic properties and is associated with decreased incidence of CRC.^{103, 104} Bacterial species such as Faecalibacterium prausnitzii from Clostridium cluster IV and Eubacterium/Roseburia species from cluster XIVa are some of the major butyrate producers found in human colon.¹⁰⁵

Proteolytic fermentation during consumption of diets rich in meat prompts generation of inflammatory and carcinogenic metabolites—notably phenols, ammonia, branched-chain SCFAs, and other nitrogen-rich metabolites.¹⁰⁶ Diets high in protein and fat content also promote growth of sulfate-reducing bacteria, such as Desulfovibrio vulgaris, and generate excess genotoxic hydrogen sulfide. High-fat diets generate bile acids, which are excellent substrates for bacteria that express 7α-dehydroxylating enzymes. These bacteria convert primary bile acids to potentially carcinogenic secondary bile acids such as lithocholic acid and deoxycholic acid.¹⁰⁷

Recent studies have focused on deciphering the link between diet, the microbiota, and CRC.^{89, 108} These studies reported lower levels of health-promoting butyrate-producing bacteria and increased metabolites (SCFA and butyrate) in meat-eating African-Americans (who are at high risk for CRC), and patients with low-fiber diets and advanced colorectal adenoma. Secondary bile acids and microbial genes that control their generation were increased in fecal samples from meat-eating African-Americans, compared to native Africans with high-fiber diets.¹⁰⁸ The gut microbiota also differed between subjects and controls in these studies. These differences in the microbiota might be attributed to differences in genetic and environmental factors between Africans and African Americans. However, they indicate the importance of using integrated approaches, including metabolomics analyses, in studies of the microbiome, to fully assess how dysbiotic microbiota contribute to cancer development.

Host Immune Recognition of Bacteria

The innate immune component of the host immune system recognizes microbial molecules such as LPS, flagellin, peptidoglycans, and other microbe-associated molecular patterns (MAMPs). Pattern recognition receptors (PRRs) are a well-studied component of the innate immune system; these include Nod-like receptors (NLRs), TLRs, and retinoic acid inducible gene I-like receptors.^{109, 110} Activation of PRRs regulates proliferation, barrier function, and inflammatory pathways in multiple cell types.^{111, 112} Several studies have associated colon neoplasias with genetic polymorphisms in, or changes in expression of, genes that control PRR pathways. For example, CRC tissues have increased levels of TLR2, 4, 7, 8, and 9 mRNAs, compared to healthy surrounding tissue (for review see¹¹³). A number of studies have investigated expression patterns of TLRs in cancer tissues, compared to healthy tissues and adenomas. Cancer tissues had increased expression of epithelial TLR4, compared to adenomas and healthy mucosa.¹¹⁴ In a meta-analysis of cDNA microarray studies, expression of genes in the NLR signaling pathways was significantly altered in CRC tissues, compared with non-tumor tissues.¹¹⁵

Although human studies have linked variants in bacterial recognition genes with CRC, there is no clear mechanism by which they alter colon cancer susceptibility. Animal studies fill this knowledge gap. Most of these studies have involved mice with disruptions of PRR pathway genes. From these studies it is evident that innate immune pathways mediate chronic inflammation in the presence of bacterial infection or chemical epithelial injury. Chronic inflammation is a major risk factor for CRC and results in immune cell recruitment and release of inflammatory mediators such as TNFα, IL6, IL1b and other cytokines.¹¹⁶ These inflammatory mediators activate NF-κB to initiate a cascade of events that culminate in colon carcinogenesis (Figure 4). In mice, TLR4 deletion protects against CRC, whereas overexpression of TLR4 results in β-catenin activation and increased colitis-associated cancer development following administration of azoxymethane (AOM, a mutagen) and dextran sulfate sodium (DSS, induces colitis).^{114, 117, 118}

Figure 4. Bacteria activate TLR signaling to promote colon carcinogenesis.

Microbial dysbiosis or mutations can activate TLR signaling, leading to phosphorylation of β–catenin, cell proliferation, and tumor progression¹¹⁴. It is not clear whether other TLRs promote colon carcinogenesis via activation of β-catenin.

However, TLR2^−/− mice develop more tumors than wild-type mice.¹¹⁹ Loss of Myd88, an adapter that transduces most TLR signals, reduces tumor development in mice following administration of only AOM, but increases colitis-associated cancer in mice given AOM and DSS.¹²⁰ This finding has been attributed to loss of IL-18 production by myeloid cells. APC^min/+Myd88^−/− mice develop fewer colon tumors than APC^min/+ mice, indicating that bacterial signaling contributes to tumorigenesis in the context of APC mutations.¹²¹ A subsequent study found that MyD88-dependent activation of ERK stabilizes β-catenin; in this way, the absence of MyD88 protects against APC-dependent tumors.¹²² Thus depending on the model used, MyD88 deficiency either protects or increases tumorigenesis.

The different outcomes of mice lacking MyD88 likely reflect nuances of the animal models and the relative dependence on inflammation and bacterial signaling. Moreover, because these models often involve total knockouts, it is difficult to know which cell type in the intestine (epithelial, myeloid, or other) responds to bacterial signals to contribute to carcinogenesis. Similar to humans, Apc^F/wt mice with a Cdx2-Cre transgene (also known as CPC-APC mice) develop tumors in the distal colon, due to allelic loss of Apc. In these mice, expression of IL23 and IL17 are upregulated in tumor tissues. Levels of IL23 and IL17 and tumor growth decrease after antibiotic treatment and deletion of Myd88.¹²³ These observations support the notion that bacterial activation of PRRs results in inflammation, immune activation, or proliferation that modifies the growth of colonic tumors.

Animal Models and Luminal Bacteria

Genetically engineered models of CRC, raised under germ-free conditions, are invaluable for studying links between colon neoplasia and gut microbes. APC^min/+ mice are commonly used to study the role of environmental factors, such as diet and intestinal bacteria, on CRC. APC^min/+ mice carry a heterozygous mutation in the APC gene and develop spontaneous tumors, mostly in the small intestine.¹²⁴ This is in contrast to humans, in which APC mutations cause colonic tumors. Germ-free APC^min/+ mice develop fewer small intestinal and colon polyps than their conventionally raised counterparts.¹²⁵

Mice with disruptions in specific genes (Il10^−/−, Tcrb/p53^−/−, Gpx1/Gpx2^−/−, and Tgfb1/Rag2^−/−) develop fewer tumors under GF than conventional conditions, as do mice exposed to AOM and DSS.¹²⁶ Studies of mice that develop colitis that promotes neoplasia support the involvement of microbes in tumor development. Infection of Rag2^−/− or Il10^−/− mice with H hepaticus leads to colitis-associated cancer.^{127, 128} However mono-association with Bacteroides vulgatus reduced colorectal tumorigenesis in Il10^−/− mice, compared to conventional Il10^−/− mice.¹²⁹ Therefore, in the same genetic background, different microbial species can have different effects on colon carcinogenesis.

The incidence and severity of colitis-associated cancer are reduced in mice given antibiotics.^{130, 131} Moreover co-housing and microbiota transplant experiments have shown that development of colitis-associated cancer can be maternally transmitted, through the microbiota in genetically susceptible mice.^{38, 130}Nod2^−/− mice have an altered microbiota, compared to wild-type mice and develop more tumors after AOM-DSS treatment.¹³⁰ The susceptibility to tumors in Nod2−/− appears to track with its microbiota since it can be diminished by treatment of mice with antibiotics or stool transplants from wild-type mice; conversely, wild-type mice can have increased numbers of tumors when colonized with dysbiotic Nod2−/− microbiota. These data suggest that manipulation of the microbiota can be used to prevent development of CRC if we could identify the principal agents driving the neoplasia risk.

Infectious Agents That Promote Colonic Neoplasia

Studies in animal models have shown that bacteria can promote tumorigenesis by releasing genotoxic compounds. Enterotoxigenic Bacteroides fragilis secretes a 20 kDa zinc-dependent metalloprotease, B fragilis toxin, which stimulates cleavage of E-cadherin and activates β-catenin. Infection of APC^min/+ mice with Enterotoxigenic Bacteroides fragilis increases tumorigenesis in the large intestine by activating STAT3, which leads to production of IL17 by T helper 17 cells; by contrast, uninfected APC^min/+ mice usually develop only small intestinal tumors.⁸⁸ Furthermore, data from epidemiologic studies indicate a higher prevalence of Enterotoxigenic Bacteroides fragilis infection among patients with CRC.

Fusobacterium has also been consistently associated with CRC.^{85, 86} Greater proportions of this bacterium have been detected in colorectal tumor and adenoma samples compared with normal tissue. In APC^min/+ mice, Fusobacterium was found to contribute to colorectal tumorigenesis by attracting myeloid immune cells to adenomas.⁸⁷ Fusobacterium adhesin A (FadA) is a surface molecule expressed by Fusobacterium. In addition to helping Fusobacterium attach to and invade cancer cells, FadA was recently shown to modulate E-cadherin signaling to β-catenin.¹³²

Some bacterial infections cause generation of reactive oxygen and nitric oxide species that damage DNA to transform cells. Arthur et al. demonstrated that an E coli strain carrying a polyketide synthase pathogenicity island that encodes colibactin increases the invasiveness of adenocarcinomas that develop in Il10^−/− mice following administration of AOM.⁷⁹ Colibactin has been shown to induce genetic instability by causing DNA double strand breaks.¹³³ Deletion of the pathogenicity island reduced the rate of invasive tumors but did not change the level of inflammation caused by this E coli strain. These data suggest that it is the synthesis by the bacteria of colibactin, rather than only the degree of inflammation, that increased the neoplasia risk. They also detected this particular strain in approximately 65% of CRC tissue specimens, compared with 20% of tissue from healthy individuals. Enterococcus faecalis has been shown to produce extracellular superoxide and hydrogen peroxide, which damage DNA in colonic epithelial cells.^{134, 135, 136} On the other hand, bacteria can express superoxide dismutase to decrease reactive oxygen damage.¹³⁷ These latter species may be beneficial for patients at risk for CRC.

Probiotics in Colon Cancer

A number of in vitro and animal studies provide evidence that consuming probiotics suppresses CRC. These studies have also proposed multiple pathways by which probiotics could inhibit colon cancer—notably by influencing innate immune pathways and apoptosis, reducing oxidative stress, and modulating intestinal bacteria and their metabolism.¹³⁸ Despite these laboratory-based studies, only a limited number of clinical trials have evaluated the potential of probiotics in suppressing CRC. Lactobacillus species are the most commonly used probiotics in clinical trials. Lactobacillus johnsonii reduced the concentration of Enterobacters and modulated intestinal immune response in CRC patients, whereas Bifidobacterium longum did not have any effect.¹³⁹ In another study, Lactobacillus casei suppressed colorectal tumor growth in patients, after 2–4 years of treatment. However these clinical trials are limited by the small number of subjects and their short duration.¹⁴⁰

Colon Cancer Summary

Development of CRC involves a combination of inherited and acquired mutations in colonic epithelial cells, along with environmental factors including the intestinal microbiota. It is not clear whether there are specific microbes that are particularly pathogenic and directly cause colorectal carcinogenesis, or the process requires specific interactions between host tissues and microbes; a variety of mechanisms are likely to be involved. It would be a challenge to identify a single type of microbe, among myriad of microbes present in the human colon, that cause cancer. Even identifying a microbial signature associated with CRC is a challenge, because the intestinal microbiota change during disease development and progression. High-throughput sequencing technologies have facilitated whole-microbiome and metagenome studies, but the absence of proper controls poses certain limitations. Moreover, it is difficult to control for genetic variation among individuals—the contribution of the microbiota to CRC is influenced by host genotype. Blood or stool samples can be screened to monitor microbes and their metabolites, and profiles of microbial metabolites may correlate with CRC development. This could be a valuable screening tool as bacterial species share an exceptional number of metabolic genes. However an in-depth profiling of bacterial metabolites is needed to identify microbial metabolites that could affect CRC development. Antibiotics and probiotics might affect CRC development, by manipulating the composition and function of the microbiota. A carefully designed standardized approach, with proper dietary and other controls, as well as follow-up markers, is needed to evaluate cancer treatment or prevention strategies.

Future Directions

Detailed analyses of the GI microbiome have associated diseases such as cancer with specific patterns of microbes. Researchers now need to investigate causes vs effects of these relationships. This could be accomplished in prospective trials in which specific microbes are altered or eliminated, and effects on disease incidence or progression are assessed. Animals with human microbiota can also be used to study the effects of different microbial populations on disease development. These studies are important, as manipulation of microbial communities has vast therapeutic potential. Microbes engineered to express specific genes or produce specific metabolites might be delivered to particular niches within the GI tract to treat or prevent disease. The GI microbiome also has the capacity to alter the metabolism of chemotherapeutic agents. Certain drugs, doses, or regimens might be selected for patients based on the composition of their microbiome, or the microbiota might be monitored to evaluate response to therapy. The microbiome is an exciting new frontier for the prevention and management of malignancies of the GI tract.