Emerging Insights into the Esophageal Microbiome

Abstract

Purpose of Review

Analysis of the esophageal microbiome remains a relatively new field of research, and most studies to date have focused on characterizing the esophageal microbiome in states of health and disease. Microbiome alterations have been implicated in the pathogenesis of inflammatory and neoplastic conditions in the colon and elsewhere in the gastrointestinal tract. The epidemiology of various esophageal conditions including Barrett’s esophagus (BE), esophageal adenocarcinoma (EAC), esophageal squamous cell carcinoma (ESCC) and eosinophilic esophagitis (EoE) point to the microbiome as a potential co-factor in disease pathogenesis, and the possibility exists that these microbiome alterations could contribute directly to the inflammatory environments necessary for the carcinogenesis or atopy involved in these conditions.

Recent Findings

The native esophageal microbiome is similar in composition to the oral microbiome, with a high relative abundance of the phylum Firmicutes and the genus Streptococcus. Limited studies to date suggest that there are certain microbiome alterations associated with esophageal diseases. Additionally, it may be possible to indirectly assess the esophageal microbiome via non-endoscopic means. This raises the possibility that non-invasive microbiome analysis could be used for disease screening and monitoring.

Summary

Further understanding of the role of the esophageal microbiome in disease pathogenesis, as well as methods for microbiome alteration may help elucidate future targets for disease modifying therapies, or minimally invasive screening tools in patients at high risk for development of various esophageal conditions.

Introduction

The esophagus previously was felt to be largely absent of bacteria. However, with the advent of culture-independent methods such as 16S rRNA gene sequencing, it was found the esophagus contains a diverse microbiome [1]. However, relatively little is known with regard to the esophageal microbiome and its role in disease pathogenesis. This is due in part to the difficulty in sampling the esophageal microbiome; specifically, upper endoscopy is required for targeted specimen collection. This is in contrast to the mouth and colon, for example, where samples can be collected from large numbers of patients non-invasively.

The incidence of certain esophageal conditions such as Barrett’s esophagus (BE), esophageal adenocarcinoma (EAC), and eosinophilic esophagitis (EoE) has risen dramatically over the past several decades. The reasons underlying these trends are incompletely understood. Coincident with the rise of these conditions has been a profound shift in the upper gastrointestinal microbiome since the mid-20^th century, around the same time that widespread antibiotic use began. During this period, the prevalence of Helicobacter pylori infection of the stomach began to decrease, which at least raises the possibility that profound, population-level upper GI microbiome shifts resulted in an increased predisposition to the development of a variety of esophageal diseases. (Figure 1) In this paper we aim to summarize findings to date on microbiome alterations in certain esophageal diseases, discuss potential clinical implications, and lay out future research directions.

Figure 1. Temporal correlation between population-level alterations to the upper GI microbiome and emergence of various esophageal diseases in western countries.

H. pylori infections began to decline and antibiotic use began to expand starting in the mid-20^th century. This was followed in time by marked rises in the incidence of esophageal adenocarcinoma and eosinophilic esophagitis.

Techniques for Sampling and Analyzing the Esophageal Microbiome

Traditionally, sampling of the esophagus has involved endoscopy with brushings or biopsies. The invasiveness and expense of this procedure has limited sample sizes of esophageal microbiome studies and has greatly hampered the ability to assess longitudinal changes. Gall et al. compared endoscopic mucosal brushings to biopsies and found that brushings gave superior yield of bacterial DNA and increased the ratio of bacterial DNA to host DNA, which is significant in light of the low bacterial concentration in the esophagus [2]. We generally collect samples using endoscopic brushes, passing the brush back and forth ten times in each of four quadrants. Within individual patients in whom two brushings were taken from the same site, we have found that taxon-level relative abundances are very strongly correlated (unpublished data), and a second brushing from a particular site may not add markedly to findings from a single, thorough brush sampling.

There are non-endoscopic techniques and devices that have been proposed for esophageal microbiome sampling. Fillon et al. found that analyzing the microbiome from the proximal segment of string from the Enterotest™ (which the authors coined the “Esophageal String Test”) provided very similar microbiome profiles compared to matched endoscopic biopsies [3]. The Cytosponge is a tethered capsule designed for cell sampling in the esophagus and has been studied as a non-endoscopic method to diagnose Barrett’s esophagus [4]. The capsule dissolves upon reaching the stomach, a spherical mesh of 3 cm diameter is exposed, and the mesh is then withdrawn through the mouth. As a tool for studying the esophageal microbiome, the Cytosponge yields higher quantities of microbial DNA than endoscopic brushes or biopsies [5]. Inflatable balloons have also been used to sample the upper gastrointestinal microbiome [6]. However, none of these techniques exclusively sample the esophageal microbiome. The oral cavity has a much higher concentration of bacteria compared to the esophagus, and thus contamination with oral flora likely overwhelms the ability to assess microbiome alterations specific to the esophagus.

Culture independent analysis of the esophageal microbiome is largely restricted to 16S rRNA gene sequencing. Since the bacterial concentration of the esophagus is low compared to other sites such as the colon or mouth, the host:bacteria DNA ratio for esophageal samples is very high. As such, more detailed profiling of the esophageal microbiome using methods such as whole genome sequencing is challenging. While post-sequencing methods have been attempted to allow for detailed microbiome profiling of samples with low bacterial abundance[7, 8], these are not widely employed at the present time.

Native Esophageal Microbiome

The esophageal microbiome is broadly similar to the oropharyngeal microbiome but with key taxonomic differences [1, 9, 10]. The first studies on the esophageal microbiota, dating back to the early 1980s and based on culture, demonstrated that the esophagus was not a sterile site and did not simply contain a transient microbial population introduced from the oral cavity [9, 11, 12].

Pei et al. performed a detailed characterization of the esophageal microbiome in healthy individuals using culture-independent methods [1]. 16S rRNA gene sequencing was performed on esophageal endoscopy samples from four patients without esophageal disease. Ninety-five species-level operational taxonomic units were identified, belonging to six phyla: Firmicutes (70%), Bacteroidetes (20%), Actinobacteria (4%), Proteobacteria (2%), Fusobacteria (2%), and TM7 (1%). The most common genera of bacteria were Streptococcus, Prevotella, and Veillonella. While the normal esophageal microbiome had many commonalities with the oral microbiome, key differences, such as the absence of Spirochaetes, suggested a unique esophageal microbiome that was not solely the product of oropharyngeal colonization. Subsequent studies have confirmed the presence of the six phyla above and overall predominance of Gram-positive bacteria with Streptococcus as a consistently highly abundant genus [3, 13–15].

Factors that Influence the Esophageal Microbiome

Proton Pump Inhibitors (PPIs)

PPIs are theorized to alter the esophageal and gastric microbiome both through increasing pH of gastric secretions and also possibly by directly targeting the bacterial proton pumps of certain bacteria that contain P-type ATPase enzymes, such as Streptococcus pneumoniae and H. pylori. Amir et al. analyzed gastric refluxate and esophageal biopsies of eight patients with GERD, before and after eight weeks of PPI treatment (lansoprazole 30mg twice daily) [14]. The authors found that the microbial communities of esophageal biopsy samples were significantly altered after PPI treatment, with decreased Comamonadaceae and increased Clostridiaceae and Micrococcaceae. The family Methylobacteriaceae, which were increased in gastric aspirates among BE/esophagitis patients before PPIs, were markedly depleted after PPI therapy. The authors speculated that PPI therapy may result in decreased availability of potentially toxic one-carbon compounds, a preferred carbon and energy source for Methylobacteriaceae. Sanduleanu et al. cultured gastric aspirates from 109 patients taking PPIs and 75 patients with untreated GERD and found that the prevalence of oropharyngeal bacteria, including Neisseria, Streptococcus and Corynebacterium, was significantly increased in patients on PPIs when compared with untreated patients [16]. These studies suggest that PPIs can have marked impacts on the upper GI microbiome, although the disease modifying effects of the microbiome changes is unknown.

Diet

Dietary factors are associated with altered risk of BE, EAC and ESCC, and dietary antigens are directly involved in EoE-related inflammation. It is possible that some effects of diet on esophageal disease risk and manifestations are mediated by changes to the local microbiome. Diet has profound effects on the colonic microbiome. High-fiber diets are generally associated with decreased colonic and systemic inflammation, whereas high-fat diets have the opposite effects [17–21]. A recent study by Mehta et al. found that “prudent” diets high in fiber were associated with decreased risks of colon cancer, but this protective effect only applied to Fusobacterium nucleatum-positive tumors [22]. These findings suggest that the effects of diet on colon cancer risk are mediated at least in part by alterations to the microbiome. However, there is little data on the effects of diet on the esophageal microbiome.

In a prospective study of 33 children with EoE, implementation of a six-food elimination diet on patients with active EoE did not have a marked impact on the esophageal microbiome [13]. However, the reintroduction of allergenic foods led to an increase in the prevalence of Granulicatella and Campylobacter[13]. Our group recently conducted a study of 47 patients who completed a 4-week dietary recall food frequency questionnaire validated for fat and fiber intake, and subsequently underwent endoscopy with esophageal brushings [23]. We found that increasing fiber intake positively correlated with increased relative abundance of Firmicutes and decreased relative abundance of Proteobacteria and Gram-negative bacteria as a whole. Interestingly, there were relatively few microbiome alterations based on dietary fat intake. However, longitudinal studies are necessary to better determine how diet alters the esophageal microbiome.

Barrett’s Esophagus (BE) and Esophageal Adenocarcinoma (EAC)

EAC is the predominant type of esophageal cancer in North America and Europe. Five-year survival from time of diagnosis is approximately 17%, as most tumors are discovered at later stages. BE is the precursor lesion that can be identified through endoscopy and followed for early detection of dysplasia and EAC. The incidence of EAC has been rising steadily and rapidly since at least the mid-1970s [24]. Between 1975 and 2009 the average annual percentage increase in incidence was 6.1% in men and 5.9% in women [25]. Data from the Connecticut Tumor Registry indicate that the age-adjusted incidence of EAC was very low (~0.4 cases per 100,000 person-years) and relatively stable until the late 1960s, and has had an overall approximately ten-fold increase since that time [26].

This pattern suggests that changes in exposures in the mid-20^th century most likely led to an increase in rates of BE with a subsequent rise in the incidence of EAC. However, the dramatic increase in incidence of EAC is beyond what would be expected just based on increased rates of GERD, obesity and smoking[27]. GERD prevalence began to rise in the 1970s, obesity in the United States began to rise markedly starting in the 1980s, and smoking rates have dropped in recent decades [28–32].

Widespread use of antibiotics began in the 1940s, which potentially led to dramatic, population-level shifts in the upper gastrointestinal microbiome. Helicobacter pylori is inversely associated with EAC risk[33]. H. pylori infection rates began to fall in the mid-20^th century, with a 26% decrease in seropositivity prevalence per decade between 1969 and 1989 [34]. The American and Western European diet also saw significant changes in the mid-20^th century, with the widespread production and consumption of processed foods and sugar-sweetened beverages [35]. The combination of the rise in antibiotic use, decreased H. pylori prevalence and significant dietary changes in the mid-20^th century, along with the inverse relationship between H. pylori infection and the development of EAC, all point to a role of the upper gastrointestinal microbiome in the pathogenesis of EAC. (Figure 1)

The Barrett’s Esophagus- and Reflux-Associated Microbiome

Studies to date suggest that BE and reflux esophagitis are associated with a modestly distinct esophageal microbiome compared to patients without esophageal disease. This microbiome appears to be enriched with Gram-negative bacteria and a broader range of bacteria overall. Yang et al. compared the microbiomes of distal esophageal biopsies in 12 patients with normal esophagus, 12 with esophagitis and 10 with BE [36]. Using clustering analyses they classified the microbiome into two types, arbitrarily named Type I and Type II. Type I was dominated by Streptococcus, and Type II contained decreased relative abundance of Streptococcus and a greater proportion of Gram-negative anaerobes and microaerophiles. Almost all (13/14) of the patients with a Type II microbiome had reflux esophagitis or BE. MacFarlane et al. cultured and sequenced esophageal biopsy and aspirate samples from 7 patients with BE and 7 patients with normal esophagus [37]. High levels of Campylobacter, a Gram-negative genus that has been linked to enteritis, periodontal infections, and tumor formation in animals, was found in 4 of 7 patients with BE but in none of the control subjects. Also, BE patients were colonized with a wider variety of both Gram-negative and Gram-positive bacteria.

Amir et al. performed 16S rRNA gene sequencing on squamous tissue and gastric aspirates from 15 normal patients and 16 patients with reflux esophagitis (11) or BE (5) [14]. There was relatively little difference in the squamous tissue microbiome between the groups. However, the gastric aspirates in patients with BE or reflux esophagitis had an increased and very high relative abundance Enterobacteriaceae, particularly of the genus Escherichia. Certain strains of E. coli promote colon cancer in mice in the presence of colonic inflammation, raising the possibility that this bacterium could play a role in the development of BE [38].

The Microbiome in Esophageal Adenocarcinoma

The microbiome associated with EAC remains poorly described. Blackett et al. used 16S rRNA sequencing of prespecified species chosen based on culture isolation, comparing controls with patients with GERD, BE and EAC [39]. This group found that in many ways the microbiome of EAC was more similar to the normal esophageal microbiome than to BE. The control and EAC samples had increased relative abundance of Bifidobacteria, Bacteroides, Fusobacteria, Veillonella, Staphylococcus and Lactobacilli and decreased relative abundance of Campylobacter when compared with BE samples. Elliott et al. compared the microbiome of tissue from patients with EAC to BE and normal controls [5]. The authors reported a decrease in microbial diversity in EAC samples and increased relative abundance of Lactobacillus fermentum. Roughly half of the tumors were dominated by a single species within the order Lactobacillales.

Zaidi et al. studied the microbiome in a surgical rat model of EAC as well as tissue from patients with and without EAC [40]. E. coli was abundant in BE and EAC tissue but not in BE-associated dysplasia. Tumor tissue from the rat model also demonstrated significant upregulation of various toll-like receptors (TLRs), which likely play a key role in recognition of and response to bacteria. Based on these findings, the authors proposed a potential model in which TLR activation by microbes could contribute to development of EAC.

However, findings from these studies should be interpreted with caution. Tumors have a markedly altered topography, which may favor colonization and growth of certain bacteria. As such, it is unclear whether microbiome alterations in EAC point to potential causal mechanism for specific taxa or merely colonization due to an altered macro- and micro-environment.

Potential Mechanisms for Microbiome-Mediated EAC Risk

Mechanistic studies on the role of specific bacteria on the development of EAC are lacking. It has been posited that Gram-negative bacteria could contribute to chronic inflammation through a number of mechanisms. Lipopolysaccharides (LPS), expressed on the surface of Gram-negative bacteria, can bind to TLRs and other cell surface receptors, inducing the pro-inflammatory NF-κβ pathway and promoting neoplastic progression [41]. Another possible mechanism for Gram-negative bacteria to promote esophageal neoplasia is through reduction of dietary nitrates to nitrites, which can then be converted into carcinogenic N-nitroso compounds by the acidic environment of the distal esophagus in patients with acid reflux [42, 43].

F. nucleatum is an anaerobic Gram-negative bacterium that is highly abundant in the oral cavity, plays a role in periodontal disease, and has been associated with the development of colon cancer in humans [44–46]. F. nucleatum can activate the beta-catenin pathway though interaction between E-cadherin and F. nucleatum-produced FadA[47]. The presence of F. nucleatum has also been associated with the CpG island methylator phenotype in premalignant colorectal lesions and colorectal carcinoma, opening up the possibility that this species may also mediate increased risk of carcinogenesis through epigenetic changes. [48]. However, a role of F. nucleatum in the development of EAC has not been described.

Esophageal Squamous Cell Carcinoma (ESCC)

ESCC accounts for roughly 90% of cases of esophageal cancer worldwide and remains the most common type of esophageal cancer in Asia, Africa, and South America and among African-Americans. In very high incidence areas such as Iran, Central Asia, and China, dietary factors such as poor nutritional status, low intake of fruits and vegetables, consumption of food and beverages at high temperatures, and consumption of carcinogenic food products are thought to account for a significant portion of cases[49, 50]. In the United States and Western Europe, rates of ESCC have been falling, with EAC now accounting for the majority of esophageal cancer cases[51].

The Microbiome in Esophageal Squamous Cell Carcinoma

Studies have demonstrated associations between diet and oral hygiene on risk for development of ESCC, suggesting a possible role of the esophageal and closely related oral microbiomes in the pathogenesis of ESCC. One study of 325 resected esophageal cancer specimens, of which 92% were ESCC, found increases in the quantity of F. nucleatum in esophageal cancer specimens compared to normal esophageal mucosa [52]. Further, the presence of F. nucleatum DNA in tumors was associated with shorter survival. These findings strongly suggest that specific bacteria such as F. nucleatum may play a role in ESCC development and may impact tumor behavior.

Yu et al. evaluated the upper GI microbiome collectively using a balloon device that sampled the stomach, esophagus and oral cavity in a Chinese population of 142 patients with esophageal squamous dysplasia and 191 controls and found that decreased alpha diversity was inversely associated with the presence of squamous dysplasia [6]. In an Iranian study of 19 patients with ESCC, 18 with squamous dysplasia, and 37 age- and sex-matched healthy controls, a greater abundance of Clostridiales and Erysipelotrichales was found in cancer and dysplasia tissue compared to normal esophagus from controls [53]. However, there were no significant differences in alpha diversity between cases and controls in this study. Collectively, these studies suggest that microbiome alterations occur in ESCC, although further studies are clearly warranted to assess causation.

Eosinophilic Esophagitis (EoE)

EoE is a chronic, immune/antigen-mediated, esophageal disease characterized clinically by symptoms related to esophageal dysfunction and histologically by eosinophil-predominant inflammation [54]. Although the pathogenesis is not fully understood, food or environmental allergens trigger a Th2-mediated response resulting in eosinophilic infiltration of the esophageal mucosa [55]. Esophageal tissue samples from affected patients are characterized by increased numbers of eosinophils, T cells and mast cells[56]. These findings, as well as the significant association between EoE and atopic conditions suggest an allergen-mediated pathogenesis of EoE [57, 58]. Elimination of specific dietary antigens can reduce or eliminate symptoms of EoE.

The first reported cases of probable EoE are from the 1960s, although it was not recognized as a unique clinical entity until about 1990 [59, 60]. Population based studies have demonstrated a progressively increasing incidence of EoE during the 1990s and 2000s [61]. One population based study that assessed the incidence of EoE in Olmsted County, Minnesota from 1976 to 2005 found significant increases in incidence from 1991 to 2005 [62]. EoE has a predilection for the more developed regions of the globe, with prevalence within these regions ranging from 4 to 400 per 100,000, and the highest prevalence reported by centers in Northern Europe and Australia[63]. One large population-based study in the United States estimated a prevalence of EoE of 57.7 per 100,000 persons and was highest in men 35–39 years old [64].

There are multiple hypotheses for the cause of the increased incidence of EoE. Allergic diseases in general have increased in developed countries in recent decades, possibly a result of a more sterile environment which results in exposure of the developing immune system to a smaller variety of antigens and a smaller range of antigens to which the body is tolerant [63]. In fact, a majority of patients with EoE have a history of other atopic conditions [57, 58].

Another theory is that changes in the esophageal microbiome may account for some of the increasing incidence of EoE. The historical trends in EoE incidence mirror those of EAC (Figure 1). Over the past half century there has been a major increase in antibiotic use and a steady decline in H. pylori infection rates, and as with EAC, H. pylori infection is associated with a decreased risk of EoE [65, 66].

The Microbiome of EoE

There is limited data on esophageal microbiome alterations associated with EoE. A prospective study was performed in 33 children with EoE, 18 of whom had active EoE and 15 of whom had inactive inflammation after dietary elimination treatment, and 35 healthy controls, all of whom underwent upper endoscopy with esophageal biopsies. [13] Patients with active EoE had increased relative abundance of Neisseria and Corynebacterium compared with controls, who had increased Streptococcus and Atopobium. This study did not find a significant effect of a six-food elimination diet on the esophageal microbiome. However, the reintroduction of certain foods led to an increase in the prevalence of Granulicatella and Campylobacter.

A study by Harris et al. used esophageal string testing to analyze esophageal samples from 70 children including 11 with untreated EoE and 26 with EoE in remission after treatment, 8 GERD controls and 25 healthy controls [15]. There was an increase in Haemophilus in the patients with EoE compared with the healthy controls. Also, there was evidence of increased bacterial load, but not diversity, in all subjects with EoE regardless of the degree of eosinophilia, compared with the healthy controls.

These studies hint at a distinct esophageal microbiome in EoE, although it is unclear whether any alterations play a role in the pathogenesis of EoE.

Potential Clinical Applications

The Oral Microbiome as a Screening Tool

The oral microbiome is broadly similar to the esophageal microbiome and very easy to sample. The oral microbiome has been studied as biomarker for various gastrointestinal malignancies, including ESCC, gastric cancer, pancreatic cancer, and colorectal cancer [67–69]. In a study of saliva samples from 87 incident and histopathologically diagnosed ESCC subjects, 63 subjects with squamous dysplasia, and 85 healthy controls, saliva in patients with ESCC had decreased microbial diversity and decreased relative abundance of several genera including Lautropia, Bulleidia, Catonella, Corynebacterium, Moryella, Peptococcus and Cardiobacterium[67]. These findings open up the possibility of using saliva to screen for ESCC in select populations.

Our unpublished data suggests that BE is associated with a unique oral microbiome including increased relative abundance of Firmicutes and decreased Proteobacteria. We created a three-taxon model that was able to distinguish patients with BE from controls based on their oral microbiome with 96.9% sensitivity and 88.2% specificity [70]. If specific taxa can be identified and validated for BE and/or EAC screening, then one could envision a rapid and inexpensive PCR-based test carried out in the primary care setting.

Can the Esophageal Microbiome be Modified to Alter Disease Risk or Outcomes?

If specific bacteria are involved in the development of esophageal diseases, then therapies aimed at modifying the esophageal microbiome could potentially lower risk or improve outcomes in those with disease. In a surgical rat model for EAC, rats given penicillin G and streptomycin had a non-significant reduction in EAC development [71]. There is relatively little research on the effects of probiotics on the esophageal microbiome. In a murine model of EoE, supplementation with Lactococcus lactis NCC 2287 significantly decreased esophageal and bronchoalveolar eosinophilia [72]. The use of systemic antibiotics may have negative effects elsewhere in the body, and topical therapies may prove more organ-selective and effective.

Given the close relationship between the oral and esophageal microbiome, another potential approach is to modify the oral microbiome. Our group is conducting a proof-of-concept trial to assess the effects of chlorhexidine mouth rinse on the esophageal microbiome (NCT02513784).

Future Directions for Research

Most research on the esophageal microbiome has involved characterizing the microbiome in disease states. Although these studies point to alterations in the microbiome associated with different diseases, data on causality are lacking. Prospective studies tracking longitudinal changes in the microbiome during disease progression and studies on the effects of various interventions on the esophageal microbiome are also needed. Such studies could both identify potential targets for risk modification and explore the clinical utility of markers for disease risk. If these changes in the microbiome are found to cause increased risk of esophageal disease, an important area of research will be to identify specific microbes that carry responsibility for conferring increased or decreased risk of diseases. Identification of antibiotics and probiotics which have an effect specific to these species, and finding best methods for local application may then be an important area of study.

Conclusions

The microbiome has been identified as a key mediator of inflammatory and neoplastic conditions of the gastrointestinal tract and the colon in particular. The epidemiology of various esophageal conditions such as Barrett’s esophagus and eosinophilic esophagitis point to the microbiome as potentially relevant players, although our understanding of the role of bacteria in the pathogenesis of these diseases remains limited. Future studies will hopefully elucidate the roles of specific bacteria in the esophagus and identify novel therapeutic approaches to modify disease risk and improve outcomes.