Abstract
The human stomach is a formidable barrier to orally ingested microorganisms and was long thought to be sterile. The discovery of Helicobacter pylori, a carcinogenic bacterial pathogen that infects the stomach mucosa of more than one half of all humans globally, has started a major paradigm shift in our understanding of the stomach as an ecological niche for bacteria. The special adaptations that enable H. pylori to colonize this well-protected habitat have been intensively studied over the last three decades. In contrast, our knowledge concerning bacteria other than H. pylori in the human stomach is still quite limited. However, a substantial body of evidence documents convincingly that bacteria can regularly be sampled from the stomachs of healthy adults. Commonly detected phyla include Firmicutes, Actinobacteria, Bacteroidetes, and Proteobacteria, and characteristic genera are Lactobacillus, Streptococcus, and Propionibacterium. In this review, we summarize the available literature about the gastric microbiota in humans and selected model animals, discuss the methods used in its characterization, and identify gaps in our knowledge that need to be addressed to advance our understanding of the bacterial colonization of the different layers of the gastric mucosa and its potential role in health and disease.
Introduction
The human stomach functions as a defensive barrier against orally ingested microorganisms. Due to the hostile conditions of the gastric environment, the gastric mucosa was long thought to be essentially sterile. This view changed with the discovery of the Gram-negative bacterium Helicobacter pylori in 1983. Helicobacter pylori chronically infects the gastric mucosa in more than one half of the human population worldwide. This infection always induces a chronic inflammation of the stomach mucosa, which causes no disease symptoms in most infected individuals. However, serious complications can arise from H. pylori infection, including gastric and duodenal ulcer disease and malignant stomach tumors. While the relationship between H. pylori and its human host has been studied in great detail over the last three decades, our knowledge about the presence of other bacteria in the stomach is quite limited, and there is even less information regarding the role of gastric microbiota in human health and disease. The principal aim of this article is to provide a comprehensive review of the literature on the gastric microbiota. We also cover H. pylori as the best-studied component of the microbiota, but limit this discussion to findings relevant in the context of its ability to colonize the gastric mucus and its impact on the presence of other bacteria in the stomach. Finally, we briefly highlight the (large) gaps in our knowledge about the gastric microbiota and emphasize future research questions.
The stomach as an ecological niche: anatomy, physiology, and compartments
The main functions of the stomach are the initiation of digestion and the inactivation of ingested microorganisms such as bacteria, viruses, fungi, or parasites to prevent them from reaching the intestine (Martinsen et al., 2005). This is achieved by the gastric juice, a complex fluid composed mainly of gastric acid and proteolytic enzymes (pepsins). The highly regulated dynamic process of gastric acid secretion is phylogenetically conserved among vertebrates and has recently been extensively reviewed (Goo et al., 2010; Schubert, 2011). The impairment of gastric acid secretion leads to hypochlorhydria (pH between 4 and 7) or even achlorhydria (pH 7) and increases the susceptibility toward infection (Martinsen et al., 2005). For example, mice deficient in the gastric H+/K+ ATPase subunit β (Scarff et al., 1999) are constitutively hypochlorhydric and show higher susceptibility to infection with different bacterial pathogens (Tennant et al., 2008). Analysis of the colonic and fecal microbiota of rats and humans with reduced gastric acid secretion showed also an increase in the abundance of several bacterial groups in the lower intestinal tract (Kanno et al., 2009). Thus, gastric acid not only prevents bacterial overgrowth in the stomach but also influences microbial composition in the intestine.
The vertebrate mucosa along the entire gastrointestinal tract from stomach to rectum consists of a single layer of columnar epithelial cells. Histologically, the human stomach can be divided into three regions: cardia, fundus/corpus, and antrum (Schubert & Peura, 2008; Fig. 1). Each region is characterized by the presence of specialized secretory cell lineages, which differentiate from multipotent gastric stem cells (Fig. 2). Fundus and corpus, which comprise about 80% of the gastric mucosa, harbor acid-secreting parietal cells, mucus neck cells (MNCs), and pepsinogen-secreting zymogenic (chief) cells. The antrum is characterized by the presence of gastrin-secreting G cells and gland cells. Mucus-producing surface mucus cells (SMCs) cover the whole gastric mucosa. The different cell types are structurally organized in glandular invaginations throughout the mucosal epithelium, which are designated as oxyntic glands in the corpus/fundus region or as pyloric glands in the antrum (Kouznetsova et al., 2011). The gastric stem cells reside in the middle part of the gland (isthmus zone) and differentiate into precursors of the different secretory cell lineages. These differ both in secretory product and in direction of migration. The SMCs (also called foveolar or pit cells) migrate apically toward the gastric lumen and populate the superficial epithelium. In contrast, the parietal and MNCs move downwards with the MNCs differentiating into chief cells as they migrate further toward the base of the oxyntic gland. A variety of neuroendocrine cells is scattered throughout the mucosa.
Schematic illustration of the human stomach.
Schematic illustration of the gastric mucosa with the main cell types of oxyntic and pyloric glands in the gastric epithelium. Gastric stem cells reside in the isthmus zone of the gland and differentiate into precursors of the different cell lineages, which migrate either apically toward the gastric lumen or downwards to the base. The superficial epithelium and the gastric glands are covered by a viscous mucus layer mainly composed of MUC5AC, secreted by the SMCs, and MUC6, secreted mainly by MNCs and antral gland cells. The mucus layer consists of an inner layer, which is firmly attached to the epithelium, and an outer loose layer. The gastric pathogen Helicobacter pylori has been shown to use the transmucus pH gradient between the acidic gastric lumen and the near-neutral epithelial surface for spatial orientation to reach its niche at the juxtamucosal epithelium. The precise location of non-H. pylori microbiota is still hypothetical.
The comparison of germ-free mice with mice colonized with either a specific pathogen-free (SPF) microbiota or an altered Schaedler flora (ASF), consisting of eight bacterial species and developed in the mid-1980s for standardized colonization of germ-free rodents (Orcutt et al., 1987; Dewhirst et al., 1999), has demonstrated that the architecture of the gastric epithelium is affected by the indigenous microbiota (Schmitz et al., 2011). In the germ-free mice, the length of the gastric glands was shortened, probably due to a decrease in the number of gastric stem cells and pepsinogen-producing zymogenic cells. Prezymogenic cells, which present an intermediate state during the differentiation of MNCs into zymogenic cells, were completely missing.
The mucosa of the entire gastrointestinal tract is covered by a viscous mucus layer (Atuma et al., 2001). The main function of this mucus layer is the separation and protection of the underlying epithelium from the luminal content. Thereby, mucus serves both as lubricant and as a physical barrier. Measurement of mucus thickness by in vivo microscopy showed that it differs both between different species and according to the gastrointestinal segment (Atuma et al., 2001; Phillipson et al., 2008). The mucus in stomach and colon is continuous and consists of two sublayers, an inner mucus layer that is firmly attached to the underlying epithelium and an outer loose layer that can easily be removed (Atuma et al., 2001; Johansson et al., 2008). How the inner layer is connected to the mucosa is presently not known. Because the composition of both sublayers is similar, it is assumed that the outer loose layer is generated from the inner firm layer, probably due to proteolytic degradation (Phillipson et al., 2008; Johansson et al., 2009). Similar findings in germ-free mice suggest that host proteases might be involved in this process (Johansson et al., 2008). In the intestine, the inner layer has been suggested to be mainly responsible for the protection of the underlying epithelium against bacterial infection (Johansson et al., 2008). Via fluorescence in situ hybridization, the outer colonic layer of mice was demonstrated to be highly populated by bacteria, whereas the inner mucus layer was virtually free of bacteria. However, in the same study, 16S rRNA gene-based semi-quantitative PCR detected bacteria also in the inner layer, even though to a lesser extent compared with the outer layer. This finding has been supported by a recent analysis of the microbiota of both mucus sublayers of rat colon, which found a 10-fold increase in bacterial abundance in the outer layer (Dicksved et al., 2012). Interestingly, the microbiota composition differed in both layers. A comparable analysis of the two gastric mucus layers has not been performed so far.
The main constituents of the mucus barrier are mucin glycoproteins (McGuckin et al., 2011). Mucins are highly O-glycosylated proteins. Their central mucin domain contains a variable number of tandem repeats rich in the amino acids proline (P), threonine (T), and serine (S) (PTS domain). The hydroxy amino acids serve as attachment sites for O-glycans, which constitute the major part of the glycoproteins. The composition of O-glycans is determined by a set of host-specific glycosyltransferases (McGuckin et al., 2011). Accordingly, O-glycosylation varies between different species but has also been shown to be highly diverse in the human population. It has been suggested that the host-specific glycosylation pattern on mucin glycoproteins in the gastrointestinal tract may specify the indigenous microbiota.
According to structural features, mucins are classified into membrane-associated and secreted mucins. Based on their ability to oligomerize, the secreted mucins can be further subdivided into gel-forming and non-gel-forming mucins. The human mucin protein family includes 17 members (10 membrane-associated and 7 secreted mucins), which differ in tissue and cell type-specific expression (Derrien et al., 2010). The main mucins in the stomach are the membrane-associated MUC1 and the secreted gel-forming MUC5AC and MUC6. MUC5AC, which is structurally highly related to the intestinal MUC2 (Lang et al., 2007), is the major structural constituent of both the inner and outer mucus sublayers (Phillipson et al., 2008). It is mainly produced by SMCs in the superficial epithelium, whereas MUC6 is expressed by mucus gland cells (Reis et al., 2000; Verzi et al., 2008; Kouznetsova et al., 2011). The membrane-associated MUC1 is also expressed by SMCs and has been demonstrated in vitro to protect the epithelium against adhesion of the pathogenic bacterium Helicobacter pylori (Lindén et al., 2009). Correspondingly, Muc1−/−-deficient mice display higher colonization levels of H. pylori (McGuckin et al., 2007; Guang et al., 2010) and more severe gastritis compared with wild-type mice (McGuckin et al., 2007).
Several studies showed an impact of bacterial infection on gastric mucin expression. Both acute and chronic H. pylori infection of mice led to a significant reduction in Muc1 expression in corpus and antrum (Navabi et al., 2013). The infection of mice with Helicobacter felis led to an increase in Muc4 and Muc5b gene expression, whereas the expression of Muc5ac remained unchanged or was even reduced (Schmitz et al., 2009, 2011).
In addition to mucins, which form the structural scaffold of the mucus barrier, mucus also contains a diverse repertoire of other proteins including antimicrobial molecules. A comprehensive proteomic analysis of the inner and outer colonic mucus layers of mice identified a variety of proteins in addition to the major structural component Muc2 (Johansson et al., 2008, 2009). However, none of the expected antimicrobials could be detected by this experimental approach. A comparably detailed analysis of the gastric mucus composition is so far not available and would be of special interest.
Antimicrobial peptides (AMPs) including defensins and cathelicidins play an important role in the mucosal defense against commensal and pathogenic microorganisms, especially in the intestine, where the bacterial load is much higher (Jäger et al., 2010). Compared with the intestine, there is much less information concerning the expression and regulation of AMPs in the gastric environment. Nevertheless, a variety of antimicrobial molecules has been identified in the stomach. A histochemical study of the rat gastrointestinal tract detected β-defensin 1 and β-defensin 2 in the gastric superficial epithelium (Yokoo et al., 2011). β-defensin 2 expression was also found in human gastric mucosa, and its expression was increased in H. pylori-infected individuals (Rogoll et al., 2011). Lysozyme, which acts on the bacterial cell wall of Gram-positive bacteria, was detected in superficial epithelial and gland cells of the antrum as well as at the base of oxyntic glands (Kouznetsova et al., 2011).
Cathelicidins comprise a group of host-defense peptides, which are highly expressed during infection, inflammation, and wound healing (Wu et al., 2010). The only human cathelicidin LL37 was found to be expressed in superficial epithelial, parietal, and chief cells in fundic glands of the normal gastric mucosa (Hase et al., 2003). Its expression was significantly increased in H. pylori-infected human gastric epithelium (Hase et al., 2003; Leszczynska et al., 2009; Rogoll et al., 2011). LL37 demonstrated an antibacterial activity against different H. pylori strains (Hase et al., 2003). The in vitro analysis of LL37 expression in cultured gastric cells showed that the increase in expression by H. pylori was dependent on a functional type 4 secretion system (T4SS) encoded by the cag pathogenicity island (cagPAI) of H. pylori. The importance of cathelicidin in the control of H. pylori infection was demonstrated using mice deficient in the murine LL37 ortholog (Zhang et al., 2013). Knockout mice displayed higher bacterial load in the stomach and increased H. pylori-associated gastritis.
The AMP hepcidin, which has been identified as a major regulator of iron homeostasis, has been suggested to link iron metabolism and host response to infection (Drakesmith & Prentice, 2012). It has been shown that inflammation and infection lead to increased hepcidin production (Wessling-Resnick, 2010). Although the liver is the major site of hepcidin production, it was also found in extrahepatic tissues including the stomach (Schwarz et al., 2012). Because the stomach plays a role in both iron absorption and bacterial defense, hepcidin might be involved in connecting these processes. Hepcidin was found to be abundantly expressed in both corpus and fundus of rats, mice, and humans (Schwarz et al., 2012). Its expression was much lower in antrum and negligible in the aglandular forestomach of mice and rats. Helicobacter pylori infection was associated with an increase in hepcidin expression both in gastric epithelial cells in vitro and in the gastric mucosa of corpus and antrum in human patients. Eradication of the human H. pylori infection reduced hepcidin levels back to normal. Opposite results were obtained in transgenic insulin–gastrin (INS–GAS) mice, which express human gastrin under control of the rat insulin I promoter (Wang et al., 1993). Infection of these mice with Helicobacter felis resulted in reduced gastric hepcidin expression level (Thomson et al., 2012). The main hepcidin expression in stomach was localized in gastric parietal cells, the site of gastric acid production (Schwarz et al., 2012). The same study showed that mice deficient in hepcidin displayed reduced expression of the H+/K+ ATPase, which was accompanied by a reduction in gastric acid secretion resulting in bacterial overgrowth in the stomach. In a Mongolian gerbil model of gastric carcinogenesis, it has been recently shown that iron depletion accelerates H. pylori-induced pathology (Noto et al., 2013). This might be of special importance because human H. pylori infection is reported to be associated with iron deficiency anemia (Tan & Goh, 2012).
Antimicrobial activity against H. pylori was also demonstrated for α1,4-N-acetylglucosamine (α1,4-GlcNAc) found as terminal residue on O-glycans of certain mucins like MUC6, which is mainly expressed in gastric glands (Kawakubo et al., 2004). α1,4-GlcNAc functioned by inhibition of the biosynthesis of cholesteryl-α-D-glucopyranoside, a major component of the cell wall of H. pylori.
One important function of the gastric mucus barrier is the protection of the mucosal epithelium against gastric acid. The pH at the epithelial surface (juxtamucosal pH) is assumed to be virtually neutral despite the highly acidic intragastric lumen. Measurements via pH-sensitive microelectrodes determined a neutral pH at the epithelial surface in the stomach of mice (Henriksnas et al., 2006), rats (Schade et al., 1994; Phillipson et al., 2002), and guinea pigs (Schreiber & Scheid, 1997).
Methods for the investigation of stomach microbiota
Sampling methods
To analyze the gastric microbial composition of small laboratory animals like mice or Mongolian gerbils, the entire stomach is generally aseptically removed (Sun et al., 2003b; Zaman et al., 2010; Lofgren et al., 2011). Afterward, the stomach is most often divided for different analyses including histopathology, RNA isolation, or determination of microbial composition. This procedure greatly minimizes the risk of cross-contamination. The stomach of rodents consists of a forestomach, which is covered by an aglandular squamous epithelium, and a glandular part, divided into antrum and corpus. Because these two compartments will probably differ in their microbial composition, it is quite important to specify the sampling location precisely, which is unfortunately not always the case.
In humans, samples of the gastric mucosa are usually obtained by biopsies taken during gastroendoscopy. Because stomach endoscopies are medical interventions that carry a small risk of complications, they are rarely performed in the absence of a medical indication. A further disadvantage of biopsy material for investigation of the human microbiota is that while both endoscopes and biopsy forceps are disinfected or sterilized before use, a contamination of the biopsy channel with throat bacteria cannot be ruled out. Given the low microbiota density in a healthy human stomach, even such low levels of instrument contamination may distort the composition of the sampled microbiota. While biopsy material might be passed through sterile medium, which should remove most bacteria introduced during passage through the biopsy channel, this problem will only be truly resolved when new sampling methods become available (for a review of recent developments, see Valdastri et al., 2012). In contrast to the sophisticated sampling techniques that have been used for experimental animal models (Schreiber et al., 2004; Bücker et al., 2012), it is currently not possible to analyze different mucus layers from human biopsy samples separately, so that the microbiota analyses always represent a mixture of bacteria in the luminal fluid, upper mucus layer (non-H. pylori microbiota), and juxtaepithelial mucus (mainly H. pylori).
In contrast to the mucosa, the gastric juice can be sampled with methods excluding contamination, although these are not often used. The relevant instruments include a bull-nosed capsule that can be guided to the stomach and then opened by applying suction via a connecting tube (Shiner, 1963; Gray & Shiner, 1967), or alternatively double or triple nasogastric tube systems, in which the outer or middle tube is closed with a membrane or cap before sampling (Rasmussen et al., 1983; Verdu et al., 1994). Other studies used conventional nasogastric tubes without closures (Milton-Thompson et al., 1982; Sharma et al., 1984). These systems can be left in the stomach for several hours and so enable repeated or continuous sampling of gastric juice.
A slightly different method was used to selectively sample the outer mucus layer (Zilberstein et al., 2007). The authors mention a ‘flexible silicone sampling probe’ to which a distal weight was added and which allowed for the sampling of mucus instead of gastric juice.
Detection and identification of microbiota
The microbiota composition in samples from the gastric mucosa can be analyzed by culture-dependent methods, and this was the only available approach before the advent of PCR-based methods for detection and identification of microorganisms in the 1980s. One obvious disadvantage of the culture-based approach is its inability to detect uncultivable microorganisms. However, one important advantage toward culture-independent methods is the selective detection of viable microorganisms. Despite this, the vast increase in our knowledge about the composition of the human microbiota would not have been possible without both PCR-based detection and deep sequencing technologies.
Culture-independent methods to investigate the microbiota usually depend on DNA-based approaches, most of which either rely on whole-genome information or focus on the 16S rRNA gene as standard phylogenetic marker (The Human Microbiome Project Consortium, 2012). An overview over a variety of methods, including more traditional DNA fingerprinting techniques such as denaturing gradient gel electrophoresis (DGGE) or terminal restriction fragment length polymorphism, can be found in a previous review (Sekirov et al., 2010). More recent methods include Sanger-sequenced 16S rRNA gene libraries (Bik et al., 2006) or microarrays targeting the 16S rRNA gene (Paliy & Agans, 2012). Because a number of high-throughput sequencing methods (454, Illumina, and Ion Torrent) now offer read lengths that enable classification of the sampled microorganisms, these have become the methods of choice (for a recent review of sequencing technologies, see Loman et al., 2012). They permit a considerably higher sequencing coverage than Sanger sequencing of 16S clone libraries (Liu et al., 2007) and a much better phylogenetic resolution than fingerprinting methods, while also allowing for the identification of unexpected or previously unknown bacteria, which could not be detected with microarray approaches.
DNA isolation
Bacteria differ widely with respect to the composition of the cell envelope, and this is reflected in different susceptibilities to protocols for lysis and DNA extraction. For all analyses that rely on DNA-based procedures, efficient and unbiased DNA isolation is critical. Available methods include enzymatic lysis with proteases, lysozyme, and mutanolysin or lysostaphin; chemical lysis with detergents or solvents; and mechanical disruption of cells. Most protocols combine several of these principles. Comparative analyses have shown that workflows including mechanical disruption by repeated beat-beating yield higher proportions of DNA from robust bacteria such as Clostridium, Veillonella, or Streptococcus than more gentle methods (Ó Cuív et al., 2011) and increase the measured diversity (Salonen et al., 2010; Maukonen et al., 2012). During the isolation and subsequent handling of DNA for stomach microbiota analysis, extra care should be taken to avoid the introduction of contaminations at these steps: as the extracted DNA will contain significantly lower fractions of bacterial DNA than many other materials, low levels of contamination might have a significant effect.
Primer selection
In all analysis, pipelines that are based on DNA amplification, including those using pyrosequencing of the 16S rRNA gene, the next critical step will be the PCR that amplifies the region to study. Because the level of sequence variability between different bacteria is not uniformly distributed over the 16S rRNA gene, the nine hypervariable regions of the gene provide different levels of taxonomic and phylogenetic informativeness. The effects of targeting different 16S rRNA gene regions are usually examined by simulating short reads based on databases of longer or full-length 16S rRNA gene sequences (Chakravorty et al., 2007; Huse et al., 2008; Nossa et al., 2010; Soergel et al., 2012). Depending on the mock dataset and the setup of the simulation, these experiments predict highly divergent levels of classifiability (Chakravorty et al., 2007; Huse et al., 2008; Nossa et al., 2010). Surprisingly, a recent study found that hypervariable regions were not necessarily more suitable for taxon identification than other sections of the 16S rRNA gene (Soergel et al., 2012), which contradicts the results of earlier simulations (Wang et al., 2007; Nossa et al., 2010). As expected, the accuracy of classification increased with the length of the sequenced region (Nossa et al., 2010; Soergel et al., 2012).
The classifiability of 16S rRNA gene stretches depends not only on the region studied, but also on the availability of closely related sequences in the databases used for classification (Soergel et al., 2012). Testing different samples of the same mock community, the Jumpstart Consortium Human Microbiome Project Data Generation Working Group (2012) found that, while the variable regions V1–V3 provided the greatest classifiability, the analysis of V3–V5 resulted in the best representation of the sequenced population. The consortium proposes that a combination of different regions might result in the most complete description of the community. By the time of this writing, this has rarely been implemented. A related approach is suggested by the results of Soergel et al. (2012), who found that the combined classification of paired-end reads in their simulation was considerably more accurate than that of single reads of the same combined length.
A further consideration in choosing the region to sequence is the specificity of the primers used. While a one-base pair mismatch in a ‘universal’ primer would usually not appreciably affect the outcome of a classical single-template endpoint PCR, the same mismatch can lead to considerable underrepresentation of the corresponding sequences in complex multispecies samples (Schloss et al., 2011; Jumpstart Consortium Human Microbiome Project Data Generation Working Group, 2012). In such situations, the number of PCR primer mismatches was shown to be significantly correlated with a reduction in PCR efficiency (Lee et al., 2012). While this can be partially counteracted by the use of degenerate primers, a recent study recommends instead to substantially reduce the annealing temperature to as low as 30 °C (Sergeant et al., 2012).
Data pre-analysis
The analysis of 16S rRNA gene datasets generated by deep sequencing is a multistep process (Fig. 3). Even within comprehensive analysis platforms such as mothur (Schloss et al., 2009), the Ribosomal Database Project (rdp) pyrosequencing pipeline (Cole et al., 2009), or qiime (Caporaso et al., 2010b), the individual analysis steps can be manually configured, and the methods to include in the workflow can be chosen individually.
Schematic overview of the workflow involved to analyze the microbiota composition by deep sequencing of 16S rRNA gene amplicons.
At the beginning of data analysis, the high-throughput data need to be preprocessed to control for low-quality sequences and for errors introduced during PCR or sequencing (Kunin et al., 2010; Jumpstart Consortium Human Microbiome Project Data Generation Working Group, 2012). The first step of these workflows is base-calling, which can be performed using proprietary programs linked to the sequencing method used. These sequences need to be trimmed for length and quality before further analysis (Sun et al., 2009; Schloss et al., 2011). Dedicated software that controls for sequencing errors based on flowgram clustering, such as PyroNoise (Quince et al., 2009, 2011) or its mothur implementation shhh.flows (Schloss et al., 2011), is designed to achieve more stringent quality control and better sequence retention. They might however introduce artifacts if the sequences clustered are not trimmed to uniform length beforehand (Gaspar & Thomas, 2013). Some authors recommend further preclustering at a later stage to remove rare variants similar to more abundant sequences in order to also control for base pair errors introduced during PCR (Huse et al., 2010; Quince et al., 2011; Schloss et al., 2011).
This quality-controlled dataset then needs to be screened for PCR chimeras, which are sequences derived from two different templates that can form in significant quantities during amplification of complex samples (Wang & Wang, 1996; Fonseca et al., 2012). These chimeras can be detected with tools such as uchime (Edgar et al., 2011) or PERSEUS (Quince et al., 2011). PERSEUS, which is also included in AMPLICONNOISE, identifies possible ‘parents’ for each sequence from all the sequences in the dataset that are at least equally abundant (Quince et al., 2011). uchime implements both a de novo algorithm similar to PERSEUS and a reference-based mode that relies on a database of high-quality chimera-free sequences sufficiently covering the phylogenetic groups in the query dataset (Edgar et al., 2011). While these methods cannot eliminate all chimeric sequences, a combination of quality control and chimera removal can reduce their rate 10-fold (Schloss et al., 2011).
After preprocessing, sequences are either classified immediately or first grouped into operative taxonomic units (OTUs) that can capture diversity patterns not well represented in the databases used for taxon identification.
Alignment methods
A widely used method to calculate OTUs relies on a multiple sequence alignment (MSA) of all the unique sequences in the dataset. The MSA is used to calculate a distance matrix, from which OTU clusters can then be inferred. Working with de novo alignments carries the disadvantage that aligning large datasets is computationally expensive, and feasible methods guaranteed to produce optimal multiple alignments do not exist (White et al., 2010; Sun et al., 2012). To obtain more reliable MSAs and to save computation time, most OTU inference pipelines implement reference-based alignments instead.
These reference-alignment-based algorithms are implemented in a variety of tools, such as pynast (Caporaso et al., 2010a), the silva Incremental Aligner (SINA; Pruesse et al., 2012), or the mothur pipeline (Schloss et al., 2009). The reference MSAs are usually precomputed and curated by dedicated long-term projects, such as silva (Pruesse et al., 2007; Quast et al., 2013) or greengenes (DeSantis et al., 2006).
For RNA genes such as 16S, other options are secondary-structure-aware de novo alignment programs like infernal (Nawrocki et al., 2009). This program is used to align large reference databases such as greengenes (McDonald et al., 2012) and the rdp database (Cole et al., 2009). infernal is also implemented in the rdp pyrosequencing pipeline (Cole et al., 2009) and has recently become fast enough to be used in microbiota analysis without the need to use a dedicated computer cluster (Nawrocki & Eddy, 2012). Reference-based MSA methods depend on the coverage of the query organisms by the reference database used: For SINA, the accuracy of the resulting alignment was shown to increase both with the average identity of the closest template sequence and with the number of sequences included in the template set (Pruesse et al., 2012).
As an alternative to MSAs, the nucleotide distance information needed for OTU computation can also be obtained from pairwise alignments. Such distances can contain considerable noise if the organisms of interest are too distantly related, but the small differences between closely related species tend to be better reflected than in MSA-based approaches. Different independent benchmark studies found that pairwise alignments were better suited for OTU computation than MSAs (Huse et al., 2010; Barriuso et al., 2011; Sun et al., 2012). However, computing all-vs-all distance matrices requires too much time and computer memory to be feasible for routine analyses, so this approach can only be used with algorithms that drastically reduce the number of necessary comparisons (Li & Godzik, 2006; Edgar, 2010; Cai & Sun, 2011).
In a study comparing different workflows, Barriuso et al. (2011) found that mothur-generated reference-guided alignments were generally faster than de novo MSA algorithms and worked well for moderately long (200–300 bp) and full-length sequences. However, mothur introduced spurious variability when aligning artificially mutated sequences. The authors attribute this to the algorithm identifying different template references for related ‘mutants’. The de novo MSA programs mafft and muscle generated more accurate alignments, especially for shorter sequences. However, even more accurate results were obtained when using esprit, which relies on pairwise alignments instead of MSAs, and the rdp pyrosequencing pipeline based on infernal alignments (Barriuso et al., 2011).
OTU computation
OTUs are groups of very similar organisms that are computationally inferred based on nucleotide sequence information. Ideally, OTUs should correspond to species or genera. In contrast to taxonomic classification, OTU computation offers a way to obtain information on both the previously described and the yet undescribed organisms present in a sample. OTUs can be obtained by different approaches.
Hierarchical clustering (HC) is a classic unsupervised learning approach that is implemented in different OTU inference programs such as mothur (Schloss et al., 2009), the rdp pyrosequencing pipeline (Cole et al., 2009), or esprit (Sun et al., 2009) and its successor esprit-tree (Cai & Sun, 2011). HC can be implemented with either average linkage (also known as average neighbor) or complete linkage (or furthest neighbor) options. Both performed well in benchmark studies (Quince et al., 2009; White et al., 2010; Sun et al., 2012), but average linkage was shown to be more robust to noise (Quince et al., 2009; Huse et al., 2010; Sun et al., 2012). While most HC algorithms require a precomputed MSA, esprit (Sun et al., 2009) and esprit-tree (Cai & Sun, 2011) work with pairwise distances.
The programs cd-hit (Li & Godzik, 2006) and uclust (Edgar, 2010) use greedy heuristic clustering instead of HC. Both tools rely on pairwise alignments and process the sequences one at a time. Basically, the first sequence is allocated to the first cluster, and each following sequence is either added to an existing cluster if it is within the similarity threshold or assigned to a new one. As this approach minimizes the number of sequence comparisons, it has significantly lower running time and memory requirements than HC. However, the initial sorting of the sequences can influence the outcome, and a recent benchmark study found the computed clusters to reflect the data structure less accurately than OTUs computed with HC methods (Sun et al., 2012).
Classification
Either OTUs or individual 16S rRNA gene sequences are usually classified based on one of the ribosomal sequence databases. These implement their own classification algorithms, although some of those can also be trained on other data, and independent methods also exist. Classifiers technically work by identifying reference sequences similar to the query and then computing both the taxon and the taxonomic level to which the query sequence can reliably be assigned.
The most widely used of these methods is arguably the rdp classifier, a naïve Bayesian classifier provided by the rdp (Wang et al., 2007). This classifier measures sequence similarity based on the occurrence of 8-nucleotide ‘words’ in the query and database sequences. While its downloadable version contains a training set based on rdp data, it can also be trained on other data such as the larger greengenes database. The rdp classifier performs best with databases containing higher sequence diversity and with the training dataset trimmed to the 16S region covered by the query (Werner et al., 2012).
Other classification programs provided by 16S sequence database projects rely on reference-based alignments to identify database sequences similar to the query. Prominent examples include the ‘search and classify stage’ of the silva database SINA tool (Pruesse et al., 2012) and the classifier provided by the greengenes project (DeSantis et al., 2006). The greengenes method also links taxonomies by different groups of expert curators and so allows for comparisons between these partially conflicting nomenclatures.
Classifiers not originally linked to specific database projects include tools based on blast-like similarity searches. rtax (Soergel et al., 2012) was originally developed for paired-end reads and uses usearch for sequence searches and determination of nucleotide identity levels. The ‘classification resources for environmental sequence tags’ (crest) tool ‘Lowest Common Ancestor Classifier’ (LCAClassifier) relies on NCBI megablast to find relevant reference sequences and provides ‘bit score’ and similarity values.
The analysis platforms qiime (Caporaso et al., 2010b) and mothur (Schloss et al., 2009) both implement a selection of classification algorithms. qiime includes blast-based classification, the rdp classifier (Wang et al., 2007), and rtax (Soergel et al., 2012). Mothur contains its own implementation of the rdp classifier algorithm as well as a ‘k-nearest neighbor algorithm’ that computes the consensus taxonomy of a defined number of most similar sequences in the database used (Schloss, 2013). Both provide several different reference databases as well.
Animal models to study gastric microbiota–host interactions
Characterization of gastric microbiota in different animals
A number of studies have characterized the gastric microbiota of diverse animal species including mice (Tan et al., 2007; Lofgren et al., 2011; Rolig et al., 2013), Mongolian gerbils (Sun et al., 2003b; Osaki et al., 2012), dogs (Garcia-Mazcorro et al., 2012), horses (Husted et al., 2010; Perkins et al., 2012), Eastern oysters (King et al., 2012), and yellow catfish (Wu et al., 2012). All studies detected a diverse repertoire of bacterial species within the stomach despite the hostile environment that was thought to prevent extensive bacterial colonization. Comparisons of the characterized microbiota between these different species and with the human situation are hampered due to apparent differences in anatomy and physiology of the stomach (reviewed in Kararli, 1995). Factors that may further complicate intraspecies comparisons include different methods for microbiota analysis and differences in the experimental setup (e.g. with respect to animal age and diets).
The low intragastric pH, which has been demonstrated to be very important for the inactivation of ingested microorganisms (Sun et al., 2003a; Friis-Hansen et al., 2006; Tennant et al., 2008), differs considerably between different species (reviewed in Kararli, 1995). Mice possess a relatively high intragastric pH (pH 3–4; Scarff et al., 1999; Icatlo et al., 2000; Aebischer et al., 2006; McConnell et al., 2008; Takahashi et al., 2011), which decreases the potential to inactivate microorganisms, thereby facilitating the gastric growth of different bacteria. In contrast, other species including Mongolian gerbils (Mollenhauer-Rektorschek et al., 2002) and guinea pigs (Schreiber & Scheid, 1997) have low intragastric pH values (pH < 2), which are similar to those measured in humans (Teyssen et al., 1995; Vakevainen et al., 2000).
Mouse models
The analysis of the gastric microbiota of mice using a variety of different methods showed a predominance of Lactobacillus spp. (Aebischer et al., 2006; Tan et al., 2007; Table 1). Variable results in the overall bacterial composition and the presence or absence of specific bacteria are probably mainly caused by both methodological differences and the specific experimental design. Concerning the latter, both the experimental diet (Sahasakul et al., 2012) as well as the age and the genetic background of the mice most likely have an impact on the indigenous gastric microbiota. For example, a study analyzing the effect of purified and nonpurified diets detected a higher abundance of Lactobacillus spp. in mice fed the nonpurified diet (Sahasakul et al., 2012). A comparison of identical mouse strains obtained from different vendors also detected differences in the relative abundance of two Lactobacillus strains (Rolig et al., 2013). The observed difference of the inflammatory responses to H. pylori infection in these mice was hypothesized to be the result of differences in the composition of the indigenous gastric microbiota.
Indigenous gastric microbiota of mice and Mongolian gerbils
| Phylum | Genus or species | Method | Reference | |
| Mice | ||||
| Actinobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011)* Rolig et al. (2013)* | ||
| Bacteroidetes | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Firmicutes | Lactobacillus spp. | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | |
| 16S rRNA gene-based qPCR | Aebischer et al. (2006) | |||
| Culture; 16S rRNA gene-based qPCR | Sahasakul et al. (2012) | |||
| Culture | Takahashi et al. (2011) | |||
| L. acidophilus | Lactobacillus species-specific qPCR | Takahashi et al. (2011) | ||
| L. gasseri | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. intestinalis | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. johnsonii | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. murinus | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. reuteri | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | |||
| Proteobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Verrucomicrobia | PhyloChip analysis | Rolig et al. (2013) | ||
| Mongolian gerbils | ||||
| Actinobacteria | Atopobium cluster | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Bifidobacterium spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Yin et al. (2011) | |||
| Bacteroidetes | Bacteroides spp. | Culture | Yin et al. (2011) | |
| Firmicutes | Clostridium coccoides group | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Clostridium leptum subgroup | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Enterococcus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Lactobacillus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Sun et al. (2003b) | |||
| 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | |||
| Culture | Yin et al. (2011) | |||
| L. gasseri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. amylovorus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. reuteri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. murinus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| Proteobacteria | Escherichia coli | Culture | Zaman et al. (2010) | |
| Kluyvera spp. | Culture | Zaman et al. (2010) | ||
| Phylum | Genus or species | Method | Reference | |
| Mice | ||||
| Actinobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011)* Rolig et al. (2013)* | ||
| Bacteroidetes | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Firmicutes | Lactobacillus spp. | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | |
| 16S rRNA gene-based qPCR | Aebischer et al. (2006) | |||
| Culture; 16S rRNA gene-based qPCR | Sahasakul et al. (2012) | |||
| Culture | Takahashi et al. (2011) | |||
| L. acidophilus | Lactobacillus species-specific qPCR | Takahashi et al. (2011) | ||
| L. gasseri | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. intestinalis | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. johnsonii | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. murinus | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. reuteri | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | |||
| Proteobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Verrucomicrobia | PhyloChip analysis | Rolig et al. (2013) | ||
| Mongolian gerbils | ||||
| Actinobacteria | Atopobium cluster | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Bifidobacterium spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Yin et al. (2011) | |||
| Bacteroidetes | Bacteroides spp. | Culture | Yin et al. (2011) | |
| Firmicutes | Clostridium coccoides group | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Clostridium leptum subgroup | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Enterococcus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Lactobacillus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Sun et al. (2003b) | |||
| 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | |||
| Culture | Yin et al. (2011) | |||
| L. gasseri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. amylovorus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. reuteri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. murinus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| Proteobacteria | Escherichia coli | Culture | Zaman et al. (2010) | |
| Kluyvera spp. | Culture | Zaman et al. (2010) | ||
These studies did not provide explicit genus- or species-level information.
Indigenous gastric microbiota of mice and Mongolian gerbils
| Phylum | Genus or species | Method | Reference | |
| Mice | ||||
| Actinobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011)* Rolig et al. (2013)* | ||
| Bacteroidetes | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Firmicutes | Lactobacillus spp. | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | |
| 16S rRNA gene-based qPCR | Aebischer et al. (2006) | |||
| Culture; 16S rRNA gene-based qPCR | Sahasakul et al. (2012) | |||
| Culture | Takahashi et al. (2011) | |||
| L. acidophilus | Lactobacillus species-specific qPCR | Takahashi et al. (2011) | ||
| L. gasseri | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. intestinalis | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. johnsonii | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. murinus | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. reuteri | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | |||
| Proteobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Verrucomicrobia | PhyloChip analysis | Rolig et al. (2013) | ||
| Mongolian gerbils | ||||
| Actinobacteria | Atopobium cluster | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Bifidobacterium spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Yin et al. (2011) | |||
| Bacteroidetes | Bacteroides spp. | Culture | Yin et al. (2011) | |
| Firmicutes | Clostridium coccoides group | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Clostridium leptum subgroup | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Enterococcus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Lactobacillus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Sun et al. (2003b) | |||
| 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | |||
| Culture | Yin et al. (2011) | |||
| L. gasseri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. amylovorus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. reuteri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. murinus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| Proteobacteria | Escherichia coli | Culture | Zaman et al. (2010) | |
| Kluyvera spp. | Culture | Zaman et al. (2010) | ||
| Phylum | Genus or species | Method | Reference | |
| Mice | ||||
| Actinobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011)* Rolig et al. (2013)* | ||
| Bacteroidetes | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Firmicutes | Lactobacillus spp. | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | |
| 16S rRNA gene-based qPCR | Aebischer et al. (2006) | |||
| Culture; 16S rRNA gene-based qPCR | Sahasakul et al. (2012) | |||
| Culture | Takahashi et al. (2011) | |||
| L. acidophilus | Lactobacillus species-specific qPCR | Takahashi et al. (2011) | ||
| L. gasseri | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. intestinalis | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. johnsonii | 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | ||
| Lactobacillus species-specific qPCR | Takahashi et al. (2011) | |||
| L. murinus | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| L. reuteri | Culture; 16S rRNA gene clone library analysis | Tan et al. (2007) | ||
| 16S rRNA gene-based PCR-DGGE and sequencing | Sahasakul et al. (2012) | |||
| Proteobacteria | 454 pyrosequencing PhyloChip analysis | Lofgren et al. (2011) Rolig et al. (2013) | ||
| Verrucomicrobia | PhyloChip analysis | Rolig et al. (2013) | ||
| Mongolian gerbils | ||||
| Actinobacteria | Atopobium cluster | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Bifidobacterium spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Yin et al. (2011) | |||
| Bacteroidetes | Bacteroides spp. | Culture | Yin et al. (2011) | |
| Firmicutes | Clostridium coccoides group | 16S rRNA gene-based qPCR | Osaki et al. (2012) | |
| Clostridium leptum subgroup | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Enterococcus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Lactobacillus spp. | 16S rRNA gene-based qPCR | Osaki et al. (2012) | ||
| Culture | Sun et al. (2003b) | |||
| 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | |||
| Culture | Yin et al. (2011) | |||
| L. gasseri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. amylovorus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. reuteri | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| L. murinus | 16S rRNA gene pyrosequencing (clone library) | Sun et al. (2003b) | ||
| Proteobacteria | Escherichia coli | Culture | Zaman et al. (2010) | |
| Kluyvera spp. | Culture | Zaman et al. (2010) | ||
These studies did not provide explicit genus- or species-level information.
The analysis of the gastric microbiota of SPF BALB/c mice depending on H. pylori infection using 16S rRNA gene clone libraries found a predominance of Lactobacillus spp. in uninfected control mice (Aebischer et al., 2006). Helicobacter pylori infection resulted in decreased abundance of Lactobacillus spp. and increased bacterial diversity due to the presence of bacteria that are typically found in the lower intestinal tract. The detection of these bacteria was not the result of lower gastric acidity, because there was no change in intragastric pH due to H. pylori infection. However, the microbiota composition was analyzed for only one mouse per experimental group, and the observed impact of H. pylori may only reflect natural variation of the gastric microbiota.
The analysis of the gastric microbiota of naïve SPF C57BL/6 mice by bacterial culture and 16S rRNA gene clone libraries also detected predominantly Lactobacillus spp. (L. reuteri and L. murinus) in the murine stomach (Tan et al., 2007). Helicobacter pylori infection did not result in a change of the gastric microbiota of these mice. The differing observations compared with Aebischer et al. (2006) can possibly be explained by the different experimental conditions, such as different mouse strains, H. pylori strains, infection times, and methods for determination of the gastric microbiota.
A more comprehensive study of the gastrointestinal microbiota of transgenic INS–GAS mice by means of 454 pyrosequencing of 16S rRNA gene amplicons demonstrated a profound effect of long-term H. pylori infection on the overall composition of the stomach microbiota (Lofgren et al., 2011). INS–GAS mice expressing human gastrin present a good model for the analysis of the development of gastric adenocarcinoma (Wang et al., 2000; Lofgren et al., 2011). Helicobacter pylori-infected mice displayed a significantly elevated relative abundance of Firmicutes and a reduction in Bacteroidetes (Lofgren et al., 2011).
A very recent study analyzed the influence of H. pylori infection on the gastric microbiota composition of female SPF C57BL/6N mice using PhyloChip arrays (Rolig et al., 2013). Applying a very narrow OTU cutoff of 0.5% sequence divergence, the authors detected about 2000 OTUs in all five noninfected control mice. These OTUs were predominantly classified as members of the Firmicutes, Bacteroidetes, and Verrucomicrobia phyla. Helicobacter pylori infection for 4 weeks did not induce a significant change in the overall composition of the gastric microbiota in these mice. Overall, the relationship between H. pylori infection and both gastric and intestinal microbiota composition still remains incompletely understood, warranting further studies.
The C57BL/6 mouse model of H. felis infection has also been used as a model for human H. pylori infection. The infection of C57BL/6 mice, colonized with either SPF microbiota or the ASF, with H. felis has emphasized the importance of the indigenous or newly acquired gastric microbiota in the clearance of this bacterial pathogen (Schmitz et al., 2011). In contrast to ASF mice, most of the SPF mice were able to clear gastric H. felis infection. The decrease in H. felis colonization in the SPF mice was accompanied by an increase in the gastric microbial diversity due to acquisition of new bacterial species including multiple Lactobacillus spp. strains.
The same model was used to investigate the effect of a eukaryotic intestinal parasite on Helicobacter-associated gastritis and gastric atrophy (Fox et al., 2000). Co-infection with the nematode Heligmosomoides polygyrus led to reduced gastric pathology, but also to a failure to clear H. felis colonization. The associated cytokine and chemokine profiles indicated that the nematode infection induced a response dominated by anti-inflammatory type 2 T-helper cells (Th2), which reduced the pro-inflammatory type 1 T-helper cell (Th1) response induced by H. felis (Fox et al., 2000). Similar effects, which were linked to different immunological processes, were identified in C57BL/6 mice co-infected with H. pylori and one of the enterohepatic Helicobacter species H. bilis, H. hepaticus, or H. muridarum (Lemke et al., 2009; Ge et al., 2011). Colonization with any of these species prior to infection with H. pylori led to a reduced Th1 response that was apparently mediated by regulatory T cells sensitized by antigens shared between the enterohepatic Helicobacter sp. and H. pylori. In case of co-infection with either H. bilis or H. muridarum, this led to a significant reduction in H. pylori-induced gastric pathology. Conversely, co-infection with H. hepaticus enhanced the pro-inflammatory Th17 response, resulting in increased H. pylori-induced gastritis (Ge et al., 2011). These important studies in mice raise the possibility that infection or colonization of nongastric parts of the gastrointestinal tract may affect the gastric microbiota and the immune reaction against microbiota components.
Despite important differences in stomach anatomy and physiology, the mouse model can be extremely useful for microbiome analyses because of the availability of genetically modified mouse strains. Several knockout mice exist that permit the elucidation of the impact of mucins or glycosyltransferases on gastric colonization of bacteria. For example, Muc1−/− mice have been used to show the importance of this surface-associated mucin in the protection of the gastric epithelium against H. pylori infection (McGuckin et al., 2007; Guang et al., 2010). Similarly, the use of Muc2−/− mice, which lack the major intestinal mucin Muc2, demonstrated the importance of mucins in protection of the intestine against bacterial infection (Bergstrom et al., 2010).
Mongolian gerbils
Mongolian gerbils have been widely used to study H. pylori-induced pathophysiology because of the similarities compared with human infection. However, comprehensive studies concerning the characterization of the gastric mucosal microbiota are rare. Most of these studies aimed to elucidate the potential influence of H. pylori infection on the indigenous gastric microbiota (Sun et al., 2003b; Zaman et al., 2010; Yin et al., 2011; Osaki et al., 2012).
These studies showed that bacteria other than H. pylori are able to colonize the gastric mucosa of Mongolian gerbils despite low pH values, which are comparable to the human intragastric pH (Table 1). Overall, the gastric microbiota of Mongolian gerbils is dominated by the genus Lactobacillus spp. similar to the findings in mice (Sun et al., 2003b; Yin et al., 2011; Osaki et al., 2012).
The impact of long-term H. pylori infection on the microbiota of gastric mucus of Mongolian gerbils was analyzed using 16S rRNA gene-based qPCR (Osaki et al., 2012). The most prevalent bacterial genera comprised the facultative anaerobes Enterococcus spp. and Lactobacillus spp. as well as the obligate anaerobe Atopobium cluster. Successful infection with H. pylori led to an increase in the abundance of the Clostridium coccoides group. Interestingly, bacteria of the Eubacterium cylindroides group and Prevotella spp. were only detected in animals, which were originally infected with H. pylori but lost the infection after 1 year. These bacteria were detected neither from uninfected controls nor from successfully infected animals.
The microbiota of the human stomach
Helicobacter pylori and related gastric Helicobacter species
The major constituent of the gastric microbiota in more than half of all humans is Helicobacter pylori, a spiral-shaped member of the Gram-negative Epsilonproteobacteria (Guisset et al., 1997; Frenck & Clemens, 2003). Infection with H. pylori profoundly affects gastric physiology and thus the properties of the gastric mucosa as an ecological niche for other bacteria (McColl et al., 2000). Helicobacter pylori infection is usually acquired during early childhood. While transient infections, spontaneous clearance, and clearance after antibiotic treatments not related to the H. pylori infection have been reported in children (Tindberg et al., 1999; Broussard et al., 2009; Duque et al., 2012), stably established infections last almost always lifelong unless treated. In old age, mucosal atrophy can also lead to a loss of H. pylori infection. Intrafamilial transmission plays a dominant role in regions with better hygiene conditions, whereas extrafamilial horizontal transmission is more common in rural areas of developing countries (Magalhaes Queiroz & Luzza, 2006; Schwarz et al., 2008). While most infections result in asymptomatic chronic gastritis, chronic infection can further progress toward symptomatic disease, including peptic ulcer, gastric adenocarcinoma, and mucosa-associated lymphoid tissue lymphoma (Suerbaum & Michetti, 2002). Helicobacter pylori has been classified as a class I (definitive) carcinogen by the World Health Organization (WHO; IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 1994). Treatment aims at a complete and lasting eradication of H. pylori infection and requires the administration of combination regimens, usually a 7-day triple therapy with two antibiotics and a proton-pump inhibitor (PPI; Suerbaum & Michetti, 2002; Malfertheiner et al., 2012). Despite intense efforts, so far there has been no success in developing a vaccine against H. pylori infection (Czinn & Blanchard, 2011).
The bacterial factors involved in the ability of H. pylori to chronically colonize the gastric niche have been intensively studied. The adaptation of H. pylori to the human stomach has become a paradigm of host–microorganism coevolution (Suerbaum & Josenhans, 2007), and lessons learned from this microorganism–host interaction may be important for future studies on mechanisms that permit the establishment of the microbiota in other ecological niches.
The natural habitat of H. pylori is the mucus layer of the stomach (Schreiber et al., 2004; Howitt et al., 2011; Bücker et al., 2012). Helicobacter pylori can also colonize areas of the duodenum where the intestinal epithelium has been replaced by gastric type epithelium (gastric metaplasia; reviewed in Dixon, 2001). The majority of H. pylori cells swim within the mucus layer, propelled by a bundle of rotating flagella (Josenhans & Suerbaum, 2002). Flagellar motility has been shown to be essential for the ability of Helicobacter spp. to colonize the stomach in several different animal models (Eaton et al., 1996; Andrutis et al., 1997; Josenhans et al., 1999).
Analysis of the precise localization of H. pylori within the gastric mucus of Mongolian gerbils detected the majority of the bacteria swimming in close proximity to the gastric mucosa (Schreiber et al., 2004). The process of the initial colonization of the mucus by H. pylori is poorly understood. To swim from the acidic stomach lumen to the mucosal epithelium, H. pylori has to pass through the mucus layer and uses a transmucus pH gradient for spatial orientation (Schreiber et al., 2004). Recent data from a Mongolian gerbil model indicate that postprandial conditions in the stomach critically influence the ability of H. pylori to migrate from the lumen to the deep mucus layer, a prerequisite for infection (Bücker et al., 2012). The influence of age-related changes in gastric physiology on initial colonization of H. pylori was analyzed by simulation of different intragastric pH profiles in the stomach of anesthetized Mongolian gerbils. The postprandial condition found in children and simulated by moderate re-acidification and concomitantly low pepsin activity was found to promote bacterial colonization in the juxtamucosal mucus layer. In contrast, faster re-acidification and higher pepsin activity simulating the gastric physiology of adults significantly diminished colonization of H. pylori. These findings provide a plausible explanation why young children have an elevated risk of H. pylori infection, while adults rarely acquire H. pylori if they have remained H. pylori-free through their childhood (or have been treated).
Bacterial orientation along chemical gradients requires an intact chemotaxis system controlling flagellar motility. Helicobacter pylori mutants completely defective in chemotaxis infected mice at significantly lower levels (Terry et al., 2005). Analyses of the four putative chemoreceptors of H. pylori showed a role for the cytoplasmic chemoreceptor TlpD, which has been implicated in energy taxis (Schweinitzer et al., 2008), in the initial colonization of mice (Rolig et al., 2012). Although H. pylori mutants deficient in TlpD were still able to colonize, the bacterial load in both corpus and antrum and in the interjacent zone was significantly diminished. Using three-dimensional confocal microscopy, it was shown that H. pylori resides deeply within the antral glands of mice (Howitt et al., 2011). This specific localization was completely depending on chemotaxis because H. pylori mutants lacking the chemotaxis regulator ChePep were no longer able to colonize the glands and were entirely displaced by wild-type H. pylori in competition experiments.
The persistent colonization of the gastric mucus layer after oral uptake of H. pylori relies on multiple bacterial factors. In addition to the essential role of flagellar motility/chemotaxis, one further important prerequisite for colonization is the urease activity of H. pylori. The cytoplasmic urease produces ammonia through hydrolysis of urea, thereby increasing the pH. The activity of urease is coupled to the uptake of urea from the environment into the cytoplasm via the pH-dependent urea channel encoded by the ureI gene. Both urease and UreI have been shown to be essential for in vitro survival at pH < 4 as well as for in vivo colonization (Tsuda et al., 1994; Skouloubris et al., 1998; Mollenhauer-Rektorschek et al., 2002).
In addition to the swimming (‘planktonic’) population of H. pylori, part of the bacterial cells adhere to the gastric epithelial cells by means of adhesins, specialized host interaction molecules on the bacterial surface. The best-studied adhesin of H. pylori is the blood group antigen-binding adhesin (BabA). BabA belongs to the Hop family of outer membrane proteins (Alm et al., 2000) and binds with high affinity to terminal fucosylated residues present on the ABO blood group antigens, especially the Lewis B (Leb) antigen of the O blood group (Boren et al., 1993; Ilver et al., 1998). These antigens are abundantly expressed in the gastric epithelium and in the mucus layer. Not all H. pylori strains express BabA, and not all BabA-expressing strains are able to bind Leb or related structures. Despite conflicting results, most data indicate an association between BabA expression and severity of H. pylori-induced gastric disease (Ohno et al., 2011). The exact role of BabA in H. pylori-induced pathogenesis is, however, not completely understood. In vitro, the BabA-Leb-mediated interaction resulted in enhanced transcription of both proinflammatory cytokines and precancer-related factors in host cells (Ishijima et al., 2011). This was dependent upon a functional cagPAI-encoded T4SS. Experimental H. pylori infection of different animal models like mice, Mongolian gerbils, and rhesus macaques showed a loss of Leb-binding ability in the course of chronic infection (Solnick et al., 2004; Styer et al., 2010; Ohno et al., 2011). Based on these findings, it has been suggested that the dynamic modulation of BabA expression in vivo might facilitate adaptation to the changing conditions in the gastric environment (Moore et al., 2011). Another member of the Hop protein family, SabA, mediates binding to sialyl Lewis X (Mahdavi et al., 2002), and several other members of the Hop protein family have also been implicated in adhesion. While mechanisms that permit H. pylori to colonize the gastric mucus are important in the context of this review, space does not permit a thorough discussion of H. pylori pathogenesis, which has been reviewed in several recent synopses (Polk & Peek, 2010; Gilbreath et al., 2011; Müller & Solnick, 2011).
Analyses of the nucleotide sequence diversity within global collections of H. pylori strains have permitted a detailed reconstruction of the evolution of this organism and its association with humans (Falush et al., 2003; Linz et al., 2007; Moodley et al., 2009, 2012). The data suggest that H. pylori was first acquired by early humans in Africa, at least 100 000 years ago, most likely through a host jump from an unknown animal source. Since that time, the species has split into two superlineages whose descendants still exist today. One lineage has given rise to the population hpAfrica2, which lacks the cagPAI and is only found in southern parts of Africa. The second lineage carries the cagPAI (although individual isolates of this lineage may lack the PAI) and has spread to all continents after the ‘out of Africa’ migrations. Due to the association of H. pylori with the human stomach for a major part of its evolutionary history and the lack of overt disease in the majority of individuals, it has been argued that H. pylori may have been a universal, possibly even beneficial component of the human microbiota (Blaser & Falkow, 2009), whose transmission has only recently been disrupted so that its prevalence has started to decline. This discussion has led to significant confusion and may have contributed to a lack of resolve by public health agencies to actively reduce the prevalence of H. pylori infection with the aim to prevent H. pylori-associated diseases, notably cancer. It thus needs to be re-emphasized that there is overwhelming evidence for a causal relationship between H. pylori infection and gastric cancer, one of the most common causes of cancer death, with at least 590 000 new cases attributable to H. pylori infection per year (Parkin, 2006), warranting further studies of the benefits of H. pylori eradication on gastric cancer prevalence.
In rare cases, the human stomach can also be colonized by Helicobacter species other than H. pylori. The term ‘Helicobacter heilmannii sensu lato’ has been proposed to describe large, long, and frequently nonculturable spiral-shaped bacteria occurring in human gastric mucosa (Haesebrouck et al., 2011). This group of organisms includes Helicobacter bizzozeronii, Helicobacter suis, Helicobacter felis, and Helicobacter heilmannii sensu stricto. Only one of these species (H. bizzozeronii) has so far been cultured from a human stomach (Andersen et al., 1996; Kivistö et al., 2010). Available evidence suggests that these infections are acquired zoonotically. An enterohepatic Helicobacter species, H. cinaedi, has also been reported to occasionally colonize the human stomach (Peña et al., 2002; Han et al., 2010). This species is otherwise known for its potential to cause serious complications such as recurrent bacteremia (Uçkay et al., 2006; Oyama et al., 2012), but can also be found in the stool of healthy carriers (Oyama et al., 2012). Overall, the pathological relevance of the colonization with non-H. pylori Helicobacter sp. remains unclear.
Non-H. pylori microbiota in the human stomach
Before the development of high-throughput sequencing methods, the microbiota of healthy human stomachs was rarely studied. The stomach was considered to be essentially sterile except for recently swallowed mouth and throat bacteria given the combined action of acid and pepsin (Anderson & Langford, 1958; Gray & Shiner, 1967). This view has changed over the last three decades, in particular because a number of more recent culture-independent studies reported a more diverse microbiota than previously assumed (Bik et al., 2006; Andersson et al., 2008; Li et al., 2009; Stearns et al., 2011; Hu et al., 2012; Delgado et al., 2013). While the gastric luminal fluid (gastric juice) can be collected more easily, samples from the human gastric mucosa can only be obtained with relatively invasive procedures (endoscopy), so that the mucosa-associated microbiota has mostly been studied in people for whom stomach endoscopy was indicated for medical reasons (Bik et al., 2006; Kato et al., 2006; Dicksved et al., 2009; Stearns et al., 2011; Delgado et al., 2013). This leads to a situation where most of the ‘healthy’ control subjects in these studies suffer from underlying gastric pathology or at least suspicion of gastric illness.
Studies from the pre-H. pylori era
When interpreting culture-based studies published before approximately 1984, it has to be noted that methods to culture H. pylori were unavailable, and stratification of results according to the presence or absence of H. pylori is not possible. The microbiota of the healthy stomach lumen was assayed as the control condition in studies addressing side effects of acid-reducing drugs (Milton-Thompson et al., 1982; Sharma et al., 1984). Even in fasting subjects with uncompromised intragastric pH, low numbers of luminal bacteria and in some cases also Candida spp. were detectable by culture (Table 2, Table S1, Supporting information). In a study investigating the effects of a 2-week treatment with a PPI, bacterial numbers increased significantly in conjunction with higher luminal pH (Sharma et al., 1984). In another study that monitored gastric juice over 24-h periods in nonfasting subjects, considerable fluctuations in both stomach pH and bacterial counts were detected: During daytime, both parameters increased substantially at mealtimes and then decreased over the course of several hours (Milton-Thompson et al., 1982). At night, short periods of higher pH values were also associated with increased bacterial counts. While the authors offer no explicit interpretation of the total bacterial numbers, they state that ‘there was no evidence of bacterial colonization’ (Milton-Thompson et al., 1982). Overall, these data supported the concept of a transient luminal microbiota that is basically composed of swallowed organisms and which in turn can seed populations of new species in the intestine (Stearns et al., 2011).
Bacteria repeatedly found in healthy human stomachs
| Phylum | Genus | Species | Sample material | Method | References |
| Firmicutes | Lactobacillus | L. antri, L. gastricus, L. kalixensis, L. reuteri, L. ultunensis | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, FISH | Sharma et al. (1984), Valeur et al. (2004), Roos et al. (2005), Zilberstein et al. (2007) and Stearns et al. (2011) |
| Streptococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007), Andersson et al. (2008) and Stearns et al. (2011) | ||
| Veillonella | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Staphylococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Enterococcus | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Bacillus | B. thermoamylovorans | Gastric juice, biopsy | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984) and Stearns et al. (2011) | |
| Actinobacteria | Propionibacterium | P. acnes, P. granulosum | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, identification using ARDRA with 16S rRNA gene Sanger sequencing | Zilberstein et al. (2007), Delgado et al. (2011) and Stearns et al. (2011) |
| Corynebacterium | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Actinomyces | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Bacteroidetes | Bacteroides | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | |
| Porphyromonas | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Flavobacterium | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) |
| Phylum | Genus | Species | Sample material | Method | References |
| Firmicutes | Lactobacillus | L. antri, L. gastricus, L. kalixensis, L. reuteri, L. ultunensis | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, FISH | Sharma et al. (1984), Valeur et al. (2004), Roos et al. (2005), Zilberstein et al. (2007) and Stearns et al. (2011) |
| Streptococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007), Andersson et al. (2008) and Stearns et al. (2011) | ||
| Veillonella | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Staphylococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Enterococcus | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Bacillus | B. thermoamylovorans | Gastric juice, biopsy | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984) and Stearns et al. (2011) | |
| Actinobacteria | Propionibacterium | P. acnes, P. granulosum | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, identification using ARDRA with 16S rRNA gene Sanger sequencing | Zilberstein et al. (2007), Delgado et al. (2011) and Stearns et al. (2011) |
| Corynebacterium | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Actinomyces | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Bacteroidetes | Bacteroides | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | |
| Porphyromonas | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Flavobacterium | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) |
NG sequencing, next-generation sequencing; FISH, fluorescence in situ hybridization; ARDRA, amplified ribosomal DNA restriction analysis.
Bacteria repeatedly found in healthy human stomachs
| Phylum | Genus | Species | Sample material | Method | References |
| Firmicutes | Lactobacillus | L. antri, L. gastricus, L. kalixensis, L. reuteri, L. ultunensis | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, FISH | Sharma et al. (1984), Valeur et al. (2004), Roos et al. (2005), Zilberstein et al. (2007) and Stearns et al. (2011) |
| Streptococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007), Andersson et al. (2008) and Stearns et al. (2011) | ||
| Veillonella | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Staphylococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Enterococcus | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Bacillus | B. thermoamylovorans | Gastric juice, biopsy | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984) and Stearns et al. (2011) | |
| Actinobacteria | Propionibacterium | P. acnes, P. granulosum | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, identification using ARDRA with 16S rRNA gene Sanger sequencing | Zilberstein et al. (2007), Delgado et al. (2011) and Stearns et al. (2011) |
| Corynebacterium | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Actinomyces | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Bacteroidetes | Bacteroides | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | |
| Porphyromonas | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Flavobacterium | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) |
| Phylum | Genus | Species | Sample material | Method | References |
| Firmicutes | Lactobacillus | L. antri, L. gastricus, L. kalixensis, L. reuteri, L. ultunensis | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, FISH | Sharma et al. (1984), Valeur et al. (2004), Roos et al. (2005), Zilberstein et al. (2007) and Stearns et al. (2011) |
| Streptococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007), Andersson et al. (2008) and Stearns et al. (2011) | ||
| Veillonella | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Staphylococcus | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Enterococcus | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Bacillus | B. thermoamylovorans | Gastric juice, biopsy | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984) and Stearns et al. (2011) | |
| Actinobacteria | Propionibacterium | P. acnes, P. granulosum | Biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene, identification using ARDRA with 16S rRNA gene Sanger sequencing | Zilberstein et al. (2007), Delgado et al. (2011) and Stearns et al. (2011) |
| Corynebacterium | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | ||
| Actinomyces | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Bacteroidetes | Bacteroides | Gastric juice, biopsy, outer mucus layer | Culture, NG sequencing of 16S rRNA gene | Sharma et al. (1984), Zilberstein et al. (2007) and Stearns et al. (2011) | |
| Porphyromonas | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) | ||
| Flavobacterium | Biopsy | NG sequencing of 16S rRNA gene | Andersson et al. (2008) and Stearns et al. (2011) |
NG sequencing, next-generation sequencing; FISH, fluorescence in situ hybridization; ARDRA, amplified ribosomal DNA restriction analysis.
Studies in H. pylori-negative individuals
Three more recent studies of the microbiota associated with healthy (and H. pylori-free) human stomach mucosa are available. In an analysis of the microbiota along the digestive tracts of volunteers, the outer mucus layer of 20 healthy study participants was investigated (Zilberstein et al., 2007). Using methods designed to minimize contamination, they could isolate 18 different bacterial taxa by culture, which were identified to genus or species level. The most prevalent of these were Lactobacillus sp., Veillonella sp., Clostridium sp., and Corynebacterium sp. (Table S1).
The mucosa-associated microbiota of healthy stomachs was also investigated in a culture-independent study based on short 16S rRNA gene sequence reads (mean, 73 bp) obtained by an early version of 454 pyrosequencing technology (Andersson et al., 2008). Three of the six 61- to 76-year-old study subjects were identified as H. pylori-negative by culture methods. The composition of their stomach microbiota was different from individual to individual, and of the 262 phylotypes found, only 33 were discovered in all three samples. The authors suggest that the most abundant of the phylotypes found were apparently swallowed organisms originating from the mouth or esophagus. Nevertheless, they identified 177 phylotypes that were found in H. pylori-free stomachs but not in the throat samples of the same study, although those originated from other study participants. While the most abundant phyla in the H. pylori-free stomachs were Actinobacteria, Firmicutes, and Bacteroidetes, the majority of the phylotypes that were not also found in throat samples were identified as Proteobacteria.
A further study included five ‘normal patients’ as controls for an investigation of the bacterial microbiota in ‘non-H. pylori and non-nonsteroidal anti-inflammatory drug (NSAID)’ (NHNN) gastritis (Li et al., 2009). These control individuals underwent routine gastroendoscopy, but were diagnosed as endoscopically normal and H. pylori-negative. Microbiota composition was assessed based on 16S rRNA gene-libraries from both antral and corpus biopsy material. The authors identified 67 and 55 phylotypes from antrum and corpus samples, respectively. The dominant genera were Streptococcus, Prevotella, Neisseria, Haemophilus, and Porphyromonas, which together accounted for 70.5% of all library clones. Of these, the two most dominant genera Streptococcus and Prevotella represented 40.6% of all sequences. While there was little difference in the taxon composition between antrum and corpus biopsies, the genus Prevotella was significantly more abundant in the antrum than in the corpus. The stomachs of the NHNN gastritis patients contained reduced numbers of sequences identified as Neisseria, ‘Haemophilus/Actinobacillus’, and Porphyromonas in either antrum or corpus biopsies, and the genus Prevotella was less abundant in the antrum but more abundant in the corpus when compared to controls. Streptococcus was significantly more abundant in gastritis patients, which was also confirmed by qPCR. The authors further determined that Streptococcus was indeed culturable from the biopsies and that 90% of its cells remained attached to the mucosa when the biopsies were washed repeatedly, demonstrating that Streptococcus appears to be a genuine part of the microbiota of the gastric mucosa.
Several other studies included an analysis of the gastric mucosa-associated microbiota from H. pylori-negative control patients undergoing medically indicated gastroscopy (Bik et al., 2006; Dicksved et al., 2009; Stearns et al., 2011; Delgado et al., 2013). While the projects by Andersson et al. (2008) and Stearns et al. (2011) explored the differences of microbial communities along the digestive tract, Bik et al. (2006) concentrated on characterizing the stomach microbiota with and without the presence of H. pylori, and Dicksved et al. (2009) compared the stomach bacteria in cancer patients and elderly controls. Delgado et al. (2013) examined the mucosa and gastric juice of dyspeptic patients without detectable gastric pathology, using culture methods for both types of samples and additional high-throughput 16S rRNA gene sequencing for the biopsy specimens. All these studies agree that a characteristic microbiota is present in the stomach (Bik et al., 2006; Stearns et al., 2011; Delgado et al., 2013; Table 2). It is however less diverse than the bacterial communities in mouth, colon, and stool (Stearns et al., 2011). As in Andersson et al. (2008), the most prominent phyla of stomach bacteria are reported to be Firmicutes, Actinobacteria, Bacteroidetes, Proteobacteria, and Fusobacteria (Bik et al., 2006; Stearns et al., 2011); in some cases, members of the candidate division SR1 are also abundant (Stearns et al., 2011). The most prominent genera according to both culture and sequencing methods were Propionibacterium, Lactobacillus, Streptococcus, and Staphylococcus (Delgado et al., 2013). While the proportions of the bacterial taxa varied between individuals, a core set of 19 genera could be identified from all four biopsies tested. Comparing the microbiota that they detected in gastric mucosa to bacterial communities from other locations along the digestive tract that were obtained from other studies, Delgado et al. (2013) found the stomach samples to cluster together in PCoA analysis. In spite of the high variability between study subjects, stomach microbiota were distinguishable from oral and throat microbiota and even more dissimilar from nose and distal gut microbiota (Delgado et al., 2013).
Low-level detection of H. pylori in non-culture-based studies
Sequence-based methods frequently detect H. pylori in samples from study subjects that were originally identified as H. pylori-negative based on conventional methods such as rapid urease test, serum antibody tests, or PCR. In one such study, the clone libraries of 7 of 11 subjects who were initially tested H. pylori-negative contained H. pylori sequences (Bik et al., 2006). However, H. pylori sequence counts in the corresponding samples tend to be substantially lower than in the samples initially tested H. pylori-positive (Monstein et al., 2000; Bik et al., 2006; Maldonado-Contreras et al., 2011), and gastric inflammation is usually associated with significantly increased H. pylori abundances (Monstein et al., 2000). While H. pylori can completely dominate a sample and can account for up to 99% of the sequences found, the average H. pylori abundance was 72% and 11% of all 16S rRNA gene library clones for H. pylori-positive and false-negative subjects, respectively (Bik et al., 2006). The widespread low abundance of H. pylori in samples tested negative for this organism in combination with the repeated detection of a mixed non-H. pylori microbiota led to the suggestion that ‘helicobacters could be part of an indigenous microbial communities in the stomach’ (Monstein et al., 2000). However, given the variability of H. pylori abundances and the confirmation of some H. pylori-free samples in sequence-based high-throughput analyses (Bik et al., 2006; Li et al., 2009), this is by no means stable across individual stomachs. Nevertheless, these results highlight that additional research into the true prevalence of low-level colonization of H. pylori is warranted, especially as the initial determination of H. pylori status in both Monstein et al. (2000) and Bik et al. (2006) was based on a combination of several tests that are considered to be highly sensitive even when applied individually (reviewed in Mégraud & Lehours, 2007). This raises many additional questions, such as to what extent the reported H. pylori reinfection rates after eradication or after spontaneous clearance in children might be due to an increase in bacterial numbers from previously undetectable levels and whether the decreased H. pylori prevalence found in industrialized countries might in part be due to H. pylori loads below the detection limit.
Influence of H. pylori infection on the composition of the gastric microbiota
In samples with an overwhelming dominance of H. pylori, the sampling depth for non-H. pylori sequences is greatly reduced, which also reduces the measured diversity (Andersson et al., 2008). If however enough non-H. pylori sequences are detected in a H. pylori-positive sample, then both the number of observed phylotypes and non-H. pylori diversity estimates can exceed the values for H. pylori-negative samples (Bik et al., 2006).
Reports on the effect of H. pylori on the composition of the stomach microbiota vary. Bik et al. (2006) found no significant correlation between phylotype distributions and H. pylori status or gastric pH levels, but instead a high level of variability between the individual study subjects. Maldonado-Contreras et al. (2011) also identified similar relative abundances of the four dominant phyla (Proteobacteria, Firmicutes, Actinobacteria, and Bacteroidetes) between H. pylori-positive samples, and those that had tested negative but still contained some Helicobacteraceae sequences. However, they could additionally correlate the relative abundance of some taxa to H. pylori status: In H. pylori-positive patients, they found increased relative abundances of non-H. pylori Proteobacteria and Acidobacteria, but reduced abundances of Actinobacteria and Firmicutes. This was corroborated by a regression analysis of PhyloChip-determined abundances, which additionally identified a positive correlation between Spirochaetae and H. pylori and negative correlations between H. pylori and Bacteroidetes, Chloroflexi, Cyanobacteria, Fusobacteria, Plactomycetes, Beta- and Gammaproteobacteria, and Verrucomicrobia. Overall, they found that 28% of the variation in microbiota compositions was explained by H. pylori status. Another 44% were explained by the origin of the patients, which is perhaps not surprising considering that this study compared Amerindian patients from rural Amazonas (Venezuela) with one Rwandan and one Bangladeshi immigrant living in the United States (Maldonado-Contreras et al., 2011).
Hu et al. (2012) combined culture-based detection with species identification using matrix-assisted laser desorption ionization time-of-flight mass spectrometry to investigate the stomach microbiota of H. pylori-positive patients with different gastric diseases. They found a significantly higher prevalence of culturable non-H. pylori bacteria in patients suffering from nonulcer dyspepsia (100%) than in gastric ulcer patients (43%), which led to the suggestion that these bacteria might be involved in the development of nonulcer dyspepsia.
Stomach microbiota under conditions of elevated pH
The acidity of the human stomach can be compromised in a variety of conditions. One of these is atrophic gastritis, characterized by the loss of acid-producing cells due to chronic mucosal inflammation, which can be immunologically mediated (reviewed in Lauwers et al., 2010) or be a result of H. pylori infection. The gastric pH can also be increased through pharmacological intervention, for example with PPIs. Long-term PPI treatment is the standard treatment for gastroesophageal reflux disease, both for acute healing and for the prevention of relapse (reviewed in Schubert & Peura, 2008). Old age has also been reported to be associated with compromised acid secretion (Husebye et al., 1992), but this is now thought to be mainly due to cohort effects related to changing rates of H. pylori infection and H. pylori-associated atrophic gastritis (reviewed in Husebye, 2005). These changes of the intragastric pH are a prominent factor influencing the microbiota of both the stomach lumen and the gastric mucosa. During atrophic gastritis, the stomach microbiota can in some cases reach densities at which its urease-producing members can cause a reaction in urea breath tests even in the absence of Helicobacter species (Osaki et al., 2008). A similar effect has been reported during H2 receptor antagonist treatment and in a patient with compromised gastric mobility (Michaud et al., 1998). The responsible taxa are reported to include Proteus mirabilis, Klebsiella pneumoniae, Staphylococcus aureus, Staphylococcus capitis subsp. urealyticus, and Micrococcus sp. (Michaud et al., 1998; Brandi et al., 2006; Osaki et al., 2008).
While we are not aware of any publication providing a more comprehensive overview over the gastric microbiota during atrophic gastritis, several authors investigated the effect of compromised gastric pH levels. In these studies, elevated pH was either caused by a variety of clinical conditions in different patients or induced pharmacologically.
The influence of the level of acidity on both the luminal and the mucosal gastric microbiota was addressed in a study on patients undergoing endoscopy for medical reasons (Kato et al., 2006). Their study population consisted of 10 children or teenagers (age 9–16 years) and 10 adults (33–79 years), with half of the patients in each age group diagnosed as H. pylori-positive by 13C urea breath test. Their culture-based assay detected non-Helicobacter bacteria in the gastric juice of all adults and three of the children. The luminal bacteria from adult stomachs were identified as Streptococcus spp., Neisseria spp., Bacillus spp., Lactobacillus plantarum, Lactobacillus salivarius, Lactobacillus fermentum, Lactobacillus gasseri, Peptostreptococcus anaerobius, ‘Bacteroides fragilis group’, and Fusobacterium spp., and the gastric juice of children was found to contain the taxa ‘Bifidobacterium group’, Eubacterium biforme, and ‘Bacteroides fragilis group’. Non-Helicobacter bacteria were also found in biopsy material from all adults, and bacterial counts in the mucosa were significantly correlated with luminal pH. In addition to the groups also found in the lumen, Staphylococcus spp., Lactobacillus casei, and Veillonella spp. were identified. Among the 10 children, mucosa-attached non-Helicobacter bacteria were only found in the Helicobacter-positive child with the highest intragastric pH (pH 2.7). Even though the lowest pH values measured in adult stomachs were within the range found in children, adults overall had significantly higher values. Moderate or severe gastric atrophy was diagnosed in 8 of 10 adult subjects but in none of the children. Although this study was based on samples from patients with various stomach disorders (Kato et al., 2006), the finding that the gastric mucosa of Helicobacter-free adults can be populated by different bacteria even in the absence of severe gastric atrophy and if the luminal pH is below 2 is significant. This result is unlikely to be influenced by contaminating mouth or throat microbiota, as the stomach mucosae of 9 of 10 children examined with the same methods were found to be free of non-Helicobacter bacteria.
PPI treatments can rapidly alter both the gastric pH and the associated microbiota. A 2-week treatment with omeprazole increased the fasting intragastric pH of healthy young volunteers by 1.2 units and led to a raise in CFU counts of culturable bacteria in the stomach contents by a factor of about 200 (Sharma et al., 1984). This effect of acid suppression therapy on the microbiota can also be detected in biopsy samples (Sanduleanu et al., 2001). Both on the mucosa and in the gastric lumen, the number of culturable bacteria increases with prolonged treatment duration (Sanduleanu et al., 2001; Del Piano et al., 2012). PPI treatment over 12 months can increase CFU counts by a factor of 106 (Del Piano et al., 2012).
Other clinical conditions
Conditions that physically alter the stomach environment can also influence its microbiota. One of these alterations is Roux-en-Y gastric bypass surgery, a surgical procedure to treat morbid obesity, which effectively splits the stomach into a small proximal pouch and a larger bypassed chamber. Investigating the same group of patients 4.5–11 years after this operation, these two stomach chambers were analyzed for microbial counts in relation to pH conditions and mucosal cytokine levels, respectively (Faintuch et al., 2007; Ishida et al., 2007). Elevated numbers of culturable bacteria and fungi were detected in both chambers. The pH of gastric juice from the proximal pouch was neutral (pH 7.0 ± 0.2), which was associated with significantly higher numbers of aerobe and anaerobic bacteria than in the bypassed chamber. Acidity in the bypassed chamber was only reduced to pH 3.3 ± 2.2, which was associated with significantly lower (albeit still elevated) bacterial numbers (Ishida et al., 2007). Differences in microbiota composition between the two chambers were not detected, presumably due to the limited number of isolates (Faintuch et al., 2007).
Another widespread alteration of the physical environment of the gut is enteral feeding, which involves the prolonged presence of a nasogastric or a percutaneous endoscopic gastrostomy (PEG) tube. In these patients, the culturable stomach microbiota was found to reflect the type of tube (Segal et al., 2006). The yeast Candida spp. was significantly more often isolated from the gastric contents of patients fed by PEG tube than from nasogastric tube-fed patients, which the authors attribute to the longer time PEG tubes usually remain in place without being replaced. From the gastric fluid of nasogastric tube-fed patients, the bacteria Pseudomonas aeruginosa, Proteus spp., Escherichia coli, Corynebacterium spp., and Streptococcus spp. of the b hemolytic type were isolated significantly more often. The authors hypothesize that the transmission of these organisms from the oropharynx is facilitated by the presence of the nasogastric tube itself, which carries its own biofilm, may disturb the function of the lower esophageal sphincter and might generally further a bidirectional transmission of organisms between mouth and stomach (Segal et al., 2006).
Conclusion and Outlook
Over the last 30 years, H. pylori infection has been studied intensively, and it may now be the best-studied chronic bacterial infection of humans in a defined ecological niche. In spite of the large number of available studies, including several that employed state-of-the-art high-throughput sequence analysis, our understanding of the non-H. pylori human gastric microbiota is still sketchy. The emerging picture is that of a low-density but diverse microbiota of distinct composition (Bik et al., 2006). It can be influenced by factors such as geographical origin of the study subject, gastritis, or infection with and abundance of H. pylori (Bik et al., 2006; Andersson et al., 2008; Maldonado-Contreras et al., 2011). Conditions of elevated pH can lead to increased bacterial densities (Sharma et al., 1984) and, in rare cases, to overgrowth with a few individual species (Osaki et al., 2008).
While studies that isolated different taxa from washed biopsy material (Li et al., 2009; Delgado et al., 2011, 2013) support the existence of attached and viable non-Helicobacter bacteria in the gastric mucus, their location in the mucus layer remains elusive.
In contrast to the viability of the attached bacteria, their in situ activity has not yet been formally demonstrated. This needs to be urgently addressed: Most bacteria are quickly killed in the harsh environment of the stomach lumen, which will not necessarily prevent them from being detected with molecular methods. Others might be able to survive in the stomach for short periods and then resume growth after being picked up during sampling, which then would not necessarily correlate with metabolic activity in situ. Studies addressing this issue should try to quantify the proportions of metabolically active and nonactive non-Helicobacter bacteria, for example by combining DNA-based with RNA-based approaches, preferably using carefully rinsed biopsies while controlling for contamination with throat bacteria. Studies using live microscopy in suitable animal models may help to better understand the ecology and host interaction of the non-Helicobacter microbiota in the mucosal layer and look for evidence of active growth. We believe that the available data summarized in this review permit to generate testable hypotheses that can be explored in both suitable experimental animal models and carefully designed clinical trials.
Acknowledgements
Work in the authors' laboratory is funded by grants from the German Research Foundation (SFB 900, SFB 621, IRTG 1273), the German Ministry for Education and Research (FBI-Zoo), the ERA-NET PathoGenoMics project HELDIVPAT, and the German Center for Infection Research (DZIF).
Authors' contribution
I.Y. and S.N. contributed equally to this work.
References
Editor: Friedrich Götz
