The Human Gastric Microbiome Is Predicated upon Infection with Helicobacter pylori

The human gastric lumen is one of the most hostile environments of the human body suspected to be sterile until the discovery of Helicobacter pylori (H.p.). State of the art next generation sequencing technologies multiply the knowledge on H.p. functional genomics as well as on the colonization of supposed sterile human environments like the gastric habitat. Here we studied in a prospective, multicenter, clinical trial the 16S rRNA gene amplicon based bacterial microbiome in a total of 30 homogenized and frozen gastric biopsy samples from eight geographic locations. The evaluation of the samples for H.p. infection status was done by histopathology and a specific PCR assay. CagA status was determined by a CagA-specific PCR assay. Patients were grouped accordingly as H.p.-negative, H.p.-positive but CagA-negative and H.p.-positive and CagA-positive (n = 10, respectively). Here we show that H.p. infection of the gastric habitat dominates the gastric microbiota in most patients and is associated with a significant decrease of the microbial alpha diversity from H.p. negative to H.p. positive with CagA as a considerable factor. The genera Actinomyces, Granulicatella, Veillonella, Fusobacterium, Neisseria, Helicobacter, Streptococcus, and Prevotella are significantly different between the H.p.-positive and H.p.-negative sample groups. Differences in microbiota found between CagA-positive and CagA-negative patients were not statistically significant and need to be re-evaluated in larger sample cohorts. In conclusion, H.p. infection dominates the gastric microbiome in a multicentre cohort of patients with varying diagnoses.


INTRODUCTION
The microbiome (the entity of bacteria, viruses, archaea, and fungi) of the human lower intestinal tract has been well characterized in a plethora of clinical and physiological studies (Levy et al., 2016;Tilg et al., 2016;Wehkamp and Frick, 2016) and its significance for medical research, diagnosis, and diseases is increasingly recognized. Correlations between specific patterns of the intestinal microbiome with specific disease entities such as chronic inflammatory bowel disease are undisputed nowadays (Patel et al., 2016;Wehkamp and Frick, 2016). Even microenvironments that were perceived to be sterile in healthy humans such as the urinary bladder or the human lung were found to harbor a diversity of microbes (Cui et al., 2014;Thomas-White et al., 2016). Information on the human gastric microbiome is increasing but still limited and large cohort data are often based on culture dependent approaches or animal models (Khosravi et al., 2014;Majlessi et al., 2017). The human stomach has been supposed to be devoid of significant microbial colonization and diversity because of its hostile environment. The very low pH and tight immune surveillance render the human stomach a gate keeper for the entrance of microbial pathogens into the intestinal tract.
However, the discovery of Helicobacter pylori (H.p.) in 1984 (Marshall and Warren, 1984) changed the view on microbial colonization of the stomach. H.p. is one of the genetically best characterized and fully sequenced organisms due to its potentially carcinogenic effect (Wroblewski and Peek, 2016) and high relevance for human health (Tomb et al., 1997;Noto and Peek, 2017;Shah, 2017). Nowadays, multiple other microbiota than H.p. have been described in gastric samples (Bik et al., 2006;Dong et al., 2016;Péré-Védrenne et al., 2017). Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, and Fusobacteria are the most abundant phyla in previous studies detected by culture dependent, mass spectrometry and sequencing approaches (Khosravi et al., 2014;Ianiro et al., 2015;Dias-Jacome et al., 2016). Although, the impact of H.p. on the non-H.p. microbiome has been studied with various techniques in animal models (Kienesberger et al., 2016) as well as human approaches (Bik et al., 2006;Schulz et al., 2016a;Yang et al., 2016;Brawner et al., 2017) the generation of human gastric microbiome and H.p. related data is still of incredible significance to investigate mechanisms of human H.p. infections.
The genome of H.p. is well annotated and pathogenicity islands (PAI) have been described (Feliciano et al., 2015). One of the most virulent of these PAI genes suspected to be a main driver of carcinogenesis is the CagA gene located in the cag PAI (Paredes-Osses et al., 2017). The cag PAI is responsible for translocation of the Cag protein into the host cell (Hatakeyama, 2014). Characterization of the CagA effect in H.p. positive samples on the non-H.p. microbiome has been performed previously in a small Colombian sample cohort. They revealed no statistically significant differences in CagA negative compared to positive infections but showed a trend of reduced H.p. abundance and reduced histopathology score in the CagA negative patients (Yang et al., 2016). Functional studies on the interaction of H.p. and the suspected gastric microbiome were described from in vitro studies or mouse models (Khosravi et al., 2016;Kienesberger et al., 2016).
Here we provide NGS based 16S rRNA gene data on the relevance of CagA positive and CagA negative H.p. infections on the gastric microbiome of a clinically very well characterized, adult human population collected at eight different geographic locations all over Austria.

Clinical Samples
Two gastric mucosal biopsy samples from the antrum of the stomach were collected from 30 patients, who were older than 18 years and did not have a gastroscopic investigation in the past 10 years. All samples were taken from the antrum region to overcome the problem of variation potentially caused by different gastric regions. The patients included in this study have never been treated for H.p. infection before and were not treated with proton-pump inhibitors for at least 2 weeks or antibiotics for at least 1 month before endoscopy. The indication for gastroscopy was in the majority of the patients upper abdominal pain and suspected gastritis (33%) and symptoms compatible with reflux esophagitis (23%). In this prospective study, samples were grouped according to their H.p. and CagA status but not according to their diagnosis. Samples were selected randomly from a cohort of more than 2000 human gastric biopsies (Bilgilier et al., 2017). A written informed consent was obtained from all participants. The study protocol was approved by the ethics committee of the Medical University Vienna (EK# 1548/2014) and the study was conducted in accordance with the Declaration of Helsinki.

Evaluation of H. p. Infection Status in Gastric Biopsies
Histomorphological evaluation of one of the two gastric biopsy samples for the presence of Helicobacter-like organisms (HLO) was done with the use of hematoxylin and eosin staining and a modified Giemsa staining followed by microscopic evaluation according to standard procedures (Fallone et al., 1997) and as described in detail in Bilgilier et al. (2017). For the detection of H.p.-specific DNA, the second fresh gastric biopsy sample from each patient was homogenized in 1 ml 0.9% NaCl solution (Sigma Aldrich) using Lysing Matrix D tubes with ceramic beads (MP Biomedicals) in an FastPrep R -24 Instrument (MP Biomedicals) at 4.5m/s for 20 s once. An aliquot of 50 µl from the homogenates were subjected to genomic DNA isolation with the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) immediately according to manufacturer's instructions. The remaining 950 µl homogenate from each sample was frozen without DNA isolation and stored at −80 • C until DNA isolation for microbiome analysis. Subsequently, 2 µl of total DNA were used in an H.p. specific PCR assay detecting the 23S rRNA gene by fluorescence detection as described previously (Schabereiter-Gurtner et al., 2004). In brief, this PCR assay was performed with 40 amplification cycles of denaturation at 95 • C for 5 s, annealing at 65 • C for 10 s and extension at 72 • C for 6 s, and read-outs were based on single fluorescence acquisition at the end of each extension step. The presence of CagA in HP positive samples was investigated with use of a PCR assay described previously (Bilgilier et al., 2016). In brief, the combination of three different reverse primers, CagA-rvP1, CagA-rvP2, and CagA-rvP3, allowed the detection of the CagA gene and also typing of the EPIYA motif encoded within this gene, which was considered an important virulence determinant. According to the results from the H.p. specific and the CagA specific PCR assays, the gastric biopsy samples were gathered in three different groups: (Levy et al., 2016) HP negative samples, (Tilg et al., 2016) HP positive samples with HP strains that do not encode the CagA gene and (Wehkamp and Frick, 2016) HP positive samples with HP strains encoding the CagA gene. Each group consisted of 10 individuals.
DNA Isolation for Microbiome Analysis, 16S rRNA Gene PCR Amplification and Sequencing Total DNA was isolated by a combination of mechanic and enzymatic lysis according to standard procedures as published recently in Klymiuk et al. (2016). Briefly, fresh biopsy samples were homogenized in 1 ml 0.9% NaCl and 950 µl of the homogenate were frozen and stored at −80 • C till DNA isolation. Samples were centrifuged at 6,000 g for 10 min and supernatant was removed. Samples were homogenized in a total volume of 500 µl MagNA Pure Bacteria Lysis Buffer from the MagNA Pure LC DNA Isolation Kit III (Bacteria, Fungi) (Roche, Mannheim, Germany) in MagNA Lyser green beads tubes (Roche, Mannheim, Germany) at 6,500 rpm for 30s repeated three times in a MagNA Lyser Instrument (Roche, Mannheim, Germany). Twenty-five Microliter lysozyme (100 mg/ml) were added to the homogenized samples, mixed and incubated at 37 • C for 30 min. Afterwards 43.4 µl Proteinase K (20 mg/ml) were added and samples were incubated at 65 • C over night. The next day after heat inactivation of enzymes at 95 • C for 10 min DNA was extracted from 250 µl lysed supernatant on a MagNA Pure LC 2.0 (Roche, Mannheim, Germany) according to manufacturer's instructions of the MagNA Pure LC DNA Isolation Kit III (Bacteria, Fungi) (Roche, Mannheim, Germany). Five microliter of total DNA were used in a 25 µl PCR reaction in triplicates using a FastStart High Fidelity PCR system (Roche, Mannheim, Germany). Each PCR reaction comprised of 1x Fast Start High Fidelity Buffer (Roche, Mannheim, Germany), 1.25 U High Fidelity Enzyme (Roche, Mannheim, Germany), 200 µM dNTPs (Roche, Mannheim, Germany), 0.4 µM primers and PCR-grade water (Roche, Mannheim, Germany). For the amplification of phylogenetic informative hypervariable regions V1-V2 the target primers 27F-AGAGTTTGATCCTGGCTCAG and 375R-CTGCTGCCTYCCGTA were used with Illumina adapters for indexing PCR reaction according to Illumina's 16 s metagenomic sequencing library preparation guide (http://www.illumina.com/content/dam/illumina-support/ documents/documentation/chemistry_documentation/16s/16smetagenomic-library-prep-guide-15044223-b.pdf; last accessed July 2017). Cycling conditions were of initial denaturation at 95 • C for 3 min followed by 30 cycles of denaturation at 95 • C for 45 s, annealing of primers at 55 • C for 45 s and extension at 72 • C for 1 min followed by a final extension step at 72 • C for 7 min and subsequent cooling to 4 • C. Triplicates were pooled and checked on a 1% agarose gel before normalization of 20 µl PCR product on a SequalPrep Normalization Plate according to manufacturer's instructions (LifeTechnologies, Germany). Fifteen microliter of the normalized PCR product were used as template in a single 50 µl indexing PCR reaction for 8 cycles; the cycling conditions were as described above for the targeted PCR. Five microliter PCR product from each sample were pooled to the final sequencing library and 30 µl of the unpurified library were loaded to a 1% agarose gel for purification with the QIAquick gel extraction kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. The purified library was quantified with QuantiFluor ONE dsDNA Dye on Quantus TM Fluorometer (Promega, Mannheim, Germany), loaded to an Agilent BioAnalyzer 2100 (Waldbronn, Germany) for quality control and the 6 pM library was sequenced on a MiSeq desktop sequencer (Illumina, Eindhoven, Netherlands) containing 20% PhiX control DNA (Illumina, Eindhoven, Netherlands) with v2 chemistry for 500 cycles according to manufacturer's instructions. FastQ raw reads were used for subsequent data analysis.

Data Analysis
A total of 2,825,234 MiSeq paired-end raw sequence forward and reverse reads were merged using ea-utils v1.1.2 (Aronesty, 2013) with standard settings, followed by a split library step from the Quantitative Insights Into Microbial Ecology (QIIME, v1.9.1) software (Caporaso et al., 2010). During this step sequence reads shorter than 200 nucleotides, reads that contain ambiguous bases or reads with an average quality score less than 30 were discarded. Chimera were removed with USEARCH v6.1 method in QIIME against 97% clustered GreenGenes reference 16S rRNA gene database (v13.8). In the second step OTU picking was done with QIIME open reference pipeline performing clustering steps at 97% sequence similarity, the taxonomy assignment with UCLUST algorithm (Edgar et al., 2011), alignment of reference sequences with pyNAST and generation of phylogenetic tree with FastTree. Finally, the OTU table was reduced by removing all OTUs present in only one sample with less than 10 reads. Downstream statistical analysis was performed in R version 3.3.3 (R Core Team, 2016) using a custom script. Principal Component Analysis (PCA) was performed using the prcomp function with default parameters. P-values for the group overlap in the PCA were calculated according to Goodpaster (Goodpaster and Kennedy, 2011) and Worley (Worley et al., 2013).

RESULTS
A total of 30 patients were included in the present study that underwent routine diagnostic gastroscopy, were never treated for H.p. infection before and did not have any gastroscopic investigation in the previous ten years ( Table 1). The cohort included patients with gastric biopsies that tested negative for H.p. by histopathology and H.p.-specific PCR (n = 10, group 1), positive for H.p. by histopathology and H.p.-specific PCR but negative by CagA-specific PCR (n = 10, group 2), or positive in all three assays (n = 10, group 3). The mean age of the 30 patients was 50 years (SD = 16.4 years; range, 23-83 years) and 53% (16 of 30) of them were female. Gastroscopy revealed pathological findings in 24 (80%) patients, including gastritis in 20 patients (67%), gastro-esophageal reflux disease in 8 patients (27%), and abdominal hernia in 4 patients (13%).   Frontiers in Microbiology | www.frontiersin.org contained the species Str. alactolyticus, Str. anginosus, Str. infantis, and Str. sobrinus. The genus Veillonella contained the species V. dispar and V. parvula and the genus Prevotella the species P. copri, P. intermedia, P. melaninogenica, P. nanceiensis, and P. nigrescens. All analyzed Helicobacter reads belonged to the species H. pylori (Supplementary Table 2).
Principal Component Analysis (PCA) on OTU and genus level data revealed a clustering (means higher similarity of samples within the group than to the samples of other groups) of gastric samples according to H.p. status with a significant difference between the H.p. negative and the H.p.+/CagA+ group (adj. p-value OTU = 0.000001, adj. p-value genera = 0.000001) and the H.p. negative and the H.p.+/CagA− group (adj. p-value OTU = 0.000009, adj. p-value genera = 0.000023), but no statistically significant differences were found between the H.p.+/CagA− vs. H.p.+/CagA+ group (adj. p-value OTU = 0.236619, adj. p-value genera = 0.095089) (Figure 2).

Effect of H.p. Infection and CagA Status on Alpha Diversity Results of Gastric Samples
Three different alpha diversity measures were used on the rarefied sequences to analyze the microbial diversity and compare the results of different methods of calculation. The highest estimated richness (Chao1) of stomach microbiota was found in the H.p.-sample group with 203 compared to 160 in the H.p.+/CagA− and 136 in the H.p.+/CagA+ group ( Figure 3A). Correspondingly, observed species as well as PD whole tree revealed the highest microbial diversity in H.p.− samples followed by the H.p.+/CagA− and the H.p.+/CagA+ group (Figures 3B,C). A decreasing alpha diversity from H.p.− to H.p.+/CagA− to H.p.+/CagA+ was observed but did not gain statistically significance (Figures 3A-C). All performed alpha diversity measures (Chao1, PD whole tree, and observed species), that were used to compare different methods of calculation, revealed significant differences between the H.p.− vs. H.p.+/CagA+ group ( Table 2). Chao1 analysis revealed statistically significant differences between the H.p.− and the H.p.+/CagA+ groups (adj.p-value = 0.028) but no significant differences in estimated richness between the other comparators (H.p.− vs. H.p.+/CagA− adj.p-value = 0.402; H.p.+/CagA− vs. H.p.+/CagA+ adj.p-value = 0.535) ( Table 2).
Power calculations were performed to estimate the samples size required for statistically significant differences in the alpha diversity of all sample groups (Supplementary Table  3). Dependent on the alpha diversity measure, 257 to 21.262 specimens per group were estimated for significant differences between the H.p.+/CagA− vs. H.p.+/CagA+ sample groups (Supplementary Table 2

Distribution Pattern of Overlapping OTUs and Genera in the Three Sample Groups
Overlap analysis on genus and OTU level was performed with Venn diagrams to provide the number of genera and OTUs specific for one, two or all investigated gastric sample groups (Figures 4A,B). From the 226 genera found in H.p.-samples 112 were in common in all three sample groups, 65 were specific for the H.p.-group, 29 in common with the with the H.p.+/CagA− group and 20 in common with the H.p.+/CagA+ sample group (Figure 4A). 7 genera were in common between the H.p.+/CagA− and the H.p.+/CagA+ group and 29 specific FIGURE 2 | Principal Component Analysis (PCA) ordination plots of relative abundance of genera (A) and OTUs (B). Statistically significant differences were found between H.p.− and the H.p.+/CagA− (adj. p-value OTU = 0.000001, adj. p-value genera = 0.000001) as well as between the H.p.− and the H.p.+/CagA+ groups (adj. p-value OTU = 0.000009, adj. p-value genera = 0.000023). No statistically significant differences were found between H.p.+/CagA− vs. H.p.+/CagA+ sample groups (adj. p-value OTU = 0.236619, adj. p-value genera = 0.095089). Ellipses denote the 95% confidence. for H.p.+/CagA− and 15 specific for H.p.+/CagA+ (Figure 4A and Supplementary Table 4). Filtering the genera for those that occur in at least 50% of the samples within one group with a relative abundance of at least 1% only 13 genera common for all sample groups were left, demonstrating that those genera responsible for the biological information are common in all sample groups. The group specific genera reveal less than 2% of the reads with exception of the sample P002Z18 in which genera specific for the H.p.-group reveal 10% of all reads in that sample. Similar distributions were found for the OTU pattern between the three groups ( Figure 4B).

Taxa Significantly Altered between the Three Sample Groups
Linear discriminant Effect Size analysis (LefSe) was performed to identify those taxa significantly different in their relative abundances in pairwise comparisons of the three sample groups from phylum to species level (Segata et al., 2011). Only taxa with a relative abundance of at least 1% in at least 50% of the samples within one group were considered.

DISCUSSION
In this exploratory study on the characteristics of the human gastric microbial pattern and its relation to H.p. infection with and without detection of the CagA gene, we analyzed gastric biopsies from a cohort of 30 patients who were prospectively evaluated for H.p. in a prospective, clinical, multicenter trial. Former studies on the human gastric microbiome indicate a distinct gastric microbial pattern with Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria as the dominating phyla and Streptococcus as the most dominant genus (Andersson et al., 2008;Khosravi et al., 2014;Ianiro et al., 2015;Llorca et al., 2016;Schulz et al., 2016a,b;Yang et al., 2016). Taxa identified in our data set correspond to these former studies from phylum to genus level. In concordance with our findings, culture dependent approaches found predominantly Streptococcus, Neisseria, Klebsiella, and Lactobacillus (Khosravi et al., 2014) with Streptococcus (8.61% over all samples) and Neisseria (1.13% over all samples) confirmed as dominant taxa. This study is distinguishing due to its prospective design, multicenter sample collection and the availability of numerous clinical data of the probands.

Technical Limitations of Gastric Microbiome Studies
The human gastric microbiome is more and more under the focus of clinical research (Majlessi et al., 2017;Noto and Peek, 2017;Péré-Védrenne et al., 2017;Shah, 2017). Recent studies on the oral, gastric and duodenal microbiome describe no significant differences at phylum level between the oral habitat and the stomach aspirates but clear differences compared to the duodenal phyla (Schulz et al., 2016a). In our opinion this is hard to believe as the oral habitat differs from the gastric so dramatically in environmental conditions. Instead, technical limitations must be considered in culture-dependent as well as culture independent approaches. DNA based studies detected all present nucleic acids. Under the permanent process of swallowing microbiota inhabiting the oral cavity or food derived microbial DNA are characterized as well. In studies on cultivable bacterial microbiota of the human stomach, samples are spread to conventional media (Vega et al., 2003) without adapting the culture conditions to those present in the natural gastric habitat (Sanduleanu et al., 2001;Delgado et al., 2013;Khosravi et al., 2014). We suggest that gastric habitat like conditions would be inevitable for serious conclusions on viability and above all proliferation capability of isolated microorganisms. In the same manner, characterization of human gastric microbiota in patients with acid inhibitors does not reflect natural stomach conditions and cannot exclude oral or intestinal sources overgrowing the natural gastric community due to the artificially increased stomach pH and therefore the disruption of its barrier function (Paroni Sterbini et al., 2016;Llorente et al., 2017) putting their physiological relevance in the healthy, acidic gastric lumen into question. Finally, the source of contamination of gastric biopsy samples through the gastroscopic instrument and sampling procedure from the upper oropharyngeal region needs to be considered. A method to discriminate between live and dead cells would be the treatment of freshly collected samples with Propidium Monoazide (PMA) and thereby masking cell free DNA derived from dead microorganisms for PCR reaction (Nocker et al., 2007). Nevertheless, we hypothesize that this approach as well as the approach to work with isolated RNA instead of DNA (Schulz et al., 2016a) does not represent an improvement as saliva is swallow continuously, therefore living microorganisms are delivered to the stomach continuously and swallowed microorganisms may tolerate the acidic conditions without proliferation for a certain period of time.
In conclusion, we describe a distinct microbial pattern of the human gastric microbiome from biopsy samples of a clinically well characterized cohort collected at eight different geographic locations in Austria. We observe a dramatic decrease of the 16S rRNA gene based detected non H.p. bacterial microbiome in H.p. infected samples and from our results, presence of the CagA gene has no statistically significant influence on accompanying microbial taxa, although a non-significant trend on the microbial alpha diversity is observed.