Patients included in this study were having various upper gastrointestinal symptoms and seeking care at the Department of Gastroenterology, Government Medical College, Trivandrum (TMC). Trivandrum is the capital of Indian state Kerala, which is located in extreme South-West part of the country (Menon, 2017). The entire state including Trivandrum has Arabian Sea to the West and Western Ghats Mountain to the East. It is suggested that humans lived in this geographical region during Neolithic Age (Menon, 2000, 2017). Later, peopling of Kerala happened during 2–3 AD through land and sea. The modern Keralite community has diverse (Negroid, Proto-Australoid, Dravidian, and Aryan) lineages (Menon, 2017).

Two gastric biopsies were collected during upper GI endoscopy. One of them was taken in 600 μl of autoclaved Brucella broth containing glycerol and the other in 200 μl of phosphate-buffered saline (PBS; 0.22 μm membrane filtered; autoclaved). A stool sample was also collected. The biopsy and stool samples were transported to the Microbiome Laboratory of Rajiv Gandhi Centre for Biotechnology (RGCB) at 4°C and were stored immediately in a −80°C freezer until further processing. Written informed consents from patients were taken. The study was approved by the Institutional Human Ethical Committee of TMC (approval number: 05/07/2016/MCT) and RGCB (approval number: IHEC/01/2017/18).

DNA was extracted from gastric biopsy as described elsewhere (Das et al., 2017). The DNA was diluted to 4 ng/μl, and 1 μl of DNA was used in 20 μl of reaction volume containing 10 μl of EmeraldAmp GT PCR Master Mix (TaKaRa) and 2 μl of forward and reverse primers specific for H. pylori ureB (Supplementary Figure S1). A patient is considered to have H. pylori infection if the collected gastric biopsy showed the presence of H. pylori either by ureB PCR or by culture (described below) or by both techniques (Table 1 and Supplementary Tables S1, S2).

The biopsies collected in Brucella broth were used for isolating H. pylori strains on Brain Heart Infusion (BHI) agar (2%) containing Dent (Oxoid), 0.4% IsoVitaleX (BBL), and 7% sheep blood. The inoculated plates were incubated in microaerobic condition (10% CO₂, 5% O₂, and 85% N₂) at 37°C. H. pylori colonies (one colony for each patient) were further propagated as pure culture. H. pylori was identified by typical translucent colony morphology, Gram staining, as well as biochemical tests like urease, catalase, and oxidase (Supplementary Figure S2).

Genomic DNA was extracted from H. pylori strains as described elsewhere (Berg et al., 1997). RAPD-PCR was carried out in 25 μl of reaction volume containing 2.5 μl of primer 1,281 (10 pmole), 4 mM MgCl₂, and 1.5 U of rTaq polymerase (TaKaRa) using previously described conditions (Berg et al., 1997; Mukhopadhyay et al., 2000). The multiplex PCR for vacA and cagA was performed using a modified protocol with 10 pmol of VA1-F/VA1-R, 10 pmol of VAG-F/VAG-R and 25 pmol of cag5c-F/cag5c-R, and 10 μl EmeraldAmp GT PCR Master Mix in 20 μl of reaction mix (Chattopadhyay et al., 2004). Characterization of other alleles of vacA and cagA by PCR and sequencing was done by previously described methods (Mukhopadhyay et al., 2000; Rhead et al., 2007; Chattopadhyay et al., 2012). The nucleotide sequences of the primers are given in Supplementary Table S3. For phylogenic analyses, PhyML 3.0 maximum likelihood trees were generated using bootstrapped (100 iterations) following estimation of an evolutionary model using full_modeltest_bootstrap genetic workflow in ETE3 python package (Guindon et al., 2010; Huerta-Cepas et al., 2016). The generated Newick tree files were used with the phylogram package for R to plot the phylogenetic trees (Wilkinson and Davy, 2018).

DNA was extracted from 200 mg of stool following a previously described protocol (Bag et al., 2016). For the preparation of metagenome library, 30 ng DNA was used to amplify the V3–V4 region of bacterial 16S rRNA genes for 26 cycles using KAPA HiFi HotStart PCR kit (KAPA Biosystems Inc., Boston, MA, United States) (Supplementary Table S3). The products were further amplified for 10 cycles by index PCR to add the Illumina sequencing barcoded adapters (Nextera XT v2 Index Kit, Illumina, United States). The products were sequenced using Illumina MiSeq following manufacturer’s protocol. The paired end V3–V4 reads (275 × 2) were demultiplexed using bcl2fastq, quality checked using FastQC, stitched using Fastq-join and analyzed using QIIME. The query sequences were clustered using UCLUST method against a curated chimera-free 16s rRNA database (Greengenes v.13.8). The taxonomies were assigned using RDP classifier to these clusters at ≥97% sequence similarity against the reference database. The generated BIOM file was used for further analysis and visualization. The box plot analysis was done by R.

For whole genome metagenome sequencing, 300 ng of genomic DNA was used after end-repairing (NEBnext ultra II end repair kit, New England Biolabs, MA, United States) and cleaning up with 1× AmPure beads (BeckmannCoulter, United States). The DNA samples were barcoded (LongAmp Taq 2× New England Biolabs, MA, United States) and cleaned up with 1.6× AmPure beads (Beckmann-Coulter, United States). The end-repairing was performed using NEBnext (New England Biolabs, MA, United States) and adapter ligation was performed for 10 min using NEB blunt/TA ligase (New England Biolabs, MA, United States). Library mix was cleaned up using 0.6× Ampure beads and finally eluted in 15 μl of elution buffer. The processed DNA samples were sequenced on GridION X5 (Oxford Nanopore Technologies, Oxford, United Kingdom) using SpotON flow cell (R9.4) in a 48 h sequencing protocol on MinKNOW 2.1 v18.05.5. Nanopore raw reads (“fast5” format) were base called (“fastq5” format) and demultiplexed using Albacore v2.3.1. The reads were compared against NCBI nr database using the diamond tool. The diamond BLASTX alignments were further converted to MEGAN readable format by using the NCBI taxonomy to summarize and order the results. MEGAN GUI is then used to estimate and interactively explore the taxonomical content by checking the read assignment from phylum to species level classification.

PCR with primers specific for Bifidobacterium species (Supplementary Table S3) was done using stool metagenomic DNA in a 20 μl PCR reaction. Similarly, PCR with DNA extracted metagenomically from gastric biopsies was also performed. We also performed quantitative PCR (qPCR) in triplicate for the stool DNA as well as the gastric biopsy DNA using Bifdobacterium species-specific primers. The qPCR was performed with Thermo Power SYBR Green Master Mix using 200 nM primers and 50 ng DNA. Standard program with annealing temperature of 55°C in Applied Biosystems QuantStudio 7 instrument was used.

For analyzing H. pylori infection status, clinical status and sex of the individual Chi-squared test was performed using Intercooled Stata 14.1 software to test the significance of the association between variables. Binary logistic regression was used to estimate the odds ratios with 95% confidence intervals. For metagenomics analysis, the statistical significances among the patient groups were calculated using the Kruskal–Wallis test (Kruskal–Wallis, p < 0.05).

The study includes a total of 375 adult (male: 181; female: 194; average age: 48.5 years) residents of Trivandrum city and suburbs. As shown in Table 1, the prevalence of H. pylori infection is remarkably low (83 of 375; 22.1%) in Trivandrum. Within the H. pylori-infected group, the total prevalence of severe gastric diseases (15 of 83; 18.1%) like gastric cancer and peptic ulcer (duodenal and gastric ulcers) are similar to other geographic regions. However, the distributions of different diseases were noticeably different from the rest of the country. For example, it is known that for most Indian states that duodenal ulcer is the major clinical outcome and gastric cancers are relatively less prevalent (Supplementary Table S4; Dhakal and Dhakal, 2018). In contrast, for Trivandrum, although the prevalence of total gastric cancer (23/375 or 6.1%) and gastric ulcer (22 or 5.9%) are high, the prevalence of duodenal ulcer (6/375 or 1.6%) is low (Table 1 and Supplementary Table S4).

Of the 83 H. pylori-infected patients, 35 were male (42.2%) and 48 were female (57.8%) (Table 2). The prevalence of severe gastric diseases like gastric cancer and peptic ulcer are significantly more in males (Table 2). Among the H. pylori-infected males, 34.3% had severe disease types, while in the corresponding female population, it was only 6.25%. The observed association between sex and disease status is statistically significant (p = 0.001) (Table 2).

Of the 83 H. pylori positive cases, 42 were positive by culture. DNA extracted from all 42 isolated H. pylori strains were used for genotyping the vacA signal sequence (s), mid (m), and intermediate (i) region alleles as well as the cagA 5′end conserved region and 3′end variable region. Of the 42 strains, 39 (92.9%) carried the toxigenic vacAs1 allele, while three strains (7.1%) carried the non-toxigenic vacAs2 allele (Figures 1A,B and Table 3). The prevalence of vacAm1 (32 of 42; 76.2%) was higher than the prevalence of vacAm2 (10 of 42; 23.8%). The cagA gene is present in 38 (90.5%) of 42 strains (Figures 1A,B and Table 3). Four strains (9.5%), which were negative for cagA gene, gave the 550 bp amplicon in cag-empty site PCR confirming that these four strains lacked the entire cag-PAI (Figure 1C and Table 3). The prevalence of vacAi1 (37 of 42; 88.1%) allele was higher than the prevalence of vacAi2 (5 of 42; 11.9%) allele (Figures 1D,E and Table 3). When the genotype data were combined, it was found that in Trivandrum, the H. pylori strains predominantly carry the most toxigenic vacAs1i1m1cagA+ genotype (73.8%), followed by the vacAs1i1m2cagA+ genotype (11.9%) (Table 3 and Supplementary Table S5). The strains that carry the less toxigenic vacAs1i2m2cagA+ (4.8%) and non-toxigenic vacAs2i2m2cagA− (7.1%) genotypes are relatively uncommon. One strain (TMC280) carried a rare vacAs1i1m1cagA− genotype (Figure 2, Table 3, and Supplementary Table S5). DNA fingerprinting analysis using randomly amplified polymorphic DNA (RAPD)-PCR for 11 representative strains of different genotypes showed unique pattern for each strain, suggesting that the Trivandrum H. pylori strains, like the H. pylori strains from other geographic regions, are highly diverse (Supplementary Figure S3). Phylogenetic analyses revealed that the vacA of Trivandrum H. pylori strains are related to the vacA of H. pylori strains isolated from South Asia (India, Bangladesh, etc.), while the cagA of the Trivandrum H. pylori strains formed cluster with the Western cagA (Figure 3).

The 3′end variable region of the cagA gene that encodes variable numbers of EPIYA motifs and spacer sequences was studied by PCR and sequencing (Figure 4A). All cagA + H. pylori strains carried Western type-specific sequence (WSS) with C-segment (Figure 4B). Of the 38 cagA + strains, 28 (73.7%) were AB-C type with three EPIYA motifs, and seven (18.4%) were AB-C-C type with four EPIYA motifs. Two strains (5.2%) did not carry EPIYA motifs at the C segment, and they were AB- type with only two EPIYA motifs. One strain (2.7%) was found to have one EPIYA motif at the A site and two EPIYA motifs at the C sites, but no EPIYA motif at the B site, and therefore, this was the A-C-C type CagA. However, no association with the number of EPIYA motifs at the C-segment was found to have any clinical correlation (Supplementary Table S6).

Likewise, no association with clinical status of the host and genotypes of the H. pylori strains could be made (Table 3 and Supplementary Tables S6, S7). Therefore, the infection with putatively pathogenic types of H. pylori strains is dominant in all patient populations, but only certain people develop the severe gastric diseases like gastric cancer and peptic ulcer. These data suggest that, in addition to H. pylori infection and associated virulence genes, other factors contribute in determining clinical outcomes of the infected individuals.

As subsets of the study population, a total of 60 patients (30 H. pylori+ with average age 50.5 years; H. pylori- with average age of 50.4 years; same male to female ratio) were included in the gut microbiota analysis using Illumina MiSeq (275 × 2) platform (Supplementary Table S8). The rarefaction curve used as a measure of depth of sequencing is shown in Supplementary Figure S4. It is evident that for most patients, the abundant gut bacterial phyla is Firmicutes, followed by Bacteroidetes, Proteobacteria, Tenericutes, and Actinobacteria (Supplementary Figures S5, S6). However, for few patients like TMC27 (a patient with GERD; infected with H. pylori), TMC50 (a patient with gastritis; not infected with H. pylori), TMC58 (a patient with non-ulcer dyspepsia; not infected with H. pylori), TMC131 and TMC156 (two patients with gastritis; infected with H. pylori). Proteobacteria is the dominant phylum in the gut (Figure 5A and Supplementary Figure S6). No patient with severe gastric diseases like gastric cancer or peptic ulcer was found to have Proteobacteria as dominant phylum in the fecal microbiome. Overall, as shown in the heat map, four phyla, Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria showed wide variation in abundance among individuals (Figure 5A). Principal Coordinates Analysis (PCoA) showed wide variations among subjects, but many H. pylori+ cases are found to be closely related to each other (Supplementary Figure S7). Likewise, many H. pylori- cases are also closely related to each other. This strongly suggests that the H. pylori+ group and the H. pylori- group may have respective unique components, which are distinct from the other group.

To this end, we looked into comparative analysis of the gut microbiota composition for the H. pylori+ (N = 30) and H. pylori− (N = 30) patients irrespective of clinical status of the host. Most of the H. pylori+ patients have more diverse gut microbiota than their respective age- and sex-matched H. pylori- counterparts (Supplementary Table S9). Collectively, the H. pylori+ group has more alpha diversity than the H. pylori- group as was discerned by the Shannon index, Simpson index, and Chao1 index analyses (Supplementary Figure S8). When the bacterial relative abundance was taken into account between two groups, it was noticed that the abundance of phylum Actinobacteria was lower and phylum TM7 (Saccharibacteria) was higher in the H. pylori+ group than in the H. pylori- group (Figure 5B). Similar analysis at the genus level showed that Bifidobacterium (belonging to the phylum Actinobacteria) and Bacteroides (belonging to the phylum Bacteroidetes) were present in lower relative abundance for the H. pylori+ group than the H. pylori- group (Figure 5C). Conversely, for the H. pylori+ group, genus Dialister (bacteria belonging to the phylum Firmicutes) and genus Prevotella (bacteria belonging to the phylum Bacteroidetes) were present in higher abundance as compared to the H. pylori- group (Figure 5C). The species of the Bifidobacterium that vary between the H. pylori+ and the H. pylori- groups are found to be B. longum (Kruskal–Wallis, p = 0.009), B. adolescentis (p = 0.03), and B. bividum (p = 0.004), although some of the species remained unidentified at this point of the analyses (Figure 5D). Similarly, for the Bacteroides and Prevotella, the species that could be identified by the 16S rRNA gene analyses are B. plebeius (p = 0.05), B. uniformis (p = 0.04), and P. copri (p = 0.003), respectively, while it was noticed that other species were also present but were not identifiable by V3–V4 regions of the 16S rRNA gene sequence analyses (Figure 5D). The species of the genus Dialister (p = 0.05) also could not be identified by this analysis (Figure 5D).

Since our analyses pointed out differences in the composition of the gut microbiota between the H. pylori+ and the H. pylori- patients, our next aim was to find the distinctiveness in the gut microbiota of the H. pylori+ patients with severe gastric disorders. Therefore, we compared H. pylori+ patients with gastric cancer or gastric ulcer (CA/GU-Hp+) with patients having milder clinical outcomes like non-ulcer dyspepsia or gastritis with H. pylori infection (NUD/GAS-Hp+), non-ulcer dyspepsia, or gastritis without H. pylori infection (NUD/GAS-Hp−) and gastroesophageal reflux disease without H. pylori infection (GERD-Hp−). The metadata related to these four patient groups are given in Supplementary Table S10. It was noticed that the CA/GU-Hp+ group has remarkably low relative abundance of bacteria belonging to the phylum Actinobacteria (Figure 6A). Further analyses revealed that the most significant uniqueness of the CA/GU-Hp+ patients is the low relative abundance of the genus Bifidobacterium (under the phylum Actinobacteria) (Figure 6B). The abundance of Bifidobacterium in the CA/GU-Hp+ group was found to be significantly lower than the GERD-Hp− (p = 0.0181), NUD/GAS-Hp− (p = 0.0117), and NUD/GAS-Hp+ (p = 0.0229) groups (Figure 6C). This finding was further confirmed by heat map (Supplementary Figure S9). Further species level analysis revealed that several Bifidobacterium species like B. adolescentis (p = 0.005 with respect to NUD/GAS-Hp−; p = 0.003 with respect to NUD/GAS-Hp+; p = 0.009 with respect to GERD-Hp−), B. longum (p = 0.002 with respect to NUD/GAS-Hp−; p = 0.008 with respect to NUD/GAS-Hp+; p = 0.02 with respect to GERD-Hp−), and B. bifidum (p = 0.01 with respect to NUD/GAS-Hp−; p = 0.01 with respect to GERD-Hp−) were present at significantly lower relative abundance specifically for the CA/GU-Hp+ group compared with the other groups (Figure 6D). The CA/GU-Hp+ group also have high Oscillospira (p = 0.007 with respect to NUD/GAS-Hp−; p = 0.002 with respect to NUD/GAS-Hp+; p = 0.02 with respect to GERD-Hp−). However, we noticed that 16S rRNA gene analyses were not able to identify all Bifidobacterium species that are present at high abundance in the gut of NUD/GAS-Hp+, NUD/GAS- Hp−, and GERD-Hp− patients but were present at a significantly lower abundance in the gut of CA/GU-Hp+ patients (Figure 6D).

The 16S rRNA gene analyses could not resolve all Bifidobacterium species. Therefore, we decided to identify them by whole genome metagenome sequencing. For this experiment, we have chosen the Oxford Nanopore technology in GridION X5 platform for its longer read length. A total of six patients were chosen. Three of them are CA/GU-Hp+ and three are NUD/GAS-Hp+. The samples are age and sex matched and all individuals are H. pylori+ to avoid any bias. Details of the metadata are given in Supplementary Table S11. The Nanopore read statistics for each sample is given in Supplementary Table S12. The abundance of each domain in each sample is shown in Supplementary Table S13. Our analysis has identified a total of eight Bifidobacterium species, which were remarkably different between the two groups and for each age- and sex-matched pairs (Table 4). Seven (B. adolescentis, B. bifidum, B. breve, B. longum, B. moukalabense, B. pseudocatenulatum, and B. reuteri) of the eight Bifidobacterium species were present at a very low abundance in the intestine of the CA/GU-Hp + individuals than the corresponding age- and sex-matched NUD/GAS-Hp+ individuals (Table 4). The lower abundance of the Bifidobacterium in CA/GU-Hp+ individuals compared with the NUD/GAS-Hp+ individuals is also confirmed with regular PCR (Supplementary Figure 10) and quantitative PCR (Supplementary Figure 11 and Supplementary Table S14).