This study included 23 children (5∼14 years old), who exhibited gastrointestinal symptoms, between January 2018 and August 2018, suggestive of peptic ulcer disease, inclusive of recurrent abdominal discomfort and pain, and dyspepsia. The children, composed of 17 (73.91%) boys and 6 (26.09%) girls, were admitted to the Children’s Hospital of Zhejiang University School of Medicine. The mean age of 23 children was 10.84 ± 3.67 years. Exclusion criteria included a history of acute onset of symptoms, the use of antibiotics, antacids, H₂ receptor antagonist, proton-pump inhibitor (PPI), bismuth-containing compounds, or nonsteroidal anti-inflammatory drugs (NSAID) in the last 4 weeks. The study protocol was approved by the Medical Ethics Committee in the Children’s Hospital of Zhejiang University School of Medicine (2018-IRB-004). Written informed consent was obtained from legal representatives of the children who participated in the study.

Patients underwent esophagogastroduodenoscopy at the Children’s Hospital of Zhejiang University School of Medicine. A total of four pieces of mucosa biopsies used for rapid urease test (RUT), culture, histology, and gastric microbiota studies were performed in the gastric antrum, respectively. Endoscopic findings were also recorded. Gastric mucosa biopsy samples taken from the antrum were preserved in the brain–heart infusion broth (Oxoid, Dardilly, France) with 5% glycerin and sent to the laboratory of Hangzhou Zhiyuan Medical Inspection Institute. The other biopsies were frozen at −80°C until DNA extraction. The homogenate of stomach biopsy specimens was inoculated onto Columbia agar plates (Oxoid) supplemented with 5 % fresh defibrinated sheep blood and kept under microaerophilic conditions (5% O₂, 10% CO₂, and 85% N₂) at 37°C for 3 days. Colonies displaying typical H. pylori morphology were selected and identified by Gram staining and urease, oxidase, and catalase activity test. H. pylori infection diagnosis was determined using the diagnostic criteria from the study “Consensus on diagnosis and treatment of H. pylori infection in children” published in the Chinese Journal of Pediatrics (Subspecialty Group of Gastroenterology, the Society of Pediatrics, Chinese Medical Association, 2015). Based on that study, one of the following four diagnostic criteria can be used to diagnose H. pylori infection: 1. positive results from gastric H. pylori bacterial culture; 2. positive results from the pathological examination of gastric mucosa biopsy and RUT; 3. in the event of inconsistencies arising between pathological examination of the gastric mucosa and RUT results, non-invasive detection, such as ¹³C urea breath test (UBT) or stool antigen test (SAT) can be performed; 4. positive results from either pathological histology of the gastric mucosa or RUT in the event of peptic ulcer bleeding.

Community microbial genomic DNA was extracted from each biopsy sample using a DNA MiniPrep kit (AXYGEN, Suzhou, China) (Bag et al., 2016). Briefly, the biopsy samples were lysed by incubating the sample in ATL lysis buffer with proteinase K overnight at 56°C and following mechanical lysis with Fastprep instrument (MP Biomedicals, Carlsbad, CA, United States) for 1 min at the level of 6.0 m/s, purified with spin columns, and eluted with 400 μl of buffer AE. The quality and quantity of DNA were measured with Nanodrop (Thermo Fisher Scientific, MA, United States). The extracted DNA was stored at -80°C before use.

The V3-V4 hyper-variable regions of the 16S rRNA gene were amplified using a universal primer set (341F: CCTACGGGNGGCWGCAG, 785R: GACTACHVGGGTATCT AATCC) with a barcode. All the template DNAs were normalized to the same concentration. PCR was carried out under conditions described by Caporaso et al. (2011). High-fidelity DNA polymerase: TaKaRa EX Taq was used in PCR. PCR products were separated by electrophoresis in 2% agarose gels, purified with a QIAGEN Gel Extraction Kit (QIAGEN, Germany), and pooled at equal concentrations. Sequencing libraries were generated using a TruSeqR^® DNA PCR-Free Sample Preparation Kit (Illumina, United States) following the manufacturer’s recommendations, and index codes were added. Library quality was assessed on the Qubit@ 2.0 Fluorometer (Thermo Scientific) and the Agilent Bioanalyzer 2100 system. The library was sequenced on an IlluminaHiSeq 2500 platform (250-bp paired-end reads) at Novogene Bioinformatics Technology Co., Ltd. (Beijing, China).

Barcodes as well as forward and reverse primer sequences were removed and raw sequences were analyzed using the Quantitative Insight into Microbial Ecology (QIIME), version 1.9 (Caporaso et al., 2010). Based on the distribution characteristics of low-quality scores of MiSeq sequencing data, quality control of original data was executed. Chimera sequences were removed using the UCHIME algorithm to yield clean tags for further analysis (Edgar et al., 2011). Sequence analysis was performed using UPARSE pipeline version 7.0.1001 (Edgar, 2013), sequences were clustered into operational taxonomic units (OTU) assuming 97% similarity. Representative sequences for each OTU were screened for further annotation. Ribosomal database project (RDP) Classifier (Version 2.2) was used to annotate taxonomic information for each representative sequence based on the Green genes 97% reference data set (McDonald et al., 2012). OTU abundance information was normalized using a standard sequence number corresponding to the sample with the fewest sequences. Subsequent analyses of diversity were performed based on this output-normalized data using QIIME.

Count data were presented as ratio or proportion, measurement data were presented as means ± SD (X ± SD). Differences were determined by t-test, chi-square test, and Kruskal-Wallis test between two groups. Statistically significant differences in the relative abundance of taxa were performed using linear discriminant analysis effect size (LEfSe). A significant alpha at 0.05 and an effect size threshold of 4 were used for all the biomarkers discussed in this study. For the co-occurrence network analysis, OTUs with relative abundance > 0.05% of the microbiome were subjected to Spearman correlation analysis of their occurrence patterns (Barberán et al., 2012; Cardinale et al., 2015), using the non-rarified sequence data. Only correlations [P < 0.05 after false discovery rate (FDR) correction] were visualized through network analysis with software Cytoscape 3.6 (Shannon et al., 2003). The functional genes of bacterial communities based on the sequencing data were analyzed by Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (Langille et al., 2013). Predicted functional genes were categorized into Clusters of Orthologous Groups (COGs) and Kyoto Encyclopedia of Genes and Genome (KEGG) orthology (KO), and compared across patient groups using Statistical Analysis of Metagenomics Profiles (STAMP) (version 2.1.3) (Parks et al., 2014). All the tests of significance were on two sides, and P < 0.05 was considered statistically significant.

In this study, 23 cases with duodenal ulcer were enrolled in the study group, including 15 cases with H. pylori infection, other 8 cases without H. pylori infection (Supplementary Table 1). Based on test results and endoscopic findings, patients were divided into two groups: 1. H. pylori-positive ulcer group (n = 15) [results consistent with H. pylori infection diagnosis, endoscopic examination showing duodenal bulbar ulcers, histopathological examination showed various degrees of inflammation]; 2. H. pylori-negative ulcer group (n = 8) [no H. pylori infection, but endoscopic examination showing duodenal bulbar ulcers, histopathological examination showed various degrees of inflammation]. The clinical information of patients including the indication for endoscopy and the main endoscopic finding was presented in Table 1.

Good’s estimator of coverage was nearly 100%, indicating that the identified reads represented the majority of bacterial sequences present in the stomach (Table 2 and Figure 1A). Based on 16SrRNA bacterial gene sequencing, microbiome profiles in the gastric mucosa of 23 individuals were dominated by phyla Proteobacteria, Campylobacterota, Actinobacteria, Bacteroidetes, Cyanobacteria, and Firmicutes. The bacterial compositions among groups were as follows: Proteobacteria (52.69%), Campylobacterota (26.00%), Actinobacteria (1.55%), Bacteroidetes (0.68%), Cyanobacteria (0.36%), and Firmicutes (0.32%) in H. pylori-positive group; Proteobacteria (82.67%), Actinobacteria (4.46%), Cyanobacteria (3.45%), Firmicutes (0.92%), Campylobacterota (0.67%), and Bacteroidetes (0.57%) in H. pylori-negative group.

Alpha diversity reflects community richness and diversity and they were presented by Chao estimator and ACE estimator and by Shannon index, Simpson index, and Heip evenness, respectively (Figures 1B–F). Diversity indices, such as Shannon and Heip evenness, were significantly decreased in H. pylori-positive group, while richness indices, such as ACE was also decreased in H. pylori-positive group. Beta diversity presented the similarity of community structure and was analyzed using principal coordinates analysis (PCoA) by bacterial abundance clustering. Significant similarity difference was found based on the Bray-Curtis distance between H. pylori-positive group and H. pylori-negative group (P = 0.004). PCoA of weighted Unifrac distances did not show statistical differences (P = 0.175). The results of PCoA in the Bray-Curtis and weighted Unifrac distances were shown in Figure 2.

Among the major phyla of gastric microbiota (when mean abundance was greater than 0.1%), clear differences were observed between H. pylori-positive group and H. pylori-negative group. In total, seven phyla differed in abundance between H. pylori-positive group and H. pylori-negative group, with Campylobacterota being more abundant in H. pylori-positive group (P = 0.009), and Actinobacteriota, Gemmatimonadota, Proteobacteria, Verrucomicrobiota, Fusobacteriota and Firmicutes being more abundant in H. pylori-negative group (Figure 3A). At the genus level, we expanded the analysis to include the most abundant genera (when mean abundance was greater than 0.1% in H. pylori-positive group and/or H. pylori-negative group). In total, sixteen genera differed in abundance, with Helicobacter being more abundant in H. pylori-positive group (P = 0.012) and Pseudomonas, Serratia, Mycobacterium, Sphingopyxis, Devosia, Pandoraea, Hydrogenophaga, Caulobacteraceae Unclassified, Rhodococcus, Gemmatimonas, Taonella, Phyllobacterium, Methylobacterium, Methylorubrum and Bosea being more abundant in H. pylori-negative group (P = 0.003–0.022) (Figure 3B).

The different bacterial compositions among groups were analyzed by the Metastata algorithm, and biomarkers (the key bacterial members) were further screened by LEfSe analysis. The linear discriminant analysis (LDA) scores showed that Campylobacterota was enriched in H. pylori-positive group, Proteobacteria and Actinobacteriota were enriched in H. pylori-negative group. At the genus level, Helicobacter was the only distinguishing biomarker in H. pylori-positive group. In H. pylori-negative group, Serratia, Pseudomonas and Undibacterium were significantly increased (Figure 4).

The overall structure of the gastric microbiota is the result of dynamic interactions between community members. To detect the relationship between different members of the gastric microbial communities, we constructed a network of co-occurrence OTU and interrogated the network for modules using weighted gene co-expression network analysis (WGCNA) (Figure 5). The correlation networks formed different bacterial clusters in the two groups, with a more complex network of interactions in H. pylori-positive group than that in H. pylori-negative group. In H. pylori-positive group, the most dominant member, Helicobacter was negatively correlated with Pseudomonas, Serratia, Mycobacterium, Achromobacter, Sphingopyxis, Devosia, Halomonas, Stenotrophomonas, Pandoraea, Brevundimonas and Hydrogenophaga, those genera demonstrated strong positive correlations. However, most of these correlations were no longer significant in H. pylori-negative group, strong negative correlations were formed among Brevundimonas, Sphingopyxis, Achromobacter, Serratia, Pseudomonas, Uncultured, Marivta, Candidatus Aquiluna and Norank.

To infer the metagenome functional content based on the microbial community profiles obtained from the 16S rRNA gene sequences we used PICRUSt (Langille et al., 2013). Overall, the gastric microbial communities present in the duodenal ulcer patients with or without H. pylori infection could be distinguished based on its functions. The predicted KEGG pathways significantly enriched in H. pylori-negative group included carbohydrate metabolism, signal transduction, amino acid metabolism, lipid metabolism, etc. (Table 3, Figure 6, and Supplementary Figure 1).

Studies have shown that nitrate-reducing bacterial species contribute to gastric malignant transformation by increasing intragastric concentrations of nitrite and N-nitroso compounds (Ferreira et al., 2018), we next compared H. pylori-positive group and H. pylori-negative group regarding the microbial functional features involved in those metabolic reactions. The results showed that the functional composition of the total H. pylori-negative microbiota had increased nitrate reductase functions, which promote the reduction of nitrate to nitrite, and nitrite reductase functions, which promote the reduction of nitrite to nitric oxide when compared with that of H. pylori-positive group (Table 4 and Figure 7). Collectively, these data provide evidence that H. pylori colonization reduces a microbial community with genotoxic potential in the gastric mucosa of children with duodenal ulcer.