Gastric microbiome in gastric cancer sequence depicts diverse microbial structures associated with cancer risk and prognosis

Abstract

Objective

Increasing evidence indicated substantial involvement of non-Helicobacter pylori microbiota in gastric tumorigenesis. We aimed to elucidate detailed relationship of microbiome dynamics between two different steps in gastric cancer (GC) such as cancer initiation and progression, and assessed their associations with clinicopathological and molecular changes.

Methods

We systemically characterized gastric microbiome during GC initiation and progression using 944 biopsies from primary GC, non-cancerous gastric mucosa from both GC and non-cancer subjects. The association between specific microbial characteristics and GC risk, prognosis and molecular changes such as TP53 mutation, DNA methylation and telomere shortening were also evaluated.

Results

Microbial α-diversity in the gastric mucosa was decreased in relation to the GC occurrence, while it increased in primary GC tissue. Such paradoxical change was also observed in specific groups of bacteria during GC occurrence and its progression. GC risk-related microbiome was associated with differentiated GC, severe intestinal metaplasia, associated DNA methylation and telomere shortening, while GC tissue-specific microbiome was associated with more aggressive features of GC and TP53 mutation status.

Conclusions

Our findings suggested the different role of non-Helicobacter pylori microbiota in GC initiation and progression steps.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07046-5.

Introduction

Gastric cancer (GC) is ranked fifth regarding the incidence rate among human malignancies and ranks fourth in mortality [1]. Despite advance in strategy for early detection, many patients still have advanced disease at diagnosis. Since the prognosis of patients with advanced tumor is poor [2, 3], deeper understanding of the pathogenesis that drives the transition from normal gastric mucosa to GC may provide new therapeutic alternatives, improving diagnostic and prognostic tools.

Helicobacter pylori (H. pylori) infection is well accepted as a critical event in the development GC and its pre-malignant conditions [4–6]. H. pylori contributes to gastric carcinogenesis by producing persistent acute-on-chronic inflammation in the gastric epithelium that lead to host somatic changes such as genetic and epigenetic anomalies [7]. However, there is considerable variations among H. pylori- infected patients regarding their degree of gastric inflammation and atrophy, which relates to their diverse GC risks [4]. Moreover, severe gastric atrophy and intestinal metaplasia, associated with strong risk for GC [4, 8] is rather associated with decreased or disappearance of H. pylori [9] but such mucosa is associated with severe host somatic changes such as DNA methylation [10]. This issue suggests some other factors may interact with H. pylori to drive gastric tumorigenesis.

Growing evidence supporting the link between non-H. pylori microbiota and gastric tumorigenesis [11–14]. A human case-controlled studies including patients with GC compared with non-cancer patients found gastric microbiome with potential cancer-promoting activities [13, 14], while other study identified GC tumor-enriched microbiome in relation to patient’s prognosis and therapy efficacy [15, 16]. These findings suggest that non-H. pylori bacteria would contribute to developing and progression of GC by synergistic or competitive effects with H. pylori or independently. However, there is a lack of detailed systemic examination of gastric microbiome in both non-neoplastic and neoplastic tissue during H. pylori-related GC sequences. Therefore, it is unclear whether GC predisposing microbiome would also accelerate GC tumor progression. Moreover, potential interaction between non-H. pylori microbiota and GC-related host somatic change have not been evaluated intensively.

The present study investigated microbiome signatures associated with GC development and progression and evaluate their clinical significance and molecular features. We characterized gastric microbiome of GC tumor tissue, paired non-neoplastic mucosa and gastric mucosa and gastric mucosa from cancer-free subjects. The association between specific microbial characteristics and GC risk, prognosis and molecular changes such as TP53 mutation, DNA methylation and telomere shortening were also evaluated.

Materials and methods

Patients and clinical samples

We examined total of 944 genomic DNA samples from gastric tissue. Those samples consisted of DNAs of primary GC (GCT) from 323 patients in addition to their paired normal mucosa (GCN) from 219 patients, and normal gastric mucosa from 402 cancer-free subjects (Supplementary Table 1). All DNA were extracted from frozen specimens by the endoscopic biopsies. All GC patients attended Fujita Health University Hospital and Kansai Medical University Hospital from January 2005 to December 2020 for the treatment of GC. Non-cancer subjects visited to the same hospitals to perform upper gastroscopy for their health check, secondary complete check-up of GC following to barium X ray examination, or for the complaint of abdominal discomfort. The samples from cancer-free subjects consisted of those from peptic ulcer disease (Ul, n = 103) and non-ulcer (Non-ul, n = 299) subjects. Of all participants, normal-appearing gastric mucosa were obtained from the greater curvature of the antrum. All GC was diagnosed histologically, and additional clinical information was also obtained according to the Japanese classification of gastric carcinoma [17]. Overall survival (OS), defined as the time from gastrectomy, or start of initial administration of chemotherapy to the date of cancer related death was determined. Progression free survival (PFS), defined as the time from gastrectomy, or start of initial administration of chemotherapy to the date of tumor progression or cancer related death was also determined. None of GC patients had been treated with chemotherapy or radiation before sample collection. Histological examination of the primary GC tissue confirmed that all biopsies contained more than 70% of cancer cells.

Both the GC and the cancer-free groups did not include patients who had previous history of H. pylori eradication therapy and gastric surgery and severe systemic diseases, current or previous history of malignancies including GC, received nonsteroidal anti-inflammatory drugs and antibiotics. This cohort was partly recruited from our previous studies investigating host somatic changes associated with GC risk and prognosis [18–21]. The Ethics Committee of Fujita Health University and Kansai Medical University approved the protocol (ID: HM18-094 and 2021046) and written informed consent was obtained from all of the subjects. All authors had access to the study data and had reviewed and approved the final manuscript.

16 S rRNA gene amplicon sequencing

Using the DNA from biopsy specimen, the 16S rRNA gene was amplified using primers 341F (5′- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′) and 805R (5′GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′), targeting the V3–V4 hypervariable regions. At the 5′ end of the primers, the universal sequence of the Illumina adapter was included. Samples underwent denaturation at 95 °C for 3 min; 35 cycles at 95 °C for 30 s, 55 °C for 40 s and 72 °C for 40 s; and a final elongation at 72 °C for 5 min. The Polymerase Chain Reaction (PCR) products were purified and the Nextera XT Index kit (Illumina Inc., San Diego, CA) was added, followed by an additional 12 cycles of PCR. Concentration of all DNA libraries were normalized and then pooled together for the next-generation sequencing using a MiSeq platform (Illumina Inc.) with MiSeq Reagent Kit version 3 (2 × 300 bp Paired End Reads, Illumina Inc.).

16 S rRNA data analysis

16 S rRNA sequence data analyses were performed using Quantitative Insights into Microbial Ecology (QIIME) pipeline (Ver.2023.2). Primer-trimmed sequences were clustered to amplicon sequence variants using the q2-dada2 plugin and sequences with anonymous bases and chimera were filtered. Bacterial alpha diversity measures were calculated using the amplicon sequence variant table in QIIME2 at the sampling depth of 1500. Taxonomic information for each ASV was aligned to the reference sequences in the Greengenes database and the mean relative abundance (percentage among all reads) was generated at the phylum and genus levels. Co-abundance network analysis was performed using the MetagenoNets (https://web.rniapps.net/metagenonets/) based on compositional data through Lasso (CCLasso) correlation algorithms. Microbial functions were predicted using PICRUSt2 in the Kyoto Encyclopedia of Genes and Genomes (KEGG, https://www.kegg.jp/) database.

Assessment of H. pylori infection status, assessment of histological gastritis and Epstein–Barr virus (EBV) status

H. pylori infection status was assessed by histological analysis of both the biopsy specimens from the greater curvature of the gastric antrum and upper corpus using antibodies to avoid false-negative results as much as possible [4]. Using biopsy specimens from the antrum, the extent of neutrophil infiltration mononuclear cell infiltration, atrophy, and metaplasia was assessed according to the updated Sydney system [22], with each factor being scored from 0 (normal) to 3 (marked). This assessment was performed by an investigator who was blinded to the participant’s clinical information. For GC patients, EBV status was screened by real-time quantitative PCRs with dual-labeled fluorogenic hybridization probes (one toward the BamHI-W region; and the other toward the EBNA-1 region) [23]. EBV‐positive cases, defined as having positive results from both assays were evaluated by the in situ hybridization (ISH) for Epstein − Barr encoding region. The result confirmed that all these cases were also positive for ISH.

Molecular study

For GC tissue, mutation status of all coding regions of TP53 gene (exon2 to exon11) were examined using the Sanger sequencing. PCR and sequence primers are shown in the Supplementary Table 2. The sequence chromatograms were evaluated with DNA Dynamo Sequence Analysis Software (Blue Tractor Software, Llanfairfachan, Wales, UK). All mutations were functionally evaluated using the SIFT (sorting intolerant from tolerant) analysis.

DNA methylation, an early phenomenon of gastric tumorigenesis [24, 25], was performed by the bisulfite pyrosequencing to quantify promoter CpG islands of five genes (IGF2, SLC16A12, SOX11, P2RX7 and MYOD1). Selection of these genes was based on the high frequency of methylation in the stomach with H. pylori–infection [10, 26]. Bisulfite-treated genomic DNA was used to evaluate the methylation status by bisulfite pyrosequencing. Pyrosequencing was performed by using a PSQ24 system and the PyroGold Reagent Kit (Qiagen). The PyroMark Q24 Software (Qiagen) was used for processing the results. We created methylation positive control DNA using SssI methylase treated DNA. We also created methylation negative control using the Whole Genome Amplification. The primers used for pyrosequencing are listed in the Supplementary Table 2. Relative telomere length, an indicator for rapid cell turnover and oxidative stress [27], was measured as abundance of telomeric template vs. a single-copy gene (T/S) by quantitative real-time PCR [28]. For the quantitative real-time PCR, the iTaq SYBR Green Supermix (Bio-Rad) and ABI Prism 7900HT Real-Time PCR System (Applied Biosystems) were used. Information about primer sequences for telomeres and single-copy genes (h-globin) are listed in the Supplementary Table 2. Detailed protocol of experiment is also available in our previous study [29].

Statistical analysis

Continuous values between two different sample groups were assessed using the Man–Whitney U test. Categorical values between two different sample groups were assessed using the Fisher’s exact test or the logistic regression analysis with adjustment for confounding factors. Correlation of continuous values between two groups were assessed using the Spearman correlation analysis. Discriminatory potential of different sample groups based on the microbial status was evaluated using the receiver operating characteristic (ROC) curve as well as area under curve (AUC). Survival analysis were performed using the Kaplan-Meier method and log-rank test to compare survival distribution. The p value < 0.05 was considered as statistically significant.

Results

Overview of the study population

We performed 16SrRNA gene amplicon sequencing of 944 gastric samples including primary GC (GCT: n = 323), their paired normal mucosae (GCN: n = 219) and normal gastric mucosae from cancer-free subjects (n = 402). The 402 samples from cancer-free subjects consisted of patients from peptic ulcer disease (Ul: n = 103) and non-ulcer subjects (non-ul: n = 299). The association between microbiome status and histological degree of gastritis, DNA methylation and telomere length were also examined. The study flowchart and patient’s characteristics are listed in Fig. 1a and Supplementary Table 1, respectively. After quality processing, we filtered samples from low read quality less than 2000 filtered reads (n = 10: Supplementary Table 3). Taxonomic information for each amplicon sequence variant (ASV) was aligned to the curated Greengenes reference database at the QIIME website.

Fig. 1.

Microbiome landscape of different groups of gastric biopsies. Flow chart showing study overview (a). Alpha diversity measures between Non-ul, Ul, GCN, and GCT groups. Statistical analysis was performed using the Man–Whitney U test (b). Differences in microbiota abundances between Ul, GCN, and GCT groups. Statistical analysis was performed using the Man–Whitney U test (c). Differences in microbiota prevalence between UL, GCN, and GCT groups. Statistical analysis was performed using the logistic regression analysis with adjustment for confounding factors such as age, gender and H. pylori infection status (d).

Microbial composition in the ulcer disease and gastric cancer

First, we compared bacterial diversity in the different groups of samples (Non-ul, Ul, GCN, and GCT groups) using three indicators (Shannon index, Faith’s phylogenetic diversity and Observed features). We found a decrease in alpha-diversity measures in the Ul and GCN groups, relative to the Non-ul group, while it showed paradoxical increase in the GCT group relative to the Ul and GCN groups (Fig. 1b).

We then searched for the relevant taxa responsible for the different groups of samples based on its relative abundance. Ul group was characterized as enrichment of Proteobacteria at the phylum level, while several taxa were also identified in the same group at the genus level including the Helicobacter, most of which are presumably the H. pylori (Fig. 1c and Supplementary Tables 4 and 7). We also identified 2 and 6 phyla, 33 and 55 genera, that were significantly enriched in GCN and GCT groups when compared to the non-Ul and GCN groups, respectively (Fig. 1c and Supplementary Tables 5, 6, 8 and 9). Both GCN and GCT enriched bacteria included previously reported bacteria such as Fusobacteria at the phylum level, and Streptococcus, Prevotella, and Peptostreptococcus, Sphingomonas and Lachnoanaerobaculum at the genus level [13, 14, 16], while it also included unreported taxa such as genus Alloiococcus and Mitsuokella. In addition, the genus Helicobacter being predominant in Ul, was decreased in the GCT compared to the GCN, which is also in line with other studies [15, 16].

To confirm the relevant bacteria in each groups of samples, we also compared the prevalence of significantly increased or decreased taxa identified according to its abundance. Logistic regression model with adjustment for possible confounders demonstrated 10, 49, and 70 genera, significantly associated with the Ul, GCN, and GCT groups, respectively. The GCN group displayed more decreased bacteria (11 increased and 38 decreased) compared to the Non-ul group. In contrast, the GCT group demonstrated more increased bacteria (51 increased and 19 decreased) compared to the Non-ul group (Fig. 1d, Supplementary Tables 10–14.

Discriminatory potential of different sample groups based on the microbial network

Identification of multiple bacteria in specific types of gastric samples suggested the existence of the community-wide virulence properties involved in gastric diseases as reported in other studies [14, 30]. We performed a correlation analysis of increasing and decreasing bacteria in the GCN and GCT groups (Fig.S1). This analysis demonstrated complex co-occurring and co-excluding interaction among both GCN and GCT-associated taxa with several sample-specific hub genera (Albidovulum, Roseburia, Faecalibacterium and Bacteroides for GCN, Brevundimonas and Treponema for GCT, respectively). This supported the synergistic contribution of bacterial network towards diseases which may also serve as potential diagnostic tissue markers. To confirm this, we next combined relevant taxa in each group and calculated the microbial index [14, 30]. The microbial index was defined as the number of increased bacteria + 1 divided by the total number of decreased bacteria detected + 1. The microbial indexes for Ul (Ul index for discriminating Ul from Non-ul), GCN (GCN index for discriminating GCN from Non-ul), and GCT (GCT index for discriminating GCT from GCN) groups were generated, respectively. The GCN and GCT indexes had favorable discriminative ability with area under curve (AUC) 0.81 and 0.87, while discriminative ability of Ul index was limited with AUC 0.68 (Fig. 2a). We then further investigated whether these indexes can discriminate different subtypes of ulcer diseases and gastric cancer such as duodenal and gastric ulcer, differentiated and undifferentiated GC. GCN index seemed to be useful to discriminate differentiated cancers from Non-ul samples with AUC 0.84, while AUC was similar for Ul and GCT indexes in relation to different disease subtypes (Fig. 2b).

Fig. 2.

ROC analysis assessing the discriminative ability of bacterial indexes for predicting different groups of samples (a) and their disease subtypes (b)

GC associated microbiome, clinicopathological factors and molecular anomalies in GC development and progression sequences

GCN and GCT indexes were further examined to investigate whether GC associated microbiome would be associated with clinicopathological factors and molecular anomalies in GC development and progression. Development of advanced histological gastritis accompanied with host somatic anomalies are thought to be an initial step toward GC development [4, 7, 8]. We found significant association between higher GCN index and higher metaplasia score such as 2 (moderate) and 3 (severe), compared to that of 0 (normal). Higher GCN index was also associated with mild degree of neutrophil infiltration (1, mild) compared to the 0 while this association was not observed for scores 2 and 3 (Fig. 3a). DNA methylation accumulation is associated with H. pylori infection and GC risk [7, 10, 24, 25]. Telomere shortening, an indicator of rapid cell turnover, and oxidative stress has also been involved in gastric tumorigenesis [20, 29]. We found significant positive correlation between higher GCN index and DNA methylation of all H. pylori infection-related genes (IGF2, SLC16A12, SOX11, P2RX7 and MYOD1, ref.10, 26;) as well as its mean Z score (Fig.S2, Fig. 3b). Higher GCN index was also significantly associated with telomere shortening (Fig. 3c).

Fig. 3.

Association between GCN index, histological degree of gastritis (a), mean Z score methylation of H. pylori infection associated genes (IGF2, SLC16A12, SOX11, P2RX7, MYOD1) (b) and telomere shortening (c). Statistical analysis was performed using the Man–Whitney U test (a) as well as the Spearman correlation analysis (b) (c)

Regarding the GCT index, we found that higher GCT index was significantly associated with more aggressive features of GC such as invasive cancer (T3 and T4), higher clinical stage, lymph node metastasis and peritoneal dissemination, while an association between higher GCT index and upper anatomical location was also observed (Fig. 4). Since GCT index presented an approximately Gaussian distribution, with over representation of high cases, we set cutoff value of 17 for the definition of higher GCT index and performed survival analysis. The result demonstrated that higher GCT index was significantly associated with worse Overall survival (OS) and Progression free survival (PFS) (Fig. 4). Since somatic changes in the primary GC would be also relevant to the GC progression [19, 21, 31], we also investigated between GCT index and TP53 mutation status, H. pylori-associated DNA methylation and telomere shortening in primary GC tissue. Summary of TP53 mutations detected in our data set was shown in Supplementary Table 15. A significant association between higher GCT index and TP53 mutant was observed (Fig. 4), while no association was found for H. pylori-associated DNA methylation and telomere shortening (Fig.S3 and S4).

Fig. 4.

Association between GCT index, clinicopathological features and prognosis of GC. Survival analysis were performed using the Kaplan-Meier method and log-rank test. All other comparisons were performed using the Man–Whitney U test

Differences of microbiome among GC development and progression sequences

Potential relationship of microbiome dynamics between two different steps in the GC, i.e., cancer initiation and progression would be of our great interest. We searched overlap of GCN and GCT-related taxa (n = 36) and found that most of them (34/36) changed paradoxically in GCN and GCT groups (Fig. 5a). The GCN and GCT indexes were also negatively correlated (Fig. 5b), suggesting that the microbiome structure and function are different between GC initiation and progression. For bacteria showing these opposed dynamics, there were 25 bacteria decreased in the GCN group and increased in the GCT group, and 9 bacteria increased in the GCN group and decreased in the GCT group (Fig. 5a). We then checked mean relative abundances of those 2 groups of bacteria and found that both were also significantly different between the Non-ul and GCT groups (Fig. 5c, d). The mean relative abundance of 25 bacteria decreased in the GCN group relative to the Non-ul group significantly increased in GCT groups in the comparison with the Non-ul group and the specific taxa associated with such change included Bacillus and Leptotrichia GCT groups (Fig. 5c). On the other hand, the mean relative abundance of 9 bacteria increased in the GCN group relative to the Non-ul group also significantly decreased in GCT groups in the comparison with the Non-ul group and the specific taxa associated with such change included Roseburia, Brevundimonas, Ruminococcus and Ruminococcaceae family (Fig. 5d).

Fig. 5.

Paradoxical dynamics of microbiome compositions in gastric cancer development and progression. Overlapping bacteria in GCN and GCT groups and their dynamics (a). Relative abundance of bacteria decreased in GCN group and increased in GCT group (c). Relative abundance of bacteria increased in GCN group and decreased in GCT group (d). Statistical analysis was performed using the Fisher’s exact test (a) and the Man–Whitney U test (b–d)

Functional analysis of microbiome

Finally, we performed predictive functional analysis using PICRUSt2 based on bacterial information related to the ulcer disease and GC. When significant correlations of metabolic pathways and enzymes were determined in relation to Ul, GCN and GCT indexes, most of them negatively correlated with GCN index, while portion of positive correlation considerably increased in relation to the GCT index (Fig. 6a, b). There was no overlap of the top five increased or decreased metabolic pathways related to Ul, GCN and GCT indexes, suggesting that microbiome compositions related to those 3 indexes are functionally different (Fig. 6c). We then searched for specific enzymes associated with those 3 indexes in the Kyoto Encyclopedia of Genes and Genomes (KEGG, https://www.kegg.jp/) database. This analysis demonstrated that superoxide reductase and superoxide dismutase were significantly negatively correlated with the GCN index. We also found that superoxide reductase and MAP kinase were significantly positively correlated with GCT index (Fig. 6d).

Fig. 6.

Predictive functional analysis of gastric microbiota associated with each group of samples. Proportions of metabolic pathways (a) and enzymes (b) correlated with each index. The top five metabolic pathways significantly correlated with each index (c). Specific enzymes found to be associated with GCN and GCT indexes. Statistical analysis was performed using the Spearman correlation analysis (c, d)

Discussion

Microbial structures associated with GC risk and progression

H. pylori has been thought to be a definite pathogen for GC development, while recent studies have demonstrated a significant role of non-Helicobacter pylori microbiota in the GC progression by altering immune and metabolic homeostasis [11, 12, 32]. Several studies have characterized the microbiota profiles of the stomach, associated with GC risk [13, 14, 33]. Other studies have also characterized the tumor enriched bacteria in primary GC tissue and those with clinical features, prognosis, and therapy efficacy [15, 16]. However, there remains a lack of research investigating the detailed relationship between two different categories of microbiome in GC, i.e., microbiome profile associated with GC risk and those associated with GC progression. Our study systemically characterized gastric microbiome during GC initiation and progression using biopsies from primary GC, non-cancerous gastric mucosa from both GC and non-cancer subjects. We showed a decrease of microbial alpha-diversity measures in the gastric mucosa in Ul and GCN groups compared to non-ul group. Instead, alpha-diversity measures paradoxically increased in GCT group compared to GCN group. Previous study showed that lower bacterial alpha diversity measures in the gastric mucosa was associated with H. pylori infection and incidence of GC [13, 14, 33], while studies comparing primary GC and matched non-cancerous mucosa showed rather increased alpha diversity measures in primary GC relative to its non-cancerous mucosa [15, 16]. These findings suggested that decreased α-diversity might be associated with an early microbial change linked to gastric cancer risk, while paradoxical increase of the α-diversity would be relevant to cancer progression process. Former studies suggested paradoxical changes of the bacterial α-diversity between cancer risk and progression, but this phenomenon has not been demonstrated systemically [13–16]. One of the strengths of our study seemed to examine both cancer tissue as well as non-neoplastic mucosa with/without gastric cancer to demonstrate landscape of microbial structures in gastric cancer sequence.

Since specific microbial changes in GCN and GCT groups were characterized as decreased (38 out of 49) and increased taxa (51 out of 70), respectively, changes of α-diversity in the specific groups of samples seems to reflect change in specific microbiome in different groups of tissues. Our findings and other studies suggest that reversal change in bacterial alpha diversity is the phenomena shown in GC initiation and progression.

Abundance screening showed Ul group presented an enrichment of phylum Proteobacteria, and several genera including the Helicobacter. The same approach demonstrated that enriched bacteria in GC patients included previously reported bacteria such as Fusobacteria at the phylum level, and Streptococcus, Prevotella, and Peptostreptococcus, Sphingomonas and Lachnoanaerobaculum at the genus level [13, 14, 16]. In addition, we also found previously unreported bacteria associated with GC patients such as genus Alloiococcus and Mitsuokella. These results supported the possible involvement of previously reported bacteria in the gastric tumorigenesis and also highlighted a contribution of population-specific bacteria in the gastric carcinogenesis that may interact with ethnically diverse host genetic, environmental and lifestyle factors. For example, the gut microbiome of the Japanese is considerably different from those of other populations [34]. Microbiome in the gastric mucosa has also been reported as highly variable even in the Asian population [35]. Therefore, it is possible that some inconsistency shown in our study and other studies may reflect racial difference and patient’s constitution.

Using both abundance and prevalence screenings, we identified multiple taxa associated with Ul, GCN and GCT samples. Moreover, there were complex co-occurring and co-excluding interaction among both GCN and CGT-associated taxa. This indicated the existence of the community-wide virulence properties that is involved in GC as reported in other studies [14, 30]. We demonstrated that the microbial index based on the sample specific bacteria well predicted both GCN and GCT groups with favorable AUC. This support the synergistic contribution of bacterial network associated with GC and combining tissue specific bacteria may also serve as potential diagnostic markers as shown in other studies [14, 36]. On the other hand, the AUC of the Ul index was below 0.7 and it was not considered to be acceptable, suggesting that the ability of Ul index for diagnosing ulcer disease is limited.

GC associated microbiome, clinicopathological factors and molecular anomalies in GC development and progression sequences

We showed that GCN index was closely associated with differentiated GC, more severe intestinal metaplasia, and host somatic changes in the gastric mucosa such as DNA methylation and telomere shortening. Severe intestinal metaplasia has been associated with an increased risk of differentiated GC [4]. Presence of intestinal metaplasia was also associated with DNA methylation and telomere shortening [20, 26]. However, induction of host somatic changes in the gastric mucosa under the H. pylori infection involves complex biological processes that are still not fully understood. Indeed, it has been reported that gastric mucosa with severe gastric atrophy presented rather low frequency of H. pylori infection, while such mucosa was associated with hyper DNA methylation [10]. Hyper DNA methylation was also remained in surrounding mucosa close to gastric cancer after successful H. pylori eradication [37]. It has been proposed that changes that occur in the stomach as a result of chronic H. pylori infection leading to decreased acid secretion allow the successful establishment of a new microbiota that contributes to malignant transformation through maintenance of inflammation and conversion of nitrates into N-nitrosamines [38, 39]. Taken together, our result suggested that non-H. pylori bacteria may have a role in maintaining host genotoxic changes and promote gastric tumorigenesis in the long-term outcome of H. pylori infection. Since the potential role of bacteria maintaining the genotoxic changes has been supported by the antibiological therapy that restore the molecular anomaly. For example, it has been demonstrated that eradication of H. pylori improves DNA methylation anomaly in the gastric mucosa [40]. Our findings also highlight the potential utility of biologic therapy against non-H. pylori bacteria for reducing the risk of GC.

We showed that GCT index, defined as the cancer-specific microbiome in GC was significantly associated with more aggressive features of GC, worse overall and progression-free survivals, suggesting that GCT index can be used as the prognostic marker for GC patients. This observation indicates that different GC subgroups may also have different microbiome status. GC is diverse in its prognosis and response to treatment; it is hypothesized that GC is a tumor in which different pathogenic mechanisms lead to various clinicopathological phenotypes. It may be therefore necessary to place greater emphasis on disease heterogeneity. In this context, our data suggest that patients with higher GCT index are associated with more aggressive phenotypes and poor prognosis, that needs careful follow up even after curative treatment.

We also showed that higher GCT index was associated with TP53 mutation positivity. Host somatic changes is deeply involved in cancer progression [19, 21, 31]. Recently, it has been proposed that intratumoral microbiome in GC patients is associated with mismatch repair status [16], suggesting that cancer-specific bacteria in GC define molecular features of GC. In this preliminary phase, we only examined most typical gene, TP53. Potential correlation between microbiome profile and mismatch repair status rases the possibility that studying microbiome status would be useful to learn the mechanisms how the ICI (Immune Checkpoint Inhibitor) provide the response against subset of gastric cancer patients [16]. It is also suggested that the ICI and molecular targeting therapy in cancers could be attributed to the specific molecular subtypes of other cancer types [41, 42]. More extensive correlation including genetic and epigenetic status and tumor-specific microbiome need to be evaluated.

Differences of microbiome among GC development and progression sequences

Potential relationship of microbiome dynamics between two different steps in GC such as cancer initiation and progression would be of our great interest, but it has not been evaluated in detail. There was negative correlation between GCN and GCT indexes. Moreover, majority of specific bacteria associated with GCN and GCT indexes moved paradoxically in cancer development and progression steps. The findings suggested that the microbiome structure and function are different between GC initiation and progression steps. Indeed, abundance of the genus Helicobacter being predominant in Ul, was decreased in the GCT compared to the GCN, which was also in line with other study [15, 16]. No association was also observed between GCT index and DNA methylation and telomere shortening, both are genotoxic changes occurs relatively earlier in gastric tumorigenesis [20, 26], suggesting that cancer predisposing microbiome may shift in the progression of GC to acquire more aggressive biological features. In the comparison of microbiome profile between non-ul and GCT groups, the mean relative abundance of bacteria, being decreased in the GCN group relative to the Non-ul group significantly increased in GCT groups compared to that of Non-ul group. On the other hand, the mean relative abundance of bacteria, being increased in the GCN group relative to the Non-ul group significantly decreased in GCT group compared to that of Non-ul group. These results indicated that the paradoxical dynamics of GC-related bacteria did not change to the normal state during cancer initiation and progression, while the biological consequence of this paradoxical change needs to be further clarified.

Different roles of GCN and GCT related bacteria were further supported from the functional perspective, which showed that the metabolic pathways related to different groups of samples are exclusive in each group. We showed that superoxide reductase and superoxide dismutase, both protect cells from oxidative stress [43, 44], were negatively correlated with the GCN index. One might be speculated that the decrease of both enzymes confers more severe oxidative stress in the gastric epithelium, leading host somatic changes towards carcinogenesis. On the other hand, MAP kinase, involved in cell proliferation in cancer [45] positively correlated with GCT index. Interestingly, there was a rather positive correlation between superoxide reductase and GCT index, suggesting that reduction of oxidative stress might promote cancer cell survival for further cell proliferation and metastasis [46]. Our study performed functional analysis through prediction by the PICRUSt2. Since the specific pathways and enzymes were not identified by the direct metabolite detection, which needs to be noted as the limitation of our study. How the specific pathways and enzymes identified in our study contribute to gastric tumorigenesis would be of interest, but should be validated by well-designed studies in the future.

Study limitations and conclusions

In conclusion, our findings suggested the different role of non-Helicobacter pylori microbiota in GC initiation and progression steps. This study used a large cohort of microbiome samples as well as detailed clinical and molecular information to clarify microbiome signatures in gastric cancer development and progression sequences. A series of our results suggested the potential utility of bacteria- based new biomarker in this cancer type and also highlighted the possibility that biologic therapy such as probiotics and prebiotics as shown in other diseases [47], may also useful for GC and its premalignant condition. However, our study has several limitations. The present study was performed in patients from few institutions in a single cultural and geographical setting (Japan). Therefore, our results need to be further confirmed in the diverse populations. Examined molecular information in our data set focused on somatic anomalies such as DNA methylation, telomere length and mutation status. Cancer occurrence as well as its prognosis would be also influenced by the environmental and lifestyle factors and host genetic status [48]. However, these issues were not evaluated in detail other than age, gender and H. pylori infection status. We examined microbiome status in the gastric tissue and correlated with the risk and prognosis of GC, while information of non-target organ such as inflammatory parameters in the blood would also reflect the pathological conditions of several diseases [49], but this issue could not be examined as well . We used 16 S rRNA gene amplicon sequencing for characterize bacterial community which may not have enough resolution to identify specific taxa associated with different types of samples. More comprehensive approaches such as shotgun metagenomic sequencing may provide more detailed information about microbiome. Finally, our findings mainly demonstrated correlative analyses without experimental validation of microbial mechanisms. Therefore, how the specific bacteria, pathways and enzymes identified in our study contribute to gastric tumorigenesis should be validated by well-designed study using in vivo and ex vivo systems. On the other hand, these limitations would be of interest with respect to risk stratification and understanding of carcinogenic mechanism in the future study.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

T.TAH., TSU.S., TAK., and J.Y. designed the study, the main conceptual ideas, and the proof outline, performed experiments and analyzed the data and wrote the manuscript.

S.K., N.N., Y.T. aided in interpreting the results and worked on the manuscript. T.TOM., Y.H, T.F., and M.N. supervised the project and proof the manuscript. All authors discussed the results and commented on the manuscript.

Funding

This work was supported by JSPS KAKENHI Grant-in-Aid for Scientists (C) (Grant Number　22K08089).

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The sequencing data generated by this study is available through the DNA Data Bank of Japan (DDBJ): BioProject Submission ID: PRJDB20660; BioSample Submission ID: SSUB033686; DRA accession ID: DRA021350-21354 (https://ddbj.nig.ac.jp/public/ddbj_database/dra). The data that support the findings of this study are also available from the corresponding author upon reasonable request.

Declarations

Consent for publication

Written informed consent was obtained from the all the participants for publication of the research findings and its related clinical information.

Declaration of Generative AI and AI-assisted technologies in the writingprocess

The authors declare that they did not use generative AI and AI-assisted technologies in the writing process of this manuscript.

Competing interests

The authors declare that they have no conflicts of interest.