Microbial Dynamics across Molecular Subtypes and Prognostic Significance of Lactobacillus in Gastric Cancer

Soo Kyung Nam; Juhyeong Park; Sujin Oh; Yoonjin Kwak; Cheol Min Shin; Kyoung Un Park; Nak-Jung Kwon; Seong-Ho Kong; Do Joong Park; Hyuk-Joon Lee; Han-Kwang Yang; Hye Seung Lee

doi:10.4143/crt.2025.449

Articles

Page Path: HOME > Cancer Res Treat > Volume 58(3); 2026 > Article

Original Article Gastrointestinal cancer Microbial Dynamics across Molecular Subtypes and Prognostic Significance of Lactobacillus in Gastric Cancer: Soo Kyung Nam^1,2, Juhyeong Park³, Sujin Oh⁴, Yoonjin Kwak⁵, Cheol Min Shin⁶, Kyoung Un Park⁴, Nak-Jung Kwon⁷, Seong-Ho Kong⁸, Do Joong Park^2,8, Hyuk-Joon Lee^2,8, Han-Kwang Yang^2,8, Hye Seung Lee^1,2,5; Cancer Research and Treatment : Official Journal of Korean Cancer Association 2026;58(3):824-845.
DOI: https://doi.org/10.4143/crt.2025.449
Published online: July 17, 2025

¹Interdisciplinary Program in Cancer Biology, Seoul National University College of Medicine, Seoul, Korea

²Cancer Research Institute, Seoul National University College of Medicine, Seoul, Korea

³Seoul National University College of Medicine, Seoul, Korea

⁴Department of Laboratory Medicine, Seoul National University Bundang Hospital, Seoul National University College of Medicine, Seongnam, Korea

⁵Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea

⁶Department of Internal Medicine, Seoul National University Bundang Hospital, Seoul National University College of Medicine, Seongnam, Korea

⁷Macrogen Inc., Seoul, Korea

⁸Department of Surgery, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Korea

Correspondence: Hye Seung Lee, Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, 103 Daehak-ro, Jongno-gu, Seoul 03080, Korea
Tel: 82-2-740-8269 E-mail: hye2@snu.ac.kr

• Received: April 25, 2025 • Accepted: July 16, 2025

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

2,626 Views
105 Download
1 Web of Science

prev next

Full Article

Download PDF

Abstract

Purpose
Recent studies have revealed a diverse gastric microbiota beyond Helicobacter pylori, suggesting a role in gastric cancer (GC). We aimed to investigate the composition and characteristics of the microbiota in GC and non-cancerous gastric mucosa (NC), with a particular focus on their relationship to molecular subtypes.
Materials and Methods
We conducted 16S rRNA sequencing and whole transcriptomic analysis on fresh-frozen GC and NC tissue samples from 192 GC patients, as well as saliva samples from 12 GC patients and 18 healthy individuals. Microsatellite instability (MSI), Epstein-Barr virus (EBV) in situ hybridization, and immunohistochemistry for p53 and E-cadherin were used to define molecular subtypes.
Results
GC tissues exhibited significantly higher diversity compared to matched NC tissues, with microbial profiles marked by decreased Helicobacter and increased Streptococcus, Prevotella, and Lactobacillus. Saliva samples predominantly contained oral bacteria and exhibited distinct microbial profiles from gastric tissues. In GC tissue, Helicobacter abundance was negatively correlated with key immune checkpoint genes (CTLA-4, PDCD1, CD274, and LAG3), whereas Prevotella, Streptococcus, and Fusobacterium were positively correlated. MSI-high and EBV-positive subtypes showed lower levels of Helicobacter but higher levels of Lactobacillus, Prevotella, and Streptococcus compared to the epithelial-mesenchymal transition–like subtype. Notably, within MSI-high GC, a subgroup characterized by Lactobacillus-enriched and otherwise microbiota-depleted profiles was significantly associated with poorer overall and disease-free survival.
Conclusion
These findings underscore distinct microbial patterns across GC molecular subtypes, suggesting potential biomarkers for GC diagnosis and treatment.
Key words: Stomach neoplasms, Microbiota, Molecular subtype, Immune microenvironment, Lactobacillus

Introduction

Numerous micro-organisms, including bacteria, viruses, or fungi, have been found to coexist in the human body sites, including oral cavity, gastrointestinal tract, lung, and skin [1]. Increasing attention has been paid to the role of microbial communities in cancer, as several studies have suggested that microbial dysbiosis may contribute to the development or progression of various cancers, including colorectal, breast, pancreatic, and lung cancers [2-5]. For gastric cancer (GC), infection with Helicobacter pylori has been recognized as a major risk factor, and the International Agency for Research on Cancer (IARC) of the World Health Organization classified H. pylori as group I carcinogen in 1994 [6]. However, traditional investigations on the gastric microbiota were limited because of difficulties in culturing it, leading to the assumption that the microbial diversity other than H. pylori in the stomach is minimal [7]. The recent advent of next generation sequencing [8] has made it possible to identify microbes without culture, revealing the presence of numerous microbiotas in the stomach, including Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and non–Helicobacter pylori Proteobacteria beyond H. pylori [9]. In addition, dysbiosis of the gastric microbiome is suggested to occur during the progression of gastritis to GC [10].

Immunotherapy, which activates the body’s immune cells to target and destroy cancer cells, has emerged as a groundbreaking approach to cancer treatment and has been approved for various cancer types [11]. The microbiome interacts with various immune cells in the gut and other sites of the human body, and has been suggested to modulate tumor immune responses and influence the efficacy of immunotherapy [12]. Specifically, the relationships between the gut microbiome and M2 macrophages or CD8-positive lymphocytes in colon cancer [13,14], natural killer T cells in liver cancer [15], and tumor-infiltrating lymphocytes (TILs) in breast cancer [16] have been investigated and demonstrated. However, research on the relationship between the gut microbiome and host immune responses in GC remains limited, and furthermore investigations of the microbial community within the tumor microenvironment of GC are rare.

GC exhibits highly heterogeneous clinical and pathological characteristics. The molecular classification of GC was proposed by The Cancer Genome Atlas (TCGA) [17], confirming distinct features associated with each subtype, including immune microenvironment. Consistent with the results from TCGA, we identified that Epstein-Barr virus–positive (EBV+) or microsatellite instability–high (MSI-H) GC were significantly correlated with programmed death ligand 1 (PD-L1)+/CD8-high subtype according to the tumor immune microenvironment subtypes classified by the co-assessment of PD-L1 expression and CD8-positive TIL densities [18]. Additionally, the expression of immune checkpoint receptors, including programmed cell death-1, lymphocyte activation gene 3 (LAG3), and T-cell immunoglobulin and mucin domain-containing-3, was associated with these specific (EBV+ and MSI-H) molecular subtypes [19]. Therefore, the molecular subtypes need to be considered in evaluating the immune microenvironment of GC.

In this study, we performed 16S rRNA sequencing using GC and paired non-cancerous gastric mucosa (NC) from 192 GC patients to elucidate the distinct characteristics of microbiome in GC patients driven by microbial imbalance, and conducted whole transcriptome sequencing (WTS) to evaluate their relationship with molecular features and immune microenvironment of GC. In addition, 16S rRNA sequencing using saliva samples from 12 GC patients and 18 healthy persons was performed to reveal and characterize oral microbiome in GC patients compared with healthy persons.

Materials and Methods

1. Study design and sample collection

We collected GC and paired NC tissues from 178 patients who underwent radical gastrectomy at Seoul National University Hospital (SNUH). The samples were stored in the Cancer Tissue Bank of SNUH and provided for this study. Additionally, GC and paired NC tissue and saliva samples, were collected from 14 GC patients, and saliva samples were also collected from 18 healthy individuals at Seoul National University Bundang Hospital (SNUBH) (S1 Fig.). To minimize microbial contamination, gastrectomy tissue samples were collected under sterile conditions during the surgical operation. Separate sterilized instruments were used for each sample, and the outer mucosal surfaces were carefully trimmed to avoid contamination from luminal contents. Samples were immediately snap-frozen in liquid nitrogen and subsequently stored in a liquid nitrogen tank (–196℃). Clinicopathologic information, including age, sex, cancer stage using the American Joint Committee on Cancer staging system (8th edition), and histologic type, was collected from electronic medical records and pathologic reports. Overall survival (OS) was recorded from the date of surgery to the date of death.

2. 16S rRNA sequencing

GC and paired NC tissue samples from 192 GC patients and saliva samples from 12 GC patients and 18 healthy individuals were processed according to the Illumina 16S Metagenomic Sequencing Library protocol to amplify the V3 and V4 regions (primers 519-806R). The barcoded fusion primer sequences used for polymerase chain reaction (PCR) amplifications were as follows: 519F, 5′-CCTACGGGNGGCWGCAG-3′; 806R, 5′-GACTACHVGGGTATCTAATCC-3′. The final purified product was quantified using quantitative PCR employing the KAPA Library Quantification kit for Illumina Sequencing platforms (Roche) according to the manufacturer’s protocol, and its quality was assessed using the LabChip GX HT DNA High Sensitivity Kit (PerkinElmer). Paired-end (2×300 bp) sequencing was performed using the MiSeq platform (Illumina) (Macrogen Inc.).

3. Whole transcriptome sequencing

Fresh frozen GC and paired NC tissue samples from 178 GC patients in the Cancer Tissue Bank of SNUH were subjected to WTS (Macrogen, Inc.). Total RNA was extracted using the RNeasy formalin-fixed paraffin-embedded (FFPE) kit (Qiagen) according to the manufacturer’s instructions. RNA integrity was assessed using a Bioanalyzer (Agilent), and tumor RNAs with RNA integrity number ≥ 6 were subjected to RNA-seq. RNA-seq libraries were generated using the TruSeq RNA Sample Preparation Kit (Illumina). mRNA was enriched with poly T oligo-attached magnetic beads, followed by mRNA fragmentation by acoustic shearing. First-strand cDNA was synthesized using reverse transcriptase and random hexamers. Second-strand cDNA was synthesized using DNA polymerase I and RNase H. Subsequently, the cDNA was subjected to adapter ligation and enriched using PCR to prepare a cDNA library, which was sequenced using a HiSeq 2000 (Illumina).

4. Data visualization and statistical analysis

Statistical analyses and visualization were performed using R software ver. 4.1.2 (R Development Core Team). A p-value < 0.05 was considered statistically significant. Detailed descriptions of the pre-processing steps and various analysis methods for 16S rRNA sequencing data and analyses using WTS data, are provided in Supplementary Methods.

5. Tissue microarray construction and immunohistochemistry

Tissue microarray (TMA) blocks were constructed using FFPE tissues (Superbiochips Laboratories). They were made from a representative 2 mm core selected from the tumor for each case [20]. Immunohistochemistry for p53 (1:1,000, DAKO) and E-cadherin (1:800, BD Biosciences) was performed using a Benchmark XT autostainer (Ventana Medical Systems). For E-cadherin, 30% or more of the tumor cells showing a loss of membranous staining or aberrant cytoplasmic staining were considered to have altered expression. Strong nuclear staining or a loss of p53 expression was defined as altered p53 expression.

6. Microsatellite instability

Microsatellite instability (MSI) status was assessed at five loci (BAT-26, BAT-25, D5S346, D17S250, and D2S123) using the National Cancer Institute panel. DNA was extracted from macrodissected tumor FFPE tissues and paired normal tissues. After PCR amplification, the products were analyzed using an ABI 3730 DNA Analyzer (Applied Biosystems). MSI-H was defined as two or more unstable markers, MSI-low as one unstable marker, and microsatellite stable as no unstable markers [18].

7. EBV in situ hybridization

EBV was detected using EBV-encoded RNA in situ hybridization. TMA slides were digested and hybridized with the INFORM EBER probe (Ventana Medical Systems) for EBV-encoding small RNAs using a Benchmark XT autostainer (Ventana Medical Systems). Staining of these sections was assessed by evaluating the positivity of the tumor cell nuclei.

8. Availability of data and materials

The 16S rRNA and WTS data generated during this study are deposited in the National Center for Biotechnology Information Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra/) under the accession number PRJNA1174735.

Results

1. Differences of microbiome GC compared with NC by 16S rRNA sequencing

S2 Table presents the clinicopathological characteristics of the enrolled patients. We analyzed the tissue-associated microbiota from GC and NC samples using 16S rRNA sequencing. All alpha diversity indices (Evenness, Shannon, and Simpson) were higher in GC tissues than in NC tissues (p < 0.001) (Fig. 1A). To compare the bacterial taxa composition between GC and NC tissues, principal coordinate analysis (PCoA) was performed using the Bray-Curtis distance, weighted UniFrac distance, and unweighted UniFrac distance. Significant differences were observed between GC and NC for each beta diversity index (p < 0.001) (Fig. 1B).

Comparing the relative abundances at the phylum and genus levels between GC and NC tissues (Fig. 1C), we observed that Helicobacter decreased significantly at the genus level (p < 0.001). However, Streptococcus, Prevotella, and Lactobacillus significantly increased in GC (p < 0.01) (Fig. 1D). Nevertheless, Helicobacter remained the dominant microbiota in GC. The functional diversity of different putative microbiomes was assessed using PICRUSt2 and visualized using Statistical Analysis of Metagenomic Profiles (STAMP) (Fig. 1E). Pathways with significant differences in mean proportions between GC and NC tissues were identified. Environmental information processing (phosphotransferase system) and metabolism pathways were increased (adjusted p < 0.001), whereas cellular processes, including bacterial chemotaxis, were decreased in GC (adjusted p < 0.001). Additionally, after dividing the microbial abundance of GC into two groups based on the optimized cut-off points for the most significant relation with OS, higher Fusobacterium abundance was associated with worse OS (p=0.046) (S3 Fig.).

2. Association between gastric microbiota and host immune responses and gene expression profiles

Using WTS data deconvolution, we analyzed immune cell interactions with seven highly abundant bacteria in GC tissue. In GC, total macrophages, M1 macrophages, and helper T cells were positively correlated with Streptococcus, Prevotella, and Fusobacterium but negatively correlated with Helicobacter, while fibroblasts showed the opposite pattern. Correlations between Helicobacter and immune cells were stronger in NC than GC, with Helicobacter positively correlated with most immune cells in NC, while most non-Helicobacter bacteria, except Leptotrichia, showed negative correlations (Fig. 2A, left).

Correlation analysis of immune checkpoint genes (CTLA-4, PDCD1, CD274, and LAG3) revealed predominantly negative correlations between Helicobacter and these genes in GC, except for CCR4. In contrast, Prevotella, Streptococcus, and Fusobacterium were positively correlated with PDCD1, PDCD1LG2, LAG3, and IDO1 but negatively correlated with CCR4. In NC, Helicobacter showed positive correlations with most immune checkpoint genes, while Lactobacillus exhibited negative correlations. Overall, correlations between bacterial abundance and immune components were weaker in GC compared to NC (Fig. 2A, right)

Differentially expressed genes between GC and NC tissues were identified from WTS data, and functional classes were assessed using Single-sample Gene Set Enrichment Analysis (ssGSEA), revealing significant correlations between these functional classes and the abundances of the seven bacteria, with host gene expression levels closely associated with bacterial proportions (Fig. 2B). Signaling pathways related to cancer cell proliferation, including the JAK/STAT signaling pathway, were enriched in GC tissues, with higher abundances of Helicobacter and Acinetobacter. However, pathways related to DNA repair systems, including base excision repair, mismatch repair, and nucleotide excision repair, were enriched in GC tissues with higher abundances of Streptococcus, Prevotella, and Fusobacterium. Helicobacter was positively correlated with pathway-related DNA repair systems in NC tissues.

3. Comparison of saliva microbiome from GC patients with tissue samples

To compare saliva from GC patients and healthy individuals and to compare differences between saliva and tissue, we analyzed microbiome characteristics in endoscopic biopsy tissue and saliva samples from the same GC patient. Alpha diversities of saliva tended to be increased in GC patients compared to healthy individuals, without statistical significance in all indices (evenness, Shannon, and Simpson) (p > 0.05). However, when comparing saliva to tissue samples, evenness and variability were greater in saliva than in tissue samples (p < 0.001) (Fig. 3A). PCoA of the Bray-Curtis distance showed no significant differences in saliva samples between GC patients and healthy individuals (p > 0.05). Nonetheless, we observed a statistically significant difference between saliva and tissue in both GC patients and healthy individuals (p < 0.001) (Fig. 3B).

When comparing the relative abundance at the phylum and genus levels between the saliva and tissue in GC patients, the saliva contained few unclassified bacteria and various bacteria with a relatively even distribution (Fig. 3C). Helicobacter, dominant in the gastric tissue, was not observed in the saliva. Saliva contained several oral bacteria, including Haemophilus, Neisseria, Porphyromonas, and Veillonella, which were absent in tissues. Compared to saliva from healthy individuals, Gemella disappeared, Prevotella decreased, Granulicatella and Oribacterium appeared, and Haemophilus and Streptococcus tended to increase in saliva from GC patients. Functional diversity of the different putative microbiomes (Fig. 3D) revealed that the ratio of cell cycle pathways in the Kyoto Encyclopedia of Genes and Genome (KEGG) database was lower in the saliva of GC patients than in healthy individuals.

4. Microbiota characteristics of each molecular subtype of GC

The molecular subtypes of GC include EBV+, MSI-H, epithelial-mesenchymal transition (EMT)–like (altered E-cadherin expression), p53-positive (altered p53 expression), and p53-negative [18]. We analyzed the microbial characteristics of GC and NC tissues in each molecular subtype. Alpha diversity of all indices (Evenness, Shannon, and Simpson) did not differ between subtypes (S4A Fig.). Beta diversity analysis using the Bray-Curtis distance in NC revealed no difference in each subtype of GC (p=0.665). However, beta diversity in GC was significantly different in each subtype (p < 0.001) (S4B Fig.). Relative abundance at the phylum and genus levels in the molecular subtypes of GC (Fig. 4A) revealed that Helicobacter was more abundant in EMT-like GC than in MSI-H and EBV+ GCs, whereas Streptococcus was less abundant in EMT-like GC. MSI-H and EBV+ GCs had lower abundances of Helicobacter but higher abundances of Lactobacillus, Prevotella, and Streptococcus than EMT-like GC (Fig. 4B).

Notable differences in microbial community structure were observed based on molecular subtypes. Beta diversity analysis revealed a clear distinction between EMT-like, EBV+, and MSI-H subtypes, with EMT-like GCs being particularly distinct from EBV+ and MSI-H (p < 0.001) (Fig. 4C). Ecological network analysis further demonstrated subtype-specific microbial interaction patterns (Fig. 4D). While all subtypes shared similar patterns of positive (red edges) and negative (blue edges) correlations, EMT-like GC displayed a greater number of edges and higher network density, indicating more complex microbial interactions compared to the other subtypes. Functional profiling using PICRUSt2 revealed distinct metabolic features among GC subtypes (Fig. 4E). EMT-like GC showed reduced enrichment in cellular process-related pathways, such as apoptosis, but exhibited enhanced activity in various metabolic pathways, particularly lipid metabolism, when compared to NC tissues. MSI-H GC was associated with decreased enrichment in genetic information processing pathways, including base excision repair and DNA replication, along with increased metabolic activity. Although EBV+ GC exhibited statistically significant functional differences relative to NC tissues, the small sample size (n=13) may have limited the detection of distinct clustering. We analyzed interactions between seven highly abundant bacteria and immune cells, as well as immune checkpoint genes, across molecular subtypes of GC. Although correlation patterns varied among molecular subtypes, no significant differences were observed (S5 Fig.).

5. Microbial cluster characteristics and prognosis in MSI-H GC

We compared bacterial abundance in five distinct clusters based on the relative abundance of the top five bacteria (Fig. 5A). The clusters were categorized as follows: group 1 (Fusobacterium-predominant), group 2 (Streptococcus-predominant), group 3 (Prevotella-predominant), group 4 (Lactobacillus-predominant), and group 5 (Helicobacter-predominant). The distribution of molecular subtypes varied across these groups; in particular, EMT-like GC was predominantly enriched in the Helicobacter-predominant group 5, whereas MSI-H and EBV+ GCs were relatively evenly distributed across all groups (S6 Table). When examining immune cell distributions within each group, no specific immune cell type was exclusively associated with a particular group (Fig. 5B, left, S7 Fig.). However, fibroblasts were notably more abundant in group 5 than in the other groups (Fig. 5B, right).

We further assessed the characteristics of MSI-H GC distributed across all groups. While there were no significant differences in bacterial abundance among MSI-H GC across groups, Kaplan-Meier survival analysis with log-rank tests revealed that group 4 exhibited significantly worse OS (p=0.002) and disease-free survival (DFS; p < 0.001) compared to the other groups (Fig. 5C, upper). To determine whether this prognostic impact was independent of clinicopathologic factors, a multivariable Cox regression analysis was conducted, incorporating age, sex, stage, Lauren classification, tumor border, and lymphatic, venous, and perineural invasion. Group 4 remained a significant independent predictor of poor prognosis, with a hazard ratio (HR) of 6.317 for OS (95% confidence interval [CI], 1.752 to 22.782; p=0.005) and 14.565 for DFS (95% CI, 2.739 to 77.462; p=0.002) (Fig. 5C, lower). To ensure the robustness of these findings, we performed Firth’s penalized Cox regression analysis (S8 Table). High Lactobacillus abundance was significantly associated with poor prognosis in MSI-H GC patients (HR, 5.94; 95% CI, 1.78 to 21.64; p=0.004). Furthermore, when MSI-H GC was stratified into two groups based on optimized microbial abundance cut-off point most strongly significant relation with OS and DFS, the Lactobacillus-abundance group was significantly associated with worse prognosis (p < 0.001) (Fig. 5D). Additionally, OS and DFS were analyzed for MSI-H GC within each group, and although no statistically significant differences were observed, except for worse DFS was noted in group 5 (S9 Fig.).

DEG analyses of immune system process gene set from gene set enrichment analysis (GSEA) database among MSI-H GC, we found that group 4 versus other groups had significantly altered gene expression (Fig. 5E, left). Notably, CCL21 was significantly increased in group 4 compared to other groups (p=0.021), while CCR1, SIT1, and IL31RA were significantly downregulated in group 4 (p=0.023, p=0.023, and p=0.008, respectively) (Fig. 5E, right).

Discussion

Our study represents one of the largest Korean GC microbiome cohorts of fresh frozen tumor and adjacent normal tissues from a Korean patient cohort. Unlike many previous studies that focused on fecal [21] or were conducted primarily in Western populations [22], our approach by directly profiling fresh frozen tissue pairs from the same patients provides a more direct assessment of the gastric microbial community. We observed significant differences in the microbial composition between GC and NC, with increased alpha diversity in GC compared to NC and marked differences in beta diversity. When comparing relative abundances of bacteria, Helicobacter decreased, whereas Prevotella and Streptococcus increased in GC compared to NC. Our analysis revealed the relationship between specific microbiomes and immune cell populations in GC and NC tissues. Streptococcus, Prevotella, and Fusobacterium positively correlated with total macrophages, M1 macrophages, and helper T cells in GC, indicating a relationship to the inflammatory response. Conversely, Helicobacter negatively correlated with these immune cells in GC. Further analysis of the relationship between the microbiome and immune checkpoint gene expression revealed that Helicobacter was predominantly negatively correlated with key immune checkpoint genes, including CTLA-4, PDCD1, CD274, and LAG3 in GC tissues. Conversely, Prevotella, Streptococcus, and Fusobacterium positively correlated with immune checkpoint genes. These results align with those of previous studies showing that the presence of Prevotella, Streptococcus, and Fusobacterium is associated with better responses to immune checkpoint inhibitors in various tumors [23]. Fusobacterium, although often linked to poor prognosis in colorectal cancer, has been reported to enhance PD‑L1 expression via STING pathway activation, thereby potentially sensitizing tumors to programmed death-1/PD‑L1 blockade [24]. However, a large pan-cancer analysis also associated Fusobacterium with resistance to immune checkpoint blockade, indicating that its role may be context-dependent [25,26]. Regarding Helicobacter, its negative correlations with immune checkpoints may reflect its known immunosuppressive role in early gastric carcinogenesis. Previous studies have shown that H. pylori can promote immune evasion through mechanisms such as induction of regulatory T cells [27] and upregulation of PD‑L1 expression [28]. Nonetheless, further experimental validation is needed to elucidate the underlying mechanisms of these associations. To our knowledge, this is the first study to provide a comprehensive comparison of the GC microbiome with that of adjacent normal tissues within the same patient, establishing a correlation between the microbiome and immune responses in GC. Furthermore, these findings highlight the complex interactions between the microbiome and the immune system in the tumor microenvironment. The associations between specific microbiomes, immune cell populations, and immune checkpoint gene expression suggest that the microbiome may influence therapeutic response via modulating the immune environment.

Our analysis revealed that distinct microbiota compositions are associated with host signaling pathways in GC. Specifically, the enrichment of proliferative pathways such as JAK/STAT in GC tissues with higher abundances of Helicobacter and Acinetobacter suggests that these bacteria may actively promote tumorigenesis by modulating inflammatory and growth-related signaling, consistent with previous reports that H. pylori activates STAT3 phosphorylation and drives oncogenic transformation [29,30]. In contrast, DNA repair pathways—including base excision repair, mismatch repair, and nucleotide excision repair—were enriched in GC tissues with increased Streptococcus, Prevotella, and Fusobacterium, indicating a potential role for these genera in maintaining genomic stability or facilitating tumor adaptation to genotoxic stress [31,32]. Interestingly, in NC tissues, Helicobacter abundance was positively correlated with DNA repair pathways, possibly protective, host response to microbial colonization [33]. These results suggest not only compositional variations in the gastric microbiome but also a potential functional role of bacterial taxa in involvement tumor-related signaling pathways.

Our study identified differences in microbiomes according to the molecular subtypes of GC, particularly EBV+, EMT-like, and MSI-H GCs. While previous microbiome studies in Korean GC cohorts have confirmed the presence of microbial dysbiosis in tumor tissues, our study further advances by demonstrating that each molecular subtype is associated with a microbial composition. This is an area that has been underexplored in Korean populations. Helicobacter, Prevotella, and Streptococcus displayed varying abundances across molecular subtypes of GC. EBV+ GC is characterized by a lower abundance of Helicobacter and a higher abundance of Streptococcus. EMT-like GC showed a decrease in the abundance of Helicobacter compared with NC, but the proportion of non-Helicobacter bacteria was maintained, resembling the microbial environment of NC. MSI-H GC had the lowest Helicobacter and higher Lactobacillus and Fusobacterium levels than other subtypes of GC. Lactic acid bacteria, particularly Lactobacillus, consistently demonstrated increased relative abundance in the gastric microbial community of GC tissue, findings from previous studies [34]. Although investigations of Lactobacillus in GC are limited, our study revealed an association between MSI-H and these bacteria.

Our study extends this context by identifying an MSI-H–specific, Lactobacillus-dominant microbial cluster (group 4). Patients in this group had significantly worse OS and DFS compared to the other groups. This association remained significant after adjusting for known clinicopathologic factors in a multivariate Cox regression analysis, suggesting that Lactobacillus abundance may act as an independent negative prognostic factor. Although Lactobacillus is generally known as a beneficial commensal microbiota, our results suggest that Lactobacillus may contribute to cancer progression in certain circumstances. A recent study found that Lactobacillus is increased in GC [35]. In addition, studies have shown that certain Lactobacillus species are associated with cancer. L. iners was more abundant in cervical cancer [36]. Lactobacillus spp., known to produce indole, has also been shown to promote cancer growth by altering macrophage function in pancreatic cancer [37]. Given that there are over 100 Lactobacillus species, it is likely that different species will exhibit different properties and functions. However, our study is limited by the fact that we used 16S rRNA sequencing, which primarily provides genus-level resolution and therefore cannot distinguish species [38]. Therefore, future studies using shotgun metagenomic sequencing are needed to accurately identify and characterize Lactobacillus species present in cancer tissues. Furthermore, the mechanisms by which Lactobacillus influence cancer remain largely unexplored, and extensive further studies are needed to elucidate their specific roles and potential implications in tumorigenesis.

Most previous studies on the gut microbiome used fecal samples containing minimal human genomic DNA [39], which do not accurately represent the gastric microbiome. Similarly, research on the microbial community in GC was mainly based on gastric fluid or saliva because of the ease of collection and non-invasiveness [40]. However, these sample types often include a high proportion of oral microbiota and may not accurately reflect the gastric microbial environment [41]. Kim et al. [42] demonstrated that antral mucosa, phloem mucosa, and gastric juice samples represent different phyla using barcode pyrosequencing of the 16S rRNA gene. This indicates that there may be differences between the microbiota of the gastric juice and the mucosa. In contrast, tissue samples obtained through surgery or endoscopic biopsy provide a more direct and potentially less-contaminated representation of microbiota [39]. Therefore, there is increasing importance in utilizing tissue samples to gain a more precise understanding of the gastric microbiota and its role. In this study, we used tissue sample from GC and paired NC tissues to characterize the gastric microbiome.

To further evaluate sampling differences, we compared gastric tissue and saliva samples collected from patients undergoing endoscopic examinations. While differences in microbiota between healthy individuals and GC patients were not significant in saliva samples, differences in alpha and beta diversity between saliva and tissue samples were significant in GC patients and healthy individuals. Comparing the relative abundance between saliva and tissue in GC, Helicobacter, which was dominant in tissue, was not observed in saliva. Moreover, a relatively even distribution of various bacteria was observed in saliva, including several oral bacteria that were not detected in gastric tissue. These findings suggest that saliva, although frequently used in gastric microbiota studies, may not accurately reflect the microbial characteristics of gastric tissues. Our results support the importance of using gastric tissue samples for more precise characterization of the gastric microbiome and its potential role in GC.

However, this study has some limitations. 16S rRNA sequencing data can only provide relative abundance, rather than absolute abundance, of the microbiota. Furthermore, our immune cell results relied on deconvoluted WTS data. Therefore, it is necessary to independently validate the immune cell composition findings to ensure reliability and accuracy.

In conclusion, our study, based on paired GC and matched NC tissue samples from a large Korean GC cohort, differs from many previous investigations that relied on fecal or saliva samples. The findings demonstrate that specific bacterial taxa are not only differentially abundant in GC tissues but also closely linked to immune cell populations and immune checkpoint gene expression. Notably, we identified a group within MSI-H GC characterized by a Lactobacillus-enriched and other major bacteria-depleted microbiota, which was associated with worse prognosis and transcriptional changes. Additionally, we observed that the saliva microbiome contains oral bacteria that differ from those present in gastric tissue, suggesting an importance of selecting appropriate sample types for microbiome research. These finding contributes valuable insights for future studies and the advancement of microbiome-based diagnostics and therapeutics.

Electronic Supplementary Material

Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).

crt-2025-449_S1_Fig.pdf

crt-2025-449_S2_Table.pdf

crt-2025-449_S3_Fig.pdf

crt-2025-449_S4_Fig.pdf

crt-2025-449_S5_Fig.pdf

crt-2025-449_S6_Table.pdf

crt-2025-449_S7_Fig.pdf

crt-2025-449_S8_Table.pdf

crt-2025-449_S9_Fig.pdf

crt-2025-449_Supplementary_Methods.pdf

NOTES

Ethical Statement

This study was conducted in accordance with protocols approved by the IRB of SNUH (IRB number: H-2016-215-1231) and SNUBH (IRB number: B-1810-499-301). Written informed consent was obtained from all participants before participation.

Author Contributions

Conceived and designed the analysis: Nam SK, Lee HS.

Collected the data: Nam SK, Kwak Y, Park KU, Lee HS.

Contributed data or analysis tools: Nam SK, Oh S, Shin CM, Park KU, Kwon NJ, Kong SH, Park DJ, Lee HJ, Yang HK, Lee HS.

Performed the analysis: Nam SK, Park J, Oh S, Kwak Y, Lee HS.

Wrote the paper: Nam SK, Lee HS.

Conflicts of Interest

Conflict of interest relevant to this article was not reported.

Funding

This work was supported by the Technology Development Program (S2638360) funded by the Ministry of SMEs and Startups (MSS, Korea), and a National Research Foundation of Korea (NRF) grant funded by the Korean government (Ministry of Science and ICT) (No. RS-2024-00337984).

Fig. 1.

Comparison of gastric microbiota between gastric cancer (GC) tissue and paired non-cancerous gastric mucosa (NC) tissue. (A) Alpha diversity metrics including Evenness, Shannon, and Simpson indices in patients with GC and NC (^***p < 0.001, Wilcoxon test). (B) Beta diversity assessing through principal coordinate analysis using Bray-Curtis, weighted UniFrac, and unweighted UniFrac distances (p < 0.001, Permutational Multivariate Analysis of Variance test). (C) Relative abundances of major taxa at the phylum (left) and genus (right) levels in NC and GC. (D) Linear discriminant analysis effect size (LEfSe) plot illustrating microbial taxa enriched in NC compared with GC tissues (upper). Cladogram plot describing from LEfSe analysis showing the relationship between microbial taxa (the levels represent, phylum, class, order, family, and genus from the inner to outer rings) (lower). (E) Differential functional pathway predictions between GC and NC using Statistical Analysis of Metagenomic Profiles (STAMP) (adjusted p < 0.001, Welch’s t test).

Fig. 2.

Correlations between immune cell expression, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway activity (Single-sample Gene Set Enrichment Analysis [ssGSEA]), and microbiota abundance in non-cancerous (NC) and gastric cancer (GC) tissues. (A) Correlation between the composition of immune cell types deconvoluted by xCell and microbiota abundance (left), the relationship between the gene expression level of immune checkpoint inhibitors estimated by whole transcriptome sequencing and bacterial abundance (right) in GC (upper) and NC (lower). Positive and negative correlations are shown in red and blue, respectively, with the darker ones being those with a stronger association (^*p < 0.05, ^**p <0.01, ^***p < 0.001). (B) Correlation between the KEGG pathway of gene expression level estimated by ssGSEA and the abundance of the microbiota compared with NC and GC. Positive and negative correlations are shown in pink and green, respectively, with the darker ones being those with a stronger association (^*p < 0.05, ^**p < 0.01, ^***p < 0.001).

Fig. 3.

Differential composition of gastric microbiota in saliva and tissue in gastric cancer (GC) Patients. Comparison between non-cancerous (NC) and GC tissues, as well as healthy individuals and GC saliva samples. (A) Alpha diversity comparison (^*p < 0.05, ^**p < 0.01, ^***p < 0.001, Wilcoxon test). (B) Beta diversity assessment using principal coordinate analysis based on Bray-Curtis, weighted UniFrac, and unweighted UniFrac distances (p < 0.001, PERMANOVA [permutational multivariate analysis of variance] test). (C) Relative abundance of bacterial communities at the phylum (left) and genus (right) levels. (D) Functional metagenomic pathway predictions comparison between GC patients and healthy individuals using Statistical Analysis of Metagenomic Profiles (STAMP) software (adjusted p < 0.05, Welch’s t test).

Fig. 4.

Variation in gastric microbiota across molecular subtypes of gastric cancer (GC). (A) Relative abundance of microbial taxa in each molecular subtype of non-cancerous (NC) and GC at the phylum (upper) and genus (lower) levels. EBV+, Epstein-Barr virus–positive; EMT, epithelial-mesenchymal transition; MSI-H, microsatellite instability–high; p53-N, p53-negative; p53-P, p53-positive. (B) Genera displaying significant differences in abundance among each molecular subtype of GC (^*p < 0.05 and ^**p < 0.01, Wilcoxon test). (C) Comparing EMT-like GC with EBV+ and MSI-H GCs using Bray-Curtis distance (p < 0.001, PERMANOVA [permutational multivariate analysis of variance] test). (D) Ecological network analysis depicting microbial interactions within each GC subtype. Red and blue edges represent positive and negative correlations. (E) Functional metagenomic pathway predictions comparison across GC subtypes using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2).

Fig. 5.

Microbial clustering in gastric cancer (GC) and its association with microsatellite instability–high (MSI-H) and prognosis. (A) Heatmap showing the normalized relative abundance (y-axis) of dominant bacterial genera in GC tissues, categorized into five distinct microbial groups: group 1 (Fusobacterium-predominant), group 2 (Streptococcus-predominant), group 3 (Prevotella-predominant), group 4 (Lactobacillus-predominant), and group 5 (Helicobacter-predominant). (B) Bar plots displaying the distribution of molecular subtypes across the five microbial groups (left), and immune cell within individual groups (right). DC, dendritic cell; EBV, Epstein-Barr virus; EMT, epithelial-mesenchymal transition; NK, natural killer; Tregs, regulatory T cells. (C) Kaplan-Meier survival analysis comparing overall survival and disease-free survival within MSI-H GC between group 4 (Lactobacillus-predominant) and other groups (upper), followed by univariate and multivariate Cox regression analyses (lower). CI, confidence interval; HR, hazard ratio. (D) Survival analysis of MSI-H GC stratified by Lactobacillus abundance using maximally selected chi-square statistics. Divided into high and low abundance groups, assessed using the log-rank test. (E) Differential gene expression analysis focused on immune system process–related gene sets, between group 4 and the other MSI-H GC groups.

REFERENCES

1. Hou K, Wu ZX, Chen XY, Wang JQ, Zhang D, Xiao C, et al. Microbiota in health and diseases. Signal Transduct Target Ther. 2022;7:135.Article PubMed PMC PDF
2. Flemer B, Lynch DB, Brown JM, Jeffery IB, Ryan FJ, Claesson MJ, et al. Tumour-associated and non-tumour-associated microbiota in colorectal cancer. Gut. 2017;66:633–43. Article PubMed PMC
3. Urbaniak C, Gloor GB, Brackstone M, Scott L, Tangney M, Reid G. The microbiota of breast tissue and its association with breast cancer. Appl Environ Microbiol. 2016;82:5039–48. Article PubMed PMC PDF
4. Pushalkar S, Hundeyin M, Daley D, Zambirinis CP, Kurz E, Mishra A, et al. The pancreatic cancer microbiome promotes oncogenesis by induction of innate and adaptive immune suppression. Cancer Discov. 2018;8:403–16. PubMed PMC
5. Yu G, Gail MH, Consonni D, Carugno M, Humphrys M, Pesatori AC, et al. Characterizing human lung tissue microbiota and its relationship to epidemiological and clinical features. Genome Biol. 2016;17:163.Article PubMed PMC
6. Schistosomes, liver flukes and Helicobacter pylori. IARC Monogr Eval Carcinog Risks Hum. 1994;61:1–241. PubMed PMC
7. Wu WM, Yang YS, Peng LH. Microbiota in the stomach: new insights. J Dig Dis. 2014;15:54–61. Article PubMed
8. Qin J, Li R, Raes J, Arumugam M, Burgdorf KS, Manichanh C, et al. A human gut microbial gene catalogue established by metagenomic sequencing. Nature. 2010;464:59–65. PubMed PMC
9. Gunathilake M, Lee J, Choi IJ, Kim YI, Kim J. Association between bacteria other than Helicobacter pylori and the risk of gastric cancer. Helicobacter. 2021;26:e12836Article PubMed PDF
10. Yang J, Zhou X, Liu X, Ling Z, Ji F. Role of the gastric microbiome in gastric cancer: from carcinogenesis to treatment. Front Microbiol. 2021;12:641322.Article PubMed PMC
11. Sharma P, Allison JP. Immune checkpoint targeting in cancer therapy: toward combination strategies with curative potential. Cell. 2015;161:205–14. Article PubMed PMC
12. Gunjur A, Shao Y, Rozday T, Klein O, Mu A, Haak BW, et al. A gut microbial signature for combination immune checkpoint blockade across cancer types. Nat Med. 2024;30:797–809. Article PubMed PMC PDF
13. Yu AI, Zhao L, Eaton KA, Ho S, Chen J, Poe S, et al. Gut microbiota modulate CD8 T cell responses to influence colitis-associated tumorigenesis. Cell Rep. 2020;31:107471.Article PubMed PMC
14. Li R, Zhou R, Wang H, Li W, Pan M, Yao X, et al. Gut microbiota-stimulated cathepsin K secretion mediates TLR4-dependent M2 macrophage polarization and promotes tumor metastasis in colorectal cancer. Cell Death Differ. 2019;26:2447–63. Article PubMed PMC PDF
15. Ma C, Han M, Heinrich B, Fu Q, Zhang Q, Sandhu M, et al. Gut microbiome-mediated bile acid metabolism regulates liver cancer via NKT cells. Science. 2018;360:eaan5931PubMed PMC
16. Shi J, Geng C, Sang M, Gao W, Li S, Yang S, et al. Effect of gastrointestinal microbiome and its diversity on the expression of tumor-infiltrating lymphocytes in breast cancer. Oncol Lett. 2019;17:5050–6. Article PubMed PMC
17. Cancer Genome Atlas Research Network. Comprehensive molecular characterization of gastric adenocarcinoma. Nature. 2014;513:202–9. Article PubMed PMC PDF
18. Koh J, Lee KW, Nam SK, Seo AN, Kim JW, Kim JW, et al. Development and validation of an easy-to-implement, practical algorithm for the identification of molecular subtypes of gastric cancer: prognostic and therapeutic implications. Oncologist. 2019;24:e1321–30. Article PubMed PMC PDF
19. Park Y, Seo AN, Koh J, Nam SK, Kwak Y, Ahn SH, et al. Expression of the immune checkpoint receptors PD-1, LAG3, and TIM3 in the immune context of stage II and III gastric cancer by using single and chromogenic multiplex immunohistochemistry. Oncoimmunology. 2021;10:1954761.Article PubMed PMC
20. Oh S, Nam SK, Lee KW, Lee HS, Park Y, Kwak Y, et al. Genomic and transcriptomic characterization of gastric cancer with bone metastasis. Cancer Res Treat. 2024;56:219–37. Article PubMed PMC PDF
21. Yu J, Feng Q, Wong SH, Zhang D, Liang QY, Qin Y, et al. Metagenomic analysis of faecal microbiome as a tool towards targeted non-invasive biomarkers for colorectal cancer. Gut. 2017;66:70–8. Article PubMed
22. Ferlay J, Ervik M, Lam F, Laversanne M, Colombet M, Mery L, et al. Global cancer observatory: cancer today (version 1.1) [Internet]. International Agency for Research on Cancer: 2024. [cited 2025 May 1]. Available from: https://gco.iarc.who.int/today
23. Routy B, Le Chatelier E, Derosa L, Duong CP, Alou MT, Daillere R, et al. Gut microbiome influences efficacy of PD-1-based immunotherapy against epithelial tumors. Science. 2018;359:91–7. PubMed
24. Gao Y, Bi D, Xie R, Li M, Guo J, Liu H, et al. Fusobacterium nucleatum enhances the efficacy of PD-L1 blockade in colorectal cancer. Signal Transduct Target Ther. 2021;6:398.Article PubMed PMC PDF
25. Matson V, Fessler J, Bao R, Chongsuwat T, Zha Y, Alegre ML, et al. The commensal microbiome is associated with anti-PD-1 efficacy in metastatic melanoma patients. Science. 2018;359:104–8. Article PubMed PMC
26. Lim MY, Hong S, Nam YD. Understanding the role of the gut microbiome in solid tumor responses to immune checkpoint inhibitors for personalized therapeutic strategies: a review. Front Immunol. 2024;15:1512683.Article PubMed PMC
27. Bagheri N, Azadegan-Dehkordi F, Rahimian G, Rafieian-Kopaei M, Shirzad H. Role of regulatory T-cells in different clinical expressions of Helicobacter pylori infection. Arch Med Res. 2016;47:245–54. Article PubMed
28. Zhu Y, Zhu F, Ba H, Chen J, Bian X. Helicobacter pylori infection and PD-L1 expression in gastric cancer: a meta-analysis. Eur J Clin Invest. 2023;53:e13880PubMed
29. Xi Y, Zhang XL, Luo QX, Gan HN, Liu YS, Shao SH, et al. Helicobacter pylori regulates stomach diseases by activating cell pathways and DNA methylation of host cells. Front Cell Dev Biol. 2023;11:1187638.Article PubMed PMC
30. Wang H, Zhao M, Shi F, Zheng S, Xiong L, Zheng L. A review of signal pathway induced by virulent protein CagA of Helicobacter pylori. Front Cell Infect Microbiol. 2023;13:1062803.Article PubMed PMC
31. Wang Q, Hu T, Zhang Q, Zhang Y, Dong X, Jin Y, et al. Fusobacterium nucleatum promotes colorectal cancer through neogenesis of tumor stem cells. J Clin Invest. 2025;135:e181595Article PubMed PMC
32. Pimentel-Nunes P, Barros A, Pita I, Miranda I, Conceicao G, Borges-Canha M, et al. Gastric microbiome profile throughout gastric carcinogenesis: beyond Helicobacter. Scand J Gastroenterol. 2021;56:708–16. Article PubMed
33. Baik SC, Youn HS, Chung MH, Lee WK, Cho MJ, Ko GH, et al. Increased oxidative DNA damage in Helicobacter pylori-infected human gastric mucosa. Cancer Res. 1996;56:1279–82. PubMed
34. Park JY, Seo H, Kang CS, Shin TS, Kim JW, Park JM, et al. Dysbiotic change in gastric microbiome and its functional implication in gastric carcinogenesis. Sci Rep. 2022;12:4285.Article PubMed PMC PDF
35. Li ZP, Liu JX, Lu LL, Wang LL, Xu L, Guo ZH, et al. Overgrowth of Lactobacillus in gastric cancer. World J Gastrointest Oncol. 2021;13:1099–108. Article PubMed PMC
36. Colbert LE, El Alam MB, Wang R, Karpinets T, Lo D, Lynn EJ, et al. Tumor-resident Lactobacillus iners confer chemoradiation resistance through lactate-induced metabolic rewiring. Cancer Cell. 2023;41:1945–62. Article PubMed PMC
37. Hezaveh K, Shinde RS, Klotgen A, Halaby MJ, Lamorte S, Ciudad MT, et al. Tryptophan-derived microbial metabolites activate the aryl hydrocarbon receptor in tumor-associated macrophages to suppress anti-tumor immunity. Immunity. 2022;55:324–40. Article PubMed PMC
38. Durazzi F, Sala C, Castellani G, Manfreda G, Remondini D, De Cesare A. Comparison between 16S rRNA and shotgun sequencing data for the taxonomic characterization of the gut microbiota. Sci Rep. 2021;11:3030.Article PubMed PMC PDF
39. Tang Q, Jin G, Wang G, Liu T, Liu X, Wang B, et al. Current sampling methods for gut microbiota: a call for more precise devices. Front Cell Infect Microbiol. 2020;10:151.Article PubMed PMC
40. Chen M, Fan HN, Chen XY, Yi YC, Zhang J, Zhu JS. Alterations in the saliva microbiome in patients with gastritis and small bowel inflammation. Microb Pathog. 2022;165:105491.Article PubMed
41. Zilberstein B, Quintanilha AG, Santos MA, Pajecki D, Moura EG, Alves PR, et al. Digestive tract microbiota in healthy volunteers. Clinics (Sao Paulo). 2007;62:47–54. Article PubMed
42. Sung J, Kim N, Kim J, Jo HJ, Park JH, Nam RH, et al. Comparison of gastric microbiota between gastric juice and mucosa by next generation sequencing method. J Cancer Prev. 2016;21:60–5. Article PubMed PMC

Figure & Data

REFERENCES

Citations

Citations to this article as recorded by

PubReader
ePub Link

Cite

CITE: export Copy Download Format; Close

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

RIS — For EndNote, ProCite, RefWorks, and most other reference management software
BibTeX — For JabRef, BibDesk, and other BibTeX-specific software

Include:

Citation for the content below
Citation and abstract for the content below

Microbial Dynamics across Molecular Subtypes and Prognostic Significance of Lactobacillus in Gastric Cancer

Cancer Res Treat. 2026;58(3):824-845. Published online July 17, 2025

DOI: https://doi.org/10.4143/crt.2025.449

XML Download

Abstract
Introduction
Materials and Methods
Results
Discussion
Electronic Supplementary Material
NOTES
REFERENCES

Microbial Dynamics across Molecular Subtypes and Prognostic Significance of Lactobacillus in Gastric Cancer

Fig. 1. Comparison of gastric microbiota between gastric cancer (GC) tissue and paired non-cancerous gastric mucosa (NC) tissue. (A) Alpha diversity metrics including Evenness, Shannon, and Simpson indices in patients with GC and NC (***p < 0.001, Wilcoxon test). (B) Beta diversity assessing through principal coordinate analysis using Bray-Curtis, weighted UniFrac, and unweighted UniFrac distances (p < 0.001, Permutational Multivariate Analysis of Variance test). (C) Relative abundances of major taxa at the phylum (left) and genus (right) levels in NC and GC. (D) Linear discriminant analysis effect size (LEfSe) plot illustrating microbial taxa enriched in NC compared with GC tissues (upper). Cladogram plot describing from LEfSe analysis showing the relationship between microbial taxa (the levels represent, phylum, class, order, family, and genus from the inner to outer rings) (lower). (E) Differential functional pathway predictions between GC and NC using Statistical Analysis of Metagenomic Profiles (STAMP) (adjusted p < 0.001, Welch’s t test).

Fig. 2. Correlations between immune cell expression, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway activity (Single-sample Gene Set Enrichment Analysis [ssGSEA]), and microbiota abundance in non-cancerous (NC) and gastric cancer (GC) tissues. (A) Correlation between the composition of immune cell types deconvoluted by xCell and microbiota abundance (left), the relationship between the gene expression level of immune checkpoint inhibitors estimated by whole transcriptome sequencing and bacterial abundance (right) in GC (upper) and NC (lower). Positive and negative correlations are shown in red and blue, respectively, with the darker ones being those with a stronger association (*p < 0.05, **p <0.01, ***p < 0.001). (B) Correlation between the KEGG pathway of gene expression level estimated by ssGSEA and the abundance of the microbiota compared with NC and GC. Positive and negative correlations are shown in pink and green, respectively, with the darker ones being those with a stronger association (*p < 0.05, **p < 0.01, ***p < 0.001).

Fig. 3. Differential composition of gastric microbiota in saliva and tissue in gastric cancer (GC) Patients. Comparison between non-cancerous (NC) and GC tissues, as well as healthy individuals and GC saliva samples. (A) Alpha diversity comparison (*p < 0.05, **p < 0.01, ***p < 0.001, Wilcoxon test). (B) Beta diversity assessment using principal coordinate analysis based on Bray-Curtis, weighted UniFrac, and unweighted UniFrac distances (p < 0.001, PERMANOVA [permutational multivariate analysis of variance] test). (C) Relative abundance of bacterial communities at the phylum (left) and genus (right) levels. (D) Functional metagenomic pathway predictions comparison between GC patients and healthy individuals using Statistical Analysis of Metagenomic Profiles (STAMP) software (adjusted p < 0.05, Welch’s t test).

Fig. 4. Variation in gastric microbiota across molecular subtypes of gastric cancer (GC). (A) Relative abundance of microbial taxa in each molecular subtype of non-cancerous (NC) and GC at the phylum (upper) and genus (lower) levels. EBV+, Epstein-Barr virus–positive; EMT, epithelial-mesenchymal transition; MSI-H, microsatellite instability–high; p53-N, p53-negative; p53-P, p53-positive. (B) Genera displaying significant differences in abundance among each molecular subtype of GC (*p < 0.05 and **p < 0.01, Wilcoxon test). (C) Comparing EMT-like GC with EBV+ and MSI-H GCs using Bray-Curtis distance (p < 0.001, PERMANOVA [permutational multivariate analysis of variance] test). (D) Ecological network analysis depicting microbial interactions within each GC subtype. Red and blue edges represent positive and negative correlations. (E) Functional metagenomic pathway predictions comparison across GC subtypes using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2).

Fig. 5. Microbial clustering in gastric cancer (GC) and its association with microsatellite instability–high (MSI-H) and prognosis. (A) Heatmap showing the normalized relative abundance (y-axis) of dominant bacterial genera in GC tissues, categorized into five distinct microbial groups: group 1 (Fusobacterium-predominant), group 2 (Streptococcus-predominant), group 3 (Prevotella-predominant), group 4 (Lactobacillus-predominant), and group 5 (Helicobacter-predominant). (B) Bar plots displaying the distribution of molecular subtypes across the five microbial groups (left), and immune cell within individual groups (right). DC, dendritic cell; EBV, Epstein-Barr virus; EMT, epithelial-mesenchymal transition; NK, natural killer; Tregs, regulatory T cells. (C) Kaplan-Meier survival analysis comparing overall survival and disease-free survival within MSI-H GC between group 4 (Lactobacillus-predominant) and other groups (upper), followed by univariate and multivariate Cox regression analyses (lower). CI, confidence interval; HR, hazard ratio. (D) Survival analysis of MSI-H GC stratified by Lactobacillus abundance using maximally selected chi-square statistics. Divided into high and low abundance groups, assessed using the log-rank test. (E) Differential gene expression analysis focused on immune system process–related gene sets, between group 4 and the other MSI-H GC groups.

Fig. 1.

Fig. 2.

Fig. 3.

Fig. 4.

Fig. 5.

Microbial Dynamics across Molecular Subtypes and Prognostic Significance of Lactobacillus in Gastric Cancer