The research team enrolled 30 patients with advanced gastric cancer who were receiving immune checkpoint inhibitors combined with chemotherapy, a regimen that has become a standard of care following landmark trials such as KEYNOTE-811 and ATTRACTION-2. To capture how the gut ecosystem changes under treatment pressure, the investigators collected fecal samples at three carefully chosen time points: before any therapy began, one month after treatment initiation, and at the end of the therapeutic course. That longitudinal design yielded 90 samples in total, each subjected to 16S rRNA gene sequencing, a technique that reads the genetic barcodes of bacteria to determine which species are present and in what proportions. The raw sequences were deposited in the NCBI Sequence Read Archive under accession number SRP508771, allowing other researchers to scrutinize and reuse the data.

To assess treatment efficacy objectively, the team stratified patients using RECIST 1.1 criteria, the internationally standardized system for measuring tumor burden on imaging scans. Patients were grouped into three categories: partial response, meaning their tumors shrank substantially; stable disease, meaning the cancer neither grew nor shrank meaningfully; and progressive disease, meaning the treatment failed to hold the cancer in check. This clinical stratification became the lens through which every microbial measurement was interpreted. The central question was deceptively simple: did patients whose tumors responded carry a different gut bacterial community than those whose tumors did not, and if so, could that difference be detected before treatment even started?

The first major finding concerned stability. Across the three sampling time points, the overall composition of the gut microbiota and its alpha diversity, a measure of how many distinct species coexist within a single individual, remained largely stable over the course of therapy. This is itself an important observation, because it suggests that chemoimmunotherapy does not indiscriminately devastate the gut ecosystem in these patients, contrary to what might be feared given the mucosal toxicity sometimes associated with cancer treatment. However, stability at the community level did not mean the microbes were inert. When the researchers looked at specific bacterial families, they found that immunotherapy was associated with enrichment of Lachnospiraceae and Bacteroidaceae, two families with well-documented roles in immune regulation. Lachnospiraceae includes many butyrate-producing organisms, and butyrate, a short-chain fatty acid, is known to nourish the intestinal barrier and modulate T cell activity. Bacteroidaceae, meanwhile, contains species whose polysaccharides can train the immune system in ways that influence tolerance and inflammation.

The most striking results emerged when the researchers compared baseline samples across the response groups. Before a single dose of therapy, the gut microbiota of patients who would later achieve a partial response already differed significantly from that of patients with stable or progressive disease. Two genera stood out in the responder group: Senegalimassilia and Lactobacillus were overrepresented in patients who went on to respond. In contrast, three other genera, Parvimonas, Intestinibacter and Catenibacterium, characterized the microbiota of patients with stable or progressive disease. The implication is profound: a fecal sample taken before treatment could, in principle, help forecast whether a given patient’s immune system will mount an effective anti-tumor response when unleashed by checkpoint blockade. If validated in larger cohorts, such microbial biomarkers could be measured alongside existing clinical markers to guide treatment selection.

Describing these associations was only the first step. The team then asked a harder question: how might gut bacteria actually influence the outcome of chemoimmunotherapy in gastric cancer? To answer it, they integrated public databases, including gutMGene, a curated resource of experimentally verified target genes of gut microbes and their metabolites, to construct two interconnected networks. The first was a microbe-metabolite-target network, linking response-associated bacterial genera to the metabolites they produce or consume and to the host genes those metabolites regulate. The second was a microbe-host interaction network, mapping direct relationships between microbial genera and human genes. These computational maps transformed a list of correlated bacteria into a testable mechanistic hypothesis about communication between the gut ecosystem and the tumor immune microenvironment.

The networks converged on a set of host genes with unmistakable immunological significance: IFNG, which encodes interferon-gamma, the master cytokine of cellular anti-tumor immunity; IL12B, a subunit of interleukin-12 that drives T cell and natural killer cell activation; CXCL6, a chemokine involved in recruiting immune cells to sites of inflammation; and DUOX2, an enzyme generating reactive oxygen species in epithelial cells that participates in mucosal defense. According to the analysis, response-associated genera may modulate the expression of these genes, and in doing so influence the infiltration of CD8-positive T cells into tumors. CD8-positive T cells are the cytotoxic soldiers of the adaptive immune system, the very cells that checkpoint inhibitors depend upon to kill cancer cells. A gut ecosystem that promotes interferon-gamma and interleukin-12 signaling, and that fosters chemokine-driven recruitment of cytotoxic T cells, could effectively prime the tumor microenvironment so that immunotherapy finds an army ready to be activated.

These findings in gastric cancer fit into a rapidly expanding body of evidence that the microbiome shapes immunotherapy outcomes across tumor types. Seminal studies published in Science in 2018 showed that the gut microbiome modulates response to anti-PD-1 immunotherapy in melanoma patients and influences the efficacy of PD-1-based immunotherapy against epithelial tumors. More recent work has illuminated specific mechanisms, including the demonstration in Nature that targeting the PD-L2-RGMb axis can overcome microbiome-related immunotherapy resistance, and reports that microbial metabolites such as trimethylamine N-oxide can drive immune activation and boost checkpoint blockade in pancreatic cancer. In gastric cancer specifically, prior studies had suggested that the gut microbiome affects responses to treatment in HER2-negative advanced disease, and work from the same Fuzhou group indicated that the butyrate-producing bacterium Eubacterium rectale inhibits gastric carcinogenesis and augments immunotherapy efficacy. The new study extends this line of inquiry by combining longitudinal sampling with network-based mechanistic prediction in the chemoimmunotherapy setting.

The study also examined enterotypes, the broad community-state groupings into which human gut microbiomes tend to cluster, and traced their associations with treatment response, adding another layer of ecological context to the genus-level findings. The work was funded by the Joint Funds for the Innovation of Science and Technology of Fujian province, and the authors report no competing financial interests. As with any study of this scale, important caveats apply: 30 patients is a modest cohort, 16S sequencing resolves bacteria only to the genus level rather than the strain level, and the microbe-host gene interactions were inferred through database integration rather than demonstrated experimentally in this paper. Correlation at baseline, however compelling, does not prove causation, and prospective validation in independent, larger cohorts will be essential before microbiome profiling enters clinical decision-making for gastric cancer.

Nevertheless, the implications are considerable. Gastr cancer remains one of the leading causes of cancer death worldwide, with GLOBOCAN estimates attributing more than 750,000 deaths annually to the disease, and chemoimmunotherapy has become its most important recent therapeutic advance. If the microbial signatures identified here, elevated Senegalimassilia and Lactobacillus at baseline signaling likely benefit, and enrichment of Parvimonas, Intestinibacter or Catenibacterium signaling likely resistance, can be validated and standardized, clinicians could one day sequence a patient’s stool before treatment and tailor therapy accordingly. Patients predicted to respond might proceed confidently with checkpoint inhibitors, while those predicted to resist could be steered toward alternative strategies, microbiome-modulating interventions such as targeted probiotics or dietary manipulation, or intensified monitoring. The study also sharpens the mechanistic agenda for the field: confirming experimentally whether the IFNG, IL12B, CXCL6 and DUOX2 pathways truly mediate microbial effects on CD8-positive T cell infiltration in gastric tumors. What began as an observational survey of 90 fecal samples may thus mark an early step toward a future in which the gut ecosystem is treated not as an incidental bystander in cancer therapy, but as a measurable, modifiable determinant of whether the immune system can be successfully mobilized against one of humanity’s most stubborn malignancies.

Subject of Research: Gut microbiota composition and microbe-host interactions associated with response to chemoimmunotherapy in gastric cancer

Article Title: Association of gut microbiota composition and microbe–host interactions with response to chemoimmunotherapy in gastric cancer

Article References: Gao, W., Wang, X., Shi, Y., Wu, G., Zhou, M., & Lin, X. (2026). Association of gut microbiota composition and microbe–host interactions with response to chemoimmunotherapy in gastric cancer. Genes & Immunity. https://doi.org/10.1038/s41435-026-00413-2

Image Credits: AI Generated

DOI: 10.1038/s41435-026-00413-2

Keywords: gut microbiota, gastric cancer, chemoimmunotherapy, immune checkpoint inhibitors, 16S rRNA sequencing, Senegalimassilia, Lactobacillus, CD8-positive T cells, IFNG, biomarkers, tumor microenvironment, microbiome