Gut Microbe Butyrate Signature Flags Early Resistance to Immunotherapy in Lung Cancer

A new study published in Frontiers in Cellular and Infection Microbiology proposes a candidate marker drawn from an unexpected source: the gut microbiome.

The work, led by Yaoguo Mei of Jiujiang Traditional Chinese Medicine Hospital with colleagues at two other Chinese institutions, took the form of a dual-cohort, patient-level secondary analysis. The team reanalyzed pretreatment fecal shotgun metagenomic sequencing data from two public cohorts, comprising 70 patients in a discovery cohort and 40 patients in an external validation cohort. All participants were being treated with anti-PD-1 therapy for advanced non-small cell lung cancer, and the researchers examined whether the microbiome measured before treatment could distinguish patients who experienced early primary resistance, defined as resistance occurring within 90 days, from those who did not.

Early primary resistance was common in both cohorts, affecting 41 of the 70 patients in the discovery cohort and 23 of the 40 patients in the external validation cohort. Using ANCOM-BC2 for differential abundance analysis, the researchers found that classical probiotic-associated taxa were less abundant in the resistant group than in the non-resistant group. A butyrate functional subscore, built from the butyryl-CoA:acetate CoA-transferase (but) and butyrate kinase (buk) genes, was also lower among resistant patients, as was the overall composite measure, which the authors named the classical probiotic–butyrate axis score, or PBAS. Notably, the buk gene alone and the species Akkermansia muciniphila showed no significant differences between the two groups, underscoring the value of the combined approach. After adjustment for age, sex, and body mass index, acetyl-CoA pathway potential was also reduced in the resistant group.

For each one-standard-deviation increase in PBAS, the odds of early primary resistance fell to an odds ratio of 0.39, and the estimate remained similar at 0.37 after adjustment for clinical variables. Discrimination was assessed with receiver operating characteristic analysis: the bootstrap-corrected area under the curve, based on 2,000 patient-level bootstrap resamples for internal validation, was 0.733, while the locked model achieved an external validation AUC of 0.726 with a 95 percent confidence interval of 0.559 to 0.894. The wide confidence interval reflects the modest size of the validation cohort and signals considerable uncertainty in the precision of that estimate.

The score was also linked to longer-term outcomes. In Cox proportional hazards models, higher PBAS was associated with a lower risk of progression or death, with hazard ratios of 0.54 in the discovery cohort and 0.61 in the external validation cohort, both statistically significant at p less than 0.001. These findings suggest that the combined microbiome signature carries information about progression-free survival as well as about the binary 90-day resistance outcome, although the analysis remains observational and cannot establish that the microbial features cause the clinical outcomes.

The authors are careful about the scope of their conclusions. They describe PBAS as an exploratory outcome risk marker for cross-cohort risk stratification among patients receiving anti-PD-1 therapy, and they explicitly state that the findings do not establish treatment-specific predictive value. In other words, the score may indicate which patients are likely to have poor outcomes, but it does not yet demonstrate that it can predict who will benefit from anti-PD-1 therapy specifically rather than from other treatments. The study is also a secondary analysis of existing data rather than a prospective trial, and the cohorts were relatively small, with 110 patients in total.

Nevertheless, the study addresses a recognized weakness in the microbiome biomarker field. As the authors note in the background to their work, existing microbiome biomarkers have largely relied on individual species and have lacked stability across populations, which has hindered their translation into clinical use. By combining taxonomic information with functional gene content, and by validating the locked model in an independent cohort, the approach represents an attempt to build a more robust and transferable signature. Whether the classical probiotic–butyrate axis score can be prospectively validated in larger, diverse patient populations, and whether it can add value alongside existing clinical predictors, remains to be tested in future studies.

Subject of Research: A shotgun metagenomic study of gut microbiome butyrate potential as a marker of early resistance to anti-PD-1 immunotherapy in advanced non-small cell lung cancer.

Article Title: Reduced classical probiotic-associated butyrate biosynthetic potential predicts early primary resistance to anti-PD-1 therapy in advanced non-small cell lung cancer: a shotgun metagenomic study

Article References: Mei, Y., Shi, L., He, S., Liu, C., Liu, M., & Hou, Y. (2026). Reduced classical probiotic-associated butyrate biosynthetic potential predicts early primary resistance to anti-PD-1 therapy in advanced non-small cell lung cancer: a shotgun metagenomic study. Frontiers in Cellular and Infection Microbiology, 16, Article 1975315. https://doi.org/10.3389/fcimb.2026.1975315

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1975315

Keywords: anti-PD-1 therapy, non-small cell lung cancer, gut microbiome, butyrate, early primary resistance, shotgun metagenomics, probiotic-associated taxa, biomarker, progression-free survival, immunotherapy, Akkermansia muciniphila, risk stratification