Fecal Transplant Success in Melanoma Hinges on Microbiome Integration, Not Diversity

For years, researchers betting on the gut microbiome as a lever for cancer immunotherapy assumed that more microbial diversity, or the arrival of specific beneficial species, was the goal. A new study from Stanford University and the University of Colorado Boulder suggests the real metric is something subtler: whether the transplanted donor community actually integrates and holds.

Donor microbiota integration, not diversity, drives anti-PD-1 response

Published in Nature Communications, the study by Justin Sonnenburg, Jessica Fessler, Matthew Olm, Edgar Engleman and colleagues used strain-resolved metagenomic analysis, a technique that tracks individual bacterial strains rather than just species categories, to dissect what separates melanoma patients who benefit from fecal microbiota transplantation (FMT) before anti-PD-1 therapy from those who do not. Their finding: successful integration of the donor microbiome was the defining feature of responders. Overall microbial diversity and the engraftment of any particular species were not.

This matters because most prior work in the field, including a meta-analysis of 12 FMT immunotherapy trials that found an overall objective response rate of 33.7%, has struggled to explain why some patients benefit and others do not. Knowing that integration itself is the key variable gives researchers a concrete, measurable target.

What integration actually looks like in responding patients

Responders in the study acquired a higher number of donor-derived strains and, crucially, maintained a stable microbial community structure after transplantation. Their gut microbiome profiles became progressively more similar to their donors over time, a pattern also observed in the Phase I FMT melanoma trial at ESMO, where 65% of 20 previously untreated advanced melanoma patients achieved an objective response and responders showed increasing donor similarity across follow-up timepoints.

Non-responders told a different story. Their microbial communities showed greater shifts in taxonomic composition after FMT, not stability, and exhibited higher estimated microbial loads. The team also found that genes associated with pathogen-linked secretion systems were more abundant in non-responders. These secretion systems are molecular machinery that many bacteria use to inject effector proteins into host cells, a characteristic more common in opportunistic or disruptive microbes than in stable community members.

Responders, by contrast, had microbiomes enriched in functions tied to community-level metabolism: the phosphotransferase system, which governs how bacteria take up sugars, and quorum sensing, the chemical signaling bacteria use to coordinate behavior within a community. Both are hallmarks of a cooperative, stable microbial ecosystem rather than a disrupted one.

Strain-resolved metagenomics across three clinical trials

What gives the findings credibility is the analytical scope. The team drew on metagenomic data from three separate clinical trials, applying strain-resolved analysis to each. Standard metagenomics assigns reads to species; strain-resolved analysis goes deeper, distinguishing between different lineages within a species. That resolution is what allowed the researchers to track whether specific donor strains were actually establishing themselves in recipients, rather than just detecting the presence of a related species that might have been there all along.

The final results of the MIMic phase 1 trial, which combined healthy donor FMT with anti-PD-1 in advanced melanoma and reported improved progression-free and overall survival alongside a 65% objective response rate, represent one node in the broader clinical picture this Stanford analysis draws on. Integrating data across multiple trials rather than relying on a single cohort strengthens the case that integration, not any trial-specific quirk, is the operative variable.

Tumor immune profiles track with microbiome restructuring

One of the more striking threads in the paper is the connection to the tumor microenvironment. Changes in the composition of tumor-infiltrating immune cells correlated with both clinical outcomes and the observed microbiome changes. That correlation does not establish causation on its own, but it is consistent with the broader model in which gut microbial signals shape systemic immune tone, which in turn influences how T cells and other immune populations respond inside the tumor.

This link between gut community structure and tumor immunology is part of what makes FMT an appealing adjunct to checkpoint inhibitors. The microbiome in advanced melanoma has been studied as a potential modulator of ICI outcomes for several years, with growing evidence that gut microbiota composition can either support or blunt the response to PD-1 blockade. The new Stanford data adds mechanistic texture: it is not static composition but dynamic integration that appears to matter.

Open questions about predicting and improving FMT integration

The study raises as many questions as it answers. The authors do not yet have a reliable pre-treatment predictor of who will achieve stable donor integration. Understanding whether host factors, including prior antibiotic exposure, baseline microbiome composition, or immune status, determine integration success will be essential for designing FMT protocols that work consistently. The three trials analyzed here also varied in design and patient population, so residual confounding cannot be ruled out.

The practical path forward involves both better donor selection and better ways to monitor integration in real time. An ongoing phase IIa trial is already evaluating FMT in immunotherapy non-responders using ICI-responders as donors, specifically to test whether microbiome modulation can convert patients who have already failed checkpoint therapy. If integration metrics can be tracked early and reliably, that trial and others like it could be enriched to select patients most likely to benefit, turning a promising but inconsistent intervention into a predictable one.