Enterotype-specific microbial biomarkers of immune checkpoint inhibitor response revealed by large-scale integrated metagenomic analysis

Immune checkpoint inhibitors (ICIs) are a powerful class of cancer drugs that harness the immune system to fight tumors, but only a subset of patients respond well to them. Growing evidence suggests the gut microbiome plays a key role in shaping those responses, yet no consistent set of microbial 'biomarkers' has emerged across studies — likely because different people harbor fundamentally different gut microbial communities. This study tackled that problem by proposing that patients should first be grouped by their 'enterotype' (a broad ecological category of gut community structure) before searching for biomarkers within each group. The researchers pooled metagenomic data from 569 stool samples collected from cancer patients across multiple tumor types who were receiving immunotherapy, then clustered them into two main enterotypes (E1 and E2), each with two sub-clusters. Within those enterotype-defined groups, they identified 166 species — including Collinsella, Blautia, and Bacteroides — whose abundance was specifically associated with enterotype membership and, by extension, potentially with ICI outcomes. A preliminary check against 19 patients with exceptional ICI responses provided early support for some of these enterotype-linked signals. Machine learning models built on the microbiome profiles showed moderate predictive power, suggesting that stratifying patients by enterotype before applying biomarker panels could be a more principled approach than a one-size-fits-all analysis.

• 569 fecal metagenomic samples from oncological patients were clustered into two primary enterotypes (E1 and E2), each containing two sub-clusters, providing a structured ecological framework for biomarker discovery.
• 166 enterotype-specific microbial species were identified as candidate biomarkers of ICI response, with notable taxa including Collinsella spp., Blautia spp., and Bacteroides spp. differentially distributed across enterotype clusters.
• Preliminary concordance assessment in 19 patients with exceptional immunotherapy responses provided initial, albeit limited, independent support for selected enterotype-associated microbial signals.
• Machine learning classifiers trained on gut microbiome profiles showed encouraging but only moderate predictive performance in the full heterogeneous dataset, underscoring the need for enterotype-stratified rather than pooled modelling.

This is a cross-sectional integrated metagenomic analysis pooling 569 fecal samples from cancer patients of multiple tumor types who were receiving immune checkpoint inhibitor therapy. Shotgun metagenomic sequencing data were aggregated from multiple cohorts and subjected to unsupervised clustering to derive enterotypes, followed by within-enterotype differential abundance analysis to identify species-level biomarkers. Machine learning models were then trained on microbiome composition features to predict immunotherapy outcomes. An independent preliminary validation was performed in a separate cohort of 19 patients classified as exceptional ICI responders.

The validation cohort of 19 exceptional responders is very small and not powered to confirm biomarker performance, leaving the enterotype-specific signals largely hypothesis-generating. Because samples were aggregated across geographies, tumor types, and ICI regimens, residual confounding from these sources cannot be fully disentangled, even after enterotype stratification. The study does not establish mechanistic causality — it is unknown whether the identified microbial taxa actively modulate immune responses or are merely correlated bystanders. A next logical experiment would be a prospective, multi-centre trial that pre-stratifies patients by enterotype at baseline, applies the identified biomarker panels, and correlates them with standardised clinical endpoints (e.g., objective response rate, progression-free survival) alongside functional readouts such as metabolomics or immune profiling.

This study extends [§100], which frames microbiome research at a translational crossroads and emphasises the ecological and causal complexity that complicates clinical biomarker development, by operationalising enterotype stratification as a concrete strategy to reduce inter-individual heterogeneity before biomarker extraction. It parallels [§86], which also applies deep metagenomic sequencing and integrated multi-omics to identify gut microbiota signatures linked to a complex systemic phenotype (post-stroke cognitive impairment), demonstrating how large-scale metagenomic integration can reveal disease-associated microbial patterns that elude smaller single-cohort studies. The enterotype-driven stratification approach also resonates with [§87], where gut microbiota composition was shown to regulate systemic inflammatory and compensatory anti-inflammatory response syndromes — reinforcing the broader principle that immune modulation by the microbiome is context- and community-structure-dependent. Finally, the moderate machine learning performance reported here aligns with cautionary notes in [§100] about premature clinical translation, suggesting that microbiome-based patient classification remains an exploratory framework requiring further prospective validation.

• The human microbiome at translational crossroads: an ecological and causal perspective
  Provides the ecological and causal theoretical scaffolding that motivates the enterotype-stratification approach used here; reading both together clarifies why pooling heterogeneous microbiome cohorts without ecological structuring undermines biomarker reproducibility.
• Integrated gut microbiota and metabolome signatures revealed by deep metagenomic sequencing in post-stroke cognitive impairment with type 2 diabetes
  Uses the same deep metagenomic integration strategy in a different complex disease (post-stroke cognitive impairment), offering a useful methodological comparator for assessing what multi-cohort metagenomic pipelines can and cannot resolve.
• Gut Microbiota Regulates Systemic Inflammatory Response and Compensatory Anti-Inflammatory Response Syndromes by Targeting PF4+ Macrophages in Acute Pancreatitis
  Demonstrates gut microbiota regulation of systemic immune phenotypes (SIRS/CARS) in a separate clinical context, providing mechanistic parallel evidence that microbial community structure shapes host immune trajectories beyond cancer immunotherapy.
• Gut Microbiota Dysbiosis Remodels the Multi-Tissue Transcriptional Landscape of the Host
  Shows that gut dysbiosis remodels transcriptional landscapes across multiple tissues, offering a functional mechanistic bridge that could explain how the enterotype-specific species identified here might translate into differential immune gene expression in ICI-treated patients.
• Westlake Frequent-sampling Cohort 2
  A dense longitudinal cohort study (Westlake Frequent-sampling Cohort) that captures within-individual microbiome dynamics, highlighting a key dimension — temporal stability of enterotype assignment — that the cross-sectional design of the index study cannot address.