Abstract
Immune checkpoint blockade has emerged as a promising form of cancer therapy. However, only some patients respond to checkpoint inhibitors, while a significant proportion of patients do not, calling for the discovery of reliable biomarkers. Recent studies reported the importance of the gut microbiome in the clinical response to PD-1 blockade against non-small cell lung cancer (NSCLC), highlighting Akkermansia muciniphila as a candidate biomarker. Motivated by the genomic and phenotypic differences across Akkermansia muciniphila strains and Akkermansia (Akk) phylogroups (AmIa, AmIb, AmII, AmIII and AmIV), we analyzed fecal metagenomic sequencing data from four publicly available NSCLC cohorts (n = 575). Encouragingly, we found that patients' responses to PD-1 blockade are significantly different across different Akk phylogroups, highlighting the relatively stronger association between AmIa and positive responses than the other phylogroups. Importantly, we also highlight the importance of considering the clinical heterogeneity among independent cohorts in across validation analysis. We built a machine learning model based on Akk gene profiles, which shed light on a group of Akk genes that may be associated with the response to PD-1 blockade. In summary, our study underlines the benefits of high-resolution analysis of Akk genomes in the search for biomarkers that may improve the prediction of patients' responses to cancer immunotherapy.