Distinct gut virome profiles are associated with response to anti-PD-1 therapy in non-small cell lung cancer

Background The gut microbiota plays a critical role in modulating the efficacy of immune checkpoint inhibitor (ICI), yet the contribution of the gut virome remains under-characterized, particularly in advanced non-small cell lung cancer (NSCLC). This study aimed to characterize the gut virome and elucidate its mechanistic involvement in response to PD-1 inhibitor therapy. Methods We performed large-scale metagenomic virome profiling of fecal samples from 338 NSCLC patients treated with PD-1 blockade, with an independent cohort (n = 30) used for external validation. Viral diversity, composition, and function profiles were analyzed. Bacterium–virus interaction networks were constructed, and random forest models were developed to predict treatment response. Results The Shannon index of gut viral diversity decreased significantly with poorer clinical response, and β-diversity analysis revealed distinct virome structures between groups. We identified 194 viral operational taxonomic units (vOTUs) enriched in non-responders (NR), predominantly from Peduoviridae and Inoviridae , and 594 vOTUs enriched in responders (R), mainly from Herelleviridae and Microviridae . Host prediction indicated that NR-enriched vOTUs frequently targeted bacterial genera such as Clostridium_M , Bacteroides , and Escherichia —previously associated with adverse ICI outcomes—while R-enriched vOTUs targeted beneficial genera, including Faecalibacterium and Roseburia . Co-occurrence network analysis demonstrated distinct, response-specific virus–bacteria interaction modules. Functional analysis revealed that NR-enriched vOTUs were significantly associated with bacterial metabolic pathways (e.g., K01689:ENO1_2_3, eno; enolase 1/2/3). Notably, a random forest model based exclusively on viral features predicted clinical response (R vs. NR) with higher accuracy (AUC = 76.8%) than a bacteria-only model (AUC = 66.4%). This performance advantage that was sustained in the external validation cohort (AUC = 74.2%). Furthermore, the presence of Akkermansia muciniphila was associated with a higher-diversity, responder-favorable virome profile. Conclusions The gut virome is profoundly reconfigured in NSCLC patients undergoing anti-PD-1 therapy, exhibiting distinct taxonomic, ecological, and functional characteristics that are strongly linked with clinical outcome. Our findings establish the gut virome as a superior predictor of ICI response compared to the bacteriome and underscore its potential as both a novel biomarker and a therapeutic target.