Lowered Abundance of Gut Bacteriophage Species Is Associated With Human Cancer Cachexia

Cancer cachexia exemplifies a high medical need condition without effective treatment. Recent studies implicated bacterial gut microbiome alterations to cancer cachexia. Whether the gut bacteriophage profile, an important microbiome component for health and disease, is also related to cancer cachexia remains unknown. We aimed to profile gut microbiome alterations in human cancer cachexia with attention on bacteriophages.

We performed shotgun metagenomic sequencing in stool samples from 78 cachectic and 42 noncachectic patients (53% male, mean age 67 ± 8 years) with newly diagnosed, advanced-stage (UICC IV) gastrointestinal cancers. Cachexia was defined according to the main criterion agreed upon international consensus (weight loss [WL] adjusted to body mass index [BMI]). Obtained DNA short-reads were used for k-mers-based, phage-inclusive matching with reference databases, de novo phage assembly and inferring microbiome-encoded functions. We replicated significance-based statistical and prediction-oriented machine-learning analyses in 2022 and 2025 generated metagenome datasets to incorporate the recent change by the International Committee on Taxonomy of Viruses (ICTV) from morphology-based (valid until 2022) to revised genome-based phage taxonomy into microbiome findings of cachexia.

Cachectic and noncachectic patients differed significantly regarding BMI (mean 20.9 vs. 26.4 kg/m2), WL (mean -6.5 vs. -0.2 kg), survival (median 5 vs. 13 months) and clinical cachexia domains (e.g., C-reactive proteine and appetite loss) (all p < 0.001) but not for other clinical covariables (e.g., cancer type) (all p > 0.05). Read-based mapping (2022/2025) identified 1.312/1.513 species (74/39 phage species), and de novo assembly resulted in 4.184/4.209 contigs (corresponding to 65/39 phage species). Concordantly, both analyses (2022 and 2025) showed that prevalent cachexia associated significantly with beta-diversity (Bray-Curtis distance, PERMANOVA, p < 0.05), but not to alpha-diversity (Shannon-Index, ANOVA, p > 0.05), reduced microbiome-encoded detoxification functions (e.g., enriched microbial β-glucuronidase and depleted bacterial efflux pumps) and lowered abundance of bacterial species with false-discovery-rate (FDR)-corrected p < 0.05 (2022: Faecalibacterium prausnitzii, Roseburia intestinalis, Streptococcus species and Lachnospiraceae species; 2025: Faecalibacterium species, Ruminococcus gauvreauii and Intestinibacter bartlettii). Further, lowered abundance of bacteriophages associated with cachexia, predominantly affecting double-stranded (2022: Caudovirales, Siphoviridae, FDR-corrected p < 0.05; 2025: Myoviridae, Siphoridae, p < 0.05) but also single-stranded (2022: Inoviridae, Microviridae, p < 0.05; 2025: Inoviridae; p < 0.05) DNA phage species. In machine-learning models, bacteriophages were top-ranked cachexia predictors (2022: Caudovirales, Siphoviridae; 2025: Myoviridae, Siphoridae). Accuracy was highest when only phage contigs were taken into account (correctly classified instances: 75.0%-85.8%; AUC: 0.703-0.916).

The previously unknown link between gut bacteriophages and human cancer cachexia expands the scope for basic, translational and clinical microbiome-targeted research in an area of significant unmet medical need.

Study Box of the German Cancer Society (Registration Number ST-U069, Date: 29 May 2018).