248 Background: As PARPi redefines treatment for HRRm mCRPC, identifying patients most likely to benefit remains a clinical priority. With predictive and prognostic biomarkers evolving, GMB composition has emerged as a potential modifier, influencing oncogenesis and PARPi response through metabolic and immune pathways. We investigated associations between GMB composition and PARPi outcomes in patients with HRRm mCRPC. Methods: Pre-treatment fecal samples were collected from pts with HRRm mCRPC. GMB composition was profiled using shotgun metagenomic sequencing (Illumina NextSeq) and analyzed with inhouse computational pipeline. GMB association with two specific response criteria were assessed independently - radiographic (RECIST v1.1) and by PSA reduction. Radiographic response (R-R) was defined as stable disease or partial response and non-response (R-NR) as progressive disease. PSA response (PSA-R) was defined as a PSA reduction of ≥30% from baseline, non-response (PSA-NR) as PSA reduction < 30%. Paired α diversity (Shannon, Simpson, Chao; Wilcoxon), between-group delta tests (Mann-Whitney (MW)), and β diversity (PERMANOVA) (Bray-Curtis, Jaccard, Euclidean) were calculated. Differential abundance at species level was assessed between response groups using MW-U and effect sizes (Cliff’s δ, Cohen’s d). Results: Median age of total 30 pts cohort at diagnosis was 64 yrs, and 40% had high-volume disease, median PSA of 8.0 ug/L at transition to CRPC. Among 1,913 species, 27 taxa were linked to response (p<0.05, |δ|≥0.33). 70% (21/30) were R-R and 47% (14/30) PSA-R. While overall microbial diversity did not differ significantly between responders and non-responders, species-level differential analysis revealed microbial signatures. Notably, Gordonibacter urolithinfaciens were enriched in R-NR and PSA-NR across endpoints (δ=0.3, p=0.03 and δ=0.29, p=0.027, respectively), indicating a consistent detrimental association. Enterocloster lavalensis (δ=0.59; p=0.009) and a previously uncharacterized Egerieimonas species (δ=0.57; p=0.0019) were more prevalent in R-NR, whereas a different Enterocloster species (δ=–0.47; p=0.038) and Alistipes onderdonkii (δ=–0.47; p=0.047) were more prevalent in among R-R. Bifidobacterium dentium (δ=0.51; d=0.63; p=0.004) and Bacteroides fragilis (δ=-0.46; p=0.027) were more prevalent among PSA-NR, while PSA-R showed higher prevalence of Faecalibacterium prausnitzii (δ=-0.42; p=0.04). Conclusions: This is the first analysis linking species-level GMB signatures to PARPi response in HRRm mCRPC. Given G. urolithinfaciens ’s role in urolithin production affecting DNA-damage/epigenetic pathways, these exploratory findings support further study of GMB and metabolites (i.e. urolithin) as potential PARPi biomarker in HRRm mCRPC.
Yip et al. (Sun,) studied this question.