Pretransplant gut microbiome signatures predict early acute rejection after kidney transplantation

Early identification of rejection remains a critical unmet need in kidney transplantation, as conventional tools detect rejection only after irreversible allograft injury. The pretransplant gut microbiome may provide novel predictive signals by modulating immune homeostasis. Pretransplant stool samples underwent shotgun metagenomic sequencing. Composition, functional profiles, and networks were compared between rejection and nonrejection (protocol biopsy ≤2 weeks). A prespecified short-chain fatty acid biosynthetic Kyoto Encyclopedia of Genes and Genomes ortholog panel was tested with false discovery rate (FDR) correction. Stepwise Random Forest models were developed with subgroup analyses and tested in a temporal validation cohort. Of 78 recipients, 26 (33.3%) developed biopsy-proven early acute rejection. Three taxa, including Phascolarctobacterium faecium, were FDR-significantly reduced. At the gene level, mcmB (a key propionate-biosynthetic enzyme) was the only Kyoto Encyclopedia of Genes and Genomes ortholog reaching FDR significance in the prespecified short-chain fatty acid panel (q = 0.018). Network analysis revealed selective microstructural reorganization. Area under the curve values improved stepwise (0.565 → 0.681 → 0.765), were preserved across rejection subtypes (T cell-mediated rejection spectrum 0.74 and antibody-mediated rejection 0.85), and reached 0.721 in temporal validation with improved reclassification (net reclassification improvement 0.11 and integrated discrimination improvement 0.055) and clinical net benefit at thresholds 0.2 to 0.5. Pretransplant gut microbiome signatures were independently associated with early acute rejection. Microbiome-augmented models outperformed clinical-only models and remained robust in temporal validation, supporting microbiome-based pretransplant risk stratification.