Multi-omic characterization of imaging invisible prostate tumors reveals microenvironmental drivers of PET visibility

Background: Up to 20% of aggressive localized prostate tumors are invisible on PSMA PET or MRI, limiting diagnostic accuracy and treatment selection. As both imaging modalities are routinely used in newly diagnosed patients, defining the molecular basis of imaging visibility is critical for imaging-driven precision medicine. Despite prior evidence implicating stromal and metabolic pathways, the biology of imaging visibility has not been resolved using directly linked multi-omic data. To address this, we analyzed prostate tumors with matched imaging and multi-omic profiles to define molecular programs underlying PET and MRI visibility. Methods: A total 71 grade group 2-3 prostate tumors were obtained from 59 patients who underwent paired PSMA PET and MRI before prostatectomy. Macrodissected tumor regions underwent bulk RNA sequencing, targeted DNA sequencing, and proteomics. Associations with PET SUVmax and MRI visibility (PIRADS≥4) were tested using multivariable models adjusting for clinico-pathologic features (FDR>0.05). Gene set enrichment was performed on ranked test statistics using preranked GSEA with MSigDB Hallmark, KEGG, and Reactome annotations. Concordant molecular features were cross-referenced across DNA and protein datasets to identify pathways consistently associated with imaging visibility. Results: MRI-invisible tumors showed downregulation of proliferative and metabolic pathways, including DNA repair and glycolysis, consistent with a quiescent phenotype. PET visibility was associated with transcriptionally and metabolically active states. Furthermore, integrative RNA, DNA, and proteomic analyses converged on extracellular matrix remodeling and immune-regulatory processes as major correlates of PSMA PET signal intensity, indicating that stromal and immune architecture shape imaging detectability. Surfaceome profiling further highlighted cell-surface proteins involved in matrix and immune interactions as candidate alternative biomarkers for PET imaging in PSMA low tumors. Conclusions: These data suggest PSMA PET visibility reflects biologically active tumor states characterized by DNA damage response and extracellular matrix remodeling, while MRI invisibility corresponds to a more quiescent and metabolically repressed phenotype. Integrated molecular and surfaceome analyses indicate that stromal and immune alterations contribute to PSMA uptake and imaging detectability. Together, these findings support imaging visibility as a biologically driven marker with potential to refine risk assessment and guide precision management in prostate cancer.