Abstract B002: Determination of tumor PSMA expression and response to Lutetium-PSMA in men with prostate cancer using a novel epigenomic liquid biopsy platform

• Published in: Clinical cancer research 2024-11, Vol.30 (21_Supplement), p.B002-B002
• Main Authors: Ravi, Praful, D'Ipppolito, Anthony, Wurster, Jenna, Savignano, Hunter, Stoltenberg, Hailey, Mary Lee, Gwo-Shu, Borges, Nuno, Kramer, Nicole, McGillicuddy, Mary, Tamakloe, Tyrone, Tran, Baovy, O'Brien, Charlene, Coyne, Michael, Painter, Corrie, Gorthi, Aparna, Eaton, Matthew, Barrett, Carl, Jacene, Heather, Berchuck, Jacob E
• Format: Article
• Language: English
• ISSN: 1557-3265; 1557-3265
• DOI: 10.1158/1557-3265.LIQBIOP24-B002

Background: 177Lu-PSMA-617 (Lutetium-PSMA) is an FDA-approved radiopharmaceutical therapy for men with metastatic castration-resistant prostate cancer (mCRPC). Treatment eligibility requires a PSMA PET scan to confirm tumor PSMA expression. With therapeutic strategies in development targeting an array of cell surface proteins, there is an emerging unmet need for minimally invasive biomarkers to quantify tumor drug target expression and predict therapeutic response. We performed comprehensive epigenomic profiling on 1 mL of plasma from 50 men with mCRPC and demonstrated an accurate cell-free (cf)DNA-based readout of tumor PSMA expression and response to Lutetium-PSMA. Methods: We collected plasma at the time of PSMA PET scan in men with mCRPC and analyzed the PET images to quantify total tumor SUVmean, the radiographic feature that most strongly correlates with response to Lutetium-PSMA. In the same patients, we collected plasma at the time of Lutetium-PSMA treatment initiation and collected detailed clinical outcomes data. We profiled genome-wide cfDNA epigenomic signals across histone modifications associated with active enhancers and promoters, and DNA methylation, on these plasma samples and identified putative gene-regulatory loci at the FOLH1 (PSMA) locus. Using these signals, we trained a model to predict PSMA PET SUVmean and validated performance in an independent set of samples held out of model training. The results below are limited to patients with cfDNA tumor fraction > 3%. Results: We present results for the initial 50 patients from a larger cohort. A predictive model trained on FOLH1 gene-regulatory activity measured by epigenomic signals in plasma from 21 patients demonstrated a significant correlation between predicted and actual PSMA PET SUVmean on both internal leave-one-out cross validation (R = 0.71; p < 0.05) and when applied to an independent validation cohort of plasma samples from eight patients (R = 0.78; p < 0.05). Assessment of plasma samples collected from 45 patients immediately prior to treatment with Lutetium-PSMA revealed that predicted PSMA PET SUVmean strongly associated with clinical outcomes, with the top tertile of patients demonstrating significantly (p < 0.05) improved PSA progression-free survival, clinical-radiographic progression-free survival, time to next treatment, and overall survival compared to patients in the bottom two tertiles. Furthermore, associations of predicted PSMA PET SUVmean with clinical outcomes