Abstract PR027: Spatial heterogeneity atlas of prostate cancer evolution (SHAPE): Spatial and genomic landscapes of advanced prostate cancer with and without Lu-PSMA therapy

Abstract Background: PSMA expression is heterogeneous across advanced prostate cancer, limiting the durability of PSMA-radioligand therapy (RLT). To characterize tumor heterogeneity and identify determinants of RLT resistance, we established the SHAPE cohort (Spatial Heterogeneity Atlas of Prostate Cancer Evolution). SHAPE is a multimodal resource integrating PET imaging, clinical treatment history, histopathology, immunohistochemistry, spatial transcriptomics, and genomic profiling across 46 patient tumor samples collected through rapid autopsy, biopsy, and surgical resection. The cohort spans adenocarcinoma, castration-resistant (CRPC), and neuroendocrine (NEPC) prostate cancer states and includes eight xenograft models. Methods: Twenty-five samples profiled using the 10x Genomics Visium spatial transcriptomics platform, comprising 195, 905 spots, were integrated and analyzed using Seurat (v5) in R. Targeted genomic profiling was performed using the Dana-Farber Cancer Institute OncoPanel platform on 36 samples derived from 7 patients and 2 animal models. Among the cohort, seven cases had prior Lu-PSMA exposure, and ten cases were untreated. Eleven clinical and preclinical samples had PSMA PET/CT, including six with both pre- and post-treatment imaging. Data integration connected structural genomic events with spatial lineage programs and imaging features. Results: RLT–treated tumors exhibited a copy-number–dominant profile compared with untreated tumors. Recurrent alterations enriched in treated tumors included PTEN deletion (p0. 01), KLLN deletion (p0. 01), AURKA gain (p0. 01), and ZNF217 gain (p0. 01). Treated tumors also showed higher per-sample copy number loss (47. 6 vs 37. 5, p0. 05) and amplification burden (7. 91 vs 2, p0. 05). In contrast, untreated tumors displayed more single nucleotide variants (14. 9 vs 6. 3, ns) and indels (4. 7 vs 1. 9, ns). Spatial transcriptomics provided complementary insights. Treated samples with PSMA SUV below the liver reference demonstrated higher copy-number loss burden (p0. 05) and contained unique transcriptomic clusters enriched for glucose metabolism. These clusters displayed variable expression of classical NEPC markers but consistently upregulated genes involved in glucose uptake and utilization. In preclinical studies, head-to-head FDG and Ga-PSMA imaging of WCM12 xenografts demonstrated that FDG, as a surrogate marker of glucose uptake, was significantly increased following relapse after PSMA-RLT (FC, p0. 05). Conclusions: The SHAPE cohort integrates genomic, spatial, and imaging data to characterize tumor heterogeneity in advanced prostate cancer. Genomic profiling identified PTEN/KLLN loss and AURKA/ZNF217 gain as recurrent alterations enriched in Lu-PSMA–exposed tumors, while spatial transcriptomics revealed PSMA-low metastases with distinct. clusters enriched for glucose metabolic programs. Together, SHAPE can inform rational strategies for combination or alternative targeted therapies. Citation Format: Martin K. Bakht, Jacob Egelberg, Yasutaka Yamada, Louise Clark, Xiao Wei, Varadha Balaji Venkadakrishnan, Anthony P. Belanger, Alice Bernard-Tessier, Sarah J. Hill, Francesca Khani, David J. Einstein, Mary-Ellen Taplin, Eliezer Van Allen, Heather A. Jacene7, Himisha Beltran. Spatial heterogeneity atlas of prostate cancer evolution (SHAPE): Spatial and genomic landscapes of advanced prostate cancer with and without Lu-PSMA therap abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Innovations in Prostate Cancer Research and Treatment; 2026 Jan 20-22; Philadelphia PA. Philadelphia (PA): AACR; Cancer Res 2026;86 (2Suppl): Abstract nr PR027.