Differences in genomic, transcriptomic and immune landscape of prostate cancer (PCa) based on site of metastasis (mets)


Umang Swami, Shayan Nazari, Georges Gebrael, Andrew Elliott, Aditya Bagrodia, Dhruv Puri, Chadi Nabhan, Rana R. McKay, Emmanuel S. Antonarakis, Neeraj Agarwal

Background: In patients with metastatic PCa, specific sites of mets are associated with differential overall survival, with lymph node (LN) only disease having the best survival, followed by bone, lung, and liver, with each showing progressively worse survival. However, the biology of tropism to specific sites and differential outcomes has not been fully explored. Herein, utilizing a large, multi-institutional, real-world dataset, we hypothesized that differences in molecular and immune landscape of PCa based on site of mets may explain different clinical outcomes.

Methods: DNA (592-gene panel or whole exome) and RNA (whole transcriptome) sequencing was performed for primary PCa (PPCa) and unpaired PCa mets sites (non-visceral: NV and visceral: V) and sequenced at Caris Life Sciences (Phoenix, AZ). Mismatch repair deficiency/microsatellite instability-high (dMMR/MSI-H) was tested by IHC/NGS, respectively. Tumor mutational burden (TMB) (high .10 mut/Mb) was assessed. Androgen receptor (AR) signaling score and Neuroendocrine Prostate Cancer (NEPC) score were calculated using previously defined RNA-based signatures. Pathway enrichment was determined by Gene Set Enrichment Analysis (GSEA). Immune cell fractions calculated by deconvolution of WTS using quanTIseq. Statistical significance was determined by chi-square and MannWhitney U tests. All analyses done with CODEai.

Results: 6074 PCa samples, including 3411 PPCa, 1634 NV (bone = 784, LN = 850) and 1029 V (liver: 468, lung: 177, bladder: 241, other: 143) were evaluated. Age at sample collection ranged 30 to 70+ years. Compared to PPCa, significantly higher mutation rates were observed for multiple genes in V (RB1 8% vs 3%, TP53 43% vs 31%, PTEN 12 % vs 8%, CTNNB1 8% vs 3%, AR 13% vs 1%, APC 15% vs 5%, AKT1 4% vs 2%; p,0.05 for all) and NV mets (AR 9% vs 1%, SPOP 7% vs 10%, TP53 38% vs 31%, CTNNB1 6% vs 3%, and KMT2C 8 % vs 5%; p,0.05 for all). Alterations of RB1, APC and AR were more frequent, and SPOP less common in liver mets compared to PPCa, while SPOP, KRAS, PIK3CA, APC and BAP1 mutations were more common in lung mets (p , 0.05 for all). Compared to PPCa, V and NV mets had higher AR signaling and interferon-g scores, while V mets had higher NEPC scores and were more frequently PD-L1+ (SP142) and TMB-high (p,0.05 for all). No differences in dMMR/ MSI-H was observed between PPCa, V and NV mets. AR-V7 variant prevalence was significantly higher in V (36%) and NV (33%) mets compared to PPCa (7%, p-value , 0.05). GSEA identified E2F, G2M checkpoint and MYC target pathways as being more upregulated in both V and NV mets compared to PPCa (q,0.05). Compared to PPCa, V and NV mets were significantly less enriched in macrophage M2, NK cells and Tregs, while V mets were also less enriched in B cells and neutrophils (p,0.05).

Conclusions: We elucidate molecular and immunologic mechanisms of metastatic tropism in advanced PCa. These data may facilitate future drug development.