Introduction:

Prostate cancer (PC) is extremely common and biologically diverse. Since treatment efficacy varies considerably between patients, efforts to better characterize PC have resulted in new molecular imaging strategies including positron emission tomography (PET) targeting the cell membrane protein prostate-specific membrane antigen (PSMA). While PSMA PET is increasingly used in clinical practice, PSMA expression can be highly variable and likely reflective of differences in tumor biology. Thus, the information from PSMA expression might be leveraged to help individualize targeted molecular tumor testing and treatment selection.

Methods:

We hypothesized that by comparing the molecular profiles of tumors with high and low PSMA RNA expression (FOLH1), we could define the biological processes associated with PSMA expression. We first correlated PSMA RNA expression with SUVmax in a prospective cohort who underwent surgery (NCT03392181; n=55). We then queried a dataset of 7,000 primary prostate tumors from the GRID database (NCT02609269) that were 1:1 grade and stage matched by tumors from the lowest and highest pentiles of PSMA expression (n=918 each). We performed Hallmark gene set differential expression (n=50) with validation in The Cancer Genome Atlas (TCGA; n=491). All differentially expressed signatures were subjected to multivariable logistic regression accounting for grade, stage, PSA, and patient age. Publicly available datasets were assessed to validate correlated pathways.

Results:

In NCT03392181, PSMA RNA expression was moderately correlated with SUVmax (Spearman ρ = 0.41). A total of 26 hallmark pathways correlated with PSMA expression in the GRID and TCGA cohorts and upon multivariable analyses (Figure). Notably, PSMA high tumors tended to be enriched for androgen response, fatty acid metabolism, and DNA repair pathways while PSMA low tumors exhibited increased immune-related and hypoxia pathways. In validating assessments, tumors from patients treated with two months of hormone therapy expressed lower PSMA compared to controls (GEO series: GSE102124). In TCGA, we validated a negative association between PSMA and tumor infiltrating lymphocytes on immunohistochemistry and a positive association with genomic markers of DNA damage. In LNCaP cells, increasing exposure to a fatty acid synthesis inhibitor led to lower PSMA expression (GSE114016), while PSMA negatively correlated with upregulation in hypoxia pathways (GSE195571).

Conclusion:

We discovered and validated several pathways associated with PSMA expression in primary PC. These findings could inform precision management approaches by optimizing interpretation of PSMA PET. Given the moderate correlation between PSMA RNA expression and SUVmax, future validation in a cohort with additional PET metrics and PSMA protein levels is warranted.

Funding: Dr. Allen and Charlotte Ginsburg Fellowship in Precision Genomic Medicine and The Simon-Strauss Foundation research award (ABW)