Prostate cancer cells acquire diverse mechanisms of castration resistance under the selective pressure of treatment. This includes expression of constitutively active androgen receptor (AR) variants. Whether AR variants drive resistance is contested, because they often co-exist with full-length AR. Yet, some tumours with AR genomic structural rearrangements (AR-GSRs) only express AR variants and not full-length AR. Therefore, our objective was to investigate how truncated variants shape the AR cistrome and responses to treatments, with or without full-length AR.
We selected patient-derived xenografts of prostate cancer from the Melbourne Urological Research Alliance (MURAL) cohort. We compared the landscapes of AR binding using chromatin immunoprecipitation sequencing and transcriptomic profiles using RNA sequencing. We also determined the responses of tumours to castration and bipolar androgen therapy in vivo.
We identified a distinct group of patient-derived models with structural rearrangements of the AR gene. These tumours all expressed ARv567es, a constitutively active AR variant. They had varying levels of full-length AR, depending on the nature of the genomic rearrangements. These tumours had distinctive AR cistrome profiles, gaining some AR binding sites and losing others compared to tumours without AR structural rearrangements. ARv567es-positive tumours also had a different profile of H3K27ac histone marks. Moreover, we defined transcriptional differences, with depletion of canonical AR-regulated gene signatures but enrichment of AR-repressed genes. Consistent with ARv567es having ligand-independent activity, ARv567es-positive tumours were resistant to castration and bipolar androgen therapy. In tumours that co-express full-length AR, this involves disruption of the autoregulatory loop that modulates AR levels.
The emergence of ARv567es through AR gene rearrangements alters the pattern of AR binding, reprograms the transcriptome, and is associated with resistance to therapies targeting the AR ligand-binding domain. Thus, AR genomic structural rearrangements and ARv567es expression are potential markers to guide treatment decisions.