Computational Analysis of Treatment Resistance in Prostate Cancer

Prostate cancer is the second most commonly diagnosed cancer in men worldwide and the most frequent cause of cancer death in men in 48 countries, making it a significant source of hardship and burden on both patients and healthcare systems. Initially, the disease depends on male hormones called androgens for growth, allowing for the eradication of cancer cells using anti-androgenic drugs. However, in many patients, new molecular changes will render cancer cells insensitive to androgens, resulting in treatment resistance. The cancer can then recur in a more aggressive, androgen-independent form called castration resistant prostate cancer. New therapies have been developed in recent years for the treatment of castration resistant prostate cancer that instead target the receptor for these hormones, the androgen receptor. However, the efficacy of these treatments is also limited by inherent and acquired resistance. Understanding these mechanisms of resistance in detail is key for improving patient survival in this ultimately lethal disease setting, both in terms of suggesting potential new molecular targets for treatment and maximizing the benefit of existing therapies.

The aim of this doctoral work was to implement and utilize computational methods to understand the impact of treatment on the prostate tumor genome, transcriptome, and epigenome, as well as investigate mechanisms through which prostate cancer cells are either susceptible or resistant to current therapies. This included investigation of the specific molecular changes that may make cancer cells insensitive to particular treatments and the mutational processes within cancer cells that can generate them. First, mechanisms of cross-resistance to castration-resistant prostate cancer treatments abiraterone and enzalutamide were studied using serial samples of cell-free DNA released by cancer cells into the bloodstream. The only recurrently observed resistance mechanism was the selection of aggressive androgen receptor genotypes. This suggests that the androgen receptor continues to play a central role after being targeted by multiple therapies, and that renewed targeting of the androgen receptor pathway even after abiraterone and enzalutamide may still benefit patients.

In the second study, single-cell transcriptomic and epigenomic sequencing was used to study the development of resistance to enzalutamide in pre-clinical models, revealing gene expression profiles associated with pre-existing and persistent cell subpopulations. These profiles could be used to stratify the treatment responses of prostate cancer patients, showing that studying single cells on the level of the transcriptome and epigenome can yield information about previously unknown predictors of treatment effectiveness and resistance.

The third study of the thesis compared the genomic and epigenomic profiles of cell subpopulations that were resistant to and eradicated by carboplatin chemotherapy to propose an explanation for their differential response to therapy. The eradicated cell subpopulation exhibited unique alterations in the DNA repair genes FANCI and EYA4, partially due to retrotransposon activity, suggesting that these changes could have facilitated their susceptibility to therapy. With further studies, this approach may be able to identify new molecular targets in patients for converting resistant cell subpopulations into eradicable ones.

Finally, recurrent mutations in the untranslated region of the prostate cancer transcription factor FOXA1 were investigated in cell-free DNA from castration- resistant prostate cancer patients. These mutations were found to likely be caused by head-on transcriptional collisions and transcriptional stalling. While the mutations were not linked to patient response to abiraterone or enzalutamide, the mechanism of head-on transcription could result in mutations associated with treatment outcome elsewhere in the castration-resistant prostate cancer genome.

Together, the work presented in this thesis demonstrates that computational analysis of genomic, transcriptomic, and epigenomic cancer cell profiles can reveal cell populations involved in treatment resistance and provide mechanistic insights into the differential behavior of cells in response to therapy. Together with further studies, these findings can help determine how to better treat each patient’s unique prostate cancer and help improve patient outcomes.