Network based analysis to identify master regulators in prostate carcinogenesis

Prostate cancer (PCa) is the second most common tumor diagnosed in man, for which robust prognostic markers and novel targets for therapy are lacking. Major challenges in PCa therapeutical management arise from the marked intra and inter-tumors heterogeneity, hampering the discernment of molecular subtypes that can be used to guide treatment decisions. For this reason, virtually all patients undergoing standard of care androgen deprivation therapy for locally advanced or metastatic cancer, will eventually progress into the more aggressive and currently incurable form of PCa, referred to as castration resistant prostate cancer (CRPC). By exploiting the richness of information stored in gene-gene interactions, I tested the hypothesis that a gene regulatory network derived from transcriptomic profiles of PCa orthografts can reveal transcriptional regulators to be subsequently adopted as robust biomarkers or as target for novel therapies. Among the 1308 regulons identified from the preclinical models, Cox regression analysis coherently associated JMJD6 regulon activity with disease-free survival in three clinical cohorts, outperforming three published prognostic gene signatures (TMCC11, BROMO-10 and HYPOXIA-28). Given its potential role in a number of cancers, in-depth investigations of JMJD6 mediated function in PCa is warranted to test if it has a driver role in tumor progression. Encouraged by the predictive abilities of the gene regulatory network inferred from transcriptomics data, I explored the possibility of integrating the regulons structure with data from the proteomes of the same preclinical orthografts studied by RNA sequencing. This approach leverages the complementarity between gene and protein expression, to increase the robustness of the statistical analysis. Similar to gene-gene co-expression profiles, protein-protein co-expression data can provide a distinct representation of the molecular alterations underlying a biological phenotype. By implementing a pipeline to integrate modules derived from transcriptomic based regulons and proteinprotein interactions respectively from matched RNA-seq and quantitative proteomic data, I obtained 516 joint modules entailing a median of four protein complexes (range 1-41) per individual transcription factor regulon, providing new insight into its regulatory mechanisms. In the final step of the analysis, a permutation-based enrichment of the genes/proteins integrative modules implicated MID1 (an E3 ubiquitin ligase belonging to the family of tripartite motif containing protein) to be a driver transcriptional regulator in CRPC. In fact, MID1 module was the only candidate for which gene-gene and proteinprotein interactions were supported (p-value < 0.05) by both differentially expressed genes and proteins obtained from the CRPC vs PC contrast. Finally, I wished to test the usefulness of a network based investigation as a tool to identify predictors of treatment response. To this end, I obtained transcriptomics data from an in vivo subcutaneous xenograft treatment experiment (namely mychophenolic acid or abiraterone/ARN-509 as stand alone treatment or in combination) and determined which regulons were inferred to be active in the tumours following treatment. The androgen receptor positive human LNCaP C4-2b prostate cancer cells were injected into mice. The effects of treatment were assessed by collecting serial tumor sizes and by performing RNAseq at the designed endpoint of the study. Noteworthy, the gene graph enrichment analysis provided novel hypothesisbehind the anti- proliferative effect of mychophenolic acid (MPA), suggesting the SET proto-oncogene to be a target for MPA mediated suppression of proliferation. Of note, standard gene-set enrichment analysis, without input on specific gene-gene interactions, was not effective in prioritising the SET protooncogene, demonstrating the usefulness of the network based investigation. Collectively, data presented in this thesis provides an alternative perspective for the analysis of multi-omics profiles from PCa and highlights the importance of gene-gene and protein protein interactions in prostate cancer growth and progression