Transcriptomic and DNA methylation alterations in prostate cancer metastasis

Epigenetic modifications remain dynamic in most somatic cells to enable flexible gene activity. Simultaneously, this creates a cellular environment prone to epigenetic mistakes‚ÄövÑvp and this is evident in most diseases, including cancers. It is well established that cancer initiation and progression is caused by a complex series of genetic and epigenetic changes encompassing altered patterns of DNA methylation, histone modifications and the physical chromatin structure. The consequence of these collective changes is an abnormal gene expression signature that can act as a catalyst for disease. Prostate cancer is a highly heterogeneous disease and owing to this, only a few highly penetrant genomic drivers have been associated with the disease. Epigenetic studies can provide additional insight into disease onset and progression, with DNA methylation being one of the most studied epigenetic alterations in prostate cancer. The transcriptomic and DNA methylation profiles of 'local' and 'metastatic' cells are typically integrated and compared to benign or 'normal' states. However, the metastasis of primary tumours is the main cause of prostate cancer related deaths, and it is important to identify gene expression and epigenetic signatures that may drive this process. To address this, the overall aim of this thesis was to identify and validate divergent transcriptomic events that may explain the biology of metastasis by drawing comparisons between cells obtained from local tumours and those of metastasised cancer. A cell line model of prostate cancer that represents 'localised' prostate cancer (22Rv1 cell line) and two prostate cancer 'metastasis' cell lines (LNCaP, lymph node; PC3, bone) was utilised, in combination with whole transcriptomic data and genome-wide DNA methylation analysis to identify candidate genes and test the impact of DNA methylation inhibitors. Putative drivers of metastasis were identified in three parts, firstly by differential gene expression; secondly, by differential alternative splicing and thirdly, by differential promoter DNA methylation. Using whole transcriptomic RNA sequencing (RNA-seq), genes of interest exhibited at least two fold difference between 22Rv1 cells and either of LNCaP or PC3 cell lines, with the genes that were commonly identified in both pairwise comparisons subject to clustering and pathway analyses. This revealed a list of 629 robust candidates, including GUCY1A2 and KISS1R, which displayed higher levels of expression in LNCaP and PC3 metastatic cell lines compared to the localized 22Rv1 cancer cells. Additionally, global alternative splicing analysis revealed that QSOX1 showed intron retention concomitant with overall higher levels of gene expression in the LNCaP and PC3 metastatic cell lines compared to the localized 22Rv1 cancer cells. DNA methylation information was generated by Illumina Infinium MethylationEPIC Beadchip array. A total of 16,866 differentially methylated regions (DMRs) were identified between 22Rv1 and PC3 comprising of 218,341 total CpG dinucleotides. These DMRs were filtered based on FDR<0.25, and mean beta fold change >0.25, and to include regions with promoter methylation patterns that tightly correlated with gene expression patterns. These stringent criteria identified 147 regions of interest spanning 2,050 CpG sites, including two DMRs overlapping the PXMP4 and SLC36A4 gene promoters. Interestingly, the DNA methylation profiles of these promoters was inverse, with PXMP4 displaying DNA hypermethylation and SLC36A4 showing DNA hypomethylation in PC3 compared to 22Rv1 cells. As predicted, this DNA methylation pattern correlated with expression patterns; PXMP4 was repressed while SLC36A4 was expressed in the metastasized PC3 compared to localized 22Rv1 cancer cells. To investigate whether these genes were regulated by DNA methylation, genes of interest were tested for responsiveness to the DNA methyltransferase inhibitor, 5-aza-2'- deoxycytidine. Drug treatment caused a significant upregulation of PXMP4 in the metastatic cell lines while no effect was seen in 22Rv1 cells. No significant change in expression was noted for SLC36A4, GUCY1A2 and KISS1R. This suggests that repression of PXMP4 in metastasis could possibly be due to the modulation of DNA methylation. Overall, this study provides insights into the genome-wide transcriptomic and methylation differences identifying a number of promising candidate genes ideal for further functional investigation in the process of prostate cancer metastasis.