Identification of Heterochromatin Markers As Epigenetic Drivers of Neuroendocrine Prostate Cancer (NEPC) Using Single Cell Tracking and Supervised Learning

Prostate cancer (PC) is the second leading cause of cancer related death of male in the US. The majority of PC related deaths can be attributed to the acquisition of drug resistance. Specifically, most PC patients receive hormonal therapy, also known as androgen receptor depletion treatment (ADT). ~20-25% patients receiving treatments relapse by developing a novel neuroendocrine prostate cancer (NEPC) phenotype that is essentially untreatable, with survival ranging from 7 months to 2 years. Understanding the transition of androgen responsive PC to NEPC is thus critically important to improve the treatment outcomes of PC patients and increase their respective survival rates. Increasing literature evidence suggests that epigenetic mechanism, including DNA methylation and EZH2 dysregulation plays a vital role in driving the transition towards NEPC. Given the intrinsic epigenetic heterogeneity within PC cells, it is extremely challenging to dissect the epigenetic driver in NEPC establishment. Enabled by novel single cell epigenetic tracking tools developed in our group, we tracked dynamic changes in 5mC, H3K9me3 and H3K27me3 of a PC cell line, LNCaP undergoing Enzalutamide (Enza, an AR inhibitor) treatment. After treating cells with 5 µM of Enza for 6 days, LNCaP cells transitioned into NEPC cells that stain positive for neuron specific enolase (NSE), During the time course, we found that global heterochromatin marker abundance decreased significantly with reduced area of condensed chromatin regions. We extracted single cell and subnuclear features for each cell to enable data-driven analysis to identify driving epigenetic features for NEPC establishment using supervised machine learning algorithm with multiclass classification. Our results suggest that loss of heterochromatin markers coupled with loss of chromatin compaction marked by H3K27me3 can potentially drive NEPC development. Collectively, we established a novel single-cell analysis workflow to monitor dynamic epigenetic modification level change, associate it with disease progression and predict potential drivers for specified disease phenotype.