Read the Full Video Transcript

Andrea Miyahira: Hi, I'm Andrea Miyahira at the Prostate Cancer Foundation. Here with me today is Dr. Raunak Shrestha of UCSF. He will review his latest paper, "An Atlas of Accessible Chromatin in Advanced Prostate Cancer Reveals the Epigenetic Evolution During Tumor Progression." This was published in Cancer Research. Raunak, thanks so much for joining me today.

Raunak Shrestha: Thanks, Andrea. Thank you for the invitation to share my work, and it's great to be here with you.

This paper is called "An Atlas of Accessible Chromatin in Advanced Prostate Cancer Reveals Epigenetic Evolution During Tumor Progression."

Androgen deprivation therapy (ADT) and androgen receptor signaling inhibitors (ARSI) are the major therapeutic interventions in advanced prostate cancer. And its prolonged use gives rise to diverse mCRPC subtypes, ranging from AR-high adenocarcinoma to AR-low neuroendocrine phenotype, and several mixed phenotypes, such as AR-positive, NE-positive, or known as double positive, and AR-negative neuroendocrine negative that is known as double negatives.

Several prior studies have categorized genomic and transcriptomic differences between the mCRPC subtypes, but very little is known about epigenomic differences, particularly chromatin accessibility between these subtypes. There have been a number of works studying chromatin accessibility and chromatin remodeling in prostate cancer, using different high-throughput technologies, and have found that AR overexpression drives genome-wide chromatin relaxation in prostate cancer.

Specifically, I would like to highlight this study published in Science 2022 by Ekta Khurana's group in Cornell. They utilized the ATAC-seq technique to study chromatin accessibility in prostate cancer, and identified four subgroups of advanced prostate cancer: the AR-high group, the neuroendocrine prostate cancer group, a group with stem cell-like features, and another enriched in the Wnt pathway. Although there are some conceptual and analytical similarities between this study and ours, this particular study, they measured chromatin accessibility in context of prostate cancer cell lines, organoids, and PDX models. And our studies used clinical samples—metastatic biopsies from prostate cancer patients.

In an effort to identify genomic and epigenomic changes in metastatic prostate cancer, the West Coast Dream Team consortium, with generous support from the Prostate Cancer Foundation and Stanford Cancer, have been able to create this unique and large collection of genomic and epigenomic profiles of mCRPC biopsies. As a part of this effort, we have recently generated ATAC-seq profiles of mCRPC biopsy to measure chromatin accessibility. We performed ATAC-seq sequencing on 70 metastatic prostate cancer biopsies, and this is the largest collection of ATAC-seq profiles of metastatic prostate cancer to date. And later, we integrated the ATAC-seq data with the RNA-seq data from matched tumors. And today, I will describe our study characterizing these ATAC-seq profiles in metastatic prostate cancer.

Now going over the results, first of all, we were interested in understanding if chromatin accessibility profiles of metastatic prostate cancer are unique. For this, we compared the ATAC-seq profile of metastatic prostate cancer with that of benign and localized prostate cancer, and performed principal component analysis. Each dot in this PCA plot represents an individual sample. Localized prostate cancer are highlighted in the blue color. Benign prostate in yellow. Red dots are the metastatic prostate cancer adenocarcinomas. And the neuroendocrine prostate cancer in green. We found that metastatic prostate cancer have distinct chromatin accessibility profiles, and among metastatic cancer, the neuroendocrine prostate cancer clustered distinctly.

Next, we were interested to find the distinct group of metastatic prostate cancer based on their chromatin accessibility profiles. Using unsupervised clustering, we found three different groups of metastatic prostate cancer samples: the cluster one, two, and three, as shown in the pairwise correlation heat map of the samples. We also measured gene expression-based AR pathway score and the neuroendocrine score, as illustrated in the box plot on the right. Interestingly, the cluster three was correlated with AR-high score and a low neuroendocrine score. And the cluster one, with high neuroendocrine score and low AR score. This suggested that the chromatin accessibility in metastatic prostate cancer is associated with subtypes linked to androgen signaling.

This led us to divide the metastatic prostate cancer cohort into five different subtypes based on AR and neuroendocrine status. And then, we measured the differences in chromatin accessibility between these subtypes. Interestingly, we found that these subtypes had quite distinct chromatin accessibility profiles, which suggested that chromatin accessibility potentially contributes towards differentiation of metastatic prostate cancer samples. This further motivated us to identify transcriptional regulators, or transcription factors associated with these metastatic prostate cancer subtypes.

ATAC-seq technique gives us an opportunity to interrogate a number of transcription factors in a single assay. Whenever a transcription factor binds to chromatin, a dip in the ATAC-seq signal is observed, known as transcription factor footprints. In this figure, this is represented by the drop in the red ATAC-seq signal. And using DNA motif matching technique, we can identify specific transcription factors through the footprints.

To demonstrate the power and utility of ATAC-seq to study metastatic prostate cancer associated with transcription factors, we measured transcription factor binding of more than 500 transcription factors in metastatic prostate cancer samples, and compared the transcription factor binding between the metastatic prostate cancer subtypes. Using this computational technique, our study identified 203 transcription factors associated with different metastatic prostate cancer subtypes. These 203 transcription factors are illustrated as a circular heat map, where each subtype is represented as individual rim of the circle. The AR-high subtype is in the outer rim of the circular heat map, and the AR-negative NEPC and double negative prostate cancer are in the inner rim of the individual rim of the heat map. The darker the color of the heat map, the stronger the association of transcription factor with individual subtypes. And here, we identified both well-studied transcription factors, as well as potentially novel transcription factors associated with different metastatic prostate cancer subtypes.

We further strive to find the influential transcription factors driving metastatic prostate cancer subtypes. For this, we correlated the chromatin accessibility profiles with RNA-seq gene expression from matched tumor samples and identified downstream target genes in the metastatic prostate cancer-associated transcription factors.

As shown in the figure, we rank the genes based on the number of downstream target genes a transcription factor potentially regulates. The top influential transcription factors are some of the well-studied transcription factors such as FOXA1 in AR-high subtypes, and ASCL1 in the neuroendocrine prostate cancer subtype. We also identified a potentially novel transcription factor ZNF263, which was identified as an influential transcription factor in the double negative subtype. And here, we further demonstrate how we can study transcription factor regulation using ZNF263 transcription factor as an example, utilizing different analytical techniques integrating ATAC-seq and RNA-seq.

ZNF263 has never been studied in context of prostate cancer. To understand its role in prostate cancer, we wanted to know if ZNF263 acts as an activator or a suppressor of its downstream target. For this, we examined changes in gene expression levels of its target in the presence or in the absence of ZNF263 binding on the promoter region. Interestingly, we found that the genes were highly expressed in the presence of ZNF263 as compared to those without ZNF263 binding. Further, we found that these genes were enriched in MYC signaling targets. Thus, these results suggest that ZNF263 potentially co-regulates a fraction of MYC target genes.

We next assessed the combined effect of ZNF263 and MYC binding on the downstream target genes. We found that the genes with concomitant binding of ZNF263 and MYC had higher expression, as compared to when either ZNF263 or MYC was bound to the promoter region. This potentially suggests that there is a potential collaborative activation of MYC targets in metastatic prostate cancer by ZNF263. So in this way, we can leverage ATAC-seq assay and transcription factor footprinting technique to study transcription factor regulation.

In conclusion, chromatin accessibility profiles in metastatic prostate cancer are distinct, associated with androgen receptor signaling, contribute towards differentiation of subtypes, and allow us to identify influential transcription factor regulators of the subtype.

With this, I would like to thank my mentors, Dr. Felix Feng and Dr. David Quigley and Dr. Mathieu Lupien, and all co-authors, and the West Coast Prostate Cancer Dream Team consortium and Prostate Cancer Foundation for generously funding my work. Thank you so much.

Andrea Miyahira: Thank you so much, Raunak, for sharing this study with us, and congratulations to you and your team on generating this incredible database.

Raunak Shrestha: Thank you.

Andrea Miyahira: So can you tell from this study which transcription factors would be best therapeutic targets, based on which gene expression programs could be disrupted?

Raunak Shrestha: That's an excellent question. So one of the major goals of this study was to map out the transcription factor dependencies of metastatic prostate cancer subtypes and their transcriptional programs. And we have identified where transcription factors bind in the chromatin and the group of target genes the transcription factors potentially modulate. And we have also ranked the influence of transcription factors in each subtype.

Now drugging transcription factors themselves is a very challenging process, and I believe that our study will be a valuable resource to nominate transcription factors for further functional experimentation, based on the ranking of influence of transcription factors. So in short, the answer to your question is, the top influential transcription factors in different metastatic prostate cancer subtypes are some of the prime candidates for therapeutic targeting, and further functional experimentation is required to pinpoint any particular transcription factor and to nominate for therapeutic interventions.

Andrea Miyahira: Okay, thank you. Can this study be used to identify plasticity programs? So how mCRPC adenocarcinomas transdifferentiate into these other subtypes?

Raunak Shrestha: Yes, certainly. Our study has categorized the chromatin accessibility states in different metastatic prostate cancer subtypes, and how chromatin rewires in different subtypes to activate or suppress different transcriptional programs. We could further investigate the differences in the chromatin rewiring to understand lineage plasticity in metastatic prostate cancer.

Andrea Miyahira: Thank you. And what are your next steps, and are there any translational plans?

Raunak Shrestha: So in the current study, we investigated chromatin accessibility from DNA obtained from the metastatic solid tumor biopsies. And this gives us a single snapshot of chromatin states at a particular time point during cancer progression. Now, we are further interested in studying how chromatin accessibility states evolve over time, and how to effectively utilize epigenomic markers to predict and understand treatment resistance. And for this, we could use cell-free DNA obtained from blood plasma, liquid biopsy samples from highly aggressive metastatic prostate cancer patients. And using whole genome sequencing of the cell-free DNA, we could measure the cell-free DNA fragmentomic patterns to measure nucleosome position in the chromatin, and infer chromatin accessibility. So we are planning to develop machine learning tools to learn the patterns of chromatin accessibility from cell-free DNA fragmentomic signal, which could be used for cancer phenotyping or subtyping, gene activity prediction, and therapeutic resistance predictions.

Andrea Miyahira: Okay. Well, thank you so much for sharing this study with us.

Raunak Shrestha: Thank you for the invitation, and thank you for having me.