2016 IMPaCT Research Vignette

Title: Divergent Clonal Evolution of Castration Resistant Neuroendocrine Prostate Cancer

Investigator: Himisha Beltran, MD; Weill Cornell Medicine

My research has focused on prostate cancer resistance and trying to understand why some patients develop aggressive disease and tend not to respond very well to the hormonal therapies that we use to treat advanced prostate cancer. And one phenotype that I’ve been particularly interested in is the small cell neuroendocrine phenotype that can occur in later stages of the disease; it tends to be very resistant to hormonal-based interventions, and I’ve been using genomic, epigenomic integrative analysis to try to better understand why, how and why these patients, some—but not all patients develop this resistance phenotype and figure out how best to diagnose it and how best to treat it.

In order to study these patients that are not responsive, we’ve been performing biopsies in patients and studying the clinical features of the patient and integrating that with biopsy features and molecular features. And what we found is that a sub-set of these patients develop features, tumor features that look like small cell or neuroendocrine carcinoma as opposed to your more traditional prostate adenocarcinoma. And through metastatic biopsies we were able to compare the extremes, these castration resistant adenocarcinomas, with these neuroendocrine prostate cancers that are clinically more aggressive.

So this work, this genomic analysis has been done in collaboration with Mark Rubin, Francesca Demichelis, and Levi Garraway. We used these data to start to plot out the DNA copy number landscape. So the X axis is chromosomes, the area above the horizontal line are genomic amplifications, and below are deletions. And what we’ve found here is just overlay the two groups, one on top of each other with the areas of gray indicating areas of overlap. And you can see that the overall spectrum and pattern of copy number alterations were very similar, almost completely overlapping which was surprising because they’re so different clinically and pathologically.

But overall, what we’ve found is that these neuroendocrine tumors seem to arise clonally from a castration resistant adenocarcinoma precursor. And I think that’s an important observation. It suggests that these evolve during the course of treatment as a mechanism of resistance rather than overgrowth of some completely different type of cancer. And I think it has implications for potentially early detection and—and early co-targeting.

As opposed to the largely similar genomic data, the epigenomic landscape was really quite different. We looked at DNA methylation across the genome and you can see that our neuroendocrine tumors are clustered all together on the left and our adenocarcinomas on the right, almost perfectly, except for these three cases, these are adenocarcinomas that grouped with our neuroendocrine tumors. Interestingly these came from patients that had clinical features suggestive of resistant disease, less hormonally driven disease progression with very low PSA levels, suggesting that there may be clues in the epigenome of advanced prostate cancer that might help us understand treatment resistance.

We started to integrate whole exome RNA sequencing data and DNA methylation data to develop the molecular classifier, which we call an integrated NEPC score. And here we’re plotting androgen receptor signaling against the integrated neuroendocrine score. So, the darker dots represent the neuroendocrine tumors. And what you can see is that the neuroendocrine score captures almost all of our neuroendocrine tumors based on molecular features. And this suggested that there was a high precision in recall of the classifier in identifying neuroendocrine tumors. And again I think this is important because it helps us identify what features are unique to neuroendocrine prostate cancer with the goal being for early detection of patients transitioning towards this phenotype. If we could identify the molecular alterations we may be able to detect them non-invasively using a blood test in the future circulating tumor DNA and that’s what we’re working on now. And using all this data, we have been analyzing the data to try and identify novel therapeutic targets which is a clinical unmet need.

I think what’s amazing about this study is that it integrated so many levels to put it all together, to understand the clinical pathology, molecular features, I don’t think we knew what we were going to find and I think we’ve learned a lot, but there’s still a lot to do. But I think that all of it has been very exciting for me. I mean I think that, you know, sometimes when you kind of delve into areas that you don’t know what you’re going to expect you end up finding the most interesting things.