Read the Full Video Transcript

Andrea Miyahira: Hi everyone, I'm Andrea Miyahira at the Prostate Cancer Foundation. Presenting today is Dr. Scott Dehm, a professor at the University of Minnesota on his recent paper, "Co-evolution of AR gene copy number and structural complexity in endocrine therapy resistant prostate cancer," published in NAR Cancer. Dr. Dehm, thanks so much for sharing this work with us today.

Scott Dehm: My pleasure. Thank you very much for the opportunity to summarize our work today. This is a collaborative study between my group, a computational biology group at Princeton University led by Ben Raphael, one autopsy program at the University of Washington led by Colm Morrissey and Eva Corey, and also a surgery cohort at the University of Umeå in Sweden, led by Pernilla Wikstrom, and our longtime collaborator at UCSF, Felix Feng. What we knew going into this study was that the gene encoding the androgen receptor, the AR gene, is the most frequently altered gene at the castration-resistant prostate cancer stage.

We know that over 85% of castration-resistant prostate cancers harbor at least one alteration in the AR gene. Specifically, those alterations include copy number gains of the AR gene body, copy number gains of an enhancer located upstream of the AR gene body, mutations or single nucleotide variants that affect the amino acid sequence of the AR protein, as well as structural variants or AR gene rearrangements that affect the structure of the AR gene and affect splicing of AR, and some of the isoforms of AR that can be expressed.

We had identified previously that approximately one quarter of castration-resistant prostate cancers had the property of simultaneously harboring rearrangements in the AR gene, denoted by these black boxes, as well as amplification of the AR gene and amplification of the AR upstream enhancer. Many of the studies that we had been performing previously involved bulk DNA sequencing approaches, and that doesn't really allow us to understand the heterogeneity of the tumor cells. The question that was driving this study was whether these co-occurring amplification and rearrangement events affecting AR were occurring in the same cells of a castration-resistant prostate cancer tumor, or was this reflecting tumor heterogeneity, and these were occurring in different cells? Knowing which was the case, of course, is important for understanding clinical heterogeneity and the evolution of prostate cancer. In our study, we used a third generation DNA sequencing technology called linked-read DNA sequencing.


This gives us much longer-range information about alterations in the DNA than we would otherwise get from using traditional short-read DNA sequencing. We applied this linked-read DNA sequencing technique to 23 clinical specimens from the University of Umeå as well as the University of Washington, and also paired patient-derived xenografts developed at the University of Washington. Using this technique, we confirmed our previous observation that a subset of castration-resistant prostate cancers had co-accumulated very high copy number of the AR gene as well as a considerable number or burden of AR gene rearrangements. This was also occurring in the patient-derived xenografts. I'll get to that in a minute.


The power of this linked-read DNA sequencing approach is, it allowed us to identify that in four of these tumors that had co-accumulated very high AR copy number, as well as a very high number of AR gene rearrangements, many of those AR gene rearrangements that we had identified were co-occurring on the same DNA molecule. This indicated they were present in the same cells, answering our original question: are they occurring in the same cell? Fortunately, these patient-derived xenografts that we were studying had captured that biology, providing us a platform to investigate this in more detail. We had these patient-derived xenografts that we could grow in intact mice, which have normal physiological levels of circulating testosterone, or in mice that had been castrated to model the androgen deprivation hormone-based therapy that prostate cancer patients receive.


In two of these models, LuCap 35 and LuCap 105, when the tumors were grown under the selective pressure of castration, they had accumulated both the increased levels of AR copy number as well as the increased numbers of AR rearrangements. The linked-read DNA sequencing told us that many of these rearrangements were co-occurring on the same DNA molecule. We prioritized investigation of these patient-derived xenografts to understand this biology in more detail. We also performed an additional orthogonal method of investigating genome structure using another third-generation DNA interrogation technique called optical genome mapping. We used the LuCap 105 patient-derived xenografts growing in the intact mice, and the LuCap 105CR xenografts grown in the castrated mice, and used optical genome mapping, which is a genome-wide approach. These circled plots tell us what the structure and copy number alterations are of every single chromosome in these tumor cells.


What we had identified using this approach was a high burden of rearrangement between the AR genome X chromosome as well as an adjacent lesion on chromosome 16, and this seemed to be occurring in both tumors. Based on these findings, as well as other findings reported in the paper, we came up with the following hypothesis that perhaps the mechanism of AR amplification in the LuCap 105 patient-derived xenograft tumor was that the AR was amplified on extrachromosomal circular DNA, which contained both AR and a segment of chromosome 16, and there was a relatively low burden of structural rearrangement on these ecDNA molecules containing AR. But then under the selective pressure of AR-targeted therapy, the castration-resistant counterpart had accumulated massive numbers of these extrachromosomal circular DNA molecules. These individual extrachromosomal circular DNA molecules had also accumulated a high degree of structural variation and structural rearrangements.


This model would be compatible with all the observations that we had made previously as well as in this paper. We turned to fairly traditional cytogenetics techniques to probe this hypothesis, so we used fluorescence, in-situ hybridization with specimens prepared from LuCap 105 and LuCap 105CR tumors. The pattern of staining was completely consistent with this model that we had. The red indicates the AR gene, so most male cells will have just one copy of the AR gene. You can see, this entire blue nucleus is absolutely full of AR gene copies, whereas the counterpart of the tumor grown in the intact mice has far fewer copies. It's consistent with that accumulation of signal.


And then we used a CO-FISH approach, where we stained for AR in red, and this region of chromosome 16 fused to the AR in green, and so the red and the green signal overlapping would make a yellow signal. We saw a significant number of yellow signals in the LuCap 105CR tumor, but fewer in the parental tumor. Again, this is consistent with these extrachromosomal circular DNA molecules, which preexisted, and then accumulated into high numbers in these advanced castration-resistant versions of the patient-derived xenografts. I would say, the most important part of this paper is that we also went back to those original clinical specimens, where we had identified this co-accumulation of AR copy number, and high burden of AR gene rearrangements, and multiple rearrangements being harbored on the same DNA molecule, and performed FISH for AR.


Again, we saw many copies of AR distributed throughout the nucleus. In some cases, these copies of AR were independent of the chromosome X centromere. In other cases, it seemed that the chromosome X centromere was also being massively amplified as well. This is a pattern that's observed now in clinical castration-resistant prostate cancer specimens as well as patient-derived xenografts that model the stage of the disease.


This is the graphical abstract that we prepared for this paper to summarize the overall results of our study. What we were able to demonstrate is that in castration-resistant prostate cancer cells, one of the ways that AR can be amplified is on extrachromosomal DNAs. These are independent of the chromosomes. This, of course, increases the AR copy number in those cells to a very high number. But what we learned, which is really unique from this study, is that those individual AR copies are structurally unstable, and that is an engine for increasing the burden of AR gene rearrangements in that cancer cell. And then I didn't show it in this presentation, but it's in the paper. We showed that this can allow the prostate cancer cell to synthesize a more diverse set of AR proteins that may contribute to the resistance phenotype in those tumors.

Andrea Miyahira: Okay. Thank you for sharing this really interesting study with us. What do you think the significance is of AR being amplified on the extrachromosomal DNA as opposed to being amplified on the X chromosome?

Scott Dehm: Well, I think when oncogenes are amplified on a chromosome, there are certain constraints to the degree to which that amplification can occur, and the number of copies that can accumulate, whereas with extrachromosomal DNA, which we've known about for decades, previously, it was called double minute chromosomes. But more recently, we've learned that they're circular, and they can harbor oncogenes. We think that this is a way that allows that particular oncogene to accumulate to a much higher copy number than could otherwise be achieved if that oncogene was still contained within the genome within a chromosome.

Not only that, but when a cell divides, or when a cell has this extrachromosomal DNA that harbors an oncogene, and divides, those two daughter cells do not need to follow the rules of Mendelian genetics. One cell can accumulate the lion's share of those extrachromosomal DNA molecules. It's a way to much more rapidly amplify the number of copies of an oncogene in the cell. If that increased inheritance of those copies of extrachromosomal DNA favors that oncogene, giving that cell a selective advantage relative to its neighbors, that cell is then going to preferentially divide, and populate the tumor. We believe it's a mechanism that enables very rapid evolution of that particular cancer cell, and the tumor, more generally.

Andrea Miyahira: What mechanisms do you think drive the formation of AR extrachromosomal DNA? Are there specific genomic sequences or interacting machinery that may be involved, and is chromosome 16 typically involved, or was it just that one sample?

Scott Dehm: Yeah. Great questions. We don't know the precise answer for AR, but we can maybe draw from knowledge of extrachromosomal DNA studied in other cancers, and other types of oncogenes that can be amplified via this mechanism. For instance, the epidermal growth factor receptor, EGFR, can be captured on extrachromosomal DNA in glioblastomas, or MEK can be captured on extrachromosomal DNA in a variety of different types of cancers. In fact, the PC3 cell line, which is of prostate cancer origin, contains MEK amplification on ecDNA. Mechanistically, we know that DNA damage can induce accumulation of extrachromosomal DNA. A therapeutic stress can drive, maybe not the formation of, but at least the enrichment of extrachromosomal DNA. Smaller molecules of extrachromosomal circular DNA are present in all eukaryotic cells, so this is not a phenomenon restricted to cancer, but it seems like the large circular fragments of DNA, multiple megabases in size, that harbor oncogenes, that is a cancer-specific phenomenon.

It is associated with very poor prognosis across many different types of cancer. When we look at the break points associated with the fusion of the two ends of the molecule to create that circle, the signature is usually associated with non-homologous end joining, and so defective homologous recombination DNA repair, which is the alternative DNA repair mechanism to NHEJ, that may also be playing a role as well. There have been many studies into the mechanisms underlying the formation of extrachromosomal circular DNA. There seems to be many pathways to get there, but DNA damage and DNA repair are two mechanisms at the heart of this.

And then regarding your second question about whether chromosome 16 is always involved, we do think that that is a specific observation in that specific tumor model, LuCap 105, and LuCap 105CR. There's really no reason to expect that that would be a recurrent fusion that's occurring. But that's another very interesting property of extrachromosomal circular DNA, as well as extrachromosomal DNA harboring oncogenes, is that they can fuse together in a cell. You can have two independent circular DNA molecules in a cell that can fuse, and then bring two oncogenes together, or an oncogene and enhancer together. These are consistently and constantly evolving within the cancer cells.

Andrea Miyahira: You mentioned MEK. Are there other amplified oncogenes in prostate cancer that can form via extrachromosomal DNA?

Scott Dehm: I do think, across multiple cancers, there have been a lot of profiling efforts to discover and characterize extrachromosomal circular DNA. There was a pan-cancer study that was performed, and prostate cancer was among the highest in terms of frequency of cancers of that type, harboring extrachromosomal DNA. I forget what the exact figure is, but it was well over 10% of prostate cancers harboring extrachromosomal DNA. MEK, of course, is one example of an oncogene that could be amplified via this mechanism. We know AR is the most frequently amplified oncogene, so it made sense to look at that one. But we can look to other oncogenes that are known to be amplified in prostate cancer, and castration-resistant prostate cancer, and then perhaps prioritize those for investigation of whether extrachromosomal circular DNA may be the underlying mechanism. We have AR, we have MEK. To a lesser extent, we have Cyclin D1, and then we also have FOXA1, as the currently amplified genes in castration-resistant prostate cancer. Those would be the ones that I would prioritize for studying whether extrachromosomal circular DNA is an underlying force.

Andrea Miyahira: Thank you. And then what are your next steps, and do you have any translational plans for these findings?


Scott Dehm: Well, I have thought a lot about circulating tumor cell technology, because with circulating tumor cells, it is fairly straightforward to use FISH as a cytogenetics technique to characterize that circulating tumor cell. I think one immediate near-term priority would be to ask whether AR extrachromosomal circular DNA FISH staining pattern in circulating tumor cells from castration-resistant prostate cancer patients is associated with adverse outcomes, or could serve as a predictive biomarker for those patients who are on AR-targeted therapies relative to those patients who do not have that staining pattern in their circulating tumor cells. Because I would envision that those prostate cancers that have accumulated AR amplification and structural rearrangement this way are going to be highly resistant to the spectrum of AR-targeted therapies that we have available. I do think this can be leveraged quickly and easily as a biomarker strategy, but of course, more work needs to be done there.

And then perhaps the nature of AR amplification on extrachromosomal circular DNA may make those copies of AR more susceptible to, say, DNA damaging agents, or something like that. That's also a therapeutic angle that we would like to investigate. And then finally, there have been studies showing that extrachromosomal circular DNA molecules can associate with each other in the nucleus in types of transcriptional hubs, where they can feed off of each other in terms of the enhancer elements that they harbor, and the transcription factors that they recruit, and the epigenetic regulators that they recruit. Epigenetic targeting drugs like the bromodomain inhibitor JQ1 have been shown to disrupt those ecDNA hubs, and perturb some of their activities. That could also be a therapeutic angle, and perhaps also a mechanistic explanation for why certain prostate cancer models have increased sensitivity to drugs like JQ1.

Andrea Miyahira: Well, thank you again for coming on, and sharing this really interesting study with us today.

Scott Dehm: My pleasure. Thank you.