Read the Full Video Transcript

Andrea Miyahira: Hi everyone. Thanks for joining us today. I'm Andrea Miyahira, here with the Prostate Cancer Foundation. Today, I'm joined by Professor Gerhardt Attard of the University College London Cancer Institute, to discuss his group's recent publication in Nature Communications, "Copy Number Architecture Defines Treatment-Mediated Selection of Lethal Prostate Cancer Clones." Thanks so much for joining me today, Dr. Attard.

Gerhardt Attard: Thank you. Thank you, Andrea, for inviting me.

My name is Gerhardt Attard. I'm a medical oncologist clinician-scientist, at University College, London. On behalf of my co-authors, it's my pleasure to present this recent data we published.


So really, where this started for me, was work we did testing plasma DNA, using a range of next-generation sequencing assays. A lot of work, in fact, funded by the Prostate Cancer Foundation. Where repeatedly, we observed alterations in the androgen receptor associating with shorter outcomes in patients treated with abiraterone or enzalutamide. And in what was quite a provocative result, we compared that to patients who had received first-line docetaxel, so not randomized to two independent cohorts, and observed that there was not that association with shorter survival and docetaxel. It suggested that potentially, AR alterations detected in plasma DNA were driving resistance to abiraterone or enzalutamide. And I use that word cautiously, of course, because these are not randomized cohorts.

But really to square the circle, I wanted to understand this in much greater detail. There's a number of publications where we started this work, including from Steve Bova's group, the PELICAN Autopsy Program, and from Pete Nelson's group, with a much larger number of patients, about 42, if I remember correctly, that were published in Nature Medicine. And those really had suggested this heterogeneity in terms of AR alterations. There's a number of questions foremost in my mind, but clearly, analyzing multiple tumors, harvested post-mortem, could start to answer some of those dilemmas.

So we worked with Shahneen Sandhu, who's based in Peter MacCallum in Australia, and has run the CASCADE program. And then a second program, which we have established here in University College London called PEACE. And from patients with advanced prostate cancer, who die from prostate cancer, we harvested a wide range of tumors that are depicted here. You can appreciate the anatomical distribution across this cohort of 10 patients.

For some patients, we had as many as 30 or 40 metastases sampled. For others, we had fewer. And the first thing we did, as implied by my introductory slide, was to focus on the androgen receptor, and really performed detailed analysis of this. And the bottom line is, the harder you look, the greater the chance of finding an alteration in the AR. And as you'll see in the next couple of slides, we detect AR copy number gain across eight out of 10 patients. And then we look, when we look in greater detail, we find AR genomic alterations in all our patients. So really, this is a ubiquitous genomic alteration.

The second observation is that, there is notable intrapatient diversity, and the skyline plots we included in our paper present the proportion of metastases that have an alteration at a specific location in chromosome X, which is the X-axis. And the dotted line here, refers to the position where the androgen receptor is cited. So you can see, for example, in CA35, the top left box, none of our patients have a copy number gain across most of the chromosome X on the left side, and then all of them have that large amplification. And when you compare to CA79, about half of them have two copies of the androgen receptor, but then, there's focal amplification occurring in all patients. So the gain of large areas of chromosome X is highly variable, involving different regions across different patients, and differentially involving different metastases in the same patient.

We then, using that information, selected specific metastases to subject to high coverage, in this case, 60X whole-genome sequencing. And we annotated the breakpoints within chromosome X that sit around the androgen receptor, and we observe highly diverse breakpoints, as you can see here. In every patient, the breakpoints are different. Which, of course, could have implications for using breakpoint tracking as a biomarker strategy, which really have to be used in individual patient-designed probes, or deep sequencing of this whole region. And others have shown this as well. And the red line refers to the androgen receptor, and the blue vertical line refers to the enhancer. So you see this accumulation or concentration of breakpoints around these two genes, which really are highly diverse. And one of the questions Andrea has asked me is, have we understood what mechanisms contribute to these structural rearrangements? And the answer is no. We've really focused on describing them at this point. But we've made an interesting observation and potentially two on the slide.

So the first is, in one of the 10 patients who's copy-number neutral, we found structural rearrangements in the androgen receptor that results in loss of the ligand-binding domain, and the constitutively, or putatively constitutively active AR transcript that lacks the ligand-binding domain. This was seen in every metastasis, and the plots show the allelic or the AR breakpoint cancer cell fraction, and then one of the prostate cores we sampled, so one of the cores from the prostate tumor. As I said earlier, the harder you look, the more likely you are to find an AR genomic alteration.

And then across all the patients, we observed that there's an association between shorter response to abiraterone or enzalutamide, and fewer unique breakpoints around the AR locus. Now all the patients in this series received AR SIs second-generation hormone treatments for first-line, or second or third-line mCRPC. None of them started them at the start of HSPC. So that's a question to address in the future, how that earlier start of treatment will impact the AR genomic landscape at death.

In two patients, out of our cohort of 10, we observed AR somatic point mutations. And interestingly, in this patient, CA36, we observed a mutation that results in the T878A protein change in every metastasis. But when you dig in deeper, and I draw your attention to the metastasis labeled 11, there's two mutations, as you can see, one resulting in a T878A change, the other in the D891 change. And when we look at all the reads, because these two mutations are close to each other, we can look at base changes on the same read, we identify reads that have only a change involving the T878A alteration, reads that have only a D891 alteration, and reads having both. It's possible that there is loss of either of those alterations, specifically the T878A alteration, although, we think that's highly unlikely. And the most likely explanation here, is that these two mutations developed independently in this metastasis, and more broadly in the manuscript, to show evidence of independent evolution of AR somatic point mutations.

Which then leads to the question, why do we only see, and repeatedly across cohorts, mutations in about 15 to 20% of patients, when we're saying that this occurs independently at multiple metastases? Which may suggest that a subset of prostate cancers are predestined to follow an evolutionary course that converges on this AR mutant resistant genotype.

We then looked much more carefully at this intrapatient AR genomic heterogeneity. And we developed and implemented an approach we called SCRATCH, which used a copy number of transition points of copy number change, that we posited contained much more evolutionarily useful information than copy number alterations. And we used this to measure the relationship of metastases to each other, and then we plotted a hierarchical clustering of metastases relationships.

And you can see that plotted here. And what was particularly intriguing, is that the AR alterations, or the chromosome X structural alterations, are topologically congruent with the autosome-assigned relationships. Which would suggest that, alterations in chromosome X occur in clones that have already been established, and predefined by their autosome copy number change. And of course, the cartoon in the middle thus, is not supported by temporally separated samples, but we start to hypothesize metastatic trajectories. And interestingly, in this patient, we identify a tumor in the metastasis that we sampled twice, or we took two cores from, and those separate cores were assigned to different clusters of metastases, and indeed, had different AR alterations. And you can see that denoted by the green and red circles in the bladder.

And we start to plot the charts, or at least show metastases that are more closely related to each other, and fall within the same clusters, as denoted by the colors seen there. And of course, we're intrigued, as I'm sure many will be, this coexistence of bonafide AR genomically wild-type normal metastases, or samples from metastases coexisting in this patient, with clearly highly AR gained mets. And intriguingly, when we measured, when we performed the expression studies, both looking at single genes and AR activity from RNA-seq data, we see that as you'd expect, and the gain metastases AR expression is much higher. But AR signaling is pretty much similar across all metastases. Suggesting mechanisms that do not involve the androgen receptor to retain that increased level of AR signaling in those mets. And that is a line of inquiry, one of many that arise from the study that my lab is now pursuing.

Andrea Miyahira: Thanks so much, Gert, for sharing this with us. So did you see any patients in your study that had AR low or negative disease, and did those patients have AR alterations?

Gerhardt Attard: So yes, AR low or negative disease, I think, you'd also be thinking here about neuroendocrine prostate cancer, what we're commonly referring to neuroendocrine prostate cancer. Which is this phenotype where PSA is not rising in a patient with clearly progressing metastases. So none of our 10 patients met that phenotype. They all had a rising PSA at progression on abiraterone and enzalutamide. So that's clearly, it would be a really interesting question to pursue in future studies.

And clearly, what was intriguing in this cohort, where we have a rising PSA and we've extensively analyzed their metastases, we're finding clusters that have AR alterations and others that don't. But in this situation, the AR altered mets had retained AR signaling.

Andrea Miyahira: Okay, thank you. And were you able to compare tumor evolutionary patterns in patients that were versus weren't metastatic at the time of initiation of anti-AR therapy? And if so, were these different?

Gerhardt Attard: I love that question. So we are asking that now with an expanded cohort of patients. So this group of patients died three, four years ago, so were treated through the mid, around 2015 to 2019. So none of them received ARSI for non-metastatic disease or at presentation. But we now have cohorts of patients coming through that did. I think that's a really interesting question, Andrea, how will we change the genomic landscape at death by earlier use of ARSI? And of course, that will then inform us on the implications for earlier use of these treatments, identify better ways to use the treatments, and potentially, the mechanisms of resistance may differ. Although, my gut feeling is they're not. We're seeing the same resistance emerging regardless of when we start ARSIs.

Andrea Miyahira: And what are your next steps for this research program?

Gerhardt Attard: The methodology we developed, we tested in archival samples, and a couple of plasma samples, and really, we started to appreciate the high value of having temporal samples in addition to multi-region sampling. So clearly, postmortem sampling is giving us this extensive diversity of samples that allow spatial resolution. But that needs to be linked to temporally separated samples, to give us a better understanding of the evolutionary trajectory. So within both the PEACE Program, and with Shahneen and the CASCADE Program, we've been collecting both plasma samples and tissue biopsies. And we have a large expanded cohort, now hitting 20 patients, who we have sampled repeatedly from diagnosis, and over the course of their treatment, and that analysis is ongoing in our groups. And I hope in the next year, we'll have some new interesting data that starts to answer at least some of the questions you asked me today.

Andrea Miyahira: Thank you again so much for coming on and sharing this with us today.

Gerhardt Attard: Thank you.