Read the Full Video Transcript

Oliver Sartor: Hi, I'm Dr. Oliver Sartor here with UroToday, and we have a very interesting discussion coming up with Jake Berchuck. Jake is a GU medical oncologist at the Winship Cancer Institute at Emory. Welcome, Jake.

Jake Berchuck: Thank you. Excited to be here. So today I'll be discussing an abstract that I presented at ESMO 2024 last week titled "Determination of Tumor PSMA Expression in Prostate Cancer from Blood Using a Novel Epigenomic Liquid Biopsy Platform." This is a collaboration between the Winship Cancer Institute, Dana-Farber Cancer Institute, Brigham and Women's Hospital, and Precede Biosciences. So the background for this study is with several therapeutic strategies for metastatic prostate cancer in development targeting an array of cell surface proteins beyond PSMA.

There's an emerging unmet need for real-time quantification of tumor drug target expression. And so the question that we sought to address in this study is: can we use a novel epigenomic profile platform to profile plasma cell-free DNA to achieve an accurate, non-invasive readout of tumor PSMA expression in men with metastatic castration-resistant prostate cancer? Here's an overview of the study. We started with a cohort of 29 men with metastatic castration-resistant prostate cancer, all of whom underwent a PSMA PET scan with the intent of going on to receive Pluvicto as standard-of-care therapy.

You can see some of the patient characteristics there. And all of these patients had plasma collected around the time of PSMA PET scan. The average time was about 60 days between plasma and PET scan. And so on that plasma we collaborated with Precede Biosciences, who ran their epigenomic cell-free DNA assay. And at a high level, what this assay does is it generates epigenome-wide or genome-wide insights into tumor gene regulation by looking at marks of active gene enhancers, gene promoters, and DNA methylation.

And so what you can see in the bottom right is that you can get a readout at a gene level on the regulatory features that are regulating the expression of that gene, with the intent of being able to then infer tumor expression of genes of interest. So the first question that we asked is: how well does this assay pick up relevant signal in the blood from men with metastatic prostate cancer?

So we first compared the blood from these 29 patients to some healthy male controls, and what the graph on the left shows is which genes had signal enriched in men with metastatic prostate cancer compared to the healthy controls. What we were really struck by is that, in this unsupervised analysis, the genes that had the strongest signal in blood from men with metastatic prostate cancer relative to controls were genes that we expect to be upregulated in prostate cancer: KLK2, KLK3, HOXB13, SPDEF.

So a really nice test too, a really nice sort of first proof of concept we were seeing, signal we expected to see. We then honed in on FOLH1, which is the gene that encodes PSMA, and looked at the signal for these different epigenetic features in blood in the men with metastatic prostate cancer and controls. And what you can see by eye in these tracks over here is that at the promoter of FOLH1, you can clearly see more signal that marks active gene enhancers and promoters in the blood from men with prostate cancer compared to the controls.

We then leveraged some AI tools to quantify PSMA PET SUV mean, which is the feature that we think most strongly correlates with Pluvicto response. And we asked the question across the entire genome in these men with metastatic prostate cancer: which genes had the strongest signal that best correlated with PSMA PET SUV mean?

And what we found is across the entire genome, in an unsupervised fashion, FOLH1 was actually the gene whose signal most strongly correlated with PSMA PET SUV mean. And you can again see this by eye. What we did is we sort of split the prostate cancer patients by the ones with highest PSMA PET SUV mean and lowest, based on above or below the mean.

And what you can see there is clearly—I'll highlight the enhancer signal—but at the FOLH1 gene, at the promoter, you can clearly see more signal in the blood from men with prostate cancer who had higher PSMA PET SUV mean than the patients who had lower PSMA PET SUV mean, sort of suggesting that we can, that the signal in the blood at the PSMA locus really does seem to correlate with tumor PSMA expression on PET scan.

So the last thing that we did that I'll highlight is we used some machine learning techniques to try to build a model to predict PSMA PET SUV mean based on the epigenomic signal in blood from these patients. We split the cohort into a training cohort and an independent held-out validation cohort. We applied machine learning tools to the training cohort and performed internal cross-validation, which achieved a correlation of 0.71 that was highly significant. And then we applied a trained model to an independent validation cohort of held-out samples.

And what we saw is that with a correlation of 0.8, which is fairly good and achieved statistically significant results in a small cohort, that we saw a fairly accurate correlation between the predicted PSMA SUV mean in the blood with what we actually observed on the PSMA PET scan. And so our conclusion from this study is that with several drugs in development targeting a wide array of tumor cell surface proteins in prostate cancer, that there really seems to be pretty exciting potential for these epigenomic liquid biopsy tools to provide a real-time and non-invasive readout of tumor drug target expression.

Obviously we focused on PSMA here, but I think sort of the potential to apply this for other drug targets is pretty clear. And then moving forward, what we're hoping to do is obviously validate these findings in bigger cohorts and independent validation cohorts, but also ask the questions of how well does cell-free DNA-inferred PSMA expression predict response to Pluvicto.

I think as additional drugs move towards regulatory approval targeting other targets, there's going to be an increased need to be able to say for this individual on this day, which cell surface targeting drug is most likely to be beneficial. And so we're hoping to sort of move towards being able to answer those questions. With that, I'll pause and be happy to take some questions.

Oliver Sartor: Yeah. Jake, that's very impressive. And the idea of a circulating marker that would correlate with a cell surface marker is a very powerful concept. Now, what you've shown is a single point in time; it would be even more powerful if you're able to show changes as a function of time and treatment. And let me get your thoughts on that concept before we go to the next question.

Jake Berchuck: I love that concept, and that's exactly another direction that we're going with this. So work that's ongoing with this cohort, building on the results we presented at ESMO, is we have a larger cohort of patients where we have the plasma at the time of PSMA PET scan. So we're planning to sort of further validate the findings that we reported. But another piece of the cohort is that we have plasma in patients at baseline right before initiation of Pluvicto and at the time of clinical progression. We have that for a large number of patients and are planning to ask that exact question.

Can we understand mechanisms of Pluvicto resistance, acquired resistance that really we haven't been able to study with traditional tumor biopsies given the challenges of obtaining tissue? And so we have a very focused hypothesis that potentially PSMA levels may be down-regulated at the time of progression on Pluvicto as a mechanism of resistance and plan to ask that question specifically. The other question that I think or what this technology enables is that, as I mentioned at the beginning, we get genome-wide data on these epigenomic features of gene regulation.

So we can ask questions like: what are other bypass pathways that may be up-regulated as tumors develop resistance to Pluvicto, to try to think about either next lines of therapy, understanding sort of resistance to sort of move the field forward in that sense. And so I think, as you alluded to, we often get molecular data at a single time point, whether it's tumor profiling or... And so I think what this enables is dynamic tracking of these targets to really understand in real time the patient's tumor molecular features.

Oliver Sartor: Yeah. Some very provocative thoughts in there. So let me sort of challenge one of the presumptions. And I initially would've also said that the down-regulation of the PSMA target would've been an important mechanism of resistance. Because if you don't express as much of the target, then you're not going to hit the target with Pluvicto. On the other hand, what we've seen is quite interesting and points to bypass pathways, which you also alluded to. A lot of the patients that are resistant to Pluvicto continue to express the PSMA.

But for reasons that are not clear—either they're radio-resistant or they're just bypass pathways—the actual Pluvicto-resistant cell has lots of PSMA, and that our ability to predict resistance by looking at down-regulation didn't work out as well on the imaging as we would've anticipated. So that's just a little bit of a comment. But let's explore the second concept, which I really do like.

You have a genome-wide assessment, and I think you can learn an enormous amount by looking at these post-Pluvicto patients, both from the responsive patients as well as the resistant patients. So let me go down that path and hear your thoughts a little bit more about a genome-wide assessment for patients either responding to resistance to Pluvicto.

Jake Berchuck: Yeah. No, I think it is a great question, and I think again highlights the, I'd say, potential of what this type of platform can help us learn about prostate cancer biology, that with our current tools we haven't been able to study as well as we'd like. So I agree. I think we have a focused hypothesis around PSMA, but certainly there are studies, as you alluded to, looking at PSMA expression post-progression on Pluvicto, and we have some information there.

What I think about PSMA levels, what I think is exciting, as you highlighted, is the opportunity for discovery of other pathways. And so I think we have really well, for example, sort of characterized association between PSMA-low disease or PSMA-negative disease being sort of a poor prognostic group. But what we don't have as much information on is what's the biology that's driving that?

And so I think what we hope to understand is Pluvicto resistance, PSMA-negative disease, really is doing a deeper dive into the molecular correlates of these radiographic subtypes that we've seen on imaging. And so I think that we have targeted hypotheses that at the time of progression on Pluvicto, we'll see enrichment of pathways in EMT, in cell cycle upregulation, pathways that we know to be sort of associated with poor prognosis and resistance in prostate cancer. So we'll go in with some targeted hypotheses. But as you alluded to, I think there's a neat opportunity for discovery to see what we don't know about these tumors as they develop resistance to Pluvicto.

Oliver Sartor: One of the things, and of course if Himisha Beltran were here, she'd be referring to the neuroendocrine subtype and the variety of neuroendocrine markers, maybe chromogranin.

And so you would have a targeted platform to be able to look at potential for neuroendocrine development. And I think that could be very interesting. There's a lot of data about the plasticity of these cells, and I love the fact that you might be able to follow them in a serial fashion post-treatment with resistance as well as the responding patients. I think that's very cool.

Jake Berchuck: Yeah, I'll highlight that we and others—Dr. Beltran, Gavin Ha from University of Washington, myself with Dr. Matt Freedman at Dana-Farber—we, several labs have independently shown that by tracking epigenomic features in the blood, you can, with quite good accuracy, detect the emergence of neuroendocrine prostate cancer. And so I think that's another thing that we should be looking at. You have to imagine as these patients become more resistant, you're selecting for the more aggressive clones that we'll see emergence of that in probably a subset of these patients progressing on Pluvicto. And that's certainly something that we'd be excited to look at.

Oliver Sartor: I'm really enjoying this conversation, so I'm going to take it in an unexpected direction. Patients with liver metastasis represent a particularly challenging subset. Even when they do respond, they often respond only for a short period of time. Do you have any subset assessments in the liver metastatic patient? And if not, I might say, "Gosh, I wish you'd look at it because I don't know what drives those patients." I'd love to understand more about them.

Jake Berchuck: Yeah. I think what you highlight is two things in my mind: a really important unmet clinical need to sort of better understand those poor prognostic, the molecular correlates of those known poor clinical prognostic features like liver mets, like osteolytic bone mets who aren't frankly neuroendocrine prostate cancer, but we know have poor prognosis. And the second thing you highlight, I think, is the potential of liquid biopsy platforms to really sort of provide new insights because tissue's hard to get.

Traditionally, our liquid biopsy platforms have really focused on genomic profiling. And I think we've done a pretty good job understanding that the features are loss of P53, RB1, sort of the tumor suppressor genes. But I think what these platforms afford is sort of asking in certain clinically relevant contexts, what are the epigenetic features that are driving these prognostic associations? And I think this is something that probably is going to be several years off, but is certainly in clinical development.

But as we develop better tools to target epigenomic drivers of cancer progression, like transcription factors—AR, HOXB13, FOXA1—these known transcription factors in prostate cancer, I think being able to understand these epigenetic subtypes and drivers is going to become increasingly important. So I think it is going to be an exciting next few years as we think about what are the unmet clinical needs that are understanding patients with liver mets that we can address with these platforms to hopefully better treat these patients down the road.

Oliver Sartor: Yeah. I'll make one of the other comments that sort of comes to my mind is we're looking at things like EZH2-targeted drugs. This is a major regulator in the environment that you're alluding to. Jake, I have a feeling we could talk for a long, long time, but I wonder if you might leave our listeners with any final thoughts or final concepts you'd like our audience to be able to be tuned in when they're thinking about this field.

Jake Berchuck: Yeah, absolutely. Again, thanks for the opportunity. Really enjoyed chatting today. I think the message that I leave is sort of a message about the potential of platforms like what Precede Biosciences and others are doing around epigenomic profiling of cell-free DNA. As more and more drugs come to the clinic targeting cell surface proteins.

And as you know, there's several in development targeting B7-H3, STEAP1, TROP2. And as these drugs come to the clinic, we're going to need biomarkers to select which patients are likely to benefit and which ones are likely to not benefit so we can prioritize other therapies. And as we've seen with molecular imaging, PSMA PET is a good biomarker for identifying which patients benefit from Pluvicto, but as we have drugs targeting four or five, six different cell surface proteins, we're going to need a tool to assess several different targets.

And I think with these epigenomic liquid biopsy tools, being able to provide insights into genome-wide regulation of genes across relevance in prostate cancer, I'm hopeful that we can move towards using these tools to really realize the potential of precision medicine in metastatic prostate cancer to select which of these drugs are going to be beneficial for which patients.

As you alluded to, I think there's a lot of exciting opportunities to better understand metastatic prostate cancer across several different unmet needs like liver mets, understanding how we can, yeah, better characterize our tumors, better treat our patients to improve outcomes. So thanks again for the opportunity and would be excited to follow up with anyone interested and with any questions offline.

Oliver Sartor: Great. Thank you, Jake. That was very enjoyable for me personally. I think our audience is really going to enjoy your insights. Jake Berchuck from Emory. Thanks for being with UroToday.

Jake Berchuck: Thanks, Oliver.