Read the Full Video Transcript

Andrea Miyahira: Hi, everyone. I'm Andrea Miyahira here at the Prostate Cancer Foundation. Today I'm pleased to welcome Dr. Marina Sharifi of University of Wisconsin-Madison. We will discuss her paper, “High-Purity CTC RNA Sequencing Identifies Prostate Cancer Lineage Phenotypes Prognostic for Clinical Outcomes,” published in Cancer Discovery. Dr. Sharifi, thanks for joining us.

Marina Sharifi: Thanks so much for having me Andrea. It's really exciting to be here to share these results with you. I'm excited to share the results of our manuscript with UroToday. This was supported by the Prostate Cancer Foundation, the Department of Defense, and really excited to be able to share these with you today. As you may know, the androgen receptor pathway inhibitors, or ARPIs, are a highly effective therapy that has improved outcomes for patients with metastatic prostate cancer. But acquired ARPI resistance remains essentially universal for our patients.

Two major mechanisms of ARPI resistance have been described. The first is reactivation of AR signaling through transcriptional or genomic changes for an AR-addicted lineage state. And the second, the tumor can transition to an AR-indifferent lineage state, which in some cases can progress to a neuroendocrine phenotype. That is the most aggressive and difficult-to-treat form of the disease.

As you can imagine, effective treatment approaches for the AR-addicted versus the AR-indifferent lineage states are quite different. But clinically, we still have very limited tools to distinguish between them, primarily because of the challenges of obtaining repeat tissue biopsies over the course of a patient's disease to identify these lineage transitions. Liquid biopsies, which leverage cancer cells shed into the blood, represent a unique opportunity to detect lineage state transitions through longitudinal sampling of patient tumor cells.

But even after CTC enrichment with most current technologies, the proportion of tumor to immune cells in these samples is typically very low, because of which it was not previously possible to use this sample type to perform the comprehensive transcriptional profiling that is needed to identify prostate cancer lineage states. We recently developed a new high-stringency CTC isolation approach that is able to remove the majority of the immune cells in these samples, leaving us with tumor content similar to tissue biopsies that is suitable for comprehensive transcriptional profiling via RNA sequencing.

So, as you can see on the left here, while CTC isolated with the standard approach have relatively low ESTIMATE tumor purity scores compared to tissue biopsy samples here in gray, with our new high-stringency approach, we're getting tumor purity or tumor content scores that are very similar to tissue biopsies, and in some cases higher than tissue biopsies—for example, bone tissue biopsies.

Importantly, this approach, which requires just one tablespoon of blood from a patient, is successful in recovering high-purity CTC RNA for sequencing in about 70% of metastatic prostate cancer samples. So having developed this approach, we then asked whether we could use CTC RNA sequencing to detect prostate cancer lineage states and transitions, which would contribute to treatment resistance. In order to do this, we assembled a multi-institutional cohort of 117 patients with histologically confirmed metastatic prostate cancer and subjected them to CTC RNA sequencing, tumor content assessment, and then from this cohort, we were able to assemble 146 samples from 70 patients with high enough tumor content that we could perform single-sample comprehensive pathway analysis.

And so what we were able to do is to perform pathway analysis for a number of different signaling pathways that are important for prostate cancer lineage states and tumor growth, and then cluster the samples by these pathways, which identified four different transcriptional phenotypes in these CTC samples.

As you can see here, clusters 1 and 2 have relatively high luminal and androgen receptor signaling scores and low neuroendocrine signaling scores, consistent with an adenocarcinoma lineage state. But we actually see two different phenotypes within these adenocarcinoma samples. In this cluster 1 that has relatively low proliferation and mixed signature scores, and a second cluster with relatively high proliferation and mixed signature scores.

And so we named these luminal A-like and luminal B-like for this low-proliferation adenocarcinoma state versus this high-proliferation adenocarcinoma state. Our third cluster had low luminal and AR signaling scores, low neuroendocrine scores, and low proliferation scores. And we did have some samples that had some higher signatures for EMT or stem cell-like phenotypes, but we named this the low-proliferation cluster.

And finally, we have a small cluster here on the end, which had low luminal AR and luminal signatures but high neuroendocrine signature scores. And you can see that all of the patients in this group had biopsy-proven neuroendocrine prostate cancer, so we classified these as the neuroendocrine CTC phenotype. We then asked whether these phenotypes were associated with prognosis or clinical outcomes in this cohort. And what we found was that indeed the samples with either the luminal A-like state or the low-proliferation state had relatively better survival outcomes, similar to those for patients who had lower-purity CTC samples that were not suitable for this pathway analysis.

But interestingly, not only did the neuroendocrine CTC phenotype samples have a very short survival, as we would expect with the neuroendocrine phenotype, but those luminal B-like CTC phenotype samples that were adenocarcinoma but had high proliferation had a very similar poor prognosis, similar to the neuroendocrine phenotype. And this is particularly important because it's well established that the amount or number of CTCs in a liquid biopsy has a strong correlation with prognosis in metastatic prostate cancer, as with many other malignancies.

But as you can see here, even among our samples with high CTC content—these high CTC samples—we're seeing a pretty stark difference in prognosis between the different CTC phenotypes we identified. And when we did a multivariate analysis, which included factors such as visceral metastasis, serum PSA, and CTC content, we found that both the luminal B phenotype and histologic neuroendocrine prostate cancer were independently associated with poor outcome, and likely the neuroendocrine CTC phenotype was not, because it's so closely linked to having a histologic diagnosis of NEPC.

And I also wanted to just highlight here that when we think about our adenocarcinoma phenotypes, the luminal A-like and luminal B-like, while they have this very different prognosis based on their CTC phenotype, they actually have very similar PSA levels and similar disease burden. So this is really picking out differences in biology that are not driven by tumor burden. We then asked whether these phenotypes would be recapitulated in tissue samples.

And we were, in fact, able to validate that these different phenotypes can be detected in tissue biopsies, in an independent cohort of 203 metastatic prostate cancer tissue biopsies, and that they had the same association with prognosis in that tissue biopsy cohort—meaning that luminal B phenotype that we identified in the tissue biopsy cohort again had this much worse prognosis compared to the luminal A adenocarcinomas.

So then we asked whether these phenotypes could be related to treatment response and resistance. And in order to do that, we had a substudy of 37 patients in our cohort who were receiving Pluvicto for advanced prostate cancer. And in these patients, we collected samples before they started Pluvicto treatment, and then longitudinally while they were receiving treatment, and then at the time of disease progression on Pluvicto.

And to highlight some of the key findings from this cohort, I would say the most striking finding was that patients with that luminal B high-risk adenocarcinoma phenotype had a much higher proportion of early progression—meaning cancer growth within the first 18 weeks of starting Pluvicto here in yellow—compared to patients with the low-risk CTC phenotypes, that low proliferation luminal A or the low CTC content phenotype. And we also saw that similarly, the luminal B phenotype was associated with shorter progression-free and overall survival in patients receiving Pluvicto.

We were also able to utilize the longitudinal samples in these patients to understand how androgen receptor signaling and neuroendocrine signature scores changed over time on treatment. And I'm highlighting here two examples of patients where, as they progress through their therapies, you're highlighted in the bottom. So you can see that patient 818 had been on chemotherapy and had relatively stable AR and neuroendocrine scores. They then started Pluvicto and actually progressed right through Pluvicto. But you can see that after the Pluvicto exposure, the neuroendocrine score went up quite a bit, and the AR score went down quite a bit. And actually, this patient transitioned into biopsy-proven neuroendocrine prostate cancer at this point.

And here on the right is an example of a patient who was on Pluvicto for quite a long period of time with relatively stable scores. But when they eventually progressed on Pluvicto, the neuroendocrine score actually increased, and the AR score decreased. And overall in our cohort, almost all of the patients with early progression showed this pattern of higher neuroendocrine scores at the time of progression. And so this highlights an early example of how we can use this longitudinal CTC transcriptional profiling to really start to understand lineage transitions here in relation to response and resistance to Pluvicto therapy. But also in the future, we hope to apply this to ARPIs and other types of therapies that are used routinely in prostate cancer.

So in terms of our take-homes, first of all, high-purity CTC RNA sequencing identified lineage states associated with prognosis in metastatic prostate cancer that were independently validated in a tissue biopsy cohort. In particular, we found two distinct luminal AR-driven CTC phenotypes with differing prognosis and differential outcomes after Pluvicto therapy. And using longitudinal CTC profiling, we were able to observe shifts in lineage status associated with progression on Pluvicto.

And I just wanted to acknowledge this was a large study with many contributors. Josh Lang and George Zhao led the study, and James Sperger and Josh Lang's group participated in developing the assay and the data analysis. Our many clinical collaborators, and just want to highlight David Kosoff, Rana McKay, and Xiao Wei who contributed samples to our multi-institutional cohort. And then our funding sources, including the Department of Defense, and then NCI and the Prostate Cancer Foundation, of course.

Andrea Miyahira: Well, thank you so much, Dr. Sharifi, for sharing this with us. So what heterogeneity of subtypes were observed in individual patients?

Marina Sharifi: That's a great question, Andrea. So in data that we didn't include in the paper but are continuing to analyze, what we find is that many patients do transition between phenotypes over time, especially as they go through different therapies in their disease course. And in particular, I think one of the big questions that we're still trying to understand is whether there's a sequential progression from luminal A to luminal B to a neuroendocrine phenotype, or whether that luminal B phenotype represents a terminal high-proliferation, very aggressive state that's distinct from neuroendocrine prostate cancer.

I will say in some of our early results, we do see that there do seem to be transitions between luminal B and neuroendocrine, suggesting that it may be the former, but we're still working through the longitudinal analysis to understand that better. I think also getting to the question of heterogeneity at a single time point, with our bulk RNA sequencing, it is challenging. We can't really determine whether transitions that we see are due to evolution pressure on a heterogeneous population versus actual plasticity of phenotypes.

But using parallel single CTC RNA sequencing and protein phenotyping approaches, we are starting to address those questions as well. And I will say that there certainly is some evidence of heterogeneity at a single-cell level that might be driving some shifts.

Andrea Miyahira: Thanks. And as far as the heterogeneity, what do you think this study revealed about implications for targeted therapies?

Marina Sharifi: Yeah, no, I think that's a really important question because I think that heterogeneity is something that underlies some of the treatment resistance that we see. For example, in the luminal B phenotype, where we found first of all a very poor prognosis overall, but also really a lack of response to Pluvicto, which may just reflect the underlying disease biology of that state. We do observe co-expression, for example, of different cell surface targets.

So there's some cells in those samples that don't have PSA expression versus some that do. And so one of the things that we look at in the paper is the heterogeneity of cell surface target expression across these phenotypes. And I think that opens up some opportunities for potential multiple dual targeting approaches that might be more effective in overcoming heterogeneity as a driver of treatment resistance.

Andrea Miyahira: Thanks. And how did the CTC subtypes identified compare with previously reported CRPC subtypes?

Marina Sharifi: Yeah, so I think in CRPC there's been a number of different publications looking at different transcriptional phenotypes. And in fact, in our clustering analysis, we use transcriptional signatures that have been previously defined as different subtypes from multiple groups. But I think what we're finding is across all of these groups—and actually, we recently published another paper looking at tissue biopsies where we're comparing all these different signatures in Molecular Oncology—I think what we're seeing is that there are some big categories that come up over time across different studies. Obviously, neuroendocrine prostate cancer is a well-established phenomenon.

But within the adenocarcinomas, there's some differences in different categorizations, but the amount of AR signaling and how that impacts lineage—I think there's been some differences between tissue and liquid biopsy studies. And I think also, one question that we've gotten a lot is, what about double negative? The double negative phenotype, which we don't really seem to see in our CTC samples. We have that group of CTC samples that has low AR and low neuroendocrine signature scores, but they don't have the poor prognosis that's been seen with double negative tissue biopsies.

So I think it's still an open question whether that phenotype may not be as prone to shedding CTCs versus a difference in the microenvironmental contribution to the transcriptional program—that our samples are free of any microenvironmental influences apart from the circulating milieu, which is obviously very different from the lung or the liver. And so whether that transcriptional phenotype that's been seen, the double negative phenotype, whether that's somehow microenvironmental-conditioned, and when those cells shed, they no longer have that phenotype. Maybe they look like luminal B in circulation. So I think we still don't know that.

Andrea Miyahira: Thank you for that. So what are your next steps? And do you have any translational plans?

Marina Sharifi: Absolutely. I think that's what we're most excited about. This paper really showed that we can do this with CTC RNA sequencing in metastatic prostate cancer. But I think the real value of this approach is going to be implementation into randomized trials, where we're able to monitor for resistance mechanisms over time and understand more about how those lineage states impact treatment response and resistance.

And so this RNA sequencing assay is now being integrated into multiple ongoing trials in metastatic prostate cancer, including the Alliance PREDICT trial and the ARCTIC trial, where both novel therapies for prostate cancer and, in the case of the ARCTIC trial, looking at ARPI response and resistance, will give us data sets where we'll be able to really explore how these lineage states impact treatment response across different types of prostate cancer therapies.

And I think also, with our initial findings in our Pluvicto substudy, we're also very excited to expand that, and we're continuing to grow that Pluvicto cohort for patients receiving standard-of-care Pluvicto, where we'll be able to understand more about how the longitudinal changes in AR and neuroendocrine signature scores are associated with response and resistance.

Andrea Miyahira: OK, well, thank you so much. And I look forward to seeing the next study.

Marina Sharifi: Thanks, Andrea. Thanks so much for having me.