Single-Cell Analysis Reveals Heterogeneity in Treatment-Resistant Prostate Cancer - Samir Zaidi
December 2, 2024
Biographies:
Samir Zaidi, MD, PhD, Genitourinary Oncologist, Memorial Sloan Kettering Cancer Center, New York, NY
Andrea K. Miyahira, PhD, Director of Global Research & Scientific Communications, The Prostate Cancer Foundation
Single Cell Analysis of Treatment-Resistant Prostate Cancer: Implications of Cell State Changes for Cell Surface Antigen Targeted Therapies.
Cell Surface Targets in mCRPC: Expression Landscape and Therapeutic Potential - Michael Haffner
STEAP1-Targeted T-Cell Engager Demonstrates Efficacy in mCRPC Patients - William Kevin Kelly
Andrea Miyahira: Hi. I'm Andrea Miyahira with the Prostate Cancer Foundation. With me today is Dr. Samir Zaidi of Memorial Sloan Kettering Cancer Center. He will go over his new paper, "Single-Cell Analysis of Treatment-Resistant Prostate Cancer: Implications of Cell State Changes for Cell Surface Antigen-Targeted Therapies," which was published in PNAS. Dr. Zaidi, thanks so much for joining.
Samir Zaidi: Thanks so much for having me. Well, great. Andrea, thanks so much for this wonderful opportunity to share our work, as you mentioned, that was published in PNAS just a few months ago. And it really is a deep characterization of prostate cancer tumors using single-cell technologies. And so I wanted to start just by laying out the land of what really happens in late-stage prostate cancer. As viewers know, the androgen receptor is the major oncogenic program that drives prostate cancer.
And as a result, there's been a tremendous amount of success using androgen receptor signaling inhibitors, such as enzalutamide or Zytiga, to block this pathway. But as a result, there's really been a plethora of different resistance patterns and mechanisms that have occurred. One is obviously reactivation of the androgen receptor. Even in the presence of the inhibitor, it can drive the tumor.
But another mechanism is really independence from the androgen receptor, and this is known as lineage plasticity. It's one of the nine Cancer Grand Challenges. And it really is a major biological and clinical problem for our field, in that it accounts for around 20% of all castrate-resistant or late-stage prostate cancer patients.
And so another way to think about this is that more of a garden-variety prostate cancer tends to be androgen receptor positive. And so one of the resistance mechanisms is, of course, a continued oncogenic addiction to prostate cancer. But we and many others have described that there are a number of tumor states that can emerge in late-stage prostate cancer, and these obviously have strong therapeutic implications. These include neuroendocrine prostate cancer, a stem cell-like program, WNT-activated program, and acquisition of a GI lineage. And this is what I'll talk about here, that we took a deeper dive into this.
The other thing I'll mention is that there are certain cells or tumor cells that are essentially in transit. And this is what we found a year and a half ago in this collaborative project, where these cells tend to be hyperactivated for JAK-STAT. And in fact, inhibiting them with a drug can reverse this process into a more androgen-responsive state. And so we first looked at this in collaboration with Michael Haffner and the Fred Hutch. And I won't go through all of this, but this is essentially an immunohistochemical evaluation of a number of tumors.
And what you can see in a single patient looking at all of their tumors is that there are certain tumors that have now transformed into this lineage plasticity and high-grade neuroendocrine prostate cancer. But then there are certain tumors within the same patient that retain their androgen receptor status and are more responsive to therapeutics. And so single biopsies and marker stains really do not provide the entire story.
And so we went into performing single-cell analyses of patient tumors. And we really tried to use PET imaging in collaboration with Mike Morris here to really pick sites that were clinically relevant. These include areas, for example, here that are highly glycolytic for tumors but may not have positivity for a marker for the androgen receptor, i.e., are becoming lineage plastic or away from the androgen receptor. And what we saw immediately, and this is UMAP where each patient is labeled by their number and color, is that there's really a diversification of gene programs within castrate-resistant prostate cancer and neuroendocrine prostate cancer.
And in fact, the schematic I showed before is far more complicated in that you can see in the androgen receptor group on the right, there are a number of gene regulatory networks that exist, including this inflammatory state and acquisition of a GI lineage. And then in the AR-negative group, there are acquisition of embryonic programs, inflammatory programs, WNT programs, among others. And then even in the transformation into neuroendocrine prostate cancer, there seem to be three major hubs, at least within our data set.
When looking even more into this single-cell data set, what you can see is that you can even identify those subclones that are in transit, that are JAK-STAT and FGFR high. And you can see those by the blue dots here, where they're now very high on the x-axis for JAK-STAT and FGFR activation but are low in androgen receptor signaling, suggesting they're actually going away from that cascade that is oncogenically addicted to the androgen receptor. And these tend to be high in EMT. And so one could think about trying to identify these through different biomarker methods and actually inhibiting these cells to now make them, again, sensitive to the androgen receptor signaling inhibitors.
Now I'll just end by telling you about two examples that clearly have implications based on these different gene regulatory networks. One is PSMA that you can see here. PSMA is shown on the y-axis and AR module score is shown on the x-axis. And if you can see, most of the gene regulatory networks that are androgen receptor positive appear to have expression for PSMA. But that's not always the case. If you have acquisition of this GI lineage, your PSMA tends to be quite low.
Also, all of the AR-negative states, at least within our cohort, and these neuroendocrine tumors also had low PSMA expression. And of course, this has implications for how patients may be responding to PSMA-directed radioligand therapy, which was, of course, approved through VISION. Now, if you look at a neuroendocrine tumor, and this is from a single patient, you can see that one of the transcription factor networks that's highlighted in green, that's marked by ASCL1, is shown here. But in another region of the liver, which is marked in pink, this is a NEUROD1-dominant transcription factor program.
And you can see that if you look at one of the therapeutic antigens that is being currently investigated, just targeting that antigen DLL3 may only have efficacy within the ASCL1 regions but may not effectively treat the NEUROD1-positive regions. And so again, understanding these transcription factor networks and their associated expression of cell surface markers or therapies, I think, is extremely critical for the field.
There are opportunities for shared cell states. Here we've looked very deeply into neuroendocrine prostate cancer and small cell lung cancer. And you can really see that if they have the same transcription factor networks, they tend to be very similar tumors and may have very similar responsiveness to therapeutics. And so I won't go over all the conclusions from this publication.
But I will say that while IHC-based approaches may yield improved insights into cell states, they really do not capture heterogeneity across or within patients and responsiveness to antigen-based therapy. And I think with advances in diagnostic liquid assays, these transcription factor GRN classifications may serve as additional tools to really refine patient selection for trials, and hopefully in the future help inform therapeutic decisions. And I'm happy to switch there and discuss this more.
Andrea Miyahira: Thank you so much, Dr. Zaidi, for sharing such a great study. So what are the therapeutic implications of this study? Do you think we are pursuing targeted therapy appropriately, or are there better targets or target combinations?
Samir Zaidi: Yep. It's a great question, Andrea. I mean, I think one of the things that I think this shows and builds upon other studies is that there's really a tremendous amount of gene program and signature diversification as you get into late-stage disease. And thinking about early-stage disease and late-stage disease, I think, is a whole different kind of bag of worms. Here, I think one of the key implications is that in late-stage disease, you are really going to have differences in expression of cell surface markers. And how to really address that heterogeneity is a major question within the field, not just in prostate cancer, but I think more globally across many tumor types.
Now, I think our field is doing quite a good job overall. PSMA continues to be a really great target for prostate cancer. But we saw in this study and other studies, for example from the Fred Hutch, that its expression can be variable across tumors. There are other targets that I think people are aggressively pursuing which are direct AR targets, which account for the large majority of prostate cancer. Those include STEAP targets and KLK2, and their specificity, I think, overall for prostate cancer hopefully will make them good targets.
I think there's also this concept of really thinking about biomarker development in this space and how to best do that. I think one of the main successes, at least for the VISION trial, is that there's a companion biomarker. There's a PSMA PET imager that goes along with a theranostic or therapeutic agent. And that really gives us a sense as to what the disease process is within a certain patient.
So I think that allows us to see it, and then treat it, which I think is a very important principle. This could hopefully be extended to liquid biomarkers. One could imagine, I think, through some work from the White and Hogg groups, that you can really try to infer transcription factor occupancy within liquid biopsies. And I think that would be a really strong way to be able to see if some of our findings, for example, are able to be detected within blood samples. And that may inform us from tumor tissue in some of the work that I've shown, exactly what therapies may be effective in that setting.
The other thing I'll say is that from some of the even IHC studies, patients can have mixed disease. And so being able to identify that through multiple PET modalities—that's been done before through FDG and PSMA—but even extending that to more specific targets, I think, will be very useful to understand the compendium of disease because we don't really know what tumor state is going to be the one that ends up being the most aggressive. We have some thoughts, obviously, as a field, but I think that's very important.
Andrea Miyahira: OK, thank you for that. And the GI lineage is not reported on widely. Can you tell us more about that lineage?
Samir Zaidi: Yeah, absolutely. There's some work from Yu Chen at MSK that originally—it's a Cancer Cell paper—that originally showed that these HNF4 factors promote a GI lineage and can confer androgen resistance. And I think the prevalence of this hasn't really been looked at more globally. But it certainly, I think, is a strong mechanism in which you can get these aberrant lineages. This is being seen more and more, I think, globally across cancers, where even in pancreatic or colon cancer or other tumors such as lung, you can get these other developmental programs that really turn on.
And so I think understanding the therapeutic vulnerabilities of these cell states is so critical, because one could imagine trying to be able to detect them through some sort of biomarker detection in that field is rapidly evolving. But then what are the therapeutic vulnerabilities that exist, and how can we delay progression of disease and improve survival? I think that's one strategy. The other strategy, obviously, from this paper and others, is really to target earlier on where there's less transcriptional diversification. But to really be able to pick up these clones early on, I think, is going to take quite a lot of work on that end in terms of biomarker development.
Andrea Miyahira: OK, thank you. And what are the next steps for your studies? Do you have translational plans?
Samir Zaidi: Yeah. So I think one of the key next steps of this study is to try to understand the dependencies of these different states more on a functional basis. This is occurring in a number of labs, but we're also very interested to pursue it in a systematic fashion. And also in that light, look at the regulators of transdifferentiation. Essentially, you need to know how the androgen receptor-positive lineages actually transdifferentiate over time into these lineages and what those regulators are. And then what the dependencies are within each of those cell states, just to keep up as biomarkers are improving the therapeutic and the mechanistic aspects. So I think that's a really important aspect.
The other thing from our JAK-STAT work that was published, we have a phase 1b/2 study that's open at MSKCC, along with Wassim Abida here, with a JAK-STAT and FGFR combined inhibitor called tinengotinib from a company called TransThera Biosciences. And to see—well, we have a lot of preclinical data for it. And that's one of the aspects that's been mechanistically understood, to see whether there are biomarkers and responsiveness that one can develop within patients in certain stages of disease. And so I think that's going to be hopefully very exciting to see the results of that trial.
The last aspect, obviously, is pairing with other labs that really have a lot of emphasis within liquid biomarker development and integrating large data sets from tumor samples that have undergone single-cell sequencing to really blood-based biomarkers. And I know, for example, the PCF is supporting a number of these groups. But to integrate that data to really try to understand how we can understand responsiveness and how we can understand different therapeutic vulnerabilities in that context.
Andrea Miyahira: OK, well, thanks so much, Samir. I really appreciate you sharing this study with us.
Samir Zaidi: Of course, my pleasure, and thanks so much for having me.