David Goodrich: Thank you, Andrea. I'm happy to be here. So this project came together with a number of people here at Roswell Park, as well as some collaborators at Dana-Farber with support from the Prostate Cancer Foundation. And we were interested in EZH2 and its role in prostate cancer because for many years, EZH2 has been implicated as a potential therapeutic target for treating advanced prostate cancer as indicated by some of the data I'm showing here from other people. So for example, EZH2 expression increases as prostate cancer disease progresses and patients who have tumors with high EZH2 expression actually fare more poorly than patients who have prostate tumors with low EZH2 expression. EZH2 has also been implicated in prostate cancer lineage plasticity, which is a form of therapeutic resistance where the prostate cancer switches from an epithelial phenotype to an alternative lineage state that's no longer dependent on androgen receptor. And you can see down here that in both human and in mouse models that the plastic forms of prostate cancer express much higher levels of EZH2. So EZH2 expression increases in advanced prostate cancer. It correlates with poor endocrine outcomes, and it's been implicated in prostate cancer lineage plasticity.
It's also important to know that EZH2 inhibitors have been developed and are being evaluated clinically for the treatment of prostate cancer. So the knowledge gap we tried to address is how does EZH2 functionally impact prostate cancer progression? And two, does EZH2 actually regulate or impact prostate cancer and lineage plasticity? And the approach we took to do this was through the use of genetically engineered mouse models, which are described here. We had previously described a series of mouse models that developed prostate cancer that exhibited lineage plasticity, and we suppressed EZH2 expression in these models by genetically knocking out either one or two EZH2 alleles. This data here simply shows you that the knockout works, that the EZH2 levels go down. And because EZH2 is the catalytic subunit for the PRC2 complex that places this particular histone mark or this post-translational modification on histones, that it's also reduced when EZH2 is suppressed. So our first finding was that when you eliminate EZH2 from these prostate cancers, that there's not a significant difference in overall survival in these mice. It's shown here. Actually, the mice with prostate cancer that's lacking EZH2 actually progresses faster.
This is a mouse that has wild-type EZH2, and this is the mouse that has one allele of EZH2 missing. So there are differences in the survival, but the loss of EZH2 does not consistently improve survival. The other thing we looked at is response to therapy. In this case, we castrated the mice to simulate androgen deprivation therapy. This is the mouse with wild-type EZH2, and this model does respond to castration, so the mice live longer with castration. The model with one EZH2 missing, there's really a poor response to castration, and the model in which both alleles are missing responds about the same to castration. So there was no significant improvement in survival in these mice, either in response to therapy or in normal cancer progression. The second main finding we made was that EZH2 does impact lineage plasticity, but it actually diversifies prostate cancer lineage state evolution. So this is a single-cell RNA sequencing analysis of tumors developing in our various mice here, color-coded by the genotype of the mice, but here are all the different prostate cancer subtypes that you can detect in these mice. The adenocarcinoma is over here in green and yellow and blue. There's an EMT variant that's inflammatory in red here, a basal variant in pink. Neuroendocrine variants, four different neuroendocrine variants here, and then a double-negative prostate cancer variant here. And you can see depending on the EZH2 status that EZH2 loss actually redirects the evolution of these variants in different ways. So when you're completely missing EZH2, you tend to favor these two variants.
A NeuroD1 variant of neuroendocrine prostate cancer develops much more frequently in these mice. And also the double-negative prostate cancer develops much more frequently in these mice. When you lose one EZH2 allele, you tend to favor these two forms. It's a variant of neuroendocrine prostate cancer that expresses neuroendocrine or neurofilament genes. And this is an amphicrine variant that expresses both NEPC markers as well as androgen receptor. This figure down here just shows you the relative distribution of these different variants in the different genotypes. And one thing I'll highlight here is the basal variant in pink seems to decline in both EZH2-deficient mice, and I'll come back to that later. So we wanted to evaluate whether what we were seeing in mice was similar to what is observed in human patients, and it is. So this is human patient single-cell RNA sequencing data from castrate-resistant prostate cancers in six different patients here. And what we've done is we've taped all the single-cell RNA data. This is color-coded by patient, but here it's color-coded by gene expression signatures that match the different variants we detected in mice. So you can see that some human patients develop this NEPC-ASCL1 variant, some develop NeuroD1 variant, amphicrine variant here, double-negative prostate cancer here, and this NEFL variant here. So all the variants that we can detect in mice are also detected in human prostate cancer. So we think the mice and humans are analogous.
And consistent with that, if you look at the different variants in humans, this is again, human data, and you look at the relative EZH2 expression, you can see that the neuroendocrine ASCL1 variant expresses the highest EZH2, which is consistent with what we're seeing in mice, but all the other variants express lower levels. So this would be consistent with what we're seeing, when we suppress EZH2 expression, we tend to favor these other variants. So the main conclusion from this paper are the following, that EZH2 loss does not restrain progression of prostate cancer that is prone to lineage plasticity, but it does alter the lineage state evolution within those cancers. So it diversifies the prostate cancer lineage variants that evolve in those tumors. We also showed that EZH2 is not required for neuroendocrine prostate cancer. In fact, EZH2 may be required for or may promote one form of neuroendocrine prostate cancer, but when you lose it, other forms of neuroendocrine prostate cancer can evolve. We've shown that EZH2 loss tends to activate alternative lineage-specifying transcription factors like KLF factors, and that's probably responsible for the different lineage variants that we see. And finally, I noted earlier that the basal prostate cancer subtype is disadvantaged by loss of EZH2. So this may be clinically relevant. So EZH2 inhibitors may yet be effective in prostate cancers that exhibit this basal-type variant, which is common in castrate-resistant prostate cancer patients.
And finally, we think we were able to answer one of the questions revolving around the origin of this more recently discovered double-negative prostate cancer. There's some speculation as to whether this is a separate evolutionary trajectory or whether it's an intermediate state on its way to neuroendocrine prostate cancer. And we were able to show that it's actually a separate evolutionary state that's distinct from the evolution of neuroendocrine prostate cancers. And we summarize all of that in this figure here. So we start with a high-plasticity intermediate state. From that state, you can either develop this basal variant or the double-negative variant or the neuroendocrine ASCL1-high variant. And then from this ASCL1-high variant, you can eventually evolve into three additional different neuroendocrine variants. So that's how we think these different variants are related, at least in the mice, and probably similarly in the human. And I'll stop there and acknowledge the people who did the work in my lab, my collaborators at Roswell Park and my collaborators at Dana-Farber, as well as funding support from the Prostate Cancer Foundation and the National Cancer Institute.
Andrea Miyahira: Wonderful study. Thank you so much, Dr. Goodrich, for sharing this with us. So what does this tell us about the role of EZH2 in lineage plasticity, and what do you think it's actually doing?
David Goodrich: Yeah, I think EZH2, its canonical function is through the PRC2 complex. There are also non-canonical functions that have been implicated EZH2 by itself outside of the PRC2 complex. But in our mouse models, I think it is actually working to drive the ASCL1 neuroendocrine phenotype. It actually promotes that. And when you lose EZH2, it diversifies the evolutionary trajectory. So it favors other alternative lineage states. So I think it's consistent with what we originally thought about EZH2's role in NEPC, but what we didn't expect was that when you inhibit it, the cells can still evolve into other states. And so EZH2 loss actually diversifies the lineage variants that you see in prostate cancers that are prone to lineage plasticity.
Andrea Miyahira: Thank you. And in the paper, you really highlight differences in biology that you see between the models where EZH2 is a wild type versus loss of one allele versus loss of both alleles. So what do you think is mediating these differences?
David Goodrich: Right. So, one of the main differences between loss of one allele and loss of two alleles is loss of one allele, obviously the protein is still there, but it's at a reduced amount. And we speculate and we have some evidence that that reduces the amount of activity of the PRC2 complex and the amount of H3K27 histone modification that you see. So we think loss of one allele is a quantitative effect. It's reducing the histone modifications, it's reducing the activity of PRC2, but loss of both alleles actually lowers that further, but it also completely eliminates the protein. So you're also losing any sort of non-canonical activity of the EZH2 protein itself. So we think that's probably the basis of the difference in phenotype that we're seeing. It's both quantitative in the amount of PRC2 activity, but it's also qualitative because when the protein's gone, it can't carry out those non-canonical functions.
Andrea Miyahira: Thanks. And based on this study, how successful do you think EZH2 inhibitors will work clinically and what considerations such as biomarkers or combination therapies might be necessary for success? And so what are your take-home messages for the pursuit of EZH2 inhibitors in the clinic?
David Goodrich: I think my take-home message is that the effectiveness of the inhibitors are probably going to be context-dependent. We know from human patient data that some prostate cancers are prone to lineage plasticity and some aren't. We also know that castrate-resistant prostate cancer often exhibits a basal phenotype. So I think in some circumstances, either when the cancer is not prone to lineage plasticity or when it exhibits that basal-type CRPC phenotype, that that would be where I would be looking to see effectiveness of such inhibitors. But if you're looking at prostate cancer that's prone to neuroendocrine prostate cancer, they're highly plastic, I think inhibiting in that circumstance is likely just to drive alternative lineage variant evolution, and then it will come down to whether that is a clinical benefit or not. So we don't know yet whether, for example, double-negative prostate cancer has a poor or better prognosis than neuroendocrine prostate cancer, but it's going to depend on results like that.
Andrea Miyahira: Thank you. And these and other recent studies in NEPC identify many variants and different factors that drive these different subtypes, which suggest context dependencies that are different. So do you think this suggests a much more complex pathway to these different subtypes and what do researchers need to do to map out this complexity?
David Goodrich: Exactly. It says that these cancers have a lot more ability to change and adapt than we originally thought, and that it's going to be a lot of hard work to map out exactly what the limitations of that plasticity are. My thinking would be given that complexity, that it may be better to understand why some prostate cancers are plastic to begin with. So in order to evolve along these complex different pathways, they have to be plastic to start with. And we know that maybe 25% of the castrate-resistant prostate cancers are likely to be plastic, but many are not. And so, I think if we could figure out why they're plastic to begin with and maybe interdict at an earlier stage, maybe we wouldn't have to worry about this complex evolution that occurs once the plasticity exists. So we're working on approaches to limit that plasticity in the beginning so we don't have to worry about some of these lineage variant evolution.
Andrea Miyahira: Thanks. And are there any drivers of lineage plasticity that you think would be most promising as therapeutic targets to prevent progression in this disease heterogeneity?
David Goodrich: Yeah, I think that still remains to be seen because the requirements for these different lineage variants may be different. They may rely on different epigenetic regulators. They obviously rely on different transcription factors. So I think it's difficult to predict given the complexity that we've already just discovered. So I'm not sure. And again, I would like to focus more on what is allowing a fraction of these cancers to be plastic to begin with. And if we can understand the molecular mechanisms there, maybe we can suppress that plasticity before it allows these cancers to start evolving into these more worse prognosis lineage variants.
Andrea Miyahira: Okay. Well, thank you so much, Dr. Goodrich, for sharing this study with us today.
David Goodrich: My pleasure. Thanks, Andrea, for the opportunity.