Read the Full Video Transcript

Andrea Miyahira: Hi everyone, I'm Andrea Miyahira here at the Prostate Cancer Foundation. Here with me today is Dr. Michael Haffner, an assistant professor at the Fred Hutchinson Cancer Center. He will discuss his group's recent paper "Assessment of TROP2, CEACAM5, and DLL3 in Metastatic Prostate Cancer: Expression, Landscape, and Molecular Correlates." This was recently published in NEJM Evidence. Dr. Haffner, thanks so much for joining us today.

Michael Haffner: Thank you very much, Andrea. It's really a great pleasure on behalf of all of my co-authors to present this work to you today. And over the past couple of years, the targeting of cell surface proteins that are preferentially expressed on tumor cells has really emerged as a very promising therapeutic approach in advanced cancers. And by using small molecules, peptides, or antibodies, one can now deliver toxic payloads, for instance, chemotherapeutics, radionuclides, but also immune cells specific to cancer cells, resulting then in a cancer-specific killing that's sparing a lot of the benign tissues. Now while these approaches are very intriguing, what all of them have in common is that they critically rely on the expression of cell surface targets and the understanding of the expression pattern of these targets becomes absolutely key.

Now, over the past decades, we've also learned that prostate cancer is a highly heterogeneous disease. And work from Peter Nelson, Colm Morrissey, Mark Rubin, and Himisha Beltran have really shown us that prostate cancer, in particular advanced metastatic prostate cancer, can be subdivided into four molecularly and clinically defined groups that are characterized by the expression of the androgen receptor and by the expression of neuroendocrine markers. So one can now subdivide castration-resistant prostate cancer into these four different flavors. Tumors that express the androgen receptor but don't show any expression of neuroendocrine markers, these are the classical adenocarcinomas. Or you can also see tumors that lose the androgen receptor axis expression but gain neuroendocrine marker expressions and are neuroendocrine prostate cancers.

We find also mixed tumors that show features of AR positive tumors and neuroendocrine marker positive tumors. And there's this emerging group of so-called double negative tumors that lack both androgen receptor as well as neuroendocrine marker expression. So a high level of complexity on the molecular level. And in more recent years, it's also become clear that even within a patient there can be a high level of intra-patient heterogeneity with individual metastatic sites showing very distinct and diverse phenotypes. So appreciating the molecular complexity and heterogeneity of castration-resistant prostate cancer, and also understanding that there is a critical need to understand the expression pattern of different cell surface targets, in this study we specifically focused on the evaluation of three targets, namely TROP2, DLL3, and CEACAM5, which all have different targeting strategies currently under clinical investigation.

So the objective of this study was to really understand the expression landscape of these targets. And the way we designed this study is we wanted to assess protein expression in a very large cohort of samples collected at rapid autopsy, and this includes more than 750 samples from 52 patients. Again, a key point from this study is that we used a rapid autopsy cohort, the University of Washington Rapid Autopsy cohort, which allowed us not to just sample a single tumor but multiple tumor sites representative of the entire tumor burden of the patients. So applying validated immunohistochemical protocols to assess the expression of TROP2, DLL3, and CEACAM5, here are some representative micrographs on the right.

You can quickly see that there is a diversity in the expression pattern of these targets across different molecular subtypes. And maybe this is better summarized here in this far graph where you can see that DLL3 expression is almost restricted to these AR negative neuroendocrine marker positive tumors and that's similar to what Himisha Beltran's group had described previously. CEACAM5 shows a little bit of a broader expression but also a very strong enrichment for high-level expression in AR negative neuroendocrine marker positive tumors in line with prior studies by John Lee. And TROP2 showed a completely different expression pattern where highest levels here were seen in AR positive neuroendocrine marker negative tumors as well as AR positive neuroendocrine marker positive and double negative tumors. So molecular subtypes are really a key determinant for the expression of these individual targets.

We also had the opportunity now to evaluate patterns of co-expression of these individual targets and what we started to see is that combinations of two targets, for instance in the context of AR positive neuroendocrine marker negative tumors, a combination of PSMA and TROP2 allowed for a potential targeting of 99% of all of the tumors in this molecular phenotype. Similarly, a combination of DLL3 and CEACAM5 would allow for targeting of more than 87% of tumors. So really thinking about potential opportunities for co-targeting strategies leveraging these markers. The unique design of this cohort that allowed for the sampling of multiple different metastatic sites from each patient allowed us now to ask the question whether the expression of DLL3, TROP2, and CEACAM5 would differ across different anatomic sites.

Shown in these box plots are now individual metastasis of each of these individual dots, color coded by their molecular phenotype across different anatomic sites. And although we saw a statistically significant lower level of TROP2 expression in liver compared to bone metastasis and the higher level of DLL3 expression in liver compared to bone metastasis, these differences were numerically very small and are likely not biologically significant. The unique sampling protocol also allowed us to assess the heterogeneity of expression of DLL3, TROP2, and CEACAM5 within an individual patient. In the box plots to the right, we see expression of each individual metastasis shown by a discrete dot color coded by its molecular phenotype for the four targets. And one can appreciate that the expression of DLL3, particularly in cases that are AR negative and NE positive, is by and large uniformly high. In TROP2, we see a strong enrichment for expression in AR positive neuroendocrine negative tumors with overall a relatively high level of homogeneity within a patient. CEACAM5, however, shows an increased level of intra-patient intratumoral heterogeneity.

Lastly, we also sought to determine genomic and epigenetic determinants of TROP2, PSMA, DLL3, and CEACAM5 expression. To this end, we correlated the protein-based expression levels with the genomic status of key driver gene alterations found in CRPC, and observed that higher PSMA and TROP2 expression was found in cases that harbored AR alterations, in particular AR [inaudible 00:09:45]. Whereas higher levels of DLL3 expression were found in tumors with bi-allelic RB1 inactivation. While these genomic associations showed some connection between the underlying genomic alteration and expression patterns, most likely there are other drivers that affect the expression of these genes. We therefore investigated the epigenetic states of DLL3, TROP2, and CEACAM5 in PDX tumors leveraging previously published ChIP-seq data for key histone marks including H3K27 trimethyl and H3K27 acetyl marks.

And we observed a very tight association between the presence of the Polycomb mark, H3K27 trimethyl, and the repression of expression for all three targets, potentially suggesting that pharmacological interventions that target these repressive marks such as EZH2 inhibitors or EED inhibitors could be used to modulate and augment the expression of these targets. In summary, we found that TROP2, DLL3, and CEACAM5 are relevant cell surface targets expressed in CRPC. While TROP2 expression was present across all molecular subtypes except for tumors that are AR negative NE positive. DLL3 and CEACAM5 expression was highly enriched in AR negative NE positive tumors. We also observed that TROP2 and DLL3 showed a relatively low intra-patient intratumoral expression heterogeneity, and we determined that certain genomic alterations, in particular AR gain and RB1 loss, and epigenetic features, in particular gain of H3K27 trimethyl marks were associated with the expression of these cell surface targets.

Andrea Miyahira: Thank you so much, Michael, for sharing that. So what are the major conclusions regarding intra-patient heterogeneity? Did you see any dramatic differences in molecular subtype and marker expression across tumors within individual patients?

Michael Haffner: Yes, it's a really good question. As I had alluded to, there's a very tight association between molecular subtype and the expression of these individual targets where we see very high expression of DLL3, mostly restricted to neuroendocrine prostate cancer or these AR negative, pardon me, AR negative neuroendocrine marker positive tumors. Conversely, TROP2 was not expressed in this particular subset of tumors. In terms of heterogeneity of expression compared to other cell surface targets that we had evaluated before, in particular PSMA, TROP2 showed a more homogeneous expression across all metastatic sites. Again, potentially suggesting a more favorable distribution in that advanced stage of the disease that was sampled here.

Andrea Miyahira: Okay, thank you. So I guess thus far from what you've seen in your studies and other studies, what do you believe are the best markers and targets for the various MCRPC molecular subtypes?

Michael Haffner: I think that's a very good question and certainly the field is evolving very, very quickly and we're learning more about the expression of these markers as well as their clinical utility. I think for AR positive neuroendocrine marker negative tumors, PSMA, but also STEEP1 and KLK2 are very, very important and relevant targets in particular because of their highly selective expression in cells derived from prostatic epithelial origin. For tumors that are AR negative neuroendocrine marker positive, the analysis that we and others have performed have really favored DLL3 as a very promising target. And there's a lot of development for different agents in this space, of course. And then if you think about other targets that are more broadly expressed, I think our studies have shown that TROP2 shows expression across certain subtypes that are usually not covered very well by other currently available targets. In particular, that subset of double negative prostate cancers that are very difficult to treat.

Andrea Miyahira: And based on these data, how would you propose designing targeted therapy studies? Would you use combinations against surface antigen targets or epigenetics? And what selection criteria or biomarkers might be most effective?

Michael Haffner: I think understanding expression of these individual targets within a patient is still very, very important. And of course, there are different approaches to do that, perform biopsies and look at tissue-based expression. For some of these targets, there are imaging agents available that allow for direct in vivo imaging of the expression pattern. And then I think there's also a great opportunity to think creatively about the use of liquid biopsy approaches that can be applied to infer the expression for some of these targets. While expression of the target I think is still highly relevant and a key determinant, we also want to think a little bit more about the resistance mechanisms that are likely present intrinsically in some tumors or that are evolving as a consequence of the therapy.

And I think right now we know very little about the common resistance pathways for these individual interventions, may that be an antibody drug conjugate or radioligand or even CAR T-cells. So I think we need to understand the biology here a little bit further to then refine our approaches and maybe come up with a strategy where we measure both the expression but also potential resistance markers simultaneously to select patients most appropriately.

Andrea Miyahira: Okay. Well, thank you again, Dr. Haffner, for coming in and sharing this with us today.

Michael Haffner: Thank you.