Novel Technology for Circulating Biomarker Development
SESSION 6: Novel Technology for Circulating Biomarker Development
Moderators:
Himisha Beltran (Dana-Farber Cancer Institute)
Alexander Wyatt (University of British Columbia)
PC-SYNERGY: Biological and Therapeutic Insights from a Metastatic Prostate Cancer Atlas at Single Cell Resolution (TACTICAL AWARD)
Peter Nelson (Fred Hutchinson Cancer Center)
Circulating Tumour DNA Genomic Correlatives in mCRPC Treated with LuPSMA or Cabazitaxel from the Randomised Phase II TheraP Trial (ANZUP 1603)
Edmond Kwan (Monash University, Australia)
View the Transcript Below:
Novel Technology for Circulating Biomarker Development
Himisha Beltran, MD [00:00:06] Good morning, everyone. We’ll get started. My name is Himisha Beltran. I am from the Dana Farber Cancer Institute. I’m excited to co-chair this session with my friend Alex.
Alexander Wyatt, PhD [00:00:18] Hi everyone, Alex Wyatt from Vancouver Prostate Center in University of British Columbia, Canada.
Himisha Beltran, MD [00:00:24] This has been a really great meeting. Thank you so much, Howard and Andrea. It’s always a fantastic event. Great to connect with you, Alex, and everyone here. Felix was mentioned a lot at this meeting. We miss him a lot and I think he would have especially loved this session as this session touches on a lot of the areas that he was really passionate about. How about you, Alex? How have you enjoyed the meeting?
Alexander Wyatt, PhD [00:00:46] Yeah, you know what? A really good meeting. Again, fantastic retreat as usual and so it’s been fascinating to hear about the latest you know, mechanisms of resistance that that we’re uncovering, the new drug targets and lead compounds that are being developed and of course see clinical trial outcomes and so forth presented.
Himisha Beltran, MD [00:01:09] Yeah, and that’s a perfect segue to this session as we start thinking about how do we bring all the science that was presented to the clinic in the form of biomarkers. And this session will touch on some areas of new technology development. How do we improve copy number estimation and ctDNA? How do we leverage transcriptomics to understand heterogeneity and resistance? What new assays are in development to capture phenotype? And also, how do we start bringing these into clinical trials and highlighting an exciting trial at the end about using liquid biopsies to assess predictive and prognostic biomarkers?
Alexander Wyatt, PhD [00:01:42] Yeah, I mean obviously I’m biased, but biomarkers are really, really promising. And we saw yesterday that so many trials have either integrated or integral biomarkers, whether it’s histology, radiology, genomics of course, and so on and so forth. And I mean Dr. Spratt presents a trial where there was a hazard ratio with .29 for a biomarker, right? And I think if we had a new drug that that gave that type of hazard ratio, we’d be jumping up and down, so they hold enormous promise for improving the way that we stratify and treat patients. And so really excited today to show you what I think is the next wave of practical biomarkers that hopefully can be working their way into phase II/III designs over the coming years. Our next speaker is Dr. Pete Nelson, who is amongst other titles, Vice President of Precision Oncology at the Fred Hutchinson Cancer Research Center.
Peter Nelson, MD [00:02:35] Thank you, Alex. Good morning, everyone. Thanks for rising so early. So, I’m the spokesperson for a very large team and I want to certainly acknowledge our gratitude to Felix for his guidance and inspiration to initially conceive and launch this project. So, this is the outline of what I’ll discuss here. I’ll give a bit of background of the initial purpose, aims, and objectives, talk a bit about the team, introduce the team, and then go through some early results of developing this metastatic prostate cancer atlas, and then I’ll leave you with a summary and some next steps. These are my personal conflicts of interest. The team has received really wonderful collaborative support from a number of industry partners. So, the challenge here, I think that we’ve certainly seen at this meeting is that the phenotypic and genomic heterogeneity in advanced prostate cancers appears to drive treatment resistance and consequent lethality of the disease. So, the hypothesis underlying this proposal is that spatial and single cell approaches in combination with computational methods would identify new targets and nominate potentially co-targeting strategies that overcome interpatient heterogeneity as well as intrapatient heterogeneity. So, for this award, there were two phases. I’m only going to talk about the first phase. The first phase comprised two specific aims and the goal here is to really understand in great depth the molecular landscape of targets in advanced prostate cancer. Aim one is to create a multidimensional atlas at the single cell level that defines heterogeneity. And then the second aim is to really look in a number of the clinical studies that are ongoing to try to understand the resistance mechanisms primarily to cell surface targeted therapeutics. So, here’s the prostate cancer synergy team. Probably half of you in the audience are members of this team, and we certainly welcome many others to join at any time. It spans multiple institutions with a great variety of expertise. This is the group that actually does the work, the young investigators, so I’ve highlighted them as well. So, here’s the challenge. We’ve really seen a major change in how we’re approaching prostate cancer as evidenced by many of the talks at this meeting in shifting our focus to lineage and specifically then cell surface targets. So, whether it’s radioligand therapy, BiTE therapy, antibody drug conjugates or CAR T-cells, these all rely on cell surface targets that have very specific lineage associations. And what we’ve started to see over the last several years is that these lineage features vary certainly from individual to individual, but also within different individuals. So, we’re probably gonna see in the future a requirement or a need for consideration of co-targeting approaches. So, you might consider one particular antibody drug conjugate has some impact on tumor viability, but you’re really gonna need to consider either on that same cell or in adjacent cells considerations for combinations. So, the first aim of this project was to create an atlas that would try to define this heterogeneity. And this initially focused on metastatic biopsies and metastatic tissue as the gold standard, but increasingly moving into noninvasive or minimally invasive methods such as liquid biopsies, circulating tumor cells, and cell free DNA. So, the approach here was to first start with bulk assessments and then layer on a number of single cell-based approaches, whether it’s single cell RNA-seq or spatial assessments, including protein-based methods and RNA based methods, and then expand those into the circulating markers. So, a lot of this work started years ago by evaluating our rapid autopsy program where different phenotypes were defined primarily by AR signaling and neuroendocrine signaling. Here you can see five different categories of metastatic prostate cancer, simply based on that RNA-seq or transcriptional-based classification. What became increasingly apparent is that in a subset of patients, within a single patient, you can have multiple phenotypes. And importantly, these phenotypes, such as AR-active or neuroendocrine, were not associated with underlying genomic changes. The genomics, the genomes were largely consistent. So, this really indicates some epigenetic drive for changing phenotype, primarily under therapeutic pressure. If you look across about 60 or so different men in this particular study, the blow-up there is for one of these cases with about nine different tumors, and you can see within one man, you see first different phenotypes as well as interestingly different proliferation rates for a number of these tumors. So really starting to identify intraindividual or intrapatient tumor heterogeneity. So, the starting point in this project was then to expand on that by building a single cell atlas. So, we’ve used in this case two different cohorts, collectively about 166 tumors, profiling a little over a million cells, of which about 800,000 are prostate cancer cells. There are certainly cells comprising the tumor microenvironment. I’m not going to get into a lot of detail about that. Here’s an interesting case. This is by no means the usual, probably more the unusual. In this one tumor, we see four different phenotypes. There’s a typical AR-active subset of cells. There are two different neuroendocrine populations, one driven by NEUROD1, one expressing ASCL1, and then a sarcomatoid subtype. These all then have different cell surface targets that may be challenging to eradicate with a single type of therapy. Here is confirmatory data from that single cell data demonstrating in this tumor you have collections of neuroendocrine cells as well as ARPC cells. So, what you can do with this atlas is also then perform some interesting biology to try to understand lineage pathways. In this particular tumor, there were three clusters of tumor cells from AR positive to a double negative to a neuroendocrine, and it does look in this case like the neuroendocrine cells have passed through a double negative state. In this case, you can infer the genomic architecture based on single nucleotide ATAC-seq, where you can infer genomic alterations to trace these lineages. So, exploiting this atlas a little bit more. Here’s an example of three neuroendocrine tumors, and this is going to be a recurring theme in the talk is these plots are a little bit complex, but you’re looking at individual cells that co-express two markers. For example, a cell expressing AR and KLK3. So, the important take home here is that there were a large number of cells in these tumors that are synaptophysin positive that some may classify as neuroendocrine. However, they don’t have the neuroendocrine lineage drivers such as ALCL1 or NEUROD1, and they continue to express the AR program. So, on the right lower is a meta program just looking at AR signaling. You can see the neuroendocrine cells have no AR activity, but the double positive cells, you could call them amphicrine cells, do retain AR activity. And as I’ll show you in a little bit, they do express the cell surface targets associated with ARPC, not NAPC. So, I’ll digress just briefly and talk about the immune composition before I get back to the cell surface targets. We had initially started the project looking at single cell-based assays as opposed to single nuclei assays. On the left here is using bioinformatics approaches to infer immune composition in metastatic prostate cancers, almost 900 patients here, and you can see very low content of any immune cells in these tumors. Whereas when we did the single cell RNA-seq, you can see a much larger percentage of immune cells that comprise these tumors. So, the question is which is correct. So, on the left side are five different orthogonal strategies to quantitate immune cells in these tumors. And you can see those methods consistently show very, very low immune cells in these tumors. On the right are four different studies using single cell-based analysis showing a much higher immune cell content. So, it’s simply a cautionary note that single cell-based assays may overestimate if you’re trying to just quantitate numbers of immune cells in these tumors compared to either spatial methods or single nuclei-based methods. Here you can map out different immune cell contributions, and I would say in the vast majority of these tumors, we do not see CD8 cells, we don’t see CD4 cells, we don’t really see neutrophils. The main immune cell component are macrophages and monocytes. Okay, so coming back to the cell surface targets and considering heterogeneity, there are large number of targets being exploited by various therapeutics. Here is an example of just taking bulk RNA-seq, trying to understand the heterogeneity of expression, in this case a full H1 PSMA. So, we’ve combined five different data sets to start to look at the variability here. Here are 12 cell surface targets that we can now start to evaluate in terms of their variability. And I’ll just make three brief comments here. So, this comes back here, we’re dividing the tumors based on their phenotype, so AR active versus neuroendocrine. Those double positive cells shown in one, you can see retain high full H1 or PSMA expression. In number two, the point being that the neuroendocrine cells that express, for example, CEACAM5, those double positive cells do not express those neuroendocrine targets. And then the third point here is that there is still within ARPC, there’s substantial variation, for example, in KLK2 or any other particular target you may wish to look at. So, coming back to the single cell atlas, here’s an example of two neuroendocrine tumors within the same patient that demonstrate substantial heterogeneity in cell surface targets such as DLL3. We can now go back with enough samples and look at what’s the dominant factor driving heterogeneity. Is it within a patient tumor or is it between tumors? And so, for each of these cell surface targets, PSMA, KLK2, STEAP1, the major component of heterogeneity is between two tumors within a patient. There’s much more consistency in general within a particular patient. We can also then look to see does tissue metastatic site have any role here. And you can see that in general in the liver, full H1 or PSMA is expressed at a lower level, whereas other targets such as MUC1, you can see in those liver metastatic tumors it’s a bit higher. So, we’ve gone on now, and that was all RNA or transcript-based. So, is there a reality check looking at protein? This is using imaging mass cytometry at single cell resolution looking at protein targets and developing a protein-based atlas that integrates with the RNA atlas. Here you can see a clustering of a little more than three million tumor cells from metastases that you can partition into different phenotypes. And I’ll give you just a couple examples here that again comes back to that point about synaptophysin or chromogranin being markers for neuroendocrine. Here’s a tumor that’s AR low, PSMA negative, synaptophysin positive, but expresses the cell surface targets that you would usually associate with an ARPC, so KLK2, STEAP1, CD46. Here’s another example. This is a mixed tumor with neuroendocrine cells mixed with adenocells, and you can see this is KLK2 negative, STEAP1 positive, CD46 positive. So, it’s just a cautionary note again about the idea of using synaptophysin or chromogranin as your major discriminator of any PC. So, thinking of the future of complementary targets, there may be multiple reasons to consider combining different therapeutics. And we can start to get into an understanding of optimal strategies by looking at spatial based transcriptomics. So, we’ve performed this, have about four million or so cells from metastatic prostate cancer. This is a representative tumor where in this tumor we’re profiling about 8,000 cells. This one does have a substantial infiltration of macrophages. So, if we start thinking about co-targeting strategies, here we’re looking at co-expression on the same tumor cells or different tumor cells within that particular tumor mass. So, you can see that individually, if we look at CD46, about 80% of the cells would express CD46 as a target, about 80% or so would express KLK2, but that combination you would essentially target 100% of your tumor cells. Here’s an example, another complementation where maybe you want to have the targets co-expressed on the same cell. And in this case, KLK2 and STEAP1, almost 85% of your cells would co-express that particular target. If you look at the converse, what would complement a PSMA low tumor? In this case, we can see that either the druggable genome or cell surface targets, we see MUC1 be contra-expressed with PSMA. So, to confirm this, you can see immunohistochemistry for PSMA low tumors expressed very high levels of MUC1. And on the right is antibody drug conjugate exposure to patient-derived xenografts expressing high MUC1, where you can see essentially eradication of those PDXs. So, the second aim for this, just very briefly, is can we understand resistance mechanisms by looking at post-treated tumors? And of course, there may be many mechanisms. The most likely or possible is that you’ve lost your target. There may be intrinsic resistance or a lineage change. So, we now have some experience, 24 patients have undergone rapid tumor removal postmortem, and these patients have all been exposed to Pluvicto. So, you can see that post-Pluvicto, we see much more substantial heterogeneity for a given target, such as PSMA, compared to the pre-treatment. And particularly in the liver, you can see a comparison of pre-treated Pluvicto patients versus post-treatment, where you see substantial loss of PSMA in the liver. Here’s an instructive case. This is a patient that had PSMA PET avid lesions at screening. The post-PET cycle one SPECT scan, the post-PSMA Lutetium treatment SPECT scan shows a large number of lesions in the liver, and then by cycle two, the PSMA avid lesions are gone, but you still have the emergence of PSMA negative lesions. You can see from the tumor sample, large regions of PSMA null mixed with some PSMA low, these do express now new targets, MUC1 and STEAP1. Due to time, I can’t get into the circulating tumor cell assays. You’re gonna hear about this in the following talk in this session, so I’ll skip this due to time. Here’s Gavin Ha will be talking a bit more about ctDNA related to the project as well as Marina Sharifi So we’re currently looking at a number of different therapeutics trying to understand resistance mechanisms that may be driven by heterogeneity or potentially antigen loss. So, to summarize and conclude, I think what the atlas has really demonstrated is that there can be profound inter and intra-tumor phenotype heterogeneity. The phenotype variation is very distinct from underlying genomic features indicating some type of epigenetic driven plasticity. The phenotype clearly does associate with cell surface targets. Importantly, these double-negative tumors categorized by synaptophysin and expression but retaining AR signaling really look like ARPC, not neuroendocrine prostate cancer. And it’s very important to have a framework for standardizing the evaluation of metastatic biopsies. Fortunately, Michael Haffner and Himisha Beltran led a working group to define how best optimally to work up a metastatic biopsy. It’s a very simple approach, and this has been embraced by the prostate cancer working group 4, so I would refer you to this study. And then finally, I would say these multiplexing methods are quite robust, and it may be quite useful to advance these into clinical diagnostics. We hope to have this atlas posted so that all of you can explore this for evaluating clinical targets. And the results do suggest that the future may require some type of combinatorial targeting. Finally, the Pluvicto resistance is multifactorial, including target loss as well as heterogeneity. And our next steps are to look more broadly in other therapeutics to assess resistance. So finally, I really want to thank again Felix for his inspiration, motivation, and certainly innovation, the patients that participated, certainly the Prostate Cancer Foundation, and then our wonderful academic and industry partners. So, I’ll stop there. Thanks.
Cora N. Sternberg, MD [00:22:18] Hi, Cora Sternberg. Thank you for an absolutely excellent talk, which really makes us think a lot of things. What do you see as the role of AI to help us in thinking about whether we should be using combination treatments or sequential treatments, not waiting to see the autopsy, but while patients are alive. Do you have any ideas on that?
Peter Nelson, MD [00:22:40] I think I’m probably not qualified to think about the AI aspect. Many others, I think we’ll probably do a better job of that. Certainly, the AR would probably help with the image analysis at least from the metastatic biopsies that maybe can help us infer even deeper aspects of heterogeneity or combining some aspect of a surface target with an intrinsic tumor feature. That’s something we haven’t explored at all. So, does underlying genomics on top of whatever your therapeutic target going to have some combinatorially influence on directing therapy? So great question. Thank you.
Himisha Beltran, MD [00:23:18] Sorry, just one more question.
[00:23:20] Oh me, me, me.
Himisha Beltran, MD [00:23:22] Okay, two more.
Unknown [00:23:27] Just a very quick question, Pete.
Peter Nelson, MD [00:23:30] Sorry, yeah.
Unknown [00:23:30] No, this was spectacular in the sense that it really helps dissect you know, what we’re looking at and trying to figure out how we’re gonna do different combinations for treatment. You know, cancer does not live by cells alone. You need something else to have those cells dealing with, you know, they just don’t float in the air. Stroma and the contributions of stroma. Was the group able to detect whether there were differences in stromal distribution, markers, MMP-9 levels, for example, all of those are the matrices or the platforms as we like to call them that, you know, hold the cells. So, were there variations that were observed?
Peter Nelson, MD [00:24:06] It’s a fantastic question. That’s an untouched I think aspect of the atlas that certainly absolutely can be explored because we are both for the spatial keeping the tissue intact so you can look at the underlying stroma in different organs and tissues, and the single cell can get you as well. But to be honest, we have not yet looked at that at all, but I think it’ll be very important.
Unknown [00:24:27] Thank you.
Peter Nelson, MD [00:24:29] That’s it?
Himisha Beltran, MD [00:24:33] Thank you, Pete, for that exciting talk. Clearly a lot of questions. Please grab him later. Our final speaker for this session is Dr. Ed Kwan. He is a clinician scientist at Monash University in Melbourne, and he did his postdoc with Alex Wyatt at the University of British Columbia and Vancouver Prostate Center. And he’ll be talking about ctDNA analysis in the therapy trial.
Edmond Kwan, MBBS, PhD, FRACP [00:25:07] Morning everyone, I’d just like to say I’m absolutely honored to be presenting data that was performed during my time in Alex’s lab during my postdoctoral fellowship. I’m also here in the capacity of the ANZUP Australian Cancer Trials Group, the cooperative group. So, I’d like to extend also my extreme gratitude to the Prostate Cancer Foundation, which none of this work would have been possible without a PCF Challenge Award in 2023. So, I think this morning you’ve heard from some really great speakers about how a single tube of blood can really push the boundaries of what we can uncover in terms of tumor biology. But I’m gonna take you back over the next 15 minutes to sort of a good old-fashioned sort of mutational calling and copy number profiling and structural variance, sort of the early generation of liquid biopsy technology. So, this group here really needs very little introduction, but the therapy study design, just a reminder, took patients with progressive metastatic castrate resistant prostate cancer who all had had docetaxel, and a high majority had had at least one AR pathway inhibitor, over 90%, and randomized to 6 cycles of Lutetium PSMA-617 versus 10 cycles of Cabazitaxel. And it was sort of courageously led by three professors, so Professor Michael Hofman, Ian Davis, and Arun Azad. And just to go through the primary results, as a reminder, what we see is that Lutetium PSMA-617 led to way better, significantly elevated sort of PSA response outcomes as well as an objective response rate and prolonged progression free survival, and this was much better tolerated compared to cabazitaxel. But overall survival in this context, as we look at these results today, was similar across both arms. So, but one could argue that therapy really wasn’t ever powered to look into this particular endpoint. So, here’s a summary of the ctDNA correlative analysis, and we have 180 patients that graciously donated both baseline and progression plasma samples, and we really came at this asking two questions. The first is can ctDNA abundance, and you’ve heard a lot about ctDNA fraction today, predict treatment outcomes, and weaved within that is genomic alteration status. The other thing we were interested in using this cohort is what are the genomic mechanisms of adaptive treatment resistance? I want to remind you as we look at these role results today that there is heavy preselection in this particular study using dual trace of PSMA as well as FDG PET. And while that allows us to look at a number of different analyzes, it isn’t rich for high PSMA expressing disease. So, this is a very late-line setting disease. So, as I mentioned, nearly all patients had at least an ARPI as well as docetaxel. And this is one of the highest ctDNA fraction cohorts that have been extensively profiled with cell free DNA. So, the median ctDNA fractions was well above 20%. And when we look at patients who were ctDNA positive, it’s closer to 30%. So, to give some context about how high that is, in the historical setting, there are cohorts that are in the treatment naive ARP MCRPC setting, which are less than 5%. And even in poor wrist disease, we see it’s around about 10%. So, as I mentioned, we had dual tracer technology in terms of looking at both PSMA PET as well as FDG PET. And we first looked at how ctDNA fraction sort of correlates with these imaging findings. And expectedly we see that patients with highly proliferate disease in the context of FDG metabolic tumor volume, they’re strongly correlated with ctDNA fraction. What was somewhat unexpected is this inverse relationship that we see, where patients with high ctDNA fraction actually had lower PSMA SUVmean. And this may suggest some features that we can discuss later about PSMA heterogeneity within the tumor. So, while we’re looking at ctDNA fraction in the context of outcomes, that’s probably one of the main important points. I want to point out the work of Dr. Nicolette Fonseca, who’s a PCF YI from several years ago. And in a very large cohort published in Nature Comms last year, she showed that ctDNA fraction was extremely prognostic. But ctDNA fraction has never been shown to guide treatment selection. So, when we were first looking at the therapy cohort, we were quite awestruck by this observation here. So, I’ll ask you to first cast your eyes at the bottom. We can see that in high ctDNA fraction, suggesting highly proliferate disease, high tumor burden, we can see that outcomes are similar irrespective of the treatment arm administered. However, as you move your eyes up forward, we can see that it’s actually patients with lower ctDNA fraction that seem to benefit much greater from Lutetium PSMA in the context of biochemical responses. This biochemical response superiority with Lutetium did translate to progression free survival, and most stark there is that green line on the on the Lutetium curve. Much better outcomes compared to use of cabazitaxel. So, we can see it has a ratio of 0.2, strongly significant. And when we put it in a multivariate analysis, taking into account one of the other strongly prognostic features, which is PSMA SUVmean high, so greater than 10 from data from James Buteau and colleagues, we can see that also that slow ctDNA fraction holds true. Adding this on, so taking just patients with high PSMA SUVmean, we can see that the detection of ctDNA fraction actually further stratifies Lutetium PSMA outcomes. Expectantly, we see that ctDNA fraction from terms of overall survival is still strongly stratified for survival outcomes in both arms. So, I think the first key message is that ctDNA fraction is a candidate predictive biomarker for differential response to these two drugs. But this is in the context of very molecularly selected patients who’ve progressed after docetaxel. And I wanted for those who were unable to attend ESMO this year, whether or not this applies in the docetaxel eligible setting remains to be seen. And our group is looking at the PR 21 study in terms of this is in the first line post ARPI setting, randomized to Lutetium or docetaxel, we can see that was no difference in RPFS, but longer overall survival in the docetaxel, suggesting that you know it’s treatment sequencing may be important in this context in terms of patient performance status. So, the Vancouver group, I just want to switch gears here. The Vancouver group has long been really interested in the use of targeted panel sequencing. And this high ctDNA fraction within the therapy cohort really allowed us to deeply dissect sort of key genes that are important prostate cancer drivers so most of our panels have this combination of both sparse probe coverage that gives an almost analogous to a low-pass whole genome sequencing, but also in very select genes, that’s denser coverage that allows us to get very focal changes with regards to copy number deletions as well as structural variants. And we did this systematically and exhaustively over a number of key prostate cancer drivers. And for the sake of time, I’ll push on. So, I want to first focus on PTEN alterations. We can see that when comparing Lutetium PSMA versus cabazitaxel, the presence of a PTEN alteration did confer greater benefit to Lutetium relatively cabazitaxel. And this was slightly more pronounced in patients who were predicted to have PTEN null status. And of course, we were pleased to see that this also translates to overall survival advantage as well in the course of about four to five months. So, what those particular forest plots don’t necessarily tell us is that what’s necessarily driving this? Is it that the alteration itself or is it the treatment that may be a mitigating factor? And these series of Kaplan Meiers show that actually it’s the differential outcomes are not so much driven by these alterations and the impact in Lutetium patients, but by extremely poor outcomes when PTEN altered patients are treated with cabazitaxel, and this is progression-free survival curve, as well as the overall survival. And as mentioned, these associations are more pronounced in the PTEN null group. Focusing on other key prostate cancer drivers, so neither TP53 or AR alteration status have provided any predictive utility in this context. And there’s a lot to unpack, but we’re happy to answer questions afterwards. And finally, from a genomic alteration status, we were really interested in DNA repair alterations. And the reason for this is quite obvious, but in the context of their role in sort of repairing single strand DNA and double strand DNA breaks. And what we can see in this albeit a smaller cohort, is that we see that there are certain DNA repair alterations that may confer sensitivity to Lutetium PSMA, one of them may be ATM. Not so much in bracket two. And the converse relationship may be present in CDK12. But I think what we need to do is that we need to amalgamate more larger cohorts to investigate this more thoroughly. So, as we sort of move forward, we have a number of DNA biomarkers at our arm’s length. And what we start to ask is can we add ctDNA fraction, PTEN alterations as well as ATM alterations as Lutetium PSMA617 biomarkers? And hopefully this expands the scope of relevant biomarkers to help us sequence treatments. Very briefly about the resistance alterations. We came at this at several angles. So, looking at these paired baseline and progression samples, we took a very sort of close look at whether or not the alterations that change over time are driven by mutations or copy number alterations, whether that be genome-wide or specifically focus the androgen receptor. And the long and short of it is that neither Lutetium or cabazitaxel really sort of reshaped the mutational landscape. And this is somewhat expected that a lot of these alterations are truncal alterations that are quite fixed throughout the patient’s disease course, unlike some of the epigenomic changes that other presenters showed this morning. We did see some new alterations in key tumor suppressor genes, but this was occurring in both arms. And importantly, we didn’t see any emergent full H1 mutations to suggest a gatekeeper mechanism of resistance. And this, as mentioned, no major changes in the genome-wide copy number profiles as well as in the AR locus. So that concludes sort of that key message that key evidence of recurrent acquired genomic alterations don’t appear to be genomically driven. So, in the last couple of minutes of my presentation, I want to highlight work. So, if genomic alterations aren’t changing in the setting of adaptive resistance, what is changing? So, I want to touch on clonal hematopoiesis, and Professor Michael Hofman touched on this before, and this is work from one of our PhD students, very highly talented PhD students in our lab, which I’ll introduce in a sec. But just a reminder that clonal hematopoiesis are these somatic mutations and in hematopoietic stem cells, and these mutations are very common in the prostate cancer population group, which is generally older, and these can expand and confer a selection advantage. And why investigate this in the context of Lutetium? I think that these agents induce single strand breaks. And the question is could beta-emitter therapy really precipitate this development of these potentially pre-malignant mutations, clinical hematopoiesis in specific genes can be a precursor for therapy related myeloid neoplasms. So, work led by PhD student Asli Munzur and was presented at ASCO this year, leveraging this particular cohort. And what she did is sort of very deep targeted sequencing using a panel that is enriched for chip alteration, genes that are enriched for chip alterations. And this is really important. And what she was able to find is that at baseline, chip alterations are very common between arms, well over 70%. But comparing at progression, she showed a threefold increase in these treatment emergent CH variants and with Lutetium, which was not evident with cabazitaxel treatment. And this is most pronounced in the conventional DNMT3, TET2 and ASXL1 genes, but also important in PPM1D, which has been in the past strongly associated with the development of therapy related neoplasms. You know, I found this data extremely striking as well. What she, Asli, showed is that in patients with preexisting chip alterations at baseline, Lutetium seemed to preferentially expand these mutant clones over time in a way that really wasn’t evident within the cabazitaxel arm. So, taking this together, these treatment-emergent clonal hematopoiesis mutations are likely to gain a little bit more relevance as we move radioligands into earlier disease phase. And this is the case in that recently presented by Scott Togawa at ESMO, a PSMAddition study, and they’ve taken extensive exploratory biomarkers to investigate this question. So finally, you know, where to next? And I think you’ve heard from a number of fantastic speakers today about the sort of information you can garner from cell-free DNA. And there’s a lot of teams, including ourselves, working on epigenetic characterization, and we’re partnering for this therapy cohort with Himisha Beltran’s lab and Francesca Demichelis about looking at the NEMO assay and then also Martin Sjöström for 5-hmC analysis. And this is work that’s going to be also led by PCF YI Sofie Tolmeijer in the class of 2022, where we start to integrate these different cohorts that are at the Vancouver lab, which really gives us an ability to sort of validate some of the findings and explore new ground. So, in conclusion, I think these results really nominate some candidate biomarkers that do need to start to be validated in bigger medic cohorts. And what in terms of resistance, I think we need to start looking at ways to focus outside genomic sort of pathways of resistance, whether that be epigenomic, phenotypic, or through the tumor microenvironment. We need to start looking at longitudinal monitoring of CH variants, especially as we move these agents in the early disease space. And finally, we hope that this serves as a bit of a framework about how ctDNA biomarkers can start to be integrated in the context of PSMA targeting therapeutics. I’d like to thank these members of the Wyatt Lab, which none of this would be possible. This data was published in Nature Medicine early this year, so I invite you to take a look at that. But all the members of this team were really instrumental in taking this forward. And finally, I’d like to thank the Prostate Cancer Foundation as well as other funders for supporting the study. Thank you very much.
Gert Attard, MD, PhD [00:40:48] That’s great, Ed. Do you want to venture a hypothesis why high ctDNA do not benefit from Lutetium?
Edmond Kwan, MBBS, PhD, FRACP [00:40:56] Yeah. So, the way I look at that, Gert, is that I think that it’s both the in that high ctDNA fraction group it’s both arms are not seemed to be benefiting. I wonder if we flip it around and say the low ctDNA infraction group, do you need the tumor to be sort of proliferating at a certain rate to really be benefiting from the taxanes? You know and does there need to be almost like a certain amount of mitotic turnover to really benefit from these cytotoxic agents. So, I think the way I’ve been seeing it is that it’s really in that high tumor fraction group, it really doesn’t matter too much about the agent that is choosing. You know, outcomes seem to be poorer in in both treatments.
Gert Attard, MD, PhD [00:41:43] Thank you.
Edmond Kwan, MBBS, PhD, FRACP [00:41:43] Thanks, Gert.
Himisha Beltran, MD [00:41:48] Thank you Ed and thank you to all speakers. This was a fantastic session. What did you think?
Alexander Wyatt, PhD [00:41:53] Yep, fantastic. Absolutely. Great organization. Goodbye, everyone.