Ashish Kamat: Hello everybody, and welcome to UroToday's Bladder Cancer Center of Excellence. I'm Ashish Kamat, urologic oncologist in Houston, Texas. And joining me today is someone who really needs no introduction to this forum. Professor David McConkey, who really has done a lot of work in bladder cancer, predictive biomarkers, biomarkers in general, all the way back from the early 2000s, when he and I interacted when he was my mentor when I was a fellow in his lab. So, David, you've given many of these presentations. This was one of the keynotes that you gave at IBCN, but of course, you've covered this topic so many times. So it's always great to have you join us and educate me and our listeners on where we are with predictive biomarkers. So please take it away.

David McConkey: Thank you, Ashish, and welcome everyone. It's great to be here. I will do anything for Ashish ever. He's one of my favorite people. And so the topic today is about predictive biomarkers, and these are my disclosures. So, just starting with a couple of definitions. So prognostic biomarkers are those that, "Risk stratify patients independent of therapy," whereas predictive biomarkers, what we're really trying to develop in my group and with our collaborators, are used to, "Predict response to or benefit from therapy." And the best predictive biomarkers, in my opinion, "Would enable us to make accurate predictions at the level of individual patients." So in 2014, our group at MD Anderson, also groups at UNC and at TCGA, reported the identification of the basal and luminal molecular subtypes of bladder cancer. And our group linked them to sensitivity and resistance to cisplatin-based combination chemotherapy.

In parallel work, groups at Dana-Farber and Memorial Sloan Kettering, led by Jonathan Rosenberg and Eli Van Allen, Gopa Iyer, as well as a group led by Betsy Plimack at Fox Chase, were using bulk DNA sequencing to identify gene mutations that were associated with response to cisplatin. And back in 2011, we had a clinical trials planning meeting, was held at MD Anderson, where we discussed what we could do to take advantage of the neoadjuvant platform in bladder cancer, namely that we have relatively easy access to tissue before and after therapy. And we decided that the best thing to do would be to validate an algorithm that Dan Theodorescu's group had discovered called CoXEN, which was based on RNA expression profiling. And we designed a clinical trial called S1314. We called it the CoXEN trial that we wanted to use to validate these observations. And unfortunately, after years of working on this, the results did not validate the original observations. I would say that the molecular subtypes remain prognostic, namely that basal tumors were associated with worse outcomes, and DNA damage-and-repair mutations enriched for cisplatin sensitivity. In fact, they probably are predictive biomarkers.

They're just not as robust as we'd like them to be. So why didn't they work better? Well, we now know based on more sophisticated studies that have been done on single-cell level, that tumors contain cells belonging to many, if not all, the molecular subtypes, and they also can contain different mutations in them. So bulk measurements don't capture this heterogeneity, and this heterogeneity is probably part of what was undermining our attempts to develop these biomarkers based on them. And then the second that a lot of us are working on now is plasticity. And we now know that with molecular subtypes in particular, probably other biological states, cancers can switch from one subtype to another. And so under pressure of therapy, it's hard to predict where the ultimate molecular subtype is going to land. A lot of us believe that we can improve on this precision by using artificial intelligence. And there's some pretty high-profile work led by [inaudible 00:04:38] that was published on this same clinical trial. But in addition, I think that all of us are very enthusiastic about liquid biopsies, which enable us to actually monitor the burden throughout the course of therapy, through longitudinal collections of blood and urine.

And maybe by leveraging all of these technologies at once, we'll be able to identify some really good strategies to predict response to therapy. So just to remind you of a couple of these liquid biopsy studies. This was led by Lars Dyrskjøt's group, published in JCO, where he demonstrated that measuring circulating tumor DNA was prognostic and predictive in patients treated with neoadjuvant chemotherapy. And a second high-profile study led by Tom Powles, which demonstrated that the ctDNA levels in patients after cystectomy were associated with benefit from atezolizumab. In other words, patients who did not have any ctDNA after surgery received no added benefit from atezolizumab, whereas those that did actually received significant benefit from atezolizumab. And for those of you who went to ESMO, you heard about the IMvigor011 study, which was positive and published in New England Journal of Medicine, which reinforced this conclusion that ctDNA can be used to identify patients who benefit from adjuvant therapy, and also, by inference, those who would not.

So we've been working together with a number of individuals, working in collaboration with Convergent Genomics, and there are other companies working on these platforms to do the same kinds of experiments using urine to measure tumor DNA. And not surprisingly, some of the same patterns emerge. We see prognostic value in urine tumor DNA pretreatment. And if we monitor the change in urine tumor DNA before and after treatment, we can actually see which patients are benefiting from, in this case, atezolizumab. So the take-home message is that we think that longitudinal measurements of tumor DNA in blood and urine can be used to monitor response to therapy. And so, by combining them with analyses performed on tissue sections, and presumably also ultimately by imaging strategies, will help us yield more accurate predictive biomarkers. So while I'm disappointed that the bulk measurements didn't work, I'm actually extremely positive and encouraged by the fact that the liquid biopsies seem to be telling us so much about what's happening systemically in patients. And so with that, I'd be happy to talk further.

Ashish Kamat: Thanks so much, David. When we talk about liquid biopsies, obviously we have blood, and then we have other liquid, and you mentioned urinary tumor DNA. Could you share with me your thoughts a little bit about the urine assays, right? I mean, obviously, we are looking at utDNA in the urine, but we also have older technology, but reliable technologies such as cytology. Have you actually seen that utDNA will help us drive the field forward ahead of these older assays, such as cytology or FISH? And if so, in which selected patients do you think it would actually make a big difference?

David McConkey: So I think that where we're seeing the most value right now, Ashish, is in patients with non-muscle-invasive bladder cancer who have CIS. And in those patients, cytology isn't great. It's not very sensitive. The urine tumor DNA assays seem to almost work better in patients with CIS for probably biological reasons. The other opportunity, I think, to make rapid impact is in upper tract urothelial cancers, where again, the biopsies are tricky, the cytology isn't great. We're starting to see pretty strong indications that in those two indications, the urine tumor DNA measurements are going to certainly outpace cytology. Now, with regard to the other measurements, I think that a lot of them are very good. I think that the difference between some of the other urine-based measurements, including usually a panel of either RNA-based or DNA-based biomarkers.

You mentioned FISH, which I know you've done quite a bit of work on yourself. Those are all great assays, but they don't tell us quite as much about the tumor biology. So, for example, we're seeing interesting prognostic implications in tumors that have specific patterns of mutation. And I think also we're at a point now where we can say that the urine tumor DNA measurements that we're making now with the next-gen sequencing platforms are more sensitive than these other assays. There are other companies entering this fray, and some of them are very famous for developing very bespoke assays that we're now using as standard of care in patients. It's going to be very interesting to see how those assays compare with the ones we're currently using, which are mostly panel-based. But the potential for these assays to get more sensitive is, I think, much higher with these genomic tests than they are with the assays that we're used to, both by the breadth of the analyses and also kind of by the bespoke nature of some of these tests.

Ashish Kamat: Yeah. No, I think you're right on target. I think when it comes to patients that have existing tumor where we're trying to see how we can personalize the therapy for that particular tumor, having utDNA with everything that we can identify that's bespoke to that tumor is clearly critical. I also think that it's going to be very relevant in patients that have, say, muscle-invasive disease that have had neoadjuvant therapy, and now we're trying to accurately define clinical complete response, right. This was something that the IBCG tackled, along with the San Raffaele Group in Milan. And I think utDNA, at least in our recommendation, was something to consider because we're actually trying to make sure that that clinical complete response truly means something for that patient. But when we're looking now at patients that have had tumor in the past, for example, and are now coming for surveillance, and you're detecting the new tumors. You've done a lot of work in this. Share with us a little bit as to your hypothesis as to new tumors, new clones, new tumors? How would they correlate with an assay that's based on a past event? Would they? Would they not? What's the significance of that?

David McConkey: So, Ashish, you put your finger on one of the concerns we have about the bespoke assays. So, because of field cancerization, which our colleague Bogdan Czerniak and others have been working on for a number of years, we're concerned about multifocal disease, about new subclones arising that will be somewhat different from the original tumors that we were studying. I think there's going to be some research that needs to be done with these newer technologies to tell us exactly how often we see a completely new clone emerge from a patient who's been treated successfully for a previous clone.

But this is, I think, one of the reasons for why some of us are actually quite positive and enthusiastic about these panel assays, because you can design a panel that covers essentially all of the mutations we know about in bladder cancer cost-effectively. And you can also do low-pass whole-genome sequencing to get aneuploidy. And by doing that, you can tell cancer, non-cancer pretty accurately without knowing exactly what the mutations were that gave rise to the tumor. So again, I think it's early days. It's a bit of the wild west out there with all these tests. We have companies making tests. We have academic centers making tests. Rik Bryan at the IBCN meeting shared a test that they're using in the UK. We just have to wait to see which one ends up being the front-runner here.

Ashish Kamat: Yeah. No, I think these tests that are popping up all over the world are a little bit concerning in many ways, right. Because I was, of course, in India, and I was in South America, and people were talking about, "Oh yeah, we use ctDNA all the time," but it wasn't the ctDNA that we've all come to recognize and that has been validated in clinical trials, for example, IMvigor011 that you just mentioned. So let me switch gears a little bit away from the urine, which, as you acknowledge, is fairly noisy. And it's an evolving field, and we clearly need to wait and see what the data shows. I really think when it comes to circulating tumor DNA, we now have a preponderance of evidence to suggest that this should be incorporated into standard care pathways, maybe with some trials still needing to read out. But in general, it's something that I use in my patients. I know many of my colleagues use. But we're using, again, tumor-informed assays, right. What's your sense, having done this for all these years? Is there a role for tumor-agnostic circulating tumor DNA? If so, where would that be? And if you don't think there's any role, of course, let me not ask you too many questions.

David McConkey: No, I think as these technologies evolve, the rules continue to change. So the reason we like the bespoke assays so much is that they've been prospectively validated, they're very sensitive and specific. They avoid a lot of the problems associated with things like field cancerization and clonal hematopoiesis, but they don't tell us much about how to treat the patient. It's, for example, very informative to know what the ctDNA is telling us about FGFR3 mutations, which a number of groups are showing can be different in metastatic lesions compared to local tumor. And in many cases, what happens is that new mutations appear in the blood that weren't seen originally in the original tumor, and those patients, therefore, would be eligible for erdafitinib.

So I think that's just one example, and there will be others, I'm sure, of how these panels will be able to tell us something about therapeutic targets. And maybe ATM and ERCC2, which were two of the DDR mutations, will turn out to be important for certain therapies that we're using right now. We still are using therapies that damage DNA in some cases, as now they're just being used as ADCs. So I think that there's a complementary role for these panels in patients, especially metastatic disease. But I think at this point there's very strong evidence that the current bespoke panels are much more accurate in terms of measuring minimal residual disease, which is what IMvigor011 did to change clinical practice, in my opinion.

Ashish Kamat: And again, you know that here at MD Anderson, Scott Kopetz and their group have the INTERCEPT Program that has really been sort of a multipronged MRD assay. And we in GU have also launched what we call the Guardian Program, which incorporates circulating tumor DNA, utDNA, AI, machine learning, everything. Essentially welcomes all the new technology to sort of come in so we can then see what works best. But I really think that what you mentioned about we should be excited, but we still need to wait for the data to mature before we make it standard of care is a critical cautionary point because we really... neither you nor I are politicians. We don't set cost. But at the same time, there are cost access issues, right. Here in North America, we are a little bit spoiled, but I am a global citizen, and I see many places in the world where patients are paying out of pocket for tests that may not actually be currently clinically validated. So David, as always, I really want to compliment you on all the work that you're doing, but also for being a very reasonable voice and not just coming on the bandwagon and saying, "Do it because I say so." So thank you for the time and thank you for your wisdom.

David McConkey: Always good to see you. Thank you, Ashish.