DNA Based Markers in High-risk Non-Muscle Invasive BCG Unresponsive Bladder Cancer - Alexander Wyatt

November 15, 2020

Alexander Wyatt from the University of British Columbia joins Ashish Kamat in a discussion on the genomic correlates of Bacillus Calmette-Guerin (BCG) responsive and relapsed non-muscle invasive bladder cancer. Dr. Wyatt shares recent study findings exploring links between common genomic driver alternations and BCG unresponsiveness and also data looking at genomic variables in the context of initial non-muscle-invasive bladder cancer and what they can tell us about clinical outcomes, while also looking at profiles upon relapse and how they differ from early tumors.

Biographies:

Alexander Wyatt, Ph.D., BSc, Assistant Professor, Department of Urologic Sciences, University of British Columbia, Senior Research Scientist, Vancouver Prostate Centre

Ashish Kamat, MD, MBBS, is a Professor of Urology and Cancer Research and Wayne B. Duddleston Professor of Cancer Research at MD Anderson Cancer Center in Houston, Texas. Dr. Kamat serves as President of International Bladder Cancer Group, (IBCG), and Co-President of International Bladder Cancer Network..


Read the Full Video Transcript

Ashish Kamat: Welcome everyone to UroToday's bladder cancer center of excellence. I'm Ashish Kamat from MD Anderson Cancer Center in Houston and it's a pleasure to welcome Dr. Alexander Wyatt today who's joining us from Vancouver, British Columbia. And Dr. Wyatt has done a lot of work in the genomics of bladder cancer in general. He's an Assistant Professor at the University of British Columbia and at the Vancouver Prostate Center. And today's going to talk about specifically the genomic correlates of BCG responsive and relapsed non-muscle invasive bladder cancer. Alex, take it away.

Alexander Wyatt: Thanks Ashish. And thanks UroToday for hosting this presentation and discussion. As you mentioned, I'm going to touch on some unpublished work from our research group that's sort of trying to look at whether genomic variables in the context of initial non-muscle invasive bladder cancer can tell us much about clinical outcomes and also looking at profiles upon relapse and how they differ to those early tumors.

I think actually, thanks to work from several different groups and consortia over the last decade, we have a pretty good understanding now of the genomic landscape of non-muscle invasive bladder cancer. And I'm showing you one encore prints here. This one is from Memorial Sloan Kettering Group using a targeted sequencing panel. But each cohort seems to tend to show approximately the same types of trends. So you see pervasive mutations in the TERT promoter. You see frequent alterations to chromatin modifiers, such as KDF6A. You see fairly frequent TP53 mutations and alterations in general, although not as high as in the muscle-invasive or the metastatic setting. And you see very common alterations to RTK genes, HER2 is frequently activated and mutually exclusive with FGFR3 hotspot mutations or rearrangements.

And there have been some hints over the last sort of decade or two of study that there is a link between the genomic variables here, these driver gene alterations and outcomes and prognosis in general in these patients with non-muscle invasive disease. I think perhaps one of the most promising ones is the concept... is the fact that ARID1A mutations appear to be associated with a higher risk of recurrence and specifically with recurrence with non-muscle invasive disease rather than progression to muscle-invasive disease. However, given the way that the transcript's academic field has moved forward, looking at signatures and patterns in data rather than relying on individual genes, I think probably the genomics field needs to also move in this direction too.

And so that's beginning to happen already. And I want to highlight a paper published just this week in Cancer Research from Bellmunt and colleagues with the Broad Institute involved as well. And they looked at 62 patients with high-grade T1 disease and performed whole-exome sequencing to look at these genomic variables and put them in context of clinical outcomes in patients with non-muscle invasive disease, predominantly initiating BCG. And they found that in fact, tumor mutational burden, ERCC2 mutations, and a few other sort of broader changes were associated with good outcomes. And we're actually able to propose a model that is preliminary in nature at the moment and clearly requires sort of testing and follow-up and validation but a model whereby certain patterns of genomic alterations might be able to classify a tumor into low or high risk and therefore be of potential use when decisions are being made early on.

And alterations that we might tend to associate with muscle-invasive disease, such as TP53 alterations or cell cycle drivers, perhaps those are the types of things where we need to think about intensive monitoring and even more radical approaches. So I think some encouraging signal coming through these types of studies that suggest that genomics may be able to help us guide management of non-muscle invasive disease.

In our study, we wanted to kind of do so actually something quite similar to the Cancer Research paper. I just highlighted and tried to explore the links between common genomic driver events and BCG response. But we also wanted to look at the post-treated tumors from the same patients in a standardized cohort as possible and try to understand the extent of genomic differences pre and post-treatment. And the reason for doing this is there has been some exploratory work in this kind of area, looking at metachronous pairs of tumors. And they'd suggested that by enlarge when you see a recurrence or progression, there is a common ancestral clone ie, the tumor is related perhaps distantly but still related to that original lesion and to find a truly independent cancer is probably very rare. However, these studies pointed to remarkable variability in the types of changes that occurs. We really wanted to try to explore this in a systematic manner.

So we collaborated in sort of equal partnership with the Universitatsspital Basel and Cyril Rentsch is the team leader within this collaboration. And we looked at 83 patients with non-muscle invasive bladder cancer, providing a TUR sample and initiating BCG. And among the patients defined as responders, we had a follow-up of 44 months. And the overall cohort was not really enriched with high-risk clinical features. It wasn't obviously as pure high-grade T1 disease as per the Bellmunt study but we suddenly had a lot of patients that had high-risk features at diagnosis. And 17 patients had recurred with non-muscle invasive disease and 14 patients have progressed to muscle-invasive disease. And we'd also had these tumors available for study.

And so a PhD student in my laboratory, Jack Bacon has subjected all of the baseline TUR samples and all of the follow-up samples, whether they're non-muscle-invasive lesions or muscle-invasive lesions to targeted DNA sequencing using a panel that captures those genes that we know already know to be associated with bladder cancer. So we're not able to do a de novo discovery for new genes, so forth but this is telling us about those established drivers of disease. And we're also doing an ongoing follow-up studies, broader profilings to try to understand genomic evolution, selecting pairs that show quite diverse profiles over time.

But prior to treatment, we can see actually probably a very reassuring landscape. And I think whenever you study a new cohort is important to look at a landscape like this and make sure that it relates quite closely to what others have seen. It can tell you about whether your cohort is representative of the general population or not. And I think here we see that again, frequent chromatin modifiers altered, common P53 mutation consistent with the high-risk clinical features of this cohort. And again, you can see FGFR3 hotspot mutations and HER2 activating alterations in many patients.

And interestingly we did observe hypermutation several tumors that went on to be long-term responders after BCG. And this is again, consistent with data from the Bellmunt study. So I think probably we're seeing a real association here. We also were able to confirm the result from Memorial Sloan Kettering Group, suggesting that ARID1A mutations are associated with recurrence and potentially skew towards non-muscle-invasive recurrence rather than muscle-invasive progression. So what we're now doing here is trying to see if we can rationally combine certain sets of alterations to try to define more of a polygenic score and try to move away from individual gene alterations. That work is ongoing.

So when we started to look at pre and post-treated pairs of tumors from the same patient, there's a number of different ways we can compare genomic alterations within these driver genes to try to understand what changes over time. And one simplistic thing to look at is to look at the presence or absence of a mutation. And so that's what this plot is showing. So every time you see something green, it means that that alteration was detected in both pre-treatment and post-treatment. As a by large you can see in these patterns suggesting that every patient which is represented by this column here has some shared alterations. And in fact, when we take into account copy number changes and broad structural rearrangements affecting chromosomes, we can see that in fact, every patient has shared events between metachronous pairs. And so that tells us that in all cases, the second tumor that we study is related in some way to the first tumor. It doesn't necessarily say that there's a direct descendant but they certainly have a common ancestor.

However, I would caution that this type of analysis is probably an oversimplification and there are many genomic variables that are not simple binary variables or at least not as categorical as we like to think of them. And I think these type of nuances have relevance across biomarker development and in cross sort of genomic clonal studies in general. So the allelic status of a mutation is actually fairly important. If we're looking at a tumor suppressor loss, is there one hit or two hits? If we're looking at an oncogene activation, is there actually imbalance across that using... Have you got extra copies of the mutation? When we're looking at copy number, a five copy level amplification is probably not the same biologically as a 50 copy number amplification. And so we shouldn't necessarily categorize those in the same camp. I think many people are familiar now and adjusting with the concept that if we're thinking about drivers of tumors, they need to be truncal. They need to be clonal rather than something that's only present in a minority of tumor cells.

But the final variable is about broader genomic context for any individual alteration. And again, I think this is beginning to be recognized, but in an example would be, if you have an FGFR3 hotspot mutation, we probably shouldn't be considering that in isolation because if that arises in the context of a tumor, which has a very high mutation count and a lot of other oncogenic alterations, then it's possible that the FGFR3 mutation is not a non-redundant driver in that case. And potentially that tumor cell has multiple escape routes from therapy. And so that might be quite different to a patient or tumor that has an FGFR3 mutation with very little else going on. That particular tumor might be more dependent on that mutation. So thinking about context around the individual alterations is going to be important.

So with that being said, I want to just show in the last couple of slides, a few examples of some of the nuances here and how we're trying to tackle them. So when we look over time, obviously we need to think about not just mutations but also copy number changes and these broader, broader events. But I'm going to show you how a focus on mutations, just as representative of the way we're looking at this. So these scatterplots on the left are showing two patients with pairs of tumors from each patient.

So on the x-axis is the pre-treatment non-muscle-invasive lesion. And on the y-axis is the post-treatment lesion, whether that's a non-muscle invasive recurrence or it's a muscle-invasive progression. And each individual dot here, each individual dot plotted is representing a driver mutation. So let's say an FGFR3 hotspot mutation or a P53 mutation. And the axis are the cancer cell fraction of that prep... that mutation is present. And so something near the top right is present in almost all cancer cells within that sum. And so the fact that in both of these boxes the dots line up along the diagonal is telling us that by enlarge, the same mutations present in the same hierarchy in both lesions. And so from a genomic perspective, if we saw this kind of pattern across all types of patients, we'd say these are genomically very similar tumors. However, I would say that these are actually the exception rather than the rule. And more often we see the types of plot shown in the middle and on the right whereby yes, there are some shared events. For example, in the middle-upper panel, you can see a shared cloud in blue and gray, but they are clouds of driver alterations that are present only pre-treatment and only post-treatment.

And so this is telling us that actually there's a remarkable level of change in some of these tumors over time. And this is in quite a stark contrast to say later in disease in the muscle-invasive or even the metastatic state where I think a lot of genomic lesions tend to get fixed. So this is talking to a high level of fluidity in this early stage. But what we can begin to do is look at the types of alterations that are enriched in certain parts of these graphs. And so in the top right-hand side here, you can see in this particular case, some alterations lining up where we would hypothesize them to. And so an FGFR3 mutation and a PIK3CA mutation that are restricted into the non-muscle invasive lesion, but at progression and this individual progressed with muscle-invasive disease. Those alterations have been lost and instead replaced with P53 mutations and BAP1 mutations.

So we can start lining these things up and try to at least have some insight into the biology that's going on here. So when we pile up TERT promoter alterations, we can see they line up down the diagonal. So that's telling us that they're shared between all pre and post-treatment lesions. These are probably early events and that fits with our understanding of TERT promoter mutations in bladder cancer as they're associated with that early tumorigenesis. Whereas for TP53 mutations, the dogma is that these kinds of alterations arise later and are associated with progression to more aggressive disease. And again, that's what we see here. These alterations enrich post-treatment.

But perhaps notes of alarm almost is that some of the alterations that we consider actionable such as FGFR3 mutation status, E0BB2 amplification mutation status tend to be very variable over time even in patients that are relapsed with just a non-muscle invasive disease and seemingly unpredictable. So sometimes these events can be shared. Sometimes they can be gained or lost. And so I think it speaks to really be sure that when we initiate treatment based on genomic alterations, that the piece of tumor material we've used to identify the genomic rearrangement is likely to be representative of the disease that's still in situ.

So in conclusion, rarely I think that the genomic driver landscape of non-muscle invasive bladder cancer prior to BCG is pretty well-defined. But I think there is a need to better understand some of those nuances that I introduced. So thinking about moving away from binarizing genomic alterations and more into a sort of continuous variables. And certainly, I think there's promise for some genomic variables particularly composite to help us at the very least understand prognosis and maybe even help predict BCG response. Unsurprisingly, perhaps post-treatment tumors are always related to the initial lesion and there's no strong enrichment of a driver alteration post-treatment. So it's not that there's one genomic alteration but we always see post-treatment or always disappears. And it's actually a high variability in the path that these lesions take into disease relapse. However, there are distinct patterns in the changes and I don't think it's a stretch to hypothesize that we'll be able to come up with biologically rational groups of tumors in terms of their evolution.

I think there are key challenges with this type of study in general. One of the most... of the big confounders is the sampling bias. So obviously one can only profile what you have and then is still likely disease in situ left or at the very least a somatic field left in the bladder of many of these patients after TUR. So I'm trying to understand how your lesion that you're studying in the lab represents what's left in the patient. It's obviously one that probably is not going to be solved anytime soon. So with that I'll end there and I look forward to discussing with you.

Ashish Kamat: Great, thank you so much, Alex. That was a very concise presentation. Lots of work that's gone into this and very deep level of not only analysis but thinking on your part is very evident. I want to ask you a couple of high level questions. Clearly you brought up the sampling bias and all of that that goes into it but leaving that aside for a moment, if you assume that what you're seeing is actually reflective of what's happening in the tumor itself, is your hypothesis that these markers or pathways are related more to the inherent biology of the tumor or are they related to the treatment in particular, and in this case we're talking about immunotherapy with BCG. What's your working hypothesis there?

Alexander Wyatt: So, I think it's a good question because obviously we're snipping out a lesion and in most cases they have negative margins by pathology. So is BCG treating, in those patients that are initiating BCG and having a long-term response, is that simply because all the tumor was removed or is it actually related to the genomics of the tumors? I think it's probably a bit of both. I know that's sitting on the fence a little bit but I don't think we can paint all tumors with a broad stroke. The fact that we see, for example, a consistent trend for high mutation burden to perhaps be associated with a better response in the context of TUR and BCG does suggest that there is a relationship between the genomics of the TUR and what's left in the bladder. But I think we have to work with the material that we have and we know from other cancers, for example, prostate that'll often... single prostate biopsy can be quite representative of the rest of the tumor, despite heterogeneity we can still see signallers linked to outcomes. And I think probably the same is true in bladder.

Ashish Kamat: Absolutely. Seeing TMB related to outcomes of BCG makes sense because we see that across immune responses in general. And along those lines, I'm sure you have done the work so do you have anything you can share with the audience about how the micro environment or at least your assessment of the micro environment in these patients and the samples that you have correlate or predict BCG response?

Alexander Wyatt: So you're talking about the stroma or infiltrating immune cells in cycles?

Ashish Kamat: Exactly.

Alexander Wyatt: So I don't have final data on that, that I can point to a specific trend but certainly I'm aware that there's been transcriptomic signatures associated with response, and they're probably part in part driven by the infiltrating immune cells, but I can't probably speak to that with this particular protocol yet.

Ashish Kamat: Okay. No worries. And then you showed some interesting data on the plot, the pre and post plot and how the different mutations correlate or don't correlate over time. One of the questions that comes up is what about the time course of these tumors? Was there any correlation between the pre-imposed alterations that you see depending on how temporarily related a tumor is to the index tumor? In other words, say the recurrence was three years later versus three months later. Did you notice a correlation between shared alterations?

Alexander Wyatt: Yeah that's a great point. And because again, one would hypothesize the bigger the time gap, the more differences you should see and there is in fact a loose relationship. It's not as strong as you might expect but certainly there is a trend for the tumors that have the longest time interval to have more differences on average. But certainly that there are cases where a relapse is only less than a year apart or a progression to a muscle invasive disease after just a few months where the lesions are remarkably different and seemingly truncal events in the initial TUR are completely absent from the invasive lesion. And so I think it does speak to the diversity of the somatic genotype within the bladder at this early stage.

Ashish Kamat: Great. Alex, this has been a great presentation and interesting discussion. In the interest of time since we do have to wrap up, any closing thoughts? High-level closing thoughts for our listeners that you want to share?

Alexander Wyatt: So I suppose a lot of my work is in the field of circulating tumor DNA in both prostate and bladder cancer and in the metastatic setting, circulating tumor DNA genotypes, they don't change as much over time as we're seeing in the non-muscle invasive setting here. And unless you're sort of targeting a specific genomic alteration which that then shifts under the influence of therapy. But I think it speaks to the potential for maybe ctDNA in urine to help us understand whether a genomic target is still present or not in the bladders of these individuals. So if there is a certain lesion identified on TUR and on follow-up one is considering whether that alteration is present or not, I think we should be looking minimally invasive analyze like urine to perhaps try to understand if that's present or not on the follow-up.

Ashish Kamat: Great. Thank you once again, Alex, for taking time off from your busy schedule. Stay safe and stay well.

Alexander Wyatt: Great. Thank you very much.