Read the Full Video Transcript

Ashish Kamat: Hello, everyone. And welcome to UroToday. I'm Ashish Kamat, Professor of urologic oncology.

And it's a pleasure to welcome once again to the stage, someone who's joined us many times before, enlightened our audience about molecular profiling, not just in non-muscle invasive, but muscle invasive and talked about MRD, Professor Lars Dyrskjøt . So Lars, thank you so much for taking the time. And the stage is yours.

Lars Dyrskjøt: Thank you very much, Ashish. And thanks for the invitation to present and share some of our recent work here from the UROMOL Consortium here, where we do a comprehensive analysis of a NMIBC. And so this work was actually initiated many years ago from an EU project started in 2008. 10 different clinical centers in Europe joined this.

And then we've been publishing on this cohort over the years and in the past identified different molecular subtypes using transcriptomic analysis of the Hedegaard paper from 2016 and the Lindskrog et al. paper from 2021. So overall here we identified four and then four different molecular subtypes in NMIBC. And these were then classified, as you can see here, where we have a high risk class here, class 2a. There's a high cell cycle and DNA replication signature.

Here we have an inflamed subtype, also the high immune score. And then we have some lower risk classes also class 1 and class 3. And this is reflected here in progression-free survival where we have the highest progression rate in class 2a and also the highest recurrence free. So the highest recurrence rate in class 2a with the lowest in class 1 also.

So we continued this work and then this recent publication here where we included the whole exome sequencing of the cohort on 438. Tumor samples and germline samples also added shallow whole genome sequencing to look at the copy number variations also. And then we used the total RNA-Seq that we have been producing early on to get a more complete overview of the genomics and transcriptomic features in NMIBC.

And here what we can see here is the genomic landscape in NMIBC. Look at all the-- we have all the different mutations here. We've seen it earlier in smaller cohorts also.

But now we have this more comprehensive overview of how many of the tumors we see these different mutations in. And as we know from other publications, we have a high frequency of FGFR3 and PIK3CA mutations. But we also see a relatively high number of TP53. And ATM mutations in NMIBC.

If we compare that to muscle invasive disease we can see that many of the tumors or many of the mutations are actually concisely present also in NMIBC identifying this continuum of the disease. We really see that TP53 and RB is really higher in NMIBC, but we still see it to a certain degree in NMIBC also.

And what we also saw from this was that we saw at relatively different mutation profiles in the different transcriptomics classes, also, as you can see from this area here. Interestingly, in this paper here, we found that 15% of the tumors actually had whole genome amplification. So that's a relatively high number. Also in this early stage disease here.

This is identified here with the red amplification errors here. And this is also found in muscle-invasive bladder cancer, where we see that this is the case in actually up to 58% of the tumors in TCGA cohort. But again here we find it in 15% of these tumors. So we characterize this a little further.

So if we look at the genes that are involved in tumors that had have whole genome doubling amplification here, we see that it is a TP53, ERBB2, and RB1. Mutations are also associated with whole genome amplification.

We see the whole genome doubling in special class 2A the highest risk, and we can see also that this is really associated with the progression free survival to a high degree here. If we look a little further into this, we can see that the whole genome doubled tumors have many mutations in the cell cycle genes-- P53, RB again here, CDKN2A changes also-- much more than we see in the diploid tumors here.

We also looked into to see what is the immune landscape of these tumors. And we did that by doing a deconvolution of the total RNA-seq from these tumors. And, and here we can see that many of the, for instance, CD4 T-cells, cytotoxic T lymphocytes, and dendritic cells and macrophages, they don't really show any significant difference between the diploid tumors and the tumors with the whole genome double genome.

But if you look at these innate cells here, the mast cells and neutrophils, we can really see that we have a higher level of these innate immune cells in tumors with the whole genome double genome. So this may actually be from some secondary effects in the tumor microenvironment in the whole genome double tumors that triggered this recruitment and activation of the innate cells, leading to a higher risk of progression in this case here. We also found that we saw a higher exhaustion score or T-cell exhaustion signal in tumors with the whole genome double genome. And this was also shown by looking at the PD1 positive cells also, where we found a higher level of these in whole genome double tumors.

So finally, another thing we did here was to instead of just look at the single genomic, transcriptomic layers, we tried to do some integrative clustering of this where we saw if we could identify novel molecular groups by looking at the copy number alterations, mutations, gene expression data in the same integrative clustering. So we have a 230 of the patients' tumors where we had this complete data set. And then we used this integrative clustering algorithm to identify four different cluster groups here in the data also, where we use them, the copy number alterations, mutations, and gene expression patterns.

And overall, if we compare these four different iClusters, as we call them, to our four original transcriptomic classes here, you can see that there are several overlaps between these, but there are also multiple differences. One thing that's striking here is that all the whole genome doubled tumors, they are actually present in the iCluster 4 here. We see many of the tumors in iCluster 4, they also have a high copy number change. And they have a high tumor mutation burden also.

Another thing we see is that many of the class 2A tumors are actually present in iCluster 4, together with some of the class 2B tumors here that have some of these features that I mentioned. And interestingly, when we look at the progression-free survival here, we see that the iCluster 4 actually harbors many of the progressions to muscle invasive bladder cancer also. So, of course, this new classification here needs to be validated in more samples. And I think this gives us some more insights into what drives tumor progression that it may not be optimal for clinical utility, but I think it gives us some other ideas about what drives aggressiveness in early stage bladder cancer.

So with that, I would like to summarize that we identified some new genomic layers. When you generate new genomic layers in the UROMOL cohort, we see 33 genes are significantly and recurrently mutated in NMIBC. And we see that this mutational landscape is actually different for the different transcriptomic classes that we see. So 15% of the samples have whole genome doubling, compared to 58% in muscle invasive bladder cancer. And this is associated with a poor outcome cell cycle checkpoint pathways and specific immune context also.

And then finally, we identify these four iClusters using the multiomic full data set we have. And this iCluster 4 really comprises a group of highly aggressive tumors. So it may be a tool to better stratify compared to transcriptomic subtypes alone. So the future will show us that in new studies.

So with that, I'd like to thank you for your attention. And this is the team that did most of the work. I would like to say thanks to all the collaborators here also in the UROMOL Consortium. So thanks for your attention.

Ashish Kamat: Thanks so much, Lars. I mean, always learn a lot every time you present. And I'd encourage all our listeners to go in and dig deep into the actual publication, because there's so much information in there that can actually help direct some of the research that you might be doing in your own labs.

And along those lines, Lars, I want to just acknowledge and thank you again for always being very supportive of young investigators throughout the world. I know you've supported several of my fellows, including on work that's been done on the UROMOL Consortium. So thank you for that.

Lars Dyrskjøt: Of course.

Ashish Kamat: Quick question about the iClusters. I mean, the distinct iClusters that you have identified, could you postulate how you might envision that being incorporated along with clinical parameters and help with personalized risk stratification for patients?

Lars Dyrskjøt: I think it's probably not going to be this multi-layer predictor that we're going to use in clinical routine. I think what we need to do now is to look into what are the most important features in these iClusters that actually drive this separation, and then try to make a more simple classification scheme. Because when we're using the iClusters, then you actually you have to have these different genomic layers, but it's also context specific in the way that you're discovering these groups from the full cohort.

So I think we're-- now I've been working on these different cluster algorithms for years now. And we have to start all over again and try to validate in new clinical data set and so on. So I think in this specific scenario here, I think that we really have to dive deeper into the most important mechanisms here and then try to validate those in new cohorts. Otherwise, I think, it becomes too difficult to actually make a clinical use of this data set or this way of doing it.

Ashish Kamat: Sure, sure. Absolutely. And looking your study, I mean, you convincingly demonstrate the association between TP53 and whole genome doubling. And again, even though the FGFR pathway is more relevant in non-muscle invasive bladder cancer, there's clearly an overlap. But substantial number of WGD-positive tumors in your cohort lacked these mutations.
So, could you elaborate on what you think are some of the drivers? Is it just the instability, TMB, ARID1? What do you think is driving the doubling in the other cohorts?

Lars Dyrskjøt: It's a very good question. We're really looking into that to see if they are not affected in the TP53 pathway, what is actually driving this amplification then? And we couldn't really find any signs of this in the data we have.

So the natural thing to think about is probably some epigenetic changes or some other things that we haven't really looked into yet that may be driving this. But it's clear that it's an important feature for cancer development to acquire this whole genome doubling. So that's also future research to identify this.

Ashish Kamat: And we've had multiple discussions. My focus tends to be immunotherapy for bladder cancers. And I was struck by the low ERBB1 expression and its role in immune invasion.

And of course, your study shows that. It also shows neutrophils associated, some T-cell exhaustion. Can you share with us some of your insights, maybe not just from this paper, but in general with all the work you've done and how you think that correlates with these molecular subtypes and clustering, this whole immune tumor-host environment interplay?

Lars Dyrskjøt: I think it's important. Of course, it's important to gain deeper knowledge about this, what is actually driving the composition of the tumor microenvironment in these tumors. But what is also driving the immune escape. What is required for the tumor to actually escape an immune surveillance here.

So I think what we can see from this is that it seems as if the exhaustion is very important. We've seen that in multiple studies earlier also. And I think it's important here to identify what is actually-- can we use this in studies when we look at the BCG checkpoint inhibitors also to see if we can actually identify patients from that may benefit from a checkpoint inhibitor by looking into these exhausted mechanisms? So I think that's extremely important to look into.

I think another thing here is that-- so now this the UROMOL cohort, it's a good cohort for looking into some of these biological signals and correlations. But it's also starting to be outdated in the way that patients are treated today, because this is-- and it is a prospective study, but it's collected from 2008 to 2012. So I think we need some more up-to-date prospective studies to really look into this with newer treatments also.

Ashish Kamat: I think, I mean, one of the key cohorts that I'm sure you're thinking of already is the ones that are being treated with TAR-210, the intravesical erda and the whole FGFR story. And of course, we're seeing trials coming out where it's all comers, those that have no FGFR alterations and then targeted on specific activations, mutations, extracellular and intracellular domain. So I think someone like yourself should definitely be someone that's going to be at the forefront of those developments.

Lars Dyrskjøt: It'll be interesting.

Ashish Kamat: I think with the cohort you have-- and I'm going to put you on the spot a little bit here because we hear this debate all the time between clinicians that say that high-grade TA tumors are behaving like any high-grade tumor. Whether it's 2 centimeters, 3 centimeters, 4 centimeters, their pathway is similar to other high-grade aggressive tumors. And then of course, the clinical guidelines look at age, 70.

They look at size, 3 centimeters. And they put patients in intermediate and high-risk tumors. I'm sure you have enough power in your cohort to be able to answer that once and for all. Do you have any insights for us on whether you could just categorically say, if you're a high-grade tumor, it doesn't matter if you're 2 centimeters or 3 centimeters, your pathway alterations are similar?

Lars Dyrskjøt: So I think we have looked into this in some of the earlier publications also on the subtypes we have, and also just in this more simple high risk, low risk subtypes and so on where we have demonstrated that these different prognostic biomarkers and subtypes are actually independent prognostic factors when we stratify for these. So now it's we haven't really stratified it using the completely new, up-to-date classification systems. But it's still, I think, we have demonstrated that it is adding additional value to this.

If it is enough for using it in the clinic, if we're adding that much additional information to really use it, that's then the next question. So I think we have that information. But then the question is, how much do we want to improve this also?

Ashish Kamat: Lars, every time I have you on the forum, I ultimately go over time. So at this time, I'm cognizant of time. And I'm going to stop. And just thank you once again for joining us.

Lars Dyrskjøt: Thank you very much for having me here.