Read the Full Video Transcript

Ashish Kamat: Hello, and welcome to your today's Bladder Cancer Center of Excellence. I'm Ashish Kamat, Professor of Urologic Oncology and Cancer Research at MD Anderson Cancer Center in Houston. And it's a distinct pleasure to welcome in a virtual manner two friends of mine from Malmö, Sweden. Associate Professor of Urothelial Cancer, Gottfrid Sjödahl, and Professor of Urology and principal investigator Fredrik Liedberg, both of them need no real introduction to our listening audience and they publish extensively in the field of bladder cancer and molecular advances. One of the recent papers that looked at different responses to neoadjuvant chemotherapy in urothelial cancer molecular subtypes has gained a lot of traction in journal clubs across the globe. And also of course discussions in social media, since we're caught and COVID. And I'm really pleased to have both of them join us today to present the highlights of their paper, and then have a quick round payable discussion. So with that Gottfrid, the stage is yours.

Gottfrid Sjödahl: Thank you very much Ashish. And first of all, thanks for the opportunity to present and discuss this latest study from our group. And the title of the study is Different Responses to Neoadjuvant Chemotherapy in UC Molecular Subtypes. So in this study, we set out to test if there is an association between molecular subtypes and outcome in patients treated with neoadjuvant chemotherapy, as compared to patients treated only with radical cystectomy. And a secondary aim in the study was to explore if we could find any other expressed genes that could potentially be associated with response. So this is a retrospective cohort study, and this slide shows the inclusion of patients in the chemotherapy treated cohort shown on the left side, and in the cohort treated with the radical cystectomy only on the right. In the chemotherapy treated cohort, we only studied patients who received at least two treatment cycles, and we excluded patients with T1 disease and there were a few samples that could not be analyzed. So eventually 125 neoadjuvant treated patients were analyzed.

And the next is showing the methods of the lab work. So starting with a TURB tissue block, we applied macro dissection and RNA extraction. And we also sampled these blocks for immunostaining by embedding tissue in TMAs. And total RNA from each tumor was then hybridized to hold transcriptome microarrays and analyzed using standard gene expression analysis tools and pipelines, including predefined single sample rule based subtype classification, according to the Lund taxonomy. And to be confident in the classification, the molecular subtypes were also assigned based on the consensus algorithm and on the basis of the immunostaining on the TMAs.

So this slide shows the result of the classification and the bars on top of this slide. The different colors denote the various Lund taxonomy subtypes. And the chemotherapy cohort is shown in the upper bars and the radical cystectomy only cohort in the lower bar. On the left is the RNA based classification, and on the right is the results with the IHC based classification. And we also show example IHC images demonstrating the typical marker profiles for tumors of these various subtypes. So given this classification, we then studied the association to pathologic response defined as pT0 and zero in the five major Lund taxonomy subtypes. So the highest pT0 rate of 52% was seen in the luminal like genomically unstable subtype GU here. And the lowest 21% was seen in the basal/squamous like subtype. And the difference between these two subtypes, the highest and the lowest was statistically significant.

So we then moved on to study cancer- specific survival of the chemotherapy treated patients. And we observed a very similar pattern as we saw for response. So on that left the Kaplan–Meier curve, and the blue curve shows the good prognosis of the genomically unstable subtype, and again, a significantly worse outcome for the red, which is basal/squamous and the purple, which is neuroendocrine like subtypes after treatment. And the second curve in my plot also shows the chemotherapy treated patients, shows in clinical stage groups and as expected, there was a big difference in survival, depending on clinical T stage. And importantly, the different survival of molecular subtypes remained significant after adjusting for clinical stage, as showed in the multi-variable analysis.

So the overlap of the Lund taxonomy RNA classification, with both the consensus classification system and IHC based classification was as expected. We obtained similar results with these methods also meaning that the significant differences in outcome were found for the corresponding groups with all three methods. And for the consensus classification, we note that the stroma-rich subtype, in addition to the luminal unstable also showed improved survival compared to basal/squamous.

And we then performed the same Lund taxonomy classification on the two largest publicly cohorts, which are from the studies by Seiler and colleagues and by Taber and colleagues. And although our study, and these are different in many ways, we still added up the responses from all three cohorts. And in aggregate the luminal like learned taxonomy subtypes, both the genomically unstable and urothelial like A, showed improved response rates compared to basal/squamous in taking all three studies together.

And this shows that the findings in our data are not reversed or rendered statistically uncertain after considering other existing data. So to conclude in molecular subtype derive different clinical benefit from neoadjuvant chemotherapy, at least in the available retrospective studies. And patients with a basal/squamous like tumor can be selected to upfront radical cystectomy instead of neoadjuvant chemotherapy, possibly. Here we note that the adjuvant chemotherapy is an option if there is adverse pathology in the case that neoadjuvant chemotherapy is not given. Thank you very much.

Ashish Kamat: Thank you so much Gottfrid, for that very succinct presentation. Since I have both of you available, I'll ask the question and either one or both of you can feel free to chime in. The first question that comes up and we'll focus on your study, but the first question that often comes up is that we've heard about molecular subtyping now for many years, right? I mean, it seems the publications were maybe in 2014 or 2013, but people have been talking about this even before then. When are these going to be ready for primetime use in your estimation?

Gottfrid Sjödahl: Well, I can answer that question, because I think from a biological perspective, what molecular subtyping is doing, is it takes the whole disease and it biologically splits it up into groups that when you have this split available, it enriches the view of the disease, and then it makes our understanding more detailed. So in that sense, it is already contributing to research and to our understanding of the disease. In terms of clinical application that may come tomorrow or a long time from now, or if ever. But I'm hopeful that this information is so central for tumor biology, that then there will be clinically useful information in this and molecular subtyping eventually.

Fredrik Liedberg: I mean, the bottom line is that we have to apply molecular subtypes in order to figure out the clinical role and the use of it. And our strategy is currently to do upfront molecular subtyping with observational study design in combination with applying molecular subtyping in randomized trials to increase the knowledge. And really succeeding in translating the subtyping into clinical practice. So, I mean we're not there, but we have to consider it while designing studies and in our daily practice in order to perform prospective observational trials. Instead of the retrospective design that we used in this trial.

Gottfrid Sjödahl: Yeah.

Ashish Kamat: Yeah. And both of you make excellent points. So I completely agree, and I don't think there's any question that the molecular subtypes, like you said, are more about biology and they help us understand the disease a lot better. And because we now have the tools and you guys have been doing this work for longer than anybody else, I think. So you can attest to this, but the tools have gotten so much better that now you can do this with smaller amounts of tissue, more reliably and more reproducibly. And that's why, of course now the clinical trials must... I mean, there's no question that if you design a clinical trial today of any intervention, it must include some analyses of subtypes moving forward to further understand it.

But let me ask you another question. Obviously I'm in an ivory tower, you are all in an ivory tower, we have this available to researchers and clinicians in our centers. But when we look at our colleagues that are taking care of patients, at least in the United States, the majority of them are not in academic centers. And they rely more on their pathologists, and of course the pathologists say, just as you showed here, that you can use IHC to make similar inferences on biology and even classify patients.

What is your level of confidence in the overlap between IHC and gene expression profiling? And do you think that one will win out over the other or that they'll always be complimentary?

Gottfrid Sjödahl: No. I think to certain extent, they will always be complimentary, because the immunohistochemistry captures in C2, for example, intratumor heterogeneity in ways that bulk sequencing cannot do. At the same time, the RNA-Seq generates a type of data that you could never get with immunostaining. So in that sense, it will be complimentary, but the overlap is there. And in this case, compared to the five groups, RNA vs IHC, agree in 67% of the times, so it's not perfect. But considering that it's five by five analysis, it is actually a pretty high degree of agreement. And then furthermore, that, it maybe does not matter so much. Maybe the disagreement is really in the threshold cases. Because as we showed the results work as well, when we apply the IHC based classification as with the RNA based. So although they don't agree perfectly, they both seem to capture the important aspects up of biology. And the core cases will be classified concurrently with both methods.

Ashish Kamat: And as we move forward into incorporating more IO agents, immuno-oncology agents into the treatment of bladder cancer patients. Clearly, the host factors are very relevant and very important. How foresee molecular subtypes and the assays of course, interacting with the host, the stroma. And how do you account for that when you're looking at tumor molecular subtypes in the setting of the host stroma either with chemo, but especially when you're looking at IO agents?

Fredrik Liedberg: Well, I think one view, or one way to try to answer your fairly difficult question is to see the story with the IMvigor trials, where the phase three was negative, and then when you sort of made the biological work in the paper by Mitathos and coworkers in nature 2018, there are multitude of factors. Intrinsic, extrinsic, tumor related, and even factors also independent on subtypes, such as TTFP type signaling and immune infiltration and so on. So I think there is a puzzling pattern without an easy answer. And I would say that, that's why we decided, to apply RNA sequencing as our base method. Because if you have the RNA-Seq data, you can always raise questions on infiltration and immune cells and stroma afterwards.

To have hypothesis generating findings within the perspective observational data. And so it's a difficult question, but I think it will be a multitude of factors contributing as well as, I mean, mutational patterns will affect whether we have other modalities in together with EOs, such as FGFR3 mutation based approaches and so on. So it's a complicated puzzle and I have no definite answer. But I think that transcriptomic data is one way to raise research questions continuously, while developing this field further, and finding and defining the role of the molecular subtypes per se.

Ashish Kamat: Yeah. And the retrospective subtype analysis is always fraught with issues, if you don't tie it in with biology. Right? If you do the biology and the biology makes sense, and your retrospective analysis makes sense, then at least we could have some level of confidence in the data. But I'm sure, many of you remember that you can even go in and look at patients based on their sun signs and come up with statistically significant response to chemotherapy. Now, of course, I've been saving one of the more provocative questions for the last. And that is how do you explain... And again, you've done this before, when we've chatted. But for our audience, how would you explain the discordant findings or seemingly discordant findings, let me not say discordant, between your basal/squamous like tumors and the response to chemotherapy and the recommendation for upfront cystectomy from what's previously been published.

Gottfrid Sjödahl: Yeah. So I can answer to this. And you're probably referring to the conclusions from the paper by Seiler and colleagues in European urology, which showed that with GSC, genomic subtyping classifier system four tier has subtyping system. The basal subtype had a survival advantage in the treated cohort compared to the reference untreated cohort. So the question here is sort of twofold. First of all, and there was no difference in response in this study. So the question is, what is this difference in survival? It could of course be due to an actual improvement in survival of this subtype in the treated patients. But it could just as well be that there is a difference between these two cohorts in outcome for this particular subtype in a way that is not directly related to treatment itself.

The second part is the different classification methods here. So actually it is a bit problematic because the subtype term basal in the GSC subtyping system, does not contain the majority of the basal/squamous subtypes with the Lund system. So again, this is an issue of nomenclature, but in this case, I would just highlight that the Lund taxonomy basal/squamous, and also the consensus basal/squamous is actually mainly classified as claudin-low or basal claudin-low as it's initially called in the GSC system. And that subtype actually does not do well in the treated group, in the Seiler et al study. So considering that basal/squamous, which does poorly in our study is actually based on claudin-low in their study, it does poorly also there. So the studies are not really in disagreement in the data. Only perhaps in the nomenclature of the subtypes, which is in that case unfortunate. I hope that, that clears it up.

Ashish Kamat: No, no it does. And that's an important point. I wanted you to highlight, which you of course did. Is that when people look at different results and different publications, it's important to recognize that the nomenclature, while it might seem to be similar, because similar sounding words are used, they may not actually be similar in the biology. The underlying biology is what your paper highlights, and that's very, very important. It's always fun chatting with the two of you and I actually lost track of time a little bit. So let me wrap this up and in closing, give both of you maybe a short 30 seconds each to highlight and leave the audience with your top level thoughts on this topic.

Gottfrid Sjödahl: Okay. So I can start. I think one of the things that I think will be important for future translational and biomarker research, is to use molecular subtypes, maybe not as the biomarker itself, but as the framework in which other biomarkers can operate. So in this sense, we are presenting here an idea of a second generation of biomarkers that operate within a specific context defined by the subtypes. So I think that could be important for the future.

Fredrik Liedberg: And for the clinical perspective. I mean, it depends on the context where you're working. I checked our last two years here in Malmö where we do a 100 cystectomies a year. And in patients below the age of 75, four out of five have neoadjuvant chemo, if they have a mass invasive tumor. So if you give a lot of patients neoadjuvant chemotherapy, this might be a way to select those perhaps should have less use of such tough therapy prior to cystectomy. So it depends on the context, if you give everyone neoadjuvant, this might be a way to reframe from the neoadjuvant, but if you only give selected patients, well, we don't know.

Ashish Kamat: Right. Well, once again, thank you, Gottfrid and Fredrik for taking the time and spending it with us today. Our audience I'm sure will really appreciate it, I certainly learn a lot every time I talk to you guys. Stay safe and stay well.

Fredrik Liedberg: Thank you very much.

Gottfrid Sjödahl: Thank you.

Fredrik Liedberg: Pleasure. Goodbye.

Gottfrid Sjödahl: Bye.