Ashish Kamat:  Welcome to UroToday's Bladder Cancer Center of Excellence. My name is Ashish Kamat. I'm a professor of urologic oncology and cancer research at MD Anderson Cancer Center. And it's a pleasure to welcome Woonyoung Choi, who used to be at MD Anderson Cancer Center, and we used to work very closely together, and we continue to work closely. But now Dr. Choi is in the Department of Urology at the Greenberg Cancer Institute at Johns Hopkins Medicine. But of course, moving from Houston, Baltimore has not stopped her from continuing her fine work. And today she's going to present to us the data that she has generated on fibroblasts' immediate resistance to BCG therapy in non-muscle-invasive bladder cancer. Dr. Choi, the stage is yours.

Woonyoung Choi:  Thank you for a nice introduction, Dr. Kamat, and thank you for having me. I'm going to talk about the BCG resistance mechanism in non-muscle-invasive bladder cancer. BCG has been our standard of care to treat the patient with non-muscle invasive bladder cancer for decades. However, within five years of the first instillation of BCG, about 50% of the tumors will recur and 20% of the tumors will progress. Therefore, to overcome BCG resistance, we need to understand the mechanisms of a BCG therapeutic effect and identify molecular correlate that can predict the BCG responses. In order to do this, firstly, we correlate molecular subtypes and BCG responses. In muscle-invasive bladder cancer, molecular subtypes haven't been thoroughly studied. However, NMIBC does not have a solid molecular classification. Therefore, we adopted a consensus molecular classification from muscle-invasive bladder cancer. Consensus molecular classification was developed by compiling six different molecular classifiers. In this consensus, molecular classification subtypes tumors into six different molecular subtypes, including three luminal tumors; luminal papillary, luminal non-specified, and luminal unstable, stroma-rich, and basal tumors and neuroendocrine tumors.

To apply this consensus molecular classification, we generated the Hopkins BCG cohort. And this one is a collaboration of work with Max Kates, a urologist at Hopkins. He identified the 30 BCG responders and 54 BCG non-responders. And after we obtained the tumors, we generated the gene expression properly and examined the tumors using consensus classification. Each block is presenting [inaudible] sub-type marker expression. And this block, they presented each molecular subtype. Out of our six consensus molecular subtypes, [inaudible] belong to three luminal tumors. And out of the three luminal tumors, the luminal papillary has the most of the BCG tumors. And luminal papillary tumor has the features of enrichment of the FGFR3 signature.

And when we looked at the relationship between molecular subtypes and response, we couldn't see any correlations. In this bar graph, the red color showed the nonresponders and the blue color showed the responders. And then as you see here, each subtype has a mixture of the responders and non-responders. Since we didn't see a good correlation in consensus classification and BCG response, we applied other molecular classifiers, MDA molecular classification, which we developed at MD Anderson. This MDA classification has three different molecular subtypes, basal, p53-like, and luminal. So when we look at the correlation between BCG response and MDA molecular classification, non-responders were enriched in p53-like tumors. And p53-like tumors are characterized by fibroblasts' signatures and immune signatures. So seemed like one BCG non-response subset has an enrichment of fibroblasts and immune signatures. And also there's another subset of nonresponders inside luminal tumors.

And these non-responders have a feature of an absence of both, either fibroblasts or immune cells. So this data suggest to us, there are three different immune phenotypes: immune inflamed, immune excluded, and immune desert subtype. So this immune inflamed phenotype has a T-cell infiltration of the tumor cells. So this one might be a good candidate for the immune checkpoint inhibitors.  And the immune excluded subtype has both immune cells and fibroblast cells.  And immune desert does not have either immune cells or fibroblast cells.  So it seems like our two BCG immune subsets in nonresponders mimicked immune excluded an immune desert subtype.

So to further investigate the relationship between BCG resistance and immune and fibroblasts signatures, I obtained the [inaudible] signature for immune cells and fibroblasts then performed the gene set variation analysis. Red color means the signature is emitted, while green color means the signature is suppressed. This data showed that our BCG non-responders were separated by immune and fibroblast signature, implying there are at least two distinct resistant mechanisms for BCG treatment and immune enrichment subtype has enrichment or fibroblasts, and this subset might have a feature for the immune excluded.  The other subtype in non-responder has the features of immune desert. Since this subtype does not have either immune signatures or fibroblast signatures, interestingly, this immune desert subtype has the enrichment of the FGFR3 signature and based on this one, this immune desert subtype, a subset of non-responders inside the luminal tumors. And this even excluded a subtype, like p53-like tumors in molecular subtypes.

To find the therapeutic targets and immune excluded subtype, we extracted the differential expressive genes in these tumors. These immune excluded tumors.  As you see here, the top differential expression genes, or fibroblasts signature. In addition to that, we found that TGF-β signaling was also enriched in immune excluded subtype by gene expression, and also by individual personal analysis. In this engineered personal analysis, they show that TGF-β one, two, and three were also activated and it's [inaudible] the three were also activated too. So TGF-β is a really interesting molecule because TGF-β shows importance in immunotherapy, resistance mechanisms. IMvigor 210 clinical trial data pieces show TGF-β expression is correlated to the probability of survivor progression and this survival curve is separated by TGF-β expression.

So when TGF-β is highly expressed, they show the poor survival outcome. And also TGF-β pathway is activated. That means the disease was progressed. So based on this data, TGF-β might have pen cancer, immunotherapy resistance mechanism roles. So including not only checkpoint inhibitors, but also BCG immunotherapy. We wanted to identify BCG the acquired, resistance mechanism. So we expanded our cohort to include the chairman's from MD Anderson and the University of British Columbia to have 53 repairs on pre and post to BCG treating tumors. First, I validated it just under immune desert and immune excluded subtypes in this extended cohort to, [inaudible] show that clearly we have two distinct subtypes, immune desert and immune excluded in this extended cohort. And then we want to know which genes were induced by BCG treatment.

And I extracted the differential express the gene between pre and post BCG tumors into different immune subtypes, immune desert and immune excluded subtype independently because, between these two subtypes, there is a huge difference in the gene expression. Therefore, there might be two distinctive immunotherapy resistance mechanisms. And interestingly immune excluded subtype does not had to many changes between after a BCG treatment. However, in the immune desert subtypes at least 500 [inaudible] were induced pre and post-treatment. And you see, here show the ratio with the post and BCG pretreated tumors. The red color means genes were induced by BCG and the green color means genes were suppressed by BCG surprisingly immune desert subtypes acquires the features of an immune excluded subtypes after BCG treatment with high expression of the fibroblasts, immune, TGF-β pathway. So it seemed like BCG induced, immune signature in the immune desert subtype, however, simultaneously BCG also induced fibroblasts and TGF-β.

So this induces the TGF-β and fibroblasts throughout the immune that infiltration into the tumor cell to make tumors resistant to the BCG treatment. So we will test this hypothesis by doing all spatial analysis by artifact's immunohistochemistry. So in summary and the future plan, we found two immunogenic cell types in BCG unresponsible tumors in both subtypes fibroblasts TGF-β might have a key role in BCG resistance mechanism. We really validated this observation in an independent dataset and develop the clinical classifiers to predict the responders and non-responders to BCG treatment. So this work is a collaboration where the John's Hopkins GBCI investigators, specialty doctors, Kates, McConkey, and [inaudible]. Also, we obtained the tumors from Dr. Black in UBC and Dr. Dinney and Kamat at MD Anderson. Thank you for listening and I will stop here.

Ashish Kamat:  Thank you so much Woonyoung, this is a lot of important data presented by you, of course, at the BCAN Think Tank and in summary today, if you had to select certain key messages that you want to reveal to our audience based on not just this presentation, but all the work that you've done, when it comes to trying to able to identify these subtypes and how it pertains to non-muscle invasive bladder cancer, what would be certain key messages that you would want the audience to remember?

Woonyoung Choi:  So TGF-β has an important role in, [inaudible] immunotherapy resistance mechanism, because initially when we look at the pre BCG pretreated tumors, and we saw that there are two different molecular subtypes based on the immune-inferior signature. So we thought that there might be two distinct resistance mechanisms among these two subtypes because it is subtype has high expression over the TGF-β. However, this subtype does not have a TGF-β and fibroblast. So this way we extracted the differential gene expression pre-and post in independently in these two distinct subtypes. But after BCG treatments, this even debt or the subtype also got the feature of the even excluded subtype. So clearly five of us mediated TGF-β has an important role in BCG resistance mechanism.

Ashish Kamat:  Great. And, when you're looking at this cohort of patients and their markers, have you also found something similar, or have you studied this when it comes to other immune agents? I know you've done a lot of work with interferon-alpha and some with pembrolizumab and things like that. Or do you think this is BCG specific?

Woonyoung Choi:  This one is immunotherapy resistant mechanism because this IMvigor 210, then clinical trial data also shows the TGF-β might have an important implication of the resistance mechanism. But this one, pre and post data set. This one is pre immunotherapies, [inaudible] So this one can just prediction, but in terms of BCG treatment tumors, where the two subsets, and even TGF-β in the beginning, they had acquired the resistance over the TGF-β.

Ashish Kamat:  Now, TGF-β has been involved and studied in multiple different pathways. What do you consider it, for example, do you think it's a targetable pathway in a realistic sense? I mean, we all know what it does, but what's your sense based on all the work that you've done on, is this a targetable pathway? Is it something that's intrinsically modified? How could we use this when it comes to harnessing the power of the data that you provided in order to improve the outcomes of patients?

Woonyoung Choi:  So I think, first of all, we have the patient we can separate the tumors based on their prognostic expression, over the EMu and fibroblasts signature in these, these are not responders. So these nonresponders might treat the TGF-β first and then do BCG like immunotherapy or like BCG immunotherapy. And also currently I am collaborating with Max Kates and try to identify the other subset. Who is it going to be responded like those, what other combination therapy? So BCG is for like the front-line therapy right now, but we need to identify the alternative therapy who is not going to respond to BCG treatment. So right now we are trying to identify other molecular,  marker, who is going to respond to chemotherapy and who's going to respond to immunotherapy. And this is an ongoing project. I'm not sure that's a good answer or not.

Ashish Kamat:  No, no, that's great. That's great because I mean, it's a complicated question that I sort of asked you and I didn't mean to put you on the spot, but I appreciate your response. One last question in the interest of time, there are groups of investigators and you've collaborated with many of them, Charles Guo, for example, that are making the case that a lot of these molecular markers are hard for the community, a pathologist or the people out in the smaller centers to do mainly because of cost and expertise and things like that. Have you also explored trying to do this marker studies using, I guess, low-cost technologies, such as IFC markers? And if so, where do you think the field is headed? Do you think we can create a panel that can be used by the average European pathologist? Or do you think it's going to move towards specialized centers, such as yours to do this sequencing and such assays?

Woonyoung Choi:  So this one, I think it would need to come from this, independent the data set first and also task school is working on histochemistry. But even histochemistry, like I know that there's a lot of development of the computational methods to analyze, histochemistry too. But so far, in our case, we would like to move this gene expression data to have a certain smaller number of the panel and to develop the clinical assay, to predict the BCG responses. And we also want to measure the TMB as may TMB as a surrogate of the BCG response, because BCG is one like immunotherapy response may be based on the TMB load. So if this one's successful, we would like to develop like a small panel of the gene expression like a classifier, but everything is an ongoing project.

Ashish Kamat:  Once again, Woonyoung, thank you so much for taking time off from your busy schedule, this is really great, a lot of exciting stuff. And of course, we continue to collaborate in, in do all this work together, but it's folks like you that are in the trenches doing this work on a day-to-day basis that understand the data so well and can present it so nicely to our audience. And we really appreciate it. Once again, thank you so much.

Woonyoung Choi:  Thank you, thank you for inviting me.