Ashish Kamat: Hello, and welcome to UroToday's Bladder Cancer Center of Excellence. I'm Ashish Kamat, Professor of Urologic Oncology and Cancer Research at MD Anderson Cancer Center, and it's a pleasure to welcome today, Dr. Phillip Abbosh, who is Assistant Professor at the Molecular Therapeutics Program at Fox Chase Cancer Center in Philadelphia, and also has a dual appointment as Assistant Professor in the Clinical Department of Urology at the Albert Einstein Medical School. Dr. Abbosh is going to talk to us today about the work that he has done. He's been involved for many years in looking at molecular correlates of different response predictors of bladder cancer. Today we are going to spend some time talking about his studies and his plans and his insights into the potential correlates related to the RETAIN study. So Phillip, with that, the stage is yours.

Phillip Abbosh: Okay, thank you, Dr. Kamat. I will just disclose a couple of conflicts that are somewhat relevant here. I will be talking about some of our work in urine biopsy and some of this work is not necessarily supported, but at least peripherally related to work that we're doing with some of these pharmaceutical companies.

Okay, so I'll just quickly jump into the RETAIN trial. So the rationale for the trial is that a good chunk of patients, 30 to 40% of patients with muscle-invasive bladder cancer, who receive neoadjuvant chemotherapy will achieve a complete pathologic response at the time of their cystectomy. You might imagine that cystectomy could or could not improve survival in patients who achieve a complete response. In other words, if someone has a pathologically normal bladder and you remove it and there's no bladder cancer left, they may have the same survival as someone who has a bladder intact and has a complete response.

Recently, biomarkers that are predictive of complete response status have emerged, which is good news for patients. And finally, radical cystectomy, obviously, is a life-altering, complicated procedure. People sometimes have to stay in the hospital for a long time, have come complications after the surgery, get readmitted, et cetera, and then, of course, urinary diversion can be life-altering in several ways. So with that, avoiding cystectomy may be a safe option, although, this hasn't really been explored very well in prospective clinical trials. So that was the impetus for the RETAIN trial.

The principal investigator for the RETAIN trial was a colleague of mine here at Fox Chase, Dan Geynisman, has been one of the medical oncologists. The protocol is actually open at multiple sites. Here on the east coast, they enrolled 78 subjects in total, but the data hasn't matured. I borrowed a couple of slides from his talk at the GU ASCO meeting earlier this spring and winter, and I'll just highlight what he's shown so far. This is from, I think, January. So the major inclusion criteria for the trial are muscle-invasive disease, T2 or T3, with no apparent lymph node involvement. Patients had to have a good performance status and predominantly urothelial histology. This was just a phase two trial, but it wasn't randomized. All patients had a TURBT and then the tissue from the pre-chemotherapy specimen underwent sequencing at a company called Caris, a NextGen-targeted sequencing panel. All patients, I should say, had chemotherapy, but there was a stratifier based on whether patients had mutations in any one of these four genes.

After chemotherapy, which again, all patients received regardless of their mutation profile, patients had a second TURBT and then were stratified into one of four treatment groups. So if this second biopsy was negative, there was no clinical T0 status and the patients had a mutation-positive status, they were invited to undergo active surveillance, which means cystoscopy, CT scans, et cetera, at regular intervals with salvage cystectomy as needed. If patients had non-muscle-invasive disease or mutation-negative, but were clinical T0, they could either receive intravesical therapy, followed by active surveillance, chemoradiotherapy, or cystectomy. Then if patients had residual muscle-invasive disease, they had definitive treatment with either chemoradiotherapy or radical cystectomy.

So the primary endpoint was actually metastasis-free survival at two years, for the intention-to-treat population. It was a non-inferiority design. You can see the alternative and null hypotheses here. And then the trial will be considered positive if the lower bound of the 95% confidence interval of the metastasis-free survival endpoint was greater than 64%, which is like a historical control to determine. Not whether this is an effective treatment, but if it is at least as safe as performing cystectomy on everybody.

The interim data point, not the metastasis-free survival, but the pathological status of patients and their recurrence, et cetera. So, they had a 71-patient intent-to-treat population. They were split about half and half between mutation-positive and mutation-negative. You can see that in the patients that had wild-type genotypes, much higher risk of recurrence, metastasis, and death than patients in the mutation-positive arm. Most of these patients in the mutation-positive arm had clinical T0 at their post-chemotherapy TUR, and most of those patients begin active surveillance. Only two have died so far of metastatic urothelial cancer. I was just talking with Dan about this project today, and he expects that, I think, the last patient will reach the two years from their treatment sometime in the spring of the next year. So stay tuned, I'm sure he'll be presenting this again.

So I'm in a really lucky spot here at Fox Chase because we have all these really neat investigator-initiated trials, and so my colleagues are generous to include me in the design of the trials so that we can do really need correlate studies in the lab. So, the correlates that I'll talk to everyone about here today are things we're doing with the immune system, particularly T-cells, these patients who are getting chemotherapy, some of the microbiological correlates we're finding in the urine, and then prognostic urine liquid biopsy-based tests that our lab is developing.

So the basis for our correlative studies in the immune setting are that we found a previous study. This is soon to be published, I hope, that neoantigens, which are the peptides that arise from mutations in coding genes, and those mutations sometimes change the amino acid sequence of those peptides, and sometimes those changes can become antigenic to the immune system. So we measured the amount of new antigens per tumor, and this graph here, each column is one single tumor that underwent whole exome sequencing. What we found was that responders, which are these patients in blue, tend to be on the right side of this spectrum. In other words, they tend to have above median neoantigen burden per tumor. So that was our first clue that maybe the immune system had something to do with chemo.

And again, these patients just got chemo, there was no immunotherapy, but there's this idea here based on this data that neoantigens may be associated with chemo response. We also highlighted ERCC2 variants because at the time we did this data analysis, ERCC2 was the main DNA repair gene that was, at the time, had been associated with chemo response, and you see the patients with those mutations also tend to cluster on the right side of the plot as well.

Using a subset of these tumors, we were able to, in collaboration with David McConkey, we were able to perform CD8 expression microarray. Then using that data, we were able to use an informatic tool to infer the transcriptional signature of specific immune cell types listed along here on the X axis. And the one that drew our attention was these CD8 T-cells, which the expression signature tended to be much higher, about double approximately, in responders than it was in non-responders. We also saw that this NK-activated NK cell signature was also statistically higher in responders than non-responders.

But because of the connection between CD8 T-cells, neoantigen, and response, we decided to look a little more closely at these CD8 T-cells, because patients are undergoing, many of these patients ... This will also include patients that were not on the RETAIN trial, but because many of these patients are undergoing cystectomy with lymphadenectomy at the same time from pelvic lymphadenectomy, we were able to procure fresh surgical lymph nodes, disaggregate these cells, and then perform flow cytometry. We were initially looking at all the T-cells and the lymph nodes, but eventually, and I'll show you why, focused on these effector and effector memory CD8 T-cells. These are the cells that are primarily responsible for eradicating tumors in your immune system. They are the cell types that secrete granzyme, perforin, et cetera. They're the ones that kill tumor cells.

So, when we started looking at these, we were looking at several markers, but the ones that, I think, became the most interesting were PD1, the exhaustion marker, as well as CD69, which is a surrogate for T-cell activation, it's an early marker that's upregulated on the surface of T-cells when they become activated. And what we started seeing was that patients who had the pathologic complete response at the time of their cystectomy. And again, this was patients on RETAIN as well as patients not enrolled in RETAIN, because many patients on RETAIN will not, certainly, responders will not have had cystectomy or lymph adenectomy.

So what we were seeing was that this population here, in particular, which is activated, but not yet exhausted, on the Y-axis, which is the measure of exhaustion, we tend to see this population in a lot of the responders. We don't really see ... Maybe they try to have that population in these patients who don't have responses, but ... It's not exactly a rule, but we do tend to see a lot of these activated, but not yet exhausted, effector T-cells, these are, again, effector CD8 T-cells in the lymph nodes, perivesical, bladder adjacent lymph nodes.

We did some additional studies here where we were looking at the repertoire of the T-cells. So the way you measure the repertoire of the T-cells, or their antigenic repertoire, is to sequence the T-cell receptor. So each T-cell has one unique T-cell receptor, and each T-cell receptor binds to one or maybe a small family of antigens. And those sequences are unique and you can perform deep sequencing on them by collecting a lymph node or a tumor, et cetera, and just sequencing the T-cell receptor within DNA or RNA from that sample.

So what we started seeing was when we plotted the T-cell- So each purple dot here is one unique T-cell receptor, and when you plot the intersection of the libraries from the tumor, or from the lymph node on the X-axis, we were seeing that there was quite a bit of sharing going on between the tumor and the lymph node. So we're seeing the same T-cells in both locations and, of course, because these are in the tumor, at least the purple ones, one might hypothesize that they're doing something to the tumor, such as either potentially trying to eradicate the tumor or acting in a way that's counterproductive and sometimes T-cells can be, for instance, anti-inflammatory and may tamp down an immune response. But because of our finding in the lymph nodes, our flow cytometry, we suspect that the T-cells in the tumor are actually pro-survival and anti-tumorigenic.

So in order to look at that a little more closely, what we've been doing now is taking these CD69 and PD1, these four quadrants, and doing RNAseq on each quadrant individually to look at: one, the activation and exhaustion profile of each T-cell population, both in responders and non-responder; and then in two, through that RNA sequencing data, we're also able to collect T-cell receptors. And we can, again, try to figure out the preponderance of tumor-infiltrating T-cell receptors in quadrant 1, 2, 3, or 4. We might hypothesize that or we do hypothesize that the tumor-infiltrating T-cell receptors will primarily be in this quadrant down here, this activated, not yet exhausted quadrant, which would, again, support our hypothesis that these T-cells are anti-tumoral.

As part of our grant, we're also in one of these specific aims, we're doing short-term cultures of these T-cell types, and we're going to incubate them with therapeutic monoclonal antibodies that are immune checkpoint blockers to determine if the antibodies reverse exhaustion states and make them more active. I mentioned this RNAseq data, and then the TCR profiling, which I mentioned as well.

So the goals of this research are to measure the potential association between immune response, which I mentioned we're measuring through a variety of methods, and pathologic response. Then hopefully, with some of our in vitro work and a mouse model that I already talked about, we'd like to develop methods to enhance immune response to chemotherapy in anticipation of chemotherapy/immunotherapy protocols. Preliminarily, we found that we have this T-cell population of interest in the lymph nodes, which I mentioned, and then we've already characterized some of the TCR overlap in between the bladder adjacent lymph nodes and the tumor. Now, as I mentioned, we're doing RNA sequencing on the four subsets of activated exhaustion T-cells to determine where the tumor-infiltrating T-cells reside.

Okay, that was that. I'm going to switch now gears to talk about microbial correlates. So in order to look at microbial correlates, we were actually able to leverage data from The Cancer Genome Atlas, the TCGA, this data, we were able to download and analyze in collaboration with a company, actually, who was able to pull out ... So typically when TCGA does analyses of their sequencing data, they pull out all the human sequences and then throw everything else away, or things that don't align to the human genome because they're garbage, but we were actually able to find data in that garbage and that is that the genomes of some of these bacteria that we were able to identify. So for instance, Bacteroides ... And I've highlighted some of these ones that are interesting. I shouldn't say just bacteria, we also found viruses.

But viruses and bacteria that are marked with a red dot are known to be, at least historically, cancer-causing either in the bladder or in other situations. So these alphapapillomaviruses are HPV viruses, which of course cause cervical cancer. We found Helicobacter pylori, which of course is associated with stomach cancer, and gastrointestinal cancer. And then these critters that are marked with a blue dot are interesting because they are associated with immune response when found in patients' stool. And there's a whole body of literature now on bacteria that's present in the gut and how it affects human beings' response to immune checkpoint therapy. Now there's actually even published a clinical trial on fecal microbiome transplant in patients receiving immune checkpoint therapy antibodies. So these were things we found in the tumor tissue, not in the gut or not in the stomach or the cervix or whatever. So to find these in the tumor is actually really eye-opening.

We also were looking at the microbes that we could detect in urine using 16S marker gene meta-genomics. We see lots of different bacteria, although, each patient, for the most part, they tend to have one predominant species. Although, there are some patients who have mixes of species. We are now starting to compare these profiles to patients who do not have bladder cancer and we're finding that there are some differences.

So now what do we do with all this DNA and urine and everything? So I'll go through what we're doing and planning to do. We collect urine sediment from patients on RETAIN and other bladder cancer trials and, as I mentioned, we perform 16S ribosomal RNA meta-genomics. And in our first aim, we're just going to compare, are there differences in the profiles of chemo responders and chemo non-responders, to perform biomarker nomination where the biomarker is the bacterial taxa of interest. We're also planning to compare the microbial profiles between the urine sediment and the tumors to determine ... Like make a Venn diagram of what's present in urine, what's present in tumor, and then what's present in both. We plan on doing a linear discriminant analysis to perform some of those analyses. Then once we have nominated and validated bacterial taxa, we proposed to test those in a mouse model of chemo response.

So our goals, as I mentioned, are to determine if there are taxa that are associated with responder status, which is basically just an informatic analysis, but then really to take it to the next level we want to see, out of that distilled list of taxa that are associated with response, do any of them actually cause response in a mouse model. Then to preliminarily describe some of the mechanisms that may be associated with response status in mouse and in human bladder cancer cell lines. Preliminarily, we found that, at least superficially, there are not taxa that are statistically associated with response, but there are organisms we've found that promote resistance to chemotherapies that are commonly used in bladder cancer. So we're following up on those leads and we've got some really neat studies planned.

Lastly, I will discuss prognostic correlates. So I mentioned previously that some of these DNA repair genes are associated with response in a predictive setting. In other words, they answer the question: who will respond? We've been using urine biopsy as a prognostic biomarker of response and that is to say: who actually responded? So DNA repair gene mutation would be something you measured before giving someone chemotherapy; urine biopsy would be something you measure either on chemotherapy or after chemotherapy. So in the RETAIN trial, we're collecting urine before and after chemotherapy, and then we isolate the DNA from that urine and perform deep sequencing. We're able to perform some intense informatic analyses to identify mutations or lack thereof. Then our idea was to correlate mutation presence or clearance in the urine with tumor presence or absence in the tumor. In other words, if the mutations are derived from the tumor and the tumor's gone, then do those mutations go away? That's a real simple basis for a urine biopsy biomarker.

So our preliminary findings are that we do indeed find that mutation clearance from the urine after chemo is associated with response and that mutation persistence after chemo is associated with persistent disease. Now this isn't one-to-one, but there is a statistically significant association between responder status and mutation ... I guess, call it pathologic responder status, and then molecular responder status. I do want to quickly just acknowledge the members of my lab: Rashida and Uttam are the primary drivers for the work I've described here. We recently had fun together downtown in Philly. And then many of our collaborators at Fox Chase: Dan Geynisman, Betsy Plimack, and Alex Kutikov, who have helped us procure clinical samples as well as mentored me along the way. And then several outside collaborators and our funding sources. Thank you.

Ashish Kamat: Thank you, [Phil 00:28:58]. That was a real tour de force of all the work that you've put in and all the potential markers. I'm glad this format on UroToday allows the viewer to go back and forth a little bit because you did cover a lot of stuff and you covered it really well-

Phillip Abbosh: Thank you.

Ashish Kamat: ... and it is fairly intense. In the interest of time, let me just ask you two questions. Number one, based on all the work that you've done personally, and of course the insight you have in the field, what do you think are your top two or three most promising markers? I mean, I know ERCC2, everybody looks at, most folks have found it useful, some haven't, and of course TMB and T-cells, but putting it all together, which two or three markers either alone or in combination you think have the most potential?

Phillip Abbosh: I think you're referring to predictive markers or prognostic markers?

Ashish Kamat: In both fields.

Phillip Abbosh: Yeah. I mean, I'm of course really hopeful with the urine biopsy. I think I don't know if I'd call it a benefit, but the benefit of urine biopsy is that it's agnostic to what happens to the predictive biomarkers because we're not looking for one mutation in a specific gene, in particular, it's whether the mutations are present or absent I think seems to make sense. The DNA repair gene story, there's been a lot of back and forth on that in the literature. I do believe in my heart of hearts that those are important predictors, but it's not exactly a one-to-one ratio, and I think anybody that works on those DNA repair genes would tell you that. I mean, certainly, some of the work we're showing suggests that it may be related to DNA repair, but there also is potentially this immune component, which is much harder to measure, but maybe just as relevant if not more so. So I guess I'd say we're most interested in the urine biopsy, at least in our lab, but I think there's lots of room for everyone to explore.

Ashish Kamat: Do you think that the data that we have, at least to date, suggests that we might be able to compliment ... Well, I know we'll be able to compliment clinical staging, but do you think we'll be able to replace clinical staging using biomarkers?

Phillip Abbosh: Yeah. Good question. Yeah, I don't ever think we'll be able to replace it with a urine test or a tissue sequencing test. It's hard for me to believe that we would replace cystoscopy, radiography, bladder biopsy, et cetera. I'm always really careful when I present, especially our work on urine biopsy, that we would enhance, I call it enhanced clinical staging. This is not really intended to trump anyone's clinical judgment. As a urologist myself, I think that ... I wouldn't believe it, let's put it that way. I think seeing is believing, but I think these markers certainly will enhance clinical staging. There's no doubt about that.

Ashish Kamat: Yeah. I'm sure you're aware of efforts overseas to do away with TURBT completely, do a cystoscopy-identified tumor, and then straight take the patient to radical surgery or radiation therapy. Obviously, you lose the opportunity to do a lot of studies, such as what you are doing, but I also sense that you lose the opportunity to fine-tune and personalize care for the patient. What are your views on that?

Phillip Abbosh: Yeah, I mean, I'm just thinking about, if I were the patient, being as informed as I am, it would be hard for me to sign up for cystectomy without ever having had a tissue, a sample, a biopsy. I mean, I guess with some certain caveats, if I were to have lymph node-positive disease or very obviously invasion on an MRI or something along those lines, but if someone has a papillary luminal appearing tumor on a cystoscopy or some excretory urography test, it would be hard for me to sign up for some extrapolated procedure without ever having had a TURBT or anything to prove that ... I'd want to be convinced that there was, for instance, muscle invasion in my specimen before someone was going to sign me up for cystectomy. So with all due respect to the groups that are ... I certainly understand the rationale. Just, I think it would be hard for patient buy-in, at least in my ... Then along those lines, as a clinician, I think I would also have a hard time making a case to remove someone's bladder without having performed a biopsy as well.

Ashish Kamat: Yeah, and ultimately we do this for the patient, right? So that's why efforts such as what, for example, the World Bladder Cancer Patient Coalition is doing, which is to actually get patient input and insight into how clinical trials should be designed, how these markers should perform, et cetera, et cetera, are really going to be very useful and obviously will help inform studies that are ongoing, including studies such as yours.

Ashish Kamat: Phil, I want to thank you once again for taking the time. This is really great. I do encourage our viewers to take the time, pause, and go back through the different slides. They were very, very informative and you did an excellent job distilling down a really complex topic into fairly understandable terms. So thank you once again.

Phillip Abbosh: Thank you. Appreciate it.