Gene Expression and Molecular Subtyping in GU Cancers Presentation - Ewan Gibb

March 16, 2023

Ewan Gibb, a PhD scientist, discusses the Decipher Bladder clinical test, a molecular subtyping assay for bladder cancer that predicts the risk of upstaging, identifies high-risk neuroendocrine-like disease and determines the benefit from neoadjuvant chemotherapy. The test breaks a patient's tumor into one of five subtypes and recognizes luminal and non-luminal categories, with neuroendocrine being a smaller subset. The Bladder Grid product includes all current subtyping models, and the assay predicts the risk of upstaging cystectomy and the benefit of neoadjuvant chemotherapy.


Ewan Gibb, PhD, Principal Scientist, Bladder Cancer Lead, Veracyte, Inc.

Read the Full Video Transcript

Ewan Gibb: Good morning. My name's Ewan Gibb. I'm a PhD scientist, the bladder cancer lead with Veracyte. I'd like to thank Dr. Necchi for the opportunity to present some of the work we've been doing at the company, and go over some of the studies we've been doing. The title of my talk is Gene Expression and Molecular Subtyping in GU Cancers. This is unfortunately a prerecorded presentation as I was unable to attend in Milan this year. The primary assay we have for the bladder cancer community is Decipher Bladder clinical test. This is a molecular subtyping assay we've spent about five years developing. This is also known as a genomic subtyping classifier in the literature. It's really designed to do three different things, and that's to predict the risk of upstaging, the benefit from neoadjuvant chemotherapy, and finally to identify high-risk neuroendocrine-like disease. The way we do this is we break a patient's tumor into one of five different subtypes.

The model that we're using currently recognizes a luminal subtype; an infiltrated luminal subtype that has luminal characteristics but it has some stromal infiltration; a basal subtype; and then a basal claudin-low subtype which has a lot of immune infiltration; and finally, this aggressive neuroendocrine-like tumor subtype, which presents as conventional urothelial carcinoma, but however has a genomic profile and clinical behavior much more consistent with small cell or neuroendocrine-like disease. This on the right here is an example of one of the patient reports that an ordering physician would get.

We lump it into two larger categories, luminal and non-luminal, with neuroendocrine being a smaller subset. Each of these reports would be different, depending on the clinical stage of the patient in question. Here we have a classic clinical T2, so you get an indication of the level risk of upstaging, and the benefit from neoadjuvant chemotherapy. If this patient was a CT1, for example, we wouldn't offer the benefit of neoadjuvant chemotherapy block. Then conversely, if it was T3 or higher, we wouldn't offer a risk of upstaging because we already know they're not organ confined. Of course, given this is a whole transcriptome assay, any interested customer can actually order a new product that's coming out soon, which we call the Bladder Grid, which includes all the current subtyping models including the consensus model TCGA and the models.

First I think to go over each of these indications, the first being predicting risk of upstaging cystectomy. This is an area where we're really heavily focused on as we start to gain some additional traction in non-muscle invasive bladder cancer as well. This remains a large clinical problem predicting non-organ confined disease at surgery. Classically, this problem is at staging where we're doing a TUR or a CT scan, and then we're finding a patient is predicted to be a clinical T1 or T2. However, often this changes at pathology, and we're finding that 40 to 50% of the time this patient can actually be upgraded non-organ confined disease, which we're defining here as PT3 plus, and or having nodal involvement.

To investigate this problem and perhaps offer a solution using genomics, we came up with a study which we call the MOL study, and MOL just is a abbreviation for molecular upstaging. This is work done with Dr. Yair Lotan and UTS, Southwestern in Texas. This cohort was comprised of 206 patients, all clinical T1, T2 treated with direct radical cystectomy with no systemic therapy. You can see with upstaging, rates of mortality are quite high compared to no upstaging/ thinking more in the long term, but I think importantly, if the patient was upstaged to non-organ confined disease and it was thought to be organ confined, we did actually miss an opportunity to treat patients with neoadjuvant chemotherapy.

Using the same cohort and actually applying a molecular subtyping assay, the GSC, to this, we found that when we broke it down to luminal versus non-luminal, and so non-luminal here includes the basals, the claudin-lows, the infiltrated luminals, and neuroendocrine tumors, we saw a significant difference in the rates of upstaging between the luminal tumors and the non-luminal tumors where luminals had a rate of about 34% and non-luminals had a 51%. Interestingly, when the patient was found to be luminal and consistent with the organ confined versus non-organ confined CS on the previous plot, we see that luminal patients actually do significantly better than non-luminal patients, which is consistent with less aggressive biology we typically see with the luminal patients. Again, this is published with Dr. Lotan and European Urology in 2019.

The second indication we're looking at is really predicting NAC benefit. I do recognize there's some controversy around using subtyping to predict NAC. We have adjusted our approach since we initially came up with the assay in 2017, and look at this problem slightly differently. Regardless, the challenge point here really is, we're seeing with chemotherapy about a 40% response rate and a 5% benefit, which isn't all that impressive even though it was enough to move the needle and get it to guidelines in the US. It still remains a major decision point whether or not you would actually want to give chemotherapy to a lot of patients.
This is where we come to the pivot point really, is over time, I think in the literature and in our own work, we've seen the luminal molecular subtype does appear to have a much less aggressive clinical presentation in general, and this was reflected by good outcomes. You can see here in the cam plot on the right that this is published Seiler in European Urology in 2017, that the outcomes for luminal patients treated only with the radical cystectomy is actually quite good compared to other subtypes. When you combine this also with the data indicates that luminal tumors are more likely to be organ confined at surgery, and non-organ confined tumors have greater benefit from chemotherapy. We start to get this idea that perhaps patients with luminal tumors could be excluded from chemotherapy and have reasonable outcomes with cystectomy alone.

To get at this problem, since we really don't have the actual clinical trial that we need to properly address this, we developed a study we called NACmeta. To date, to my knowledge, this is probably one of the largest NAC versus non-NAC cohorts assembled to date. To do this, we combined four different cohorts, some treated with NAC, some with without. NAC I, NAC II are two in-house cohorts. We have both treated with neoadjuvant chemotherapy. CirGuidance was a clinical trial with Jos Formans in the Netherlands. Some patients received NAC, but the majority did not. MOL is an upstaging cohort we were working with, but we're only here using patients with actual pathological muscle invasive disease. Once we exclude a number of patients for various features, including past stage, after they just received non-cisplatin-based NAC, et cetera, we ended up with a large cohort of 601 patients, 247 receiving NAC, 354 not.

Using inverse probability weighting, we matched these cohort as best we could to our two groups as best we could for clinical parameters. So in the subsequent plots you're going to see some data that are percentages as opposed to action numbers, and that just reflects the balancing that we needed to do to make this work. The first checkpoint was really a sanity check to see what the data looks like when we look at NAC versus no NAC exclusively. Importantly, in the study we did exclude the neuroendocrine patients because they tend to have poor outcomes regardless of therapy, and we feel they represent an entirely different disease. When you look at the cohort as large, we see a net difference about 7% at three years for NAC versus no NAC, which is very much in line with the clinical trial data which reports about a 5% benefit at the same time.

Knowing this, we looked then again at luminal versus non-luminal, since this is our two groups. In this case, for the luminal patients we see that there is actually a negative difference of 2% in overall survival at three years, representing really no benefit or no change for these patients upon receiving NAC. This was consistent with UBA and MBA as well. Interestingly, we look at non-luminal patients, this is when we actually see a much different plot. In this case, we see a net difference about 10% in overall survival at three years. This was significant, and again, holds up on both UBA and MBA analysis. This work was published by Dr. Lotan again in Journal of Urology in 2022.

Taken together, we start to see the clinics and biology actually starting to align. When we're looking at non-organ confined disease, we see that they're more likely to be non-luminal in nature, have a much higher rate of upstaging, 10% benefit from the neoadjuvant chemo at three years. Conversely, at the luminal tumors, we see a much lower rate of upstaging at 34%, really no benefit of chemotherapy, even less benefit of chemotherapy at three years using the current data, suggesting that these patients really could be cured the cystectomy alone. This may be reflected in the data we're seeing now.

The next bit that we have is identifying a high-risk neuroendocrine-like disease. Again, these are the patients who presented with conventional urothelial carcinoma, but really have a disease more consistent with small cell or neuroendocrine-like tumors. I'm not going to go over the whole model development, this has being published previously, but this is a validation study we did. You could see that these neuroendocrine tumors really robustly expressed neuronal markers that are depleted for both basal and luminal markers. The outcomes for these patients is quite poor. In this study, we had about eight events in a year, and this is with systemic chemotherapy. They're doing quite poorly and we really think that they should be treated a little bit differently than with standard chemotherapy regimens.
This presents a large problem because although they only appear about 6% of the time, they would not be identified from routine pathological screening. So they really represent sort of the bad apple in the muscle invasive bladder cancer space, or the worst of the bad apples is in the muscle invasive bladder cancer space, and should really be treated differently. I think that NCCN current guidelines do recommend etoposide plus cisplatin for any tumors of neuroendocrine features. I would make the argument that neuroendocrine-like is a neuroendocrine feature and should be considered as such.

Finally, I think maybe this is the most exciting data and most recent data, is predicting benefit from neoadjuvant pembrolizumab in clinical cancer research where we're starting to see a very interesting and a robust pattern emerging. When we use the consensus classifier, TCGA classifier, we sort of get a nod of outcomes with not a lot of meaningful implication how we would use this data clinically, with the exception of course neuroendocrine or neuronal-like diseases doing quite poorly. But when you use the Decipher classifier, we see this very interesting pattern where the claudin-low tumors actually have almost an exquisite benefit with only a single patient having an event in three years. We found this to be significant after collecting the three-year long-term outcome data. We feel this is very exciting and speaks to the potential of actually stratifying patients correctly to neoadjuvant pembrolizumab when they present with muscle invasive bladder cancer.

To put it all together, really the question that remains, and there's a lot of work to be done still, but could 2020s be the decade of precision medicine for bladder cancer? With further validation, of course, the suggested therapy is based on the data we have so far, could be luminal patients having lower risk of upstaging, limited benefit of chemo, could be candidates for immediate cystectomy. Claudin-lows much higher risk of upstaging, maybe less benefit of neoadjuvant chemo, maybe great candidates for neoadjuvant pembro, given the data we have. Basal luminal infiltrated, again, continue back until we can figure out what to do with these patients. Neuroendocrine-like, again, should be treated as neuroendocrine disease. With that, that's the end of my presentation. I'm happy to answer questions via email, if that's the appropriate platform.