Precision Medicine and Advanced Bladder Cancer: Where Are We Today?

What I would like to do in the next 20 minutes or so is give you a bit of a vignette, an overview on precision medicine or personalized medicine in advanced bladder cancer, and again it’s a massive field, personalized medicine is a massive field, and that’s why I’m saying I’m going to give you vignettes. I don’t have any extra conflicts of interest since an hour ago.

The learning objectives; basically go over the principles of precision medicine and I guess it’s nice that you just heard the previous talk because some of the things I want to mention now are going to be prefaced on the previous talk, so I’m not going to repeat them. The hope precision medicine can impact urologic oncology through a couple of examples, and then some of the caveats and challenges.

This is basically an evolution in medicine, and all the way from empiric medicine where we had basically the physician with minimal tools, stethoscope here, we went over to stratified medicine, and stratified medicine is the bulk of the medicine we practice today, which is stratified by virtue of groups. You’ve seen almost every talk has had a Kaplan Meier curve, and that’s essentially stratified medicine. Precision medicine, because of the advent of the gene, the Human Genome Project driven in great part by this gentleman here, the NIH Director Francis Collins, basically brought us to the era of personalized or precision medicine, which is truly individualized medicine on a patient level, so the right patient, the right treatment, at the right time. Those are really the three, the triad of precision medicine.

You’ve seen these boxes in the previous slide, and I just wanted to lay them out with a concrete example. These various markers where you have the predisposition markers, which are really the SNPs, telling you basically risks, and again there are other markers of risk. Some are not molecular, like do you smoke or you don’t smoke, in terms of bladder cancer for example. So, that identifies the case if you will.

And then you have prognostic markers, basically is the disease aggressiveness. That’s what a prognostic marker is, and by knowing the disease aggressiveness it tells you, it answers the question of who do you want to treat, and treat in sort of a very large sort of definition of treat. That’s means more aggressive or less aggressive follow up. Okay? I think of that as the treatment if you will.

Low risk/high risk if you go into the high-risk element or grouping then you basically invoke another set of markers, which are predictive markers. Predictive markers are basically therapeutic response markers. And those markers can be what commonly are now more and more companion biomarkers to drugs. So, those companion biomarkers can either look at the mutation that a drug is targeted against, like you saw the Gleevec idea, or could actually look at genes that are unrelated to the actual target of the therapy, but these genes inform or report on the biological capability of the cell to resist the treatment. So, there are different types and I’ll give you some examples of those.

And finally, you have a subset of the predictive markers, which are targeted with companion diagnostics to tell you that the target that you’re driving your drug against is present in Mrs. Smith.

This is a calculation showing that the informed response approach leads theoretically to a better treatment success because you’re basically only selecting patients that will respond to the therapy, and therefore your treatment response is enhanced. The idea being the ones that are not predicted as responders go onto other therapies or to clinical trials. So, in theory then, in practice this is really the pathway idealized for precision medicine.

What are the challenges in bladder cancer specifically? Well, to improve primary diagnosis by identifying highly at risk patients, surveillance of non-muscle invasive disease, in other words predicting recurrence and progression, and for invasive disease is really reducing the mortality. That’s what you want for invasive disease. And you could identify patients with nodal metastases that are occult and treat them, and I’ll show you an example of that. High-risk patients that are also responders because it’s nice to know you’re high risk, but if you can’t do anything about it it’s not so good. So, predictive markers for either chemotherapy, targeted therapy, or immunotherapy, or new targets and new drugs, and I’ll give you vignettes of all of these.

These are some of the example. I gave you the GWAS example for bladder earlier on in the prior talk, FGFR3 as well in the previous talk. I’m going to show you this example of identifying nodal metastases with a gene signature as well as predictive markers for chemo sensitivity and how a new therapeutic avenue was developed using basic science principles of drug discovery.

Neoadjuvant chemotherapy is definitely of benefit in patients with muscle invasive disease. The problem is that it’s not been widely adopted in the community, and that’s primarily because we can’t really stratify very well responders and non-responders. In addition, we don’t know before cystectomy which patients have occult nodal disease, which by most criteria would get adjuvant chemotherapy in most cases. The idea is then can we develop prognostic markers that will tell us which patients have nodal disease occultly, which means they have normal CTs etcetera at cystectomy, but have occult disease, which is about 25%, or predictive markers of those that would get chemotherapy which of those are likely to benefit from chemotherapy with the idea that by focusing the patients that mostly would benefit from chemotherapy would have better adoption of the approach.

This is a study that was done that identified using gene arrays that I showed you earlier, a gene set, a model using a discovery set in a Canadian cohort in a validation or test set from Memorial Sloan Kettering, and came up with a 20-gene signature that predicted occult nodal disease. This signature had an ROC curve here, and this ROC curve, just to give you an example, is similar to the one for Oncotype DX for breast cancer. So, then when you take this information you would stratify into various groups of high risk versus low risk, and what we did is we validated this on a completely independent dataset from Europe, from actually a prospective clinical trial, and the validation is shown here. The idea is now to develop this clinically for in practice to identify patients that have node-positive disease and therefore potentially could benefit from chemotherapy.

The other thing that our laboratory has done is really develop the COXEN principle, and this is a cartoon here. This is a fairly complicated idea but the overarching, the 30,000-foot sort of perspective on it is quite simple. So, what it is is basically a Rosetta Stone between the genetic signatures of cells growing in culture to a human tumor. And by doing this Rosetta Stone we can test cells in culture with drugs, and figure out by joining with the patient tumor what the patient is sensitive to. So, we did this and we looked at it in multiple data sets in multiple cancers, not just bladder, and this is a typical example from one of the studies showing that the genes that were picked by the computers as stratifying patients as sensitive and resistant to in this case chemotherapy did stratify actual patients. And this was done retrospectively, like I said in multiple tumors and tumor types, including bladder.

But there is now a SWOG trial, S1314 that many of you, many folks here in the audience are involved, and is led by Tom Flaig at the University of Colorado, and I’m the translational science person that is going to evaluate the gene signatures. So, if this works then this is a technology that can be applied to virtually any cancer that we see a drug signal in culture. So, hopefully that will be impacting significantly across the field.

This is the design of the trial. The patients are coming in here. They’re randomized actually to two of the most common chemotherapy regimens, MVAC in dose dense variety, and gem-cis, and we’re going to evaluate the COXEN biomarkers, the endpoint being the P0 rate. In addition, there’s going to be a molecular analysis done by other groups on these samples that are obviously going to be very valuable because they’re going to be out of a clinical trial. Even if COXEN does not work there is a tremendous repository here of tissues well characterized from a prospective trial that could be used to develop other biomarkers and predictive biomarkers.

And now let me shift gears and give you some examples of things that have been translated if you will from the genomic information that we’ve acquired over the past five years from a number of articles and studies that did next generation sequencing of bladder cancer. As you can imagine, this has been a trove of findings, and I’ll give you just one example here. This is a study that has looked at next generation sequencing in patients, but also looked at there’s a, let me step back for a second. There’s a couple of times of next generation sequencing. There is whole genome, which looks at everything basically DNA wise, and then there’s exome sequencing that only looks at exomes. This is a study that combined exome sequencing to actually looking at the telomerase promoter mutations. So, this is also called TERT, and this is a gene that is involved. It’s more than one but this is the major gene that’s involved in maintaining cellular immortality. So, this is why a cancer cell, this is one of the genes or one of the major genes why a cancer cell does not die. And so, the reason why I have this cartoon here is this is the classic hallmark of cancer modification or the latest version of it by Hanahan and Weinberg, and interestingly enough, as you can see maybe not at the back but you’ll take my word for it, that many of the genes that are altered in bladder cancer are very intrinsically related to the hallmarks of cancer idea, so telomerase being one of them. Bladder cancer turns out has the highest incidence of telomerase promoter mutations of any cancer, or close to, among the top. This is to say that a lot of the targets are likely going to be relevant and effective in therapy.

Let me show you a couple of examples. So, therapeutic targets; one of the things that many of the studies, this is from the TCGA study, one of the things that has come up in almost every single sequencing study in bladder cancer is a very significant prominence of cell cycle genes that were altered at the DNA level, mutated. So, this is an example of the pathways that are involved in this, and the good news is is that there are a number of inhibitors that could be targeting these pathways. Now, I won’t get into the specificity issues, and obviously cell cycle there are more than just the cancer that has a cell cycle in your body, so there are issues. But nevertheless, there are inhibitors potentially for these pathways. Here there are MDM2 inhibitors, a number of them.

The other thing that came completely in parallel to these findings is a study that we did looking at basically the genes that are biomarkers of progression across multiple cancers, and this is basically focusing on bladder cancer and other tobacco-related cancers. And the thing that came out among those three tobacco-related cancers most prominently was the cell cycle. And so, a 15-gene model was constructed and it turns out that when you looked at it in to validate its ability to stratify patients in four different studies in bladder cancer the cell cycle gene signature, just like the nodal signature I showed you earlier, the same idea, was able to stratify multiple datasets. Again, very consistent and supporting the idea that cell cycle is really important in bladder cancer.

The other thing that came out very importantly in bladder cancer sequencing was the involvement or the deregulation if you will of the receptor tyrosine kinase, Ras, and PI3 kinase pathways that are seen here. And again, there are a number of inhibitors that can be, or are currently either in trial or in development against most of these pathways. The big one that’s not been yet inhibited, but you’ll see there’s avenues that could be dealt with in this, is the RAS pathway. But you have inhibitors for the others.

Here is one example that I’m going to show you that highlights the value of identifying the molecular lesions in clinical studies. And this is this paper from Sloan Kettering, and what they did is they really identify, they really looked at patients that were in a phase II trial of an mTOR inhibitor, one of the pathways in the box earlier, and looked at patients that were very, very responsive in those patients, in that trial. And although, the really fascinating thing here of this paper that it showed that if you look at molecular, if you will molecular characterization of the responders and nonresponders that gives you a lot of information, even though the trial is actually negative. This was a negative trial. But even though it was overall negative what it gave the insight of was that they figured out that certain mutations were tracking with the response, highlighting again the need to do molecular investigation on all clinical trials.

Another example of how this is useful is this example. So, we know that the RAS pathway is important in bladder cancer, but the problem with RAS is that it has three canonical downstream effectors. PI3 kinase/MEK and then RAL, and the problem is that only two of those pathways are blocked, and RAS is also highly relevant in a number of the other cancers, pancreas being one of them, colon, and lung cancer, KRAS mutant lung cancer and colon cancer are particularly adverse molecular lesions in terms of prognosis.

It was interesting to sort of contemplate what to do about this problem, so we actually undertook this challenge and what we did is instead of trying to drug RAS itself what we thought we would do is try to shut it down by drugging its effectors, two have already been drugged but we wanted to go after RAL. And what RAL is is a GTPase, and although the nice thing about RAL is that it’s not mutated in cancer, but it’s active in cancer because of RAS being above it it’s really active. And the way RAL works is that as a GTPase, which is a molecular switch, all GTPase, there are about 60 of them, RAS being one of them, they’re molecular switches. They go back and forth, back and forth, and so the idea was instead of trying to block the active form of the molecule what we would do is we would try to block the inactive form. People tried to block the active form unsuccessful, so we tried a different avenue, and the idea then was to take a small molecule and try to block it by binding to the inactive form because the cycle is nonstop if you block the inactive form eventually you’re going to run out of activation molecules. So, you’re going to run out of this thing, and therefore you’re never going to be able to activate it. That was the premise. But the hypothesis of this was that the molecule in the inactive form was basically like an open Pacman, Pacman with the open mouth and the active form was the closed Pacman. So, if that hypothesis was not true this would have not worked. Okay?

The good news is is that after I was able to convince our structural biologist to look at the structure of RAL in the inactive and active form. We were able to actually find such a pocket, and that was the part of science where it’s good to have a good idea, but unless you get a little lucky it’s not going to work. So, there is the pocket that then collapses, and then what we did is we screen a million compounds using super computers to find different molecules that go into this pocket, but then bounce off when the pocket is closed. And to make a long story short, we had various stratifications here in terms of screening. We went from computerized a million to 88, the best hits, then we did biochemical assays, cell biology assays, in vitro and cell culture. And then in animals RAL comes in two forms, they’re both shown here. We actually as of last week were able to crystalize this form, but this is a threaded, this is a slide from a month ago, so this is actually a computer generated threaded structure of RAL. But the reason why I’m showing those is because both RALs are bad, and we did a lot of the screening and work on RalA, but it turns out that the drug also inhibits RalB, and we knew that biochemically, but structurally now we know the same site is present in RalB because you really want to knock out both of them.

This is a summary of the work showing that the drug itself is tumor static, and it turns out that when we look at it in a metastatic setting it is a lot more powerful in the metastatic setting, which is not, which is consistent with the idea of RAL driving most of the metastatic disease, advanced disease as opposed to primary tumors.

This has been licensed to a company, and what we hope to do is whether this being, this is a cystectomy but it could be Whipple, or it could be a lobectomy, etcetera, for lung cancer. Those are going to be likely the two major cancers that we’re going to study once the drug comes to clinical use. In trials we’re going to stratify them by immunohistochemistry with RAL, and then gene expression signature that I didn’t show you, but we have a signature that looks at RAL activity, and then enroll patients to either standard of care or one of the molecules called RBC8, and this is a fictitious Kaplan Meier curve that I’m hoping will happen. But anyway, it’s a pretty exciting story.

And before–I’m going to give you a little of the negative of precision medicine and some of the caveats now. Obviously, it’s the wave of the future, but there are some pitfalls of course and potholes. P53 is a pitfall and a pothole that some of you in this room know very well. Seth Lerner was very much involved with this study, and this was a study that was premised on a very, very, very compelling observation out of USC in 1991, and here it is. That p53 is a very powerful prognostic marker of progression after cystectomy and of chemosensitivity. So, Seth and others put together an international trial with an enormous amount of effort. I think it took eight or nine years to accrue, and unfortunately despite this the trial itself, the results were like this. What this shows you is that, and one of my friends, George Sledge, is a head of medical oncology at Stanford, he wrote a very nice article recently about the ten rules of cancer doctors. One of those rules are nobody is smarter than a phase III trial. I like that rule. And this is nobody is smarter than a phase III trial because don’t let retrospective data, this is why we’re doing a COXEN trial, a prospective trial, because it doesn’t matter what the retrospective data. The retrospective data only tells you this could work. So, this is a cautionary tale here using retrospective tools or studies to validate biomarkers.

Finally, the challenges in implementing personalized and precision medicine there are a lot. Obviously, there are regulatory ones, there’s insurance issues, there’s a lot of issues that need to be taken care of. Obviously, for us that are on the science side it’s fine to discover all sorts of cool things, but eventually you want to get them to people, to help people. So, there are a lot of other things that have to happen, and clearly no single person can push that through.

So, with that I thank you for your attention.