Functional Precision Medicine: Testing Cancer Drugs Before Treatment

•Do you have any comments on functional precision testing?

•The value of functional precision testing seems intuitive, yet it’s not widely practiced. What are the barriers or objections to it? The information and opinions expressed on this website or platform, or during discussions and presentations (both verbal and written) are not intended as health care recommendations or medical advice by Cancer Patient Lab/Prostate Cancer Lab, its principals, presenters, participants, or representatives for any medical treatment, product, or course of action. You should always consult a doctor about your specific situation before pursuing any health care program, treatment, product or other course of action that might affect your health. — Meeting Notes Brad Power: We're very honored to have Tony Letai of Dana Farber with us. This is very timely. We've been having a conversation over the last weeks about combination therapies with Ally Perlina of CureMatch and Bob Gatenby of Moffitt Cancer Center. Bob Gatenby adds other treatment strategy considerations that combinatorially make having evidence from randomized clinical trials pretty problematic for some of the more personalized strategies.Tony has ideas about functional testing that might provide a way out of this conundrum of personalized treatments that include dosing and sequencing of treatments and so little or no evidence from the batches of patients that are required for randomized clinical trials. Tony Letai: I'm a clinician, but I mainly run a cancer biology lab at Dana Farber. I've been at Dana Farber since roughly 1995 when I was an intern, and I have been there ever since. I'm going to talk about an alternative way of personalizing therapy or choosing the right drug for the right patient. We can have a very open-ended conversation, which can certainly include how you can incorporate this into making novel therapy combinations. There's always going to be a definite tension between personalizing therapy and assembling novel combinations, which we know are going to be important for long term disease-free survival in a lot of cases on the one hand, and on the other hand verifying safety ahead of time. We can't solve everything, but I acknowledge this tension. This is usually how this goes. This is all about getting information to choose a therapy for the right patient. I like this metaphor, so I'm going to present it to you. I find it analogous to this question. Say you're walking along the street with a six-year-old kid. Maybe it's your niece, your nephew, or your grandkid. I see this little dog, and you want to know, what would happen if I poked that dog with a stick? How would the dog respond? Would it run away? Would it bark? Would it wet itself? Would it bite me? What on earth would it do? How do we answer that question? If you're a conventional omic person in today's cancer biology world, it's an analogous problem where they're trying to figure out, using genomics, how a complicated system – in their case a tumor, in this case a little dog – how it's going to respond to a perturbation – that's treatment with a drug or some other sort of treatment. The way omics or modern cancer genomics addresses this problem is: the first thing you do is you kill the dog. Then you sequence the DNA. Or now we're getting more mature: you're sequencing the RNA. You're looking at proteins, metabolites, all these, what I would call static phenomena associated with the tumor at time equals zero – initial condition measurements. You get big, big data sets. These enormous data sets that people like this gentleman live to analyze. They're often housed at big universities and they take these enormous data sets and do all sorts of classifications and whatnot. And every now and then you actually spit out a suggestion for whether or not a drug works or not; not always, but sometimes you do. What is an alternative to this approach? I'll summarize what I think the success of this kind of approach is in a little bit, but an alternative approach, if you want to know what happens, if you poke this dog with a stick, how would a six-year-old solve this problem? The six-year-old would poke the dog with a stick and then watch what happens. I am a firm believer that we do not do enough of that in cancer biology. Of course, what I'm going to get to with regard to poking it with a stick is actually taking the patient's living tumor cells, exposing them to drugs, and seeing what happens. Why don't we do more of that? You might ask why we call this functional precision medicine. The omics approach looks purely at static biomarkers. I think of this as looking at dynamic biomarkers. I was originally trained as a physicist. One thing we know is that you can learn a lot from a complicated system by making perturbations to the system, rather than just measuring initial conditions. But there's already a way to solve it, right? If you've listened to anything from the NCI, from big cancer centers around the world, including my own at Dana Farber, this problem of choosing drugs for patients, we've kind of got it licked, right? We know that we do precision medicine: we take the patient's tumor, we sequence it, and it's going to spit out a mutation that tells us to use a particular drug. As you can see, in this famous graphic the NCI uses, this is how you do precision medicine. You see it's very persuasive because there's this DNA chain. You can see for these different patients, there's a blue asterisk or purple asterisk there, then a green dot there, and an orange dot there, all these different parts on a DNA double helix. You could see if you look at all the patients, they're all served by this. Everyone gets a drug out of all this. This approach was tested in the NCI-MATCH trial. You may have heard that they took people with advanced, solid tumors, all across the US, and did a genomic test, some kind of sequencing, to find mutations, then assigned them to different treatment arms. It started out with around 10 treatment arms, and then grew to something like 30 as new knowledge increased. It's worth considering: How well did this work? This was the National Cancer Institute using a general approach for advanced cancer patients doing the sequencing that we've been told is the way to do precision medicine. Let's talk about what we mean by “work''. What I mean by “work'' is: how many patients got a drug that helped them? I could tell you if you look at a lot of the publicly available information from the NCI- MATCH trial – and I encourage you to look into this yourselves – you won't find a lot of information about how many drugs were found that actually helped patients. You find a lot of information about logistics, how many people were sequenced, how many sequences there were, and so forth, but not so much about patient response. But I'll show you what I know and what I could find in the publicly available information. As you see, they enroll big numbers. They were able to sequence like 7,000 people. But what's really important to recognize is how many of those patients after sequencing made it to a treatment arm. That is how many patients had a mutation identified that actually was relatable to an arm that had a drug assigned to it. Only 16% of the patients even made it to a treatment arm of these 30 treatment arms. How did those patients do? Well? It depends on the treatment arm, and for the most part, there's no publicly available information on what I would consider the most important result of this trial. Maybe it'll come out eventually, but here's what we do know. Here's maybe about 10 different treatment arms. You can see they have different letters as they sort of advanced through the alphabet and then needed more. The “ORR” is the overall response rate. That's an objective response rate based on shrinkage of the patient's tumors. This means a patient went to a group of patients who had a mutation identified, went to a treatment arm, and then got a drug. If you look at these numbers, you might have anticipated that number would be a lot higher, based on what we know about genomically- directed precision medicine. Recognize that this overall response rate is only for the 16% that even made it to a treatment arm. My point in presenting this is not to tell you that genomics is useless. In a lot of patients it can be very helpful, but the ceiling on how many patients actually benefit from this is lower than I think many of us consider. And if you take in all cancer patients getting genomic tests, it's not that nobody benefits. Some people certainly benefit. I can list EGFR-mutant lung cancers, or BRAF- mutant tumors, whether it's melanoma or colon cancer or something like that, or patients with TRK mutations in a wide variety of diseases, that certainly benefit, but that's not most cancer patients. In fact, we're probably down in the single digits percents for patients who actually benefit from these genomic analyses. I'm saying this just to indicate that, yes, there is room for improvement. There is a need for additional methods to assign drugs to patients. If we rely purely on genomics, we're tying one hand behind our back. It's not a useless test, but it definitely needs help. So what's my proposal? Something that we talk about called cancer functional precision medicine. It's a very simple idea. You take a drug, you put it on a bunch of tumor cells, and then you figure out a smart assay to see what happens. This is what we use even today as the gold standard in microbiology, except for in that case, you are usually trying to kill a bacterium. If you have a blood infection,urine infection, or pneumonia, your doctor will try to culture the bacteria causing that infection. We'll grow it in a Petri dish, really fast, treat it with all the drugs, and let's see which one works best. And that's how we choose. That's still the gold standard for what the best antibiotic is. Today, in 2022, you might ask, this is so common sense: Why don't we do this in cancer? People did try ideas like this. The heyday was something like 30 years ago. There were so- called ex-vivo chemosensitivity assays, and most of the leaders in the cancer biology world decided they weren't ready for prime time. Back then there were very few drugs to choose among. We had maybe cisplatin and something else. It wasn't very useful information to know whether or not you could receive cisplatin if there wasn't a great alternative available to you. Culture conditions by today's standards were exceedingly rudimentary. Very often what happened over these extended cultures, the tumors were overgrown with normal fibroblasts and the assay provided no information. All the assays available back then, what we call bulk assays, meaning there was no single cell information obtained. You had this heterogeneous mix of God knows what, and you'd get some readout that had some metabolic value that wasn't clearly associated with tumor cell death, maybe vaguely correlated. There were a lot of problems with every step here. We had very few drugs, it was very difficult to culture the tumors in a way that was informative, and we didn't have good assays. 30 years have passed, and all of those things have gotten way better. We have many, many drugs, we have much better ways of culturing tumor cells, and we have much better ways of analyzing them. Brad Power: What is the correlation between test results and clinical results? How has that been progressing? As you said, maybe it wasn't so good before, and maybe it's better now? Tony Letai: I'll give you some examples, and also some challenges. It is very hard to answer sometimes. There is a property of cells called “apoptosis”. Every cell in our body has built within it a suicide mechanism so that that cell can decide to kill itself. The cells are programmed to do this. If they go bad, if they're doing something they shouldn't, like growing in a way they shouldn't, or metabolizing something in a way they shouldn't, they can activate apoptosis to eliminate themselves. It's good for the body as a whole and virtually every multicellular organism that's around that it has this process in itself. It's a threshold effect. Once it gets triggered, the cell rapidly commits to cell death within minutes. I like to use a cliff metaphor to model it. You can imagine that there are some cells that are really close to the edge of the cliff. And there are other cells that are really far away from the cliff. It turns out we've shown that the cells that are really close to the edge of the cliff are really sensitive to chemotherapy and are easy to kill by our treatments. And the ones that are further from the edge of the cliff are harder to kill. They might endure the chemotherapy and then live to fight another day. We wanted to figure out ways that we could selectively provoke this type of signaling in cancer cells to move cells from far from the edge of the cliff to close to the edge of the cliff, what we call “priming them for apoptosis”. That required an understanding of the apoptotic pathways. All of your cells have mitochondria. You may be familiar with them as power plants of the cell. For our purposes, they exist as the decision mechanism around whether or not a cell progresses to cell death. And that involves the membrane around the mitochondria. And if that permeabilizes, it releases a bunch of factors that causes the cell to chop up its DNA, chop up its protein, and even tag itself with signals so that it gets eaten by neighboring cells. That's regulated by this complicated protein family called the BCL2 family. There are some proteins in that family that cause this permeabilization event. Those are the ones we call “pro-apoptotic”. And there are other proteins that oppose this permeabilization event. Those are the ones that we call “anti-apoptotic''. They function by binding up the pro-apoptotic ones so that they can't make the holes in the membrane. Using knowledge of this pathway, we developed a way to measure how close the cell is to the threshold of apoptosis. We call it “BH3 profiling”. The way we do this all takes place at the mitochondria. We probe the mitochondria of the cells, usually cancer cells, but you can do it on normal cells, any cells you want. We probe them with these synthetic things we call “BH3 peptides”. These are parts of the pro-apoptotic molecules I just told you about. If we need to put a lot of these BH3 peptides on a mitochondrion, then we know it's very far from the cliff's edge, and it's very unprimed for apoptosis, in order to make it permealizable. If we only need to add a little bit of the BH3 peptides to make it permeabilized, then we know it's very close to the threshold of apoptosis. I may use this term “MOMP” once or twice. That's “mitochondrial outer membrane permeabilization”. That's what we measure in our assay. If you need a little bit of BH3 peptide to cause permeabilization, then it's pretty primed, if you don't need a lot, it's pretty unprimed.