Combination Screens for Cancer Vulnerabilities: A Conversation with Kris Wood
, by Jim Hartley
A native of Kentucky and trained as a chemical engineer, Kris Wood is now an Assistant Professor in the departments of Biomedical Engineering and Pharmacology & Cancer Biology at Duke University. We spoke to him recently about his lab's approaches to finding weaknesses in cancer cells.
Please describe the rationale behind your recent paper on the "landscape of therapeutic cooperativity" in RAS-mutant cancers.
Sure. Our lab has been thinking about RAS in the context of resistance to therapy, and there are two sides to that coin. Of course we are well aware that there are no drugs available to target RAS directly in patients, and unfortunately it is also true that even though the RAF-MEK-ERK cascade downstream of mutant RAS seems to be the dominant driver of proliferation and survival, inhibition of that cascade is insufficient to provide robust therapeutic responses. So one flavor of the work we've been doing is asking, what other pathways in the cell, or what other signaling events, are supporting that intrinsic resistance, if I can use the term.
The flip side of our work is that RAS also plays a really important role in acquired resistance. In cancers that are driven by other cancer genes, JAK2 is one example that we've looked at and EGFR is another example, RAS mutations or hyperactivation of the RAS pathway often emerges as a dominant mechanism of acquired resistance to those therapies. So we've been trying to think about new ways of overcoming RAS-driven resistance that involve either the downstream pathways that are being turned on by RAS that support resistance, or the kind of collateral effects of RAS activation. Are there things that happen when RAS gets turned on in the setting of resistance that enable you to target some vulnerability that maybe you didn't know about before? So overall our studies primarily focus on RAS in the setting of resistance, intrinsic or acquired, and that's the overarching theme or lens through which we tend to view RAS.
So your "landscape" study was applicable to both intrinsic and acquired resistance?
Yes, that was the goal. It was known that if you block signaling through MEK or ERK in RAS-driven cancers that that can lead to feedback activation of receptor tyrosine kinases and the PI3 kinase/AKT pathway and that those effects will blunt the activity of the MEK/ERK inhibitors. What was not known was what the broader constellation of signaling pathways that could communicate with MEK and ERK looked like. And so we did a large number of CRISPR loss-of-function screens which essentially did a few things. Number one, they confirmed the RTK-PI3 kinase feedback loops that we knew about from the literature, and they helped us to realize that those are certainly very strong mechanisms of intrinsic resistance to MEK/ERK inhibition. But what our studies also showed was that there seems to be a fairly diverse constellation of signaling pathways that also specifically cooperate with MEK/ERK signaling to support RAS-driven cell growth and survival. And those include things like chromatin modifiers or direct activators of transcription like CDK7. We also found metabolic pathways that seem to specifically cooperate with MEK/ERK inhibition to support growth of these cells. And we also found a number of other signaling pathways, p38, SRC, and other survival-related pathways that seem to specifically cooperate with MEK/ERK inhibition.
And number two, our studies showed that the pathways that cooperate with MEK/ERK inhibition vary as a function of tissue of origin. They are different in KRAS-driven lung cancers than they are in KRAS-driven pancreas cancers, and they also vary in a manner that reflects secondary mutations in the tumors. In our paper we described one particular combination therapy, which is a SRC inhibitor / MEK inhibitor combination, that works in KRAS/PI3 kinase mutant colon cancers, but not in colon cancers that have only KRAS mutations or only have PI3 kinase mutations. You have to have both mutations. And so the landscape of combination therapies varies as a function of both tissue of origin and secondary mutations.
Can you describe why there seem to be so many vulnerabilities in KRAS-mutant cancers when you inhibit MEK/ERK signaling?
I think that biologically what that implies is that in order for MEK/ERK signaling to drive proliferation and survival in these cells, there have to be inputs from multiple aspects of the cell's biology. The cell cycle has to be intact, metabolic pathways to support proliferation have to be intact, certain anti-apoptotic mechanisms have to be intact, and the disruption of any one of those things creates a vulnerability to MEK inhibition. So I think that what this reflects is the idea that proliferation and survival are multifaceted processes that involve metabolic pathways, that involve cell cycle and cell division mechanisms, that involve inhibition of apoptosis and many other aspects of the cells' functions.
Why should these vulnerabilities be specific for a KRAS-mutant cells?
We find empirically that if we take, we haven't done this exhaustively, but when we have taken combination therapies that are synergistically active in KRAS-driven cells and we test them in KRAS-wild type cells, whether they're cancer cell lines or normal cell lines, we find that the synergy is typically much more pronounced in the KRAS-driven setting. But it's not obvious that it should always be that way or that it always has to always be that way. And so what we're trying to do in the lab right now is systematically credential these various combinations that arose from our screen with respect to their mechanistic biology, and their degree and breadth of activity, and their toxicities. In the context of the broader physiology of a living mammal, which pairs of nodes can you get rid of in a way that is tolerable? But I think the main significance of our study was that it taught us that there are lots of things that you can combine with MEK/ERK inhibitors that can give synergistic responses. You're not just stuck with PI3K/AKT or a few different RTKs.
Your study was a sort of triple synthetic lethal screen, correct? Mutant KRAS plus drug inhibition of MEK/ERK signaling plus gene knockouts?
Exactly. I think the reason we found a variety of vulnerabilities had to do both with the design of the screen and the tools we used. We used drugs to inhibit MEK or ERK because we wanted to be able to dial in precisely the amount of inhibition we got, at a level that the cells could still proliferate at a slow rate, so that we could get results from the screens. The downside is that drugs are promiscuous and maybe some of the cooperative effects that we observed in our screens could have been related to off-target activities of the MEK and ERK inhibitors, and I think we can't rule that out. But we did try to use drugs that are thought to be pretty selective, especially on the MEK inhibitor side, because the MEK inhibitors are allosteric. And we used the CRISPR loss of function system because it allowed us to very precisely interrogate these various signaling nodes and to do it at scale, across many cell lines. And so where an RNAi screen or a chemical screen typically gives you a number of hits, but you can really only believe those that you extensively validate, the CRISPR screen, even at the level of the primary screen, can allow you to reliably visualize this landscape simply from the primary data because of the high fidelity of these loss-of-function tools.
And I think the other thing that we benefitted from was instead of trying to go really deep on perturbations with a genome-scale screen, instead we made a relatively small library of only 2000 guide RNAs targeting about 400 key signaling genes. Then we used that small library to screen across many, many different cell lines, treated with various MEK, ERK, and PI3 kinase inhibitors. As a result of that we were able to gain insight into what the landscape of therapeutic cooperativity looks like, not only within a given cell line, but also across dozens of cell line/drug combinations, around 70 I think. And so I think when you combine the sort of high fidelity of the CRISPR system with a small customized library that allowed us to screen across many cell lines and drug treatments, it's the combination of those things that allowed us to get kind of a landscape view that otherwise might have been a little more difficult to achieve.
You also have a recent paper on codon bias in KRAS. How does that play into the establishment of cancers?
That study was quite serendipitous. We've been working for years on a technology we developed that allows us to quickly find what gain-of-function mutations, or activated signaling pathways, drive resistance to drugs, and two of the constructs that just happened to be in our library were a G12V mutant form of HRAS and a G12V mutant form of KRAS. We transduced our mutant pathway activating library into various cell lines and did hundreds of screens, and something like 96% of the time, HRAS would score as a hit, and virtually every time HRAS was the number 1 overall hit in the entire library. In sharp contrast, KRAS rarely scored as a hit in our screens. That, of course, was not what we expected given that KRAS drives so many more cancers than HRAS. We actually published the first paper with that library, but we worried that maybe our KRAS construct was somehow non-functional. However, we knew it was signaling correctly, and we knew the sequence was correct. We literally had a bit of an ethical crisis. I sat there and thought, "I want to exclude this data from the paper but I don't have a reason to do that." And so of course we included it. And of course some of the reviewers said "Why doesn't KRAS score, what's going on, is your construct broken?" And we had to write back and say "We don't know why." So this weird observation was sort of hanging out in the background in our lab, and within the next year I was at a retreat for my department and learned that one of my colleagues, Chris Counter, had been doing what I think is just fascinating work on the role of codon bias in RAS-driven tumorigenesis. His group had shown that KRAS, the most famous of the RAS oncogenes, seems to be the one that is most poorly expressed in its native genomic context, and on the other end of the spectrum HRAS, the least prevalent clinically, is the most highly expressed and most highly active form of the protein. And they were able to attribute some of this back to codon bias. And so we put two and two together and said well, maybe codon bias explains why our KRAS construct doesn't drive resistance very well but our HRAS construct does.
We next went back into the lab and we proved that, and that was all well and good, but up to that point in the study all we had was an intellectually satisfying explanation for our unexpected experimental data. But then we started thinking about the clinical contexts in which KRAS or RAS drives resistance, and probably the most famous example is in EGFR-driven colon cancers, which are treated with cetuximab and other EGFR inhibitory antibodies. And in that case we know clinically that RAS mutations almost always drive resistance, or at least very, very frequently. We teamed up with a colleague of mine, Alberto Bardelli in Torino in Italy, and we started asking whether this codon bias phenomenon might have important implications. And through that effort we ended up finding several examples that were quite interesting. We found, for example, that the phenomenon of codon bias likely explains why it is that patients who develop resistance to EGFR inhibiting antibodies develop the Q61 site mutations in KRAS much more frequently than patients with de novo disease. They're probably doing that because those mutant alleles signal more strongly, enabling them to overcome the effect of codon bias and effectively drive resistance.
Because they're so highly loaded with GTP?
Exactly. And the other thing that the study suggested to us is that when patients’ resistance is driven by a G12 or G13 site mutation, that that alone is not sufficient to drive resistance, you have to also find a way of overcoming the protein translation barrier imposed by codon bias. And so what we found in models of resistance that were driven by the G12 or G13 site KRAS mutations is that in addition to the KRAS mutations, they also have upregulated translation globally. As a result, those resistant cells are hypersensitive to drugs that inhibit protein translation, for example inhibitors of the mTOR/4EBP1 axis. So that was a really interesting example of a study that no one ever planned on doing, but that has interesting clinical implications.
And it's interesting too to think about how that model might fit in with the concept of oncogene-induced senescence. Chris Counter’s work suggests that if the early mutational event that happens in a cell is an HRAS mutation, at G12 or Q61, that protein's expressed really highly, you get massive signaling, and it's so much that you just go into senescence, and you never get a tumor from that. With KRAS you get this very mild signaling effect that keeps the cell alive so that it can incur additional mutations. Then once enough of the barriers to transformation have been overcome, then the cell can ratchet up KRAS expression to the level that it's actually sufficient can drive tumor growth. Chris' feeling I think is that you get KRAS mutations, those kind of hide out until enough other transformation barriers have been overcome. And once you have everything you need for a tumor, you start selecting for cells within that mass that have amplified the locus or otherwise found ways to express higher amounts of KRAS protein, because at that point cells can tolerate it.
You've also published on mitochondrial dynamics in the context of mutant KRAS. How did you get there?
There is a part of my lab that is interested in trying to discover novel vulnerabilities in tumor types like high-grade serous ovarian cancers that are difficult to target because they have very few recurrent mutations other than p53. One interesting characteristic of the cancer cells from these tumors is the presence of dysregulated mitochondrial fission-fusion dynamics. And briefly what we know about mitochondrial fission and fusion is that in all the cells in our bodies mitochondria are not static organelles, but rather they're constantly undergoing processes of fission, where a single mitochondrion will break into multiple smaller mitochondria, or fusion, which is the opposite, where smaller mitochondria will fuse into a larger mitochondrion. And it turns out that mitochondria that are more fused have different energetic, biosynthetic, and survival regulating properties than those that are fizzed. The idea that is emerging in the cell biology community is that by shifting mitochondrial dynamics more toward fusion, or more towards fission, a cell can exert control over its metabolism and its proliferation and survival, because so many important metabolic, biosynthetic, and survival functions are localized in that organelle. There's an interesting body of work that's being led by a number of people, David Chan at CalTech is one that comes to mind, on the basic biology of mitochondrial dynamics.
With that said, we observed that in cancers there frequently exists dysregulation of mitochondrial dynamics. So you have mitochondria in tumors that often times don't look or act like mitochondria in normal cells; they're either way more fused or way more fragmented. And if you go into the TCGA you can actually see that some of the genes that canonically drive fusion or fission are amplified recurrently in certain tumor types. So it makes sense that if altering fission-fusion dynamics can alter cell survival and growth, that of course cancers would want to coopt that. This is the sort of emerging picture in the literature. And so we asked if there could be vulnerabilities that result from that, from dysregulating mitochondrial morphology. We took ovarian cancer cells and we made genetic manipulations to those cells that forced their mitochondria to either be highly fused or highly fragmented, by manipulating some of the canonical fission-fusion proteins. And once we established that these cells had the appropriate mitochondrial structural changes, we started doing drug screens, comparing the effects of thousands of different drugs on the parental line versus isogenic lines in which we made the mitochondria highly fused or highly fragmented.
We found that there are classes of drugs that seemed to have much more activity if you have either more fused mitochondria or more fragmented mitochondria. And perhaps not surprisingly, some of the drugs that were most active in these screens were drugs that target the mitochondrial apoptosis network. That kind of makes sense because mitochondrial apoptosis is governed by things that are happening right at the mitochondrial membrane, and if you are dramatically altering the morphology of those membranes you might be altering apoptotic proclivity. And so we ended up writing a story about one particular set of drugs, the SMAC mimetics, that seem have highly differential activity when you have altered mitochondrial dynamics.
So that's basically where we're going.
Now how does that relate to KRAS? Chris Counter and Dave Kashatus showed a few years ago that KRAS directly regulates mitochondrial dynamics, by driving the phosphorylation of a protein called DRP1, which is a protein that helps to execute mitochondrial fission. The biological reason for this makes sense, because altering mitochondrial dynamics can make cells grow better, and survive in different environments better. We showed that in KRAS-driven cancer cells you can add an ERK inhibitor to dysregulate mitochondrial dynamics, and as a result the cells become hypersensitive to SMAC mimetics. You can do the same thing in BRAF-driven cells and EGFR-driven cells and so on if you treat with the appropriate targeted therapies. And what's interesting is this cooperation that we're seeing, this sensitization to SMAC mimetics, is not because of some general effect, it's specifically being caused by the targeted therapy-induced changes in mitochondrial dynamics.
Can you tell us about your path to research in cancer biology?
Well, I grew up in a rural town in Kentucky, on a farm. I really loved math growing up, and as a freshmen at the University of Kentucky, 30 miles down the road from my hometown, I took a chemistry course. I really liked the course, and I eventually decided to study chemical engineering, simply because I liked chemistry and I liked math and it seemed like a reasonable combination. That got me into science from a formal coursework perspective. But the really transformative thing that happened to me was that I started doing research in a lab. My thermodynamics professor my sophomore year stopped me after class one day and said I think you should be doing research, and I would like to invite you to come to my lab and work. And so I said yes, and for two years he and I worked together. That was really the transformative thing for me because I realized that I really loved research, I loved the idea that I could apply analytical thinking to try to invent what the future would look like.
After college I moved Boston and went to MIT, where I got my PhD in chemical engineering, and became more and more interested in biomedical research, albeit through an engineer's lens. I did my PhD with two people, Paula Hammond, who's a very well-known polymer scientist, and Bob Langer, who's sort of the founder of biomaterials and drug delivery. When I finished my PhD work I wanted to continue working on biomedical problems, but I recognized that I didn't know enough biology. Around that time, I met a young professor at MIT, David Sabatini, and David and I struck up a friendship and he was nice enough to invite me to come to his lab as a postdoc. I always tell people that at my thesis defense my thesis committee asked me what I was going to do after graduation, and I said I'm going to do a post doc with David Sabatini. And one of my committee members said, "Oh, David Sabatini, you must be interested in mTOR." And I said, "What's mTOR?" And he said, "It's a kinase." And I replied, "What's a kinase?" So I really didn't know any biology at the time. But I went to David's lab and I jumped in the deep end, and I really got the bug for doing basic biology. I've been at it ever since.