Conversation with Dr. Shiva Malek and Dr. Christiaan Klijn

December 6, 2018, by Jim Hartley

Hartley: The contraption on the cover of Cancer Cell that contained your article refers to some of your findings, correct?

Klijn: The concept for the cover was created by Allison Bruce who’s a graphics illustrator, we have extensive experience with her and she’s great. She thought of the RAS pathway that we were examining as a Rube Goldberg machine, and she has captured the basic function of the RAS pathway and the feedback mechanism that we and many others have described by having RAS and RAF be a series of gutters and the gray lines represent the paths of the balls without inhibitors. The balls from the RAS and RAF gutters fall down onto the MEK trampoline, and then bounce into the ERK receptacle, where they are deposited into a bucket that then raises the RAS gutter so the balls stop flowing. One of the points of the paper is that MEK inhibition alone, represented by the bat, is not sufficient to stop the pathway, since when you introduce the bat into the system you hit the balls but they bounce back on the wall back up to the RAS gutter and basically keep the signal flowing. Experimentally we see this as an increase in phospho-MEK and RAS-GTP in the presence of a MEK inhibitor. And the introduction of the RAF inhibitor, which is in this case the mitt, makes sure that the balls that get hit out of the system by MEK inhibitor are taken completely out, by depositing them into the funnel down below. So the presence of both MEK and RAF inhibitors basically causes the system to be completely blocked off.

Malek: Allison did a really great job, because the Rube Goldberg at its heart is a system that you build that does something very simple using a very over-complicated process. And I thought that was kind of central to this, because the MAP kinase pathway is meant to do something very simple, regulate cell growth, regulate cell proliferation. But it is highly complicated, it’s convoluted which is what this Rube Goldberg illustrates, and to interrogate this pathway with small molecule inhibitors or think of therapeutically approaching this pathway, I think we have to consider the complexity of the signaling when we’re thinking about it.

Hartley: Is it your view that the increase in BRAF-CRAF dimers that is driven by [the MEK inhibitor] cobimetinib is driven only by increased RAS-GTP?

Malek: That’s the clear link that we have, and what we found interesting in this, because there are several downstream effector pathways, including PI3K, including RAL-GDS. We asked ourselves this very question: Why is it that when you elevate RAS-GTP levels, why aren’t we seeing signaling through AKT or any of the downstream effector pathways, why is it that we’re seeing this effect of B- and CRAF. Several thoughts about that. First, it might be context dependent, so maybe in the context we looked at we happened to be looking in places where it’s primarily rewiring the pathway to depend on BRAF and CRAF dimerization. The other thought I had is if you look in the literature there is published data showing that there is a preference for RAS-GTP to interact with RAF, it’s about an order of magnitude greater affinity, versus PI3K. And so the other hypothesis was, maybe this is essentially a biochemical preference that RAS-GTP has to interact with RAF dimers versus activate PI3K, and perhaps that’s one of the reasons we’re seeing what we’re seeing. But we don’t know exactly why this pathway, why if it’s really about RAS-GTP you would imagine any of the effector pathways could be impacted. And certainly I think the PI3K pathway, AKT signaling plays a role, but even in our chemical biological screen that we did, the synergy we saw with RAF and MEK inhibitors was far greater than anything else.

Hartley: Early in the Discussion you mentioned the possibile role of a kinase-independent function of CRAF.

Malek: We are obviously actively working on understanding kinase-independent functions of CRAF. Mariano Barbacid’s paper that was published earlier this year really showed nicely that in the KRAS-driven GEMMs model in an intervention setting that CRAF depletion inhibited tumor growth. And it’s nicely supported, the earlier papers from Dave Tuveson and Mariano in a tumor initiation context. What we think is going on is, we think that in the presence of MEK inhibition it’s really RAF kinase dependence that is the pathway driver, which is why you see the synergistic activity with RAF kinase inhibitors. The kinase-independent function of RAF appears to be, at least in our early data, MEK independent, and we’re now trying to understand exactly what the molecular mechanism of that is. And there may be context dependence again in places where you might see RAF kinase independent functions vs kinase dependent functions.

Klijn: I think also this is a way to explain why single agent RAF inhibition isn’t as effective as you might have expected if everything would have been kinase dependent in that context. I think that one thing that we also tried to say here is that, depending on which of the RASs are mutated, there might be a differential dependency on the kinase function of CRAF, without any other inhibition. I think MEK really drives the dependence on RAF kinase function, but in other contexts those dependencies might fluctuate depending on which RAS is mutated or what the cell context is.

Malek: That’s an excellent point, because I think, that’s kind of how we started the paper, where we did this broad cell line profiling with the RAF kinase inhibitors, across hundreds of cell lines, Figure 1 of the paper, Figure 1b, and looked at BRAF-mutant, KRAS-mutant, NRAS-mutant, and wild-type cells. And clearly the RAF inhibitors are quite effective in the V600E cell lines but they were not effective in the majority of KRAS-mutant lines. From there our thinking was, either these cell lines simply don’t care about RAF kinase activity, because we’re dosing the drugs up and we’re just not seeing much of an effect, or maybe our RAF kinase inhibitor is not sufficiently potent and it could be a pharmacology thing. But we know that kinase inhibitor is actually active enough to work in the BRAF V600E setting. In fact if you look at Figure 1b you can see that the RAF inhibitors, the RAF dimer inhibitors anyway, type II, work quite well in NRAS mutant lines. So we think there is a context dependency where there will be subsets of RAS mutants that are going to be more dependent on RAF kinase activity, and there will be others that won’t. And the NRAS data, we didn’t extensively go into them in this paper, but we have data with multiple RAF dimer inhibitors showing greater single agent activity in NRAS-mutant tumors that looks almost equivalent in potency to what we see in the context of BRAF V600E. So we think probably RAF dimers are more active, perhaps it’s because the NRAS mutant lines harbor the Q61 mutation which we know is a more dominant mutation, so perhaps it’s about elevated RAS-GTP levels in that context. I’m actually going to talk about this at the AACR RAS meeting in December and get into this concept a little bit more.

Hartley: I was struck by the fact that in your screen of several hundred cell lines you found synergy between your type II RAF inhibitor and your MEK inhibitor in some cell lines and not in other cell lines, and that was sometimes independent of KRAS status.

Klijn: Broadly the synergy tracks with RAS mutation status, that’s Figure 4 I believe, Figure 4A, there’s clear and statistically significant enrichment of synergy in the RAS mutant context. So broadly, it tracks with RAS mutations. But there’s still a spread within that group, there are RAS mutants that show less synergy than others and similarly in the wild type situation you can see that there are RAS and BRAF wild type lines that show synergy but are not mutant for those genes. So it’s not a cut and dried, black and white difference, but the enrichment in RAS mutation group is very clear and significant. What I think we say in 4F and G and also the quantification in H is that the synergy is clearly seen in those lines whether mutant or not if they’re able to increase phosphoMEK after MEK inhibition.

Hartley: Is this combination therapy clinic worthy, is it phase I worthy?

Malek: I think so, we currently have a phase I trial ongoing with a company called Hanmi, they have a RAF inhibitor that we are partnering with them on, we’re combining with cobimetinib. Novartis is also pursuing this combination, they’re combining both with trametinib, and they’re also running a combination with their ERK inhibitor in development. The key to this again is, can you achieve a therapeutic index with this combination.

Hartley: Are you stratifying patients?

Malek: Yes, obviously the RAS mutant patients we would expect would be enriched for responses and certainly the KRAS G13Ds in colon might be a place that we would be really interested in testing this combination. And I think the synergy hypothesis leads us to believe that you could dose down the MEK inhibitor and re-wire the pathway and confer RAF kinase dependency under those conditions. So that’s why we think there’s a potential to get a combination dose that’s tolerated in the clinic.

Hartley: I’ve heard that toxicity has always been the problem with the MEK inhibitors. Could you go down by a factor of two or four, do you think?

Malek: That’s what we’re going to find out. It’s always a challenge developing these combinations. But from our preclinical data it does not look like we have to use cobimetinib at fully efficacious exposures. You see this pathway re-wiring at very low doses, down to 50, 100 nanomolar cobi. So our thinking about this has been, you’re better off reducing the dose of the MEK inhibitor, because that’s basically the strategy, to rewire, and then try and get the RAF inhibitor as high in dose as possible to inhibit RAF dimers. We’re using the MEK inhibitor essentially to confer the dependency and the RAF inhibitor is the real drug you’re using to get efficacy.

Hartley: Have you looked for resistance yet?

Malek: It will be interesting because most of the mechanisms of resistance to the existing BRAF inhibitors, whether it’s complementing a dimer, driving dimerization, or it’s a splice variant, a RAS mutation, BRAF amplification, they’re basically driving RAF dimers. So I think it will be interesting in this case to see what pathway escape looks like.

Hartley: Chris, what was your role in this paper?

Klijn: The pure computational part of the paper is the way we handled the drug screening, as well as some RNA seq. I think the major optimization was allowing us to do synergy screening over hundreds of cell lines. You can see that there are some full dose-matrix Bliss plots in here but those are laborious, and you can’t really scale those to the numbers that we wanted to scale them to. So in collaboration with the our functional genomics group, which is our screening core led by Scott Martin, we used co-dilution screens, where you basically screened a single agent over a dose range, and then at co-dilutions with two single agents. We developed ways to make sure that those curves are all comparable and that we get the right statistics out of them, so that you could see the biological effect as you see in the dot plots like the one in Figure 4a, represents quite a bit of work by very many many people, and also a lot of computational development. And then, obviously, the association of genomics afterwards, where we identified the G13Ds as being especially synergistic. All those things are basically grounded in the fact that the synergy statistics are quite robust.

Hartley: Chris, did you start out in the lab, or at a computer?

Klijn: I was in the lab in the early stages of my career, I did my Ph.D. in the Netherlands Cancer Institute, working on genomics in mouse models, mostly. That’s where I made the switch to computational biology. And after my Ph.D. I moved to Genentech as a postdoc, and after I finished my postdoc I stayed on as a scientist.

Malek: And it’s been a really effective way to do discovery science, with this close partnership where Chris is suggesting experiments, looking at the data so we have this iterative process going back and forth, and same thing with his team, he has a team of computational biolgists that work with several members of our department. Basically every scientist is partnered up with a computational biologist. And I think that’s basically essential now, because an individual scientist generates millions, billions of data points, and you need somebody who really understands how to view those big data sets, you need that close interaction to do the work at the scale that we want to do.

Hartley: Shiva, is there anything you’d like to say about women in science?

Malek: I take my role as a scientific leader in this field really seriously, both obviously from a scientific standpoint to make contributions to the community, but also as a mentor for women. I’m co-organizing the upcoming AACR RAS meeting with Karen Cichowski, Frank McCormick, and Gideon Bollag, and we put a lot of thought into the gender balance, making sure that was part of the discussion. I think we have a long ways to go the in the scientific community, I think we need more women at the table, in positions where we can influence, like on organizing committees, where you can open the discussion a little more. Genentech’s been a fantastic place, I talk a little bit about this in my Portrait, and there is a strong network of women that really support each other that’s front and center for me every single day. We have senior leaders like Wendy Young, Lori Friedman, Sara Kenkare-Mitra , there are many that have done a lot to raise other women in the organization. I would like for us to build a network like that in the scientific community and particularly in the RAS community. With folks like Karen Cichowski, Sharon Campbell, Debbie Morrison and Dafna Bar-Sagi, there’s no shortage of great women in the field.

Hartley: Chris, do you have any closing thoughts?

Klijn: As a junior scientist in the field, as well as a computational scientist in the field, I think it is important to realize that the future of science is probably team-based science. It’s not going to be individuals in individual labs working on a given problem. The problems are too big and the amount of data being generated is too large. A lot of the projects we have here at Genentech are very collaborative and that’s the thing that I really enjoy.