A Conversation About RAS Dimers
April 20, 2018, by Chiara Ambrogio, Pasi Jänne, and Frank McCormick
There are conflicting data on whether KRAS protein molecules form dimers. For example, two 2018 papers are titled "K-Ras4B Remains Monomeric on Membranes over a Wide Range of Surface Densities and Lipid Compositions"1 and "The Biological and Therapeutic Impact of KRAS Dimerization in KRAS-Mutant Cancer"2. Two authors of the latter paper, Pasi Jänne and Chiara Ambrogio, recently discussed the issue of RAS dimers with Frank McCormick, who is advising the NCI RAS Initiative at the Frederick National Lab. Jim Hartley of the RAS Initiative moderated the discussion.
Hartley: Pasi and Chiara, your recent paper2 has stimulated quite a lot of interest3,4. Would you start by summarizing your results?
Jänne: Really the paper has two observations, and started from the fact that there's previous literature suggesting that wild type KRAS has a growth inhibitory effect on mutant KRAS. Chiara brought the RASless MEFs from Mariano Barbacid's lab, and they allowed us to study mutant KRAS in the presence or absence of wild type KRAS. Using this model system we were able to recapitulate the growth inhibitory effect of wild type KRAS on mutant KRAS. And that got us asking questions about dimerization using the cell line system, and our collaborator Ken Westover at the UT Southwestern using a FRET assay. And key was generating this mutant, D154Q, that Ken hypothesized would be in an interface that could potentially dimerize. The mutant in both the FRET assay as well as in the cell line phenotypic assays, in fact lost the growth inhibitory effect of wild type KRAS. So that was the first observation that this inhibitory effect was somehow mediated by interaction of wild type with mutant KRAS.
And then the second part of the paper was to ask, Forget about wild type KRAS altogether, what about just oncogenic KRAS, does it also need to be in some dimeric or multimeric complex to have a transforming activity? And using the same [MEF] system, when we take out the wild type KRAS and put in either oncogenic mutant KRAS or oncogenic KRAS that has the D154Q mutation, we found that in fact the D154Q mutation has a tremendous phenotype both in vitro and in vivo. This suggests that dimerization is important for the full oncogenic transformation mediated by RAS.
In the paper we go into detail also to show that the mutation doesn't affect some of the other known properties of RAS, membrane localization or nucleotide exchange or GTP hydrolysis, etc., those were not impacted, meaning that we believe that the finding here is truly due to dimerization, and opens some hopefully new therapeutic avenues for RAS mutant cancers.
Hartley: That was a very clever use of the RASless MEFs. Frank, would you like to summarize the data from your lab, and from the RAS Initiative here at the FNL, that have influenced your thinking about how RAS proteins might interact?
McCormick: First of all, congratulations on the paper, as Jim just said we've been interested in this phenomenon for a long time, largely because Allan Balmain, who first discovered the interference of wild type RAS on mutant RAS, is right next door to my office and we discuss this sort of stuff a lot. But until your paper there really haven't been many demonstrations of this interference in a tractable system. A couple of papers have shown that overexpression of wild-type RAS can interfere with mutant in some ways, but it's always been very complicated to interpret because obviously the wild type protein can signal as well, so seeing interference of the mutant is difficult. Your paper is the first to clearly show that interference in a nice sort of tractable system.
We started hypothesizing that the most obvious interpretation of interference was from subunit poisoning by wild type proteins with mutant proteins into complexes, where maximum signal output would require both the proteins to be in the active state persistently, and that having a wild type protein would be like a sort of flat tire and would be a much less efficient complex than a sort of double barreled mutant RAS. But demonstrating that biochemically has been so far very challenging. As you refer to in the paper you can certainly see that proteins appear to dimerize using fluorescent tags, and at Frederick we've used two or three different approaches to try to verify that kind of apparent dimerization in living cells or in cell membranes, using step photobleaching or FRET or BRET or visualizing fusion fluorescent proteins coming together in space. And there's definitely something that looks like dimerization going on, but our complication has always been, as we first found in the Nan paper5 and as Tommy Turbyville and colleagues at Frederick have repeated in much more depth, just the hypervariable regions of the RAS proteins are sufficient to give the same appearance as dimerization. That's something that has been seen in multiple systems and verified by other people including Channing Der and John Hancock and others. So it seems like what we're measuring in BRET or FRET-type assays for visualizing is mostly driven by the C-terminal hypervariable region that doesn't have any required input at all from the G domain.
And the hypervariable regions themselves are most unlikely to dimerize, since they have polybasic sequences and so on, so we've been moving in the direction more of the hypervariable regions forming very small local clusters of proteins in the membrane, where the concentrations become very high. Then obviously it becomes difficult to distinguish between a true dimer with a protein-protein interface, or two proteins that are very close together that functionally are close enough to activate two kinases, for example, or two different effector proteins. And so I think we, and the whole field, have struggled to demonstrate in a biochemical way, or at the atomic level, a real protein-protein interface that would be a sort of card-carrying interface that would satisfy hard core biochemists. I think that's where we are in trying to summarize how these proteins interact. They definitely come very close together, but we are not convinced that we are seeing true dimers as opposed to very highly proximal proteins that are brought together through lipid micro domains into very high local concentrations.
You might say, well who cares, really, if they function like dimers and show subunit poisoning and so on as you have demonstrated very nicely in your paper. So whether those helices form proper protein-protein interaction dimer interfaces or something close to it, may be a semantic issue, may not be so important. But if you think about that as an interface for targeting for drug discovery then I think it would be really nice to know if there is a real protein-protein interface that contains multiple interactions, involving multiple side chains and so on. And that hasn't been seen in proteins in solution, as you're well aware, and may only be possible to see in very very high local concentrations in very small particular patches of the membrane.
So we are thinking of trying biophysical methods to try and visualize direct protein-protein interface interactions that would more closely fit the sort of formal definition of a dimer than the more functional definition you've used very elegantly in your paper. That's one of the challenges we're trying to take on in Frederick.
Hartley: Can I ask how the D154Q mutation could have such a potent phenotype if RAS molecules only form transient clusters at the membrane? We don't detect RAS-RAS protein interactions with purified proteins, and the imaging techniques show us a pretty dynamic situation.
McCommick: It's technically very hard to see them.
Jänne: I think the interesting thing is that even in the presence of this mutation, even in this clean system, we don't completely inhibit ERK signaling. There's a reduction, and obviously in the presence of what we believe is the monomeric version of RAS, you can still interact with RAF, and so these are membrane bound, you still get some kind of inefficient or occasional RAF dimerization and you see some activation of MAP kinase signaling but not full activation. But then the dimerization is somehow required to really make this process efficient, to make the RAF dimerization efficient. But how to dissect that out is more of a challenge.
Ambrogio: I can add that we are working on a mouse model in which we will be able to form a tumor driven by dimer-proficient KRAS and then turn on the D154Q mutant. We know that the D154Q mutant impairs tumor initiation, so we hope to see a phenotype when the tumor has already formed.
McCormick: I'm very glad to hear you're going in that direction, because in fact our approach at Frederick is to go in the opposite direction and try and become more reductionist and more biochemical and biophysical, and hopefully that approach will complement the more biological approaches that you guys are taking.
McCormick: I think the two approaches are always mutually informative, and I think from our perspective trying to understand the precise mechanisms of RAF activation by mutant RAS is a critical part of the whole Frederick RAS Initiative. So we sort of presume in normal cells that RAS GTP levels are rate limiting in that whole process. But when you have mutant RAS which is always loaded up with GTP, maybe the dimerization step becomes rate limiting, and then very close proximity, whether it is formally dimers or close dimers, may have a much bigger impact. We don't really understand enough about the whole biochemistry of that process to know as yet. And also there may be other proteins involved in the activation complex that could be affected by the 154Q mutation as well. And Debbie [Morrison] thinks there could be other contact points between RAF and RAS proteins, and we know there are other contact points between RAS and the plasma membrane. What's actually happening in that little micro world of RAS and RAF dimerization we just don't know at the sort of molecular level. Clearly the region of the protein around 154, that whole alpha helix, seems to be really critical in that whole process, as your very nice work has shown, and also the work of John O'Bryan6, as you refer to in the paper, showing that the monobody that binds close to that site inhibits KRAS localization in the membrane correctly, seems to say disrupting that whole machinery in that region is a potential Achilles heel in the whole RAF activation process.
Hartley: Does this make drug screening difficult? We have interactions of an uncertain number of molecules with an uncertain set of interaction partners in an transient domain in the membrane ...
McCormick: We've done a lot of BRET- and FRET-type assays in Frederick, it really looks like that interaction is amenable for screening for compounds that prevent interaction. We have collaborations going on to try and find compounds that prevent that interaction, even though we don't necessarily know the precise molecular details of the interactions that are going on. It's a very simple assay for looking at things that disrupt association of RAF proteins, or RAS-RAF complexes. So you can also use it to screen for genes which regulate that process, which might be more of a discovery area to understand more about the actual biochemistry of the process. But it's a simple assay, we might find compounds that do what we want even before understanding how the whole thing works.
Jänne: Absolutely, and we're certainly taking that direction as well. It's a straightforward thing to do and hopefully it will lead us in new targeted directions. It remains to be seen whether that interface, whether you can get a small molecule, or you need a large molecule like the monobody, to disrupt that. But I think that can now start to ask those questions.
Hartley: It would really be better, wouldn't it, if we really knew what other proteins were in the same neighborhood as RAS molecules in the membrane of a living cell?
Ambrogio: The recent paper1 from the Jay Groves group very elegantly showed that KRAS lacks intrinsic dimerization properties, and we agree, so we're trying right now to evaluate proteins that might act as scaffolds for KRAS.
McCormick: I think that is the way the field is moving, it's really difficult biochemistry. As we know, as soon as you solubilize the membrane all these complexes fall apart, and protein that's there because it's hanging out in the same neighborhood of the membrane is going to dissolve. But capturing the complexes in membranes in a way that you can actually do biochemistry and actually look at specific interactions, that's another whole challenge which is much more complicated than looking at proteins interacting in solution. But it's what we've got to do.
Hartley: What's the path forward from here? Pasi?
Jänne: I think on the biology side one of our big questions is, Is this interaction required for tumor maintenance? We want to demonstrate, or hope that it is, or has a role in that, because that obviously opens a link, or provides biologic information on the therapeutic approach. For us, that's an important question. Number 2, we're thinking about the different type screens to try to see, can we disrupt the dimerization. So they're sort of complementary approaches, but of course we're keenly interested in this from the therapeutic angle, obviously.
McCormick: We have the same interests. Again, our approach to get there is to try and use some biophysical methods to actually measure interactions of the hypervariable region of RAS with synthetic membranes or nanodiscs, where we can actually pin down exactly which residues are involved in membrane associations and RAF associations and in this potential high concentration dimerization or clustering. If that works, we might be in a better position to try and make the kind of mutations which you guys are already testing. We'll see. I think we all have the same overall goals, thinking that disrupting these complexes is going to be an efficient way of reducing oncogenic RAS signaling. Just as the wild type proteins are very strong suppressors of the mutant proteins in mouse models, as we discussed at the beginning. I'm sure other people will be moving in the same direction also, because it's fascinating biology, and also opens up other issues like, if RAS proteins form these clusters or dimers with each other, what else can be in that complex as well?
Jänne: It's nice that, as you mentioned earlier, we're taking complementary approaches, as I'm sure other people are as well, so hopefully there'll be a body of evidence both from the biochemical and from the biological side that will continue to support this and hopefully lead us somewhere.
McCormick: Some work was presented from Geoff Wahl's lab at the RAS symposium, showing using a split readout system that RAS proteins formed these tight associations, and they found that even proteins like RAC1, which is another member of the RAS family with a fairly basic hypervariable region, can actually team up with KRAS and other RAS proteins in the same assay. So again, they have the idea of a sort of zip code where different RAS family member proteins can end up in the same signaling space, literally, and probably very interesting biological properties of sort of mixed complexes. Another whole level of biology which hasn't been tapped.
Hartley: May all your experiments be illuminating! Thank you.
About the participants
Chiara Ambrogio is a staff scientist in the Jänne lab at the Dana Farber Cancer Institute in Boston. Pasi Jänne is a medical oncologist at the Dana-Farber and a Professor of Medicine at Harvard Medical School. He is the director of the Lowe Center for Thoracic Oncology and the scientific co-director of the Belfer Center for Applied Cancer Sciences.
Frank McCormick is a Professor in the UCSF Helen Diller Family Comprehensive Cancer Center and holds the David A. Wood Chair of Tumor Biology and Cancer Research. He founded Onyx Pharmaceuticals and led the development of the first RAF inhibitor, sorafenib.