Kras and Cancer Metabolism: A Conversation with Carla Martins
January 23, 2017, by NCI Staff
Editor's note: Carla Martins is a Programme Leader at the MRC Cancer Unit in Cambridge, United Kingdom. Carla's research has focused on the identification and pre-clinical targeting of tumour progression mechanisms. Her work helped define the therapeutic potential of p53 targeted cancer therapy; and more recently, the metabolic heterogeneity and metabolic susceptibilities of mutant Kras lung tumours. Her group’s interests include the identification of therapeutic vulnerabilities in high grade lung tumors driven by mutant Kras.
Tell us a little about your background. I started my scientific career at home in Portugal, where I was awarded a Ph.D. fellowship from the Gulbenkian Ph.D. program in medicine and biology. Through that program I went to the Netherlands Cancer Institute where I worked with Anton Berns. Ton is a world leader in mouse cancer models, and although I had no experience in mouse genetics, I was a geneticist by training and really wanted to do cancer research. I felt that the type of questions I wanted to ask required in vivo models and was fascinated by what the lab could do, so I decided to join the group for my Ph.D. And that’s how I became a mouse modeler. We used insertional mutagenesis in vivo to identify the different driving mutations needed to turn a normal cell in to a tumor, focusing on cells that were already deficient for different cell cycle inhibitors (p21, p27 and p15). I ended up focusing mostly on a synergism between p27 and Myc during lymphoma development, but our in vivo insertional mutagenesis screen triggered my interest for the mechanisms that drive tumor evolution and how to link that information to therapy, from a mouse genetics perspective.
That eventually led me to the lab of Gerard Evan at UCSF as a postdoc. What drove me there was both Gerard's enthusiasm for cancer research, and the p53ER knock-in model they were developing that allowed us to switch p53 function "on" or "off" in vivo. I was truly fascinated by the potential of that model for tumor therapy studies and after joining the lab I used it to determine the therapeutic potential of targeting p53 in lymphoma and lung cancer [Cell 127, 1323, 2006; PMID 17182091; Nature 468, 567-71, 2010; PMID: 21107427].
Please explain the model and what you found.
It had been known for a long time that when Myc is overexpressed in transgenic mice under the regulation of an immunoglobulin enhancer, the mice develop B-cell lymphomas. If this is done in a p53 heterozygous background, there is a strong selective pressure for the loss of the wild-type p53 allele and the lymphomas that develop are deficient for p53. I knew this model well from my Ph.D. and thought it was ideal to test the impact of p53 restoration in tumors. In p53ER mice p53 is expressed from its native locus but as a fusion with the ligand binding domain of estrogen, which was modified to respond only to 4-hydroxytamoxifen. So in this model, the p53ER gene is transcribed as the wild-type but the protein is inactive unless the ligand is administered, and this switch is reversible and works well both in vitro and in vivo. So I crossed the two models, allowed tumors to form and then restored p53 function in lymphomas for a week. The results were dramatic, with the tumors being cleared after treatment and survival was significantly improved. These data were so motivating that I thought, if we can kill aggressive lymphomas by restoring p53 let's test it in solid tumors. I decided to use Tyler Jacks' KrasG12D, p53 null lung model and that's pretty much where my current scientific interests began.
It all started with a somewhat negative result. We tested the potential of p53 restoration therapy in well-established, late stage lung tumors and after one week of treatment we saw that the lungs were still full of tumors and pretty similar to those of control animals. After a careful analysis we understood that actually some tumors did respond but others not. In fact, within the same tumor, some cells would respond and others not, and high-grade cells were the only ones responding. This was a clear case of tumor heterogeneity driving resistance to treatment and it was really puzzling. Why would p53 be activated in some cells but not others, even within a single tumor? Eventually, we identified the mechanism, which was really gratifying. In short, increased Kras oncogenic signaling is selected for during lung tumour progression and therefore, high grade tumor cells express high levels of the oncogenic stress-responsive gene p19/ARF, which is also a p53 activator. Interestingly, this also means that low grade Kras mutant cells do not have high levels of oncogenic stress and therefore, they can be maintained in the presence of functional p53, because p53 is literally "blind" to them. Since then I have been trying to understand how oncogenic stress drives tumor progression and if advanced tumors can be associated with unique vulnerabilities that can be exploited therapeutically.
I soon realized little was known about how Kras-mutant lung tumors evolve but we knew that cells from aggressive tumors have distinct morphologic characteristics, that's why pathologists can identify them as high grade cells. So I thought that we should be able to find molecular differences between low and high grade mutant Kras lung tumors, and that this may help us define ways to target them more efficiently. Around this time I moved from UCSF to Cambridge UK and joined Dave Tuveson's group for a short period before starting my lab. Already in Cambridge, I was able to show that high grade lung tumor regions were associated with Kras allelic imbalance (gains of the mutant allele and loss of the wild type) and these imbalances were also present in human tumors. So once I started my lab, one of our focuses was to determine the impact of mutant Kras copy gains in vitro and in vivo.
We began by using MEFs [mouse embryonic fibroblasts], for their genetic homogeneity and because mutant Kras G12D/G12D mice are embryonic lethal. We compared MEFs that were p53-null and either wild type for Kras or heterozygous or homozygous for Kras G12D. To our surprise, there was not much difference in proliferation between the heterozygous and homozygous [Kras G12D] MEFs. However, we identified a striking metabolic phenotype in the homozygous cells. They underwent a metabolic switch, becoming very glycolytic and their glucose metabolism was rewired towards the production of TCA metabolites and glutathione. We were able to validate this metabolic switch in mutant homozygous murine and human lung tumor cells and also in high grade lung tumors, demonstrating that mutant Kras tumors came in at least two metabolic "flavours". Importantly, thanks to their distinct metabolic signatures, these two genetic mutant Kras groups also exhibited distinct therapeutic vulnerabilities, which is something that our lab is quite interested in.
There has been a long-running speculation that wild-type Ras genes act as tumor suppressors, and you're seeing quite a change when you lose the wild-type Kras gene. Do you have an opinion?
In the lung cell lines we generated you can find mutant gains more frequently than wild type loss. But we do see all types of scenarios. In tumors we see both, although wild type loss detection requires the use of laser-capture microscopy to avoid contamination from other cells. That's why I think it's been underreported in humans, but there are cases published. I do think wild type Kras is contributing to some of the phenotypes we see, but we don't know yet if it's truly tumor suppressive in this setting. So far, we have mostly tested the impact of mutant gains but are now also analyzing the effects of wild type loss and hopefully, will be able to clarify this soon.
Could you use Mariano Barbacid's Ras-less MEFs in some way? You could manipulate the copy numbers of Kras.
Those MEFs would indeed be very interesting to study the impact of a different number of copies of Kras on the phenotypes we described. But in order to translate those findings to cancer we will also need epithelial cells/tumor models. But those are indeed questions we would like to address. Also important is to ask whether the metabolic rewiring we observe can only take place if extra copies of mutant Kras are present. In other words, are the metabolic changes that we see in advanced tumors that acquire extra copies of Kras a feature of all advanced tumors, including those without extra mutant Kras? It would be truly fascinating if we could use the phenotype rather than the gene alteration to single out advanced tumors. In that case, our work may not only enable us to develop therapeutic strategies that are efficacious against mutant Kras advanced tumors (which in itself would be great!) but also against advanced tumors in general.
But back to the MEFs, I think they are a great tool. We tend to underrate their use in cancer research projects but it is important to recognize the merits of different models. With the right question, MEFs can be a powerful tool. In fact, the identification of the metabolic rewiring phenotype we described would have been more difficult if we would have started with cancer cells, as they are significantly more genetically heterogeneous and proliferate at different rates. Something else that was really cool and surprising is that our mutant Kras homozygous MEFs cells were able to form lesions in the lung of a mouse (whereas the heterozygous didn't). These are fibroblasts, right? This was particularly unexpected since we did typical in vitro assays for aggressive phenotypes, such as the scratch and invasion assays, and nothing was different between homozygous and heterozygous cells (except colony formation capacity). We were really puzzled, as we expected a more invasive phenotype in homozygous cells and wondered if it could only be manifested in vivo. We didn't know whether a metastatic assay could work with MEFs but we put them in the blood stream and suddenly we had large lung lesions from homozygous cells. We couldn't believe it at first, but it worked every single time. Now we know, some MEFs can "metastasize".
Isn't it likely that dosage of mutant Kras protein is acting through changes in some transcriptional program?
We do see a large number of transcriptional changes in mutant Kras homozygous cells relative to heterozygous, both in MEFs and lung tumour cells. That's how we identified the glucose metabolism phenotype. Of course we often ask ourselves what's the transcription factor. But I doubt it can all be explained by a single transcriptional factor, as Ras talks to many pathways and that's why its phenotypes are so complex. So although our brains like to explain things in linear ways, it's not usually that simple and we have to consider other options. On the other hand, it can all be down to a master regulator, say Myc. But if it is, do we then feel that we answered everything by substituting mutant Kras as an answer by Myc? Or do we still need to ask, what part of the Myc program actually matters? Again, I am more interested in the biology. We are trying to find therapeutic vulnerabilities in these cells and in the end, the target doesn't have to be a gene or protein, it can be a phenotype.
And cells dependent on mutant Ras are also more sensitive to reactive oxygen, correct?
I think that depends on the experiment itself. If you add a lot of mutant Kras to cells, you will probably see them starting to generate a lot of ROS [reactive oxygen species]. But if you are looking at mutant Kras cells that are happily growing, they were probably selected for mechanism that enable them to buffer, or "ignore" ROS, because they wouldn't have made it otherwise. They have adapted to a better way of coping with ROS and they will probably do better when it comes to metastasis, because they can take that extra stress. So we don't actually see more ROS, we see less in mutant homozygous cells, but we also see selection for mechanisms that reduce ROS. But if you challenge them by targeting those ROS-coping mechanisms they will indeed show high ROS levels and decreased survival.
I'm more interested than I would have been because we are also seeing a mitochondrial connection in one of our drug screens using the MEFs. We developed these MEFs to use for cell-based drug screens, and we're screening using a MEF rescued with Kras G12D and comparing drugs that hit on those compared to cells rescued with wild type Hras. And MEFs rescued with BRAF V600E are truly Ras-less and are an excellent tool for finding molecules that hit Kras directly.
That's great. Do you think you are close to getting a Kras-specific compound?
I wouldn't say that, but we do have productive collaborations with pharmaceutical companies, and those are in various stages.
I feel that there is a new optimism in the Ras field, isn't there, people seem to be thinking they're close, so that's why I'm asking you. It feels like the mood has finally changed.
I think that is true, I think part of it of course is Kevan Shokat coming up with a G12C-specific tool compound and we're also looking at that sort of thing. We've dipped our toe a little bit into metabolism, and I've seen cartoons in our group meetings with a dotted line between the mitochondria and mutant Kras, but we don't know what that dotted line is.
Can I finish by asking you if you have any philosophical thoughts on approaching better therapies for Ras-mutant cancers?
I think the Kras field is making amazing progress and we will eventually succeed in finding strategies to inhibit mutant Kras. But now that we know that there are tumors with more mutant Kras and tumors with less mutant Kras, if we have the opportunity to actually inhibit some of that mutant Kras, which cells are going to respond better? We don't actually know. You could say the ones with more Kras will be more addicted to it, but maybe they will be more difficult to inhibit. Personally, my focus isn't on targeting a single mutant protein. That can work really well but often leads to resistance to therapy. I want to know if we can use other approaches. By looking at what tumors do differently from normal cells we were able to find a way to kill tumor cells in the clinic, by targeting their proliferation properties. That has been successful across all tumor types, just not very specific. We want to find targetable phenotypes that are truly specific to Kras cancers. For that, we focus on advanced tumors, and want to know what makes them tick, what keeps them aggressive, what do they depend on to be what they are. Hopefully, the metabolic signatures we identified will help us with that task, and the answer may not even be unique to Kras tumors, who knows. There are still many questions to answer, but I really think that we will soon see clear improvements in mutant Ras cancer treatment. But the solutions will need to keep evolving because the tumors will for sure. So we need to keep working together and explore different ways to outsmart them.