I’m very interested in driving combination therapies because I think that’s going to be the important next step that we need to make in cancer treatment. In fact, to be honest, we’ve been doing combination therapies for decades in cancer therapy but with the advent of targeted therapies you’re seeing a number of drugs approved as monotherapy but wherever you’ve seen a targeted therapy approved as monotherapy resistance is a universal phenomenon occurring in those patients. So if we start to understand what drives resistance we can start to devise combination therapies that are more likely to produce either more durable outcomes or, hopefully, increase the actual long-term survival rate and potentially even the cure rate in some cancers.

What were some of the key points of your talk?

When you’re thinking about combinations it can be overwhelming to think of all the things that you could do, the permutations and combinations become not quite infinite but close to that. So the key thing is to start with a drug that you already know to be active because you’ve got some level of clinical validity, then, that that target, that pathway, is relevant and important in a significant number of the patients in the disease that we’re trying to treat. So I’d start with where you already know something is active, as I’ve said, then the next step is to understand what’s driving resistance to that individual drug. If you can understand that you can either design a better drug that overcomes that resistance, we’ve got one example of that that I talked about with the Azadine 9291 drug which is a mutant selective inhibitor of EGFR which was designed to overcome a known resistance mechanism that’s common in the current EGFR directed therapies in lung cancer. So that’s one example of what that information can then give you but what we’re also doing with that is then trying to predict what the resistance mechanisms are going to be. So actually we only started dosing with that drug just over a year ago but we’ve already started in the lab creating Azadine 9291 resistant cell lines and models and we’re understanding that actually what happens then is that as you develop resistance you get up-regulation of the MAP kinase pathway. So it’s a different mechanism than you see generating resistance to the currently available EGFR therapies. What that insight has given us is the opportunity to then test a combination of 9291 with a MEK inhibitor that we have, selumetinib, and we’ve shown pre-clinically that that has activity and generates tumour regressions in vivo. So we’re going to then take that into the clinic and we’re starting a trial in the next month to look at that mechanism of resistance. So that’s one example.

The second example is looking at the interaction between, say, the PI3 kinase pathway and androgen receptor and oestrogen receptor signalling. So the point I’m making here is that the lineage of the cell, the cancer cell, where it’s actually derived from is actually important. We know that AR remains important throughout the life of prostate cancer; we know that ER signalling remains important throughout the life of ER positive breast cancers and that is different, fundamentally, in the biology from how other types of breast cancer respond. We also know that in both of those endocrine driven lineages of cancer, if you like, that there’s an interaction between either AR signalling or ER signalling in the PI3 kinase pathway. That’s important when you’re thinking about potential combinations so we’ve given another example of interaction between both androgen receptor and PI3 kinase inhibitors and oestrogen receptor inhibitors and PI3 kinase inhibitors in two different settings.

It also brings me on to the third point which is that schedule matters and actually when we’re thinking about combinations too often we’ve put two drugs together, trying to dose both of them continuously and simultaneously rather than thinking about the schedule of interaction of the two drugs. Actually what you’re trying to do with that combination which is generally not just to stop cells proliferating but to actually kill them. If you think about that in terms of what schedule do we need to actually kill the highest proportion of cancer cells then you often come up with a different regimen. So, for example, we’ve tested multiple PI3 kinase inhibitors on more intermittent schedules rather than continuously and that generates a higher degree of apoptosis. When you dose continuously you inevitably get pathway reactivation and you get inhibition of proliferation for a time but you don’t get the same level of apoptosis. If you dose intermittently, hit the pathway hard, you allow the pathway to then be reset during the break and then you can come in with another dose, hit it hard again and get another wave of apoptosis. What that does, first of all, it kills a higher proportion of cancer cells so we think it drives better efficacy but it also happens to be better tolerated, particularly when you’re thinking about doing that in combination. So I’ve given a couple of examples of the data that we’ve got that we’re deriving across our portfolio for looking at that.

So those are the basic principles that I’ve highlighted which I think, particularly as you start to hear about talks of triplet combinations, start to be really important because continuous combinations of two or three drugs often run into tolerability issues.

Is there a risk that the more combinations you use the more likely cells are to find a way around them?

I’m not sure. I think we need to go back to this fundamental principle of how many cells are we killing. I’m actually trained as a clinical oncologist, radiation oncology, where you do clonogenic assays and you look at the proportion of cell kill. If you think about it in that way, I’m not sure we’ve applied that thinking to how we try to generate combinations. In fact, you need to produce many logs of cell kill if you’re actually going to produce the situation where you’ve got the potential for cure. So I’m not sure that in that scenario if you’ve killed a high enough proportion of the cancer cells that you’re going to get multiple pathways coming up because actually the range of options is reduced if the total pool of cancer cells is much smaller and the mutations tend to be spread across that pool. So if you’re killing cells more effectively you’re actually going to reduce the potential number of resistance mechanisms that will come up, not increase them.

What features make certain drugs suitable or not suitable for combination?

You’ve got to understand both the interaction of the pathway and the potential overlapping tolerability issues. People have tried to combine, for example, MAP kinase inhibitors with PI3 kinase inhibitors multiple times and those trials have generally run into tolerability issues. Again, if you actually look at those schedules, most of them were tried to be combined on a continuous regimen for both drugs and what you ended up doing is having to compromise the dose of one or the other. So thinking about this in this way, which is what’s the schedule and interaction of the combination, might make possible drug-drug combinations that have currently been abandoned because they were felt to be too toxic. But if you’re thinking about combinations in general you want a drug that has the lower propensity for causing drug-drug interaction; ideally you’d want one that’s relatively well-tolerated as monotherapy because it makes it easier then to create a tolerable regimen in combination. But we’re rejecting too high a proportion of them because we haven’t actually thought about the optimal schedule of how you might combine them rather than just going for the straight combination and ruling things out that can’t be tolerated when combined continuously.

Are there any examples of a potentially good drug combination?

Again, we’re going to need to think about this within the segments that have already occurred in terms of understanding of cancer. So, for example, in lung cancer you need to think about it in the context of EGFR driven disease, RAS driven disease, ALK driven disease and then think about what are the mechanisms that are coming up as resistance within those. So I’ve given a couple of examples, it may be an EGFR and a MEK inhibitor are going to be possible to combine if we think about that rationally. The other piece that’s both an exciting opportunity and a real challenge at the moment is to think about how you combine some of the small molecule targeted therapies with the immuno-oncology therapies, where the models to look at that potential interaction pre-clinically are often challenging because the models that are used for immuno-oncology are often xenogeneic, don’t include the actual genetic relevant mutations or background that’s relevant for the small molecule. Of course, the small molecule preferred models tend to be xenografts that are immunosuppressed. So some of the genetically engineered mouse models can help you by having the relevant genetic question in a context of an immune competent host but even then I’m not sure you’ve got the same level of genetic diversity in the tumour as you would have in a human tumour. We really need to think about how we’re going to test systematically those kinds of combinations. There was already some data presented at the meeting which suggested ways forward in terms of how we can prioritise that so I’m encouraged that we’re starting to think those things through.

What is your take home message from your talk?

We’re at a very exciting point where we’ve got such a level of information about the genetic background, transcriptional profiling, microRNA profiles, together with a large array of potential therapeutic targets that the opportunities for combinations are coming through and we’re starting to systematically address them. So that’s very encouraging from my perspective.