Skip Navigation

103 ESMO Mills Transcript

Kristine Crane: Personalized medicine is becoming increasingly important in oncology, as researchers home in on patients’ genetic mutations that may help determine which therapies, especially targeted therapies, might benefit them. What are the challenges to having so much information at hand, and how will drugs and other therapies address this genetic variability? Dr. Gordon Mills, chairman of the department of systems biology at the MD Anderson Cancer Center in Houston spoke about molecular personalized medicine at the ECCO/ESMO cancer congress in Stockholm. I spoke to Dr. Mills after his lecture. So why did you say that breast cancer is leading the way for stratified cancer therapies?

Gordon Mills: So breast cancer, about ten to 15 years ago, was thought of microscopically as a single disease, at least the most common form of ductal breast cancer. We knew that there were some differences in terms of the estrogen receptor and there were some hints that we should treat those differently, but when transcriptional profiling appeared and took a look carefully, it soon became clear that not only were there different expressions of different genes, but actually, these were different diseases; that breast cancer should be separated, at least initially, in the studies into luminal breast cancers, which are ER positive, sensitive to ER targeted therapy, basal breast cancers, and then HER2 positive breast cancers that can occur on both of those backgrounds. Further analysis of that data has begun to show that there are additional subsets. There are luminal A and luminal B, well behaved early breast cancers. The luminal As that don’t need chemotherapy, luminal Bs that probably benefit from chemotherapy, much more active cancers. And then you could split the basal cancers into basal and claudin-low. And although initially there were arguments about this, they turned out to clearly be separate diseases.

Gordon Mills: If you add on top of that some of the changes such as differential mutations, mutations in BRCA1, 2, mutations in P10, mutations in PI3 kinase, there’s at least eight or nine different subtypes clinically, in terms of management. Now what we know that is very exciting is for the last 15 years, there has been a 1 to 2 percent improvement in patient outcome. And what we can show is that at least half of that can be attributed to improved therapy. That improved therapy, really, I think is based on treating the right subtypes with the right drugs. Now that comes with a little different challenge. We now have so many types of breast cancers that it’s really a set of orphan diseases, and that it is sometimes difficult to mount clinical trials to treat, say for example, claudin-low breast cancers as a separate subset, because these are now uncommon. However, it’s clear that the message from breast cancer is that not only are these different subtypes, they really are different diseases that histologically look the same.

Gordon Mills: And if we continue in other diseases to ignore that message and treat all bowel cancers on a single trial, all ovarian cancers on a single trial, we’re doomed for failure. And indeed, what we can show already is that basal breast cancer has more in common with ovarian cancer than it has-- high grade, serious ovarian cancer-- than it has with ER positive breast cancers. So you’d be better off to do a clinical trial, combining ovarian cancer and basal breast cancer, than just doing breast cancer trials. And we believe that this is the same situation in most other cancer types. What we see histologically is a group of similar diseases of similar cell origin, but potentially very different genetic events, different behavior, different outcomes and different types of therapy are needed.

Kristine Crane: So is this going to mean potentially that different types of cancer could be treated in the same way? For example, neuroblastoma has more in common with pancreatic cancer, or the very aggressive tumors.

Gordon Mills: Sure. So ten years ago, many of us in the community were saying that we’re going to have cancer clinics related to the underlying genetic defects. For example, we would have a P53 clinic. Wouldn’t matter what tumor type you had, it was a P53 clinic, a Ras clinic, a PI3 kinase clinic. I think some of the newer data is beginning to argue that that’s not going to be the way of the future. It turns out that the effects of the genomic aberrations in tumors is dependent on something else. Now what that something else is, is a little controversial, but if we start off with a very simple statement: if you have an EGFR mutated tumor, you are likely to benefit from an EGFR targeted therapy, unless you have a Ras mutation. So there’s a combination of more than one mutation that determines the outcome. So we would have to start thinking about combination clinics. The other thing that we find is that when we do very large screens of cell lines, those cell lines segregate in their response by genetic aberration and by lineage. So I think that what we’re going to see is that there will be different effects of some drugs, dependent on the lineage of the tumor. Now we already have a great example of that, and that is, the V600E mutation in RAF. In melanoma, that signals a response rate of 40 to 60 percent. In bowel cancer, a response rate of under 5 percent, exactly the same mutation.

Gordon Mills: So either that means that the intrinsic gene expression pattern of the tumor lineage alters the context of the mutation, or alternatively, it may mean that the set of co-mutations that happen in bowel, with V600E mutations, are different from those that happen in melanoma. And indeed, we have some very early modeling data that says that that is actually the case, that different co-mutations occur in different cell lineages, and that is what determines why some will work. On the other side, there are examples where this does work. For example, we first showed that targeting HER2 in breast cancer was highly effective. A recent study now has shown that if you have that same lesion in gastric cancer, you will also benefit from targeting HER2. ______ ovarian cancer is a little different. It doesn’t look like you benefit as much, although that is very soft data, because it’s hard to find enough patients. So many people are looking at the concept, that yes, there may be some signal across lineage, but the best results are going to be combining lineage with the genetic aberrations.

Kristine Crane: So what are some of the challenges that that poses?

Gordon Mills: So I think that there are some early messages that we’re receiving that we need to start paying attention to. The first is, is that many types of cancer are intrinsically, genetically unstable, bowel cancer, endometrial cancer, melanoma. In melanoma and lung cancer, we know that at least a lot of that is due to the insults that cause the tumor. That is, cigarette smoke in lung, and UV light in melanoma, both mutagens, you would expect many mutations. The question that we have in those tumors are which ones of them matter? And which ones of them matter comes in sort of two definitions. One is the statement and term that people use, what’s a driver? Now, I don’t like that term, but at least it’s a conceptual one we can work with. A driver is a genetic event that is driving the behavior of the tumor at that time, and if you do something about it, it should make a difference, passengers, everything else. And so that operational definition is very useful, and it may well be that there’s a similar number of drivers in all of those tumor types, but they’re messed up or mixed up or hidden, like needles in a haystack, amongst a whole lot of passengers. So the challenge that we’re going to have is, what do we do with the information from these larger scale sequencing programs?

Gordon Mills: And indeed, for a while, much of our effort is focusing in on what we’re calling actionable events, that is, genetic events that we’ve seen enough times, other people and we have studied them, to determine what they do and show that they really are drivers. That, I think, is going to be much more useful. So I think that we have to start very quickly distinguishing drivers from passengers and which drivers are actionable. Now some of them just aren’t. Right now, we don’t know what to do with a P53 mutation, whether it’s a driver or not. We don’t know what to do with many of the other tumor suppressors. So we also are putting a large amount of effort on where we have drugs that would either make a difference, or that a patient would be entered on a trial to fill that trial quickly. That’s in the patient management side. The research side, unrelated question here. We clearly want to go in and characterize tumors in depth to understand why each patient is responding or not responding. And you’re going to get that information from looking more deeply. But the transition from when this is researched through the patient management, I think is a little further off than we would like, because of the complexity.

Kristine Crane: So when you mentioned also that the literature on the data are not exactly lining up.

Gordon Mills: So one of the things that we have found in our early studies that is very exciting, is that when you look at patients who are going on clinical trials, and that means that they have aggressive disease that has not responded to conventional therapy, that was not cured by surgery, they have many fewer actionable mutations than we would have predicted, _____ mutations in PI3 kinase pathway, the Ras pathway. And so we had been challenged and concerned as to why that was. What we now know is that when you look at early low grade disease, that is frequently cured by surgery, that is chock full of these targetable mutations in the PI3 kinase pathway, even in the Ras pathway. So those patients don’t go on clinical trials, for good reason. We cure them by other mechanisms. So when we look at who’s going on clinical trials, it’s frequently patients with high grade, ugly tumors that seem to be driven by P53 and by genomic instability. Now I don’t want to say this is black and white. I’m saying there’s an enrichment. There’s still a large number of patients with these actionable mutations going on clinical trials. There’s just fewer than we would predict, and what that means is, we have to redesign how we identify those patients and power our trials.

Kristine Crane: That was Dr. Gordon Mills, chairman of the department of systems biology at the MD Anderson Cancer Center in Houston, speaking from the ECCO/ESMO cancer congress in Stockholm.

#### End of Mills.mp3 ####