Media Center Banner


2013 OCRP Investigator Vignette

What Determines Ovarian Cancer Patient Response to Treatment and Survival?

David Bowtell, PhD; Peter MacCallum Cancer Centre, Melbourne, Australia

Can I say in the introduction; I can't believe that it's 10 years-10 years since I've been here. When we were coming onto the Base and it had changed a bit, I said oh, it must be five or six years since I've been here, but 10 years-.

Karen thank you very much for that introduction and-and the-the invitation to speak here. As you can tell, the program has been really very important to us and-and it's a great pleasure to come back here and-and to share some of the successes of that work and-and in fact, in the first few slides I'm just going to just show you a thumbnail of-of some of the things that have come out of that original program grant.

So yes; it's a great, great pleasure to be here.

The talk today is about what determines survival and response to treatment in women with ovarian cancer. And I think this is something that we're trying to understand not just for ovarian cancer but for solid cancers in general. On the one hand, this is obviously very important in terms of-of being able to manage patients better and improve their outcomes, but also patients in a sense act as a-sort of an experiment of nature, in that those that do well or those who do poorly give us insights into the biology of the disease and help us-help take us to the genes that-that drive these cancers and ultimately we think to-to the targets that will allow us to treat the diseases more effectively.

So I'm going to focus virtually all the talk on that but as I said I just want to start with a couple of slides about this. This was the award that was made back in-in 2000 and started in 2001 or late 2001, once we had got the ethics in place. It's by far and away the most important grant that I've ever received. It's not actually the largest, although it was pretty large by-by anyone's standards and certainly by our standards at the time. And it was made even larger by the fact that in those days the exchange rate was almost two to one. So there was a sort of fantastic multiplier for grants that we received from the US.

Happily for us, if we're traveling in the States that it's closer to-in fact is that parity now but certainly back then it was a lot of money. But really what was so important about this grant was as Karen mentioned, it allowed us to create the Australian Ovarian Cancer study. And these were the aims of the study as we had written them back in 2000. We said that we would recruit 1,000 women who were presenting for surgery for suspected ovarian cancer based on their clinical presentation and that we would-we would aim to collect bio-specimens on-or tissue samples on about 60% of those cases, bloods on-we hoped about 90%; we would also recruit a similar number of women as controls from the electoral registry. One of the advantages of-of the fact that-Australia is one of the few countries where it's compulsory to vote, and I think there is only one other country that does that. I'm not sure which one it is. And as a result there are very extensive demographic registries that you can with approval gain access to and so we were able to identify a group of demographically matched controls that we recruited to the study, obviously not collecting their tissues but actually collecting their blood and dietary and epidemiological information.

And what we intended to do with these resources was really three things. One was to look for molecular subtypes; the work from Pat Brown and Chuck Pro and others identifying the molecular subtypes of breast cancer had only just come out a year or two earlier and we hypothesized that perhaps it would be similar molecular subtypes in high-grade serous cancers or in ovarian cancer and that was one of the things that we'd look at using micro-arrays which we at that time manufactured so this is going back quite a ways. We'd look at an-epidemiological risk factors which of course had been looked at previously for ovarian cancer but we wanted to sub-type them by histotype to see whether there are associations that might have been missed if things were lumped together. And we wanted to analyze candidate low-risk genes.

So really three parts; look at the biology of tumors, particularly molecular subtypes, the epidemiology and link them to histotypes and look for candidate low-risk genes. And it amazes me to sort of stand here or particularly you know a few years ago when all of this work kind of came to fruition to find that-that it actually worked. It was one of those grants where everything that we said that we planned to do in terms of assembling the cohort and the things that we aimed to do with it actually turned out. And I-and I say that not because we normally don't deliver on grants or something like that but you know it's research and you don't know how things are going to turn out. And to sort of hit the bullseye on all these different parameters still is a surprise to me and I think it was really a testimony to really strong collaborative spirit in Australia particularly in our ability to put together this cohort.

And so this is-this is sort of a graphic representation of it. These are only some of the clinical centers that we interacted with. Basically what we were able to do with the-with the OCRP grant was to place research nurses in the major treatment centers that would see women presenting for surgery for ovarian cancer in the capital cities. And they're sort of shown on this inter-group here, so even though Australia is about the same size as the lower 48 States, you can cover around about 65% of the population by targeting about six or seven cities. It's a very highly urbanized population. And so by placing research nurses in these centers for the surgery we could-we could cover a very large portion of the population. About 30% of those women then went back to rural and regional areas, and so we needed all these other centers and I think in the end it was over 100 centers that allowed us where we had ethics in place and allowed us to follow these women through the course of their disease.

And with some extra help from OCRP we extended the study a little bit by about 12 months and then rather than recruiting 1,000 cases, we recruited 2,500-two and a half thousand cases presenting for surgery through this period of 2002 and 2006. As we promised, we collected blood on over 90% of the cases, fresh frozen tissues on about 65% and we collected the controls. And the-the clinical outcome for these women we're continuing to track that. About 40% of the women have died now but the remaining women we're continuing to follow and we have-we have lost very few of those women to follow-up.

I don't think you'll be able to read this but on this slide and on this slide the publications that have come from the use of AOCS as a resource, so there's about 120 of them so far and we're publishing, we're-not just us as investigators but people making use of AOCS as a resource for ovarian cancer research are publishing about two papers a month at the moment, about 20 to 25 papers a year. And these are some of the findings that just sort of kind of are the headline findings that have come out of it. It's become to our surprise I guess, probably the largest molecular epidemiological study in the world for ovarian cancer. As I said about 120 papers since 2007 and that's how long it took for the-the study to sort of mature and to create a cohort and a resource. Not only has it supported the research programs that were originally envisioned in the-in the program grant, but it's supported a large number of other projects by other investigators in five or six countries around the world through a structured process.

So AOCS is not a private resource; it's a resource that anyone can access through a structured application process and as long as they can show, you know, that it's a good use of the resources. AOCS has become a major contributor and really a driver particularly for the Foundation of OTTA and OCAC. These are some of the international cohorts of cohorts if you'd like to-for ovarian cancer research and some of the things that we've done are-obviously I won't have time to go through the papers, but we have done a comprehensive mapping of the epidemiological risk factors as I suggested and looking at them by histotype. That's work that's been led by Penny Webb in particular. Identification of multiple low-risk alleles; you will have seen in the most recent edition of Nature Genetics, there were a whole series of papers from iCOGS and AOCS I think contributed to five of those papers and was really a lead driver or Georgia Chenevix-Trench was one of the lead drivers on several of those papers. She's one of the founding AOCS investigators.

As I'll describe in a moment we have identified novel molecular subtypes of high-grade serous cancers as we had hoped we would find. We were discussing earlier about this stuff actually getting through to the clinic and we have just completed-Dana DeFazio has just completed her first investigator initiated clinical trial of looking at host genotype and its relationship to the handling of chemotherapy. It's a very interesting study. And with the Dana Farber Institute, we're commencing a collaborative clinical trial we hope in early 2014 on our studies on Cyclin E which we have developed as a therapeutic target identified as a therapeutic target in high-grade serous cancer and Karen you mentioned the-the grant that we were awarded in 2007 and I'll touch on that as well briefly. But that's played a major role certainly in Australia and I think in other countries in changing the genetic testing guidelines for ovarian cancer patients in both the timing of the-of the-of the testing and who should be offered the testing.

So I didn't really want those slides to be there as an advertisement for AOCS but rather just to acknowledge the tremendous impact that the grant back in 2000 has had not only on our research but a whole program of research for ovarian cancer. And I'd also particularly like to acknowledge Pat and Teresa who I see out the back and the rest of the team for their contribution to the study. They've been incredibly helpful and they've believed in this group that we're way out in the boonies down under a long way away and helped us at various times with the study, so thank you very much for that. It's been extremely important to us.

Okay, so let's get into the science. Oh, I should just say that where is AOCS going now. We-we kind of talk about AOCS too and-and if we could we'd roll this out on a much larger scale, but I think that the-the focus now has really shifted to trying to collect samples and enable research in the relapse setting. This is such an important area and I'll touch on it during the talk. And so for a number of years we've been collecting samples from women with recurrent disease, who were originally enrolled as primary surgical patients and now have returned with progressive disease, and we've collected particularly their ascites which has to be drained for palliative purposes and that's often full of tumor cells and I'll describe some of those studies. And I'm also just a PI for this Rapid Autopsy Study that we just started last year and have had our first few cases which has been a very challenging study to get off the ground but with-we now have a very-I think very sophisticated process in place for doing rapid autopsy-rapid autopsies on women with ovarian-who have obviously had failed treatment to understand the evolution of their cancer. And I think this is a very, very important area of research for solid cancers in general and-and something that's become really to the force and particularly Charlie Swanton's paper on renal cancer in the New England Journal this year and on the evolution late last year and the evolution of renal cancer and solid cancers.

Okay; so let's get into the science. So here-here are the CA125 plots from four women, each with high-grade serous cancer, each with Stage-either Stage 3 or Stage 4 disease so there's-on the Y-axis here the levels of this biomarker on a log scale from 1 to 10,000 here, and this is their-sort of their-their cancer journey over in this patient over a couple of years. And you can see this pattern, these little dots here represent cycles of chemotherapy and-and the-they're color-coded for the different types of chemotherapy and you can see this patient is as close I suppose to a typical high-grade serous cancer patient as you might get where there's a rapid fall in CA125 associated with the chemotherapy and the de-bulking surgery and the cancer falls-the biomarker and other clinical symptoms put the woman into remission for a period of 18-12 to 18 months. But then the cancer comes back and although she doesn't have surgery she again has a round of platinum and taxane-based chemotherapy and you can see that the biomarker drops very nicely. But it doesn't fall as far or for as long and the cancer returns. And she then goes onto other lines of chemotherapy. Again it falls; again not as long or as-as far, returns-more chemotherapy; now the cancer is resistant to treatment and unfortunately the woman dies here.

And as I said, that-that in many respects is a-sadly a very common pattern for women with high-grade serous cancers. Here's another patient here who has-also has surgery and chemotherapy and the marker falls but then it-this is-sorry the scale is missing off of here, but instead of it being down for a period of a year or two it's only down for a couple of months. And it returns and although it's not shown here, she has more chemotherapy and she has progressive disease and dies of her disease. And so she's actually primary resistant to treatment and has rapid progression and rapid death.

Here's a very rare case. These are sort of so unusual that you wonder whether they're actually really what they appear, but they are in fact high-grade serous cancer patients who are-have what some people call one and done. They are treated with chemotherapy and surgery, and they have these really remarkable responses to treatment and they can span many years. And-and we have found some of these patients through AOCS when you recruit two and a half thousand women you can find the handful of women who have these extraordinary responses and I'll-I'll talk about them a little bit in a moment.

And then you have a somewhat similar patient in the sense that they have a very long survival compared to the norm but instead of having this kind of pattern of one and done, they relapse but they continue to respond to platinum and over and over again rather than becoming resistant.

So four women, four very different responses; at the moment-

Q: Are these women routinely given antibiotics during platinum treatment?

A: Antibiotics? No; no, they're not. So four different responses; upfront, we don't have any really clear guidelines or very few guidelines to tell whether a woman is going to turn out like this or like this or like this or like this. So this is really what the talk is about. How do we try and understand what determines this outcome? How can we use that information to help the patients? And what can it tell us about the biology of the disease? So these next two slides if you remember nothing else about the talk, I think, try to take home this message that if we've learned anything over the last few years it's that ovarian cancer is an extremely heterogeneous disease, both between and within histotypes and what's really bedeviled the field I think is that it's sort of figuratively we have treated it you know as a-as a single disease but in fact it's sort of a fruit salad here of a whole lot of different diseases. When we do our omic analysis on it as a single entity, it's very, very difficult to find consistent patterns and in fact what we have to do is we have to split it out into its various types and then to ask the question in a much more refined way. And when we do that, we actually find associations that we had missed before. And I'm just going to give you two examples of that.

Q: Pardon me-we went over the 4 different charts (sic; actually CA125 plots) from four women with high-grade serous cancer-I was only curious about one thing. Do you know what the percentage of your total group falls into each of the four categories?

A: Yeah; so-so these-the headline figure for women with high-grade serous cancer would be of the order of 20 to 30%. These women are vanishingly small, less than a couple of percent. These are a bit more common, a few percent, and these make up the majority of them.

Q: The majority-70-80%?

A: About 70 or 80%. So okay so here's this first example and it's from Martin Kobel's study and essentially what they did was that they took a number of biomarkers and then they analyzed them against a mixed cohort or histotype specific cohort and they looked at clinical outcome. And I'll just show you the data for WT1 which is a very useful biomarker for high-grade serous cancers and this is the analysis of survival by WT1 expression. This is high-level-high-expression or-or positive for WT1 is in the red or negative for WT1 is in the blue. And you can see and-and this is cumulative survival over a period of time, and the women who express WT1 have a poor outcome versus women who don't express it who have a relatively good outcome.

And so when you look at the entire cohort you would say that WT1 is a-a marker of-of bad actors-that it's a marker-that it's an indication of-of an adverse disease. Yet when you refine the question and you just look at the high-grade serous cohort, it actually switches around. So now women who are WT1 positive within this cohort have a good outcome. And so what that's saying is that you cannot ask a question like that with a heterogeneous cohort. You have to refine the question. And here's a second example. This is actually our work and we were-one of the things we were interested in was the role of p53 as the potential marker of response to chemotherapy. Now there's a literature of over 70 publications and I-I know this because I've actually looked through all of them, spanning over two decades of-of associating p53 mutations with clinical outcome particularly chemotherapy, primary chemotherapy resistance.

And what these studies showed was a mutation frequency of 35 to 80% and about half the studies found an association with clinical outcome and half the studies found that p53 was not an indicator of how the women were likely to do.

Now that has-there are two things here that sort of should ring alarm bells. The first thing is that the mutation frequency is extremely wide. So how can that be? And the second thing is that half the studies agree-feel that there is an effect and half of them don't, so you know one of them has got to be right. And when you look through the studies they have a number of limitations; they're particularly limited by sample size. Ovarian cancer tends to be an uncommon disease. And so samples and numbers are often small. Because it's uncommon, people kind of throw a whole lot of different cases of different types into the study design so they usually are very heterogeneous. They often lacked comprehensive clinical information and just from a technical point of view, often they-the way that the mutations were analyzed were not optimal-for example, relying on immunohistochemistry.

So just like the Huntsman study but for p53 we refined the question and we said okay now if we just ask that question about high-grade serous cancers and we design the study so that we explicitly include just an unselected group plus we seed it with patients who we might expect to see an effect for p53, so primary resistant cases, we do it properly in terms of the way that the-the mutational analysis is done by sequencing the entire gene and all the splice sites and now we find a very different answer. We find that in fact the mutation frequency is essentially 100%; 97.7% of the cases had pathogenic p53 mutations and the remaining 2.5% were probably actually not even high-grade serous cancers but misclassified as high-grade serous cancers.

And so that says two things right away; first of all there can be no association between p53 mutation and outcome or at least p53 mutation per se, right because they all have p53 mutations. And then secondly it tells us something about the biology of the disease so that this mutation which turns out to be very, very early on in the genesis of these cancers is absolutely fundamental to their biology. So I won't-don't have time to go into it today but that's taken us into a direction of understanding how this is one of the key pieces of the puzzle in understanding what drives these cancers.

So those are two examples of when you ask a general question you get a misleading answer and when you ask a much more refined question you get a much more useful answer.

So what our work and particularly work from Chris Crum and Bob Kurman and a number of other investigators have shown over the last 5 or 7 years is that as I said, ovarian cancer is a very heterogeneous disease. And essentially anatomy has been very misleading in this case. To call it ovarian cancer has been a misleading term. We probably really should call it pelvic cancer except it would be very confusing I think to-to switch the nomenclature for a lot of the public and that these really are a whole collection of very different and distinct diseases that essentially just share an anatomical location. And as I'll show you in a moment, these cancers individually have more in common with other cancers at other anatomical sites than they do with each other.

So to give you an example of that, this is some work that we published just a couple of years ago and followed some very nice work from Mike Birrer's group that was published a few years earlier than that-excuse me which showed that ovarian clear cell cancers molecularly resemble renal clear cell cancers in terms of their biology marker shown, that in terms of expression profiling-I won't go through the details but we showed that there is a pathway of pseudo-hypoxic drive in these tumors where there's a constitutive activation of the HIF pathway which drives a proangiogenic response that's very similar to what's seen in renal clear cell cancers and this information turns out to be actually very useful-that a cancer-a therapy that's used in renal cancer, at least in a small number of patients that we're able to evaluate seems to be very effective in ovarian clear cell cancer patients. So these are cancers that are typically resistant to standard of care which is platinum/taxane treatment and the response rate is of the order of 10 to 15%.

And here's a woman who was treated with platinum and taxane, and you can see her CA125 fell very sharply here, probably mostly associated with the de-bulking surgery rather than the chemotherapy but almost as soon as the chemotherapy is finished, she has a rapid return and growth of the cancer as evidenced by the biomarker. It just blows straight through a second line or a second line of treatment and through chemo she has a very short response, but she essentially has very rapidly progressing disease that's resistant to conventional chemotherapy. She was started on Sunitinib, which is a drug that targets VEGF signaling which is constitutively activated in these tumors and is used in renal-renal cancer and as you can see she's had a very rapid fall in her CA125. It's rebounded during the 2 weeks off the drug and then fallen again when she's come back on the drug, rebounded again 2 weeks off, and then it's fallen again and stayed off for a period of 2 years while she's remained on the drug and then she's progressed after that. But then has gone onto another treatment, an mTOR inhibitor that has also been used in renal cancers and had a very good response to that.

And so this is a very substantial and durable response in this particular patient. Not all these patients respond to Sunitinib-using a drug regimen that is not in ovarian cancer but is used in renal cancer.

Here's another example of a molecular parallel; this is some work that we published associated with the p53 study and-and back in a Nature Review's cancer and also have been highlighted in an analysis of triple negative breast cancers that was published by TCJ and Nature last year. And what-and the parallel here is between high-grade serous cancers and basal like cancers. And they both share this very high frequency of p53 mutation of RB loss, of Cyclin-E amplification which I'll talk about in more detail in a moment and this very high-frequency of-of disruption in the BRCA pathway and some work that is just under review at the moment shows that this parallel crosses between high-grade serous cancers and basal cancers particularly around BRCA 1 mutations where there are specific chromosomal amplifications particularly involving in 8q24 and some very-and some loci on the X-chromosome that is shared between these two diseases.

They also share an intense immune infiltration which is probably important in terms of-in patient outcome.

Q: Are they myloid suppressor cells?

A: That is particularly a T-cell infiltration that we see in those. So-so this is a very shorthand way of-of I think making another point that as I said ovarian cancer is very heterogeneous. The disease shares an anatomical location; those diseases are distinct, and we need to look to other types of cancers to find molecular parallels because they are similar-more similar to other diseases both in terms of their molecular biology and their therapeutic response.

Host genotype also makes a major contribution to clinical outcome. This is the 2007 DoD award and so what we did was we dipped into the AOCS resource and we wanted to look at BRCA mutations in an unselected cohort of patients who were presenting for ovarian cancer. For many years now we have known what the risk of developing ovarian and breast cancer is if you're carrying a mutation. What has been less clear is what is the probability that a patient would carry one of these mutations if they were simply presenting with ovarian cancer whether they had a family history or didn't have a family history, and how would the presence of a mutation affect their response to first and subsequent lines of chemotherapy.

So we went ahead and tested 1,000 women from the AOCS cohort that were population-based incident cases. We focused on the invasive cancers. We excluded borderlines and cancers of mucinous histology and focused on serous, clear-cell, endometrioid, and took those where we had blood samples available and some of these had already been to a Familial Cancer Center and we directly mutation-tested about 950 of them by Sanger sequencing. And this is what we found; we found that the mutation frequency was certainly much higher than the conventional wisdom of 5 to 10%. It was actually 17%. Almost all the cancers were in high-grade serous cancers. We did find some in clear cell and endometrioid cancers but when we went back and re-reviewed these cases they turned out to be actually high-grade serous cancers that had been misclassified as these two histologies. So it's essentially like the p53 story; it tells us something about the biology of the disease. It also taught us something very useful clinically in that about 44% of the women who carried these mutations did not have a sufficient family history that would be sufficient to suggest that they be offered genetic testing. So the current genetic testing guidelines in Australia would miss around about half the carriers. And-and it's been these results that have really driven the reform of the genetic testing guidelines in Australia for women presenting for ovarian cancer and any women that have these histologies now are offered genetic testing to pick up these mutations.

Q: Are these somatic mutations or-?

A: These are both; these are germline mutations. These are germline mutations. There's another 7% of somatic mutations on top of that.

So how does that affect the outcome of these women? It has been known for some time particularly from studies on Ashkenazi Jewish women but also around the time of our study in larger population-based cohorts that although women who carried these mutations had the disadvantage that they would-are at-risk of developing breast and ovarian and some other cancers they had the advantage that they had a relatively favorable survival compared to non-carriers. And this is just showing graphically in the data from that cohort the BRCA 1 and the BRCA 2 as compared to the overall survival of the non-carriers.

One of the strengths of this study was that we had actually determined the mutation status of a very large number of women including and excluding women from-who might have carried a mutation.

This is a very sort of busy slide and it's been published just last year in JCO so I'll go through it relatively quickly. And hopefully you'll get the flavor of it. Basically the way it's color coded is blue means response and red means resistance. And in this group here, these are the women who carry a mutation and these are the women who don't carry a germline mutation. This is their response to primary treatment and this is their response to-to second line treatment after they have relapsed. And what you can see is that the carriers all have more blue than the wild types. So both to initial treatment and to first line responses, carriers respond better to chemotherapy.

They also respond better to non-platinum based chemotherapy in a relapse setting, so you can see there's more blue here in the carriers than in the non-carriers. This turns out to be important because clinical trials are sometimes designed, for example, a recent GOG trial was designed on the basis of the predicted response rate in a-in a second line setting of women with high-grade serous cancers since the majority of those are non-carriers. The response rate in carriers is estimated on the basis of the response rate largely in the non-carriers but you can see they're quite different and if you do that you can be caught out and they were caught out in a GOG study in that their response rate in-in a second line setting is different in carriers versus non-carriers. So basically knowing this can influence the way that you design clinical trials.

Carriers also respond-or even though they respond better to non-platinum than non-carriers, they actually respond best to platinum. And so this-these findings have helped make the case that it's important to determine the carrier status for a woman at the start of treatment, not at the end of her chemotherapy because it influences the-the degree to which clinicians persist with platinum, including in an unusual group of women-this is a group of women who are carriers who technically are regarded as resistant to first line treatment because they have shown some level of relapse within 6 months of the end of chemotherapy but in fact these women normally would not be treated with platinum in a relapse setting because they're regarded as primary resistant but in fact some of these women were treated with platinum and the vast majority of those showed a resist response in a second line setting. So again this has underscored the importance of persisting with platinum in carriers and determining their status at the outset and not at the end of their chemotherapy and-and so it's influenced their care.

If we now look at not just first line and-and relapse but right out to the third line you can see in carriers again more blue. They continue to respond to-this is now just platinum-they continue to respond to platinum at least in a proportion of them-a quarter of them are still responding to platinum out to the third line of treatment compared to the non-carriers who only about 14% of them have continued to respond to platinum. So clearly having those BRAC mutations which inhibited homologous recombination, repair, and interfered with the removal of the bulky adducts that platinum causes is really pivotal to the reason that high-grade serous cancers respond to-to platinum-based agents. And so it sort of begged the question for us why did these-who were the non-carriers that unusually kept on responding to platinum out here? And when we go through and we look at these we find that in fact these cases are enriched for not germline mutations-they're non-carriers, not for germline mutations in BRCA 1 or BRCA 2 but actually somatic mutations in either BRCA 1 or BRCA 2 or methylation of the BRCA 1 promoter.

So these are women who have phenotypically arrived at the same space as the carriers by virtue of a somatic disruption of their pathway.

One of the things that we've wondered about is if-if carriers do well why don't they all do well. This is the data from the study showing the overall survival in carriers versus non-carriers. This is the median overall survival for the non-carriers at around about 36 months; you can see that the carriers are doing better than median survivals closer to 60 months but there's still a decent proportion of women out here who are carriers who do poorly. And-and so again because we have got such a large study we can start to try and pick this apart.

One of the things that we wondered about was whether the type of mutation, the type of germline mutation might affect their response to treatment or their overall survival. But we don't see this in-at least at this level of analysis of breaking down the mutations by the particular type and what we were particularly interested in were these large deletions. And the reason for that is that work from Allen Ashworth and more recently from Liz Swisher has shown that there is reversion of the germline allele in a substantial proportion of women who carry these germline mutations to restore some activity to the BRCA 1 or BRCA 2 genes. And what we wondered was whether the type of mutation influences the-how readily they revert and become resistant to platinum and we expected that large deletions which in principle should be very difficult to revert would be-would result in longer survival in these women, longer periods of responsiveness to platinum but we don't see this. And so I think this is a real conundrum of why we have this disconnect between this high reversion rate which seems to be associated with resistance and yet we don't see a genotype/phenotype relationship at least in our analysis of the BRCA mutations.

What other things might affect outcome? One of them is de-bulking. It's non-it's probably one of the most reliable indicators of outcome in women with high-grade serous cancers, women who are non-optimally de-bulked at primary surgery have a worse outcome than women who are optimally de-bulked. And we can see that here's in red carriers who are optimally de-bulked versus non-carriers who are optimally de-bulked you can see there's this large survival difference between them, but what this analysis shows is that being a carrier does not remove the requirement to be optimally de-bulked because a carrier who is sub-optimally de-bulked is pushed back almost to the status of a non-carrier. So if there is any interaction between these different parameters that affect the outcome-this one a clinical one and a molecular one that can intersect.

But even when we exclude the type of mutation and the degree of de-bulking there are still differences between the women, and now we're really starting to get down to very small numbers because you know we started with 1,000 women and we went to the carriers and then we went just now to the women who carry exactly the same mutation. These are two different founder mutations in the Ashkenazi Jewish population. This is overall survival in months. So this is one mutation, and this is a different mutation in a different group of women and you can see even though when we control through optimally de-bulking and the type of mutation there are still differences in the outcome. So there are still things to be found when we sort of peel the onion; you know we go from ovarian cancer to high-grade serous cancers to carriers which is affecting outcome to removing de-bulking, to removing the influence of the type of mutation, there are still things in these patients that are influencing their outcome that we need to understand.

We can also learn from the outliers, these long-term survivors and this is a study that we're just-that it's still just in progress and I'll just share some non-published results with you. And it's actually part of yet another DoD grant that Lee Pierce and Malcolm Pike are the leaders on and that AOCS has been very happy to contribute to of-of just starting now to look at these long-term survivors. And so this is some work that we've been doing just before the DoD grant was awarded. This was some work that goes back now a couple of years with Anna DeFazio particularly who is one of the investigators for AOCS.

And so these are the-these two unusual types of patients that I described at the beginning of the talk there's extraordinary patients who have responses almost like what you see for testicular cancer and then what we call these relapse responders, the ones that don't have a particularly short-particularly long progression free survival but continue to respond to platinum based treatment. So we've-we have delved into AOCS and controlled for all the variables that we can-know of; we-we included women who had macroscopic residual disease so in other words, we didn't want women in this category who had been cured surgically. So we wanted women who-who had to be having unusually good responses to their chemotherapy. And we used clinical criteria to-where they were far outside the norm.

One of the first things you wonder is whether these are really not high-grade serous cancers, so we've reviewed them back and forward in all sorts of different ways and we find that their pathology resembles that of a typical high-grade serous cancer. So they are high-grade serous cancers. And what we find in the multiple responders what is here is a-is a graphic showing that these are enriched for disruption of the BRCA pathway so a little over half of them have either germline or somatic mutations either point mutational methylation in parts of the BRCA pathway but surprisingly in this extremely sensitive cohort they have the same frequency of disruption of the BRCA pathway as women who have a more typical course of their disease.

And so there is something additional going on in there, something really unusual that is giving these women these incredibly good responses to chemotherapy. And we're now as part of the ICGC project in the midst of sequencing these women. We're just getting sequencing results back from them, both their tumors and their germline to try and understand what might be driving this extremely good outcome in these-and this data will play forward into the DoD grant.

Differences in outcome are associated with molecular subtype and I'll run through this fairly quickly because time is getting on and-and much of this work has been published. As I mentioned, back in 2008 we identified these four molecular subtypes which we termed C1, 2, 4, and 5 of high-grade serous cancers. In the TCGA study that we're involved in back in 2011 of the 500 high-grade serous cancers completely independent set of samples, the-the four subtypes were really beautifully recapitulated in the TCGA data where reference 25-I think they could have said total et al but anyway we were reference 25 and-and despite our efforts to try and change some of the investigators' minds those were renamed. I actually still stick with the C-subtypes not just because we discovered them first but I think when you give them a name it sort of fixes in your-in your mind an idea about what the biology of these things are and I think we still don't really understand the biology of these different subtypes and so that's the reason why we've kind of stuck with these rather neutral names for the subtypes. But nevertheless, just like in breast cancer there are four very distinct subtypes of high-grade serous cancers and like breast cancer, these are associated with differences in clinical outcome, three different data sets; this is Mike Birrer's data set; this is the TCGA data set, the AOCS data set and we can see that C2 patients as a group always do the best. C1 and C5 patients do the worst, and this is the sort of overall analysis of only 1,000 cases and you can see there are very distinctly different outcomes nearly double the length of survival between the best and the worst subtypes for these two groups-these four groups.

I'm not going to go through it in detail but this is another good example of if you take the bowl of fruit and break it into its components now breaking the high-grade serous cancers into these four subtypes you can find something that was not apparent before. So this is some work that Aslaug Helland did in our lab now about 18 months ago showing that this signaling pathway in one way or another is activated in the C5 molecular subtype. We see MYC amplification and MYC over-expression-sorry NMYC over-expression, NMYC amplification and a downstream signaling pathway that fits with this very, very specifically in C5 tumors. And so NMYC, which is-was first discovered in neuroblastomas-pediatric neuroblastomas has never really been associated with high-grade serous cancers before. But when you break apart the groups and you isolate them and you ask a narrow question then suddenly you can see that this signaling pathway pops out as being a key driver and that there are a number of therapeutic target options for this group. And they would represent a sizable group to be targeted, certainly you know as a subset, a decent subset as say some of the molecular subgroups in non-small cell lung cancer.

We have been very interested in determinants of primary treatment failure; these are the women who have that sort of initial response and then it sort of pops back up. And so again making use of the AOCS cohort, Dariush Eternalmogham in the lab did a copy number analysis and a gene expression analysis of women who were matched for all parameters-histotype, de-bulking, numbers of cycles of treatment, and so on, except the differences in their response to initial treatment, and found that this amplicon at 19q12 was the dominant amplicon associated with primary treatment failure and you can see that women who have amplification do very poorly compared to those that have-that are unamplified and the women who are-have a gain at that locus fall into the intermediate group.

And to cut a long story short we stepped across the-narrowed the-the region down and went to cell lines that had this focal amplification and stepped across it with an siRNA screen and found that a key driver in that region was Cyclin E1 a cell cycle protein so that-and this is a very clear oncogene addiction so in unamplified lines knock down of Cyclin E1 does nothing; in amplified lines it's a very strong attenuation of clonogenic growth and we were also able to validate in this unpublished work its protein partner CDK2-is the key partner for Cyclin E and we see the knockdown of CDK2 essentially phenocopies the knockdown of Cyclin E in terms of this amplification dependent sensitivity to attenuation. And there are drugs to CDK2; this is a Pharmacia but subsequently a Pfizer compound showing a gain in amplification-dependent attenuation of growth with the CDK2 inhibitor.

Interestingly, in the TCGA analysis of the high-grade serous cancers they-we found that Cyclin E amplification is mutually exclusive with BRCA inactivation, not perfectly. There's an overlap of about 8% but there's a very distinct enrichment of Cyclin E amplification in patients who do not have inactivation of the BRCA pathway. And that's important again for a couple of reasons. One because it gives us insights into the biology but also it's this group of patients out here who are very unlikely to benefit from platinum or PARP agents and who need other therapies and so identification of CDK2 is particularly important in this group in terms of providing a therapeutic strategy for a group that-that's unlikely to benefit from some of the molecular therapies. And in the AOCS data set, we essentially find the same is in the TCGA data. The amplifications are associated with the women who lack the BRCA mutations.

So we wondered why-why should it be that BRCA mutations and Cyclin E amplification should be mutually exclusive. One possibility is that these both cause chromosomal instability and they may be two different ways of becoming a high-grade serous cancer, so there's no selective advantage of having both mutations in a sense. And so you never really see them together. And you see this sort of thing in low-grade cancers; you very rarely see different-you know an N-RAS mutation with a K-RAS mutation or a B-REF and a K-RAS mutation because these are different ways of activating the pathway and-and for that reason are mutually exclusive.

The other possibility is though that there may be actually deleterious when they occur together. Because both cause chromosomal instability that they may both together be deleterious for the tumors. If I had to-well I did bet on this and I would have bet the former-just two different way of making high-grade serous cancers-and I was wrong. So it turns out when this is a collaborative study that we did with Bill Hahn and Barbara Weir at the Dana Farber and it turns out that BRCA 1 mutations are synthetically lethal with Cyclin E amplification. And I won't go through this in detail because we're running out of time, but basically highly statistically significant and-and a significant biological effect for inactivation of BRCA 1 in the presence of the Cyclin E amplification.

The reason-and interestingly we also saw a larger effect with members of the ubiquitin pathway. And I'm going to go through this quickly but basically that brought together two streams of observation to suggest that inhibiting the BRCA-inhibiting the proteasome pathway which turns out to be an effective way of inhibiting homologous recombination and repair work from Alan D'Andrea and others is potentially a therapeutic approach in women with Cyclin E amplification. It's a way of targeting this synthetic lethality and in fact that's what we see. So we see this amplification dependent sensitivity to Bortezomib which is a proteasome inhibitor in these tumors that have Cyclin E amplification. So we believe that this opens up a second therapeutic option; one is targeting CDK2 and the other is targeting the proteasome for these patients with Cyclin E amplification and that's the basis of the clinical trial that we believe will start in the first quarter of next year.

So I just want to finish in the last couple of minutes with what determines response to therapy in the relapse setting and to me this is actually in many respects sort of the elephant in the room for high-grade-well particularly for high-grade serous cancers, for ovarian cancers, in general, but particularly for high-grade serous cancers. We have a very good therapy as a first line therapy, platinum works in the majority of women and they get very good responses. They're likely to respond to platinum in their-in a relapsed setting but resistance almost invariably develops and-and so this is really-and when it develops we have essentially no guidelines, no biomarkers to decide whether a woman should be treated with Topotecan, Gemcitabine, or Liposomal Doxirubicin and it's essentially a dealer's choice. We don't know whether she's likely to respond; we don't know which drugs she's most likely to respond to. And so that's actually a very difficult clinical situation and we need biomarkers in this recurrent setting. And so-and-and in fact this sort of-actually this slide, I forgot that I put this in-really underscores what I was just saying. These are patients across the top. These are lines of treatment. So this is the first line treatment; this is sort of probably 12 or 18 months later; this is the second line, third line, fourth line, and-some-one women ended up having in this series had 8 lines of treatment.

And what you can see is in the first line setting there is-there is great consistency and essentially women are treated with platinum plus or minus Taxol or just different types of platinum and sometimes with or without Taxol but basically it's very consistent. You can see in the second line setting it's mostly platinum and Taxol but then it starts to switch to other things, Caelyx for example. But this is what I was talking about; once we get into the subsequent lines it's essentially a shotgun. They could be almost treated with anything here because there are so few guidelines to treat-to decide how to treat these patients.

So this is some work we've been doing for the last few years of trying to look at changes in space and time, in the changes that occur in the genome across the tumor in different locations around the body removed at primary surgery-I won't show that data-but it's just published last year and also in time. What happens that allows this tumor to evolve from this point where it was sensitive to platinum and taxane to out at this point where it's resistant.

Now what we've done here is we've collected samples pre-treatment and in this recurrent setting and-and this is the summation of a whole lot of different chemotherapy here. And we're looking at the evolution of this genome and-and we're wanting to understand how far it has evolved and what are the genes that have changed. And one of the things that you might-don't worry too much about the colors here, but you might imagine that the longer the period of time between the initial treatment and when we collected the relapse sample might affect how far the genomes evolved. But in fact we don't see that; if we were to-if that was the case we'd see a nice line on this axis. This is the months between primary and secondary and this is the number of copy number changes between those two samples.

And it's essentially-there's not a relationship.

There is also not a relationship on the number of-by the-sort of a similar parameter the number of cycles of chemotherapy. These are DNA damaging drugs; you could imagine that each line of chemotherapy causes more genomic change but again rushing through the data here a little bit we don't see a relationship between the number of lines of treatment and the extent of change. What we do find is that the distance the tumor travels depends on the sensitivity to first line treatment. So patients who are resistant from the get-go, their tumors when you collect them at relapse irrespective of how long that's been have traveled less far; they've evolved less distance than patients who originally were sensitive to treatment and now have become resistant. And we see this particularly in sensitive-resistance patients who are not BRCA mutated. So those that carry BRCA mutations don't travel as far, perhaps because there is an easier route to developing resistance possibly by reversion of the BRCA allele. But those that don't carry the mutation travel the furthest and that kind of makes sense if you think that these tumors are evolving over time.

I guess more importantly what we've been able to do is to look through these patients to find genes that have changed between prior to treatment and the development of resistance and one of the genes that we focused on in this study that was published last year in Cancer Research is LRP1B. We find that in sensitive resistance, so initially sensitive and become resistant cancers there is down-regulation of LRP1B at the RNA level and we see this in an independent cohort. We don't see this in patients who are resistant initially and remain resistant; we don't see it in ascites samples that are compared to the corresponding tissue samples, so it's not a function of-of-of LRP1B being low in ascites material. And then functionally what we find intriguingly is that when we overexpress this-this gene we see an increased sensitivity to liposomal Doxirubicin but not to Doxirubicin. So this is a Doxirubicin derivative that has been chemically modified to increase its uptake into tumor cells. What's happened here is that when we overexpress this protein which is associated with lipid transport we see an increase in sensitivity to this drug and then mimicking what we see in these patients when we lose LRP1B we have a reduced sensitivity to liposomal Doxirubicin, but not to Doxirubicin. So this is sort of the first example of where the resistance that is developed is actually not to the drug per se, but actually to the carrier of the drug.

So I'll finish at that point-I see my time is momentarily up-by saying you know what are the determinants of clinical outcome? This is clearly still a work in progress from our lab and many other labs. I think it's now really firmly established that these are very different diseases. Anatomy has been misleading; the term ovarian cancer is misleading. The treatment of you know kind of one size fits all that day is passing rapidly now. And we are certainly moving to histotype-specific treatments for these different types and we will eventually move like we have in other solid cancers to molecularly based treatments.

We still-we need to put together a pie diagram that maps all the different ways-all the different things that can affect how a woman fairs in terms of her outcome. De-bulking status we have known about for a long time; germline mutation is very important, not just in terms of overall survival but in terms of the patterns of response to chemotherapy, the molecular subtypes I mentioned-have not gone into any detail for a very large and important field of immune response. Work from George Coukos and-and a number of other labs show us that this is very important and it drives one of the molecular subtypes of high-grade serous cancers, C2s. But there are still only important known and unknowns-of why some of the BRCA carriers don't do well and why we have these extraordinary responses in a very small number of women to chemotherapy. We need to understand a lot more about what's happening in a recurrent setting. These tumors are tumors particularly the high-grade serous cancers are driven by copy number and they are evolving a very long distance and to collect samples at surgery and make treatment decisions 2 years later based on those samples is really, really flawed and it has to become standard of care that we're collecting samples in a relapse setting and we're learning from those and we're identifying biomarkers like LRP1B and work from Hani Gabra and others have-involving AKT, so that we again can map all these different determinants of treatment, determinants of treatment failure, and so that we can develop therapeutic strategies to target those and bring sort of ovarian cancer into sort of much more of a modern era where we've got these molecularly based therapies that are targeted to the underlying molecular defects in the cancers.

And I'll finish just by acknowledging the people. I don't think there's time to read out all the names; the ones in red are those that particularly contributed to the studies that are highlighted in blue, and of course to thank our funders of these studies and particularly again to thank the program and the members of the program for your wonderful support for this research; thank you. [Applause]

Q: Are there any thoughts about combining the chemotherapy within the immunotherapy during the time of remission?

A: Yes; so I'm not across them but there are trials of drugs like ipilimumab, a CTL4 inhibitor and you know an AdCTLA4 inhibitor and conventional chemotherapy. I don't know how those trials are going but just like other solid cancers those things are being done at the moment.

Q: BRCA 1 mutation-if you go back to the cancer genome map it's one of the things that they talk about is that BRCA 1 is-that there is no real correlation-do you see the same-?

A: It behaves-?

Q: Mutation in the triple negative breast or the basal line breast cancer but it's-but it's like a-it's BRCA 1 nulled.

A: Right.

Q: And we know why that is but hypoxia for instance will actually drive BRCA 1 down-. I'm wondering if that's been what you're seeing here with the different outcomes, just hypoxic?

A: Oh why do they do better? Look; I think it actually is more direct than that. I mean platinum causes these bulky adducts and-and BRCA 1 is part of a complex that is needed to resolve those. And so I think that there's probably quite a direct link between BRCA mutations and the response to platinum and that's certainly what we're able to map in that large study. We-these patients responded extremely well to platinum and that's why it's so important to-to persist with platinum. When-when conventionally they might be regarded as being-having a short PFS; progression free survival. You should stick with platinum until they're clearly progression-progressing on platinum. So I think that that's part of it.

The BRCA 1s we also see particularly enriched for this C2 molecular subtype. That was the red one that had the generally good outcome and they have this very florid T-cell response, this intra-tumoral T-cell response that George Coukos I think mapped first as being associated with particularly good outcomes. So it may be that-that causes a degree of chromosomal instability and generates neo-antigens that leads to an improved immune response. So that might be part of the-the picture as well.

Q: So you have a tremendous resource with all your-with all these tumors and I know that you said a lot of times that there's a lot of differences across the tumor types. There's a lot of tumor-but what about within one patient? I think a lot of the times one drug, one mutation doesn't pan out in these types of tumors. So what about the intra-tumoral heterogeneity? Do you have any insight into that?

A: Yes; I-I guess I was sort of running through that fairly quickly at the end. I mean this is a way of graphically showing it. I have other slides I could show you afterwards that actually show you the primary data. This is-this is the number of copy number change events between the primary and the secondary; it's an extraordinary number. In some of the patients we see 60, 70% of the genome different between them. That's particularly true of women where we collected them in time, so where there's been a couple of years so there's been a lot of opportunity for evolution. But in the slide that I'm thinking of there's also a lot of heterogeneity right there at the get-go. So there's a-you know there's a patient where we've collected five samples and one of them is a mesenteric lymph node sample and you can see very specific copy number events that are associated with that sample that are not present in the others.

And you know this is affecting hundreds of genes, thousands of genes, so you know it's fundamental Darwin evolution, you know you need variation and you need selection right and then you get evolution. And so there's heaps of variation present in many of these patients at the get-go and so that's a great substrate for drug selection. So I think it plays to exactly what you're talking about that we either have to target trunk mutations so you know the analogy is like a tree that there are-you know there are mutations that occur right at the beginning and they're in all the samples and then there are mutations that occur late in the evolution there and there are a few of you know the deposits in a patient. We got to be targeting things that are down here in the trunk.

That restricts us in terms of our therapeutic options. So you know identifying trunk mutations is important but whether that will narrow things I'm not sure. So it's-it's daunting actually to find-

Q: So then the disease is already-?

That's right. Yeah; to find-I mean we're sort of-if you think about it sort of the journey in the whole field has been on and you know we've kind of realized okay this is really not one disease-that pathologists were right. The number of different histotypes, different diseases; okay well actually these even in histotype there are multiple different ones. Okay even in you know within these you know molecular subtypes there is variation between patients-oh well hang on and even in the patients there are maybe 10,000 genomes in these things you know. It's very hard; it's going to be very hard I think.

Q: But if we target something like N-MYC what kind of percentage theoretically we would-?

So the C5 is around about 15% of the high-grade serous cancers so you know 15% of the most common histotype is still is not a very common disease. It makes the design of clinical trials difficult you know. That's been possible in lung cancer because lung cancer is a very common disease so you know something like that might only be in a few percent, but the numbers add up.

I think it's going to be a real challenge to-to design the clinical trials and I think the way forward with that is going to be molecularly guided trials. So not just say a high-grade serous cancer with N-MYC but other cancers that have N-MYC amplification. The problem with that is that I think the context will still be important. We have learned from BRAF mutations that you know there isn't a BRAF-oma; you know even though BRAF mutations is very predictive of a response to BRAF mab in melanoma, but that drug does-is not predictive-is not effective in colorectal cancer when there's a BRAF mutation present because of-of a HGF receptor activation. So-so we're going to subset all these things down and that's going to create problems in terms of designing clinical trials; we can maybe get around that by joining up with other anatomically distinct but molecularly similar diseases and have molecularly guided trials but the catch 22 there may be that although they're molecularly related there may still be context cell lineage dependent effects that will only see effects in subsets. But that's cancer. I mean if we've learned one thing over the last decades is that it's a heterogeneous disease and if you try and treat it as a single entity you are destined to failure. And you have to find ways of working around the fact that just about every cancer is a rare cancer at the end of the day. I don't know; I mean what do you think?

Q: I don't know because you know if we believe Jim Watson over-have you read his article?

A: Which-which particular one?

Q: Jim Watson, the Double Helix-?

A: Yes; yes, I know who Jim Watson is, yeah.

Q: So he suggested the push that it all started-because it's c-MYC across all human types. So but it may be a valid suggestion only if we can target this thing.

A: Well we need to be able to target it. There's actually some-

Q: No; there are chemical problems because it's-there's no drug-

A: Right; there are some interesting agents for the c-MYC inhibition but-but I mean I think this-I don't want to just sort of belabor the point but this is the trouble with saying that c-MYC is amplified in-have I gone past it-yeah I did. C-MYC is amplified in all tumors; you have to refine the question you know. I mean look at-this is AQ24; that's in the MYC locus and that's the frequency of-you're-the conventional wisdom in high-grade serous cancer is 80% of high-grade serous cancers have aberration of the MYC locus. Okay; you'd say oh that's great. It's a fantastic target and almost high-grade serous cancers but look at the data. I mean the actual amplification is really particularly restricted to the BRCA mutation in toted cases particularly BRCA 1. So I doubt that there is going to be-I doubt that c-MYC is actually-even within c-MYC it'll be sub-setted.

Q: I may be you know-have been brainwashed because I am coming from an immunology program. But I think that's probably you know the best-is combination of immune therapy with this-

A: I truly believe that and I think-

Q: -and one of the targets which is interleukin10-which is elevated in the majority of ovarian cancer-.

A: Look; I think-I think immune-immune therapy is probably going to be very important in-in ovarian cancer and-and clearly women who have these florid T-cell responses do better whether that's the florid T-cell response as a biomarker or BRCAness and that's they're doing better for other reasons like I alluded to in terms of the platinum response or they're doing better because the BRCA mutation is generating this florid T-cell response which in turn is attenuating the tumor-we don't know. But you know I mean the recent data with anti PD1 and anti CTLA4 monoclonals is incredibly impressive, so I think everything at the moment is pointing to the fact that we need combinations of therapy and not just combinations of targeting signaling proteins but immunotherapy and chemotherapy I think to get durable responses; yeah.