IMPAKT Breast Cancer Conference 2013
The Cancer Genome Atlas (TCGA): comprehensive molecular portraits of breast cancer
Prof Charles Perou - University of North Carolina School of Medicine, USA
Charles, you’re an expert in TCGA and you’ve got some studies of The Cancer Genome Atlas particularly. Can I ask you, first of all, about your study using The Cancer Genome Atlas of ER positive, HER2 negative breast cancer? What were you doing there?
So The Cancer Genome Atlas wanted to try and characterise a large portion of the molecular heterogeneity of breast cancer. We actually did all breast cancer types – ER positives, HER2 positives, triple negatives – but the story is so rich for each of those clinical subtypes that at the IMPAKT meeting here I focussed mostly on the ER positive HER2 negatives.
It’s a big story and there’s a lot to say, isn’t there?
Yes there is and there’s a lot to say in part because in the TCGA project we did at least five different omic technologies including whole exome sequencing, gene expression analysis, DNA methylation analysis, protein analysis, protein analysis for about 200 proteins, and microRNA.
It sounds comprehensive. It sounds potentially confusing as well though.
Very confusing, well at times very confusing. We generated this data now probably a year ago and in some ways we’re only beginning to get the comprehensive picture as we compare data across the different platform types and try and synthesise them together into one picture.
But from the smile on your face I can tell you’ve got things to say, though, that could be beneficial. So perhaps we can start with the ER positive HER2 negative, what have you found so far there?
For me one of the surprising and fascinating features there was that the list of recurrently mutated genes was quite long and extensive for ER positive HER2 negatives, contrasted with HER2 positives or triple negatives where the list was very short. Three or four genes might be recurrently mutated in those groups whereas it’s 20-30 genes in the ER positives.
What difference might that make to therapy?
Well the difference it will probably make is that many of the genes that are on the ER positive list are therapeutic targets and many of the genes that are on the short list, particularly for triple negatives, are not. The top gene in the triple negative list was p53.
Which is hard to target.
Which is hard, very hard, to target. But the top gene on the ER positive list was PI3 kinase and we have a whole session here at IMPAKT focussed on therapeutically targeting the PI3 kinase pathway; there are some promising results so we’re quite hopeful that PI3 kinase mutations may turn into a predictive biomarker.
Now I’d like to ask you about the PI3 kinase and then move on to the other fascinating factors. What’s in the pipeline for PI3 kinase, then, that doctors should take note of?
There are many drugs from multiple pharmaceutical companies being developed that target PI3 kinase. PI3 kinase is actually a family of genes, there are three or four of them, and so some of these drugs hit all of them, some of them are more selective for one than the other. I think there were some promising results presented yesterday on what are called the alpha specific inhibitors and it is the alpha subunit that is the one and the only one that’s recurrently mutated in breast cancers. Then other PI3 kinase inhibitors hit PI3 kinase and a related kinase mTOR and so we have a lot to work out yet. But it does make some sense that it seems like our best success in cancer therapy has been for targeted agents that target the mutant kinase. So in this case the alpha specific drugs that target PI3 kinase alpha, I think, sound the most logical and probably the most promising at this time.
A lot of other factors, though, things like DNA methylation and mutations actually affecting gene expression subtype, as it’s called, and then something called copy number as well. Can you fill me in on what’s happening in these subcategories of factors?
From the DNA methylation that was done on the TCGA we took the genome-wide methylation patterns and looked for subtypes defined by methylation and then compared that to some of the subtypes that we had from, say, the gene expression platform. What we could see there was two very interesting findings: number one, the gene expression subtype that we call basal-like, which is the majority of triple negative breast cancers, have a un-methylated or a hypo-methylated phenotype and a subset of the luminal B gene expression defined subtype are hyper-methylated. So now within, for example, luminal B breast cancers, which tend to be the poor outcome ER positive breast cancers, some of them have a diverse mutation repertoire and are not so methylated and then another subset have a broad scale, genome-wide methylation. Interestingly enough, those seem to have less mutations than the ER positives that aren’t methylated suggesting somehow the methylation, the global methylation, is phenocopying what is occurring in other tumours by a direct mutation mechanism.
What might this imply in clinical terms?
In clinical terms there are a number of emphases to try and reverse methylation with things that interfere with the DNA methylation or interfere with concomitant modification. So it might suggest this group of patients who have the hyper-methylated phenotype would be the candidates for that class of drugs. Again, those classes of drugs are typically still in clinical development and so we don’t know if, say, the methylation is a biomarker for those drugs. But certainly now we know we should at least be looking for that.
And in the molecular portraits of breast cancer that you’ve been painting here, you did name triple negative breast cancer as being a bit black, a bit grey, not very hopeful. Do you see any hope emerging?
Absolutely. So it was grey from the perspective of the diversity of frequently mutated genes where it was basically p53 and almost nothing else. However, you can contrast that with the DNA copy number landscape where the basal-like subtype, which is 75% of the triple negative breast cancers, have a very large number of chromosomal changes, specific DNA losses that only occur in this subtype, specific gains that only occur in that subtype and some of those may be therapeutically targetable. So that’s another avenue that is being explored as we speak.
Clearly The Cancer Genome Atlas is an absolute wealth of rich data that you’re just brimming with ideas at the moment, much of it still needs to be exploited. Could you make a few predictions about what could float to the top clinically and what you advise doctors to make of all of this rich data?
That’s the tough question, that’s the $64,000 question. I think the most obvious and promising is the PI3 kinase mutations because in ER positive HER2 negative breast cancer 40% of patients have these so they’re very frequent. There are many drugs in clinical development and so that one is particularly exciting. There are actually other mutations in the pathway, including p10 loss, AKT mutation, there’s a regulatory subunit of PI3 kinase that is sometimes mutated and so whether those are biomarkers of sensitivity or resistance to these same drugs is also something I discussed and could be of great interest. I didn’t discuss it today or yesterday, but the cyclin dependent, the CDK4/6 inhibitors, also we can see from The Cancer Genome Atlas data would suggest that there would be many ER positive HER2 negative patients who would benefit from these and I think the early clinical data on those is quite exciting as well. So I think the PI3 kinase/mTOR pathway and probably the CDK4/6 inhibitors that are very promising for ER positive HER2 negatives.
And what about those less easy to deal with, more difficult to treat cancers like triple negatives. You said things like copy number might be targetable.
What we could see from the triple negative copy number landscape was that there were a number of growth factor receptor tyrosine kinases that were amplified but not mutated. So that’s a little paradoxical because usually they’re both amplified and mutated but not in this case. So there still we need to test whether the amplification of it was in particular MET, EGFR and actually PI3 kinase itself was frequently amplified but not mutated in basal cancers. So are those biomarkers of sensitivity to drugs that target those receptor tyrosine kinases is something to figure out. There will also be some presentations today from a colleague of mine, Gary Johnson, on how triple negative breast cancers respond to a particular kinase inhibitor, in this case it was a MEK inhibitor. What we could see was that when you give the single kinase inhibitor the cell moves the signal around that and adapts to the presence of that inhibitor and it reprograms the cell in a particular way. So now if you know how it reprograms you can target the reprogramming simultaneous with the initial target and that looks to be more effective. So now we’re going to need to basically empirically determine how this reprograming occurs. Hopefully it is common across multiple individuals or common within a subtype, which I think it will be, and then we’ll have some rationally selected combinations of these targeted agents. I think that’s what will be the therapeutic advances for the triple negatives.
So as far as doctors are concerned, do you foresee that getting the genomic landscape is going to be a prerequisite of treatment in the future?
Yes. And we kind of get a little bit of it with ER, PR, HER2. Clearly those are biomarkers of drug sensitivity, they also tell us about the biology of those tumours and definitely the genomics, the gene expression, the DNA sequencing, the DNA copy number is going to make a more complete portrait and better individualise the therapy for those classes of patients who have similar tumours.
And how do you advise doctors to fit in that knowledge with the clinical pattern that they’re getting from a patient?
We’re still not there yet. Much of this sequencing is still in the research phase but it’s getting very close to the clinical implementation. We still have to demonstrate clinical utility for many of these, for example, PI3 kinase inhibitors today, none are approved for breast cancer use. It looks quite promising and so I’ll say people need to watch closely because in the next few years many of these could get approved and suddenly we’re going to go from three to ten drugs that oncologists will need to be aware of. So they should just begin to familiarise themselves with many of these pathways that are now whole sessions here and as these drugs come in they’ll know where to fit them in.
Well we’ll all be waiting for them. Charles, thank you very much indeed.
My pleasure. Thank you for having me.