Triple negative breast cancer accounts for about 20% of breast cancer. It’s probably the most malignant and it affects younger women and is described as what it isn’t – that is it is not ER positive, it’s not PR positive and it’s not HER2 positive. But what we and others are now finding that within this subgroup is actually heterogeneity. What I’m going to describe is how that heterogeneity can be abstracted into very characteristic genomic configurations that has clinical meaning as well.
How would you go about this?
It began with an analysis of whole genome sequencing, by the way, not exome sequencing but whole genome. We were specifically interested in structural mutations in breast cancer. What we were struck by was a number of breast cancers had a unique configuration that was multiple single tandem duplications. These are segments of DNA that are duplicated in a head-to-tail fashion but only duplicated once. But that happens all throughout the genome – your genome, other cancer genomes. But one thing we found very unusual was that this occurred very frequently, in the hundreds, and they were spread out throughout the entire genome as opposed to concentrated in a hyper-mutable area. This was so unique because you see this and it becomes obvious that something is wrong. We decided to do it in a computational mathematical manner because otherwise it’s like asking is this picture like Matisse or like Rembrandt? Unless you can de-convolute it into a mathematical principle it’s hard to look at it across this complexity across all tumours.
So we made a very simple measure of the number of tandem duplications and how evenly distributed they are in a genome. The thing that surprised us is that it was not a Gaussian distribution, it was actually a subgroup of cancers that had a very high mutational rate based on the single tandem duplications across the entire genome. When we looked at it carefully, now we extended that study to about 2,700 tumours that had been sequenced, we found that they are uniquely clustered in certain tumour types. In breast cancer only triple negative breast cancer, not the other types, had this and triple negative breast cancer was 50% had this kind of many tandem duplications. 50% of uterine, 50% of ovarians had this. Equally important was that prostate, colon, leukaemias, thyroid cancers and brain tumours were completely devoid of this tandem duplicator phenotype. Let me make it clear – tandem duplications occur but in the frequency and the distribution that it has it’s a very unique subset.
Then we started to dig down deep in terms of the characteristics and found uniquely that these tandem duplications can be further segregated into the size of the tandem duplicon, what we call span size. Instead of saying it could be one, ten, fifteen, twenty, a hundred, a thousand kilobases we found that they were clustered in individual tumours. So some tumours that are tandem duplicator phenotype tumours, TDP, that had many of these clustered around span sizes of ten kilobases. The hundreds that they have are all around ten kilobases plus or minus five and ten. There was another class that clustered around 230 kilobases. Then the third class, which is rare, that clustered around 1.7 megabases. That obviously told us there is unique biology in terms of this restriction in the tumour types and unique biology in terms of the clustering of the tandem duplicons.
So we began to look at it carefully and found that, of the TDP tumours coming out, 15% of all tumours that we looked at, of the 2,700 tumours, had a TDP which was enriched in triple negative breast cancers, ovarian and uterine. When we looked at those we found that individual tumours can have either this 10kb variety, the 230kb, the 1.7Mb but they also frequently mix but they mix only in binary fashion. So the 10kb and the 230kb occurred within a single tumour; the 230kb with the 1.7Mb occurred in one tumour but these combinations, either the unique ones or combinations accounted for almost 100% of all the tandem duplicator phenotype tumours.
Now, it was very important for us to find out by using this, this approach separating out the different TDP types, that we were able to then look at the originating mutations that generate these tandem duplicator phenotypes. So we found that the short tandem duplicator phenotype, which we call group 1, were exclusively BRCA1 mutant and p53 mutant. The important part of this is that it is not just a germline carrier, in fact, over 50% of the type 1 TDP, the one with the 10kb span, were somatic mutations or promotor methylation of the BRCA1. For the 230kb variety, that had predominantly cyclin E aberrations inducing augmentation of cyclin E expression but always with a p53 mutation as well. Then in ovarians CDK12 mutation was also a cause of the large span DNA. So we actually generated the hypothesis that specific mutations generated these chromosomal instability configurations.
We confirmed this by looking at mouse models. Mouse models that have knockout of p53 get breast cancers; knockout p53 and BRCA1 get breast cancers; BRCA2 and p53 get breast cancers. What we found was of the mammary tumours of these mice only the ones with p53 and BRCA1, not p53 alone, not p53 and BRCA2, generate this tandem duplicator phenotype of exactly the same configuration that we saw in the human. In a Nature publication that just came out we worked with our colleagues at Beth Israel Deaconess Medical Center where they used an in vitro indicator system that they can score tandem duplication formation and found that in the absence of BRCA1 and not BRCA2 in conjunction with replication fork stalling and not double strand breaks you induce tandem duplications of the same size.
So we, in a very short order, identified the mechanism for this, surrounded by BRCA1 deficiency and p53 deficiency, causing a clinical syndrome; it’s a carcinogenic process but causing a clinical syndrome that’s relevant in the human. A very important part of this is that we found using patient derived xenografts and breast cancer cell lines, triple negative breast cancer cell lines, that the TDP score is correlated with response to cisplatinum chemotherapy. This is actually very important because triple negative breast cancer, unlike all other breast cancer types, is responsive to platinum. The induction of pathological complete remission is the best indicator for long-term survival. The big problem is how do you know? Pathological complete remissions only occur in between 30% and 50% of patients and the 50% is when they use platinum and taxane up front. Now we think we have a tool to be able to segregate out those who will respond to this therapy versus those who won’t. This allows us in the first instance to personalise therapy within the triple negative subset. We’re in the process of moving this into the clinic to get that validated. We’re very excited about those possibilities.
What are the consequences of these repeats?
From this it’s actually very important – what are the downstream consequences of these hundreds of tandem duplications? We studied that; fascinating is that that probably is the systems induction of a cancer. No single TDP tumour looks alike because these tandem duplicons aren’t in exactly the same places. But what we found is that there is a tendency for these tandem duplications to augment the expression of oncogenes and kill the expression of tumour suppressors but not one, multiple ones.
The challenge with that is that some of the oncogenes and tumour suppressors that they augment and otherwise would not be detected by classical diagnostic approaches. Let me tell you why. Normally you look for HER2 amplification. The tools we use actually screen for amplifications at the five- and six-fold, that’s the threshold. We’re talking about a single tandem duplication. There’s increasing data to suggest that in the absence of amplification that there are HER2-like triple negative breast cancers that may be responsive to lapatinib. Now we need to test this but this could be exactly the reason why some individuals respond in the triple negatives.
The other aspect of it is that, for example, p10 is mutated and Rb is mutated but these mutations are not scored by exome sequencing because they are not point mutations, they are actually tandem duplications that interrupt the appropriate transcripts. So what this tells us is that if we’re going to look at triple negative breast cancers we’re probably going to need to have a different set of tools to address the mutational profile because they are not point mutations, they are actually structural mutations which are more difficult to discern using the standard approaches that we have at this time.
Any other key points to mention?
One of the key issues here is an issue of how you look at the data. The unique aspect of this is that we’re not looking at individual genes that are mutated, we’re actually looking at a chromosomal configuration. That just is an indication of the complexities that we will have to grapple with when we analyse cancer genomes.