ASCO 2013

Effects of The Cancer Genome Atlas

Dr Matthew Ellis - Washington University in St Louis, USA

This is an interesting story and illustrates the power of the TCGA data. As we were looking through our own sequencing data we noticed one or two mutations in HER2 that were occurring in cases without HER2 gene amplification. We began to study these and found that these mutations were activating, in other words they switched on the HER2 kinase.

What’s the significance of seeing them without amplification?

You’d miss them by the normal testing because there isn’t increased signal with immunohistochemistry and there’s the FSH test will be negative. But this is a cryptic activation of HER2, you have to do sequencing to find these things and interestingly working closely with the TCGA, I was sitting in a meeting last March and the analyst announced they’d found another seven HER2 mutations. So then we studied the data coming out of other people’s sequencing efforts and ended up with about 25 which we reported in Cancer Discovery the first part of this year. Now we have a clinical trial, interestingly not with the drugs you’d expect. You’d expect that we’d be studying trastuzumab or you’d expect we’d be studying lapatinib but in fact the careful molecular pharmacology indicated the best drug would be a drug called neratinib, which is a suicide inhibitor of HER2. So instead of binding HER2 and then the drug having a reversible binding, which is how lapatinib works, these drugs bind to HER2 in a covalent way and inactivate the enzyme. And, for whatever reason, in comparative studies neratinib looked to be the best drug broadly against all the mutants we were studying.

Will effects be seen in the clinic?

Working with the pharmaceutical company involved we’re now opening sites throughout the United States. Interestingly, and this is a little nice trick with next gen sequencing, we started off by doing screening with Sanger sequencing but it’s insensitive and laborious and I said to my next gen sequencing people, look, I don’t want 150 genes on one patient, I need 150 patients analysed for one gene. So now we use the next gen sequencing in a multiplex fashion so that when you run the sequencing reaction you can run forty or fifty patients in parallel and screen lots of patients for this mutation that is occurring at about a 2%, 3%, 4%, prevalence. So it’s a rare mutation but the point about breast cancer is it’s so common that just a few per cent can translate into thousands of patients. The incidence of chronic myeloid leukaemia is about 4,000 patients a year and, I guess, HER2 mutant breast cancer could be in that ball-park. So from that perspective it’s not that rare in terms of the diseases that we regularly treat with targeted tyrosine kinase inhibitors.

Is an inhibitor possible for all of these mutations?

It looks like most of them would be responsive to neratinib so that’s the nice thing about it. Only some of them look responsive to lapatinib. So what this illustrates is that when you have to functionalise the TCGA data you have to do careful molecular pharmacology and that increases your chance that you’ll end up with a positive clinical trial.

This project has also turned up information on ER?

Interestingly, it didn’t. Let me explain to you about this and we did some reporting of this at AACR. We’ve been looking at patients with advanced ER-positive breast cancer that had developed hormone refractory characteristics; in other words, they’d been treated with various endocrine drugs that had stopped working. Our investigation style was to grab these tumours and make them grow in mice so that we could study them and we found in one of these there was actually an oestrogen receptor translocation where the ER was spliced into another gene from a different chromosome. This construct would be resistant to all endocrine therapies. So we didn’t see that kind of thing in the TCGA data so what I’m beginning to think is that in metastatic breast cancer ER is becoming mutant. Interestingly there were also some recent reports from a clinical trial called BOLERO2 where they did some sequencing of metastatic breast cancer samples and they found additional point mutations in ER around the ligand binding domain that allow the oestrogen receptor to function in a hormone independent manner. Again, those mutations are not seen in the primary breast cancers that were sequenced by TCGA. So the point I want to make is we need another cancer genome atlas, not of primary breast cancers but of resistant metastatic breast cancers so we can understand what the genomic changes are that generate resistance. So in the case of ER the tumours evolved in a bath of oestrogen so they didn’t bother to mutate their oestrogen receptor. But then we took those patients and took their oestrogen away and what we’re beginning to suspect is the ER becomes mutant as the tumour’s ability to cope with the low oestrogen levels drives the accumulation of mutations in the oestrogen receptor itself.

How far away from the clinic are these ideas?

I think what you’ll see in the next year, two years, is a lot of sequencing of metastatic breast cancer and from that some new thinking about resistance. Once you’ve understood resistance we’ll be able to develop the appropriate pharmacology to prevent it. Quite a radical thought really. But what might we do in the clinic if this oestrogen receptor mutation story is real? Well interestingly it might tell you what to do with the patient. So, for example, if this translocation mechanism is something that is clearly prevalent and worthwhile looking for, well maybe you would look for it because no endocrine therapy would work in the presence of one of these translocated oestrogen receptors. You’d choose some chemotherapy or some other type of drug because you don’t want the patient to be on an ineffective treatment. So that’s the power of prediction, in this case predicting resistance so you can give the patient something else and not bother about second or third line endocrine therapy if they’ve already developed this kind of resistance mechanism.

Is this type of sequencing a difficult thing to do?

The difficult thing to do is to get clinicians to biopsy metastatic disease and make sure you’ve got a genotype of the metastatic tumour at the time you want to make a treatment decision. We think we can go back to the primary and sequence that and that will tell us what to do with the patient but it won’t tell us about the acquired resistance mutations. This is a story that we see in CML all the time, you treat with imatinib, they develop resistance and you can treat with a second drug that targets the resistant mutation.

What do doctors need to do in the clinic?

Lots of co-operation; the willingness to biopsy metastatic breast cancer, get those samples into discovery pipelines and then we can translate that into useful tools for the practising clinician.

What is your take home message?

Well, we’re in this fantastic new world. For a while it’s going to be chaotic because we don’t know what all the information means but I think what we’ll be finding will be there will be very nice new tools to personalise therapy, maybe just starting with our two favourite genes, ER and HER2, with additional tools related to the somatic genetics of those two genes and then moving on into many other of the very interesting targets that we’ve fingered in these large genome atlas projects.