Oestrogens and breast cancer

29 January 2019

Interview with

Andrew Holding, Cambridge Institute, Cancer Research UK

CHEMICALS-BURETTES.jpg

Oestrogen response in breast cancer cells is not cyclical, as previously thought.

Share

Arguably one of the most important discoveries so far in the field of breast cancer is the finding that many of these tumours are sensitive to oestrogens; indeed, the subsequent use of hormone therapy has had a remarkable impact on breast cancer outcomes. But when scientists first looked at how tumour cells respond to oestrogens there appeared to be a very specific and conserved pattern to the response, suggesting that it must be critical to the tumour cells. But, as Andrew Holding explains to Chris Smith, when he tried to reproduce the results in order to understand them at the level of the whole genome, he began to think maybe he should reevaluate his career aspirations as a scientist…

Andrew - So I was hired five years ago to look at how the oestrogen receptor drives breast cancer. So in 70 per cent of breast cancer cases, which is the most common form of cancer, the hormone oestrogen drives the growth and proliferation of tumours. We have this body of literature that said that when you actually supply the hormone to the tumour, the way the tumour turns sort of turns on in pulses, in 45 minute pulses, so these are a bit like daily rhythms - circadian rhythms - but happening on a 45 minute interval. So really kind of rapid response, and we thought this would be a really interesting thing to model and if we could try throwing lots of things at the system and see how it changed those timings, mybe this could have a really big impact on how we develop therapeutics for breast cancer patients.

Chris - Now when you say it comes in cycles, literally the hormone goes to the cell; goes inside the cell - because the way the steroid hormones work is that they physically engage with the genome and turn on and off different segments of the genome - and you're saying that the literature had counselled us that this happens with these funny pulses: it goes on for 45 minutes of activity, then it goes off again. That sounds a bit extraordinary! Why did people draw that conclusion?

Andrew - Yes. So when it first came out, this was extraordinary because, as you mentioned, these receptors - oestrogen receptors - they actually bind onto the genes they turn on. And what someone showed, was that one particular gene, that was studied before we had all these genome-wide technologies, was pulsing on and off; and this everyone was astounded by. And then there've been several papers since where people said "look, we see the same thing!" And this is a really tightly controlled mechanism, and it was sort of seen as a dogma of the field that this is how the signaling works.

Chris - And, clearly, if oestrogen plays such a big role in cancer, and it's got this very rigidly-controlled cyclical activity, that must be important; so, therefore, it might be an important therapeutic avenue - is that your thinking?

Andrew - Exactly. And when we came to approach this, genome-wide technologies had happend. So instead of looking at one gene, we thought we could look at all 20,000 places in the genome where the oestrogen receptor turns on genes.

Chris - So you are hired five years ago; you're going to try and unpick what's going on. How did you approach the problem. How were you trying to study it.

Andrew - So the first thing I did was try and approach it repeating those original results, so looking at that single gene using these techniques which shows whether the receptors engaged at that gene and activating it. And what I found was on some days or weeks I got something similar and other days I found it was very different. And I kind of assumed it was me and I took it very personally that I was failing to reproduce this work because I wasn't as good in the lab as I kind of hoped I was.

Chris - Oh, ouch! So in other words this whole grand scheme of unpicking this cyclical activity didn't look like it was going to pan out?

Andrew - No; it looked like we had a really big challenge on our hands. And originally, you don't go around assuming that everyone else is wrong; and the other challenge was the methods we used didn't have very good controls; and there were steps the process that you didn't know whether the engagement on the gene wasn't there because you did the experiment wrong, or because it wasn't there. So I had to start thinking about whether it's me getting it wrong or actually the oestrogen receptor isn't doing what we think it's doing.

Chris - So how did you build in the right sorts of controls that would enable you to to unpick this problem and work out whether there really was this on off cyclical activity, or whether - as you put it - the field had got it wrong?

Andrew - So the technique we developed we actually decided to look for something else in the genome that was engaged at positions on the DNA that we could use as a control that we thought would always be there. And the thing we decide to pick on was this protein called CTCF. CTCF is one of the most stable things you can find in the genome. So we then went and checked that this was true in breast cancer. So we had a really robust solid sort of marker saying had we done the experiment right.

Chris - And this means that you can ask, well if I go in and if that control marker is not there some of the time, with a similar sort of frequency that we keep seeing the oestrogen signal going on and off apparently, maybe what's actually happening is that our system doesn't work very well. And in fact people were being misled before?

Andrew - Yes. So the idea was that we'd have this sort of benchmark to know what's going on. The other thing we did was we decided right, we know that people are going to be slightly concerned that we've gone and used a whole new technique, so what we did was we decided previously people would use one - maybe two - replicates in their studies, only doing it once or twice before publishing. We went, well we have the resources here to try and do this six times at 10 different time points. So we're talking about 60 genome-wide experiments and these are big experiments. And the idea of that was we had also the reproducibility to show that whatever you got was consistent every time we did it.

Chris - And what is the bottom line; does this cycle in the way that the field had suggested, or does it not?

Andrew - So the answer seems to be that, if you put oestrogen as a hormone onto these breast cancer cells they turn on and stay on.

Chris - So this is completely different to what people had said. The whole kind of there is this interesting cycle going on, which might give us a new therapeutic avenue, that's just not true?

Andrew - No! it seems to be that, actually, the simple answer is the right one. The only sort of caveat we had to apply to that, when we looked at our control, that was really really nice and stable; when we looked to the oestrogen receptor binding after 10 minutes, so we had a zero - nothing happened; then, at 10 minutes, we found it was really noisy. The signal had quite a lot of variability in it; some replicates there would be a lot of binding and other replicates there wouldn't be very much. But it wasn't tightly controlled. And when you averaged that out across all the genes and across all the time points, you found it was pretty static.

Chris - So does this mean then that what people were seeing and interpreting as these cyclical changes was the noise in the system; it's just random and they were interpreting that as this is this this cyclical binding activity?

Andrew - So there are a thousand papers that cite the original paper, and I went through all 1000 of them, found the studies that did exactly the same conditions - which turned out to be only four - some of them have bar graphs where I read off with a ruler, because the original data - I mean these are 20 year old papers, people didn't put the raw data out then - or they had pictures of these representing this activity and I was measuring how dark the image was using image analysis software, and we reanalyzed this data and what we found was exactly that. And that was just the noise in the system: the variability in how it responds.

Comments

Add a comment