IMPAKT Breast Cancer Conference 2013
Single cell sequencing techniques in early and late breast cancer
Dr Nicholas Navin - MD Anderson Cancer Center, Houston, TX, USA
Nicholas, your lab at the MD Anderson is actually doing something a bit new because you’re looking at sequencing in a single cell. Tell me, would you please, what you’ve achieved so far.
Sure. We’ve developed a new technology, it’s single cell sequencing. In the past, or up to now, most of these genome sequencing tools require millions of cells and so it’s a hunk of tissue and there’s an admixture of different stromal cells and immunocytes and tumour cells and it’s a big mixture of different tumour cells. So you’re looking at an average signal from those cells and here you can look at individual cells. One of the big advantages is that you can look at very rare populations of tumour cells such as cancer stem cells or circulating tumour cells.
Why has this been difficult in the past?
Because of the amount of material you need going into the sequencing reactions. It requires micrograms of DNA; a single cell only has 6 picograms so it’s a very small amount. So these technologies allow you to amplify the DNA and then prepare libraries and then do the sequencing and you get genomic information on individual cells.
It sounds clever stuff, what do you actually have to do to make this work then?
The procedure is… there are many different ways to isolate the cells so we use flow sorting, it’s been around for a long time and it’s very accurate. We actually isolate nuclei instead of individual cells and that’s because they’re less sticky and they tend to sort better.
They’ve got the DNA, most of it?
They have the DNA, they have all the DNA. And then we go into the amplification reaction which we limit to reduce technical error and we use a transposase system to make libraries.
You amplify by what means?
It’s a polymerase from a bacteriophage, it’s called phi29 and it does something called multiple displacement amplification where it creates very large strands and they network and polymerise off previous strands and so it amplifies the DNA. We do that about a thousand-fold and then we go into the sequencing reaction.
So then is it any easier, is it just standard sequencing from then on?
From then on we use a transposase system which can operate on very small amounts of DNA. We go from picograms to nanograms and then we make libraries. So it’s not quite as straightforward as doing a normal sequencing reaction where you use a microarray of lots of DNA micrograms. So we still work with small amounts of libraries and going into the sequencing reactions we also start with much lower material. But after that we generate data and then we have algorithms that are a bit different to detect variants and to eliminate technical error in our datasets.
Now I believe that with this technology you can look at things like circulating tumour cells and also putative stem cells as well. What have you been able actually to do so far?
We’re starting to apply these tools to circulating tumour cells and to cancer stem cells but in the past so far we’ve worked mostly on triple negative breast cancers which have a lot of genetic heterogeneity. We don’t sequence just one single cell but we’ll try to sequence as many single cells as possible. Then we compare the mutations in those cells and we reconstruct phylogenetic trees of how those cells evolved and we can identify whether early mutations or the late mutations that occurred during tumour progression. Interestingly, what we’ve found so far is that certain cancer driver mutations tend to be present in every single cell but in addition to that there are lots of new mutations that are unique to individual cells and there are substructure and subpopulations within the tumour.
Now, at this conference here in Brussels we’ve been hearing about taking more than one biopsy, more than one sample, but you’re talking about taking many, many samples, perhaps from the same tumour?
In a way it’s similar, instead of taking samples, though, you’d be taking single cells. It’s a sampling problem so you want to sample as many cells as possible to get a good picture of the diversity within a tumour.
What advantage could this be giving you clinically?
The things we’re most excited about, number one is being able to non-invasively monitor the tumour. So you can take a blood sample, you can get the CTCs and if the CTCs accurately represent the genomes of the primary and the metastatic tumours then you can just take blood samples and sequence that DNA and you’ll be able to track the tumour and also how it changes in response to certain therapies.
So you can expect to have CTCs in the blood at any time and you can monitor what’s happening. Is that on a rapid timescale?
We would like it to be. Currently we’re limited a little bit by the sequencing time itself which takes quite a bit of time. Data processing also takes a day or two so it’s not quite ready for the clinic but it’s more useful in a research setting. We’re starting to look at what type of clinical applications there would be. The other thing we’re really excited about is early detection and scarce clinical samples such as fine needle aspirates where you only have maybe 10,000 cells available, or so, for analysis. Early detection, we hope that one day everyone will just go to their general physician once a year or so, get a blood sample drawn, they can see if there are any tumour cells in the blood and then they can sequence those cells and see if they have any driver mutations. If they do that might indicate that somewhere in that patient there’s a tumour and then they can follow it up with imaging or other analysis. But that’s way in the future.
Potentially could you look for free DNA in the bloodstream too?
That’s the other. So there’s a lot of excitement now also about circulating free DNA and these are fragments of DNA that are extra-cellular and they’re usually from one base to a thousand bases. It’s still being investigated how similar or how accurately those represent the primary tumour versus the circulating tumour cells. So we’ll have to see, as it pans out, which are better markers or better tools for us to monitor the primary tumours.
You’ve been talking here in Brussels at the Breast Cancer Conference about single cell sequencing in early and late breast cancer. What are the stages of the disease that this is best suited to and how are you best going to use it? This is for monitoring recurrence or for looking at the progress of metastatic disease?
Those are both things we’re interested in. We’ve been looking mostly at late stage breast cancer so far, triple negative, usually grade 3 invasive carcinomas. We’re trying to understand just how much diversity is there within the tumour. What we’re seeing is that the level of DNA copy number or chromosome counts, the tumour is actually very stable and we think that early on in tumour evolution you have lots of chromosomal rearrangements but then the genome restabilises and those cells are very stable as they expand to form the tumour mass. But on top of that we see lots of new point mutations and indels that arise in individual cells and that seems to be a more gradual or continuous evolution. So copy number evolution seems to be punctuated but point mutations and indels evolve gradually over time. Occasionally there’s a clone that has a strong selective advantage and there’s a sweep across the population but those mutations continue to evolve gradually.
So where is this giving you the best yield clinically, potentially?
We would like to have a diversity index for every tumour and then we can see a scale of zero to one how diverse, genomically diverse, is this tumour and we can correlate with clinical parameters like does that indicate poor survival in the patient? What’s the probability of metastasis? What about invasion? So we can try to correlate clinical parameters.
So early on in the disease?
We’re looking early too; this would be for more late stage patients but in theory it could be applied early on too. So with the early tumours it’s most advantageous for looking at things like ductal carcinoma in situ which are very small, confined lesions and they’re only a few hundred cells.
You could decide not to do anything.
You could take a look at them and see how diverse they are and try to… well, right now it’s just a basic research question is how related are those in situ populations to the invasive populations. That’s one thing that we’re looking at as well.
Could you potentially monitor, for instance, things like neoadjuvant therapy?
You could try to, for example, have a biopsy before neoadjuvant therapy, have a few thousand cells, monitor them and then after neoadjuvant therapy you could take a look at… usually there’s a lumpectomy that follows that, in breast cancer at least, and you could look at those cells and see how the populations have changed in response to the neoadjuvant therapy, whatever it might be.
So how do you advise doctors, then, to view this technique? I should ask you, first of all, has it been production lined yet? Is it easy to do?
No, it’s not, it’s not streamlined yet. It’s still really in the research phase. What we can look at, some important implications we have so far for therapy is that there are some mutations that are present in every single tumour cell and they seem to have a very strong selective advantage. Those are really mutations you want to go after with your therapies, targeted therapies. The other mutations that appear in subpopulations within the tumour, if you target those you’re not going to eradicate the whole tumour.
And how much better is single cell than regular sequencing in this? Is it just a little bit better or do you think it could make a whole sea-change in your approach.
It could certainly tell you a lot more about the sub-structure of the tumour and help you identify which subpopulations to target in the tumour. So when you sequence a bulk tumour you don’t get that information, that’s mainly because of cost. So if you could have a bulk tumour and you could sequence it down to a million-fold coverage, then in theory you would get the same information as you get from the single cells but in the single cells you also know what combinations of mutations are present within the cell, you don’t get that from bulk sequencing. But currently it’s too expensive to sequence a tumour to completion so single cells you can randomly sample a few cells and you get a spectrum of that diversity.
And a quick take home message from your work, what would that be?
I would say that with sequencing so far we’ve looked at the tip of the iceberg, we have identified some of the very common mutations that are frequent across cancers but as you sequence deeper and deeper in the tumour you’ll find a lot more rare mutations. That has very important implications, for example, for chemotherapy in suggesting that a lot of resistant mutations are probably going to be pre-existing within tumours instead of being spontaneously induced.
Thanks very much for coming in.
Sure, great. Thank you.