Historically by conventional testing one uses a large group of animals to be controls, no treatment, versus those that are treated to see if a drug has a statistically significant effect against tumour growth. In the case of paediatric cancers the reason we do this is because there are relatively few clinical trials that can be conducted on kids to test new drugs, in part because paediatric cancers are rare. Secondly, the five year overall survival, event free survival, is now approaching, or exceeding, 80%, certainly in the United States. So there are very few patients that are eligible for experimental drugs. So using preclinical models gives us the option of looking at a large number of drugs and to prioritise what goes further into the clinic.
The reason we’ve taken the approach of single mouse is that one of the failures of conventional testing has been that because we’re using the resources, using a large number of animals, we could only look at a few tumour lines. So if, for example, one took Wilms’ tumour, a tumour of the kidney, we could only look at, say, three different Wilms’ tumours in mice. That certainly doesn’t adequately represent the genetic diversity of those tumours.
So what we did was to analyse about ten years of data where we had done conventional testing. This is 83 xenograft models, that is human tumours grown in immune deficient mice, and 67 drugs and asked the really simple question if we could just use one mouse instead of ten in the treatment group would we have come up with the same answer? The result was quite astounding – 80% of the time it would have been absolutely correct. If you allowed for a slight deviation of plus or minus one response category, which didn’t make any difference to the outcome, the results were 95% concordant.
This analysis allowed us to say, ‘Let’s just use one mouse per tumour line and instead of looking at five tumour lines let’s look at 50 tumour lines of this particular cancer type.’ Now we can incorporate the diversity, the genetic and epigenetic diversity, that we know exists within these paediatric cancers.
What specific methods were used to test this?
As I said, we did a retrospective analysis of over 2,100 studies. That gave us the confidence that these models were quite reproducible and that one mouse would work. So what we presented at the EORTC symposium, we did five prospective tests, three were done in leukaemia models by Richard Knox’s group in Australia, and two were done in solid tumours.
In the case of the leukaemias we used between 70 and 90 different leukaemias in mice. So each leukaemia was from a different patient. In the case of the solid tumours we used 34 and 50 different models from different patients to represent their diseases. The concordance is about 95% again.
So in a prospective test it shows that this single mouse study identifies tumours that are very sensitive to these drugs and those that are essentially resistant to these drugs. That allows you now to probe the molecular characteristics of these tumours because all of these tumours are linked to RNA-Seq and whole exome sequencing so we can look for certain mutations that confer sensitivity or resistance or certain expression profiles that again relate to drug sensitivity of a tumour model.
What key results can you highlight?
The results that I mentioned are really the key highlights in the sense that as we go forward, and we ran a very large programme in the United States to identify drugs that are really very active and should be prioritised to move into paediatric clinical trials, that we can now do this on a much larger scale and, perhaps over the next five years, not only identify drugs that are really very active but they are very active in specific subsets of cancers and be able to identify the characteristics of those cancers so that we can stratify patients that will ultimately benefit from these new drugs.
How might this affect testing in the future?
In the United States we just passed the so-called RACE Act which is an act by which the FDA is asked to prepare a list of drugs in development and their targets that they interact with. If that target is relevant to the growth of paediatric cancers the FDA can mandate that that drug is moved forward into testing in paediatric cancers of the appropriate molecular type.
Where I think the single mouse study is going to be advantageous is that we can select models that are driven by specific signalling pathways or by specific cancer genes where these drugs are reportedly going to interact. We can take those models and test the drugs to see whether those drugs are really effective in the molecular subtype that they are being developed to test. The reason for this is that we know that certain mutations confer sensitivity to, say, signal inhibitors such as BRAF inhibitors which are very active in melanoma. But we know that the same mutation in colon cancer does not confer sensitivity to these drugs. So even though you have a common genetic defect or a mutation that activates an oncogene, still the response of the tumour is very context specific. So just because it works in an adult cancer does not mean to say that the same mutation in a paediatric cancer will confer the same level of sensitivity.
So these are the sorts of questions we can ask and, if the drugs are active, to move them quickly into the appropriate molecular subtype of paediatric cancer. This idea that we categorise tumours by their histology, by their site and such like, is still valuable but as we go forward we’re going to be subtyping tumours at a molecular level and hopefully our drugs will be specific for that molecular subtype and have very marked anti-cancer activity without the toxicity that we have from current cytotoxic therapies.