DR. CARROLL: This is open for questions, and please identify yourself when you ask a question. It helps the people recording the session.

PARTICIPANT: I am curious about the TPMT gene work that you showed and the various other malignancies that those patients get. Is that a phenotype, do you think, for a cancer risk in and of itself in a mutant?

DR. RELLING: We don't think so. It has been phenotyped in quite a few large patient populations, and there is no significant difference in the frequency of the TPMT defect in patients with de novo cancer. We don't think that it is related to overall cancer risk.

We have recently made a mouse model of the human polymorphism by knocking out the TPMT gene, and these mice are perfectly fine. They have no phenotype in the absence of drug stressor.

So, we think it is an interaction between the very high concentrations of thioguanine nucleotides achieved in these patients treated with thiopurines, and the other cancer stressors that are, of course, much more evident in treated cancer patients than they are in the general population. The general population isn't exposed to high doses of drugs and irradiation.

DR. LOOK: So, it sounds like what you are really getting toward is sort of an analysis of all single nucleotide polymorphisms that affect amino acids in expressed regions of every gene, and then do a sort of unbiased look at impact on phenotype.

DR. RELLING: That would be lots of polymorphisms, though, wouldn't it. There are 3.2 million SNPS, and a third of them are likely to be functionally important.

We have struggled with, is it better to take a totally unbiased approach and do some kind of genome-wide scan, or is it better to look at these genes that have bubbled up through traditional biochemistry and clinical epidemiologic studies and say, for these things that we know have functional consequence and have pretty penetrant phenotypes, what is going on in children with ALL.

You know, just looking at the expression array, those genes that differentiate phenotypes of interest, we are finding fairly frequently that there are common polymorphisms that are determining their level of expression, even in the tumor cells.

So, there are going to be way too many polymorphisms for the number of patients with acute lymphoblastic leukemia in the world. There is no way we will ever have enough patients to statistically justify looking at all these functionally important genetic variants.

I don't really see any alternative other than replication by independent studies and follow up laboratory models to get at the possible mechanisms of functional consequences.

Clinically, if we are going to decide up front to genotype and then modify therapy based on genotype, we have got to know which ones are important.

TOP