SLIDES & TRANSCRIPTS
Monday, May 12, 2003

Pharmacogenomics

Mary Relling, Pharm.D.

Of course, when we are talking about ALL or any cancer, we really have theoretically two genomes, the host genome that will affect the risk of drug availability to the tumor through pharmacokinetic variability, tumor sensitivity, as we have just been hearing about through intrinsic mechanisms, and host toxicity.

The acquired genetic changes of the ALL-blast or the tumor genome also have some influences on those three phenotypes, with obviously drug sensitivity of the tumor being the largest impact of the host genome.

However, the impact of germ line genetic variability does affect the tumor susceptibility to drug. We feel that this impact has been somewhat overlooked and may, in fact, even affect the acquired genetic changes themselves in the ALL blasts.

TOP

In ALL, we have the situation where we are using multiple drugs, and we have got a phenotype -- for example, a toxicity of neutropenia -- that may be influenced by many drugs, and it may also be influenced by non-drug therapies. It will also be influenced by non-chemotherapy supportive care measures.

So, it is a complicated situation in which to evaluate how germ line polymorphisms affecting these multiple drugs and non-drug influences may impact on a single phenotype.

TOP

Of course, every drug interacts with multiple gene products, each of which is subject to polymorphism. For us, the most interesting lesson of the human genome project has been that essentially every single gene in the human population is polymorphic at reasonable frequencies.

This is a simplified diagram just showing a few of the targets, italicized for methotrexate, one of the backbones for ALL therapy. All of those individual genes targeting for products with which the drugs interact are subject to polymorphisms.

TOP

Nevertheless, despite this complexity, we feel that it is worthwhile to investigate the contribution of germ line genetic polymorphisms to clinical phenotypes in childhood ALL.

TOP

There are sort of two ways of thinking about the approaches. One is to take a target gene approach and look at known functional polymorphisms in genes whose products are likely to interact with the anti-cancer drugs that we use to treat childhood ALL, and may impact on the important phenotypes, those phenotypes being event-free survival, for example, second cancers, and toxicities that we would like to avoid, such as avascular necrosis, or long-term neurotoxicity.

Of course, that is looking with the lights turned on and where we are smart enough to think about looking, but the other part of that is to identify additional genetic targets that may be polymorphic and may influence the phenotypes of interest in childhood ALL, and I will address that a little bit more this afternoon in one of the breakout groups.

TOP

So, a couple of examples of looking at how germ line genetic polymorphisms affect ALL outcomes, kind of taking a one gene at a time approach.

Obviously, those genotype phenotype associations have to be highly penetrant associations in order to be observable in the background noise of treating ALL.

TOP

You are all familiar with thiopurine methyl transferase. It is subject to a common genetic polymorphism and low TPMT increases more captive purine availability for activation down the thioguanine nucleotide pathway and, therefore, incorporation into DNA.

I am presenting this mostly as an example.

TOP

In the population, if we look at the frequency distribution of the activity of this single gene product, TPMT, it is trimodal with about 10 percent of the population being heterozygous for TPMT activity, and there is an inverse relationship between TPMT activity and accumulation of thioguanine nucleotides, with the rare homozygous mutant patients having extremely high thioguanine nucleotide levels, and the wild type patients, who make up the majority of the population -- and that is the population that has really contributed to the average dose of 6MP that we use to treat childhood ALL, actually have very low concentrations of these thioguanine nucleotides.

TOP

That is translated into, because it is a monogenic penetrant effect, clinically significant phenotypes. For example, this is looking at the cumulative incidence of the need for dosage reduction versus time in the patients who are homozygous mutant.

They absolutely can't tolerate normal population doses of 6-mercaptopurine, and require severe dose reductions over 10-fold from that tolerated by the population.

Heterozygotes, approximately one third require a dosage decrease, whereas the vast majority of wild type patients don't require a dosage increase, using the somewhat higher population dose of 75 milligrams per meter squared that has been used at St. Jude for many years.

TOP

That is translated into a difference in overall leukemia free survival with those individuals having at least one defective TPMT allele and, therefore, higher thioguanine nucleotide concentrations tending to have better event-free survival than those individuals who are wild type for TPMT.

In fact, what we observed is that the failures that we observed in this study where a relatively high proportion of patients receive prophylactic cranial irradiation after intensive thiopurine therapy, were not ALL relapses but, in fact, secondary brain tumors.

TOP

We looked very hard to see host factors that predisposed to the development of brain tumors in individuals who had at least one defective allele for TPMT, had almost a 50 percent cumulative incidence of secondary brain tumors after receiving irradiation.

TOP

We have been sensitized to look because TPMT was also associated with a higher risk of topoisomerase II inhibitor induced therapy-related AML, and this was subsequently confirmed by the NOPHO group the next year.

TOP

In fact, high TGNs had been associated many years ago by Lynn Leonard in renal transplant patients receiving the thiopurine azathioprine, with the risk of secondary skin tumors as well.

TOP

So, a couple of lessons from the TPMT germ line polymorphism. The interaction between the target drug and the polymorphism -- in this case, 6 mercaptopurine and TPMT, that there may be an interaction between the thiopurine and genetic polymorphism and other therapy, for example, topo II inhibitors or irradiation.

There can be short term and long-term consequences of the polymorphism on acute toxicity, overall leukemic-free survival and secondary cancers.

Of course, it illustrates a couple of ways in which we are trying to address type I error. We are looking at lots of genes with relatively small numbers of patients.

The fact that the findings can be replicated in multiple studies may be important in decreasing the risk of type I error, and also the follow up with laboratory models which our laboratory is pursuing to get at the molecular basis of mechanism for how low thiopurine methyl transferase activity can contribute to the risk of secondary tumors, for example.

TOP

A couple of other examples, this sort of one gene at a time approach. Thymidylate synthase is the target for antifolate therapy. It is subject to a very common polymorphism of two versus three 28-base pair repeats in the enhancer with a very high allele frequency.

Those individuals with two tandem repeats have lower expression of TS, and their tumors have had better anti-tumor response; with 3-10 tandem repeats, worse anti-tumor response.

This makes sense. If there is more of the target that needs to be inhibited, then the patients may have a worse prognosis, given the same drug therapy.

TOP

So, the TS high activity allele with three repeats has been associated with worse outcome, event-free survival. In patients treated with both high dose and low dose methotrexate by one group, published a couple of years ago.

TOP

However, this has not been replicated in all studies, and just a month or so ago the BFM looked at a case control design and couldn't see any impact of this TS polymorphism in their population.

TOP

Likewise, another folate related polymorphism, a common SNP in MTHFR that results in low activity associated with the T allele, that is about 10 percent of the populations.

Those patients who are homozygous for T have a higher mucositis index when treated with methotrexate than the patients who are wild type,

TOP

but again, not confirmed in all studies.

In our study using high dose methotrexate in several hundred patients receiving this therapy, there was no difference in acute toxicity among patients who were wild type at the MTHFR 677, heterozygote versus mutant.

TOP

This brings up sort of another principle, that the differences in prognostic impact of these germ line polymorphisms may relate to differences in therapy, just like any prognostic factor does.

So, it is possible that, by using high doses of methotrexate one may ameliorate or abrogate the prognostic significance of polymorphisms in folate-related genes.

Of course, the opposite is also true. If the therapy is very aggressive with other non-folate-based therapy, then those folate-related polymorphisms may not shine through as prognostically significant.

TOP

Of course, pharmacogenetic traits are likely to be multigenic. So, when we are looking at multiple polymorphisms versus phenotypes, there are several possibilities.

One is to combine genotypes within an individual based on pathways, to make patients sort of high folate responsive versus low folate responsive germ line status, to combine genotypes based on a mathematical approach using decision trees, to estimate haplotypes where applicable.

Obviously, most of these have more than one polymorphism. So, there might be an important contribution of haplotype in some of these gene association studies.

TOP

Just to give you a flavor for what we are doing at St. Jude, we have genotyped about 250 patients for over 16 common polymorphic loci, all in gene targets that are putatively related to the efficacy or toxicity of the common anti-cancer agents that we use in childhood ALL.

We are taking various approaches, trying not to take a one gene at a time approach to combining these polymorphisms to see what are the important pharmacogenetic determinants of phenotypes in ALL.

TOP

Using this multivariate approach, I was allowing event free survival to be estimated in a time dependent fashion.

We do, in fact, find that combining along the glutathione transferase pathways, that individuals who are defective in the three common GST (glutathione S-transferase) polymorphisms, tend to have better event free survival than those individuals who are wild type at all three of those GST loci.

TOP

Looking at a decision tree type analysis where, again, the gene polymorphisms are taken into account one at a time, in a Cox proportional hazards model type approach, again, in these about 250 patients, the most prognostically important polymorphism was the GSTM1, those individuals being null, having a better prognosis.

Then, when we look within those who are non-null or wild type for GSTM1, we can see which polymorphisms then tend to be important.

In fact, it turns out that the GSTpi1 is prognostically significant in the direction we anticipated, and also that the TS polymorphism I described earlier is prognostically significant, again, in the direction that we anticipated.

So, these sorts of multivariate approaches may be helpful in determining which polymorphisms are the most important in patients with leukemia.

TOP

I will just close by saying, the disadvantage, of course, one of the disadvantages of this target gene approach is, we certainly may be looking at the wrong genes.

So, this afternoon we will talk a little bit more about how we are using expression array and other tools to discover new genetic targets whose polymorphisms may be important for outcome in childhood ALL. Thank you. I will stop there.

TOP