At A Glance
Name: Daniel Nebert
Title: Professor in the Department of Environmental Health, University of Cincinnati Medical Center — 1989-present; Professor in the Department of Pediatrics and Developmental Biology, Division of Human Genetics, Children’s Hospital Medical Center Professor — 1991-present; Professor in the Division of Human Genetics at the Department of Pediatrics and Molecular Developmental Biology, University of Cincinnati Medical Center — 1990-present
Background: Designated US Co-ordinator for the Medical Genetics Program of the United States-People’s Republic of China Cooperative Medical Health Protocol — 1982-1989; Director, Center for Environmental Genetics, University of Cincinnati Medical Center — 1992-1997
Education: BA in biology and chemistry from Wesleyan University, Connecticut — 1959; MS in biochemistry from University of Oregon Medical School — 1964; MD from University of Oregon Medical School — 1964; Pediatric internship and residency at the University of California at Los Angeles
In one of the opening lectures of the Third-Annual Meeting of the International Society of Pharmacogenomics in Santorini, Greece, in late September, Daniel Nebert urged a cautious and skeptical approach to pharmacogenomics.
He said drug responses are not thought of generally as monogenic traits, but as complicated problems requiring the use of more than genetic methods.
He is the author of two recent pharmacogenomics reviews, the first in the November 2003 issue of the American Journal of Pharmacogenomics, and the other in the October 2004 issue of the European Journal of Pharmacology.
Nebert is board qualified in both pediatrics and human genetics, and is the author or co-author of more than 520 peer-reviewed reports, invited reviews, and book chapters. His research interests include: drug metabolism and pharmacology; human genetics; environmental toxicology; evolutionary genomics; mouse genetics; and gene nomenclature.
He has been a member of the London-based Human Genome Organization-sponsored Gene Nomenclature Committee since 1999.
Can you discuss the high points of your talk in Santorini?
The reason for that talk and for the November ‘03 [American Journal of Pharmacogenomics review] and the October ‘04 European Journal of Pharmacology, was that we had invited Allen Roses [senior vice president of genetic research at GlaxoSmithKline] up to Cincinnati in April of 2002, and he got my attention. Immediately after that day, I started jotting down paragraphs of all the reasons why he couldn’t possibly be right, and it just started snowballing.
What were some of those reasons?
These reviews list roughly two dozen reasons why it’s almost impossible to get an unequivocal genotype, just from the number of enzymes that are used on any one drug. All the enzymes that contribute, all the genes that contribute to the metabolism — transport distribution channels, receptors, transcription factors, and everything. And all the ethnic differences in the world. Even the HapMap’s attempt to find tag SNPs that will define all haplotype blocks. They continue to spend money, but I think that they realize it’s pretty useless.
I was looking forward to that meeting because Allen Roses was supposed to speak just before me or just after me. He had a review in September, where he’s back pedaled just a little, saying that pharmacogenomic testing could cut down on the total number of people during Phase I clinical trials.
There’s just so much we don’t know about the genome, and every month or two, there comes another discovery that makes it even more complicated.
Pharmacogenomics has taken a few small steps into the clinic. Do you expect it to go much farther in the next 5 to 10 years?
With some things that are predominantly monogenic — it’s been realized in the last decade that there’s really no such thing as a complete monogenic trait, everything has modifier genes and contributions from things in the environment — but for something that is predominantly monogenic and overwhelming, like CYP2D6 or CYP2C19 polymorphism, that could play a major role.
If you did a test and found that a patient was an extensive metabolizer, you could give a prescribed dose and be right perhaps one out of three times, but not nine out of 10 or 20 out of 20.
So, as far as individualized drug therapy, it doesn’t look like we’re any closer to that than we are right now with clinical pharmacology, where you give a dose and check the blood levels sometime later and titrate the dose up or down.
Even if a person is high or low for a major metabolism of a drug, there are still enough times that when you give the drug, you find that the plasma levels are unexpectedly high or unexpectedly low because of other genes.
Do you think there is a chance of eventually reaching a nine-out-of-10 success rate for dosages?
I can’t see that happening anytime in the near future.
What should pharmacogenomics be shooting for, given that outlook?
As I alluded to at the end of my reviews, the field of metabolomics sounds really promising. These techniques are really being refined, and maybe metabolomics — and possibly proteomics — can complement what we find out about genomic paneling of a drug.
There are some examples that the drug companies are looking at now, and they’ve been able to predict which patients will respond favorably to a particular drug just by looking at the metabolite profile in the urine. These are thousands of low-molecular weight molecules, and they don’t really know what these molecules are, just that the pattern looks like this for somebody who is going to respond.
In my own field of environmental health rather than clinical pharmacology, I can see where occupationally exposed people, or people smoking cigarettes, or even people with heart disease or renal disease in the family — if they’re followed every year or five years with a metabolomic profile, I think the day will come down the road that you’ll be able to predict, ‘This person is within a few years of having angina, or he needs an angiogram.’
Do you feel the same way about gene expression profiling?
The problem there is, it’s great for tumors — it’s been a major breakthrough in tumor therapy and prognosis of what drug to use, what drug might fail — just due to patterns in expression profile.
You can look at tumor or blood cells or a tissue biopsy — but it pretty much depends on excreta. In other words, you can’t follow — with brain biopsies — a person who’s on a drug for bipolar disorder. So the microarrays have the limitation of being pretty much [used on] discarded tumors, tissue, excreta, or blood.
And how about using expression profiling as a person is undergoing drug treatment?
Except for chemotherapeutic agents for cancer — if you looked at the white cells in blood, that would be one thing. But if you are giving a drug that is being metabolized in the liver, but it has its effect on the lungs, or has its effect on the brain, there is most likely not going to be anything in the white cells in blood or any easily obtainable tissue that you can do a microarray on.
Then is this all good money after bad?
I was invited to write a review for Nature — for the pharmacogenomics special in September — but my review was too sober and was turned down. It was reviewed and they said that part of the money being paid to [produce] this special issue was from pharmaceutical companies, and it would not be nice to bite the hand that feeds you, that sort of thing. That review was turned down.
I turned around and put it in the European Journal of Pharmacology pretty quickly — “[Advances in Pharmacogenomics and Individualized Drug Therapy] Exciting Challenges That Lie Ahead.”
How would you summarize it — what’s exciting?
It’s exciting how much we don’t know about the genome. It seems to be almost like a living breathing organism.
If you knock a gene — we knocked out the mouse CYP1A2 gene, and everyone pretty much agrees that this takes care of arylamines and does little else. It handles maybe five to 10 drugs. When we knocked it out, we tested some muscle relaxant that is metabolized by 1A2 and also 2E1, and without this gene the mice were paralyzed for 12 hours instead of 20 minutes, it was real dramatic.
Then we did a microarray of liver of untreated 1A2 knockout versus 1A2 [treated] littermates, and there were about 30 genes that were up- and down-regulated quite dramatically in the knockout. And these genes were involved in fatty acid and cholesterol biosynthesis and cell cycle. I guess it sort of told us that each gene is contributing a number of things to the genome — if you take it out, they get a committee together and they decide who has to go up and who has to go down to compensate for the loss of a gene.
I mentioned in the review that Bill Evans had a great study from St. Jude’s in the New England Journal [of Medicine] just recently. Methotrexate [or another drug] was given to acute leukemia patients — children — and they drew blood, and there were about 110 genes that were up- and down-regulated in the white cells, with a microarray. Here’s a drug that you anticipate is going to be metabolized by three or six or 10 enzymes, but all kinds of things are happening in the genome.
What is needed for the future of pharmacogenomics?
The last table [of the EJP review] is what is still needed from genomics. The first thing is that we have to identify every exon, and we’re not even close. Every time I go into the database, I find enormous numbers of errors — genes fused together and exons existing where they don’t and vice versa.
I think in this review I mention examples of where the regulatory region of a gene is several thousand or even a million bases away from the gene that it regulates. These are all critical things.
And there are conserved non-genic sequences now that we don’t even understand. [Numbering] two-and-a-half times the number of genes are these non-genic sequences. Whatever they’re doing, we assume they’re altering gene expression in some way. This is just another unknown.
Long conserved variable regions — it turns out that two independent labs I know [from 2004 wrote papers on] regions between 100 kilobases and 2 million bases in size, and they’re duplicated. On average, individuals have at least 11 of these in their genomes. So the question comes back — whose genome do you consider normal? If you have a region of two genes or ten genes that are duplicated in a person, this is going to turn up metabolizing or receptor capability at least two fold.
All the SNP databases are almost completely exclusive with Caucasians, Africans, and Asians. Oceanians and Amerindians make up at least half of the intercontinental variability on this planet. Rather being Caucasio-centric, we should start considering that the world has other races, and they are all really mixed into the very homogeneous European or American population. We have to consider these other SNPs to be important.
What has to happen before these things are cleared up?
Over the next decade I think we will go about as far as we can go. We need at least 20 or two dozen people from all five major geographically isolated subgroups to discover all the important SNPs. And from that we can use software to tease out the tag SNPs, then use tag SNPs on another 500 individuals to determine the biallelic frequencies.
Then when you have all that, you have the problem of the unequivocal phenotype. Phenotype, genotype, association studies — I said a large cohort should be 500, but somebody in the audience in Santorini said, ‘You’re talking 5,000 or 10,000 if you consider the degree of penetrance or expressivity of many of these traits.’
There was a proposal in the May, 2003, Nature Genetics proposing the human phenome project, just to document all possible phenotypes into a large file.