Use of Pharmacogenetics in Clinical Medicine – Tristan Sissung

Use of Pharmacogenetics in Clinical Medicine – Tristan Sissung


Tristan Sissung:
Okay, well, thanks for having me today. I think I saw in the — I was billed as an M.D.
Ph.D., so while I appreciate the honorary degree, I’m only a Ph.D. and an M.S, and I’m
basically a bench scientist who does a lot of translational research, so that’s my perspective.
I’m going to talk a lot about molecular pathways and that sort of thing today. So, anyway, let me launch in. I’m going to
talk about pharmacogenetics today, and I think probably one of the best compelling stories
for pharmacogenetics is this paper right here. It’s a case report about a 2-year-old boy
who went — underwent a tonsillectomy. After his surgery, everything went fine. He was
in outpatient, he went home, took codeine to manage the pain, and died a couple of days
later of respiratory depression. So didn’t really understand why this had happened.
According to the paper, because he had taken the correct amount of pills, there was nothing
— no overdose due to the pills. But somebody got smart and genotyped him, and found out
that he has, instead of two copies of a gene that activates codeine into morphine, called
CYP2D6, right here, he has three copies of this gene. So everybody inherits one chromosome
from Mom and Dad. In general, you have two genes. This kid had a duplication on one of
his chromosomes that gave him three copies, so he was considered an ultra-rapid metabolizer
of codeine into morphine, and while there is probably about 5 percent of the U.S. population
has this ultra-rapid metabolizer genotype, this probably indicates that this kid had
some other confounding issues. It wasn’t just that. Nonetheless, had he been genotyped,
codeine would have been avoided, and he probably would still be alive today. On the other side of the coin, people go to
the dentist, they get dental work done, they are given codeine to manage the pain when
they get home. When they get home, sometimes they take the codeine, and it doesn’t really
have much of an effect. This is because there’s a lot of people in the population who are
deficient in CYP2D6 and cannot turn codeine into morphine. Codeine has very little analgesic
effect. It’s really the morphine conversion that is needed for that. So the same gene
can cause inefficacy, and it can cause severe toxicities. So I work at the NCI, so a lot of my slides
have cancer drugs on them, so this is no different. There is a lot of variation in most drug therapies,
especially in cancer, but, you know, you can see 3- to 50-fold variation in certain drug
therapies, and this variability is partially, oftentimes, attributed to genetics, but not
always, which leads to the next slide. I’m sure people in this room can probably think
of more sources of variability, but I’m just going to go through them each here. So drug specific, dose schedule, the dosage
form, how the drug is formulated, et cetera, can affect the variability. Body size, body
composition, demographic variables such as age, race, sex, can affect drug therapies.
Physiologic, especially disease states, hepatic and renal function, can affect how drugs are
handled in the body. Environmental interactions, like drug-drug interactions, drug-food interactions,
these sorts of things, can affect. And genetics is just one sort of these many variables that
will affect drug therapy. So we see this more as a useful tool and not the “end all, be
all” of determining variability in drug therapies. Now sometimes the genetics is extremely important,
and sometimes it’s not important, barely at all. So today I’m going to be primarily talking
about cases where the genetics really contributes a lot to the variability, and it’s actually
useful for making clinical decisions. There are several types of pharmacogenetic endpoints
that we use at the NCI. We’re — like I said, I’m more of a wet bench kind of guy so I’ve
been, you know, handling a lot of the mining, a lot of the samples in our clinical pharmacology
program, primarily from cancer patients, and we will notice from time to time that there
is an association between a gene SNP and some sort of clinical outcome that we can then
go and figure out why this is happening. So here we had a group of patients with prostate
cancer treated with docetaxel. We found that a polymorphism in a gene, CYP1B1, was related
to the outcome. So men carrying the wild-type *1 SNP had a double overall survival compared
to people carrying the *3 SNP. This gene does not metabolize docetaxel, so we had to do
a little investigative work to figure out what was going on. We found out that estradiol
is actually metabolized by CYP1B1. CYP1B1 is also upregulated in almost every single
prostate tumor. Those carrying the *3 allele turn estradiol into a very reactive metabolite
that binds to docetaxel and adducts it. And this form of docetaxel is not very potent
at all. It also interferes with microtubule polymerization, because this reactive form
of estradiol will bind to practically everything in the cell, and it really likes the sulfinyl
groups on tubulin. So we would have never found this interaction without the use of
pharmacogenetics, so we use it in the discovery capacity. We’re also doing a lot of clinical trials
at the NIH, as I’m sure you all know, and so we’re often looking at variation in phenotypes.
So there’s a molecular pathway that feeds into a variation of phenotypes. So here we
were studying an investigative drug that was shown to cause QT prolongation. We knew the
drug was handled by a transporter that existed in the heart, and basically, the transporter
functioned so that when the drug got into the heart, it was pumped back out. Patients
who were not able to pump the drug out as effectively, because of a genetic polymorphism,
are shown here, they had QT prolongation, whereas patients who were more effectively
able to pump the drug out had barely any, if at all, QT prolongation. So here we’re
looking at a variation of phenotype. We had the molecular pathway sort of characterized. Now both of these feed into clinical trial
inclusion and exclusion criteria. You can take people who are responders, non-responders,
or you’re going to get significant toxicities. You can take them out of your population and
treat them with other sorts of drugs, and subset your population for people where you
think the drug is going to be more effective. And all of this, of course, leads into actual
translation of these findings into clinical practice. So today, the objectives are: to review the
molecular and physiological basis for drug-drug — or gene-drug interactions; to appreciate
the impact on drug therapy; to discuss the future of pharmacogenetics and drug development
and treatment. So basically, I’m going to sort of give you a bird’s eye view of pharmacogenes
and what they do. I’m going to talk about how the molecular pathway will alter phenotype,
which will then alter drug therapy. And then, at the very end of the talk, the NIH has instituted
a pharmacogenetics program where a patient comes into the hospital, they are genotyped,
and that genotype follows them around the hospital in our computer system. So if they’re
given — if the doc wants to give them a drug, they will have to put it into the system.
The system will flag if there is a genetic issue with administering that particular drug.
So I’m going to talk about the drugs that we’ve flagged as important at the NIH at the
end. So just launching with the types of pharmacogenes.
So probably the — when people think of pharmacogenetics, they tend to think of these sorts of interactions
where you have Phase I metabolism, which tends to be just redox reactions that oxidize drugs.
Sorry, the arrow got scooted over there. So here you just have the drug that’s oxygenated
and becomes more polar. Now this can have two effects. One, it can activate drugs, like
with codeine, but, in general, it deactivates drugs and makes them more soluble, readily
excretable. Phase II metabolism is also the chemical modification of a drug. Here, you
take a polar R group and add it onto the drug. So you have drug, drug R; the R is polar,
it’s more soluble, and easier to detoxify the drug. Before I go into the CYPs that I’m going to
talk about, it’s helpful to think about what are the major CYPs that metabolize most of
the pharmaceutical armamentarium. In general, it’s CYP3A4. 3A family probably metabolizes
40 to 60 percent of the drugs that are available right now. This is an old slide, but little
has changed in the last 13, 14 years. This gene really does not have very many genetic
polymorphisms that are very predictive, so I’m not going to really talk about CYP3As
today; however, the next two most frequent metabolizers of drugs, CYP2C9 and CYP2D6,
do have some very important genetic variants that will alter their activity. So I’m going
to talk about those today. Phase II metabolizing enzymes tend to be the
UGTs. You have — you have these UGTs in the liver, the glucuronidate drugs, and make them
more readily excretable in the bile and urine. Then your sulfotransferases, and then a host
of others that are more or less important in the major metabolism of multiple drugs. I’m going to talk about TPMT today. Even though
this is a very small sliver, this particular gene is quite important in pharmacology. So
I’m going to give you the first example here, CYP2D6 and tamoxifen. It’s already mentioned
codeine will activate — or, I’m sorry, CYP2D6 will activate codeine. CYP2D6 actually also
activates tamoxifen. When tamoxifen was developed, people were thinking, “I believe that the
NDM, or the 4-hydroxy, were the major metabolites that were actually active.” Relatively recently,
some studies at Georgetown proved that it was endoxifen that’s really the active compound
of tamoxifen. Endoxifen is formed through N-desmethyl tamoxifen, which I’m going to
call NDM, and it forms this compound which is 3- to 100-fold more active than tamoxifen
or NDM alone. Also, when tamoxifen was being used, people
noticed that SSRIs actually inhibited the hot flashes that people would experience when
they were, you know, undergoing tamoxifen therapy, and I don’t think people really understood
why until recently, when they found that really, what they were doing, was inhibiting the enzyme
that formed the active metabolite. So you had less active metabolite and less hot flashes
due to that. So it’s kind of useful to think about, “How does the population break down
in terms of CYP2D6 genetics?” We would expect, just to back up, that people who were deficient
in this would have more N-desmethyl tamoxifen to endoxifen ratio. People who were very were
rapid would have more endoxifen to NDM. The poor metabolizers who do not form as much
of the active metabolize comprise probably about 10 percent of the population, roughly,
and they’re at the top right here. On the bottom right, you’ll see about maybe another
5 to 10 percent who are ultra-rapid metabolizers; they form a lot of endoxifen and the drug
is actually, probably, more effective in these people, especially. When the drug was developed,
though, these two extreme ends of the genetic spectrum here were not the general population.
The drug was really developed for people sort of in the middle. And the people at the ends,
unfortunately, don’t benefit as well from the drug, or they have more hot flashes, more
toxicity to deal with. So, you know, this is how the population breaks down. Go ahead
and talk about the plasma concentrations of the drugs now. So if you look at the endoxifen to NDM ratio,
and you take the population, look at their plasma concentrations, you get — they put
it on a Normit plot, which just is sort of a statistical method to figure out what groups
of people comprise this population. You’ll find four bell-shaped curves that are very
distinct of endoxifen-to-NDM ratio. These people on the left end here have little endoxifen
to NDM, these have high endoxifen to NDM, so we would expect then, that if CYP2D6 was
really an important genetic predictor of endoxifen concentration, that you would see this curve
enriched for poor metabolizers and this one enriched for rapid metabolizers. And that’s
exactly what you see. Draw your attention to the right-hand side of the table here.
The poor metabolizers over here are the major constituents of Group 1, which has low endoxifen.
The ultra-rapid, or extensive metabolizers, are those that comprise Group 4, which have
high endoxifen. Here’s another way to look at the data, and
I want to point something out here. The poor metabolizers tend to cluster low on the endoxifen-to-NDM
ratio, whereas the extensive metabolizers are high on it. However, you’ll notice how
much the data really spread here. There’s several extensive metabolizers that look like
poor metabolizers. This is because this gene is not a perfect predictor of anything. However,
it is still a very useful predictor. So if you look at patients with extensive metabolizing
versus poor metabolizing, how long it takes them to have recurrent breast cancer, you’ll
see this, where patients with extensive metabolism are benefitting much more from tamoxifen than
patients who are poor metabolizers. So, basically, we think that the poor metabolism group here
really is not benefitting as much from tamoxifen. They should probably be given another drug,
such as an aromatase inhibitor or something else, whereas people who are extensive metabolizers
probably benefit more from tamoxifen than they do from other drugs. So when you think about, you know, this issue
in terms of, “How does tamoxifen stack up with one of these aromatase inhibitors,” for
example. Tamoxifen is causing a little bit more recurrence, however, this part of the
Kaplan-Meier analysis here is composed of a lot of poor metabolizers who are sort of
dragging down the efficacy of tamoxifen. And right now, studies are really trying to compare
these two curves to see if taking poor metabolizers out of here and moving them to here will actually
improve this curve. And some early data from one of these trials is indicating that poor
metabolizers that are switched to anastrozole after two years of tamoxifen experience no
increase in breast cancer recurrence. So the poor metabolizers who are switched are actually
doing better than they would have done on tamoxifen, is really the idea. So I talked about a Phase I metabolizing enzyme,
CYP2D6; now I’m going to switch gears and talk about Phase II metabolizing enzymes.
We’ll talk about thiopurine methyltransferase and 6-mercaptopurine and its analogs. So the
thiopurine methyltransferase just simply methylates drugs and deactivates them through methylation.
6-mercaptopurine and its analogs are used to treat ALL, inflammatory bowel disease,
and autoimmune disorders. They’re fairly heavily used in the transplant community as well,
especially azathioprine and the transplant community, I’ll mention that in a minute.
These drugs basically just incorporate cytotoxic thioguanine nucleotides into the DNA, which
causes the cell to die. However, they also do a second thing. They inhibit de novo purine
synthesis, so the cell is not as able to synthesize DNA and divide it as it otherwise would be.
So they’re very good drugs. 6-mercaptopurine was heavily used in childhood ALL, and some
of the initial pharmacogenetics studies actually were very concerned with this drug because
this drug can cause severe hemotoxicity, in childhood patients can cause death, so St.
Jude was very interested in it, and it was heavily developed at St. Jude. So the TPMT, which basically functions to
take azathioprine, which is converted into 6-MP, right, and then it goes into one of
two fates, inhibiting de novo purine synthesis or incorporating it into DNA and leading to
cytotoxicity. But before it can do that, it will see a lot of TPMT in the blood and other
tissues, where it just gets methylated and inactivated. So when the drug was developed,
the dosing was based off of people who were very able to metabolize mercaptopurine through
TPMT and inactivate it. So the metabolism of these mercaptopurine
drugs is decreased with polymorphic TPMT variation by up to 200-fold. So 200-fold is a very large
number in any therapy, and it has a lot of cytotoxicity in patients who are not able
to methylate it and get rid of it, and these are the kids that are really experiencing
very severe toxicity from 6-MP, so I’ll talk about the SNPs in a second. The rapid metabolizers
are resistant to the drug, the slow metabolizers are at risk. So the rapid metabolizers are
these wild-type individuals who have functional TPMT. They’re about 80 to 98 percent of the
population, depending on which population you’re looking at. The intermediate metabolizers
are — they carry one wild-type allele, and one allele that’s not functional. And they’re
about 65 — they need about 65 percent of the dose, but they’re — they have some toxicity
but it’s not nearly as severe as this group down here of slow metabolizers, who carry
two copies of these two TPMT-deficiency alleles. And they carry about 10 to 15 percent of the
original dose. And if you’re talking about kids, these people are also at risk for secondary
malignancy; so you give them these drugs in childhood, they can develop cancers later
on because they were just administered too much for what they needed. I’m — just got some results back from the
largest pediatric cohort treated with azathioprine, and the results are very positive. The exact
same thing is going on with azathioprine as it is with 6-MP, and the results should be
published within the next year. So it’s not only 6-MP that’s affected, it’s these other
drugs as well, and it’s not just pediatric patients, it’s also adult patients. Oh, by the way, I wanted to mention one other
thing: The genetic variation in TPMT explains 95 percent of these hemotoxicity issues with
6-MP. So all of this information is high level of
evidence. We’ll talk about high levels of evidence in a minute, but it’s made it into
the package insert of 6-MP, at least, and the package insert says that “substantial
dosage reductions may be required to avoid the development of life-threatening bone marrow
suppression in these patients.” Now I’m not a clinician, but I’ve heard that there is
not a lot of genotyping in these patients going on, and this is something that probably
needs to be translated clinically to avoid some of these severe toxicities, especially
in children. So I’m going to switch gears again, to talk
about UGT1A1. This is also a Phase II metabolizing enzyme, very important. It is involved — first,
let me talk about the SNPs. So you have these TA repeats in the promoter of UGT1A1. Normal,
functioning UGT1A1 has six TA repeats. A gene that carries seven TA repeats is expressed
much less effectively in the liver. And if people carry two copies of this allele, called
UGT1A1*28, they have a decreased expression and function of UGT1A1. UGT1A1 is the primary
glucuronidator of bilirubin, so these patients have a slight jaundice phenotype, known as
Gilbert’s syndrome, and this is about 10 percent of the U.S. population has this deficiency.
There are some other SNPs that also that are predictive, I’m not going to go through them,
though. These SNPs explain about 40 percent of the variability in glucuronidation reactions
as a whole. Glucuronidation is absolutely key in irinotecan
toxicity. So irinotecan is administered IV, goes into the blood. These carboxylesterases
cleave certain groups off of irinotecan that turn it into its active metabolite, called
SN38. SN38 is rapidly glucuronidated by UGT1A1, and is completely detoxified when that happens.
If a patient is unable to glucuronidate their SN38, the drug becomes very toxic, and you
can see some severe ADRs again. However, this is very dependent on the irinotecan dose.
This is really what I wanted to bring up. At high dose, almost 100 percent of the patients
who carry this SNP get a severe hemotoxicity, whereas, you know, a moderate amount of patients
with wild-type alleles get the hemotoxicity. However, if you go down to 125 mgs/meter squared,
the — this SNP no longer really matters at all. So this is a very dose-dependent situation,
and so sometimes when we think of pharmacogenetics association, we have to consider other issues
other than just the gene. Yeah, let me go on. So irinotecan toxicity through glucuronidation
reactions has made its way to the package insert of the drug. The package insert — this
one says that the glucuronidation of bilirubin, such as those with Gilbert’s syndrome, people
with that will be at a greater risk of myelosuppression. I think the updated one actually does list
UGT1A1*28 now. Switch gears from Phase II metabolizing enzymes
to transporters. Going to talk about one transporter in particular that’s been very highly studied
in the past five years, and I think is on its way to making it into pharmacogenetics-directed
therapy. It’s this OATP1B1 here. So a patient receives a statin, it goes into the gut, goes
through the gut wall into the portal blood. It can be metabolized in the gut wall by CYP3A4,
or pumped back into the gut wall by MDR1 and MRP2. Once in the portal blood, it basically
needs to see an OATP. OATP1B1 is the primary transporter of simvastatin. There are some
other OATPs that are very important, but unless this statin sees an OATP, it does not very
effectively get into the liver cell. Once in the liver cell, it’s metabolized and eliminated.
Some of it makes it into the bloodstream, and, you know, you have varying levels of
AUC exposure in these patients. What happened there? Here’s a slightly more complex version of
what’s going on in the liver cell. There is a SNP in this gene, a single nucleotide polymorphism,
SNP, in this gene that affects how much statin actually gets into the liver cell. The SNP
is what’s called a non-synonymous transition. You have N in most people, those are the wild-type
allele, at position 130 gets changed to a D, and this actually has a great effect on
AUC exposure of statins. We knew this back in 2006; a very good paper was published showing
that thing is heavily linked to the AUC of statins. Now, greater exposure to statins
can lead to statin-induced myopathies. So in patients carrying the SNP that can’t get
their statins into the liver cell as well, you worry that they’re overexposed and they’re
going to get a myopathy. Another study was published more recently,
looking at 500,000 alleles in the genome. I love this study. It shows that only one
polymorphism was associated with statin-induced myopathy, and not only was it associated,
it was several orders of magnitude over the association threshold, which was just denoted
by that brownish line there. This SNP is almost in 100 percent complete linkage, meaning it’s
co-inherited with that NI30D SNP. So this SNP is probably just a passenger that’s riding
along with the N130D SNP, causing overexposure to statins and statin-induced myopathies.
This group also took these data into a validation cohort, where they had cumulative percentages
of myopathy, and they found that, again, they see the same SNP is — about 20 percent of
the patients are getting statin-induced myopathy, and about 60 percent of statin-induced myopathy
cases could be attributed to this SNP. So this is a very predictive allele, and the
present SNP has a 15 percent representation in the U.S. population. So this is a very
frequent SNP. There’s a lot of people getting statins that are probably at risk for myopathy,
just due to this issue alone. At this point, the FDA has not really weighed in on whether
or not we should genotype for this one yet, but I think it’s coming soon, and at the NIH,
we are genotyping for this. I’m going to talk about targets today as well.
So, you know, drugs are designed to bind to something in the body, and, you know, so these
are drug targets. Most people, when they think of drug targets, think of, you know, your
Imatinibs of the world where they — it’s targeted to a somatic mutation in something
like BCR-ABL. I’m not really going to talk about that today because I’m really concerned
more with the germ-line variation, the DNA that Mom and Dad gave us, not mutations in
tumors. There are other types of targets that are subject to germ-line variation, and I’m
going to talk about that instead. So before I get to the targets, here are two
cytochromes, P450, that take warfarin and convert it into an inactive form of warfarin.
So these — more hydroxylation through 2C9 and CYP4F2 leads to less active warfarin in
the bloodstream. But I’m not really going to focus on the CYP story, I’m going to focus
over here. Warfarin is designed to bind to the vitamin K oxidoreductase C1. By doing
so, it reduces the amount of reduced vitamin K, which reduced vitamin K is pro-clotting
— has a pro-clotting function. So warfarin binds to this target. There is a SNP in this
target gene, VKORC1, that has the — causes the expression of the gene to go down by many-fold.
So if a patient lacks sufficient expression of VKORC1, warfarin will bind it all up and
cause bleeding events. Brief aside on CYP4F2, it was fairly recently
discovered, using a platform I’m going to talk about in a minute, called the DMET platform.
Here’s the association, it’s very strong. The FDA has, again, not weighed in on this
one, but I think it’s going to be up and coming. So here is the incidence of warfarin sensitivity
— I like this paper a lot — showing basically what causes warfarin sensitivity in the general
population. And you can see this sort of red/pink piece of the pie chart and this yellow piece
of the pie chart, correspond to CYP2C9 and VKORC. So about 40 percent of warfarin sensitivity
in the general population can be attributed to these polymorphisms alone. Incidentally,
this CYP2C9 polymorphism, which metabolizes warfarin, is about 1 to 15 percent of the
U.S. population. VKORC variants are more frequent, especially in Caucasians; about 40 percent
of us carry these SNPs that lower VKORC1, and it’s about 12 percent in African Americans. If you look at the package insert, you’ll
find this little table which gives you a warfarin starting dose based on these two SNPs, or,
actually, it’s three SNPs, in VKORC1 and CYP2C9. There’s even a neat little iPhone app that
allows you to put this information in and get a warfarin starting dose. It’s pretty
neat. In this case, if the warfarin was already — dose was already decided upon based on
INRs, then, obviously, you don’t need this information, but it still is useful as a starting
dose — to decide on a starting dose. Okay, I’m going to switch gears again. So
I’ve talked about targets, now I’m going to talk about genes that have effects that are
not necessarily related to the target but are sort of ancillary, you know, targets themselves.
Okay, so, you know, I’ll show you what I’m talking about in a second if that doesn’t
make sense. So you have a tumor lysis syndrome. You have cellular breakdown, which spills
out a lot of DNA. This DNA is catabolized into a lot of purines. These purines can cause
hypouricemia. This uric acid can precipitate in renal tubules and cause renal failure,
so this is known as tumor lysis syndrome. A drug is given to avoid this — actually
two drugs, Allopurinol and Rasburicase can be used. Rasburicase, here, takes uric acid
and converts it into a readily excretable form of uric acid called allantoin. Here is
the actual reaction up here. When urate is converted into allantoin it produces a lot
of hydrogen peroxide. This hydrogen peroxide is clear by glucose-6-phosphate dehydrogenate.
There is a group of people that do not have functional G6PD. They tend to be Mediterranean
in origin, and it’s the same group that cannot eat fava beans, which is why I have the broad
bean up here, because the toxin in fava beans will actually cause the exact same thing to
happen. They’ll get severe hemolysis due to too much hydrogen peroxide. Just an interesting
aside, it’s thought that this population has this deficiency because they want to produce
a lot of peroxide in the bloodstream because they want to combat malaria. It’s a kind of
interesting idea. So, anyway, genotyping for G6PD is a very, very good predictor of G6PD
function and so this is a genetic test as well. And the last type of gene-drug interaction
I’m going to talk about are these hypersensitivity reactions which are becoming increasingly
important, I think, in pharmacotherapy. So a drug like abacavir goes into an antigen-presenting
cell where it sees one of these major histocompatability complexes. These image C proteins are encoded
by human leukocyte antigen, which is called HLA. These are the genes in the genome, so
I’m going to say HLA referring to these proteins here, the genes for these proteins, anyway.
These proteins will bind to your drug, go out and start to amount an immune response
to the drug itself which causes hypersensitivity. And it’s really — it’s a Stevens-Johnson
syndrome in general. And here’s a kid with Stevens-Johnson. This is really considered
— it’s starting to be considered malpractice to not genotype for this before you give some
certain drugs, especially abacavir. There are similar results with carbamazepine and
Allopurinol, still only recommended by the FDA, but it’s still extremely predictive of
hypersensitivity reactions. Just a simple genotype test can really tell you who’s going
to get it and who will not. About 5 percent of patients get abacavir hypersensitivity.
If they have one of these HLA loci, you can have up to 103-fold odds ratio of risk of
getting hypersensitivity reactions. It’s 100 percent positive predictive value; if the
patient has this genetic background, they are almost certain to get a hypersensitivity.
It also has a 97 percent negative predictive value; if they don’t have the SNP, you can
be 97 percent sure that they’re not going to get hypersensitivity. And here is one of
the — the conclusion of one of the Sentinel papers investigating this, I’m just going
to read it. “In our population” — Australians — “withholding abacavir and those with HLA-B*5701
or these other HLAs should reduce the prevalence of hypersensitivity from 9 to 2.5 percent
without inappropriately denying abacavir to any patient.” And I think that’s really a
very good summation of the power of these HLA genotypes. So I have sort of given you the bird’s eye
view of all of the pharmacogenes that are currently out there and are probably moving
towards the translation side. Now, I’m going to just briefly mention one of the platforms
that we use to actually get the genotypes in these patients, just talk to you a little
bit about it. This ChIP, it’s an array-based technology called DMET, which stands for drug
metabolizing enzymes and transporters. It has 2,000 variants and 235 PK/PD genes, so
you can see all of these Phase I enzymes, you’ll see the ones I mentioned in there;
the Phase II enzymes, you’ll see the ones I mentioned in there; transporters you’ll
see the ones, again, the SLCO1B1 is in here. And then these other genes that can have effects
on PK/PD, so here’s G6PD, for example, cytidine deaminase, which is important for certain
other drugs, et cetera. This ChIP is actually — it only costs about
$500 to do the ChIP, and if you batch a lot of samples, as we’ve learned, it actually
costs only about $50 a patient. So it’s not some outrageously costly thing to do. However,
it does have one major deficiency that we’ve identified, and that is that it takes three
days to actually get data out of this, and that’s a fast turnaround time. So for a lot
of these drugs, if you need the information right away, you cannot get it, it’s just not
possible. This isn’t CSI Miami; we can’t just genotype something in 15 minutes. So, basically, what we’ve done at the NIH
to combat this issue is we have made a policy where a patient gets admitted, and then they
get this genotyping test done; the information follows them around so that if a clinical
decision has to be made rapidly, then this information is there and available, and will
be flagged to the clinician who is going to give them the drug. We’ve based the — talking about our experience
with PG testing at NIH, we’ve sort of used this website called PharmGKB, which is run
by a lot of the pharmacogenetic experts in this country. They are all a part of a network
called PGRN, and they have really curated the pharmacogenetics literature very well.
So if you’re interested in this, PharmGKB is an excellent resource for learning more.
They’ve published levels of evidence, so we have only selected those that have the highest
levels of evidence that are available; published control studies of good quality relating to
phenotype and/or genotype patients; healthy volunteers having relevant pharmacokinetic
and clinical influence. Pretty much everything I’m going to discuss today has that high of
a level of evidence. It also has a very high level of clinical relevance, so even though
maybe you have a high level of evidence that a SNP is associated with some outcome, that
outcome may not be that clinically important. So they’ve also curated the clinical importance
of this, and all of these genes I’m about to talk about have a high level of clinical
importance as well. So I’m just going to go through the list,
because I think, you know, you may see some of your favorite drugs on this list, and I’m
going to keep it short so that I don’t keep you here for too long, but here we go. Abacavir,
I already mentioned this one, HLA, B57O1; this one is recommended, so if an investigator
will get flagged, this says you really should — you really need to get this genotype before
you can administer abacavir. And even though it says — the test says TBD, our laboratory
medicine branch actually runs this test all of the time so we’re currently processing
this SNP through that branch, anybody treated with abacavir. Allopurinol, another drug with
hypersensitivity reactions. Same story, it’s recommended and can be run through the lab
right now. Azathioprine or any of these mercaptopurine drugs, I already mentioned these so I won’t
go through the mechanism; this is also a very, very highly, strongly-recommended SNP to test
before administering any of these drugs, and we can actually use the DMET platform to do
so. Carbamazepine is another HLA. The FDA recommends testing this in Asian populations. Now, this is an issue here. So I have a friend
in — I’m from California — I have a friend who’s grandfather is — was one of the original
Japanese immigrants to the United States, and he doesn’t look at all Asian, but he has
a significant part of his genome that is Asian. He wouldn’t identify himself as Asian, he
would identify himself as a Caucasian. If he was treated with this drug, because he
wasn’t Asian and we decided not to genotype him, then he could potentially experience
some severe reaction here. So we’ve decided that really looking at a person’s self-identified
race is not the way to go about this. We really need to actually genotype every patient to
find out if they have this SNP or not. So this one is actually very recommended; test
is, again, through the laboratory branch. Clopidogrel, Plavix, the poor metabolizers
have non-responsiveness to clopidogrel. Higher doses may be needed in these patients, or
there’s new anti-platelet agents out that can be used instead of clopidogrel. This one
we consider optional or available, but we assume that since the information’s already
available to the clinician, that they will just opt for one of those other anti-platelet
agents. Codeine, I already mentioned it; we don’t use a lot of codeine at the NIH. This
one’s still is optional or available; the DMET will give you the information. Fluoropyrimidine’s
metabolized by DPYD. Patients with deficiencies of DPYD will have some potentially fatal toxicities,
so this test is recommended, and it’s already available to the clinician by the DMET ChIP.
Interferon alpha has an association with IL28-β SNP. This is — one SNP is very predictive
to who is going to respond well to this drug, and then another is predictive of who will
not respond well to the drug. We consider this optional or available. We have to go
outside of NIH to LabCorp to really do this one. Irinotecan, I already mentioned it. We
— DMET ChIP already tests UGT1A1 so this one’s already being used. Isoniazid with NAT2;
NAT2 is a phase two conjugating enzyme that acetylates isoniazid and gets rids of a very
reactive intermediate metabolite. If people are slow acetylators, they’re have a threefold
increase in drug-induced liver injuries. This one is considered optional or available; the
DMET tests it. CYP2D1, similar story, go through it optional or available. Phenytoin: difficult
drug to dose. There is some variance in CYP2C9, which affect the toxicity and efficacy. This
information will be available for dosing phenytoin. Phenytoin also causes some hypersensitivity
reactions, and there’s an HLA that’s predictive, so this one’s strongly recommended, and the
test is done through the laboratory branch. Rasburicase, which I already mentioned; G6PD
genotyping is already available through DMET. Statins and OATP1B1, mentioned it; test is
available through DMET. Tamoxifen 2D6; test is available through DMET. Warfarin: same
SNPs, DMET test. And then we have the molecular pathology laboratory who is already doing
all of the somatic mutations for these targeted agents, so I’ll just run through the targeted
agents and not mention much about them. Trastuzumab, lapatinib, imatinib, dasatinib, and nilotinib.
And imatinib also affects KIT, so we have the molecular pathologies test KIT for us.
Gefitinib, renotalib [spelled phonetically], and these others. BRAF inhibitors, EGFR inhibitors,
RET inhibitors; alkylating agents, and that’s it. So those are all the drugs that we have
implemented at this point at the NIH in the PG testing arena. So just a couple of final thoughts. How many
drugs have pharmacogenetic markers in the label? Well, at this point, there are 114
of these drugs, and if you go on to this website at the FDA, you can look at all of these drugs.
How many drugs had FDA recommendations that are actually actionable? Seven have boxed
warnings that — where the testing is very important; 29 have indications and usage information;
and 24 will give you information about the dosage. So a subset of those are actionable.
And the last slide here, just considering the prevalence of use of pharmacogenetically-affected
drugs. There’s about 24 million people — this was in 2008 — using drugs that are — that
have pharmacogenetic information that’s available that you can just genotype them and know what
— know more information, anyway, about what to do to make clinical decisions. There’s
a lot of people using these drugs. This number is just ever increasing, and eventually, they
think this stuff is really going to be important in clinical medicine. And Doug Figg, my boss, always ends his talk
by saying, one day, he envisions a child is born, the child gets a DMET ChIP-like genetic
test, and that test can be carried with them through life on a thumb drive, and they can
go hand it to their doctor one day, doc put it into a database, it’ll tell them, “Don’t
give this drug, do give this drug.” So, that seems to be the way that things are going.
And so that’s all I have to say, and thank you very much. [applause] Male Speaker:
Comments or questions? Yes. Male Speaker:
If I want to start a patient on Clopidogrel — Tristan Sissung:
[affirmative] Male Speaker:
— how do I find out if it’s going to be effective, what do I actually do? Male Speaker:
Could you — could you paraphrase the question [inaudible]? Tristan Sissung:
Yeah. So the question was how do you find out if a patient is at risk for Clopidogrel
inefficacy? And you can use a few options. First option is you can send it off to have
it genotyped by a private company. There are several private companies out there right
now doing this. The test really needs to have, I think, three different alleles, and each
one of those alleles can cost a certain amount of money. We’ve found that it’s actually cheapest
to just have the DMET ChIP run on people. You can take the blood sample, you can send
it to the Coriell Institute, they will give you the information back. A guy named Norman
Gerry there is the guy we run through; he’s doing all of the NIH studies. You can get
this information back, and then make the decision based on that. Male Speaker:
Yes? Male Speaker:
Thank you for a great talk. You’ve raised a lot of important issues. I’m sure I see
at least one patient a week that’s either slow or rapid metabolizer that’s not doing
well clinically. Tristan Sissung:
[affirmative] Male Speaker:
There was a Dr. Flockhart [spelled phonetically] in prior practice that used to do consults,
so how can we get consult in terms of private practice to help us, because these two issues,
one is a specific drug, one is — metabolizers slower that might affect many, many drugs,
and that might be beyond the expertise of the private practice doctor. Tristan Sissung:
That’s absolutely right. I know there is some agencies that are — that are springing up
that offer pharmacogenetic consulting to clinicians. It’s a very new thing, you can Google search
it, I know that Doug Figg was approached by one of these agencies, I forget the name of
it, but we’re also at the NIH, and I’m sure we can — we can direct you in the right direction.
I think my email is up here. And if we can’t help you, I’m sure we can put you in touch
with somebody who can at this point. Male Speaker:
Those of you who are entrepreneurs, it sounds like that’s an opportunity. Tristan Sissung:
It is definitely. [laughter] Male Speaker:
I want to reiterate the excellent nature of this program, and quite timely and relevant
to private practice. Interestingly enough, just from an historical point of view, the
6-MP discoverers won the Nobel Prize, you may be aware of that, [inaudible] in the 1980’s,
but to take that a step further, there have been some recent guidelines that have been
published by a national — our national organization suggesting that HLA-B5801 profiles be obtained,
and that certain groups of patients who are going to be admitted Allopurinol, and happen
to be the Kahn [spelled phonetically] Chinese and certain Thai subgroups. But getting back to your California story,
you wonder how many of these particular groups may be here and vulnerable because this is
so important for the Allopurinol hypersensitivity syndrome. So from bench to bedside, this is
recommended, we’re looking at the economics of this as we speak, and to the practicality
and bench to bedside we are told that this HLA-B5801 is now available commercially. Is
this in the area that you are — have studied more than your slides? Tristan Sissung:
I’m not an expert on HLAs by any stretch of the imagination, but I do know the Allopurinol
story, and I agree with the sentiment that we really need to genotype everyone. So I’m
not sure exactly — can — is there — did that answer your question, or? Male Speaker:
Well, a statement and a question, just to point out the relevancy of this discussion
relevant to clinical practice. Tristan Sissung:
Yeah. So, yeah, I think that this needs to be genotyped in clinical practice; it absolutely
needs to be done because it’s so predictive of who’s going to get these toxicities, it’s
very important. Male Speaker:
[inaudible] the national organizations who are suggesting it. This may entertain another
low culpability by not doing it. Tristan Sissung:
That’s true. I actually — I looked up before I came here — I always look to see if there
has been yet a lawsuit for malpractice about one of these things popping up. Nobody has
yet sued anybody and won, as far as I can tell from Google, for not doing one of these
HLA tests. However, I have found — you mentioned Allopurinol — a woman was misdiagnosed with
gout, was given Allopurinol, got Stevens-Johnson, sued, and won $6 million. So, clearly, it
is, it is something that needs to be addressed clinically. Male Speaker:
Well, you’ll see [unintelligible] on television very quickly on this matter, I think. [laughter] Tristan Sissung:
Lawyers are entrepreneurs, too. [laughs] Male Speaker:
I’m reminded of Norman Shumway in response to a congressional question at a hearing made
the observation that none of us are purebreds. Tristan Sissung:
That’s definitely true, especially in America. We are very admixed. Female Speaker:
Hi. Thank you — Tristan Sissung:
Hi. Female Speaker:
— for a wonderful talk. Tristan Sissung:
Thank you. Female Speaker:
I’m curious — I remember what you mentioned about package inserts having warnings about
genomics, and you also talked about [unintelligible] and how that’s not really helpful, how you
haven’t actually genotype everyone. So I wanted to know if you have an opinion or if you’d
offer your perspective, considering translation, what role or lack of role do you think these
package inserts are playing right now in the translation of this pharmacogenomic information
as to actual use in practice. Tristan Sissung:
Yeah, thank you. So there was a paper published by the people at St. Jude who came up with
the TPMT observation, and they talked about genetic excellence, that the genetic tests
are held to a higher standard than your standard clinical assays just because they’re — people
want them to be so predictive of everything, although they never really will meet that
benchmark. So I think that there is a lot of resistance out there right now to implementing
a lot of this stuff because of that issue. Secondarily, the CYP2D6 tamoxifen story has
been recently stalled by two published studies that came out at the San Antonio Breast Cancer
Symposium showing no relationship between CYP2D6 and tamoxifen outcome. Now, these two
studies were fundamentally flawed. There’s a editorial by Mark Ratain in Cancer Letters
talking about how these two studies both violate a fundamental law of nature: the random sorting
of alleles amongst populations. And the reason for this is that these folks genotyped tumors
and did not genotype the germ-line DNA. The tumors get mutated, and it’s not an accurate
reflection of what’s going on in the liver, how much endoxifen is actually being formed.
So these studies have a lot of impediments to them that are outside the control of a
lot of us who are doing the science, so… Male Speaker:
Thank you. Male Speaker:
Yes, again, I’d like to thank you for an outstanding talk. [inaudible] Ph.D. initiative of the
hospital, so when I hear something like this, you know, my mouth waters a bit. And I wondered
is the Institute interested or thinking about perhaps doing some test drives in community
hospitals in terms of typing individuals coming in and seeing its impact since you all put
[spelled phonetically], certainly at NCI. Where are you with that? Tristan Sissung:
I think that, you know, our group would be partially interested in — Doug Price here
has come for some moral support, he’s a fellow staff scientist in our lab, so, I mean, I
think we could probably talk to Doug Figg about that, maybe doing some of those studies.
Juan Lertora is the guy that runs the PG program right now at NIH, and I think you could definitely
approach him and ask. He would be — he’s always interested to talk about this sort
of information. Male Speaker:
Would you comment on the role of — the traditional role of pharmacists in protecting patients
and how you see that evolve? Tristan Sissung:
Well, I mean, for this, I think — pharmacists are not geneticists, and I know that very
well because I am a geneticist and I have to deal with pharmacists all of the time.
I think that what needs to really happen here on the pharmacy side is that we need to have
some very good curated databases where you can just put in genotype information, and
the people who are experts in genetics and all of the other fields that are needed to
really understand this information, that this database just spits out a clinical decision
that should be made, rather than having the pharmacist do it all. Male Speaker:
So, in fact, at the end of the day, one could conceive of a system that doesn’t lead to
alarm fatigue, which happens now a lot in pharmacies, I think. Get a bunch of interaction
messages, and eventually a pharmacist ignore them. It’s going to take a lot of work it
seems. Tristan Sissung:
Yeah. Male Speaker:
There is a small, but significant incidence of — sorry — small but significant incidence
of fatal malignancies, lymphomas, I believe, in inflammatory bowel patients, and maybe
rheumatoid arthritis patients [inaudible]. Any data on genotyping those? Tristan Sissung:
I don’t know of any, but I’m more of a cancer researcher so I can’t say that there is not.
I was actually recently diagnosed with psoriatic arthritis, and my doc actually mentioned that
to me when I went to him. So — Male Speaker:
Could you repeat the question? Could you repeat — Tristan Sissung:
Oh, I’m sorry, the question was basically there’s secondary malignancies in certain
diseases like arthritis, inflammatory bowel disease, and the question was, do you see
secondary malignancies that are related to those diseases, I think is basically what
you’re saying, right? Male Speaker:
Or is there a genotype that would be predisposed? Tristan Sissung:
Or a genotype that’s predisposed. So that’s more of a risk allele, less of a pharmacogenetic
allele. I could see maybe that if you were treated with azathioprine for inflammatory
bowel disease, that you might see secondary malignancies in patients with certain variants,
but the disease alleles, I just don’t know much about. Male Speaker:
You raised an important issue in terms of clinical trials, and that is, you know, maybe
we should lower the patient population to people most likely to benefit. One example
that I see every day is glucosamine chondroitin works in a subset of the population, but it’s
said ineffective when you look at the whole population. Tristan Sissung:
Interesting. Male Speaker:
Are we any closer to using genetics in clinical trials to make drugs more effective? Tristan Sissung:
There are several out there in the literature right now. They’re finally doing this, which
is exciting. I mean we really needed the prospective side of this. Now, I know that there is some
resistance to drug companies to do — from drug companies to do these sorts of studies
because they want their drug to work in the whole population and in any one subset. So
oftentimes you’ll see these prospective studies already being done on approved drugs. I’m
not aware of any drugs that are being developed at this point with pharmacogenetics in mind,
but I also don’t work for drug companies, so I don’t really know for sure. [laughs] Male Speaker:
Other comments or questions? Yes, sir? Male Speaker:
In the world of saving a few bucks, have you ever noticed any difference between a generic
drug and a — from a genetic point of view — the same drug produced generically versus
the standard drug? Tristan Sissung:
I don’t think anybody has ever done a study like that. I think we primarily assume that
a generic and an on-label, or, I’m sorry, I forget the name, you know, a drug that’s
produced by a drug company are the same compound. So I don’t think we ever look at generics
versus the drug companies’ drugs. Male Speaker:
So the American College of Physicians did a survey on something like 500 of their fellows
and members, and asked a bunch of questions about this sort of thing, and found that,
a) each of us believe that this is a really important field for future practice of medicine;
and b) felt very incompetent in being able to use it. And it seems to me that revolves
around competency rather than knowledge. And one of the reasons we were very interested
in having pharmacogenetics talk here is that this one is very, very close to the clinic
on the bedside. And it seems like maybe we ought to do some more of this. What do you
think? I see heads nodding, maybe we should do a bit more of it. I want to thank you very
much, Dr. Sissung. Tristan Sissung:
Yes, thank you very much.

Leave a Reply

Your email address will not be published. Required fields are marked *