 |
Pharmacogenetic Targeting of Cardiovascular Drugs
by Michael R. Bristow, MD, PhD
S. Gilbert Blount Professor of Medicine
Co-Director, Colorado University Cardiovascular Institute
University of Colorado Health Sciences Center
Denver and Aurora, Colorado
Lecture related to:
Chapter 34: Antiarrythmic Drugs
Slide 1: Title and Conflicts of Interest
I need to make you aware of some conflicts. I'm no longer 100 percent academic, I'm 50 percent. At the University of Colorado, I co-direct a cardiovascular institute there as part of those activities. But I spend half my time in biotech now, and obviously there are conflicts related to that. This biotech company and another one we just sold to Gilead has licensed all the intellectual property (IP) from our laboratories and so I have to make you aware of that conflict, and I'll flag that conflict when we get close to the line on that.
Slide 2: Conceptual Models of Heart Failure
Over the years in heart failure there have been various frames of references that have been proposed to sort of frame the debate about heart failure and to serve as an infrastructure for developing new therapies and new ideas. Most of them actually have been triggered by the introduction of a new idea and then a framework developed around that, and then additional ideas were developed.
And so I'm not going to go through the stages of some of these — our laboratory has been involved in several of these — but I just want to highlight where we think we are at the present; that is in 2007, in terms of a conceptual model of heart failure. We, as the oncologists, are really thinking in terms of altered gene expression in the failing heart as being the conceptual model and basically figuring out ways to normalize or improve that altered gene expression, identify the abnormalities, figure out how to get it back to normal. And that is going to be one of the underlying themes of this lecture, and I'll show you how we're attempting to do that or make use of that in the development of one type of therapy.
Slide 3: Altered Gene Expression in Heart Muscle Disease/Heart Failure
Shown here is the concept of this conceptual framework as has been laid down in some publications. In terms of alterations in gene expression accounting for at least disease progression in heart failure, if not the establishment of an altered phenotype that ultimately develops into the clinical syndrome of heart failure, there are three general ways in which that can happen.
The most obvious way — and this is not any mystery to folks at the University of Utah who have kind of invented this at the Hughes Institute — is single gene mutations leading to altered phenotype. And there are lots of examples of that, in terms of the cardiomyopathy world, both hypertrophic, which was the original gene mutation identified, and the codon 403 beta myosin heavy chain by the Seidmans, but also in the phenotype that causes the majority of heart failure, which is a dilated cardiomyopathy phenotype.
There are all kinds of genes now that have been identified as being associated with that phenotype. Some of them are shown here, lamin A/C, desmin. There's a whole long list. Now, sarcomeric gene mutations and so forth.
And this is good for diagnosis in families. And some information about prognosis derives from that. We have a molecular genetics operation, as most academic centers do, in our cardiovascular institute that specializes in this. But, of course, we are dealing with things that are less than 1% prevalence by definition. In terms of serving as some nidus for therapeutic development, it's somewhat limited in that sense, although one can learn a lot about the biology and natural history by studying these gene mutations. It's typically how they alter the biology and then figuring out something that might be more generalizable from that information.
What actually has more general promise, of course, is a second way in which gene expression can be altered, and that is by holomorphic variation and wild-type genes, either single nucleotide polymorphisms, changes in copy number, etc. And there are lots of examples now in the literature of things that may alter the natural history of heart failure and actually may interact with therapy of heart failure, and I'll be showing you one example of that. In fact, we're going to emphasize that in this lecture.
Then finally a third way in which the dilated cardiomyopathy phenotype can be altered via gene expression is a change in the expression of wild type genes, if you will, regulated changes in expression. Lots of examples of that. We've been instrumental or at least active in the area of an abnormal gene program called the Fetal Gene Program and have actually launched an entire therapeutic development program designed to develop drugs to change that back to normal or prevent its development. That's all partnered with Novartis via Gilead Colorado and so forth, and I'm actually not going to talk about this, but this is a whole other area of therapeutic potential. We're going to concentrate actually on polymorphic variation in this talk.
Slide 4: Chronic Heart Failure Phase III Clinical Trials
So one of the things, in fact, the thing that has led me into this area is our general recent lack of success in phase III heart failure clinical trials, in trying to come up with new therapies that will be effective in the general population of patients with heart failure. We had a tremendous run beginning in the late 80s up until about 1999/2000 in terms of developing new therapies. And then the whole area of drug development in heart failure sort of struck a rock — not I-R-A-Q, but A R-O-C-K — and if one actually looks at the scorecard for what's happened in phase III clinical trials in heart failure over the past few years, you see something like this.
From 1998 to 2006, in terms of just dividing the trials into those that are positive on the primary endpoint versus negative on the primary endpoint, since 2001 the failure to success ratio is 10:4. There's a 29% success rate. And each one of these trials costs upwards of a hundred or more million dollars. So a 29%/30% success rate is an untenable model in drug development when the cost of your phase III program is $100,000,000. So, necessarily we've got to get smarter in terms of how we perform clinical trials, how we approach all of this in terms of drug development in heart failure.
Slide 5: Current Clinical Trial Design in CHF
On the other hand - so this is a summary of that notion before we get to the 'on the other hand.'
Current heart failure clinical trials, whether they are designed by industry or the National Heart, Lung, Blood Institute (NHLBI) are too costly, they take too long, and they're too risky from an efficacy standpoint. Success rate is 30% or less.
And the reason is that the typical agent, the typical effective agent — ones that actually have won in phase III clinical trials — the typical effective agent only works in 30% to 50% of subjects. We've got to find a way to increase the efficacy, the response rate, if you will, up to at least a doubling of that response rate so that we can take down the sample sizes of these trials and guarantee success or at least increase the probability of success.
There are various ways to approach that. One of them is pharmacogenetic. The NHLBI convened a conference that Jeff Tobin and I ran in 2002 on this whole issue, and we tried to convince the NHLBI to actually get involved in this. They declined, feeling it was beyond the scope of their interests. And it was at that point that I began to put together the company (that is listed as a conflict on the introductory slide) to do this on the industry side since it clearly wasn't going to happen at the level of the NHLBI.
The problem in terms of clinical trial design is that the way clinical trials are looked at — the way the clinical trialists see the world is, at least in cardiovascular clinical trials, the Richard Peto model. That model, as summarized in another venue by the great Spanish American philosopher Santayana, is this: keep it simple. Round up thousands and thousands even tens of thousands of patients to the point where you'll be able to detect a small difference and just gun away with a simple primary endpoint and hope something good happens. That is not the way to do clinical trials as far as I'm concerned. It may be attractive to the human brain as alluded to here by Santayana, but it's not really the way to develop drugs or new therapies.
Slide 6: Placebo Cartoon
So we need to do something different. Now here's a possibility. One of the things that are a problem, of course, is how good the placebo is in clinical trials. So if we just had a placebo that was less effective or at least had a worse adverse event profile we might do better. But I think the FDA is probably not going to buy that approach.
Slide 7: Response Rates in MOCHA
Just to give you an idea of the heterogeneity of clinical response in heart failure, here's a drug that's literally dear to the hearts of Mike Gilbert and me. We did all the early work on carvedilol here at the University of Utah in the 1980s. And this is one of the pivotal trials that we did in the early 90s, the so-called MOCHA trial.
This is not the primary endpoint of MOCHA, but this is the change in left ventricular ejection fraction, which is a good surrogate marker for clinical outcome with this type of therapy. If you had to pick a single agent in heart failure, carvedilol would be at the top of your list in terms of agents that are efficacious. So this is sort of a best case scenario looking at responses. Response is defined as what happens to ejection fraction, which is an easy metric to measure, and you get a more quantitative idea of the response rate when you do this.
This trial had three different doses of carvedilol, had a low dose, 6.25 b.i.d., mid/medium dose, and a high dose. And the data are divided. It was a six-month trial. The change in ejection fraction is what's looked at here, and a response is defined as an improvement by at least 5 ejection fraction (EF) units. From 20 to 25, for example, would be a positive response. Negative response is a decline in ejection fraction by 5 EF units, and no response is anything from -5 to +5.
You can see that there's a nice dose-related response rate here, but of course there is a placebo response rate in this trial. Here are the negative responders and here are the no responders. We know that it's the positive responders that ultimately are associated with the clinical response. If you actually subtract the placebo response rate, which you should do and nobody ever does except for the hypertension people, you should subtract the placebo response from the active drug-related responses.
Here are the response rates at low dose, and many patients are on this low dose in the community, 30% response rate, 38% at mid dose, and only 45% at high dose.
And this is the best case scenario for a drug that ultimately got approved and is effective, and I would argue that these numbers need to be higher than that going forward for any new drug going into phase III.
Slide 8: Carvedilol in Pre-Tx Patients
Just to give you an idea, this drop in ejection fraction as a surrogate for negative response rate is just not a stylistic issue. People on beta blockade can actually get into trouble and do worse. This applies particularly to the more advanced patients.
This is a trial that was mandated — the carvedilol component of this clinical experience at least — was mandated by the German regulatory authorities for the approval of carvedilol, since that drug had not been in European patients and had not been in sick patients when they tried to register it in Europe.
What happened was that a protocol was developed in Europe, in Germany and other European countries, in very sick patients. And they went to patients that were listed for transplant, who, by definition, are very sick patients. Most of these patients were treated with carvedilol, but there were a few other beta-blockers, like atenolol in this experience.
They published what happened in this experience, and what happened to these patients is shown here. Many of these patients (122 essentially, out of 318) the doctors wouldn't even touch with a beta-blocker for various reasons: COPD, blood pressure too low, heart rate too low, and so forth. One hundred and ninety-six were given beta blockade, and I think 170 of them got carvedilol. Only 126 patients tolerated beta blockade out of 196.
So in total, 36 patients responded unfavorably to beta blockade. And that's a real issue with this response. It's not just a matter of who responds and the rest of the people have sort of a neutral effect and who cares, we can just empirically give this drug. There is actually a subset of patients that don't do well on this therapy, and that's another thing that needs to be considered up front - how do you identify such patients so you don't expose them to serious adverse events, the most important of which is death.
Slide 9: CRT and Mortality in Advanced Heart Failure
Here's the idea of how to improve things: don't target the general population of patients but target a subpopulation that has a higher probability of response. Figure out a way to do that.
Here's one way to do that. This is actually with a device. The device is cardiac resynchronization therapy (CRT) that Dr. Freedman et al., (University of Utah Division of Cardiology faculty) has lectured on numerous times here. And the way you select patients who are at high likelihood of responding to this therapy is you basically do an EKG, and select those who have prolonged QRSs or lengthened QRSs —intraventricular conduction delay. Basically, those with a high probability of having desynchrony of contraction due to the delay in activation of the ventricle getting to the left ventricular free wall. You can also do this echocardiographically, but the easiest way to do it is by selecting those with a QRS wider than 120 mSec. You select out the patients that are apt to respond to this therapy, synchronizing depolarization and improving contraction of the heart and ultimately improving remodeling of the heart. And when you do that you win, and you actually win fairly easily.
These are two trials in which that win is apparent. One is COMPANION. Here's cardiac resynchronization therapy. This is all cause mortality. So here is CRT alone. Here's a combination of CRT and an ICD. And you pick off a few sudden deaths here and improve the survival benefit when you have the combined device. But clearly CRT is winning in these patients that you've selected for being responders, essentially, to this therapy.
Here's another trial that was published a year later that went longer, and had less sick patients. The curves had an opportunity to separate more than in the other trial with sicker patients. And this is a marked reduction in mortality in the heart failure world, highly statistically significant, a reduction in mortality of over 30%.
So this is an idea of how you can win in a clinical trial by selecting, up front, a responsive subpopulation.
Slide 10: CHF Phase III Clinical Trials
It turns out if you look at trials in which somehow a subpopulation of patients likely to respond has been achieved by study design, there are only winners in the last several years. Here are the winners on the left. There are no losers on the right. Here's COMPANION. Here's CAREHF that I just showed you. Here's the CHARM trial with an ARB (angiotensin receptor blocking agent), basically selecting out patients that couldn't tolerate ACE inhibitors and, therefore, would be more sensitive essentially to any agent that might inhibit the renin-angiotensin system.
And here's the AHab trial with BiDil, a vasodilator combination previously shown to have a heightened response in African Americans. This trial only enrolled African Americans and was highly positive, so these are basically studies selecting out a responsive subpopulation, and clearly this is something that needs to occur going forward.
Now it has been pointed out by numerous pundits, including Francis Collins, it really isn't a great idea to select patients out based on things like race. Race-based therapy is something that is built in at the major discomfort level in at least Americans and certainly American physicians. We need to find a better way to select subpopulations and we need to select based on biology or based on biologic markers or ideally based on something very simple that's a biomarker, such as a genetic test, as opposed to things like skin color. So that is a challenge going forward.
Slide 11: Putting Pharmacogenetics into Practice
One way to do this, of course, is pharmacogenetically. The idea that you can select responsive populations pharmacogenetically with a simple DNA sample has been around for a long time. It's been talked about as the wave of the future for at least a decade and beyond. There's an enormous amount of activity in this sector. This slide basically summarizes the activity. If anybody's interested in this area and you want a review of what's happening, this is the place to find a general overview of what is happening.
There are 264 public research groups worldwide, with 73 in the U.S. engaged in pharmacogenetic research. There are 15 big pharmas with pharmacogenetic capability either in-house or with alliances with tool companies. There are 64 or 65 small to medium-sized firms engaged in commercial pharmacogenetics. ARCA Discovery is one of those, and it's one of the 27 different therapeutic companies engaged in this activity.
Now, with all this activity something good eventually is going to happen. If for no other reason, blind dumb luck will prevail. It's not happening very fast — and I'm going to be talking about why that's true — but it is happening, and I'm going to give you the examples of how it is happening and give you some sense of how close we are to being successful in this area.
Slide 12: Human Genome
Of course, all of this is made easier by the Human Genome Project and the elucidation of genetic variation, which is now available to a large extent on public databases. I don't have to talk about that here at a place where some of this literally was invented.
Slide 13: Approaches to Pharmacogenetic Therapeutic Targeting
Now, there are two ways basically to engage in pharmacogenetic targeting of a drug or a therapy.
The first bullet here is typically what big pharma does. It's essentially a data dredging, retrospective series of activities. What big pharma typically will do is set up a DNA bank and a phase III clinical trial. They typically do whole genome scanning. They generate terabytes of data, typically in collaboration with a tool company, such as Perlegen, for example, and they essentially carpet-bomb the genome. They generate enormous amounts of data. The advantage of this is that you don't have to have any preconceived idea of what you might find, and so inevitably you do find things that are associated with efficacy as well as adverse events.
Of course with terabytes of data involved, this can be chance alone. In fact, statistically it most likely is, and so you have to do another study where you prospectively identify in a hypothesis-driven way what you are looking for, hopefully, with some idea of what the biology of that might be and then prove it in a second study. All of this is very costly and very time consuming and so forth. But there are many big pharmas that actually are committed to this now, most notably Glaxo Smith Kline (GSK) under the direction of Allen Roses.
Another way to do this (and this is the way that we've done it) is to start with your pharmacogenetic target. The way we've done it and the way this is probably going to be done by other groups, is that you become interested in polymorphic variation within the gene or the protein that you study in your group, within some mechanistic pathway you're interested in that you know or think might be important.
You start with a target. You spend years characterizing the target. You examine then the interaction of this candidate polymorphism with a treatment effect in a clinical trial. You set up a DNA bank. You roll it into the trial with a prospective hypothesis saying that this gene variant may be or is associated with an interaction with treatment that's either favorable or unfavorable. In other words, it's prospectively done.
The advantage of this is that it's a lot less costly. The disadvantage is that it takes a long time. There's a big challenge in identifying a gene variation that might be important.
Slide 14: Criteria
This slide shows the challenge, laid out. If you're interested in developing a new therapy, you really want it to have a certain amount of prevalence to be practical. You have to generate the funding to be able to do this. So, if you are a private company operating on investment capital, nobody is going to fund you if there is no market at the other end. And the same thing occurs in a big pharma. You're not going to get the budget from your central budgeting apparatus unless you've got a market associated with this. And so you've got to have a certain amount of prevalence here of whatever genetic variation you are after to be practical, and we set that number at 10%. We want to see something that has at least 10% prevalence.
The second thing is that your gene variation has to be functionally important. It can't just be a SNP that doesn't mean anything, that doesn't produce a change in amino acid, for example, and it has to be something that actually produces a change in function. Moreover, it has to reside within a critical, disease-altering pathway. It can't be in some redundant pathway that doesn't mean anything. It has to be in something that really makes a difference in the disease indication that you are studying. And finally, there has to be a potential to intervene at that level therapeutically. You have got to have a drug or a treatment essentially that's going to alter the function of this genetic variant, and ideally that treatment would do it uniquely so that you would have exclusivity against other drugs in the class.
This is the challenge, and if you sort of look at the odds — and these are just estimates of the probabilities of each of these as individual events, and then you have to multiply them to have them all come together — you end up with something that has very long odds indeed, less than a 0.1% probability of being able to satisfy all these criteria. This is why you're not seeing a lot of this done. This is actually very difficult to do.
Slide 15: Fasten Your Seatbelts...
All right. We're going to launch into the data that we have for a particular treatment, so fasten your seatbelts.
Slide 16: Cardiac Adrenergic Receptors
I'm going to be talking about two polymorphisms within the adrenergic neuroeffector signaling pathway in the heart, and they are shown in cartoon form here.
The first is a polymorphism in the alpha-2C receptor. The alpha-2C receptor is a receptor that is present pre-junctionally on adrenergic neurons. It inhibits the release of norepinephrine in the heart and in other areas.
The other polymorphism is in the beta-1 receptor. This is the business end of the adrenergic pathway in the heart. This is the dominant receptor. This is the receptor that's responsible primarily for support of the heart. It's also responsible for progression of cardiomyopathy as adrenergic drive is used to support the failing heart excessively and chronically.
It turns out that the polymorphism in the alpha-2C receptor that we are interested in is a deletion polymorphism. We'll go into more detail in a minute. It's present in 16% of the population. Eighty-four percent of the population has the wild type form of that receptor. This deletion polymorphism, as we'll show you, affects the release of norepinephrine and affects the interaction of norepinephrine release with a therapeutic agent.
The beta-1 receptor, shown down here, comes in two flavors. One is a high-functioning flavor: the arginine flavor at codon 389. When there are two alleles of this - the homozygous state of this receptor - present, the function of that receptor is quite high, much higher than if any glycine is present, either heterozygously at 389 or in the homozygous state. So both of those together are a carrier state for the glycine at that position, and this is distributed roughly 50:50 in the population. So both of these, sort of, fit the prevalence criteria, if you will, for being of interest therapeutically.
Slide 17: Adrenergic Receptor Polymorphisms
Here is further definition of these two polymorphisms, starting with the 389 beta-1 receptor polymorphism. This is an important area of the beta-1 receptor functionally. It's down towards the carboxy terminus, but it's in an area of the receptor that interacts with G proteins.
This polymorphism actually is maldistributed or non-evenly distributed by race. The high-functioning form of this receptor, the allele frequency, is less in blacks than in nonblacks. So nonblacks have a higher frequency, 72 to 57 percent, of the high functioning form of this receptor, which will be important a little bit later.
Here is the alpha-2C deletion polymorphism. This is not a SNP. This is a SNP here. This is not a SNP; this is a four-amino acid deletion, which ordinarily from the standpoint of what happens to a sequence would be considered a mutation. It's a big deal sledgehammer effect on this protein. But because it's so prevalent in the population it's considered a polymorphism. When this occurs, when this four-amino acid deletion occurs, it basically destroys the function of this receptor.
Now, this is really, really distributed with race, and the allele frequency of this deletion polymorphism is ten times higher in blacks than nonblacks. It's almost a racial marker, if you will.
So, just as a preview of what's to come here: When you have the high functioning form of the beta-1 receptor, there's a nearly fourfold greater efficacy of the beta-blocking agent bucindolol than if you have a glycine at this position. So the arginine form of this receptor essentially increases the response rate or efficacy.
When you have the deletion polymorphism here in the alpha-2C receptor, there's an increase in adverse events which amounts to an increase in mortality in the presence of bucindolol, a 1.6-fold increase in mortality compared to wild type.
So, you have to put these two polymorphisms together essentially to target this drug, and we'll talk about how that's done.
Slide 18: Gly389 in Humans
We're going to talk first about the 389 polymorphism, the Arg/Gly389. Human is the only species actually that has significant numbers of individuals with a glycine at this position. All sequenced other species have an arginine, the high-functioning form of this protein. So this is a highly-conserved region, and this is very unusual, only humans having this glycine variation.
Slide 19: Arg vs. Gly Functional Effects
This polymorphism makes a huge difference in the function of the beta-1 receptor. This was originally described by Steve Liggett. We knew about this in 1998 based on collaborations with his laboratory, and this is the publication that first defined this in 1999.
These are recombinant receptors, Arg versus Gly, cyclic AMP or adenylate cyclase activity measured with the Arg homozygous or the Gly homozygous transfected into cells. You can see that there is a much greater response to the beta agonist isoproterenol in the arginine form of this receptor. It's about a four-fold difference in response no matter how you normalize the data.
Slide 20: Allele-Specific Features of Arg389 and Gly389
Liggett went on to express the Arg versus the Gly in transgenic mice using a cardiac specific promoter, alpha myosin heavy chain. He basically found that when you express the arginine, the higher functioning receptor, you end up getting the greater amount of cardiomyopathy shown here. You still get some cardiomyopathy with the Gly. We had done this previously with the Gly, and you do get cardiomyopathy. But when you do the Arg versus the Gly at the same level of protein expression in the heart, it's a much more vigorous cardiomyopathy related essentially to the much more active signaling occurring through that receptor, which if done chronically proves to be adverse to the myocardium in the phenotype.
Slide 21: Effect of Gly
In collaboration with Steve, we went on to look at this, beginning in the late 90s, in isolated human heart. And these are data we recently published with Steve in that regard. These receptors are very different in the isolated human heart as well, and so this is non-failing heart. These are hearts from organ donors, removed for organ donation, that can't be used for transplant. Here's the Arg homozygous versus Gly carrier state, and there's a big difference in responsive obviously to isoproterenol.
In the failing heart the difference is maintained, although it may be slightly reduced, mitigated somewhat. There is still a much greater response to the arginine homozygous. In fact, the arginine homozygous in the failing heart has a greater total maximum response here than the glycine in nonfailing heart.
And there appears to be a gene dose effect of the glycine if you view this by genotype as opposed to Gly carrier state. So here's arginine homozygous in the failing hearts, here's the heterozygous state, and here's glycine homozygous, and there's progressive right shift in sensitivity to isoproterenol.
This is something that's not just seen in recombinant expressed receptors. This is actually seen in the human heart.
Slide 22: Anti-Adrenergic Agents/Treatments with Phase II or III HF CT Data, or in Development
Here is the drug that interacts with this polymorphism. This is bucindolol. This is another drug the phase II work of which was all done here at the University of Utah by Mike Gilbert and in our laboratory over on the fourth floor of the hospital in the cardiology space.
This is a drug that is very similar in molecular fit and some other properties to carvedilol. It's a beta-blocker vasodilator, but it has some unique properties compared to carvedilol as well. One of them is — and all these are all beta-blockers, looking at various pharmacologic properties here — all of them block beta-1 receptors. That's the class effect of a beta-blocker. Some of them also block beta-2s and alpha-1s and affect beta-3 receptors and so forth, but bucindolol is unique in that it produces norepinephrine lowering. So it not only blocks the receptor, it lowers norepinephrine much as does moxonidine or clonidine shown here.
This is a sympatholytic effect unique to bucindolol in terms of having something that's easily demonstrable. Something that we knew about here based on our phase II work; we didn't pay that much attention to it. It turns out this was very important as we went to phase III, and we'll get back to why that was important and show you that it was important.
So that's a unique property. Another unique property relevant to this discussion and the interaction with the beta-1 389 polymorphism is inverse agonism, the ability to inactivate active-state receptor, particularly for the Arg-Arg present with bucindolol possibly, although it's controversial; present for carvedilol to a lesser extent; and not present for other beta-blockers, at least the ones that we've looked at. And we'll show you that data right now.
Slide 23: Responses to Bucindolol, Carvedilol, and Xamoterol
This is isolated human heart again, so it's not just agonists that produce a differential response in the arginine versus glycine version of this receptor. Antagonists also can produce a differential response.
This is isolated human heart, response to various antagonists or an antagonist with an intrinsic sympathomimetic activity, xamoterol, also known as a partial agonist. Let's start with that.
This was used as a positive control to see if we could pick up weak partial agonist effects of these ligands in this system. We were interested in the differential effect by genotype. So this is xamoterol. This is increase in systolic tension just like with isoproterenol in the previous slide. And the arginine homozygous in the solid lines here has a much more consistent positive inotropic response which is statistically significant, in other words has a positive slope compared to a slope of zero, whereas the glycine has a very inconsistent response, and it's not statistically significant compared to a slope of zero.
The arginine homozygous as you might expect basically sets up response to this partial agonist. You are able to see that response in the higher functioning receptor. That would make sense.
Here's bucindolol, which in some systems including rat, has some partial agonist properties, and the issue was, does this have partial agonist properties in human? The answer is no. But here's what happens by genotype.
So here's the Arg homozygous with bucindolol, and it has an inverse agonist effect. It has a negative inotropic effect, and this based on model system work we know is due to inactivation of active state receptors, whereas the glycine is not affected at all. It's a neutral antagonist for the glycine carrier preparations, but it's an inverse agonist for the arginine homozygous.
Here's carvedilol, which does not exhibit evidence of an inverse agonist effect in the arginine homozygous in isolated human heart. There's a recent paper in the JCI suggesting it does have this in a model preparation using fluorescent resonance energy transfer (FRET), but at least in isolated human heart it does not appear to have that.
Slide 24: Arg vs. Gly Expressed in Fibroblasts
This is sort of how this works pharmacologically, now bridging to therapeutic response. This is an experiment done in Liggett's lab where beta-1 389 Arg versus Gly are expressed in fibroblasts.
The first point on this competition curve is the response to 5-micromolar norepinephrine, and of course it's much greater for the arginine version of this receptor compared to the Gly, again around a four-fold difference in maximal efficacy in response to the beta agonist norepinephrine.
Then bucindolol is added in increasing doses of the competition curve, and not surprisingly in both cases there is complete antagonism or competition essentially of the agonist effect on norepinephrine, getting to more or less the same level. So the affinity of bucindolol is not different for the arginine versus glycine, the IC50 of this and the Ki calculated from that is exactly the same. But there is a greater amount of absolute inhibition just by virtue of the fact that the curve starts at a higher level. So this is the absolute amount of inhibition possible in this signal transduction system which essentially would translate, one would predict, into a therapeutic potential.
There'd be more therapeutic potential going from here to here than from here to here. That was the hypothesis that we generated to put into this clinical trial, and that hypothesis, as you will see, was confirmed by the data in that trial.
Slide 25: BEST Trial Hazard Ratios for Mortality by Subgroups
In this slide we see the trial. University of Utah was heavily involved in this as we were in Colorado. This was the first. This is the Beta-Blocker Evaluation of Survival Trial (BEST). It was a public/private partnership engineered by some of us with a partnership among the NHLBI, the VA Cooperative Studies Program, and a corporate sponsor, Incara (Pharmaceuticals). It also had in it the world's (at least the heart failure world's) first DNA bank, which was an incredibly big pain in the ass to set up in 1995, but we did that.
And so, this is the results in the entire cohort, and these are just point estimates of the hazard ratios by subgroups. And here's the entire cohort.
The primary endpoint of this trial was all cause mortality, which is what is shown here as an endpoint, and there ended up being an 11% — not 10% as published in the New England Journal because of a rounding error — but 11% reduction in mortality with a p-value of .10, not quite significant.
The trial was stopped early for loss of investigator equipoise. As two other trials starting after BEST but finishing earlier had reported positive, investigators basically said we can't continue these patients on placebo. We have to stop this trial and will stop with 87% of the total amount of information having been obtained and the p-value not quite statistically significant.
This is what happened among subgroups, and you can see just by pattern recognition that there are two subgroups that are on the wrong side of the line here. In other words, they have an increase in mortality in the bucindolol treated patients versus placebo (although it overlaps 1.0, it is not significant) the very advanced heart failure patients who have only been studied in this trial (no other trial looked at very advanced heart failure class IV patients) and then blacks.
Because BEST used NHLBI money — it was the only beta-blocker phase III trial that was sponsored by the NIH — of course, Building One demands that you enroll minorities, and we did a very good job of that. We enrolled 23% minorities.
Here's what happened in Blacks versus nonblacks. This is not significant either, but it's a 17% increase in mortality, and the test for interaction against nonblacks is statistically significant. So this clearly catches your eye, and you might remember that I pointed out that there is a racial difference in the distribution of two of these adrenergic receptor polymorphisms. It turns out that this is in someway or partially explained by these receptor polymorphisms.
So the hypothesis in the adrenergic receptor substudy of Liggett's, Bill Abraham's, and mine in this out of the DNA bank was that the beta-1 389 homozygous patients would respond better to bucindolol. That substudy was submitted in 1998 before the trial ended. Eventually the DNA found its way to Liggett's lab. We saw the data early in 2003, and these are the data for all-cause mortality.
Slide 26: Effects of 389 Polymorphism on Mortality in BEST
These are Kaplan-Meier curves of survival, and you can see that three of them are sort of superimposed without any obvious difference, and there's one that stands out. The one that stands out is the arginine homozygous bucindolol cohort with a reduction in mortality of 38%, which is statistically significant compared to its placebo group.
The glycine has a reduction of 10% that's not statistically significant, and there's no difference in the placebo groups between the Arg and the Gly. In other words, there is no effect on natural history here in terms of mortality within the timeframe of this trial, but there is an interaction with treatment effect.
Slide 27: Best: Hazard Ratios for Clinical Endpoints
It turns out that all clinical endpoints that are meaningful have a favorable interaction with the arginine variant at 389, and this is just some of the data.
I just showed you the mortality data, which are shown here. So this is the 38% versus the 10% point estimate of the hazard ratio for mortality. Here's mortality in heart failure hospitalization, which is the typical endpoint now used in heart failure clinical trials and will be used for the approval of bucindolol. Highly significant in the entire cohort, but showing a pharmacogenetic enhancement by the beta-1 polymorphism with nearly a 40% reduction in the Arg and less in the Gly. Again, not positive test for interaction, but clearly a greater therapeutic response.
Slide 28: Alpha-2C Deletion Mutation
Here's heart failure hospitalization alone censoring mortality. Again enhancement.
I'm not going to show you this, but the greatest effect of pharmacogenetic enhancement is length of stay in heart failure hospitalized patients, which goes from 7.5 days in the entire DNA substudy down to 3 days in the arginine homozygous patients, over a 50% reduction with several zeros and a one on the p-value. So clearly the patients that are arginine homozygous respond better to bucindolol.
All right. Now on to the alpha-2C. I've shown you this already. This is the four-amino acid deletion. This is the third intracytoplasmic loop of this receptor protein, and this is an area known to be important for G protein coupling.
Again, this has a racial predilection, a roughly tenfold difference in allele frequency, higher in blacks compared to nonblacks.
Slide 29: Functional Effect of Alpha-2C Deletion Mutation
This is the function of this receptor, wild type versus the alpha-2C deletion in Liggett's lab expressed as ability to inhibit adenylate cyclase in response to epinephrine. You can see there's a huge difference in response. The alpha-2C deletion virtually destroys the function of this receptor.
Slide 30: Change in Systemic NE at 3 months by Bucindolol or Moxonidine vs. Placebo in BEST Or MOXCON, Compared to Mortality
This is the interaction of bucindolol with norepinephrine, the effect of bucindolol on norepinephrine and then the interaction with clinical endpoints represented here by mortality. So again, as I said earlier, we knew about bucindolol's reduction in norepinephrine. We didn't think that much of it. We thought it was just another thing that provided an antiadrenergic profile to the drug. We weren't completely sure that it was unique to bucindolol, although it turns out it was. And we weren't sure how active or how substantial this result was until we did the BEST trial.
Here are the data from the BEST trial. This is change in norepinephrine at three months in the placebo group and the bucindolol group. In contrast I've graphed out the change in norepinephrine in another clinical trial that used a pure sympatholytic agent, moxonidine.
So the reduction in norepinephrine in the bucindolol group in BEST at three months compared to placebo was by 19%. In MOXCON the reduction was by 23%.
Here's what happened with mortality in these two trials. Here's the 11% reduction in BEST by bucindolol. Not significant.
And the moxonidine trial was stopped after nearly 2,000 patients were enrolled because of a dramatic increase in mortality apparently related to the norepinephrine reduction, over a 60% increase in mortality.
And then realizing that this might be an issue, we went on to look at what happened within BEST in patients that had a very large reduction in norepinephrine versus a less large or even an increase in norepinephrine. Long story short, in the patients that had a very large reduction in norepinephrine, which was 18% of the cohort defined as (we did this by computer modeling) 223 pg/mL, there was a 70% increase in mortality.
Lowering norepinephrine by a very large amount is not a good thing in a heart failure patient. You basically lose adrenergic support to the heart. This essentially increases mortality, and most of that increase in mortality is sudden death. This is not a good thing. And it turns out this is under genetic control via this alpha-2C polymorphism.
Slide 31: Alpha-2c-AR Genotype, Systemic NE Response at 3 Months and Survival by Treatment Group in BEST
Here is the alpha-2C polymorphism. The deletion carriers are in the hatch bars, and solid bars are the wild type homozygous form of the alpha-2C.
And here's what happened in norepinephrine at three months in a DNA substudy in patients on placebo versus bucindolol. So on placebo there's a small decrease in norepinephrine in the wild type alpha-2C that's statistically significant compared to placebo, but in the alpha-2C deletion carriers there's a huge reduction in mortality. It's around 200 pg/mL; getting into the danger range, if you will. It's statistically significantly greater than in the wild type. Greater, of course, than placebo. And this basically sets up the patients for an increase in mortality.
When you look at mortality in these patients you see a 30% reduction in the patients with the alpha-2C wild type and a 9% increase in the patients carrying the alpha-2C deletion polymorphism.
So you can eliminate essentially at least most of the patients with a big sympatholytic potential simply by genetic testing with this drug.
Slide 32: Effects of Genetics in BEST
This is a summary slide kind of marching through all of this, and that is—this is the entire cohort data of BEST in all 2,708 patients. The 11% reduction.
If you eliminate class IV in blacks, and these are the patients studied in all other heart failure beta-blocker trials, there is a 25% reduction in mortality, highly statistically significant.
If you go to the alpha-2C wild types that I've just shown you alone, it's 30% reduction. If you go to the beta-1 Arg389 it's a 38% reduction. If you combine the alpha-2C wild type and the beta-1 Arg-Arg, it's a 40% reduction, and there's enhancement of other endpoints even greater than that.
If you go to the glycine beta-1 carriers and the alpha-2C deletion carriers, the so-called unfavorable diplotype, you end up with this, and that's a 35% increase in mortality.
So by simple genetic testing you can select out patients with a better probability of response and a high probability of a serous adverse event. And the difference between the very favorable diplotype here and the unfavorable diplotype here in terms of mortality is a 2.25-fold difference in hazard ratio.
Slide 33: Regulatory Strategy and Status
All right. So what's going on with this? We've licensed this drug to this "newco" that we have, and we're working with the FDA basically to get this drug approved. I'm not going to go into the details of this, but the new drug application (NDA) will be submitted later this year. It will go to panel. We're hopeful that this will be approved and will be available within a couple of years combined with a genetic test that we are developing as well.
Slide 34: DeCode
So there is other activity in this sector.
Here's DeCode, that I'm sure many of you are familiar with. This is the pharmacogenetics or molecular genetics company that has genotyped the entire country of Iceland, and that has led to some interesting things, including identifying a haplotype of the 5-lipoxygenase activating protein (FLAP) associated with an increased MI rate in Iceland as well as in U.S./European patients.
They basically then in-licensed a drug from Bayer that is a FLAP inhibitor. The haplotype that they have with FLAP increases activity of this and increases inflammation in arteries if you will, including the coronary artery, to make things simple. And, they in-licensed a drug that had been abandoned as an antiasthmatic agent from Bayer that actually is a FLAP inhibitor. The news releases of this are shown here.
Slide 35: FLAP Inhibitor Trials
They then did a biomarker phase II study published in JAMA, shown here and basically showed they could take down these biomarkers of FLAP activity very nicely, and that was big news. They then went into phase III. That phase III trial began in May of '06. It was put on hold in October of this year because of a "formulation problem". Hopefully, this trial will get back on course once they solve their formulation issue.
Just parenthetically, this is some of the risk of small pharma or biotechs doing drug development. It's inconceivable that a big pharma would end up with a formulation problem in phase III. These issues should be well sorted out before you hit phase III. Nevertheless, that's another sort of pharmacogenetic targeting that's, at least in theory, in phase III.
Slide 36: Genetic Determinants of Response to Drug Therapies for Heart Failure
This is the next to last slide. So, smart trial design based on polymorphic variation in genes that modify drug response is the key to the future of drug development. We've been saying this for years. There are two ways in which this sentence can end. "... and it always will be the future. We'll never get there." That's kind of the way it's been looking.
The other way this can end is "... the future is now.", and I'm arguing here that the future is now.
Slide 37: Timing Is Everything
But timing is everything. Just look at this for a second. This is the USC cheerleader as Texas scores the winning touchdown in the Rose Bowl [laughter]. Yeah. So timing is everything, and hopefully the timing has converged here to make this possible in the cardiovascular space.
Thank you very much.
|
|
|
