The success and fall of blockbuster drugs
offers insights into the long road toward personalized medicine.
| Sep
1, 2005 |
| By:
Lukas
K. Buehler |
| Pharmaceutical
Discovery |
|
Lukas K. Buehler
|
Are drugs safe? This is an obvious question given that widely popular
drugs are being taken off the market at a steady rate. Some of these drugs
are highly recognizable products — since they are so heavily advertised
— making problems all the more obvious and relevant for a large number
of people. Vioxx (Merck & Co. Inc., Whitehouse Station, New Jersey,
USA), an anti-inflammatory; Baycol (Bayer AG, Leverkusen, Germany), a
cholesterol lowering drug; Rezulin (Parke-Davis/Warner-Lambert; now
Pfizer, New York, New York, USA), an anti-diabetic and Fen-Phen (Wyeth-Ayerst,
Madison, New Jersey, USA), appetite suppressants, all have been taken off
the market after years of enjoying a striking increase in amount and
duration of use. What these drugs have in common is their extreme
popularity, which may have contributed to the problem. They were being
used by a very large number of people and over increasingly longer time
periods to treat chronic problems, or even as lifestyle choice (Fen-Phen
was used for weight loss, most likely by people who were merely overweight
rather than morbidly obese). Rare but severe toxicity problems injured and
killed people, though these were problems that were found to affect fewer
than 1 in 1000 individuals. Clinical trials are not just designed to catch
such rare events.
This column is not the place to discuss risk assessment. The key
question rather is this: how can the pharmaceutical industry avoid
developing drugs for chronic ailments like arthritis and obesity that
within a few years of use turn out to be unnecessarily dangerous? In
addition to negotiating what constitutes acceptable risk, the obvious
solution to readers of this column is to reduce risk by finding ways of
anticipating complications before a new molecular entity is brought into
clinical trial stages. In other words, solve the problem during the early
stages of drug design.
|
Predictive toxicology and toxicogenomics will be leading the research
efforts in this area, assessing risk based on genetic predisposition.
Expectations for both new sciences fly high based on the progress being
made in molecular modeling, bioinformatics and proteomics. Predictive
toxicology goes beyond the familiar turf of understanding a single
structure–function relationship of a drug molecule. The goal is to know
in advance who will benefit from a drug, who is affected by side effects
and who isn't. That is to say, treat the problem, not from the drug's
point of view but from the patient's point of view.
Speaking of patients, I think the current debates on the safety of
blockbuster drugs are particularly useful not only to improve our
understanding of personalized medicine but also to better inform the
public about a very real and relevant aspect of science: how drugs are
made and what they can and cannot do. Let's take a closer look at the
Cox-2 inhibitors, the recent withdrawal of Vioxx and why we just might
have reached the limits of rational drug design. Vioxx was approved by the
FDA in 1999 as a non-steroidal anti-inflammatory drug (NSAID) for the
relief of the signs and symptoms of osteoarthritis, the management of
acute pain in adults and the treatment of menstrual symptoms; it was later
approved for the relief of the signs and symptoms of rheumatoid arthritis
in adults and as a painkiller for children. Vioxx is a member of a
chemical class of anti-inflammatory drugs that lessen gastrointestinal
morbidity and mortality symptoms (e.g., ulcers, bleeding) often associated
with the non-specific NSAIDs ibuprofen, naproxen and aspirin. Turns out,
we may have traded one problem for another one. In September 2004, Vioxx
was voluntarily pulled off the market because of a demonstrated increased
risk in heart disease when administered at high dosage and for more than
18 months (clinical trials for FDA approval usually last no more than 11
months).
Cox-2 inhibitors are without a doubt
marvels of modern biotechnology. They stem from our understanding of the
biological mechanism of NSAIDs and provide a stunning example of the
successful application of rational drug design principles. NSAIDs target
the enzyme cyclooxygenase (Cox) that converts omega-3 and omega-6 fatty
acids into prostaglandins and thromboxans. This enzyme comes in five
variants: Cox-1 and Cox-1b (Cox-3) are two constitutive enzymes important
for controlling fluid balance and platelet aggregation. In addition, our
body has three Cox-2 isoforms, two of which are inducible (1). The
inducible isoforms produce the inflammatory prostaglandin PGE2 involved in
pathophysiological reflexes that cause fever and pain — mechanisms that
alert the brain to imminent tissue damage due to infection and injury. The
steroidal anti-inflammatory cortisol interferes with the pain pathway by
inhibiting the gene expression of Cox-2 (2).
Traditional NSAIDs like aspirin
(introduced in 1897) inhibit all cyclooxygenase isoforms. In particular,
they have been shown to suppress thromboxan A2 formation via Cox-1
inhibition, thereby contributing to gastrointestinal bleeding in some
patients. Instead of trying to understand why some but not all patients
are afflicted by gastrointestinal side effects, the problem was solved by
bypassing Cox-1 altogether — thus the search for Cox-2 specific
inhibitors. In the early 90s, only the constitutive Cox-1 and one
inducible Cox-2 isoform were known. Armed with the high-resolution
structure of these cyclooxygenase proteins, it became clear that
increasing the size of the inhibitor would prevent them from binding to
Cox-1, which unlike Cox-2 has a hydrophobic amino acid in the binding
site. Thus, the bulkier Cox-2 inhibitors are prevented from binding to the
active site of Cox-1. This is rational drug design at its best. The next
step involved modifying functional groups around the pharmacophore of the
new molecular entity (3) with the goal of improving Cox-2 specificity and
strength of binding to allow the administration of lower dosages. This
resulted in several new Cox-2 inhibitors being put on the market. However,
this step did not address the molecule's pharmacokinetics (i.e., how the
drug behaves systemically). While all Cox-2 inhibitors do what all NSAIDs
are doing — relieve pain — there remained differences in drug-specific
risk profiles. For example, while Vioxx was pulled from the market,
related Cox-2 inhibitors were found to have less pronounced side effects.
This raises two questions that I'd
like to address briefly. First, how can one out of several drugs in the
same class targeting the same enzyme be more toxic than the others? Why
Vioxx and not Celebrex (Pfizer)? Second, could it be possible to foresee
cardiovascular risk due to prolonged use of Cox-2 inhibitors — or of any
drug, for that matter? Answering these questions brings us back to a
familiar topic — the complexity of biological systems.
It's easy to invoke complexity in
biology, of course, and one might be tempted to use it as a proxy to sound
smart when in fact we do not understand actual mechanisms — or worse we
don't even know what kind of questions to ask. Still, the complexity is
real and I want to take the opportunity to chart basic physiological
principles that, in my view, foil any attempt of accurately predicting the
action of an untested drug, if the only thing we know about it is its
molecular structure (4). For instance, a small change in a ligand
structure can make an enormous difference in its physiology. Consider the
cholesterol-derived hormones estradiol and cortisol — one a female sex
hormone, the other a stress hormone that regulates energy metabolism,
blood pressure and cardiovascular function. In hindsight, knowing that
cortisol acts by reducing Cox-2 levels, one might suspect that Cox-2
inhibitors affect cardiovascular function as well — the apparent problem
for Vioxx.
There is nothing new or unusual,
really, in these observations. What should be surprising to researchers in
drug development is how much we still depend on trial and error stages –
from high-throughput screening to animal tests and clinical trials. Part
of the problem may simply be a lack of knowing all the physiological
details. Cox-2, for instance, comes in an unforeseen constitutive isoform
with wide tissue distribution, in addition to its inducible response
during inflammation, as originally found. This constitutive use of an
enzyme isoform likely explains the increased risk for heart disease now
associated with Vioxx. But this does not explain why Vioxx increases the
cardiovascular risk more than other Cox inhibitors. True, Vioxx may bind
more strongly to the target or bind to other proteins because of a
structural difference (e.g., carrying a methylsulfonyl group
instead of the more common sulfonamide [as in Celebrex], a big difference
in the local charge). A solution might come from structural genomics
through ab initio prediction of ligand binding sites from DNA
sequences. However, protein structure prediction at this detailed level is
not always reliable.
I strongly believe that not lack of
facts but the inherent hierarchical organization of complex systems is the
real roadblock preventing accurate predictions of higher-level functions
(here, for how an organ responds to a drug) from molecular structures
alone (5). No matter what the structure, any particular drug has the
following biological rules that always apply. To keep it simple, the use
of the term "drug" refers to any ligand, natural or synthetic,
including physiological metabolites and signaling molecules. The term
"target" refers to any protein, be they enzymes, transporters,
structural proteins or receptors.
1. Know your competitors:
several drugs exist for the same target (competitive and non-competitive;
agonist and antagonist).
2. Know your target: a single
drug has multiple targets that play unique roles in different
physiological contexts; for instance, the hormone epinephrin either
constricts or dilates blood vessels, depending on the adrenergic receptor
type found in the smooth muscle cells lining the vessel.
3. Know your target's
distribution: a target may be found in just one or several
tissues/organs, contributing to one or more (patho-) physiologies. For
instance, gap junction protein Cx26 is expressed in both the inner ear and
liver; a mutant Cx26 causes deafness (6), but has no apparent effect on
liver function.
4. Know your pathways: a
target may be part of different metabolic or signal transduction pathways
interacting with different proteins. In the inner ear, Cx26 is paired with
Cx30, but in the liver with Cx32, possibly explaining that
protein–protein interactions play a crucial role in determining the (dys)function
of the Cx26 mutant in the inner ear.
5. Know your drug interactions:
situations 1 through 4 explain why drugs interact with other drugs in
unforeseen ways.
6. Know your genes: allelic
variants (mutations) of targets may alter an individual's response to a
drug along situations 1 through 5.
The list above is not meant to be
all-inclusive, and many subtle effects — some still to be discovered —
play crucial roles in physiological processes. For instance, a target may
be involved in more than one cellular function (dual-function proteins),
as demonstrated by a recent publication exploring the link between the
Krebs cycle enzyme aconitase and mitochondrial DNA maintenance (7).
Research into the controls of signal transduction pathways by metabolic
enzymes (rather than the other way around) is an exciting new development
in biochemistry and will change the way we think about the role of
metabolic enzymes in general.
The point here is that a large number
of endocrine and nervous signaling mechanisms are involved in maintaining
a healthy body and may do so differently in different people due to
allelic variations. In the short term, lowering risk based on a patient's
genetic background can contribute to better use of many drugs, improve
their safety profile and strengthen public acceptance of the drug approval
process.
Having said all this, the problems
facing Cox-2 inhibitors is not acute toxicity but long-term effects. Their
advertised purpose, of course, is to treat chronic pain, and therein lays
an entire new challenge for predictive toxicology. Unlike interventive
medicine (e.g., using antibiotics to fight a bacterial infection or
chemotherapy to destroy tumors), our body responds to persisting
conditions by adjusting its internal states. Weight gain, insulin
resistance and drug tolerance all are examples of such adjustments. I
shall come back to this issue in later columns.
Lukas K. Buehler is an adjunct
professor at Southwestern College in Chula Vista, California USA, and the
founder of SciScript Inc., in San Diego, California of San Diego's
Extension Bioscience Program. He can be reached at lbuehler@ucsd.edu
References
1. T.D. Warner and J.A. Mitchell, PNAS
99(21), 13371–13373 (2002).
2. M.T. Rae et al., J. Clin.
Endocrinol. Metab.89(9), 4538–4544 (2004).
3. A. Palomer et al., J. Med.
Chem.45(7), 1402–1411 (2002).
4. L.K. Buehler, Pharma DD 3(5),
20–21 (2003).
5. L.K. Buehler, Pharma DD 4(6),
24–26 (2004).
6. D.P. Kelsell et al., Nature 387(6628),
80–83 (1997).
7. X.J. Chen et al., Science 307(5710),
714–717 (2005).
|
