This page uses the Jmol java applet. Be sure your browser is java-enabled. Be patient--it may take several seconds for the structure to load! After it loads, you can drag the structure to rotate it or shift-drag to zoom. If you have a 2-button or 3-button mouse, the right-button may give you more display options.
In the beginning, plants and other natural products were the source of most medicinal substances. For example, the foxglove plant can be used to treat congestive heart failure. As the science of medicinal chemistry evolved, it was discovered that foxglove contains digitalis and other cardiotonic glycosides which are responsible for the therapeutic effect. It became possible to isolate the active components, so that dosage could be more accurately regulated.
Other medicines came about because of accidental observations and discoveries. Penicillin is one good example. Alexander Fleming noticed that some of his bacterial cultures were ruined because of contamination with mold. Wherever the mold grew, the bacteria were apparently killed. It was found that the mold (Penicillium) produces a substance called penicillin, which kills many kinds of bacteria.
The discovery of penicillin led to a large-scale screening effort, in which thousands of soil microorganisms were grown and tested to see whether they could produce other substances which kill bacteria. Antibiotics such as streptomycin, neomycin, gentamicin, erythromycin, and the tetracyclines resulted from these efforts. Chemical modification of known drugs can often lead to improved drugs. Naturally occurring penicillin G (below left) is broken down by bacterial beta-lactamases. Addition of two -OCH3 groups produces methicillin (below right), which is resistant to lactamase.
Another example of chemical modification: opiate analgesics. The parent compound is morphine, which occurs in opium poppies. Morphine is a powerful analgesic, but it has serious side effects: respiratory depression, constipation, and dependence liability. Thousands of analogs (related chemical structures) have been synthesized in an effort to find analgesics with lower incidence of side effects. In some cases, small changes in chemical structure may have a big influence on the activity: nalorphine is a partial agonist (shows some morphine-like activity, and at higher concentration, antagonizes morphine effects), and naloxone is an antagonist.
Considerable simplification of the molecule is possible. Meperidine has only two rings instead of four, but it maintains strong analgesic activity. It has better oral absorption than morphine, and shows less GI side effects. Methadone is an analgesic in which the original piperidine ring (6-membered ring containing a nitrogen atom) is completely absent. It retains analgesic activity, has good oral activity, and has a long half-life in the body. Dextromethorphan is constructed on a mirror-image of the morphine ring system. It has no opiate analgesic effects or side effects, but is a useful anti-tussive agent.
Quinine is a useful antimalarial drug, but during times of war it has not always been accessible. Structural modification has led to synthetic analogs such as chloroquine:
For a pharmaceutical company, finding a new blockbuster drug (>$1 billion/year sales) is a little like winning the lottery. And not just with respect to the big money. In order to be successful, a new drug must satisfy several criteria:
If you don't match all six, you don't win big. Let's see why.
Safe.
All drugs have side effects. You would like to have a large therapeutic index, the ratio between the toxic dose and the therapeutic dose. Start with acute toxicity tests in mice. Check for mutagenicity and teratogenicity. Chronic effects over several generations. Safety tests in several species of animals before you try it out on humans. Careful monitoring of adverse reactions in all phases of clinical trials. In the U.S., there are three phases of clinical trials. In Phase I, the compound is tested in healthy volunteers at a range of doses to establish that there are no immediate adverse effects. In Phase II, the compound is tested in a small number of patients in order to establish that there is an effect and to find what dosage may be required to achieve this effect. In Phase III, large-scale double-blind trials are carried out. Usually the compound is compared to an established drug or to placebo. This allows an objective measure of the drug's effectiveness. And after approval, there is continued monitoring for adverse effects. This is sometimes refered to as "Phase IV." Your drug can get stalled for more testing, or recalled after approval.
Effective.
This is the obvious one. But it's not cheap. You have to do some sort of screening to find active compounds. You have to do some in vivo testing in animals to show that it works. You have to do lots of animal studies before you can try it out in humans. Then you have to go through Phase I, Phase II, and Phase III clinical trials, send your paperwork to FDA, and wait for approval. Each stage gets progressively more expensive. Your drug candidate can fail to move along at any of these stages.
Stable.
Chemical stability is necessary. Aspirin breaks down over a period of time when exposed to humidity (sniff an old bottle to see if you can detect the acetic acid [vinegar] odor). You need to do stability testing on your dosage form so that you can put a reasonable expiration date on it. Chemical stability also extends to the acid environment of the stomach for oral medications. Metabolic stability is also important. Will your drug survive the various metabolic enzymes of the GI tract? If not, you may need to use a parenteral dosage form. Will your drug be rapidly destroyed in the liver? Will it induce liver enzymes which affect the breakdown of other drugs, causing drug interactions?
Soluble.
How water-soluble is your drug? You will need at least some water-solubility in order to get absorption and distribution of your drug throughout the body. But if it's too water-soluble, it will rapidly leave via the kidneys. Lipid-solubility is also important. If the lipid-solubility is very low, the drug will not be absorbed well across lipid membranes. If it is too lipid-soluble, it may partition into fat stores and not reach the intended site of action. The balance between these two properties is measured roughly by the octanol-water partition coefficient. When octanol and water are mixed, they form a two-phase system (like an oil-and-vinegar dressing). Octanol roughly approximates the hydrophobicity of membrane lipids. So shake your drug in a mixture of octanol and water, let the layers separate, and measure the amount of your drug in each layer. The ratio [conc. in octanol]/[conc. in water] is the partition coefficient, P. The logarithm of P (log P) should ideally be less than 5.
Synthetically feasible.
It's one thing for a chemist to spend six months in a lab to make 10 mg of a compound to test. But when you're going to market with a new drug, you have to be able to make it in kg quantities. This means it can't be a 14-step procedure, it can't use exotic solvents/reaction conditions, and starting materials must be readily available, cheaply, in large quantity. If it's going to cost $500 per tablet, you won't sell very many. The HIV protease inhibitor saquinavir, shown below, is on the difficult end of the spectrum, having 6 chiral centers and, therefore, 64 possible stereoisomers:
Novel.
Your drug has to be sufficiently novel that you can patent it. You need patent protection in order to justify the $100 million you will spend to get approval for your drug. And it has to be sufficiently novel that it doesn't infringe someone else's patent.
A lead is any chemical compound which shows the biological activity you are looking for. A lead is not the same as a drug--it's just a clue that you are on the right track. We can distinguish two broad tasks in drug discovery. The first is lead-finding. Here the task is to find a chemical compound which has a desired bioactivity. The second is lead-optimization, modifying your lead structure to build in all those other desirable properties (safety, solubility, stability, etc.). Lead-finding There are many ways to find lead compounds:
Lead-optimization
Once you have an active compound, you can begin to fine-tune the activity. Here is where chemical modification of the lead structure is important. Can you simplify the structure and retain activity? Do you need to increase or decrease the partition coefficient or improve solubility? Do you need to modify the structure to prevent metabolic breakdown? Do you need to decrease the molecular weight to improve absorption? Chemical modification of the lead compound may provide a drug candidate which looks good in animal models, but ultimately human testing must take place before a drug reaches the market.
The receptor is the target (usually a protein) to which a drug molecule binds in order to cause a biological effect. Emil Fischer (in 1894) proposed the analogy that a drug molecule is like a key which fits into a lock. The specificity of a drug molecule for its target is like the specificity of a key for its lock; if you make similar keys, they may also "fit" into the "lock." The term "receptor" was coined by Ehrlich in 1909 to describe the target. But it was many decades later when X-ray crystallography finally let us see the three-dimensional structure of the receptor. Eventually we also got to see crystal structures of ligands bound into the active sites of proteins, too.
X-ray crystallographic structure of HIV-1 protease. The protein is a dimer; one subunit is shown in gray, and the other black. The large open space (top center) is the active site.
X-ray crystallography requires, first, that you isolate some pure protein. Then, you must find conditions under which the protein crystallizes. You can now buy kits which let you try hundreds of crystallization conditions, to speed up the process; still, growing protein crystals is something of an art, and it doesn't always work. Membrane proteins are particularly difficult to crystallize--when you purify them, removing the membrane lipids, they usually do not fold properly. There are other limitations, too. The resolution is not high enough to see the positions of hydrogen atoms. In the early days of protein crystallography, there was concern that crystal packing might distort the protein conformation, but this is not a problem except in small local regions where protein molecules touch neighboring molecules. In fact, protein crystals have a high water content (~50%), and the conformation observed is generally about the same as in aqueous solution.
NMR spectroscopy also can provide information about the 3-D structure of proteins. Hydrogen atoms, 13C, and 15N can produce NMR signals. The spectra are quite complex since there are so many of these atoms in a protein molecule. You need large amounts of protein to get started. You usually don't get just a single structure, but a family of possible conformations. NMR is limited to fairly small proteins. But the experiments are carried out in aqueous solution, so you can manipulate properties such as pH and ion concentrations.
Nowadays it's not uncommon to identify the gene coding for a target protein; clone, express, and purify the protein; crystallize the protein; and then use the three-dimensional structure to guide the design of ligands. We will discuss one well-studied example: the HIV protease.
HIV infects cells and directs the cellular machinery to make viral proteins and RNA. Several of the proteins are synthesized in one continuous chain (polyprotein). The polyprotein is cleaved into smaller chains which can then assemble to form new virus particles. The cleavage, which takes place at specific sites on the polyprotein, is carried out by the virus' protease enzyme. If you can block the activity of the protease, you can prevent the synthesis of new virus.
In 1988 the role of the HIV protease was established, and its amino acid sequence was determined. It was shown to have considerable homology to Rous sarcoma virus protease, the structure of which had been determined crystallographically. Thus, the first model of HIV protease was produced by homology modeling. The sarcoma virus protease was used as a starting point, and side chains were modified to match those of the HIV protease. This can be done because protein tertiary structure (3D-folding) is highly conserved (even more conserved than primary structure or sequence). In 1989 the structure of HIV protease was determined by X-ray crystallography, and the homology-modeled structure was shown to be essentially correct.
Homology-modeled HIV protease (blue, PDB file 1HVP) and crystallographic structure (red, PDB file 1HSG)
There are numerous proteases throughout all living organisms: digestive enzymes and blood clotting factors are just two examples in humans. The catalytic mechanism provides a way to classify proteases into three large families. Aspartic proteases contain two aspartic acid side chains which participate in cleavage of the peptide bond. Serine proteases use a serine-histidine-aspartate triad. Metalloproteases use a metal ion (e.g., Zn ) to help catalyze proteolysis.
Some proteases are fairly non-specific, and others have very tight specificity. The figure below shows, in schematic form, how the amino acid side chains may interact with "specificity pockets" on the protease to establish the protease's specificity. S1,
In HIV protease, S3 and S3' are essentially non-existent; the other sites are specific for large hydrophobic side chains such as phenylalanine and proline. Therefore, in designing an inhibitor, there should be substituents which look something like these side chains. Another consideration in designing an inhibitor is that the inhibitor should not be hydrolyzed by the protease, so amide bonds should be avoided. In fact, in the central part of the inhibitor, it is useful to mimic the transition state or intermediate in the hydrolysis reaction:
This is because enzymes work by facilitating the development of the transition state, which is on the path from substrate to product. They do this in part by binding more favorably to the transition state than to either substrate or product.
It was known from work on renin, another aspartic protease, that replacement of the peptide C=O with a hydroxy (-OH) group makes a good transition state mimic. But renin has different specificity sites than HIV protease, so known renin inhibitors weren't very good HIV protease inhibitors. It's just as well--if HIV protease inhibitors also inhibited renin, they would have side effects involving hypotension.
Several examples of successful protease inhibitors are illustrated below. We will discuss these in class. It is not important or even useful to memorize structures. The structures are shown here to illustrate the principles involved in designing a drug molecule to fit into a receptor site. These include hydrogen bonding interactions in the active site as well as hydrophobic interactions in the specificity sites. There is one particularly important feature which is routinely observed in crystal structures of HIV protease with inhibitors bound in the active site. This is the presence of a tightly bound water molecule near the top of the active site. It forms two hydrogen bonds from water hydrogens to oxygen atoms on the inhibitor (in the natural function of the protease, it hydrogen bonds to two peptide backbone C=O oxygen atoms). In turn, the water oxygen atom forms two hydrogen bonds to peptide backbone N-H atoms of the protease.
Saquinavir interactions with HIV protease active site. Blue = important hydrogen bond interactions. Red = interactions with specificity sites of the protease.
Indinavir interactions with HIV protease active site. Blue = important hydrogen bond interactions. Red = interactions with specificity sites of the protease.
Ritonavir interactions with HIV protease active site. Blue = important hydrogen bond interactions. Red = interactions with specificity sites of the protease.
A DuPont Merck inhibitor's interactions with HIV protease active site. Blue = important hydrogen bond interactions. Red = interactions with specificity sites of the protease. Note that the water molecule has been replaced by an oxygen atom in the drug molecule! Currently available HIV protease inhibitors suffer some limitations. Viruses are notorious for their ability to develop resistance to drugs; this is why combination therapy is usually indicated. Some of the available inhibitors are too rapidly metabolized or too poorly absorbed, so that large doses must be used. This is expensive, and it increases the likelihood of adverse side effects.
Below is a 3D structure of HIV protease which was crystallized with an inhibitor bound into the active site. HIV protease .
Many receptors are not readily amenable to receptor-based drug design. For example, many important receptors are membrane-bound proteins, which are notoriously difficult to crystallize. In such cases, a lead compound or active ligand must be found, and then the structure of the ligand guides the drug design process. We will illustrate this case with the story of the discovery of sildenafil, better known under the trade name Viagra. Some of this material is adapted from Modern Drug Discovery, 1998, 1: 31-38.
The story starts in Pfizer's labs in Sandwich, England. Atrial natriuretic peptide (ANP) causes the kidneys to release sodium and increase urine flow, making it a natural diuretic. It also causes the smooth muscle cells of blood vessels to relax, which increases blood flow and lowers blood pressure. It was believed that increasing natural ANP activity would be a new treatment for high blood pressure. ANP acts by binding to receptors on the external surface of cells, which activates guanylate cyclase, which, in turn, converts guanosine triphosphate (GTP) to cyclic guanosine monophosphate (cGMP). cGMP lowers intracellular levels of calcium, which can lead to sodium release in kidney cells, or to smooth muscle relaxation in blood vessels. cGMP levels are controlled by phosphodiesterase (PDE), an enzyme which converts cGMP to GMP. The rationale for the Pfizer project, then, was to inhibit phosphodiesterase, which would lead to higher cGMP levels, which would mimic the effect of ANP. So the work that led to discovery of sildenafil arose from the search for an antihypertensive drug.
At this time (1986) it was known that there were at least 5 different PDEs (at least 9 are known now). Some of these PDEs control breakdown of cyclic AMP, an important second messenger in other types of cells. A non-selective PDE inhibitor would affect many tissues and, therefore, potentially have many side effects. A PDE inhibitor with the wrong selectivity would hit the wrong PDE and not have the desired effect. But the early assays were not very good, and it was difficult to tell if compounds were selective or not. Later assays showed that sildenafil inhibits PDE-5, which relaxes vascular smooth muscle (as hoped) and, in platelets, inhibits blood coagulation. Pfizer scientists realized that this could be a good profile for a drug to treat angina: relax vascular smooth muscle in the heart to increase blood flow, and inhibit thrombosis, which can precipitate heart attack. On the other hand, it was discovered that PDE-5 is not present in the kidney. So the research changed its focus from antihypertensives to anti-angina drugs. Pfizer scientists looked for compounds which would inhibit PDE-5 but not PDE-1 or PDE-3, and one of the compounds they discovered was sildenafil
The figure below shows the structures of cGMP and sildenafil. It is likely that the SO2 group of sildenafil mimics the phosphate in cGMP. It is possible to superimpose these two structures with quite good matching of gross shape, in such a way that the SO2 group aligns roughly with the phosphate, and the 4-piperidone rings (six-membered ring with N at one end and C=O at the other) roughly superimpose. The match isn't perfect. An inhibitor doesn't have to bind in exactly the same way as substrate. It just has to bind tightly enough so that it inhibits substrate from getting in. Remember that the natural substrate should bind reasonably well, but not too tightly or it will never release; it's the transition state which should bind most tightly. This is why many drugs are designed to look like transition states (see protease inhibitors, above)
Below is a 3D superposition of the two structures:
Sildenafil didn't look very good in the first clinical trials. Safety looked fine, but cardiovascular effects were much less than with nitroglycerine. There were a few reports of side effects, including penile erections. These reports came along just when the role of nitric oxide (NO) in penile erection was being worked out. NO, like ANP, signals via cGMP. It was found that, in men with erectile dysfunction, sexual stimulation releases insufficient levels of NO from the penile nerves. As a result, cGMP was not released, and vascular muscles in the penis did not relax enough to allow blood to fill the erectile chambers, causing an erection.
Pfizer scientists use an analogy of water going into a bathtub with an open drain to explain the drug's effect, likening NO to the faucet, cGMP to water, and PDE 5 to the open drain: "In a normal patient, sexual stimulation turns the faucet on high enough that the bathtub fills quickly, for the water that drains away isn't leaving fast enough to drain the water in the tub. An impotent man can't turn the faucet on completely." So, rather than turning the impotent man's faucet wide open, sildenafil plugs the drain, allowing the tub to fill.
The timing of the discovery was fortuitous. Up until the 1970's it was generally believed that most male impotence was psychological in nature; after that time, it became increasingly clear that disease states such as heart disease and diabetes could cause impotence as well. The potential size of the market was not well understood until much later. In its first quarter on the U.S. market there were 2.9 million Viagra prescriptions, far beyond Pfizer's predictions.