ÐÓ°ÉÔ­´´

DESIGNING drugs has always been a process of trial and error. Pharmaceuticals companies used to start from a natural compound, often from a local plant that people were already using in traditional medicine. Now they are more likely to make chemicals and test them at random – a process called screening – to see if any of them have useful biological effects. Then they choose the best candidate, or ‘lead’ compound, and modify it systematically to enhance its activity and give it extra properties that will turn it into a useful drug. For example, they might make the drug more long acting, stable enough for patients to take orally, or they might eliminate harmful side effects. Pharmaceutical chemists have designed many useful drugs in this way, but for each successful drug, they have had to make and test thousands of compounds – a time-consuming business that usually costs tens of millions of pounds.

Some researchers have suggested a way of improving this hit-and-miss procedure. Their technique relies on the latest developments in biochemistry, molecular biology and genetics. Instead of starting with a random search, they describe in minute detail the target that the drug must seek out in the human body. Once they have done this, they can begin to design a drug that fits the target, by changing its structure, studying how it interacts with the target, improving it again, and studying its interaction with the target. The use of such ‘design cycles’ in the pharmaceuticals industry over the past 10 years has been an important advance.

The target for the drug is usually a large molecule, often a protein. Advances in biochemistry and molecular biology have given us a much better understanding of the chemistry and biology of such targets. Some targets turn out to be enzymes that speed up important steps in biological processes. When a drug binds to an enzyme it stops the enzyme from working, so that a particular biological process cannot continue.

For example, we now know the structure – the sequence of amino acids – of one such target. This is the protein renin, an enzyme that is important for increasing blood pressure. If we could find a compound to replace the one that binds to renin naturally – its substrate – we can stop renin from working. We may then be able to develop the compound as a drug to treat people with high blood pressure.

Other possible targets for drugs are protein molecules on the surface of particular kinds of cells. These proteins, or receptors, are designed in nature to bind hormones. Hormones are important and complex molecules that act as chemical messengers between different tissues in the body. The hormone adrenaline, for example, is secreted by the adrenal glands and travels through the bloodstream to stimulate the heart to beat faster. During the past decade, researchers have found out the sequence of amino acids of many complicated proteins that act as receptors for molecules such as adrenaline, acetylcholine and insulin.

Knowing the structure of a receptor is not enough, however. We also need to describe how the hormone recognises and interacts with its receptor, or, if the drug target is an enzyme, how the enzyme binds to its substrate. Only then can we begin to understand how a drug molecule might interfere in this process. Advances in techniques such as X-ray analysis and developments in computers, especially graphics, gave chemists the chance to make real headway (‘Computers picture the perfect drug’, New ÐÓ°ÉÔ­´´, 16 June 1988). With these techniques, chemists can describe the shapes of these complex molecules and picture them in three dimensions.

More than 25 years ago David Phillips and his colleagues at the Royal Institution in London showed that the enzyme lysozyme has a cleft, called a binding site, into which its substrate fits. Since then, researchers have found that other enzymes and receptors have such a cleft. But would knowing the structure of these binding sites enable chemists to design drugs? To be effective, the drug would have to bind specifically and tightly, in place of the natural substrates.

Here the news is not so good. Even with the best techniques of X-ray crystallography, biological chemists still know little about the three-dimensional shapes of most targets for drugs. Often, they have to infer information about the shapes of binding sites by comparing these with molecules whose shapes they know. In general, researchers know most about the structure of relatively simple enzymes such as renin, and much less about the structure of receptors for hormones and the complex protein molecules that transmit signals in the nervous system. These receptors are also often attached to membranes, which makes them difficult to isolate and almost impossible to crystallise.

But if we can define the three-dimensional structure of a drug target and create a model of how a drug interacts with it, we can try to improve the drug. For example, we might suggest changing the drug molecule so that it fits better into the target. This may involve changing its electronic charge, its oiliness, or simply its shape. We can then make a new drug molecule that incorporates all the improvements and test it to see how well it binds to the target.

This is not as simple as it sounds. We are not yet experts at designing new molecules and it makes sense for us to study the structure of the complex that forms when the new drug binds to the target. Usually, these studies reveal that there are further improvements to be made, so we go round the same cycle again: we synthesise a new molecule with the changes suggested, test its binding and study the structure of the complex again. To make a new drug we must go round the design cycle many times, hoping to make a small improvement in the molecule each time. There are several types of design cycle, depending on what process chemists use to modify the drug. For example, if the drug is a protein, the design cycle will be a protein engineering cycle. Drug designers who exploit this type of cycle use genetic engineering techniques to modify the protein.

Researchers used the protein engineering cycle to develop the hormone insulin as a drug. Nearly 50 years ago, Fred Sanger, working at the Medical Research Council’s Laboratory for Molecular Biology in Cambridge, showed that insulin is a small protein made up of 51 amino acids arranged in two chains. Twenty years ago at the University of Oxford, Dorothy Hodgkin, Guy and Eleanor Dodson, M. Vijayan and I worked out its shape using X-ray analysis. Today, many people with diabetes use insulin to lower the levels of sugar in their bloodstream. Their bodies do not produce enough of the hormone and they must inject insulin once or twice a day. In the past five years, scientists at Novo Nordisk in Denmark, working with Guy Dodson at the University of York, used protein engineering to modify insulin molecules so that they are absorbed much more quickly, a useful improvement for diabetics. Improving a drug in this way is a complex but rational process. The researchers began the process many years ago by isolating insulin from the pancreas of humans, cows, pigs, and other animals, where it is produced. They studied how these natural insulins bind to receptors in cells. Then they changed these natural insulin molecules in various ways, so that they could find out which amino acid fragments must be present in insulin before it will bind to its receptor. Since 1969 they had had an X-ray picture of the three- dimensional structure of insulin. Could they now chart the position of the essential amino acids in it? The problem was that insulin molecules do not only bind to receptors. They also stick to each other. If they stick strongly to each other, a person with diabetes absorbs the insulin much more slowly. What the scientists did, with the help of computer graphics, was to design insulin molecules that would still bind to the receptor but would not stick together as normal insulin does.

Their next task was to isolate the gene – the DNA that carries the coded instructions for making the protein insulin – and transfer it to a bacterium such as Escherichia coli. Then these genetically engineered bacteria would express the gene and make insulin. In many cases, this is done quite straightforwardly by taking the natural gene. Recently, protein engineers have been able to alter specific sites in the protein by modifying the gene that codes for it. Biochemists can then work out how the changes made to the structure of the protein affect its behaviour. This brings the operation to full cycle.

The modified protein has to go round the cycle several times to give the researchers further clues about how to improve its design. The team of academic and industrial researchers from the University of York and Novo is now moving towards testing its newly engineered insulins on diabetic patients.

Medical research is not the only area to benefit from this type of design cycle. Researchers in many companies have used protein engineering to produce enzymes for soap powders that are more stable at the temperatures of modern washing machines. Industrial chemists also use protein engineering to make enzymes that catalyse difficult steps in reactions, including those for making new drugs.

Other chemists take a somewhat different approach. They are interested in drug molecules smaller than insulin that bind to large protein receptors. The molecule may be the substrate of an enzyme; or it may be a hormone such as adrenaline. Provided chemists know enough about the three-dimensional structure of the complex formed by the drug and its protein receptor, they should be able to improve the drug by means of a receptor-based design cycle.

An example of receptor-based drug design is the inhibition of the angiotensinogen cascade. This sounds complicated but is simply a consecutive series of reactions in the body that increases blood pressure by narrowing the arteries and increasing the flow of water and salts into the bloodstream. Researchers working for the pharmaceuticals company Squibb, in New Jersey, found that they could treat people with high blood pressure if they blocked the second enzyme in the cascade. Pandi Veerapundian, Jon Cooper and other colleagues at Birkbeck College in London are interested in finding a drug that will inhibit the first and more specific enzyme, renin.

Renin is related to enzymes in the digestive system, such as pepsin. Pepsin breaks down proteins into smaller units called peptides. Researchers, including Peter Hobart at the pharmaceuticals company Pfizer in Connecticut, first discovered that the enzymes were related when they sequenced the human renin gene. The similarity between amino acid sequences of renin and pepsin that this work implied gave Lynn Sibanda and I at Birkbeck College the basis for predicting renin’s three-dimensional structure back in 1984. Since then, other researchers have genetically engineered and purified human renin in sufficient quantities for Michael James and Anita Sielecki at the University of Alberta, Canada, to study it by X-ray diffraction. In 1989 they found out its three- dimensional structure. Their experiments confirm the predictions of the structure in general but have added more detail to our models.

For some time, Michael Szelke and his colleagues, first at the Hammersmith Hospital in London, and more recently at the Ferring Institute at Southampton, have been designing analogues, or variants, of the renin substrate angiotensinogen. Renin normally acts on this molecule to create angiotensin I, a precursor of the active molecule angiotensin II that eventually leads to an increase in blood pressure. (The researchers at Squibb and at Pfizer studied how angiotensin converting enzyme, or ACE, brings about the second step, from angiotensin I to angiotensin II.) The analogues differ from angiotensinogen in one important way – they resist attack by renin.

Some of the first analogues that Szelke and his colleagues made were effective as inhibitors of human renin. Unfortunately, the researchers could not use them as drugs because the analogues were too large to be absorbed through the gut and were digested when patients took them orally. The researchers needed to produce a smaller analogue, which probably would not be a protein, that could survive in the gut. It would also have to stay in the circulation for several hours. Chemists in many pharmaceuticals companies approached this problem in the classical way by systematically varying the chemical structure of angiotensinogen. Here, receptor-based drug design could make an important contribution to the search for an effective inhibitor of renin.

At Birkbeck, we studied crystals of the complexes that form between an enzyme related to renin and analogues of angiotensinogen. Michael Szelke and some of his industrial competitors from the pharmaceuticals giants such as Pfizer, Warner Lambert and Merck Sharp & Dohme produced the analogues. Our studies gave us detailed images of the complex between the renin-like enzyme and the analogues of angiotensinogen that inhibit it. From these we defined the structure and interactions of these inhibitors when they bind to renin itself. Working with various companies, most importantly with Dennis Hoover at Pfizer in Connecticut, we then looked for likely groups of atoms that we could attach to inhibitors to make them bind renin even better. Then chemists working for the pharmaceuticals companies made inhibitors with such groups attached. We looked at the structure of the inhibitors and how well they bound the renin-like enzyme. These studies told us a lot about how we could improve the molecule that inhibits renin to turn it into a drug. We were able to make more informed decisions about the design of useful drug molecules and we hope that patients will soon be testing some of them.

The renin story is a good example of how to design drugs using a receptor-based design cycle. It shows that we must first find out in detail the chemistry and the structure of the complex between the drug and its target. Then we can use this knowledge to suggest better groups to attach to the inhibitor, often with the aid of computer graphics and computer modelling. Medicinal chemists then make the improved inhibitors and the cycle begins again. As we gain more knowledge about receptors that are useful targets for drug molecules, such design cycles will become increasingly central to the pharmaceuticals industry.

Chemists should find similar methods useful for designing drugs that will ‘kill’ viruses. Last year Maria Miller and Alex Wlodawer at the National Cancer Institute in Washington and Risto Lapatto and I at Birkbeck unravelled the detailed three- dimensional structure of the enzyme – a proteinase like renin – that causes HIV to mature. In this process the proteinase cuts proteins of a larger, inactive precursor into smaller ones. The knowledge of the proteinase structure gives researchers the essential information they need to design inhibitors of the enzyme, and of HIV itself. Both inhibitors are likely to be useful weapons in the war against AIDS. Industrial chemists will be able to apply similar design cycles to the making of insecticides and herbicides for agricultural use.

Vaccines to order

We can also use design cycles to improve vaccines. Vaccines are suspensions of viruses or bacteria that have been inactivated so that they are no longer virulent or infective. Despite this, they still stimulate the body’s immune system to produce antibodies against them. If the body encounters the real virus or bacteria, it can defend itself because it recognises them and produces antibodies. Traditionally, medicinal chemists made vaccines by killing the real virus or bacteria. But this is not foolproof. Inevitably, some people become infected because the agent was not properly inactivated. In Europe, such active vaccines have been a major cause of diseases such as polio in humans or foot-and-mouth disease in cattle. We need an alternative approach, and the vaccine design cycle could help to solve this problem.

At the simplest level, chemists can begin to look at the problem of improving vaccines by identifying the proteins that are on the surface of the bacteria or virus. In the case of the foot-and-mouth disease virus, Fred Brown and his colleagues at Wellcome Biotechnology in London have genetically engineered proteins on the ‘coat’ of the virus to turn them into pure and non-virulent proteins for vaccines.

Another approach is to try to identify which bits of the proteins on the surface of the virus bind to an antibody. We can do this for foot-and-mouth disease virus, thanks to the work of David Stuart at the University of Oxford, with Brown and his colleagues at Wellcome. Two years ago, they announced that they had worked out the complete three-dimensional structure of the foot-and-mouth disease virus. They can make one part of the protein that sticks out from the viral coat and it makes a very effective vaccine. We should be able to use a similar approach for the viruses that cause polio and the common cold. This can be tricky, however. Sometimes viruses can change their protein coat very quickly and so avoid the antibodies that the vaccine stimulated. Unfortunately, this appears to be the case for the common cold and for HIV, the AIDS virus.

Design cycles will undoubtedly play an important role in the design of novel proteins, drugs and vaccines in the future. But this does not mean that pharmaceutical chemists will abandon the old methods of screening and randomly modifying existing molecules. For a pharmaceuticals or biotechnology company, the rational approach to designing more effective drugs has one major shortcoming: if one company can do it, so can a competitor. Most drugs companies feel happier with a ‘lead compound’ that they have chanced upon because others are unlikely to have found the same molecule. The best solution may be for pharmaceutical chemists to combine the two approaches: find a lead molecule by random selection, study how it interacts with its receptor and then improve the molecule by rational design.

Tom Blundell is a professor in the University of London at Birkbeck College and honorary director of the Imperial Cancer Research Fund Unit of Structural Molecular Biology.