Dr Jenny Martin is a pharmacist who followed a research career path. She has a Master of Pharmacy from the Victorian College of Pharmacy for drug design studies and a DPhil from the University of Oxford for protein crystallographic studies on glycogen phosphorylase. After postdoctoral work at Rockefeller University in New York, she was awarded a Queen Elizabeth II Fellowship to establish the first protein crystallography laboratory in Queensland. Her research at the Centre for Drug Design and Development at the University of Queensland focuses on protein structure and drug design. In 1999, Dr Martin was awarded an ARC Senior Research Fellowship for 'Applications of Protein Crystallography to Drug Design'.

Molecular structure and recognition

Designer medicines – molecules of the future
by Jenny Martin
J.Martin@mailbox.uq.edu.au

What are medicines? Most drugs are relatively small chemicals. For example, aspirin has only ~10 non-hydrogen atoms. But drugs exert their pharmacological effect by interacting with much larger species, 1000 to 1 million atoms, depending on the protein. For example, the target of aspirin, an enzyme called COX, is a large biological molecule comprising a few thousand atoms. Ideally, drugs should have a very specific biological effect, or else side effects may result.

How are new drugs developed? The process of developing a drug from a chemical on the laboratory bench to a product on the pharmacy shelves is called the drug pipeline. The pipeline has many stages and a candidate drug can fail the process at any one of these. Estimates are that the chances of a drug making it through the pipeline are 1 in 10,000, the costs are US$300-500 million and the time required is 10-20 years.

How can we reduce the costs and increase the chances of finding new and better drugs? The later stages of the pipeline are regulated by the Therapeutic Goods Administration, to ensure that drugs making it to market are safe and effective. So it is not possible to reduce the costs or time at these stages. But we can reduce them at the early stages of the drug pipeline, specifically at the drug discovery step (identification of lead compounds).

Historically, we have identified lead compounds in a variety of ways:

• folklore use of natural products (eg, the use of foxglove to treat dropsy led to the discovery of digitalis);

• serendipity – a chance observation meeting a prepared mind (eg, Fleming and Florey’s discovery of penicillin);

• random screening, searching through thousands of chemicals until one is found with the desired biological activity, a bit like finding a needle in a haystack with odds of 1 in 10,000.

However we can be smarter about the way in which we discover new drug candidates, using a technique called structure-based design, which reduces the odds to 1 in 300.

So why isn’t structure-based design used all the time? Because there are certain prerequisites for this to work. The disease biology must be understood, a target protein must be identified, and its three-dimensional structure must be known in order to use this technique.

There are currently only a handful of diseases which meet these criteria, because only relatively few protein structures are known. The protein structure allows visualisation of hot spots in the protein, active sites where the enzyme does its work. Once we know what the hot spots look like, we can design drugs that match their size, shape and physical and chemical properties.

Recent examples of structure-based drug design are Relenza and Celebrex.

Relenza is an Australian-designed drug to treat influenza, for which there was previously no effective treatment. The work was a collaboration between the research groups of Peter Colman, Mark von Itzstein and Graeme Laver. It began by profiling the influenza virus, and identifying the key protein, neuraminidase. The three-dimensional structure of this protein was then determined and an inhibitor was designed to fit the structure. The drug was approved for use in humans in 1999.

Structure of the influenza virus neuraminidase enzyme (purple and yellow) showing how the influenza drug Relenza (green) blocks the function. From the work of research groups of Peter Colman, Mark von Itzstein and Graeme Laver.

Celebrex is a new treatment for arthritis. Previous treatments such as aspirin and non-steroidal anti-inflammatories have side effects because they don’t discriminate between two related proteins COX-1 and COX-2. COX-2 causes the inflammation of arthritis and is the real target for anti-arthritic drugs. Scientists determined the structures of both COX-1 and COX-2 and then designed inhibitors selective for COX-2. The new drug was approved for use in 1999 and was the most prescribed drug for treating arthritis.

Genomics will help us profile more diseases that are currently untreatable. As the album of three-dimensional structures grows, it will provide pictures of more and more disease targets. Structure-based drug design will then provide us with the tools for designing better drugs, and to combat diseases that are currently untreatable. These methods herald a new era in drug discovery.

Discussion

What proportion of new drugs will employ structure-based design?

Jenny Martin. That requires some crystal ball gazing. It takes 10 to 20 years to get a new drug onto the market and we are only seeing the first trickling through of structure-based designed drugs now. The sky’s the limit. If you have all the prerequisites it is possible and the odds are better. Most drug companies are using these techniques.

John Shine. It should also be noted that the genomic database is highlighting individual DNA differences that may contribute to disease. So it is likely that drug design will also become more individualised.

What is driving the search for drugs to cure diseases? The first world?

Jenny Martin. Pharmaceutical companies do have a responsibility to their investors. But universities are focusing on diseases of developing countries such as malaria and dengue fever.

Peter Colman. The problem is not limited to this technique.

How are the structures derived? Is it automated?

Jenny Martin. Protein structures are generally determined using X-ray crystallography. High-throughput crystallography is being developed to increase the speed.

Peter Colman. It is an imaging technique that can be used if you can crystallise the protein.

How many strains of influenza virus does Relenza treat?

Peter Colman. The susceptible site is the same on all wild strains of influenza. There is always the possibility of resistant strains developing, but we haven’t seen that happen yet.

Is it possible to model the structure of a protein rather than using X-ray crystallography for determining structures?

Jenny Martin. It depends on the problem. X-ray crystallography is the experimental method of choice for most proteins because it is not limited by the size of the protein. It is possible to model a protein structure from a related protein structure if the sequence identity is high. When there are no related structures known, it becomes a very difficult problem. As we determine more experimental structures using X-ray crystallography, the modelling problem will become easier.