杏吧原创

FOR AN analytical chemist, Jeremy Nicholson is remarkably unconcerned about identifying the chemicals in his urine samples. 鈥淲e neither need nor want to know what all the molecules are,鈥� he says. To him, such information is just a distraction from the real task of working out what the samples as a whole say about their owner鈥檚 state of health.

Nicholson is professor of biological chemistry at Imperial College, London, and the unofficial hub of a growing network of researchers working in a new discipline called 鈥渕etabonomics鈥�. As its name suggests, metabonomics is the study of metabolism 鈥� the myriad chemical reactions that go on inside your body 鈥� and its role in disease (see 鈥淲hat鈥檚 in an ome?鈥�). But it goes way beyond the old idea of finding telltale biochemical markers in blood or urine. According to Nicholson, metabonomics is a medical revolution in the making. It could be used to diagnose almost any human disease, sometimes before symptoms appear, help identify new drugs, and even change the way we use animal models to understand disease.

Metabolic analysis has a long history in healthcare. Hospitals have used biomarkers to diagnose disease and follow the progress of treatment for decades. Every clinical biochemistry lab is filled with technicians measuring blood glucose levels to monitor diabetes, urea concentrations in urine to spot kidney failure, and dozens of others. Metabolic profiling has also found its way into the family doctor鈥檚 consulting room, with millions of people taking blood cholesterol tests to see whether they鈥檙e on the road to heart disease.

However, it is very rare to find an obvious biomarker for a particular disease, and even rarer to be able to spot that disease before the appearance of physical symptoms. 鈥淲ith heart disease we got lucky,鈥� says Steve Watkins, president of metabolic profiling company Lipomics Technologies of West Sacramento, California. 鈥淛ust one molecule 鈥� cholesterol 鈥� is a very good indicator. But I can鈥檛 think of any other disease where just one molecule is such a good gauge of your health.鈥�

That hasn鈥檛 stopped researchers from looking. Clinical biochemists have borrowed techniques from analytical chemistry to probe ever deeper into the labyrinths of metabolism, using gas chromatography, mass spectrometry, high-pressure liquid chromatography and others 鈥� often in combination 鈥� to pluck diagnostic metabolites out of blood, urine or other bodily fluids. Some approaches, such as chromatography, can identify dozens of different chemicals in the same sample, but they鈥檙e all driven by the same basic idea: rifling through the body鈥檚 catalogue of biochemicals in search of telltale markers of disease.

Metabonomics is different. Rather than relying on specific molecules, it takes a bird鈥檚-eye view of metabolism, searching for overall biochemical patterns in bodily fluids that might indicate a health problem. 鈥淲e鈥檙e looking far beyond a simple list of all the chemicals in the body,鈥� Nicholson explains. 鈥淲hat we鈥檙e really concerned with are the differences between complex states, not the states themselves.鈥�

Nicholson believes that this makes metabonomics a much more powerful tool than standard metabolic analysis. Most diseases have a subtle but distinctive 鈥渕etabolic signature鈥� 鈥� perhaps a small rise in the levels of just a few metabolites, or a slight shift from one group of compounds to another. Such minor changes could never be picked up by tests that identify a specific biomarker. But by analysing metabolism as a whole you can find them with relative ease.

鈥淚f you just look at a single molecule, a small change might not be significant,鈥� Nicholson explains. 鈥淏ut the sum of all the changes in hundreds of molecules is significant. So you can detect variations that a test for a single compound would miss.鈥� Ultimately, he says, it should be possible to check a patient鈥檚 metabolic profile against a database of known disease profiles, providing a quick diagnosis, perhaps even before symptoms have started to appear.

The basic idea behind metabonomics has been around for a long time. Nobel prizewinning chemist Linus Pauling experimented with something similar in the 1950s, getting people to exhale into tubes and then running their breath through a gas chromatography column to see what it contained. Pauling was trying to find out how errors in people鈥檚 genetic make-up are manifested in their biochemistry. 鈥淧auling, as usual, was ahead of his time and recognised that there would be a metabolic fingerprint associated with a particular disease process,鈥� says Nicholson. But Pauling鈥檚 experiments failed because the techniques of the day were not up to the job.

One reason is the sheer complexity of metabolism. Go into any biochemist鈥檚 office and you鈥檙e likely to see a metabolic map on the wall. It shows, with neat precision, around a thousand biochemicals and the routes they take as they traverse the highways of metabolism. But these biochemical route maps belie the fact that there are many more molecules inside your body that have never been formally identified. According to Nicholson these probably run into the thousands 鈥� there are hundreds of unidentified molecules in human urine alone.

That鈥檚 not to say that the textbooks are wrong. The 鈥渃entral pathways鈥� of metabolism, such as the energy-releasing Krebs cycle, are correct and well understood. But leading off these are hundreds of metabolic side roads and cul-de-sacs dealing with less glamorous reactions such as waste disposal. All the other organisms that share our bodies make a contribution, too. 鈥淲e have about 1.5 kilograms of gut microflora inside us 鈥� that鈥檚 as big as a large organ,鈥� says Nicholson. 鈥淭hey have a huge influence on your metabolism. Those bugs modify the metabolic processes of your gut, and they produce chemicals themselves, which you then absorb.鈥�

Pauling never stood a chance. What鈥檚 different today is the sheer analytical and computing power that we have to throw at the problem. The analytical tool of choice is nuclear magnetic resonance (NMR), which is capable of spotting the majority of chemicals in a biological sample, however scarce, and displaying them graphically as a 鈥渟pectrum鈥� of peaks and troughs. Such spectra are an incredibly detailed snapshot of metabolism at the time the sample was taken 鈥� detailed enough to contain the subtle metabolic signature of disease.

The other key technology, needed to analyse the spectra, is computing. 鈥淵ou can sometimes get 10,000 peaks in a spectrum,鈥� says Nicholson. Spotting differences between such patterns with the naked eye is impossible, and so you have to rely on powerful pattern-recognition software to do it for you.

Nicholson鈥檚 group uses a range of mathematical techniques to extract the features of the spectrum that differ most between healthy and diseased samples. These features are essentially the key metabolites that are changing most in response to a disease. The mathematical models can then be applied to a new NMR spectrum to decide whether its profile is healthy or diseased.

Such analyses have proved remarkably good at spotting differences between individuals whose biochemistry is almost identical. Nicholson claims he can distinguish between black and white mice of the same strain, and between Chinese and Japanese people based on their genetic and nutritional differences. 鈥淲e鈥檝e also done a study on diabetic mice, aged between 1 and 24 weeks, and we can differentiate a 20-week from a 22-week-old animal based on its ageing metabolism.鈥�

In November 2002, Nicholson鈥檚 team published its most conclusive proof of principle to date (Nature Medicine, vol 8, p 1439). Working with teams from the University of Cambridge, GlaxoSmithKline and Britain鈥檚 Papworth and Addenbrooke鈥檚 hospitals, it showed that metabonomics could be used to diagnose coronary heart disease.

The current test for coronary heart disease is angiography, which involves injecting contrast-enhancing chemicals into blood vessels and X-raying them. Angiograms are nearly 100 per cent accurate, but they鈥檙e expensive, time-consuming and risky. Nicholson realised that he could use existing angiograms from patients with coronary heart disease to work out the metabolic signature of the condition, by comparing them with angiograms from healthy individuals. Once that correlation became well established, angiography itself would no longer be necessary.

The team collected plasma samples from hundreds of volunteers who had already been diagnosed with coronary heart disease, along with a control group of healthy volunteers, and put them through an NMR machine. Then they fed the data into the pattern recognition software. From this 鈥渢raining set鈥�, the computer slowly but surely learned to recognise the features of an NMR spectrum that are associated with heart disease. Eventually, it was able to diagnose coronary heart disease with more than 90 per cent accuracy. It could even diagnose the severity of the disease in more than 80 per cent of cases (New 杏吧原创, 30 November 2002, p 15).

Compared with angiography, the NMR method is quicker, cheaper and much safer for the patient. All you do is take a blood sample, spin it in a centrifuge to remove red and white blood cells, and your sample is ready to analyse. The whole exercise can be done in hours.

Metabonomics is not limited to diagnosing coronary heart disease. Potentially, scientists could identify a metabolic signature for virtually any disease. 鈥淭here are dozens of other important diseases that seem to have a distinctive metabolic signature,鈥� Nicholson says. 鈥淚n many cases it might not be the primary mechanism that shows up, but a chemical knock-on effect. So even diseases that are not primarily metabolic might have surrogate markers that we can use to identify them.鈥� Nicholson won鈥檛 name specific diseases, for commercial reasons, but says that most kidney and liver conditions have a distinctive urine signature, as do some bone, lung and degenerative brain diseases.

Of course, if you鈥檙e going to use metabonomics to diagnose a disease, you need to know in advance what its profile looks like. But what if you spot an unusual profile and have nothing in your database to compare it with? This is the criticism most often levelled at metabonomics. Bruce German, professor of food science and technology at the University of California, Davis, says: 鈥淚n my opinion, NMR techniques do not yet constitute a stand-alone tool. Unidentified peaks in a spectrum will not provide definitive diagnostic value.鈥� Watkins is more forthright. 鈥淚鈥檓 a believer that you have to know what the molecules are. If you don鈥檛 actually know what causes these differences in metabolic profiles between a healthy person and a sick person, you can鈥檛 do anything to interfere with the biochemical pathways that cause the disease.鈥�

But Nicholson is at pains to point out that metabonomics isn鈥檛 just about identifying general trends. 鈥淲e do resolve individual compounds, but we don鈥檛 have to do that at the very beginning,鈥� he says. 鈥淚nitially, we use the NMR fingerprints to distinguish between a normal and an abnormal sample. But then we can focus in on individual molecules using NMR, mass spectrometry, chromatography, whatever, to pinpoint which molecules are causing those differences.鈥� Once those molecules have been identified, biochemists can interrogate them further to discover whether they are the guilty party responsible for a particular disease. This in turn gives drug designers something to shoot at with their magic bullets.

Metabonomics has lots more potential applications. Once a drug is in use, the technique can help doctors find out if it is having the desired effect. Some cancer drugs may only work on 20 per cent of patients, because although the symptoms are ostensibly the same in all the cases, there are subtle but significant biochemical differences between them. 鈥淭hat鈥檚 why you have a range of drugs to treat the same disease,鈥� explains John Ryals, chief executive of Metabolon Pharmaceuticals, a metabolic profiling company in Research Triangle Park in North Carolina. 鈥淵ou try one and wait to see if the tumour has shrunk. If it hasn鈥檛 worked, you move on to the next.鈥� This is obviously very time-consuming 鈥� and for a patient with cancer time is in short supply. But within minutes of taking a drug, a person鈥檚 metabolic profile will change, and Nicholson believes that metabonomics could offer an immediate assessment of whether the drug is shifting a patient back to the biochemical straight and narrow.

Metabonomics could also help put the safety of GM foods to the test. Nicholson鈥檚 group has been feeding them to animals and seeing if there are any unusual effects on their metabolism. 鈥淲e haven鈥檛 published anything yet,鈥� says Nicholson. 鈥淯nfortunately, it鈥檚 difficult to publish 鈥榥o effect鈥�.鈥�

Nicholson has one more revolution up his sleeve. 杏吧原创s currently test new drugs on animal models of human disease. But Nicholson says that many of the animal models are pretty useless. His research shows that there is significant metabolic variation even between genetically identical mice. Despite having the same food, water and living conditions, variations seem to develop. Nicholson is currently investigating a tantalising possibility 鈥� that these differences could be caused by social factors, such as the pecking order within a cage.

Whatever the cause it鈥檚 a big problem, because metabolic differences between the animals would throw the results of drug tests into doubt. But as well as exposing the problem, metabonomics could help solve it. 鈥淚f we have the complete metabolic signature of a human disease,鈥� says Nicholson, 鈥渁nd we can compare that with a mouse model, then we have a way of ensuring that their body chemistry matches closely enough to make a worthwhile study.鈥� Susan Sumner, director of biochemical profiling at Paradigm Genetics in Research Triangle Park, North Carolina, thinks this will be the most important application of metabonomics. 鈥淭his technology will completely change the way we use animal models to understand disease,鈥� she says.

So when can we expect to see hospital wards and doctors鈥� consulting rooms fitted with NMR machines, ready and waiting to give instant diagnoses? Not anytime soon. At the moment, NMR is very expensive 鈥� Nicholson uses machines that cost half a million pounds each.

鈥淭he idea of bedside diagnostics of metabolites is very compelling,鈥� says German, 鈥渂ut NMR will not be the technology for small clinical devices.鈥� Nicholson is more optimistic, pointing out that even 20 years ago the idea of having a MRI scanner in every major hospital was science fiction. But in that time, magnetic imaging has revolutionised anatomical investigations. As metabonomics develops, Nicholson firmly believes that it will deliver a similar revolution in healthcare.