When Merck & Co. withdrew the painkiller Vioxx from the market last September after the drug was linked to increased risks of heart attack and stroke, more than one metabolomics researcher shook their heads and thought, "If only ...."
"The removal of Vioxx raises the question: Couldn't it have been possible to know earlier what the behavior of these drugs in vivo would have been?" says Steve Martin, senior vice president and chief technology officer of BG Medicine, a systems biology company in Waltham, Mass. "There is a real drive to apply [metabolomics] to get a much higher-resolution view much earlier of humans' responses to drugs," says Martin.
Metabolomics is the systematic study of the unique chemical fingerprints that specific cellular processes leave behind. The field is relatively new, but the basic concept is not: Physicians have long understood that metabolites can reveal pathology.
To get a sense of just how much that is so, consider the theorems of metabolic control analysis, says Douglas Kell, a chemist at the University of Manchester, UK, and a leading metabolomics researcher. The suite of equations developed in the United Kingdom and Germany in the 1970s "tells you that small changes in fluxes through pathways can lead to large changes in the concentration of metabolites," meaning that metabolomics is bioscience's proverbial canary in the coal mine, signaling tiny but crucial disturbances that other analyses might not perceive as quickly.
That makes it an indispensable element of systems biology, metabolomicists argue. Last year, scientists at TNO Pharma in Zeist, the Netherlands, and BG Medicine showed that atherosclerosis in lab mice could be diagnosed earlier and more clearly using metabolomics combined with genomics and proteomics than with just the latter two alone.1 "This will be the year in which people show that they have put together microarrays, proteomics, and metabolic profiles all in the same run from a single sample," predicts Pedro Mendes, a research associate professor at the Virginia Bioinformatics Institute in Blacksburg.
But first metabolomics will have to overcome some teething pains. Researchers have matched thousands of metabolites to cellular processes, but thousands more remain unidentified. And software's analytical powers are sinking fast beneath the rising tide of new information.
Nevertheless, the field is growing, and Big Pharma has embraced metabolomics. Technology developers are pouring resources into the breach, looking for alternatives to nuclear magnetic resonance (NMR) and mass spectrometry, the traditional tools of the trade. And, the National Institutes of Health has earmarked some $70 million over the next five years to advance research and technology development.
Courtesy of Oliver Fiehn
Part of the
Interest among drug makers has become "huge" over the past two years, according to Kell. Many global pharma companies are conducting their own small projects in-house or are collaborating with the growing number of academically linked private consulting firms.
SIDMAP, a Los Angeles company, helped one pharma giant determine whether an experimental compound for treating a metabolic disorder was able to force fatty acids into mitochondria more effectively than current drugs can. "We could quantify how much more effective the new compound is," says Laszlo Boros, SIDMAP's chief scientific advisor and codirector of the Los Angeles Biomedical Research Institute's Stable Isotope Research Laboratory. "Testing the rest of the metabolic network, we didn't see any undesirable side effects," an analysis that could speed drug development and approval.
Novartis has invited Boros to talk with its scientists about some patients' resistance to the antileukemia drug, Gleevec. "Patients can be resistant because the disease can switch from a metabolic pathway Gleevec controls to one it doesn't," he says, also noting that SIDMAP's tracer-based metabolomics technology predicted that problem five years ago.
Metabolon, a metabolomics consulting firm in Durham, NC, has found distinctive biochemical markers in blood and cerebrospinal fluid, signaling the onset of amyotrophic lateral sclerosis (ALS). "We've clearly recognized a number of different [metabolic] patterns within the ALS population," says Christopher Beecher, the company's chief scientific officer. "We assume that this represents different pathological origins," indicating a promising direction for research.
More importantly, Metabolon has also analyzed the cellular effects of riluzole, the only drug approved for treating ALS. Riluzole can provoke an array of side effects from nausea to impaired lung function. "We see a gross disturbance in biochemistry," Beecher notes. "Many of the changes are not necessarily part of the original state that differentiated the ALS patient from the control group," pointing to the drug as the culprit.
Beecher and his team are working with pharmaceutical companies to develop alternatives to riluzole, and they can now compare the drug's metabolic effects with those of early-stage candidates to determine which one is least likely to harm patients. Metabolon has begun similar research into Parkinson and Alzheimer diseases, and in March announced collaboration with Massachusetts General Hospital to search urine and plasma samples for biomarkers of diabetic nephropathy.
TOOLS OF THE TRADE
To create these metabolic profiles, metabolomicists use mass spectrometry and, to a lesser extent, NMR, which serves as a measuring tool. "If a peak in your sample is twofold higher than in a control, then you know that there is twofold more of the compound," says Oliver Fiehn, a molecular biologist who helped coin the term "metabolomics" and recently moved from the Max Planck Institute for Molecular Plant Physiology in Pots-dam, Germany, to the University of California, Davis. "But a single metabolite can give many different signals in NMR, so you have problems identifying the metabolites."
That's where mass spectrometry comes in. Gas chromatography with time-of-flight mass spec analysis can identify as many as a thousand compounds in a sample instead of the usual few dozen. Recently, liquid chromatography has yielded more precise results when coupled to two-capillary electrophoresis. Fourier-transform techniques reveal chemical fingerprints by comparing the amount of infrared light that compounds absorb.
"In metabolomics, you go from very polar things to very hydrophobic things, so you need multiple methods," Kell says. "Then compare the results to get a clear picture."
Equipment suppliers are working to make those comparisons easier. Last year Bruker BioSpin, based in Billerica, Mass., introduced its Metabolic Profiler, which separates a sample in a liquid chromatograph, splits the fractions between an NMR instrument and a mass spectrometer, and then correlates the resulting data using software. "The advantage is convenience and speed," says Werner Maas, Bruker's vice president of research and development. "Labs might have these same machines, but when you're screening or doing clinical validations, samples need to be handled a number of times. If you can put a well plate into our device and say, 'OK, I'll come back tomorrow,' that's useful for busy people."
SIDMAP, an acronym for stable isotope-based dynamic metabolic profiling, has developed its own addition to the metabolomics toolbox. The company attaches carbon-13 isotope tracers to molecules, such as glucose, that make their way into virtually every cellular reaction. By then seeing where and how the tracers are arranged in a cell's metabolites, researchers can deduce the reactions and pathways that led to those results. That knowledge can not only map new biochemical knowledge, but also trace cells' reactions to drugs or disease.
At the moment researchers have no metabolomic equivalent to the DNA microarray, says Fiehn, yet he remains hopeful. "It will take five or 10 years, because it's so difficult," he says. "But eventually it can be done."
For now, however, metabolomics researchers are largely content with their existing instrumentation. Researchers always crave more sensitivity and dynamic range, but "on the hardware side we have more or less everything we need," says Robert Hall of the private, nonprofit Plant Research International in Wageningen, the Netherlands. "In software, that's definitely not the case."
COMBATING DATA OVERLOAD
Indeed, metabolomicists are being swamped by data. Fiehn says he can name about 400 metabolites in 25 minutes from a single run. "But typically there are another 500 to 1,000 compounds there that we don't recognize. If you're working with the human genome, you see 20,000 genes and you're done. But there are about 200,000 known metabolites, and it's a good guess to say that there are at least a million out there."
Fiehn's research group is working to identify them, a task he can complete at the rate of about one metabolite per day. The group invested two years in writing and perfecting algorithms to search mass spectra for unknown metabolites and record them in the lab's library. The program can sift through 500,000 spectra in 24 hours, compared to the 500 that Fiehn could manage manually.
Of course, recurring metabolic patterns can still yield statistically useful indicators even if the compounds themselves can't be identified. Commercially available pattern-recognition software does a competent job, although most researchers still write their own programs to move data from their equipment into their statistical analysis software.
Now dedicated commercial metabolomic software products are coming online. Plant Research International has begun publicly offering its MetAlign program, which screens time-related and other noise from mass spec traces. And Phenomenome Discoveries in Saskatchewan, Canada, has developed software that speeds metabolic profiling. Applied Biosystems, one of the largest mass spectrometer manufacturers, has formed a software marketing collaboration with Phenomenome, in which the software developer will customize the program for compatibility with Applied Biosystems' instruments and then offer it to the company's customers.
"Right now, we're in a period similar to the early days of DNA sequencing," says Metabolon's Beecher. "In the beginning, the sequencing was done by hand. As the process was better and better understood, sequencing was not only more and more automated, but there also was more confidence that the software was doing it correctly. In metabolomics, we're still spending a fair amount of time checking the machines and software. A few years from now, the process will be automated and very straightforward."
NIH KICKS IN
The National Institutes of Health has fielded two initiatives to help move this development forward. The first is the Lipid Metabolites and Pathways Strategy (LIPID MAPS). Under a five-year, $35 million grant, 18 universities, companies, and research institutes are collaborating to chart the ways of fats in the RAW264.7 macrophage cell line. "We're trying to globally and comprehensively identify, characterize, and determine the function of all lipids present in that cell," says project director Edward Dennis, a biochemist at UC-San Diego.
The project has three goals: to separate and detect all the lipids in the cell and characterize any novel lipids present; measure each of the lipid metabolites in the cell and changes in their levels and locations during cellular processes; and define and map biochemical pathways and interaction networks for each lipid.
"We're making good progress in identifying the major lipids present," Dennis says, "and in setting standards and developing technology for measuring them. We're getting initial data on the lipids and measuring changes in their amounts over time. Our third goal of mapping pathways and interaction networks will come later."
The project has a larger objective as well, however. "We see this as a prototype that other people in nonlipid areas of metabolomics could use to structure their own research," Dennis says. He sees the Human Genome Project, which studied DNA from only a few individuals, as an analogy. "We hope that as we establish this knowledge for this cell, other researchers will be interested in studying other aspects of the same cell. By understanding one organism, you develop methods that you can use to look at others."
The second initiative is part of the NIH Roadmaps project charting critical paths of medical research for the new century. It targets metabolomics technology development with as much as $35 million in grants spread over three years. The project made nine awards in September 2004 and another five in March 2005, the two rounds totaling $10.8 million. Projects showing progress will share up to $14 million in development funding in each of the program's second and third years.
"We asked not just for projects related to classical NMR or mass spec technologies, but also for entirely new technologies that would allow us to measure concentrations of molecules in living tissue with attention to the spatial and temporal location of those molecules," says Maren Laughlin, the project's coleader and the senior advisor for integrative metabolism at the National Institute of Diabetes and Digestive and Kidney Diseases.
One grant supports better probes to speed NMR throughput; others fund efforts to improve NMR's sensitivity for metabolomic applications and improvements in sample introduction and measurement by mass spectrometry. But the program is also backing attempts to modify the "riboswitch," an RNA that regulates gene transcription and can be engineered to bind to a targeted compound, as well as efforts to create genetically encoded nanosensors that can zero in on "certain membrane-bound components of the cell such as the nucleus," Laughlin says. "They will go straight to the targeted compartment and report only on the concentration of certain molecules there."
The Roadmaps program is also underwriting attempts to help metabolomics researchers beat data overload. One grant is helping bioinformatics specialists find more sophisticated ways to deal more accurately with experimental results but still do so at high throughput. "Now, the predominant method for analyzing large data sets is pattern recognition and binning," Laughlin notes. "Our granted project is trying to use a lot of
In July the NIH will host a workshop to examine metabolomicists' needs and concerns in bioinformatics and, many hope, make a plan to address them systematically.
Taken together, the tumult of activity in metabolomics indicates that the new science is coming of age or, at least adolescence. "A lot of pharmaceutical companies have been slow to jump into the latest 'omics technology," says Julie Wingate, Applied Biosystems' mass spectrometry product manager. "It's a good approach: They want to make sure that metabolomics will prove its utility. In the near future, companies have to start seeing metabolomics actually provide benefits. When this happens, metabolomics will really take off."