There are several reasons why some patients may or may not respond to a drug, or may exhibit a certain side effect that other patients do not. Some of those reasons are genetic, as pharmacogenomics researchers have shown — certain alleles can predict response to a drug or the likelihood of an adverse reaction. But pharmacogenomics has been unable to explain all the variability in drug response, so metabolomics researchers have stepped in to see whether their discipline can help explain why some patients respond to drugs the way they do.
While metabolomics researchers look at metabolic profiles in plasma, serum, or urine to determine the differences between people with a certain disease and healthy people, pharmacometabolomics is an extension of that, says Imperial College London's John Lindon. "Once you've got the biomarkers of the disease — these are the metabolites — you can go back and look for the mechanism by looking at the enzyme pathways, to see which pathways are involved in using up those metabolites," Lindon says. "We look at a group of people's urine and we look for metabolic differences in the pre-dose, which would then be predictive of what happened post-dose."
Like pharmacogenomic researchers, pharmacometabolomic researchers look for signals in a person's biology that may indicate why a drug affects a person the way it does. But instead of looking at genetic differences, these researchers look at differences in enzymes, metabolites, and small molecules. Using nuclear magnetic resonance spectroscopy and different kinds of mass spectrometry, "we look for the metabolic fingerprint that says this person would process this drug differently — it might be more toxic in that person or more beneficial in that person. The idea would be to go towards personalized medicine," Lindon adds. His group published the first pharmacometabolomic study on pharmacometabolomic phenotyping and its potential use as a personalized medicine tool, in Nature in April 2006.
The benefit of looking at drug response on a metabolic level rather than a genomic level, Lindon says, is that while genomics reveals everything about a person's DNA, it says nothing at all about a person's environment. "Epigenetics tells you about your environment, but genetics and genomics people are largely blind to the environmental influences," he adds. "Metabolism is the endpoint of all the processes of the body, and is exquisitely sensitive to environment."
A comprehensive view
Indeed, says Duke University's Rima Kaddurah-Daouk, pharmacometabolomics is a way of looking at drug response in a top-down, systems biology way, where everything from patients' environments, to what they eat, to their genetic makeup and the composition of their gut microbiome comes into play. "At the end of the day, all of these factors interacting together is going to define our health," she adds. "If we can measure all of these chemicals in the blood or in the tissues, we're going to be able to understand better how the environment, the gene, the gut microbiome — all of these interacting factors — were able to impact our health in a more global way."
Kaddurah-Daouk's team published a study in PLoS One last October on the pharmacometabolomics of simvastatin treatment for high cholesterol. The team used a gas chromatography-mass spectrometer to measure a panel of metabolites within the bile acid formation, cholesterol synthesis, and dietary -sterol absorption pathways to see if changes in metabolites other than cholesterol could explain the variability of patients' responses to the drug. They identified three bile acids derived from gut bacteria that can contribute to the magnitude of some patients' responses to the drug, confirming other findings that had suggested the gut microbiome plays a role in cardiovascular health.
They further found that simvastatin not only affected cholesterol levels, but also the levels of more than 200 different lipids — some of which, in turn, affect other molecules, including neurotransmitters in the brain. Such studies, Kaddurah-Daouk says, illustrate the importance of adding pharmacometabolomics to drug research, particularly as it relates to personalized medicine. "We want to use metabolic status to determine, 'Am I a good patient for simvastatin? Should I be on a simvastatin or something else?'" she adds. "By getting the biochemical information, maybe there is additional information that a clinician can glean. The pharmacogenomics has been good but a limited number of success stories have been delivered. After all sorts of GWAS studies, we realize that the genetics cannot be everything — it does not explain all of the variation of response. If we add this extra layer of information — the metabolic signature — can we really start to glean insight about who is going to respond and why? Early data suggests this is very possible."
The pharmacometabolomics approach could, then, allow researchers to create new drugs, to more carefully target existing drugs to certain sub-populations of patients, or even to re-purpose existing drugs. "Maybe we will find something in [the metabolic makeup] that will help us determine who is a responder and who is not, and, more importantly, we will be able to determine what are the biochemical changes that can lead to response or no response within an individual," Kaddurah-Daouk says.
This is not to say that pharmacometabolomic researchers want their field to supplant pharmacogenomics — in fact, many see pharmacometabolomics as a tool to help pharmacogenomic researchers by giving them additional information as to where to look for SNPs that may cause some patients to react to a drug differently than others.
The Pharmacometabolomics Research Network, an NIH-funded project run by Kaddurah-Daouk, started about five years ago with the purpose of integrating metabolomics into pharmacology and pharmacogenomics. "We said, 'Look, we can push the envelope, not only to better understand our health by scaling up biochemistry, but we can also give a drug and try to understand what this drug does to metabolism in ways not possible before,'" Kaddurah-Daouk says. "And can I take this information and go to the pharmacogenomicist and say, 'Maybe this information can help guide your effort.' Instead of looking at millions of SNPs and variations to pick up a very faint signal or signals in the middle of a big haystack, maybe the metabolomics can focus this effort and give insight that could have been missed."
Richard Weinshilboum's lab at the Mayo Clinic in Minnesota has collaborated with Kaddurah-Daouk's team to conduct many of the pharmacometabolomic-related studies that combine that approach with pharmacogenomics. "So now you have one institute sponsoring pharmacogenomics and pharmacometabolomics and trying to bring them together. And to take the same patients that we're studying in the Pharmacogenomics Research Network and utilize samples from those patients in the metabolomics network and see how the two 'omics might inform each other," Weinshilboum says. "The problem is that they've operated somewhat in isolation from each other and the idea is, 'How do you bring these things together?'" He adds that this approach of combining pharmacogenomics and pharmacometabolomics is "generalizable" and could be used in a multitude of ways.
Indeed, in a Clinical Pharmacology and Therapeutics paper published in January 2011, the team used a pharmacometabolomic approach to implicate the glycine pathway in the response of patients with major depressive disorder to SSRI medications. Metabolic plasma assays of 20 responders and 20 non-responders to the drug escitalopram showed that glycine was negatively associated with treatment outcome. The researchers then genotyped tagged SNPs for genes encoding glycine synthesis and degradation enzymes, and identified a specific SNP in the glycine dehydrogenase gene that was associated with the treatment response phenotype.
"This pathway was not seen by the geneticists because if you're looking at 2 million or 3 million SNPs, you can easily miss it," Duke's Kaddurah-Daouk says. "We took biochemical information back to the pharmacogenomicists, and we were able to highlight for them all the genes that are in the pathway. And now we told them, 'Instead of looking at all 30,000 genes, please go back and look at these 14 genes. These are the genes that seem to be implicated and maybe you can find something if you look in a targeted way.' They looked at the SNPs within these genes and the task became more manageable." The pharmacogenomics researchers were able to confirm that a glycine dehydrogenase SNP was associated with response or non-response to SSRIs, which the researchers missed when they did a random genetic screen.
In a follow-up study published in Pharmacogenetics and Genomics this February, the team obtained GWAS data on those same samples — -allowing them to skip the genotyping step — and used that information as a scaffold to impute SNPs for genes in pathways identified through metabolic panels and GC-TOF mass spectrometry. "The reason for doing that was not the make the point that you can impute. That's fairly obvious," Weinshilboum says. "It was that once you've done this sort of study, and you have the clinical phenotype, if you have the GWAS data, you can now go in depth into any pathways that might be identified by performing metabolomics on many different platforms."
This follow-up study confirmed the results of the first, and indicated that the use of GWAS data to impute SNPs was a viable, rapid, and cost-effective way to identify SNPs in pharmacogenomic studies. "The [various metabolomic] platforms see different metabolites, they will identify different pathways, and we can very quickly, in a matter of weeks — once a pathway is identified by metabolomics — integrate that information into genomics without ever genotyping, and then replicate any of the top SNPs that we see," Weinshilboum adds.
This finding is of particular importance to researchers studying drug response in people with neuropsychiatric disorders, he says. One difficulty with studying such disorders has been that genomic techniques that have successfully identified SNPs and mutations associated with disease risk and drug response in the body have not been as successful with the brain. "There are a lot of possible explanations, but one that we suspect as contributing is diagnostic heterogeneity — that is we don't have a biomarker, a blood test, that we can use to diagnose depression, so we diagnose it on the basis of a series of validated questionnaires that clinicians will apply," Weinshilboum says. "That's quite different than what we have in other areas of medicine where we will have a blood cholesterol level or a scan that will show us the calcium in the coronary arteries." So one reason that there is so much variability in drug response in patients with these disorders is that there may be multiple differing pathophysiological reasons why people show psychiatric symptoms that are currently diagnosed as one disorder.
"What metabolomics offers us is the opportunity to sort out differing pathways that might contribute to those different pathophysiologies," Weinshilboum says. "Metabolomics may begin to be one tool that will help us to find different pathways in different patients which we can then rapidly query to see if there are underlying genetic polymorphisms or mutations that might be contributing to that particular pathway going astray." Once these pathways and genes are identified, pharmacometabolomics combined with pharmacogenomics could aid researchers in creating more effective and targeted drugs for these patients.
"We call this approach pharmacometabolomics-informed pharmacogenetics," Kaddurah-Daouk adds. "Basically at the heart of it is really the systems approach in trying to understand variation within a disease like depression. … This is really the beginning of bringing the 'omics together, and the birth of a new field."