by Jeffrey M. Perkel
For all the hype that the genome and (to a lesser extent) proteome receive, when it comes to understanding the biology of disease, development, and differentiation, there's another 'ome to which researchers should pay attention: The metabolome. Metabolomics refers to the study of the complete set of metabolites a given cell, tissue, organ, or organism produces under a given set of conditions, i.e. its metabolome. But unlike its macromolecular counterparts, the metabolome is highly responsive to changing conditions, says Steve Fischer, senior applications scientist for metabolomics at Agilent Technologies.
"Metabolomics is the metabolite equivalent of genomics," Fischer says. "The only difference is the genome is essentially immutable, whereas the metabolome changes in response to stimuli. It's like the genome is the program and the metabolome is the output result the genome generates given input conditions."
As a result, researchers in both academia and industry are actively searching the metabolome for molecules that can act as canaries-in-the-mine-shaft for their diseases of interest. They are looking, in short, for biomarkers.
"Very small changes in protein levels can have a significant downstream effect [on the metabolome]," explains Gary Siuzdak, Director of the Center for Metabolomics and Mass Spectrometry at the Scripps Research Institute in La Jolla, Calif. "So we find that metabolites are really very effective at telling us what's happening biochemically and what's being dysregulated." Two classic cases in point: phenylalanine/tyrosine for Phenylketonuria (PKU) disease, and cholesterol/sphingomyelin for Niemann-Pick's.
Those researchers interested in mining the metabolome have no shortage of tools available to help them in their search, from chromatography equipment and mass spectrometers to downstream analytical software. That doesn't mean it's easy, however. Scientists can detect hundreds and even thousands of metabolites per mass spectrometry run. But that doesn't mean they can necessarily identify those molecules, or figure out which ones are truly important.
One issue in metabolomics is that of scale. Just how big is the metabolome? Researchers haven't pinned down the exact number, but it's surely high. "Virtually every naturally occurring organic molecule is a metabolite," says Robert Gates, a Market Segment Manager at Sigma Life Science, the biological products and services research business at Sigma-Aldrich, a company that has hundreds of mass spec-ready metabolite standards in its collection. Fischer estimates the human metabolome comprises some 2,500 or so primary metabolites and maybe 15,000 secondary metabolites.
Unlike nucleic acids and proteins, these molecules have a range of physicochemical properties that preclude comprehensive one-shot analysis. Glucose is small, water soluble, and polar, for instance, whereas sphingolipids are highly "greasy" and hydrophobic. "The challenge in metabolomics," Fischer says, "is how you analyze these different things or get full coverage." Fischer divides the metabolome in his work into three: hydrophobic, polar, and very polar compounds. Analyzing those three fractions covers about 90% of the metabolome, he estimates. But most researchers likely never achieve anywhere near that; a broad coverage metabolomics study in the literature might hit only 5,000 metabolites at best.
Mike Milburn, Chief Science Officer at Metabolon, a metabolomics technology developer and service provider in Durham, NC, says his scientists typically identify between 500 and 600 metabolites from serum (of which 300 - 400 are "named biochemicals," and the rest novel), and between 800 and 1000 (50 - 60% named) from urine. "We can easily profile 500 metabolites in a 50 ul sample, and we can do thousands of samples," he says.
To get those profiles, the company employs a pipeline comprising liquid and gas chromatography coupled to single-stage "robust" mass spectrometers: Thermo Fisher Scientific's LTQ linear ion trap (for LC-MS) and DSQ single quadrupole (for GC-MS).
"We've found that the . . . instrument that best enables the technology is a simple linear ion trap that can perform at very high scan speed over a wide mass range and can also do fragmentation," says Milburn.
Your individual mass spectrometry needs may differ, though, depending on your application, explains Yingying Huang, Strategic Marketing Manager for Metabolism and Metabolomics at Thermo Fisher Scientific. "There are two types of metabolomics research," she says. Untargeted profiling is "a systems-wide approach," trying to dissect metabolic differences between healthy and diseased samples, for instance, or between fasting and fed subjects. In short: biomarker discovery.
The other type of research is "targeted" metabolomics, quantifying metabolites you have already identified as important. "The main challenge for metabolomics resides in the untargeted approach," she says, because of the complexity of the spectra, the concentration of metabolites at the low mass range, and the challenges in data processing and identifying all those molecules.
"In the ideal world," says Fischer, you'd have instruments for both needs: an accurate mass-capable machine that can measure masses down to low ppm levels for discovery work, and a triple quadrupole for quantification.
For metabolomics work Thermo Scientific recommends its high-resolution Orbitrap mass spectrometer, available in three versions, says Huang. These include the Thermo Scientific LTQ Orbitrap (a hybrid linear ion trap-Orbitrap system), the Thermo Scientific Q-Exactive (a lower-cost quadrupole-Orbitrap hybrid), and the new Thermo Scientific Orbitrap Elite, with a mass resolution of over 240,000.
"With 240,000 resolution, you can look at fine isotopic resolution of the molecules, which means you can do a lot of new experiments that weren't possible before," says Huang, such as "isotope-tracing experiments" and "fluxomics" experiments, which track the molecular fate of isotopically labeled precursor compounds in vivo.
At Agilent Technologies, Fischer recommends either a standard time-of-flight or hybrid quadrupole-TOF instrument for metabolomics research, as both provide the mass resolving power and accurate mass measurements that are key for deducing the precise elemental composition of novel compounds -- with affordability and ease-of-use.
Agilent has four such instruments in its portfolio, the newest of which, the 6550 qTOF, is 10-time more sensitive than its closest counterpart, the 6540 qTOF. Fischer's lab, though, mostly uses the workhorse 6224 TOF, relying on mass and retention time alone for identification. "With the 6550," he says, "we might rethink our processes, because 10-times more sensitivity creates new possibilities." But, he adds, such sensitivity is only necessary in cases where sample is scarce. "The reality is, if I have plenty of sample I don't need the most sensitive qTOF, because with enough sample I can have enough to see the rare metabolites."
Deriving biological insight
Of course, getting the data and being able to make sense of it are two different things. According to Milburn, it isn't the instrumentation and separation technologies Metabolon uses that make the company's technology so powerful. The real secret ingredient behind the company's process – which it plans to apply to some 50,000 samples and 500 projects this year alone -- is the "million-plus lines of code we've written" to extract compound identification and biological insights from the complex mass spectra and chromatography profiles that result from a metabolomics experiment.
Each sample the company runs, Milburn explains, produces between 10,000 and 15,000 "ion features" – a composite of ion mass, intensity, and retention time. That's some 50,000-odd features total for all three mass spec runs each sample undergoes. According to Milburn, the company has spent years developing software to identify those features, highlight the ones that appear interesting but which it cannot identify, and then extract a measure of biological meaning – which pathways are undergoing changes in metabolic flux, for instance.
The real challenge, Milburn says, is to move one step beyond that to the 30,000-foot level, to be able to look at the global metabolite profile and infer that this is a cell undergoing mitochondrial malfunction or oxidative stress, for instance. Milburn calls that process the "triangulation of biochemistry," and it is still the most laborious, manual part of his company's entire process, and it's one reason they have 20 PhD-level biochemists on staff.
Mass spectrometry vendors are developing software for their customers to help them sort through metabolomics data, too, such as Thermo Scientific's SIEVE 2.0 software. "What SIEVE gives you is a very nice data-mining tool and dramatically decreases your false positives," says Huang.
Alternatively, researchers can use the free XCMS and metaXCMS software developed in Siuzdak's lab. According to Siuzdak, XCMS, which takes LC/MS data on metabolic profiles of, say, diseased vs normal samples and identifies "the molecules that are most interesting in terms of statistically significant p-values and fold-changes," has been downloaded some 50,000 times since 2006. metaXCMS offers the same capabilities, but across multiple sample populations. In one 2011 study, his team used the software to reduce the problem of pain-associated metabolites from about 1800 dysregulated candidates to three.
"The biochemical aspect of data analysis is quite complicated, so if you can reduce the number of peaks through second-order 'meta analysis' of multiple populations, it helps identify the most relevant metabolites, the ones directly associated with the disease or mutants," Siuzdak says.
Yet for all its power, the original XCMS is not for novices, says Siuzdak; it is a command-driven application that requires expertise in the statistical language, R. Recently, though, his team launched a menu-driven, web-based version called XCMS Online . Users upload their LC/MS data to the Scripps server, and it processes the data and extracts the most interesting molecules. The inclusion of both mass and retention time data are key, Siuzdak says, as metabolite masses often overlap, but the combination of both features is generally diagnostic of a given compound.
Based on those features, the software makes a provisional identification of the compound with a link to the METLIN database which includes detailed reports of the mass spectrometric behavior of some 45,000 individual metabolites as well as MS/MS data on thousands of those metabolites (in both positive and negative ion mode, over a range of collision energies.) According to Gates, Sigma submits 100 or so compounds periodically for inclusion in the database, which then links back to the Sigma product page for ordering information.
"XCMS and METLIN's online MS/MS database has made the whole process of doing comparative analysis much more straightforward," Siuzdak says.
Advice
For all the tools available to researchers, metabolomics should not be entered into lightly, Fischer says. Proteomics is comparatively simple, as peptides as a whole have more in common chemically than they differ; the complexity comes from the sheer numbers and need for detection sensitivity. But metabolites are all over the place, chemically speaking, meaning no one sample preparation method, analytical approach, or instrument can handle everything in one pass.
He offers four bits of advice to any researcher contemplating metabolomics. First, "Find a colleague who's doing metabolomics and have a conversation with them: They will have lived through the trials and tribulations."
Second, "recognize that there are physicochemical differences in metabolites, so you'll need to do multiple analyses." Alternatively, unless you're interested in the entire metabolome, consider how to selectively extract the class of molecules you actually care about.
Third, make sure you have someone in the lab (or as a collaborator) who can do the statistics and data mining required of metabolomics.
Finally, be aware that "compound identification is an ongoing challenge." New metabolites are being identified and characterized regularly – Metabolon adds about 90 new compounds a month to its libraries, Milburn says – but many remain unidentified. It's entirely possible that the most interesting compound to emerge in your own metabolomics research could be something no one has ever seen before, which means you can look forward to some detailed analytical experimentation to work out its precise structure.
"Recognize that identification will be a challenge," Fischer says. "You cannot make biological sense out of something unless you know what it is."
The image at the top of this page is Thermo Fisher Scientific's Orbitrap Elite.