by Jeff Perkel
The call comes in from the doctor following a routine blood draw. Your cholesterol is too high, she says, watch your diet. Or maybe, serum hormone levels have just provided the happy news: you're pregnant! In either case, your physician gained a window into your physiology with the help of biomarkers—molecular canaries in the proverbial mine shaft.
As Eric Fung, chief scientific officer at diagnostics development company Vermillion, puts it, "The most generic definition of a biomarker is any analyte that correlates with a phenotype."
Traditionally, biomarkers have been limited to such standard clinical laboratory fare as cholesterol, human chorionic gonadotropin (hCG), and glucose. But the definition need not be so restricted; gene expression, protein, and metabolic profiles can herald Alzheimer's disease progression, drug toxicity, and even early cancer.
Key to unearthing such biomarkers is mass spectrometry. Says Jeremy L. Norris, director of mass spectrometry at Protein Discovery, a MALDI imaging and biomarker discovery service provider in Knoxville, Tenn., mass spec "is the engine of discovery, if you will."
Fundamentally, the process is simple. Say you wanted to find a marker for cardiovascular disease. You would collect samples from both diseased and control individuals, fractionate the samples chromatographically, collect mass spectra on each fraction (either online, using electrospray, or offline, using matrix-assisted laser desorption/ionization (MALDI), and then use software to sift through the data that emerges, looking for proteins, peptides, or metabolites that correlate, either positively or negatively, with disease. The resulting short-list must then be validated, to show, for instance, that the biomarker is robust, and that it holds up in other populations.
Vermillion uses two kinds of mass spectrometers in its research and development. The primary is a variant of MALDI mass spectrometry, called SELDI. "The SE in SELDI is 'surface-enhanced,' which refers to the chip surface," explains Diane McCarthy, biomarker research centers manager for Bio-Rad Laboratories in Malvern, Pa., which offers both SELDI instrumentation and contract services. Unlike in MALDI, where, because the sample and matrix are plated on an inert surface, all the molecules in the sample are captured more or less equivalently, SELDI arrays are derivatized to selectively capture molecules based on charge, hydrophobicity, or metal-affinity, providing on-chip cleanup and an extra measure of separation.
(Bio-Rad and Vermillion have a shared interest in SELDI; Bio-Rad obtained its SELDI business last year from long-time developer Ciphergen, which shifted its focus to diagnostics, and earlier this year changed its name to Vermillion.)
The other system Vermillion uses is a SELDI-adapted Applied Biosystems QSTAR MALDI quadrupole-time-of-flight hybrid, which the company uses for protein sequencing and identification, says Fung.
SELDI offers two advantages for biomarker discovery, says Fung: throughput and selectivity. Based on a 96-well plate format, SELDI chips enable high-throughput research, while their derivatized surfaces enable a degree of on-chip fractionation. "To me, proteomics is about throughput and fractionation, and SELDI allows you to do both," he says.
The combination has evidently paid off for Vermillion, which has in clinical trials a SELDI-based diagnostic for ovarian cancer triage and hopes to win US Food and Drug Administration approval in 2008.
Protein Discovery also employs a MALDI-based workflow, using a Bruker Daltonics autoflex MALDI time-of-flight instrument to identify biomarkers of drug efficacy, including in a recent project with Novartis to identify indicators of Gleevec function.
"The autoflex is the central cog in the type of biomarker discovery we do here," Norris says, noting that Bruker instruments "are superior in terms of ability to measure both proteins and peptides at high resolution and mass accuracy, and speed." Add to that the company's software for mass spec-based imaging, and the result is "the ideal instrument for us," he says.
Katianna Pihakari, Fourier transform mass spectrometry (FTMS) field marketing manager for Varian FTMS Systems, notes there are two alternative strategies for protein biomarker discovery. In the "bottom-up" approach, samples are digested with protease (such as trypsin) prior to analysis, producing a peptide mixture. In the "top-down" approach, the proteins are analyzed while intact, making it possible to detect, for instance, differentially modified forms of the same protein, as well as any peptides in the sample.
For either strategy, she asserts, high resolution, high mass accuracy instruments, such as Varian's 901MS electrospray FTMS system, are ideal. That's because the initial phase of biomarker discovery must cast as wide a net as possible. The human proteome contains some 28,500 proteins, amounting to more than three million tryptic peptides, some of which are so similar in size, she says, "only FTMS can separate those in a reasonable time scale."
On the other hand, she continues, "If you want to do top-down, FTMS is also great, because it has the most versatile fragmentation tools available." Varian's 901MS, for instance, as a hybrid triple-quad FTMS, supports MS/MS (using the quadrupole collision cell for fragmentation), as well as MS3 and MS4 in the FTMS, using the triple-quad for ion filtration.
(Note that FTMS is the not the sine qua non of high mass accuracy; Q-TOFs, such as the 6520 from Agilent Technologies, and Thermo Fisher Scientific's Orbitrap systems, will also work.)
Once a candidate biomarker has been identified, says Mary Lopez, director of Thermo Fisher Scientific's biomarker research in mass spectrometry center in Cambridge, Mass., focus switches from discovery to validation, which typically involves quantification of one or a few candidate markers in large numbers of samples using a triple-quadrupole with single- or multiple-reaction monitoring (SRM/MRM).
Thermo Scientific TSQ™ Quantum triple-quads, she says, "allow us to specifically measure very accurately the quantification of a particular target peptide."
MRM is a targeted process that involves selecting the intact analyte in the first quadrupole (Q1), and its specific fragmentation products in the third, or Q3, quadrupole. The specificity arises from the relative unlikelihood that any ion other than the desired one will produce the diagnostic pair of intact Q1 and Q3 fragment ions.
But just because this is unlikely, does not make it impossible. Mark Garner, senior manager for translational sciences at Applied Biosystems, says given the complexity of peptide mixtures in biological samples, even MRM can be fooled. Applied Biosystems/MDS SCIEX has developed a workflow for its 4000 QTRAP hybrid triple-quad/linear ion trap system, called MIDAS (MRM-initiated detection and sequencing), specifically to address this problem (and also to help develop more robust MRM assays in the first place).
MIDAS, Garner says, couples the quantification and specificity of MRM with the surety of peptide sequencing: "When that one Q3 mass comes through, you can trigger a high-sensitivity full MS/MS spectrum. So you can sequence the peptide, and can confirm that the peptide you are quantifying is in fact the peptide you think it is."
Of course, biomarkers need not be peptides; most liquid chromatography-based mass spec systems will work just as well for small molecules and metabolites, and the workflow is basically identical. The difference is in the software.
That's because, while identifying a protein from a peptide sequence is almost routine, the same cannot be said of metabolites, polysaccharides, and so on. Thus, the primary difference between Waters Corp.'s Identity High Definition Proteomics System and its Metabonomics System, both of which are based on the company's new Synapt mass spectrometer, is in the software used to interpret the data.
Still, for small molecules and glycans at least, software can only get you so far. Says Keith Waddell, LC/MS applications solutions manager at Agilent Technologies, "there's no software from an MS vendor to sequence glycans. You have to analyze MS/MS spectra manually to figure out what the likely sequence is."
