Biomarker development, from detection to validation, has become a cornerstone in the field of quantitative proteomics. The ability to decipher the presence of signature peptides as surrogates for low- to mid-abundance proteins indicative of disease states carries enormous inherent potential for cancer diagnostics, prognostics, and therapeutics. Consider the deluge of papers published in the last 12 months suggesting candidate biomarkers: a PubMed search for papers associated with the term "potential biomarker" in 2009 yields more than 7,800 results. Just less than half of those results are also categorized by the search term "cancer."
The methods through which proteins of interest are detected and quantified — and potential biomarkers validated — are constantly evolving to meet the requirements of both basic and clinical applicability. From traditional immunohistochemical approaches to recent advances in mass spectrometry-based techniques, researchers are developing standardized protocols to produce higher throughput, specificity, and multiplexing capabilities, all the while saving time and cutting costs.
It's a lofty task representing millions of dollars in federal funding supplied through thousands of grants. Several consortia and partnerships have been established to accomplish these goals; the principal international proteomics research establishment, the Human Proteome Organization, has dedicated an entire initiative to biomarkers for disease, with a sub-initiative focused solely on cancer. Lee Hartwell, the president of the Fred Hutchinson Cancer Research Center, established the International Biomarker Research Consortium in 2004 after a discussion with Reudi Aebersold at the Institute for Systems Biology, observing the early biomarker successes of colleagues at the Fred Hutchinson Cancer Research Center in Seattle, and meetings with researrchers in his travels along the Pacific Rim. The IBRC currently represents proteomics researchers in Korea, Taiwan, China, and the US; there are plans to establish connections within Singapore, Canada, Australia, and France as well.
But researchers say that while they have seen improvements in mass spec-based quantitative proteomics, there is still much progress to be made.
"Methods exist for the generation of large numbers of candidate biomarkers," says Shalini Makawita, a graduate student in Eleftherious Diamandis' lab at the University of Toronto. "However, comparable methods for the verification of generated candidates are lacking. … As a result, there is a lag in moving candidates from the bench to the clinic."
Amanda Paulovich, director of the Early Detection Initiative at Fred Hutch and a member of the IBRC steering committee, is funded through two National Institutes of Health grants for biomarker discovery using quantitative proteomics. The grants are worth more than $4.8 million combined, and one of them is a highly selective Grand Opportunity award. She uses multiple reaction monitoring mass spec in her lab to distinguish signature peptides in human plasma.
Indeed, MRM-mass spec is now viewed as a highly specific alternative to the gold standard ELISA approach in protein quantification. Neil Kitteringham and colleagues at the University of Liverpool and Applied Biosystems describe MRM-mass spec as "an essential element in the quantitative proteomics toolbox" in their 2009 Journal of Chromatography review.
"I think the principle motivations for pursuing the mass spec-based approaches are that it's much easier to create the assays … and the specificity of the assay is much greater with mass spec because you can achieve absolute structural specificity for the peptide that you're measuring," says Leigh Anderson, founder of the Plasma Proteome Institute. "[Another] major advantage is that it's very easy to multiplex — in other words, measure multiple targets — in the mass spec assay at one time."
Michael Pisano, CEO of NextGen Sciences, an Ann Arbor, Mich.-based firm devoted to protein biomarker discovery and validation services for clients in the pharmaceutical/biotechnology industries as well as government and academic labs, says, "MRM assays allow for rapid development." He notes that it "certainly buys time compared to ELISA."
The Hutch's Paulovich says that in addition to time, cost is an important factor to consider in choosing a quantification platform. "[There are] a lot of really good things about ELISAs — I don't mean to trash on those," she says. "The problem is it takes over a year to develop one ELISA and it costs hundreds of thousands to several million dollars, depending on what you're going to use it for." Faced with thousands of candidate biomarkers, she adds, the cost of running an ELISA for each would be impractical.
It's also nearly impossible to screen so many candidates, according to Daniel Martin at the Institute for Systems Biology, considering that running an ELISA requires a highly specific antibody to each protein of interest and "there's just no throughput there with current technologies."
MRM-mass spec is not a new concept; it first appeared in the literature more than 30 years ago. The term was coined following the publication of a 1977 paper by Baty and Robinson, who used the technique to observe the levels of an anti-epileptic drug in plasma. There are several mass spec platforms amenable to MRM analysis, and the use of triple-quadrupole spectrometers has essentially become the standard for specific protein assays in biomarker research.
"This is a different type of mass spectrometer than has typically been used in proteomics," Anderson says. "It's typically been used in pharmaceutical industries to quantitate small molecules like drugs. We use triple-quadrupole mass spectrometers — [we're] essentially adapting a very well-understood type of mass spectrometric instrument for the measurement of peptides, which, as it turns out, are not that different from drugs or other small molecules in terms of measurement."
As with all technologies, there has been — and will continue to be — room for optimization. Researchers have devised a variety of MRM-mass spec-based approaches suited to the challenges they've faced in their own research. While they diverge somewhat in their methods, researchers agree that no one-size-fits all platform exists and it may be quite some time before it does.
Enter the acronyms
Ian Blair at the University of Pennsylvania uses stable isotope labeling by amino acids in cell culture, or SILAC, in his analyses of proteins in plasma secreted by colon cancer cells. He emphasizes that the process of searching for potential biomarkers can be convoluted by the distracting appearance of high--abundance, naturally occurring proteins in serum.
"Immunoaffinity purification methodology can be employed to remove many of the high-abundance proteins that are present," Blair says. "But there is always the concern that important secreted proteins might bind to the high-abundance proteins and they would also be removed."
SILAC protocols, originally proposed by Ong and Mann in 2006, call for two groups of cells to be cultured in separate media — one culture is exposed to a 13C-labeled amino acid, another to a 15N-labeled one, and the last is exposed to only unlabeled essential amino acids. Upon incorporating the cultures and performing mass spec, Blair says, their differential expression can be compared. From there, Blair and his colleagues generate a stable isotope labeled proteome (SILAP) standard to act as a carrier to prevent losses of low-abundance proteins that bind to high-abundance proteins in plasma.
"We currently have 15 SILAC-labeled cell lines that have provided more than 5,000 labeled proteins for use as SILAP standards," Blair says. "It is relatively easy for individual laboratories to generate their own."
In Seattle, the ISB's Martin is identifying potential biomarkers for prostate cancer. He and his collaborators are using QCAT, or synthetic proteins, and xenografting with mice to do so. "With proteomics, it's possible to distinguish the human peptides from the mouse peptides so it reduces the level of difficulty of detecting the cancer-associated peptides," Martin says. QCAT takes the specificity one step further, he says; by creating a QCAT protein with specific peptides in silico and expressing the unnatural product in E. coli or yeast, one can develop a concatemer protein and purify it.
"When it boils down to it, ideally what you want is a heavy peptide," Martin says. "And QCAT is one way of doing it. Or you just get out your checkbook and companies will synthesize them for you."
Martin explains that there are pros and cons to purchasing synthetic proteins versus developing your own in-house. "If you want 96 peptides, you can call up a company and they'll charge you anywhere from $50 to $100 a pop," he says. To produce QCATs in a lab requires time to culture a substrate, synthesize and clone genes, and test for specificity. "In the end, I don't really know how [the costs] stack up, but we're working with QCAT right now just because we like it," Martin says, adding, "It's kind of fun to play with."
Leigh Anderson and the Hutch's Paulovich are pioneering work on what they say is the most promising MRM-mass spec-based method yet.
"The essence of the SISCAPA method is that we decide in advance which peptides we want to measure and then we make a specific antibody against that peptide … to fish it out of a complicated digest" like that of human plasma, Anderson says. He and colleagues introduced SISCAPA — stable isotope standards and capture by anti-peptide antibodies — in 2003 in a highly cited paper in the Journal of Proteome Research. Anderson says SISCAPA enriches the peptide of interest by 100,000-fold allowing researchers to fish the target peptide out of a much larger sample than was possible to load directly onto the spectrometer before.
Paulovich, who has collaborated with Anderson since 2003, is hopeful that SISCAPA might translate to biomarker success in the clinic. "The most exciting [part] is the ability to have a simple workflow for sample processing that still gives you high-sensitivity measurements," she says. "It's really quite dramatic, the success we've had in generating these antibody-based assays and how much [SISCAPA] simplifies the workflow needed to process a plasma sample and make a highly precise and sensitive measurement."
"That, being coupled to a targeted MRM mode of mass spec, provides a very powerful combination of tools, which will minimally provide a bridging technology for biomarker verification and, in the best case scenario, these could end up being the end-clinical assay[s]," Paulovich says.
She and Anderson are both investigators in the National Cancer Institute's Clinical Proteomic Technology Assessment for Cancer initiative that aims to advance the study of proteins in personalized medicine. Anderson says that a current CPTAC project uses the SISCAPA methodology in an effort to develop quantitative assays for human proteins.
"We're developing about 200 assays for 200 targets which [covers] about 1 percent of the human proteome," Anderson says, adding that plans for the second phase of this project are to build 2,000 such assays, covering approximately 10 percent of all proteins. "A final effort would be to create 20,000 assays, covering all of them," he says, acknowledging that only with such a suite of assays could researchers systematically discover and qualify biomarkers in a high-throughput manner.
Anderson adds that, aside from developing technologies, he and his colleagues are also interested in "trying to get a complete picture of where we are in the protein diagnostics business ... which, by the way, is not in a good place."
He says that although hundreds of candidate biomarkers are elucidated every year, the rate at which they are approved for clinical use is alarmingly disproportionate.
"When you look at what the FDA has approved in terms of biomarkers, the rate is one and a half new proteins per year — flat — for the last 15 years," he says. "There's been no improvement in that rate and in my mind that pretty much proves that there's something wrong in the biomarker pipeline."
Bottlenecks in the pipeline
The University of Toronto's Makawita, along with principal investigator Diamandis, recently published a review article in Clinical Chemistry outlining the clog in the pipeline. Makawita says that while improvements to MRM-mass spec-based methods provide a desirable alternative to ELISA-based quantification, the inability to verify candidate biomarkers is "the greatest bottleneck in the discovery pipeline."
She says that while the past few years have seen improvements in the multiplexing capabilities of MRM assays and target enrichment through immuno-extraction, depletion of high-abundance proteins, and serum fractionation to reduce sample complexity, "unfortunately, we are not at the stage where candidate verification technologies can keep up with discovery-phase studies."
ISB's Martin echoes that sentiment, adding that several potential biomarkers are incorrectly classified — even when all precautionary steps are taken to prevent misidentification. "If you do discovery, like we or other people have done, where you take serum and look to see which peptides are there in disease and in health, and you say 'Well, I see this peptide, it indicated that this protein is a biomarker,'" Martin says. "This isn't necessarily true."
A grand opportunity
The Hutch's Paulovich says that the bottleneck observed in the validation pipeline can be attributed to two things: a lack of high-quality biospecimens and the need for more sensitive quantitative assays. To that end, she's poised to push the limits imposed by current technologies in a project funded through the American Recovery and Reinvestment Act, along with co-principal investigator Steve Carr at the Broad Institute, and Anderson as a consultant.
Paulovich and Carr have been awarded a total of $4.8 million over two years to assess the feasibility of scaling assays for every human protein. In a pilot study, Paulovich and her colleagues are examining proteins expressed in breast cancer. They are "using genomic data to help prioritize proteins that we can detect with the mass spectrometer and that appear as though they may have some clinical significance down the road," she says. They plan to build 400 assays to 200 of those proteins, "and multiplex them as high as we possibly can."
The outcomes of these trials are likely to indicate whether there is a future for MRM-mass spec in the clinical setting. "Part of the demonstrating the robustness of the technology is testing what the difficulties [might be] in scaling [and] developing assays for the proteome," Paulovich says.
At ISB, Martin is investigating the use of affinity capture methods to sort out peptides of interest post-MRM. And the University of Pennsylvania's Blair is examining the possibility of analyzing up to 250 proteins on 1,000 timed MRM transitions.
Makawita maintains an optimistic view in her study of biomarkers for pancreatic cancer using MRM, saying the technique's "utility will likely increase as improvements to instrumentation are made, and with [the] implementation of online fractionation techniques and increased automation of sample processing."