A team led by researchers at the Swiss Federal Institute of Technology Zurich and the Institute for Systems Biology has developed selected-reaction monitoring mass spec assays for more than 1,000 cancer related proteins in plasma and urine.
The researchers also demonstrated the potential of SRM-MS to efficiently verify candidate protein biomarkers, using the technique to evaluate the usefulness of 30 target proteins added to the existing five-protein panel of Vermillion's OVA1 ovarian cancer diagnostic.
Detailed in a paper published this week in Science Translational Medicine, the work provides a resource for researchers interested in using SRM-MS in biomarker verification and validation studies and offers a framework for integrating genomics-based biomarker discovery to improve and streamline clinical proteomics efforts, Ruedi Aebersold, an ETH Zurich researcher and one of the leaders of the study, told ProteoMonitor.
The researchers set out to build a library of SRM assays in urine and plasma to all proteins that have been implicated in cancer. They based this library on a list of 1,261 cancer-related proteins originally compiled by Malu Polanski and Leigh Anderson. Of these 1,261 proteins, they selected 1,130 that, they wrote "could be unambiguously associated with UniProt identifiers." They also added any US Food and Drug Administration-approved biomarkers not included in the original list, for a total of 1,172 proteins.
Using an AB Sciex 5500 QTRAP instrument, the researchers developed SRM assays for each of the 1,172 proteins in both urine and plasma. They then applied the assays to actual depleted plasma and urine samples, detecting in plasma 302 peptides corresponding to 182 proteins and in urine 661 peptides corresponding to 408 for a total of 473 unique proteins. They detected proteins across five orders of magnitude and at limits of detection below 10 nanograms per mL.
The assays are collected at the PeptideAtlas SRM Experiment Library, Passel, providing a resource for researchers looking to do targeted protein biomarker analysis and validation, Aebersold said. The intention, he noted, is that such a resource will help ease the gap between biomarker discovery and translation that has long been an issue in proteomics (PM 2/11/2011).
"There are a large number of papers that have suggested biomarkers … but it usually stops there because it [requires] a huge amount of effort to test these potential markers in large cohorts," Aebersold said. "Part of the limitation, of course, is that one needs suitable samples. But the other part is that one needs suitable assays to test the proteins and quantify them in these samples. We wanted to demonstrate that we could take a large number of potential biomarkers and very quickly use them to generate specific mass spec assays and then use them in a variety of samples."
He added that another intent is to "make these data accessible [via Passel], so that others would be able to test in their specific disease if any of these biomarkers actually have any merit."
To demonstrate such a disease-specific application, the researchers used the SRM assays they generated to compare the levels of 34 proteins in the plasma of 68 ovarian cancer patients and 16 patients with benign pelvic masses. These 34 proteins included four proteins from Vermillion's OVA1 ovarian cancer diagnostic – which is designed to distinguish between benign and malignant ovarian tumors – along with proteins that researchers had previously identified as potential ovarian cancer biomarkers or predicted to be functionally related to genes altered in ovarian cancer.
Their analysis found for the four OVA1 proteins significant differences in expression that tracked with differences identified in previous immunoassay-based measurements of these proteins. It also identified 15 other proteins that showed significant abundance change between the benign and malignant patient sets, including three that had been previously identified as potential ovarian cancer markers; nine that network analyses had identified as linked to genes altered in ovarian cancer; and three that were chosen based on both literature and network analyses.
The results, Aebersold said, suggest that rather than continuing to focus heavily on shotgun proteomics-based biomarker discovery, researchers might have more success using targeted proteomics to verify and validate candidates culled from genomics data and network-based analyses – a notion that he has promoted in a number of forums (PM 2/11/2011 and PM 1/20/2012 and PM 2/17/2012) in recent years.
"I'm convinced that a lot of useful information is coming out of transcriptomics and genomics and that one would possibly be able to skip over the rather tedious proteomics discovery [process]," he said. "I'm not saying that it isn't a good strategy to also do proteomic measurements, but I think increasingly there is a wealth of information coming from genomics."
"That's what we wanted to demonstrate with this [ovarian cancer] example," he added. "That we can use genomic data and generate potential biomarkers by filtering and rapidly testing [via SRM]."
Aebersold said that, in fact, he was surprised at how well the genomic-based discovery effort worked in the ovarian cancer example, noting that he hadn't expected such tight overlap between the genomic and proteomic markers. He added that this suggests that not only can genomics inform protein biomarker discovery, but that proteomics might aid in genomic marker discovery, as well.
"Genomics has basically a problem of richness," he said. "[Researchers] generate hundreds of cancer genome sequences, compare each one to healthy tissue from the same person, and in each case there are hundreds if not thousands of genetic variants. They all have probably some merit, but not equal merit. And I think that by going into these lesions and trying to organize them on the level of networks and then testing whether they have any weight by proteomics is a very promising way to go."
Mary Lopez, director of Thermo Fisher Scientific's Biomarkers Research Initiatives in Mass Spectrometry Center said the paper provided further confirmation of SRM-MS's usefulness in biomarker research and potential applicability as a clinical technique. In particular, she noted as "very nice" the study's global integration of spectral library data with pathway information.
However, she told ProteoMonitor, the assays as presented appear more suited to a research than clinical environment. "The caveat with this paper is that it is very clearly in the research bin in the sense that it’s a nanoflow-based assay, and it depends on things like depletion that add a huge source of irreproducibility and time," she said.
Challenges also remain in terms of achieving the levels of sensitivity required for measuring low-abundance protein markers via mass spec. Aebersold noted that for the foreseeable future some form of enrichment will be required to enable SRM-based quantitation of low-abundance proteins.
Such enrichment steps present their own challenges in terms of throughput and reproducibility. Progress, though, is being made on these fronts (PM 6/8/2012), and, said Aebersold, techniques like the immunoenrichment method SISCAPA and his lab's glycoprotein-based enrichment technique are "fairly mature" and can offer robust and reproducible results.
Perhaps a more significant challenge to his goal of a well-integrated genomic-proteomic biomarker pipeline is selling genomics researchers on the idea. Ideally, Aebersold noted, in such research the same samples would be used for the genomic and proteomic portions of the work. However, achieving such levels of cooperation has thus far been difficult, he said.
The National Cancer Institute has taken steps in this direction through the second stage of its Clinical Proteomic Technologies for Cancer initiative, which is currently ongoing (PM 8/26/2011). In that project, proteomics researchers are doing biomarker discovery in samples that have been characterized genomically by the NCI-funded Cancer Genome Atlas.
Aebersold said that he is currently trying to arrange similarly coordinated efforts in Europe but that "for a while at least it will be an uphill battle."
In general, the cancer genomics community has shown relatively little interest in proteomics, he said. "They are, of course, very much focused on their own progress, which is amazing, and I think also that the substantial progress in proteomics hasn't been widely enough brought to the attention of the cancer genomic community."
"I think it's like anything else," Lopez said. "People are in their specialized fields and they tend not to venture out if they don't have to."
She noted, however, that it depended largely on the individual researchers, citing epigenetics work her group published recently with Buck Institute researcher Victoria Lunyak as an example of a fruitful collaboration between genomics and proteomics.
Lopez added that "in a lot of core facilities as well you're getting a lot of interaction between genomics groups and proteomics groups because they need information from each other to submit grants or write papers."
"I think it's our task to convince [genomics researchers] that there has been a lot of progress [in proteomics] and that there are a lot of synergies," Aebersold said, noting that he has established several small collaborations with genomics researchers with the "intent of maybe generating mutual trust and exchange of information."
"It's natural, of course, that genomics people are not super keen on just sending out samples," he said. "They need to be convinced that the data that come out of [proteomics experiments] are useful, and that is one of the intents of this paper."