A multi-institutional team of researchers has used genetically engineered mouse models to validate a method to quantitatively analyze proteins in the plasma proteome at the picomolar and even sub-picomolar level, according to a recent paper.
The goal of their research is to identify potential disease biomarkers, therefore addressing a major bottleneck in protein biomarker research.
In their study, which appears in the June 10 online version of PLoS Medicine, the researchers describe a GEM that they used to identify a panel of biomarkers linked to early-stage pancreatic cancer.
Their work also corroborates for the first time protein changes associated with early-stage pancreatic tumor development in GEMS with early-stage, presymptomatic pancreatic cancer in humans. But for proteomics the message is that, contrary to belief, technology now exists to perform an in-depth analysis of plasma in a quantitative, unbiased manner, according to senior study author Sam Hanash, head of the Molecular Diagnostics Program at the Fred Hutchinson Cancer Research Center.
Using their technique on the GEM, Hanash and his colleagues developed a five-biomarker panel, which potentially could be combined with tests based on the CA19.9 glycoprotein to detect early-stage, pre-symptomatic pancreatic cancer and to differentiate the disease from pancreatitis.
CA19.9 is currently in clinical use as a marker for pancreatic cancer. While 80 percent of patients newly diagnosed with the disease have elevated levels of CA 19.9, it is not linked to early-stage onset of the disease.
The technique Hanash and his colleagues developed is also being used to explore biomarkers for other cancers including lung and breast, he said, and has applications for non-cancer diseases such as neurological disorders, cardiovascular disease, diabetes and other illnesses in which mouse models are used as a proxy for human subjects.
“The findings from this paper go beyond mouse models, they go beyond cancer, and … basically validate the utility of proteomics to search for markers in plasma,” Hanash said. “What this study demonstrates is that we’re able to assemble technology that allows us to de-complex the plasma proteome to less complex fractions and analyze each of those fractions in depth and be able to find potential markers.”
The conventional approach to biomarker discovery in plasma is to take the plasma, digest it, and feed it onto a mass spectrometer. This, Hanash said, “is completely overwhelming to the mass spectrometer” and results in only the most abundant proteins being detected.
“What this study demonstrates is that we’re able to assemble technology that allows us to de-complex the plasma proteome to less complex fractions and analyze each of those fractions in depth and be able to find potential markers.”
The key to his approach, he said, was that rather than subjecting plasma directly to mass spectrometry, he and his fellow researchers fractionated the plasma “extensively” and analyzed each fraction by itself. They prepared more than 100 fractions and analyzed each one by high resolution LC-MS/MS, resulting in more than 100 gigabytes of data.
“That contributed to a substantial depth of analysis,” Hanash said. To be sure, their approach is more time-consuming and demands greater resources than the conventional approach, but in the end more and better data will be generated, Hanash said.
“So analyzing maybe fewer plasmas and going much deeper has a much greater utility than superficial analyses of many more specimens,” he said.
For a laboratory with the requisite resources in instruments, personnel, and funds, reproducing their method will take weeks, he said. “The approach that we followed cannot be described as a quick approach in terms of the steps that are involved. It’s a complex approach.”
While the method may at this point not be practical for all laboratories, “I’m hoping that [as] with so many aspects of technology that seem to be complex and take a lot of time today, tomorrow with advancements … you can do it much faster,” said Hanash. “I have reasons to be optimistic that in-depth analysis to the extent that we have accomplished, instead of taking months to do like it took us may just take days.”
Commercial vendors as well as academic researchers including those at the Hutch are working on ways to do fractionation in a multiplex fashion, which would simplify and speed up the approach the authors developed, Hanash added. Further work, however, needs to be done to speed up the bioinformatics part of the analysis.
“The challenge will still remain [that] if you are generating a tremendous amount of data points, hundreds of gigabytes and terabytes of data, then the informatics has to be just as fast,” he said.
A GEM of a Mouse Model
The PLoS Medicine article is notable also for its validation of GEMs as a strategy for cancer biomarker research. In it Hanash and his colleagues said that one goal of cancer biomarker research is to develop a non-invasive diagnostic for early cancer detection.
Genomic analyses of human and mouse cancer indicate a “significant concordance in chromosomal aberrations and expression profiles,” making them useful stand-ins for humans. GEMs, they said, reduce biological and non-biological heterogeneity through defined stages of tumor development and standardized blood sampling and other conditions.
In theory GEM blood samples should yield robust proxy biomarkers for human cancer, and recent mouse models have shown “broadly conserved tumor biology and molecular circuitry similar to” human pancreatic ductal adenocarcinoma, the most common form of pancreatic cancer, the authors said. However, the theory had not been tested and no markers using existing models and technologies have been demonstrated.
For their work, the researchers chose pancreatic cancer because it is one of the most aggressive and deadly. According to the National Cancer Institute, pancreatic cancer is the fourth-leading cause of cancer deaths in the US. In 2008 there will be 37,680 new cases of the disease in the country, and an anticipated 34,290 deaths.
They also chose pancreatic cancer because prior work has indicated that genetically engineered mice “closely recapitulate the histopathogenesis of the human disease,” they write in the article.
Hanash and his colleagues obtained PDAC and pancreatic intraepithelial neoplasias and their respective controls from male mice. The three most abundant proteins, albumin, IgG, and transferring, were immunodepleted. Protein fractionation resulted in 720 individual fractions, which were then tryptically digested in solution. The digests were analyzed with a Thermo Fisher Scientific LTQ-FT mass spectrometer coupled to a nano-Acquity chromatography system made by Waters.
Data was processed by the Computational Proteomics Analysis System with searches done “considering cysteine alkylation with the light form of acrylamide as a fixed modification and heavy form of acrylamide as a variable modification,” the authors said. Search results were analyzed using the PeptideProphet and ProteinProphet programs.
For quantitative analysis, they differentially labeled peptides containing cysteine with acrylamide isotopes. Using a script developed in-house dubbed “Q3” they obtained the relative quantitation of each pair of peptides identified by tandem mass spectrometry containing cysteine residues. Proteins that were presented as “cancer only” had only detected peptides labeled with the heavy form of acrylamide.
LC-MS analysis yielded 1,095 candidate proteins from the PanIn and PDAC experiments, and mRNA analysis of pancreatic tissue yielded an additional 347 proteins. The quantitative analysis identified 621 proteins of which 165 were found to be up-regulated in cancer samples compared to controls.
Using more specific criteria, they further reduced the number of candidate proteins to 45 proteins.
Hanash and his colleagues then compared the up-regulated mouse proteins with serum from 30 individuals with a confirmed diagnosis of pancreatic adenocarcinoma. They also chose five proteins, LCN2, TIMP1, REG1A, REG3, and IGFBP4, based on their increased level of at an early stage of tumor development in the mouse and tested them in a blind study in 26 humans from the Carotene and Retinol Efficacy Trial cohort study involving 18,314 individuals with increased risk for cancer.
Based on their data, he and his co-authors concluded that a plasma proteomic analysis of GEM models of cancer is a “useful” strategy for identifying biomarkers for human cancer. “The strong concordance between mouse and human pancreatic cancer in both tissue and circulating markers is striking,” the authors say.
As they build on the research, the researchers will be developing high-throughput assays for additional candidate markets that may be identified. Additional validation studies will also be done looking at specific applications such as implementing a panel-based test that can distinguish pancreatic cancer from pancreatitis, and the utility of a panel for the early detection of pancreatic cancer.
An early-detection test would be of particular use, according to Hanash and his co-authors, because of a five-year survival rate of just 3 percent. Because of “limitations in diagnostic methods and a lack of specific symptoms at an early stage,” the disease is usually diagnosed at late stages, Hanash and his colleagues said in the article.
They also hope to use not only biomarkers that they have discovered but also those identified by other researchers, Hanash said.