A team led by scientists at the Fred Hutchinson Cancer Research Center has identified plasma protein profiles for several molecularly distinct forms of lung cancer.
The study, which was published in the current issue of Cancer Cell, found 39 proteins present at elevated levels in mice with lung adenocarcinoma, as well as protein signatures linked to EGFR and Titf1/Nkx2-1, a known lineage survival oncogene in lung cancer.
The researchers also measured five of the identified proteins in human samples from newly diagnosed and pre-diagnosis lung cancer patients, observing concordance with their behavior in the mouse models, which, the authors noted, suggests their potential usefulness as early detection biomarkers.
Beyond its findings, the study is significant for the biomarker discovery and validation workflow that it lays out, said George Mason University researcher Emanuel Petricoin, author of a commentary in Cancer Cell accompanying the paper.
Calling it "a showcase example" of how protein biomarker work would ideally be done, Petricoin – who was not part of the Hutch project – told ProteoMonitor that the study "is one of the first that really shows what could be a roadmap for other efforts."
In particular, he noted, the combination of using a series of well-studied mouse models to establish the biochemical underpinnings and specificity of the identified biomarkers along with validation in samples from well-characterized cohorts "addressed a lot of the issues that [researchers] have been facing and saying we need to overcome for a number of years."
"It's not like our biomarker discovery efforts are just pumping out [US Food & Drug Administration]-approved biomarkers left and right, Petricoin said. "Many people who have been working in the area for a decade may have never actually had a biomarker make it all the way to FDA. So, it's tough, and what [the study] did was try to address some of the critical bottlenecks."
In an e-mail to ProteoMonitor, Samir Hanash, Fred Hutchinson researcher and leader of the study, seconded Petricoin's assessment, saying that "perhaps the novelty in the paper is in the experimental design." He noted the researchers' use of extensive fractionation of their samples to achieve deep proteome analysis as well as the use of a large number of mouse models to identify signatures specific to lung cancer and the integration of that information with additional data from mouse tumors and human cancer cell lines.
"Although rather cumbersome and labor intensive, actually such a strategy enhances the chances of success in the validation process," he said.
For discovery, the researchers collected plasma from three mouse models of lung adenocarcinoma – Lung-EGFR, Lung-Kras, and Lung-Urethane – as well as a small-cell lung cancer mouse model and mouse models for breast, prostate, colon, ovarian, and pancreatic cancer, quantifying the relative concentrations of proteins in cases and age-matched littermate controls via mass spectrometry on Thermo Scientific LTQ-Orbitrap and LTQ-FT instruments.
They measured the plasma proteomes of two mouse models with inflammatory disease, as well, to determine which of the observed changes were due to general inflammatory processes as opposed to the cancers themselves.
To identify proteins specific for adenocarcinoma, they then analyzed the quantitation data for proteins elevated in at least two of the three adenocarcinoma models but not in the other mouse models, finding 13 such proteins. They also found three additional proteins that were elevated in both the adenocarcinoma and SCLC models, suggesting that they might be linked to lung cancer more broadly, and 16 proteins with altered expression across several cancer models but not in the confounder models, suggesting a link to epithelial tumors in general.
The researchers followed this work by analyzing the proteomes of 21 human lung adenocarcinoma cell lines to determine which of the proteins found elevated in the mouse adenocarcinoma models are present in lung cancer cells and could be released into the extracellular space. They identified in the conditioned media 25 of the 39 proteins that were elevated in the mouse models as well as an overlapping 26 proteins in the cell surface compartment of the tumor cell lines, suggesting that tumor cells were likely contributors to the increased levels of these proteins in the mouse plasma.
To determine the affect of drug treatment on protein levels, the researchers compared plasma profiles in Lung-EGFR before and after treatment with erlotinib, a kinase inhibitor targeting EGFR signaling. The differentially expressed proteins, they found, returned to baseline upon treatment, demonstrating that their abundance reflects tumor progression and regression.
They investigated the proteins' potential usefulness in humans by measuring five of them – SFTPB, WFDC2, ANGPTL3, and EGFR for NSCLC cases and ROBO1 for SCLC cases – via ELISA using samples from the Carotene and Retinol Efficacy Trial. The NSCLC work examined 28 newly diagnosed NSCLC cases and 39 controls matched for age, sex, smoking status, ethnicity, and plasma collection protocol as well as plasma collected from 26 subjects up to 11 months prior to their diagnosis with NSCLC and 26 matched controls who remained cancer free over a four-year follow-up period. The four-protein panel yielded an AUC of .882 for the newly diagnosed set and .808 for the pre-diagnosis set. The SCLC study compared 10 SCLC cases to 39 controls.
The study, Petricoin said, "checked off many of the boxes that a lot of us would hope to be checked off," demonstrating, in the process, the level of rigor and resources that could be required to move protein biomarkers into the clinic. As Petricoin noted, in addition to the Hutch team led by Hanash, the study's co-authors included researchers like Nobel laureate and National Cancer Institute director Harold Varmus, who, he said, "have great expertise with these specific mouse model systems."
"It isn't something that just anybody could do," Petricoin said. "I think that's an important aspect for anyone thinking about getting into biomarker discovery. It takes a village. It's not something that you can just pull off a shelf and do. It speaks to the collaborative consortium-based efforts that you need to happen in biomarker research. It's very difficult for laboratories to have all the expertise in one lab to do everything."
The key to protein biomarker studies, Hanash said, "is in the experimental design, depth of quantitative studies, integration with other data sets, and, perhaps most importantly, deriving a mechanistic understanding of why this or that protein is considered as a potential marker – not just reliance on statistical analysis."
Determining whether proteins identified by the Hutch study will, in fact, be useful for early detection of lung cancer will require additional studies, Petricoin said, noting that any markers that emerge from the effort will likely be most useful in combination with CT screens. Given that, he said, it will be necessary to determine their performance in that context, which the Cancer Cell paper didn't do.
Protein markers for use in combination with CT scans are particularly desirable in light of recently released results from the National Lung Screening Trial – a National Cancer Institute study involving more than 53,000 current and former heavy smokers – that found that CT scans reduced lung cancer mortality by 20 percent compared to standard chest X-rays. Given these results, CT scans will likely become a more common procedure moving forward, heightening the need for non-invasive methods – like protein biomarker-based tests – for determining the status of nodules detected via the scans.
Validation studies for some of the proteins identified in the paper are currently underway to determine the additive value of a biomarker test as a companion to CT scans, Hanash said.
In addition to Hanash's team, several other parties have been working on lung cancer biomarkers that could be used for this purpose, including researchers at Celera, which is now a part of Quest Diagnostics, and Somalogic (PM 4/22/2011).
Have topics you'd like to see covered in ProteoMonitor? Contact the editor at abonislawski [at] genomeweb [.] com.