A team led by researchers at George Mason University has completed a phosphoproteomic analysis of the National Cancer Institute-60, a set of 60 human cancer cell lines.
The analysis provides a means of linking protein pathway activation information to other data generated via study of the NCI-60 set, including data from previous omics experiments and drug-sensitivity studies, said Emanuel Petricoin, co-director of GMU's Center for Applied Proteomics and Molecular Medicine and one of the leaders of the effort.
The team's results also suggest limitations to genomic approaches to guiding cancer therapies, Petricoin told ProteoMonitor, noting that they found essentially no concordance between cancer pathway activation and genomic data, such as gene expression and DNA mutation profiles.
In the study, published this month in Molecular Cancer Research, the researchers used reverse phase protein arrays to measure phosphorylation of 194 proteins and isoforms in the 60 cell lines making up the NCI-60 set.
They then fit this signaling data into models of six key cancer signaling pathways: AKT, mTOR, EGFR, IGF-1R, integrin, and apoptosis signaling, calculating for each pathway an activation score consisting essentially of aggregations of the activation levels of the individual proteins in a given pathway. Using these scores, the researchers were able to investigate links between protein pathway activation and other measures like RNA, DNA, metabolomic, and drug-sensitivity profiles.
Established in 1990, the NCI-60 consists of human cell lines representing leukemia, melanoma, and cancers of the lung, colon, brain, ovary, breast, prostate, and kidney. Nearly 300 research groups have performed molecular characterizations of the cell set, with more than 300,000 measurements taken.
"The beauty of the NCI-60 is that they've measured so many things," Petricoin said. "So this allows us to ask the question: If we measure the activation of a signaling pathway at a network level, what is linked with that?"
Somewhat surprisingly, Petricoin and his colleagues found that while their pathway activation scores correlated significantly with response to drugs targeting those pathways and to a lesser extent with microRNA data, the activation scores didn't correlate with genomic and transcriptomic profiles of the cell lines.
"There was very little interlinkage between the genomics side of the story and the real functional [protein signaling] end of the biochemistry," he said. "And that's a very interesting finding because it says that if you want to know what signaling pathways are activated, you have to measure [the protein phosphorylation levels] – you can't infer pathway activation from gene expression or DNA mutational analysis alone."
This, Petricoin said, suggests that proteomics, and phosphoproteomics specifically, could prove more effective than genetics for guiding use of therapies targeting these cancer signaling pathways – despite the fact that genetic approaches currently dominate companion diagnostic research.
While Petricoin and his colleagues found no correlation between the genomic data and their pathway activation scores, genomic analyses of the NCI-60 lines have established correlations between genetic data and drug response. However, Petricoin said, pathway activation has the advantage as a biomarker of being directly tied to mechanisms of drug response.
Researchers "are finding drug-sensitivity relationships with genomic data, no doubt about it," he said. "But it's not always completely clear why those sets of genes would be predictive for why a drug works or not, and that can become difficult to validate."
This notion is not a new one for Petricoin, who, for years now, has been investigating and promoting phosphoproteomics as a technique for personalizing cancer therapy. In addition to his position at GMU, he is a founder and chairman of the scientific advisory board at Theranostics Health, a phosphoproteomics-based diagnostics firm that next month will launch its first test, a 14-protein assay for profiling activation of the HER signaling network (PM 5/3/2013). The assay is intended as a supplement to conventional HER2 testing for guiding therapy in breast cancer patients.
The utility of phosphoproteomics for guiding drug therapy "is something that we postulated," Petricoin said. "But the NCI-60 dataset really allowed us to investigate it methodically."
The MCR study isn't the first to do a phosphoproteomic analysis of the NCI-60 line, though it is the most extensive to date. In 2010 a team of researchers from MD Anderson Cancer Center published a paper in Molecular Cancer Therapeutics using reverse phase protein arrays to profile 222 unique protein features including a number of phosphoproteins.
In an email to ProteoMonitor, MD Anderson researcher Gordon Mills, an author on the 2010 MCT paper, said that the MCR study – of which he was not a participant – "greatly extend[s] the earlier analysis of the same cell line set" done by he and his colleagues.
The GMU "team is to be congratulated for making a high-quality data set of functional proteomics analysis for the NCI-60 cell line set available," Mills said, adding that the researchers' inclusion of an "accompanying list of validated antibodies will be of utility to both the RPPA community and the cancer research community in general."
"The new integrative data has the potential to greatly facilitate the identification of biomarkers of benefit for novel and existing therapeutic approaches," he said.
Indeed, Petricoin said that he is now working to apply the findings of the MCR study to his work in clinical studies like the I-SPY 2 breast cancer trial and collaborations with pharmaceutical companies.
"We're going to be cross-walking the NCI-60 drug-sensitivity and [pathway-activation] relationships to our clinical trials and partnerships with pharma where we are actually treating patients with drugs to see if they respond," he said.
Petricoin noted that he was particularly interested in looking further at the integrin pathway as well as the AKT, mTOR, and IGFR pathways. "Those are areas where we found some really interesting drug-sensitivity relationships that we want to investigate in clinical trials," he said.
Of course, identifying these relationships in cell lines is considerably more straightforward than doing so in actual clinical samples. On the one hand, as Petricoin noted, this is part of the NCI-60 set's appeal.
"You don't have to control for how the tissue is sent out, for handling the tissue correctly," he said. "It allows us to clean all that up and just ask the question: If we measure activation of a signaling pathway at a network level, what is that linked with?"
On the other hand, as Petricoin recently observed in a separate interview with ProteoMonitor, the labile nature of many phosphoproteins and the difficulty of controlling for such pre-analytical variation in a clinical environment is perhaps a key reason why phosphoproteomics-based personalized medicine – despite its potential – lags well behind its genomic counterpart (PM 5/10/2013).
"In some ways it may be hurting proteomics because there are people on the other side of the coin who say that the only thing that can be believed is DNA analysis because there's no impact of degradation," he said. "I think that's wrong, but it does permeate, especially with all the genomic mutational analysis being done. And I think this is why a lot of times pharma is more hesitant to bring phosphoprotein markers into companion diagnostics – because they feel it's too difficult."
"So, you have all the caveats [regarding the NCI-60 work] – this is a cell line series, not clinical tissue specimens," Petricoin said. "But they still give you a great sandbox to do this cellular biochemical analysis."