Researchers at the Technical University Munich have completed the first global proteomic analysis of the National Cancer Institute's NCI-60 cancer cell line panel.
In the study, detailed in a paper published last week in Cell Reports, the TUM scientists used mass spec to generate quantitative proteome and kinome profiles of the 59 cancer cell lines in the NCI-60 collection and model proteome and drug-resistance profiles for 108 US Food and Drug Administration-approved cancer therapeutics.
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.
As the authors noted, this includes proteomic measurements – particularly protein signaling assays conducted via reverse phase protein array. However, no one had previously performed a broad proteomic analysis of the full cancer cell collection, said Hannes Hahne, a TUM researcher and, with his TUM colleagues Amin Gholami and Zhixiang Wu, co-first author on the paper.
Using one-dimensional PAGE followed by LC-MS/MS, the researchers identified 10,350 proteins including 375 protein kinases across the 59 cell lines. Their data consisted of two levels of mass spec analysis – a set of standard proteomic experiments, which they ran on all 59 lines using a Thermo Fisher Scientific LTQ Orbitrap XL and then a set of "deep" proteomic experiments, in which they used a Thermo Scientific Orbitrap Elite to profile one cell line from each of the nine tissue types – brain, blood and bone marrow, breast, colon, kidney, lung, ovary, prostate, and skin – represented in the NCI-60 panel.
This two-tier approach stemmed from the researchers' acquisition of the Orbitrap Elite late in the project, Hahne told ProteoMonitor.
"We started this project two years ago, and the Elite wasn't available at that time," he said. "But at the end [after obtaining the Elite], we decided that since we had the cell lines clustered by tissue, we would use [the Elite] on one characteristic cell line per tissue to go a bit deeper into the cancer proteome."
The researchers' Orbitrap LTQ analysis of the 59 lines identified a total of 8,113 proteins, while their deep analysis of the nine representative lines using the Orbitrap Elite yielded 8,443 proteins. Of these, 5,578 proteins were identified in at least one cell line from every tissue group, suggesting, the TUM team said, that these proteins can be considered the "core cancer proteome."
Using unsupervised hierarchical clustering, the researchers grouped the cell line proteomes based on their proteomic profiles, finding, in general, that these groupings reflected the cancers' tissues of origin. For instance, seven out of seven colon cancer lines, five out of six leukemia lines, five out of six central nervous system cancer lines, and six out of eight melanoma lines clustered together based on their proteomic data.
Analysis of the kinome data, on the other hand, revealed more heterogeneity with clusters diverging more from the tissues of origin. This, Hahne noted, was to be expected given that kinase expression is known to be very often dysregulated during cancer development.
Indeed, previous proteomic analyses of the NCI-60 using RPPA to measure protein phosphorylation levels across the different lines have found even less clustering based on tissue of origin. For instance, a recent phosphoproteomic analysis of the NCI-60 headed by George Mason University researchers (PM 5/17/2013) found that "similarity [based on site of origin] fell away once you looked at the functional output of the kinase activity," Emanuel Petricoin, co-director of GMU's Center for Applied Proteomics and Molecular Medicine and one of the leaders of that effort, told ProteoMonitor this week.
That could be due to the fact that the TUM team looked at "all the kinases, and not every kinase participates in intracellular signaling and not every kinase is deranged in cancer," said Petricoin, who was not involved in the TUM work. "Because they cast a wider net, that may have in some ways diluted out the specific signaling changes that occur in the cell lines themselves."
He suggested that a potentially interesting follow-up to the TUM analysis might be to look specifically at "subsets of kinases that have been implicated specifically in cancer tumorigenesis," perhaps integrating it with phosphorylation data like that generated by his group's effort.
"I would love to now look at kinase expression levels themselves versus phosphorylation of their substrates," Petricoin said. "How well does the expression of any given kinase correlate with the actual activation of the pathway that kinase resides in?"
The ability to explore such questions using data like that generated by the TUM group is one of the great opportunities afforded by molecular profiling initiatives like the NCI-60 project, he said. "People can just deposit their data and it becomes kind of a sandbox for people to collaborate."
The global proteome and kinome profiles added by the TUM team "represent a great addition" to the NCI-60 data, adding "what has been a missing component" of this collection, Petricoin said.
In addition to generating proteome and kinome profiles for the cell set, the researchers combined their data with previous analyses of mutational status and drug response to identify molecular signatures potentially predictive of drug sensitivity or resistance.
Using elastic net modeling, they identified 20,743 protein-drug associations. Of those, they identified 1,801 associations, involving 97 different drugs, as highly significant.
Among these associations the researchers found several proteins – including the anti-apoptotic regulator Bcl-2 and the helicase CHD4 – associated with increased sensitivity to drugs across a variety of classes, as well as a number of proteins – including 14-3-3 zeta/delta and several members of the Rab protein family – linked to drug resistance.
Such findings, Hahne suggested, could facilitate future experiments exploring proteins linked to drug response. For instance, "one could knock out or target proteins involved in drug resistance and then test whether the cell lines are still resistant to that particular drug," he said.