NEW YORK (GenomeWeb) – A team led by researchers at the Wellcome Trust Sanger Institute has completed a proteogenomic analysis of colorectal cancer cell lines, identifying the impact of somatic genomic variants on proteins networks.
In a paper published today in Cell Reports, the researchers demonstrated how genomic variations can directly affect both protein expression levels and larger networks of protein-protein interactions. They also tested 265 existing anti-cancer drugs against the cell lines, identifying a variety of responses at the proteomic level not predicted by genomic and transcriptomic data.
The study used isobaric labeling combined with mass spec analysis to characterize the proteomes of 50 colorectal cancer cell lines from the COREAD collection that had previously been analyzed using whole-exome sequencing, gene expression profiling, and copy number and methylation profiling. Via the mass spec analysis, the researchers quantified an average of more than 9,000 proteins and 11,000 phosphopeptides per cell line.
Using this data, they built protein co-variation networks, generating a set of 284 proteins modules ranging from groups of three to groups of 1,012 proteins. The also generated de novo predictions of kinase-substrate interactions, identifying co-phosphorylated groups of proteins. These co-variation and co-phosphorylation networks, the authors wrote, revealed "the higher-order organization of cellular functions" in colorectal cancer, and "constitutes a reference point for the better understanding of the underlying biological networks in the COREAD panel."
The researchers also investigated the effect of known colorectal cancer driver mutations on protein abundance, finding that, generally, "mutations in canonical tumor suppressor genes" led to a decrease in protein abundance. A broader analysis looking at 4,658 genes with somatic single-amino-acid substitutions as well as 20 recurrent copy number alterations found that, while a small number of these variants led to changes in protein expression, "only for a small portion of the proteome can the variation in abundance be directly explained by mutations and DNA copy number variations," they added.
However, as the authors noted, tight control of protein abundance is key to protein complexes and protein-protein interactions, raising the possibility that downregulation of a particular protein due to a mutation can lead to altered regulation of its interaction partners. To explore this question further, they used their protein covariation data to construct networks of proteins linked to loss-of-function mutations, building what they termed "mutation-vulnerable protein networks."
The findings "indicate that an additional layer of protein variation can be explained by the collateral effects of mutations on tightly co-regulated partners in protein co-variation networks," the team wrote.
The investigators also explored whether proteomic-based signatures might provide additional insights into drug response beyond those provided by genomics and transcriptomic data. Screening the 50 cell lines against 265 agents, including 48 drugs currently on the market, 76 in clinical development, and 141 experimental compounds, they found that by combining genomic, proteomic, and phosphoproteomics models they could predict response to 81 of the 265 compounds tested, with protein expression levels of drug efflux pump proteins particularly useful for predicting response.
While "predictive genomic biomarkers may still be discovered, the importance of proteomic associations with drug response should not be underestimated," the authors concluded.