An Integrative Analysis of Tumor Proteomic and Phosphoproteomic Profiles to Examine the Relationships Between Kinase Activity and Phosphorylation

Abstract

Phosphorylation of proteins is a key way cells regulate function, both at the individual protein level and at the level of signaling pathways. Kinases are responsible for phosphorylation of substrates, generally on serine, threonine, or tyrosine residues. Though particular sequence patterns can be identified that dictate whether a residue will be phosphorylated by a specific kinase, these patterns are not highly predictive of phosphorylation. The availability of large scale proteomic and phosphoproteomic data sets generated using mass-spectrometry-based approaches provides an opportunity to study the important relationship between kinase activity, substrate specificity, and phosphorylation. In this study, we analyze relationships between protein abundance and phosphopeptide abundance across more than 150 tumor samples and show that phosphorylation at specific phosphosites is not well correlated with overall kinase abundance. However, individual kinases show a clear and statistically significant difference in correlation among known phosphosite targets for that kinase and randomly selected phosphosites. We further investigate relationships between phosphorylation of known activating or inhibitory sites on kinases and phosphorylation of their target phosphosites. Combined with motif-based analysis, this approach can predict novel kinase targets and show which subsets of a kinase's target repertoire are specifically active in one condition versus another.

Keywords: Breast cancer; Cancer Biology; Computational Biology; Ovarian cancer; Phosphoproteome; Phosphorylation; iTRAQ.

Graphical abstract

Fig. 1.

Protein phosphosite and intraphosphosite correlation. Distribution of correlations between (A) protein and cognate phosphosite abundance and (B) phosphorylation of phosphosites (co-phosphorylation) on the same protein, for ovarian (red) and breast (blue) tumors. Dashed vertical lines indicate the means of the distributions.

Fig. 2.

Kinase known substrate correlation. Frequency distribution of correlations of kinases with their known substrates in ovarian (red) and breast (blue) cancer. Dashed lines indicate means.

Fig. 3.

Kinase mean substrate correlation. Mean correlation of kinase abundance with known substrate abundance for ovarian and breast tumors.

Fig. 4.

Kinase phosphosite correlation - known versus all. Comparison of correlations of kinase abundance with the phosphorylation of known substrates against the complete set of phosphorylation sites profiled. A, Boxplots comparing the correlations of the entire kinase set with their known substrates against the background of correlations of the kinases with all of the phosphosites in the data. B, Distributions of correlations for individual kinases with known substrate target sites against all phosphosites in the data.

Fig. 5.

Prediction of CDK1 target phosphosites. An example of kinase-specific target substrate prediction for the kinase CDK1 by correlation analysis of kinase protein abundance with phosphorylation levels of phosphosites. The top panel shows the distributions of correlations of the kinase CDK1 with known kinase substrate phosphosites against all phosphosites in the data. The threshold is used for new phosphosite prediction from the data. Phosphosites above the threshold were used to predict kinase targets. The vertical yellow bars below the density plot mark where the known substrate phosphosites line up. The bottom panel is a sequence logo representation of the identified phosphorylation motif from the predicted target substrate phosphosites of the kinase CDK1. Position zero indicates the phosphorylation site.

Fig. 6.

Kinase functional phosphosite known substrate correlation. Boxplots of distribution of correlations of kinase functional (activating and inhibitory) phosphosites with known substrates in the (A) ovarian and (B) breast cancer data sets.