A large-scale method to measure absolute protein phosphorylation stoichiometries

Abstract

The functional role of protein phosphorylation is impacted by its fractional stoichiometry. Thus, a comprehensive strategy to study phosphorylation dynamics should include an assessment of site stoichiometry. Here we report an integrated method that relies on phosphatase treatment and stable-isotope labeling to determine absolute stoichiometries of protein phosphorylation on a large scale. This approach requires the measurement of only a single ratio relating phosphatase-treated and mock-treated samples. Using this strategy we determined stoichiometries for 5,033 phosphorylation sites in triplicate analyses from Saccharomyces cerevisiae growing through mid-log phase. We validated stoichiometries at ten sites that represented the full range of values obtained using synthetic phosphopeptides and found excellent agreement. Using bioinformatics, we characterized the biological properties associated with phosphorylation sites with vastly differing absolute stoichiometries.

Figure 1

Principle of the method for phosphatase-based, absolute stoichiometry measurements. Two identical aliquots of a proteolyzed protein lysate are either mock- or phosphatase-treated followed by differential chemical labeling with stable isotopes. After mixing, fractional occupancy is encoded in a single ratio comparing the nonphosphorylated form of a peptide with and without phosphatase treatment. Peptide sequences obtained here are then examined against a database of known sites from the literature. We term these overlapping, nonphosphorylated forms “occupancy-determining peptides” (ODPs). Based on the ratio of the ODPs, stoichiometries are calculated.

Figure 2

Absolute site stoichiometries for 5,033 events in exponentially-growing yeast. Three biological yeast replicates were grown to mid-log phase. Lysates were proteolyzed with endoproteinase lys-C. An identical aliquot of each proteolyzed lysate was either mock- or phosphatase-treated as in Fig. 1. Occupancy-determining peptides (5,033) were identified based on the overlap with sites in five published studies^,,–. (a) Scatter plot of all ODP ratios. Ratios directly encode fractional occupancies (orange lines). (b) Example of an ODP from protein Bbc1. Phosphatase treatment increased the amount of the heavy version of the ODP by the amount that existed in the phosphorylated form (27%). (c) Site stoichiometry distribution for 5,033 events from WT yeast undergoing exponential growth.

Figure 3

Validation of ten sites with varied stoichiometries by AQUA. Synthetic peptides were generated representing heavy phosphorylated and nonphosphorylated versions of the peptide. The synthetic peptides were spiked into proteolyzed lysates and separated by LC-MS/MS techniques (see Methods). Using the heavy peptides as internal standards, the light versions were quantified and stoichiometries were calculated. (a) An example of the measured amount of peptide (RIIEHSDVENENVK) and phosphopeptide (RIIEHS*DVENENVK) by AQUA. The calculated stoichiometry was 69.2%. (b) An example of phosphopeptide identification (RIIEHS*DVENENVK) by MS/MS.

Figure 4

Bioinformatic analyses of site stoichiometry with respect to kinase motifs and gene ontology. (a) Phosphorylation events containing an acidic (i.e., casein kinase II-like) motif, but not basic or proline-directed ones, have higher absolute site stoichiometries on average. (b) The number of phosphorylation events in secondary structure predicted to be ordered is inversely proportional to absolute site stoichiometry. The inset shows that most sites are predicted to be in disordered regions. Clustering of the phosphoproteins based on their enrichment in specific biological processes (c) and cell compartments (d). Phosphoproteins were grouped into four classes according to their highest stoichiometry site and examined for enrichment by the DAVID software. Categories without a P-value were assigned a conservative value of 1. The P-values were log transformed and then z-transformed. Phosphoproteins were then grouped based on their z-scores via hierarchical clustering. Full versions of these figures are found in Supplementary Fig. 2.

Figure 5

Evolutionary conservation of the site residues across 25 yeast species. Sites were subgrouped by fractional occupancy into “high,” “medium,” and “low” sets and then clustered based on overall conservation levels. Each column represents a single site residue. If a homolog was identified and the site residue was conserved, the corresponding cell is yellow otherwise it is blue. If no homolog was identified, the cell is black. Most sites were conserved only within the closest neighboring species.