Isoelectric point-based fractionation by HiRIEF coupled to LC-MS allows for in-depth quantitative analysis of the phosphoproteome

Abstract

Protein phosphorylation is involved in the regulation of most eukaryotic cells functions and mass spectrometry-based analysis has made major contributions to our understanding of this regulation. However, low abundance of phosphorylated species presents a major challenge in achieving comprehensive phosphoproteome coverage and robust quantification. In this study, we developed a workflow employing titanium dioxide phospho-enrichment coupled with isobaric labeling by Tandem Mass Tags (TMT) and high-resolution isoelectric focusing (HiRIEF) fractionation to perform in-depth quantitative phosphoproteomics starting with a low sample quantity. To benchmark the workflow, we analyzed HeLa cells upon pervanadate treatment or cell cycle arrest in mitosis. Analyzing 300 µg of peptides per sample, we identified 22,712 phosphorylation sites, of which 19,075 were localized with high confidence and 1,203 are phosphorylated tyrosine residues, representing 6.3% of all detected phospho-sites. HiRIEF fractions with the most acidic isoelectric points are enriched in multiply phosphorylated peptides, which represent 18% of all the phospho-peptides detected in the pH range 2.5-3.7. Cross-referencing with the PhosphoSitePlus database reveals 1,264 phosphorylation sites that have not been previously reported and kinase association analysis suggests that a subset of these may be functional during the mitotic phase.

Figure 1

Experimental workflow applied to perform HiRIEF-based phosphoproteomics analysis. Protein extracts from HeLa cells untreated (4 replicates), pervanadate treated or arrested in mitosis (3 replicates each) were digested to peptides with trypsin. For Phospho HiRIEF analysis, samples were enriched for phosphorylated peptides with titanium dioxide (TiO₂), labeled with Tandem Mass Tags (TMT) and pooled. The pooled sample was split in two and fractionated by HiRIEF using immobilized pH gradient (IPG) strips with pH range 2.5–3.7 (ultra-acidic strip) and 3–10 (wide-range), prior to LC-MS analysis. For standard proteomics analysis, peptide samples were labeled with TMT, pooled, fractionated by HiRIEF using a wide-range (3–10) IPG strip and analyzed by LC-MS.

Figure 2

Identifications by and reproducibility of HiRIEF based phosphoproteomics. (a) Pie chart representing the number of unique peptides broken down by number of phosphorylations. (b) Distribution of unique peptides across HiRIEF fractions, broken down by number of phosphorylations. Fraction numbering proceeds from the acidic end towards the basic end of the strips. (c) Venn diagram showing the overlap between unique phospho-peptides identified with the two HiRIEF pH ranges, 2.5–3.7 and 3–10. (d) Total number of identified unique phospho-sites, displayed by modified residue (serine, threonine or tyrosine). (e) Correlation of phospho-sites log₂ transformed ratio values for each replicate pair; Pearson correlation coefficient is displayed. Ratio values were calculated using as a denominator the average of the four untreated samples, and normalized to total protein levels (see Methods).

Figure 3

Functional characterization of the identified phospho-sites. (a) Heatmap representing complete linkage hierarchical clustering based on Euclidian distance of the unique phospho-site ratios. Row color-coding represents the modified residue (S,T or Y). (b) Venn diagram showing the overlap between the identified phospho-sites, the ones previously reported in the PhosphoSitePlus database and the HeLa phospho-sites reported by Sharma et al. in 2014. (c) Proportion of novel and known phospho-sites per modified residue. (d) Distribution of protein precursor areas for the novel and known phospho-sites per type of modified residue. Two-sided t-test was performed to assess significance.

Figure 4

Novel phospho-sites have putative roles in mitotic regulation. (a) NetworKIN score distributions for the subsets of known functional phospho-sites or phospho-sites with no known function (Other sites). Two-sided t-test was performed to assess significance. A threshold score of 3 is used to separate a population of putatively functional phospho-sites and GO Biological Process enrichment analysis is performed on the score separated populations. (b) Number of genes identified for each of the examined classes in the phospho-HiRIEF analysis. Genes ascribable to more than one class are indicated in separate groups, with the classes separated by a “/” character. (c) Relative distribution of genes across classes for the subsets of genes corresponding to functional, putatively functional or low-scoring phospho-sites. Class enrichment in each subset compared to the class distribution for all genes identified in the Phospho-HiRIEF analysis is evaluated by Fisher exact test. Genes attributable to more than one class were excluded from this analysis. (d) Heatmap representing hierarchical clustering based on Euclidian distance and complete linkage of putatively functional novel phospho-sites. Row color-codings represent: “Class”, class of the protein carrying the indicated phospho-site (color coding as in Fig. 4b); “Residue”, modified residue; “Significant regulation in mitotic samples”, red – significant, grey - not significant. Detailed list of the phospho-sites included in the heatmap is provided in Supplementary Data S5.

Figure 5

Cell cycle related novel putatively functional phospho-sites are highly connected in a protein-protein interaction network. The network represents interactions between proteins containing novel putatively functional phospho-sites annotated with the GO term “Cell cycle process”. Interactions of these proteins with other “Cell cycle process” annotated proteins are also included (see Methods for description of the network generation process). Square shaped nodes symbolize proteins (annotated by gene symbol) while round shaped nodes symbolize phosphorylation sites (detailed list available in Supplementary Data S5). Phosphorylation sites are color-coded by the average log₂ normalized ratio value across the three mitotic arrested samples. The network was generated using the Cytoscape software platform and the PhosphoPath plugin. Protein interactions are retrieved by PhosphoPath based on the BioGRID database.