Introduction

The reversible phosphorylation of tyrosine, serine, and threonine residues is an important mechanism for modulating biological processes such as cellular signaling, differentiation, and growth. A comprehensive understanding of these dynamic cellular processes at the molecular level requires the simultaneous detection of changes in the sites and levels of phosphorylation across numerous proteins over time and through space within the cell. We employ fully automated, highly selective new methodologies that allow for the simultaneous assignment of the temporal and spatial pattern of phosphorylation sites from exceedingly complex mixtures derived from whole cell lysates. This technological infrastructure is applied to the analysis of complex signaling networks in diverse biological and pharmacological contexts.

Phosphoproteomics

Many cellular processes are directly controlled through the reversible phosphorylation of protein tyrosine residues. These regulatory functions are ultimately achieved through the coordinated phosphorylation and dephosphorylation of numerous tyrosine residues across multiple proteins over time. Clearly, benefits arise from individually characterizing specific components of a particular pathway, such as identifying a site of phosphorylation on a given protein, the kinase responsible for the modification, or the phosphatase responsible for its removal, or the identity of proteins that subsequently interact. Ultimately though, a thorough understanding of these signaling pathways at the molecular level requires the thorough, simultaneous evaluation of all phosphorylation and dephosphorylation events – changes in phosphorylation state – over time.

Emerging methodologies in mass spectrometry derive their utility through application to persistent, difficult problems in pharmacology and cellular biology. The ability of the mass spectrometer to assign sites of phosphorylation and ubiquitination in complex mixtures of peptides represents an opportunity to greatly accelerate the elucidation of signaling pathways. Identification of novel signaling pathway members will provide targets for rational development of drugs that selectively inhibit these pathways. Recent successes with the BCR/ABL kinase inhibitor STI571 (Gleevec) in the treatment of chronic myelogenous leukemia and with the HER2 receptor protein tyrosine kinase inhibitor Herceptin in the treatment of advanced breast cancer, illustrate the viability of targeting signaling pathway members to treat disease.

Development of Bioinformatic Tools for Proteomics

The amounts of data generated in proteomics experiments is staggering. Current mass spectrometers routinely can generate over 10,000 spectra per hour. Database search algorithms such as SEQUEST, and Mascot make the assignment of genomic sequence from spectral data a possibility. Unfortunately, assignment of sequence from phosphopeptide spectra is often hampered by poor spectral quality due to abundant neutral loss of phosphate or from their low abundance. Therefore, manual validation of database assigned sequences is essential for confident assignment of phosphorylation sites. This process is extremely time consuming and tedious. The need for bioinformatic tools to more rapidly process large sets of proteomic data is obvious. Our lab is interested in the development of tools which not only allow for more rapid assignment of peptide sequence from MS/MS spectra but also integrate available resources such as Scansite or the Human Reference Protein Database (HPRD). To this end, we have developed a phosphoproteomic database called PeptideDepot which has the ability to present spectra in a way that makes their manual validation efficient and intuitive. Furthermore, we developed a new SEQUEST or Mascot rescoring algorithm called the logistic spectral score based on variables designed around concepts often employed in the expert manual validation of spectra.  This logistic spectral score more than doubles the yield of confidently assigned MS/MS spectra at a 1% false discovery rate compared to SEQUEST with mass error thresholding alone.

Within PeptideDepot, the temporal and spatial pattern of phosphorylation is quickly discerned by a time course view which allows for easy quantitative comparison between experimental conditions while simplifying analysis by collation of biological replicates and graphical representation of replicate variation. Direct integration of protein information resources such as HPRD directly within our database also enables automated discovery of known interactors and sites of post translational modification which correspond to new proteins discovered in our experiments. The quantitative data within PeptideDepot can also be visualized in the context of canonical signaling pathways and protein-protein interaction databases with a new tool devloped in collaboration with Dr. David Laidlaw in the Brown Computer Science department.  This ProteinNetVis tool allows proteomic researchers to efficiently construct interactome networks which integrate both quantitative proteomic data and the scientific literature. In collaboration with the lab of Dr. Ben Raphael,  a new tool to discover phosphorylation motifs within phosphoproteomic data was built.  These discovered motifs in combination with knowledge of preferred motifs of kinases (i.e. Minimotif Miner, NetPhos, Scansite) help proteomic researchers to predict signaling networks.