Protein phosphorylation is a post-translational modification implicated in a diverse variety of cellular processes, spanning from proliferation and differentiation to apoptosis. Site-specific phosphorylation events can function as molecular switches that either activate or inhibit protein activity, dictate sub-cellular localization or act as recruitment platforms for interacting proteins with special domains (such as SH2, PTB, BRCT, 14-3-3 and FHA domains). Cellular protein phosphorylation is tightly controlled by protein kinases and phosphatases, and as these enzymes have differential expression levels across tissues, protein phosphorylations are dynamic events with restricted spatial and temporal distribution. The activity of kinases and phosphatases are themselves fine-tuned by phosphorylation events, thereby interconnecting signalling pathways outlining a complex regulatory pattern. Phosphorylation events have been implicated in the pathophysiology of several severe diseases, such as cancer, diabetes and neuropsychiatric disorders1 2 3 4 5 6. For instance, in leukemia, activating mutations in kinases such as flt3 (ref. 7) and bcr-abl8 are often the oncogenic drivers of cell transformation. The fact that deregulated signalling is a hallmark of many diseases highlights the importance of developing techniques that allow for rapid, comprehensive and quantitative determinations of tissue phosphoproteomes. Quantitative mass spectrometry (MS)-based phosphoproteomics is currently the most powerful technique for analysis of cellular signalling networks9. Advances of the methodology have mainly been driven by the introduction of robust methods for phosphopeptide enrichment10 11 12 in combination with stable isotope labelling techniques13 14 and high-resolution hybrid mass spectrometers15. We and others have previously described methods to study global phosphorylation site changes as a function of specific stimuli16 17 18 19. However, these investigations were typically the results of huge efforts requiring hundreds of hours of mass spectrometric analysis and were all conducted in cell lines. So far, there have only been limited attempts to analyse phosphoproteomes of tissues and organs on a systems-wide scale20 21 22 23. Such attempts have all been based on extensive fractionation by ion-change chromatography to reduce sample complexity and low-resolution tandem MS, necessitating days of mass-spectrometric measurement time per tissue sample. Rodent models exist for many human signalling diseases and to date phosphoproteomes of nine mouse tissues has been analyzed in-depth20. However, the rat has important advantages relative to mouse for the study of cardiovascular diseases, diabetes, arthritis and many autoimmune, neurological, behavioural and addiction disorders24 as well as for testing pharmacodynamics and toxicity of potential therapeutic compounds25. Therefore, we aimed to quantify the rat organ phosphoproteome in an in-depth and reproducible manner. Here we quantitatively map phosphoproteomes of 14 rat tissues and present a large data set of 31,480 phosphorylation sites from 7,280 proteins as a resource to the scientific community. We combine an effective tissue phosphoproteome preservation and homogenization protocol with a simple, single-step phosphopeptide enrichment method followed by higher-energy collisional dissociation (HCD) fragmentation26 on an LTQ-Orbitrap Velos instrument27. This approach allows for in-depth investigation of tissue phosphoproteomes in single-shot liquid chromatography (LC)-MS analyses using a gradient of just 3 h, thus significantly reducing the time required for determination of a tissue phosphoproteome. In addition, HCD provides higher data quality covering the full mass region without a low-mass cut-off combined with high-resolution and accurate mass fragment ion measurements, which makes it a potent fragmentation technique for phosphopeptides28. Further underscoring the general applicability and translational aspects of the developed method, we validate the rat skeletal muscle phosphoproteome in human skeletal muscle biopsies. For each tissue, we systematically analyse the physical interactions of phosporylated proteins in silico to generate first drafts of spatial molecular networks regulated by tissue-specific phophatase and kinase dynamics. Results Phosphoprotein identification from 14 rat tissues To investigate phosphoproteins across tissues, organs were harvested from Sprague Dawley albino rats (Crl:SD) and they were all immediately snap frozen. We pooled organs from four rats to account for biological variation. The tissues isolated were: brain (dissected into cerebellum, cortex and brainstem), heart, muscle, lung, kidney, liver, stomach, pancreas, spleen, thymus, perirenal fat, intestine, testis and blood (Fig. 1). To preserve the in-vivo state of the phosphoproteome and reduce post-mortem effects in dissected tissue samples, we eliminated endogenous enzymatic activity by thermal protein denaturation of the snap frozen samples using a Stabilizor T1 (Denator, Sweden). This procedure effectively abolish the activity of protein phosphatases, kinases, proteases and other enzymes that can change the protein modification site abundance during sample handling29. Next, we carefully homogenized the tissues in a urea buffer using ceramic beads on a Precellys 24 (Bertin Technologies, France). Following brief sonication, protein concentration was determined and from each tissue extract 10 mg protein was digested in solution with endoproteinase Lys-C and trypsin. Two rounds of phosphopeptide enrichment by titanium dioxide chromatography were performed, and the enriched phosphopeptide fractions were analysed by 3 h LC-MS gradients on a high-performance LTQ Orbitap Velos mass spectrometer, where all tandem mass spectra were recorded in the orbitrap analyser with high-resolution using the HCD technology. All LC-MS/MS raw files were processed together, peptide sequences were identified by Mascot and phosphoproteins were quantified using the MaxQuant software suite's label-free algorithm based on peptide extracted ion chromatograms. All raw files and annotated MS/MS spectra are provided as a resource (see Supplementary Data 1). In total, 876,203 high-resolution HCD-MS/MS events were collected of which 43% were identified with high confidence, which resulted in 28,733 unique phosphopeptides corresponding to 31,480 phosphorylation sites from 7,280 proteins. Tables with all identified phosphoproteins and phosphopeptides are provided in Supplementary Data 1, 2, 4, and evaluation of the high-quality MS data is shown in Supplementary Fig. S1. Furthermore, we have set up a web-accesible MySQL database named the CPR PTM Resource containing all identified phosphoproteins making it easy to search for identified phosphorylation sites on any given protein of interest: http://cpr1.sund.ku.dk/cgi-bin/PTM.pl. The confidence of phosphorylation site localization was evaluated for each site on every phosphopeptide and for annotation of a specific phosphorylation site a localization score ≥75% combined with a ΔPTM ≥5 was required. More than 70% of the sites we identified localized to a specific amino acid with a median localization score >99.9%. Thus, the combination of an efficient protein extraction procedure with high-accuracy mass spectrometric measurements allowed us to identify a large number of phosphorylation sites in a very limited time frame. The method outlined makes it possible to determine a tissue phosphoproteome in less than 2 days. Phosphoprotein expression pattern across tissues We first focused on the phosphoproteins identified in each LC-MS run and used normalized phosphoprotein intensities derived from summation of measured phosphopeptide extracted ion chromatograms to perform a comparative analysis. Hierarchical cluster analysis visualizes the experimental specificity and reproducibility (Fig. 2a). Reassuringly, the duplicate enrichments from each tissue cluster together, as does functionally related tissues, as for instance heart and muscle and the three brain regions investigated. Phosphoproteins are colour coded according to their MS signal intensities, which is a relative measure for protein abundance30 31, and the highlighted yellow areas thus indicate that the majority of tissues have abundant expression of a specific cluster of phosphoproteins. It is evident that the identified phosphoproteins vary in expression pattern as well as in phosphorylation site abundance among the tissues reflecting the physiological differences of the tissues. A few clusters of phosphoproteins are present in all tissues investigated, which is the pattern expected for instance for house-keeping proteins. Only few phosphoproteins identified in blood are also identified in other tissues, illustrating that our perfusion of the animals during euthanization was effective. The total number of phosphoproteins identified from each of the two enrichment steps is comparable within each tissue, but a slight gain in coverage is obtained with the second incubation resulting in an increased total number of phopshoproteins when merging the two data sets (Fig. 2b). For each tissue, the duplicate enrichments yield reproducible normalized phophopeptide intensity results with Pearson correlation coefficients in the range 0.77 pyro-Glu, and phosphorylation (STY). Search parameters were set to an initial precursor ion tolerance of 7 p.p.m., MS/MS tolerance at 0.02 Da and requiring strict tryptic specificity with a maximum of two missed cleavages. Label-free peptide quantification and validation was performed in the MaxQuant software suite49 50. Phosphopeptides were filtered based on Mascot score, PTM (Andromeda) score51, precursor mass accuracy, peptide length and summed protein score to achieve an estimated false discovery rate 0.75 and ΔPTM score >5 (class 1) for the analysis. To search for over-represented motifs, we used sequence windows of ±6 residues adjacent to all serine and threonine phosphorylation sites and matched these against ten known linear protein kinase motifs (http://www.phosida.com) representing PKA, AKT, PKD, CAMK2, CHEK, CK2, PLK, CDK, ERK and ATM/ATR kinases, as well as proline-directed substrate sites. The frequency of all kinase motifs were extracted for each tissue individually and compared with the median occurrence in all 14 tissues. To identify tissue enriched as well as under-represented motifs, we calculated the percentage difference between the individual tissues and the median occurrence for each kinase motif and clustered this matrix in Perseus using correlation-based two-way hierarchical clustering. Sequence pattern analysis We performed sequence pattern analysis using iceLogo54 with percentage difference as scoring system and a P-value cut-off of 0.05. For details see Supplementary Data 3. CPR PTM resource The CPR PTM Resource (http://cpr1.sund.ku.dk/cgi-bin/PTM.pl) is a web-based data repository that integrates all of the high-confidence in-vivo post-translational modifications sites such as site-specific phosphorylation that we have identified by MS-based proteomics in different tissue samples from various species. It is based on a MySQL database and developed with the Perl/CGI language. Modified proteins can be visualized based on their Uniprot identifiers. For each modified site in a protein, we list matching kinase motifs and use the Reflect (http://reflect.cbs.dtu.dk/index.html) service to add additional information about the modified proteins. Selected screen shots from the website are displayed in Supplementary Fig. S12. Protein–protein interaction networks Protein–protein interaction networks were built using a previously described up to date interaction network of quality controlled predicted and measured human protein interactions (InWeb, Lage et al.34). Detailed description is provided in Supplementary Data 4. Author contributions A.L., A.S. and J.V.O. designed experiments. A.L. and A.S. performed sample preparation and the proteomics experiments. A.S. performed rat, C.L. and N.B.N. human experiments. K.L. generated PPI networks. A.D. developed the PTM database. A.L. and J.V.O. analysed data and wrote the paper. Additional information How to cite this article: Lundby, A. et al. Quantitative maps of protein phosphorylation sites across 14 different rat organs and tissues. Nat. Commun. 3:876 doi: 10.1038/ncomms1871 (2012). Supplementary Material Supplementary Information Supplementary Figures S1-S12 Supplementary Data S1 All modified rat peptides identified. Supplementary Data S2 All quantified rat phosphoproteins. Supplementary Data S3 All quantified rat phosphorylation sites. Supplementary Data S4 All quantified rat phosphorylation sites that are localized. Supplementary Data S5 P-values for all tissue-specific protein interaction networks. Supplementary Data S6 Human skeletal muscle phosphopeptides. First tab contains all phosphorylation sites, second tab all modified phosphopeptides identified and third tab a list of all phosphoproteins.
