Portal de Periódicos da CAPES

Machine Learning of Global Phosphoproteomic Profiles Enables Discrimination of Direct versus Indirect Kinase Substrates

2017; Elsevier BV; Volume: 16; Issue: 5 Linguagem: Inglês

10.1074/mcp.m116.066233

1535-9484

Evgeny Kanshin, Sébastien Giguère, Jing Cheng, Mike Tyers, Pierre Thibault,

Fungal and yeast genetics research

Mass spectrometry allows quantification of tens of thousands of phosphorylation sites from minute amounts of cellular material. Despite this wealth of information, our understanding of phosphorylation-based signaling is limited, in part because it is not possible to deconvolute substrate phosphorylation that is directly mediated by a particular kinase versus phosphorylation that is mediated by downstream kinases. Here, we describe a framework for assignment of direct in vivo kinase substrates using a combination of selective chemical inhibition, quantitative phosphoproteomics, and machine learning techniques. Our workflow allows classification of phosphorylation events following inhibition of an analog-sensitive kinase into kinase-independent effects of the inhibitor, direct effects on cognate substrates, and indirect effects mediated by downstream kinases or phosphatases. We applied this method to identify many direct targets of Cdc28 and Snf1 kinases in the budding yeast Saccharomyces cerevisiae. Global phosphoproteome analysis of acute time-series demonstrated that dephosphorylation of direct kinase substrates occurs more rapidly compared with indirect substrates, both after inhibitor treatment and under a physiological nutrient shift in wt cells. Mutagenesis experiments revealed a high proportion of functionally relevant phosphorylation sites on Snf1 targets. For example, Snf1 itself was inhibited through autophosphorylation on Ser391 and new phosphosites were discovered that modulate the activity of the Reg1 regulatory subunit of the Glc7 phosphatase and the Gal83 β-subunit of SNF1 complex. This methodology applies to any kinase for which a functional analog sensitive version can be constructed to facilitate the dissection of the global phosphorylation network. Mass spectrometry allows quantification of tens of thousands of phosphorylation sites from minute amounts of cellular material. Despite this wealth of information, our understanding of phosphorylation-based signaling is limited, in part because it is not possible to deconvolute substrate phosphorylation that is directly mediated by a particular kinase versus phosphorylation that is mediated by downstream kinases. Here, we describe a framework for assignment of direct in vivo kinase substrates using a combination of selective chemical inhibition, quantitative phosphoproteomics, and machine learning techniques. Our workflow allows classification of phosphorylation events following inhibition of an analog-sensitive kinase into kinase-independent effects of the inhibitor, direct effects on cognate substrates, and indirect effects mediated by downstream kinases or phosphatases. We applied this method to identify many direct targets of Cdc28 and Snf1 kinases in the budding yeast Saccharomyces cerevisiae. Global phosphoproteome analysis of acute time-series demonstrated that dephosphorylation of direct kinase substrates occurs more rapidly compared with indirect substrates, both after inhibitor treatment and under a physiological nutrient shift in wt cells. Mutagenesis experiments revealed a high proportion of functionally relevant phosphorylation sites on Snf1 targets. For example, Snf1 itself was inhibited through autophosphorylation on Ser391 and new phosphosites were discovered that modulate the activity of the Reg1 regulatory subunit of the Glc7 phosphatase and the Gal83 β-subunit of SNF1 complex. This methodology applies to any kinase for which a functional analog sensitive version can be constructed to facilitate the dissection of the global phosphorylation network. Virtually all cellular behavior is influenced either directly or indirectly by protein phosphorylation (1.Pawson T. Scott J.D. Protein phosphorylation in signaling—50 years and counting.Trends Biochem. Sci. 2005; 30: 286-290Abstract Full Text Full Text PDF PubMed Scopus (497) Google Scholar). The extensive phosphorylation-based signaling network of the cell (2.Hunter T. Signaling—2000 and beyond.Cell. 2000; 100: 113-127Abstract Full Text Full Text PDF PubMed Scopus (2261) Google Scholar, 3.Hunter T. Plowman G.D. The protein kinases of budding yeast: Six score and more.Trends Biochem. Sci. 1997; 22: 18-22Abstract Full Text PDF PubMed Scopus (403) Google Scholar) is governed by a highly interconnected network of protein kinases, phosphatases, and phospho-dependent recognition modules. A central problem in modern biology and drug discovery is therefore to identify the direct in vivo targets of protein kinases. Recent advances in mass spectrometry (MS) have allowed the profiling of 10,000s of phosphosites from few milligrams of cell lysate (4.Macek B. Mann M. Olsen J.V. Global and site-specific quantitative phosphoproteomics: Principles and applications.Annu. Rev. Pharmacol. Toxicol. 2009; 49: 199-221Crossref PubMed Scopus (345) Google Scholar, 5.Rigbolt K.T. Prokhorova T.A. Akimov V. Henningsen J. Johansen P.T. Kratchmarova I. Kassem M. Mann M. Olsen J.V. Blagoev B. System-wide temporal characterization of the proteome and phosphoproteome of human embryonic stem cell differentiation.Sci. Signal. 2011; 4: rs3Crossref PubMed Scopus (363) Google Scholar). However, the diversity of kinases in the cell and the organization of kinases into complex regulatory networks precludes the effective deconvolution of large-scale phosphoproteomic datasets into defined signaling hierarchies (6.Levy E.D. Landry C.R. Michnick S.W. Cell signaling. Signaling through cooperation.Science. 2010; 328: 983-984Crossref PubMed Scopus (44) Google Scholar). A particular difficulty arises in isolating the effects of a specific kinase from downstream kinases or phosphatases that form a local signaling network. This situation is further complicated by the propensity of kinases to form a globally connected network (7.Breitkreutz A. Choi H. Sharom J.R. Boucher L. Neduva V. Larsen B. Lin Z.Y. Breitkreutz B.J. Stark C. Liu G. Ahn J. Dewar-Darch D. Reguly T. Tang X. Almeida R. Qin Z.S. Pawson T. Gingras A.C. Nesvizhskii A.I. Tyers M. A global protein kinase and phosphatase interaction network in yeast.Science. 2010; 328: 1043-1046Crossref PubMed Scopus (524) Google Scholar). Thus, even highly specific perturbation of a given kinase may trigger many ancillary events not associated with the primary set of target substrates. In addition, many phosphorylation events may arise from nonspecific kinase activity, which may contribute substantial nonfunctional noise to the phosphoproteome (8.Landry C.R. Levy E.D. Michnick S.W. Weak functional constraints on phosphoproteomes.Trends Genetics. 2009; 25: 193-197Abstract Full Text Full Text PDF PubMed Scopus (216) Google Scholar, 9.Levy E.D. Michnick S.W. Landry C.R. Protein abundance is key to distinguish promiscuous from functional phosphorylation based on evolutionary information.Phil. Trans. Royal Soc. London B Biol. Sci. 2012; 367: 2594-2606Crossref PubMed Scopus (73) Google Scholar). Protein kinases must correctly discriminate their cognate substrates and phosphorylation sites within these substrates from a nonspecific background of ∼700,000 potentially phosphorylatable residues in the proteome. There are multiple mechanisms responsible for this exquisite specificity, including the structure of the catalytic site, local and distal interactions between the kinase and substrate, the formation of complexes with scaffolding and adaptor proteins that spatially regulate the kinase, systems-level competition between substrates, and error-correction mechanisms (10.Ubersax J.A. Ferrell Jr., J.E. Mechanisms of specificity in protein phosphorylation.Nature Rev. Mol. Cell Biol. 2007; 8: 530-541Crossref PubMed Scopus (995) Google Scholar). The responsibility for the recognition of substrates by protein kinases appears to be distributed among a large number of independent, imperfect specificity mechanisms, many of which are not fully recapitulated by in vitro assays. Conversely, the interpretation of in vivo data is confounded by the complexity and cross-connections of the global signaling network, in which perturbation of a single kinase can alter phosphorylation not only of cognate direct substrates but also through indirect effects on many other substrates that may reside in distant regions of the network. Due to these difficulties, the identification of direct kinase substrates requires both in vivo and in vitro studies with multiple levels of cross-validation that is time consuming and often not comprehensive (11.Galan J.A. Geraghty K.M. Lavoie G. Kanshin E. Tcherkezian J. Calabrese V. Jeschke G.R. Turk B.E. Ballif B.A. Blenis J. Thibault P. Roux P.P. Phosphoproteomic analysis identifies the tumor suppressor PDCD4 as a RSK substrate negatively regulated by 14–3-3.Proc. Natl. Acad. Sci. U.S.A. 2014; 111: E2918-E2927Crossref PubMed Scopus (54) Google Scholar, 12.Jeffery D.C. Kakusho N. You Z. Gharib M. Wyse B. Drury E. Weinreich M. Thibault P. Verreault A. Masai H. Yankulov K. CDC28 phosphorylates Cac1p and regulates the association of chromatin assembly factor I with chromatin.Cell Cycle. 2015; 14: 74-85Crossref PubMed Scopus (15) Google Scholar). Genetic approaches enable the manipulation of kinase activities in vivo through either simple gene deletion or the targeted introduction of inactivating or activating point mutations. However, the interpretation of the effects of such mutations on the phosphoproteome is heavily confounded by compensating effects in the network (13.Bodenmiller B. Wanka S. Kraft C. Urban J. Campbell D. Pedrioli P.G. Gerrits B. Picotti P. Lam H. Vitek O. Brusniak M.Y. Roschitzki B. Zhang C. Shokat K.M. Schlapbach R. Colman-Lerner A. Nolan G.P. Nesvizhskii A.I. Peter M. Loewith R. von Mering C. Aebersold R. Phosphoproteomic analysis reveals interconnected system-wide responses to perturbations of kinases and phosphatases in yeast.Sci. Signal. 2010; 3: rs4Crossref PubMed Scopus (250) Google Scholar). In a more elegant approach, it is possible to engineer many protein kinases to accept cell-permeable bulky ATP analogs that are not able to bind or inhibit wild-type (wt) kinases (14.Bishop A.C. Shah K. Liu Y. Witucki L. Kung C. Shokat K.M. Design of allele-specific inhibitors to probe protein kinase signaling.Current Biol. 1998; 8: 257-266Abstract Full Text Full Text PDF PubMed Scopus (182) Google Scholar). These engineered kinases, which are referred to as analog-sensitive (as) alleles provide several key advantages over other genetic techniques for kinase inactivation. In particular, the chemical inhibitor can be applied on short timescales to avoid compensation mechanisms that would otherwise buffer the loss of the target kinase. Both tyrosine and serine/threonine kinases are amenable to this approach, which has been used for selective inhibition of mutant Src-family kinases (14.Bishop A.C. Shah K. Liu Y. Witucki L. Kung C. Shokat K.M. Design of allele-specific inhibitors to probe protein kinase signaling.Current Biol. 1998; 8: 257-266Abstract Full Text Full Text PDF PubMed Scopus (182) Google Scholar), Abl-family kinases (15.Liu Y. Witucki L.A. Shah K. Bishop A.C. Shokat K.M. Src-Abl tyrosine kinase chimeras: Replacement of the adenine binding pocket of c-Abl with v-Src to swap nucleotide and inhibitor specificities.Biochemistry. 2000; 39: 14400-14408Crossref PubMed Scopus (25) Google Scholar), mitogen-activated kinases (16.Bishop A.C. Ubersax J.A. Petsch D.T. Matheos D.P. Gray N.S. Blethrow J. Shimizu E. Tsien J.Z. Schultz P.G. Rose M.D. Wood J.L. Morgan D.O. Shokat K.M. A chemical switch for inhibitor-sensitive alleles of any protein kinase.Nature. 2000; 407: 395-401Crossref PubMed Scopus (860) Google Scholar), cyclin-dependent kinases (CDKs) (16.Bishop A.C. Ubersax J.A. Petsch D.T. Matheos D.P. Gray N.S. Blethrow J. Shimizu E. Tsien J.Z. Schultz P.G. Rose M.D. Wood J.L. Morgan D.O. Shokat K.M. A chemical switch for inhibitor-sensitive alleles of any protein kinase.Nature. 2000; 407: 395-401Crossref PubMed Scopus (860) Google Scholar), p21-activated kinases (17.Weiss E.L. Bishop A.C. Shokat K.M. Drubin D.G. Chemical genetic analysis of the budding-yeast p21-activated kinase Cla4p.Nature Cell Biol. 2000; 2: 677-685Crossref PubMed Scopus (113) Google Scholar), and Ca2+/calmodulin-dependent kinases. Despite the power of the as-allele approach, the functional integrity of the mutant kinase and the actual target specificity of the inhibitor are crucial considerations. It has been reported that the catalytic activity of some protein kinases is affected by as mutations (15.Liu Y. Witucki L.A. Shah K. Bishop A.C. Shokat K.M. Src-Abl tyrosine kinase chimeras: Replacement of the adenine binding pocket of c-Abl with v-Src to swap nucleotide and inhibitor specificities.Biochemistry. 2000; 39: 14400-14408Crossref PubMed Scopus (25) Google Scholar) and unintended off-target effects of the inhibitor on other enzymes (18.Bishop A.C. Buzko O. Shokat K.M. Magic bullets for protein kinases.Trends Cell Biol. 2001; 11: 167-172Abstract Full Text Full Text PDF PubMed Scopus (202) Google Scholar). Given that kinase inactivation will shift the equilibrium of cognate substrates to the dephosphorylated state, due to the action of constitutive phosphatases, a combination of quantitative phosphoproteomics and selective chemical inhibition can in principle be used to discover in vivo kinase targets on a large-scale. This approach has for example been used to identify Cdc28 substrates in S. cerevisiae (19.Holt L.J. Tuch B.B. Villén J. Johnson A.D. Gygi S.P. Morgan D.O. Global analysis of Cdk1 substrate phosphorylation sites provides insights into evolution.Science. 2009; 325: 1682-1686Crossref PubMed Scopus (664) Google Scholar). Based on criteria of at least twofold dephosphorylation in response to kinase inhibition and the presence of a full CDK consensus motif, a total of 547 phosphorylation sites have been assigned to 308 candidate Cdc28 substrates (19.Holt L.J. Tuch B.B. Villén J. Johnson A.D. Gygi S.P. Morgan D.O. Global analysis of Cdk1 substrate phosphorylation sites provides insights into evolution.Science. 2009; 325: 1682-1686Crossref PubMed Scopus (664) Google Scholar). Unexpectedly, a significant proportion of the quantified sites on these substrates were up-regulated following Cdc28 inhibition, which suggests the influence of feedback loops that may activate additional kinases and/or inactivate phosphatases. Notably, some of these up-regulated phosphosites were even located within the CDK consensus sequence. Additional criteria are thus needed to resolve the spectrum of direct in vivo substrates for any given kinase. To address the above issues with kinase substrate identification, we set out to develop a method for unbiased large-scale identification of direct in vivo kinase substrates based on a combination of specific kinase inhibition, large-scale phosphoproteomics, and machine learning approaches (Fig. 1a). The point of departure for our strategy is the selective inhibition of an as-form of a kinase of interest by the chemical inhibitor 1NM-PP1 (20.Papa F.R. Zhang C. Shokat K. Walter P. Bypassing a kinase activity with an ATP-competitive drug.Science. 2003; 302: 1533-1537Crossref PubMed Scopus (197) Google Scholar). Upon treatment with the inhibitor, we expect changes in phosphorylation in response to the inhibitor itself (kinase independent) and changes that result from inhibition of the as kinase (kinase dependent). The latter may arise from either dephosphorylation of direct substrates or phosphorylation/dephosphorylation of indirect substrates. The latter substrates would represent downstream targets that are not directly phosphorylated by a kinase, for instance when the targeted kinase modulates the activity of another kinase or phosphatase, which in turn affects the phosphorylation of other proteins. To discriminate between kinase-independent, direct, and indirect substrate classes, we developed and validated a machine learning strategy that uses the support vector machine (SVM) (21.Cortes C. Vapnik Vladimir Support-vector networks.Machine Learning. 1995; 20: 273-297Crossref Scopus (0) Google Scholar) and ideas from positive and unlabeled learning (22.Elkan, C., and Noto, K., (2008) Learning classifiers from only positive and unlabeled data. Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining,Google Scholar). A proof of principle using the well-studied Cdc28 kinase in yeast (23.Malumbres M. Cyclin-dependent kinases.Genome Biol. 2014; 15: 122Crossref PubMed Scopus (828) Google Scholar) allowed us to identify many novel substrates and obtain a deeper understanding of kinase–substrate relationships. To demonstrate applicability to other less-characterized kinases, we then use our approach to identify direct targets of Snf1, a kinase that is required for transcription of glucose-repressed genes, thermotolerance, sporulation, and peroxisome biogenesis in yeast (24.Conrad M. Schothorst J. Kankipati H.N. Van Zeebroeck G. Rubio-Texeira M. Thevelein J.M. Nutrient sensing and signaling in the yeast Saccharomyces cerevisiae.FEMS Microbiol. Rev. 2014; 38: 254-299Crossref PubMed Scopus (379) Google Scholar, 25.Kayikci O. Nielsen J. Glucose repression in Saccharomyces cerevisiae.FEMS Yeast Res. 2015; 15Crossref PubMed Scopus (167) Google Scholar). Our integrated approach should enable the identification of direct substrates for any kinase for which a functional as allele can be engineered. The systematic application of this method should help reveal the true structure of the global phosphorylation-based network and provide insights into biological regulation. Inhibition of CDK1 1The abbreviations used are: CDK1, cyclin-dependent kinase;Cdc28, cyclin-dependent kinase 1;as, analog-sensitive alleles;1NM-PP1, 1-(1,1-dimethylethyl)-3-(1-naphthalenylmethyl)-1H-pyrazolo[3,4-d]pyrimidin-4-amine;OD, optical density;SILAC, stable isotope labeling with amino acids in culture;TRIS, tris(hydroxymethyl)aminomethane;FDR, false discovery rate;SVM, support vector machine;GS, generic string;YPD, yeast extract peptone dextrose;AGC, automatic gain control;PPI, protein-protein pnteraction(s);GO, gene ontology;STRING, (Search Tool for the Retrieval of Interacting; Genes/Proteins);SCB-binding factor, SBF;MCB-binding factor, MBF;SCB, Swi4/6 cell cycle box;MCB, MluI cell cycle box;SNF1, Sucrose NonFermenting 1. and SNF1 with 1NM-PP1 was performed as six independent treatments, every time using reverse triple SILAC labeling (see supplemental Fig. S1). To find significantly affected phosphosites, t test pValue threshold was chosen based on false discovery rate (FDR) <5% using permutation-based correction. For all dynamic (time course) experiments, only one replicate was used for analysis and significance was calculated based on the shape of the entire kinetic profile rather than individual measurements (26.Kanshin E. Kubiniok P. Thattikota Y. D'Amours D. Thibault P. Phosphoproteome dynamics of Saccharomyces cerevisiae under heat shock and cold stress.Mol. Syst. Biol. 2015; 11: 813Crossref PubMed Scopus (41) Google Scholar). The S. cerevisiae strains used in this study were isogenic with the S288C background. Single deletion strains, in which open reading frames were replaced by the kanMX cassette, were obtained from H. Bussey (McGill University, Montreal, Canada), Research Genetics and EUROSCARF. For SILAC labeling, the wt strain YAL6B (MATa his3Δ leu2Δ met15Δ ura3Δ lys1::KanMX6 arg4::KanMX4) was provided by Ole Jensen (University of Southern Odense, Denmark). To construct the cdc28-as1 strain used for SILAC analysis, a mutation resulting in an F88G substitution was introduced at the CDC28 locus in strain YAL6B by PCR amplification from a CDC28::Tadh1::HIS3MX6 selection cassette (27.Chylek L.A. Akimov V. Dengjel J. Rigbolt K.T. Hu B. Hlavacek W.S. Blagoev B. Phosphorylation site dynamics of early T-cell receptor signaling.PloS One. 2014; 9: e104240Crossref PubMed Scopus (45) Google Scholar). The presence of the F88G mutation and absence of secondary mutations were confirmed by sequencing the entire CDC28 open reading frame. The snf1-as strain for SILAC labeling was constructed by integration of an snf1(I132G)-LEUMX cassette into an snf1::kanMX deletion strain (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) followed by crossing to the YAL6B strain and selection for a spore clone that was unable to grow on -Lys or -Arg medium. The snf1-as mutation and flanking regions of genomic DNA were sequence verified. Phosphorylation site mutations were introduced at the genomic loci by homologous recombination-mediated seamless integration in a two-step acceptor/donor protocol. First, an acceptor strain was generated by transformation of a linear DNA fragment that contained 45 bp upstream and 45 bp downstream of the mutation site flanking the URA3MX cassette into parental strain BY4741 and selection on -Ura medium. Second, the acceptor strain was transformed with a donor DNA fragment that contained 150 bp upstream and 150 bp downstream of the mutation site, which substituted the natural codon for an alanine codon, followed by selection on 5-FOA medium. All phosphorylation site mutant strains were verified by sequencing of at least 200 bp upstream and downstream of the mutation site. Three independent isolates of each strain were assessed for mutation-associated phenotypes. An SNF1(S391A) snf4Δ double mutant strain was constructed by PCR-based disruption of the SNF4 locus with an His3MX cassette in the parental SNF1(Ser391Ala) strain. Yeast cells were grown in synthetic dextrose (S.D., 0.17% yeast nitrogen base without amino acids, 0.5% ammonium sulfate, and appropriate amino acids) supplemented with either light (Lys0/Arg0), medium (Lys4/Arg6), or heavy (Lys8/Arg10) lysine (30 mg/l) and arginine (20 mg/l) (Cambridge Isotope Laboratories). For experiments with the cdc28-as allele, cultures were grown in S.D. in the presence of 2% glucose. For experiments with snf1-as, cultures were grown in S.D. media containing 2% ethanol. All cultures were grown to a optical density of OD600 = 0.7. Kinase inhibition was performed by adding the 1NM-PP1 to a final concentration of 10 μm, whereas control cultures were treated with vehicle (DMSO). After 15 min, TCA was added to final concentration of 10%, cells were collected by centrifugation a 2,000 × g for 10 min, washed with ice-cold PBS, and the SILAC channels were combined before cell lysis. For nutrient shift experiments, snf1-wt strain was grown in S.D. media containing 2% EtOH, concentrated glucose solution was added to light SILAC culture to final 2% (w/w), whereas heavy SILAC culture was used as a control. Probes from both SILAC channels were taken every 2 min for a time course of 30 min. For growth assays, mutant strains were cultured in the S.D. medium containing the appropriate carbon source (2% glucose or 2% galactose or 3% glycerol) or XY medium (2% bactopeptone, 1% yeast extract, 0.01% adenine, 0.02% tryptophan) containing 3% glycerol, as indicated. For size analysis, strains were inoculated from fresh YPD plates into indicated media and cell size profiles determined after 8, 16, 24, and 40 h on a Beckman–Coulter Z2 particle sizer. Daughter cell size was defined as the half maximal peak on the left-hand side of the cell size distribution, and mode size was defined as the value at maximum peak height. For growth curve measurements, cultures were seeded from an overnight culture in 2% glucose medium at density 2 × 106 cells per ml in 100 μl of media in 96-well plates. Culture growth was measured in triplicate on a Tecan Sunrise automated shaker/reader and analyzed using Magellan V7.1 software. Growth parameters were calculated using custom cell growth curve software (http://sysbiolab.bio.ed.ac.uk:3838/my-app/GCA.shiny/). Cells were lysed by bead beating for two cycles of 5 min each in lysis buffer (8 m urea, 50 mm Tris, pH 8.0, supplemented with HALT phosphatase inhibitor mixture, Pierce). Samples were centrifuged at 40,000 × g for 10 min, and the supernatants were transferred into clean tubes prior to determination of protein concentrations by bicinchoninic acid assay (Thermo Fisher Scientific). Disulfide bonds were reduced by adding dithiothreitol to a final concentration of 5 mm, and samples were incubated at 56 °C for 30 min. Reduced cysteines were alkylated by adding iodoacetamide to 15 mm and incubating for 30 min in the dark at room temperature. Alkylation was quenched with 5 mm dithiothreitol for 15 min. Samples were diluted sixfold with 20 mm TRIS, pH 8, containing 1 mm CaCl2 prior to overnight digestion at 37 °C with trypsin (Sigma-Aldrich) using an enzyme to substrate ratio of 1:50 (w/w). Tryptic digests were acidified with 1% formic acid (FA), centrifuged (20,000 × g 10 min) and desalted on Oasis HLB cartridges (Waters) according to manufacturer instructions. Peptide eluates were snap-frozen in liquid nitrogen, lyophilized in a speedvac centrifuge, and stored at −80 °C. Tryptic digests were subjected to enrichment on TiO2 beads as described previously (28.Kanshin E. Michnick S.W. Thibault P. Displacement of N/Q-rich peptides on TiO2 beads enhances the depth and coverage of yeast phosphoproteome analyses.J. Proteome Res. 2013; 12: 2905-2913Crossref PubMed Scopus (27) Google Scholar). Sample loading, washing, and elution steps were performed using custom StageTips (29.Rappsilber J. Ishihama Y. Mann M. Stop and go extraction tips for matrix-assisted laser desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics.Anal. Chem. 2003; 75: 663-670Crossref PubMed Scopus (1794) Google Scholar, 30.Ishihama Y. Rappsilber J. Mann M. Modular stop and go extraction tips with stacked disks for parallel and multidimensional peptide fractionation in proteomics.J. Proteome Res. 2006; 5: 988-994Crossref PubMed Scopus (224) Google Scholar) made from 200 μl pipette tips containing an SDB-XC membrane (3 m) frit and filled with TiO2 beads. We equilibrated TiO2 material in 250 mm lactic acid 70% acetonitrile 3% TFA; the same buffer was used for sample loading. After extensive washing steps, retained phosphopeptides were displaced from TiO2 with 500 mm phosphate buffer at pH = 7. Peptides were desalted in 50 μl of 1% FA directly on SDB-XC frits and subsequently eluted using 50 μl of 50% acetonitrile 1% FA. Eluates were dried in a speedvac and stored at −80 °C. To increase phosphoproteome coverage prior to MS analysis, phosphopeptides were fractionated offline by strong cation exchange chromatography. Peptides were solubilized in 100 μl of loading buffer (0.2% FA 10% acetonitrile) and loaded on StageTips containing 10 mg of poly-sulfoethyl-A strong cation exchange phase (5 μm 300A, Canada Life Science). Columns were washed with 50 μl of loading buffer, and peptides were eluted in five separate 100 μl salt steps of 25, 50, 80, 150, and 250 mm NaCl. All fractions (including flow-through) were collected, dried in a speedvac, and resuspended in 10 μl of 4% FA prior to MS analysis. Enriched phosphopeptide extracts were analyzed by LC-MS/MS using a Proxeon nanoflow HPLC system coupled to a tribrid fusion mass spectrometer (Thermo Fisher Scientific). Each sample was loaded and separated on a reverse-phase analytical column (18 cm length, 150 μm inner diameter) (Jupiter C18, 3 μm, 300 Å, Phenomenex) packed manually. LC separations were performed at a flow rate of 0.6 μl/min using a linear gradient of 5–30% aqueous acetonitrile (0.2% FA) in 106 min. MS spectra were acquired with a resolution of 60,000. The TopSpeed (maximum number of sequencing events within 5 s window) method was used for data-dependent scans on the most intense ions using high energy dissociation. AGC target values for MS and MS/MS scans were set to 5e5 (max fill time 200 ms) and 5e4 (max fill time 200 ms), respectively. The precursor isolation window was set to m/z 1.6 with a high energy dissociation normalized collision energy of 25. The dynamic exclusion window was set to 30 s. MS data were analyzed using MaxQuant (31.Cox J. Mann M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification.Nature Biotechnol. 2008; 26: 1367-1372Crossref PubMed Scopus (9141) Google Scholar, 32.Cox J. Neuhauser N. Michalski A. Scheltema R.A. Olsen J.V. Mann M. Andromeda: A peptide search engine integrated into the MaxQuant environment.J. Proteome Res. 2011; 10: 1794-1805Crossref PubMed Scopus (3445) Google Scholar) software version 1.3.0.3 and searched against the SwissProt subset of the S. cerevisiae uniprot database (http://www.uniprot.org/) containing 6,630 entries (November 2013). A list of 248 common laboratory contaminants included in MaxQuant was also added to the database as well as reversed versions of all sequences. The enzyme specificity was set to trypsin with a maximum number of missed cleavages set to 2. Peptide identification was performed with an allowed initial precursor mass deviation up to 7 ppm and an allowed fragment mass deviation of 20 ppm with subsequent nonlinear mass recalibration (33.Cox J. Michalski A. Mann M. Software lock mass by two-dimensional minimization of peptide mass errors.J. Am. Soc. Mass Spectrom. 2011; 22: 1373-1380Crossref PubMed Scopus (108) Google Scholar). Phosphorylation of serine, threonine, and tyrosine residues was searched as variable modification; carbamidomethylation of cysteines was searched as a fixed modification. The FDR for peptide, protein, and site identification was set to 1% and was calculated using decoy database approach. The minimum peptide length was set to 6, and the “peptide requantification” function was enabled. The option match between runs (1-min time tolerance) was enabled to correlate identification and quantitation results across different runs. In addition to an FDR of 1% set for peptide, protein, and phosphosite identification levels, we consi