A Novel Proteomic Screen for Peptide-Protein Interactions

Regulated interactions between short, unstructured amino acid sequences and modular protein domains are central to cell signaling. Here we use synthetic peptides in "active" (e.g. phosphorylated) and "control" (e.g. non-phosphorylated) forms as baits in affinity pull-down experiments to determine such interactions by quantitative proteomics. Stable isotope labeling by amino acids in cell culture distinguishes specific binders directly by the isotope ratios determined by mass spectrometry (Blagoev, B., Kratchmarova, I., Ong, S.-E., Nielsen, M., Foster, L. J., and Mann, M. (2003) Nat. Biotechnol. 21, 315–318). A tyrosine-phosphorylated peptide of the epidermal growth factor receptor specifically retrieved the Src homology domain (SH) 2- and SH3 domain-containing adapter protein Grb2. A proline-rich sequence of Son of Sevenless also specifically bound Grb2, demonstrating that the screen maintains specificity with low affinity interactions. The proline-rich Sos peptide retrieved only SH3 domain containing proteins as specific binding partners. Two of these, Pacsin 3 and Sorting Nexin 9, were confirmed by immunoprecipitation. Our data are consistent with a change in the role of Sos from Ras-dependent signaling to actin remodeling/endocytic signaling events by a proline-SH3 domain switch. Regulated interactions between short, unstructured amino acid sequences and modular protein domains are central to cell signaling. Here we use synthetic peptides in "active" (e.g. phosphorylated) and "control" (e.g. non-phosphorylated) forms as baits in affinity pull-down experiments to determine such interactions by quantitative proteomics. Stable isotope labeling by amino acids in cell culture distinguishes specific binders directly by the isotope ratios determined by mass spectrometry (Blagoev, B., Kratchmarova, I., Ong, S.-E., Nielsen, M., Foster, L. J., and Mann, M. (2003) Nat. Biotechnol. 21, 315–318). A tyrosine-phosphorylated peptide of the epidermal growth factor receptor specifically retrieved the Src homology domain (SH) 2- and SH3 domain-containing adapter protein Grb2. A proline-rich sequence of Son of Sevenless also specifically bound Grb2, demonstrating that the screen maintains specificity with low affinity interactions. The proline-rich Sos peptide retrieved only SH3 domain containing proteins as specific binding partners. Two of these, Pacsin 3 and Sorting Nexin 9, were confirmed by immunoprecipitation. Our data are consistent with a change in the role of Sos from Ras-dependent signaling to actin remodeling/endocytic signaling events by a proline-SH3 domain switch. Important protein-protein interactions in cell signaling are frequently mediated by short, unstructured sequences, which specifically interact with peptide binding domains (1Pawson T. Scott J.D. Science. 1997; 278: 2075-2080Crossref PubMed Scopus (1900) Google Scholar). Typical examples are the binding of tyrosyl-phosphorylated peptides to proteins containing Src homology domain 2 (SH2) 1The abbreviations used are: SH2, Src homology domain 2; SH3, Src homology domain 3; 12C6-Arg, 12C6 or "normal" arginine; 13C6-Arg, 13C6-arginine; EGF, epidermal growth factor; EGFR, EGF receptor; Grb2, growth factor-associated protein 2; LeuD3, l-leucine-5,5,5-d3; Pacsin, protein kinase C and casein kinase substrate in neurons; SILAC, stable isotope labeling with amino acids in cell culture; Snx9, Sorting nexin 9; Sos, Son of Sevenless; MS, mass spectrometry. or phosphotyrosyl binding domain (2Bradshaw J.M. Waksman G. Adv. Protein Chem. 2002; 61: 161-210Crossref PubMed Scopus (103) Google Scholar, 3Yaffe M.B. Nat. Rev. Mol. Cell Biol. 2002; 3: 177-186Crossref PubMed Scopus (286) Google Scholar). Peptides with certain proline motifs constitutively bind to proteins containing Src homology domain 3 (SH3) at lower affinities (4Mayer B.J. J. Cell Sci. 2001; 114: 1253-1263Crossref PubMed Google Scholar, 5Cesareni G. Panni S. Nardelli G. Castagnoli L. FEBS Lett. 2002; 513: 38-44Crossref PubMed Scopus (122) Google Scholar). Despite the central importance of motif-dependent interactions in cell signaling, there are few methods to identify peptide-binding partners. Phage display (6Kay B.K. Kasanov J. Yamabhai M. Methods. 2001; 24: 240-246Crossref PubMed Scopus (78) Google Scholar), array technologies (7Espejo A. Cote J. Bednarek A. Richard S. Bedford M.T. Biochem. J. 2002; 367: 697-702Crossref PubMed Scopus (138) Google Scholar), and peptide libraries (8Elia A.E.H. Cantley L.C. Yaffe M.B. Science. 2003; 299: 1228-1231Crossref PubMed Scopus (571) Google Scholar) are powerful strategies to learn about general features of optimal peptide sequences binding to a given domain or protein. Information from these experiments has in turn been helpful in the development of bioinformatic algorithms to predict occurrence of these sequences in proteins (9Yaffe M.B. Leparc G.G. Lai J. Obata T. Volinia S. Cantley L.C. Nat. Biotechnol. 2001; 19: 348-353Crossref PubMed Scopus (464) Google Scholar). However, these methods cannot efficiently identify specific in vivo interaction partners of a given modified peptide sequence. The yeast two-hybrid system (10Legrain P. Selig L. FEBS Lett. 2000; 480: 32-36Crossref PubMed Scopus (128) Google Scholar) has been successfully applied to large scale protein-protein interaction mapping (11Uetz P. Giot L. Cagney G. Mansfield T.A. Judson R.S. Knight J.R. Lockshon D. Narayan V. Srinivasan M. Pochart P. Qureshi-Emili A. Li Y. Godwin B. Conover D. Kalbfleisch T. Vijayadmodar G. Yang M. Johnston M. Fields S. Rothberg J.M. Nature. 2000; 403: 623-627Crossref PubMed Scopus (3915) Google Scholar, 12Ito T. Chiba T. Ozawa R. Yoshida M. Hattori M. Sakaki Y. Proc. Natl. Acad. Sci. U. S. A. 2001; 98: 4569-4574Crossref PubMed Scopus (2946) Google Scholar, 13Walhout A.J. Vidal M. Nat. Rev. Mol. Cell Biol. 2001; 2: 55-62Crossref PubMed Scopus (119) Google Scholar), but it is not well suited to determine the interaction partners of short and modified peptide sequences. In addition, the yeast two-hybrid system is a heterologous system for studying interactions of mammalian or plant proteins. The most straightforward method for determining interaction partners of a peptide is to use it as bait in affinity pull-down experiments followed by direct detection of binding proteins. General use of this approach has been limited by two factors. First, the affinity of a given binding domain to the peptide bait is often weak and can further be reduced because the peptide bait is used outside of the context of the whole protein. Second, binding partners of interest in signal transduction are typically of low abundance. Thus, small amounts of protein binding specifically to the bait are masked by more abundant but nonspecific binders to both the peptides and the matrix employed in the pull-down. This has often prevented unbiased methods, such as peptide sequencing, to be employed for the identification of interacting proteins, and synthetic peptides are usually only employed in the verification of postulated protein-protein interactions by competitive disruption of binding. Recently our group described a proteomic method to detect signaling complexes formed upon cell stimulation (14Blagoev B. Kratchmarova I. Ong S.-E. Nielsen M. Foster L.J. Mann M. Nat. Biotechnol. 2003; 21: 315-318Crossref PubMed Scopus (607) Google Scholar). Using stable isotope labeling by amino acids in cell culture (SILAC) (15Ong S.-E. Blagoev B. Kratchmarova I. Kristensen D.B. Steen H. Pandey A. Mann M. Mol. Cell. Proteomics. 2002; 1: 376-386Abstract Full Text Full Text PDF PubMed Scopus (4581) Google Scholar, 16Ong S.-E. Kratchmarova I. Mann M. J. Proteome Res. 2003; 2: 173-181Crossref PubMed Scopus (375) Google Scholar), all the proteins in one protein population were metabolically labeled with a heavy isotope- form of arginine (13C6-Arg, leading to a difference of six mass units). In this way, two forms of each arginine-containing peptide are measured in subsequent mass spectrometric experiments. The two forms differ only in their mass and can therefore be used to directly quantify the ratio of the proteins from one population in relation to the other. By using the SH2 domain of Grb2 as an affinity probe, tyrosine-phosphorylated EGF receptor and associated proteins were purified from labeled, stimulated cells and non-labeled, non-stimulated cells. After trypsin digestion of eluted proteins, arginine-containing peptides were observed as light (the normal 12C6 form of arginine) and heavy (13C6-arginine) forms in mass spectra. One-to-one ratios between the isotope forms indicated proteins unaffected by EGF stimulation, whereas ratios greater than 1 indicated proteins recruited to the signaling complex upon EGF stimulation. A related quantitative proteomics strategy has independently been described (17Ranish J.A. Yi E.C. Leslie D.M. Purvine S.O. Goodlett D.R. Eng J. Aebersold R. Nat. Genet. 2003; 33: 349-355Crossref PubMed Scopus (311) Google Scholar), where isotope-coded affinity tag technology (18Gygi S.P. Rist B. Gerber S.A. Turecek F. Gelb M.H. Aebersold R. Nat. Biotechnol. 1999; 7: 994-999Crossref Scopus (4350) Google Scholar) was used to distinguish two cellular states, leading to the identification of specific members of the yeast polymerase II complex in a large background of nonspecific proteins. Our previous work allowed characterization of multiprotein complexes formed as a result of cell stimulation. Here we employ the principle of SILAC in affinity pull-down experiments for a different purpose: the determination of the interaction partners of short unstructured and potentially modified amino acid sequences, which requires a completely different strategy. We find that synthetic peptides can efficiently be used as baits in such experiments. In contrast to the previous approach, the current screen is designed to primarily reveal direct binders, rather than multiprotein complexes. Furthermore, cloning and expression of domains, or prior knowledge of the signaling events leading to the interaction, are not required, making it more easily adaptable to a large scale approach. The strategy was validated in a classical part of growth factor signaling, namely in the interactions of the adapter protein Grb2. Its role in EGF signaling is to bring the guanine exchange factor Sos to the activated receptor, so that Sos can initiate signaling through Ras at the plasma membrane. We selected a phosphotyrosine-containing peptide of EGFR, which is known to bind the SH2 domain of Grb2, and a proline-rich motif-containing peptide of Sos1 and Sos2 that is known to bind the SH3 domains of Grb2. Both proline-rich Sos peptides retrieved only proteins containing SH3 domains as specific binding partners, demonstrating exquisite specificity in the affinity pull-downs from whole cell lysates. Two of the identified proteins binding to the proline-rich Sos peptide represent novel interactions and were confirmed by co-immunoprecipitation. These proteins, Pacsin3 and Snx9, also have potential roles in growth factor signaling. Metabolic Labeling in Cell Culture—Human HeLa cells were grown in arginine-deficient Dulbecco's modified Eagle's medium with 10% dialyzed fetal bovine serum. One cell population was supplemented with normal isotopic abundance l-arginine (Sigma) and the other with 99% isotopic abundance 13C6-arginine (Aldrich) as described (16Ong S.-E. Kratchmarova I. Mann M. J. Proteome Res. 2003; 2: 173-181Crossref PubMed Scopus (375) Google Scholar). Each population was grown for at least five population doublings. Four plates (14 cm) of each cell population were used per experiment. In some experiments, l-leucine-3,3,3-d3 (Aldrich) and Leu were used for metabolic stable isotope labeling in leucine-deficient Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (15Ong S.-E. Blagoev B. Kratchmarova I. Kristensen D.B. Steen H. Pandey A. Mann M. Mol. Cell. Proteomics. 2002; 1: 376-386Abstract Full Text Full Text PDF PubMed Scopus (4581) Google Scholar). Peptide Synthesis and Pull-downs—Biotinylated peptides were synthesized on a solid-phase peptide synthesizer using amide resin (Intavis, Germany). Peptides were designed as 15-mers bearing an N-terminal biotin on the tetrapeptide linker SGSG (19Ward C.W. Gough K.H. Rashke M. Wan S.S. Tribbick G. Wang J.X. J. Biol. Chem. 1996; 271: 5603-5609Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar). Biotinylation was done using biotinyloxysuccimide (Novabiochem) or sulfo-NHS-SS-biotin (Pierce), the latter allowing reductive cleavage of the 15-mer from the biotin group. The identity and purity of the synthesized peptides were confirmed by mass spectrometric analysis. Peptides were synthesized as pairs as "active" and "control" forms. For affinity pull-downs, 30 nmol of immobilized peptide was added to ∼6 mg of cell lysate. Under such conditions, sub-picomole amounts of specific binding proteins were detected during mass spectrometric analysis, indicating a 1000-fold excess of bait molecules over endogenous binding proteins. Thus, no competition for binding sites is expected. Immobilized streptavidin beads (Pierce) were loaded with biotinylated peptide prior to incubation with cell lysates. Cells were lysed in 1% (v/v) Nonidet P-40, 150 mm sodium chloride, 50 mm Tris-HCl, pH 7.5, protease inhibitors (Complete Tablets, Roche Applied Science), and 1 mm sodium orthovanadate as phosphatase inhibitor. Equal amount of protein was incubated with the respective immobilized peptides at 4 °C for 6 h. After extensive washes, bound proteins were eluted from the immobilized peptides by boiling in SDS sample buffer. Alternatively, the bait peptide with its bound proteins was cleaved off the beads by using 50 mm dithiothreitol. Eluates from the active and control bait peptide pull-downs were combined for further analysis. Immunoprecipitation—HeLa cells were lysed in 1% (v/v) Nonidet P-40, 150 mm sodium chloride, 50 mm Tris-HCl, pH 7.5, 1 mm sodium orthovanadate, and protease inhibitors (Complete Tablets, Roche Applied Science) and incubated at 4 °C with a rabbit antibody against Sos isoforms (Santa Cruz Biotechnology). Protein A-Sepharose beads were added after 4 h and incubated for additional 4 h. After extensive washes, co-precipitated proteins were eluted in sample buffer for SDS-gel electrophoresis. Liquid Chromatography/Mass Spectrometry, Data Base Searching, and Quantitation—Combined eluted proteins were resolved by SDS-PAGE (15% w/v). After silver staining, the entire lane was excised in up to six equally spaced slices and digested enzymatically with trypsin prior to LC-MS analysis. Tryptic peptide mixtures were then desalted (20Rappsilber J. Ishihama Y. Mann M. Anal. Chem. 2003; 75: 663-670Crossref PubMed Scopus (1811) Google Scholar) and loaded onto reversed phase analytical columns for liquid chromatography (21Ishihama Y. Rappsilber J. Andersen J.S. Mann M. J. Chromatogr. A. 2002; 979: 233-239Crossref PubMed Scopus (260) Google Scholar). Peptides were eluted from the analytical column by a multistep linear gradient running from 5 to 60% acetonitrile in 60 min and sprayed directly into the orifice of a QSTAR-Pulsar quadrupole time-of-flight hybrid mass spectrometer (ABI-Sciex). Proteins were identified by high mass accuracy tandem mass spectrometry (liquid chromatography/mass spectrometry) by information-dependent acquisition of fragmentation spectra of doubly, triply, or quadruply charged peptides. Acquired spectra were then searched against the human International Protein Index Data base (www.ebi.ac.uk/IPI) by using the Mascot algorithm (22Perkins D.N. Pappin D.J.C. Creasy D.M. Cottrell J.S. Electrophoresis. 1999; 20: 3551-3567Crossref PubMed Scopus (6776) Google Scholar) with search parameters as described (23Foster L.J. de Hoog C. Mann M. Proc. Natl. Acad. Sci. U. S. A. 2003; 100: 5813-5818Crossref PubMed Scopus (729) Google Scholar). All protein identifications were manually verified against the raw mass spectrometric data by using in-house software. For quantitation, ratios between the centroids of the MS peaks of "heavy" and "light" forms of the peptide were calculated and averaged over consecutive MS cycles for the duration of the respective MS peaks in the total ion chromatogram (16Ong S.-E. Kratchmarova I. Mann M. J. Proteome Res. 2003; 2: 173-181Crossref PubMed Scopus (375) Google Scholar). Heavy and light forms of peptides co-elute for 13C-based labels, and the quantification is based on the average of a number of independently determined ratios for each peptide as the peak elutes from the chromatography column. Labeled and non-labeled peptide pairs were identified by their charge state and mass difference, for example doubly charged peptide pairs were identified through a difference in mass-to-charge ratio of 3. In-house developed software allowed automation of the extraction of corresponding MS peaks. Each ratio was subsequently calculated from the extracted ion chromatograms after manually verifying that the MS spectra containing the respective peaks were at a sufficient level above background and separated from the interfering peaks of other peptides. Ratios obtained from different peptides identifying the same protein were averaged. A final standard deviation for protein ratios was calculated from the ratios of the individual peptides or, when the identification was based on a single peptide, from the ratios obtained from the different mass spectra of this peptide pair. We have employed custom-made software for our instrument. However, the quantification task is very similar to those encountered in the popular isotope-coded affinity tag method (18Gygi S.P. Rist B. Gerber S.A. Turecek F. Gelb M.H. Aebersold R. Nat. Biotechnol. 1999; 7: 994-999Crossref Scopus (4350) Google Scholar) or other quantitative proteomics schemes. Our peptide hit verification, and quantitation software has been made open-source at Sourceforge (sourceforge.net). Furthermore, whereas SILAC has certain advantages in this screen, it would also be possible to perform it with other quantitative proteomic approaches, for which software is already available. The Proteomic Peptide-Protein Interaction Screen—First, two cell populations were metabolically labeled with normal arginine (12C6-Arg) and arginine in which the carbon atoms were replaced by the stable 13C isotope of carbon (13C6-Arg) (Fig. 1). HeLa cells were adapted to the medium for a minimum of five cell doublings to ensure complete labeling of even long lived proteins. This "SILAC" labeling (15Ong S.-E. Blagoev B. Kratchmarova I. Kristensen D.B. Steen H. Pandey A. Mann M. Mol. Cell. Proteomics. 2002; 1: 376-386Abstract Full Text Full Text PDF PubMed Scopus (4581) Google Scholar, 16Ong S.-E. Kratchmarova I. Mann M. J. Proteome Res. 2003; 2: 173-181Crossref PubMed Scopus (375) Google Scholar) makes arginine-containing peptides distinguishable between the two populations without introducing any other chemical or biological differences. In particular, both forms of each peptide are detected with equal efficiency by the mass spectrometer so that the measured ratio between the two peptides accurately reflects the ratio of the proteins in the two populations. Cell lysates from the 12C6-Arg-labeled and the 13C6-Arg-labeled cells were employed in affinity purification using synthetic peptides, representing the non-phosphorylated and the phosphorylated form of a peptide, respectively. The unphosphorylated peptide was exposed to the 12C6-Arg-encoded cell lysate, and the phosphorylated peptide was exposed to the 13C6-Arg-encoded lysate. Lysates were then mixed and analyzed together (Fig. 1). Background proteins (for example, ribosomal proteins binding to the beads to which the peptides are attached) are present in equal amounts in both pull-down experiments. Upon analysis by mass spectrometry, peptides from such proteins in the two forms with equal abundance. a protein specifically to the and not to the non-phosphorylated peptide, then it only be in the cell lysate. in this we only the form of the a protein specifically but not to one of the two forms of the then this to a ratio between the of peaks that is related to the specificity of the binding to one form over the other. the above any other that can be synthesized can be for protein binding as of peptides be by using peptides as the of as pull-down experiments to the strategy in a bait peptide was of EGFR, which is known to bind the adapter protein Grb2 upon of the (19Ward C.W. Gough K.H. Rashke M. Wan S.S. Tribbick G. Wang J.X. J. Biol. Chem. 1996; 271: 5603-5609Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar, D. J.M. J. Mol. Cell. Biol. PubMed Google Scholar, T. J. Biol. Chem. 1996; 271: Full Text Full Text PDF PubMed Scopus Google Scholar). The peptide containing a biotin for to beads and a SGSG (19Ward C.W. Gough K.H. Rashke M. Wan S.S. Tribbick G. Wang J.X. J. Biol. Chem. 1996; 271: 5603-5609Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar), was synthesized in the and Lysates of 13C6-Arg-labeled HeLa cells were incubated with the phosphorylated peptide, whereas lysates of normal amino HeLa cells were incubated with the unphosphorylated control proteins were digested to peptides, and analyzed by mass spectrometry as described The peptide bait Grb2 from the cell (Fig. indicating a sufficient excess of bait of proteins only Grb2 pairs of peptides with a ratio as for those proteins specifically binding to the In only the forms of the Grb2 peptides were detected above (Fig. indicating a ratio of more than between labeled and forms. In binding proteins, such as ribosomal proteins, were detected in labeled and forms with ratios to In addition, were present in amounts that to ion during mass spectrometry (Fig. This that it would have been very to identify the low abundant specific in a affinity pull-down through of proteins after electrophoresis. the a was using of arginine for metabolic and Grb2 was detected as the only specific identified in peptide affinity pull-downs using a tyrosine-phosphorylated peptide of or a proline-rich of Sos1 and Ratios are as to 12C6 for the pull-down and as 12C6 to for the experiments. Ratios that only one isotopic form was detected at the given Proteins in were verified in experiments. no arginine-containing not in were verified by in a of experiments using a of EGF receptor as mass of a doubly charged 13C6-Arg-labeled peptide of Grb2 as identified through the interaction with the tyrosine-phosphorylated peptide of mass of a doubly charged peptide of ribosomal protein a nonspecific isotopic ratios of proteins identified also in experiment. Proteins with ratios of were with the form incubated with the isotopic ratios of proteins identified in pull-down and experiments by using a specific binding protein was detected in this experiment. to any potential in the binding of background proteins, experiments were to the strategy described in In these experiments, the active peptide is incubated with the cell of the light amino acid containing proteins, whereas the control peptide is incubated with the heavy amino acid containing proteins. proteins that specifically bind to the bait have to 12C6 ratios equal to the 12C6 to ratios in the (Fig. were only such proteins because most of the background proteins were identified only in one of the two experiments. Grb2 was detected with a ratio of greater than in both and is therefore confirmed as a specific to the other proteins large ratios in one but not the other experiment. In particular, proteins these to these proteins to in We have that the for ratios in the is a for specific binding and background proteins in all of binding to SH2 domains is determined by the amino acids to the and it is known that exchange of the amino acid in the of the peptide used is sufficient to binding J. P. T. M. H. H. Yi T. Mol. Cell. Biol. PubMed Scopus Google Scholar). a further for the specificity of our binding we synthesized a peptide where the was to an This single amino acid change to the complete of Grb2 in the of identified proteins, and of the identified proteins an isotopic ratio of a specific (Fig. We cannot that other specific binding proteins that were present at amounts detection However, the that Grb2 was in experiments with different peptides and with a ratio of more than to that such proteins would have to be present at lower abundance than Grb2. Peptides of Sos1 and Sos2 as to its SH2 Grb2 has two SH3 domains that can bind proline-rich of the Ras guanine exchange factor In a the specificity of the screen for lower affinity interactions was by using one of the proline-rich motifs of the two isoforms Sos1 and The peptides representing amino acids of Sos1 and amino acids of Sos2 the classical II for SH3 domain M. D. F. M. F. B. C. J. Biol. Chem. Full Text Full Text PDF PubMed Scopus Google Scholar). Both peptides bind Grb2 with an affinity of about 4 which is than the interaction of the SH2 domain of Grb2 with the peptide used above T. J. Biol. Chem. 1996; 271: Full Text Full Text PDF PubMed Scopus Google Scholar, M. D. F. M. F. B. C. J. Biol. Chem. Full Text Full Text PDF PubMed Scopus Google Scholar, L. D. J. Biol. Chem. Full Text Full Text PDF PubMed Scopus Google Scholar). control baits to the strategy in were replaced with to F. M. F. B. C. Nat. Biol. 1: PubMed Scopus (103) Google Scholar). Both proline-rich peptides specifically a small number of proteins of more than identified proteins. Grb2 was identified in the labeled indicating that the screen well with motifs of SH3 domain interactions (Fig. In addition, Pacsin3 and Snx9, which a