Combinatorial Approach for Large-scale Identification of Linked Peptides from Tandem Mass Spectrometry Spectra

Abstract

The combination of chemical cross-linking and mass spectrometry has recently been shown to constitute a powerful tool for studying protein–protein interactions and elucidating the structure of large protein complexes. However, computational methods for interpreting the complex MS/MS spectra from linked peptides are still in their infancy, making the high-throughput application of this approach largely impractical. Because of the lack of large annotated datasets, most current approaches do not capture the specific fragmentation patterns of linked peptides and therefore are not optimal for the identification of cross-linked peptides. Here we propose a generic approach to address this problem and demonstrate it using disulfide-bridged peptide libraries to (i) efficiently generate large mass spectral reference data for linked peptides at a low cost and (ii) automatically train an algorithm that can efficiently and accurately identify linked peptides from MS/MS spectra. We show that using this approach we were able to identify thousands of MS/MS spectra from disulfide-bridged peptides through comparison with proteome-scale sequence databases and significantly improve the sensitivity of cross-linked peptide identification. This allowed us to identify 60% more direct pairwise interactions between the protein subunits in the 20S proteasome complex than existing tools on cross-linking studies of the proteasome complexes. The basic framework of this approach and the MS/MS reference dataset generated should be valuable resources for the future development of new tools for the identification of linked peptides. The combination of chemical cross-linking and mass spectrometry has recently been shown to constitute a powerful tool for studying protein–protein interactions and elucidating the structure of large protein complexes. However, computational methods for interpreting the complex MS/MS spectra from linked peptides are still in their infancy, making the high-throughput application of this approach largely impractical. Because of the lack of large annotated datasets, most current approaches do not capture the specific fragmentation patterns of linked peptides and therefore are not optimal for the identification of cross-linked peptides. Here we propose a generic approach to address this problem and demonstrate it using disulfide-bridged peptide libraries to (i) efficiently generate large mass spectral reference data for linked peptides at a low cost and (ii) automatically train an algorithm that can efficiently and accurately identify linked peptides from MS/MS spectra. We show that using this approach we were able to identify thousands of MS/MS spectra from disulfide-bridged peptides through comparison with proteome-scale sequence databases and significantly improve the sensitivity of cross-linked peptide identification. This allowed us to identify 60% more direct pairwise interactions between the protein subunits in the 20S proteasome complex than existing tools on cross-linking studies of the proteasome complexes. The basic framework of this approach and the MS/MS reference dataset generated should be valuable resources for the future development of new tools for the identification of linked peptides. The study of protein–protein interactions is crucial to understanding how cellular systems function because proteins act in concert through a highly organized set of interactions. Most cellular processes are carried out by large macromolecular assemblies and regulated through complex cascades of transient protein–protein interactions (1Robinson C.V. Sali A. Baumeister W. The molecular sociology of the cell.Nature. 2007; 450: 973-982Crossref PubMed Scopus (415) Google Scholar). In the past several years numerous high-throughput studies have pioneered the systematic characterization of protein–protein interactions in model organisms (2Uetz 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. Vijayadamodar G. Yang M. Johnston M. Fields S. Rothberg J.M. A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae.Nature. 2000; 403: 623-627Crossref PubMed Scopus (3914) Google Scholar, 3Giot L. Bader J.S. Brouwer C. Chaudhuri A. Kuang B. Li Y. Hao Y.L. Ooi C.E. Godwin B. Vitols E. Vijayadamodar G. Pochart P. Machineni H. Welsh M. Kong Y. Zerhusen B. Malcolm R. Varrone Z. Collis A. Minto M. Burgess S. McDaniel L. Stimpson E. Spriggs F. Williams J. Neurath K. Ioime N. Agee M. Voss E. Furtak K. Renzulli R. Aanensen N. Carrolla S. Bickelhaupt E. Lazovatsky Y. DaSilva A. Zhong J. Stanyon C.A. Finley R.L. White K.P. Braverman M. Jarvie T. Gold S. Leach M. Knight J. Shimkets R.A. McKenna M.P. Chant J. Rothberg J.M. A protein interaction map of Drosophila melanogaster.Science. 2003; 302: 1727-1736Crossref PubMed Scopus (1917) Google Scholar, 4Rual J.-F. Venkatesan K. Hao T. Hirozane-Kishikawa T. Dricot A. Li N. Berriz G.F. Gibbons F.D. Dreze M. Ayivi-Guedehoussou N. Klitgord N. Simon C. Boxem M. Milstein S. Rosenberg J. Goldberg D.S. Zhang L.V. Wong S.L. Franklin G. Li S. Albala J.S. Lim J. Fraughton C. Llamosas E. Cevik S. Bex C. Lamesch P. Sikorski R.S. Vandenhaute J. Zoghbi H.Y. Smolyar A. Bosak S. Sequerra R. Doucette-Stamm L. Cusick M.E. Hill D.E. Roth F.P. Vidal M. Towards a proteome-scale map of the human protein-protein interaction network.Nature. 2005; 437: 1173-1178Crossref PubMed Scopus (2284) Google Scholar). Such studies mainly utilize two techniques: the yeast two-hybrid system, which aims at identifying binary interactions (5Fields S. Song O. A novel genetic system to detect protein protein interactions.Nature. 1989; 340: 245-246Crossref PubMed Scopus (4860) Google Scholar), and affinity purification combined with tandem mass spectrometry analysis for the identification of multi-protein assemblies (6Rigaut G. Shevchenko A. Rutz B. Wilm M. Mann M. Séraphin B. A generic protein purification method for protein complex characterization and proteome exploration.Nat. Biotechnol. 1999; 17: 1030-1032Crossref PubMed Scopus (2286) Google Scholar, 7Collins M.O. Choudhary J.S. Mapping multiprotein complexes by affinity purification and mass spectrometry.Curr. Opin. Biotechnol. 2008; 19: 324-330Crossref PubMed Scopus (102) Google Scholar, 8Gingras A.-C. Gstaiger M. Raught B. Aebersold R. Analysis of protein complexes using mass spectrometry.Nat. Rev. Mol. Cell Biol. 2007; 8: 645-654Crossref PubMed Scopus (559) Google Scholar). Together these led to a rapid expansion of known protein–protein interactions in human and other model organisms. Patche and Aloy recently estimated that there are more than one interactions to R.A. Aloy P. high-throughput identification of the 2008; 8: PubMed Scopus Google Scholar). rapid most current one to proteins which is the understanding how proteins A more from the of protein which that how interactions and the the the methods to protein are and and there have been several to S. in to the of protein complexes H. K. H. The a of 2007; PubMed Scopus Google Scholar, Scholar). there has been of for protein in the in years H. K. H. The a of 2007; PubMed Scopus Google Scholar), the of for protein complexes has R.A. Aloy P. high-throughput identification of the 2008; 8: PubMed Scopus Google Scholar). their large transient and of have complexes from approaches in the development of with which to the structure of large protein complexes to D. T. characterization of proteins and complexes using Biol. PubMed Scopus Google Scholar, C. H. Y. C. T. in methods for molecular Rev. 2008; PubMed Scopus Google Scholar, K. protein with mass spectrometry to the structure of macromolecular Rev. PubMed Scopus Google Scholar, H. T. of the and current Biol. 2008; PubMed Scopus Google Scholar, A. cross-linking and mass spectrometry to map protein and protein-protein Rev. PubMed Scopus Google Scholar, T. S. P. for the and analysis of protein-protein 2007; PubMed Scopus Google Scholar). have the analysis of protein and interaction by cross-linking and tandem mass spectrometry A. cross-linking and mass spectrometry to map protein and protein-protein Rev. PubMed Scopus Google Scholar, F. A. D. R. A. T. A. M. N. L. Aebersold R. of a protein by chemical cross-linking and mass PubMed Scopus Google Scholar, H. N. of protein-protein interactions and in with chemical cross-linking and mass 8: PubMed Scopus Google Scholar, A. L. C. S. T. M. L. M. P. J. of the complex by cross-linking and mass J. PubMed Scopus Google Scholar, S. E. A. D. structure by systematic of J. PubMed Scopus Google Scholar, N. M. of by mass and Scopus Google Scholar, F. Aebersold R. C.V. and cross-linking with the mass spectrometry of PubMed Scopus Google Scholar). The basic this is to capture and identify of that are to these linked of are from the protein that one the of protein of from proteins how proteins with one cross-linking a N. G. protein identification by using from and mass 2000; Scopus Google Scholar, Scholar), in the complex MS/MS generated from linked peptides this approach and therefore it not in mass spectrometry there has been in this to protein and identify protein–protein interactions. However, most studies have been on protein complexes. mass of of thousands of spectra in a it is to this approach to the analysis of complex have to this using and computational a of chemical cross-linking are for these and are to for peptide identification in concert with analysis H. N. of protein-protein interactions and in with chemical cross-linking and mass 8: PubMed Scopus Google Scholar, studies of the complex by chemical cross-linking using the of for of peptide and their PubMed Scopus Google Scholar, analysis of a using a novel and mass PubMed Scopus Google Scholar). have been several to computational methods for the identification of linked peptides from MS/MS spectra P. A. C. R.A. P. of protein mass spectrometry and an 2008; PubMed Scopus Google Scholar, O. J. T. L. M. A. M. Aebersold R. of cross-linked peptides from large sequence 2008; PubMed Scopus Google Scholar, F. a for identification of cross-linked PubMed Scopus Google Scholar, F. J. E. L. mass analysis of and in proteins and protein J. 2005; Scopus Google Scholar, H. Zhang L. and characterization of in proteins and peptides from tandem data by of the MS/MS 2007; PubMed Scopus Google Scholar, S. P. P. cross-linked peptides by a of cross-linked peptide PubMed Scopus Google Scholar, G. J. S. K. F. R. A. for in protein interaction of The for Google Scholar, B. M. J. Zhang K. Li S. H. Li Hao Z. S. K. M. of cross-linked peptides from complex PubMed Scopus Google Scholar). However, because of the lack of large annotated most approaches to fragmentation from peptides the fragmentation from data of O. J. T. L. M. A. M. Aebersold R. of cross-linked peptides from large sequence 2008; PubMed Scopus Google Scholar, S. J. S. H.Y. E. algorithm for the identification of on fragmentation in tandem mass Scopus Google Scholar), which not In it is to generate large it is and of B. M. J. Zhang K. Li S. H. Li Hao Z. S. K. M. of cross-linked peptides from complex PubMed Scopus Google Scholar). disulfide-bridged peptides an we propose a novel method that a peptide to efficiently generate a large mass spectral reference dataset for linked peptides and these data to automatically train new which can efficiently and accurately identify linked peptides from MS/MS spectra. We peptide libraries with the sequence The in that were at in and were at the The sequence patterns were to several to identification. one for there is in the The of is known to MS/MS spectra with fragmentation F. for peptide tandem 2003; PubMed Scopus Google the mass of disulfide-bridged peptide that can be from the peptides and the sequence patterns were to for the of peptides with the in with to the of of the peptides using peptide to generate peptides and disulfide-bridged peptide This a large dataset to the fragmentation patterns of disulfide-bridged peptides at the a that MS/MS spectra from disulfide-bridged peptides can be using a in which linked peptide are the peptide libraries were in that allowed the of disulfide-bridged and were using tandem mass of peptides were with for at were using a in of and on an were an for by on a system were a with a of and a of to A is and is with a analysis of were an with a of and were using an mass were in a mass at MS/MS in the with the in with the most for MS/MS spectra from the disulfide-bridged peptide libraries were using a in an using a model from peptides E. F. Cell 2008; PubMed Scopus Google a peptides. peptides are a specific of linked peptide in which the peptide is linked to a of a The of the identification of peptides are J. J. J. J.R. N. A approach for identification of complex Scholar). set of MS/MS spectra from disulfide-bridged peptides were at a mass these a model specific for disulfide-bridged peptides and to the spectra from peptide libraries one more to a of MS/MS spectra from disulfide-bridged peptides. with were with a mass and mass were for a mass of than and an of using a approach for in protein by mass spectrometry.Nat. 2007; PubMed Scopus Google tandem mass spectrometry In to the between a cross-linked peptide and an MS/MS we a linked peptide a of two a at the cross-linked with mass to that of the mass of the other peptide the mass of the In a one to how a peptide to an MS/MS cross-linked peptides one how a of peptides to an MS/MS In J. N. identification by of tandem mass Scopus Google Scholar), we a model to how a of peptides to a MS/MS from peptides. The framework that of by the specific fragmentation of linked peptides. an MS/MS is a of a mass of on MS/MS is a the from most to in a of a peptide is a the of the with mass in that The model peptide fragmentation by using a set of annotated MS/MS spectra to the that of an with a a model can be using in the the The function for a peptide is the of the that the is generated from the peptide the that the is generated from This is a model to that by in S. N. N. peptide with of tandem mass 8: PubMed Scopus Google and in S. N. and of tandem mass a 2008; PubMed Scopus Google and S. N. N. L. S. The function of and of tandem mass to PubMed Scopus Google Scholar). Because linked peptides are of we can a linked peptide two The from the and the from the a of we the the peptide that for the most in the MS/MS and we the other peptide the This is to for in fragmentation patterns between the and peptides. The of a to a of peptides is is to model the between the two peptides. from and the the model the with using the to the identification of linked peptides. not of peptide In to capture the fragmentation of cross-linked the set of linked and linked are that are linked to a for that is to we linked in in the current the were for the of a linked peptides the were to the that can be generated from linked peptides. from to the of the MS/MS were these new the fragmentation specific to linked peptide can be from a set of linked peptide and can be for linked and a function that the fragmentation of linked we can how a cross-linked peptide an MS/MS However, one still to cross-linked peptide in the sequence to the the protein sequence is it is not to peptide in the In this is to the problem of identifying spectra from peptides N. Li M. S. B. Aebersold R. an of identifying peptides from a by a tandem mass 2005; PubMed Scopus Google Scholar, J. J. P. N. identification from tandem mass PubMed Scopus Google Scholar, J. N. A. R.A. Mann M. a peptide the PubMed Scopus Google the mass for of the peptides is known in the of which can the of peptides. linked peptides the combined mass of the two peptides is therefore the identification of linked peptides can be of identifying two peptides with an at the P. A. C. R.A. P. of protein mass spectrometry and an 2008; PubMed Scopus Google Scholar, F. a for identification of cross-linked PubMed Scopus Google Scholar). to approaches P. A. C. R.A. P. of protein mass spectrometry and an 2008; PubMed Scopus Google Scholar, O. J. T. L. M. A. M. Aebersold R. of cross-linked peptides from large sequence 2008; PubMed Scopus Google Scholar, F. a for identification of cross-linked PubMed Scopus Google Scholar), a to the cross-linked peptide Because a linked peptide is a of of which has a of the mass of the the other we on the that for a linked peptide that the at one peptide should have a to the by In in the of the peptide is the by in the the peptides are to the peptide be a with mass and be a peptide with mass be a peptides. A peptide is peptide with a mass of at the peptide in the we of is the mass between the mass of the and the mass of the peptide and for the peptide in the of disulfide-bridged are peptides are the and by shown in the data were a of peptide with the E. one of the peptides in the peptides and the other peptide in the in more than of than peptide can peptide and still the in more than of In the is because peptide are that their combined the mass of the MS/MS the to has been the are using a to from in the are (i) in by the of in the (ii) of MS/MS by the in the MS/MS of of in the by the of from the peptide of the of and and mass between and that we can the for of the linked peptides. the combined that from cross-linked peptides in constitute the set of in the The of the were from the MS/MS spectra from the of peptides in the the were data and from the were were to their and were their a The to a a set of the is estimated using a approach for in protein by mass spectrometry.Nat. 2007; PubMed Scopus Google Scholar, T. M. A. F. S. F. M. Aebersold R. for cross-linked peptides by mass spectrometry.Nat. PubMed Scopus Google Scholar). because has two peptide it can one of the in which peptide are from the in which one peptide is from the and is from the in which peptide are from the we the of of a is the of of we can the for We the data from a cross-linked of and proteasome from a study A. R. T. F. S. F. Aebersold R. the chemical cross-linking by the of and by PubMed Scopus Google Scholar). The from were from the A. R. T. F. S. F. Aebersold R. the chemical cross-linking by the of and by PubMed Scopus Google Scholar), and the were by the with a mass and the mass for spectra. The were with a mass and for the proteasome complex were from A. R. B. S. E. H. R. M. protein 2005; Google by proteins from the that the using the function of the cross-linked we of the proteasome complexes of from the J. Z. G. H. The protein data 2000; PubMed Scopus Google Scholar). Because the for and S. are not the S. proteasome were to the structure of the Saccharomyces proteasome L. J.R. Y. Hill for the of by 2000; PubMed Scopus Google and the proteasome were to the structure of proteasome M. R. W. M. M. and proteasome in and PubMed Scopus Google Scholar). The by of from the proteasome complex of the two using the sequence function in system for and PubMed Scopus Google Scholar). The of the sequence of known structure were to of the sequence with the between the of the in cross-linked peptide of than were to be the of the on the that can be by the the of the of a and the between the of protein the we peptides by the data protein using with the on on and and on identification of peptides with a and cross-linked there are mass because the and were at a and proteins at one peptide were proteins for the of the peptides proteins in proteins are computational in the identification of linked peptides. the of two peptides the of the peptides and that fragmentation patterns than by existing for peptides. spectra from linked peptides from two MS/MS tools that from a The of two peptides in the a for peptide are therefore to this there are a of spectra that are to the fragmentation for linked significantly the development of approaches for peptides S. N. N. L. S. The function of and of tandem mass to PubMed Scopus Google Scholar, A function for 8: PubMed Scopus Google Scholar, for peptide data analysis by of and PubMed Scopus Google Scholar). In to address these we and peptide with a at the peptides. The peptide libraries were allowed to disulfide-bridged and were with an mass MS/MS spectra were using a A of MS/MS spectra from disulfide-bridged peptides were to this is the MS/MS dataset of linked peptides for the development of new tools for the identification of linked peptides. these the fragmentation patterns of disulfide-bridged peptides were We from linked peptides linked and are that linked to a peptide and were to have fragmentation patterns than are in linked from a are in from the we that linked and from disulfide-bridged peptides fragmentation patterns than from peptides. In in the MS/MS spectra than the of from peptides In for linked highly to be more than in peptides because linked are to peptide that and peptides have at the that have a affinity for it that of linked peptides to have fragmentation patterns the fragmentation patterns of disulfide-bridged peptides and of peptides E. F. Cell 2008; PubMed Scopus Google Scholar), which are a of linked peptide in which the of the peptide is to a of peptide are in disulfide-bridged peptides and are in the in the it is to a model for of linked peptide in to the sensitivity of their identification from MS/MS spectra. In to for the fragmentation of linked for linked and linked and were to linked and a linked peptide an MS/MS address the that an MS/MS from linked peptides from two a fragmentation model to that in J. N. identification by of tandem mass Scopus Google to for the of two peptides In a in which it one peptide and the other peptide in the linked and This to the of peptide by to of still identifying the linked peptides and for In to in the identification of disulfide-bridged the MS/MS spectra from disulfide-bridged peptide libraries were peptide with an model from we an set of MS/MS spectra from disulfide-bridged peptides. this specific for disulfide-bridged peptides were and to identify an to 60% more MS/MS spectra from peptide This that of linked peptides have fragmentation patterns and that these fragmentation patterns the sensitivity of the identification of linked peptides from mass spectra. of approaches for the identification of linked peptides is to the identification of linked peptides in more complex we the to large sequence We this by E. and human protein databases in their to the of peptide and the MS/MS spectra from the peptide libraries these The of on sensitivity in identifying disulfide-bridged peptides is shown in The that we the of the by a of which to a in the of linked there on a to in In it is that in the of the in the sensitivity of we a yeast dataset J. Zhang B. protein identification and data in Biol. Scopus Google using S. N. N. L. S. The function of and of tandem mass to PubMed Scopus Google Scholar), a tool for peptides. and spectra with and mass were using a mass this to a to the that the of peptides do not mass This there is an in sensitivity the of the is in sensitivity is than that in tools for this is by the in the of linked peptides the for peptides. with the for in these demonstrate to identify linked peptides proteome-scale sequence the of in we it to the identification of cross-linked peptides on two MS/MS one from the S. proteasome and from the 20S proteasome complex A. R. T. F. S. F. Aebersold R. the chemical cross-linking by the of and by PubMed Scopus Google Scholar). complexes are for the of proteins the and the complexes is known because of their D. and of by the Rev. PubMed Scopus Google Scholar). model complexes for cross-linking linked peptides can be using known we can in in able to identify significantly more cross-linked peptides than the two current tools for cross-linked O. J. T. L. M. A. M. Aebersold R. of cross-linked peptides from large sequence 2008; PubMed Scopus Google and B. M. J. Zhang K. Li S. H. Li Hao Z. S. K. M. of cross-linked peptides from complex PubMed Scopus Google Scholar). This in allowed to identify 60% more pairwise interactions between the protein subunits in the proteasome complex This that the fragmentation from disulfide-bridged peptides are generic for the identification of linked peptides. the of the the cross-linked peptides were to proteins with in the H. K. H. The a of 2007; PubMed Scopus Google Scholar). The between the cross-linked and is shown in The of the cross-linked peptides of the of the cross-linked peptides than which is to the of the This is with the that in the using the approach for in protein by mass spectrometry.Nat. 2007; PubMed Scopus Google Scholar). The be to the on the structure of a protein complex with in between identify linked peptides large sequence the proteasome dataset the databases of proteasome protein to the E. and protein sequence shown in still able to identify a large of cross-linked there a in sensitivity with in to improve the sensitivity of one the with the However, in with the analysis of the proteasome complexes are in studies the identification of protein the of the protein subunits be address this on the that in cross-linking there are peptides in the that can be to the protein of the we a in which a of proteins a for peptides and peptides with a using a to identify cross-linked peptides this set of shown in this able to to of the cross-linked peptides in the proteasome proteome-scale with the a between the protein of the and the sensitivity of identification of linked peptides. This is to high-throughput protein interaction studies in which protein complexes are from cellular using affinity purification cross-linking by tandem mass spectrometry is a for the analysis of protein and protein–protein interactions. However, there are several that to be it can be on a large The development of novel studies of the complex by chemical cross-linking using the of for of peptide and their PubMed Scopus Google Scholar, analysis of a using a novel and mass PubMed Scopus Google has to linked peptides from a large of other in complex and their analysis using mass the on these novel how the linked peptides in mass Here we show that computational methods that capture the fragmentation patterns of linked peptides can improve to identify from MS/MS spectra. recently shown that peptides that are of fragmentation patterns in MS/MS and these fragmentation can improve the identification of peptides from MS/MS spectra S. N. N. L. S. The function of and of tandem mass to PubMed Scopus Google Scholar). Because of the of linked the development of fragmentation is more crucial for their identification. However, this is because there are reference to their fragmentation an and computational on peptide we have shown that it is to efficiently generate a large mass spectral reference dataset for linked peptides at low reference data that linked peptides do have fragmentation patterns than most current tools from peptides are for the identification of linked peptides. specific fragmentation and in we have shown that it is to identify disulfide-bridged peptides proteome-scale sequence the in the development of and methods for the identification of linked on peptide can be to of linked and and the development of computational tools for the identification of other of linked peptides. The data generated in this study have been in the and we to be a valuable for the future development of for the identification of linked peptides. with