Automated and High Confidence Protein Phosphorylation Site

Oct 12, 2012 - and Electron Transfer Dissociation Tandem Mass Spectrometry. Thomas A. Hansen, Marc Sylvester,. †. Ole N. Jensen, and Frank Kjeldsen*...
0 downloads 0 Views 1MB Size
Download PDF
Technical Note pubs.acs.org/ac

Automated and High Confidence Protein Phosphorylation Site Localization Using Complementary Collision-Activated Dissociation and Electron Transfer Dissociation Tandem Mass Spectrometry Thomas A. Hansen, Marc Sylvester,† Ole N. Jensen, and Frank Kjeldsen* Protein Research Group, Department of Biochemistry and Molecular Biology, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark S Supporting Information *

ABSTRACT: Reversible protein phosphorylation plays a critical role in cell signaling and is responsible for the regulation of many biological processes in most living organisms. The low stoichiometry of protein phosphorylation requires sensitive analysis by tandem mass spectrometry. However, incomplete peptide fragmentation and the loss of labile phosphate groups complicate identification of the site of the phosphate motif. Here, we have implemented and evaluated a novel approach for phospho-site localization by the combined use of peptide tandem mass spectrometry data obtained using both collision-activated dissociation and electron transfer dissociation, an approach termed the Cscore. The scoring algorithm used in the Cscore was adapted from the widely used Ascore method. The analytical benefit of integrating the product ion information of both ETD and CAD data are evident by increased confidence in phospho-site localization and the number of assigned phospho-sites at a fixed false-localization rate. The average calculated Cscore from a large data set (>7000 phosphopeptide MS/MS spectra) was ∼32 compared to ∼23 and ∼17 for the Ascore using collision-activated dissociation (CAD) or electron transfer dissociation (ETD), respectively. Compared with the Ascore using either CAD or ETD, the Cscore identified up to 88% more phosphorylation sites. Using a phosphopeptide library revealed that the score threshold for obtaining a false-localization rate of 0.5% was lower for the Cscore than either the Ascore (CAD) or the Ascore (ETD).

P

estimates the correctness of the peptide sequences, not the position of phosphorylations or other PTMs. If two potentially phosphorylated amino acid (AA) residues exist in a singly phosphorylated peptide and no detectable fragment ions resulting from the cleavage between these residues are present, it is unlikely, if not impossible, to determine the true phosphorylation site regardless of the quality of the remaining mass spectrum. This is exemplified in Figure 1, which shows the collision-activated dissociation (CAD) product mass spectrum of the peptide with the sequence S(pST)PLPTLSSSAENTR. From this MS/MS spectrum, the single phosphorylation could be restricted to the AA residues Ser2-Thr3. However, since no product ions were recorded from the backbone bond cleavage between Ser2-Thr3 it is not possible to correctly assign the phosphorylated AA residue with any confidence. In many biological and clinical studies, the incorrect assignment of a phosphorylated AA residue can jeopardize the biological conclusions since proposed phosphorylation sites identified in large-scale studies are submitted into databases for the use of other scientists. It is therefore essential that

ost-translational phosphorylation of proteins is a key mechanism in the regulation of many cellular processes.1 Most signal transduction pathways eventually involve protein phosphorylation or dephosphorylation by kinases and phosphatases2 that dictates the cellular response to internal or external cues. Hence, identification and quantification of protein phosphorylation can improve our understanding of cellular mechanisms, diseases, and biology, in general. Tandem mass spectrometry (MS/MS) has become the method of choice for these studies because of its high sensitivity, speed, and accuracy.3 Recent developments in mass spectrometry (MS) instrumentation as well as strategies for phosphopeptide enrichment have made large-scale studies of protein phosphorylation feasible. In several studies, thousands of phosphorylation sites have been identified and quantified.4−6 The size and complexity of these results, however, call for computer algorithms to fully exploit the obtained information.7 One major issue regarding the large phosphopeptide data sets is the uncertainty of site-localization of protein phosphorylations. Protein database search engines assign a score to the identified phosphorylated peptides but do not present any score to denote the confidence level of the individual phosphorylation site determination. For instance, estimating the false discovery rate using a decoy database8 only © 2012 American Chemical Society

Received: August 16, 2012 Accepted: October 12, 2012 Published: October 12, 2012 9694

dx.doi.org/10.1021/ac302364r | Anal. Chem. 2012, 84, 9694−9699

Analytical Chemistry

Technical Note

Figure 1. CAD MS/MS giving rise to a confident assignment of a phosphorylated peptide S(pST)PLPTISSSAENTR (mascot ion score of 74). The site of phosphorylation could not be assigned with confidence since no fragment ions representing backbone cleavages between Ser2-Thr3 were recorded.

tryptic peptides.23 Therefore, fragmenting peptides using CAD and ETD sequentially leads to higher overall peptide sequence coverage and confirmation of phosphorylation sites by complementary ion types. This has the potential to improve the confidence in phosphorylation-site localizations obtained. The relations between the information obtained from phosphopeptides of both CAD and ETD has not been automated into a scoring strategy. Herein, we introduce the Cscore (Complementary score), a scoring algorithm for automatic phosphorylation site validation based on the Ascore algorithm, which uses the combined mass spectral data of both CAD and ETD. The increase in information obtained by using two orthogonal fragmentation techniques increases the confidence in the site-localization. The features explored with Cscore should be a complement to typical bottom-up workflows and attractive in many phosphoproteomics studies.

confidence in phosphorylation site localization is high. Manual validation of the phospho-site assignments has been used to surmount this problem. This analysis, however, is biased by the assessment of the person doing the validation. Moreover, the large size of the data sets makes this approach impractical because of the high workload involved. To overcome that problem, several scoring algorithms have been developed to identify the correct phosphorylation site and determine the confidence in its localization.9−13 The Ascore11 is widely used in the phosphoproteomics field. The Ascore is based on the calculation of a cumulative binomial probability determined using exclusively the informative phospho-site determining ions. This implies that the Ascore would be 0 for the spectrum shown in Figure 1 since no site determining ions were recorded. Therefore, using the site determining ions, the Ascore gives rise to one score per phosphorylation site, not per spectrum or sequence, because the site determining ions are a different set of ions for the different phosphorylation sites. Although CAD has been applied successfully in many largescale phosphopeptide studies, it has some analytical limitations. The primary disadvantage of CAD is that one of the predominant fragmentation channels of phosphopeptide ions is that resulting in loss of phosphoric acid.14−16 This can complicate the possibilities for assigning phospho-site localizations. On the contrary, application of the milder fragmentation techniques of electron capture dissociation17 (ECD) and electron transfer dissociation18 (ETD) leave the phosphate group attached to the phosphorylated AA residue during peptide backbone fragmentation. This feature in combination with a general higher sequence coverage in ETD/ECD than CAD improve the ability to identify the correct phosphorylation site.19 However, the fragmentation efficiency of ECD/ETD is highest if the charge state of the precursor ions is 3+ or higher.20,21 ECD/ETD and CAD are complementary to each other as evidenced by examining the differences in the product ion types formed,17,22 differences in preferred peptide charge state, as well the poor correlation between the cleavage frequency between specific AA residues in

■

EXPERIMENTAL SECTION

Sample Preparation. All chemicals were purchased from Sigma-Aldrich unless stated otherwise. The human lung carcinoma cell line A549 was cultured in RPMI1640 (with stabilized glutamine, Invitrogen Ltd.) with 10% fetal bovine serum (Sigma-Aldrich) and left untreated or treated with 40 ng/mL human hepatocyte growth factor (R&D Systems). Cells were washed with cold PBS and lysed on ice with 20 mM triethylammonium bicarbonate (TEAB, pH 8.5), 5% sodium deoxycholate (SDC), 20 mM dithiothreitol, protease inhibitor mix III (Calbiochem), phosphatase inhibitor mix 2, and 1 mM sodium orthovanadate. Cell lysis was completed with tip sonication, followed by heating to 60 °C for 10 min. Cleared lysate corresponding to 300 μg of protein was applied to a micro spin filter with 10 kDa cutoff (Nanosep Omega, Pall Corporation), washed with 20 mM TEAB, and alkylated with iodoacetamide. Digestion with trypsin was performed for 4 h at room temperature in the presence of 0.5% SDC. Peptides were collected by centrifugation, and SDC was precipitated with 0.5% trifluoroacetic acid. Phosphopeptides were enriched using essentially the SIMAC procedure as described.24 Briefly,

9695

dx.doi.org/10.1021/ac302364r | Anal. Chem. 2012, 84, 9694−9699

Analytical Chemistry

Technical Note

Figure 2. Example demonstrating the use of mass spectral information gained from both CAD and ETD data to calculate the Ascore and the Cscore. For RGSPTTGFIEQK, the two sites with the highest peptide score (the score based on all peaks) were S3 and T5.

Implementation. The algorithms to calculate the Ascore and the Cscore were implemented using the Java software tool PhosphoSiteLocalizer, which takes Mascot search result XMLfiles as input and reports the Ascore and the Cscore of all assigned phosphopeptides in a table. PhosphoSiteLocalizer is freely available at http://composition.sdu.dk/

peptides were applied to a Fe2+ chelate resin, flow-through and acidic eluates were further enriched for phosphopeptides using TiO 2 beads while the basic eluates were used after neutralization with formic acid. LC−MS/MS. The peptide samples were separated using reversed phase chromatography (ReproSil-Pur C18-AQ, in an 18 cm long column, 100 μm i.d.) during a 100 min gradient with an Easy-nLC system (Thermo Scientific). MS measurements were performed on an LTQ-Orbitrap XL instrument interfaced with an ETD module (Thermo Scientific). The LTQ-Orbitrap XL was operated using the unattended datadependent acquisition mode, automatically switching between MS (60 000 resolution at m/z 400) and MS/MS using a threshold of 15 000 for ion selection (the automated gain control (AGC) was 20 000 ions in each MS/MS). The three most intense peptide ions were fragmented consecutively by CAD/ETD and recorded in the linear ion trap. CAD was performed as a multistage activation (MSA) experiment25 (normalized collision energy 35; MSA on the loss of 98 Da from precursor ions in charge states 2+, 3+, and 4+) if a phosphate neutral loss was detected. The ETD event lasted for 100 ms for the 2+ ions (charge state dependent activation time as well as supplementary activation was enabled) using a 5-fold excess of radical anions. Data Processing. Data processing was performed using Proteome Discoverer. Phosphopeptides were identified using Mascot version 2.2 via Proteome Discoverer against the IPIHuman database (version 3.66). The following search parameters were used: enzyme, trypsin; instrument, ESITRAP (CAD/MSA) and ETD-TRAP (ETD); mass accuracy precursor, 7 ppm; mass accuracy fragments, 0.7 Da; fixed modifications, carbamidomethyl (C); variable modifications, oxidation (M), phosphorylation (STY); max. missed-cleavages, 2. The data were exported as XML files from Mascot and processed using the PhosphoSiteLocalizer tool. The MS/MS tolerance was set to 0.6 Da.

■

RESULTS AND DISCUSSION The Cscore is a scoring algorithm constructed to assign phosphorylation sites in phosphopeptides and provide an estimate of the confidence of this assignment using the information from both CAD and ETD MS/MS spectra of the same precursor ions. The underlying algorithm is the same as that used to calculate the Ascore.11 Briefly, for each phosphopeptide sequence identified by the search engine, the two most likely sites of phosphorylation were identified by calculating a peptide score. The peptide score reflects the probability that the peptide-spectrum-match is a random event by using a cumulative binomial distribution. The identities of site-determining ions are theoretical fragment ions separating the two highest scoring phosphopeptide analogues. The cumulative binomial distribution was then used to calculate the probability of matching at least the observed number of site determining ions from all theoretical site-determining ions. This was done for both the first and second most likely phosphorylation site(s). The final Ascore was then obtained by subtracting these probabilities from each other. Unlike the Ascore, the Cscore includes both b,y-type and c,z•-type ions (z• ions are the most frequent occurring C-terminal ion type in ETD)26 from CAD and ETD spectra, respectively. Figure 2 shows an example of the benefit of integration of the information from both CAD and ETD data to calculate the Cscore. The two most likely phosphopeptide candidates (determined by the calculated peptide scores) were RGpSPTTGFIEQK and RGSPpTTGFIEQK. The Ascore algorithm identified the number of site determining ions supporting either site from a single spectrum, whereas the 9696

dx.doi.org/10.1021/ac302364r | Anal. Chem. 2012, 84, 9694−9699

Analytical Chemistry

Technical Note

spectra. However, 711 phosphosite localizations were only identified by using the Cscore while there were 306 and 466 unique localizations identified by using Ascore (ETD) and Ascore (CAD), respectively. The Cscore increased the number of phosphorylation site localizations by 25% and 88% relative to the Ascore (CAD) and Ascore (ETD), respectively. In this regard, we investigated the extent of phospho-site disagreement between the Ascore (CAD), Ascore (ETD), and Cscore. This analysis revealed that the least disagreement was found between the Ascore (CAD) and the Cscore data, while the greatest disagreement was found between Ascore (ETD) and Ascore (CAD) data (Figure S2 in the Supporting Information). The number of disagreements naturally depends on the score. When assignments were above 19, the maximum number of disagreements was very low (less than 32), but with lower thresholds, the assignments were less confident and hence disagreements more abundant. By considering only phospho-sites significantly determined by at least one of the three scores, we found that the Ascore was the highest in 865 and 1759 cases for the ETD and CAD spectra, respectively, whereas the Cscore resulted in the highest score in 3730 cases. The Ascore was higher than that of the Cscore in cases where one of the two spectra used in Cscore provided no or very limited additional information. The A- and Cscores are based on the probability of finding the actual number n out of x potential site determining ions. Using two spectra, x will be twice as high, so if n is not increased, the score will be lower. As an example, finding 3 out of 6 possible ions with the Ascore yields a better score than finding 3 out of 12 with the Cscore. Figure 3 also shows that in a few (9) cases the Ascore from both CAD and ETD was higher than the corresponding Cscore. This is unexpected as two good spectra should lead to a high Cscore. In those cases, an inconsistency was found in the localization of the phosphorylation sites provided by the Ascore depending on if it was calculated using the ETD or CAD MS/MS spectra. One possible explanation for this could be gas-phase rearrangements of the phosphate groups during CAD fragmentation.27,28 Therefore, assigning those localizations as unreliable (as the resulting Cscore) is an additional strength of the Cscore. The lower confidence levels for phospho-site localization obtained by the Ascore (ETD) relative to the Ascore (CAD) could result from the lower quality mass spectrum of the ETD spectra. This relates to the typical 2+ charge of many tryptic peptides, which is optimal for CAD fragmentation but not for ETD.29 Phosphopeptide identification by Mascot was dependent on the use of either ETD or CAD fragmentation, and the two techniques had very little overlap. To demonstrate, CAD and ETD spectra resulted in the same Mascot assignment for only 3 246 of 10 315 phosphorylation sites in the data set, while 4 245 phosphosites were identified using only CAD and 2824 were identified using only ETD. Some ETD/CAD spectra were of poor quality, but the information content often proved valuable for improving the Cscore. This was true for 2823 phospho-sites where the Cscore was higher than any of the Ascores, although only one of the spectra led to confident phosphopeptide identification. To determine if the Cscore was as significant as the Ascore, we used a publically available data set containing a library of synthetic phospho-peptides fragmented using both CAD and ETD.27 The correct phosphorylation sites are known for this data set and can be compared with the results predicted by the Ascore and Cscore algorithms. In addition, it also allowed us to

Cscore used information from both spectra, which increases the confidence in the identification. This increased the number of matched theoretical site determining ions to 4 out of 8 (Cscore) from 2 out of 4 (Ascore). Although both the Ascore and the Cscore agree on the phospho-site localization, confidence value was ∼47 for the Cscore versus ∼25 for the Ascore. To test the performance of using the Cscore to predict phospho-site localization, we applied it to a large data set. A tryptic digest from a human cell lysate was enriched for phosphopeptides, fractionated, and analyzed using 32 LC−MS/ MS on an LTQ Orbitrap. Selected precursor ions were fragmented using both CAD (as a multistage activation experiment) and ETD with low resolution in the linear ion trap. The MS/MS spectra were searched using Mascot. Assigned phosphopeptides with a score above the homology threshold at a 0.05 significance threshold were included in the analysis. This resulted in an FDR of 3.0% estimated by Mascot’s incorporated decoy search. A CAD/ETD spectrum pair resulted in a phospho-site identification if the Mascot score was above the homology score for the CAD spectrum, the ETD spectrum, or both. The Mascot outcome of this analysis resulted in identification of 9233 phosphopeptides of which 2590 were nonredundant phosphopeptides comprising 3441 phosphorylation sites. In the subsequent analyses, we considered all spectra that gave rise to phosphopeptide identifications excluding those that were unambiguous. This gave 1504 spectra where the number of possible sites was equal to the assigned number of phosphorylations. This left a total of 7729 spectrum pairs comprising 10 315 phosphorylations for further analysis of the scores (Ascore (CAD), Ascore (ETD), and Cscore (CAD/ ETD)). A score threshold of 19 was used to reflect a high level of confidence in potential phosphorylation site localization since a previous study showed that this threshold should provide >99% confidence in site localization.11 Overall, the average score increased from 23.37 and 17.02 for the Ascore (CAD) and Ascore (ETD), respectively, to 31.47 for the Cscore, and it identified more potential phosphorylation sites (5591) than either the Ascore (CAD) (4485) or Ascore (ETD) (2980). These results were consistent with an overall broader score distribution shifted toward higher confidence scores for the Cscore compared to that of the Ascore (CAD) and Ascore (ETD) (Figure S1 in the Supporting Information). Figure 3 depicts the overlap and differences between the numbers of phosphorylation sites of Ascores or Cscores ≥ 19. Confidence in assignments calculated with the Cscore were similar to those found using the Ascore with either CAD or ETD MS/MS

Figure 3. Venn diagram showing the number of phospho-site localizations (confidence score ≥ 19) using Ascore (CAD), Ascore (ETD), and Cscore (CAD/ETD) mass spectra. 9697

dx.doi.org/10.1021/ac302364r | Anal. Chem. 2012, 84, 9694−9699

Analytical Chemistry

Technical Note

score threshold for the Ascore (CAD) was found to be 17.5 (which is close to the 19 suggested by the Gygi group providing 1% FLR). Lowering the Cscore from 19 to 7, we compared the number of confidently assigned phospho-sites of the A549 samples and found an increase in phospho-site localizations from 5591 to 8754 (+ 57%), which corresponds to a total increase of 95% when compared to the results obtained with the Ascore (CAD) using a score threshold of 19. For further comparison, the performance of the Cscore to that of the Ascore using CAD or ETD at score cut-offs 3, 5, 7, 10, 15, and 20 is displayed in Figure S3 in the Supporting Information. At all cut-offs, the Cscore was superior. Cscore is therefore useful in large-scale phosphoproteomics studies when both ETD and CAD fragmentation are applied. Similarly, decreasing the Ascore (ETD) is also warranted (Table 1), which will also result in more phospho-site localizations (data not shown).

calculate the false localization rate (FLR), which is the number of wrong localizations divided by the total number of localizations determined using a score threshold of 19 for both the Ascore and Cscore. We identified the score threshold needed to obtain an FLR of 0.5% and counted the number of phosphorylation sites assigned by the Ascore and Cscore. A summary of the results are shown in Table 1 that demonstrate Table 1. Comparison of the Ascore (ETD), Ascore (CAD) and Cscore (CAD/ETD) Determined for the Synthetic Phosphopeptide Library [LREA][FKDG][TS]GH[PRDAF][EKG][pST]LER and [LREA][FKDG][pTS]GH[PRDAF][EKG][ST]LER31

FLR at score ≥ 19 no. of phosphorylation site localizations at score ≥ 19 score threshold resulting in FLR = 0.5% no. of phosphorylation site localizations at FLR = 0.5%

Ascore (ETD)

Ascore (CAD)

Cscore (CAD/ ETD)

top hits

0.13% 6232

0.42% 7828

0.21% 8681

0.36% 8916

9.8

17.5

7.0

13.7

7173

7978

9035

9019

■

CONCLUSIONS Complementary sequence information has previously been demonstrated to facilitate higher confidence in protein identification.32 Similarly, the Cscore algorithm takes advantage of the complementarity between CAD and ETD and can increase confidence in the number of assigned phosphorylation sites in MS/MS based phosphoproteomics studies. Because the Cscore is dependent on both an ETD and CAD MS/MS spectrum of the same phosphopeptide, it will lead to an increase in the duty-cycle of the MS/MS analysis. However, this cost is compensated for by the resulting higher confidence in phospho-site localization. With a few modifications, the Cscore could be suitable for the analysis of phosphopeptides fragmented with electron detachment dissociation (EDD)33,34 or NETD35 in the negative ion-mode in conjunction with another complementary fragmentation technique. Indeed, the Cscore could also be applied to the study of other types of posttranslational protein modification such as acetylations, methylations, formylations, sulfations, or glycosylations.

the FLR of the Cscore was lower than that of the CAD-based Ascore at a threshold of 19. As a consequence of that was an observed increment of significantly identified phosphorylation sites obtained using the Cscore (+ 39%) compared to that of the Ascore (CAD). The Ascore, based on ETD data, was found to be more significant than the corresponding Ascore based on CAD data. The general observation was that the confidence achieved by each scoring approach was high. At the score threshold of 19, the FLR was much lower than the 1% estimated in the original Ascore paper.30 This might, however, be biased since the distance between the potential phosphorylation sites was five residues in all the synthetic peptides, which in most cases simplifies unique phospho-site localization assignment. Even though the Cscore outperformed both the Ascore (CAD) and Ascore (ETD) on most parameters, there were cases as in Figure 3 where the confidence in phospho-site localization was only above the threshold with one of the three scores. To maximize confidence in the phospho-site localizations, we formed a metric termed “Top hits” that included only the highest scoring phospho-sites obtained by either the Ascore (CAD), Ascore (ETD), or Cscore (CAD/ETD). Compared to the Cscore, the “Top hits” resulted in more (2.7%) phospho-site localizations with scores above 19. However, the FLR of the “Top hits” was higher than that of the Cscore because of the contribution of the less confident results of the Ascore (CAD). Including all scores in “Top hits” did not provide additional phospho-site localizations when a fixed FLR of 0.5% was applied. As a consequence, an alternative strategy would be to exclude the Ascore (CAD) results from the “Top hits”, which then would lower the FLR and most likely increase the number of confident phospho-site localizations compared to the Cscore alone. Although the phosphopeptide library does not represent the great diversity of naturally occurring phosphopeptides, future phospho-site localization studies could be performed using a Cscore threshold less than 19. Indeed, it is reasonable to assume that the obtained Cscore of 7 is within a narrow range of what could be applied to biological samples given that the

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: +45-6550-2351. Fax: +45-6593-2661. Present Address †

Institute of Biochemistry and Molecular Biology, Rheinische Friedrich-Wilhelms University, Bonn, D-53115, Germany. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by grants from the Danish Council for Independent Research, Natural Sciences to F.K. (Grant FNU FK272-08-0044) and O.N.J. (Grant FNU 09-062299) and a generous grant from the Danish National Research Foundation to the Center for Epigenetics (O.N.J.).

■

REFERENCES

(1) Cohen, P. Trends Biochem. Sci. 2000, 25, 596−601. (2) Graves, J. D.; Krebs, E. G. Pharmacol Ther. 1999, 82, 111−121.

9698

dx.doi.org/10.1021/ac302364r | Anal. Chem. 2012, 84, 9694−9699

Analytical Chemistry

Technical Note

(3) Pandey, A.; Mann, M. Nature 2000, 405, 837−846. (4) Zhang, Y.; Wolf-Yadlin, A.; Ross, P. L.; Pappin, D. J.; Rush, J.; Lauffenburger, D. A.; White, F. M. Mol. Cell. Proteomics 2005, 4, 1240−1250. (5) Olsen, J. V.; Vermeulen, M.; Santamaria, A.; Kumar, C.; Miller, M. L.; Jensen, L. J.; Gnad, F.; Cox, J.; Jensen, T. S.; Nigg, E. A.; Brunak, S.; Mann, M. Sci. Signal. 2010, 3, ra3. (6) Thingholm, T. E.; Jensen, O. N.; Robinson, P. J.; Larsen, M. R. Mol. Cell. Proteomics 2008, 7, 661−671. (7) Kumar, N.; Wolf-Yadlin, A.; White, F. M.; Lauffenburger, D. A. PLoS Comput. Biol. 2007, 3, e4. (8) Moore, R. E.; Young, M. K.; Lee, T. D. J. Am. Soc. Mass Spectrom. 2002, 13, 378−386. (9) Martin, D. M.; Nett, I. R.; Vandermoere, F.; Barber, J. D.; Morrice, N. A.; Ferguson, M. A. Bioinformatics (Oxford, England) 2010, 26, 2153−2159. (10) Schlosser, A.; Vanselow, J. T.; Kramer, A. Anal. Chem. 2007, 79, 7439−7449. (11) Beausoleil, S. A.; Villen, J.; Gerber, S. A.; Rush, J.; Gygi, S. P. Nat. Biotechnol. 2006, 24, 1285−1292. (12) Savitski, M. M.; Lemeer, S.; Boesche, M.; Lang, M.; Mathieson, T.; Bantscheff, M.; Kuster, B. Mol. Cell. Proteomics 2011, 10, M110.003830. (13) Olsen, J. V.; Blagoev, B.; Gnad, F.; Macek, B.; Kumar, C.; Mortensen, P.; Mann, M. Cell 2006, 127, 635−648. (14) Hoffert, J. D.; Knepper, M. A. Anal. Biochem. 2008, 375, 1−10. (15) Heck, A. J. R.; Boersema, P. J.; Mohammed, S. J. Mass Spectrom. 2009, 44, 861−878. (16) Reid, G. E.; Palumbo, A. M.; Smith, S. A.; Kalcic, C. L.; Dantus, M.; Stemmer, P. M. Mass Spectrom. Rev. 2011, 30, 600−625. (17) Zubarev, R. A.; Kelleher, N. L.; McLafferty, F. W. J. Am. Chem. Soc. 1998, 120, 3265−3266. (18) Syka, J. E.; Coon, J. J.; Schroeder, M. J.; Shabanowitz, J.; Hunt, D. F. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 9528−9533. (19) Swaney, D. L.; Wenger, C. D.; Thomson, J. A.; Coon, J. J. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 995−1000. (20) Swaney, D. L.; McAlister, G. C.; Coon, J. J. Nat. Methods 2008, 5, 959−964. (21) Good, D. M.; Wirtala, M.; McAlister, G. C.; Coon, J. J. Mol. Cell. Proteomics 2007, 6, 1942−1951. (22) Syka, J. E. P.; Coon, J. J.; Schroeder, M. J.; Shabanowitz, J.; Hunt, D. F. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 9528−9533. (23) Savitski, M. M.; Kjeldsen, F.; Nielsen, M. L.; Zubarev, R. A. Angew. Chem., Int. Ed. 2006, 45, 5301−5303. (24) Larsen, M. R.; Thingholm, T. E.; Jensen, O. N.; Robinson, P. J. Mol. Cell. Proteomics 2008, 7, 661−671. (25) Schroeder, M. J.; Shabanowitz, J.; Schwartz, J. C.; Hunt, D. F.; Coon, J. J. Anal. Chem. 2004, 76, 3590−3598. (26) Sun, R. X.; Dong, M. Q.; Song, C. Q.; Chi, H.; Yang, B.; Xiu, L. Y.; Tao, L.; Jing, Z. Y.; Liu, C.; Wang, L. H.; Fu, Y.; He, S. M. J. Proteome Res. 2010, 9, 6354−6367. (27) Aguiar, M.; Haas, W.; Beausoleil, S. A.; Rush, J.; Gygi, S. P. J. Proteome Res. 2010, 9, 3103−3107. (28) Palumbo, A. M.; Reid, G. E. Anal. Chem. 2008, 80, 9735−9747. (29) Good, D. M.; Wirtala, M.; McAlister, G. C.; Coon, J. J. Mol. Cell. Proteomics 2007, 6, 1942−1951. (30) Beausoleil, S. A.; Villen, J.; Gerber, S. A.; Rush, J.; Gygi, S. P. Nat. Biotechnol. 2006, 24, 1285−1292. (31) Aguiar, M.; Haas, W.; Beausoleil, S. A.; Rush, J.; Gygi, S. P. J. Proteome Res. 2010, 9, 3103−3107. (32) Nielsen, M. L.; Savitski, M. M.; Zubarev, R. A. Mol. Cell. Proteomics 2005, 4, 835−845. (33) Budnik, B. A.; Haselmann, K. F.; Zubarev, R. A. Chem. Phys. Lett. 2001, 342, 299−302. (34) Kjeldsen, F.; Horning, O. B.; Jensen, S. S.; Giessing, A. M. B.; Jensen, O. N. J. Am. Soc. Mass Spectrom. 2008, 19, 1156−1162. (35) McAlister, G. C.; Russell, J. D.; Rumachik, N. G.; Hebert, A. S.; Syka, J. E. P.; Geer, L. Y.; Westphall, M. S.; Pagliarini, D. J.; Coon, J. J. Anal. Chem. 2012, 84, 2875−2882. 9699

dx.doi.org/10.1021/ac302364r | Anal. Chem. 2012, 84, 9694−9699