Ultra-High-Pressure RPLC Hyphenated to an LTQ-Orbitrap Velos

Mar 9, 2011 - Clifford Young , Alexandre V. Podtelejnikov , and Michael L. Nielsen ... Yuxin Li , Bing Bai , Xusheng Wang , Haiyan Tan , Tao Liu , Tho...
0 downloads 0 Views 1MB Size
Download PDF
ARTICLE pubs.acs.org/ac

Ultra-High-Pressure RPLC Hyphenated to an LTQ-Orbitrap Velos Reveals a Linear Relation between Peak Capacity and Number of Identified Peptides Thomas K€ocher,* Remco Swart, and Karl Mechtler Research Institute of Molecular Pathology (IMP), Vienna, Austria Dionex Corporation, Amsterdam, Netherlands Institute of Molecular Biotechnology (IMBA), Vienna, Austria ABSTRACT: Currently, unbiased protein identification is mostly performed by directly coupling reversed-phase liquid chromatography (RPLC) via electrospray ionization to a mass spectrometer. In contrast to the innovations in mass spectrometric instrumentation, cutting-edge technology in RPLC has generally not been well adopted. Here, we describe the effects of increased peak capacities on the number of identified proteins and peptides in complex mixtures utilizing collisioninduced dissociation on an LTQ-Orbitrap Velos, providing a rationale for using advanced RPLC technology in LC-MS/MS. Using two different column lengths and gradient times between 1 and 10 h, we found a linear relation between the obtained peak capacities and the number of identified peptides. We identified on average 2516 proteins in the tryptic digest of 1 μg of HeLa lysate using an 8 h gradient on a 50 cm column packed with 2 μm C18 reversed-phase chromatographic material.

M

ass spectrometry has become an indispensable tool in biological research.1 The identification of more than 1000 proteins in a single experiment has become possible due to significant advances in all relevant areas such as mass spectrometric instrumentation,2 data interpretation,3 and high-performance liquid chromatography (HPLC).4,5 In parallel, further characterization of identified proteins regarding their quantities6 and post-translational modifications7 has become feasible. In the most commonly used route to protein characterization, referred to as shotgun proteomics,8 proteins are first in-solution digested by a specific protease such as trypsin. The proteolytic peptides are then separated via liquid chromatography online coupled to a mass spectrometer and analyzed by tandem mass spectrometry (LC-MS/MS).9 Although there are other options such as HILIC (hydrophilic interaction chromatography),10 chromatography of peptides is performed almost exclusively on a reversed-phase column by gradient elution using a mobile phase optimized for electrospray ionization.4 Typically, linear gradients are used with an increasing percentage of an organic solvent such as acetonitrile. Gradient times are commonly restricted to 30 min up to 2 h, mainly because longer gradients would require longer acquisition times. Very complex mixtures such as those generated by tryptic digests from whole cell lysates require maximum separation efficiency, currently only achievable with two-dimensional LCMS/MS.11 However, sample demands for two-dimensional LCMS/MS are higher and the technology is significantly more difficult in its operation and is time consuming; therefore, it r 2011 American Chemical Society

should be only considered when one-dimensional LC-MSMS fails to solve the analytical problem. As a consequence, most samples such as protein mixtures from protein complex purifications via tandem affinity tags or immunoprecipitations are studied with one-dimensional LC-MS/MS.12 It is a crucial feature of mass spectrometry-based protein characterization that it does not require detection of all proteolytic peptides of a protein; instead, even a single MS/MS spectrum of a peptide might be enough to unambiguously identify a protein.3,13 In addition, the concept of MS/MS allows for sequencing coeluting peptides by isolating a specific m/z (mass to charge) window containing the isotopic cluster of only one peptide.9 Consequently, separating all peptides from each other is not absolutely required in LC-MS/MS. Successful application of this strategy is only limited by the duty cycle, the sensitivity of the instrument, and the ion suppression effects inherent to electrospray ionization. Resolving power in chromatography can be measured in peak capacity,14,15 defined as the number of ideal Gaussian peaks that can be separated from one another over the total time frame in which the peptides elute. Peak capacity is the most commonly used metric to measure the performance of gradient RPLC separations, with values between several hundreds up to about Received: December 13, 2010 Accepted: February 21, 2011 Published: March 09, 2011 2699

dx.doi.org/10.1021/ac103243t | Anal. Chem. 2011, 83, 2699–2704

Analytical Chemistry 1500.16 In HPLC, peak capacity can be improved either by increasing the column length or by reducing the plate height of the column, which can be achieved by reducing the particle diameter. Both approaches to increase peak capacity are limited by the resulting pressure drop over the column.5 Another approach for improving separation capabilities is to adjust the temperature17 or expanding the total gradient time.18 In general, longer gradient times produce higher peak capacities, although limits are reached at longer gradient times due to peak broadening.18 The applied flow rate also influences the obtained peak capacity;19 however, routine set ups in nano-HPLC use very similar flow rates of about 250 nL/min because the applied flow rate strongly affects the ionization efficiency.20,21 From a chromatographic point of view, aiming for high peak capacities is paramount, especially when analyzing complex mixtures. It also seems intuitive that an increase in peak capacity would result in a higher number of peptides, which are detected, sequenced, and identified in LC-MS/MS. However, it is not clear how much benefit is actually obtained in the coverage of peptide mixtures from an increase in peak capacity because the analytical strategy is based on the above-mentioned ability of MS/MS to isolate coeluting peptides from each other. Consequently, it would be conceivable that the total acquisition time contributes decisively to the number of identified peptides and increasing the peak capacity at constant gradient times would not have significant positive consequences. Factors influencing peak capacity in LC such as gradient time or reduced peak width were reported to be beneficial for LC-MS/MS, such as by increasing the sequence coverage of a single protein.22 However, the relation between peak capacity and the number of identified peptides in complex samples was not studied extensively. Additionally, there are not many published results reporting the performance of ultralong gradient times in LC-MS/MS and even less publications systematically investigating their effects for protein identification.16,23,24 Here we report that by using long gradients we identified very large numbers of proteins and peptides from a HeLa (human epithelial carcinoma cell line) cell lysate using collision-induced dissociation (CID) with an LTQ-Orbitrap Velos. In addition, we demonstrate by modulating the peak capacity via the column length and gradient time that the number of identified peptides showed a linear response to the achieved peak capacity.

’ EXPERIMENTAL SECTION HPLC and Mass Spectrometry. Nano-HPLC-MS/MS analysis was performed on an UltiMate 3000 RSLCnano LC system (Dionex). Peptide separation was carried out on a C18 column with a length of either 25 or 50 cm (Acclaim PepMap RSLC C18, 25, or 50 cm  75 μm  2 μm, 100 Å, Dionex) using the following solvent system: (A) 2% acetonitrile, 0.1% formic acid; (B) 80% acetonitrile, 0.08% formic acid, and 10% trifluoroethanol. The sample was first loaded on a trap column (Acclaim PepMap C18; 2 cm  100 μm  5 μm, 100 Å) with 0.1% TFA. After washing with 0.1% TFA, the trap column was switched online with the separation column. For separation on the analytical column a gradient from 100% A to 40% B (starting length 55 min increased in 1 h steps to 595 min, omitting the 7 and 9 h gradient) was used, followed by a second gradient from 40% B to 90% B (5 min). The HPLC was directly coupled to a nanoelectrospray ionization source (Proxeon, Odense, Denmark).

ARTICLE

The LTQ-Orbitrap Velos mass spectrometer (Thermo Fisher Scientific) was operated in positive ionization mode. The MS survey scan was performed in the FT cell recording a window between 350 and 2000 m/z. The resolution was set to 60 000, and the automatic gain control was set to 1 000 000 ions with a maximal acquisition time of 400 ms. Minimum MS signal for triggering MS/MS was set to 500, and m/z values triggering MS/MS were put on an exclusion list for 240 s. In all cases one microscan was recorded, and a maximum of 20 MS/MS experiments were triggered per MS scan. The lock mass option was enabled, and polydimethylcyclosiloxane (protonated (Si(CH3) 2O))6; m/z 445.120025) was used for internal recalibration of the mass spectra.25 CID was done with a target value of 3000 in the linear ion trap, maximal acquisition time 200 ms, collision energy of 35%, Q value of 0.25, and an activation time of 10 ms. Sample Preparation. HeLa proteins were isolated and proteolytically digested as described.26 In brief, Hela Kyoto cells were treated with nocodozole for 16 h. The cells were harvested with a scraper and washed two times with 1 PBS. The cell pellet was suspended in an equal amount of lysis buffer, and the cells were disrupted by pulling the cell suspension through a thin needle. Soluble proteins were reduced with dithiothreitol, alkylated with iodoacetamide and digested with trypsin essentially as described.27 Data Interpretation. Peak widths for five randomly chosen peptides at different retention times were measured in their extracted ion chromatograms at 4σ (13.5% of the maximum peak height for a Gaussian peak) in the triplicates of all experiments, and an average peak width for each LC run was calculated. Peak capacities were calculated using the equation nc (peak capacity) = tG (gradient time)/W4σ (peak width at 4σ). Fragment ion data were interpreted with Mascot (Matrix Science) using Mascot Daemon (version 2.2.2). Data were searched against the (human) IPI database (release 2, 13, 2010). Fragment ions considered were singly and doubly charged b and y ions. A precursor ion mass tolerance of 5 ppm was used. Fragment ion masses were searched with a 0.5 Da mass window. Three missed cleavage sites for trypsin were allowed. Carbamidomethyl at cysteine residues was set as a fixed modification, oxidation of methionine and phosphorylation at serine, threonine and tyrosine as variable modification. Mascot search results were imported into ProteinCenter (Proxeon, Odense, Denmark) for further analysis. False discovery rates for proteins were calculated by the equation FDR(%) = 100  (Hits Decoy)/(Hits Target). We calculated the false discovery rates by searching a combined database consisting of the (human) IPI database and its reversed form, searching the mgf files separately. The reversed version of the IPI sequence entries was generated with MaxQuant.28

’ RESULTS AND DISCUSSION There is a positive relation between gradient time and the number of proteins and peptides identified in a RPLC-MS/MS experiment.16 However, with increasing gradient times the negative influence of increased peak broadening results in reduced ion currents and ultimately in the failure to identify eluting peptides of less abundance. Main factors in this relation are also the sequencing speed and the sensitivity of the mass spectrometer, ultimately limiting the number of sequenced peptides. In this study we want to address a number of related 2700

dx.doi.org/10.1021/ac103243t |Anal. Chem. 2011, 83, 2699–2704

Analytical Chemistry

ARTICLE

Figure 1. Exemplary UV chromatograms of a tryptic digest of 1 μg of HeLa lysate are shown.

questions using a LTQ-Orbitrap Velos, a mass spectrometer with very fast MS/MS capabilities.29 What is the maximum gradient time still leading to identification of additional peptides and proteins with the used LC-MS/ MS setup? Second, is the optimal gradient time depending on the column length? Third, is there a relation between the peak capacity and the number of proteins and peptides identified, and if such a relation exists, do peptides of highly abundant proteins behave differently to peptides of proteins with low abundance? Fourth, is the protein identification rate maximized when utilizing longer gradient times or is it advantageous to repeat the analysis with shorter gradients? In order to study the influence of prolonged gradient time on the number of identified peptides and proteins, we analyzed 1 μg of a tryptic digest of a HeLa cell lysate. We used various gradients ranging from 1 to 10 h using a column with a length of 25 cm and later 50 cm, packed with 2 μm chromatographic material (Figure 1). The time of the linear gradient from 0 to 36% organic phase was increased from initial 55 min in steps of 1 h, each followed by a 5 min short gradient to 90% organic phase. The rationale behind choosing a HeLa lysate as a model analyte was that it has been used in many studies assessing the performance of LC-MS/MS instrumentation and provided us with a theoretical number of a few hundred thousands of peptides present at a wide dynamic range. We analyzed the tryptic peptide mixture on both columns and the various gradient times in three technical replicates. As expected we found that the number of peptides (Figure 2A) and proteins (Figure 2B) identified increased with extending gradient times using a cutoff for the Mascot peptide score of 20. Both the number of proteins and the number of peptides followed a logarithmic relation with the gradient time (R2 = 0.98) in the studied time range. After a gradient time of 8 h both values converged to a maximum. The average number of proteins identified in the triplicates of both the 8 h and the 10 h gradient was 1946 using the 25 cm column. The false positive rate was less than 1% at the protein level. We analyzed the whole data set using a cutoff of 15 for the peptide score in order to assess eventual deviations from above result. We identified approximately 15% more peptides with the lower threshold but observed an almost identical response of the peptide identification rate to the applied gradient times (data not shown). Next, we repeated the analysis of the tryptic digest using a 50 cm column applying the same gradients. The operating pressure was around 750 bar, well below the pressure limit of the

Figure 2. (A) Number of peptides identified using the two different columns are plotted against the gradient time. The data points for both the 50 cm (pink) and the 25 cm (blue) column follow a logarithmic relation (dashed line; R2 = 0.97). (B) The number of proteins identified based on these peptides is shown. (C) Relation between the ratio of the number of peptides identified and the gradient time is plotted against the gradient time and follows a power law (dashed line; R2 = 0.98).

instrument of 800 bar. Interestingly, we identified a very similar number of peptides and proteins using a gradient of 1 h. However, with increasing gradient times the number of peptides (Figure 2A) and proteins (Figure 2B) identified was significantly higher, suggesting that the potentially higher performance of the longer column depends on the applied gradient conditions. We observed again that the number of peptides converged to a maximum value after 8 h, however leading to identification of an average number of 2516 proteins, 659 proteins more than by using the 25 cm column. We also analyzed the time efficiency of the analysis by calculating the ratio between the number of identified proteins and the gradient time (Figure 2C). Following a power law this ratio decreased with increasing gradient times for both columns, 2701

dx.doi.org/10.1021/ac103243t |Anal. Chem. 2011, 83, 2699–2704

Analytical Chemistry

Figure 3. Peak widths (A) and resulting peak capacities (B) plotted against the applied gradient time for the 25 (blue) and 50 cm (pink) column. (C) Ratios between the peak capacities achieved for the 25 (blue) and 50 cm (pink) column and the respective number of identified peptides reveal a linear relation (R2 = 0.84).

resembling the results obtained for peptides. In the case of the 50 cm column the protein identification rate started with a very high value of 19.1 identified proteins per minute for the 1 h gradient, dropping to 4.1 proteins per minute for the 10 h gradient. Plotting the peak width against the gradient time (Figure 3A) we found an approximately linear relationship between gradient time and peak widths, possibly limiting a further increase of identified peptides. We concluded that very long gradient times can be exploited for identification of very large numbers of proteins in a single LC-MS/MS experiment. Relation of the Peak Capacity vs the Number of Peptide Identifications. We calculated the achieved peak capacities for all triplicates of the different gradient lengths and columns, obtaining peak capacities ranging between 199 and 707 (Figure 3B). The peak capacities for the two columns start with very similar values for the 1 h gradient but clearly separate after

ARTICLE

4 h of gradient time and differ by an average factor of 1.4, corresponding to the square root of the ratios of the two different column lengths (Figure 3B). Next, we investigated a possible correlation between the number of peptides and the peak capacities. We plotted the number of peptides identified in all experiments using a minimum peptide score of 20 against the obtained peak capacities (Figure 3C). Interestingly, we obtained an almost linear relationship between the peak capacity and the number of peptides for both columns. The data points corresponding to the 10 h gradient times showed a slight deviation from the observed trend. Calculating the constant k for a linear relation npep = knc between peak capacity nc and the number of identified peptides npep assuming an intercept at zero, we obtained a constant k = 23.7 with a correlation of R2 = 0.83 (Figure 3C). Analyzing the data for each column separately we obtained even better correlations. For the data from the 50 cm column alone we obtained a constant k = 22.4 (R2 = 0.92). Using other cutoff values such as a minimum peptide score of 15 the observed linearity was not affected but the obtained factor increased to k = 27.3 (R2 = 0.81). We also tested the hypothesis whether the number of protein identifications also depends linearly of the peak capacity. In this case again assuming an intercept at zero, the constant of a linear relation was k = 4.3 with a considerable weaker correlation (R2 = 0.69). In summary, the data suggests a linear relationship between peak capacity and number of peptide identifications in LC-MS/ MS, regardless of whether an increase in peak capacity is achieved by increasing gradient time or by a change in the length of the analytical column. It should be noted that this correlation was found for a mass spectrometer with a high sequencing speed and would likely be different for instrumentation with weaker MS/ MS capabilities. We observed also a positive relation between protein identifications and peak capacity, but this relation was not strictly linear in the studied range. However, the results demonstrate that in conjunction with fast sequencing capabilities increasing peak capacities are a prerequisite for obtaining optimal results with LC-MS/MS. Relationship between the Peak Capacity and the Observed Dynamic Range. After concluding that higher peak capacities result in higher numbers of identified peptides, we addressed the question of whether this improvement also results in a higher dynamic range of the analysis and the improved coverage of less abundant proteins. Using the data sets acquired for the 50 cm column, we divided the identified proteins in groups according to their relative abundance in the mixture. We used the concept of exponentially modified protein abundance index (emPAI)30 which allows for a semiquantitative relative quantification by expressing the molar fraction of proteins in mixtures. First, we analyzed the subgroup of proteins present at a mole fraction of at least 0.1% in the mixture. Although only a semiquantitative method, the number of proteins in this subgroup remained relatively constant, regardless of the total number of identified proteins, the gradient length, or the cutoff value used for the peptide score. Using the 50 cm column and a peptide score of 20 or better we found on average 298 proteins with emPAI values corresponding to this relative molar percentage with a standard deviation of 32 (Figure 4A). Interestingly, except for the shortest gradient the number of peptides corresponding to this subgroup stayed relatively constant over different gradient lengths. Plotting the number of peptides against the gradient time clearly demonstrates this trend (Figure 4B). This 2702

dx.doi.org/10.1021/ac103243t |Anal. Chem. 2011, 83, 2699–2704

Analytical Chemistry

Figure 4. Plotting different subsets of the number of identified proteins (A) and peptides (B) against the gradient time. Highly abundant proteins and peptides present at a molar fraction at least 0.1% in the mixture (pink) behaved differently than less abundant proteins and peptides. Data points are shown for proteins present at a molar fraction of at least 0.05% (light blue), at least 0.01% (yellow), or all proteins (blue) based on their emPAI values.

observation showed that for a very complex sample at least a 2 h gradient is required to cover the subset of the proteotypic31 peptides which belong to the most abundant proteins. For gradient times of 2 h and more, the additionally sequenced peptides preferentially led to identification of novel proteins present at lower abundance in the sample mixture. In order to investigate this effect further we also plotted the number of proteins with emPAI values corresponding to more than 0.05 relative molar percent of the mixture, observing again constant values after a gradient time of 2 h (Figure 4A). In contrast, plotting the proteins present with more than 0.01% and the total number of proteins displayed a constant increase with increasing gradient times. From this data one could hypothesize that after a gradient time of 2 h all highly abundant proteins were identified and the vast majority of the proteins additionally identified after 2 h were lesser abundant components of the mixture. In order to challenge the above hypothesis, we also analyzed the sequence coverage of two high molecular weight proteins with very different emPAI values over all experiments. The sequence coverage of a highly abundant protein, clathrin heavy chain, which accounted for about 0.15 mol % in the mixture did behave exactly as predicted according to the above model. The initial sequence coverage of 40.6% rose to 53% in the 2 h gradient but stayed relatively constant at this value in all other experiments with a standard deviation of 2.5%. Next, we tried to falsify the hypothesis with a minor component of the mixture, the protein RanBP2, which was not even detected in any of the 1 h gradient

ARTICLE

experiments. This protein had an emPAI value which corresponded to approximately 0.005 mol % in the mixture. In this case the average sequence coverage increased continuously with gradient time up to 5%, which corresponded to an average number of 11 unique peptides. We conclude that improving the peak capacity leads to increased sequence coverage of low abundant proteins but does not significantly increase the sequence coverage of highly abundant proteins in complex mixtures. Efficiency of Long Gradient Times for Protein Identification. A common strategy to increase proteome coverage achieved with LC-MS/MS is, despite its shortcomings,31 the repetitive analysis of an identical sample.32 Using the acquired data, we asked whether it is more time efficient to analyze a sample twice with a shorter gradient or once using a longer gradient. For the evaluation we ignored the time needed for washing and equilibrating and the higher sample demands for running two experiments. First, we compared the number of proteins uniquely identified in the combined data set of two 1 h replicates to the data of one 2 h gradient obtained using the 25 cm column. For the three possible combinations of the data acquired in the 1 h experiments we identified on average 1272 proteins, whereas for running one 2 h gradient we could identify on average 1400 proteins. Even combining all three 1 h runs we only could identify 1391 proteins, slightly less than one 2 h gradient and on average 175 proteins less than with one 3 h gradient. Comparing the performance of one 6 h gradient with either two 3 h gradients or three 2 h gradients we obtained on average 1857 proteins with a single 6 h gradient, 1752 for three times a 2 h gradient, and 1803 for combining two 3 h gradients. This data suggested that longer gradients are more efficient than repeating the experiment twice with two shorter gradients, as long as longer gradient times concur with higher peak capacities. Analyzing the data acquired with the 50 cm column, we obtained a similar but even more pronounced effect. For example, in the 6 h run we obtained on average 2366 proteins, but by combining all three 2 h runs we obtained only 2032 unique proteins. Combining two of the three 2 h gradients we identified on average 2169 proteins. We assume that this effect might be caused by the tendency of the mass spectrometer to sequence preferentially abundant peptides.31 We compared the absolute numbers of unique peptides in these three runs to the number of unique peptides in each of these runs. For the 6 h experiment we obtained on average 13 314 unique peptides, whereas combining the data from the three 2 h runs only 11 745 unique peptides were identified. Of these peptides, 46.1% were identified in all three experiments, confirming the above assumption. We calculated the percentage of shared peptide identifications between the three replicates for all gradient times. For the 50 cm column the percentage of shared unique peptides only changed marginally from 44.5% for the 1 h gradient to 39.1% for the 10 h gradient. For comparison, the absolute number of unique peptides detected three out of three times changed from 3221 to 8068. The same trend was observed with the unique peptides shared by two runs out of the three replicates. In this case, 51.5% of the totally identified unique peptides were shared between two runs with a standard deviation of 2.4%. This observation suggested again that increasing gradient times in conjunction with high peak capacity has the potential to identify unique peptides which are not easily identifiable with shorter gradient times or weaker chromatographic resolution. 2703

dx.doi.org/10.1021/ac103243t |Anal. Chem. 2011, 83, 2699–2704

Analytical Chemistry

’ CONCLUSIONS This study demonstrates a clear and linear correlation between the chromatographic peak capacity in reversed-phase liquid chromatography and the number of peptides identified with tandem mass spectrometry using a LTQ-Orbitrap Velos. Against common belief, we also showed that repetition of LC-MS/MS experiments with the same gradient time is less efficient in terms of identified peptides and proteins than running a longer gradient. For defining optimal gradient times, peak capacities should be measured and for complex samples gradient times should be optimized accordingly. Using an 8 h gradient and the 50 cm column system, we report the identification of on average 2516 proteins based on 14 292 peptides with a false positive rate of 0.4% at the protein level. To the best of our knowledge, this is the highest number of proteins reported so far for a single-run HPLC-MS/MS experiment. ’ AUTHOR INFORMATION Corresponding Author

*To whom correspondence should be addressed. Phone: 0043 1 79044 4283. Fax: 0043 1 79044 110. E-mail: Thomas.Koecher@ imp.ac.at.

’ ACKNOWLEDGMENT This work was funded by the Austrian Proteomics Platform (APP) within the Austrian Genome Research Program (GEN-AU) and the Institute for Molecular Pathology (IMP). We thank the other members of the Mechtler group, especially Ines Steinmacher and Michael Schutzbier for experimental help and Peter Pichler for critical reading of the manuscript.

ARTICLE

(21) Wahl, J. H.; Goodlett, D. R.; Udseth, H. R.; Smith, R. D. Electrophoresis 1993, 14, 448–457. (22) Fairchild, J. N.; Walworth, M. J.; Horvath, K.; Guiochon, G. J. Chromatogr. A 2010, 1217, 4779–4783. (23) Xu, P.; Duong, D. M.; Peng, J. M. J. Proteome Res. 2009, 8, 3944–3950. (24) Eeltink, S.; Dolman, S.; Detobel, F.; Swart, R.; Ursem, M.; Schoenmakers, P. J. J. Chromatogr. A 2010, 1217, 6610–6615. (25) Olsen, J. V.; de Godoy, L. M. F.; Li, G. Q.; Macek, B.; Mortensen, P.; Pesch, R.; Makarov, A.; Lange, O.; Horning, S.; Mann, M. Mol. Cell. Proteomics 2005, 4, 2010–2021. (26) Poser, I.; Sarov, M.; Hutchins, J. R. A.; Heriche, J. K.; Toyoda, Y.; Pozniakovsky, A.; Weigl, D.; Nitzsche, A.; Hegemann, B.; Bird, A. W.; Pelletier, L.; Kittler, R.; Hua, S.; Naumann, R.; Augsburg, M.; Sykora, M. M.; Hofemeister, H.; Zhang, Y. M.; Nasmyth, K.; White, K. P.; Dietzel, S.; Mechtler, K.; Durbin, R.; Stewart, A. F.; Peters, J. M.; Buchholz, F.; Hyman, A. A. Nat. Methods 2008, 5, 409–415. (27) Gregan, J.; Riedel, C. G.; Petronczki, M.; Cipak, L.; Rumpf, C.; Poser, I.; Buchholz, F.; Mechtler, K.; Nasmyth, K. Nat. Protoc. 2007, 2, 1145–1151. (28) Cox, J.; Mann, M. Nat. Biotechnol. 2008, 26, 1367–1372. (29) Olsen, J. V.; Schwartz, J. C.; Griep-Raming, J.; Nielsen, M. L.; Damoc, E.; Denisov, E.; Lange, O.; Remes, P.; Taylor, D.; Splendore, M.; Wouters, E. R.; Senko, M.; Makarov, A.; Mann, M.; Horning, S. Mol. Cell. Proteomics 2009, 8, 2759–2769. (30) Ishihama, Y.; Oda, Y.; Tabata, T.; Sato, T.; Nagasu, T.; Rappsilber, J.; Mann, M. Mol. Cell. Proteomics 2005, 4, 1265–1272. (31) Kuster, B.; Schirle, M.; Mallick, P.; Aebersold, R. Nat. Rev. Mol. Cell Biol. 2005, 6, 577–583. (32) Durr, E.; Yu, J. Y.; Krasinska, K. M.; Carver, L. A.; Yates, J. R.; Testa, J. E.; Oh, P.; Schnitzer, J. E. Nat. Biotechnol. 2004, 22, 985–992.

’ REFERENCES (1) Walther, T. C.; Mann, M. J. Cell Biol. 2010, 190, 491–500. (2) Han, X. M.; Aslanian, A.; Yates, J. R. Curr. Opin. Chem. Biol. 2008, 12, 483–490. (3) Nesvizhskii, A. I.; Vitek, O.; Aebersold, R. Nat. Methods 2007, 4, 787–797. (4) Sandra, K.; Moshir, M.; D’Hondt, F.; Verleysen, K.; Kas, K.; Sandra, P. J. Chromatogr. B 2008, 866, 48–63. (5) Jorgenson, J. W. In Annual Review of Analytical Chemistry; Annual Reviews: Palo Alto; Vol. 3, 2010; pp 129 150. (6) Wilm, M. Proteomics 2009, 9, 4590–4605. (7) Zhao, Y. M.; Jensen, O. N. Proteomics 2009, 9, 4632–4641. (8) Wolters, D. A.; Washburn, M. P.; Yates, J. R. Anal. Chem. 2001, 73, 5683–5690. (9) Steen, H.; Mann, M. Nat. Rev. Mol. Cell Biol. 2004, 5, 699–711. (10) Alpert, A. J. J. Chromatogr. 1990, 499, 177–196. (11) Fournier, M. L.; Gilmore, J. M.; Martin-Brown, S. A.; Washburn, M. P. Chem. Rev. 2007, 107, 3654–3686. (12) Kocher, T.; Superti-Furga, G. Nat. Methods 2007, 4, 807–815. (13) Gupta, N.; Pevzner, P. A. J. Proteome Res. 2009, 8, 4173–4181. (14) Davis, J. M.; Giddings, J. C. Anal. Chem. 1983, 55, 418–424. (15) Neue, U. D. J. Chromatogr. A 2008, 1184, 107–130. (16) Shen, Y. F.; Zhang, R.; Moore, R. J.; Kim, J.; Metz, T. O.; Hixson, K. K.; Zhao, R.; Livesay, E. A.; Udseth, H. R.; Smith, R. D. Anal. Chem. 2005, 77, 3090–3100. (17) Lee, S. T.; Olesik, S. V. Anal. Chem. 1994, 66, 4498–4506. (18) Gilar, M.; Daly, A. E.; Kele, M.; Neue, U. D.; Gebler, J. C. J. Chromatogr. A 2004, 1061, 183–192. (19) Wang, X. L.; Stoll, D. R.; Schellinger, A. P.; Carr, P. W. Anal. Chem. 2006, 78, 3406–3416. (20) Wilm, M.; Mann, M. Anal. Chem. 1996, 68, 1–8. 2704

dx.doi.org/10.1021/ac103243t |Anal. Chem. 2011, 83, 2699–2704