Anal. Chem. 2005, 77, 8179-8184

High-Throughput Identification of In-Gel Digested Proteins by Rapid, Isocratic HPLC/MS/MS Yue Chen, Sung Chan Kim, and Yingming Zhao*

Department of Biochemistry, University of Texas Southwestern Medical Center at Dallas, Dallas, Texas 75390-9038

The high sensitivity and accuracy of mass spectrometry for identifying proteins has led to an explosive expansion of proteomics research, necessitating rapid procedures for HPLC/MS/MS analysis. Current HPLC/MS/MS analysis usually relies on elution of peptides from the HPLC column with a gradient that takes a total of 45-70 min for each cycle, limiting sample throughput and the speed of protein identification. Here we report a simple method for high-throughput protein identification, using isocratic, either methanol- or acetonitrile-based buffer systems, HPLC elution into an LTQ mass spectrometer. This procedure allows each cycle of highly sensitive HPLC/ MS/MS analysis to be completed in 5 min, thus boosting the efficiency of HPLC/MS/MS analysis 9-14-fold. Using this method, each operator can acquire HPLC/MS/MS data for 96 in-gel proteolytic digests in one 8-h working day. The method can easily be implemented in any laboratory with an LTQ mass spectrometer. This protocol should find wide application in mass spectrometry laboratories that require high-throughput analysis but are limited by inefficient use of machine time. HPLC/tandem mass spectrometry (MS/MS) has become the method of choice for identifying proteins, largely due to the unparalleled sensitivity with which fragment mass fingerprints can be generated.1 In a typical protein identification experiment, the protein of interest is digested with a proteolytic enzyme, usually trypsin, and the resulting peptides are subjected to HPLC/MS/ MS analysis. Each resulting MS/MS spectrum contains the masses of the parent peptide and its fragment ions. This mass information is used in an automated search of a protein sequence database using one of several commercial algorithms (e.g., SEQUEST,2 PepSea,3 Mascot,4 Sonar,5 ProbID,6 Popitam,7 or Tandem8) to find the peptide that most closely matches the observed spectrum. * To whom correspondence should be addressed. E-mail: yzhao@ biochem.swmed.edu. Fax: (214) 648-2797. Tel: (214) 648-7947. (1) Aebersold, R.; Mann, M. Nature 2003, 422, 198-207. (2) Eng, J. K.; McCormack, A. L.; Yates, J. R., 3rd. J. Am. Soc. Mass Spectrom. 1994, 5, 976-989. (3) Mann, M.; Wilm, M. Anal. Chem. 1994, 66, 4390-4399. (4) Perkins, D. N.; Pappin, D. J.; Creasy, D. M.; Cottrell, J. S. Electrophoresis 1999, 20, 3551-3567. (5) Field, H. I.; Fenyo, D.; Beavis, R. C. Proteomics 2002, 2, 36-47. (6) Zhang, N.; Aebersold, R.; Schwikowski, B. Proteomics 2002, 2, 1406-1412. (7) Hernandez, P.; Gras, R.; Frey, J.; Appel, R. D. Proteomics 2003, 3, 870878. (8) Craig, R.; Beavis, R. C. Rapid Commun. Mass Spectrom. 2003, 17, 23102316. 10.1021/ac051468t CCC: $30.25 Published on Web 11/05/2005

© 2005 American Chemical Society

Because of the high sensitivity of HPLC/MS/MS, it is commonly combined with biochemical purification methods (e.g., TAP purification9) to dissect protein-protein or protein-ligand interactions. Proteomics analyses have seen an explosion in popularity, with protein quantification via 2D gel electrophoresis generating large numbers of protein spots. Efficient processing of such large numbers of protein samples in either core facilities or research laboratories requires rapid experimental protocols for each step of the analysis, including in-gel digestion, HPLC/MS/MS analysis, and automatic protein sequence database searching. HPLC/MS/MS is typically carried out using an HPLC gradient from a solution with low organic content (e.g., 2-5% acetonitrile) to high organic content (e.g., 60-90% acetonitrile). The gradient takes 10-45 min for a simple protein band or spot, with each spot usually containing a limited number of proteins. Using such a protocol, it usually requires 45-70 min to complete each cycle of analysis, including equilibration of the HPLC column. This lengthy procedure limits instrument efficiency and represents a bottleneck in the protein identification process. Here we report a simple procedure that uses nano-HPLC/LTQ mass spectrometry for rapid protein identification. Our method takes advantage of the speed with which the LTQ instrument acquires spectra: ∼200 ms per MS spectrum and ∼250 ms per MS/MS spectrum, with a maximum injection time of 100 ms. Such fast data acquisition allows use of isocratic HPLC elution to separate peptides from tryptic digests. Eluting isocratically with a solution of high organic content, we are able to generate peptide peaks with widths between 4 and 6 s, which are sufficient for both MS and MS/MS analysis. This procedure allows us to complete each HPLC/MS/MS in 5 min, 9-14-fold faster than procedures reported in the literature. We anticipate this protocol will find wide application in core facilities and research laboratories that demand rapid protein identification. MATERIALS AND METHODS Materials. Bombesin, myoglobin, and HPLC grade acetic acid were purchased from Sigma-Aldrich (St. Louis, MO); purified synthetic peptide M with sequence EYRVAIK was made by the Protein Chemistry Core Center of the University of Texas Southwestern Medical Center; SilverQuest Silver Staining Kit was from Invitrogen (Carlsbad, CA); HPLC grade water, HPLC grade methanol, and HPLC grade acetonitrile were from EMScience (Gibbstown, NJ); and modified porcine trypsin was from Promega (Madison, WI). (9) Rigaut, G.; Shevchenko, A.; Rutz, B.; Wilm, M.; Mann, M.; Seraphin, B. Nat. Biotechnol. 1999, 17, 1030-1032.

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005 8179

HPLC/MS/MS Analysis. HPLC/mass spectrometric analysis was performed in an LTQ 2D ion trap mass spectrometer (Thermo Electron, Waltham, MA) equipped with a nanoelectrospray ionization source. The source was coupled online to an Agilent 1100 series nanoflow LC system (Agilent, Palo Alto, CA). Peptide sample solutions (bombesin and peptide M both at 2 fmol/µL) were prepared in HPLC buffer A (see below for composition of HPLC buffers). Two microliters of peptide in buffer A was manually injected onto a capillary HPLC column (35 mm length × 75 µm i.d., 5-µm particle size, 100-Å pore diameter) packed inhouse with Luna C18 resin (Phenomenex, Torrance, CA). Peptides were eluted either using a gradient or isocratically. Under gradient elution conditions, peptides were eluted from the column with a gradient of 25-90% buffer B over 8 min. Under isocratic elution conditions, peptides were eluted with 90% buffer B. The eluted peptides were electrosprayed directly into the LTQ ion trap mass spectrometer. The MS/MS spectra were acquired in a datadependent mode, such that the masses and fragmentation patterns of the four strongest ions in each MS scan were determined. Two buffer systems were used for HPLC. In the methanol system, buffer A was 2% methanol/97.5% water/0.5% acetic acid (v/v/v) and buffer B was 99.5% methanol/0.5% acetic acid (v/v). In the acetonitrile system, buffer A was 2% acetonitrile/97.9% water/0.1% acetic acid (v/v/v) and buffer B was 90% acetonitrile/ 9.9% water/0.1% acetic acid (v/v/v). Protein Silver Staining and In-Gel Digestion. Two micrograms of purified mouse nuclear proteins was separated on a 12% SDS-PAGE gel. The gel was stained using a SilverQuest staining kit (Invitrogen). The protein band of interest was excised and washed with acidic buffer (acetic acid/methanol/water, 10:50:40, v/v/v) twice for 1 h each time and then swollen in water twice for 20 min each time. The gel band was destained using freshly made destaining buffer (30 mM potassium ferricyanide and 100 mM sodium thiosulfate, 1:1, v/v10) for 5 min. The destained gel band was washed with water twice for 10 min each time and then cut into small pieces. The gel pieces were washed for 10 min with 100 mM ammonium bicarbonate in 50% acetonitrile solution (acetonitrile/water, 50:50, v/v) and then completely dehydrated with 100% acetonitrile and dried in a SpeedVac (Thermo Electron, Waltham, MA) for 20 min. Twenty microliters of trypsin solution containing 10 ng/µL modified porcine trypsin (Promega, Madison, WI) in 50 mM ammonium bicarbonate was added to the gel pieces followed by overnight incubation at 37 °C. Tryptic peptides were extracted with acetonitrile/water/trifluoroacetic acid (50:45:5, v/v/v) and then acetonitrile/water/trifluoroacetic acid (75:24.9:0.1, v/v/v). Extracts were dried in a SpeedVac, desalted with a µC18 ZipTip (Millipore, Billerica, MA) according to the procedure described by the vendor and analyzed by LC/MS/MS as described above. Protein Sequence Database Search and Manual Verification. Tandem mass spectra were used to search the NCBI-nr database with the Mascot search engine (version 2.0, Matrix Science, London, U.K.). Mass tolerance was set to (4 Da for parent ion masses and (0.6 Da for fragment ion masses. Peptides with Mascot score above 40 were considered potentially positive identifications and were manually verified. (10) Gharahdaghi, F.; Weinberg, C. R.; Meagher, D. A.; Imai, B. S.; Mische, S. M. Electrophoresis 1999, 20, 601-605.

8180

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

Strict manual analysis was applied to validate protein identification results. The following criteria were used for manual verification: y, b, and a ions as well as their water loss or amine loss peaks were considered. For doubly charged or triply charged ions, it was necessary to match all the major isotope-resolved peaks with fragment masses of the identified peptide. The isotoperesolved peaks were emphasized because single peaks can arise because of electronic sparks and are less likely to be relevant to peptide fragments. The major isotope-resolved peaks for doubly charged or triply charged ions were defined as (1) those isotoperesolved daughter ions with m/z higher than the parent m/z and intensity greater than 5% of the maximum intensity or (2) those isotopically resolved peaks with intensities greater than 20% of the maximum intensity and m/z values between one-third of the parent m/z and the parent m/z. Typically, more than seven independent (amine and water loss not counted), isotope-resolved peaks were matched to theoretical masses of the peptide fragments. For manual validation of a singly charged parent ion, the Mascot score was required to be equal to or above the identity score threshold of the peptide. For peptides ending with arginine or lysine, at least four consecutive amino acids in the peptide sequence had to be confirmed by the isotopically resolved b-ion and y-ion series. RESULTS AND DISCUSSION Duty Cycle of Rapid, Isocratic HPLC/MS/MS Analysis. HPLC/MS/MS analysis to identify gel-separated proteins is usually carried out in a relatively short HPLC gradient (8-30 min) using a methanol or acetonitrile buffer system. In this procedure, the pressure of the C18-HPLC column (typically 75-200-µm i.d.) should be equilibrated after each cycle of HPLC/MS/MS analysis to maintain reproducible HPLC performance. Under our experimental conditions of an 8-min gradient from 25 to 90% buffer B, each cycle of HPLC/MS/MS analysis takes ∼45 min. Under these conditions, and using a constant flow rate of 0.6 µL/min, the peak widths for two standard peptides (synthetic peptide M, with low hydrophobicity, and bombesin, with high hydrophobicity) were 6 and 40 s, respectively (Figure 1A, B). Using dynamic flow elution, with flow rate dropping from 0.6 to 0.2 µL/min at 13 min, the peak widths for peptide M and bombesin were 12 and 90 s, respectively (Figure 1C, D). The peak widths with these flow rates are much wider than the time required to acquire one MS and four MS/MS spectra in the LTQ mass spectrometer, a total of less than 1.2 s for one MS and four MS/MS spectra. It takes ∼15 min to equilibrate the column and balance its pressure when an 8-min HPLC gradient is applied under our optimized flow rate of 0.6 µL/min (Figure 1). To reduce the time for each HPLC/MS/MS analysis, we tested the possibility of using isocratic elution and a high flow rate for HPLC/MS/MS analysis. Isocratic elution using 90% methanol buffer B and 10% methanol buffer A allowed much faster equilibration of the HPLC column, such that each HPLC/MS/MS analysis could be completed in 5 min (Figure 2). Thus, the duty cycle for each HPLC/MS/MS analysis was 9-14 times shorter. Using isocratic elution, column equilibrium is much faster because the flow path always contains the same high-organic elution buffer and the column does not reach complete equilibrium in aqueous phase, reducing the time required before and after sample loading.

Figure 1. Base peak and back pressure chromatograms of conventional HPLC/MS analysis in the methanol buffer system. (A) Base peak chromatogram and (B) HPLC back pressure chromatogram with flow and gradient diagrams, eluting with an 8-min gradient from 25 to 90% buffer B under constant flow (0.6 µL/min). (C) Base peak chromatogram and (D) HPLC back pressure chromatogram with flow and gradient diagrams, eluting with an 8-min gradient from 25 to 90% buffer B, where flow rates were set to 0.6 µL/min from 0 to 13 min, 0.2 µL/min from 13 to 25 min, and 0.6 µL/min from 25 to 45 min. For this analysis, 4 fmol each of bombesin and peptide M in HPLC buffer A was injected. The compositions of buffer A and buffer B are given in Materials and Methods.

Figure 2. Base peak and back pressure chromatograms of isocratic HPLC/MS analysis in the methanol buffer system. (A) Base peak chromatogram and (B) HPLC back pressure chromatogram with flow and gradient diagrams, using isocratic elution and constant flow (3.2 µL/min). (C) Base peak chromatogram and (D) HPLC back pressure chromatogram with flow and gradient diagrams using isocratic elution and dynamic flow, where flow rates were set to 3.8 µL/min from 0 to 1.4 min, 0.8 µL/min from 1.4 to 3.0 min, and 3.8 µL/min from 3 to 5.0 min. For this analysis, 4 fmol each of bombesin and peptide M in HPLC buffer A was injected.

Peak Width and Intensity of Isocratic Analysis. Under our isocratic elution conditions, we observed that the peak widths for synthetic peptide M and bombesin were 1.8 and 4.8 s, respectively, using a constant flow rate of 3.2 µL/min (Figure 2A, B), and 4 and 6 s, respectively, using dynamic flow elution with the flow rate dropping from 3.8 to 0.8 µL/min during dynamic elution (Figure 2C, D). We compared the sensitivity of gradient elution with that of isocratic elution. Under optimized conditions and using dynamic flow elution, a bombesin peak was compared to noise signal level at m/z 750-760 and showed a signal-to-noise ratio of ∼500 using either the 8-min gradient elution or the isocratic elution (Figure 3A, B, respectively). We rationalize that comparable sensitivity is observed with isocratic elution because the narrow width of the eluted peak and reduced sample diffusion compensate for the peptide loss in early loading, the latter being expected because the column is equilibrated with high-organic buffer. Next, we checked whether isocratic elution would provide a time window wide enough for MS and MS/MS analysis. In the LTQ, it takes ∼200 ms to acquire an MS spectrum and 250 ms to

acquire an MS/MS spectrum when maximum trapping time is set to 100 ms and one microscan is specified for both MS and MS/MS. For comparison, these times are shorter than those required by the commonly used LCQ XP instrument, in which it takes ∼1.2 s to acquire an MS spectrum and ∼1.8 s to acquire an MS/MS spectrum when maximum trapping time is set to 110 ms and three microscans are specified for both MS and MS/MS. Using isocratic elution into the LTQ, the two standard peptides generated signals with peak widths varying from 1.8 to 4.8 s under a constant flow rate of 3.2 µL/min, and from 4 to 6 s under dynamic flow elution, long enough to acquire one MS and four MS/MS analyses. In contrast, under the traditional 8-min gradient condition and using the LCQ XP, the two standard peptides generated signals with peak widths of 12 and 90 s. It should be noted that the wider peaks observed with the 8-min gradient do not result in more useful MS/MS information, as only one MS/MS spectrum is needed for a peptide identification. Modulation of Peak Widths for Hydrophilic Peptides with Dynamic Flow Rate and Methanol Buffer System. Under a Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

8181

Figure 3. Comparison of bombesin sensitivity under 8-min gradient and isocratic elution using dynamic flow in methanol buffer system. MS spectra in the range of m/z 750-850 are zoomed in 10-fold to show the relative signal-to-noise ratio of the bombesin peak detected during (A) gradient elution and (B) isocratic elution in methanol buffer (bombesin relative peak intensity ) 100). HPLC conditions for panels A and B were as described in Figures 1 and 2, respectively, using dynamic flow elution.

constant flow rate of 3.2 µL/min, the peak width for some hydrophilic peptides is close to the minimum time required for each MS and MS/MS analysis in the LTQ. When analyzing real samples, peak widths of peptides would be much shorter than the observed peak widths of standard peptides, due to increased noise level, low peptide concentration, and higher sample complexity. These short peak widths could compromise the quality of MS/MS spectra and the sensitivity of peptide identifications, especially for hydrophilic peptides that elute early. To address this problem, we first applied a dynamic flow rate to widen the peaks for hydrophilic peptides. The flow rate was dropped from 3.8 to 0.8 µL/min at 1.4 min, right before the hydrophilic peptides started to be eluted. The low flow rate (0.8 µL/min) was maintained for 1.6 min. Then, the flow rate was raised to the initial flow rate (3.8 µL/min) at 3.0 min, when more hydrophobic peptides started to be eluted. Using this scheme, one cycle of analysis could be finished in 5 min, as indicated by back pressure reaching equilibrium at the end of each analysis (Figure 2D). The acetonitrile buffer system has an elution capability different from the methanol buffer system. We compared the methanol and acetonitrile buffer systems described in Materials and Methods in terms of their performance for HPLC/MS in the LTQ. The sensitivity of the methanol buffer system was a little lower than that of the acetonitrile buffer system (within 40%, see Figure 4B, D). However, peaks tended to be wider with the methanol buffer 8182 Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

Figure 4. Comparison of HPLC/MS results using the acetonitrile (A, B) and methanol (C,D) buffer systems, using isocratic elution conditions. Panels A and C show the HPLC/MS base peak chromatograms, and panels B and D show the signal-to-noise ratio for bombesin peptide (bombesin relative peak intensity 100). For these analyses, 4 fmol each of bombesin and peptide M in HPLC buffer A (containing methanol or acetonitrile, as indicated in Materials and Methods) was injected.

system (Figure 4A, C), and therefore, this system may be more suitable under high flow rate conditions. Our results indicate that the methanol buffer system is better able to generate suitable peak

Figure 5. Dynamic range of isocratic HPLC/MS/MS in the methanol buffer system. (A) Base peak chromatogram of 100 fmol of myoglobin digested with trypsin and mixed with 20 amol of bombesin. (B) Full mass spectrum at 2.34 min showing the doubly charged bombesin peptide. Tandem mass spectrum of the labeled peak led to MS/MS identification of bombesin (data not shown). The peptides were loaded in buffer A.

widths for MS and MS/MS analysis for both hydrophilic and hydrophobic peptides. Dynamic Range of HPLC/MS/MS Analysis under Isocratic Elution Conditions. We tested the dynamic range of our optimized isocratic elution HPLC procedure by adding 20 amol of bombesin to a tryptic digest of 100 fmol of myoglobin (Figure 5). In the LTQ mass spectrometer, we were able to identify and obtain tandem mass spectra for bombesin with a dynamic range of ∼5000:1. We have routinely applied our 5-min/cycle isocratic HPLC/ MS/MS method to identify gel-separated proteins. One example is shown in Figure 6. Protein database searching using spectra generated from isocratic HPLC coupled to MS/MS analysis in the LTQ mass spectrometer led to identification of seven proteins in one silver-stained SDS-PAGE band (between 35 and 38 kDa). Only 2 µg of mouse liver nuclear extracts was loaded onto the lane. All the protein identifications were manually verified. In case the 5-min/cycle isocratic elution protocol with dynamic flow cannot balance column pressure at the end of each duty cycle due to the limitations of HPLC pump or columns, a more reproducible, 10-min/cycle isocratic elution HPLC method with dynamic flow (using the same methanol/water buffer system) can be applied. The flow and gradient diagrams of 10-min/cycle isocratic elution HPLC method is shown in Figure 7. We realize that this procedure is not applicable to samples containing a complicated protein mixture (e.g., more than 10

Figure 6. Identification of proteins from a weakly silver-stained band using isocratic elution with the methanol buffer system. (A) Silverstained gel image showing the excised band, band A, containing proteins with molecular weights between 35 000 and 38 000. Only 2 µg of mouse nuclear extract was loaded into the lane. (B) HPLC base peak chromatogram and (C) HPLC back pressure chromatogram with flow and gradient diagrams of tryptic peptides from band A, eluted with 5-min isocratic HPLC/MS/MS into the LTQ mass spectrometer. Seven proteins were identified, as listed in panel D.

proteins). For such samples, elution with a slow gradient over 10-60 min provides greater dynamic range and allows identification of the maximum number of proteins. However, the procedure described here will be especially useful for efficient analysis of samples from 1D or 2D gels SUMMARY In summary, we developed a procedure for rapid HPLC/MS/ MS analysis in the LTQ mass spectrometer. The procedure involves isocratic elution of peptides from the HPLC column with Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

8183

Figure 7. Flow and gradient diagrams for 10-min isocratic HPLC elution with dynamic flow in methanol buffer system. Flow rate was set to 1.9 µL/min from 0 to 2.7 min, 0.4 µL/min from 2.7 to 7.0 min, and 1.9 µL/min from 7 to 10.0 min, while %B was set to 90% under isocratic elution condition.

high-organic buffer. However, the peptide digest is solubilized and loaded in 2 µL of low-organic buffer. During loading, a small portion of peptides might flow right through the column because it is equilibrated with high-organic buffer. However, this sample

8184

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

loss is compensated for by improved sensitivity due to narrow peaks (because peptides are eluted with high-organic buffer) and low peptide diffusion. Our method generates peptide peaks with widths between 4 and 6 s, wide enough for MS/MS analysis in the LTQ mass spectrometer. Under our experimental conditions, 1.2 s is sufficient for each cycle of one MS and four MS/MS experiments in the LTQ. The protocol allows us to complete each HPLC/MS/MS analysis in 5 min, thus increasing efficiency of the analysis 9-14fold. The method is routinely used in our laboratory for HPLC/ LTQ mass spectrometry. Using this procedure, each operator is able to generate data for 96 in-gel digested samples within one working day (8 h). ACKNOWLEDGMENT Y.Z. is supported by The Robert A. Welch Foundation (I-1550) and NIH (CA 107943). We are grateful to George Kemp for technical help. Received for review August 15, 2005. Accepted October 7, 2005. AC051468T