Capillary Chromatography-Coupled Mass Spectrometry with Column Switching for Rapid Identification of Proteins from 2-Dimensional Electrophoresis Gels David C. Delinsky and Kenneth D. Greis* Procter & Gamble Pharmaceuticals, 8700 Mason-Montgomery Road, Mason, Ohio 45040 Received February 22, 2002
Abstract: An improved capillary liquid chromatography procedure, incorporating column switching in combination with mass spectrometry, is reported. The dual column system allows for rapid inject-to-inject cycle times to improve the speed of protein identification for proteomics applications. Full gradient elution of peptides from either of the two C18 columns can be achieved in less than 17 min while maintaining sufficient resolution for the peptides to be detected and fragmented by the mass spectrometer for protein identification. Importantly, the use of two columns for subsequent injections is reproducible and without carry-over. The limit of detection for the system is between 25 and 50 fmol per injection. This fully automated system is capable of analyzing and identifying proteins from an entire 96-well plate in about 27 h. Keywords: LC-ESI-MS/MS • protein identification • column switching • proteomics • mass spectrometry
Introduction Protein identification by mass spectrometry has become the method of choice for proteomics studies. Increasingly in the pharmaceutical and biotechnology industry, the need to improve the throughput of protein identification has led to several innovations in instrument design and automation.1 Given that it is often necessary to analyze hundreds of samples in a short period of time, system throughput is a critical issue for many proteomics mass spectrometry groups. The two primary modes of protein identification include (1) generating peptide “mass fingerprints,” typically by MALDI-TOF-MS and (2) MS/MS fragmentation sequence tags using LC-ESI-MS/MS. Details of these methods have recently been reviewed.1 The challenge in the latter case is to increase the sensitivity of detection by going to smaller capillary columns and lower flow rates while maintaining a reasonable sample throughput. This is no simple task given that each decrease in flow rate results in additional delay times due to the inherent dead volumes in any capillary LC system. A review of the capillary LC-MS/MS literature for proteomics studies reveals that most of the method development has been directed toward achieving additional resolution and peptide * To whom correspondence should be addressed. Tel: (513) 622-2670. Fax: (513) 622-1196. E-mail:
[email protected]. 10.1021/pr020004y CCC: $22.00
2002 American Chemical Society
coverage for tryptic digests of complex protein mixtures either by using “peak parking” methods2 or multidimensional chromatography.3 On the other hand, high throughput, parallel LCMS applications with column switching have been reported for small molecules, bioanalytical applications, and intact proteins.4-7 Many of the small molecule approaches have the advantage of either having greater sample quantities available and/or knowledge of the expected mass for the compound of interest. For example, for many bioanalytical applications, the mass spectrometer can be focused solely on the expected compound by predetermined fragmentation characteristicss an approach that greatly increases the sensitivity of detection. Since there is usually no information about the expected peptides from unknown proteins in proteomics samples, the mass spectrometer must be tuned to effectively fragment the wide range of peptides produced by protease digestion. Thus, the increase in speed and sensitivity for proteomics samples must be achieved by a combination of capillary flow and fast gradient separation rather than through selective fragmentation in the mass spectrometer as in bioanalytical applications. In this paper, we demonstrate greater than 2-fold improvement in the throughput of capillary LC-ESI-MS/MS analysis for protein identification by using a two-column system where one column is washed and injected with sample while the other column is in line with the mass spectrometer during gradient elution. The system was tested for sensitivity, robustness, sample carry-over, and overall utility for common proteomics samples.
Experimental Section Chemicals and Reagents. Bovine serum albumin (BSA), rabbit erythrocyte carbonic anhydrase (CA), and horse skeletal muscle apomyoglobin (Myo) were all purchased from Sigma. Ultrapure formic acid was from EM Science, and sequencing grade modified porcine trypsin was from Promega. Equipment. The capillary chromatography system was from LCPackings USA (San Francisco). The system consisted of an Ultimate gradient pumping unit, Switchos II dual 10-port valve switching unit with an auxiliary isocratic pumping unit, and a FAMOS microtiter plate autosampler.8 The mass spectrometer used was a Finnigan LCQDeca quadrupole ion-trap mass spectrometer equipped with a custom-built liquid junction microspray ionization source containing a 50 µm i.d. tapered fused silica spray tip (catalog no. TT360-50-50-N-5) from New Objectives (Boston, MA). Journal of Proteome Research 2002, 1, 279-284
279
Published on Web 04/19/2002
technical notes
Rapid capLC-ES-MS/MS via Column Switching
Figure 1. Dual column configuration on a 10-port switching valve. Two capillary C18 columns were configured with the gradient pump and the auxiliary pump/autosampler as described in the Experimental Section: panel A, column one in line with gradient pumps and detector; panel B, column one in line with auxiliary pump and autosampler.
Sample Preparation. Standard digests of BSA, CA, and Myo were prepared by incubation at a 20:1 molar ratio of protein to trypsin at 37 °C for 16 h. The digestion reaction was quenched by acidification with TFA, and the digests were dried and then suspended in mobile phase A to the desired concentration. In-gel digestion of silver-stained protein spots was conducted as described previously9 except for the addition of an initial destaining step in a solution of freshly prepared 50 mM sodium thiosulfate and 15 mM potassium ferricyanide.10 Protein spots were digested with 20 ng of sequencing grade porcine trypsin (Promega, Madison, WI) in 50 mM ammonium bicarbonate overnight to release tryptic peptides. Tryptic peptides were extracted from the gel spot with 50% CH3CN/ 0.3%TFA. Recovered peptides were dried in a SpeedVac concentrator in 96-well plates to remove residual ammonium bicarbonate prior to capillary LC-ESI-MS/MS. Capillary Liquid Chromatography. All chromatography was performed on 0.3 i.d. × 50 mm PepMap C18 columns (LCPackings, USA) at 4 µL/min flow rate. Mobile phase A consisted of 98:2 water/CH3CN with 0.1% formic acid; mobile phase B consisted of 2:98 water/CH3CN with 0.1% formic acid. The final optimized gradient proceeded at 4 µL/min from 0 to 39% B over the first 10.5 min; then to 90% B at 11 min; held constant for 1 min; ramped back to initial conditions over 1 min; and then held constant for 3.5 min to allow all mobile phase B to exit the gradient pump. All other gradient conditions were as indicated in the text or figure legends. A second column was integrated into the system using a 10-port valve on the Switchos II module of the LC system. This configuration allowed the peptides from one column to be separated and detected (by UV or mass spectrometry), while the other column was allowed to reequilibrate via the auxiliary pump. The columns were attached to the 10-port valve as depicted in Figure 1. A connecting capillary between ports 4 and 9 was used to allow flow from the autosampler to reach both columns depending on the valve position. The auxiliary pump flowed at a constant 5 µL/min in 100% mobile phase A. During initial work, 15 picomoles of trypsin-digested BSA per injection was used with UV detection at 215 nm. For mass spectrometric detection, typically between 25 and 500 femtomoles of digest was injected. Capillary LC-Coupled Mass Spectrometry. The dried protein digests from the 96-well plates were analyzed by capillary LCESI-MS/MS with an LCPackings (San Francisco, CA) Ultimate 280
Journal of Proteome Research • Vol. 1, No. 3, 2002
capillary LC system equipped with a FAMOS micro-autosampler and a 0.3 i.d. × 50 mm PepMap C18 column coupled to a Finnigan LCQDeca ion trap mass spectrometer (ThermoFinnigan, San Jose, CA). Data were collected and analyzed using Finnigan Xcalibur v. 1.2 software in data-dependent scan mode such that any peptide signal over 1 × 105 intensity triggered the automated acquisition of a higher resolution “ZoomScan” followed by an MS/MS fragmentation spectrum for each detected peptide. The mass spectrometer was also set to exclude the fragmentation of any peptide that had already been analyzed and an MS/MS spectrum collected. Dual Column Switching. To coordinate multiple injections and column switching, user-defined autosampler programs (UDP) were written for the FAMOS autosampler to control the sample re-suspension, loading, injection and column switching through the Switchos column-regenerating pump and the 10port valve (see the Supporting Information for complete UDPs). To initiate analyses from a 96-well plate of samples, the first UDP (UDP-Start) simply instructed the autosampler to resuspend and inject the first sample onto column 1. The 10-port valve was then switched to bring the gradient pump in line with column 1 (Figure 1A). In parallel with the gradient elution of peptides from column 1, column 2 was equilibrated and preloaded with sample (UDP-1). Upon completion of the gradient on column 1, the next UDP (UDP-2) instructed the 10-port valve to switch and thus bring column 2 in line with the gradient pump (Figure 1B) and the detector, while column 1 was reequilibrated and preloaded with the next sample. The alternating cycle continued to switch from column 1 to column 2 until the last sample was loaded. The last gradient was then initiated without loading a new sample onto the alternating column by using the final UDP (UDP-End). Note that the dried digests (in 96-well plates) were reconstituted in 5 µL mobile phase A by the autosampler just before injection to minimize solvent evaporation. After reconstitution, the sample was center loaded onto a 10 µL loop with mobile phase A as the makeup solvent.
Results and Discussion Gradient Optimization. As a starting point for gradient optimization, the separation conditions supplied by the vendor upon installation of the capLC system were usedsa linear
technical notes
Figure 2. Gradient optimization with UV detection at 215 nm. 15 pmol injections of trypsin-digested BSA were used to optimize the gradient elution time: panel A, gradient rate at 1.8% B/min; panel B, gradient rate at 3.7% B/min.
Figure 3. Column switching with a gradient rate at 3.7% B/min. 15 pmol injections of trypsin-digested BSA were analyzed using a dual column configuration as shown in Figure 1: panel A, UV detection at 215 nm on column two with a 300 µL of static mixer in the system; panel B, UV detection on column two after removal of the static mixer.
Figure 4. Consecutive injections with column switching. BSA digests and gradient elution were as in Figure 3B: panel A, UV detection at 215 nm on column one; panel B, UV detection on column two.
gradient from 5 to 60%B in 30 min. These conditions resulted in a good separation of the BSA tryptic digest within 24 min (Figure 2A). However, the inject-to-inject cycle time for this method was 52 min including a high organic wash and column reequilibration. Thus, the rate of change of the gradient was optimized from 1.8%B per min over 30 min (as above) to 3.7%B per min over 10.5 min. The net result of this optimization was complete peptide elution within 15 min and an inject-to-inject cycle of 27 min (Figure 2B). Importantly, sufficient resolution
Delinsky and Greis
Figure 5. Timeline of simultaneous events on columns 1 and 2: panel A, overall time to cycle the 10-port valve from one column to the other; panel B, the timing of events on column one; panel C, the timing of events on column two.
of the peptides was maintained to establish a 6 min elution window for the peptides. Given that the mass spectrometer can collect a full series of data-dependent automated MS/MS spectra every 0.07 min, the 6 min elution window provides sufficient peptide resolution to collect fragmentation data on up to about 85 different peptides in a single separation. Whereas an increase in the gradient rate ultimately produces some overlap or coelution of peptides, the mass spectrometer can more than compensate for the coelution by incorporating the dependent scan features and the exclusion of peptide masses previously used to produces MS/MS fragmentation spectra. Column Switching. As described above using a single column and a linear gradient at 3.7%B/min, the peptides could be eluted within 15 min, but it still required a total of 27 min before the next run could be initiated. These additional 12 min were primarily the result of the time required to wash the column up to 90% B to strip any residual peptides and to reequilibrate the column back to initial conditions. Thus, by adding a second column to initiate another separation while the first column was being reequilibrated with an auxiliary pump, a portion of this time could be eliminated. Unfortunately, attempts to switch to the second column at the 15 min mark resulted in no retention of peptides on the second C18 columnsthey simply eluted through in the void volume of the column (Figure 3A). After some troubleshooting, it was concluded that this was due to the high concentration of organic solvent from the previous gradient that was still present in the transfer line between the gradient pump and the columnswitching valve. Additional investigation determined that the primary culprit was the large volume (300 µL) of the static mixer of the system in conjunction with the volume of the transfer line that in turn caused a delay of about 4.0 min from the time the gradient was formed until the gradient reached the column. Upon removal of the static mixer, the gradient delay was minimized and the peptides in successive runs were bound to the column and eluted within 15 min (Figure 3B) as was previously determined for a single column configuration (Figure 2B). Additional reproducibility studies, on successive runs in the absence of the 300 µL static mixer, indicated that sufficient mixing was occurring to prevent significant changes in the peptide retention times (Figure 4). Furthermore, since this system was being optimized to use a mass spectrometer as the detector, slight variations in retention times are inconsequential Journal of Proteome Research • Vol. 1, No. 3, 2002 281
Rapid capLC-ES-MS/MS via Column Switching
technical notes
Figure 6. Assessment of detection limits by LC-MS and MS/MS. The indicated amount of digested BSA was loaded onto the dual column system and eluted directed into the mass spectrometer as in Figure 4: panels A-C, the full scan mass spectra of a doubly charged peptide at m/z 582.6 at 100, 50, and 25 fmol of digested protein injected; panels D-F, the automated MS/MS fragmentation spectra collected in the scan following the full scan spectra in panel A-C, consecutively.
for protein identification via automated peptide fragmentation and database searching using sequence tags. However, future enhancements to the system might include the addition of a low volume dynamic mixer to improve peak resolution and thus peak height and sensitivity of detection in the mass spectrometer. Concurrent with optimizing the gradient speed and the timing of the column switching, it was necessary to address the amount of time it took the autosampler to advance through its injection routine prior to the start of the gradient. Since our proteomics samples were provided as dried tryptic peptides in a 96-well plate, the autosampler method was defined to resuspend the samples in 5 µL of mobile phase A and then load 4 µL (80%) of the sample into the sample loop. The remaining sample (20%) was saved for subsequent analysis, as needed. Finally, the autosampler routine included a thorough washing of the syringe and capillary tubing to prevent sample carry-over into the next run. Overall, the injection routine 282
Journal of Proteome Research • Vol. 1, No. 3, 2002
averaged about 2.5 min/sample. To minimize this dead time, we took advantage of the user define programming (UDP) feature of the FAMOS autosampler to control both the injection method and the switching valves. As shown in Figure 5, this routine allowed for column 1 to be in line with the gradient pump so that the peptides were eluted to the detector (as in Figure 1A), while column 2 was reequilibrated and loaded with the next sample in line with the auxiliary pump and the autosampler. Once the valve was switched (as in Figure 1B), the gradient was applied to column 2 while column 1 was reequilibrated and loaded with the next sample. Thus by preloading the next sample onto the alternate column, a time savings of 2.5 min/sample was realized which extrapolated to about 4 h saved per 96-well plate. System Validation. After optimizing the system for gradient speed, column switching, and preloading of samples as described above, the system was validated by testing for robustness, sensitivity, sample carry-over, and with real protein spots
technical notes
Delinsky and Greis
Table 1. Assay for Sample Carry-Over from Injection to Injection run
digest
fmol injected
column
1 2 3 4 5 6 7 8 9 10 11
BSA blank blank BSA CA CA BSA Myo Myo Myo BSA
500 0 0 500 50 50 500 50 50 25 25
1 2 1 2 1 2 1 2 1 2 1
a
protein detecteda
Table 2. Identification of Proteins from a Silver-Stained 2-D Gel
carry-over
spot no.
protein identificationa
no. of peptides
column
carryover
none none none none none none none none none none none
1 2 3 4 5 6 7 8 9 blank
hemopexin precurser serum albumin precurser (rat) pyruvate kinase Enolase 1, R preprohaptoglobin aldehyde dehydrogenase 3 phosphoglycerate kinase R-1-macroglobin aldehyde dehydrogenase 3 autotryptic peptides only
4 10 12 15 4 3 4 4 3 3
1 2 1 2 1 2 1 2 1 2
none none none none none none none none none none
BSA BSA CA CA BSA Myo Myo Myo BSA
a
Based on MASCOT search of entire capLC-ES-MS/MS run.
Based on MASCOT search of entire capLC-ES-MS/MS run.
from a 2-D gel. First, Figure 4 shows the similarity of separations of successive injections from one column to the other, even with the predicted variability that resulted from removing the static mixer. Next, the system was attached to the ion-trap mass spectrometer to determine sample carry-over and sensitivity in data-dependent MS/MS experiments. Thirty consecutive injections were made using trypsin-digested BSA, Myo, or CA. The resulting MS/MS fragmentation spectra were searched against a protein database for protein identification. Consistent, protein matches for each of these samples confirmed the robustness of the system (not shown). As part of the 30 consecutive injections, several 500 fmol injections of BSA were followed by two blank injections or two consecutive 50 fmol injections of CA or Myo in order to test for any sample carry-over (Table 1). It was necessary to run two blank or two low level samples to determine carry-over since the first was to check for BSA carryover in the sample loop and tubing before the columns and the second was to check for residual BSA that remained on the column from the previous run. BSA was not detected in any of the subsequent injections indicating that there was no carry-over detected in the system. The sensitivity of the system was tested by injection of trypsin-digested BSA at 100, 50, and 25 fmol per injection. The limit of detection was based on the number of peptides identified and their signal strength. Figure 6 shows the full scan mass spectra (panels A-C) and MS/MS data (panels D-F) for a doubly charged peptide at m/z 582.6 at the 100, 50, and 25 fmol injection levels. Note that the quality of the MS/MS data for all three samples was nearly the same (panels D-F) even though the full scan peak intensity for the 25 fmol injection (panel C) is less than the height of some background noise. Thus, the absolute signal-to-noise for the peptides is not a good indicator of sensitivity when data-dependent MS/MS data are collected. The best indicator of sensitivity is the production of MS/MS fragmentation spectra suitable to identify the protein via a database search. For BSA, this level of detection was 25 fmol injected, while results for Myo and CA ranged from 25 to 50 fmol as the low limit of detection. While it is likely that each protein may have a different minimal level of detection by MS/ MS fragmentation, based on these three model proteins, we determined that the limit of detection for this system configuration is between 25 and 50 fmol. To demonstrate the usefulness of the dual column system, nine random spots and a blank were taken from a silver-stained 2-D gel. The proteins were digested with trypsin and prepared as described in the Experimental Section. The results from the
analyses of these samples are displayed in Table 2. Each protein was identified with an average of six peptide fragmentation spectra (minimum of three) with no sample-to-sample carryover noted. Subsequently, three additional plates with greater than 75 samples each have been analyzed in automated fashion on this dual column switching system. Results were similar to those obtained previously on a single column system (not shown). Thus we have demonstrated that column switching can be effectively used to improve throughput for proteomics samples by capillary LC-ESI-MS/MS without compromising the sensitivity and quality of the protein characterization.
Conclusions A dual column capillary LC system is demonstrated to enhance the throughput of proteomic protein identification. The optimized system produced greater than 2-fold increase in the speed of analysis without compromising sensitivity of detection, thus allowing for the complete analysis of a 96 well plate of samples in just over 26 h versus over 60 h prior to optimization of the column switching and integration of the injection system. Importantly, the system was shown to be effective for real proteomic samples from silver-stained 2D gels. Finally, the system has been reliable enough that we have integrated it into our standard proteomic protein identification regimen.
Acknowledgment. We thank Scott Brittain for assistance with the mass spectrometry experiments and Larry Thompson for critical review of the manuscript. Supporting Information Available: Tables containing four user-defined programs (UDPs) used to control the autosampler, auxiliary pump, and 10-port switching valve. This material is available free of charge via the Internet at http:// pubs.acs.org. References (1) Hoke, S. H., II; Morand, K. L.; Greis, K. D.; Baker, T. M.; Harbol, K. L.; Dobson, R. L. M. Int. J. Mass Spectrom. 2001, 212, 135196. (2) Davis, M. T. and Lee, T. D. J. Am. Soc. Mass Spectrom. 1997, 8, 1059-1069. (3) Link, A. J.; Eng, J.; Schieltz, D. M.; Carmack, E.; Mize, G. J.; Morris, D. R.; Garvik, B. M.; Yates, J. R. Nat. Biotechnol. 1999, 17, 676682. (4) De Biasi, V.; Haskins, N.; Organ, A.; Bateman, R.; Giles, K.; Jarvis, S. Rapid Commun. Mass Spectrom. 1999, 13, 1165-1168. (5) Foltz, D. J.; Bornes, D. M.; Peng, S. X.; Baker, T. M. Proceedings of the 47th ASMS Conference on Mass Spectrometry and Allied Topics, Dallas, TX, June 13-17, 1999.
Journal of Proteome Research • Vol. 1, No. 3, 2002 283
technical notes
Rapid capLC-ES-MS/MS via Column Switching (6) Van Pelt, C. K.; Corso, T. N.; Schultz, G. A.; Lowes, S.; Henion, J. Anal. Chem. 2001. 73, 582-588. (7) Feng, B.; McQueney, M. S.; Mezzasalma, T. M.; Slemmon, J. R. Anal. Chem. 2001, 73, 5691-5697. (8) Chervet, J.-P.; Van Ling, R.; Evans, K.; Salzmann, J.-P. Am. Lab. 1999, 31, 44, 46, 48, 50.
284
Journal of Proteome Research • Vol. 1, No. 3, 2002
(9) Shevchenko, A.; Wilm, M.; Vorm, O.; Mann, M. Anal. Chem. 1996, 68, 850-858. (10) Gharahdaghi, F.; Weinberg, C. R.; Meagher, D. D.; Imai, B. S.; Mische, S. M. Electrophoresis. 1999, 20, 601-605.
PR020004Y
