Anal. Chem. 2001, 73, 2985-2991
Very High Pressure Gradient LC/MS/MS Luke Tolley,† James W. Jorgenson,† and M. Arthur Moseley*,‡
Department of Chemistry, University of North Carolina at Chapel Hill CB 3290, Chapel Hill, North Carolina 27599-3290, and Structural Chemistry Department, Glaxo Research Institute, P.O. Box 13398, Research Triangle Park, North Carolina 27709-3398
A very high pressure liquid chromatography (VHPLC) system was constructed by modifying a commercially available pump in order to achieve pressures in excess of 1200 bar (17 500 psi). A computer-controlled lowpressure mixer was used to generate solvent gradients. Protein digests were rapidly analyzed by reversed-phase VHPLC with linear solvent gradients coupled to either a tandem mass spectrometer using electrospray ionization or a UV/visible detector. The separations were performed at pressures ranging from 790 (11 500 psi) to 930 bar (13 500 psi) in 22-cm-long capillary columns packed with C18-modified 1.5-µm nonporous silica particles. A digest of bovine serum albumin (BSA) was analyzed by the VHPLC system connected to a mass spectrometer in MS mode. An analysis of 12.5 fmol of sample gave signal-tonoise ratios of tryptic peaks greater than 10:1 in the base peak plot mass chromatogram. This system was also used to analyze a proteolytic digest of a rat liver protein excised from a 2-D gel separation of a liver tissue lysate. For this analysis, the mass spectrometer was set up to perform data-dependent scanning (automated switching from MS mode to MS/MS mode when a peak was detected) for peptide sequencing and protein identification by database searching. The results of this analysis are compared to an analysis performed on the same sample using the nanoelectrospray-MS/MS technique. Though both techniques were able to identify the unknown protein, the VHPLC method gave twice as many sequenced peptides as nanoelectrospray and improved the signal-to-noise ratio of the spectra by at least a factor of 10. Direct comparisons with nanoelectrospray for MS and MS/MS data acquisition from a BSA digest were made. These comparisons show enhancements of greater than 20-fold for VHPLC over nanoelectrospray. In addition, the VHPLC/MS/MS data acquisition was accomplished in an automated manner. The use of smaller packing material in HPLC columns can produce separations with better resolution in less time than a column using larger particles.1 Using reduced parameters and an equation for flow through a packed bed allows calculation of the effects of particle size on column performance. According to these equations, the minimum plate height achievable is equal to twice †
University of North Carolina. Glaxo Research Institute. (1) Snyder, L. R.; Kirkland, J. J. Introduction to Modern Liquid Chromatography; John Wiley and Sons: New York, 1979.
the particle diameter. As the particle size decreases, the column can therefore generate more plates per meter. A decrease in particle diameter also causes the optimum linear velocity to increase as the inverse of the particle diameter, permitting a faster separation. The main drawback to using smaller packing material is that the pressure required for optimum linear velocity increases as the inverse of the particle diameter cubed.2 The transition from 5-µm particles to 1.5-µm particles in the same length of column should both reduce analysis time and increase the number of theoretical plates by a factor of 3.3 at optimum flow rates, but the required pressure would increase by a factor of 37. Though columns packed with 1.5-µm particles are commercially available, the packed beds are typically only 3 cm long, which is much shorter than conventional columns packed with 5-µm particles. The short column length is used to decrease the pressure requirements for optimum mobile-phase flow rates to levels that are achievable with conventional HPLC systems. These columns do provide a much faster analysis, but with a decrease in resolution and separating power due to the short length of the packed bed. An increase in the maximum pressure available from an HPLC system would make longer columns with small particles practical, providing faster separations with better separation efficiency. Assuming an analyte with a diffusion coefficient of 6 × 10-6 cm2/s, a 25-cm column packed with 1.5-µm particles running at optimum flow rates would have a dead time of ∼200 s and would produce 83 000 plates in an isocratic separation. However, the required pressure to achieve the optimum flow rate is ∼900 bar (13 000 psi). Building a pump system to use such a column with flexible gradient programming capability was the goal of this research. The increased separation power possible with smaller packing material in a long column, coupled with a shorter analysis time, makes a very high-pressure system well suited to difficult analytical problems, such as the analysis of drugs, drug metabolites, and protein digests. To gain further information about the sample, the system can be connected to a hybrid quadrupole/ time-of-flight tandem mass spectrometer (MS/MS). The very high pressure liquid chromatography (VHPLC) system would provide rapid separation of the sample and the tandem mass spectrometer would allow for fragmentation and identification of the compounds present. Protein digests were used to test the system due to their applicability to other research conducted in the laboratories.
10.1021/ac0010835 CCC: $20.00 Published on Web 05/10/2001
© 2001 American Chemical Society
(2) MacNair, J. E.; Lewis, K. C.; Jorgenson, J. W. Anal. Chem. 1997, 69, 983989.
Analytical Chemistry, Vol. 73, No. 13, July 1, 2001 2985
One approach used to analyze complex peptide mixtures is to infuse the sample directly into the mass spectrometer using nanoelectrospray.3 Nanoelectrospray-MS/MS has proven to be an extremely useful tool for protein identification and characterization. The utility of this technique is based on the very low flow rates from nanoelectrospray needles (20-50 nL/min), the concentration-dependent response of electrospray, and the ability of mass spectrometers to resolve complex mixtures of peptides based on their mass-to-charge ratios. As the spectra are recorded and summed on the computer, the user manually selects which ions will be sequenced by MS/MS. While this technique is very useful, it requires constant user input during the analysis, which limits its practicality for high throughput. Moreover, nanoelectrospray has a low duty cycle since it is an infusion techniquesall analytes are flowing into the mass spectrometer all of the timeswhereas MS/MS spectra must be acquired sequentially. In addition, the time required for manual selection of precursor ions further decreases the duty cycle of the analysis. In the work presented here, fused-silica capillaries were slurry packed with 1.5-µm C18 bonded nonporous silica particles. Separations at pressures over 685 bar (10 000 psi) of known protein digests as well as an unknown sample are shown. The separations were performed using linear gradients of water/acetonitrile with either 0.1% trifluroacetic acid (TFA) or 0.5% formic acid by volume. The VHPLC system was coupled to a hybrid quadrupole time-offlight (Q-TOF) mass spectrometer using an electrospray ionization source. EXPERIMENTAL SECTION System Setup. A block diagram of the experimental setup is shown in Figure 1. The liquid chromatography system is based on a Waters model 6000 pump (Waters Associates, Milford, MA) which has been modified to operate at up to 1375 bar (20 000 psi). Several modifications were performed including the replacement of the existing check valves in the pump with check valves designed to operate at 1030 bar (15 000 psi) (Analytical Scientific Instruments, Richmond, CA), replacing the pressure sensor with one rated to 1375 bar (20,000 psi) (Omegadyne Inc., Sunbury, OH), and changing the gear ratio in the drive train of the pump to produce a higher force on the reciprocating pistons. The outlet check valves are connected to a 1/16-in. stainless steel tee (Valco Instrument Co. Inc., Houston, TX) by standard 1/ -in. stainless steel tubing with an inner diameter (i.d.) of 0.020 16 in. Steel (1/16-in.) Valco compression fittings were used to connect the tubing to the tee and were able to withstand these pressures without modifications. Vespel ferrules (Alltech Associates Inc., Deerfield, IL) were used wherever a connection to a fused-silica capillary was made. The pump was not originally designed for the low flow rates that are required in capillary separations. After the modifications to the pump, the lowest practical flow rate for gradient generation was 300 µL/min. A flow splitter with an inner diameter of 50 µm and length of 200 cm was needed to divert most of the flow from the pump, leaving only that needed for the column. A flow rate of 330 µL/min from the pump at a pressure of 790 bar was split to generate a column flow rate of 0.5 µL/min. (3) Wilm, M.; Mann, M. Anal. Chem. 1996, 68, 1-8.
2986 Analytical Chemistry, Vol. 73, No. 13, July 1, 2001
Figure 1. Block diagram of the gradient VHPLC system interfaced to a Q-TOF mass spectrometer. When the system was set up for UV detection, the Q-TOF and electrospray interface were replaced by a Linear UVIS 200 positioned immediately after the outlet frit of the column.
Column Preparation. All columns were 150 µm × 22 cm and were slurry packed4 with 1.5-µm C18 bonded nonporous silica particles (MICRA Scientific, Northbrook, IL). Slurries were made with 10 mg/mL particles in a 33% acetone/67% hexane solution. The capillaries had outlet frits made with 5-µm bare silica particles which had been packed ∼3 mm into the end of the capillary and then sintered with an electronic arcing device.5 Slurry packing was performed using a stirred stainless steel pressure reservoir at 5000 psi. The columns were slurry packed to within several centimeters of the inlet end and then dried before use. No frits were put on the inlet ends of the column. Though the open capillary at the inlet greatly distorts the injection profile and decreases the efficiency of an isocratic separation, the use of a solvent gradient greatly reduces the effects that the injection profile has on the separation. Since the separations used solvent gradients exclusively, the separations were not adversely affected by the open section of capillary preceding the packed bed. Gradient Generation. The pump generates a binary gradient by using low-pressure mixing as shown in Figure 1. There are two solenoid valves (BioChem Valve Corp., Boonton, NJ), proportioning the flow of the aqueous and nonaqueous solvents into the mixing chamber. The only wetted material in the valves is Teflon, which gives excellent solvent compatibility. These two valves are connected to a stirred low-pressure mixing chamber which has an internal volume of less than 200 µL. This volume (4) Kennedy, R. T.; Jorgenson, J. W. Anal. Chem. 1989, 61, 1128. (5) Hoyt, A. M.; Beale, S. C.; Larmann, J. P.; Jorgenson, J. W. J. Microcolumn Sep. 1993, 5, 325-330.
was calculated to be large enough to mask the effects of the finite cycle time of the proportioning valves and yet be small enough to allow fairly rapid changes in the composition of the solvent at the flow rates used. A computer program written in-house using LabVIEW (National Instruments Corp., Austin, TX) running on a Power Macintosh (Apple Computer, Cupertino, CA) controls the switching of the valves to generate the gradient. The computer control of the valves gives control over the gradient profiles. Injection. Standard loop or six-port injectors are not able to operate at pressures above 690 bar (10 000 psi), so injections were instead performed using a pressure reservoir technique.6 To inject a sample, the head of the capillary column was removed from the tee and inserted in an 0.5-mL microcentrifuge tube (Fisher Scientific, Pittsburgh, PA) containing the sample inside of a stainless steel pressure vessel. Gas pressure (28 bar) from a nitrogen tank was applied to the vessel for a calculated length of time to inject a plug of sample on to the column. The analytes were chromatographically focused onto the head of the stationaryphase bed since the column was equilibrated with aqueous buffer prior to injection and the sample was in an aqueous solution. To determine the amount of sample injected, the flow rate through the column was measured and this number was multiplied by the total injection time to obtain the actual volume of sample on the column. After the injection the column was reconnected to the tee for the remainder of the analysis. UV-Visible Detector. Prior to connecting to the mass spectrometer, several runs of protein digests were performed using a Linear UVIS 200 (Linear Instruments, Fremont, CA) UVvisible detector to ensure that the separation system was functioning properly. Detection was performed through the capillary immediately after the outlet frit to minimize postcolumn band broadening. The path length was the same at the internal diameter of the column which was 150 µm. The peptides were detected at 215 nm. Electrospray Interface. Two different interfaces were evaluated for use with VHPLC: a coaxial nanospray interface7 and a nebulized nanoflow electrospray interface supplied by Micromass (Manchester, U.K.). The coaxial nanospray interface uses a 1/16-in. tee (Valco) which has been drilled out to permit the 360-µm-o.d. fusedsilica column (Polymicro Technologies, Phoenix AZ) to pass through the tee and into the fused-silica tip. Fused-silica tips were drawn out from 180-µm-i.d., 360-µm--o.d. fused-silica capillaries (Polymicro Technologies) to give an tip inner diameter of 8-12 µm. These were sealed in the tee with a short length of 0.020in.-i.d., 1/16-in.-o.d. PEEK tubing (Upchurch Scientific, Oak Harbor, WA) using a stainless steel ferrule and nut. The coaxial makeup solution (50:50 methanol with 5% formic acid in water) was delivered to the tee by a 30-µm-i.d., 150-µm-o.d. fused-silica capillary 60 cm long. This column was sealed in the sidearm of the tee with a short length of Teflon tubing 0.010-in. i.d., 1/16-in. o.d. (Upchurch Scientific) using a PEEK ferrule and fingertight PEEK nut (Upchurch Scientific). (6) Moseley, M. A.; Deterding, L. J.; K. B., T.; Jorgenson, J. W. Anal. Chem. 1991, 63, 1467. (7) Moseley, M. A.; Lewis, K. C.; Opiteck, G. J.; Ramirez, S.; Jorgenson, J. W.; Anderegg, R. J. 45th ASMS Conference on Mass Spectrometry and Allid Topics, Palm Springs, CA, 1997.
To accurately deliver makeup solutions to the coaxial interface in a pulse-free manner at low nanoliter per minute flow rates, stainless steel pressure vessels were used. The nebulized nanoflow interface uses a low internal volume tee to connect the column to a 20 µm i.d. × 5 cm spray capillary. This capillary was biased with a voltage of 3.6 kV and had nitrogen as a sheath nebulizing gas at 20 psi. These settings resulted in a stable electrospray during the entire gradient at a column flow rate of ∼500 nL/min The spray capillary tip was positioned ∼1 cm away and orthogonal to the inlet orifice of the Z-Spray interface (Micromass) of the mass spectrometer.8 Mass Spectrometer. A Micromass Q-TOF hybrid quadrupole time-of-flight mass spectrometer was used. In MS mode, a spectrum of mass range of 300-2500 was obtained every 0.5 s. To gather MS data, the quadrupoles of the mass spectrometer are set to rf-only mode and the ion beam is analyzed by the orthogonal time-of-flight section in the mass spectrometer. For MS/MS analysis the data-dependent scanning mode of the instrument was used (automatic switching from MS mode to MS/ MS mode).9 In MS/MS mode, the first quadrupole is set to isolate the precursor ion selected and argon is introduced into the hexapole collision cell. After collisionally activated dissociation of the tryptic peptides, the peptide fragments are mass analyzed by the time-of-flight region of the mass spectrometer. Automated collision energy switching was used so that each precursor ion was fragmented at three different collision energies, and the resultant product ion spectra were summed. Enzymatic Protein Digests. The tryptic digest of ovalbumin was prepared by dissolving chicken egg ovalbumin (Sigma, St. Louis, MO) in 100 mM borate buffer (pH 8.5) at a concentration of 10 mg/mL. This sample was placed in boiling water for 5 min to denature the protein. After the solution had cooled, 0.33 mg/ mL of trypsin (Sigma) was added, giving a 30:1 weight ratio of ovalbumin to trypsin. This solution was incubated at 37 °C for 24 h after which it was filtered through a 0.2-µm filter and stored frozen to slow further reactions. The digest of bovine serum albumin (BSA) was prepared by standard methods. Briefly, the protein was denatured with a solution 8 M urea in 0.40 M ammonium bicarbonate. The cysteine residues were reduced with dithiothreitol (1:5 v/v with 45 mmol of dithiothreitol (Pierce, Rockford, IL)) and alkylated with 5 µL of 100 mmol of iodoacetamide (Pierce) RESULTS UV Analysis of Ovalbumin Digest. A tryptic digest of ovalbumin was analyzed by gradient UHPLC using a UV detector as described. Several different pressures and gradients were evaluated to obtain the highest peak capacity in an analysis lasting less than 30 min. Two picomoles of digest was injected onto the column to give adequate UV detector response. This amount of protein digest did not overload the column. Figure 2 is the UV chromatogram for an analysis at 930 bar (13 500 psi) with a 20-min linear gradient from 1 to 60% acetonitrile in water with 0.1% (by volume) TFA. This separation produced (8) Voyksner, R. D.; Lee, H. Anal. Chem. 1999, 71, 1441. (9) Bateman, R. H.; Carruthers, R.; Hoyes, J. B.; Gilbert, A. J.; Langridge, J. I.; Malone, K.; Bordoli, R. S. 46th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 1998.
Analytical Chemistry, Vol. 73, No. 13, July 1, 2001
Figure 2. UV chromatogram of a tryptic ovalbumin digest detected at 215 nm using a 22 cm × 150 µm column packed with 1.5-µm C18 nonporous silica. A 20-min linear gradient from 1 to 60% acetonitrile in water with 0.1% TFA was used.
peaks with an average base width of 10 s over a period of 30 min, giving a peak capacity of 187.10 MS and MS/MS of BSA Digest. A tryptic digest of BSA was analyzed to determine the limits of detection of the VHPLC/MS system. The digest samples were run with a mobile-phase gradient from 1 to 45% acetonitrile in water with 0.5% formic acid over a period of 30 min. Formic acid was used instead of trifluoroacetic acid due to the ion suppression effects of trifluoroacetic acid11 which decrease the sensitivity of the MS analysis. Figure 3 is a base peak mass chromatogram (BPI) of 12.5 fmol of BSA digest injected on column under the chromatographic conditions listed using the nebulized nanoflow interface. Figure 4 illustrates that the S/N ratio for the selected ion chromatogram of the BSA peptide T39 (+2) ion is greater than 25. In Figure 5 the percent protein coverage obtained by increasing the sample amount to 2 pmol is compared to the coverage with 12.5 fmol of sample injected. The percent protein coverage within the scanning range of the mass spectrometer decreases from 93.7% at 2 pmol to 83.8% at 12.5 fmol. The average base peak width for resolved peaks in this separation is ∼20 s, or twice the width of the peaks observed on a similar column with a UV detector. The increase in peak width is possibly due a combination of effects, including the substitution of formic acid for the TFA, and to peak broadening in the nebulized nanoflow interface. Figure 6 compares the VHPLC/MS/MS spectrum for peptide T39 to the spectrum obtained with nanospray-MS/MS. The VHPLC/MS/MS spectrum was acquired in 7.5× less time than required for the nanospray-MS/MS acquisition and yet the ion count was the same, even though the nanoelectrospray-MS/MS used 24 times as much sample.
MS/MS Analysis of a Protein from a 2D Gel Separation. Rat liver tissue lysate proteins were separated by 2D gel electrophoresis. Gel spots from multiple 2D gels were combined and subjected to an in-gel digest using the method of Mann and Wilm.12 After digestion, the solution of proteolytic peptides was divided into aliquots representative of one gel spot. This was done to facilitate direct comparisons of the data obtained by nanospray and VHPLC. The nanoelectrospray-MS/MS technique developed by Wilm and Mann,3 has proven to be an extraordinarily useful technique for proteomic studies. The very low detection limits (low femtomoles) and long analysis times (∼1 h/µL of sample) obtained via nanoelectrospray has proven to be useful for the identification of proteins. However, nanoelectrospray is a labor-intensive process. The high sample throughput requirements for large-scale proteomics will benefit from the automation of the process of acquiring MS/MS data from proteolytic peptides. The sample was preconcentrated and desalted after in-gel digestion using a 200-nL bed of POROS R2 (Perseptive Biosystems, Framingham, MA) and eluted in 2 µL of 70% methanol/(5% formic acid in water). The nanoelectrospray analysis generated signal for ∼60 min, giving a flow rate of ∼30 nL/min. During the nanoelectrospray analysis of an aliquot of the digest mixture representative of a single gel spot, 14 precursor ions were interrogated by MS/MS, 4 of which identified the protein as ATP synthase β chain. An additional four ions matched those expected for this protein by molecular weight (MS data only). The analysis by VHPLC/MS/MS used half as much sample as was used for the nanospray analysis (equivalent to half of one gel spot). The separation employed a 60-min linear gradient from
(10) Lan, K.; Jorgenson, J. W. Anal. Chem. 1999, 71, 709-714. (11) Apffel, A.; Fischer, S.; Kuhlmann, F. E. J. Chromatogr., A 1995, 712, 177.
(12) Wilm, M.; Shevchenko, A.; Houthaeve, T.; Breit, S.; Schweigerer, L.; Fotsis, T.; Mann, M. Nature 1996, 379, 466-469.
Analytical Chemistry, Vol. 73, No. 13, July 1, 2001
Figure 3. Base peak index chromatogram of 12.5 fmol of BSA digest, using a 22 cm × 150 µm column packed with 1.5-µm C18 nonporous silica. A 30 -min linear gradient from 1 to 45% acetonitrile in water with 0.5% formic acid was used.
Figure 4. Selected ion trace showing BSA peptide T39 from the data in Figure 3. Also included is a mass spectrum from peptide T39 illustrating the signal-to-noise ratio achieved.
1 to 60% acetonitrile in water with 0.5% formic acid added to both solvents. The nebulized nanoflow interface was used during this analysis. Under these conditions, 8 peptides from rat ATP synthase were identified out of the 26 peptides interrogated by automated datadependent scanning MS/MS, including the 4 that were identified by nanoelectrospray. Moreover, one of the major advantages of this technique is the absence of required user input during the data collection period. Data-dependent scanning and mode switching removes the need for additional input from the operator.
Collision energy switching helps to ensure that the peptide will be fragmented at an energy where useful information will be acquired as illustrated in Figure 7. VHPLC concentrates each peptide into a narrow band so that each peptide elutes in a 10-20-s-wide peak. This results in a ∼4fold higher peptide concentration being electrosprayed into the mass spectrometer compared to the concentration in nanoelectrospray analysis. Moreover, the individual nanoelectrospray-MS/ MS spectrum was acquired over 4.2 min even though the analyte was flowing into the source for ∼60 min, since nanoelectrospray Analytical Chemistry, Vol. 73, No. 13, July 1, 2001
Figure 5. LC/MS coverage of tryptic digest of BSA. Comparison of protein coverage for analyses performed with 2 pmol and 12.5 fmol of sample injected. The percent coverage for the two analyses increase to 93.7 and 83.8%, respectively, when the fragments that were out of the scanning range of the mass spectrometer are excluded from the calculations.
Figure 6. Comparison of signal-to-noise ratios and acquisition times for peptide T39 in nanoelectrospray-MS/MS and UHPLC/MS/MS.
is a constant-infusion technique. The duty cycle for acquisition of MS/MS data from this peptide was ∼7% (4.2 min/60 min), which means that 93% of the peptide flowed into the source while other peptides were being interrogated by MS/MS. For a completely separated peptide, the duty cycle for LC/MS/MS analysis should be over 75%. The combination of improvement in concentration sensitivity and improvement in duty cycle for the VHPLC/MS/ MS data acquisition provides the improvement in sensitivity observed in the VHPLC/MS/MS analysis over the nanoelectrospray technique. 2990
Analytical Chemistry, Vol. 73, No. 13, July 1, 2001
Currently, every aspect of the VHLPC/MS/MS analysis is automated except for the sample injection. Injection is a manual process since no autosampler valve is currently available that can withstand the high mobile-phase pressures used in this experiment. With the inclusion of an autosampler and an automatic injection method that can function under the high pressures used, the system could be used for totally automated protein analysis giving short analysis times and good sensitivity as is currently possible with nanoscale capillary LC with conventional mobilephase pressures.
Figure 7. MS/MS spectra at three different collision energies: 32, 28, and 23 V. The collision cell pressure is ∼5 × 10-5 Torr. A different fragmentation pattern is observed at each collision energy.
The high pressures involved in VHPLC do not require extra safety precautions due to the small amount of stored energy in the liquid mobile phase. The quantity of stored energy in the liquid is based on the compressibility of the mobile phase. Aqueous mobile phases were used in this series of experiments, which are very incompressible, resulting in very little stored energy in the system. CONCLUSIONS We have seen that the combination of a very high pressure LC system with a hybrid quadrupole/time-of-flight mass spectrometer makes a good system for the analysis of complex samples, as illustrated by the analysis of protein digests. By employing data-dependent scanning capabilities of the mass spectrometer, 26 different eluting peptides from a protein digest were interrogated for identification of the original protein. Compared to nanoelectrospray-MS/MS, this system demonstrated a higher duty cycle and better signal-to-noise ratio and operated in a fully automatic manner. (13) Hopfgartner, G.; Bean, K.; Henion, J.; Henry, R. J. Chromatogr. 1993, 647, 51-61.
Electrospray is a concentration-sensitive ionization technique,13 which is why the use of smaller inner diameter chromatography columns can give significant improvements in detection limits. The increase in concentration sensitivity is proportional to the inverse square of the column diameter; thus, a 75-µm-i.d. capillary column should offer a 4-fold improvement in sensitivity over the 150-µm columns used, assuming the same amount of sample is injected. A higher chromatographic efficiency results from the use of small stationary-phase particles, which translates into narrower chromatographic peaks and a higher peak capacity. This increased resolving power makes the resulting data easier to analyze since more of the peaks consist of a single compound. Future research will include using smaller diameter columns and smaller diameter packing material at high pressures, which should allow subfemtomole analyses of protein digests.
Received for review September 12, 2000. Accepted March 15, 2001. AC0010835
Analytical Chemistry, Vol. 73, No. 13, July 1, 2001