On-Line MALDI-TOF MS Using a Continuous Vacuum Deposition

Anal. Chem. 1998, 70, 5278-5287

On-Line MALDI-TOF MS Using a Continuous Vacuum Deposition Interface Jan Preisler, Frantisek Foret, and Barry L. Karger*

Barnett Institute and Department of Chemistry, Northeastern University, Boston, Massachusetts 02115

In this work, a new interface for continuous on-line MALDI-TOF MS is presented. The sample, mixed with a suitable matrix, was transported into the evacuated source chamber of the mass spectrometer at liquid flow rates of 100-400 nL/min. The liquid sample matrix was deposited on a rotating quartz wheel and transported to the repeller, where laser desorption took place. Rapid evaporation of the solvent (water or methanol) on the surface of the wheel resulted in formation of a thin, ∼50-µm-wide, sample trace. Scanning electron microscopic photographs of the vacuum-dried trace revealed the deposited material to consist of an amorphous film. Furthermore, sample uniformity along the trace, in conjunction with its narrow width, resulted in excellent signal reproducibility, with detection limits in the attomole range. The interface permitted the on-line coupling of microcolumn separation techniques with MALDI MS, as demonstrated in the capillary electrophoresis MALDI-TOF MS analysis of a 12peptide mixture. The approach offers the potential for rapid separation and trace analysis of complex mixtures. The application of mass spectrometry in the biological sciences often demands the ability to handle and analyze picomole and lower amounts of sample. The introduction of microionization techniques, such as micro- or nanoelectrospray1 and matrixassisted laser desorption/ionization (MALDI),2 has provided powerful approaches for the analysis of such small sample amounts. Since liquid samples are sprayed continuously at atmospheric pressure and the resulting ions are directly introduced into the mass spectrometer, electrospray ionization (ESI)3 is often used for on-line coupling with separation techniques such as liquid chromatography (LC).4 In contrast to ESI, MALDI has typically been operated as an off-line ionization technique. Here the analyte, mixed with a suitable matrix, is deposited on a target surface, followed by formation of dried mixed crystals and placement of the target in the source chamber of the mass spectrometer. MALDI MS analysis often requires finding the “sweet spot” on the sample target in order to obtain a reasonable signal.5,6 Although a motorized x-y stage may be incorporated for automated searching (1) Wilm, M.; Mann, M. Anal. Chem. 1996, 68, 1-8. (2) Karas, M.; Bachmann, D.; Bahr, U.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Processes 1987, 78, 53-68. (3) Fenn, J. B.; Rosell, J.; Meng, C. K. J. Am. Soc. Mass Spectrom. 1997, 8, 1147-1157. (4) Balogh, M. P. LC-GC 1998, 16, 135-144. (5) Perera, I. K.; Perkins, J.; Kantartzoglou, S. Rapid Commun. Mass Spectrom. 1995, 9, 180-187.

5278 Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

of the optimum position on the target, this procedure can be timeconsuming. The off-line approach has recently been improved using microscale targets for automated high-throughput MALDI analysis.7,8 In these designs, picoliter to nanoliter sample volumes are deposited into microfabricated wells with dimensions similar to the spot size of the desorption laser beam (∼100-µm diameter). Thus, the entire sample spot is irradiated and the search for the “sweet spot” eliminated. While the miniaturization of the sample target simplifies static MALDI analysis, the approach is not directly compatible with continuous analysis, including chromatographic or electrophoretic separations. There have been numerous attempts to achieve MALDI analysis of flowing liquid samples. In off-line arrangements, the sample components exiting a separation capillary at atmospheric pressure were spotted9-11 or continuously streaked12 on a target, followed by MALDI-TOF MS analysis. While these approaches are useful, on-line analysis would be more effective for ultratrace analysis, where fewer sample handling steps would reduce the chance for sample losses. Furthermore, speed and throughput should, in principle, be higher in the on-line approach. With respect to on-line MALDI-TOF, the liquid samples have been analyzed directly in the vacuum of the mass spectrometer using a variety of designs. For example, a nebulizer interface has been developed for continuous sample and matrix introduction, with MALDI being performed directly on rapidly dried droplets.13 In another design, a continuous probe, similar to a fast atom bombardment (FAB)14 interface, has been used for the analysis of a flowing sample stream with a liquid matrix, with ethylene glycol or glycerol added to prevent freezing of the sample.15,16 Other attempts at flowing liquid sample desorption have made use of fine dispersions of graphite particles17,18 and liquid (6) Strupat, K.; Karas, M.; Hillenkamp, F. Int. J. Mass Spectrom. Ion Processes 1991, 111, 89-102. (7) Jespersen, S.; Niessen, W. M. A.; Tjaden, U. R.; van der Greef J.; Litborn, E.; Lindberg, U.; Roeraade. J. Rapid Commun. Mass Spectrom. 1994, 8, 581-584. (8) Little, D. P.; Cornish, T. J.; O’Donnell, M. J.; Braun, A.; Cotter, R. J.; Ko ¨ster, H. Anal. Chem. 1997, 69, 4540-4546. (9) Stevenson, T. I.; Loo, J. A. LC-GC 1998, 16, 54-58. (10) Bergman, A. C.; Bergman, T. J. Protein Chem. 1997, 16, 421-423. (11) Walker, K. L.; Chiu, R. W.; Monnig, C. A.; Wilkins, C. L. Anal. Chem. 1995, 67, 4197-4204. (12) Zhang, H. Y.; Caprioli, R. M. J. Mass Spectrom. 1996, 31, 1039-1046. (13) Murray, K. K. Mass Spectrom. Rev. 1997, 16, 283-299. (14) Nichols, W.; Zweigenbaum, J.; Garcia, F.; Johansson, M.; Henion, J. LC-GC 1992, 10, 676-686. (15) Nagra, D.; Li, L. J. Chromatogr. 1995, 711, 235-245. (16) He, L.; Li, L.; Lubman, D. M. Anal. Chem. 1995, 67, 4127-4132. (17) Dale, M. J.; Knochenmuss, R.; Zenobi, R. Anal. Chem. 1996, 68, 33213329. 10.1021/ac9807823 CCC: $15.00

© 1998 American Chemical Society Published on Web 11/04/1998

Figure 1. Schematic of the on-line MALDI-TOF MS instrument (top view). Components are listed in the figure and described in detail in the Experimental Section.

matrixes.2,19-23 More recently, the exit of a capillary electrophoresis column was placed directly in the vacuum region of the TOF mass spectrometer.24 The sample ions, eluting in a solution of CuCl2, were desorbed by a laser irradiating the capillary end, and on-line spectra of low-molecular-weight peptides separated by capillary electrophoresis (CE) were demonstrated. In addition to MALDI, ESI attempts to introduce liquid samples directly into the evacuated source of a mass spectrometer have also been reported.25-27 Each of the above on-line approaches has met with limited success; however, there is currently no universal on-line MALDI interface for simple and sensitive analysis of minute sample amounts. The aim of the present research was to develop such an interface for continuous introduction of the liquid sample into the TOF mass spectrometer. The interface was designed to resemble the current MALDI process as closely as possible, to take advantage of the significant experience already accumulated (18) Sunner, J.; Dratz, E.; Chen, Y. C. Anal. Chem. 1995, 67, 4335-4342. (19) Overberg, A.; Karas, M.; Bahr, U.; Kaufmann, R.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1990, 4, 293-296. (20) Strobel, F. H.; Solouki, M. A.; White, M. A.; Russell, D. H. J. Am. Soc. Mass Spectrom. 1991, 2, 91-94. (21) Cornett, S.; Duncan, M. A.; Amster, I. J. Anal. Chem. 1993, 65, 26082613. (22) Williams, J. B.; Gusev, A. I.; Hercules, D. M. Macromolecules 1996, 29, 8144-8150. (23) Zo ¨ llner, P.; Schmid, E. R.; Allmaier, G. Rapid Commun. Mass Spectrom. 1996, 10, 1278-1282. (24) Chang, S. Y.; Yeung, E. S. Anal. Chem. 1997, 69, 2251-2257. (25) Sheehan, E. W.; Willoughby, R. C. In Proceedings of the 45th ASMS Conference on Mass Spectrometry and Allied Topics, Palm Springs, CA, June 1-5, 1997; p 115. (26) Dohmeier, D. M. Ph.D. Thesis, University of North Carolina, Chapel Hill, 1995. (27) Gamero Castano, M.; Aguirre DeCarcer, I.; deJuan, L.; delaMora, J. F. J. Appl. Phys. 1998, 83, 2428-2434.

in off-line approaches. In addition, a requirement in the design was the facile handling of very small (low microliter to submicroliter) sample volumes, in which sample losses would be minimized and low mass detection limits achieved. The interface consisted of the deposition of the sample liquid stream at flow rates of 100-400 nL/min onto a moving surface inside the vacuum region of the TOF instrument. The analyte stream, premixed with a suitable matrix, was deposited onto a rotating quartz wheel by means of liquid flow through a narrow fused silica capillary. The solvent rapidly evaporated, leaving a narrow trace deposited on the wheel. Rotation then transported the sample trace to a slit in the repeller plate, where the nitrogen desorption laser irradiated the trace. It is to be noted that this approach is different from that of a conventional moving belt interface,28 where the deposition occurs at atmospheric pressure. Among other factors, deposition in a vacuum reduces significantly the demands on pumping. The on-line MALDI interface, because of optimized deposition characteristics, was able to achieve attomole detection limits, even using a simple linear TOF instrument. In addition, high-performance separation in the CE-MALDITOF MS analysis of a 12-peptide mixture is presented. This interface appears to be generally applicable to on-line MALDITOF MS with low mass detection limits. EXPERIMENTAL SECTION Mass Spectrometer. A linear, Wiley-McLaren-type TOF mass spectrometer29 with a 1-m-long drift region was constructed, as shown in Figure 1. The 20-cm3 source chamber, sample load mechanism, acceleration optics, 10-cm-diameter flight tube, and (28) McFadden, W. H.; Schwartz, H. L.; Evans, S. J. Chromatogr. 1976, 122, 389-396. (29) Wiley, W. C.; McLaren, I. H. Rev. Sci. Instrum. 1955, 26, 1150-1157.

Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

5279

detector were purchased from R. M. Jordan Co. (Grass Valley, CA). The original sample load mechanism was used only to analyze conventional MALDI samples (prepared by the dried droplet method). The distances between the repeller plate and the first grid as well as between the first and the second grids were each 12.7 mm. Ion transmission for each of the two grids, which were grounded, was 90%. The voltage on the repeller plate (+15 kV) was controlled by a power supply (model CZE1000R/ X2263, Spellman, Hauppauge, NY). A 40-mm dual-microchannel plate (MCP) with an extended dynamic range (Galileo, Sturbridge, MA) served as ion detector. The ion transmission of the detector input grid was 82%, leading to a total ion transmission of the three grids of 66%. The instrument was evacuated by a diffusion pump (model VHS-6, Varian, Lexington, MA) with a maximum pumping speed of 2400 L/s. A refrigerated recirculator (model CFT-75Neslab, Portsmouth, NH) was used to cool the diffusion pump. The diffusion pump was backed by a two-stage mechanical pump (model Pascal 2015, Alcatel, Annecy, France) equipped with a molecular sieve trap (model KMST-150-2, MDC, Hayward, CA). Oil contamination of the mass spectrometer was prevented by a liquid nitrogen cryotrap (model 326-6, Varian) and an electropneumatic gate valve (model GV-8000V-ASA-P, MDC). The cryotrap also removed condensable vapors, such as water or methanol, that were infused to the chamber. A vacuum controller with two convectron gauges and two ion gauges (model 307, Granville-Phillips, Boulder, CO) was used. The convectrons were located in the foreline and in the source chamber, while the working ion gauge was in the detector region. The lowest pressure in the flight tube was 5 × 10-8 Torr, with typical pressures in the low 10-6 Torr range during solvent deposition. A laboratory-built TOF MS controller operated the diffusion pump, electropneumatic gate, and high-voltage power supplies. The controller protected the instrument and its components from damage due to an accidental pressure increase or a cooling malfunction. It also contained the voltage supply for the MCP detector. A 337-nm, 30-Hz nitrogen laser (model VSL-337ND-S, Laser Science, Franklin, MA) was used for MALDI. The laser beam was attenuated with a stepped neutral density filter (Edmund Scientific, Barrington, NJ) and focused with a quartz lens on the sample target. The angle of incidence of the desorption beam (defined by the beam and the flight axis) was 60°. If not otherwise stated, the laser was used at the repetition rate of 20 Hz. On-Line MALDI-TOF MS Interface. Initial experiments with vacuum sample deposition were conducted in a small cylindrical vacuum cell made of polycarbonate. The basic arrangement was similar to the actual interface described below. Solutions of methyl green were deposited on an acetal resin wheel (Delrin, DuPont, Wilmington, DE) propelled by a 3-V dc motor. The small cell did not contain high-voltage electrodes because it was designed simply for monitoring of the deposition process. The cell was evacuated by a mechanical pump (model DD 20, Precision Scientific, Chicago, IL). For the actual on-line MALDI-TOF MS instrument, a mixed solution of analyte and matrix was deposited via a fused silica capillary (Polymicro Technologies, Phoenix, AR), 20 µm i.d., 150 µm o.d., and 12.0 cm in length, on a quartz wheel (Optikos, 5280 Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

Figure 2. Detail of the design of liquid deposition process on the rotation wheel in the vacuum of the mass spectrometer (side view). Also shown is the design of the repeller region. See text for details.

Cambridge, MA), see Figures 1 and 2. The capillary was positioned in a probe made of a stainless steel tube, 9.51 mm o.d., 6.7 mm i.d., 7 cm length. A pipe adapter with a PEEK ferrule, 0.4 mm i.d., was attached to the outer (atmospheric) side of the tube, and a Delrin cap with a center hole, 0.25 mm i.d., covered the inner (vacuum) side of the tube. The probe was inserted into the source chamber via quick coupling, 9.51 mm i.d., in the center of the interface flange to a position such that the end of the capillary was slightly bent while touching the wheel. The outlet of the capillary was tapered, using fine sandpaper. The diameter of the quartz wheel was 5.0 cm and the thickness 1.0 cm; the perimeter surface of the wheel was unpolished. The wheel, which was perfectly balanced on a stainless steel shaft, was propelled by a geared stepper motor (model ABS, Hurst, Princeton, IN), with rotation speeds ranging from 0 to 12 rpm and 1800 steps for one full rotation. The original repeller plate (R. M. Jordan Co.) with a center hole for the sample probe was used only for initial analysis of conventional MALDI samples; modification of the repeller plate was required to incorporate the wheel. A rectangular hole of 12 × 30 mm was cut in the center of the repeller plate, and two flat pieces of stainless steel foil (25 × 25 × 0.05 mm) were glued to the repeller with an electrically conductive glue to form a slit (12 × 0.25 mm) in the center of the plate. It was important to adjust accurately the position of the wheel to gently touch the repeller at the slit position. Scanning Electron Microscopy (SEM). The morphology of dried MALDI samples was examined with a scanning electron microscope (model AMR 1000, Amray, Bedford, MA). The samples were sputtercoated with gold/palladium (60/40) in a sputtercoater (model Samsputter 2a, Tousimis, Rockville, MD). Under the coating protocol, the thickness of the metal coating was estimated to be 10-15 nm. Experiment Control and Data Acquisition. A digital delay generator (model 9650A, EG&G, Princeton, NJ) triggered the desorption laser as well as a laboratory-built digital divider. The output of the divider drove the stepper motor controller (model

EPC01, Hurst) for precise synchronization of the rotation of the wheel with the laser pulses. By setting the divider ratio, the number of laser pulses applied to each sample spot could be appropriately adjusted. The controller of the stepper motor was modified such that external pulses could be received to propel the motor and reset the controller counter. A 500-MHz, 1 G/s digital oscilloscope (model 9350AM, LeCroy, Chestnut Ridge, NY) allowed real-time measurement and/or transfer of mass spectra to the computer. A computer program, operating under DOS, transferred multiple files from the oscilloscope to a PC via a GPIB interface. Approximately 50 single-shot mass spectra (each spectrum consisting of 2000 data points of ion signal over the selected m/z range) could be transferred to the computer memory in 1 s. Capillary Electrophoresis. Capillary electrophoresis was performed using 75-µm-i.d. and 375-µm-o.d. fused silica capillaries (Polymicro Technologies) coated with poly(vinyl alcohol)30 to eliminate electroosmotic flow, with 10 mM solution of citric acid (electrophoresis grade, Schwarz/Mann Biotech, Cleveland, OH) as running buffer. Electrophoresis was driven at 500 V/cm by a high-voltage power supply (model PS/EH30, Glassman, Whitehouse Station, NJ). The sample was injected from unbuffered solution either by electromigration at 50 V/cm or by pressure at 250 Pa. For UV detection, the total length of the capillary was 15 cm and the effective length 10 cm. Absorbance at 220 nm from a CE detector (model Spectra 100, Spectra Physics) was recorded using Chrom Perfect (Justice Innovations, Mountain View, CA). For on-line MALDI-TOF MS, the separation capillary (10 cm length) was connected to a liquid junction31,32 made of polycarbonate, which contained 10 mM R-cyano-4-hydroxycinnamic acid matrix as cathodic buffer. PEEK liners (models FS1L.15 PK and FS1L.4PK, Valco Instruments Co., Houston, TX) were inserted into holes in the polycarbonate block to center precisely the separation and infusion capillaries, with a gap of approximately 100 µm. The sample was injected into the separation capillary, and the stepper motor was activated (0.33 rpm) within 5 s. Chemicals. Solutions of methyl green (Sigma Chemical Co., St. Louis, MO) in methanol and distilled water were initially used to explore deposition in a vacuum. R-Cyano-4-hydroxycinnamic acid (RCHCA), 2,5-dihydroxybenzoic acid, 4-hydroxy-3-methoxycinnamic (ferulic) acid (all from Sigma Chemical Co.), and 3-hydroxypicolinic acid (Aldrich Chemical, Milwaukee, WI) were used as matrixes for MALDI, each consisting of 0.1 M stock solutions in methanol. Angiotensins and angiotensinogens (see Table 1), heptapeptide EDPFLRF, and bovine insulin (BI) were purchased from Sigma Chemical Co. and made as 1 mg/mL stock solutions in water. A 1 mM stock solution of BI was prepared by dissolving the protein in 0.1% trifluoroacetic acid (J. T. Baker Inc., Phillipsburg, NJ). Methanol, ethanol, and acetonitrile (all HPLC grade) were purchased from Fisher Scientific (Fair Lawn, NJ). RESULTS AND DISCUSSION Preliminary Studies. To examine in detail the vacuum deposition process on the wheel, experiments were initially (30) Goetzinger, W.; Karger, B. L. PCT Int. Appl. W09623220, August 1996. (31) Lee, E. D.; Muck, W.; Henion, J. D.; Covey, T. R. Biomed. Environ. Mass Spectrom. 1989, 18, 844-850. (32) Foret, F.; Kirby, D.; Karger, B. L. Proceedings of the 44th ASMS Conference on Mass Spectrometry and Allied Topics, Portland, OR, May 12-16, 1996; Poster WPH 147.

Table 1. List of Angiotensins Used for CE-MALDI-TOF MS peptide angiotensin I

human bullfrog goosefish salmon des-Asp1 [Val5] angiotensin II human fragments 1-7 fragments 3-8 angiotensin III human angiotensinogen human porcine

structure

solute no.

MW

DRVYIHPFHL DRVYVHPFNL NRVYVHPFHL NRVYVHPFNL RVYIHPFHL DRVYVHPFHL DRVYIHPF DRVYIHP VYIHPF RVYIHPF DRVYIHPFHLVIHN DRVYIHPFHLLVYS

1 2 3 4 5 6 7 8 9 10 11 12

1296.5 1259.4 1281.5 1258.4 1181.4 1282.5 1046.2 899.0 774.9 931.1 1760.0 1759.0

performed in a small vacuum cell at the 10-100 mTorr range. The process could be observed visually, and any variations in conditions could be easily implemented. Depositions at pressures in this range were assumed to be similar to those under high vacuum. Solutions of 10 mM methyl green in methanol, 50% (v/v) methanol, or water were deposited on a plastic wheel rotating at a continuous rate of ∼1 rpm. It was found that solvent freezing at the capillary exit did not occur if the capillary outlet was in contact with the wheel. The capillary was bent to ensure good contact between the tip and the wheel surface inside the chamber, see Figure 2. In addition, the end of the capillary was tapered to prevent accumulation of the deposited solution on the outer capillary wall, with subsequent clogging of the capillary. Teflon was first examined as the material of the wheel; however, this polymer was not effective, as the deposited dye trace formed a series of stains rather than a continuous streak. On the other hand, when Delrin (an acetal resin) was employed, a regular trace of the dye was found. This result was likely due to the lower hydrophobicity of Delrin relative to Teflon, and perhaps to the somewhat rougher surface of Delrin. Importantly, no spraying of the dye inside the chamber was observed, indicating that virtually all of the sample adhered to the wheel. To determine the liquid flow rate in the capillary, water was infused on the Delrin wheel rotating continuously at ∼1 rpm. A 6-cm portion of the outer protective polyimide coating of the infusion capillary (20 µm i.d., 150 µm o.d., 12 cm length) was removed from the column in the region outside the vacuum chamber. A short plug of 10 mM aqueous methyl green solution was injected into the water stream, and the time required for the color zone to travel over 5.0-cm distance in the capillary was measured. The velocity of water in the infusion capillary was determined to be 10.6 ( 0.7 mm/s, and the corresponding flow rate was 200 ( 20 nL/min. The flow rate was also calculated from the Poiseuille equation, F ) π∆pr4/(8ηl), where the pressure difference, ∆p ) 101 kPa, capillary radius, r ) 10 µm, viscosity at 25 °C, η ) 0.89 mPa‚s, and capillary length, l ) 0.12 m. The calculated flow rate of 220 nL/min was in good agreement with the measured value. The correspondence of the measured and calculated values also implied that water evaporation cooled only the very end of the infusion capillary and that the water in the capillary was close to room temperature. Sample Morphology. The preparation of the MALDI sample (mixture of analyte and matrix) is known to play an essential role Analytical Chemistry, Vol. 70, No. 24, December 15, 1998

5281

in achieving high performance in MALDI-MS analysis. As a result, a number of different sample preparation methods have been introduced to improve sample homogeneity and hence reproducibility of the MALDI analysis. In addition to the common dried droplet method,33,34 procedures include slow growing of large crystals,35 preparation of a microcrystalline matrix substrate by crushing matrix crystals36 or by fast evaporation,37 crystallization under a stream of nitrogen38 or under vacuum,39 as well as other methods.5,40-43 Moreover, the recent work in deposition of nanoor picoliter volumes of samples7,8,44 resulted in fast evaporation of solvent, yielding small crystals of matrix. Many of these methods attempt to overcome the known discrimination effects of the dried droplet method, in which signal is dependent on spot position.45-47 It was of interest to compare the current approach of vacuum deposition on the rotating wheel with conventional procedures. We first examined by SEM analysis the morphology of a MALDI sample prepared by the dried droplet method. One microliter of a mixed solution of 1 µM angiotensin III and 10 mM RCHCA in 50% (v/v) methanol was deposited on an aluminum sample holder and dried at room temperature under atmospheric pressure. SEM photographs (Figure 3A) revealed that this typical preparation procedure of a MALDI sample yielded individual crystals of 2-5 µm in size. Similar sizes and shapes of RCHCA crystals were reported by others.45 The results in Figure 3A were then compared with those of the on-line approach adopted in this work. The preparation of a MALDI sample on the wheel for SEM analysis required the use of a short piece of self-adhesive copper tape placed on the Delrin. The identical solution of angiotensin III and RCHCA was deposited on the modified wheel rotating at a speed of 1.0 rpm in the evacuated cell. After deposition and release of the vacuum, the copper tape was removed and placed on the SEM sample holder. Figure 3B presents the SEM photographs of the MALDI sample deposited on the wheel. A smooth trace on the copper surface with a width of only 40 µm was observed. With such small dimensions, the desorption laser could irradiate a full segment of the deposited sample trace, since the laser spot size could be made wider than the trace. Subtle grooves, perpendicular to the trace, due to the copper surface, are also seen in the SEM photograph. (33) Karas, M.; Hillenkamp, F. Anal. Chem. 1988, 60, 2299-2301. (34) Beavis, R. C.; Chait, B. T. Rapid Commun. Mass Spectrom. 1989, 3, 432435. (35) Xiang, F.; Beavis, R. C. Org. Mass Spectrom. 1993, 28, 1424-1429. (36) Xiang, F.; Beavis, R. C. Rapid Commun. Mass Spectrom. 1994, 8, 199-204. (37) Vorm, O.; Roepstorff, P.; Mann, M. Anal. Chem. 1994, 66, 3218-3287. (38) Chan, T.-W. D.; Colburn, A. W.; Derrick, P.; Gardiner, D. J.; Bowden, M. Org. Mass Spectrom. 1992, 27, 188-194. (39) Weinberger, S. R.; Boernsen, K. O.; Finchy, J. W.; Robertson, V.; Musselman, B. D. In Proceedings of the 41st ASMS Conference on Mass Spectrometry and Allied Topics, San Francisco, CA, May 31-June 4, 1993; p 775a-b. (40) Hutchens, T. W.; Yip, T. T. Rapid Commun. Mass Spectrom. 1993, 7, 576580. (41) Mock, K.; Sutton, C. W.; Cottrell, J. S. Rapid Commun. Mass Spectrom. 1992, 6, 233-238. (42) Axelsson, J.; Hoberg, A.-M.; Waterson, C.; Myatt, P.; Shield, G. L.; Varney, J.; Haddleton, D. M.; Derrick, P. J. Rapid Commun. Mass Spectrom. 1997, 11, 209-213. (43) Yao, J.; Dey, M.; Pastor, S. J.; Wilkins, C. L. Anal. Chem. 1995, 67, 36383642. (44) Allmaier, G. Rapid Commun. Mass Spectrom. 1997, 11, 1576-1569. (45) Amado, F. M. L.; Dominiques, P.; Santana-Marques, M. G.; Ferrer-Correia, A. J.; Tomer, K. B. Rapid Commun. Mass Spectrom. 1997, 11, 1347-1352. (46) Dai, Y.; Whittal, M.; Li, L. Anal. Chem. 1996, 68, 2494-2500. (47) Cohen, S. L.; Chait, B. T. Anal. Chem. 1996, 68, 31-37.


Figure 3. Scanning electron micrographs of deposited MALDI samples1 µM angiotensin III and 10 mM RCHCA in 50% (v/v) methanol. Preparation of MALDI sample: (A) dried droplet method, (B) and (C) trace of sample deposited at low pressure. SEM characteristics: acceleration voltage, 10 kV; sample tilt, 0° (A) and 60° (B, C). See text for deposition details.

In addition, one groove in the middle of the trace, resulting from the capillary tip scratching the copper strip, is observed. Assuming a uniform profile of the 40-µm wide trace, a flow rate of 300 nL/min, and a matrix density of 1.2 g/cm3, the thickness of the sample film was calculated to be ∼70 nm. In fact, however, the sample tended to accumulate at the edges, as the solvent evaporated, with the dimensions of the mound being 1-2 µm in width and several hundred nanometers in height, see Figure 3C.

Importantly, however, the sample was distributed regularly along the trace; i.e., the sample properties did not vary as the wheel rotated. The sizes of the small structural features in the film in Figure 3C were estimated to be e40 nm, i.e., 100 times smaller than the crystals of a conventional MALDI sample (Figure 3A). This amorphous or microcrystalline sample trace was undoubtedly the result of the rapid evaporation of the small volume of solvent per unit time flowing into the vacuum. Such uniform sample morphology would appear to be ideal for MALDI analysis. The next step was to implement this sample preparation method into the on-line MALDI-TOF mass spectrometer. On-Line MALDI-TOF MS Characteristics. As described in the Experimental Section, we constructed a linear time-of-flight (TOF) mass spectrometer, incorporating the rotating quartz wheel in the source region, see Figure 1. The sample was transported via a 20-µm-i.d. capillary with a tapered tip to the wheel, where deposition took place. The process of infusion and deposition could be conveniently monitored via pressure measurement with an ion gauge. However, while the gauge was calibrated against air, the measurement was dependent on the ionization efficiency of the gas in the system. Because the relative ionization efficiencies of methanol (1.85) and water (1.12) are higher than that of air (1.00), readout of the gauge included a positive bias during infusion of these solvents. In addition, since the composition of the background gas in the chamber varied during an experiment, the specific ionization efficiency of the gaseous mixture at a given time was not known; therefore, the ion gauge signal in volts was used as an approximate measure of pressure in the vacuum. An increase of ion gauge voltage of 1 V, for example, corresponded to a 10fold increase of pressure at constant composition of background gas. In the case where the actual composition of the background gas was known, pressure values were used. Using a tapered fused silica capillary (20 µm i.d., 150 µm o.d., 12 cm length), see Figure 2, infusion and deposition of 50% (v/v) methanol on the quartz wheel (0.5 rpm) was examined and the pressure change monitored, as shown in Figure 4A. Initially, only air flowed through the capillary, and equilibrium between the infused air and its removal by the diffusion pump yielded a pressure of ∼3 × 10-7 Torr. At the time marked with the asterisk in Figure 4A, the capillary inlet was placed in a microvial containing the 50% (v/v) methanol. Within 1 s, the pressure decreased to below 2 × 10-7 Torr as the air from the capillary was sucked into the chamber. Next, while the solvent filled the capillary, neither air nor solvent eluted from the capillary outlet, and the pressure thus remained low. When the solvent reached the capillary outlet, the liquid was deposited on the wheel as a thin film. Evaporation of the solution from the film and tip was extremely fast, as evidenced by the sharp increase in the ion gauge signal in Figure 4A. The rate of infused solvent was quickly equal to the rate of solvent that evaporated and/or sublimated. Steady state for solvent evaporation, adsorption on the walls of the mass spectrometer, and solvent removal by pumping was reached within several seconds, resulting in a pressure plateau (∼2 × 10-6 Torr). The microvial with 50% (v/v) methanol was next removed from the capillary inlet at the time marked with the plus sign in Figure 4A. Air flowed into the capillary, and the flow rate of solvent

Figure 4. Ion gauge signal (volts, related to pressure) during deposition of 50% (v/v) methanol on quartz wheel rotating at 1 rpm (A) and vacuum infusion of methanol (B), 50% (v/v) methanol (C), and 10% (v/v) methanol (D). Start of sample infusion is indicated by the asterisk (/) and end of run by the plus sign (+).

increased as the length of the solvent plug in the capillary decreased. This enhanced liquid flow resulted in a pressure spike (10 rpm). The entire cleaning procedure including evacuation took less than 5 min. In future designs, an on-line cleaning procedure will be implemented, e.g., using laser ablation on the downstream surface region of the wheel. The wheel, made of nonconductive material (quartz), insulated the electrodes inside the mass spectrometer from the outside system, e.g., a CE column connected via a liquid junction. Furthermore, in the present design, decoupling the deposition region from the desorption laser region, as a result of mechanical (48) Beavis, R. C.; Chait, B. T. Rapid Commun. Mass Spectrom. 1989, 3, 233237.


Figure 5. 100-shot average MALDI-MS spectra of bovine insulin with RCHCA as matrix. Sample preparation: dried droplet (A) and on-line vacuum deposition (B). See text for details.

transport of the sample, was advantageous. The solvent, which evaporated in the deposition region, did not cause any local overpressure in the desorption laser region, which could lead to electrical discharge in the deposition region. It should be noted, however, that a wheel made of a conductive material was also briefly tested. It was found that arcing in the source region did not occur during the deposition of a matrix solution on the wheel held at 15 kV. While either a conductive or nonconductive wheel can thus be used, future research will elucidate which type of material is more advantageous. Finally, it should be noted that a high-speed diffusion pump has been utilized in this instrument design; however, separation of the deposition and desorption regions would, in principle, permit use of differential pumping, in which the deposition would be maintained under rough vacuum. Alternatively, the deposition area could be surrounded by a small refrigerated trap to remove solvent molecules. On-line MALDI Performance. (a) Comparison to Conventional MALDI. Conventional off-line MALDI-MS analysis was first carried out using a commercial sample load probe and repeller (R. M. Jordan Co.). Figure 5A presents the result obtained from the average of 100 spectra produced from a dried droplet sample of a 10-µL volume of 100 µM bovine insulin and 100 mM RCHCA solution in a mixed solvent (acetonitrile/ethanol/water ) 36:60:4 (v/v/v)) deposited on a stainless steel probe tip, 6.3 mm in diameter. Figure 5B shows the result obtained from the on-line interface for 1 µM bovine insulin with 10 mM RCHCA aqueous solution deposited for 30 s on the quartz wheel rotating at 0.33 rpm with two shots applied to each segment. A comparison of parts A and B of Figure 5 reveals that MALDI spectra prepared conventionally and by vacuum deposition are similar. However, importantly, the on-line sample vacuum deposition provided a higher insulin signal in the mass spectrum, even though the concentration and volume of the deposited sample were 2 orders of magnitude lower than those for the off-line sample preparation. Note that the low resolution shown in Figure 5A and B can be

Figure 6. Average ion signal of angiotensin II, fragments 1-7 (liquid deposited on wheel per rotation), vs segment number. Solution of 1 µM peptide with 10 mM RCHCA in 50% (v/v) methanol deposited on the quartz wheel at 0.33 rpm.

significantly improved by incorporation of time-lag focusing29 and use of a reflectron.49 (b) Regularity of Sample Trace. A necessary prerequisite for good reproducibility of MALDI spectra and on-line coupling of separation methods is the regularity of the sample trace over the surface of the wheel. To determine this, a mixed 50% (v/v) methanol solution of 1 µM angiotensin II, fragments 1-7, and 10 mM RCHCA matrix was deposited on the quartz wheel rotating at 0.33 rpm. The deposition was halted after 1 min, and then, over a period of 50 s, 1000 single-shot spectra were collected at a laser repetition rate of 20 Hz and a wheel rotation speed of 0.066 rpm, i.e., 10 shots/segment. The average of 10 single-shot spectra from each of 100 segments was calculated, and peak areas corresponding to the analyte ion (m/z ) 900) in the averaged spectra were plotted vs segment number, as shown in Figure 6. Although the relative standard deviation in the signal was (18%, the ion signal of the peptide was relatively constant in comparison to the signal obtained from a conventional MALDI sample.42 In addition, and importantly, the peptide peak was present in every single-shot spectrum, whereas in conventional MALDI it is well known that some shots do not yield spectra. The regularity in Figure 6 is a consequence of the good alignment of the laser beam with the sample trace, which itself is quite uniform. Note that absolute ion signals have been presented in Figure 6 in order to provide a clear picture of the regularity of the signal. Of course, the fluctuation in the signal can be improved by using an internal standard. (c) Type and Concentration of Matrix. The present on-line interface can be used with matrixes typically employed in MALDI. Conventional matrixes, such as R-cyano-4-hydroxycinnamic acid, 2,5-dihydroxybenzoic acid, 4-hydroxy-3-methoxycinnamic (ferulic) acid, and 3-hydroxypicolinic acid were successfully tested with peptide samples (results not shown). It is thus expected that existing off-line MALDI experience should be transferable to the present interface. (49) Mamyrin, B. A.; Karataev, V. I.; Shmikk, D. V.; Zagulin, V. A. Sov. Phys.JETP 1973, 37, 45-48.

Figure 7. Single-shot MALDI mass spectra of a methanolic solution of 1 µM heptapeptide (EDPFLRF) with (A) 10 or (B) 1 mM RCHCA matrix deposited on the quartz wheel at 0.33 rpm.

We next examined the effect of the concentration of the matrix, known to be an important parameter in MALDI,50 on analyte signal intensity. A methanolic solution of 1 µM heptapeptide EDPFLRF with 1, 10, or 100 mM RCHCA was deposited on the quartz wheel rotating at 0.33 rpm. It was first found that 100 mM RCHCA caused clogging of the capillary. On the other hand, as shown in Figure 7A, a relatively intense analyte peak could be observed for a single-shot MALDI mass spectrum with a matrix concentration of 10 mM. At 1 mM RCHCA concentration, Figure 7B, only low-intensity analyte peaks were observed. Thus, intermediate concentrations of matrix of roughly 10 mM seem best for the present on-line interface. It can be further seen in Figure 7A and B that peaks of adduct ions, such as [M + Na]+, [M + K]+ and [M + 2Na]+, were present in the spectra; indeed, they were the major signal for the analyte ion at the lower matrix concentration. Since most of the alkali metal ions originated from the peptide sample, on-line desalting will be essential to obtain good quality MALDI mass spectra, especially with such thin sample films, as produced by this on-line interface.51 (d) Sample Consumption and Signal Level. To determine how many laser shots will produce mass spectra of a given sample segment, a mixed solution of 1 µM angiotensin III with 10 mM RCHCA matrix in methanol was deposited on the quartz wheel rotating at 0.33 rpm. Infusion was interrupted after ∼30 s, and the wheel was rotated to transport the sample trace to the desorption region. The power density was adjusted to ∼20% above the threshold, and 100 shots were applied to a segment. The ion signal of the analyte (m/z ) 932) was found to gradually decrease, as the sample was removed, to 10% of the maximum signal intensity after 20 shots, with virtually no signal after 40 laser shots. Of course, the optimum number of laser shots per segment will (50) Hillenkamp, F.; Karas, M.; Beavis, R.; Chait, B. T. Anal. Chem. 1991, 63, 1193A-1203A. (51) Zhang, H.; Andren, P. E.; Capriolli, R. M. J. Mass Spectrom. 1995, 30, 17681771.


5285

Figure 8. CE-UV electropherogram of angiotensin mixture. See Table 1 for identification of solutes; ∼400 pg of each peptide injected, 10 mM citric acid solution used as buffer; see details in text.

depend on many factors, such as the analyte and matrix concentration, the laser power density, and the rotation speed of the wheel. While a detailed study on mass detection limits will be presented later, it was decided to conduct a preliminary experiment to examine the levels possible. A mixed solution of 0.1 µM of the heptapeptide EDPFLRF with 10 mM RCHCA matrix in methanol was deposited on the quartz wheel rotating at 0.33 rpm. It was found that the peptide ion peak yielded a S/N of ∼13 from a single desorption spectrum from one segment. Based on the solution flow rate on the wheel, the amount of peptide in each segment was estimated at approximately 50 amol. Spectral averaging and incorporation of time lag focusing should reduce the detection limit to at least the low attomole level. Such low detection levels obviously require careful manipulation of the small volumes of samples with this on-line MALDI-TOF interface. CE-MALDI MS. The on-line vacuum deposition of mixed solutions of sample and matrix offers a facile approach to coupling separations directly to MALDI MS. In this work, we demonstrate the on-line coupling capability using CE-MALDI MS to separate and analyze the mixture of the 12 angiotensins listed in Table 1. On-line µLC-MALDI MS could be easily implemented, as well. A liquid junction was employed at the exit of the capillary column to connect the CE column to the infusion capillary of the MALDI MS interface. The liquid junction reservoir contained the MALDI matrix that was mixed with the separated analytes in the infusion capillary for deposition on the wheel. Initially, conventional CE-UV (absorbance detection at 220 nm) was examined for the separation of the peptide mixture. Citric acid solution (10 mM) was selected as running buffer because the pKa and ion mobility of this acid were similar to those of RCHCA.52 A 10 mM RCHCA solution in 50% (v/v) methanol was placed in the reservoir of the liquid junction for CE-MALDI MS. Approximately 400 pg (∼8.3 µg/mL) of each peptide was then electrokinetically injected at 50 V/cm for 5 s. Figure 8 shows the resulting rapid separation in which 9 of the 12 peaks were observed. The distorted peak shape was caused by the relatively large amount of sample injected in order to achieve a signal, since impurities in the citric acid solution caused a large baseline drift. (52) Pospichal, J.; Gebauer, P.; Bocek, P. Chem. Rev. 1989, 89, 424-430.


Figure 9. On-line CE-MALDI MS of the angiotensin mixture in Table 1. Single-shot mass spectrum at normalized migration time ) 62.9 s is shown in the inset. See details in text.

For on-line CE-MALDI MS detection, the capillary and the anodic reservoir were filled with 10 mM citric acid solution, and the cathodic reservoir (at the liquid junction) contained 10 mM RCHCA solution in 50% (v/v) methanol. The identical angiotensin mixture was again electrokinetically injected at 50 V/cm for 5 s. The stepper motor was set at 0.33 rpm, and the laser repetition rate was adjusted to 10 Hz, i.e., 1 shot/segment. Figure 9 presents the results of this on-line CE MS experiment in terms of surface and contour graphs. As can be seen, 11 of the 12 peptides were resolved and identified in the MS electropherogram. Angiotensins 1 and 6 overlapped because of similar migration times, and, since their molecular masses differed by only 14 Da, no peak separation could be seen in the contour graph. However, as the inset at migration time ) 62.9 s demonstrates, the two peptides could be easily identified in the MS spectra. The countour graph also shows satellite peaks at higher m/z values for the largest peaks. These satellite peaks are alkali metal ion adducts, which are nevertheless significantly less than in the infusion experiment, see Figure 7. The CE separation provided a desalting step (as would also be true for µLC); the remaining salt is likely from the matrix. In future work, the MALDI matrix will also be desalted. In Figure 9, the average number of segments across a CE peak was 50, based on the electrophoretic velocity of the analytes and

the wheel rotation rate. For the current software and hardware system, each mass spectrum was an average of two single-shot spectra from two consecutive segments, i.e., on average, 25 mass spectra across the peaks. Figure 9 displays every fourth spectrum in this sequence. Because of the uniformity of the deposition and the large amount of sample injected, the above measurement sequence is seen to be an effective approach. For samples that contain significantly lower quantities of material, the rotation of the wheel can be slowed by a factor of 2-3. Moreover, the laser can be operated at a higher repetition rate to enhance S/N. Since the timing of the experiment was done manually, the migration times in CE-MALDI MS were corrected to peak 5 in the CE-UV (migration time of 52.5 s). It is also to be noted that the migration time omitted the constant time interval for sample transport from the liquid junction to the desorption region (approximately 100 s). It was further found that the normalized half-widths of the CE-MALDI MS peaks were lower than those found by UV. CONCLUSIONS This paper has presented a new approach to on-line coupling of liquid flow to MALDI-TOF MS. The method is based on vacuum deposition of the liquid on a rotating wheel. After evaporation and/or sublimation of the solvent, the sample is transported on the wheel to the repeller, where laser desorption takes place. Since the liquid flow rate is in the range of 100-400 nL/min and solvent removal is rapid in the vacuum, the sample trace, as revealed by SEM photographs, is shown to be amorphous and regular in the direction of rotation. Furthermore, each

segment, i.e., the surface film for each stepper motor position on the wheel, is smaller than the laser spot size, leading to relatively uniform signal intensity for each shot, as determined by peptide sample infusion. Detection limits in the attomole level are achievable. The interface is demonstrated to be effective in the on-line coupling of capillary electrophoresis to MALDI-TOF MS. Improvements in the current design will include implementation of time lag focusing for high-resolution mass spectra. In addition, the interface will be upgraded for uninterrupted operation. Work is also underway to couple the MALDI-TOF MS to capillary liquid chromatography. For all capillary separation methods, high-speed separation MALDI-TOF analysis will be possible with this interface design. Finally, the present approach offers the possibility of true multiplex separation/MS analysis, i.e., multiple capillary array/MS. These and other improvements will be subsequently reported. ACKNOWLEDGMENT The authors acknowledge NIH for support of this work (GM15847). The authors also thank Dr. William Fowle, Northeastern University, for his operation of the scanning electron microscope. This article is contribution no. 739 from the Barnett Institute.

Received for review July 15, 1998. Accepted September 29, 1998. AC9807823


5287

On-Line MALDI-TOF MS Using a Continuous Vacuum Deposition

Recommend Documents