An Integrated Microfluidics-Tandem Mass Spectrometry System for

Alberta Microelectronic Centre, #318, 11315-87 Avenue, Edmonton, AB, Canada, T6G 2T9 ... ibility of the system with the automated analysis of proteins...
52 downloads 0 Views 116KB Size
Anal. Chem. 1998, 70, 3728-3734

Articles

An Integrated Microfluidics-Tandem Mass Spectrometry System for Automated Protein Analysis Daniel Figeys,*,† Steven P. Gygi,‡ Graham McKinnon,§ and Ruedi Aebersold‡

Institute for Marine Biosciences, National Research Council Canada, 1411 Oxford Street, Halifax, Nova Scotia, B3H 3Z1, Department of Molecular Biotechnology, University of Washington, Box 357730, Seattle Washington 98195-7730, and Alberta Microelectronic Centre, #318, 11315-87 Avenue, Edmonton, AB, Canada, T6G 2T9

We describe an integrated analytical system consisting of a microfluidics device micromachined using photolithography/etching technology, a panel of computer-controlled high-voltage relays, and an electrospray ionization tandem mass spectrometer. Movement of solvents and samples on the device and off the device to the mass spectrometer was achieved by directed electroosmotic pumping induced by the activation of a suitable constellation of high-voltage relays. The system was used for the sequential automated analysis of protein digests. We demonstrate low femtomole per microliter sensitivity of detection and compatibility of the system with the automated analysis of proteins separated by two-dimensional gel electrophoresis. The shift from the analysis of single genes and proteins to the comprehensive analysis of biological systems and pathways has been one of the most dramatic recent developments in biological research. It is a direct consequence of the development of automated, high-throughput genomic technologies which are now used for whole-genome sequencing (Haemophilus influenzae Rd1 and Saccharomyces cerevisiae2 as examples) and for the establishment of comprehensive mRNA expression maps.3-5 Currently, no equivalent technology is available for the analysis of biological systems on the protein level. However, proteins are the most significant class of biological control and effector molecules, and

a complete model of a biological process cannot be established without knowledge of the identity, function, and state of activity of the proteins involved. Protein analytical technology has been revolutionized by the development of powerful mass spectrometric (MS) techniques and the development of computer algorithms for the identification of proteins by correlating protein and peptide mass spectral data with sequence databases.6-10 Current technology has reached a level of sensitivity that permits the identification of essentially any protein which is detectable by conventional protein staining.11-13 However, protein sample throughput is orders of magnitude lower than the throughput of DNA-based technologies, and some of the most sensitive protein analysis techniques developed to date11-15 are difficult to automate. Here we describe a novel approach to automated highsensitivity, high-throughput protein analysis. It is based on the integration of electrospray ionization tandem mass spectrometry (ESI-MS/MS) with a computer-controlled microfluidics device into a single system. The microfluidics devices are modules of specific function and design which are micromachined by photolithography/ etching of glass. Samples, typically tryptic digests of gel-separated proteins were applied to specific reservoirs and directed by computer-controlled electroosmotic pumping through a network of etched channels to the MS detector where selected peptide

* Corresponding author, (phone) 902-426-0558; (fax) 902-426-9413; (e-mail) [email protected]. † National Research Council Canada. ‡ University of Washington. § Alberta Microelectronic Centre. (1) Fleischmann, R.; Adams, M.; White, O.; Clayton, R.; Kirkness, E.; Kerlavage, A.; Bult, C.; Tomb, J.; Dougherty, B.; Merrick, J.; et al. Science 1995, 269, 496-512. (2) Bussey, H.; Storms, R. K.; Ahmed, A.; Albermann, K.; Allen, E.; Ansorge, W.; Araujo, R.; Aparicio, A.; Barrell, B.; Badcock, K.; et al. Nature 1997, 387, 103-105. (3) Velculescu, V. E.; Zhang, L.; Zhou, W.; Vogelstein, J.; Basrai, M. A.; Bassett, D. E.; Hieter, P.; Vogelstein, B.; Kinzler, K. W. Cell 1997, 88, 243-251. (4) Shalon, D.; Smith, S. J.; Brown, P. O. Genome Res. 1996, 6, 639-645. (5) Pietu, G.; Alibert, O.; Guichard, V.; Lamy, B.; Bois, F.; Leroy, E.; Mariage Samson, R.; Houlgatte, R.; Soularue, P.; Auffray, C. Genome Res. 1996, 6, 492-503.

(6) Eng, J.; McCormack, A. L.; Yates, J. R., III J. Am. Soc. Mass Spectrom. 1994, 5, 976-989. (7) Yates, J. R., III; Eng, J. K.; McCormack, A. L.; Schieltz, D. Anal. Chem. 1995, 67, 1426-1436. (8) Yates, J. R. In High Resolution Separation And Analysis Of Biological Macromolecules, Part B; Karger, B. L., Hancock, W. S., Eds.; Academic Press: San Diego, 1996; Vol. 271, pp 371-377. (9) Clauser, K. R.; Hall, S. C.; Smith, D. M.; Webb, J. W.; Andrews, L. E.; Tran, H. M.; Epstein, L. B.; Burlingame, A. L. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 5072-5076. (10) Mann, M.; Wilm, M. Anal. Chem. 1994, 66, 4390-4399. (11) Figeys, D.; Ducret, A.; Yates, J. R., III; Aebersold, R. Nature Biotechnol. 1996, 14, 1579-1583. (12) Figeys, D.; Aebersold, R. Electrophoresis 1997, 18, 360-368. (13) Wilm, M.; Shevchenko, A.; Houthaeve, T.; Breit, S.; Schweigerer, L.; Fotsis, T.; Mann, M. Nature 1996, 379, 466-469. (14) Tomlinson, A. J.; Benson, L. M.; Guzman, N. A.; Naylor, S. J. Chromatogr., A 1996, 744, 1-2. (15) Wilm, M.; Mann, M. Anal. Chem. 1996, 68, 1-8.

3728 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

S0003-2700(98)00320-5 CCC: $15.00

© 1998 American Chemical Society Published on Web 08/13/1998

Figure 1. Diagram of integrated analytical system. A nine-position microfabricated device was coupled to an ITMS instrument via a transfer capillary and a micro-ESI ion source. The inner surface of the transfer capillary (15 cm long, 75-µm i.d., 150- µm o.d.) was derivatized with (3-aminopropyl)silane. The etched channels were 30 µm deep and 72-73 µm wide. The diameter of the reservoirs was 1 mm. Sample flow was controlled by an array of computer-controlled high-voltage relays which are also schematically represented.

ions were subjected to collision-induced dissociation (CID). The resulting fragment spectra unambiguously identified the protein from which the peptides originated. Both, the selection of peptide ions for CID and the identification of the target protein by correlating the peptide CID spectra with sequence databases were performed automatically by the system. Applying the system to the automated identification of proteins separated by high-resolution two-dimensional gel electrophoresis (2DE), we demonstrate concentration limits of detection (LOD) in the low femtomole per microliter range and the possibility for sequential, automated analysis of samples concurrently present on the microfabricated device. The general approach described here is adaptable to the many analytical problems beyond protein analysis which require fast, unambiguous, automated, and sensitive analysis of complex mixtures of analytes. EXPERIMENTAL SECTION Chemicals, Materials, and Instrumentation. Bovine serum albumin (BSA), carbonic anhydrase (CA), β-lactoglobulin (βlac), horse myoglobin (Mb), and human haptoglobin 2-1 (Hg) were from Sigma Chemical Co. (St. Louis, MO). Sequencing grade modified trypsin (porcine) was from Promega (Madison, WI). Acetic acid, acetonitrile, and methanol were obtained from J. T. Baker (Phillipsburg, NJ). The fused-silica capillary tubing was from Polymicro Technologies (Phoenix, AZ). Teflon tube dual shrink and stainless steel tubing were from Small Parts (Miamilakes, FL). Rapidly hardening epoxy glue was from Devcon Corp. (Danvers, MA), UV-curable glue was from Norland Products (New Brunswick, NJ), and Epoxies were from ETC (Greenville, RI). Distilled water was deionized (18 MΩ) using a Milli-Q system from Millipore (Bedford, MA). CE high-voltage power supplies were purchased from Spellman (Plainview, NY). The LCQ ion trap MS (ITMS) was a product of Finnigan MAT (San Jose, CA). βlac concentration was calibrated by amino acid analysis. 2DE and Sample Preparation. 2D gel electrophoresis of total lysate of yeast strain S288C was performed exactly as

Figure 2. Sensitivity of detection measurement. (A) Signal-tobackground ratio in MS mode for three peptides from βlac digest plotted against analyte concentrations (log scale). (B) Cross-correlation factor (Xcorr) from Sequest database searches for βlac peptides plotted against analyte concentration (log scale). Database, bovine protein database. Xcorr values exceeding 2 are considered significant. The legends indicate the m/z and charge state of the peptides.

described,11, except that 40 µg of total protein per gel was applied. Proteins were visualized by a silver stain method modified from Blum et al.17 Using an identical 35S-methionine-labeled gel, protein levels in spots were determined by amino acid analysis of intense spots and scintillation counting to give an average specific activity normalized per methionine. Proteins were excised from the gel and digested with trypsin.18 Each digest was pressure loaded on a small C18 bed inserted in a gel loading tip. The beads were rinsed with 10 mM acetic acid/methanol (90:10 v/v), and the peptides were eluted using 65% acetonitrile. Dried samples were resuspended in 15 µL of 10 mM acetic acid/methanol (90:10 v/v). Construction of the Microfabricated Device. The microfabricated devices (Figure 1) were made by the Alberta Microelectronic Centre (Edmonton, Alta, Canada) from a piece of glass (540 µm thick) following procedures described previously.19 The device was connected to a micro-ESI source via a 15-cm-long fusedsilica transfer capillary (75 µm i.d. × 150 µm o.d.) which was derivatized on the inner surface with (3-aminopropyl)silane.20 The (16) Garrels, J. I.; Futcher, B.; Kobayashi, R.; Latter, G.; Schwender, B.; Volpe, T.; Warner, J. R.; McLaughlin, C. S. Electrophoresis 1994, 15, 1466-1486. (17) Blum, H.; Beier, H.; Gross, H. J. Electrophoresis 1987, 8, 93-99. (18) Shevchenko, A.; Wilm, M.; Vorm, O.; Mann, M. Anal. Chem. 1996, 68, 850-858. (19) Figeys, D.; Ning, Y.; Aebersold, R. Anal. Chem. 1997, 69, 3153-3160. (20) Figeys, D.; Ducret, A.; Oostveen, I. v.; Aebersold, R. Anal. Chem. 1996, 68, 1822-1828.

Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3729

Figure 3. Automated analysis of calibrated samples. MS spectra obtained for (A) BSA tryptic digest flowing from reservoir 1 at 182 fmol/µL, (B) myoglobin tryptic digest from reservoir 2 at 237 fmol/µL, and (C) haptoglobin tryptic digest from reservoir 3 at 222 fmol/µL are shown. Numbers indicate the measured m/z ratio and the charge state of the peptide ions. 3730 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

Figure 4. 2D gel electropherogram of yeast proteins. Aliquots of total yeast (S. cerevisae) lysate containing 40 µg of protein were separated by 2DE (pI range 4-7). Separated proteins were detected by silver staining, and selected spots were subjected to tryptic digestion and automated analysis. The spots analyzed are labeled on the 2D gel electropherogram and summarized in Table 1.

link between the device and the capillary was made perpendicular to the plane of the device by inserting a 200 µm i.d. × 350 µm o.d. sheath capillary into the 350-µm hole at the end of the main channel. The transfer capillary was inserted into the sheath capillary so that its end reached into the etched channel. The whole transfer assembly was sealed and stabilized by a “dualshrink” Teflon tube placed over the sheath capillary. The liquid junction micro-ESI ion source was as described.11,21 The spraying potential applied to the ion source was supplied by the power supply of the ITMS and was held constant at +1.3 to +1.7 kV for the duration of an experiment. Construction and Control of the High-Voltage Relay System. The electronic circuit that controlled the voltage applied to each reservoir consisted of nine high-voltage relays (Kilovac) under the control of optorelays. These were connected to a digital analog converter board (National Instruments) and activated by a program written in Labview (National Instruments). The Labview program was triggered through the LCQ software using the driving voltage of the divert valve. This voltage was applied to a 2.4-kΩ resistor in series with an optorelay to ground. The optorelay connected +5 V to a digital input/output from the DAC board. All the electronic and computer control procedures are transparent to the user. The only interventions required are setting of the proper parameters on the LCQ software for automated data acquisition and generation of CID spectra. Sample Mobilization and Analysis. At the beginning of an experiment all the reservoirs, channels, and capillaries were filled with electrophoresis buffer (10 mM acetic acid/methanol 90:1, (v/v) pH 3.0). The buffer was electroosmotically pumped toward (21) Figeys, D.; Ducret, A.; Aebersold, R. J. Chromatogr., A 1997, 763, 295306.

Table 1. Proteins Identified from a Yeast 2D Gel Using the Nine-Position Microfluidic Device spot no.

protein identified

calcd pI

calcd MW

1 2 3 4 5 6 7 8 9a 10 11 12 13 14 15

G3P1 ALF ADH1 PGK ADH1 METE EF1A IPYR GBLP TPIS MPG1 MT17 HS72 HS75 SCP1

8.3 5.5 6.3 7.1 6.3 6.1 9.1 5.4 5.8 5.7 6.0 6.0 4.9 5.3 6.0

35 618 39 490 36 692 44 607 36 692 85 728 50 032 32 184 34 805 26 664 39 566 48 540 69 338 66 470 61 468

a Does not agree with the pI and molecular weight measured on the gel.

the MS and the micro-ESI source was aligned with the entrance of the MS using an xyz translation stage. Once an optimal sprayer position was achieved, electrophoresis buffer in the reservoirs was replaced with sample solutions. The samples were mobilized from the reservoir by applying -5 kV to the respective reservoir and +1.3 to +1.7 kV to the micro-ESI ion source. Each reservoir was connected to a specific high-voltage relay, and the potential at each reservoir was controlled by the Labview program. The ITMS was used essentially as described12 except that the selection of peptide ions for CID was automated and that the computer controlling the ITMS also controlled the high-voltage relay system. In MS mode, the target number of ions was set to 9 × 107, and in CID mode, to 1 × 107. In a typical experiment, Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3731

Figure 5. MS spectrum obtained for the tryptic digest of spot 14 from Figure 4. Numbers indicate the measured m/z ratio and the charge state of the peptide ions. Table 2. Results of the Search of the Yeast Composite Sequence Database with CID Spectra from Tryptic Peptides Derived from Spot 14a [MH]1+

Xcorr

∆corr

sequence

1041.5 1062.5 1167.6 1185.7 1218.6 1242.7 1459.8 1478.8 1729.9

1.7 2.5 2.7 2.6 2.2 1.7 3.4 3.0 3.6

0.2 0.3 0.3 0.3 0.2 0.2 0.5 0.4 0.7

(K)LLSDFFDGK (K)TGLDISDDAR (R)FEDLNAALFK (K)DAGAISGLNVLR (K)SSNITISNAVGR (R)STLEPVEKVLK (K)SQIDEVVLVGGSTR (R)VTPSFVAFTPQER (R)IINEPTAAAIAYGLGAGK

a All the masses are presented in their MH1+. The sequence of the peptides is indicated.

the ITMS triggered the Labview procedure which in turn applied the voltage to the appropriate reservoir. The first 14.3 min of acquisition were done in MS mode only, scanning the range from 500 to 1400 Da. This left sufficient time for the analyte mixture to reach the MS. The ITMS then automatically acquired MS and MS/MS data in three segments of the spectrum. Peptide ions in the first segment (510-710 Da) were fragmented at 35% of collision energy. Ions in the second segment (710-920 Da) were fragmented at 45% collision energy and ions in the third segment (920-1120 Da) were fragmented at 50% of collision energy. The segments were subdivided into sections of 10 Da, and the ion with the highest intensity was automatically detected and selected for CID using a 3-Da window centered around the mass of the selected ion. Each 10-Da window required a cycle time of 0.26 min. Since the program for the analysis of each section constituted a complete method file, the high voltage was automatically turned off while the method for the analysis of the next segment was loaded in the computer. The CID spectra generated were recorded and used to determine the identity of the protein from which the peptide originated.6,7,20 Before storage, the device was extensively cleaned with methanol and acetonitrile. RESULTS AND DISCUSSION System Description and Mode of Operation. The system is schematically shown in Figure 1. It consists of a microfabricated device for sequential sample delivery, an ESI-MS/MS instrument for the structural analysis of analytes, an array of high-voltage relays controlling direction, origin, and magnitude of the electroosmotic flow, and a computer workstation which controls the relays, the MS instrument, and analyzes the generated data. 3732 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

The microfabricated device consists of nine reservoirs connected via channels to a common transfer capillary. Reservoirs and channels were etched in glass using a photolithographic mask and an isotropic etching process with hydrofluoric acid. The samples were applied to the reservoirs, and one sample at a time was mobilized by inducing an electroosmotic flow from the specific reservoir through the transfer capillary to the MS. The flow was controlled by a high-voltage electrode which was connected via a computer-controlled relay to a high-voltage power supply. The micro-ESI ion source at the end of the transfer capillary interfaced the sample delivery module with the MS in which selected peptide ions were subjected to CID. The resulting CID spectra were recorded and searched against a protein or DNA database using the Sequest software.6,7 This cycle was automatically repeated by mobilizing the sample in the next reservoir until all samples initially applied to the device were analyzed. Limit of Detection. To evaluate the concentration LOD attainable with the system described in Figure 1 we applied tryptic digests of βlac calibrated to concentrations between 160 amol/ µL and 2.6 pmol/µL to reservoir 2 and analyzed the sensitivity of detection by monitoring selected peptides by ESI-MS (Figure 2A) and ESI-MS/MS (Figure 2B). The results shown in Figure 2A indicate that peptides exceeding a concentration of 5 fmol/µL were detectable in MS mode. To determine the LOD in MS/MS mode we used the cross-correlation factor (Xcorr), which was calculated by the Sequest software when the generated CID spectra were searched against a bovine protein database. The factor indicates the quality of the correlation with a value of 2 empirically being considered significant. The results shown in Figure 2B indicate that Xcorr values exceeding 2 were achieved at sample concentrations between 160 amol/µL (m/z ) 771.7) and 1 fmol/µL (m/z ) 458.8; m/z ) 533.3). The higher LOD achieved in MS/MS compared to the MS mode has been observed before12 and is a consequence of the ability of the ITMS instrument to accumulate ions of a specific m/z ratio prior to CID. In summary, the results show that βlac was conclusively identified at subfemtomole per microliter concentration. This sensitivity is comparable to the sensitivity achieved by nano-ESIMS,13 by sample delivery through fused-silica capillaries, and from a simple micromachined device.19 Sample-to-Sample Cross Contamination. Automated, sequential analysis of samples deposited concurrently on the microfabricated device is only possible if sample-to-sample cross contamination can be minimized. Tryptic digests of βlac, CA, and BSA were applied to the adjacent reservoirs 1-3, respectively, and sequentially mobilized by manually switching the high-voltage relays. To make even minor sample-to-sample cross contamination detectable, these analyses were done at concentrations of ∼200 fmol/µL, which exceeded the LOD by a factor of at least 50. One sample at a time was mobilized toward the MS, and CID spectra at every m/z value were manually generated in the range between 450 and 1800 Da. Data acquisition was started 14 min after mobilization of a new sample to allow for the new sample to replace the previous one in the shared segments of the flow path. Analysis of the resulting CID spectra indicated that in the case of βlac sample 13 peptides were identified as derived from βlac. No peptide from either BSA or CA was detected. For the CA sample, five CA-derived peptides were identified while one βlac peptide

Figure 6. MS/MS spectra obtained for (A) ion 865.92+ from spot 14 (see Figures 4 and 5) and (B) ion 7622+ from the low-abundance spot 9 (see Figure 4). The full line double arrows indicate the spacing between the y-ion series and the broken line double arrows indicate the spacing between the b-ion series.

was detected. No BSA peptides were detected. In the BSA sample, 12 BSA peptides were identified and no CA or βlac peptides were detected (data not shown). In 17 experiments of this type, we found an average of 0.7 ( 0.7 contaminant peptides. Furthermore, contaminating peptides were usually present in low amounts. These results indicate that samples concurrently present on the device could be sequentially analyzed with minimal sample-to-sample cross contamination. The number of crosscontaminating peptides, while already low, could be further reduced by flushing the common segments of the flow path with electrophoresis buffer present in one of the reservoirs, by moving the samples back into their respective reservoirs after analysis, or by chemically derivatizing the wetted surfaces to minimize nonspecific peptide absorption. Automated Protein Identification. The manual sample mobilization and analysis procedure was automated by implementing a computer program that coordinately controlled the sample flow from the appropriate reservoirs to the ITMS and the generation and analysis of CID spectra of selected ions (see Experimental Section). Human intervention was thus limited to inserting the sample into the reservoirs and starting the experiment.

Automated Analysis of Standard Proteins. Results from the automated analysis of calibrated protein samples are shown in Figure 3. Tryptic digests of BSA, Mb, and Hg were applied to reservoirs 1-3 and 5-7, respectively. Reservoirs 4, 8, and 9 were filled with electrophoresis buffer. The samples were automatically sequentially mobilized and analyzed. The number of peptides identified varied from 12 to 18 for the proteins tested. This unambiguously identified the respective proteins. Results from the analysis of the samples present in reservoirs 1-3 (Figure 3AC) were indistinguishable from the results obtained from the same samples placed in reservoirs 5-7, which were analyzed in the same experiment (data not shown). Furthermore, as with the manual procedure, sample-to-sample cross contamination with the automated procedure was minimal. Automated Analysis of Yeast Proteins Separated by 2DE. To evaluate the performance of the automated system at higher sensitivity and under realistic experimental conditions, proteins separated by 2DE were analyzed (Figure 4 and Table 1). Protein identity was established by searching the yeast sequence database with the generated CID spectra using the Sequest search algorithm. One cycle of the system took 30 min, and the search of the yeast database took 15-30 min. This means that the Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

3733

database search was finished before the generation of the data for the next sample was done. The amount of protein present in the spots was estimated to be in the range from 12 pmol (spot 1) to 180 fmol (spot 3) and sample concentrations ranged from an estimated 800 (spot 1) to 12 fmol/µL (spot 3). As a representative example, the MS data obtained from the analysis of spot 14 (from Figure 4) are shown in Figure 5. The nine peptides found in Figure 5 unambiguously identified the protein as heat shock protein 75. Table 2 presents the sequences of the peptides identified and the Sequest correlation parameters. The quality of the MS/MS spectra is demonstrated in Figure 6 for ion 865.92+ from Figure 5 and also for ion 7622+ obtained from the lowabundance spot 9 (see Figure 4). These results demonstrate the capability of the system to automatically and routinely identify protein samples at high sensitivity. CONCLUSIONS We have described an integrated, automated analytical system consisting of a module microfabricated by photolithography/ etching technology, an ESI-MS/MS system, an array of computercontrolled high-voltage relays for directing sample flows, and a data system for automated collection and analysis of MS/MS data. Results from the application of the system to the automated, sensitive identification of proteins separated by 2DE are also presented. The coupling of simple microfabricated devices to MS has been reported before.19,22,23 However, this is the first demonstration of a truly integrated system in which the sample flow as well as the analysis of the MS data has been automated under computer control. The approach demonstrated in this report is general, versatile, and expandable and addresses some of the most difficult challenges in the analysis of trace amounts of biomolecules including sample contamination and sample loss. The use of microfabri(22) Xue, Q. F.; Foret, F.; Dunayevskiy, Y. M.; Zavracky, P. M.; McGruer, N. E.; Karger, B. L. Anal. Chem. 1997, 69, 426-430. (23) Ramsey, R. S.; Ramsey, J. M. Anal. Chem. 1997, 69, 1174-1178.

3734 Analytical Chemistry, Vol. 70, No. 18, September 15, 1998

cated flow modules of specific function is not limited to protein identification. By adding affinity modules for the enrichment of analytes containing a specific structural determinant, the system will be expanded for the analysis of protein-ligand complexes and for the determination of protein modifications. The system is equally suited for the conclusive analysis of analytes other than proteins and peptides, as long as the compounds are soluble in solvents compatible with ESI-MS/MS and produce diagnostic CID fragmentation patterns. We expect further improvements in the sensitivity of the system by the use of transfer lines and ES ion sources of smaller inner diameter and by implementing solid-phase extraction online with the microfabricated device. Furthermore, we expect that polymeric variants of such devices will be cost effectively mass produced as routine sample feeds for high-sensitivity ESI-MS/ MS. The integration of sequential steps into a complete analytical process and the automation of process control and data analysis are essential features of a practical analytical chemistry of the future. The system described in this paper is an important step toward this goal. Abbreviations: 2DE, two-dimensional gel electrophoresis; ESI, electrospray ionization; MS, MS/MS mass spectrometer/try and tandem mass spectrometer/try; CID, collision-induced dissociation; βlac, β-lactoglobulin; LOD, limit of detection; BSA, bovine serum albumin, CA, carbonic anhydrase; ITMS, ion trap mass spectrometer, Mb, horse myoglobin; Hg, human haptoglobin 2-1. ACKNOWLEDGMENT This work was supported by the National Science Foundation’s Science and Technology Center for Molecular Biotechnology. The 2D gel of yeast was provided by Y. Rochon and B. Franza, Department of Molecular Biotechnology, University of Washington. Received for review March 19, 1998. Accepted July 8, 1998. AC980320P