Picoliter Sample Preparation in MALDI-TOF MS Using a

Department of Analytical Chemistry, Lund University, P.O. Box 124, 221 00 Lund, ... Research and Development, Astra Draco AB, Box 34, 221 00 Lund, Swe...
0 downloads 0 Views 149KB Size
Anal. Chem. 1998, 70, 4755-4760

Picoliter Sample Preparation in MALDI-TOF MS Using a Micromachined Silicon Flow-Through Dispenser Patrik O 2 nnerfjord,† Johan Nilsson,‡ Lars Wallman,‡ Thomas Laurell,‡ and Gyo 1 rgy Marko-Varga*,§

Department of Analytical Chemistry, Lund University, P.O. Box 124, 221 00 Lund, Sweden, Department of Electrical Measurements, Lund Institute of Technology, Box 118, 221 00 Lund, Sweden, and Bioanalytical Chemistry, Preclinical Research and Development, Astra Draco AB, Box 34, 221 00 Lund, Sweden

This paper presents a picoliter sample preparation technique utilizing the flow-through principle, allowing on-line coupling of chromatographic systems to be made. The work was performed in order to investigate the characteristics and the physicochemical properties of the sample preparation using typical mobile phase conditions from µ-CLC (column liquid chromatography) separations. The device presented here is a pressure pulse-driven dispenser, formed by two silicon structures processed by conventional micromachining. The pressure pulse is generated in the flow-through channel by a piezoceramic element. Depending on the orifice size, the droplets ejected range between 30 and 200 pL. The maximum ejection frequency is 500 Hz, limited by resonances within the unit. A pyramid-shaped nozzle improves the directivity of the droplets since it reduces the wetting of the orifice front surface area. The risk of particles sticking close to the orifice is also minimized. The analyses of the deposited sample spots were carried out on a matrix-assisted laser desorption/ionization time-of-flight mass spectrometer with delayed extraction. It was possible to detect attomole amounts (159-248 amol) of various proteins (cytochrome c, ribonuclease A, lysozyme, and myoglobin) from a single droplet of matrix:analyte 1:1 (drop volume ≈ 110 pL). Additionally, it was found that sample enrichment could be carried out using multiple depositions on the same spot; i.e., 31 nM of insulin was easily detected when more than four depositions were made on the same spot, while no detection was possible without sample enrichment. Size optimization of the MALDI sample spot gave target zones of 100-500-µm diameter that matched the size of the laser focal point and resulted in a considerably increased sample throughput. Droplet dispensers for sample handling in the nanoliter to picoliter range are now making their way into analytical chemistry. This development is fueled by rising demands for speed in highthroughput screening systems and the overall trend toward miniaturized analysis systems. As the dimensions of fluidic systems are reduced to capillary sizes in the range of 10-50 µm, †

Lund University. Lund Institute of Technology. § Astra Draco AB. ‡

10.1021/ac980207z CCC: $15.00 Published on Web 10/17/1998

© 1998 American Chemical Society

there is an increased need for new tools that efficiently and precisely can handle sample volumes in the nano-/picoliter domain. Microdispensing is, therefore, a potential technique for picoliter pipetting and sample injection. Ink-jet printing technology was originally the basis for this development. However, the drawback of the conventional ink-jet print heads is that they are generating droplets from an enclosed volume of liquid. This makes it impossible to insert them in the flow line of flow-through systems for on-line sample deposition. Further, ink-jet print heads are not easily adaptable to fluids that change their physical properties with time, like in most chemistry applications. Manufacturers such as Microdrop (Hamburg, Germany), Microfab (Plano, TX), Cartesian Engineering (Irvine, CA), and GeSim (Grossermansdorf, Germany) supply droplet dispensers aimed to a more general use, and applications have been presented where such dispensers are used for the handling of biological samples,1 for sample introduction in CE,2 for fabrication of sensors,3 and for sample deposition in MALDI.4,5 However, the droplet dispensers in these applications are all of the single-end type, where the sample is either filled from a stationary reservoir or aspirated through the orifice of the dispenser. Since these procedures involve a certain a risk of sample contamination, when running multiple samples, special attention to washing procedures must be considered. Additionally, coupling to chromatographic systems is only possible off-line where, e.g., a fraction collector from an HPLC separation could serve as a sample rack for the dispensing unit. Nilsson et al. have previously presented a solution to this problem, where a dispenser is designed as a flow-through cell and the droplets are ejected perpendicular to a flow passing the cell.6 This flow-through dispenser can be used with a flowing as well as a stationary liquid within the cell. However, the prototype was manufactured in rather crude fashion using conventional (1) Schober, A.; Gu ¨ nther, R.; Schwienhorst, A.; Do¨ring, M.; Lindemann, B. F. Biotechniques 1993, 15, 324-329. (2) Sziele, D.; Bru ¨ ggeman, O.; Do¨ring, M.; Freitag, R.; Schu ¨ gerl, K. J. Chromatogr. A 1994, 669, 254-258. (3) Nilsson, S.; Lager, C.; Laurell, T.; Birnbaum, S. Anal. Chem. 1995, 67, 30513056. (4) Little, D. P.; Cornish, T. J.; O’Donnell, M. J. Anal. Chem. 1997, 69, 45404546. (5) Allmaier, G. Rapid Commun. Mass Spectrom. 1997, 11, 1567-1569. (6) Nilsson, J.; Szecsi, P.; Schafer-Nielsen, C. J. Biochem. Biophys. Methods 1993, 27, 181-190.

Analytical Chemistry, Vol. 70, No. 22, November 15, 1998 4755

machining and showed a limited performance regarding droplet directionality with a large internal cell volume, 35 µL. Silicon micromachining is now a well-established method for fabricating minute-volume flow-through systems.7-10 The method offers a lot of interesting properties, such as the simplicity of fabricating well-defined and small flow channels, minute orifices, and thin membranes, all combined with a high mechanical strength. Not only its mechanical properties but also its chemical robustness make silicon a useful construction material in chemical microsystems. Monocrystalline silicon is chemically compatible with most biochemical buffer systems as well as with organic solvents used in ternary mobile phases for biomolecule determinations. These are some of the reasons for the rapid development of new chemical microanalysis systems seen today. By using micromachining, we have developed a flow-through droplet dispenser that circumvents the potential problems associated with the older version6 and at the same time opens up the possibility of mass fabrication at a reasonable cost. The dispenser presented here has also been used for sample enrichment in capillary electrophoresis.11 MALDI-TOF has been shown to be a very powerful MS technique for molecular mass determinations of peptides, proteins, and oligonucleotides,12-15 especially since the delayed extraction technique was made commercially available.16 To fully utilize the low detection limits and the analytical speed of MALDI-TOF MS, new means of rapid and low-volume sample handling must be introduced and developed. In this context, the use of microdispensing technologies is emerging as an efficient method to deposit small sample volumes onto MALDI plates.4,5,17 The use of small sample spots generated by optimized microdispensing technology circumvents the need to search for “hot spot” areas, and thus the time needed for data collection can be reduced. As a first step toward high-speed determination of proteins with an integrated on-line separation/dispenser/MALDI MS system, this paper describes the development, optimization, and characterization as well as the application of the flow-through microdispenser to deposit protein solutions onto standard gold MALDI plates with subsequent analysis by mass spectrometry. EXPERIMENTAL SECTION Chemicals. Bovine insulin (lot 86H0674), cytochrome c (lot 84H7145), ribonuclease A (lot 16H7161), myoglobin (lot 76H7210), and human insulin, synthetically derived from porcine insulin (I1507), were all purchased from Sigma Chemical Co. (St. Louis, MO). Lysozyme (lot 56518) was purchased from ICN Chemicals (7) Proceedings of the 2nd International Symposium on Miniaturized Total Analysis Systems µTAS-96, Basel, 19-22 November 1996; Widmer, H. M., Verpoorte, E., Barnard, S., Eds.; In Anal. Methods Instrum. 1996, 4. (8) Proceedings of the First International Conference on Microreaction Technology, Frankfurt, 23-25 January 1997, In Microreaction Technology; Ehrfeld, W., Ed.; Springer: Berlin, 1998. (9) Chuong-Ruo, W. Sens. Actuators A 1993, 35, 181-187. (10) Manz, A.; Harrison, J.; Verpoorte, E.; Widmer, H. M. In Advances in Chromatography; Brown, P. R., Grushka, E., Eds.; Marcel Dekker Inc.: New York, 1993; Vol. 33. (11) Peterson, M.; Johansson, J.; Nilsson, J.; Wallman, L.; Laurell, T.; Nilsson, S. J. Chromatogr. B 1998, 714, 39-46. (12) Mann, M.; Talbo, G. Curr. Opin. Biotechnol. 1996, 7, 11-19. (13) Karas, M. Biochem. Mass Spectrom. 1996, 24, 897-900. (14) Fenselau, C. Anal. Chem. 1997, 69, 661A-665A. (15) Harvey, D. J. J. Chromatogr. A 1996, 720, 429-446. (16) Vestal, M. L.; Juhasz, P.; Martin, S. A. Rapid Commun. Mass Spectrom. 1995, 9, 1044-1050. (17) Wallace, D. B. Lab. Automation News 1996, 1, 6-9.

4756 Analytical Chemistry, Vol. 70, No. 22, November 15, 1998

(Costa Mesa, CA); acetonitrile was from Merck (Darmstadt, Germany); and trifluoroacetic acid (TFA, catalog no. 91699) and sinapinic acid (catalog no. 65430) were obtained from Fluka Chemicals (Buchswald, Switzerland) and used without any further purification. The MALDI matrixes were made in acetonitrile/ water containing 0.3% TFA (60:40, v/v) at a concentration of 10 mg/mL. Stock solutions of the proteins were made by dissolving them in phosphate buffer (0.01 M), and working solutions were made by diluting the stock solution in 0.1% TFA. A mass standard kit (no. 2-3143-00, lot no. 7022801, Perseptive Biosystems, Framingham, MA) was used on a daily basis for external mass calibration. The water was purified using a Millipore apparatus (Bedford, MA). Dispensing Apparatus. Sample deposition was accomplished by manual adjustment of the target plate using an x-y stage (Micro-Controle, Evry, France). Visualization of the sample spotting was obtained by using a microscope (Leitz, Wetzlar, Germany) equipped with a videocamera (CCD-72EX, DAGE-MTI Inc., Michigan City, IN). A stroboscopic light source was pulsed at the same frequency as the piezoelectric element using a stroboscope (model 4913, Bru¨el & Kjaer, Naerum, Denmark) to enable viewing of ejected droplets. Typical pulse conditions, obtained using a pulse generator (model 8111A, Hewlett Packard, Palo Alto, CA), were 50-100 Hz, using an applied voltage of 1020 V given by a dc power supply (model E3612A, Hewlett Packard). Mass Spectrometry. The MALDI-TOF MS instrument was a Voyager DE-RP (Perseptive Biosystems Inc.) with built-in delayed extraction and a linear flight path of 1.1 m. It is also equipped with a videocamera system to ensure precise focusing of the 337-nm nitrogen laser, having a focused laser spot diameter of approximately 100 µm. The sample plate can be moved with high precision using an x-y stage with a minimal increment step of 3.2 µm. An accelerating voltage of 25 kV was used for the protein samples using sinapinic acid as the sample matrix. The delay time used was 150-200 ns, and or average up to 128 scans were recorded for each mass spectrum. RESULTS AND DISCUSSIONS Construction of the Flow-Through Dispenser. The microdispenser was assembled from two microstructured silicon dice. The top die comprised the flow connections and the pressuregenerating membrane, and the bottom die included the flow channel and the pyramid-shaped nozzle. A cross section of the microdispenser structure is presented in Figure 1a. The thin actuating silicon membrane adjacent to the flow channel was prepared by n-doping the die followed by a pn-etch stop process,9 which provides a membrane of approximately 10 µm. The die comprising the nozzle was constructed by anisotropic etching of the flow channel as well as an inverted pyramid which defined the shape of the nozzle. The flow channel dimensions were 2 × 15 × 0.1 mm, which corresponds to a cell volume of 3 µL. The die was subsequently n-doped and submitted to a pnetch stop process yielding the pyramid-shaped nozzle, orifice size ) 40 × 40 µm, extending 70 µm above the bulk silicon. The two dice were glued together, and cylindrical plexiglass supports were glued to the top die to enable the mounting of the actuating piezoceramic bimorphous element. Silicone rubber tubes with i.d. fitting to 1/16-in. standard LC tubings were glued to the inlet and outlet.

Figure 1. (a) Schematic cross section of the dispenser construction. (b) Schematic drawing of the actuation principles for the microdispenser.

Operating Characteristics of the Microdispenser. The microdispenser was actuated by applying a short voltage pulse of 10-20 V, 10 µs duration, to the piezoceramic bimorph element. As the actuator bends, the thin silicon membrane is pushed into the flow-through channel, generating a pressure pulse that ejects a droplet through the nozzle, as shown in Figure 1b. To obtain stable droplet formation, it is essential that the orifice front surface is kept free from liquid deposits, crystals, and particles. The pyramid-shaped nozzle helps to ensure this since the front surface area is only about 8 µm wide. Liquid and particles that might be deposited there are more likely to stick to the side walls of the nozzle, with no adverse effect on droplet formation. It is also important that the pressure buildup in the flowthrough channel is rapid in order to create high acceleration of the liquid in the nozzle. If the velocity of the liquid at the exit of the nozzle is too low, the liquid will wet the front area due to surface tension, causing reduced stability and directional control of droplet formation (Figure 2a, showing droplet formation from the first version of the dispenser6). A rapid pressure buildup is facilitated by making the physical dimensions of the dispenser small. Droplet ejection using the nozzle results in a more stable droplet formation, where the droplet is unlikely to wet the outer surface of the orifice (see Figure 2b). Depending on the amplitude and duration of the applied pulse, the tail may get relatively long, and one or more satellite droplets may be formed. This is illustrated in Figure 2c, where a good droplet formation (left) is shown together with a tailing droplet formation at the nozzle (middle) and a satellite droplet (right). The velocity of these droplets is hard to predict, but it is normally less than that of the main droplet. Since the satellite droplets normally are much smaller than the main droplet, they are further more retarded by air resistance and get a trajectory different from that of the main

Figure 2. (a) Droplet formation with old cell construction, resulting in wetting of the orifice, leading to less favorable droplet formation and directivity. (b) Droplet formation with new flow-through cell construction and optimized nozzle design. (c) Good droplet formation (left) and satellite drop formation at two different stages (middle and right), resulting when a too-high voltage is applied.

droplet. This is true unless the droplets are ejected downward. To get satellite-free droplet formation, the pressure pulse shape must be adjusted to the liquid parameters. An important and novel feature offered by the microdispenser is the flow-through option, which enables connection to any analytical microseparation or flow-through system. This allows on-line interfacing to any instrument situated downstream and reduces otherwise costly and tedious procedures of fraction collection and handling prior to the final analysis. A most powerful application would be the connection of the microdispenser to a liquid chromatography system, allowing high-resolution, highspeed separations in line with high-frequency picoliter sample dispensing onto MALDI plates for subsequential analysis. The separation system to be connected should not be limited to pressure-driven separations such as CLC in capillary sizes. Also, electrically driven liquid phase separation techniques such as capillary electrophoresis, capillary electrochromatography, or Analytical Chemistry, Vol. 70, No. 22, November 15, 1998

4757

Figure 3. Illustration of the flow-through cell as an interface to a MALDI-TOFMS target plate. Figure 5. Variation of spot diameter with mobile phase composition and number of dispensed droplets.

Figure 4. Surface tension (4) and drop volume (3) variations at different mobile phase compositions. Data are from ref 18.

simply sample injection for high-throughput screening purposes should also be considered. In this paper, we used the microdispenser according to the sample injection principle, which is illustrated in Figure 3, where the dispenser is filled with the sample solution and subsequently sample droplets are ejected onto the MALDI plate. Physicochemical Properties of Sample Deposition onto Target Surfaces. We investigated the droplet formation, generated drop sizes, and wetting properties which are related to the liquid viscosity and surface tension. The viscosity for water is 0.890 cP, and that for acetonitrile is 0.369 cP at 25 °C.18 The liquid viscosity thus decreases as the content of acetonitrile increases. The surface tension also decreases with increasing amounts of acetonitrile,18 as illustrated in Figure 4. Surface tension and viscosity influenced the voltage (or force) required to eject droplets from the dispenser. The sample spotting was optimized to obtain good droplet formations, whereby the applied voltage had to be adjusted upon variations in the acetonitrile content. Additionally, the droplets varied in size according to Figure 4. It is obvious that the droplet volume decreases with increasing amount of organic modifier but becomes constant at acetonitrile levels of 30% or higher. The resulting variation of generated droplet volume ranging from 270 to 110 pL was due to changes in viscosity and surface tension. Another factor of significant importance is the spot area generated by the dispensed droplets on the MALDI target. This is illustrated in Figure 5, where the spot area at various amounts of droplets is plotted against the organic modifier content of the sample. These sample spot surfaces reach a minimum diameter (18) Handbook of Chemistry and Physics, 63rd ed.; CRC: Boca Raton, FL, 1982.

4758 Analytical Chemistry, Vol. 70, No. 22, November 15, 1998

at approximately 50% of acetonitrile, governed by both the viscosity and the surface tension factors. To make an efficent sample preparation for the MALDI-TOF analysis, it is of utmost importance to prepare the matrix spot as well as the sample spot with high precision. This can be achieved by depositing sample and matrix on the same spot separately or premixing the solution and spotting the mixture directly onto the plate. A matrix solution of 10 mg/mL sinapinic acid with 50% acetonitrile is recommended when mixing sample (0.1-10 pmol/ µL) and matrix solution in a 1:1 (v/v) ratio. However, due to problems with crystallization at the orifice of the sample dispenser, this matrix solution was not suitable for direct spotting. Variation in the percentage of organic modifier was investigated using 20100% of acetonitrile. At percentages below 30%, little or no crystallization occurred, and at levels of 30-50%, crystallization occurred at the nozzle rim. At higher percentages, g70%, the matrix crystallized on the outside of the nozzle due to different wetting properties of the liquid, making it impossible to obtain stable droplet formation with high directivity precision over a long time period. At a level of 60% acetonitrile, crystals formed at the nozzle rim were immediately redissolved, allowing the longest operational time. Additionally, crystals with good laser desorption properties were formed when deposited onto the target plate. Thus, this matrix composition was selected as an optimum and used either to spot matrix alone or sample premixed with matrix. The sensitivity of the MALDI-TOF instrument is not sufficient for analyzing samples at low nanomolar, and picomolar concentrations, such as those present in many biological applications. One way to overcome this drawback would be to use multiple series of droplets spotted onto the target surface, resulting in a preconcentrating effect on the MALDI gold plate before adding the matrix. In Table 1, the increase in signal response versus the volume deposited for insulin at 31 nM is illustrated. It was found that a S/N ratio of 230 was obtained using 16-fold enrichment, while no detectable peaks were found using a 2-fold enrichment. The reason for the large increase between 8 × 100 (eight bursts of 100 droplets each) and 16 × 50 droplets might be a combined effect between increased number of sample depositions and smaller spot surface area. The drying time for a burst of 1-500 droplets was found to vary between 3 and 140 s at various organic modifier levels, which is considered as a relatively long waiting time before the next series of droplets can be dispensed. A plot of drying time versus

Table 1. Sensitivity Variations Using Multiple Depositions (50-200 Droplets) of a 31 nM Insulin Sample on the Same MALDI Target Spot no. of drops

total amount (fmol)

S/N ratio

1 × 200 2 × 200 4 × 200 8 × 100 16 × 50

0.78 1.56 2.73 2.73 2.73

30 60 230

Figure 6. Varation of drying time (s) using various amounts of acetonitrile (%) and various number of droplets (1-500).

the acetonitrile content (typically achieved in HPLC gradient separations) and dispensed droplets is shown in Figure 6. Consequently, to avoid time-consuming sample deposition steps, the drying time of a burst needs to be minimized. Two alternative methods were considered in order to reduce drying time: (i) turning on a small stream of nitrogen/air between spotting series or (ii) heating of the gold plate up to a temperature of 40 °C. The latter approach was considered to be most efficient, and experiments revealed that the drying time was reduced 5-fold at 0% and 2-fold at 100% ACN. Comparisons with Other Systems. The current dispenser was compared with the conventional Interplate sample preparation system (Perseptive Biosystems), which is used for biopolymer samples fractionated at microliter volumes from capillary LC separations. The Interplate is built upon the principle of a small capillary touching the surface of the MALDI plate whereby the sample is deposited. This technique has some disadvantages, such as sample volume variations during gradient separations due to varying liquid properties in the gradient. Additionally, there is an obvious risk of changing the surface structure of the spot, especially if a thin layer of matrix is predeposited. Our dispenser also has sample volume variations, but these are adjusted by varying the number of dispensed droplets. To avoid satellite droplet formations, the applied voltage has to be corrected according to the variations in liquid properties in the mobile phase. The split ratio between waste flow rate and sample

Figure 7. (a) A 4-fold enrichment of protein mixture. 4 × 200 droplets were deposited on the gold-plate, and subsequently 250 droplets of matrix (sinapinic acid, 60% ACN, 0.3% TFA) were applied onto the same spot. Scans averaged: 75. (b) Single droplet, 110 pL of premixed protein sample/matrix solution. Scans averaged: 60. (c) 1 × 200 droplets of a spiked neutrophile sample and subsequently 250 droplets of matrix. The two peaks in the inserted mass spectrum originate from bovine insulin (Mw ≈ 5734) and human insulin (Mw ≈ 5808) Scans averaged: 128.

deposition rate can easily be adjusted by varying the frequency applied (1-500 Hz) on the piezoelectric element. MALDI-TOF Analysis. Optimization of proteins mixed with the deposited matrix was done by using the dried droplet technique19 with manually prepared spots of 0.5-1 µL sample volumes. Other matrixes were investigated, including R-cyanohydroxycinnamic acid (CHCA) and 2,5-dihydroxybenzoic acid (2,5DHB), but sinapinic acid was selected due to its superior resolution and higher sensitivity. However, the problems with crystallization at the nozzle were not found when using 2,5-DHB (water-soluble), while they were even more pronounced when using CHCA. Additionally, glycerol (30% v/v in methanol) was successfully tested in order to investigate the possibility of using liquid matrixes.20 Another water-soluble matrix, 3-hydroxypicolinic acid (3-HPA), which is mainly used for oligonucleotides, also works well. (19) Karas, M.; Hillenkamp, F. Anal. Chem. 1988, 60, 2299-2301.

Analytical Chemistry, Vol. 70, No. 22, November 15, 1998

4759

The amount of matrix needed for good laser desorption properties and high sensitivity were in the order of 10 mg/mL (saturated solution). No significant difference was found when sample (10% acetonitrile) was applied on the plate first and matrix second or vice versa. However, mixing sample and matrix before application onto the gold plate resulted in better sample homogeneity as determined by microscope. The determination of a diluted protein sample is illustrated in Figure 7a, where we deposited 4 × 200 drops (88 nL) at a concentration of 17.8-22.5 nM (1.6-2.0 fmol). The mass spectra of singly charged cytochrome c, ribonuclease A, and lysozyme are shown with resolution of 1900, 1100, and 1400, respectively. The resulting spot diameter is included in Figure 5 (y-axis). Typical values ranging from 100 to 500 µm were found for protein samples containing 1-200 droplets, with organic modifier levels ranging from 0 to 90%. These small areas are optimal for achieving fast data collection, circumventing the need for “hot spot” searching as experienced on a conventionally prepared MALDI target. Additionally, small sample spots have been reported to give significantly enhanced sensitivities.21 In Figure 7b, the mass spectrum obtained from a single droplet of 110 pL is shown at concentrations of 1.6-2.3 µM, corresponding to 225, 186, 178, and 159 amol of cytochrome c, ribonuclease A, lysozyme, and myoglobin, respectively. All proteins are detected at the [M + H]+ ion, and the [M + 2H]2+ ion was detected for cytochrome c and lysozyme. The higher S/N ratio in comparison to Figure 7a is due to a 100-fold higher concentration and a 25fold smaller spot size area (Figure 5). These two examples were chosen to illustrate the analysis from a single series of a 200drop-generated sample (sample enrichment) as well as the smallest sample volume that we can handle today. (20) Dale, M. J.; Knochenmuss, R.; Zenobi, R. Anal. Chem. 1996, 68, 33213329. (21) Jespersen, S.; Niessen, W. M. A.; Tjaden, U. R.; Greef, J. v. d.; Litborn, E.; Lindberg, U.; Roeraade, J. Rapid Commun. Mass Spectrom. 1994, 8, 581584.

4760 Analytical Chemistry, Vol. 70, No. 22, November 15, 1998

Cell lines incubated and stimulated with the calcium ionophor A23187 were lysed, after which the cell supernatant was spiked with a protein mixture. This sample was dispensed (1 × 200 droplets) using the microdispenser, resulting in sample spots containing bovine insulin (38 pmol), human insulin (38 pmol), cytochrome c (18 pmol), ribonuclease A (16 pmol), lysozyme (15 pmol), and myoglobin (13 pmol). The resulting mass spectrum is shown in Figure 7c, where the two peaks at m/z values of 3374 and 3445 originate from the neutrophile sample. The high resolution obtained when using delayed extraction makes the identification of the two insulins from the various species possible, the m/z for singly charged bovine insulin being 5735 while human insulin has an m/z value of 5809, i.e., ∆M ) 74 Da. CONCLUSIONS We have shown, for the first time to our knowledge, a picoliter flow-through dispenser interface used for protein determinations at attomole levels. Preconcentration of samples has successfully been achieved by using multiple sample depositions at the same spot before matrix addition. Investigations of the dispenser characteristics have shown a potential for on-line coupling to CLC separations. The mobile phase variations in fluid properties can be accounted for in the dispensing process; e.g., droplet formation can be controlled by varying the applied voltage to the piezoelectric element and droplet frequency during a gradient run, resulting in constant volume depositions. Current work is focusing on optimization of the dispenser design with respect to cell volume and maximum dispensing rate. Furthermore, parallel dispensing systems are under construction, allowing multiline analysis. Automation for unattended operation is also a major goal in these developments. Received for review February 23, 1998. Accepted August 23, 1998. AC980207Z