Anal. Chem. 1995, 67,4549-4556
A Microscale Electrospray interface for On-Line, Capillary Liquid Chromatographyrrandem Mass Spectrometry of Complex Peptide Mixtures Michael T. Davis, Douglas C. Stahl, Stanley A. Hefta, and Terry D. Lee* Beckman Research Institute of the City of Hope, Duarte, Califomia 91010
A microcapillary liquid chromatography (HPLC) system designed for the gradient elution of peptide and protein samples at flow rates < 1 pWmin has been coupled to a triple-sector quadrupole mass spectrometer via a simple sheathless electrospray interface (microspray). The microspray interface used a flame-drawn, uncoated, fused silica needle with tip outer diameters in the range of 1520 pm and an opening less than 5 pm in diameter. Online sample filtration to prevent clogging of the drawn needle was accomplished by using a hydrophilic PVDF membrane flter integrated into the needle assembly. The spray potential (0.5-1 kv) was applied directly to the sample stream through the capillary union. Stable electrospray conditions were obtained over the full range of the gradient (0-90% acetonitrile in water) and was generally independent of flow rate. Both off-line and online analyses of proteins and peptide digest mixtures were performed at sample levels less than 10 fmol. HPLC parameters could be optimized for either rapid LC/MS analysis or enhanced performance in LC/MS/MS experiments by modulation of the eluting peak widths. Additionally, flow could be greatly reduced as selected componentspass through the interface to prolong the time available to collect mass spectral data. The reduced spectral background and peak width manipulation facilitated the acquisition of peptide product ion spectra (MS/ MS)using real-time, automated instrument control procedures. The advent of electrospray ionization techniques' has fulfilled the promise of tandem liquid chromatography/mass spectrometry analyses of large biomolecules and complex mixtures. In its conventional form, stable electrospray performance requires either postcolumn or coaxial addition of an organic modifier to lower the surface tension of the LC effluent. The organic modifier reduces the concentration of peaks eluting from the column. Additionally, even at low microliter per minute flow rates, only a small percentage of column effluent actually enters the vacuum region of the mass spectrometer. These factors combine to significantly limit the sensitivity of the coupled technique. By scaling down the dimensions of the electrospray needle assembly and operating at flow rates below 1pL/min, the need for sheath gas or sheath liquid is eliminated. The lower potential needed to achieve stable electrospray makes it possible to position the needle (1)Whitehouse, C.M.; Dreyer, R N.; Yamashita, M.; Fenn, J. B. Anal. Chem. 1985, 57,675-679. 0003-2700/95/0367-4549$9.00/0 0 1995 American Chemical Society
within the vortex of gas drawn into the mass spectrometer. This ensures that nearly 100%of the sample enters the first vacuum region. Stable sheathless microelectrospray (microspray) analysis of aqueous peptide and protein samples has been demonstrated by a number of groups but usually is confined to either off-line applications,2-*capillary electrophoresis or pseudo on-line techniques utilizing bulk elution of samples concentrated onto a reverse-phase packing? We present an on-line capillary HPLC microspray interface capable of high-sensitivity LC/MS and LC/MS/MS analyses at submicroliter per minute flow rates. EXPERIMENTAL SECTION
Materials and Suppliers. All fused silica capillary tubing (FSC) was purchased from Polymicro Technologies (Phoenix, AZ). The Vydac reverse-phase packing was obtained from The Separations Group (Hesperia, CA), and the hydrophilic PVDF was purchased from Millipore (Bedford, MA). The myoglobin and cytochrome c proteins were obtained from Sigma (St. Louis, MO), and the sequencing grade endoproteinase-Lys C was purchased from Boehringer Mannheim (Indianapolis, IN). The reagent grade ammonium bicarbonate was obtained from Mallinckrodt (Paris, KY). Reagent grade water (Milli Q, Millipore, Bedford, MA) was used in all aqueous buffers, and the HPLC grade TFA and acetonitrile were obtained from Pierce (Rockford, IL) and EM Science (Gibbstown, NJ) , respectively. Micro-HPLC Design and Column Construction. All LC/ ESI/MS analyses were performed using a microscale HPLC system built by the authors7and recently described in detaiL8This system is distinguished by its use of a preformed gradient for column elution and the use of pressure programing to regulate column flow. Specific modifications to the system for enhanced ultralow flow performance included a reduced volume gradient loop (25 pL, 0.007 in. id.), reduction of critical volumes using 0.007 in. i.d. transfer lines, and 25 mL Hamilton gas-tight syringes for gradient formation. The combined gradient formation flow rate was 50 pL/min. ModAcation of the pressure program from its (2) Wilm, M. S.; Mann, M. Int. J. Mass Spectrom. Ion. Processes 1994, 136, 167-180. (3) Emmett, M. R;Caprioli, R. M. J. Am. SOC.Muss Spectrom. 1994,5, 605613. (4)Kriger, M. S.; Cook, K. D.; Ramsey, R S. Anal. Chem. 1995,67,385-389. (5) Wahl, J. H.; Goodlett, D. R; Udseth, H. R; Smith, R D. Anal. Chem. 1992, 64,3194-3196. (6)Wahl, J. H.;Gale, D. C.; Smith, R. D. J. Chromatogr. A 1994, 659,217222. Stahl, D. C.; Swiderek, K. M.; Lee, T. D. Methods 1994, 6, (7)Davis, M.T.; 304-314. (8) Davis, M. T.; Stahl, D. C.; Lee, T. D. J. Am. SOC.Muss Spectrom. 1995,6, 571-577.
Analytical Chemistty, Vol. 67, No. 24, December 15, 1995 4549
conventional format! included an extension of the loading pressure (4000 psi, 2-3 pL/min) for 1-2 min beyond the injection of the preformed gradient, followed by a 2 min ramp down to the running pressure (200-400 psi, 200 nL/min). With these modifications, there is no signiiicantdifferencein the performance of the HPLC system at these lower flow rates from that previously demonstrated at higher flow rates8 Conventional LC/MS analyses were done with the standard Finnigan MAT electrospray source at a flow rate of 2 pL/min, using a 15 cm x 250 pm i.d. column packed with Vydac C18 reverse phase support. To stabilize the spray over the course of the gradient, sheath liquid (2-methoxyethanol, 2 pL/min) and nitrogen sheath gas (50 psi) were used. The general construction and packing of the microcolumn and the fabrication of the on-column flow cell have been described? Speciiically, the column (10-15 cm in length) and the on-column flow cell were constructed from 100 pm i.d. x 350 pm 0.d. FSC. The transfer lines utilized 20 pm i.d. x 90 pm 0.d. FSC which was butted with a 3 cm length of 100 pm x 350 pm FSC at the end for coupling to the microspray union. The reverse-phase packing (Vydac C18, 5 pm particle, 300 A pore) was retained by a hydrophilic PVDF frit. The column effluent was monitored across a 0.6 mm gap between the transfer limes within the oncolumn UV window. Samples were analyzed using linear gradients of various slopes from 2 to 92% buffer B (A, 0.1%TFA in water; B, 90%acetonitrile, 0.07%TFA in water, v/v) at 100-200 nL/min. Column flow rates were measured by the timed collection of the column effluent into a calibrated disposable glass micropipet (1-5 pL). The partial filling of the microcapillary was measured using an 8x jewelers' loop and compared against the calibrated volume to determine the flow rate. On-column UV detection was performed at 200 nm using an AB1 759A UV/vis spectrophotometer (Applied Biosystems, Inc., Foster City, CA) equipped with a capillary flow cell holder. A Harvard Apparatus Model 44 syringe pump (South Natick, MA), fitted with a 25 pL Hamilton gas-tight syringe (Reno, NV), was used for off-line analyses. Microspray Needle. The microspray needle assembly (Figure 1) utilized a flame-drawn 150 pm id. x 350 pm 0.d. FSC approximately 50 mm in length. Capillary tips were drawn manually using a vertically suspended section of FSC to which a small weight (-30 g, HPLC tee) had been attached. The capillary was slowly heated to melting using the smallest "blue" flame attainable from a Micronox torch (Alltech Associates, Deerfield, IL), which was then quickly withdrawn once the capillary began to draw. All tips were drawn to separation. TIPpatency was conhned by immersion of the tip into a vial of methanol, followed by visual inspection for drawn solvent using an 8x jewelers' loop. Visibly reproducible tips could be obtained with practice. A discontinuous 150pm 0.d. x 25 pm id. capillary was placed within the 150 pm i.d. needle to both minimize the dead volume and retain a hydrophilic PVDF membrane filter (Figure lA). The polyimide coating was removed from the leading end (1 mm) of the first section of transfer line to facilitate insertion and to minimize the dead volume at the tip. Once inserted, this section was cut to its final length (1cm). The PVDF filter was cut using the blunt face of the needle capillary as a punch and positioned by the insertion of the second transfer line. The second transfer line was fully inserted to seat the first transfer line and the frit at the tip of the needle and then withdrawn slightly (< 1mm) before (9) Davis. M. T.; Lee, T.D.Protein Sci. 1992,3, 935-944
4550 Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
Drawn 350 um OD x 150 um ID Fused Silica Capillary
Poiyimide Coating Removed
/
Monolayer Hydophllic PVDF Frit
A
Discontinous?50 um OD x 25 um ID Fused Silica Capillary
B
Valco S t a i n l e s s Steel 1132" ZDV U n l o n wlth Graphlte F e r r u l e s Platinum Sheath
C
Supelco Stainless S t e e l Capillary U n l o n wlth Vespel F e r r u l e
Figure 1. Details of microspray interface design. The microspray interface is a coaxial arrangement of a drawn 350pm x 150pm FSC outer tip containing an internal 150 pn x 25 ym transfer line with an integrated membrane filter. (A) FSC microspray tip. (6)Valco union with graphite ferrules, potential applied directly to union. (C) Supelco union with Vespel ferrule, potential applied to platinum sheath on the transfer line.
being cut to its final length. The withdraw1 of the second transfer line served two functions. First, it was easiest to achieve a flush final assembly by cutting the exposed transfer line slightly short, followed by reinsertion until flush. Second, the PVDF frit lasted longer (before clogging) if it was left decompressed. The final assembly was easiest to perform under a dissecting microscope. Two different methods were used to couple the microspray needle to the transfer line during the course of this work. Most of the data were obtained using a Valco stainless steel 1/32 in. ZDV with 0.4 mm i.d. graphite ferrules (Figure 1B). Later work was done using a Supelco stainless steel capillary union equipped with a single Vespel ferrule (Figure IC). Because the Vespel ferrule is electrically nonconductive, the transfer line was sheathed in a 25 mm platinum tube (0.4 mm 0.d. x 0.19 mm id., 27 gauge) (Hamiton Co., Reno, NV). When it was necessary to replace the transfer line, the platinum tube was removed from the old transfer line by burning the epoxy away with a torch. Microspray Positioning and Imaging System. The positioning of the microspray interface was controlled using an X n micropositioner-optical rail assembly (Newport, Newport Beach, CA), which had been mounted onto the Fmnigan ESI source and observed using a rigid microborescope (Karl Storz GMBH & Co., Tuttlingen, Germany) (30" offset view), coupled to a color CCD video camera (World Precision Inst., Sarasota, FL) and monitor (Figure 3). Unless otherwise indicated, the microspray tip was centered in the plane of the heated, metal capillary of the mass spectrometer, which is defined as the zero position. Standard Sample Preparation. The endo-Lys C enzymatic digestion of horse heart cytochrome c was performed in 100 mM ammonium bicarbonate, pH 7.8, at an enzyme to substrate ratio of 1:100 and a substrate concentration of 20 pmol/pL. Incubation was overnight at 37 "C. The digest mixture was quantitated by amino acid analysis and brought to a final concentrationof 5 pmol/ p1 in 5%aqueous acetic acid for storage. The myoglobin solution was prepared from a sequencing standard, (Sigma, St. Louis, MO,
collision cell was pressure was 2-3 mTorr throughout the run. When selected for MSfMS analysis, the precursor peak width was increased to -5 mfz,and the collision cell voltage was set to 50 V. The second quadrupole analyzer was scanned at a rate of 600 m f z over the range of 50-2000 m fz. Ion selection for MSf MS analysis was based on the signal-tenoise ratio (SNR), defined as the intensity of a given mass divided by the standard deviation of all the signals in the spectrum. In the data shown, the threshold SNR was set at 8. Fragment ion peak widths ranged from 1.5 to 2.5 mfz. RESULTSAND DISCUSSION
Figura 2. Electronphotomicrograph of a manually drawn microspray tip. The white reference bar is scaled to 10 pm.
Figura 3.
System FSC Column De'ector Component diagram for on-line microspray analysis.
70 nmolfvial) by dilution with 5% aqueous acetic acid to a final concentration of 5 pmolfpl.
Mass Spectral Analysis. Mass spectral analyses were performed on a Finnigan MATTSQ 700 triplesector quadrupole mass spectrometer equipped with a Finnigan MAT (San Jose,CA) electrospray source. Conventional electrospray spectra were obtained with the standard Finnigan MAT electrospray source operating at flow rate of 2 pLfmin and a potential of 3.5-4.5 kV. Both sheath gas (nitrogen, 50 psi) and sheath liquid (2-methoxyethanol, 2pLfmin) were used to achieve stable electrosprayover the course of the gradient elution. The electrospraysource highvoltage interlock was bypassed when the microspray interface was used. Mass spectra were obtained over a 1500 amu mass window (250-1750 or 500-2000 amu) using a 3 s scan time. MSfMS analyses were performed automaticallyusing real-time spectrum analysis and instrument control programs developed in our For LCfMSfMS runs, n o d mass spectra were acquired by scanning the first quadrupole analyzer at a rate of 600 m f z per second over the range 500-2000 mfz. The D. C.: M d m , P. A: Swiderek, K M.: Davis, M. T.; Lee, T. D. Pmcrrdiw of the 4Gth ASMS Conference on Mas Spdmomehy and Allied Topics, Washington. DC,May. 1992:p 1801-18M.
(10) Stahl.
System Componentsand Assembly. The microelectrospray assembly developed by Wilm and Man$ utilized drawn borosilicate glass tips (1-3 pm orifice) typically used as microcellular electrodes. The outside of the glass needle was sputter coated with a thin layer of gold to provide the highvoltage contact W i l e these needles are easily fabricated using conventional methods and instrumentationavailable in many biology departments, they are not readily adapted to on-line LCfMS analysis due to their large intemal dead volume. Also,the impermanent nature of the gold coating may compromise an extended analysis in an on-line application. A multistep method for the construction of durable, goldcoated, fused silica capillary microspray needles has been described.' Presently,this method is limited to manually formed tips with rather large (25pm) orifice diameters. Previously it has been shown that the potential may be conducted to the m i ~ o ~ p r a y tip via the liquid stream at a liquid-metal interface,'Z thus avoiding the need to metal coat the glass needle. The design for our interface Flre IA)features an uncoated, drawn fused silica tip with an orifice diameter between 1and 5 pm. It has a reduced dead volume and physical dimensions which afford direct coupling to our LC format without loss of resolution. An integral membrane filter has been included to prevent obstruction of the tip by particulate matter in the liquid stream. Tip durability has been variable but has improved as we have refined our techniques. Some tips have been used for more than 50 h without failure, while a few may last only a few hours. The average l i f e h e is in the range of 8-12 h. Tip failure is generally due to clogging. The needle should be assembled in a clean environment with careful attention to avoiding particulate contamination on the needle side of the frit With particularly dirty samples, the membrane filter may become overloaded, obstructing the flow. Tips that are used for more than 1day are routinely cleaned by wiping with a methanol-wetted tissue or by sonication in ethyl acetate. Needle assembliesare mounted in a commercial capillary union, and replacement requires only a few minutes. Manual drawing of the thin-walled FSC produced a rapid taper down to an outside diameter of -25 pm, followed by a more gradual taper to 15-20 pm 0.d. over a length of 2-3 mm. The region just in front of the orifice has a wall thickness of 2-3 pm and an inside diameter of -10 pm. The end of the needle is a rounded structure with a 1-5 pm ori6ce (Figure 2). TIPS with these dimensions consistently yielded more stable electrospray operation and more intense signals compared to other geomehies. (11) MI. D.C.; Swiderek, K M.: Davis, M. T.; Lee, T.D. b"eedi#@ offkc 4 3 d A S M S Confncncaon Mas SpeclmmetqondAliied Topies.Atlanta, GA
1595. (12) Gale. D.C.; Smith, R D. Ropid Connun Mas Spectmn. 1993. 7, 1017-
1021.
Analytical
Chemistry, Vol. 67, No. 24, December 15, 1995 4551
Because tip dimensions are not exactly reproducible,there is some variability in the operational parameters, particularly the voltage required to achieve stable electrospray throughout the course of the gradient and the pressure needed to achieve a given flow rate. These parameters were readily adjusted, and comparable results were obtained using dif[erent tips. Typical spray potentials were within the range 0.5-1 kV, and the optimal voltage for a given tip was usually 100 V above the spraying threshold, as observed through the borescope. Predictably, decreased spray thresholds were observed to coincide with the smaller orifices. The microspray interface was mounted, through a custom insulating lexan clamp, to an XYZ multiaxis translational stage assembly to allow precise manipulation of the tip position (Figure 3). A rigid microborescope,coupled to a color CCD video camera and monitor, was employed for remote observation of the tip in real time. While not essential, the borescope is very useful for properly positioning the needle, determining the voltage threshold, and monitoring the electrospray process. For the majority of the work reported here, the microspray tip was coupled to the transfer line using a stainless steel union and graphite ferrules (Figure 1B). The graphite ferrules are electrically conductive, and even with the capillary ends butted closely together, there is enough of a gap to provide the electrical contact to the fluid. From time to time, intense background ions were observed in the mass range of m/z 700-1000. Eventually it was determined that these ions were derived from the graphite ferrule. When the capillary ends were closely coupled, there was virtually no background in the mass range above m / z 500. However, if the capillary ends were not cut flush or there was a gap between the ends, exposing the graphite ferrule to the liquid stream, the background ions were observed. This problem was eliminated by switching to a Vespel ferrule and making the electrical contact with a platinum sheath on the transfer line (Figure 1C). The use of other metals such as stainless steel resulted in intense metal adduct ions with many of the peptides. While such adducts can be useful for determining charge states in low-resolution spectra, signal intensity is divided over more species, reducing sensitivity. Also, in those cases when multiple adducts are observed, there can be some question as to whether the MH+ ion is observed and, consequently, uncertainty about the mass of the peptide. These problems become more severe with larger protein structures. Off-Line Microspray Interface Performance. Optimization of the microspray interface with respect to tip position, spray potential, and source parameters was performed using an off-line (continuous infusion) analysis of an aqueous solution of equine apomyoglobin at flow rates from 5 to 100 nL/min. Equivalent spectra were obtained throughout a range of distances (0-5 mm) between the microspray needle and heated metal capillary, each distance having a different optimal potential (data not shown). Thus, the positioning of the interface was not particularly critical. Signal stability was somewhat better at the zero position, presumably due to pneumatic assistance from the high gas flow at the o d c e of the capillary. Spectral intensity was independent of flow rate (data not shown). Consequently, the greatest sensitivity in terms of amount of sample consumed was achieved at the lowest flow rate. At a flow rate of 5 nWmin (as set at the syringe pump, not actually measured), spectra of apomyoglobin could be obtained with signal-to-noiseratios > 6O:l (for charge states > 50%relative 4552 Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
1001
8
9
E+07 8.23
211
8o 60
f 40
I 20
500
'Oo0
m/z
1500
2000
Figure 4. Single-scan microspray mass spectrum of horse heart apomyoglobin. The spectrum was obtained by continuous infusion of 5 pmol/pL equine apomyoglobin at a rate of 5 nUmin, for a total 1.3 fmol consumed for the 3 s scan. Microspray needle potential was 700 V.
abundance) with a total of 1.3 fmol of sample consumed for a single scan (Figure 4). The tip could be inserted beyond the zero position (up to 0.5 mm penetration) without a signifcant effect on signal intensity. On-Line LC/MS Performance. A standardized endo-Lys C digest mixture of equine cytochrome c8 was used to evaluate the on-line performance of the microspray interface. Performance comparisons between the conventional electrospray format (utilizing sheath gas and liquid) and the microspray interface were done at both the low picomole and low femtomole levels. The comparison of analyses done at the 2 pmol level (total amount loaded on the column) illustrates a number of points (Figure 5). With both sources, similar chromatograms were obtained monitoring either the W absorbance or the base peak intensity in the collected mass spectra. There was no loss in chromatographic resolution as the result of using the smaller column format and the microspray interface. In both instances, the electrospray process was stable over the course of the gradient. Base peak intensities were only a factor of 2 or 3 greater for the microspray run, despite the fact that concentrations of eluting peaks were greater by at least a factor of 20. A number of factors conspire to make the electrospray response decidedly nonlinear at higher sample concentrations. These factors include reduced sample ionization, space charge effects, and sample cluster ion formation. Spectra obtained using the microspray source at sample levels of 1pmol and 100 fmol are remarkably different (Figure 6). At 100 fmol (Figure 6B), most of the signal was due to the 2+ charge state. At the higher concentration (Figure 6A), the If charge state was more intense. Additionally, there were more salt adducts (Na, K, and Fe) and dimer clusters formed (ions > m / z 1600). With the signal spread over a number of different m / z values, the base peak intensity increased by only a factor of 2.3 for a 10. fold increase in sample concentration. The expected sensitivity advantage of the microspray source was realized with sample amounts in the low femtomole range. Useful chromatograms could not be obtained from either the UV trace or the base peak trace from the microspray analysis of 2 fmol of the standard digest (data not shown). However, major peptides gave good selected ion mass chromatograms (Figure 7A), comparable to results obtained using the conventional source using 40 fmol of sample (Figure 7B). The advantages of microspray are readily apparent in the comparison of spectra obtained by averaging scans over the peak for mlz 736 (peptide
A “‘1
UV TRACE
A
I
2+ loo,
404.0
1, .. 807.2 I
E+07 3.43
KYIPGTK 1
picomole
E+07 8.415
ml z
200
400
600
800
1000
6
B
ii2+
100,
404.4
E+07 1.51
100
9
femtomoler
60
2 40
I
I
E+07 4.103
4
1+ 806.6
20
1600
0
800
1200
ml z
100
200
300
400
500
Scan
Figure 5. Comparative on-line LC/MS analysis of 2 pmol of cytochrome c standard digest mixture. (A) Sheathless microspray interface and (B) conventional electrospray (sheath gas and sheath liquid). The microspray analysis was performed using a gradient from 2 to 92% B over 100 min at a flow rate between 100 and 200 nUmin, with a 700 V potential. Conventional analysis was performed using a 250 p m i.d. capillaty column eluting at 2 pUmin, with an equivalent sheath flow of 2-methoxyethanol. The gradient profile was from 2 to 92% B over 45 min. Needle potential was 4.5 kV.
d, Figure 5A) using microspray (2 fmol sample, Figure 7 4 inset) and the conventional source (40 fmol sample, Figure 7B, inset). Although signal-to-noiseratios are comparable for the two sets of data (microspray 15:1, conventional source 13:1), the microspray spectrum was much cleaner, with none of the solvent cluster ions derived primarily from the sheath liquid. The basic microspray design of Wilm and Mann2provides for an extended analysis of both simple samples and complex mixtures due to its extremely low flow rate (25 nWmin). Online LC/MS has the advantage that dilute samples are concentrated on the column, peptide and protein components are separated from impurities such as salts and detergents that would compromise the mass spectrometry, and retention time and elution order information is obtained. The principal disadvantage of an LC/MS analysis is that each component can be analyzed for only a short time. To some extent, the time available for the analysis of each component can be increased by increasing the time scale of the chromatography, as illustrated by the analysis of 1 pmol of the standard digest mixture using two different gradient profiles Figure 8). For a rapid LC/MS analysis ( ~ 2 0 min including sample and gradient injection), a representative component (peptide e, Figure 6A) yielded five scans, with a base
Figure 6. Effect of sample concentration on the mass spectrum of the peptide with the sequence KYIPGTK (peak a, Figure 5A). (A) Averaged mass spectrum over the peak using 1 pmol of sample. (B) Averaged mass spectrum over the peak using 100 fmol of sample. Chromatographic parameters were unchanged from Figure 5A. A 750 V potential was applied to the needle assembly.
peak intensity >5 x 106 F i e SA). For a slower analysis Figure BB), spectra for the same component with the same intensity threshold were spread over nine scans. The increase in analysis time is important if automated programs are used to switch between normal MS and MS/MS modes. There is more time for the computer to analyze the spectrum and make the necessary decisions. The principal disadvantages of increasing the time scale of the gradient elution are long run times and very large data files. If all or most of the components in a mixture are important, a long run time with the best chromatography possible is desirable. However, if only one or a few components in a mixture are of interest, other approaches may be more efficient. Rather than decreasing the slope of the gradient, peak elution times can be increased by decreasing the flow rate. This can be done in a discontinuous fashion, with parts of the elution profile collected at a high flow rate to save time and computer disk space and others collected at a low flow rate in order to allow more time for analyzing specific components. The quality of the chromatographic separation will suffer in this mode of operation, but not as much as might be expected. As a test of the feasibility of this approach, two components of the standard digest mixture were selected for extended MS analysis (Figure 9). When ions for peptide d and e iirst appeared, pressure at the head of the column was set to 20 psi. Because flow through our HPLC system is regulated by controlling pressure, nearly instantaneous adjustment of flow rate can be accomplished. The syringe pump actually reverses direction to achieve the lower pressure rather than just stopping. The actual Analytical Chemistry, Vol. 67, No. 24, December 75,7995
4553
A
738
g8 loo:8 0 TGQAPQFNDANK
0
9
6040-
E+04 2.57
200
600
400
400
800
800
1200
1600
mlz
100
1 mlz
649
1 100
1
I\ mlz 804
I 100
200
300
400
500
600
Scan Numbrr Flgure 7. Selected mass chromatograms for three of the peptide components of the standard Lys C digest of cytochrome cat the level of (A) 2 fmol of injected sample using microspray and (9) 40 fmol of sample using the conventional source. Insets are spectra obtained by averaging spectra contained in the peak for the mlz 736 ion. Refer to Figure 5 for chromatographic and mass spectral parameters.
flow at the microspray needle changes with time as the result of decaying pressure in that part of the system. We estimate that the residual flow was on the order of a few tens of nanoliters per minute on the basis of the volume calculated from the peak width in the UV chromatogram and the duration of the signal when normal flow is halted. Mass spectra were collected until the signal decayed significantly (-50 scans), and then the normal running pressure was reestablished and the chromatography continued. The effect on the chromatography can be assessed by comparing the UV and reconstructed ion chromatograms @ICs) (Figure 9, panels A and B). Peptide d was widely separated from what 4554
Analytical Chemistry, Vol. 67, No. 24, December 15, 1995
follows, and no effect in the UV chromatogram was observed as a result of “parking”on the peak. By the time the peak reached the mass spectrometer, it had already passed the UV flow cell. The situation was somewhat different for peptide e. It was followed closely by peptide f, and both had passed the flow cell when the flow was reduced. During the interval that spectra were being collected for peptide e, the band for peptide g entered the flow cell, and the first part of peak g in the UV chromatogram was broadened considerably. Thus, both peptides f and g had eluted from the column and were in the transfer line and flow cell when normal flow was halted. Slowing the flow for -2.5 min
A 100 uv
1
I
I
!
,
,
100
01
I
,
(
I
l
l
e l
I
I
Scan
115
2+ 649.5 TGPNLffiLFM
50
E+07
736.2
1.203
.
1297.5
.
D i
40
400
,
800
1200 1600
m/ z
I
I
200 01
II I
200
Time (minutes)
I
d
l
,
Scan
1
400
Time (minutes)
1
~
(
I
I
600
132
Figure 8. LC/MS analysis of the standard cytochrome c digest mixture using different gradient profiles. (A) The reconstructed ion chromatogram (RIC) from a fast microspray analysis of 1 pmol of injected sample (20 min gradient from 2 to 92% B, 20 min to complete analysis, including sample injection). Inset is the spectrum obtained from five scans averaged over peak e (marked with an asterisk). (B) The RIC from a slower analysis performed at the same sample level (100 min gradient from 2 to 92% B, 40 min total with sample injection). Inset is the spectrum obtained from nine scans averaged over peak e (marked with an asterisk). In both instances, scans with base peak intensities > 5 x lo6 were used for averaging. The potential applied to the needle assembly was 750 V.
had only a minimal effect on the chromatographic resolution for the latereluting components. Even peaks f and g were minimally disturbed, as evidenced by the mass spectrometer RIC. Minor scan-to-scan fluctuations in ion intensities were evident in the mass chromatograms for the m / z 1471 and 1297 ions (Figure 9, panels C and D). However, all spectra over the course of the peak were of similar quality. Figure 9 insets are single-scan mass spectra selected at random for each of the periods when the flow was slowed. The peak parking procedure has great significance for LC/ MS analyses. The mass spectrometer is no longer totally subservient to the chromatographic time frame. By extending the time that a sample peak is available to the mass spectrometer by a factor of 5-10, the number of analysis options increases dramatically. Narrow mass range higher resolution scans could be done over specific ions to obtain charge state information. MS/ MS spectra could be obtained on different charge states of the same component or on ions of different coeluting components. MS/MS spectra can be obtained for a component using a variety of collision energies. There is sufficient time for multiple stages
Figure 9. Analysis of the standard cytochrome cdigest mixture with two peptides selected for extended MS analysis. (A) UV chromatogram of the effluent from the HPLC. (B) Chromatogram reconstructed from the total ion signal at the detector of the mass spectrometer (RIC). (C) Mass chromatogram for the m/z 1471 ion. (D) Mass chromatogram for the m/z 1297 ion. Region between the two sets of dotted lines indicates those scans collected while the flow was halted. Insets in panels C and D are randomly selected single-scan mass spectra taken from each of the two sets of slow flow scans.
of MS/MS on instruments such as ion traps and ET-ICR mass spectrometers. It should be emphasized that this mode of operation is possible only with the microspray interface. With the normal electrospray interface, spray stability is too dependent on the flow rate, and nearly stopping the flow would result in loss of the signal. Microspray LC/MS/MS Analysis. The cumulative benefits of the lower baseline noise, higher sensitivity, and demonstrated flexibility of microchromatography/microspray analysis greatly facilitate automatic real-time analyses of low sample amounts. Using the present version of our automated LC/MS/MS programs," normal mass spectra are collected by scanning Q1 and collecting ions at the detector after they pass through the octopole collision cell and Q3 quadrupole analyzer. A constant collision cell pressure is maintained throughout the course of the analysis. Once a specified SNR threshold has been crossed, the data system locks Q1 on a precursor ion mass, while Q3 is scanned to collect the resulting fragment ions. After the SNR has dropped below the threshold value, the mass spectrometer reverts to the original scan mode for the detection and/or selection of the next parent ion. Some intensity in the normal MS scans is lost due to the pressurized collision cell, and microspray ionization helps to maintain adequate SNRs ratios at lower sample amounts. Additionally, the reduced level of background means less time is devoted to collecting MS/MS data on background ions. Analytical Chemistry, Vol. 67, No. 24, December 15, 7995
4555
A
Ix2 100 B 286.9
~ + 2 2 + 738.0
I
T
500
I00
;y
1x15
1
11136
!+OS 1.14
TGQAPGFMDANK 1
plcomok
1500
1000
E+05 1.49
11136
286 9
100
358.0
lemtomolas
2*
M+2H
01.3
716.7
185.5
500
1000
1500
ml z
Figure 10. Automatic capillary LC/microspray MS/MS analysis at the 1 pmol and 100 fmol levels. (A) Averaged spectra of the automatic MS/MS analysis over peak d obtained from 1 pmol of injected standard (eight scans averaged). The Roepstorff nomenclature is used to identify fragment ions.13(5)Averaged spectrum over peak d obtained from 100 fmol of the same standard (five scans averaged). In both instances, the applied potential at the electrospray needle was 700 V.
Automated LC/MS/MS analyses were done on the standard digest mixture at levels of 1pmol and 100 fmol. For the purpose of this comparison, the same peak selection criteria were used in
each case. At the 1 pmol level, nearly all of the known components of the mixture were selected for MSIMS analysis. At the 100 fmol level, only the six peptides having more intense signals were selected for MS/MS analysis. In each instance, the fragment ion spectra were of good quality, as illustrated by spectra (Figure 10) for the ion at mlz 736 (peptide d, Figure 5). At the 1pmol level, a total of eight fragment ion spectra for the m/z 736 precursor were collected (Figure lOA), the most the program will allow. At the 100 fmol level @@re lOB), only five spectra were collected. Nevertheless, there was still a good correspondence between the two spectra, and there was no question concerning the identity of the peptide. CONCLUSIONS The benefits of on-line capillary LC/microspray analysis over the conventional electrospray methodology are evident in reduced spectral background, decreased sample consumption, and enhanced sensitivity. Precise real-time positioning of the microspray needle is useful but not critical. Equivalent spectral quality could be obtained with use of a broad range of intercapillary distances and corresponding needle potentials. Gradient profiles can be either tailored for rapid analysis at high sensitivity or extended to prolong the acquisition of fragment ion information in an automatic LC/MS/MS analysis. Additionally, with the microflow interface, flow can be reduced to levels of a few tens of nanoliters per minute while a peak is in the mass spectrometer interface. This greatly extends the available analysis time for selected components without seriously affecting either the electrospray signal intensity or the chromatographic separation. Good quality MS/MS data can be obtained at the level of 100 fmol of injected sample using automated procedures. ACKNOWLEDGMENT This work was supported in part by the Public Health Service (NIH RR 06217 and NIH CA33572). Received for review July 10, 1995. Accepted September 21, 1995.m AC9506894 ~
(13) Roepstorff, P.: Fohlman, J. Biomed. Mass Spectrom. 1984, 11, 601-602.
4556 Analytical Chemisrty, Vol. 67,No. 24, December 15, 1995
~~~
@Abstractpublished in Adoance ACS Abstracts, November 1, 1995
