Anal. Chem. 1994,66, 3688-3695
LWESI-MS Determination of Proteins Using Conventional Liquid Chromatography and Ultrasonically Assisted Electrospray J. Fred Banks, Jr.,* John P. Quinn, and Craig M. Whitehouse Analylica of Branford, Inc., 29 Business Park Drive, Branford, Connecticut 06405
Conventional unassisted electrospray ionization mass spectrometry (ESI-MS) has severe limitations as a liquid chromatography (LC) interface due to the few solvent compositions which can be electrosprayed without some type of assistance. LC mobile phases which have high flow rates (over 5 pL/min), high conductivity, or high surface tension are normally unsuitable for use with ESI. For this reason, an ultrasonic nebulizer has been developed which mechanically creates the fine spray of droplets needed for ESI and thus functions independently of the solvent composition. This device also operates at considerably higher liquid flow rates, up to several hundred microliters per minute. To characterize the system’s performance, the ultrasonic nebulizer frequency, source electrode potentials, and drying and focusing gas flow rates were studied and optimized. Also, droplet size measurements were taken using a phase-Doppler anemometer, which showed consistent nebulizer performance up to a liquid flow rate of several hundred microliters per minute. Finally, the ultrasonic nebulizer was used for an LC applicationinvolving the separation of proteins on 1.0 and 2.1 mm i.d. columns. LC flow rates of up to 200 rL/min of mobile phases containing 0.1% trifluoroacetic acid could be easily nebulized. With the 1.0 mm i.d. column, as little as 32 fmol of cytochrome c could be detected in selected ion monitoring mode. Electrospray ionization (ESI) possesses the proven ability to both desolvate and ionize fragile chemical species, such as proteins, peptides, nucleic acids, and pharmacologically active compounds, from solution for transport to a mass analyzer. The types of mass analyzers which have already been interfaced successfully with ESI include quadrupole,’“ magnetic,’-I2 Fourier transform,13-15ion and time-of-flight.18 Ion ( I ) Yamashita, M.; Fenn, J. B. J . Phys. Chem. 1984, 88,4451. (2) Yamashita, M.; Fenn, J. B. J . Phys. Chem. 1984, 88, 4671. (3) Whitehouse, C. M.; Dreyer, R. N.; Yamashita, M.; Fenn, J. B. Anal. Chem. 1985, 57, 675. (4) (5) (6) (7)
Meng, C.; Mann, M.; Bennett, F. J. Z . Phys. D 1988, 10, 361. Mann, M.; Meng, C.; Fenn, J. B. Anal. Chem. 1989, 61, 1702. Wong, S.F.; Meng, C.;Fenn, J. B. J. Phys. Chem. 1988, 92, 546. Meng, C.K.;McEwen, C. N.; Larsen, B. S. Rapid Commun. Mass Spectrom.
spray, which incorporates a high-velocity nebulizing gas in the same direction as the liquid flow is an alternative method for liquid nebulization.19-22 While ESI has become increasingly popular as an interface between liquid chromatography (LC) and mass spectrometry (MS), severe limitations have greatly restricted the types of LC applications which are amenable to ESI and stunted the growth of this very promising biochemical analysis tool. The most significant limitations of ESI primarily revolve around the chemical and physical properties of the supporting solutions (and not the analytes themselves) that are amenable to the electrospray ionization process. At the current time, conventional unassisted ESI is not compatible with LC mobile phases or sample solutions which (1) exceed a volumetric flow rate of approximately 5 pL/min, (2) have high conductivity. or (3) have high surface tension. It is for this reason that optimal electrospray performance is usually reported to be obtained from methanol/water solutions at flow rates of 1-3 pL/min. While not unreasonable for flow injection analysis (FIA) studies of pure compounds soluble in this particular solvent system, these conditions are unrealistic for the majority of LC applications of interest. This incompatibility between optimal LC mobile phases and “electrosprayable” solutions has prompted us to pursue a single solution to address simultaneously all these aforementioned problems. In order to best accomplish this task, we have sought a means to decouple as much as possible the spray formation and ionization processes, so that the generation of the fine spray of droplets required for successful ion evaporation in the source, as originally described by Iribarne and Th0mpson,2~ could be treated as a single problem. The primary requirement was that the spray formation process be relatively independent of LC mobile-phase physical properties and flow rate. This challenging goal suggested that the spray should be produced by some mechanical means, Le., ultrasonic vibration, instead of through the application of electric fields. The work described here summarizes our efforts in this area and characterizes an ultrasonic nebulization device for use with
1990, 4, 15 1.
(8) Meng, C.; McEwen, C. N.; Larsen, B. S.Rapid Commun. Mass Spectrom. 1990, 4, 147. (9) Gallagher, R. T.; Chapman, J. R.; Mann, M. Rapid Commun. MassSpectrom. 1990, 4, 369. (10) Larsen, B. S.;McEwen, C. N. J . Am. SOC.Mass Spectrom. 1991, 2, 205. (11) Wada, Y.; Tamura, J.; Musselman, 8.D.; Kassel, D. B.; Sakari, T.; Matsuo, T. Rapid Commun. Mass Spectrom. 1992, 6, 9. (12) Cody, R. B.; Tamura, J.; Musselman, B. D. Anal. Chem. 1992, 64, 1561. (13) Henry, K. D.; Williams, E. R.; Wang, B. H.; McLafferty, F. W.; Shabanowitz, J.; Hunt, D. F. Proc. Natl. Acad. Sci. U.S.A. 1989, 86, 9075. (14) Henry, K. D.;Quinn, J. P.; McLafferty, F. W. J . Am. Chem. Soc. 1991,113, 5447.
3688 Analytical Chemistry, Vol. 66, No. 21, November 1, 1994
(15) Loo, J. A.; Quinn, J. P.; Ryu, S.I.; Henry, K. D.; Senko, M. W.; McLafferty, F. W. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 286. (16) Van Berkel, G. J.; Glish, G. L.; McLucky, S.A. Anal. Chem. 1990,62,1284. (17) Lin, H.-Y.; Voyksner, R. D. Anal. Chem. 1993, 65, 451. (18) Boyle, J . G.; Whitehouse, C. M. Anal. Chem. 1992, 64, 2084. (19) Bruins, A. P.; Weidolf, L. 0. G.; Henion, J. D.; Budde, W. L. Anal. Clem. 1987, 59, 2647. (20) Bruins, A. P.; Covey, T. R.; Henion, J. D. Anal. Chem. 1987, 59, 2642. (21) Lee, E. D.; Muck, W.;Covey,T. R.; Henion, J. D. Anal. Chem. 1989,18,844. (22) Nichols, W.; Zweigenbaum, J.; Garcia, F.; Johansson, M.; Henion, J. LC-GC 1992, I O , 676. (23) Iribarne, J. V.; Thompson, B. A. J . Chem. Phys. 1976, 64, 2287.
0003-2700/94/0366-3688$04.50/0
@ 1994 American Chemical Society
LC/ESI-MS. A previous paper24described results obtained at low flow rates for capillary LC with aqueous mobile phases having high surface tension. Here, however, we have made further investigations that now show the use of high flow rate and high conductivity as well as high surface tension LC mobile phases. The application chosen for examination was the detection of proteins separated on 1.0 and 2.1 mm i.d reversed-phase LC columns. Proteins and peptides are of course widely studied because of their biological activity. Recent developments in total synthesis as well as recombinant production of these species have further increased demands for improvements in instrumental analysis tools. ESI-MS has been a natural choice for advanced detection due to its ability to form multiply charged ions. The resulting mass spectra not only reduce the observed m / z values to levels that standard quadrupole mass analyzers can detect but also give an accurate molecular weight of the protein when d e c o n ~ o l u t e d .Unfortunately ~~ for ESI, LC analysis carried out on reversed-phase columns with 0.1% trifluoroacetic acid (TFA) added in the mobile-phasegradients is one of the predominant methods of protein separation. This inclusion of TFA in the mobile phase, which is mandatory for adequate reversed-phase chromatography, raises the conductivity of the mobile phase to a level unsuitable for ESI. Additionally, the use of mobile phase flow rates of up to 1 mL/min, which conventionalLC mandates, is again unsuitable for ESI. While the use of organic sheath liquids has been previously demonstrated to help generate ions for ESI from both highly conductive solution^^^,^^ and solutions with high surface t e n ~ i o n ,this ~ ~ .aid ~ ~does not solve the restriction on the use of high flow rates. The ultrasonic nebulizer, however, can easily generate a spray from a conductive mobile phase at flow rates of hundreds of microliters per minute, thus falling well within the range of optimum performance for 1.O and 2.1 mm i.d. LC columns, and possibly within the range of optimum performance for 4.6 mm i.d. LC columns. In the work described here, the use of an ultrasonic nebulizer to assist in the formation of a spray for LC/ESI-MS is described. The purpose of this development is to expand the realm of LC applications currently amenable to ESI and thus help bring to fruition the promise of ESI as an interface for on-line LC/MS analysis.
EXPERIMENTAL SECTION Apparatus. The pump used for mobile-phase delivery was a dual-syringe type, Model AB1 140A from Applied Biosystems, Inc. (Foster City, CA), equipped with 10 mL capacity cylinders. The 1 mm i.d. c18 LC column used was from Microtech, Inc. (Saratoga, CA) and was 15 cm in length. The 2.1 mm i.d. column used was from Vydak (Hesperia, CA) and was 25 cm in length. Sample injection was accomplished (24) Banks, J . F., Jr.; Shen, S.; Whitehouse, C. M.; Fenn, J. B. Anal. Chem. 1994, 66, 406. (25) Labowski, M.; Whitehouse, C.; Fenn, J. Rapid Commun. Mass Spectrom. 1993, 7, 71. (26) Smith, R. D.; Barinaga, C. J.; Udseth, H. R. Anal. Chem. 1988, 60, 1948.
(27) Mylchreest, I.; Hail, M. Presented at the Proceedings of the 39th Conference on Mass Spectrometry and Allied Topics, Nashville, TN, 1991. (28) Parker, C. E.; Perkins, J. R.; Tomer, K. B.; Shida, Y.;OHara, K.; Kono, M. J . Am. SOC.Mass Spectrom. 1992, 3, 563. (29) Hunt, D. F.; Henderson, R. A,; Shabanowitz, J.; Sakaguchi, K.; Michel, H.; Sevilir, N.; Cox, A. L.; Appella, E.; Engelhard, V.H. Science 1992,255,1261,
with a Rheodyne (Cotati, CA) 8125 LC injector with a 5 pL sample loop. The electrospray ionization source was similar to that developed at except that an ultrasonic n e b u l i ~ e r ~ ~ * ~ ~ J ' was substituted for the original needle assembly. The ultrasonic nebulizer was driven using a Wavetech (San Diego, CA) Model 90 function generator and an EN1 (Rochester, NY) Model 240L power amplifier. More complete details of the nebulizer itself are given in a previous p~blication.2~ The source design allowed independent control of the potentials on the nebulizer needle ( V n d ) , cylindrical electrode (VcYl), nosepiece (Vn-e), and capillary entrance (Vent). With this configuration, the system was always operated with V n d = ground dc offset potential in order to protect the user from electrical shock and isolate the LC system from high voltage. Drying-gas temperature was always set to 300 OC. The mass spectrometer (HP89A) was from Hewlett-Packard (Palo Alto, CA). Materials. Protein samples were obtained from Sigma (St. Louis, MO). At the time of injection, these samples were dissolved at the appropriate concentration in a mobile-phase composition that was identical to the beginning of the LC gradient. All water was obtained from a Barnstead (Boston, MA) NANOpure I1 system. Acetonitrile (ACN) was purchased from Mallinckrodt, Inc. (Paris, KY). All solvents were filtered through nylon 66 membranes from the Anspec Co., Inc. (Ann Arbor, MI). TFA was also obtained from Mallinckrodt, Inc.
RESULTS AND DISCUSSION Effect of Ultrasonic Nebulizer Operating Frequency. The initial experiment for effective use of the ultrasonic nebulizer was to find and maintain its resonant frequency. This frequency is determined largely by the physical dimensions of the resonant body of the nebulizer itself. To optimize the system for its natural resonant frequency, cytochrome c (1 pmol/pL) was dissolved in a 1:1 (v/v) mixture of water and ACN in 0.1% acetic acid and the resultant mixture was infused directly into the source at a flow rate of 50 pL/min. The driving frequency was then varied over a given range while the signal for all of the appearing multiply charged peaks was summed to give a final signal strength. The maximum ion signal occurred at 21 1.2 kHz, so this frequency was used for all future studies. While all of the work here was accomplished with manual control of the ultrasonic nebulizer using a frequency generator and power amplifier, recent developments now allow for fully automated driving of the nebulizer from a single electronic control device. Effect of VeYlon Signal Strength. Although the data just shown were encouraging, original results obtained with the use of the ultrasonic nebulizer at high sample flow rates were poor. In these early studies, the electrospray source potentials were operated in a manner consistent with conventional electrospray with Vned = ground, Vcyl = -3000 kV, V,,, = -4000 kV, and Vent= -5000 kV. Under these conditions, the observed ion signal at a sample flow rate of 50 pL/min was (30) Banks, J. F., Jr.; Shen, S.; Whitehouse, C. M.; Fenn, J. B. Presented at the Pittsburg Conference, Atlanta, GA, 1993; paper 726. (31) Banks, J. F., Jr.; Shen, S.;Whitehouse, C. M.; Fenn, J. B. Presented at the Pittsburg Conference, Atlanta, GA, 1993; paper 728.
Analytical Chemistty, Vol. 66, No. 21, November I, 1994
3089
generally over 1 order of magnitude less than conventional electrospray yields when the flow rate is 1 pL/min. After noting that the droplets from the nebulizer were formed and thus charged from the side of the nebulizer tip, we investigated increasing further the negative potential on the cylindrical electrode in order to (1) increase the electric field strength around the side of the needle tip and (2) attract some of the charged droplets away from the capillary center line, thus reducing the drying requirements in the acceptance region in front of the capillary. Intuitively, it seems that diverting droplets, and ultimately ions, would tend to decrease the observed ion signal. However, knowing that many more droplets and ions are produced than can be sampled (due to space charge effects and drying limitations), we assumed that directing a fraction of the ion current to the side would not substantially affect the ion signal. This assumption is supported by our previous observation with conventional ESI, where the maximum ion signal is generally observed at liquid flow rates in the nanoliter per minute range. Increasing the electric field in the region of the tip was facilitated in this case by the presence of a constant layer of liquid on the ultrasonic needle tip which acted as a natural electrical insulator. A conventional electrospray needle, such as that normally used with this source, will usually produce a corona discharge when Vcylis increased much beyond -3000 kV. The ultrasonic needle, however, allows VCylto be ramped to at least -8000 kV beforea discharge occurs. It is important to note that some other minor physical changes in the source were necessary to prevent electrical discharge in other vulnerable areas not originally designed to sustain voltage differentials of this magnitude. The effect of increasing the electric field strength through Vcyl was then investigated. By use of the same test solution and flow rate as before, Vcylwas varied from -3000 to -8000 kV, and the ion signal for cytochrome c was measured. These data are summarized in Figure 1 and show that, with Vcyl at the optimum potential of -7000 kV, the relative ion signal observed has been increased nearly 20-fold over that when a more conventional potential of -3000 kV is used. The use of these higher electric fields in the ion source, then, was found to be a critical point for obtaining optimum performance, especially at high liquid flow rates. Effect of Drying and Focusing Gas Flow Rates. A final consideration in the design and subsequent successful operation of this new probe was the inclusion of a focusing gas flow countercurrent to the usual drying gas flow. We had anticipated that the somewhat larger droplets produced from mechanical vibration would require the use of additional drying gas. To compensate for a higher drying gas velocity, which has the potential to blow droplets and ions away from the capillary entrance, the use of a “focusing gas” introduced around the needle tip and oriented 180’ from the drying gas was explored to aid in the transport of droplets and ions toward the capillary entrance. Based on previous experience with iow flow rate the relative effects of drying and focusing gas flows were investigated by independently varying these two flows while directly infusing the test solution used above at 50 pL/min. The results of this experiment are shown in Figure 2, and several observations on these data are appropriate for discus3690
Analytical Chemistty, Vol. 66, No. 21, November 1, 1994
’ T 0.9
0.8 0.7
11
/
--
0.6 --
-m
E, 0.5 --
i7j
0.4
--
0 4
3000
I 5000
4000
7000
6000
8000
VCyl (-kv)
Figure 1. Relative cytochrome c slgnal vs V,.
0
200
400
600
800 1000 1200 1400 1600
Focus Gas (mumin) 1-9
+13
+17
+21
1
Drying Gas (L/min)
Flgure 2. Relative cytochrome c signal vs drying and focusing gas flow rates.
sion. First, as the drying gas flow rate was increased from 9 to 21 L/min, the initial signal with no focusing gas applied consistently decreased. This observation was as expected, confirming that the increasing drying gas velocity blows droplets and ions away from the capillary entrance. Also, as
1000000 1
120000100000-
100000
00000.
60000-
10000 40000-
-
6
1000
3
d o 3""w."ce
L,
idoo
11100
ld00
1100
.
,
, 12100 , , , , 1300 , , , ,
.
, , , , , , , , 12100 iioo
120000
100 10000080000-
10 60000-
40000-
1
20000-
0
-.-
200
400
600
800
1000
-
Flow Rate (uUmln)
Conventional Electrospray
-
-9 b O A
I,
,
Ultrasonically-assisted Electrospray
Flguro 3. Cytochrome c slgnal vs sample flow rate with conventional eiectrospray and ultrasonicaily assisted electrospray.
the drying gas was steadily increased, the amount of focusing gas that was required to maximize signal increased as well. This result was again reasonable, since the ever increasing drying gas velocity should require an increasing focusing gas velocity to recover the ion signal by mechanically transporting the droplets back toward the capillary entrance. Finally, the optimum signal was obtained at very modest drying gas and focusing gas flow rates, 13 L/min and 900 mL/min, respectively. Further increasing the drying gas flow rate only initiated a losing battle requiring theuse of increasing focusing gas which never restored the signal to its maximum level. Finally, the use of excessive focusing gas, it seems, always resulted in sudden signal loss, most likely due to the transport of large partially dried droplets into the capillary. Ultrasonic Nebulization with Mobile Phases Having High Flow Rates. As discussed above, the use of an ultrasonic nebulizer was explored for this application in order to overcome some of the severe limitations associated with conventional electrospray. These limitations result in the complete loss of ion signal observed when mobile phases are used that are associated with high liquid flow rates, high conductivity, or high surface tension, and thus greatly restrict the range of LC/MS applications that are amenable to electrospray ionization. To asses system performance relating to the first of these limitations, the use of high liquid flow rates, ion signal vs sample flow rate, was compared when both conventional electrospray and ultrasonically assisted electrospray were used. For this pair of experiments, the same cytochrome c solution was again directly infused into an electrospray source, and the total ion signal (sum of multiply charged peaks) was recorded across a range of sample flow rates. With conven-
Flguro 4. (A) Cytochrome c spectrum wlth conventional ESI at 1 pLlmln in 1:1 (vlv) ACNlwater wlth 0.1 % acetic acM, (B) cytochrome c spectrum wlth ultrasonic nebullzatlon at 50 pL/mln in 1: 1 (vlv) ACNl water with 0.1 % acetic acid, (C)cytochromecspectrum wlth ultrasonic nebulization at 50 pL/mln In 1:l (v/v) ACNlwater with 0.1 % TFA, and (D) cytochrome c spectrum with ultrasonic nebulization at 50 pL/mln In water wlth 0.1 % acetic acid.
tional electrospray, a standard needle assembly (100 pm i.d.) was used to generate the spray, but for ultrasonically assisted electrospray, the ultrasonic nebulizer was substituted in its place. Figure 3 shows the relative signal arising from cytochrome c when both conventional electrospray and ultrasonically assisted electrospray were used at sample flow rates ranging from 1 to 1000 pL/min. The data show that the ion signal observed from conventional electrospray begins to fall quickly after 1 pL/min and is essentially reduced to zero by 20 pL/min. However, the Analytical Chemistty, Vol. 66,No. 21, November 1, 1994
3691
A
2.48k
1000000
1 I
P
I
100000 --
-a
10000
-Diameter (pm)
.-cn v)
1000
--
100
--
lo
--
I
I
signal observed when the ultrasonically assisted electrospray is used indicates that a much broader range of operation is possible, up to 1 mL/min, with only modest losses in signal occurring with increasing flow rate. It should be noted that, because of the tremendousvolume of liquid entering the source at flow rates exceeding a few hundred microliters per minute, reoptimization of the drying and focusing gases as well as the operating frequency was necessary to maintain performance. Also, with flow rates in excess of approximately 300 pL/min, the atmospheric side of the source sometimes accumulated liquid. This problem was remedied by inserting a simple drain in the bottom of the source compartment. Nonetheless, the data show that at flow rates from 50 to 200 pL/min, optimum for the 1 and 2.1 mm i.d. LC columns used in this work, the use of an ultrasonic nebulizer produced signal that was essentially equivalent with the best signal obtained with either conventional electrospray or ultrasonically assisted electrospray at 1 pL/min. This signal level can be directly compared to a signal level of zero, which is the observed result when a flow rateof50pL/minisused without the helpoftheultrasonic nebulizer. Finally, panels A and B of Figure 4 show cytochrome c scans (m/z 600-1400 ) obtained at 1 and 50 pL/min with electrospray and ultrasonically assisted electrospray, respectively. The appearance of the spectra, and the magnitude of the signal and background noise, are nearly identical. Ultrasonic Nebulization with Mobile Phases Having High Conductivities. A second critical failure of conventional electrospray, the inability to be used with mobile phases containing highly conductive additives, was also solved by the use of this ultrasonic nebulizer. With conventional electrospray, the formation of the Taylor cone, which is directly responsible for generating the observed "spray" of liquid, is 3692
Analytical Chemistry, Vol. 66, No. 21, November 1, 1994
2.16k
$-
1.62k
-
Diameter (pm)
Flgure 6. Droplet size distributions at (A) 1 and (8) 400 pL/min of 0.1 '30 TFA in water.
prevented when solution conductivity is high. Under these conditions, liquid is observed to jet out from the spray needle orifice in a continuous stream and no visible "spray" results. This phenomenon prevents the use of conductive modifiers in LC mobile phases destined for ESI. Unfortunately, this use of high conductivity mobile phases is deeply entrenched into LC methods, one of the most well-known being the addition of TFA into mobile phases used for peptide and protein separations. With the TFA composition most frequently employed (0.1% TFA in various ACN/water mixtures), the solution conductivity is far too high to allow the use of conventional electrospray. No ion signal can beobserved under these conditions. The conductivity problem is then further compounded by the loss of signal at the high mobile-phase flow rates appropriate for conventional LC methods. To demonstrate the performance of ultrasonic nebulization for ESI with highly conductive mobile phases (TFA in this case), cytochrome c (1 pmol/pL) was dissolved in a solution of 1:l ACN/water, with 0.1% TFA and directly infused at a flow rate of 50 pL/min. Shown in Figure 4C is a scan of this sample, again from m / z 600 to 1400, obtained under these conditions. In this spectrum, the center of the multiply charged peak envelope was somewhat higher, falling at m/z 95 1. Also, the signal magnitude was approximately 20 times less than the signal in Figure 4A,B. The cause of this second and unfortunate result is most likely a chemical effect from the strong ion-pairing behavior of the TFA anion as well as the low pH of the solution. This side effect is purely chemical in nature; the process of ultrasonic nebulization itself is a physical phenomenon and is unaffected by it. Nonetheless, the signal obtained was still adequate enough to give impressive LC/MS sensitivity, as will be shown later, and is considerably better than conventional electrospray, which yields no signal under these conditions. Ultrasonic Nebulization with Mobile Phases Having High Surface Tension. A final, and just as significant, failure of
A
cytochrome c
1600000]
'ime-->
5:OO
10'.00
15'.00
21
B
I
2500-
2000-
I
1500-
1000-
500-
,.n/z
- ->
0
u 800
900
1000
1100
0
1300
L400
1500
1600
1700
Flgure 7. (A) Separation of cytochrome c, albumin, and myoglobin (20 pmol each) on a 1.O mm i.d. LC column. (B) Average of scans under albumin peak.
conventional electrospray is its inability to be used with mobile phases that are high in surface tension and thus mostly aqueous. Again, this shortcoming results from the failure to generate a Taylor cone and subsequent fine spray of droplets when solutions of this type are needed. The use of highly aqueous solvents is also regularly incorporated into many LC applications involving the separation of polar, low molecular weight analytes on reversed-phase columns. To test the ultrasonic nebulizer's performance with these types of solvent systems, cytochrome c (1 pmol/pL) was dissolved in water (with 0.1% acetic acid) and directly infused through the ultrasonic nebulizer at a flow rate of 50 pL/min. A scan from this experiment is shown in Figure 4D. The spectrum appears to be similar to those in Figure 4A,B, but with a slight signal loss of approximately 20%. Following these results, performance over a wide range of organic/water solvent mixtures was investigated as a figure of merit. For this study, cytochrome c was dissolved at 1 pmol/pL in mixtures of ACN in water varying from 10 to 90% (with 0.1% TFA) and directly infused into the source. This range was chosen to simulate conditions that would encompass nearly any reasonable LC gradient program for the separation of proteins. Figure 5 gives the ion signal,
measured as before, across the range of solvent compositions studied. The ion signal was quite steady, indicating that the ultrasonic nebulizer should perform well over an organic/ water gradient program. This result is especially encouraging since both the surface tension and the vapor pressure of the mobile phase changed substantially across the solvent composition range studied here. Droplet Size Measurements. As a final test of this system, the droplet size distribution from an ultrasonically generated spray of a "worst-case" LC mobile phase was determined. For this experiment, a solution of 0.1% TFA (high conductivity) in water (high surface tension) was directly infused into the nebulizer and droplet size measurements taken using a phaseDoppler anemometer. The mean droplet diameter obtained as the liquid flow rate was varied from 1 to 400 pL/min was found to be 11-66pm, with a standard deviation of 0.38. From these data, it is clear that the average droplet size remained relatively constant over this flow rate range. Additionally, panels A and B of Figure 6 show that actual droplet size distributions at 1 and 400 pL/min, respectively, are narrow and nearly identical. Maintaining this narrow distribution over such a broad flow rate range is critical for success, since it is largely the well-known increase in droplet size and droplet Analytical Chemistry, Vol. 66,No. 21, November 1, 1994
3693
size distribution with increasing flow rate that cause the loss of ion signal in conventional ESI-MS. Signal Stability and Injection Reproducibility with Ultrasonic Nebulization. Signal stability was also investigated as a figure of merit. For the first of these experiments, cytochrome c was dissolved in 1:l ACN/water (with 0.1% acetic acid) as before (1 pmol/pL) and directly infused at a flow rate of 50 pL/min for a period of 2.5 h. The total ion current (TIC) for these successive scans varied less than 4% from scan to scan and should be sufficiently stable for any reasonable length of LC gradient program executed. The lack of signal decay with time under the punishing conditions of continual sample influx also suggests that the vacuum region of the source containing the lenses remains clean and free of sample or charge buildup. This system was used for 6 months without any cleaning of the glass capillary tube or vacuum components and therefore strongly implies that the ions are substantially desolvated by the drying/focusing gas system before entering the capillary. To test reproducibility in a manner more meaningful for chromatographic evaluation, 5 pL (5 pmol of sample) of the solution above was injected 10 times as a 1:l (v/v) mixture of water and ACN (with 0.1% acetic acid) was directly infused into the source. For this flow injection analysis experiment, the signal from the m / z 825 peakof cytochrome c was recorded in selected ion monitoring (SIM) mode. For a series of 10 injections, the relative standard deviation of the peak heights was calculated to be 0.7364%. LC/MS Applications: Analysis of a Protein Mixture. In order to test this system for use with LC, a mixture of proteins containing cytochrome c, albumin, and myoglobin (20 pmol each component) was injected onto a 1 mm i.d. C I Scolumn, operated at its optimum flow rate of 50 pL/min, and detected by ESI-MS. For this separation, the gradient was ramped from 30 to 90% ACN in water (with 0.1% TFA) over 30 min following a linear program. Figure 7A shows the TIC for this example. The m / z scale was scanned from 600 to 1800 for each data point shown. Figure 7B shows the average of scans which appear underneath the albumin peak in the TIC from Figure 7A. Sensitivity and linear dynamic range (LDR) were then investigated with this system by injecting cytochrome c alone in increasing amounts and under the same chromatographic conditions as listed above. Both scan and SIM mode were studied, and sample quantities ranging from 32 fmol to 12 500 pmol, increasing by factors of 5, were used. For scan mode, the MS conditions were identical to those for the protein application discussed above. For SIM mode, the m / z 825 peak of cytochrome c was detected with a dwell time of 100 ms. Using scan mode, as little as 4 pmol of sample could be detected above the TIC background noise (S/N = 3). Signal response linearity (0.98 correlation) was maintained until 12 500 pmol was injected, at which time the column became overloaded, exhibiting wide poorly shaped peaks with splitting and ghost shadows. Using SIM, the lowest amount of cytochrome c which could be detected was 32 fmol at S/N = 3. Signal response was again nearly linear with a 0.98 correlation until 12 500 pmol was injected, when the column was again overloaded. These scan and SIM data are combined and shown in Figure 8. 3694
Analytical Chemistry, Vol. 66, No. 21, November 1, 1994
100000000
10000000
m
6
m
looooo 10000
1000
10
I t
/
1
+
/
/
i. 001
01
10
1
-
100
I000
10000 100000
Amount Injected (pmol) SIM
--C
SCAN
Figure 8. Signal (as peak height) of cytochrome c injected on a 1.O mm i.d. column from scan and SIM acquisition modes vs amount injected.
Finally, in order to demonstrate the capabilities of LC/ MS with ultrasonically assisted ESI at even higher LC flow rates, the same protein mixture (20 pmol each component) used above was injected on a 2.1 mm i.d. LC column operated at its optimum flow rate of 200 pL/min. For this column, the gradient was ramped from 30 to 90% ACN in water (with 0.1% TFA) in 15 min to optimize the separation. Panels A and B of Figure 9 show the TIC and scans for the albumin peak as before. The sensitivity for the 2.1 mm i.d. column was somewhat less than that observed for the 1.0 mm i.d. column. This is as expected and is due to sample dilution in the larger column and increased axial diffusion effects.
CONCLUSIONS This work has demonstrated the performance of a new type of ultrasonic nebulizer designed specifically for use with LC/ESI-MS. The device is capable of producing the fine spray of droplets necessary for electropspray ionization by a purely mechanical means, which is less dependent on the influence of applied electrical fields than is unassisted ESI. For this reason, LC mobile phases, which are not normally amenable to electrospray, may be used for LC/ESI-MS, thus extending the range of suitable LC mobile-phase compositions to include high flow rate, high electrical conductivity, and high surface tension solutions. This set encompasses the bulk of LC applications that are currently impossible or very difficult to accomplish with conventional unassisted ESI. Operational variables such as the nebulizer’s vibrational frequency, the drying and focusing gas flow rates, and the electrical potential on the cylindrical electrode were systematically investigated to achieve optimum performance. The effect of high potentials on the cylindrical electrode was of particular interest and only possibledue to the insulating effect of a liquid layer present on the tip of the nebulizer. We hope that this effect may find additional use by allowing the
A
bundance
a 1bumi n
800000 1
700000
1
600000
-
500000
-
myog 1obin
cytochrome c
400000 -
300000
-
100000 -
200000
o
m
ime-->0.00
B
, Z.'OO
-
4.'00
-00
6.bO
16.00
ZEidance J
900
-
0
L L500
1600 1700
Flgure 9. (A) Separation of cytochrome c, albumin, and myoglobin (20 pmol each) on a 2.1 mm i.d. LC column. (B) Average of scans under albumin peak.
application of higher fields in general with ESI. One particular area where this would be of value is with the generation of negative ions with ESI, where corona discharge frequently inhibits stable spray formation. Performance of the ultrasonic nebulizer was evaluated across a wide range of organic solvent compositions and sample flow rates and found to be fairly constant in terms of providing consistent MS signal levels for a common protein, cytochrome c. Also, a series of loop injections in a flow injection analysis experiment gave MS peak heights with RSDs of less than 1%. In an application demonstration, LC/MS examples of protein separations on 1 and 2.1 mm i.d. columns (with no splitting) were shown. In the 1 mm Column case, the minimum detectable amount of cytochrome c was found to be 32 fmol in SIM mode.
Finally, although largely successful for its intended purpose, this system posed additional challenges which must now be addressed. Specifically, mobile phases with TFA added (for protein separations on reversed-phase columns) could easily be electrosprayed, but a signal loss of approximately 20-fold was observed.
ACKNOWLEDGMENT This work was supported by Grant 2 R44 RR07528-02 from the National Institues of Health. Received for review March 23, 1994. Accepted JUIY 7, 1994." e Abstract
published in Advance ACS Abstracts. September 1, 1994.
Analytical Chemistry, Vol. 66, No. 21, November I , 1994
3695
