ionization time-of-flight


Graphite surface-assisted laser desorption/ionization time-of-flight...

9 downloads 380 Views 730KB Size

Anal. Chem. 1995, 67,4335-4342

Graphite Surface-Assisted Laser Desorption/ Ionization Time-of-Flight Mass Spectrometry of Peptides and Proteins from Liquid Solutions Jan Sunner,* Edward Dratr, and Yu=Chie Chen

Department of Chemistry and Biochemistry, Montana State University, Bozeman, Montana 59717

Laser desorption time-of-flightmass spectra of peptides and proteins, as well as of lower molecular weight analytes, have been obtained by using a pulsed nitrogen UV laser (337 nm) to irradiate mixtures of 2-150 pm graphite particles and solutions of the analytes in glycerol. Protonated analytes as well as abundant alkali cation adducts were observed. Carbon cluster ions, Cn+,typically had a low abundance but dominated the mass spectrum at elevated laser powers. In spectra of a cytochrome c tryptic digest, all but one of the tryptic peptides were easily observed. Spectra of low molecular weight analytes dissolved in glycerol are very similar to FAB spectra of the same glycerol solution with added alkali salts. However, in many peptide and protein spectra, glycerol ion abundances are very low, and the alkali ions dominate the spectra at low mass. These spectra may correspond to wet and dry surface desorption conditions, respectively. The best spectra of the larger molecules were observed under dry conditions. In these initial experiments, we have obtained a sensitivity in the pico- to nanomole range and a mass resolution of about 300. The signal intensity is as good as that in conventional MALDI, and under optimal conditions, few background peaks appear, even at low mass. In the last few years, matrix-assisted laser desorption/ionization

(MALDI), l ~ zand electrospray (ES)3have revolutionized biochemical mass spectrometry. However, the history of the efforts to extend mass spectrometry to involatileand thermally labile species is by now a very long one. Very early steps were taken with field ionization and field de~orption.~These efforts continued with P D M W and static SIMS.7-9 The introduction of a liquid solution (glycerol) into the SIMS experiment, Le., FAB,'O more a g propriately called liquid SIMS (LSIMS), started the biochemical (1) Karas, M.; Bachmann, D.; Bahr, U.; Hillenkamp, F. Int. J. Mass Spectrom. Ion. Processes 1987,78, 53-68. (2) Karas, M.; Ingendoh, A; Bahr, U.; Hillenkamp, F. Biomed. Enoiron. Mass Spectrom. 1989,18, 841-843. (3) Whitehouse, C. M.; Dreyer, R. N.; Yamashita, M.; Fenn, J. B. Anal. Chem. 1985,57, 675-679. (4) Beckey, H. D. Principles of Field Ionization and Field Desorption Mass Spectromety; Pergamon Press: Oxford, 1977. (5) Torgerson, D. F.; Skowronski, R P.; MacFarlane, R D. Biochem. Biophys. Res. Commun. 1974,60, 616. (6) Sundqvist, B. U. R.; MacFarlane, R D. Mass Spectrom. Reo. 1985,4 , 421. (7) Benninghoven, A 2. Phys. 1970,230,403. (8) Grade, H.; Cooks, R G. /. Am. Chem. SOC.1978,100, 5615. (9) Pachuta, S. J.; Cooks, R. G. Chem. Rev. 1987,87, 647-669.

0003-2700/95/0367-4335$9.00/0 0 1995 American Chemical Society

mass spectrometry revolution. The upper mass limit of LSIMS restricted its use to peptides and very small proteins. The upper mass limit was then increased by more than an order of magnitude by MALDI and ES. As a result, these latter ionization methods have had a tremendous impact on biochemical mass spectrometry in the last few years. Of the two methods, MALDI follows directly in the tradition of the earlier desorption/ionization @I) methods, namely PDMS, SIMS, and FAB. The difference, of course, is that a laser is used to energize the matrix instead of a particle beam. In MALDI, the matrix must fulfill all the (poorly understood) requirements for a successful desorption/ionization process in addition to absorbing the laser radiation. These combined requirements have severely limited the number of useful matrix compounds. Indeed, the matrix selection is largely a matter of trial and error, and no clear guidelines seem to have emerged." At this point, it seems that all commonly used MALDI matrices are solids.12 The use of solid matrices has several undesirable consequences. For example, the drying and crystallization process is harsh and may result in protein denaturation, aggregation, and subunit dissociation. Further, crystallization causes the unpredictable phenomenon of sweet-spots in MALDI.l2 Finally, the use of a solid matrix makes it more difficult to interface electrophoresis or HPLC to MALDI.13-16 For these reasons, it is worthwhile to develop laser desorption of liquid solutions. Several possible approaches to such a goal have been demonstrated. First, W-absorbing liquid matrices may be used. For example, 3-nitrobenzyl alcohol (3-NBA) has been used for this p u r p o ~ e . ' ~ - ' ~ Problems with this method seem to be low mass resolution and a high chemical b a c k g r o ~ n d . ' ~Also, , ~ ~ the number of different W-absorbing liquid matrices seems rather limited. A second, possibly more promising approach, is to dissolve UV-absorbing dyes, such as Rhodamine 3G or 3-NBA, in a non-UV-absorbing (10) Barber, M.; Bordoli, R. S.: Sedgewick, R D.; Tyler, A N. J. Chem. SOC., Chem. Commun. 1981,325. (11) Fitzgerald, M. C.; Parr, G. R: Smith, L. M. Anal. Chem. 1993,65, 32043211. (12) Vorm, 0.;Roepstorff, P.; Mann, M. Anal. Chem. 1994,66, 3281-3287. (13) Warren, W. J.; Cheng, Y.-F.; Fuchs, M. LC-GC 1994,12,23-28. (14) Vanveelen, P. A; Tjaden. U. R.; Vandergreef, J.; Ingendoh, A; Hillenkamp, F. J. Chromatogr. 1993,647, 367-374. (15) Weinmann, W.; Parker, C. E.; Deterding, L. J.; Papac, D. I.; Hoyes, J.; Przybylski, M.; Tomer, K B. J Chromatogr. A 1994,680, 353-361. (16) Elicone, C.; h i , M.; Geromanos, S.; Erdjumentbromage, H.: Tempst, P. /. Chromatogr. A 1994,676, 121-137. (17) Li, L.; Wang, A. P. L.; Coulson, L. D. Anal. Chem. 1993,65, 493-495. (18) Yau, P. Y.; Chan, T. W. D.; Cullis, P. G.; Colbum, A W.; Derrick, P. J. Chem. Php. Lett. 1993,202, 93-100. (19) Chan, T. W. D.; Thomas, I.; Colbum, A W.; Derrick, P. J. Chem. Phys. Lett. 1994,222, 579-585.

Analytical Chemistty, Vol. 67, No. 23, December 1, 1995 4335

solvent like glycerol; with this method, a mass resolution of about Indeed, the graphite particles used are as large as or larger than 50 has been reported.20 The use of dyes like Rhodamine allows the laser focus. As we will see, the resulting mass spectra show for the use of visible laser lightz1 In a third approach, a liquid clear characteristics of a surface desorption process. We here matrix was dripped onto a fine fibrous paper to obtain highquality refer to this method as graphite surface-assisted laser desorption/ laser desorption mass spectra of porphyrins.22 It is remarkable ionization (of liquid solutions), or graphite SALDI. that one of the liquids (15crown-5) used by the authors does not EXPERIMENTAL SECTION absorb the UV light but still gave good spectra. A fourth approach The experiments were performed on two different timeof-flight is to use IR-MALDI.23-25For example, the 3 pm light from an mass spectrometers: a Perceptive LaserTec Benchtop I1 reflecting Er-YAG laser is absorbed by the 0-H stretch in liquids such as instrument with a 1.3 m vertical fight tube and a Vestec 2000 water and glycerol. Thus, Hillenkamp et al. demonstrated a linear instrument with a 1.4 m horizontal flight tube (both from spectrum of lysozyme at 14 300 Da from a glycerol solution.24 Perceptive Biosystems, Framingham, MA). Each mass spectromA fifth approach toward laser desorption of liquids was eter uses a 337 nm nitrogen laser (CVI Laser, Newton, MA) for presented by Tanaka et al. in a noteworthy paper as early as 1988.26 desorption. The LaserTec instrument was used in the linear mode These researchers suspended 300 A diameter cobalt particles in only. The acceleration voltage was 20 kV, and the vacuum glycerol and obtained spectra of proteins as large as chymotrypsipressure in the ion source was -(1-2) x Torr. All spectra nogen (25 717 Da). The mass resolution in the spectra was shown were obtained in the positive ion mode. The spectra were relatively poor (about 20), and as many as 500 nitrogen laser shots mass calibrated and plotted using Perceptive software and were required to obtain peaks with a relatively low signal-to-noise GRAMMS (Galactic, Salem, NH). Unless otherwise noted, mass ratio. (However, as with other liquid desorption results discussed calibration was internal, and two known peaks in each spectrum here, it is d a c u l t to decide to what extent low mass resolution were used as calibration points. and low sensitivity were due to inherent problems with the Glycerol (Gl) and diethanolamine @EA) were obtained from desorption process or to problems with the mass spectrometer Aldrich; bradykinin, angiotensin, cytochrome c, and myoglobin design.) The microscopic metal particle approach to laser defrom Sigma; and TPCK-treated trypsin (88.3%)from Worthington. sorption does not appear to have been developed further. The G1 and DEA were degassed by pumping over the liquid while Tanaka et al. suggested that the desorption mechanism in their it was immersed in a bath sonicator for about 1 h. Two types of experiments involved a heating of the extremely small metal graphite flakes were tested in these experiments. Unless otherparticles, followed by heat conduction to the surrounding glycerol wise specified, a graphite preparation used for lubrication and liquid,26 Le., a thermal desorption/ionization m e c h a n i ~ m . ~ ~ - ~ ~ machining purposes @ixon No. 635; Lubricating Flake Graphite, Indeed, for these experiments, there seems to be no reasonable Jersey City, NJ) was used. A synthetic graphite sample obtained alternative mechanistic explanation. The Tanaka results therefore from Aldrich was found to give equally good mass spectra and indicate that proteins can be desorbed in a purely thermal was used for one of the experiments presented here. The Dixon desorption process. graphite particle size distribution was determined by optical The experiments reported here were performed in an effort microscopy to be in the 10-150 pm range, whereas the Aldrich to improve mass resolution and sensitivity in UV laser desorption flakes were considerably smaller, -2 pm. Both graphite samples from liquid solutions. We report results on neat glycerol and were studied by scanning electron microscopy, and both samples diethanolamine and on biomolecules dissolved in glycerol. Relawere found to contain reasonably smoothly shaped flakes. Both tively large graphite particles were used to couple the W laser samples also contained a large fraction of particles with a much energy into the liquid solution. The large size of the graphite more complex appearance and with surface structures in the particles used is significant. Tanaka et aLZ6used particles of nearsubmicrometer range. The size of the highest-intensity focal spot atomic dimensions, whose diameter was smaller (300 A) than the of the laser beam was determined by measuring bum size with a wavelength of the light used. When these are suspended in scanning electron microscope. A small piece of poly(viny1idine glycerol, a “bulk desorption process is expected to result upon ditluoride) (PVDF)membrane was attached to the sample plate, UV laser irradiation. In contrast, we are using particles that are and the laser was fired as the membrane was moved in order to many orders of magnitude larger, Le., in the 10-150 pm size separate individual laser spots. The surface damage region was range, except that 2 pm particles were used in one experiment. clearly seen in the scanning electron microscope when higher (20) Comett. D. S.; Duncan, M. A; Amster, I. J. Anal. Chem. 1993,65, 26082613. (21) Comett, D. S.; Duncan, M. A; Amster, I. J. org. Mass Spectrom. 1992,27, 831-832.

(22) Kim, Y. L.; Zhao, S.; Sharkey, A G.: Hercules, D. M. Mikrochim. Acta 1994, 113, 101-111. (23) Overberg, A.; Karas. M.; Bahr, U.; Kaufmann, R.; Hillenkamp, F. Rapid Commun. Mass Spectrom. 1990,4,293-296. (24) Overberg, A,: Karas, M.; Hillenkamp. F. Rapid Commun. Mass Spectrom. 1991,5, 128-131. (25) Strupat, K: Karas, M.; Hillenkamp, F.; Eckerskom, C.; Lottspeich, F. Anal. Chem. 1994,66. 464-470. (26) Tanaka, IC;Waki, H.; Idao, Y.; Akita, S.; Yoshida, Y.: Yoshida, T. Rapid Commun. Mass Spectrom. 1988,2,151-153. (27) Beuhler, R. J.: Friedman, L. Int. J. Mass Spectrom. Ion. Processes 1987,78, 1-15

(28) MacFarlane, R. D.; Torgerson, D. F. Phys. Reo. Lett. 1976,36,486. (29) Sunner, J.; Ikonomou, M. G.; Kebarle, P. Int. J. Mass Spectrom. Ion. Processes 1988.82, 221-237.

4336 Analytical Chemistry, Vol. 67, No. 23, December 1, 1995

laser powers were used and the spot diameter was -10 pm. Two different sample preparation methods were used, and they will be referred to as methods A and B. In both methods, the peptides or proteins were first dissolved in a volatile solvent like methanol or water. In method A, 0.5 p L of G1 was applied to the well or probe tip, and 1p L of the peptide or protein solution was added from a pipet. The graphite powder was then dropplaced on top of the liquid with a spatula in an amount sufficient to cover the liquid surface. For best results, the graphite particles should not be immersed in the liquid but should remain on the surface such that the surface of the powder has a dry appearance. Method A was used throughout this paper, unless otherwise noted. In sample preparation method B, 5 vol % of G1 was added to methanol. An approximately equal volume of (dry) graphite powder was mixed with the Gl/methanol solution using a vortexer.

01

220

420

703

mlz Figure 1. Graphite surface-assisted laser desorption mass spectrum (337 nm) of glycerol. Average of 82 laser shots.

One microliter of the slurry was pipetted into the probe well, where the methanol evaporated after a few seconds. One microliter of a solution of the analyte in either methanol or water was subsequently added to the well. M e r 1-2 min in air at room temperature, the slurry had a dry appearance, and the probe was inserted into the ion source. Though very similar spectra were obtained, sample preparation method B was found to have important advantages over method A Fist, in method A, good analyte signals tended to be obtained only in localized “sweet-spots” that were dif6cult to find. In method B, the sweet-spots phenomenon was much less pronounced, it was much easier and quicker to find good spectra, and the ion signals were more persistent. Second, on the Vestec 2000, the probe tip surface is vertical, and part of the liquid sample prepared by method A was often lost by gravity as residual dissolved air caused bubble formation in the liquid sample. Occasionally, the liquid also flowed into a narrow space between the cylindrical probe tip and a surrounding electrode by capillary action. m e r e was no such problem with the Laseflec instrument because it has a horizontal sample tray.) In contrast, the sample prepared by method B does not flow. Third, with method B, the slurry aggregated into a thin, solid mass, and even after drying there was little or no loose powder. In contrast, method A resulted in a dustlike sample after drying. This difference may be significant because of the possibility that small graphite particles will detach from the sample and stick to electrodes or other critical components in the ion source and cause discharges or other problems. It is important to note, however, that we have not yet experienced any adverse effects of introducing the graphite powder into the ion sources, though we have mostly used method A. Method B is a more recent development, and protein spectra have so far been obtained only with sample preparation method A.

The laser power required to obtain analyte spectra varied considerably. Though signal intensities generally increased with increasing laser power, the signal threshold was not as pronounced as that typically observed in MALDI. All the spectra shown here were obtained at a laser power setting close to that typically used in MALDI, unless otherwise stated. RESULTS AND DISCUSSION The low-mass region of a graphite surface-assisted laser desorption/ionization (SALDI)spectrum of G1 is seen in Figure 1. All major peaks are easy to identify. Thus, the ion at m/z = 93 is protonated G1. This ion and the fragments at m/z = 75, 57, 45,31,29, and 19 are all prominent in the FAB spectra of G1. The intense peaks at m/z = 23, 39, and 133 are due to Na+, K+, and Cs+, respectively. Ubiquitous alkali salt contaminants were observed in all of our spectra. However, the abundances of the different ions, in particular of Cs+, varied considerably between different experiments. The peaks at m/z = 115,131, and 225 are due to alkali ion/G1 adducts, Na+(Gl), K+(Gl), and Cs+(Gl), respectively. A series of lower-intensity peaks due to carbon clusters, C,H,+, with n = 1-25 and beyond and with m = 0 to about 4, is also observed. The spectrum in Figure 1 is, with the exception of the carbon clusters, very similar to FAB spectra of solutions of alkali halide salts in GL30 One major exception is that the G1 dimer at m/z = 185 was not observed in the graphite SALDI spectrum of G1. Though a low dimer abundance may result from high analyte or salt concentration^,^^ it is likely that the plasma temperature in the laser desorption experiment is higher than that in FAB and that the G1 ions are formed with a higher internal excitation energy. In this context, it is noteworthy that there is a pronounced asymmetry in the K+ and Cs+ ion (30) Sunner, J. 1.Am. SOC.Muss Spectrom. 1993,4, 410-418. (31) Sunner, J.; Morales, A.; Kebarle, P. Anal. Chem. 1988,60,98-104.

Analytical Chemistry, Vol. 67, No. 23, December 1, 1995

4337

8oool 6000

4000

I1

R

1 I

i

I *OOOj/

63

163

308

500

mh Figure 2. Graphite surface-assisted laser desorption mass spectrum (337 nm) of diethanolamine. Average of 67 laser shots.

peaks. This peak asymmetry can be explained by extensive metastable decomposition, in this case declustering of cationized species,

M+(Gl) = M+ + G1

(1)

Graphite SALDI can be used with liquids other than G1. As an example, Figure 2 shows a graphite laser desorption spectrum of DEA. As in the previous spectra, Na+ and K+ are abundant ions. In contrast, Cs- ions are absent. Further, DEA ions are seen at m/z = 106 @EM’), 128 (Na+@EA)), and 144 @EA)), and DEA fragment peaks are seen at m/z = 56,74, and 88 that are also observed in FAB. Again, the graphite laser desorption spectrum is very similar to what is observed in FAB for a DEA alkali halide solution. Figure 3 shows a graphite SALDI spectrum of a 0.25% DEA solution in G1 . Despite the relatively low analyte concentration, protonated DEA at m/z = 106 is seen to dominate over protonated G1 at m/z = 93. In FAB, proton transfer from G1 to a more basic analyte occurs during the desorption e ~ e n t . The ~ ~ ,gas-phase ~~ basicity (at 300 K) of DEA at 220 kcal/mol is significantly higher than the gas-phase basicity of G1 at 196 kcal/m01.~~Thus, the proton transfer reaction,

e+-

G1H’

+ DEA = DEAH’ + G1

(2) is exoergic. The same reaction should occur in SALDI. However, in the SALDI spectrum, the abundance of DEAH+ relative to GIH+ is about 10 times that observed in FAB for the same DEA concentration. This is not surprising, considering that the volume of the high-density selvedge (desorption plume), where reaction 2 occurs, is likely much larger in laser desorption than in FAB, (32) Sunner, J.: Kulatunga, R.; Kebarle, P. Anal. Chem. 1986,58, 1312.

4338 Analytical Chemistry, Vol. 67, No. 23, December 7, 7995

resulting in a more complete reaction. In contrast, cationized G1 ions [Na+(Gl) at m/z = 115 and K+(Gl) at m/z = 1311 are much more abundant than cationized DEA ions [Na’OEA) at m/z = 128 and K+@EA) at m/z = 1441 as expected in the absence of reactive coupling. One reason for this is likely that the Na+ and K+ ion affinities of G1 and DEA are close. In the low-mass region of the SALDI spectra, close to unit mass resolution is obtained. In addition, SALDI spectra are considerably less congested in the low-mass region than MALDI spectra. Unit resolution, together with a relatively low chemical background, would seem to make S U I suitable for the analysis of low molecular weight compounds, including smaller peptides in proteolytic digests. Figure 4 shows a spectrum of the same DEA/Gl solution as in Figure 3, except that the laser power used is significantly higher. The most notable feature of this spectrum is the high abundance of the C,H,+ carbon cluster ions. In this spectrum, the cluster sequence can be followed up to n x 30, and peaks that differ only in the number of hydrogens are resolved up to n x 13. It is noteworthy that the larger carbon clusters have fewer hydrogen atoms and also that every fourth cluster (C11+, CIS’, C19+,etc.) is somewhat more abundant. A number of G1 peaks are also seen in Figure 4, including GlH+ at m/z = 93 and the Cs+(Gl) peak at m/z = 225, as well as the usual G1 fragments. All these ions need not coexist. The “cold” glycerol ions may well originate from a different region (peripheral) within the laser focus than the “hot” carbon cluster ions, where the laser pulse energy density is lower. Graphite SALDI is suitable for the analysis of biomolecules. M solution of Figure 5 shows a spectrum of a 1.0 x bradykinin in G1 (1 pmol sample). The inset shows the highmass region in greater detail. The major peaks are due to

30000

20000 Q)

$CI

z

3

9

10000

0 0

58

146I

1

273I

II

440

m/z Figure 3. shots.

30000

iraphite surface-assisted laser desorption mass spectrum (337 nm) of a solution of 0.25% (vh) DEA in GI. Average of 107 laser

a,

a

20000 Q)

$ 0

z

II

9

i

10000

0 58

148

zn

440

I

m/z Figure 4. Graphite surface-assisted laser desorption mass spectrum (337 nm) of 0.25% DEA in GI at an elevated laser power setting. The spectrum is dominated by carbon clusters, see text. Average of 96 laser shots.

protonated bradykinin at m/z = 1061 and to adducts of the peptide with Na+, K+, and Cs+ at m/z = 1083,1099, and 1193. The mass resolution for these peaks, measured at half-height, is about 300. In the low-mass region, the alkali and G1 ions are present. In

addition, a few even-mass ions, notably at m/z = 60, 70, and 112, now stand out. Presumably, these are nitrogen-containing, evenelectron fragment ions derived from the peptide. Mass spectra of several other peptides and their mixtures have been obtained, Analytical Chetnistty, Vol. 67, No. 23, December 7 , 1995

4339

0

20000

r

15000

Q)

0

6

10000

r

c

2a n

a

a

r

g 1

500C

T

(

Figure 5. Graphite surface-assisted laser desorption mass spectrum of bradykinin. Average of 121 laser shots. Internal mass calibration using Cs+ and Cs+(bradykinin). Inset shows the bradykinin pseudomolecular region on an expanded scale.

260

460

660

ab

loo0

1io0

-~ 1400

1do0

Mass (mh)

Figure 6. Graphite SALDI spectrum of tryptic digest of cytochrome c. Average of 51 laser shots and external mass calibration using bradykinin. Twelve out of sixteen dipeptides or larger are seen in the spectrum. Inset shows part of spectrum of the same sample obtained using sample preparation method B and 2 p m graphite flakes, see text.

i.e., angiotensin, methionine enkephalin, and melittin, as well as cytochrome c tryptic peptides, as shown next. Figure 6 shows a graphite SALDI mass spectrum of a tryptic digest of cytochrome c (800 pM). Sodiated peptides, Na+(M), dominate the mass spectrum due to the addition of a sodium salt. Of 16 tryptic peptides with two or more amino acids, 12 are easily identified in this spectrum. The inset shows part of a spectrum 4340

Analytical Chemistry, Vol. 67, No. 23, December 1, 1995

obtained with sample preparation method B. The two peaks are due to the sodiated and potassiated 80-86 peptide. The 14-22 peptide with the bound heme at -1634 Da, as well as the other two missing peptides, appears in the same spectrum. The mass accuracy and sensitivity of any new or modified mass spectrometry method are fundamental measures of the usefulness of the method. In our experience, the mass accuracy in graphite

1

20000

F:

i3

3

I 804

3288

7453 -J

132SS

~

I

m/z Figure 7 . Graphite surface-assisted laser desorption mass spectrum of cytochrome c. Average of 21 laser shots. Internal mass calibration using Cs+ and Cs+(cytochrome c).

SALDI is roughly equal to that of MALDI and -0.1% or better for an external mass calibration. As with MALDI, the mass accuracy can probably be improved by using improved sample preparation method^.^^,^^ With regard to sensitivity, close to 1nmol of peptide was used for the peptide spectra shown in this paper. In MALDI, it is common to use about 1pmol (1p L at 1pM) of peptide, but it is often possible to decrease this amount further.12 By decreasing the peptide concentration in our SALDI experiments, we have found that the amount of peptide could be decreased to -10 pmol by using 10-5 M peptide solutions without any other changes in the sample preparation procedure. At lower peptide concentrations, the signal-to-noise ratio of the molecular ion peaks becomes too low. Experiments are currently underway both to explore the factors that determine the SALDI sensitivity and to further increase the sensitivity. It is possible that the ionization efficiency in SALDI is inherently lower than that in MALDI, but at this point that would be pure speculation. However, with sample preparation method B, the SALDI sample area can easily be decreased because peptide signals are reliably found everywhere on the sample. Further, it is likely that most of the analyte is trapped in the -40 pm thick graphite layer. For these reasons, there is a very good potential to increase SALDI sensitivity considerably in the future by ensuring maximum liquid solvent access to the laser focus site. Graphite SAZDI can be used to obtain also protein spectra. An example is shown in Figure 7 for a 800 pM cytochrome c solution in glycerol. In the pseudomolecular ion region, the Cs+protein adduct gives a clearly separated peak. However, the mass peaks due to the protonated protein and to the Na+ and K+ protein adducts are not resolved, and the resulting composite mass peak (33) Vorm, 0.;Mann, M. J. Am. SOC.Mass Spectrom. 1994, 5, 955-958.

is rather broad. The mass resolution, measured on the Cs+ adduct peak, is again about 300. Though we have repeatedly obtained graphite SALDI spectra of cytochrome c and myoglobin using sample preparation method A, one usually has to search over the sample surface for a considerable time to obtain spectra such as that shown in Figure 7. In addition, such spectra are obtained for only about 1s. Clearly, very special conditions on the sample surface must be fulfilled in order to obtain good protein spectra. We do not yet understand what those special conditions are. It seems likely that once these conditions are understood, it should be possible to create them reproducibly in order to obtain protein spectra from liquid solutions. In the mass spectrum of cytochrome c in Figure 7, the alkali ions, particularly Cs+, dominate at low mass, and the G1 peaks are either absent or hidden in the noise. Indeed, we have observed a clear correlation between good peptide and protein spectra and low-intensity G1 peaks. It seems unlikely that nearly complete ionization transfer from solvent (Gl) to analyte (protein) would occur. For example, in analogy with FAB, one would expect, in such a case, to still see the low-mass G1 fragment ions at m/z = 29 and 3L31 The explanation for the absence of G1 ions is likely different. Instead of being solvated by G1, the protein and peptide molecules may be deposited on the graphite surface after G1 has evaporated. The macromolecules would then be laserdesorbed in a more or less dry state directly from the graphite surface. This may also explain why the protein signals are shortlived, since the intense UV laser light may well destroy the protein or desorb all the protein available on the surface. To confirm that both GI and graphite are essential to obtain SALDI spectra, we conducted a number of control experiments. We were indeed able to desorb bradykinin, deposited from a Analytical Chemistry, Vol. 67, No. 23, December 1, 1995

4341

methanol solution, from dry graphite without adding G1 (results not shown). The same protonated and cationized bradykinin ions appeared, but ion intensities were much lower and the mass resolution was poor as compared to when G1 was present. In addition, the ion signals were very short-lived. With a G1 solution, only on the stainless steel probe, i.e., in the absence of graphite, we were not able to obtain either G1 or peptide spectra. With peptides deposited on the stainless steel surface from a methanol solution, we were occassionally able to obtain low-quality, shortlived spectra of bradykinin and angiotensin I. These results indicate that both G1 and graphite are indeed important to the SALDI desorption process(es). It is not easy to see how the G1 (or a similar liquid solvent) can be that important if, as argued above, the peptides and proteins are desorbed from a (nearly) dry state on the graphite surface. We can only speculate that a small residual amount of G1, associated with the peptide or protein, is beneficial for the desorption process. It is also possible that the ability of the liquid G1 to mediate analyte transport may play an important role, as is the case in Desorption of macromolecules directly from solid surfaces may have important advantages. For example, the sensitivity is potentially very high, since all the analyte molecules could possibly be deposited on the surface and desorbed. Also, the mass resolution can potentially be very high, since all the ions can be made to originate from a well-defined plane with a well-defined electrical potential. For example, a mass resolution of over 10 0o0 (34) Kriger, M. S.; Cook, K. D.; Short, R. T.; Todd, P. J. Anal. Chem. 1992, 64, 3052-3058.

4342 Analytical Chemisfry, Vol. 67, No. 23,December 7, 7995

can be achieved in SIMS, where ions do originate from a welldefined, flat surface. In conclusion, we have demonstrated that laser desorption spectra of smaller analytes, peptides, and proteins can be obtained from liquid (glycerol) solutions of the analytes. Protonated analytes are observed, but the most intense peaks are typically due to alkali ion adducts. The method seems particularly suited to the analysis of peptide mixtures (for example, proteolytic digests). However, more work is required to improve sensitivity. Curiously, two distinct desorption mechanisms seem to be involved. Smaller analytes appear to be desorbed directly from the solution, and the similarity with FAB spectra is very striking. Proteins and peptides are originally dissolved in the glycerol; however, they seem to actually be desorbed from the graphite surface with at the most a very small amount of liquid (glycerol) present. It may well be that separate optimum experimental conditions can be found for either of these processes, and either process holds the promise of unique advantages. ACKNOWLEDGMENT

The present work was supported by NIH-idea and NSFEPSCoR We thank Joe Sears for valuable help. Received for review March 15, 1995. Accepted July 1995.B

AC950263P Abstract published in Advance ACS Abstracts, October 15, 1995

20,