Anal. Chem. 2004, 76, 2958-2965
Technical Notes
Suppression of r-Cyano-4-hydroxycinnamic Acid Matrix Clusters and Reduction of Chemical Noise in MALDI-TOF Mass Spectrometry I. P. Smirnov,* X. Zhu, T. Taylor, Y. Huang, P. Ross, I. A. Papayanopoulos, S. A. Martin, and D. J. Pappin
Applied Biosystems, 500 Old Connecticut Path, Framingham, Massachusetts 01701
Progress in high-throughput MALDI-TOFMS analysis, especially in proteome applications, requires development of practical and efficient procedures for the preparation of proteins and peptides in a form suitable for high acquisition rates. These methods should improve successful identification of peptides, which depends on the signal intensity and the absence of interfering signals. Contamination of MALDI samples with alkali salts results in reduced MALDI peptide sensitivity and causes matrix cluster formation (widely reported for CHCA matrix) observed as signals dominating in the range below m/z 1200 in MALDI spectra. One way to remove these background signals, especially for concentrations of peptides lower than 10 fmol/µL, is to wash matrix/sample spots after peptide cocrystallization on the MALDI plate with deionized water prior to analysis. This method takes advantage of the low water solubility of the CHCA compared to its alkali salts. We report here that the application of some ammonium salt solutions, such as citrates and phosphates, instead of deionized water greatly improves the efficiency of this washing approach. Another way to reduce matrix cluster formation is to add ammonium salts as a part of the MALDI matrix. The best results were obtained with monoammonium phosphate, which successfully suppressed matrix clusters and improved sensitivity. Combining both of these approachessthe addition of ammonium salts in the CHCA matrix followed by one postcrystallization washing step with ammonium buffers provided a substantial (∼3-5-fold) improvement in the sensitivity of MALDI-MS detection compared to unwashed sample spots. This sample preparation method resulted in improved spectral quality and was essential for successful database searching for subnanomolar concentrations of protein digests. In the past 15 years, mass spectrometry has become a major analytical tool for the analysis of biomolecules. In addition to the information about molecular mass of the analyte, it is able to * To whom correspondence should be addressed. E-mail: smirnoip@ appliedbiosystems.com. Fax: (508) 383 7883.
2958 Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
perform analysis at high speedswith the rate of a few seconds per sample. The demand for high analysis speed is especially important in the field of proteomics, where one of the major tasks is identification of proteins by peptide mass fingerprinting, as described in the review by Jungblut and Thiede.1 The success of this step is defined mostly by the quality of the mass spectra obtained. In its turn, spectral quality not only depends on the mass spectrometer itself but is also impacted greatly by the method of sample preparation, MALDI matrix composition, and possible sample impurities. For example, traces of metal salts or the presence of surfactants may complicate MALDI analysis. Generally, however, MALDI-MS is less demanding with respect to the purity of the sample than electrospray ionization techniques.2 The presence of metal salts decreases the sensitivity of MALDI and often reduces spectral resolution (especially for oligonucleotide analysis). As a result, methods of sample purification are the subject of careful investigation by many groups in the mass spectrometry field. There are a number of available desalting methods described in the literature, such as desalting concentration on reversed-phase microcolumns3,4 or macroporous polystyrene beads,5 as well as various immobilization techniques on different membranes,6 ion exchange polymers (PEI and PAA7,8), and nitrocellulose.9 Because of the difference in solubility, salts mostly crystallize around the outer surface of the matrix crystals, whereas molecules of analyte are more or less evenly distributed throughout the crystals. Because of this difference in distribution between analyte and salts, the quality of the spectrum can often be improved by simple washing of the sample-matrix crystals on the MALDI plate (1) Jungblut, P.; Thiede, B. Mass Spectrom. Rev. 1997, 16, 145-162. (2) Roepstorff, P.; Klarskov, K.; Andersen, J.; Mann, M.; Vorm, O.; Etienne, G.; Parello, J. Int. J. Mass Spectrom. Ion Processes 1991, 111, 151-172. (3) Gobom, J.; Nordhoff, E.; Mirgorodskaya, E.; Ekman, R.; Roepstorff, P. J. Mass Spectrom. 1999, 34, 105-116. (4) Courchesne, P.; Patterson, S. Biotechniques 1997, 22, 244. (5) Doucette, A.; Craft, D.; Li, L. Anal. Chem. 2000, 72, 3355-3362. (6) Blackledge, J.; Alexander, A. Anal. Chem. 1995, 67, 843-848. (7) Smirnov, I. P.; Haff, L. A.; Ross, P. L.; Hall, L. R.; Zhu, X. In Proceedings the 50-th ASMS Conference on Mass Spectrometry and Allied Topics, Orlando, FL, 2002. (8) Xu, Y. D.; Watson, J. T.; Bruening, M. L. Anal. Chem. 2003, 75, 185-190. (9) Landry, F.; Lombardo, C.; Smith, J. Anal. Biochem. 2000, 279, 1-8. 10.1021/ac035331j CCC: $27.50
© 2004 American Chemical Society Published on Web 04/13/2004
Figure 1. (A) Spectrum of the CHCA matrix after crystallization from its solution in 50% acetonitrile/water + 0.1% TFA. (B) A shift in mass of the matrix clusters occurs because of the addition to the matrix an equimolar mixture of Na and K acetate (5 mM total acetate salt concentration).
using deionized water.10-12 The advantage of this method is speed and simplicity, since this procedure involves simple addition of a small volume of water on the crystal surface followed by fast removal of the supernatant together with the most-contaminated layer of the crystallized matrix. This procedure takes advantage of the low solubility of many matrixes (for example, R-cyano-4hydroxycinnamic acid, CHCA) in water, compared to salts and surfactants. However, there is probably a loss of some sample due to the partial dissolution of matrix during the washing step. Even when the peptide analyte solution is relatively pure, MALDI analysis of the sample is often complicated by appearance in the spectrum of matrix clusters caused by traces of alkali metals (Figure 1), especially when the analyte molar fraction is of the order of 1 ppm.10 The signals from these clusters are observed below 1.3-1.5 kDa, and they interfere with useful signals in that mass range. The interest in suppressing matrix clusters has been elevated recently, because this mass range is important for proteomics analysis. It has been recently shown that peptides bind to the CHCA matrix surface when the peptide sample is applied to a previously (10) Keller, B.; Li, L. J. Am. Soc. Mass Spectrom. 2000, 11, 88-93. (11) Beavis, R.; Chaudhary, T.; Chait, B. Org. Mass Spectrom. 1992, 27, 156158. (12) Vorm, O.; Roepstorff, P.; Mann, M. Anal. Chem. 1994, 66, 3281-3287.
crystallized matrix.12,13 This suggests that it may be possible to partially redissolve some matrix/peptide crystals while keeping peptides still absorbed on the matrix surface, which could potentially improve the peptide signal. The result would be a concentration of the peptides on the matrix surface together with desalting of the sample surface. In our study, we investigate this possibility by applying several ammonium salt buffer solutions at different concentrations as washing media or as matrix additives. The idea behind this approach has been reported by us earlier.14,15 The results presented here are sufficient to show that application of the ammonium salt solutions substantially increases MALDI MS sensitivity and improves overall performance (mass accuracy, signal-to-noise ratio, and reduced data acquisition time). EXPERIMENTAL SECTION Mass Spectrometry. All mass spectra in this study were acquired on both a Voyager DE STR MALDI-TOF instrument and (13) Gobom, J.; Schuerenberg, M.; Mueller, M.; Theiss, D.; Lehrach, H.; Nordhoff, E. Anal. Chem. 2001, 73, 434-438. (14) Smirnov, I. P.; Zhu, X.; Taylor, T.; Huang, Y.; Ross, P.; Papayanopoulos, I. A.; Martin, S.; Pappin, D. J. In 51st ASMS Annual Conference on Mass Spectrometry and Allied Topics, Montreal, Canada, 2003. (15) Smirnov, I. P.; Zhu, X.; Papayanopoulos, I. A.; Pappin, D. J. In ABRF2003 Translating Biology Using Proteomics ans Functional Genomics, Denver, CO, 2003.
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2959
a 4700 protein analyzer (Applied Biosystems, Framingham, MA) in positive ion reflectron mode. Accelerating voltage was 20 kVwith 150-ns delay time for the Voyager system and 400 ns for the 4700 protein analyzer. The pressure in both ion sources was ∼5 × 10-7 Torr. Mass spectrometers were equipped with either a 337-nm N2 gas laser or a 355-nm Nd:YAG solid-state laser. Typical spectra were obtained by averaging 200 acquisitions from single laser pulse (20 Hz, Voyager, N2 337-nm laser) or 1000 acquisitions (200Hz laser, Nd:YAG 355-nm laser, 4700 analyzer) with the minimum possible laser energy in order to maintain the best resolution. MALDI Matrixes. For all mass spectrometry experiments, recrystallized CHCA was used as the MALDI matrix at a concentration of 25 mM in 50% water/acetonitrile and 0.1% TFA. When necessary, monoammonium phosphate (pH 4.4) and diammonium citrate (pH 4.8) were added to the matrix to produce a final concentration that ranged from 1 to 20 mM. Matrix and sample were then mixed in a 1:1 (v/v) ratio prior to spotting. Spotting on the MALDI plates was performed either manually or using the 16-probe SymBiot XVI Sample Workstation (Applied Biosytems). A total volume of 0.8 µL of the analyte/matrix mixture was used for each spot deposited. SymBiot spotted samples were deposited at a density of 192 spots/plate. Preparation of Samples Using the MALDI Spot-Washing Procedure. Spot sample onto the MALDI plate, and after it is dried, place 1.5 µL of the ammonium washing buffer on the top of the sample spot. After 3-5 s, blow off all supernatant solution retained on each spot with compressed air or remove the supernatant by the use of any pipet or the pipet tip attached to vacuum aspirator pump. Perform MS analysis. RESULTS AND DISCUSSIONS The matrix clusters of CHCA appear in MALDI spectra as periodic sets of signals, equally spaced by approximately 200220 Da, with gradually declining intensity (Figure 1A). Each of these sets of clusters consists of a certain number of CHCA molecules and Na+/K+ ions. Such complexes naturally group together, depending upon the number of matrix molecules involved in the cluster formation. In the typical MALDI spectra of pure CHCA matrix, it is possible to observe clusters formed by matrix octamers, although their intensity decreases with an increasing cluster mass. Interference of the clusters with peptides occurs predominantly in the 650-1300-Da mass range. An example of CHCA matrix clusters is shown in Figure 1A, where clusters of matrix trimers (m/z ∼650) to hexamers (m/z ∼1250) are observed. The observed pattern of matrix clusters varies depending upon many experimental conditions, some of which are unclear. The primary determinant, however, is the molar ratio between the matrix and alkali salts, especially the concentration of sodium and potassium cations in the solution at the moment of matrix crystallization. As the concentration of sodium or potassium becomes higher, the cluster signals shift toward higher mass. This is shown in the Figure 1B, where a 5 mM equimolar mixture of sodium/potassium acetates has been added to the 25 mM CHCA solution. In Figure 1A, the most intense signals for a tetramer of matrix molecules are m/z 833.08 ([4M + 2K - H]+), 855.05 ([4M + 2Na + K - 2H]+), and 871.03 ([4M + 3K - 2H]+), while in Figure 1B the main adducts are [4M + 3Na + 2K 4H]+, [4M + 2Na + 3K - 4H]+, [4M + Na + 4K - 4H]+, (899.04, 915.01, and 930.99 m/z, respectively). In the last case, there are 2960 Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
Table 1. Observed and Theoretical Masses of Some Matrix Cluster Signals, and Their Composition molecular weight cluster composition
observed
theoretical
3M + H 3M + K 3M + Na + K - H 3M + 2K - H 3M + Na + 2K - 2H 3M + 3K - 2H 4M + 2K - H 4M + Na + 2K - 2H 4M + 3K - 2H 4M + 4Na + K - 4H 4M + Na + 3K - 3H 4M + 3Na + 2K - 4H 4M + 2Na + 3K - 4H 4M + Na + 4K - 4H 4M + 5K - 4H 5M + Na + 2K - 2H 5M + 3K - 2H 5M + Na + 3K - 3H 5M + 4K - 3H 6M + 3K - 2H 6M + Na + 3K - 3H 5M + 4K - 3H
568.15 606.09 628.07 644.04 666.02 682.00 833.08 855.06 871.03 883.08 893.05 899.04 915.01 930.99 946.96 1044.10 1060.07 1082.06 1098.03 1249.10 1271.09 1287.06
568.13 606.09 628.07 644.05 666.03 682.00 833.09 855.07 871.04 883.06 893.03 899.03 915.01 930.99 946.95 1044.11 1060.09 1082.07 1098.04 1249.13 1271.09 1287.08
five metal cations per four CHCA molecules. This shows that the presence of the free CHCA carboxylic group is not the only factor defining cluster formation. One possibility could be that alkali metals are partially retained by interaction with the π-electron system of the CHCA aromatic ring, resulting in the formation of sandwich-type structures. The compositions of the most common peaks observed in this study are listed in the Table 1. Some observed signals apparently have complex origin, such as the 887 m/z signal (Figure 1), which can be attributed to an oxidized form of the 871 m/z complex ([4M + 3K + O - 2H]+). It is not obvious whether it is due to peroxide formation, such as COOOH, or due to another oxidation event without additional research. During the last several years, Nd:YAG lasers (355 nm) with 200-Hz pulse frequency have become available and are expected to replace shorter-lifetime N2 lasers (337 nm) on MALDI mass spectrometers. A higher pulse rate allows faster data acquisition, which is an essential requirement for high-throughput proteomics studies. The shift of laser wavelength from 337 to 355 nm, however, results in more intense CHCA matrix clusters. To verify this, a simple experiment was designed. A dilution series of desArg1-bradykinin from 200 to 25 nM was prepared in 25 mM CHCA. All samples were analyzed on two different instruments equipped with either a 337-nm N2 or a 355-nm Nd:YAG laser. Each instrument was optimized for the best sensitivity. Figure 2B shows typical spectra obtained for 1 ppm mole fraction of analyte for both types of the lasers. To compare these results, the intensities of observed signals were normalized using the signal intensity of the bradykinin ion. Figure 2C demonstrates the normalized signal of the [4M + Na + 2K - 2H]+ (m/z 855.1) peak at both laser wavelengths as a function of analyte/matrix ratio. With the Nd: YAG laser, cluster intensities are approximately twice as high at 1 ppm analyte, and at 8 ppm, their intensities are still comparable to the signal of bradykinin. Cluster intensities with the 337-nm laser decline much faster and, at 8 ppm, have only 15% of the bradykinin intensity. The ratio of molar extinctions of CHCA (in
Figure 2. Stronger cluster signals obtained with a Nd:YAG laser (355 nm) in MALDI MS compared to a N2 laser (337 nm). (A) UV spectrum of a CHCA water solution. The extinction coefficient of CHCA at 355 nm is 2.2 times lower than at 337 nm. (B) A typical spectrum is shown for 1 ppm of bradykinin using 337- and 355-nm lasers. (C) Dependency of the intensity of matrix clusters vs various (1-8 ppm) analyte/matrix ratio (des-Arg1-bradykinin/CHCA) for both laser wavelengths.
solution), 337/355, is ∼2.2, as shown in the UV spectrum of a CHCA aqueous solution (Figure 2A). Our study (discussed below) suggested that the only matrix clusters observed are those that form during desorption of the outer layers of the crystalline surface. In this case, lower adsorptivity at 355 nm, at the same laser pulse intensity, will result in less absorbed energy of surface layers. This may help to slow the cluster dissociation rate and, as a result, increase their observed intensity. Thus, employment of high-repetition Nd:YAG lasers at 355 nm in newer, highthroughput MALDI systems creates a need to develop more efficient protocols for matrix cluster elimination. The advantage of limited CHCA solubility has been used with postcrystallization washing with deionized water.10-12 To avoid the potential loss of material, the addition of 0.1% TFA has also been used. Addition of ammonium salts to the wash solution may seem contradictory to logic, because it will definitively dissolve more CHCA matrix (although it may be interesting to investigate because of earlier reports on nonspecific absorption of peptides from supernatant onto CHCA crystals).13 To study whether such effects may take place, a solution with 6 nM des-Arg1-bradykinin, 25 mM CHCA, 0.1%TFA, and 5 mM mixture of sodium and potassium acetates (1:1 mol/mol) was
prepared. Sodium and potassium salts were added to the matrix solution in order to get reproducible cluster formation. Prior to acquiring spectra, sample spots on a MALDI plate were washed with either deionized water or ammonium buffers of various concentration. The best results were achieved when either diammonium citrate (pH 4.8) or monobasic ammonium phosphate (pH 4.4) dissolved in deionized water was used. To estimate the washing efficiency, three signals were monitored at m/z: 893.1 [4M + Na + 3K - 3H]+, 915.1 [4M + 2Na + 3K - 4H]+, and 904.5 [des-Arg1-bradykinin + H]+. Results for diammonium citrate washings are presented in Figure 3A-C, with the intensities of matrix cluster signals normalized by bradykinin (presented in Figure 3C). The concentration of the applied citrate buffer was in the range of 5-20 mM. A typical spectrum with no postcrystallization treatment is shown in Figure 3A, where the monitored peaks are circled. The two cluster signals are 1.7-2 times more intense than bradykinin, as shown by corresponding bars in Figure 3C (bars labeled NW). Application of deionized water leads to a modest reduction of cluster signals (bars labeled WHOH in Figure 3) and more than an 8-fold increase of the bradykinin signal/noise ratio (Figure 3D). Washings with diammonium citrate solutions lead to much greater Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2961
Figure 3. Effect of the concentration of diammonium citrate as a washing buffer on the elimination of matrix clusters and improvement of the analyte signal/noise values. All spectra were obtained for 10 nM des-Arg1-bradykinin in 25 mM CHCA together with 5 mM Na/KOAc. Washings were performed with various concentrations (0-20 mM) of diammonium citrate. Plate A shows a MALDI spectrum obtained with no washing. Plate B shows the result of washing of a similar spot with 20 mM diammonium citrate. Plate C shows the ratio between the signal intensities of cluster components [4M + Na + 3K - 3H]+ (m/z 893), [4M + 2Na + 3K - 4H]+ (m/z 915), and bradykinin (m/z 904.5). Plate D presents the signal-to-noise ratio for bradykinin itself vs washing conditions.
cluster reduction and to even higher bradykinin S/N (W5ACW20AC, panels C and D), with the best results achieved for the maximum applied concentration (20 mM). A representative spectrum is shown in profile B, where a substantial, 40-fold sensitivity improvement (50 000 counts for bradykinin signal versus 600 counts on the profile A) is evident. The 1-Da periodic chemical noise that was totally masked by high-frequency noise on the original spectrum (A) is now clearly visible. Unfortunately, under 20 mM ammonium citrate wash conditions, the loss of the material due to dissolving of the CHCA crystals during the washing step was severe under microscope inspection, making the wash under these conditions unpractical. For typical operation, a “less aggressive” wash by 5-10 mM of the same buffer is recommended. Microphotographs of a sample spot before and after the washing step with 10 mM diammonium citrate are shown in Figure 4A and B, respectively. Although a small reduction of the size of the crystals in the profile B is evident, most of the material is still on the plate. Interestingly enough, the ratio between the signal intensity of bradykinin and its sodium and 2962 Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
potassium adducts does not depend very much upon the washing conditions employed (data not shown). Intensities of sodium adducts stay within 7-15% of the bradykinin signal (m/z 904.5) and potassium adducts within 12-26%, even when the alkali matrix cluster signals are practically eliminated (Figure 3B). This contradiction between presence of peptide alkali adducts and absence of matrix clusters suggests a difference in the mechanism of their formation. Generally, it may be assumed that alkali metal salts are formed inside of the matrix crystal by both analyte and matrix molecules. With the substantial molar excess of matrix versus analyte, if matrix clusters are stable and are not prone to dissociation, clusters should be formed across the entire volume of ablated material and should be observed with or without washing the surface of crystals. Since the experimental data do not support this assumption, it is plausible that the only surviving matrix clusters observed in MALDI spectra are associated with surface-type processes, and not the volume-type, and their formation happens in the early stages of laser ablation and plume generation. Thus, being on the outer layer of the expanding plum,
Figure 4. Microscope photograph of CHCA matrix crystals on the MALDI sample plate (90× magnification). Profile A: after crystallization of matrix was completed. Profile B: the same surface after washing with 10 mM diammonium citrate for 5 s.
Figure 5. Analysis of the mixture of three peptides: des-Arg1-bradykinin, angiotensin I, and [glu1]-fibrinopeptide B (human). Concentration of each peptide is 2 fmol/µL. Concentration of CHCA 25 mM. Profile A: a typical spectrum obtained from CHCA matrix without any postcrystallization treatment and without any additives. Profile B: same as (A), with 10 mM (NH4)H2PO4 added to the matrix solution. Profile C: same as (A), except the sample spot was washed with 0.1% TFA in deionized water. Profile D: same as (A), except the sample spot was washed with 10 mM diammonium citrate. Profile E: 25 mM CHCA, 10 mM (NH4)H2PO4. After drying, the spot was washed with 10 mM diammonium citrate for 3 s. Profile F: the signal-to-noise ratio of des-Arg1-bradykinin, angiotensin I, and [glu1]-fibrinopeptide B (human) depending on sample preparation method.
clusters avoid multiple collisions, which would otherwise lead to their dissociation. The formation of alkali adducts of peptides depends on the sodium/potassium concentrations within the CHCA crystal and occurs also in the later stages of desorption (ablation), when noncovalent matrix clusters (despite the presence of alkali ions) cannot survive multiple collisions (Figure 3B). On another note, the results suggest that there may be some retention of peptides on the surface of the CHCA crystals during the washing step, which improves the overall signal/noise ratio of bradykinin (Figure 3D), supporting our original assumption.
In the MALDI analysis of DNA, suppression of alkali adducts is often achieved by the incorporation of ammonium salts in the matrix solution.16-19 This approach is evaluated here for peptide analysis. Among the ammonium salts tested, monobasic ammonium phosphate provided the best results. It efficiently suppressed matrix cluster formation and resulted in better analyte signal. A mixture of three peptides, des-Arg1-bradykinin (m/z 904.5 [M + H]+), angiotensin I (m/z 1296.7 [M + H]+), and [Glu1]fibrinopeptide B (human) (m/z 1570.7 [M + H]+), was prepared at a concentration of 2 fmol/µL of each component and subseAnalytical Chemistry, Vol. 76, No. 10, May 15, 2004
2963
Figure 6. Dependency of angiotensin I S/N vs concentration of monobasic ammonium phosphate (0-20 mM). Spectra were recorded for 1.5 nM angiotensin in 25 mM CHCA with 0-20 mM ammonium phosphate in the matrix solution. Curve A: CHCA/ammonium phosphate with no postcrystallization treatment. Curve B: CHCA/ ammonium phosphate and postcrystallization washings with water. Curve C: CHCA/ammonium phosphate and postcrystallization washings with 10 mM ammonium phosphate Curve D: CHCA/ammonium phosphate and postcrystallization washings with 10 mM diammonium citrate.
quently analyzed under different conditions (various matrix additives and postcrystallization washings) in 25 mM CHCA (0.1% TFA). Figure 5A shows a typical spectrum accumulated using a standard MALDI protocol (no matrix additives, no washing procedure, 400 laser shots). Addition of 10 mM monobasic ammonium phosphate to the matrix not only suppressed cluster formation but also led to a 50-100% stronger analyte signal (Figure 5B). A similar spectrum resulted when another spot (without any additives in the matrix solution) was washed after crystallization with deionized water (Figure 5C). To understand the effects of ammonium phosphate additives, a systematic study was performed on solutions of the same peptides (1.5 fmol/µL of each peptide) in CHCA containing monobasic ammonium phosphate with a concentration range of 0-20 mM. The signal-to-noise ratio for angiotensin I as a function of the ammonium phosphate concentration is shown (curve A of the Figure 6). Signal/noise ratio improves substantially in the range of 1-4 mM concentration of the additive. Higher additive concentrations led only to modest improvements of the angiotensin I signal. Good cluster elimination can be achieved in the case of moderately contaminated samples using 5-10 mM ammonium phosphate. A further increase in concentration leads to problems with matrix crystallization, which became evident with 20 mM ammonium phosphate and higher. Generally, postcrystallization washings with ammonium salts have a stronger impact (16) Currie, G. J.; Yates, J. R. I. J. Am. Soc. Mass Spectrom. 1993, 4, 955-963. (17) Nordhoff, E.; Ingendoh, A.; Cramer, R.; Overberg, A.; Stahl, B.; Karas, M.; Hillenkamp, F.; Crain, P. Rapid Commun. Mass Spectrom. 1992, 6, 771776. (18) Pieles, U.; Zurcher, W.; Schar, M.; Moser, H. E. Nucleic Acids Res. 1993, 21, 3191-3196. (19) Zhu, Y. F.; Chung, C. N.; Taranenko, N. I.; Allman, S. L.; Martin, S. A.; Haff, L.; Chen, C. H. Rapid Commun. Mass Spectrom. 1996, 10, 383-388.
2964
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
on S/N than employing additives in the solution of the matrix (Figure 5B). For example, Figure 5D shows a MALDI spectrum of a spot subjected to postcrystallization washing with 10 mM diammonium citrate, and this spectrum displays approximately twice the signal than the addition of phosphate to the CHCA solution. Interestingly enough, the combination of both stepss ammonium salt addition to the matrix solution and postcrystallization washingshas a cumulative effect and provided the best results with 3-5-fold overall improvement in sensitivity (Figure 5E) versus untreated sample (Figure 5A). A systematic study of the impact of washing and addition of ammonium phosphate to the matrix solution is presented in the Figure 6. For pure CHCA, washings with ammonium citrate (curve D) and phosphate (curve C) are much more efficient compared to those with deionized water (curve B) or the addition of ammonium phosphate in the matrix solution without washings (curve A). The difference in washing by water or by ammonium-containing buffers in respect to signal/noise ratio of angiotensin I with an increase of ammonium phosphate concentration became less apparent, but generally, the best results were still obtained by washing with 10 mM diammonium citrate. In summary, the combination of 5-10 mM monobasic ammonium phosphate additive in the CHCA matrix solution followed by postcrystallization washing with 5-10 mM dibasic ammonium citrate results in the optimum conditions for achieving the best MALDI MS signal/noise ratios of diluted peptide solutions (see Figure 5F as illustration). Two diluted tryptic digests have been analyzed by the optimized method described above: β-galactrosidase (bovine) and hemoglobin (human) at concentrations of 2 and 0.7 fmol/µL, respectively. Panels A and C of Figure 7 show the spectra of hemoglobin and β-galactosidase in pure CHCA matrix. Spectra shown in the lower panels (B and D) were obtained by an addition of 10 mM ammonium phosphate in the matrix solution followed by washing with 10 mM diammonium citrate. Such treatment totally eliminated signals from matrix clusters, such as [4M + Na + 2K - 2H]+ (m/z 855), which were observed on both the spectra with pure CHCA on Figure 7A (S/N 372) and on the Figure 7C (S/N 321). Furthermore, significant improvement of the MALDI in signal is observed for the peptide peaks. For example, for three hemoglobin digest fragments, LLGNVLVVVLAR (m/z 1265.8), LLVVYPWTQR (m/z 1274.7), and FLANVSTVLTSK (m/z 1279.7), the observed signal/noise ratio was 20, 45, and 10, respectively, for pure CHCA matrix (Figure 7A). The corresponding S/N values were 169, 106, and 31 on the typical spectra presented on profile B. In the lower spectra, chemical noise with 1-Da periodicity can be observed more clearly. Such sample preparation methods have a similar effect in the case of a diluted β-galactosidase tryptic digest mixture. For example, in pure CHCA, the signal-to-noise ratio for HQQQFFQFR (m/z 1265.6) is 34 (Figure 7C), compared to 85 when our suggested procedure was used. The sample preparation techniques described here help both to eliminate matrix cluster signals and to improve the signal intensities of analyzed peptides. It should allow better detection of peptides in LC-MALDI MS applications. This washing methodology in combination with ammonium phosphate addition to
Figure 7. Analysis of protein digests. Profiles A and B: hemoglobin (bovine) tryptic digest, 0.7 fmol/µL. Profiles C and D, β-galactosidase tryptic digest (2 fmol/µL). Both upper MS spectra (A and C) were recorded in pure CHCA with no additives and no extra treatment. For profiles B and D, sample were mixed with 25 mM CHCA + 10 mM (NH4)H2PO4 solution and then dried and followed by washing of crystallized sample/ matrix spots with 10 mM dibasic ammonium citrate. Similar results were obtained with 5 mM dibasic ammonium citrate as washing buffer (not shown).
the matrix solution is becoming a valuable approach in every day sample preparation in our laboratory. CONCLUSIONS CHCA cluster formation is more apparent in MALDI spectra acquired using 355-nm Nd:YAG laser than for a 337-nm N2 gas laser. It is possible, that CHCA matrix clusters observed in the spectra are formed on the early stages of the desorption process. Application of ammonium-containing buffers as washing media of previously crystallized sample spots on the MALDI plate offers total elimination of the alkali metal matrix cluster signals and increases the intensities of peptide signals. The best results were obtained using 5-10 mM diammonium citrate buffer. Application of ammonium salts as matrix additives provide a similar, although
weaker effect on the quality of MALDI spectra. The best results were obtained with 4-10 mM monobasic ammonium phosphate. Combining both of the above stepssaddition of ammonium salts of phosphates (4-10 mM concentration) in 25 mM CHCA matrix following 5-10 mM diammonium citrate (or phosphate) buffer washing of the corresponding sample spotsprovides a substantial, severalfold increase in signal intensity, in addition to the total elimination of matrix cluster signals.
Received for review November 10, 2003. Accepted March 3, 2004. AC035331J
Analytical Chemistry, Vol. 76, No. 10, May 15, 2004
2965
