Ionization Mass Spectrometry

---

Anal. Chem. 2004, 76, 7288-7293

Matrix-Implanted Laser Desorption/Ionization Mass Spectrometry Alexey Novikov,† Martine Caroff,‡ Serge Della-Negra,† Yvon Lebeyec,† Miche`le Pautrat,† J. Albert Schultz,§ Agne`s Tempez,§ Hay-Yan J. Wang,| Shelley N. Jackson,| and Amina S. Woods*,|

Institut de Physique Nucle´ aire, CNRS-IN2P3, 91406 Orsay, France, Equipe “Endotoxines”, UMR 8619 du CNRS, IBBMC, Universite´ de Paris-Sud, 91405 Orsay, France, Ionwerks, Inc., 2472 Bolsover Street, Suite 255, Houston, Texas 77030, and NIDA IRP, 5500 Nathan Shock Drive, Baltimore, Maryland 21224

The implantation of low-velocity massive gold clusters is shown to be a method of choice for homogeneous incorporation of a metallic matrix into the near-surface region of a solid biopolymer for subsequent laser desorption/ ionization (LDI) MS analysis. Matrix implanted (MI)LDI spectra from cluster-implanted pure test peptide or tissue exhibit molecular ion peaks similar to those observed by matrix-assisted LDI. Moreover, the ion emission is very reproducible from any spot on the surface of these test samples. MILDI promises to be a powerful technique for mass spectrometric analysis of native biological samples as demonstrated by the first results on rat brain tissues. Mass spectrometry has become a major tool for identifying and characterizing biological macromolecules with femtomole sensitivity. Matrix-assisted laser desorption/ionization (MALDI)1,2 is one of the most efficient methods for analyzing molecules from a solid sample. The preparation of MALDI samples consists either in cocrystallization of analyte molecules with small light-absorbing organic molecules1 or in dissolving analyte molecules in suspensions of solid particles (metal, carbon, silicon, oxide, nitride, etc.) of micrometer or nanometer sizes.2-5 The dried droplet technique usually produces a nonuniform spatial distribution of analyte within the dried thin film.6,7 During MALDI analysis, widely varying spectral intensities occur as the laser position is moved from one focal spot to the next. This presents a severe limitation for laser microprobe molecular imaging of tissue samples during which organic matrix solution is deposited on the surface of a biological tissue.8 Derivatized fullerene naonparticulate have recently been shown to be effective as matrixes when solutions are deposited * Corresponding author. Tel: 410-550-1507. Fax: 410-550-6859. E-mail: awoods@ intra.nida.nih.gov. † Institut de Physique Nucle´aire. ‡ Universite´ de Paris-Sud. § Ionwerks, Inc. | NIDA IRP. (1) Karas, M.; Hillenkamp, F. Anal. Chem. 1988, 60, 2299-2301. (2) Tanaka, K.; Waki, H.; Ido, Y.; Akita, S.; Yoshida, Y.; Yoshida, T. Rapid Commun. Mass Spectrom. 1988, 8, 151-153. (3) Sunner, J.; Dratz, E.; Yu-Chie, C. Anal. Chem. 1995, 67, 4335-4342. (4) Dale, M.; Knochemuss, R.; Zenobi, R. Anal. Chem. 1996, 68, 3321-3329. (5) Schurenberg, M.; Dreisewerd, K.; Hillenkamp, F. Anal. Chem. 1999, 71, 221-229. (6) Garden, R. W.; Sweedler, J. Anal. Chem. 2000, 72, 30-36. (7) Luxembourg, S. L.; McDonnell, L. A., Duursma, M. C.; Guo, X.; Heeren R. M. A. Anal. Chem. 2003, 75, 2333-2341.

7288 Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

onto rat brain tissue sections.9 Nevertheless, the uniform incorporation of organic MALDI matrix remains probably the greatest difficulty for a successful MALDI microprobe image analysis from a biological surface. Matrix solution deposition onto a tissue sample surface can also easily perturb the spatial distribution of the targeted biomolecules. We describe in this work a newly developed alternative method for the homogeneous incorporation of matrix metal clusters into a bioorganic solid material. Very recently it was shown that largesize cluster projectiles Aun at a total energy of 5-20 keV/cluster electric charge (typically Au4004+) could be produced with a liquid metal ion source (LMIS).10 Due to the large number of atoms in the clusters, the energy per atom is small (∼100 eV) and corresponds to an estimated depth of penetration of several molecular layers in an organic material. Recently, Au4004+ cluster beams were used in SIMS experiments to analyze a variety of neat peptide samples demonstrating the nondestructive character of the very heavy cluster projectiles.11 We observed11 a slight increase in the intensity of the gramicidin S MH+ secondary ion during the implantation of up to 2.5 × 1013 Au400/cm2. This remarkable result motivated the present work which shows that the same LMIS cluster ion source can be used to implant gold clusters into the top layers of biological samples in order to achieve a matrixlike effect in UV laser desorption mass spectrometry. In these first experiments, the performance of matrix-implanted laser desorption/ionization (MILDI) is demonstrated for a variety of samples: dynorphin 1-7, thymic factor, bovine insulin B chain, and slices of rat brain tissue. We show that implanted gold atoms not only function as a MALDI matrix but that excellent homogeneity of ion emission from the entire implanted surface is achieved as well. The influence of the implantation dose and of the laser fluence on the ion emission from the implanted samples is also reported as well as a discussion of some anticipated limitations of the technique. (8) Todd, P. J.; Schhaff, T. G.; Chaurand, P.; Caprioli, R. M. J. Mass Spectrom. 2001, 36, 355-369.M. V. (9) Ugarov, M.; Egan, T.; Khabashesku, D. V.; Schultz, J. A.; Peng, H.; Khabashesku, V. N.; Furutani, H.; Prather, K. S.; Wang, H.-W. J.; Jackson, S. N.; Woods. A. S. Anal. Chem. In press. (10) Bouneau, S.; Della-Negra, S.; Depauw, J.; Jacquet, D.; Le Beyec, Y.; Mouffron, J. P.; Novikov, A.; Pautrat, M. Nucl. Instrum. Methods Phys. B. In press. (11) Tempez, A.; Schultz, J. A.; Della-Negra, S.; Depauw, J.; Jacquet, D.; Novikov, A.; Le Beyec, Y.; Pautrat, M.; Caroff, M.; Ugarov, M.; Bensaoula, H.; Gonin, M.; Fuhrer K.; Woods, A. S. Rapid Commun. Mass Spectrom. 2004, 18, 371-376. 10.1021/ac049123i CCC: $27.50

© 2004 American Chemical Society Published on Web 11/06/2004

EXPERIMENTAL SECTION Cluster Beams. The irradiation of samples with massive gold cluster ions was performed at the Institut de Physique Nucle´aire in Orsay (IPNO), France. The cluster beams were produced with a (Au-Si) LMIS used since the early 1990s at the IPNO to investigate secondary emission from various solid samples.12,13 It was shown recently for the first time that beams of mass-selected large metal clusters containing hundreds of atoms can be produced by this LMIS.10,11 A Wien filter is used for mass selection after acceleration of the clusters by a 10-20-kV electric potential. The Aun+ beams are well separated up to n ) 9. Above n ) 9, massive cluster beams can be selected with broader mass distributions but fixed values of n/q (q is the cluster charge). In the experiments reported here, we accelerated cluster ions by 10 kV and adjusted the Wien filter to n/q ) 100. In another study, the mean charge of these clusters has been estimated as 〈q〉 ) +4.10 Therefore, the mean number of atoms in a cluster is 〈n〉 ) 400 and the mean total energy 〈E〉 ) 40 keV. Sample Preparation. Three peptides were used: i.e., dynorphin 1-7 (YGGFLRR, MW 868.0), thymic factor (EAKSQGGSN, MW ) 858.9), and bovine insulin B chain (FVNQHLCGSHLVEALYLVCGERGFFYTPKT, MW ) 3493.6). Dynorphin 1-7 and thymic factor were purchased from Bachem, and bovine insulin B chain was from Sigma. Dynorphin 1-7 and thymic factor were selected to examine any effect of aromatic side chains in the absence of gold matrix. Droplets containing 2 µL of a 1 nmol/ µL solution of these pure analytes in water or in 0.1% TFA/H2O (for insulin) were deposited on a stainless steel plate and dried. The resulting dried droplet samples have a diameter of 2-3 mm. These samples were used either for implantation with Au400 clusters or directly for laser desorption/ionization (LDI)MS analysis. For each substance, MALDI reference samples (with R-cyano-4-hydroxycinnamic acid (CHCA) as the matrix) were also prepared by the dried droplet method. To establish the role of cationic impurities (Na, K), a series of dynorphin 1-7 samples were prepared with and without desalting the peptide solutions prior to dried droplet deposition. The desalting was done by adding a few grains of Dowex 50 (Me2NH+). The Sprague-Dawley rat brain tissue slices were 14 µm thick; they were cut, placed on stainless steel targets, and dried in a vacuum chamber for 5 min. The dried tissue slices were then implanted with gold and analyzed within 48 h. Vacuum drying establishes a standard surface that can be easily shipped among our three laboratories. Implantation Procedure. Peptide samples were irradiated with 40-keV Au4004+ clusters at a normal incidence. The energy per gold atom was only 100 eV. This means that the depth of penetration into an organic sample is limited to ∼10 nm.11 The implantations were performed in a small vacuum chamber ((12) × 10-6 mbar) connected to the end of the gold cluster beam line. A beam of uniform fluence over a section of 3-mm diameter was used to homogeneously implant the gold cluster projectiles into the sample. The implantation fluences varied from 2.5 × 1011 to 2.5 × 1013 Au400/cm2. To control the beam intensity during (12) Della-Negra, S.; Joret, H.; Le Beyec, Y.; Blain, M.; Schweikert, E. A. Phys. Rev. Lett. 1989, 63, 1625-1628. (13) Benguerba, M.; Brunelle, A.; Della-Negra, Depauw, J.; Joret, H.; Le Beyec, Y.; Blain, M.; Schweikert, E.; Ben Assayag, G.; Sudraud, P. Nucl. Instrum. Methods B 1991, 62, 8-22.

Figure 1. (a) LDI MS spectrum from an unimplanted control dynorphin 1-7 sample obtained with 50 laser shots. (b) MILDI MS spectrum from a dynorphin 1-7 sample after implantation of 1 × 1013 Au400/cm2 obtained with two laser shots. The laser fluence necessary to observe molecular ion peaks is substantially higher for the control than for the implanted sample.

irradiation, the sample was replaced from time to time by a 3-mm diaphragm (sample size) and the current of the beam passing through the diaphragm was measured by a Faraday cup. For a beam intensity of ∼100 pA the required implantation time varied from 2 to 200 min. After gold cluster implantation the samples were transferred to a mass-spectrometer equipped with a nitrogen UV laser to acquire MILDI mass spectra. Laser Desorption/Ionization MS Analysis. LDI MS analysis of the implanted samples was done using both a Voyager DESTR and a DE-Pro MALDI-TOF (Applied-Biosystem, Framingham MA) equipped with a 337-nm N2 lasers and a 20-kV acceleration voltage. Comparison spectra were acquired in parallel from virgin (unimplanted) reference samples as well as from standard MALDI matrix preparations of the same substances. Three characteristics were compared: (i) signal-to-background ratio for the molecular ion peak, (ii) reproducibility of the molecular peak intensity obtained from different spots of the sample surface, and (iii) number of laser shots at an individual irradiated spot before the molecular ion peak disappears. MILDI and MALDI spectra were acquired using only 2 laser shots, whereas up to 50 shots and higher laser fluence were necessary for LDI mass spectra of the virgin substances. All spectra were acquired in positive and negative ion modes. The laser fluence was adjusted for each sample. Moreover, using dynorphin 1-7 as a model, we studied the effect of the implantation dose, the laser fluence, and the presence of metal cations in the sample, on the resulting MILDI mass spectra. RESULTS AND DISCUSSION Gold Cluster Implantation Enhancement of LDI Molecular Ion Yield. Figure 1 compares a positive ion LDI MS Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

7289

Figure 2. Decay of the molecular peak intensity as a function of the number of laser shots for two dynorphin 1-7 MILDI samples (implantation doses 2.5 × 1012 and 2.5 × 1013 Au400/cm2). The data from six spots on each sample were recorded and averaged. Standard deviations over each series of measurements characterize the reproducibility of the MILDI data.

spectrum from an unimplanted dynorphin 1-7 control sample obtained with 50 laser shots to a MILDI MS spectrum from a gold-implanted dynorphin 1-7 sample (1 × 1013 Au400/cm2) obtained with 2 laser shots. Both spectra were acquired under the same conditions except for laser fluence. To acquire LDI spectrum from the control sample, it was necessary to increase the laser fluence (from “2700” to “3100”, according to the laser attenuator graduation). At this higher fluence, the LDI spectrum from the unimplanted control sample shows weak molecular ion peaks visible above a sloping background of chemical noise. By contrast, the MILDI spectrum shows intense peaks of molecular ions with single or double potassium adducts. Similar results are obtained from thymic factor and insulin B chain samples. These results clearly demonstrate that Au400 implantation acts as matrix in laser desorption mass spectrometry. The matrix effect of Au400 implantation prior to the LDI MS analysis is easily observed for an implantation dose as low as 2.5 × 1011 Au400/cm2. However, the optimal implantation dose for the peptide samples is between 5 × 1012 and 2 × 1013 Au400/cm2. Uniformity and Reproducibility of the Molecular Ion Signal as an Implanted Sample Is Scanned by the Laser. Both the molecular ion spot-to-spot intensity variation and the persistence of the positive and negative molecular ion intensity as a function of the number of laser shots at the same spot were measured from three dynorphin 1-7 samples implanted with the following dose range of gold cluster ions: 2.5 × 1011, 2.5 × 1012, and 2.5 × 1013 Au400/cm2. Figure 2 shows the decrease of the MNa+ molecular signal intensity (averaged from the six different spots on each sample) as a function of the number of shots. The top curve, obtained from the 2.5 × 1013 Au400/cm2 implanted sample, shows the least spot-to-spot variation. The error bars on the first point of this curve (after two laser shots) represent the 15% variation between the spectral intensity from the six virgin spots on the sample. Even after 20 shots, the variation of spotto-spot intensity is still less than 50%. A much longer “lifetime”, of an individual sample spot is observed for the sample implanted with this higher dose than for the sample dosed with 2.5 × 1012 Au400/cm2. This lower curve from the sample implanted with 2.5 7290 Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

Figure 3. Negative ion mode (a) MILDI and (b) MALDI spectra of thymic factor obtained using two laser shots, the laser fluence being optimized each time to obtain the best results. Implantation dose for the MILDI sample is 2.5 × 1012 Au400/cm2.

× 1012 Au400/cm2 has been normalized by the intensity of the molecular peak resulting from the first two shots and plotted in Figure 2 so that it can be easily compared to the curve from the higher dose. The spot-to-spot variation is clearly much higher for this lower dose, and the signal persists for only 12 shots. For the lowest implantation dose of 2.5 × 1011 Au400/cm2 (data not shown), the molecular ion signal disappears after two laser shots at each of the six spots and the spot-to-spot variation of the intensity is 60%. MILDI and MALDI Mass Spectra and the Influence of Cations. We have first compared mass spectra of both dynorphin 1-7 and thymic factor, obtained from the implanted Au cluster matrix and CHCA MALDI matrix. In each experiment, the laser fluence was adjusted to obtain the best results. The intensity of the molecular ion signal, the molecular ion peak-to background ratio, and the mass resolution were very similar between the MILDI and matrix preparations of either of these two test peptides. Figure 3 shows MALDI and MILDI mass spectra of thymic factor obtained in negative ion mode using only two laser shots. The laser fluence required for MILDI peptide samples is higher than for MALDI samples. However, there is a substantial difference between positive-ion MILDI and MALDI mass spectra. We observed that the (M + K)+ ion peak is the most intense in MILDI (see Figure 1b) while it is the (M + H)+ ion peak in MALDI. Preferential cationization by sodium and potassium in positive mode and deprotonation in negative mode were observed earlier in LDI experiments using suspensions of nanometer and micrometer particles as matrixes2,3,4,9 and is very similar to the behavior we see here in MILDI. It is difficult to derive efficiencies of the Au cluster matrix from the data in Figure 3. However, from estimates of penetration depth in previous work11 it is clear that the 2.5 × 1012 Au400/cm2 dose (which represents approximately one “monolayer”) is distributed

Figure 4. MILDI spectra of nondecationized (a) and decationized (b) dynorphin 1-7 samples. Both samples were prepared with the same implantation dose (1 × 1013 Au400/cm2) and analyzed with the same laser fluence using two laser shots.

throughout ∼10 nm of the near-surface region of the peptide sample. Unfortunately the depth of the laser crater is very difficult to measure, and thus, the total amount of peptide consumed from the 100-µm-diameter focal volume of the laser spot is not known, although simple calculations show that only ∼1.5 × 1010 Au clusters are in the illuminated volume. To study the influence of metal cationic impurities on the ionization processes in MILDI, we compared MILDI spectra from two dynorphin 1-7 samples prepared before and after decationization of the solution (see Figure 4). Both samples were implanted with the same Au400 dose (1 × 1013 Au400/cm2) and analyzed with the same laser fluence using only two laser shots. The (M + H)+ is the main ion peak in the decationized sample spectrum. However, its intensity is three times lower than that of the (M + K)+ peak in the nondecationized sample spectrum. In addition to the cationized adducts, there are other peaks in the spectra that are consistent with Au adduct attachment as well. In Figure 4b (decationized sample), weak peaks at around m/z 1064 and 1019 are consistent with MAu+ and (M - COO) Au+. In Figure 4a (nondecationized sample), both of these peaks are even weaker, but strong peaks appear at m/z 1086 and 1002 that are consistent with (M + K - NH2)Au+ and (M - COO - NH2)Au+. These assignments are yet to be confirmed by higher resolution spectra and MS/MS structural work, but they, nevertheless, point to the increased complexity of the resulting MILDI spectra compared with MALDIsparticularly where significant amounts of alkali cations are present. Influence of the Implantation Dose and of the Laser Fluence in MILDI. If the initial dried droplet samples are chemically homogeneous, then the quality of MILDI spectra (i.e., intensities of the peaks, resolution, and chemical background) depends mainly on the implantation dose and on the laser fluence. Figure 5 shows the evolution of the negative ion MILDI spectrum

Figure 5. Evolution of the negative ion mode MILDI spectrum of a dynorphin 1-7 sample with increasing laser fluence. The implantation dose of this sample is 1 × 1013 Au400/cm2. All spectra were recorded using two laser shots.

Figure 6. Dependence of the deprotonated molecular ion signal on the laser fluence for two dynorphin 1-7 samples implanted with doses of 2.5 × 1012 and 1 × 1013 Au400/cm2. The peak area is taken as a measure of the molecular ion signal intensity.

of a dynorphin 1-7 sample implanted with 1 × 1013 Au400/cm2 while increasing the laser fluence. The latter is given in arbitrary units corresponding to the scales of the laser attenuator. After a certain laser fluence threshold, the molecular ion peak appears and grows with increasing laser fluence. However, above the laser fluence of 2800, the molecular ion peaks become broader and the signal-to-noise ratio is reduced. It is also observed in the positive mode spectra that the proportion of cationized ions increases with the laser fluence and that the doubly cationized ions increase more rapidly than the singly cationized ions. This behavior is observed for both dynorphin 1-7 and thymic factor samples implanted with different Au400 doses. We also observed that the lower doseimplanted samples required higher laser fluence to generate good Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

7291

Figure 7. MILDI MS spectrum of an insulin B chain sample after implantation of 1 × 1013 Au400/cm2.

quality spectra. This is illustrated in Figure 6, which compares the evolution of the deprotonated molecular ion signal (peak area) for two dynorphin 1-7 samples implanted with doses of 2.5 × 1012 and 1 × 1013 Au400/cm2. In Figure 6, the regions of optimal fluence are also noted where we simultaneously obtain good intensity and good linear mode resolution; these regions are indicated with double arrows inserted next to the fluence curves for both implantation doses. A caveat is worthwhile at this point. The signal intensities from the gold-implanted peptides depend not simply on the even distribution and mixing of peptide analyte and implanted gold clusters but on the even distribution of alkali within the dried droplet as well. This even spatial distribution of alkali within the dried peptide droplet may not always be achieved, and it will almost certainly never be the case in a practical surface such as brain tissue. Thus. additional study and characterization of cationized test samples is clearly going to be necessary. Insulin B Chain. The results with insulin chain B were obtained using the optimal implantation dose (∼1 × 1013 Au400/cm2) found from the study with the two peptides: dynorphin 1-7 and thymic factor. Both positive and negative molecular ions

were detected. Figure 7 shows a negative ion mode MILDI spectrum of insulin B chain (implantation dose 1 × 1013 Au400/ cm2). No molecular ions were found in the unimplanted control sample at any laser fluence. Further work is in progress to establish optimal conditions for analyzing higher molecular weight molecules. Brain Tissue. To estimate the ability of this new technique to analyze more complex biological samples, we carried out the first MILDI experiments on Sprague-Dawley rat brain tissue sections. Spectra were acquired in both linear and reflectron modes. In the reflectron mode, high ion signal intensity was obtained in the mass range 100-1000 Da. Figure 8 shows a MILDI mass spectrum in the mass range 300-1000 Da obtained in the reflectron mode from rat brain tissue implanted with a fluence of 5‚1012 Au400/cm2. Conditions similar to those established for the peptide samples were used. The group of peaks in the mass range 800-900 Da are attributed to cationized lipid molecular ions mixed with isobaric peptides. Analysis of gold-implanted tissues is not, however, restricted to small molecules. We also measured linear mode signals corresponding to higher masses, one of which was assigned to H4 histone (m/z 11394) (see Figure 9). This assignment was based both on the known mass of H4 histone and on its expected presence in the brain tissue. Other unassigned proteins were seen in the mass region up to ∼30 kDa and will be discussed in more detail elsewhere as part of an ongoing study to define the upper mass limits of the gold cluster matrixes. CONCLUSION Implantation of massive metallic cluster ions (Au4004+, 100 eV/ at.) is a new method for incorporation of a metal matrix into the top layers of solid bioorganic samples. The implantation is not destructive for organic material and makes it amenable to UV laser desorption/ionization analysis. The MILDI method in the current stage is shown to be efficient in the mass range from some hundreds to some thousands of daltons for peptides and small proteins. It can be applied to the analysis of nonchemically

Figure 8. Positive ion MILDI spectrum in the mass range 300-1000 Da obtained in the reflectron mode from a rat brain tissue slice implanted with a fluence of 5 × 1012 Au400/cm2. Peaks of gold clusters were used for the internal mass calibration. 7292 Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

in response to laser irradiation makes the MILDI method particularly attractive for further applications such as molecular imaging of biological tissues with laser microprobes. However, as with other particulate matrixes, the desorbed bioanalytes are often highly cationized and the upper end mass limit for the detection of proteins is not yet known.

ACKNOWLEDGMENT

Figure 9. Positive ion MILDI spectrum in the mass range 10 00013 000 Da obtained in the linear mode from a rat brain tissue implanted with a fluence of 5 × 1012 Au400/cm2. The peaks MW matched that of a common brain protein, histone H4.

modified biological samples such as tissues and cells. The excellent homogeneity of ion emission from the implanted surface

J.A.S. thanks Marjorie Schultz for the use of personal funds to support Ionwerks’ participation in this work. The authors also thank the reviewers for several very useful suggestions that have been incorporated into the manuscript. A.S.W. thanks ONDCP for instruments funding, without which this and other projects could not have been done.

Received for review June 14, 2004. Accepted September 16, 2004. AC049123I

Analytical Chemistry, Vol. 76, No. 24, December 15, 2004

7293