RAPID COMMUNICATIONS IN MASS SPECTROMETRY Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

New aspects in matrix-assisted laser desorption/ ionization time-of-¯ight mass spectrometry: a universal solvent-free sample preparation S. Trimpin1, A. Rouhanipour1, R. Az2, H. J. Ra¨der1* and K. Mu¨llen1 1

Max-Planck-Institute for Polymer Research, Ackermannweg 10, 55128 Mainz, Germany Clariant GmbH, PTRF Pigmentanalytik, 65926 Frankfurt, Germany

2

Received 13 February 2001; Accepted 21 May 2001

SPONSOR REFEREE: Professor Dr. J. Grotemeyer, Institut fu¨r Physikalische Chemie der Universta¨t Kiel, Germany

A method of solvent-free sample preparation is shown to be of universal applicability for matrixassisted laser desorption/ionization mass spectrometry (MALDI-MS). Results obtained were compared with those of traditional solvent-based sample preparation for MALDI-MS in order to demonstrate their similarities with respect to accuracy, sensitivity and resolution for polymers such as polystyrene and poly(methyl methacrylate) in a mass range from 2 to 100 kDa. The results revealed that there is fundamentally no difference in the quality of the obtained mass spectra, and we conclude that the mechanism of desorption and ionization remains unchanged. However, the solvent-free sample preparation turned out to have some advantages over the traditional method in certain cases: quick and easy applicability is shown for polyetherimide avoiding time-consuming optimization procedures. In particular, industrial pigments that are insoluble in common solvents were characterized without interfering signals from fragments. The method even showed improvements with respect to reproducibility and mass discrimination effects in comparison to traditional sample preparation. Additionally, this contribution provides new insight regarding the analyte/matrix preorganization for the desorption step which now appears to be independent of crystallinity. Copyright # 2001 John Wiley & Sons, Ltd.

Since the late 1970s, lasers have been used to desorb ions from samples deposited on metallic surfaces.1,2 This approach coupled with mass spectrometric detection permitted the analysis of ions derived from organic and inorganic compounds up to 9 kDa, most often accompanied by high levels of fragmentation.3 The development of matrix-assisted laser desorption/ionization (MALDI)4,5 allowed the detection of intact, soluble molecules of substantially higher masses (>1 million Da),6 while minimizing fragmentation. Hence, MALDI-MS is considered as a very `soft' desorption/ionization process in mass spectrometry.7,8 It is assumed that this `soft' process is due to preorganized crystal formation, where the analyte is built in the matrix crystal during solvent evaporation, enabling a proper matrix assistance during the desorption step.9,10 Evidently, one of the key factors that greatly influences the success and the quality of MALDI mass spectrometric analysis is the sample preparation step. Prerequisites for a common sample preparation method require soluble analytes and matrices and the compatibility of the solvents employed for both the matrix and analyte systems, subsequent solvent removal and the homogeneity of the crystallized analyte/matrix mixture

*Correspondence to: H. J. RaÈder, Max-Planck-Institute for Polymer Research, Ackermannweg 10, 55128 Mainz, Germany. E-mail: [email protected] DOI:10.1002/rcm.372

on the surface of the MALDI sample holder. The search for an appropriate preorganization of sample and matrix led to the development of various sample preparation methods in order to modulate the crystal growth. Homogeneity of the final analyte/matrix mixture can be reached through the proper choice of solvents,11 matrices,12,13,14,15 mixed matrices,16 co-matrices17,18 and surfactants,19,20 pH,21,22 presence of metal salts23 and counter anions,24 concentration in solution and the matrix-to-analyte ratio,25 sample deposition density and the deposition protocol such as dried droplet,26 thin-27 and seed-layered,28 spin-coated29 and electrospray30 deposition. These factors can considerably affect the ion yield and can also introduce severe mass discrimination effects. Drawing a conclusion for the conventional solvent-based sample preparation for MALDI-MS (solvent-based MALDI-MS), the influencing key parameters are too many. Hence, there is a need for a more general method that is easily applicable and has less critical parameters. Attempts to realize this were recently published with a solvent-free pressed pellet sample preparation31 that analyzed polyamide. Another method of solvent-free MALDI sample preparation for MALDI-MS (solvent-free MALDI-MS) characterized intractable macromolecules such as giant polycyclic aromatic hydrocarbons (PAHs),32 polydithiathianthrene,33 as well as polyfluorene,34,35 that were difficult to analyze by any known analytical method. In a previous study36 we reported the advantages of Copyright # 2001 John Wiley & Sons, Ltd.

Solvent-free sample preparation in MALDI

1365

Figure 1. Structures of the analytes investigated (excluding bovine insulin).

solvent-free preparation in comparison to conventional solvent-based MALDI-MS whereas the aim of the present contribution is to prove the anticipated universal character of the new preparation method. The MALDI mass spectra obtained by this solvent-free sample preparation method are thus compared with mass spectra obtained by conventional solvent-based MALDI-MS and by LD-MS under strictly identical experimental conditions. Special emphasis is given to the sensitivity, resolution and shot-to-shot reproducibility utilizing the solvent-free MALDI-MS method. Further tests to evaluate the `universal' character of the method include the upper molecular weight limit and the applicability to various kinds of different analytes (e.g. organic compounds, biopolymers). Finally, the implications of this solvent-free sample preparation method with respect to the hypothetical model of MALDI-MS is also briefly discussed.

EXPERIMENTAL Materials Analytes

Polystyrene (PS; 2, 7, 27, 44 and 96 kDa) and poly(methyl methacrylate) (PMMA; 2, 7, 28, 52 and 100 kDa) were obtained from Polymer Standards (Mainz, Germany). Polyetherimide (PEI; 30 kDa) was purchased from Polysciences Europe (Eppelheim, Germany). The pigments such as C.I. Pigment Red 144 were provided by Clariant GmbH (Frankfurt, Germany) and originated from industrial production. Bovine insulin was purchased from Fluka (Buchs, Switzerland). The structures of the analytes are depicted in Fig. 1. Copyright # 2001 John Wiley & Sons, Ltd.

Matrices

1,8,9-Trihydroxyanthracene (dithranol) and 7,7,8,8-tetracyanoquinodimethane (TCNQ) were obtained from Aldrich (Steinheim, Germany). a-Cyanohydroxycinnamic acid (HCCA) was obtained from Sigma (St. Louis, USA). Silver trifluoroacetate and lithium trifluoroacetate were purchased from Aldrich (Steinheim, Germany) and used without further purification.

Solvents

Tetrahydrofuran (THF) was obtained from Fluka (Buchs, Switzerland) and o-dichlorobenzene from Aldrich (Steinheim, Germany). Water used was of deionized quality. Dichloromethane was obtained from Fisher Chemicals (Loughborough, UK). Acetonitrile was purchased from Riedel-DeHaen (Seelze, Germany).

Sample preparation General procedure for sample preparation for solventfree MALDI-MS

For a direct comparison of solvent-free and solvent-based MALDI-MS the components, i.e. analyte, matrix and cationizing salt, were mixed in a ball mill in the solid state with total exclusion of solvents. Mixing of the powders was performed using a ball mill MM 2000 (F. Kurt Retsch GmbH & Co. KG, Haan, Germany): analyte and matrix were mixed initially in a molar ratio 1:50. Appropriate cationizing salts were added to the analyte/matrix mixture in a molar analyte/salt ratio of 1:10 also without solvents. The sample/matrix/salt mixture (10:50:10) was ground for 10 min at room temperature. The matrix dilution (up to 1:5000) was carried out by adding matrix to the pre-prepared Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

1366 S. Trimpin et al. Table 1. Various aspects investigated: analytes and detailed preparation conditions Investigated aspect Optimal molar ratios

Calibration, sensitivity and resolution High molecular weight analysis Application to PEI Application to pigment red 144 Application to bovine insulin

Molar mixing ratios of analyte/matrix/salt PS (2 kDa)/dithranol/Ag: 1:0:10; 1:50:10; 1:500:10; 1:5000:10 PMMA (2 kDa)/dithranol/Li: 1:0:10; 1:50:10; 1:500:10; 1:5000:10 PMMA (7 kDa)/dithranol/Li: 1:0:10; 1:500:10 PS (2 kDa)/dithranol/Ag: 1:300:10; 1:500:10 PMMA (28, 52, 100 kDa)/dithranol/Li: 1:500:10 PS (7, 27, 44, 96 kDa)/dithranol/Ag: 1:500:10a),b) PEI (30 kDa)/dithranol: 1:250:0c) Pigment red 144/TCNQ: 1:0:0; 1:500:0 Bovine insulin/HCCA: 1:500:0

Solvent THF THF THF THF THF THF CH2Cl2 THFd) Water/acetonitrile (50/50)

a)

PS 96 kDa with dithranol as matrix appeared to give higher quality mass spectra when the solvent-free sample mixing was carried out by cooling with a liquid nitrogen cooling device for the ball mill container. b) Mortar and pestle treatment could also be applied alternatively for homogenization of the sample mixtures and was successful for PS (7, 27 and 44 kDa). c) Ground for 20 min. d) Suspension.

1:50:10 analyte/matrix/salt mixture and was then further milled for 10 min at room temperature to produce a MALDI powder mixture of 1:500:10. The same procedure was applied to produce a 1:5000 dilution from the 1:500 mixture. The powder mixtures are stable after storage for several months.

Detailed preparation conditions

The sample preparation protocols for the individual analytes, as well as notes of any slight changes to the general procedure, are given in Table 1.

Systematic approach for the comparison of solvent-free MALDI-MS with different sample preparation methods

To ensure exactly the same conditions while comparing solvent-free MALDI-MS with various MS sample preparation methods, the samples which had been prepared under exactly the same conditions (e.g. molar ratios, matrices, salt) were measured at the same time and with the same instrumental settings of the mass spectrometer (e.g. laser power, extraction voltage, detector settings).

Comparison of solvent-free and solvent-based MALDI-MS

The MALDI powder mixtures obtained were divided into two parts: one part was crushed on the target as a very fine powder to produce a very thin coverage of the analyte/ matrix mixture, the other one was completely dissolved in an appropriate solvent respectively and applied to the sample holder as a solution. After rapid evaporation of the solvent, both samples were measured immediately.

Comparison of different solvent-free MALDI-MS preparation methods

The pre-mixed analyte/matrix/salt powder mixture was divided into two parts: the first one was applied directly as sample powder to the target (solvent-free sample preparation), the other was first pressed to a very thin disc analogous to a KBr pellet (pressed pellet sample application) conventionally used in IR spectroscopy and then fixed on the target by double-sided conductive tape. Copyright # 2001 John Wiley & Sons, Ltd.

Comparison of solvent-free and solvent-based MALDI-MS with LD-MS

In contrast to conventional LD-MS sample preparation, where the sample is first dissolved in an appropriate solvent, the sample powders were applied after ball mill mixing with the appropriate salt directly to the target as a powder to allow direct comparison with solvent-free MALDI-MS. The LD-MS comparison with solvent-based MALDI-MS was obtained by dissolving with the appropriate solvent. It should be noted that with both the solvent-free MALDIMS and the LD-MS sample preparation methods there is a risk of contaminating the ion source of the mass spectrometer by losing powder from the target. To avoid this, excess powder with insufficient adhesion at the surface of the target is removed by air-blowing. Furthermore, only very thin powder films were created in order to avoid calibration errors and loss of sensitivity.

Mass spectrometry MALDI-TOF instrumentation

LD-TOF and MALDI-TOF mass spectra were recorded using a Bruker Reflex II2 MALDI-TOF mass spectrometer (Bremen, Germany) equipped with a N2 laser (l = 337 nm) operating at a pulse rate of 3 Hz. The ions were accelerated with pulsed ion extraction (PIE2 design from Bruker) by a voltage of 20 kV. The analyzer was operated in reflection mode and the ions were detected using a microchannel plate detector. For negative ion mode measurements, all the parameters were kept identical except that the polarity of the acceleration was reversed. MALDI-TOF mass spectra were recorded in the m/z range 25±100 kDa, using a Bruker Reflex I2 MALDI-TOF mass spectrometer (Bremen, Germany) equipped with a N2 laser (l = 337 nm) operating at a pulse rate of 3 Hz. The ions were accelerated by a voltage of 28.5 kV. The analyzer was operated in linear mode and the ions were detected using the Bruker HIMAS2 detector.

Calibration in MALDI-MS and LD-MS

Calibration was carried out before each measurement. In positive ion mode, analytes up to 10 kDa were calibrated Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

Solvent-free sample preparation in MALDI

1367

Figure 2. Mass spectra of PMMA (2 kDa) obtained with different mixing ratios of PMMA/dithranol/ Li-salt i) by solvent-free MALDI-MS: (a) 1:50:10, (b) 1:500:10, (c) 1:5000:10; ii) by solvent-based MALDI-MS: (d) 1:50:10; (e) 1:500:10; (f) 1:5000:10.

using 2 kDa polystyrene (PS), while, above 10 kDa, 3.5 kDa PS was used. PS was dissolved in THF together with dithranol and silver salt (1:500:10). A multipoint calibration of the isotopic resolved PS-oligomers was carried out. In negative ion mode a two-point calibration is obtained using the lowest of the isotopes of the resolved C60 and C70 fullerene.

RESULTS AND DISCUSSION Optimal molar analyte/matrix ratios

For easy comparison of results obtained by conventional and solvent-free MALDI with respect to sensitivity and resolution, well-characterized synthetic polymer standards such as polystyrene (PS) and poly(methyl methacrylate) (PMMA) were used which are available with high purity and narrow polymer distributions at different molecular weights. The mass spectra of PMMA (2 kDa, Fig. 2) show almost identical results for both preparation methods with respect to the most important information such as the absolute molecular weight values of the resolved signals and the shape of the molecular weight distribution. Additionally, there is no significant change in the shape of the molecular weight distribution at different molar mixing ratios of the analyte and matrix from 1:50 to 1:500. Hence, the overall appearance of the mass spectra does not change and remains virtually the same for solvent-free and solvent-based MALDI-MS. Solvent-free MALDI-MS, however (Figs 2(a), (b) and (c)), shows advantages over solvent-based MALDIMS (Figs 2(d), (e) and (f)) with better signal intensities (ca. 3fold stronger), thus resulting in an improved signal-to-noise Copyright # 2001 John Wiley & Sons, Ltd.

ratio in the mass spectra for all the molar analyte/matrix ratios examined. Interference from the matrix is less intense as well as reduced baseline shifting with increasing matrix dilution applying solvent-free MALDI-MS. When solventbased MALDI-MS was applied, a more complex mass spectrum was obtained due to an undesirable competition between pseudo-molecular ion formation with different cations (in particular, between lithium and sodium) which was more pronounced for increased matrix dilution (Figs 2(e) and (f)). This disadvantage is negligible when the solvent-free sample preparation is employed and this is most apparent at a dilution of 1:5000 where the quality of the mass spectrum is greatly improved (Figs 2(c) and (f)). The optimum mixing ratio of the analyte and matrix (Fig. 2) is found at 1:500 for both preparation methods; however, solvent-free MALDI-MS shows an improved tolerance for other mixing ratios, since the appearance of the spectra in Figs 2(a)±(c) remains virtually the same, in contrast to the spectra from solvent-based MALDI-MS shown in Figs 2(d)± (f). It should also be mentioned that cationizing metal salts (e.g. lithium trifluoroacetate, Table 1) are found to be suitable in solvent-free MALDI-MS, similar to conventional MALDIMS. Metal ion attachment to form pseudo-molecular ions is rather important since polymeric analytes commonly depend on cationizing agents other than proton sources. Similar investigations were carried out on PS (2 kDa) as for PMMA and similar conclusions could be drawn. There is strong overall conformity in the results of both preparation methods that can be traced back to the narrow polymer distribution of these polymer standards. Solvent-free MALRapid Commun. Mass Spectrom. 2001; 15: 1364±1373

1368 S. Trimpin et al.

Figure 3. Isotopic distribution of a polystyrene (20mer) with molar ratio of 1:500:10 (PS/ dithranol/Ag-salt): (a) computer simulation (resolution 5000); (b) solvent-based MALDI-MS (resolution 4000); (c) solvent-free MALDI-MS (resolution 5000); (d) solvent-free pressed pellet sample application for MALDI-MS (resolution 3000).

DI-MS shows no disadvantages in comparison to solventbased MALDI-MS in the analysis of PMMA and PS and we assume that this will also hold for other analyte/matrix systems. To demonstrate the significant difference between direct excitation of PMMA (2 kDa) in laser desorption/ ionization in contrast to measurements with matrix assistance, the investigations were supplemented with LD-MS

measurements. These LD-MS results indicate that the LD mass spectra of these analytes were only obtainable at high laser power, with low signal intensities and significantly different curve shapes of the polymer distribution. When identical investigations were performed for PMMA (7 kDa) it was shown that both MALDI-MS sample preparation methods allowed the characterization of the intact molecular

Figure 4. Comparison of polystyrene (2 kDa) MALDI mass spectra obtained by different solvent-free sample applications with respect to different polystyrene/dithranol/Ag-salt ratios: (a) pressed pellet application for MALDI-MS (1:300:10); (b) solvent-free MALDI-MS (1:300:10); (c) pressed pellet application for MALDI-MS (1:500:10); (d) solvent-free MALDI MS (1:500:10). Copyright # 2001 John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

Solvent-free sample preparation in MALDI

1369

Figure 5. Mass spectra of polystyrene (96 kDa) smoothed and baseline corrected: (a) solvent-free MALDI-MS; (b) conventional solvent-based MALDI-MS.

ions whereas LD-MS failed completely. Thus, desorption and ionization in this molecular weight range is not accessible without matrix assistance for PMMA.

Calibration, sensitivity and resolution

A representative indication of mass resolution and calibration accuracy can be obtained from PS (2 kDa) obtained for the two different methods of sample preparation using the optimum mixing ratio of 1:500:10 (analyte/matrix ratio). When isotopic masses were evaluated for a PS 20-mer (2246 Da, Fig. 3) using external calibration, the molecular weight deviation was <30 ppm for both sample preparation methods. The resolutions obtained are compared with a simulated resolution of 5000 (Fig. 3(a)), which is the isotopically resolved distribution of the 20-mer. The MALDI-MS resolution is about 4000 for solvent-based (Fig. 3(b)), 5000 for solvent-free sample preparation (Fig. 3(c)), and 3000 when a pellet is pressed (Fig. 3(d)). Thus, direct powder application showed better resolution and sensitivity than the pressed pellet application. The higher density of the pressed pellet matrix/analyte mixture seems to hamper the desorption step, probably because of an increased surface energy. Subsequent investigations varying the analyte/matrix/salt molar ratios of the two different solvent-free powder application methods to the target however showed remarkable differences in the quality of PS (2 kDa) mass spectra comparing molar ratios of 1:500:10 and 1:300:10. When the 1:300:10 ratio sample was investigated, the mass spectra of the solvent-free sample preparation method (Fig. 4(b)) showed a 3-fold increase in signal intensity and improved mass resolution compared to the pressed pellet application (Fig. 4(a)). Dilution by the matrix to 1:500:10 then led to a strong increase in quality in the case of solvent-free MALDIMS (Fig. 4(d)) whereas, in contrast, a large decrease in quality was observed with the pressed pellet application (Fig. 4(c)), appearing as low signal intensities and signal-tonoise ratios. The pressed pellet application gave good MALDI-MS results but showed disadvantages over direct Copyright # 2001 John Wiley & Sons, Ltd.

powder application to the target: a time-consuming pellet preparation and a decreased tolerance of less optimum molar analyte/matrix ratios leading to a rapid decrease in quality in the MALDI mass spectra with higher matrix dilution. Similar results have been reported for the pressed pellet sample preparation of a <5-kDa polyamide.31 The rather unusual analyte/matrix ratio of the pressed pellet application indicates less effective matrix assistance and is in strong contradiction to solvent-free and solvent-based MALDI-MS. It would appear that pressing pellets from the sample powder can be avoided as long as care is taken to ensure good adhesion of the analyte/matrix/salt powder mixture to the target.

High molecular weight analysis

A further aspect of universality is to prove the potential of this method in the analysis of high molecular weight samples. A mass range up to 100 000 Da was selected for study by the new and traditional methods of sample preparation. Figure 5 shows the mass spectra of PS (96 kDa) as a representative example for the results obtained by the two different methods for PMMA (28, 52 and 100 kDa) and PS (7, 27, 44 and 96 kDa) samples. Solvent-free MALDIMS yielded mass spectra with almost equal quality to solvent-based sample preparation. Slight advantages were observed for solvent-based MALDI-MS with respect to signal intensity. It should be emphasized that, in the particular case of PS, increasing molecular weight increases the waxy nature of this polymer which hampers the homogenization of the matrix and analyte during grinding. Thus, external cooling of the ball mill container was applied to increase the brittleness of the PS (96 kDa) sample facilitating the formation of a homogeneous powder of the analyte/matrix mixture. Increasing quality of mass spectra obtained by solvent-free MALDI-MS was in the following order of different mixing methods: mortar treatment < ball mill < cooled ball mill. The strongest correlation to the quality of the mass spectra is only indirectly dependent on Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

1370 S. Trimpin et al.

Figure 6. Mass spectra of polyetherimide: (a) solvent-free MALDI-MS; (b) conventional solvent-based MALDI-MS.

the molecular weight but given by the brittleness of the analyte and matrix themselves and their ability and tendency to form homogenous mixtures with the chosen matrix. Thus, ensuring homogeneity in the analyte/matrix mixture is the key step in the success of solvent-free MALDIMS. The final particle size after ball mill treatment is described to be in the range of one micrometer37 but strongly depends on the nature of the analyte. It can be easily enhanced when necessary by extension of the grinding time and/or external liquid nitrogen cooling of the ball mill container during milling. This enhancement reflects the advantages of a ball mill over a mortar and pestle treatment: preparative conditions were more easily adaptable to the situation whereas time demands for preparation were almost identical for both grinding methods. High grinding efficiency to mm size is obtained with the ball mill treatment and insures homogeneity and high spot-to-spot reproducibility even for high mass polymers, whereas mortar and pestle treatment is valuable but is inadequate in special cases even when applied extensively.

Application of solvent-free MALDI-MS to granula PEI

Polyetherimide (PEI, 30 kDa; Fig. 1) has not been the subject of MALDI-MS investigations so far. Solvent-free MALDI-MS allowed direct acquisition of MALDI mass spectra with high quality, good signal-to-noise ratio and sensitivity (Fig. 6(a)). A conventional solvent-based MALDI mass spectrum could not be obtained in equal quality and especially equal signal intensity using CH2Cl2 (Fig. 6(b)) which is the best-known solvent for PEI.38 An optimum solvent matrix system could presumably be found for this sample, although this approach was not pursued. This example indicates the additional advantages of solvent-free MALDI-MS, such as stronger signal intensities as well as unnecessary timeconsuming optimization procedures. The well-known suppression effects39,40 of higher molecular weight components Copyright # 2001 John Wiley & Sons, Ltd.

in MALDI-MS still remain the same independent of the kind of sample preparation (Figs 6(a) and (b)).

Application of solvent-free MALDI-MS to pigment red 144

The most important evidence of matrix assistance in solventfree MALDI-MS can be expected for substances that require matrix assistance for the production of molecular ions and, hence, are not accessible by LD-MS. Promising candidates in the low mass range are substances that absorb at the laser wavelength (337 nm in case of a N2 laser). Fragmentation is increased when the analyte has significant absorption at the wavelength of the desorption laser, thus releasing the absorbed energy during desorption and ionization by molecular decay, whereas MALDI causes a large decrease in the fragmentation tendency.41,42 In the comparison of the LD-MS and MALDI-MS techniques, 20 pigments of industrial relevance were chosen and the results for pigment red 144 are shown in Fig. 7. Pigment red 144 (for structure, see Fig. 1), which up to now has only been characterized by conventional methods such as thin layer chromatography (TLC) and IR spectroscopy, is only soluble in solvents with high polarity (partially soluble in hot o-dichlorobenzene) or acidity (concentrated sulfuric acid, trifluoroacetic acid). All these solvents are less advantageous for common MALDI matrices and the acidic solvents create additional problems due to the oxidation tendency of pigment red 144. Pigment red 144 has a relatively low molecular weight but only a fragment is accessible by LD-MS. It should be stressed that pigment red 144 is a mass-produced industrial product and consists of a product mixture which in general can cause substantial suppression effects and equivocal interpretation. In the LD mass spectra of pigment red 144 (Fig. 7(a)) no intact molecular ion was observed; however, a signal at m/z 809 was, and this suggests a loss of OH. Post-source decay (PSD) MALDI experiments showed that the fragment was not formed in the drift-tube of the TOF mass spectrometer Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

Solvent-free sample preparation in MALDI

1371

Figure 7. Mass spectra of pigment red 144: (a) solvent-based LD-MS; (b) solventbased MALDI-MS; (c) solvent-free MALDI-MS.

(data not shown). The signal at m/z 809 can therefore be attributed to an in-source fragmentation (ISD) or an unknown side-product coupled with a high suppression effect of the molecular ion. These unsatisfactory results demonstrate the limitation of LD-MS. In contrast, the solvent-free sample preparation enables MALDI mass spectra that indeed show the intact molecular ion as a radical cation (Fig. 7(c)) and sodiated pseudo-molecular ion as well as traces of its dechlorinated side-product as a radical cation (Fig. 1, R = H). The solubility limitations of pigment red 144 only allow a suspension of the sample in THF for solvent-based MALDI-MS. Nevertheless, solvent-based MALDI-MS (Fig. 7(b)) also gives an intact molecular ion and therefore indicates a certain degree of matrix assistance. We

assume that the poor performance of conventional solventbased MALDI-MS in this case is due to the inhomogeneity in the matrix/sample system caused by the limits of the liquid suspension and by additional segregation during the evaporation of the solvent in the crystallization process. Thus, the comparison of LD-MS, solvent-based and solventfree MALDI-MS methods of such a UV-absorbing compound shows the latter method as giving an indisputable matrix assistance, high sensitivity and indicates a high sample homogeneity reflected in a high shot-to-shot reproducibility. Hence, solvent-free MALDI-MS clearly shows the best results of the methods under comparison in this investigation. It should be emphasized that this new solvent-free MALDI-MS method overcomes not only the

Figure 8. Mass spectra of bovine insulin: (a) solvent-free MALDI-MS (positive ions); (b) conventional solvent-based MALDI-MS (positive ions); (c) solvent-free MALDI-MS (negative ions); (d) conventional solvent-based MALDI-MS (negative ions). Copyright # 2001 John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

1372 S. Trimpin et al.

incompatibility of analytes/matrices/solvents due to the very low solubility of this analyte in organic solvents, but also the tendency of this analyte to oxidize during the dissolving step that is required for solvent-based MALDIMS with acids. Additionally, this contribution demonstrates for the first time the characterization of pigment red 144 using a MALDI-MS approach.

Application of solvent-free MALDI-MS to bovine insulin

To evaluate a method as universal, general investigations should include a variety of very different analytes. MALDIMS has its broadest application in the area of biopolymer analysis. The feasibility of this solvent-free sample preparation method was therefore briefly tested for bovine insulin, a biopolymer used as a calibration standard in solvent-based MALDI-MS. This analyte should also allow an assessment of its applicability to negative ions.43 The solvent-free MALDI mass spectra obtained for bovine insulin detected in both positive and negative ion mode (Figs 8(a) and (c)) were of equal quality. Slight advantages are found with respect to signal intensities in both detection modes for the solventbased MALDI mass spectra (Figs 8(b) and (d)). Solvent-free MALDI-MS showed no restriction for this biopolymer sample and seems to have no differences in the MALDI mechanism in general. Also no principal disadvantages in ionization and detection for solvent-free MALD-MS were observed. Even though the solvent-free sample preparation is straightforward, handling of the analyte/matrix mixture has one limitation with respect to the minimum sample amount for the ball mill container (1.5 mL in volume): up to now, the smallest analyte/matrix mixture that can be conveniently handled is about 3 mg. This limitation especially affects bioanalytical applications where sample amounts are much smaller. However, this limitation is only due to the size of the commercially available equipment and can be overcome, e.g., by specially developed smaller ball mill containers.

Implications of solvent-free MALDI-MS for the hypothetical model of the MALDI-MS mechanism

The overall evaluation of solvent-free MALDI-MS, basically a mechanical mixing of analyte and matrix without any solubilization procedures, provides true matrix assistance in the desorption/ionization process while preventing inhomogeneity and fragmentation. Solvent-free MALDI-MS showed no principal disadvantages over the conventional sample preparation step (e.g. preparation time) or in the MS analysis step (e.g. desorption, ionization or detection) up to 100 kDa but is of course still governed by common MALDIMS limitations (e.g. polydispersity). The variety of different analytes analyzed in our work indicated that there seems to be no restriction due to the nature and the origin of the analyte (e.g. synthetic polymers, biopolymers, organic compounds). Solvent-free MALDI-MS gave mass spectra of equal or slightly improved quality in comparison to solventbased MALDI-MS. This is most clearly demonstrated in the mass range up to 10 kDa. Important drawbacks for high Copyright # 2001 John Wiley & Sons, Ltd.

molecular weight analysis are not found up to 100 kDa. Generally, a higher shot-to-shot reproducibility was found for solvent-free MALDI-MS. Taking all these experiments and its new aspects into account, the theoretical model of MALDI-MS has to be discussed in a new light to allow these experimental MALDI-MS results to be explained. Recently Hillenkamp et al. presented results showing that it is not a prerequisite for the analyte to be uniquely embedded in a matrix crystal but that this kind of preorganization on a molecular level leads to better results in matrix assistance.44 The solvent-free MALDI-MS results presented here confirm that there is no demand for a highly crystalline analyte/matrix preorganization on a molecular level; the original hypothesis of an analyte embedded in matrix crystal seems to be indeed a misconception.9,10 This conclusion is supported as follows: it has to be assumed that solvent-free MALDI-MS sample preparation procedures produce particles with grain sizes of about one micrometer.37 A micrometer scale is at least several orders of magnitude larger than molecular dimensions. Thus, with the solvent-free MALDI-MS procedure, we do not expect to obtain a truly homogeneous mixture between matrix and analyte in molecular dimensions but we assume to achieve at least a partial homogenization on a molecular level which is sufficient for successful measurements, as experimentally verified. Vice versa, the results suggest that in solvent-based MALDI-MS the homogeneity obtained in the solid state is much poorer than expected traditionally. To visualize the generality of segregation in solvent-based sample preparation, it has to be remembered that (re)-crystallization is used as a cleaning process in synthetic chemistry. Hence, any crystallization procedure of two or more compounds involves separation phenomena in differing degrees. Intrinsically, any conventional solventbased MALDI-MS procedure has to show inhomogeneity to a certain degree due to the crystallization step. This limitation cannot be attributed to solvent-free MALDI-MS at all. However, it cannot be excluded completely that the mechanical treatment causes changes in the morphology on a molecular level. Further assumptions to the theoretical aspect of solvent-free MALDI-MS are that an adsorption model between analyte and matrix might be considered which does not depend on crystallinity.

CONCLUSIONS This contribution underlines the similarities between the traditional method of solvent-based and the new solventfree MALDI-MS methods. Proof of true matrix assistance in solvent-free MALDI-MS has been demonstrated for any analyte investigated and showed in general higher signal intensities and better signal-to-noise ratios. Hence, this allows solvent-free MALDI-MS to be established in principal as an additional and complementary preparation method for MALDI-MS. The advantages of solvent-free over solvent-based MALDI sample preparation shown in this investigation indicate its potential: additional advantages of solvent-free MALDI-MS can be expected for problematic analytes such as compounds where segregation effects due to solvent removal play a Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373

Solvent-free sample preparation in MALDI

major role when MS measurements fail. Other problems during sample preparation such as, e.g., miscibility, segregation and suppression effects, solubility and incompatibility restrictions, etc., can also be minimized. Time-saving aspects for complete characterization of industrial product mixtures are useful evaluation goals, too. Solvent-free MALDI-MS also allows investigations regarding completely new matrix and analyte combinations since the choice is no longer dependent on the compatibility of the solvent system. The most useful and promising advantage, however, is the capability of solvent-free MALDI-MS to characterize insoluble samples in general. This might also be valuable with respect to the analysis of intractable proteins such as, e.g., membrane proteins that show strong solubility restrictions. Concerning the theoretical background of MALDI-MS we can conclude with these results that the hypothetical model of an analyte embedded in a crystalline matrix is indeed a misconception and that the most important factor influencing the quality of the MALDI mass spectra is the highly homogeneous character of the analyte/matrix mixture.

Acknowledgements

Dr. T. P. Knepper (ESWE-Institute, Wiesbaden, Germany) is gratefully acknowledged for his contribution to these investigations. The authors would also like to thank Dr. Brychcy (Clariant GmbH, Frankfurt, Germany) for helpful discussions and donation of samples.

REFERENCES 1. Posthumus MA, Kistenmaker PG, Meuzelaar HLC. Anal. Chem. 1978; 50: 985. 2. Stoll R, Rollgen FW. Org. Mass Spectrom. 1979; 14: 642. 3. Zenobi R. Int. J. Mass Spectrom. Ion Processes 1995; 145: 51. 4. Karas M, Bachmann D, Bahr U, Hillenkamp F. Int. J. Mass Spectrom. Ion Processes 1987; 78: 53. 5. Tanaka K, Waki H, Ido Y, Akita S, Yoshida Y, Yoshida T. Rapid Commun. Mass Spectrom. 1988; 2: 151. 6. Schriemer DC, Li L. Anal. Chem. 1996; 68: 2721. 7. Chan TWD, Colburn AW, Derrick PJ, Gardiner DJ, Bowden M. Org. Mass Spectrom. 1992; 27: 188. 8. Doktycz SJ, Savickas PJ, KruÈger DA. Rapid Commun. Mass Spectrom. 1991; 5: 145. 9. Chan TWD, Colburn AW, Derrick PJ, Gardiner DJ, Bowden M. Org. Mass Spectrom. 1992; 27: 188. 10. Doktycz SJ, Savickas PJ, KruÈger DA. Rapid Commun. Mass Spectrom. 1991; 5: 145. 11. Cohen SL, Chait BT. Anal. Chem. 1996; 68: 31. 12. Latourte L, Blais J-C, Tabet J-C. Anal. Chem. 1997; 69: 2742. 13. Karas M, Bahr U, Strupat K, Hillenkamp F. Anal. Chem. 1995; 67: 675.

Copyright # 2001 John Wiley & Sons, Ltd.

1373

14. Juhasz P, Costello CE, Bierman K. J. Am. Soc. Mass Spectrom. 1993; 4: 399. 15. Chou CW, Williams P, Limbach PA. Intern. J. Mass Spectrom. 1999; 193: 15. 16. Dale MJ, Knochenmuss R, Zenobi R. Anal. Chem. 1996; 68: 3321. 17. Vandell VE, Limbach PA. Rapid Commun. Mass Spectrom. 1999; 13: 2014. 18. Simmons TA, Limbach PA. J. Am. Soc. Mass Spectrom. 1998; 9: 668. 19. Kassis CM, DeSimone JM, Linton RW, Lange GW, Friedman RM. Rapid Commun. Mass Spectrom. 1997; 11: 1462. 20. Schnaible V, Michels J, Zeth K, Freigang J, Welte W, BuÈhler S, Glocker MO, Przybylski M. Int. J. Mass Spectrom. Ion Processes 1997; 169: 165. 21. Lin SH, Cotter RJ, Woods AS. Proteins (Suppl 2) 1998; 12. 22. Jensen C, Haebel S, Andersen SO, Roepstorff P. Int. J. Mass Spectrom. Ion Processes 1997; 160: 339. 23. Tang XD, Dreifuss PA, Vertes A. Rapid Commun. Mass Spectrom. 1995; 9: 1141. 24. Hoberg AM, Haddleton DM, Derrick PJ, Jackson AT, Scrivens JH. Eur. Mass Spectrom. 1998; 4: 435. 25. Bolbach G, Riaki K, Spiro M, Brunot A, Breton F, Blais JC. Analusis 1993; 21: 383. 26. Karas M, Hillenkamp F. Anal. Chem. 1998; 60: 2299. 27. Vorm O, Roepstorff P, Mann M. Anal. Chem. 1994; 66: 3281. 28. Onnerfjord P, Ekstrom S, Bergquist J, Nilsson J, Laurell T, Marko-Varga G. Rapid Commun. Mass Spectrom. 1999; 13: 315. 29. Perera IK, Kantartzoglou S, Dyer PE. Int. J. Mass Spectrom. Ion Processes 1996; 156: 151. 30. Hensel RR, King RC, Owens KG. Rapid Commun. Mass Spectrom. 1997; 11: 1785. 31. Skelton R, Dubois F, Zenobi R. Anal. Chem. 2000; 72: 1707. 32. Przybilla L, Brand JD, Yoshimura K, RaÈder HJ, MuÈllen K. Anal. Chem. 2000; 72: 4591. 33. Leuninger J, Trimpin S, RaÈder HJ, MuÈllen K. Macromol. Chem. Phys in press. 34. Setayesh S, Grimsdale AC, Weil T, Enkelmann V, MuÈllen K, Meghdadi F, List EJW, Leising G. J. Am. Chem. Soc. 2001; 123: 946. 35. Trimpin S, RaÈder HJ, MuÈllen K. Anal. Chem submitted. 36. Trimpin S, Keune S, RaÈder HJ, MuÈllen K. J. Am. Soc. Mass Spectrom submitted. 37. Retsch GmgH & Co. KG (Haan, Germany), operating instructions. 38. Polyscience Europe GmbH (Eppelheim, Germany). 39. Martin K, Spickermann J, RaÈder HJ, MuÈllen K. Rapid Commun. Mass Spectrom. 1996; 10: 1471. 40. Montaudo G, Montaudo MS, Puglisi C, Samperi F. Rapid Commun. Mass Spectrom. 1995; 9: 453. 41. RaÈder HJ, Spickermann J, Kreyenschmidt M, MuÈllen K. Macromol. Chem. Phys. 1996; 197: 3285. 42. Remmers M, MuÈller B, Martin K, KoÈhler W, RaÈder HJ. Macromolecules 1999; 32: 3190. 43. Morozov V, Merron B, Williams ID, Davies RJH, Hughes IG. Rapid Commun. Mass Spectrom. 1998; 12: 97. 44. Horneffer V, Dreisewerd K, Ludemann HC, Hillenkamp F, Lage M, Srupat K. Int. J. Mass Spectrom. 1999; 187: 859.

Rapid Commun. Mass Spectrom. 2001; 15: 1364±1373