Anal Bioanal Chem (2014) 406:2261–2278 DOI 10.1007/s00216-014-7646-6

REVIEW

Recent methodological advances in MALDI mass spectrometry Klaus Dreisewerd

Received: 6 December 2013 / Revised: 17 January 2014 / Accepted: 21 January 2014 / Published online: 21 March 2014 # Springer-Verlag Berlin Heidelberg 2014

Abstract Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) is widely used for characterization of large, thermally labile biomolecules. Advantages of this analytical technique are high sensitivity, robustness, highthroughput capacity, and applicability to a wide range of compound classes. For some years, MALDI-MS has also been increasingly used for mass spectrometric imaging as well as in other areas of clinical research. Recently, several new concepts have been presented that have the potential to further advance the performance characteristics of MALDI. Among these innovations are novel matrices with low proton affinities for particularly efficient protonation of analyte molecules, use of wavelength-tunable lasers to achieve optimum excitation conditions, and use of liquid matrices for improved quantification. Instrumental modifications have also made possible MALDI-MS imaging with cellular resolution as well as an efficient generation of multiply charged MALDI ions by use of heated vacuum interfaces. This article reviews these recent innovations and gives the author’s personal outlook of possible future developments. Keywords MALDI-MS . Halogenated matrices . Laser wavelength . Laser spot size . Liquid matrices . Highly charged ions

K. Dreisewerd Institute for Hygiene, Biomedical Mass Spectrometry, University of Münster, Robert-Koch-Str. 41, 48149 Münster, Germany K. Dreisewerd (*) Interdisciplinary Center for Clinical Research (IZKF) Münster, University of Münster, Domagkstr. 3, 48149 Münster, Germany e-mail: [email protected]

Introduction Remarkable improvements have been achieved since the first reports of the use of matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) for the analysis of peptides and proteins were published in the second half of the 1980s [1–3]. MALDI mass spectrometers now routinely enable the analysis of femtomole amounts of basic (e.g., tryptic) peptides with mass accuracies in the low-ppm range. Most current instruments are also equipped with dedicated means for tandem mass spectrometry, typically achieved by collision-induced dissociation (CID) of selected precursor ions [4–6]) or a combination of CID and post-source decay (PSD [7]) processes. In addition, in-source decay (ISD) fragmentation, a fast radicaldriven dissociation of analyte ions, is increasingly being used for top-down sequencing and, in fact, has partly replaced the classical Edman degradation approach [8]. However, because all ions are simultaneously subjected to the ISD process, the analysis of complex samples usually requires prior fractionation, e.g., by liquid chromatography (LC) [9]. For these purposes, robotic platforms, with which automated off-line spotting of LC eluent and matrix can be achieved, are marketed by several companies. Besides for instruments that are dedicated to specific tasks (e.g., large-scale high-throughput screening of DNA assays [10]) essentially all current instruments also allow performing MALDI-MS imaging (MALDI-MSI) experiments. The main advantage of MALDI-MSI over other (complementary) imaging techniques, such as fluorescence microcopy and immunohistochemistry, is the simultaneous label-free detection of a wide range of biomolecules and administered drugs in “histological” tissue slices [11]. Currently, the standard spatial resolution of MALDI-MSI is somewhere in the range 20– 100 μm, depending on the instrument and analytical task. Much effort has recently been made to reduce these values to the cellular level. For example, various procedures have

2262

been developed to achieve a homogenous coating of tissue with small crystals of sub-10 μm-dimensions. These methods range from applying the matrices by pneumatic, vibrational, or electrospray means to use of sublimation under vacuum [12] for generation of ultrafine matrix coatings. In many applications, pretreatment of tissue is another important step, e.g., to remove abundant lipids for enhanced analysis of peptides and proteins [13]. Many excellent overview papers have recently been published that discuss the different sample preparation approaches as well as possible problems (e.g., [11, 14–17]). The second major factor determining the spatial resolution, the implementation of laser beams with focal diameters of 10 μm and below [18], is discussed in detail within the present paper. The quality of the MALDI-MS analysis is affected by numerous other factors. For example, the physicochemical properties of the MALDI matrix determine the extent of cocrystallization with the analyte, their co-desorption on application of the pulsed laser light, and, not least, the efficiency of analyte ionization [19, 20]. All these factors contribute to the achievable limit of detection (LOD). Although a large number of MALDI matrices have been identified over the years, most analysts seem to rely on a relatively small set of say one to two dozen established compounds. α-Cyano-4-hydroxycinnamic acid (HCCA), sinapinic (or sinapic) acid (SA), and 2,5dihydroxybenzoic acid (DHB) are three examples of widely used matrices. These substances are particularly suitable for the analysis of peptides (HCCA/DHB) and/or proteins (DHB/SA); DHB is also widely employed for the analysis of carbohydrates and lipids. Another example of a matrix that is particularly suitable for the analysis of polar lipids, for example triacylglycerides, is 2,4,6-trihydroxyacetophenone (THAP) [21]; this compound is also an example of a less acidic, noncarboxylic acid matrix. More “specialized” matrices that can enhance specific features of the MALDI process or enable the analysis of specific types of analytes are also available . For example, because of its almost “neutral” pH, 6-aza-2thiothymine (ATT) is particularly well suited to analysis of non-covalently bound complexes [22, 23], and 1,5diaminonaphthalene (DAN) is a matrix with which high ISD yields are achieved [24, 25]. While most of the established MALDI matrices were, presumably, discovered in a “guided trial-and-error way”, a different innovative approach was recently pursued by Jaskolla and colleagues [26]. Based on density functional theory (DFT) calculations these researchers synthesized halogenated cinnamic acid derivatives in an effort to produce matrices with reduced proton affinities (PA) compared with standard HCCA. Sizable improvements of LOD and sequence coverage were demonstrated in proteomics applications when the results obtained with 4-chloro-αcyanocinnamic acid (ClCCA) and HCCA were compared [27]. The ion yield can also be optimized by adjustment of the irradiation conditions. The strong effect that the laser fluence (energy per irradiated area) exerts on the MALDI signal

K. Dreisewerd

intensities has been thoroughly investigated (e.g., [28–30]). A few of these studies also included measurements of ion yields (i.e., the ratio of ejected ions to the overall number of particles ejected) [31–33]. This work revealed that maximum ion yields are generally obtained in a laser fluence range ~2–4 times above the matrix and wavelength-dependent iondetection thresholds. Experiments that systematically investigated the UV MALDI-relevant wavelength range from approximately 280 to 370 nm were only conducted recently [32, 34]. These studies emphasized the importance of sufficient absorption by the matrix at the excitation laser wavelength. Because the solid-state absorption maxima of many matrices differ from the two standard laser wavelengths of 337 nm (N2 laser) and 355 nm (frequency-tripled Nd:YAG lasers), and also from the emission wavelength of the sometimes employed frequency-tripled Nd:YLF laser at 349 nm, it must be assumed that even many classical matrices are not optimally excited under standard MALDI conditions. Other irradiation parameters (e.g., the pulse duration, focal beam size, and focal beam intensity profile) have been investigated in the context of MALDI in a few experimental and theoretical studies [29, 31, 35–41], but their effect on the MALDI mechanisms is still only partly understood. This is somewhat surprising because, for example, the lateral resolution in MALDI-MSI experiments is co-determined by the laser spot size. The availability of lasers with pulse repetition rates in access of 1 kHz, together with fast sample stages and spectra acquisition protocols, has substantially reduced the data-acquisition times. This facilitates a high-throughput analysis in standard MALDI applications and allows scanning of relatively large sample areas in a reasonable time in MALDIMS imaging experiments [42, 43]. Laser pulse duration and laser penetration depth into the matrix material determine whether thermal or stress confinement conditions are achieved [44, 45]. Under thermal confinement conditions, energy dissipation by heat transport is negligible during the laser pulse. Because of the shallow laser penetration depths into the MALDI matrices of ~30–200 nm (defined as the 1/e-decrease in photon flux) thermal confinement is only partially fulfilled for standard UV-MALDI. As a consequence, some homogenization of the initially exponential temperature profile with sample depth z will occur during excitation with ns-long laser pulses [19, 46]. The typical pulse duration range for N2 and Nd:YAG lasers used for MALDI is between 0.5 and 6 ns and, thus, covers approximately an order of magnitude. In general, no large differences in the MALDI performance characteristics are notable when lasers of the same type but with slightly different pulse durations are used; in early work by this author, this was verified experimentally by comparing the MALDI results obtained with two N2 lasers emitting 0.6 and 3 ns long pulses, respectively [35]. Stress confinement does not occur in standard UVMALDI, because it takes no longer than a few tens of

Recent methodological advances in MALDI‐MS

picoseconds for an acoustic wave to propagate through the shallow excited matrix layers. Therefore, more sizable photoacoustic stress does not build up during excitation with a nslong laser pulse [30]. Picosecond and femtosecond lasers, with which stress confinement would be partially or even fully achieved, have been used in a few studies addressing the MALDI mechanisms [36, 46–48]. In addition to leading to different material ejection mechanisms, the use of excitation pulses in the ps and sub-ps range also affects the ionization pathways [49, 50], e.g., by switching from ladder-switching to ladder-climbing processes [46]. A few more interesting findings were obtained in these works, for instance concerning different upper mass ranges and different fragmentation pathways for ns and ps-excitation, respectively. However, few groups have later reported the use of ps-pulsed lasers for specific analytical purposes [51]. Because laser-penetration depths are generally greater in the near infrared (IR)—depending on the matrix and excitation wavelength typical values are in the 0.5–10 μm range [52]— stress confinement can have substantial effects in IR-MALDIMS [53–55]. In particular, build-up of sizable tensile stress amplitudes (up to several tens of MPa) on propagation of a bipolar stress wave through the material can contribute to explosive ablation of the material [44, 45, 56, 57]. Because, under these conditions, material ablation starts at lower temperatures, particularly labile ions such as phosphorylated and/or glycosylated peptides [58, 59] or non-covalent complexes [60] may be advantageously analyzed by IR-MALDI-MS. It remains to be investigated whether similar advantageous properties may eventually also be achieved if UV-MALDI was performed under stress-confinement conditions which were “equivalent” to those realized in these IR-MALDI experiments. MALDI is generally not regarded as a particular quantitative technique. For example, ion suppression effects and variation of sample morphology can both substantially affect ionization efficiency [61, 62]. Several studies have, however, demonstrated that by using particularly controlled experimental protocols and/or purification strategies a sufficiently quantitative response can often be obtained with standard crystalline matrices [63–65]. In particular, a significantly improved quantification of specific target compounds becomes possible if an internal standard is used [66]. Ideally, this would be an isotopically labeled analogue of the analyte of interest, but eventually the use of compounds with similar chemical composition and physicochemical properties may also suffice [67]. Because of their intrinsic homogeneity, quantification is easier to achieve if liquid MALDI matrices are utilized. One way to produce a liquid MALDI matrix with sufficient desorption and/or ionization properties is to dissolve the active matrix component in a vacuum-stable polar liquid, for example glycerol [68, 69]. To achieve sufficient solubility of the typically acidic matrices (e.g., HCCA, DHB) by deprotonation, a base (e.g., 3-aminoquinoline, pyridine, or

2263

triethylamine) is usually added as a third component. Excellent quantification of peptides over 2–3 orders of magnitude has been achieved by use of glycerol-based liquid matrix systems [70]. In some cases, the bi-component cation–anion system alone may form a vacuum-stable room temperature ionic liquid (RTIL) such that the glycerol support liquid may be omitted. With these matrices also, similar good quantification of peptides has been reported [71, 72]. A disadvantage of RTIL systems is that they are more tricky to use because some RTIL matrices tend to precipitate in the vacuum of the MALDI ion source. Glycerol-based systems are, therefore, probably more promising as liquid matrix systems. With the exception of very large biomolecules, such as proteins, singly charged ions are primarily produced by standard MALDI. Although this feature simplifies interpretation of spectra, it also has a few disadvantages. For example, more informative tandem mass spectra are typically obtained if multiply charged peptide ions are subjected to CID tandem MS [59, 73]. Also, some dissociation techniques, electron capture and electron transfer dissociation (ECD and ETD [74, 75]) cannot be applied to singly charged ions. Although increased generation of multiply charged MALDI ions of relatively small peptides has occasionally been reported [76, 77], due to a limited reproducibility and high LODs these findings did not lead to greater adoption of the underlying concepts. However, a few novel concepts, introduced more recently, enable the generation of multiply charged MALDI ions under more reproducible experimental conditions [78, 79]. These have attracted much interest, because they foster the coupling of MALDI ion sources with high-performance ESI mass spectrometers, for example QTOF and Orbitrap instruments that have only a limited m/z range, and improve tandem MS applications. One particularly promising concept is the use of liquid glycerol-based matrices along with an atmospheric pressure (AP) MALDI ion source and a heated (vacuum interface) ion transfer capillary [80]. In the following paragraphs, each of these five developments (new derivatized matrices, use of wavelength-tunable lasers, generation of small laser spot sizes for MS imaging, liquid matrices, and generation of highly charged MALDI ions) and their potential for improving the MALDI performance characteristics will be discussed in more detail. The article focuses on recent methodological innovations and, furthermore, on MALDI with UV lasers. The reader who is interested in further advances in standard MALDI MS and MALDI-MSI applications is referred to a comprehensive collection of current reviews and textbooks (e.g., [11, 14, 15, 81–89]). The article also focuses entirely on technical and experimental studies. Many excellent theoretical and combined theoretical and experimental papers have recently contributed to a better understanding of the MALDI processes and also led to a quite lively debate, in particular with regard to the complex MALDI ionization mechanisms [39, 41, 50, 90, 91].

2264

Novel designed MALDI matrices All commonly used solid-state MALDI matrices share a few important properties: they co-crystallize well with the specific type of analyte under investigation, absorb the laser light efficiently by excitation of electronic singlet states, and produce reactive ionic species that form the basis for an efficient ionization of analyte molecules by protonation, deprotonation, and/or adduct formation with alkali and/or halogen ions. Although the proton-transfer step is explained by different models [50, 92], all assume an active role of the matrix in this process. One major route is proton transfer from a protonated matrix molecule to the neutral biomolecule. The efficiency of this process generally increases with the increasing difference between the PA of the two reactants [62]. This feature, together with a sufficient optical absorption in the near UV, explains why aromatic carboxylic acids constitute a major class of MALDI matrices. The matrix PA can be reduced by incorporation of electronwithdrawing moieties (e.g., halogens) into the core structure (Fig. 1a) [26]. On the basis of high-level density functional theory (DFT) calculations, Jaskolla et al. recently introduced a set of halogenated cinnamic acid derivatives. In the case of the ClCCA matrix a chlorine atom substitutes for the hydroxyl group of HCCA in the para position [26]. This substitution results in a decrease in PA of ~24 kJ mol−1, due to the modified molecular electrostatic potentials (Fig. 1a) [34]. An even stronger reduction of approximately 29 kJ mol−1 is obtained for 2,4-difluoro-α-cyanocinnamic acid (DiFCCA), the first multiply halogenated compound to be introduced [93]. Because of their reduced PA, the halogenated matrices result in enhanced protonation of weakly basic, neutral, and/or acidic peptides. When a tryptic digest of bovine serum albumin was used as test system, an increase in the LOD of more than one order of magnitude and significant improvement in sequence coverage was observed when results obtained with ClCCA and HCCA matrices were compared (Fig.1b) [27]. Moreover, this effect was found to be widely independent of the specificity of the chosen protease (Fig.1c). Heavily halogenated CCA derivatives (e.g., 2,3,4,5,6pentafluorocinnamic acid) were tested in a subsequent study [34]. One problem with these less polar compounds is a change in morphology: needle-like crystals are formed, instead of compact crystals with dimensions in the 10-μm range generally regarded as the ideal MALDI morphology for all CCA derivatives [34]. The morphology can eventually be improved by mixing two halogen-substituted compounds of similar PA (to prevent reactivity losses), e.g., 4-bromo-αcyanocinnamic acid (BrCCA) and ClCCA (2:8, n/n) [94]. Another problem is related to the blue shift in the optical absorption profile that accompanies substitution with electron-withdrawing moieties. This can lead to a substantial reduction in absorbance at the standard laser wavelengths of

K. Dreisewerd

337 and 355 nm and, as a consequence, deteriorates the LODs [34]. To partly compensate for this effect, HCCA may be added to the derivative [94]. This approach allows one to make use of the extremely low PA of some multiply halogenated CCA derivatives, even if they are not directly excited by the laser light. This effect is probably related to electrontransfer processes from neutral halogenated matrices to excited matrix radical cations of HCCA. In addition, neutralization reactions of protonated analytes with neutral matrices on collision in the early dense MALDI plume become less probable. Although some groups have recently reported the advantageous use of the halogenated compounds (in particular ClCCA) [95], a wider exploitation has so far been hampered by a limited availability. The halogenated compounds are protected by patents [96, 97] and no agreement between the patent assignee and a potential distributor seem to have been reached for a longer time. In June 2013, Sigma–Aldrich announced the licensed distribution of selected singly and multiply halogenated CCA derivatives. CCA derivatives with other substituents have recently also been synthesized (e.g., (E)-2-cyano-3-(naphthalen-2-yl)acrylic acid (2E)3-(anthracen-9-yl)-2-cyanoprop-2-enoic acid [98]. These matrices were found to be useful for MALDI analysis of small molecules. Another electron-withdrawing moiety is NO2. Recent experiments in the author’s laboratory, in which an NO2-carrying CCA derivative (3-OH-4-NO2-CCA) was investigated, produced an unexpected result. Although this matrix does not improve LODs compared with standard matrices, it induces a sizable abstraction of one or more hydrogen atom from peptides, carbohydrates, and lipids, and even from synthetic polymers, for example as poly(ethylene glycol) (PEG) [99]. Such reactions are eventually also found for some of the classical matrices (e.g., for DHB), albeit to a much lesser extent. As a result of the hydrogen abstraction, analyte radical species might undergo new fragmentation pathways, which could, in turn, result in structural information. In the future, other synthesized compounds could yield more such surprises.

Laser wavelength The relevance of the excitation laser wavelength for the MALDI performance characteristics was already recognized in the initial phase of the MALDI development [100, 101]. Later, however, few follow-up studies were conducted to investigate this important parameter in more detail. In this work distinct wavelengths, e.g., provided by N2 (337 nm), frequency-tripled and quadrupled Nd:YAG (355/266 nm), or XeCl excimer lasers (308 nm) were compared [102–104]. In addition, Chen et al. used a wavelength-tunable Ti:sapphire

Recent methodological advances in MALDI‐MS

2265

a

b

c

Fig. 1 Physico-chemical and MALDI-MS properties of the classical HCCA matrix and of halogenated CCA derivatives. (a) Molecular electrostatic potentials of neutral HCCA (= CHCA), ClCCA, and DiFCCA matrices. Regions with low electron density are shaded in red, regions with high density in blue. The matrix structures were energetically optimized by DFT calculations using the extended B3LYP/6-311++G(3df, 3pd) data set. (b) MALDI mass spectra of a tryptic in-solution digest of BSA acquired from a ClCCA (top) and an HCCA (bottom) matrix; 50 fmol of digested analyte was spotted; details

are given in Ref. [26]. (c) Achievable sequence coverage of proteins, digested in solution by use of different proteases. All values refer to the average of three independent digestions. Cyt. C, cytochrome C; Serotransfer., serotransferrin; details are given in Ref. [94]. The MALDI mass spectra were acquired with a Voyager TOF mass spectrometer (AB Sciex) at a laser wavelength of 337 nm. Fig. 1a courtesy of Thorsten W. Jaskolla (University of Münster); Fig. 1b adapted from Ref. [26] (© 2008, National Academy of Sciences of the USA); Fig. 1c compiled from Ref. [27] (© 2009, American Chemical Society)

laser to scan the wavelength region from 375–435 nm [105]. Although two CCA derivatives (HCCA and SA) were investigated in that study, which exhibit a sizable tailing of their absorption bands toward wavelengths slightly in excess of 400 nm, neither the peak absorption of these compounds nor the emission lines of the two standard MALDI lasers of 337 and 355 nm were covered. Characterization of the more relevant wavelength range between 280 and 355 nm was recently achieved by use of a dye laser that was tuned in small wavelength increments of 5 nm [34]. The signal intensities of molecular analyte and matrix ions, and fragments thereof, were recorded as a function of laser wavelength and fluence. This work, which focused on DHB and five CCA derivatives as matrices, was amended by two follow-up studies. In the first, MALDI imaging conditions were approximated by irradiating fixed sample positions on dried-droplet sample preparations with consecutive laser pulses [33]. Ion yields per laser pulse and integrated ion yields per irradiated pixel were compared at approximately the individual absorption maxima of HCCA, ClCCA, DiFCCA, and DHB, and/or at the MALDI standard wavelengths. In the second follow-up study, the overall amount of ejected material was recorded as a function of wavelength and fluence by use of a photoacoustic detection

principle [32]. Parallel recording of both neutral and ionic species under identical irradiation conditions is extremely helpful for obtaining a better understanding of the convoluted desorption and ionization processes, because analyte ion yields in MALDI are only of the order of 10−4–10−3 [39, 106, 107]. Hence, by far the largest fraction of material is ejected in the form of neutral particles rather than as ions. The main finding of these combined investigations is that the highest ion yields are generally obtained for laser wavelengths at or close to the individual absorption maxima of the matrices in the solid state (Fig. 2 [32]). A moderate reduction in absorbance (up to approx. 50 %) can typically be compensated for by elevating the laser fluence, such that the same energy per unit sample volume is deposited into near surface volume elements and similar MS signal intensities are obtained [32, 105]. This relaxes requirements for the choice of the laser and enables use of matrices for which the peak absorption does not fully match the laser emission line. Beyond this point, however, the MALDI performance deteriorates rapidly, i.e., a further reduced optical absorption cannot usually be compensated for simply by increasing the laser power [34, 105]. A few more noteworthy fine features were revealed in the course of these wavelength studies. For example, the dependence of ion-yield on wavelength for all the CCA derivatives

2266

K. Dreisewerd

a

b

Fluence / Jm-2

300

ejected material

200

100

0

300

325

350

375 max

c Fluence / Jm-2

300

analyte ion intensity

200

100 0

0

Fluence / Jm-2

300

300

325

350

375

325

350

375

analyte ion yield

200

100

0

300

Wavelength / nm Fig. 2 Effect of the laser wavelength in MALDI. (a) Wavelength course of the absorption profiles of four MALDI matrices (DHB, HCCA, ClCCA, and DiFCCA) in the solid state, determined by diffusive reflection spectrophotometry; details are given in Ref. [34]. (b) Heat maps showing the wavelength-fluence course of the overall material ablation from a DiFCCA matrix (top; recorded using a photoacoustic detection scheme), that of the [M + H]+ signal intensities of the co-desorbed molecular peptide ions (middle; these data were acquired with an oTOF mass spectrometer under essentially the same irradiation conditions as used for recording the overall material ejection), and that of the analyte ion yield (bottom; calculated by division of the two above data sets; in rel. units). Solid white lines represent the inverse of the solid-state absorption profile of DiFCCA; further details are given in Ref. [34]. (c) Top: “ion

chromatograms” representing molecular peptide ion intensities [M + H]+, recorded as a function of number of successive laser pulses applied to single positions of a dried-droplet DiFCCA sample preparation and for three laser wavelengths (corresponding to the optimum wavelength of 305 nm and the two standard MALDI wavelengths of 337 and 355 nm). Each data point represents the ion signal generated by approximately one laser pulse. Solid lines represent the same data sets smoothed with an adjacent average function. Bottom: “Integrated ion yields” after application of 900 laser shots to one sample position, as a function of wavelength and laser fluence; further details are given in Ref. [33]. Fig. 2a and c from Ref. [33] (© 2012, John Wiley and Sons), Fig. 2c adapted from Ref. [32] (© 2013, Springer)

investigated was clearly “asymmetric” [32, 34]. Despite the equal optical absorptivity at the “red” (short-wavelength) and “blue” (long-wavelength) sides of the peak absorption,

substantially reduced abundances of the molecular peptide ions were consistently generated when the shorter, more energetic, laser wavelengths were used. This finding was mainly

Recent methodological advances in MALDI‐MS

attributed to increased (thermal) fragmentation of both analyte and matrix ions on the “blue” shoulder [34], which is, in turn, indicative of an increased energy content (temperature) within the system. In a simple picture, all energy exceeding the lowest possible electronic S0–S1 transition may be rapidly converted into heat by internal intramolecular conversion and/or excitation of the matrix lattice [108]. Qualitatively, these findings are corroborated by a recent study of Ahn et al., who recorded MALDI PSD ion yields at six discrete wavelengths between 307 and 357 (10 nm step size) and observed an increase for the short wavelengths [109]. For the CCA derivatives, this effect is intensified by their pronounced thermal instability and photo-reactivity, which can result in both decarboxylation and dimerization [110, 111], leading, respectively, to hypsochrome (decarboxylated compounds) or bathochrome (CCA dimers) shifted optical absorption bands. As a consequence, the absorption at the regular laser wavelength can become so low that only low abundances (or even no MALDI ions at all) are generated [33, 34]. As was pointed out by Knochenmuss, such cumulative effects should be included in comprehensive analysis of wavelength–fluence effects in MALDI [112]. The largest fraction of the laser pulse energy is deposited at sample depths that do not contribute to ejection of the material. Under standard MALDI conditions a few to a few tens of monolayers are ejected per laser pulse, only [19, 33] whereas the 1/e-laser penetration depth is between ~30 and 200 nm, depending on matrix and laser wavelength. For photolabile and/or thermolabile compounds, for example the CCA derivatives, subsequent irradiation leads to gradually accumulating “radiation damage”. As a consequence of the complex interplay between material and laser parameters, ion signal intensities vary with the number of laser pulses applied onto a distinct sample position (Fig. 2c [33]). Moreover, the exact course of the decline in ion signals with increasing number of pulses depends on the wavelength of the laser (Fig. 2c). An elegant way of partly recovering the ion signals is by increasing the laser pulse energy for one or two shots to ablate the modified material and produce a fresh unmodified surface [113]. However, despite the wide use of the HCCA matrix this possibility does not seem to be implemented in any standard data acquisition routine of current mass spectrometers. As a consequence of the reactivity of the CCA derivatives, somewhat “red-shifted” wavelengths (by 10–15 nm) relative to the absorption maxima are advantageously used for a soft, fragmentation less analysis, whereas excitation of the samples with higher photon energies can increase the yield of informative fragments [32, 109]. In contrast to the CCA derivatives, some of which are frequently described as “hot” MALDI matrices in the sense that, e.g., HCCA induces high PSD yields [114, 115], neither a sizable wavelength asymmetry nor any radiation damage is observed for the “cooler” DHB matrix [32–34, 112]).

2267

Laser spot size and focal intensity profile The easiest way of focusing the initially parallel laser beam on to the MALDI sample is with a single spherical lens (e.g., a plano-convex lens). However, the use of a set of at least three lenses is generally recommended for intermediate beam expansion, to achieve more flexibility concerning the generation of variable spot sizes, and for mode filtering (e.g., with an aperture that is placed in the plane of an intermediate image) [89, 116]. The dimensions of the focal spot are largely determined by the focal length of the last lens, the diameter of the input beam, its divergence, and the angle of incidence. Due to the presence of the ion-extraction electrodes, lenses with focal lengths smaller than ~50 mm (such lenses are, e.g., utilized in some Bruker instruments) can generally not be used. If the lens is not positioned inside the vacuum of the ion source, this minimum distance typically increases to 120 mm and above (e.g., in the MALDI Synapt instrument from Waters). Therefore, the maximum usable numerical aperture (NA) cannot be much larger than ~0.1–0.15. As a consequence, the minimum focal spot size is, per se, limited to diameters of no less than a few micrometers. Taking into account the imperfect beams of the “low-cost” lasers typically used, lens aberration and diffraction-derived effects, and the angle of incidence of the beams (typically ≥30°), minimum spot sizes between 10 and 20 μm (Bruker, Flex series) and ≥30–50 mm (most other instruments) in diameter are typically achieved. A comprehensive list of commercially manufactured MALDIMS(I) mass spectrometers, with benchmark figures for the best achievable lateral resolution, can be found at http://www.maldi-msi.org. Kaufmann and Spengler designed an objective that enables irradiation of MALDI samples at a high NA of ~0.6 at normal incidence [117–119]. This aspheric objective, consisting of five large-diameter stacked lenses, was placed in close proximity (~16 mm in the original set-up [118]) to the target. The objective images a pre-focused spot (of ~10–30 μm diameter [118, 119]) on to the sample plate. Taking diffraction patterns into account, the minimum effective spot diameter is close to 1 μm [119]. Ions are extracted using electrodes and electrostatic lenses and pass through a metal capillary that is mounted in a central bore through the objective. Using this arrangement the distributions of a variety of biomolecules (phospholipids, neuropeptides, proteins) and administered drugs were imaged in brain and kidney tissue slices with a lateral resolution of 5– 10 μm [37, 120–122]. In contrast to the vast majority of MALDI-MS instruments, an AP-MALDI ion source is used in the most recent version of this scanning MALDI (SMALDI) set-up; this ion source is also commercialized by the Giessen group. A beam-shaping approach is adopted in TOF instruments of the Bruker Flex series. In addition to implementing a telescope that enables adjustment of focal

2268

spot sizes between 10 to 20 and ~100 μm in diameter, the so-called “smartbeam” technology uses a diffractive optical element. A micro lens array or a Dammann grid in combination with a spherical lens are two suitable modulators [37, 123, 124]; the type of device actually implemented is not disclosed by the company. Imaging of the profile generated by the modulator on to the sample results in the formation of distinct beam patterns that are containing a series of hot spots (near-Gaussian-shaped peaks distributed over the totally illuminated area) (Fig. 3b). To ensure that the full focal area is, on average, irradiated “uniformly” after several hundred laser shots have been applied, the diffractive element is continuously rotated (or alternatively moved in perpendicular direction to the beam) such that statistically the hot spots hit the full irradiated area. In most other current MALDI instruments the laser beams are delivered without more extensive beam shaping. In these instruments the focal intensity profile is, however, still affected by the spatial coherence of the impinging laser beam. For example, focusing an ideal TEM00 beam (e.g., from a highquality Nd:YAG laser) with an aspheric plano-convex lens results in a Gaussian intensity profile in the focal plane (far field). Moderate defocussing, such that larger spots are obtained, results in “broadened” near-Gaussian profiles (an example is shown in Fig. 3b). A stronger modulated focal profile with hot spot areas is obtained when a low-coherent N2 laser is used [37]. For some experimental conditions, at least, e.g., for specific matrix preparations, the focal intensity pattern of the laser beam (e.g., Gaussian vs. flat-top vs. modulated) can, in fact, have a notable effect on the ion yield. By using thin-layer HCCA sample preparations Holle et al. demonstrated that higher ion yields were obtained if these samples were irradiated with a smartbeam profile rather than the Gaussian profile of a non-modulated Nd:YAG laser [37]. Similarly improved results were obtained if the samples were irradiated with the modulated profile of an N2 laser. For standard dried-droplet HCCA preparations—these produce larger crystals, with diameters in the low tens of microns range, than the thin-layer preparations—a similar difference was, however, not notable; a trend toward rather better results was observed after irradiation with the Gaussian beam. For other matrices and/or sample preparations, for example dried-droplet preparations of DHB and 3-hydroxypicolinic acid (3-HPA), the standard matrix for the analysis of oligonucleotides, again a strong improvement was found if the smartbeam (or the N2 laser) was used. Given these different findings, both types of irradiation profile (Gaussian and modulated) should, ideally, be available as options. Technically, this may be achieved simply by moving the modulator in-and-out of the beam path followed by some adjustment of lens positions. Another convenient way of producing a shaped and/or modulated laser beam is by mode mixing inside a step index

K. Dreisewerd

multimode fiber (Fig. 3a). The illuminated end surface of the fiber core can be conveniently imaged on to the sample with a telescope [29]. Care should, however, be taken to use highquality aspheric lenses if the steep flanks of the initial profile across the exit face of the fiber are to be reproduced in the sample plane. The extent of mode mixing depends on the spatial and temporal coherence lengths of the laser, the length of the fiber, and its core diameter. For example, use of a short fiber will generally produce Speckle patterns across the end surface (which are caused by interferences between temporally coherent rays; a Speckle pattern is, for instance, visible in the middle image of Fig. 3a). This effect can be exploited to produce a focal intensity profile similar to that obtained with the smartbeam approach (Fig. 3b). If a sufficiently long fiber is used for extensive mode mixing, a close to flat-top intensity profile is produced at the end surface of the fiber. The extent of mode mixing can be significantly improved by twisting the fiber such that shorter fibers can do the job [31]. Particularly uniform flat-top profiles can also be produced with special fibers, e.g. jacketed air-clad fibers. These light guides even enable the generation of square spots (Fig. 3b, right image [125]), a feature that is highly relevant for MALDI-MSI applications. Fibers with very small, e.g. 10 μm, core diameters are a special case; these light guides can transmit few modes only and are, therefore, closer in performance to mono-mode fibers (Fig. 3 a, left image). In cases where residual mode patterns are obtained, continuously “shaking” the fiber (e.g., with a vibrating device [31, 42]) can still produce an “averaged uniform” profile, again similar to the smartbeam approach. Thus, using an optical fiber is a means of obtaining well defined MALDI irradiation conditions. This has a few advantages in application-driven as well as in fundamental studies: First, the intensity profile is uniform across the laser spot diameter, provided that sufficient mode mixing occurs (or, otherwise, the profile may contain a desired modulation). For the flat-top case, this means that the optimum fluence can be applied homogeneously over the whole of the irradiated area. In contrast, when a Gaussian laser beam is used, a significant portion of the laser energy is contained in the “wings” of the beam—which in a MALDI setting leads to ablation of material in these areas, but not to significant ion formation, because of the well-known threshold behavior and the strong dependence of the ion signals on laser fluence [29]. On the other hand, in the beam center the local energy density may be substantially higher than the threshold for ion generation, such that increased fragmentation of labile biomolecules may occur [34]. Qiao et al. showed that a flat-top beam profile of a Nd:YAG laser, generated by mode mixing in a quartz fiber 30 m long, produced better results in terms of peptide ion yields (determined from HCCA thin-layer sample preparations) than the near-Gaussian intensity profile of the non-

Recent methodological advances in MALDI‐MS

2269

a

b

c

„smartbeam“

Fig. 3 Beam shaping and effect of laser spot size. Two methods for shaping a MALDI laser beam: (a) by using an optical fiber (this concept is, e.g., used in the oMALDI2 ion sources of QStar instruments) [31] and (b) by implementation of a diffractive optical element (modulator) (as used in “smartbeam” Bruker instruments [37]). L1–L5, lenses; A, attenuator; M, modulator; O, round aperture; T, sample target; components in green are needed for measurement of laser profiles only; planar mirrors for axial beam adjustment are not shown; further details are given in the text and in Refs. [31, 37]. Examples for focal intensity profiles: (a), bottom: beam profiles produced by mode mixing of an N2 laser beam inside two 2 m-long optical fibers with core diameters of 10 (left) and 100 μm (middle) [31]. A closer to uniform flat-top profile would be obtained by using a longer and/or stronger twisted fiber. An example of an almost ideal flat-top profile is shown on the right; the profile was produced by use of a square core jacketed air-clad fiber of 400 μm core

size and 3 m length; the fundamental beam of a single-mode Q-switched Nd:YAG laser (λ= 1.064 μm) was propagated [125]. (b), bottom, “smartbeam” profile (left) containing a series of high-intensity peaks [37]. The image on the right shows the profile of a near-Gaussian beam, produced with the same Nd:YAG laser by omitting the modulator and under slight defocussing conditions. (c) Dependence of MALDI ion yields on laser spot size (diameter) and fluence. Plotted are the integrated signal intensities of protonated substance P molecules obtained by irradiating fixed positions on a thin-layer HCCA preparation with a flat-top profile, in accordance with a. Fibers with five different core diameters were used to produce the indicated spot sizes [31]; further details are given in the text and in Ref. [31]. Figures (a) and (c) were adopted from Refs. [31] (© 2008, John Wiley and Sons) and [125] (© 2006, the Optical Society of America), and (b) from Ref. [37] (© 2006, John Wiley and Sons)

2270

modulated Nd:YAG laser [31]. The second advantage is that the area from which material ablation (or ionization) occurs does not change with applied laser power. Both features (uniform profile and defined spot boundaries) can be advantageously exploited to improve the lateral resolution of MALDI-MSI by “oversampling”. In the oversampling mode, all analyte-containing matrix material is either fully ablated at a given irradiation site (or at least irradiated long enough until the ion signals decline to an acceptable level) before the sample plate is moved a small increment. As a result, the lateral resolution achieved may be less than the beam diameter [126, 127]. A disadvantage of using optical fibers for laser beam delivery is related to their relatively high NA (>0.1). The correspondingly large aperture angle of the exiting beam renders it difficult to reduce the spot size by use of telescope optics to values much below the fiber core diameter. This problem is aggravated by the limitations set by the damage threshold of the fiber. For this reason, fibers with cores diameters below ~50–100 μm are increasingly difficult to handle. Therefore, use of small-core fibers for high-resolution MALDI-MSI, for which spot sizes in the 10 μm range are desired, is limited. The use of (fiber-generated) flat-top laser beams has proved exceedingly useful in fundamental studies of the MALDI mechanisms. For instance, the controlled laser energy supply assisted greatly in characterizing the effects of laser fluence and laser spot size on overall material ablation, MALDI ion yields, and fragmentation patterns [29, 31–35, 128, 129]. With regard to spot size and fluence, a few effects have been revealed in these studies that have consequences for both general MALDI and MALDI-MSI applications. For example, at fluences not far from the ion detection threshold the signal intensities obtained follow a cubic dependence on the irradiated area, A, i.e., over-proportionally more ions are generated from larger spots. This effect is demonstrated in Fig. 3c (top) which displays the yield of protonated substance P peptide molecules obtained from fixed positions on a thin-layer HCCA preparation as a function of spot diameter and fluence. To visualize the cubic dependence the data were divided by A3. Qiao et al. showed that this behavior changes, however, when, instead of laser energies close to the ion detection threshold, “saturation” fluences a factor of approximately 3 above the spot-size-dependent ion detection threshold are evaluated [31]. Figure 3c (bottom) illustrates this finding, by dividing the same data set by A, and demonstrates that for fluences of this magnitude the dependence of the ion yield on irradiated area is closer to linear. In tendency, even more ions per unit area are generated from small spot sizes. This could be related to an improved ion yield at high fluences, because the dependence of overall material ejection on fluence is weaker than that for ion generation [29–33].

K. Dreisewerd

Similar results were obtained in a study by Günther et al. in which near-Gaussian beams with sub-10 μm-diameters were used [119]. Exact comparison of the different data is difficult, however, because different laser spot diameters were used on different (high-vacuum, fine-vacuum, AP-MALDI) instruments and moreover, different laser profiles were used. Further experiments, which should also include additional matrices and analytes, must be conducted to evaluate the extent to which the beneficial increase of ion yield per unit area with decreasing spot size can be exploited on different instruments. For instance, it could be speculated that instruments that are equipped with fine vacuum or AP ion sources may be most suitable for high-fluence MALDI-MS with μm-spots. In addition to providing efficient collisional cooling of the desorbed ions [129–131], in these instruments mass resolution and accuracy are not affected by the laser fluence. Because of decoupling of ion generation and mass analysis the laser fluence can, hence, be increased without degradation of these analytically important properties. In contrast, the laser fluence is a rather critical parameter in MALDI axial-TOF mass spectrometry [37, 132]. Even with laser fluences only moderately elevated, the calibration can be shifted because of the ejection of a dense MALDI plume and variations in the initial ion velocity distributions. For similar reasons, optimum mass resolving power is generally obtained only at fluences that are within a factor of approximately two above the ion-detection threshold. Another way of producing MALDI-MS images with sub10 μm resolution is to use a stigmatic ion microscope [133–135]. In this “microscope mode” the sample is irradiated with a uniform beam of larger diameter (in the 100 μm-range) and the initial lateral distribution of the ejected ions is imaged with magnifications up to 100× on to a position-sensitive ion detector. The principle is similar to that used in secondary electron microscopy (SEM) but includes time-of-flight measurement of the ions. The technique requires ion generation and extraction under high-vacuum conditions but, because of the large spot diameters, only moderate laser fluences are necessary. The Heeren group demonstrated that a lateral resolution close to 5 μm can be achieved in the analysis of matrix-coated tissue slices with a TRIple focusing time-offlight (TRIFT II) mass spectrometer [133, 136]. Despite ejection of a substantial amount of material during the MALDI process, the information about the origin of the detected ions is largely preserved. However, a major complication of the method is the need to use a detector that simultaneously records the spatial and time-of-flight information while also having a reasonable dynamic range. Building on innovations from nuclear physics, promising progress in this regard was recently achieved. By using a micro channel plate (MCP) detector with large cross-section for ion-electron conversion and amplification and a 512 × 512 timepix

Recent methodological advances in MALDI‐MS

system for position-sensitive data handling, MS images could be recorded from mouse testis tissue slices in an m/z window of 500–1500 [136]. Comparing microscope mode with the typical scanning “microprobe” MALDI-MSI mode, a substantially faster data acquisition is theoretically possible, because all of the irradiated large area is sampled instantaneously instead of pixelwise. Neighboring areas can be stitched together in a mosaic approach. Currently, however, the microscope instruments are, at best, in the development phase. Major limitations are still set by the single-stop detection principle and the low time resolution of the timepix detector (of ~16 ns). Future improvements toward a fast multi-stop detection scheme and a faster time resolution of the timepix system (≤ 1 ns) will be required to make these instruments more compatible with standard microprobe mass spectrometers. As an exciting “side effect”, a notably improved detection sensitivity for particularly large MALDI-generated ions, for example IgG/IgA, was recently observed by Jungmann et al. in a study in which the MCP/Timepix combination was used in lieu of the standard MCP ion detector [137]. At the same time, the spatial distribution of the ion beam, in that work generated with an Ultraflex III instrument (Bruker) operated in the linear TOF mode, could be recorded. This feature is clearly useful for the design of time-of-flight instruments. Another way of producing laser spots with ultra-small diameters is by use of near-field optics. Coupling of scanning near-field optical microscopy (SNOM) with MS was developed by the Zenobi group [138, 139]. In their instruments, the sharp SNOM tip serves both for visualization of the sample surface with ultrahigh lateral resolution and for material ablation upon application of an intense laser pulse. Material ejection occurs if the energy absorbed from the evanescent wave exceeds the material ejection threshold. As proof of concept, the analysis of anthracene (from neat samples) was demonstrated with a lateral resolution of the MS analysis in the low-μm range [139]. However, DHB matrix samples did not seem to produce mass spectra, although ablation craters were clearly visible. In addition to an ion yield which might be too low, the difficult “sidewards” ion-extraction geometry of this arrangement could result in losses under the AP conditions which are too large. So far this concept has, therefore, not found a practical use. To improve the ion extraction, co-axial irradiation of sample and ion extraction through a 100 nmsized aperture of a hollow atomic force microscopy (AFM) tip was also proposed [140]. This theoretical concept has, however, not yet been investigated experimentally.

Liquid matrices (and quantitative MALDI-MS) MALDI with solid state matrices is often associated with the occurrence of so-called sweet-spot effects, i.e. a position-

2271

dependent variation of the ion yield caused by an inhomogeneous analyte–matrix co-crystallization [141]. The extent of such effects varies substantially with the matrix, the type of sample preparation (e.g., dried-droplet vs. thin-layer preparation) and, possibly, with the type of analyte also (e.g., with regard to the polarity [142]). Although, on a macroscopic scale, quite uniform sample preparations are readily achieved with CCA-based matrices and by use of both the dried-droplet and thin-layer preparation protocols, the aforementioned dynamic alteration of the chemical structure by preceding laser pulses adds a notable complication for quantitative MALDIMS with these compounds. On a microscopic scale, fine structures become visible, moreover, for example with regard to non-uniform incorporation of different analytes with sample depth z [33]. For many other matrices (e.g., DHB) both macroscopic and microscopic variations are even much more pronounced than for CCA-derived compounds [33, 143, 144]. In MALDI-MS imaging applications variable matrix crystallization on different types of tissue and/or a different extent of analyte extraction from the tissue may also have to be taken into account [65]. Generally, in addition to the physicochemical properties of matrix and analyte, the solvent system, the properties of the sample plate surface, and the way the samples are applied (e.g., by thin-layer, dried droplet, or spray preparation) will all have a notable effect on the type of sample morphology generated. As already discussed, different procedures were developed over the years to overcome these effects and to produce samples with uniform morphology and homogeneous analyte–matrix co-crystallization[66, 145–147]. Unfortunately, a quantitative MALDI-MS analysis is complicated by further factors. Among these are the different ionization efficiencies for individual compounds (e.g., peptides or metabolites with different basicity) [62, 148], and the possible dependence of the ionization efficiency on the analyte-to-matrix ratio [149] as well as on the presence of charge-competing compounds (“ion suppression effects” [150]). Therefore, fractionation of complex samples, e.g., by liquid chromatography (LC) will usually improve quantification, albeit at the cost of additional work. For less complex samples, simple purification procedures, e.g., by using C18 columns (for example ZipTips C18 pipette tips) may help to improve the analysis. More sophisticated immunochemistry principles [151] can be used to isolate (concentrate) specific compounds of interest from a complex background. Another complication is the exponential effect of the laser power on the ion signals (Fig. 3c) [28, 29, 31]. At least partially driven by the increasing demand for pulse energy stability for reliable data acquisition in MALDI-MSI, lasers with improved stability have become available in recent years. Nevertheless, no commercial instruments have yet been equipped with a means of monitoring laser pulse energy. Tongue-in-cheek, one could compare the lack of such a device

2272

with driving a car without a tachometer. Taking all these factors into consideration, a semi-quantitative MALDI-MS analysis, at least, is often feasible, but careful calibration strategies are usually required. One possible approach to obtain a semi-quantitative result, even for “sweet spot” matrices or non-uniform matrixcoatings, is the use of an internal standard. In fact, this simple approach can substantially improve quantification, even for sweet-spot matrices such as DHB [66, 67] or for MALDIMSI [65]. At least some of the effects described above that are hampering a more straightforward quantitative analysis by MALDIMS are much less of a concern if liquid matrices are used. For example, if a liquid HCCA–glycerol system was used, almost constant signals of the co-desorbed peptide ions were achieved over ten thousand laser shots applied onto a ~1-μL sample drop (Fig. 4a) [70]. Quantification was demonstrated with this (and with a liquid ClCCA matrix system) over more than two orders of magnitude (Fig. 4b). Notably, laser-induced modification of the matrix-properties, as is prominent for crystalline CCAderived samples (cf. Fig. 2 upper plot), was not noticeable. Peptide quantification over two orders of magnitude by use of RTILs (e.g., triethylammonium α-cyano-4-hydroxycinnamate) as matrix has also been reported [152]. Although somewhat conflicting figures-of-merit have been claimed [72, 153], taking all the experimental evidence into consideration it must be assumed that liquid matrices are less sensitive for most applications. In fact, when high concentrations of matrix and base are mixed with glycerol such that a fraction precipitates on the MALDI sample plate, significantly higher peptide signal intensities were consistently obtained from the crystalline part than from the liquid phase [154]. A few studies have shown that different results are eventually obtained in the analysis of low-molecular-weight (LMW) compounds, such as amino acids. In particular, RTIL matrices with HCCA or DHB as active matrix compound were found to produce high-quality quantitative results for several LMW analytes or even complex mixtures thereof. A prerequisite for this success was preparation of the samples in rather low matrix-to-analyte ratios [72, 148, 155]. Generation of abundant Na+/K+-adduct ions from RTIL matrices has also been reported to be beneficial for the analysis of carbohydrates [156]. In an attempt to improve the LODs for peptides, several glycerol-based matrices containing halogenated CCA derivatives (e.g., DiFCCA) were recently tested [154]. In contrast to solid-state MALDI-MS with these compounds, this has not yet led to significant improvement of LODs. This negative result is presumably caused by the presence of the base, another species competing for charge in the complex ionization pathways [154]. An interesting “side observation” made in the study by Cramer et al. [154] was that crystalline MALDI preparations also can eventually be improved, if the samples are overlaid

K. Dreisewerd

with a thin film of glycerol. Interestingly, this approach not only resulted in lower LODs for the analysis of peptides but also led to significantly reduced levels of matrix clusterderived background ions in the mass spectra. In this study, Cramer et al. also investigated whether the reduced energy deposition into halogenated matrices (because of their blueshifted absorption profiles) could be compensated for by adding a highly absorbing base, for example 3-aminoqinoline (3-AQ). The negative outcome of this experiment revealed that energy absorption by this less reactive partner alone is not sufficient to induce the formation of reactive [M+H]+ species of the active, weakly absorbing CCA component.

Multiply charged ions Glycerol-based liquid matrices can also be used for generation of highly charged MALDI ions. This phenomenon was first observed when pure glycerol was used in the AP-IR-MALDIMS analysis of peptides and proteins [78]. In this study by König et al. apo-myoglobin was, for instance, recorded with charge states between 9 and 13. Use of a heated capillary interface between the AP region and the employed Paul ion trap mass analyzer was found to be a prerequisite for generation of these ionic species. It is well known that many smallanalyte-containing droplets are produced in IR-MALDI with a glycerol matrix [157, 158]. Their explosive disintegration in the heated air-vacuum interface likely forms the basis for generation of the highly charged ions [159]. This phase explosion is presumably initiated by a combination of effects, including the rapid increase in temperature with concomitant drop in pressure [44] and, possibly, collisions with other particles or the inner surface of the capillary. Sampson et al. demonstrated that highly charged peptide ions can also be generated from aqueous samples containing a dissolved classical MALDI matrix, for example DHB, if a high electrical potential is applied between the sample plate and the heated inlet capillary of the mass spectrometer [160]. A pulsed UV laser beam (applied with a relatively high fluence) served in this case for ablation of material. However, it remained somewhat unclear whether the highly charged ions were essentially produced in an ESI-like process (as suggested by the authors, who propose that a high charge density on the sample surface contributes substantially to the ion formation) or whether other effects (for example, corona discharge effects under the AP conditions) may not also add to the ion-formation pathways. Two recent studies by Pirkl et al. and Witt et al. demonstrated that multiply charged ions of peptides [59] and heavily sulfated glycosaminoglycans (GAGs) [161] can also be generated from a water ice matrix when this is excited with a Qswitched Er:YAG laser at 2.94 μm (pulse duration, ~150 ns). Notably, in these two studies the multiply charged ions were

Recent methodological advances in MALDI‐MS

2273

a HCCA

b ClCCA

Fig. 4 Liquid glycerol-based MALDI matrices for improved quantification. (a) Intensity of the molecular ion [M + H]+ signal of the peptide angiotensin I as a function of the number of laser pulses applied to the liquid glycerol-based matrix. 100 fmol of peptide was prepared on the target in a liquid matrix containing HCCA–3-aminoquinoline–glycerol– 50 % MeOH in 10 mmol L−1 ammonium phosphate (1:3:5:5 w/w/v/v), which was further diluted by a factor of 30 with 50 % MeOH in 10 mmol L−1 ammonium phosphate. For analysis, the matrix solution was mixed 1:1 with analyte solution and 1 μL was deposited on the MALDI target. Spectra were acquired as the sum of 100 single-shot

spectra at 50 Hz from a single desorption position with 1 s between each accumulation, for a total of 10,000 laser pulses. (b) Ion signal intensity as a function of sample amount (5–1,000 fmol spotted) for a ClCCAcontaining liquid matrix. Four spots were measured for each data point (500 laser pulses each). Data are shown for angiotensin I (diamonds), substance P (squares) and Glu-fibrinopeptide B (circles). All R2 values are better than 0.99. Data were recorded with an Ultraflex axial-TOF mass spectrometer (Bruker); further details are given in Ref. [70] (© 2010, American Chemical Society)

produced under fine vacuum conditions in an oMALDI2 ion source. This source was coupled to a QStar instrument and, hence, lacks a heated vacuum transfer stage. Work by McEwen, Trimpin, and co-workers extended the concept of rapid evaporation ionization to solid-state matrices. In their initial approach, which they denoted “laser spray ionization” (LSI) [79], a crystalline 2,5-dihydroxyacetophenone matrix, prepared on a transparent glass slide, was irradiated from the rear using a rear side illumination geometry [162]. In contrast to conventional MALDI mass spectrometry, for which focal laser fluences of approximately 50 to a few 100 J m−2 are most typical, very high photon fluxes corresponding to fluences of a few tens of kJ m−2 are applied for LSI-MS [79, 159]. Presumably, this leads to the explosive ablation of larger sample volumes. A fraction of the material will be ejected either directly as liquid droplets or in form of solid particles that melt rapidly when they are sucked into the heated capillary. In a series of experiments the two groups demonstrated that the concept of

generating highly charged ions in a heated vacuum interface can be extended even further. For example, it is possible to generate highly charged ions from the effluent of a capillary (e.g., from a micro-HPLC system) if this is placed inside the heated vacuum interface [163], or even by applying crystalline analyte–matrix material directly inside the capillary interface [164]. Although these phenomena are rather interesting from a fundamental point of view and also the coupling to further types of mass spectrometers (e.g., a high-resolution Fourier-transform ion cyclotron resonance (FTICR) instrument [165]) was recently realized, all variants of these ionization techniques seem to suffer from a restricted reproducibility and generally low sensitivity. In the recent study by Cramer et al., liquid glycerol-based UV-MALDI matrices that contained either HCCA or DHB as active matrix component were used for AP-UV-MALDI mass spectrometry [80]. Also in this work, a heated transfer capillary (T ~200 °C), forming the vacuum interface was found to be a requirement for generation of highly charged peptide

2274

K. Dreisewerd

operated under standard UV-MALDI conditions and solidstate matrices were used. Finally, Fig. 5c shows the second major advantage of generating MALDI ions with high charges states, i.e. the improved fragment ion yield and sequence coverage that is obtained on subjecting the multiply charged precursors to low-energy CID.

ions (Fig. 5a,b). However, in contrast to the LSI-work, fluences in the range of 200–2000 J m−2, only, were applied, which are typical of those used for UV-MALDI-MS with liquid matrices. As a consequence, only a small amount of material was ejected per laser shot and it became possible to generate an almost constant analyte ion current (similar to that shown in Fig. 4a) over several thousand to tens of thousands of laser pulses applied onto a 1-μL drop of sample [80]. LODs in the 10 fmol range were achieved for small peptides. These values are within one order of magnitude of those obtained if the second-generation MALDI QStar pulsar i instrument (AB Sciex) employed was

a

Perspectives Despite 25 years of successful fundamental and applicationdriven research, novel concepts for improving the MALDI

c 100

Precursor ion: 2+ [M+2H]

2+

[M+2H] 660.41

Intensity

80

T~225 °C

y8 y7

y1

MKRPPGFSPFR

60

40

y3

b3

y1 175.12

b3 416.26 y3 419.25

0 100 200

400

a7 a8,b8 a10

y7 807.48 a 8 a7 873.54 786.53

2+

y8 452.78

20

y8 904.53

600

a10 1117.60

800

1000

1200

m/z y10 y9 y8 y7 y6 y5 y4 y3 y2 y1

b [M+3H]

120

450

3+

[M+Na]

2000

+

100

Precursor ion: 3+ [M+3H]

2+

y10 594.87 3+

Intensity

[M+H]

[M+2H]

+

+ +

[M-H+2Na+DHB-H2O] + [M+Na+DHB-H2O]

2+

1000 0 1200

1400

1500

60 y1 175.12

600

800

1000

1200

1400

1600

m/z

Fig. 5 Highly charged MALDI ions produced in an AP-UV-MALDI ion source. (a) Layout of the AP-UV-MALDI ion source as used in Ref. [80] for generation of highly charged peptide ions. Highest yields of multiply charged ions were obtained with the temperature of the vacuum interface capillary set between 200 and 250 °C. (b) Mass spectrum of MKbradykinin (sequence: MKRPPGFSPFR) desorbed and ionized from a DHB-based liquid matrix containing approximately 20 % glycerol before

2+

y9 530.82

0 100

393.74

b3

2+

a8

b8

437.26

451.26

2+

2+ a7 404.22

416.27

0

2+

380 390 400 410 420 430 440 450 460

y4

b6 y6 y8 y5 667.42 710.40 904.55 653.39 y7 807.46

y2 322.19

200

b6 a7,b7 a8,b8 2+

y7

506.31

253.65

20

b3

30

y4

2+

y3

210.13

1600

500

0 400

419.26

40 1300

y3

80 Intensity

[M-H+2Na]

1500

[M+3H] 440.93

MKRPPGFSPFR

300

400

500

600

700

800

900

m/z

evaporation of volatile solvent components. (c) Low-energy CID MS– MS spectra of the doubly (top) and triply (bottom) protonated MKbradykinin ions. The CID collision potentials were set to 35 V and 20 V, respectively. Mass spectra were acquired with a QStar pulsar i o-TOF instrument (AB Sciex); further details are given in Ref. [80]. Fig. 5a courtesy of Alexander Pirkl (University of Münster), Figs. 5b, c adapted from Ref. [80] (© 2013, John Wiley and Sons)

Recent methodological advances in MALDI‐MS

technology and for extending its applications continue to be developed. The synthesis of MALDI matrices selected on the basis of quantum-mechanical calculations and further rational design considerations is one such innovation. Because the matrices comprise the central elements in the complex MALDI process, enhancement of their physicochemical properties has a direct effect on achievable LODs, can improve specific features of the analysis, and even enable the analysis of new types of analyte. It seems fair to expect that the synthesis of further new matrix derivatives (e.g., with a focus on the negative-ion mode) will in the future lead to more exciting findings, if not large surprises. Judging from tabulated optical absorption profiles (e.g., http://maldimatrixinfo.wikispaces.com; accessed on December 3, 2013; NIST webbook, http://webbook.nist.gov/) it is obvious that under standard UV-MALDI settings even many classical matrices are not excited at “their optimum” wavelength. For many compounds there is even a sizable gap between the absorption maximum and the classical MALDI laser wavelengths of 337 and 355 nm (e.g., THAP and ATT for which absorption maxima in the solid state are at approximately 293 and 262 nm, respectively; the spectrophotometric measurements were made in the author’s laboratory). This suggests that the MALDI performance characteristics could probably be improved for several of these matrices by adjustment of the excitation wavelength. Wavelength-tunable optical parametric oscillator (OPO) or dye lasers can in principle be used for this purpose. However, in addition to their higher price and increased technical complexity, current disadvantages of these lasers are a generally lower pulse-to-pulse energy stability than for solidstate Nd:YAG and N2 lasers and a typically lower beam quality; both factors would affect their use for MALDI-MS imaging. Compared to industrial and some medical applications, in which precise laser beam shaping is paramount for exact material processing, with the exceptions of the smartbeam approach by Bruker and the use of fiber optics by AB Sciex in their previous QStar instruments only few similar efforts seem to have been made in the MALDI field. As a consequence of the immense interest in MALDI-MS imaging, a stronger borrowing of industrial concepts, e.g. from the areas of laser-based material processing, to obtain optimum laser spots and profiles might be expected for the future. This could not at least help to improve the lateral resolution in MALDI-MSI to a truly cellular level. To push this boundary even further, the combination of MALDI and secondary ion mass spectrometry (SIMS) is, moreover, likely to become a future standard in applications in which both a cellular to sub-cellular resolution (provided by SIMS, although at the cost of extensive fragmentation of biomolecules and the generation of exceedingly complex mass spectra [166]) and information on the overall composition of intact biomolecules in the tissue (provided by MALDI) must be retrieved. Selected examples of the combined use of SIMS and MALDI-MSI are available elsewhere [167–169].

2275

There has been a revival of interest in the use of liquid MALDI matrices. Although the improvement of quantitative MALDI is clearly the strongest driving force, liquid matrices have other potential advantages. In particular, the flexibility in composing liquid matrix systems is an interesting feature. For example, the pH can be adjusted more readily than for solidstate matrix preparations and enzymes or other reactive compounds may be added. Chemical or enzyme-driven reactions may eventually even be monitored in real time by use of APMALDI-MS. So far, only a fraction of these options seems to have been investigated in any detail. The generation of abundant multiply charged MALDI ions is another exciting application of liquid MALDI matrices. It should be straightforward to couple AP-UV-MALDI ion sources similar in design to that used by Cramer et al. [80] to essentially all ESI-mass spectrometers. For some of these instruments, for example FTICR and Orbitrap instruments, mass resolution and accuracy are extremely high, and sophisticated means of achieving ETD and/or ECD are also available. For the first time, multiply charged ions of relatively low molecular weight (e.g., peptides and glycoconjugates) produced under controlled UV-MALDI-like irradiation conditions could be analyzed using these dissociation techniques. At the same time, advantageous features which are making out the “beauty” of MALDI, such as an “on-demand” ion generation, high throughput capacity, and greater tolerance of buffers and contaminants (compared with ESI) might be maintained. In an ideal world, the advantageous features of both contemporary techniques for biomolecular MS (MALDI and ESI) would thus be combined. Acknowledgments The author thanks the members of his workgroup, especially Thorsten W. Jaskolla (TWJ) and Jens Soltwisch, for numerous helpful discussions. Financial support by the German Science Foundation (grants DR416/8-1, DR416/8-2, DR416/9-1, and DR16/10-1 to KD, and JA2127/1-1 to TWJ) and the Interdisciplinary Center for Clinical Research (IZKF) of the Münster University Medical School (grant Z03) is gratefully acknowledged.

References 1. Karas M, Bahr U (1986) Trends Anal Chem 5:90–93 2. Karas M, Bachmann U, Bahr U, Hillenkamp F (1997) Int J Mass Spectrom Ion Processes 78:53–68 3. Karas M, Hillenkamp F (1988) Anal Chem 60:2299–2301 4. Medzihradszky KF, Campbell JM, Baldwin MA, Falick AM, Juhasz P, Vestal ML, Burlingame AL (2000) Anal Chem 72:552–558 5. Pittenauer E, Allmaier G (2009) J Am Soc Mass Spectrom 20: 1037–1047 6. Kubo A, Satoh T, Itoh Y, Hashimoto M, Tamura J, Cody RB (2013) J Am Soc Mass Spectrom 24:684–689 7. Suckau D, Resemann A, Schürenberg M, Hufnagel P, Franzen J, Holle A (2003) Anal Bioanal Chem 376:952–965 8. Suckau D, Resemann A (2003) Anal Chem 75:5817–5824 9. Reiber DC, Grover TA, Brown RS (1998) Anal Chem 70:673–683

2276 10. Stanssens P, Zabeau M, Meersseman G, Remes G, Gansemans Y, Storm N, Hartmer R, Honisch C, Rodi CP, Böcker S, van den Boom D (2004) Genome Res 14:126–133 11. Norris JL, Caprioli RM (2013) Chem Rev 113:2309–2342 12. Hankin JA, Barkley RM, Murphy RC (2007) J Am Soc Mass Spectrom 18:1646–1652 13. Lemaire R, Wisztorski M, Desmons A, Tabet JC, Day R, Salzet M, Fournier I (2006) Anal Chem 78:7145–7153 14. Zemski Berry KA, Hankin JA, Barkley RM, Spraggins JM, Caprioli RM, Murphy RC (2013) Chem Rev 111:6491–6512 15. Chugtai K, Heeren RMA (2011) Chem Rev 110:3237–3277 16. Goodwin RJA (2012) J Proteomics 75:4893–4911 17. Kaletas BK, van der Wiel IM, Stauber J, Dekker LJ, Guzel C, Kros JM, Luider TM, Heeren RMA (2009) Proteomics 9:2622–2633 18. Lanni EJ, Rubakhin SS, Sweedler JV (2012) J Proteome Res 75: 5036–5051 19. Dreisewerd K (2003) Chem Rev 103:395–425 20. Zenobi R, Knochenmuss R (1998) Mass Spectrom Rev 17:337–366 21. Stübiger G, Belgacem O (2007) Anal Chem 79:3206–3213 22. Zehl M, Allmaier G (2003) Rapid Commun Mass Spectrom 17: 1931–1940 23. Woods AS, Huestis MA (2001) J Am Soc Mass Spectrom 12:88–96 24. Fukuyama Y, Iwamoto S, Tanaka K (2006) J Mass Spectrom 41: 191–201 25. Debois D, Smargiasso N, Demeure K, Asakawa D, Zimmerman TA, Quinton L, DePauw E (2013) Top Curr Chem 331:117–141 26. Jaskolla TW, Lehmann WD, Karas M (2008) Proc Natl Acad Sci U S A 105:12200–12205 27. Jaskolla TW, Papasotiriou D, Karas M (2009) J Proteome Res 8: 3588–3597 28. Ens W, Mao Y, Mayer F, Standing KG (1991) Rapid Commun Mass Spectrom 5:117–123 29. Dreisewerd K, Schürenberg M, Karas M, Hillenkamp F (1995) Int J Mass Spectrom Ion Processes 141:127–148 30. Rohlfing A, Leisner A, Hillenkamp F, Dreisewerd K (2010) J Phys Chem C 114:5367–5381 31. Qiao H, Spicer V, Ens W (2008) Rapid Commun Mass Spectrom 22:2779–2790 32. Soltwisch J, Jaskolla TW, Dreisewerd K (2013) J Am Soc Mass Spectrom 24:1477–1488 33. Wiegelmann M, Soltwisch J, Jaskolla TW, Dreisewerd K (2013) Anal Bioanal Chem 405:6925–6932 34. Soltwisch J, Jaskolla TW, Hillenkamp F, Dreisewerd K (2012) Anal Chem 84:6567–6576 35. Dreisewerd K, Schürenberg M, Karas M, Hillenkamp F (1996) Int J Mass Spectrom Ion Processes 154:171–178 36. Knochenmuss R, Vertes A (2000) J Phys Chem B 23:5406–5410 37. Holle A, Haase A, Kayser M, Höhndorf J (2006) J Mass Spectrom 41:705–716 38. Günther S, Römpp A, Kummer W, Spengler B (2011) Int J Mass Spectrom 305:228–237 39. Knochenmuss R, Zhigilei LV (2013) Anal Bioanal Chem 402: 2511–2519 40. Knochenmuss R, Zhigilei LV (2010) J Mass Spectrom 45: 333–346 41. Knochenmuss R (2009) Int J Mass Spectrom 285:105–113 42. Trim PJ, Djidja MC, Atkinson SJ, Oakes K, Cole LM, Anderson DMG, Hart PJ, Francese S, Clench MR (2010) Anal Bioanal Chem 397:3409–3419 43. Spraggins JM, Caprioli RM (2011) J Am Soc Mass Spectrom 22: 1022–1031 44. Vogel A, Venugopalan V (2003) Chem Rev 103:577–644 45. Zhigilei LV, Leveugle E, Garrison BJ, Yingling YG, Zeifman MI (2003) Chem Rev 103:321–347 46. Chen Y, Vertes A (2003) J Phys Chem A 107:9754–9761

K. Dreisewerd 47. Demirev P, Westman A, Reimann C, Håkansson P, Barofsky D, Sundqvist B, Cheng Y, Seibt W (1992) Rapid Commun Mass Spectrom 6:187–191 48. Luo GH, Marginean I, Ye L, Vertes A (2008) J Phys Chem B 112: 6952–6956 49. Knochenmuss R (2002) J Mass Spectrom 37:867–877 50. Knochenmuss R (2006) Analyst (Cambridge, U K) 131:966–986 51. Wichmann J, Lupulescu C, Wöste L, Lindinger A (2009) Eur Phys J D 52:151–154 52. Menzel C, Dreisewerd K, Berkenkamp S, Hillenkamp F (2001) Int J Mass Spectrom 207:73–96 53. Rohlfing A, Menzel C, Kukreja LM, Hillenkamp F, Dreisewerd K (2003) J Phys Chem B 107:12275–12286 54. Dreisewerd K, Berkenkamp S, Leisner A, Rohlfing A, Menzel C (2003) Int J Mass Spectrom 226:189–209 55. Menzel C, Dreisewerd K, Berkenkamp S, Hillenkamp F (2002) J Am Soc Mass Spectrom 13:975–984 56. Rohlfing A, Menzel C, Kukreja LM, Hillenkamp F, Dreisewerd K (2003) J Phys Chem B 44:12275–12286 57. Zhigilei LV, Garrison BJ (2000) J Appl Phys 88:1281–1298 58. Cramer R, Richter WJ, Stimson E, Burlingame AL (1998) Anal Chem 70:4939–4944 59. Pirkl A, Soltwisch J, Draude F, Dreisewerd K (2012) Anal Chem 84:5669–5676 60. Dybvik A, Norberg AL, Schute V, Soltwisch J, Peter-Katalinić J, Vårum K, Eijsink V, Dreisewerd K, Mormann M, Sørlie M (2011) Anal Chem 83:4030–4036 61. Duncan MW, Roder H, Hundsucker SW (2008) Brief Funct Genomics 7:355–370 62. Breuker K, Knochenmuss R, Zhang J, Stortelder A, Zenobi R (2003) Int J Mass Spectrom 226:211–222 63. Hatsis P, Brombacher S, Corr J, Kovarik P, Volmer DA (2003) Rapid Commun Mass Spectrom 17:2303–2309 64. Sleno L, Volmer DA (2005) Rapid Commun Mass Spectrom 19: 1928–1936 65. Pirman D, Kiss A, Heeren RMS, Yost RA (2013) Anal Chem 85: 1090–1096 66. Muddiman DC, Gusev AI, Hercules DM (1995) Mass Spectrom Rev 14:383–429 67. Jimenez CR, Li KW, Dreisewerd K, Mansvelder HD, Brussaard AB, Reinhold BB, Van der Schors RC, Karas M, Hillenkamp F, Burbach JPH, Costello CE, Geraerts WPM (1997) Proc Natl Acad Sci U S A 94:9481–9486 68. Sze ETP, Chan TWD, Wang G (1998) J Am Soc Mass Spectrom 9: 166–174 69. Cramer R, Corless S (2005) Proteomics 5:360–370 70. Towers MW, McKendrick JE, Cramer R (2010) J Proteome Res 9: 1931–1940 71. Lia YL, Gross ML (2004) J Am Soc Mass Spectrom 15:1833–1837 72. Tholey A, Heinzle E (2006) Anal Chem Bioanal Chem 386: 24–37 73. Cramer R, Corless S (2001) Rapid Commun Mass Spectrom 15: 2058–2066 74. Zubarev RA, Kelleher NL, McLafferty FW (1998) J Am Chem Soc 120:3265–3266 75. Syka JEP, Coon JJ, Schröder MJ, Shabanowitz J, Hunt DF (2004) Proc Nat Acad Soc USA 101:9528–9533 76. Frankevich V, Zhang J, Dashtiev M, Zenobi R (2003) Rapid Commun Mass Spectrom 17:2343–2348 77. Kononikhin AS, Nikolaev EN, Frankevich V, Zenobi R (2005) Eur J Mass Spectrom 11:257–259 78. König S, Kollas O, Dreisewerd K (2007) Anal Chem 79:5484–5488 79. Trimpin S, Inutan E, Herath D, Thushani N, McEwen MC (2010) Mol Cell Proteomics 9:362–367 80. Cramer R, Pirkl A, Hillenkamp F, Dreisewerd K (2013) Angew Chemie Int Ed 52:2364–2367

Recent methodological advances in MALDI‐MS 81. Wieser A, Schneider L, Jung JT, Schubert S (2013) Appl Microbiol Biotechnol 93:965–974 82. Lee J, Soper SA, Murray KK (2009) Anal Chim Acta 649:180–190 83. Gemoll T, Roblick UJ, Habermann JK (2011) Mol Med Reports 4: 1045–1051 84. Amantonico A, Urban P, Zenobi R (2010) Anal Bioanal Chem 398: 2493–2504 85. Cole RB (Ed) (2010) Electrospray and MALDI mass spectrometry: fundamentals, instrumentation, practicalities, and biological applications 2nd edition Wiley 86. Benk AS, Roesli C (2012) Anal Bioanal Chem 404:1039–1056 87. Alley WR, Mann BF, Novotny MV (2013) Chem Rev 113:2668– 2732 88. Sandrin TR, Goldstein JE, Schumaker S (2013) Mass Spectrom Rev 32:188–217 89. Hillenkamp F, Katalinić J (Eds) MALDI MS (2013) A Practical Guide to Instrumentation, Methods and Applications 2nd edition Wiley. 90. Karas M, Krüger R (2003) Chem Rev 103:427–439 91. Knochenmuss R, Zhigilei LV (2005) J Phys Chem B 109:22947– 22957 92. Jaskolla TW, Karas M (2011) J Am Soc Mass Spectrom 22:976– 988 93. Teuber K, Schiller J, Fuchs B, Karas M, Jaskolla TW (2010) Chem Phys Lipids 163:552–560 94. Jaskolla TW (2013) Analytix Newsletter (Sigma-Aldrich). Issue 4: 17–19 95. Leszyk JD (2011) J Biomol Tech 21:81–91 96. Jaskolla T, Karas M, (2009) Univ. of Frankfurt/Main WO2009026867– A1 97. Jaskolla T, Karas M (2011) Univ. of Frankfurt/Main WO2011107076– A1 98. Porta T, Grivet C, Knochenmuss R, Varesio E, Hopfgartner G (2011) J Mass Spectrom 46:144–152 99. Somasundaram P, Dreisewerd K, Jaskolla TW (2013) 46th Annual Meeting of the German Society for Mass Spectrometry, Berlin, March 10–13. 100. Karas M, Bachmann D, Hillenkamp F (1985) Anal Chem 57:2935– 2939 101. Beavis RC, Chaudhary T, Chait BT (1992) Org Mass Spectrom 27: 156–158 102. Horneffer V, Dreisewerd K, Lüdemann HC, Hillenkamp F, Läge M, Strupat K (1999) Int J Mass Spectrom 185(186/187):859–870 103. Allwood DA, Dreyfus RW, Perera IK, Dyer PE (1996) Rapid Commun Mass Spectrom 10:1575–1578 104. Wu KJ, Steding A, Becker CH (1992) Rapid Commun Mass Spectrom 7:142–146 105. Chen X, Carroll JA, Beavis RC (1998) J Am Soc Mass Spectrom 9: 885–891 106. Puretzky AA, Geohegan DB (1997) Chem Phys Lett 286:425–432 107. Mowry C, Johnston M (1993) Rapid Comm Mass Spectrom 7:569– 575 108. Dlott DD (1990) J Opt Soc Am B 7:1638–1652 109. Ahn SH, Bae YJ, Kim MS (2012) Bull Korean Chem Soc 33:2955– 2960 110. Hoyer T, Tuszynski W, Lienau C (2007) Chem Phys Lett 443:107– 112 111. Tarzi O, Nonami H, Balsells RE (2009) J Mass Spectrom 44:260–277 112. Knochenmuss R (2014) Analyst (Cambridge, U K) 139:147–156 113. Fournier I, Beavis RC, Blais JC, Tabet JC, Bolbach G (1997) Int J Mass Spectrom 169:19–29 114. Raska CS, Parker CE, Huang C, Han J (2002) Glish Gl, Pope M, Borchers CH. J Am Soc Mass Spectrom 13:1034– 1041 115. Karas M, Bahr U, Strupat K, Hillenkamp F, Tsarbopoulos A, Pramanik BN (1995) Anal Chem 67:675–679

2277 116. Zavalin A, Yang J, Caprioli R (2013) J Am Soc Mass Spectrom 24: 1153–1156 117. Spengler B, Hubert M, Kaufmann R (1994) Proceedings of the 42nd Annual Conference on Mass Spectrometry and Allied Topics, Chicago, Illinois, May 19–June 3, p. 1041 118. Spengler B, Hubert M (2002) J Am Soc Mass Spectrom 13:735–748 119. Günther S, Köstler M, Schulz O, Spengler B (2010) Int J Mass Spectrom 294:7–15 120. Römpp A, Günther S, Schober Y, Schulz O, Takats Z, Kummer W, Spengler B (2010) Angew Chem, Int Ed 49:3834–3838 121. Römpp A, Günther S, Takats Z, Spengler B (2011) Anal Bioanal Chem 401:65–73 122. Chaurand P, Schriver KE, Caprioli RM (2007) J Mass Spectrom 42: 476–489 123. Haase A, Kayser M, Höhndorf J, Holle A, Bruker Daltonik GmbH (2006) DE1020040044196A1 124. Haase A, Kayser M, Höhndorf J, Holle A, Bruker Daltonik GmbH (2006) DE102005006125A1. 125. Hayes JR, Flanagan JC, Monro TM, Richardson DJ, Grunewald P, Allott R (2006) Opt Express 14:10345–10350 126. Jurchen JC, Rubakhin SS, Sweedler JV (2005) Anal Chem 16: 1654–1659 127. Snel MF, Fuller M (2010) Anal Chem 82:3664–3670 128. Westmacott G, Ens W, Hillenkamp F, Dreisewerd K, Schürenberg M (2002) Int J Mass Spectrom 221:67–81 129. Soltwisch J, Souady J, Berkenkamp S, Dreisewerd K (2009) Anal Chem 81:2921–2934 130. Krutchinsky AN, Loboda AV, Spicer VL, Dworschak R, Ens W, Standing KG (1998) Rapid Commun Mass Spectrom 12:508–518 131. O'Connor PB, Costello CE (2001) Rapid Commun Mass Spectrom 15:1862–1868 132. Ingendoh A, Karas M, Hillenkampf F, Giessmann U (1994) Int J Mass Spectrom Ion Processes 131:345–354 133. Luxembourg SL, Mize TH, McDonnell LA, Heeren RMA (2004) Anal Chem 76:5339–5344 134. Hazama H, Aoki J, Nagaoa H, Suzuki R, Tashima T, Fujii K, Masuda K, Awazu K, Toyoda M, Naito Y (2008) Appl Surf Sci 255:1257–1263 135. Azama H, Yoshimura H, Aoki J, Nagao H, Toyoda M, Masuda K, Fujii K, Tashima T, Naito Y, Awazu K (2011) J Biomed Opt 16: 046007 136. Jungmann JH, Smith DF, MacAleese L, Klinkert I, Visser J, Heeren RMA (2012) J Am Soc Mass Spectrom 23:1679–1688 137. Jungmann JH, Smith DF, Soltwisch J, Heeren RMA (2013) Angew Chemie Int Ed 52:11261–11264 138. Stöckle R, Setz PD, Deckert V, Lippert T, Wokaun A, Zenobi R (2001) Anal Chem 73:1399–1402 139. Schmitz TA, Gamez G, Setz PD, Zhu L, Zenobi R (2008) Anal Chem 80:6537–6544 140. Knebel D, Amrein M, Dreisewerd K, JPK Instruments AG (2004) WO2004017019A1 141. Garden RW, Sweedler JV (2000) Anal Chem 72:30–36 142. Qiao H, Piyadasa G, Spicer V, Ens W (2009) Int J Mass Spectrom 281:41–51 143. Luxembourg SL, McDonnell LA, Duursma MC, Guo X, Heeren RMA (2000) Anal Chem 72:30–36 144. Horneffer V, Reichelt R, Strupat K (2003) Mass Spectrom 26:117– 131 145. Wetzel SJ, Guttman CM, Flynn KM (2004) Rapid Commun Mass Spectrom 18:1139–1146 146. Gusev AI, Wilkinson WR, Proctor A, Hercules DM (1995) Anal Chem 67:1034–1041 147. Soltwisch J, Berkenkamp S, Dreisewerd K (2008) Rapid Commun Mass Spectrom 22:59–66 148. Vaidyanathan S, Gaskell S, Goodacre R (2006) Rapid Commun Mass Spectrom 20:1192–1198

2278 149. Wang BH, Dreisewerd K, Bahr U, Karas M, Hillenkamp F (1993) J Am Soc Mass Spectrom 4:393–398 150. Knochenmuss R, Dubois F, Dale MJ, Zenobi R (1998) Rapid Commun Mass Spectrom 10:871–877 151. Nelson RW, Krone JR, Bieber AL, Williams P (1995) Anal Chem 67:1153–1158 152. Li YL, Gross ML (2004) J Am Soc Mass Spectrom 15:1833–1837 153. Armstrong DW, Zhang LK, He LF, Gross ML (2000) Anal Chem 73:3679–3686 154. Cramer R, Karas M, Jaskolla TW (2014) Anal Chem 86:744–751 155. Zabet-Moghaddam M, Krüger R, Heinzle E, Tholey A (2004) J Mass Spectrom 39:1494–1505 156. Bungert D, Bastian S, Heckmann-Pohl DM, Giffhorn F, Heinzle E, Tholey A (2004) Biotech Lett 26:1025–1030 157. Leisner A, Rohlfing A, Berkenkamp S, Dreisewerd K, Hillenkamp F (2004) J Am Soc Mass Spectrom 15:934–941 158. Fan X, Little MW, Murray KK (2008) Appl Surf Sci 255:1699– 1704 159. Trimpin S, Wang BX, Inutan ED, Li J, Lietz CB, Harron A, Pagnotti VS, Sardelis D, McEwen CN (2012) J Am Soc Mass Spectrom 23: 1644–1660

K. Dreisewerd 160. Sampson JS, Hawkridge AM, Muddiman DC (2008) Anal Chem 80:6773–6778 161. Witt L, Pirkl A, Dreisewerd K, Mormann M (2012) 45th Annual Meeting of the German Society for Mass Spectrometry, Posnań, March 10–13 162. Schürenberg M, Schulz T, Dreisewerd K, Hillenkamp F (1996) Rapid Commun Mass Spectrom 10:1873–1188 163. Wang B, Inutan ED, Trimpin S (2012) J Am Soc Mass Spectrom 23: 442–455 164. McEwen CN, Pagnotti VS, Inutan ED, Trimpin S (2010) Anal Chem 82:9164–9168 165. Nyadong L, Inutan ED, Wang X, Hendrickson CL, Trimpin S, Marshall AG (2013) J Am Soc Mass Spectrom 24:320–328 166. Tyler BJ, Brüning C, Rangaranjan S, Arlinghaus HF (2011) Biointerphases 6:135–141 167. Benabdellah F, Seyer A, Quinton L, Touboul D, Brunelle A, Laprevote O (2010) Anal Bioanal Chem 396:151–162 168. Touboul D, Roy S, Germain DP, Chaminade P, Brunelle A, Laprevote O (2007) Int J Mass Spectrom 260:158–165 169. Monroe EB, Annangudi SR, Hatcher NG, Gutstein HB, Rubakhin SS, Sweedler JV (2008) Proteomics 8:3746–3754