Protein and peptides in pictures: Imaging with MALDI mass spectrometry

Transcription

1 Proteomics 2008, 8, DOI /pmic REVIEW Protein and peptides in pictures: Imaging with MALDI mass spectrometry Richard J. A. Goodwin 1, Stephen R. Pennington 2 and Andrew R. Pitt 1 1 Institute of Biomedical and Life Science, University of Glasgow, Glasgow, UK 2 UCD Conway Institute, University College Dublin, Belfield, Dublin, Ireland Imaging using MS has the potential to deliver highly parallel, multiplexed data on the specific localization of molecular ions in tissue samples directly, and to measure and map the variations of these ions during development and disease progression or treatment. There is an intrinsic potential to be able to identify the biomarkers in the same experiment, or by relatively simple extension of the technique. Unlike many other imaging techniques, no a priori knowledge of the markers being sought is necessary. This review concentrates on the use of MALDI-MS for MS imaging (MSI) of proteins and peptides, with an emphasis on mammalian tissue. We discuss the methodologies used, their potential limitations, overall experimental considerations and progress that has been made towards establishing MALDI-MSI as a routine technique for the spatially resolved measurement of peptides and proteins. As well as determining the local abundance of individual molecular ions, there is the potential to determine their identity within the same experiment using relatively simple extensions of the basic techniques. In this way MSI offers an important opportunity for biomarker discovery and identification. Received: April 10, 2008 Revised: May 26, 2008 Accepted: May 27, 2008 Keywords: Biomarkers / Imaging mass spectrometry / MALDI / Tissue 1 Introduction and overview Molecular imaging is one of the cornerstones of the experimental methods available for understanding the spatial complexity of biology at the tissue and cellular level. Ideally imaging methods should provide detailed data on a number of tissue components and provide a resolution that approaches that of individual cells. Most current methodologies can only be performed on a limited number of components, for example fluorescence (see [1] and references in the corresponding special issue of Nature Methods), radiochemical Correspondence: Dr. Andrew R. Pitt, Joseph Black Building, Institute of Biomedical and Life Science, University of Glasgow, Glasgow G12 8QQ, UK a.pitt@bio.gla.ac.uk Fax: Abbreviations: MSI, MS imaging; OCT, optimal cutting temperature medium; QqTOF, quadrupole-tof; SIMS, secondary ion MS tracers or positron emission tomography [2], or can only measure bulk properties, such as NMR spectroscopy [3]. In addition, and more significantly, these often require a priori knowledge of the molecules of interest, and so are less effective as a true discovery tool. Imaging using MS fulfils many of the requirements for effective imaging, with the ability to perform a complex analysis of multiple components, encompassing the molecular size scale from metabolites and drugs to medium sized proteins, with a spatial resolution of 50 mm or better. With commercial, off-the-shelf packages becoming readily available, MS imaging (MSI) is increasingly within the reach of the broader scientific community, and can now be applied to translational research, especially clinical studies. However, there is still a level of expert intervention needed to make the best of the technique, and some way to go before its full potential is achieved. Imaging using MS is a new and rapidly growing area, as evidenced by a number of recent reviews of different aspects of the technique [4 12], although the number of primary papers reporting real biological applications is still relatively limited.

2 3786 R. J. A. Goodwin et al. Proteomics 2008, 8, In this review, we focus on the application of MALDI-MSI to peptides and proteins. Special consideration will be given to sample preparation, instrument design and data analysis, with an emphasis on mammalian tissue. We will also try to highlight some of the current shortcomings and assess what advances are needed to overcome them. In essence, MALDI-MSI comprises the collection of a mass spectrum at each of a regular series of points across a section of tissue. Using these spectra it is then possible subsequently to plot the relative intensity of the individual m/z peaks in the mass spectrum across the tissue, and thus to visualize the distribution of the individual molecular ions. This provides a pixilated molecular intensity map which can be viewed as an intensity or pseudocolour heat map. The outline of the steps in this process is shown in Fig. 1. This can be performed for each mass peak of interest in the spectrum, generating 2-D distribution maps of any or all of the observed components. Reproducible sample preparation, mass spectral acquisition and data processing allows these maps to be compared between related samples to assess what changes occur, for example, during tissue development or disease progression or treatment. These two dimensional images can also be taken in series through a number of serial sections from the tissue and combined to generate a 3-D map [5, 13], reconstructing the volumetric data. MALDI-MSI has been demonstrated to be applicable to many tissues, from brain [9], which is the most commonly studied tissue, to whole animals [14]. Brain is the most studied tissue, not only because it is a relatively robust tissue that sections well, but also because it has a complex architecture where the overall function is very dependent on regiospecific changes. It is also appealing as MSI data can be integrated with other imaging data such as functional magnetic resonance imaging. The basic methodology of MALDI-MSI is simple: MALDI matrix is applied to the tissue section, and the section is then analysed by MS to allow the spatial, spectral composition to be plotted. The sample preparation and analysis methodologies originally pioneered by Caprioli, Gile, Stoeckli and Chiraud [15 17], and those that continue to be developed by these and other groups are, in practice, much more complex, with a number of technical challenges limiting MALDI-MSI from becoming a routine technology. In addition, although the intrinsic mass range of MALDI- MSI mass spectrometers may be quite broad, the actual effective mass range in imaging mode is currently substantially more limited. The lower mass region (,600 Da) is partially obscured by ions from the matrix required for biomolecular ionization, leading to the requirement for relatively high concentrations of a particular analyte in order for it to be detectable. Furthermore, sensitivity drops off very rapidly at higher mass, making the current practical upper limit around kda. The steps in MALDI-MSI workflow, and the considerations needed at each one, are shown in Fig. 2. For each step in the basic process there is an increasing array of reported alternatives, modifications and addi- Figure 1. The basic premise for mass spectrometric imaging. (a) An optical image of the tissue section mounted on the MALDI target is taken, complete with registration points, so that the areas to be images by MS can be delineated. Shown is an optical image of unstained central coronal brain section from mouse captured using a standard macro lens CCD system. (b) After coating of the sample with matrix, the sample is loaded into the mass spectrometer and the area to be imaged is defined. A Cartesian array of points at which imaging is to be performed is then overlaid on this area of the image. The separation of the imaging points determines the resolution at which the image is obtained. A mass spectrum is then obtained at each of the points on the grid. (c) The intensity of any mass window in the mass spectrum can then be plotted to show the intensity of the signal in the mass spectrum. The image shown is an intensity plot of m/z Da, from a50mm raster, with some of the fine features being only mm (1 2 pixels) across.

3 Proteomics 2008, 8, Technology 3787 Figure 2. Schematic diagram of the MALDI-MSI workflow showing the key steps as outlined in this review and the factors that need to be considered to minimize variation and optimize performance. tions that are currently being explored in order to address the current limitations. Therefore, this review will also attempt to cover the major factors affecting MSI and will address some of the possible routes for technological improvement that could be considered for further improving MSI. Finally, as with any primary discovery-based method, it is still necessary to validate the results using a complementary method, and then assess the results for biological significance. 2 Collection and storage of samples The first stage of the MALDI-MSI process is the collection and storage of tissue. The considerations are as important here as they are in any biomarker discovery programme using biological samples. These issues have been discussed in detail elsewhere (see for example [18, 19]). It is worth noting that most studies have sought to minimize sample handling artefacts, and relatively few have been directed towards a systematic evaluation of the critical factors in sample preparation, especially by directly assessing what protein degradation has occurred. In general, sample handling, from the point of collection to the point of imaging, is one of the most critical points in the overall MSI process, and it needs to be performed in such a way as to maintain the integrity and spatial localization of the molecules in the original tissue. As with all biological tissue samples, once the tissue has been excised molecular degradation and modification is initiated or will continue. While any alteration, modification or evolution at the proteomic level may not always be as rapid as on the metabolomic level, they are still rapid and varied enough that all tissue preservation or processing should be performed to rigorous and reproducible standards. In addition, it is important to emphasise that at this stage that subtle experimental variability in tissue handling can produce variations in tissue biomarker profiles that may dramatically affect downstream MALDI-MSI analysis over the time course of an experiment. Many of the basic requirements are taken from well-established histological methods (see for example to the point where the tissue would be fixed. The simplest, and most commonly employed, protocol for tissue storage is for material to be immediately snap-frozen postharvesting, and then stored at 2807C until required for analysis. However, even the freezing process needs to be considered as a potential source of sample variation. Dropping fresh tissue into liquid nitrogen usually leads to cracking and fragmentation as parts of the tissue cool at different rates, making sectioning, and comparison of biological replicates difficult, therefore the use of ethanol or isopropanol at temperatures of 2707C or lower is recommended. It is also important to try to maintain the shape of the tissue at this stage, so the common practice of wrapping the tissue in foil or plastic film, or containment within a centrifuge tube, is not recommended due to the resulting deformation of the tissue as it follows the contours of the container. If multiple sections are required, or material will be required for downstream proteomic analysis or validation experiments, it is best to section all the required tissue at the same time to avoid repeated handling of the tissue and the potential for localized warming, although storage of the tissue sections may increase potentially deleterious effects, such as oxidation. Finally, it is essential when samples are acquired to ensure that the appropriate meta-data is captured at this stage in the experiment, especially if clinical materials are being analysed. Such data will include biological and geographical sources of samples, experimental protocols, dates and deviations from standard operating protocols, but should be as comprehensive as possible. Compliance with currently accepted proteomic standards initiatives is increasingly required for publication of results, and is valuable for data exchange (see for example the following: HUPO Protomic Standards Initiative site and the associated MIAPE (minimum information about a proteomics experiment) information and the

4 3788 R. J. A. Goodwin et al. Proteomics 2008, 8, PRIDE (proteomics identifications database) site at the European Bioinformatics Institute, pride/), as well as following the guidelines of the major publishing houses. Detailed descriptions of the sample, its history and the processing and analysis applied to the data are necessary, as it is routine for slight variations to the protocol to be made on-the-fly during an experiment. For example, one seemingly simple modification to standard protocols that we have found valuable and worth recording is direct measurement of the amount of matrix deposited by weighing each slide before and after spray coating. 3 Tissue preparation and matrix application Although considerable research has been undertaken to produce standardized methodologies for tissue preparation and matrix application, there is still considerable scope for sample-dependent validation, optimization and basic tissuespecific characterization. Clearly much of the tissue preparation and matrix choices for MALDI-MSI are based on wellestablished methods used in related fields, but it is notable that this is one area in which novel research has recently lead to a vast improvement of the quality of the information obtained from MALDI-MSI, as described in detail in the following sections. The final step of sample preparation covers matrix selections and application. This used to be based on personal experience and trial and error, but with the ever increasing popularity of MSI several companies have designed matrix applicators to automate this process, allowing greater confidence in inter-run reproducibility. Assuming that tissues have been collected and stored using standardized methods, it is perhaps fair to note that tissue preparation is the area where most of the improvement to the quality of data collected can be made. Poorly considered MALDI-MSI experiments are likely to produce data of an inadequate standard. This section will now follow the order of a typical MALDI-MSI workflow, detailing stages where experimental variables need to be considered, and will highlight a number of additional treatments that can be used to help minimize variation. As with any emerging field, the protocol improvements reported in the literature are sometimes rather subjective, often with no associated rigorous characterization or validation. However, these observations often do seem to improve the data collected. Therefore, even where such observations are reported without significant published data or validation, details on the recommendations will still be provided in this review, as they provide an overview of the concerns being considered to improving the quality of MSI data. It is worth noting that as MSI develops it is also likely to be applied more and more in a way that is best described as MSI targeted proteomics, where data collected from MSI is used to inform and direct more traditional downstream proteomic analyses, such as LC-MS or 2-DE, either on whole tissue sections, or using material collected by laser microdissection or microcapture to enrich samples prior to traditional proteomics [20 25]. Therefore it is often prudent to collect material for this purpose, and for any required immunohistochemistry, at the time of cutting material for MSI. This also has the benefit of allowing the experimenter to take adjacent sections and hence increase the ease by which results from different experiments can be collated. 3.1 Section mounting Fixed versus fresh tissue To date, the vast majority of MALDI-MSI has been performed on fresh, snap-frozen, chemically unmodified tissue sections, although potentially important inroads are being made slowly into the analysis and interpretation of the complex crosslinked proteins that are present in tissue sections that have been preserved specifically for study by pathologists by formalin fixing and paraffin embedding [26, 27]. The more routine use of such tissue would allow access to the huge resource of tissues that are stockpiled in numerous tissue banks around the world. It is impressive in this regard that, following protocols to first remove embedding paraffin, and then perform in situ enzymatic digestion, information can now be gleaned from such highly modified tissue. However, questions still remain about the validity of information obtained by MALDI-MSI of such tissues. These include what effects the widely varying fixing and unfixing protocols make to imaging reproducibility, what effects the processing has on sensitivity and hence the ability to investigate the less abundant molecular features and what variations in the protein modification and oxidation additionally present in these tissues that may cause changes in the robustness of quantitative or qualitative comparisons. Further experiments to demonstrate that the fixing and subsequent steps in the process do not significantly alter the imaging of specific biomarkers will be required. Protocols to overcome the complications of integrating MSI data with the optical features observed using traditional section staining histology/pathology have been described [28] and a new matrix solution fixation protocols has recently been reported that incorporates simultaneous matrix deposition and tissue fixation, allowing histology-compatible tissue preparation for MSI [29]. Alternative fixative methods have reported that rely on heat stabilization of post mortem tissues, which rely on protein denaturation to effectively preventing proteomic cellular activity [19], and we have recently demonstrated that the commercially available system from Denator can be applied to tissue for MSI (Fig. 3). However, such treatments may themselves have an affect on the MSI of tissues, although in our hands this seems to be much reduced over any other protocols. It is also worth considering that fixing the tissue removes the opportunity to use

5 Proteomics 2008, 8, Technology 3789 Figure 3. Degradation of molecular species during a time course of incubation of the tissue at room temperature, showing rapid degradation of markers on warming of the samples and stabilization by heat treatment of the tissue using the Denator system. (a) Coronal sections of mouse brain 14 mm thick thaw mounted onto indium-tin oxide coated MALDI target slides. A 50 mm laser spot and 150 mm raster centre-to-centre was used. Samples were manually spray coated with 10 mg/ml CHCA matrix using a standard tlc sprayer. Sections were warmed on the target slide for 0, 0.5, 1, 1.75, 2.75 and 5.25 min (labelled 1 6, respectively). Upper and lower series are from two separate technical replicates. The mass plotted is m/z Da and can be seen to decrease substantially across the time course. (b) Parasagittal sections of mouse brain 14 mm thick thawmounted onto indium-tin oxide coated MALDItargetslides. A50 mm laserspot and 150 mm raster centre-to-centre was used. Samples were manually spray coated with 10 mg/ml CHCA matrix using a standard tlc sprayer. All sections were incubated for 5 min and the plotted mass is again Da, corresponding with time point 6 in the times course in panel (b). For sample 1 the tissue was heat treated (Denator system) to denature proteins immediately the tissue had been excised. Sample 2 is the corresponding untreated tissue and sample 3 was heat treated on the slide after sectioning and thaw mounting. It can be seen that the marker is retained in the immediately treated tissue whereas for the tissue that was allowed to warm for any length of time the marker was significantly reduced in intensity. intrinsic degradation as an additional piece of information on the tissue that additionally brings biomarkers into the range accessible by MS Embedding Traditional pathology techniques usually require the embedding of tissue in a supporting material, such as paraffin wax, to provide support during the microtoming of the sections. For nonfixed tissues that are being cut on a cryostat microtome, the support media is replaced with optimal cutting temperature media (OCT). In MALDI-MS, this material is found to ionize easily and acts as a significant ion suppressor, as well as cluttering the spectrum with contaminant peaks [8, 30] and, therefore, embedding of tissue should be avoided and care taken to avoid contamination of the tissue with OCT. Thorough cleaning of the cryostat microtome should be performed before sectioning for MALDI-MSI if the

6 3790 R. J. A. Goodwin et al. Proteomics 2008, 8, equipment is also used by traditional pathologists. The standard protocol for cutting sections adopted by most researchers is to mount the base of the tissue to be sectioned for MSI in the minimum amount of OCT, leaving the part to be sectioned standing well proud of the embedding medium. Where embedding is essential, some success has been achieved using polymeric resin [14] or simply in ice as the embedding medium, but much more needs to be done to optimize these protocols Tissue thickness The appropriate thickness for tissue sections is a much debated area. MSI initially employed relatively thick sections ( mm) [15], although tissue section thickness was soon optimized to mm [9, 12]. More recent publications [13, 31 33] continue to show high quality data from these medium thickness sections, which seem to provide a good compromise between optimum performance of the MALDI-MSI and having adequate material present to ensure a good S/N. There is a tendency for thicker sections to distort during the washing and drying process, which can result in complications, especially where coregistration with other images is required, such as in 3-D reconstruction applications [5]. Tissue thicknesses of 3 5 mm has been recommended to obtain higher quality MS data, especially in the higher m/z regions [34, 35], where it has been proposed that thicker sections produced poorer quality data as a result of biological compounds that interfere with the MALDI process being insufficiently washed away during ethanol wash stages, and that thinner sections therefore had lower charging effect on tissue as during the matrix application the surface of thicker sections become more isolated from the conductive region of the MALDI target Thaw mounting There are two approaches to attaching tissue sections to MALDI targets, the use of adhesive tape to stick the section to the target [36], or the thaw mounting of the section directly off the cryostat microtome cutting plate, with successive sections mounted by warming the reverse of the target to produce a localized warm patch. The latter method is the approach most commonly employed as the thawing of the section onto the target adheres the section well enough, and with lower risk of sample contamination, that subsequent required washing stages rarely result in loss of any of the tissue. However, a concern is that this thaw mounting stage, and any variation in the time taken to process the sections once the target has been removed from the cryostat, can cause significant variation in the mass profiles, as rapid degradation occurs [19]. Hence all experimental protocols need to be standardized if inter-run comparisons are to be made. Additionally, thaw mounting can limit what processing can be performed downstream of the mounting step, as the tissue risks being washed off the surface by any vigorous solution-based treatments, especially those containing polar solvents. 3.2 Sample pretreatment prior to matrix application Washing protocols Following sectioning of the tissue washing is required to remove small molecules such as salts present in the tissues from interstitial fluids or from cells that are ruptured either during freezing and thawing or during the cutting process. These small molecules form adducts with the proteins and peptides, making the spectra more complex, and also suppress ionization. Some of the principal exponents of MALDI- MSI recommend immediate drying of the sectioned sample in a desiccator, followed by freeze drying prior to washing [13, 37], although many others omit this freeze drying step with satisfactory results. Significantly improved signal intensity has been reported following the washing steps; typically a brief 70% ethanol wash can be used to wash away both cell salts and debris, followed by a % ethanol wash to help complete tissue dehydration and to temporarily fix the tissue and prevent subsequent degradation of the proteome [28, 30, 38]. For many tissues, lipids are in high concentration and can themselves often cause suppression of ionization, so the use of an organic solvent treatment, such as chloroform washes, have been used to remove them [33]. Increasing the number and duration of washes, as well as alteration of the composition of the solvent, may also improve the MSI where signal intensity seems low. However, such additions to the number of steps in the protocols increase the possibility of loss of the more soluble target analytes into the wash buffer or may result in alterations in their abundances that significantly skew the mass spectra obtained [5]. Alternative approaches under evaluation that may help to combat the loss of sensitivity due to ion suppression include the use of sputter coated gold monolayers [27, 39, 40] On tissue digestion (spot and spray) It is not generally possible to detect larger proteins by MALDI-MSI; the upper limit for reasonably sensitive detection is currently kda, depending on the instrumentation used. However, the localization of larger proteins may be enabled by the use of in situ digestion on the tissue section. The smaller peptide fragments generated by such enzymatic cleavage are much easier to detect by MALDI-MSI [41], and should still represent the abundance and localization of the parent proteins, as the spatial positions of the parent protein can be inferred from the localization of the fragment. In addition further fragmentation of the peptides in the mass spectrometer also allows a better possibility of identifications of the unknown proteins directly from the tissue, as in general the methods used for MS/MS identification only work

7 Proteomics 2008, 8, Technology 3791 well for smaller peptides. A significant concern is that in situ on-tissue digestion requires that the biological material has to remain moist to maintain enzyme activity. This increases the possibility of analyte (peptide) diffusion during the course of the incubation. Enzyme application, either through spray-coating or spotting, also requires careful consideration, and is probably more of an issue than application of matrix alone, especially for hydrophilic peptides and proteins. Despite this the standard technologies used for enzyme coating are the same as those used for matrix application (see Section 3.3). Where high resolution spatial information is not required, spotting of the digestion enzyme has the benefit of controlling the incubation conditions and analyte diffusion and is often the method of choice. Spray coating covers the entire surface of the sample, so the opportunity for diffusion of the analyte is significant. However, if this is controlled then good resolution imaging can be achieved [41]. Application of enzyme by discrete spotting results in discrete areas that can be analysed, with a minimum diameter of around 150 mm and centre-to-centre distance of 250 mm, limiting the resolution achievable. 3.3 Choice and application of matrix The process for obtaining high quality MALDI-MSI mass spectra relies fundamentally on effective MALDI matrix deposition on the sample. The matrix application method must extract the maximum amount of analyte from the sample while minimizing diffusion and ensuring proper crystallization of the matrix. This can best be achieved by balancing the requirement of applying the correct matrix in the right solvent composition in conditions wet enough that the solution is able to draw out the proteins and peptides and crystallize with them, but not so wet as to allow excessive diffusion of the target analytes Choice of matrix For MALDI-MSI of peptides and proteins, as for standard MALDI-MS, the choice of matrix is normally determined by the mass range over which the experiment is to be performed. Therefore, for lower molecular weight masses (peptides), typically using a TOF mass spectrometer in reflectron mode or when coupled to non-tof analysers, CHCA is used. For higher molecular masses (proteins), typically using the mass spectrometer in linear mode, sinapinic acid (3,5-dimethoxy-4-hydroxycinnamic acid, SA) is normally used. A summary of the crystallization characteristics of the common matrixes on tissue sections was reviewed early in the development of MSI protocols [30] and is reported in detail in another review [42]. As described above, the key is to generate crystals of matrix that are small enough to be below the resolution of the imaging itself, while still ensuring good inclusion of analyte into the crystals and minimum diffusion during deposition and crystal formation. Alternative ionic liquids composed of matrix mixtures, for example consisting of CHCA acid with 2- amino-4-methyl-5-nitropyridine and CHCA with N,N-dimethylaniline, have been used to reportedly give better spectral quality in terms of resolution, sensitivity, intensity, noise and number of species detected [40] Matrix application methods Matrix application is the point in most protocols where poor experimental method results in a failure to produce high quality data, and which requires the greatest amount of expert feel for the technique. The main methods of matrix application used today are manual or automated spotting [38, 43, 44], where the matrix is placed discrete spots and any diffusion is therefore contain within the spot, and spraying coating [45] which aims to coat the whole of the surface of the sample, resulting in a continuous, even layer of matrix. Both spot and spray coatings can be achieved manually, with varying success, but the choice of method depends on a number of factors, including the type of information required in the analysis and the availability of automated systems. Manual spotting using a pipette, while not particularly suitable for MSI, has been reported to be excellent for MS profiling, the forerunner of MSI where high resolution mapping of the distribution of the molecular species in the sample was not attempted. Up to 1000 distinct mass signals were observed in the spectrum from each individual spot, and, more significantly, an upper m/z limit exceeding Da could be obtained [7]. As manual spotting results in the dispensing of ml volumes of matrix solution compared to the pl volumes that can be robotically spotted, the resulting low spatial resolution make it less suitable for MSI, and it has been superseded in the main by automated piezoelectric or acoustic spotting, and spray coating techniques. However, as large clinical studies are undertaken manual spotting may well once again be useful in higher throughput studies, following more detailed analysis using the modern matrix application techniques, as tissue profiling can overcome the limitations in throughput of other methods. Robotic spotting systems that enable the deposition of discrete pl droplets of matrix, giving matrix spot sizes on the tissue of around mm, and hence a minimum resolution of around 200 mm spot-centre-to-spotcentre, based around ink jet [43, 44] and acoustic deposition technologies [30, 38] have been reported and are commercially available. This is generally seen to be the method of choice if MS/MS sequencing of the peptides directly from the tissue if identification of the parent protein is to be performed, as extraction appears more efficient, giving better S/N, and the spot provides a larger homogeneous area over which to raster the laser to collect the large number of laser shots usually required. Another advantage of these systems is that they can also be used for spotting other materials, such as proteases, on to the tissue surface for on-tissue digestion.

8 3792 R. J. A. Goodwin et al. Proteomics 2008, 8, Manual spray coating is typically achieved using a pneumatic airbrush or TLC sprayer. This method probably still provides the best approach for matrix application for sections that are either to be analysed individually or compared to other sections on the same target, but only in the hands of an experienced and capable operator, as the technique requires a significant amount of training, practice and experimental confirmation of operator ability. Significant advantages are the speed of the process, which can be completed rapidly compared to the long periods of time automatic coating systems take (minutes compared to hours), and the significantly reduced up-front investment needed. The risk of analyte spreading is, perhaps, the major concern associated with the coating methods of matrix application. However, excluding dip coating methods which were soon ruled out, there has been little evidence that the application of matrix, even with manual pneumatic spray coating if this is done well, induces analyte spreading that affects the image quality at the resolution at which the majority of MSI experiments are being performed (.50 mm). Analyte spreading does become a concern when spray coating to dispense solutions for on-tissue chemistry or digestion is being performed, especially if the tissues are to be incubated in a humid atmosphere. During matrix coating any alteration in matrix coverage, thickness, wetness upon application, and reproducibility across the sample, may produce significantly disparate results. Automated matrix application equipment has, therefore, recently been developed which aims to standardize the application of matrix on the sample, although we have observed an uneven distribution of matrix towards the extremities of the slide with some of the robotic coating systems. In practice, most researchers will start in the field by using manual spray coating, but it can be anticipated that as the automated systems decrease in price and increase in robustness, these may become the method of choice. The human control of the coatings may be a rather dark art, but in our hands can often produce the highest quality data. Alternative methods for matrix coating that seem to show some advantages, especially in the size of matrix crystals formed, but have been less widely used, include sublimation [46], which forms very small crystals (or an amorphous layer) that could be useful for improved resolution methods, but is generally seen to give lower quality spectra [47], and electrospray deposition [15, 30, 45, 48]. In-depth analyses of the quality of the crystal formation formed in all of these spray processes has been performed [47], but this is one area that arguably should receive additional attention. In addition, matrix preseeding techniques, where finely ground dry matrix is lightly brushed on to the tissue before coating, generating nucleation centres for crystallization, have been described to give more uniform crystallization, and to generate smaller crystals for higher resolution imaging [15, 38, 49, 50]. As with much of the MSI workflow, the choice of matrix application is often experiment and sample dependent, and typically determined by operator ability and equipment availability. However, matrix application is a critical factor that can vastly alter the quality of spectral data obtained, and the requirement of inter-run reproducibility and operator independence is a driving factor in the development and use of robotic and mechanical systems Alternative matrixes and target surfaces A number of alternatives to the classical UV absorbing matrices used in MALDI have been described for MALDI-MSI, although these have usually been developed for standard MALDI applications and are not always easy to implement for imaging. Desorption-ionization on silicon (DIOS) has been described as a matrix-free method for MALDI for some time for small molecule analysis where interference from matrix limits the sensitivity [51 54], and for improved peptide coverage in PMF MALDI applications [55]. This has now been implemented in MSI, by transferring the analytes to the DIOS surface by contact with the tissue of interest [51]. This has the potential to increase the sensitivity of the MALDI, especially at low mass; however, further optimization is required before this technique is robust enough to be generally applied. Colloidal graphite-assisted laser desorption/ionization MS, another surface-based method, has also been demonstrated to work in MSI application by spraying graphite nanoparticles onto the surface of the tissue, although there are still significant issues with sensitivity, suppression of ionization and optimum distribution of the particles [56, 57]. Carbon nanotubes, generated in situ by chemical vapour deposition, have also been demonstrated to work as a matrix which gives minimal interference, although this has not yet been shown to work on proteins or peptides [58]. Clathrate nanostructures were recently reported as a method of obtaining spatial resolutions below the micrometer level, similar to that observed in secondary ion MS (SIMS), while maintaining the soft nature of MALDI ionization that maintains the integrity of the labile biological analytes [59]. While this is technically demanding and still in the very early days of development, it demonstrates that there is still significant scope for improving the quality of data obtained from high resolution MSI methods. In conclusion, it is important to note that for all stages in an MSI experiment, a degree of experiment specific optimization is still required, and no clear optimum recommendation has been reached. 4 Analysis hardware The idea of using MS to image tissue has been around since the early 1990s, when the well-established SIMS approach was employed [60]. MALDI-TOF methods became available in the late 1980s [61], but it is only relatively recently that technical advances in the use of MALDI-MS have enabled MSI to become useful for studying labile biomolecules in the

9 Proteomics 2008, 8, Technology 3793 laboratory, with Caprioli et al. [15] reporting the first use of MALDI for tissue imaging in Hardware improvement include changes to the mass spectrometers themselves, especially the introduction of high repetition-rate frequencytripled solid-state Nd:YAG lasers with performance close to that of the traditional nitrogen lasers [62], improvements in the source optics, and the availability of software suites specifically tailored for data acquisition, visualization and analysis in imaging, for example FlexImaging from Bruker Daltonics, or the BioMap software of Stoeckli et al. [17]. Previously, researchers had to assemble the necessary instruments, software and related equipment separately, but now MALDI-MSI packages are available commercially, making them more generally accessible. The recent availability of automated systems for matrix application, be it nebulizerbased systems such as electrospray or acoustic spraying, or droplet based, as described in Section 3.3, have reduced the expert steps to cryostatic microtoming and mounting of the sample on the MALDI target to generate the thin sections for analysis. However, as for most advanced technologies, there is no substitute for expert knowledge of every step of the process for obtaining the best from the methodology. The key instrument dependent stages in the process that will influence data quality are ionization, specifically the laser, ion extraction and source geometries, and the mass analysers. This review concentrates on MALDI-based ionization, which is the most commonly used method for tissue imaging, although developments allowing biomolecule analysis using SIMS [63, 64], which offers significantly better spatial resolution, will soon provide competition, and desorption ESI is a new soft matrix-free ionization method that has great potential for imaging [65]. SIMS methodology is reviewed elsewhere in this issue. MALDI sources have now been fitted to most types of mass spectrometer, including trap [66 68] and Fourier transform [29, 31] machines, but the workhorse machines use TOF and TOF TOF or quadrupole-tof (QqTOF) geometries. 4800) which is very fast but has limited positional accuracy (around 25 mm), and mechanical or piezoelectric methods used in most instruments, which can have a 5 mm or better accuracy, but where the movement is slower. The laser beam spot size has traditionally been in the mm range, which gives relatively low resolution when compared to a typical mammalian cell size of 10 mm, but this has been sufficient to generate images of larger tissue regions with adequate visualization of structures (see Figs. 1, 3 and 4 for examples). This limitation of spot size can be overcome to a certain extent by using a larger laser spot (e.g. 150 mm) with a smaller centre-to-centre step of the raster (say 50 mm). If all of the sample is ablated at each raster position before moving on to the next, an apparent resolution less that the size of the laser beam can be achieved. However, this complete ablation of the sample is not always easy to obtain, and edge effects around the laser ablation site, especially on medium to thick tissue sections, mean that in practice this is not so easy to accomplish. The laser spot size has been reduced sub-stantially to a spot size of around 20 mm, albeit with the trade 4.1 Lasers In a typical MSI experiment the laser beam is rastered across the surface of the tissue, collecting data at each raster point. In this process it is important to consider the resolution at which imaging is required, and the speed with which this can be achieved. The final image resolution is determined by the sample preparation, specifically the size of the crystals of matrix and diffusion of the analyte, as discussed in Section 3, and the technical attributes of the instrumentation. On the technology side, the obvious key features are the accuracy of locating the laser beam on the target and the size of the laser beam spot. The raster of the laser beam is usually achieved by moving the target itself rather than the laser, as the positioning of the laser beam is fixed in an optimized position for the operation of the source. Movement of the sample has to be rapid and accurate, and common methods include both magnetoelectrical devices (e.g. as on the Applied Biosystems Figure 4. Intensity plot images of parasagittal sections of mouse brain imaged at 100 mm (left image) and 75 mm (right image), demonstrating no significant loss of information on moving to the less dense raster, which results in an approximate halving of acquisition time and data file size. The complete image for the two sections together collected some raster points, took 18 h to run and generated a data file of 35 Gb. The images shown are of (A) m/z Da, (B) Da, (C) Da and (D) Da. The distribution and localization of the individual masses is easily observed, for instance m/z is localized to the cerebellum, whereas m/z 4941 is virtually absent from the same tissue.

10 3794 R. J. A. Goodwin et al. Proteomics 2008, 8, off of significantly reduced sensitivity, by modification of commercial instrumentation [42]. Focussing of the laser on an optical bench has been coupled to smart modulation of the laser beam to obtain a laser pulse that works well for all matrices and maximizes ionization of the sample [69] (now marketed as Smartbeam by Bruker Daltonics). This provides characteristics close to that of the slower N 2 lasers that are optimized for the range of matrices usually employed in MALDI. High resolution scanning with very focussed laser beams in the conventional microprobe mode, where the beam is scanned over the sample surface, have been achieved [47, 70], but both sensitivity and mass range are seriously compromised. Higher resolutions, down to a pixel size of 500 nm or less and a spatial resolution of 4 mm on rat pituitary have been achieved using mass spectrometers in the microscope mode, where the ion optics are set up to retain the spatial distribution within the laser spot area, and then spread out onto an area detector using ion optics in a similar way as conventional microscopy to achieve high resolution [4, 71], although it is important to emphasize that this is limited to detection of a single mass in any one image. Cellular resolution (10 mm) MSI across the usual mass range (up to 30 kda) and with reasonable sensitivity has very recently been reported, and this was, as discussed in Section 3.3, achieved using a coaxial laser beam and in situ focussing optics based around the source of Spengler [70]. These systems are still in the development phase and not generally available, but they demonstrate that the limitations of this method now seem to be based around the inability to deposit matrix with the required homogeneity and small crystal size across the sample while still extracting sufficient analyte. The repetition rate of the laser is another key feature. To put this into context, a coronal section of mouse brain is well over 1 cm 2, imaged at 50 mm centre-to-centre resolution this would require raster points, and between 100 and 1000 laser shots will need to be collected at each point. With a standard 20 Hz N 2 laser, the time taken to collect the data alone would be over 55 h, and when the time for movement of the stage, data transfer and storage is taken into account, the overall data acquisition time becomes impractical. The introduction of Nb:YAG (niobium yttrium aluminium garnet) lasers with repetition rates of Hz on commercial instruments, has made MALDI-MSI on tissue sections a practical proposition, and laser rates up to 2 khz have been reported in the literature [72]. Even so, the parasagittal section of mouse brain imaged at a combination of 75 and 100 mm resolution raster with a 200 Hz laser shown in Fig. 4 took around 20 h to collect, and this experiment generated a 35 Gb data file. Figure 4 shows that little information is lost in moving from a75to100mm raster, and that both give images of high quality. However, the 100 mm raster approximately halves the time for data acquisition and the size of the data file. Clearly, the best balance between the resolution required and the time and data file sizes needs to be carefully considered before embarking on data collection. 4.2 Source design There are two main types of MALDI source, one that operates at close to atmospheric pressure and the other that operates at high vacuum. Atmospheric pressure sources have been used for MSI [73, 74], but have no great advantage over vacuum sources and have had relatively limited uptake in the community. Improvements in the source geometries of commercial MALDI instruments have lead to substantial gains in sensitivity, for instance by removing the grids that were typical of previous generation of MALDI sources, and having the laser hit the target as coaxially with the ion extraction optics as possible. This results in the ion cloud generated by the laser ablation of the surface being directed along the axis of extraction from the source, maximizing ion capture and increasing the length of time that the source performance remains optimal. In MSI applications, it is not uncommon for a single experiment to use over laser shots in under 24 h, ablating high microgram quantities of matrix into the source. Clearly this puts much more strain on the system than normal use, and even with modern sources, regular cleaning of the source is necessary. This represents a significant issue for reproducible data acquisition. 4.3 Mass analyser technologies As noted before, MALDI-MSI has been coupled to a wide range of different mass analysers, including TOF, TOF TOF, QqTOF, trap (both linear and spherical ITs) and Fourier transform ICR (FT-ICR). Each of these has their own merits, some of the more important of which are discussed here. TOF is a robust methodology that covers a broad mass range and is ideally suited to MALDI ion generation. However, it requires that the sample is relatively thin and flat. In addition, because of the steep field gradients in the source that are needed for ion acceleration for TOF measurement, the sample must be mounted on an electrically conductive surface to ensure that the electric field across the sample is as even as possible, and additionally to prevent charge build-up on the surface that would distort the field. Therefore, the limitations imposed by the use of tissue samples means that mass resolution often falls well short of the capabilities of the instrument, resulting in limited discrimination between species close in mass. In modern instruments TOF is generally run in two different modes depending on the m/z range of the ions being detected. Instruments are generally run in linear mode for imaging so that ions across the whole mass range can be seen in a single experiment. This is the best mode for ions in the kda range. In theory, the upper range using MALDI-TOF is infinite, although sensitivity and resolution drop off rapidly with increasing mass above 20 kda, and while signals from very abundant proteins in the kda range have been observed from tissue, most imag-