Serial 3D Imaging Mass Spectrometry at Its Tipping Point - Analytical

Mar 28, 2015 - Essential tools and protocols were developed, in particular to improve the reproducibility of sample preparation, speed up data acquisi...
0 downloads 16 Views 2MB Size
Perspective pubs.acs.org/ac

Serial 3D Imaging Mass Spectrometry at Its Tipping Point Andrew D. Palmer†,‡ and Theodore Alexandrov*,†,‡,§,∥ †

European Molecular Biology Laboratory, Meyerhofstrasse 1, 69117 Heidelberg, Germany Center for Industrial Mathematics, University of Bremen, 28359 Bremen, Germany § SCiLS GmbH, 28359 Bremen, Germany ∥ Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California, San Diego, San Diego, California 92161, United States ‡

ABSTRACT: Since biology is by and large a 3-dimensional phenomenon, it is hardly surprising that 3D imaging has had a significant impact on many challenges in the life sciences. Imaging mass spectrometry (MS) is a spatially resolved labelfree analytical technique that recently maturated into a powerful tool for in situ localization of hundreds of molecular species. Serial 3D imaging MS reconstructs 3D molecular images from serial sections imaged with mass spectrometry. As such, it provides a novel 3D imaging modality inheriting the advantages of imaging MS. Serial 3D imaging MS has been steadily developing over the past decade, and many of the technical challenges have been met. Essential tools and protocols were developed, in particular to improve the reproducibility of sample preparation, speed up data acquisition, and enable computationally intensive analysis of the big data generated. As a result, experimental data is starting to emerge that takes advantage of the extra spatial dimension that 3D imaging MS offers. Most studies still focus on method development rather than on exploring specific biological problems. The future success of 3D imaging MS requires it to find its own niche alongside existing 3D imaging modalities through finding applications that benefit from 3D imaging and at the same time utilize the unique chemical sensitivity of imaging mass spectrometry. This perspective critically reviews the challenges encountered during the development of serial-sectioning 3D imaging MS and discusses the steps needed to tip it from being an academic curiosity into a tool of choice for answering biological and medical questions.

I

submicrometer resolution in all dimensions with subcellular precision.8 3D SIMS images have been collected from serial tissue sections (see Figure 2c) but this is less commonly encountered. Motivated by challenges encountered during the early experiments, novel tools and protocols for serial 3D imaging MS were developed during the past decade, in particular, for improved sample preparation, accelerated data acquisition, and computationally intensive analysis of the big data generated. At present, serial 3D imaging MS analysis is still a complex but feasible technique and is becoming more accessible as illustrated by a growing body of literature on this approach. This advanced and recently emerged technique is now at the tipping point of its development: its future depends on whether specific applications will be found where the benefits brought by this advanced technique will outweigh its complexity. Serial 3D imaging MS inherits advantages of 2D imaging MS. As compared to the established molecular 3D imaging techniques (such as 3D immunohistochemistry, magnetic resonance imaging with targeted contrast agents, magnetic resonance spectroscopy, positron emission tomography, and single-photon emission computed tomography), 3D imaging MS is label-free (does not require tracers or dyes), untargeted

maging mass spectrometry (imaging MS) is a spatially resolved mass spectrometry-based analytical technique for in situ molecular analysis. Localization of intact biomolecules, including metabolites, lipids, peptides, and proteins, in thin tissue sections has been a particularly high impact area of application.1 In the past decade, imaging of intact biomolecules has expanded to the 3D analysis to determine volumetric molecular distributions within tissue specimens,2,3 agar plates,4 and 3D cell cultures5 so newcomers to the field have access to a range of successful experimental protocols from a diverse range of sample types. While the technique is not yet mainstream the number of citations related to 3D imaging MS is rapidly increasing; see Figure 1. Currently, the most common approach to 3D imaging MS includes serial sectioning of a sample, analyzing each section in 2D with imaging MS and then reconstructing a final 3D imaging MS data set using computational methods.6 Serial sectioning-based 3D imaging MS (later in the manuscript referred to simply as 3D imaging MS) was introduced in 2005 for matrix-assisted laser desorption ionization (MALDI)-time of flight (TOF) imaging MS,2 and later its feasibility was shown under ambient conditions using desorption electrospray ionization (DESI).7 3D MALDI- and DESI-imaging MS allow the detection of intact biomolecules over a wide field of view. Secondary ion mass spectrometry (SIMS) uses a focused ion beam and provides much higher spatial resolution at a cost of a decreased field of view. In SIMS, 3D images are usually created by sputtering away the surface layer of a sample to provide © XXXX American Chemical Society

Received: December 11, 2014 Accepted: March 28, 2015

A

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry

Figure 1. Number of citations (a) from publications (b) related to 3D imaging MS returned by a ISI Web of Knowledge search as of September 2014 for topics: (3d “imaging mass spectrometry”) OR (three-dimensional “imaging mass spectrometry”) OR (three-dimensional “mass spectrometry imaging”) OR (3d “mass spectrometry imaging”); timespan, all years.

Figure 2. Examples of 3D images from serial section imaging MS studies: (a) Reprinted with permission from ref 7. Copyright 2010 John Wiley & Sons, Inc. (b) Reprinted with permission from ref 4. Copyright 2013 Nature Publishing Group. (c) Reprinted with permission from ref 16. Copyright 2012 Springer. (d) Reprinted from ref 25. Copyright 2012 American Chemical Society. (e) Show reconstructed ion volumes. Reprinted with permission from ref 29. Copyright 2012 Elsevier. (f) Shows 3D ions overlaid on a whole body MRI volume. Reprinted with permission from ref 30. Copyright 2012 Elsevier. (g) Reprinted with permission from ref 31. Copyright 2009 Springer. (h) Show single sections registered with 3D tissue volumes acquired with optical imaging. Reprinted with permission from ref 17 Copyright 2014 Elsevier. (i) Reprinted with permission from ref 27. Copyright 2011 Royal Society of Chemistry. (j) Use multivariate analysis and segmentation to divide the volume into discrete regions. Reprinted with permission from ref 3. Copyright 2013 Elsevier.

based research and for such specific applications as biomarker discovery, characterization of host−pathogen molecular interactions, discovery of novel drug targets, ADME (absorption, distribution, metabolism, and excretion) studies in drug development, and metabolite or protein profiling in preclinical

(can detect hundreds of molecules from different chemical classes in the same experiment), sensitive (femtomole to zeptomole limits of detection9), and specific (can discriminate molecules whose molecular weights differ by less than 0.01 Da). This positions 3D imaging MS as a perfect tool for discoveryB

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry

trometer as of the beginning of 2015, the spectra acquisition rate in imaging mode is 1−2 pixels per second. As an example, a cubic sample of 5 mm × 5 mm × 5 mm with 50 serial sections imaged with a 50 μm pixel size sums up to 70−140 h of instrument time even without considering the additional overhead of sample preparation. Adding the preparation steps including sectioning of a specimen into tens of serial sections, washing of sections, application of matrix to each serial section (when a MALDI source is used), and staining sections post image acquisition extends the realistic data acquisition for a single specimen to several weeks, if not longer. Serial 3D imaging MS sets high demands on the reproducibility of all steps making the acquisition of one 3D imaging MS data set an adventure and sets a high barrier for those who decide to enter the field. In some sense, 3D imaging MS is a frontier of the imaging MS field where many requirements are pushed to the limit.13 It creates the need for more robust protocols for sample preparation, sectioning, and imaging MS, as well as for faster mass spectrometry and efficient algorithms for analysis of big imaging MS data. Following the first 3D imaging MS experiments 10 years ago, many of these challenges have been addressed. Technology has been improved; protocols developed that are tailored to the needs of 3D imaging MS; and novel computational tools produced, including commercial products. In the next subsections, we are highlighting the developments which either turned out to be helpful in our research or, in our estimation, had a significant positive impact onto the field. Sample Preparation. Sample preparation for 3D imaging MS generally follows the protocols for 2D imaging MS but the following aspects should be considered: First, in the case of multimodal 3D imaging, when a nondestructive imaging such as MRI is followed by 3D imaging MS,3,34 one needs to ensure the chemical and morphological stability between the modalities. Second, 3D imaging MS sets high demands on the reproducibility of data acquired from tens of serial sections. Several recent developments in sample preparation techniques and protocols have either been applied for 3D imaging MS or have high potential to help in these experiments. Of note are prevention of post-mortem degradation by stabilizing tissue using heating,35 support of sectioning by embedding tissue specimens,3,34 use of fiducial markers for improving subsequent computational alignment,25 sectioning by using tape transfer systems,36 and novel protocols or instrumentation for matrix application.37 In this subsection, we review these techniques in more detail. Not much progress has been seen in improving sample preparation of agar samples and 3D cell cultures, maybe because these techniques are less mainstream as compared to tissue imaging. Tissue Stabilization. Tissue stabilization using heating is a promising technology for preventing post-mortem molecular degradation of tissue proteins, peptides, and metabolites without introducing fixative agents or solvents.35 Preventing molecular degradation is particularly important for 3D imaging MS where tissue specimens or tissue sections are held at room temperature longer than for a single-section analysis either during multimodal imaging or during steps of acquisition, transfer, and analysis of a large number of sections. However, using heating for tissue stabilization led to mixed reports. Our experience has been that it negatively affects the tissue sectioning process, probably due to tissue drying, which makes it hard to get high-quality consecutive sections.36 Micro cracks occur over the whole section and often the

and clinical studies. The most similar technique in terms of resolution and sample handling is the reconstruction of 3D volumes from histologically stained tissue sections that on the contrary provides only morphological information or localization of few proteins only when combined with immunohistochemistry. Serial histological imaging, although challenging, proved to be useful for creation of 3D atlases such as the volumetric gene expression contained within the Allen Brain Atlas, which has received over five million cumulative visits as of 2015.10 Serial sectioning 3D imaging MS data contains distributions of hundreds if not thousands of molecular species, immediately exceeding the scope of what is achievable by applying many stains to serial tissue sections. 3D imaging MS avoids the difficulty that has been noted in aligning (or registering) stains that show different features11 as MS collects all the molecular signals from the same tissue section.12 This perspective paper on the serial 3D imaging MS starts with a review of the challenges of the past and how they were addressed, followed by a summary of the current challenges. Then we project our expectations into the future by discussing the emerging technologies in this field and potential applications and conclude with future perspectives.



CHALLENGES OF THE PAST Since 2005, the potential of serial 3D imaging MS was illustrated in multiple studies and has been highlighted in several review articles.13,14 3D imaging MS has been applied for analysis of mammalian tissues, mouse brain,2 rat brain,15 mouse kindey,3 rat heart,16 whole mouse;17 invertebrates, crustacean central nervous system;18 plant tissue, contaminants in lemon peel,19 nutrients in a wheat grain;20 microbial colonies grown in agar;4,21 and cell cultures.22−24 The potential of 3D imaging MS for biomedical applications has also been demonstrated on a diverse types of tissue, including breast tumor xenograft,25 human ovaries,26 mouse mammary tumor,27 as well as drug imaging along a coronary stent,28 see Figure 2 for examples. However, despite the growing number of publications on 3D imaging MS, using this technique still comes with its own challenges. It is a complex multistep technique requiring highlevel expertise in different areas such as sample preparation, imaging MS, and bioinformatics. Specialist skills, software, and algorithms are also required for spatial alignment of individual imaging MS data sets into a 3D volume data and for processing the large data sets which can exceed 100 GB in size. 3D imaging MS is time-consuming. In our experience, with exploiting the latest-generation commercially available imaging mass spectrometer, the data acquisition is the slowest step accounting for over 60% of the overall time necessary to produce a 3D imaging MS data set for a sample. The time necessary for the data acquisition depends on many factors. The most influential factor is the MS analyzer: the sampling rate of advanced TOF-analyzers reach up to 10−50 Hz,32 whereas the sampling rate of the Orbitrap and Fourier transform ion cyclotron resonance (FTICR) analyzers is limited by 0.3−1 Hz dependent on the resolution required.33 Other mass spectrometry characteristics influencing the acquisition time are performance of the ion source with and performance of the hardware stage controlling the spatial positioning of the sample. Other important factors are the sample size and the required spatial resolution because decreasing the pixel size quadratically increases the number of pixels and thus the acquisition time. As a rule of thumb for a current-generation commercially available imaging mass specC

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry sections break apart. However, recent reports38 have reported positive experience in using the Stabilizor T1 heating device (Denator, Sweden), so it seems that more studies are necessary to evaluate this approach for 3D imaging MS. Tissue Embedding. Surrounding a tissue block with an embedding medium serves to provide protection and support during handling and processing. This can extend the lifetime of fresh tissue during storage13 and enable high-quality sections to be reproducibly collected. Several media for embedding have been evaluated including water,34 carboxy-methylcellulose,2 gelatin,18 and PAXgene.3,13 On one hand, water is the most accessible medium and generates no unwanted MS signals, on the other hand, other media may better support sectioning even if they can produce interference in the resulting spectra. Additionally samples embedded in water cannot be stored at room temperature. PAXgene is a special medium developed and evaluated in the EU FP6 COMPUTIS project which simultaneously fixes and embeds tissue without cross-linking proteins. We proved it to be compatible with 3D MALDI imaging MS of proteins and MRI3 making it a great choice for multimodal protein imaging. Fiducial Markers. Data produced from serial sections need to be aligned (or “registered” as it is called in image analysis) to reconstruct original spatial relations between sections. The most common strategy is to align the imaging MS data with optical tissue images and then align the set of optical images based on similar features between sections.15 As these can vary between modalities and tissue sections, artificial fiducial markers can be included within the embedding medium that are identifiable and provide the basis for alignment. This can be in particular helpful to estimate the level and compensate for shrinkage of sections during drying, to evaluate the registration methods by providing ground truth, and to solve the so-called “banana problem” when performing registration. The banana problem refers to the artificial straightening that occurs during registration of a series of slices from a banana without additional information on the curvature.39 Including these markers within the embedding media25 could replace the use of optical images for registration or can compliment it by providing additional information. Sectioning Using Tape Transfer Systems. Use of tape transfer systems (such as Leica CryoJane) is an established approach to improving the quality of sections, in particular, for full-body sectioning in the pharmaceutical industry.40,41 We experimented with using it for sectioning of a mouse brain and mouse kidney and did not notice significant improvement in sectioning, maybe because both tissue types are relatively easy to section. Overall, so far sectioning using tape transfer systems found only limited applications for imaging MS and 3D imaging MS. We hypothesize that this is due to relative complexity of this technique and due to the widespread use of relatively simple tissue specimens where using this system does not bring a significant advancement. However, it has been reported be helpful particularly for large or fragile tissue specimens and fullbody analysis.42 Matrix Application. Development of novel protocols and techniques for matrix application is an active field of research and technology in imaging MS. (Here we do not cover development of novel matrixes where many reports have been published in the recent years, such as van Hove et al.43). Several companies provide commercial solutions for matrix application including ImagePrep (Bruker Daltonik, Germany), SunCollect (SunChrom, Germany), ChIP 1000 (Shimadzu, Japan), and

TM-Sprayer (HTX Technologies). There also exists an open source project from whom it is possible to purchase a spraying robot kit (iMatrixSpray, Tardo, Switzerland).44 In the traditional approach, matrix dissolved in an organic solvent is deposited on the tissue where the solvent also aids in extraction of endogenous metabolites which are then cocrystallized with the matrix. As the molecules are mobile within the solvent, care must be taken not to overwet the sample and dislocate molecules of interest. Protocols for “dry” matrix deposition through sublimation37 or sieving45 have also been developed to avoid the issue of overwetting. Automated methods have reduced the operator involvement in preparing large numbers of samples but are far from being unsupervised so still require substantial operator time. For analysis of serial sections, the ongoing challenge is to ensure high-quality and consistency across tens of sections. Data Acquisition. For acquisition of data from serial sections for 3D imaging MS, conventional imaging mass spectrometers are used. The main challenge is that analysis of one specimen requires acquisition of tens of imaging MS data sets which can take several weeks. The measurements are later combined into one data set so a high requirement is set on the stability of measurements over time. In the past decade, mass spectrometry experienced significant evolution that was minimally motivated by challenges of 3D imaging MS. This involved development and improvement of sources, analyzers, sensitivity, resolving power, mass range, etc., but for 3D imaging MS probably the most significant improvement was the increase of the acquisition speed. Data Acquisition Speed. The mass spectrometers used for the first 3D imaging MS experiments (e.g., Voyager DE-STR, Applied Biosystems2) were equipped with a 20 Hz nitrogen laser and a rather slow positioning stage meaning that each spectrum took 10−30 s to generate, depending on the number of laser shots averaged per spectrum. Whereas, in 2015 most imaging mass spectrometers have acquisition rates of 1−2 pixels per second. This equates to an approximately 10-fold speed up in data acquisition. An experiment with a hundred sections would take a year in 2005 and now takes a few weeks that shifts it from the category of nearly impossible and certainly impractical to complex but feasible. Reducing this by another order of magnitude, assuming stability of measurement sensitivity could be maintained, could shift the time from weeks to days. Software. 3D imaging MS requires computational methods in two key areas: first, for registration of individual data sets into a 3D volume data and second, for analysis, interpretation, and visualization of the data produced. A common approach is to extract the intensities for a particular mass-to-charge (m/z) and employ a general 3D visualization software (e.g., 3D Doctor,7 ImageJ,2 MATLAB,4 Slicer,46 or Paraview47) where registration and then visualization are performed. The advantages of this approach are no need for specialized 3D imaging MS software (especially important at the early stages of 3D imaging MS development), flexibility (analysis and visualization can be adjusted specifically to particular needs), free or low costs (when using open-source software such as ImageJ or Paraview), and the use of advanced visualization already provided by powerful 3D visualization toolboxes. Some examples of the results for individual ion images are illustrated in Figure 2a−d. The main disadvantages are the need for a bioinformatician or software developer to program import of imaging MS data into a general visualization framework not D

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry adapted to imaging MS and finally the lack of interactive analysis that is often requested in discovery-based studies. The second approach is to use a software package oriented at 3D imaging MS. Currently, only one software package dedicated to 3D imaging MS is available, namely, SCiLS Lab 3D (SCiLS GmbH, Germany) that includes tools for registration, 3D visualization, and data analysis.

Computational analysis and interpretation of imaging MS data is an ongoing challenge, but various approaches for identifying distinct tissue compartments have been produced for 2D image analysis (e.g. refs 53−55) and these are being transferred into the 3D domain.6,46 The best method for representing these results in 3D for interpretation and analysis, particularly in publications, remains an open question. As the field moves beyond single sample analysis and into comparative studies, a specific combination of expertise in MS, image analysis, and bioinformatics will be required to extract meaningful information.



CURRENT CHALLENGES AND A WISH LIST After overcoming the major challenges that arose in the first years of its development, the proof of principle studies performed in the past decade demonstrated the feasibility of 3D imaging MS. From an only one-lab-can-do-it technique it became a complicated-but-achievable task. There is still a number of challenges associated with performing serial 3D imaging MS, and in this section we discuss them. Collection of individual MS images is becoming routine so many of the experimental challenges revolve around scaling the approach to larger sets. Producing 3D imaging MS data requires the handling of large numbers of tissue sections. Performing this task manually increases the risk of sample-to-sample variability and restricts the scaling. Works has been started to develop automated loaders and preparation devices to reduce the number of manual stages in the preparation workflow and ensure consistent handling.48 If robust autosamplers for imaging MS were standardized and compatible with the matrix application robots, then the most labor intensive portions of the process could be performed automatically. By and large these technological advances go hand-in-hand with large scale 2D imaging MS studies, so it is likely that the natural progression of the technique will also provide benefit for the 3D imaging MS. The key challenge for 3D imaging MS is the accurate reconstruction of the spatial coordinates of each pixel. Future studies will need to move beyond manual alignment or heavily curated approaches, as these methods cannot scale sufficiently. Improvements of the sampling procedures will reduce the distortions introduced and so make the task of aligning images simpler. Another alignment challenge is that of aligning spectra between images as small fluctuations in calibration can significantly affect the final data rendering (particularly in the case of FT based MS). The spread of the 3D imaging MS technique and the increase of the number of users will probably require continued development of novel software packages. The importance of the software will probably only grow, and in particular the scalable software solutions will be required to handle volumes of data greater than a few gigabytes. Large projects such as 3D imaging typically require the involvement of several partners and sharing the data produced is not trivial. Efforts such as imzML49 have generated an accessible open format that is slowly being adopted by manufacturers and the community. However, imzML does not provide an efficient route for compressing and transmitting data. A data format specifically designed for processing 3D data sets has been developed50 and showed potential for visualization of ion volumes. Strategies that take advantage of the spectral redundancy within 2D and 3D imaging data sets to compress the spectra while preserving the spatial and spectral resolution have been presented but has not yet been standardized.51 A recent effort has provided a set of 3D imaging MS data sets in multiple formats as an open resource for improving access to the field and benchmarking the performance of newly developed algorithms.52



EMERGING TECHNOLOGIES AND FUTURE PERSPECTIVES The field of 3D imaging MS is young and dynamic. Recently, a number of technologies were proposed that can have impact onto the field of 3D imaging MS by speeding up the collection of data, increasing the knowledge generated by an experiment and expanding its areas of application. In the following, we discuss those emerging or future technologies which we believe can transform 3D imaging MS by opening novel niches of applications and by building the capacity to address biomedical problems. 2.5D and 3D Surface Imaging. Data fusion is a process of merging the information collected from the same sample with different modalities providing a more holistic picture of a single sample.17 For example, 3D anatomical or morphological views of intact samples can be generated by mature 3D imaging technologies, such as MRI or micro-CT. After nondestructive 3D imaging of a specimen is performed, a single tissue section is collected with its location within the tissue block recorded and imaged with imaging MS, then these two modalities can be overlaid to link morphological and molecular information as illustrated in Figure 3b. Examples of this from a barley grain31

Figure 3. Illustration of the emerging modes of 3D imaging acquisition using surface imaging (a) serial imaging MS (the main focus of this paper), (b) 2.5D multimodal chemical overlays where a single section imaged with MS is localized within the tissue volume, (c) imaging over nonflat surfaces with 3D coordinates preserved for each data point, and (d) ablation of sample to reveal the next sampling site.

and a bone tumor17 are shown in Figure 2. Performing this with a single section avoids the complexities of multimodality registration and so can be called a 2.5D molecular view. Preserving the original 3D coordinates of the sections would allow a bridge between the 2D and 3D worlds to be made. An alternate approach to melding the 2D and 3D worlds is to acquire samples over a 3D surface of the intact specimen while recoding the coordinates of collection sites followed up by mass spectrometry analysis of each sample individually. Mapping the mass spectra intensities back onto a 3D model of the surface allows one to visualize spatial distribution of detected molecules across the surface as illustrated in Figure 3c. Although this general approach can be applied to a large variety of samples on various spatial scales, in our preliminary experiments we found it to be the most useful for analysis of biological surfaces of the E

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry

individual pixels but from groups of randomly selected pixels.65 As our simulations show, using compressed sensing for imaging MS would speed up the data acquisition up to 1 order of magnitude. However, it requires substantial new hardware. For a laser-based desorption technique, each laser shot should be directed to several randomly selected pixels merging the signals from the ions produced for these pixels into one mass spectrum. Visualization and Interaction. Extracting information from 3D imaging MS data remains a conceptual and technical challenge. Recent revolutions in immersive 3D technology such as digitally enabled eyewear (such as Google Glass, Microsoft Hololens, or Oculus Rift) present opportunities for augmented reality approaches or virtual handling of the data into the lab so that researchers can interact and manipulate their data in virtual reality. These technologies do not make the data compatible with traditional publication through journal articles. The increased access of journals through online portals, rather than traditional print, offers opportunities to take advantage of new media standards for embedding interactive content such as 3D models so the reader can interact with the data to better understand the biomedical results that are being presented.

scale of centimeters where various extraction techniques could be applied.56 Both of these approaches sacrifice complete volumetric imaging for a practical compromise which provides 3D molecular fusion. Ablation Mode. Ablation of samples to reveal fresh material for imaging has a long history in SIMS where dynamic mode has been used to delve into sample volumes (see Fletcher et al.57). The application of rapidly developing portable MS for 3D imaging MS can create opportunities for novel applications. Integration of an iKnife58 or another intraoperative MS with a position tracking system would allow one not only to detect tumor cells directly during surgical resection but also provide molecular data for 3D spatial profiling. The downside of this technique is that the sample is destroyed during the collection but the data could be merged or overlaid with 3D images collected prior to ablation to improve spatial registration, for example, MRI or CT. Block-Face Imaging MS. Serial block-face microscopy is an established approach to perform 3D imaging by iterating serial sectioning with light or electron microscopy imaging of the block face surface of the sectioned specimen.59 Integration of imaging MS into a cryostat or with a microtome would pave the way to block-face imaging MS. This could be achieved by using atmospheric imaging MS requiring no sample preparation such as DESI. Similar to serial block-face microscopy, this would allow one to achieve perfect registration quality by preventing deformations of individual tissue sections and to make the acquisition process automatic or semiautomatic by avoiding the phase of mounting tissue sections onto the glass slides. Serial block-face imaging MS would be challenging for frozen tissues, because it would require constructing a specialized cryostat with integrating the imaging MS source and adapting mass spectrometry to the analysis of frozen tissue. However, this could find its application in analysis of embedded tissue at room temperature. Ultrafast Imaging Mass Spectrometry. The field of mass spectrometry will certainly produce novel developments in the coming years that will be advantageous in 3D imaging MS. In 3D imaging MS, we are especially interested in developments reducing the acquisition time. When using modern MS with fast analyzers (such as TOF or quadrupole time-of-flight (QTOF) technologies) and high-frequency lasers of 2−5 kHz, the pixel-by-pixel way of data acquisition requiring highacceleration high-precision positioning stage becomes a bottleneck. Advanced high-performance imaging MS was reported that employs 5−40 kHz lasers providing an acquisition speed of approximately 20 pixels per second.60 At these speeds, data acquisition from single sections of sagittally sectioned mouse brain took 10−20 min with 100 μm spatial resolution.32,61 However, as each pixel is still acquired sequentially, a restriction remains. An alternative to the pixel-by-pixel microprobe acquisition approach is the microscope mode where ions are desorbed from an area in one shot and are detected with a spatially resolved detector. The success of Ron Heeren’s team in developing this technique has recently been confirmed by reports of other groups using this approach independently.62,63 The microscope imaging MS could revolutionize the field by accelerating data acquisition by 2 orders of magnitude. Compressed Sensing. The idea of applying the compressed sensing approach to imaging MS has existed for only a short time. Motivated by the proof-of-principle demonstration of a “single pixel” photo camera,64 it has recently been formulated for imaging MS as acquisition of spectra not from



CONCLUSION 3D imaging MS has been demonstrated to be a capable tool for exploring molecular distributions throughout sample volumes and impressive efforts have been made in showcasing the technology. We believe the technical challenges of throughput and reproducibility are being solved in tandem with development of traditional 2D imaging MS. However, 3D imaging MS is still a technique looking for a niche in which to truly excel. We hope that this review will help to increase the knowledge of this technique within the analytical community and accelerate the update of this molecular imaging tool.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare the following competing financial interest(s): Theodore Alexandrov is the scientific director of SCiLS GmbH which develops and markets software for imaging mass spectrometry.



ACKNOWLEDGMENTS We thank the reviewers for their detailed comments, the inclusion of which has strengthened the article. The authors have received funding from the 3D-MASSOMICS project (European Union Seventh Framework Programme Grant No. 305259).



REFERENCES

(1) Norris, J. L.; Caprioli, R. M. Chem. Rev. 2013, 113, 2309−2342. (2) Crecelius, A. C.; Cornett, D. S.; Caprioli, R. M.; Williams, B.; Dawant, B. M.; Bodenheimer, B. J. Am. Soc. Mass Spectrom. 2005, 16, 1093−1099. (3) Oetjen, J.; et al. J. Proteomics 2013, 90, 52−60. (4) Watrous, J. D.; Phelan, V. V.; Hsu, C.-C.; Moree, W. J.; Duggan, B. M.; Alexandrov, T.; Dorrestein, P. C. ISME J. 2013, 7, 770−780. (5) Li, H.; Hummon, A. B. Anal. Chem. 2011, 83, 8794−801. (6) Trede, D.; et al. Anal. Chem. 2012, 84, 6079−6087. (7) Eberlin, L. S.; Ifa, D. R.; Wu, C.; Cooks, R. G. Angew. Chem., Int. Ed. 2010, 49, 873−876. F

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry (8) Nygren, H. K.; Malmberg, P. Trends Biotechnol. 2007, 25, 499− 504. (9) Rubakhin, S. S.; Jurchen, J. C.; Monroe, E. B.; Sweedler, J. V. Drug Discovery Today 2005, 10, 823−37. (10) Allen Institute Website, http://www.alleninstitute.org/newsevents/news/articles/2014-year-review. (11) Song, Y.; Treanor, D.; Bulpitt, A. J.; Magee, D. R. J. Pathol. Inform. 2013, 4, S7. (12) Abdelmoula, W. M.; Skrásǩ ová, K.; Balluff, B.; Carreira, R. J.; Tolner, E. A.; Lelieveldt, B. F. P.; van der Maaten, L.; Morreau, H.; van den Maagdenberg, A. M. J. M.; Heeren, R. M. A.; McDonnell, L. A.; Dijkstra, J. Anal. Chem. 2014, 86, 9204−9211. (13) Seeley, E. H.; Caprioli, R. M. Anal. Chem. 2012, 84, 2105−2110. (14) Weaver, E. M.; Hummon, A. B. Adv. Drug Delivery Rev. 2013, 65, 1039−55. (15) Andersson, M.; Groseclose, M. R.; Deutch, A. Y.; Caprioli, R. M. Nat. Methods 2008, 5, 101−108. (16) Fornai, L.; Angelini, A.; Klinkert, I.; Giskes, F.; Kiss, A.; Eijkel, G.; Amstalden-van Hove, E. A.; Klerk, L. A.; Fedrigo, M.; Pieraccini, G.; Moneti, G.; Valente, M.; Thiene, G.; Heeren, R. M. A. Anal. Bioanal. Chem. 2012, 404, 2927−2938. (17) Seeley, E. H.; Wilson, K. J.; Yankeelov, T. E.; Johnson, R. W.; Gore, J. C.; Caprioli, R. M.; Matrisian, L. M.; Sterling, J. A. Bone 2014, 61, 208−216. (18) Chen, R.; Hui, L.; Sturm, R. M.; Li, L. J. Am. Soc. Mass Spectrom. 2009, 20, 1068−1077. (19) Nielen, M. W. F.; van Beek, T. A. Anal. Bioanal. Chem. 2014, 406, 6805−6815. (20) Wu, B.; Andersch, F.; Weschke, W.; Weber, H.; Becker, J. S. Metallomics 2013, 5, 1276−1284. (21) Traxler, M. F.; Watrous, J. D.; Alexandrov, T.; Dorrestein, P. C.; Kolter, R. mBio 2013, 4, No. e00459-13. (22) Rabbani, S.; Fletcher, J. S.; Lockyer, N. P.; Vickerman, J. C. Surf. Interface Anal. 2011, 43, 380−384. (23) Keithley, R. B.; Weaver, E. M.; Rosado, A. M.; Metzinger, M. P.; Hummon, A. B.; Dovichi, N. J. Anal. Chem. 2013, 85, 8910−8918. (24) Ahlf, D. R.; Masyuko, R. N.; Hummon, A. B.; Bohn, P. W. Analyst 2014, 139, 1−8. (25) Chughtai, K.; Jiang, L.; Greenwood, T. R.; Klinkert, I.; Van Hove, E. R. A.; Heeren, R. M. A.; Glunde, K. Anal. Chem. 2012, 84, 1817−1823. (26) Chansela, P.; Goto-Inoue, N.; Zaima, N.; Hayasaka, T.; Sroyraya, M.; Kornthong, N.; Engsusophon, A.; Tamtin, M.; Chaisri, C.; Sobhon, P.; Setou, M. PLoS One 2012, 7, 1−11. (27) Reindl, W.; Bowen, B. P.; Balamotis, M. A.; Green, J. E.; Northen, T. R. Integr. Biol. 2011, 3, 460−467. (28) Fisher, G. L.; Belu, A. M.; Mahoney, C. M.; Wormuth, K.; Sanada, N. Anal. Chem. 2009, 81, 9930−9940. (29) Jones, E. A.; Shyti, R.; van Zeijl, R. J. M.; van Heiningen, S. H.; Ferrari, M. D.; Deelder, A. M.; Tolner, E. A.; van den Maagdenberg, A. M. J. M.; McDonnell, L. A. J. Proteomics 2012, 75, 5027−5035. (30) Attia, A. S.; Schroeder, K. A.; Seeley, E. H.; Wilson, K. J.; Hammer, N. D.; Colvin, D. C.; Manier, M. L.; Nicklay, J. J.; Rose, K. L.; Gore, J. C.; Caprioli, R. M.; Skaar, E. P. Cell Host Microbe 2012, 11, 664−673. (31) Bollenbeck, F.; Kaspar, S.; Mock, H. P.; Weier, D.; Seiffert, U. Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics) 2009, 5462 LNBI, 126−138. (32) Spraggins, J. M.; Caprioli, R. M. J. Am. Soc. Mass Spectrom. 2011, 22, 1022−1031. (33) Makarov, A.; Denisov, E.; Kholomeev, A.; Balschun, W.; Lange, O.; Strupat, K.; Horning, S. Anal. Chem. 2006, 78, 2113−2120. (34) Sinha, T. K.; Khatib-shahidi, S.; Yankeelov, T. E.; Mapara, K.; Ehtesham, M.; Cornett, D. S.; Dawant, B. M.; Caprioli, R. M.; Gore, J. C. Methods 2009, 5, 57−59. (35) Goodwin, R. J. A.; Lang, A. M.; Allingham, H.; Borén, M.; Pitt, A. R. Proteomics 2010, 10, 1751−1761.

(36) Goodwin, R. J. A.; Nilsson, A.; Borg, D.; Langridge-Smith, P. R. R.; Harrison, D. J.; Mackay, C. L.; Iverson, S. L.; Andrén, P. E. J. Proteomics 2012, 75, 4912−4920. (37) Hankin, J. a.; Barkley, R. M.; Murphy, R. C. J. Am. Soc. Mass Spectrom. 2007, 18, 1646−1652. (38) Cazares, L. H.; Vantongeren, S.; Kenny, T.; Lane, D.; Panchal, R.; Bavari, S. Heat Fixation Inactivates Viral and Bacterial Pathogens and is Compatible with Downstream MALDI Mass Spectrometry Tissue Imaging. 62nd ASMS Conference on Mass Spectrometry and Allied Topics, Baltimore, MD, June 15−19, 2014; poster WP010. (39) Malandain, G.; Bardinet, E.; Nelissen, K.; Vanduffel, W. Neuroimage 2004, 23, 111−127. (40) Khatib-Shahidi, S.; Andersson, M.; Herman, J. L.; Gillespie, T. A.; Caprioli, R. M. Anal. Chem. 2006, 78, 6448−6456. (41) Goodwin, R. J. A. J. Proteomics 2012, 75, 4893−4911. (42) Goodwin, R. J. A.; Nilsson, A.; Borg, D.; Langridge-Smith, P. R. R.; Harrison, D. J.; Mackay, C. L.; Iverson, S. L.; Andrén, P. E. J. Proteomics 2012, 75, 4912−4920. (43) Amstalden van Hove, E. R.; Smith, D. F.; Heeren, R. M. J. Chromatogr. A 2010, 1217, 3946−3954. (44) Stoeckli, M.; Staab, D.; Wetzel, M.; Brechbuehl, M. Chimia 2014, 68, 146−149. (45) Cristoni, S.; Molin, L.; Crotti, S.; Bernardi, L. R.; Seraglia, R.; Traldi, P. J. Mass Spectrom. 2009, 44, 1579−1586. (46) Eberlin, L. S.; Norton, I.; Orringer, D.; Dunn, I. F.; Liu, X.; Ide, J. L.; Jarmusch, A. K.; Ligon, K. L.; Jolesz, F. A.; Golby, A. J.; Santagata, S.; Agar, N. Y. R.; Cooks, R. G. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, 1611−1616. (47) Palmer, A.; Oetjen, J.; Becker, M.; Trede, D.; Heldmann, S.; Berger, J.; Strehlow, J.; Thiele, H.; Ernst, G.; Guntinas-Lichius, O.; Maass, P.; von Eggeling, F.; Alexandrov, T. Molecular Analysis in 3D Using Imaging Mass Spectrometry and Randomised Methods, A presentation at the MSACL conference, San Diego, CA, March 1−5, 2014, https://www.msacl.org/2014_program/, accesssed April 7, 2015. (48) McDonnell, L. A.; van Remoortere, A.; van Zeijl, R. J. M.; Dalebout, H.; Bladergroen, M. R.; Deelder, A. M. J. Proteomics 2010, 73, 1279−82. (49) Schramm, T.; Hester, A.; Klinkert, I.; Both, J. P.; Heeren, R. M. a.; Brunelle, A.; Laprévote, O.; Desbenoit, N.; Robbe, M. F.; Stoeckli, M.; Spengler, B.; Römpp, A. J. Proteomics 2012, 75, 5106−5110. (50) Klinkert, I.; Chughtai, K.; Ellis, S. R.; Heeren, R. M. a. Int. J. Mass Spectrom. 2014, 362, 40−47. (51) Palmer, A. D.; Bunch, J.; Styles, I. B. Anal. Chem. 2013, 85, 5078−5086. (52) Oetjen, J. et al. GigaScience Database, 2015, http://www. gigasciencejournal.com/, DOI: 10.5524/100131, in press. (53) Alexandrov, T.; Becker, M.; Deininger, S.-O.; Ernst, G.; Wehder, L.; Grasmair, M.; von Eggeling, F.; Thiele, H.; Maass, P. J. Proteome Res. 2010, 9, 6535−6546. (54) Fonville, J. M.; Carter, C. L.; Pizarro, L.; Steven, R. T.; Palmer, A. D.; Griffiths, R. L.; Lalor, P. F.; Lindon, J. C.; Nicholson, J. K.; Holmes, E.; Bunch, J. Anal. Chem. 2013, 85, 1415−1423. (55) Jones, E. A.; van Remoortere, A.; van Zeijl, R. J. M.; Hogendoorn, P. C. W.; Bovée, J. J. V. M G.; Deelder, A. M.; McDonnell, L. A. PLoS One 2011, 6, e24913. (56) Bouslimani, A.; Porto, C.; Rath, C. M.; Wang, M.; Guo, Y.; Gonzalez, A.; Berg-Lyon, D.; Ackermann, G.; Moeller Christensen, G. J.; Nakatsuji, T.; Zhang, L.; Borkowski, A. W.; Meehan, M. J.; Dorrestein, K.; Gallo, R. L.; Bandeira, N.; Knight, R.; Alexandrov, T.; Dorrestein, P. C. Proc. Natl. Acad. Sci. U.S.A. 2015, DOI: 10.1073/ pnas.1424409112. (57) Fletcher, J. S.; Vickerman, J. C. Anal. Chem. 2013, 85, 610−639. (58) Balog, J.; Sasi-Szabó, L.; Kinross, J.; Lewis, M. R.; Muirhead, L. J.; Veselkov, K.; Mirnezami, R.; Dezso, B.; Damjanovich, L.; Darzi, A.; Nicholson, J. K.; Takats, Z. Sci. Transl. Med. 2013, 5, 194ra93. (59) Denk, W., Horstmann, H. PLoS Biol. 20042. G

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX

Perspective

Analytical Chemistry (60) Trim, P. J.; Djidja, M. C.; Atkinson, S. J.; Oakes, K.; Cole, L. M.; Anderson, D. M. G.; Hart, P. J.; Francese, S.; Clench, M. R. Anal. Bioanal. Chem. 2010, 397, 3409−3419. (61) Steven, R. T.; Race, A. M.; Bunch, J. J. Am. Soc. Mass Spectrom. 2013, 24, 1−4. (62) Jungmann, J. H.; MacAleese, L.; Buijs, R.; Giskes, F.; de Snaijer, A.; Visser, J.; Visschers, J.; Vrakking, M. J. J.; Heeren, R. M. a. J. Am. Soc. Mass Spectrom. 2010, 21, 2023−2030. (63) Bamberger, C.; Renz, U.; Bamberger, A. J. Am. Soc. Mass Spectrom. 2011, 22, 1079−1087. (64) Duarte, M.; Davenport, M.; Takhar, D.; Laska, J.; Sun, T. S. T.; Kelly, K.; Baraniuk, R. IEEE Signal Process. Mag. 2008, 25, 83−91. (65) Bartels, A.; Dülk, P.; Trede, D.; Alexandrov, T.; Maaß, P. Inverse Probl. 2013, 29, 125015.

H

DOI: 10.1021/ac504604g Anal. Chem. XXXX, XXX, XXX−XXX