Correlative mass spectrometry imaging, applying TOF-SIMS and AP-MALDI to a single tissue section.

RATIONALE: Mass spectrometry imaging (MSI) is a powerful tool for mapping the surface of a sample. Time-of-Flight Secondary Ion Mass Spectrometry (TOF-SIMS) and Atmospheric Pressure Matrix-Assisted Laser Desorption/Ionization (AP-MALDI) offer complementary capabilities. Here, we present a workflow to apply both techniques to a single tissue section and combine the resulting data on the example of human colon cancer tissue. METHODS: Following cryo-sectioning, images were acquired using the high spatial resolution (1 µm pixel size) provided by TOF-SIMS. The same section was then coated with a para-nitroaniline matrix and images were acquired using AP-MALDI coupled to an Orbitrap mass spectrometer, offering high mass resolution, high mass accuracy and MS/MS capabilities. Datasets provided by both mass spectrometers were converted into the open and vendor-independent imzML file format and processed with the open-source software MSiReader. RESULTS: The TOF-SIMS and AP-MALDI-MS mass spectra show strong signals of fatty acids, cholesterol, phosphatidylcholine and sphingomyelin. We showed a high correlation between the fatty acid ions detected with TOF-SIMS in negative ion mode and the phosphatidylcholine ions detected with AP-MALDI in positive ion mode using a similar setting for visualization. Histological staining on the same section allowed the identification of the anatomical structures and their correlation with the ion images. CONCLUSIONS: This multimodal approach using two MSI platforms shows an excellent complementarity for the localization and identification of lipids. The spatial resolution of both systems is at or close to cellular dimensions and thus spatial correlation can only be obtained if the same tissue section is analyzed sequentitially. imzML-based data processing allows a real correlation of the imaging datasets provided by these two technologies and opens the way for a more complete molecular view of the anatomical structures of biological tissues.


INTRODUCTION
Mass spectrometry imaging (MSI) is the major and very active analytical method among the various techniques aiming to map the surface of the sample, capable of providing simultaneously the spatial distribution of a wide range of molecules directly from biological samples in a single run. 1  These two techniques allow access to the distribution of several classes of biomolecules from the surface of a tissue section. 2,3,4,5 Time-of-flight-SIMS (TOF-SIMS) consists of the bombardment of the sample by a focused beam of mono-or polyatomic ions, which induces desorption/ionization of secondary ions from the surface of the sample. 6,7,8 It also offers the possibility of localizing various ions produced from molecules, mainly lipids, drugs, xenobiotics and metabolites, with a mass-tocharge ratio up to m/z 1000 -1500, good mass resolution M/ΔM = 8 000 (FWHM) at m/z > 500, and a high lateral resolution from 400 nm to 1 -2 µm. This makes TOF-SIMS a method of choice for the micrometric scale analysis of lipids or other kinds of small molecules in biological samples. 5 Moreover, no matrix coating is required, i.e. no surface modification is made. One of main important breakthroughs in TOF-SIMS during the fifteen last years concerns polyatomic ion sources. The introduction of the polyatomic ion sources and, in particular, ion guns providing metal clusters (e.g. bismuth and gold clusters) has improved the desorption/ionization of intact ions from molecules, significantly expanding the application of TOF-SIMS from a mapping tool of elements or small mass fragments to a powerful molecular microscope used in various fields, ranging from materials characterization to biological tissue imaging. 6,8,9,10 Despite this improvement, two main limitations still exist: the high fragmentation rate induced by the high collision energy from the primary ion beam and the lack of tandem mass spectrometry capabilities. 11,12,13 After the first attempts a few years ago, 14,15 the latter issue is going to be addressed in the near future, with the recent advent of SIMS instruments with TOF/TOF and/or high resolution Orbitrap mass analyzers. 16,17 MALDI imaging was described initially by Spengler et al 18,19 and tissue imaging was first shown by Caprioli et al. 20 Until recently, the main limitation of the MALDI method for MSI was its spatial resolution, which was typically in the range of 50 -200 µm. The Spengler group developed an efficient atmospheric pressure scanning microprobe MALDI (AP-SMALDI) with a focused laser beam providing a high spatial resolution of 1.4 µm. 21 Moreover, coupling this with an orbital trapping mass spectrometer offers high mass resolution, mass accuracy and MS/MS capabilities. 21 In addition, an atmospheric pressure MALDI ion source is perfectly suited to investigate biological samples and allows the detection of a wide range of biomolecule classes, including metabolites, 22,23 lipids 24 and peptides/proteins. 25 Mass spectrometry imaging is now widely used in many applications, mainly in biological sciences and medical research, 26,27,28,29 , but also in cultural heritage research. 30 Correlated imaging has become an emerging strategy to combine complementary information from different analytical techniques. 31 The Cooks group combined desorption electrospray ionization (DESI) and MALDI imaging using a single tissue section, 32 achieving lipid and protein imaging by DESI-MS and MALDI-MS, respectively. 32 Despite the improvement concerning the polyatomic ion sources, Brunelle et al showed the need to combine molecular information from TOF-SIMS and MALDI-MS imaging, and the possibility of performing a MALDI imaging experiment on the same sample after TOF-SIMS imaging. 33 Eijkel et al combined MALDI and SIMS imaging datasets applied to the human cerebellum tissue. 34 Touboul et al also combined these two imaging techniques to study skin and kidney biopsies of patients suffering from Fabry disease by mapping globotriaosylceramides and digalactosylceramides, 35 showing good complementarity between the two techniques based on the identification and localization of biomolecules. In addition, Chughtai et al combined the elemental and small molecular distribution provided by high lateral resolution SIMS with the specific distribution of the lipids and peptides/proteins provided by MALDI for the study of musculoskeletal tissue. 36 Imaging dataset processing is a great challenge. The main difficulty for biologists or clinicians is to analyze, merge, compare and correlate data provided by different instruments on the same platform. Moreover, MSI data comprises a complex and huge dataset containing all relevant properties correlated to the mass spectral data. Vendors of MS instruments and many bio-informatics groups have come up with several pieces of software to analyze MSI datasets. Consequently, a common data format known as imzML has been developed over the past few years (www.imzml.org). 37 The vendor-neutral data format imzML facilitates the flexible sharing of MSI data and its visualization into various software tools available without restriction to a proprietary vendor. 37 Additional details are provided in a book chapter. 38 One of the most relevant examples is the data processing of a multicenter study. 39 The authors analyzed adjacent sections of mouse brain in five laboratories situated mainly in Europe and the USA. Five different instruments were used including MALDI-TOF/TOF, Orbitrap, QTOF, FT-ICR and TOF-SIMS. The imaging dataset was converted into imzML format using the appropriate converter tools (www.imzml.org) and displayed in a common opensource software to facilitate the exchange and the comparison. 39 In the current study, we defined a workflow based on the investigation of lipids combining TOF-SIMS and AP-MALDI-Orbitrap. In addition, this multimodal approach using these two imaging methods offers a strong complementarity, due, on the one hand to the precise localization of biomolecules by the high spatial resolution provided by TOF-SIMS, and on the other to the identification/confirmation of molecular structures by the high mass accuracy, high mass resolution, high lateral resolution and MS/MS capability of the AP-MALDI-MS setup. Imaging data were converted to the standard imzML format and MS images generated using an open-source software. The workflow was applied to only one tissue section of human colon tumor to correlate information.

Tissue samples
Serial cryosections of human colon cancer (thickness: 12 µm) were cut at -20 °C using a CM1950-S cryostat (Leica, Wetzlar, Germany) and deposited on glass slides coated with indium tin oxide (ITO). The samples were dried in vacuum under a pressure of a few hectopascals for 15 min before the SIMS analyses. Optical images were recorded with an Olympus BX51 microscope (Olympus, Rungis, France) equipped with ×1.25 to ×50 lenses and a Color View I camera, monitored by Cell B software (Soft Imaging System, GmbH, Münster, Germany).

TOF-SIMS imaging
The experiments were performed using a commercial TOF-SIMS IV mass spectrometer (ION-TOF GmbH, Münster, Germany). This mass spectrometer, described in detail elsewhere, 8 is fitted with a bismuth liquid metal ion gun delivering Bin q+ bismuth cluster ions (Bi3 + ions were selected). A low-energy electron flood gun was activated between two primary ions pulses to neutralize the sample surface, causing only minimum damage. 40 Only one mode of operation of the primary ion column was used during the experiments, which is called a "high-current bunched mode", 34 Images of the human colon with a field of view of 500 µm × 500 µm containing 512 × 512 pixels were recorded, leading to a pixel size of 1 µm. Consequently, in this mode the pixel stepsize was smaller than the beam diameter (2 µm), leading to oversampling. Another mode of operation could be used, which combines a higher spatial resolution of ~400 nm and a mass resolution of M/ΔM = 8 × 10 3 , thanks to a delayed extraction of the secondary ions. 42 However in the present case the "high current bunched mode" was preferred because it ensures the fastest acquisition time. Under these conditions, the fluence (also called the primary ion dose density) was maintained at 5.0 × 10 11 ions/cm 2 , which is below the so-called static SIMS limit. 43 Because of the very low initial kinetic energy distribution of the secondary ions, the relationship between the time-of-flight and the square root of the m/z value is always linear over the whole mass range. The calibration was always internal and the signals used for the initial calibration were those of H + , H2 + , H3 + , C + , CH + , CH2 + , CH3 + and C2H5 + ions in positive, and H -, C -, CH -, CH2 -, C2 -, C3 -, C4Hin negative ion mode. The mass calibration could eventually be refined by adding well-identified ions of higher mass, such as fatty acid carboxylates and deprotonated vitamin E, to further improve mass accuracy. 44,45 The data acquisition software used was SurfaceLab 6.2 (ION-TOF GmbH).

AP-MALDI-MS imaging
After the static SIMS imaging experiments, a uniform matrix layer (pNA, 10 mg/mL in 1:1 acetone/water, 0.1 % TFA) was applied to the section using a pneumatic sprayer. 46 The MALDI MS imaging analyses were performed using a high lateral resolution atmospheric pressure imaging ion source (AP-MALDI10, TransMIT GmbH, Giessen, Germany) coupled to an orbital trapping mass spectrometer (Q Exactive, Thermo Fisher Scientific GmbH, Bremen, Germany). 21 The mass spectrometer was operated in positive ion mode at a mass resolution of 140,000 at m/z 200 over a mass range of m/z 700 to 900. The ion source was equipped with a nitrogen laser (λ = 337 nm), operating at a repetition rate of 60 Hz, for desorption/ionization. A useful spatial resolution from biological tissue down to a pixel size of 5 µm has been reported using this ion source. 2,24 Internal mass calibration was performed using a lipid ion signal as a lock mass [PC(34:1) + K] + ion at m/z 798.54096 in positive ion mode, resulting in a mass accuracy better than 2 ppm. Positive lipid ion fragmentation was performed to identify and confirm some molecular structures of lipids by high-energy collisional dissociation (HCD). The isolation window for the precursor was set to ± 0.5 u.
The mass resolution for MS/MS was set to R = 70,000 (@ m/z 200).

Data processing
Image dataset from TOF-SIMS (.ITM files from ION-TOF were exported into .GRD by SurfaceLab software) and AP-MALDI-MS (.RAW file from Thermo Fisher Scientific) were converted to imzML using the "toImzmlModule" converter developed by Commissariat à l'Energie Atomique et aux Energies Alternatives (CEA, Saclay, France), and the "RAW to imzML" converter developed by Justus Liebig University, respectively. 37 As a result, the imzML files were processed using MSiReader, a free open-source MSI software. This vendor-neutral interface was built on Matlab by Robichaud et al. 47 The ion selection bin width (m/z window) of the images generated from the MALDI MS dataset was Δm/z = 0.01, and Δm/z = 0.2 for the TOF-SIMS dataset. Additional details of the imzML conversion and processing are available at www.imzml.org. Note that the TOF-SIMS and AP-MALDI-MS images have not been normalized or interpolated.

Histological staining
The section imaged was stained after MSI measurement to compare the histological features. Using the same section, hematoxylin and eosin (H&E) staining was performed after removing the pNA matrix with 100 % ethanol.

RESULTS AND DISCUSSION
The workflow presented for correlating MSI combining TOF-SIMS and AP-MALDI-MS on a single biological section consists of several steps and is presented in Figure 1 Table 1 in the supporting information.
Colon cancer spreads through the mucosa layer to the submucosa layer. Cancer cells infiltrate the submucosa and modify the cellular and extracellular composition. 52 Figure 3A shows an optical image of the submucosa infiltrated. The necrotic structures, tumor microenvironment (desmoplastic tumor stroma) and tumor cells were assigned after H&E staining and were labeled accordingly. The displayed area was mapped by AP-MALDI corresponding to a field of view of 2550 µm × 3000 µm, with a step size of 10 µm (255 × 300 pixels). The black square indicates the region that was scanned by TOF-SIMS. This second area corresponds to the combination/juxtaposition of 16 adjacent images of 500 µm × 500 µm each, with a pixel size fixed at 1 µm (512 × 512 pixels). The imzML conversion allows access to a large choice of software tools and the corresponding key features (www.imzml.org). In this study we used the open-source software MSiReader, 47 which offers the opportunity to overlay ion images with an optical picture, such as the histological staining. Moreover, the main advantage of using a unique software is the possibility of using the same settings for visualization, such as the color panel/interpolation settings, leading to a better combination and correlation / comparison of the imaging dataset. Figure 3B  Moreover, cholesterol and sphingomyelin ions seemed to be co-localized in the necrotic areas. Formation of these areas could be the consequence of the apoptosis mechanism. The Setou group showed the distribution of sphingomyelin in colon cancer liver metastasis tissue by MALDI-MS imaging. 53 The workflow presented using only one tissue section has shown, for the first time, a high image quality, due to the very high spatial resolution provided by TOF-SIMS (~1 µm) and the very competitive spatial resolution provided by the efficient AP-MALDI (10 µm).
The reproducibility of the workflow is demonstrated using another kind of tissue section. Mouse brain was used for this purpose and showed a high correlation. The mass spectra showed a similar lipid profile in negative ion mode using the two MSI techniques ( Figure S2, supporting information). In both cases, mass spectra were dominated by the sulfatides (ST).
A common lipid ion at m/z 888 tentatively assigned to the sulfatide [ST(42:2) -H] -, was chosen and used to generate the TOF-SIMS and AP-MALDI images, showing the same distribution in the hippocampus area ( Figure S3, supporting information). This demonstrates that the described workflow results in a high reproducibility and can be applied to other tissue types.

CONCLUSIONS
The workflow applied to a single human colon cancer sample, combining the most commonly used mass spectrometry imaging technologies TOF-SIMS and AP-MALDI-Orbitrap, showed high spatial correlation and complementary molecular information.
Improvements in the MALDI imaging spatial resolution allowed a much better spatial correlation with TOF-SIMS. This requires the sequential analysis of a single tissue section, in contrast to the parallel investigation of adjacent tissue sections as carried out in previous investigations combining MALDI-TOF and TOF-SIMS. This fact has to be considered in the sample preparation procedure concerning, for example, sample support, sample handling and histological staining. The emerging technology of multimodal imaging significantly expands the capabilities for revealing the molecular complexity in tissue of both healthy and diseased state. Consequently, this approach could be used to obtain a more compete and detailed understanding of pathological changes on a molecular level, for example, by combining fast SIMS measurements of highest spatial resolution with high molecular information MALDI measurements.    using the "parula" colormap.