High resolution tissue imaging on an orbitrap mass spectrometer by desorption electro-spray ionization mass spectrometry (DESI-MS)

To the editor,

The major methods for examining tissue samples include histology, autoradiography of radiolabeled compounds, and fluorescence microscopy in conjunction with labeled antibodies. Mass spectrometry (MS) imaging has the potential to complement these methods and is advantageous for reasons that include the fact that more chemical information can be gleaned during a single experiment and that labeling is not required. Some of its limitations include ion suppression effects, the fact that targeted experiments are often more challenging, and, most importantly, that it is an invasive experiment. The principal methods for MS imaging are secondary ion mass spectrometry (SIMS)^{1, 2}, which has the advantage of high spatial resolution, and matrix assisted laser desorption/ionization (MALDI)^{3, 4}, which allows for the detection of higher molecular weight species and has been used for high resolution/high mass accuracy imaging of lipids from tissue⁵. More recently, attention has been given to imaging by ambient ionization methods, which allow for the examination of samples in situ, outside the vacuum of the mass spectrometer; reports in the literature of ambient imaging include use of DESI-MS⁶, surface-sampling probe electrospray ionization⁷, and laser ablation electro-spray ionization (LAESI)⁸. The first ambient imaging method, DESI-MS, analyzes condensed phase samples using a pneumatically assisted stream of charged microdroplets directed at the surface⁹. A thin liquid film collects on the surface, and the impact of incoming droplets on this thin film causes the ejection of small secondary droplets containing the dissolved analyte¹⁰. Reports in the literature include examples of imaging of thin tissue sections^{6, 11, 12}, seaweed¹³, TLC plates¹⁴, ink on paper¹⁵, and latent fingerprints¹⁶.

One defining feature of ambient analysis is that, because samples are analyzed in their native state, little or no sample preparation is used. Any sample preparation or treatment is typically done concurrently with the analysis and not as a separate step. The principal benefits of ambient analysis, therefore, are a reduction in the time required for analysis and the ease of operation. One drawback is that untreated samples are often chemically complex, and this is particularly true in tissue imaging applications. High selectivity is therefore important for the successful implementation of ambient analysis. Increased selectivity in DESI over that available by unit resolution, single stage mass spectrometry has been achieved by tandem mass spectrometry, the use of in situ chemical reactions¹¹, and by high resolution mass spectrometry¹⁷. High resolution DESI experiments were reported in 2006 by Fernandez, Muddiman and their coworkers using FT-ICR. Volny and coworkers recently extended this work by performing high m/z resolution and mass accuracy DESI imaging on an FT-ICR, but the spatial resolution was lower than previously reported for DESI imaging¹⁸. In this letter, we describe DESI-MS imaging of rat brain tissue sections using an orbitrap mass spectrometer (Thermo Scientific, San Jose, CA), which has mass accuracy and resolving power in the same range as that of the FT-ICR.

All of the experiments were performed on an LTQ Orbitrap mass spectrometer (Thermo Scientific). Imaging experiments were done using the 30k mass resolution setting, while the 100k setting with lock masses was used for exact mass measurements to calculate molecular formulae. Mass standards were delivered directly to the mass spectrometer via an electrosonic spray ionization (ESSI) source concurrently with the DESI analysis. The [M+H]⁺ of reserpine and the [M+Na]⁺ ion of cyclosporine A, exact masses of 609.28121 and 1224.83114, respectively, were used as standards in the positive ion mode; the [M-H]^- ion of taurocholate and the C₁₅H₁₅O₆N₃P₃F₂₀(C₂F₄)₂ ultram]1621 peak, with theoretical exact masses of 514.28385 and 1005.97266, respectively, were used in the negative ion mode. Frozen rat brain was purchased from Pel-Freez Biologicals, cut into 10 micron sections, and thaw mounted on a glass slide. The pixel size in the imaging experiments was 230 × 230 microns. The acquisition time per pixel on the LTQ-Orbitrap was 1.06 s for a total image collection time of 48 minutes for the coronal section and 78 minutes for the sagittal sections. Full profile mode was turned on to keep the orbitrap mass list constant. The MS images were viewed in BioMap (www.maldi-msi.org) by first importing the MS data into MATLAB (MathWorks) for conversion into the Analyze 7.5 format and scaling of the intensities to a maximum of 32767, the upper intensity limit imposed by BioMap. A commercially available software package, FireFly, was also used for the image conversion (Indigo BioSystems, Inc., Indianapolis). No denoising of the data was performed

Full scan mass spectra for the direct analysis of rat brain at 100k resolution are shown in Figure 1. The polar lipid profile was similar to that obtained previously from direct analysis of rat brain tissue by DESI-MS on a unit resolution ion trap^{6, 19}, except that a number of new peaks are identifiable due to the increased resolution. In the positive ion mode, most of the abundant species were potassium and sodium adducts of diacyl-glycerosphosphocholines (GPCho), while some of the less abundant peaks were identified as glycerophosphoethanolamines (GPEtn), sphingomyelin (SM), and glycosphingolipids like cerebroside (GlcCer). In the negative ion mode, deprotonated fatty acids such as palmitate, oleate, and docosahexaenoate were detected; the [M–H]⁻ ions of the relatively acidic polar lipids were also observed, including glycerophosphoserines (GPSer), glycerophosphoinositols (GPIns), and sulfatides.

Figure 1.

Typical mass spectra in the negative ion mode (A) and the positive ion mode (B) for the analysis of rat brain sections. The region around m/z 834.5 in the negative ion mode is shown in (C).

In both the positive and the negative ion modes, several pairs of peaks differing by less than 0.1 m/z units (Thomson - Th) were observed. The measured m/z for these ions and the proposed chemical formulae and molecular identifications are shown in Table 1. Molecular identification on the LTQ orbitrap was performed by a combination of exact mass measurement, tandem mass spectrometry, and database searching using the LIPID MAPS Structure Database (www.lipidmaps.org). The mass accuracy at 100k resolution was generally sufficient to correctly calculate the molecular formula. Tandem mass spectrometry (MS/MS) was used to confirm the molecular identity of the ions; the collision induced dissociation (CID) fragmentation patterns of polar lipids have been reported on extensively²⁰. Briefly, fragmentation in the negative ion mode is usually informative, commonly yielding fragment ions corresponding to fatty acid side chains and the polar head group. MS/MS of phospholipids was less informative in the positive ion mode. GPCho, SM, and GPEtn gave a characteristic neutral loss from the headgroup, but fragment ions resulting from the radyl groups were generally low in intensity or absent.

Table 1.

Examples of polar lipids detected in the positive and negative ion modes differing by less than 0.1 Th

negative ion mode
measured m/z	proposed formula	error (ppm)	molecular identity
834.5284	C46H77O10NP	0.575	40:6 diacyl GPSer
834.5766	C44H84O11NS	0.72	C20 Sulfatide
863.6105	C4513CH88O11NS	−0.104	C22 sulfatide, 13C
863.5609	C45H84O13P	−3.555	36:1 diacyl GPIns
886.5527	C4613C H82O13 P	0.609	38:4 diacyl GPIns, 13C
888.6226	C48H90O11NS	−0.28	C24:1 sulfatide
909.6562	C4513C3H92O12NS	6.711	C24-OH sulfatide, 13C
909.5487	C49H82O13P	−0.035	40:6 diacyl GPIns
positive ion mode
768.5308	C41H80O7NP+K	0.391	36:1 diacyl GPEns
768.5882	C42H84O7NP+Na	0.418	34:1 plasmalogen GPCho
784.5622	C42H84O7NP+K	0.461	34:1 plasmalogen GPCho
784.5834	C42H84O8NP+Na	0.714	34:0 GPCho
806.5103	C43H78O8NP+K	0.637	38:4 diacyl GPEtn
806.5700	C46H81O8NP	0.569	38:6 diacyl GPCho
832.5835	C46H84O8NP+Na	0.966	38:4 diacyl GPCho
832.6645	C48H91O8N+Na	0.973	C24:1 GlcCer
848.5577	C46H84O8NP+K	1.269	38:4 diacyl GPCho
848.6386	C48H91O8N+K	1.147	C24:1 GlcCer

Eight pairs of selected ion images, four pairs each from the negative (Figure 2A–H) and the positive ion mode (Figure 2I–L), are shown in Figure 2. In five of the ion pairs shown, two unique molecular species have nearly the same m/z, such as 36:1 GPEtn and 34:1 plasmalogen GPCho in Figure 2I and J. In the other three pairs, a relatively abundant isotope peak obscures a minor component, such as the ¹³C isotope peak of 38:4 GPIns and the [M–H]⁻ ion of C24:0 sulfatide in Figure 2E and F. In this case, only one of the peaks is of unique interest. However, without the use of a high resolution mass spectrometer, the minor component would be obscured by the more abundant isotope peak.

Figure 2.

Selected ion images in the negative (A–H) and positive (I–L) mode

The selected ion image of m/z 834 from a coronal section of rat brain imaged on an LTQ ion trap in the negative ion mode is shown in figure 3A⁶. This previous study identified the ion as 40:6 GPSer. Imaging on the LTQ orbitrap, however, indicated that this ion actually consists of two different molecular species (figure 1C) with distinct distributions: 40:6 GPSer and C20 sulfatide (Figure 3B and C). This example illustrates the difficulty with using MS/MS alone to identify ions of almost identical mass. First, interpretation of tandem mass spectra resulting from multiple molecular ions is challenging. Second, fragment ions resulting from precursor ions with lower relative abundance may be missed. Third, and most importantly, some overlapping species may only be present in isolated regions of the tissue. In the case shown in Figure 3, the minor component, C20 sulfatide, is detectable only in the white matter (e.g. corpus callosum) but is absent in the gray matter, which makes up the bulk of the tissue surface area. A tissue section would need to be exhaustively analyzed (i.e. imaged) in the MS/MS mode for each peak to identify overlapping ions; N peaks would therefore require N number of MS/MS images, a significant problem because N times more sample and time would be required compared to collecting a single image. Imaging using a high resolution MS solves this problem. The improved resolution of the orbitrap compared with an ion trap resolves most nearly isobaric species and increases the amount of chemical information obtained from complex samples. Using an orbitrap at a resolution of 30k does increase the amount of time required to collect an image by approximately 1.5 times compared with a unit mass resolution ion trap.

Figure 3.

Selected ion images in the negative ion mode of (A) m/z 834.5 on an LTQ ion trap. B) m/z 834.53 and (c) m/z 834.58 collected on an orbitrap

Accurate mass data, however, was often not sufficient to unambiguously identify unknown peaks for two reasons. First, there are a number of isomeric lipid species with the same exact mass. Second, even when operated at 100k resolution, some peaks in the rat brain positive and negative mode spectra are not fully resolved, reducing the accuracy of the mass/charge measurement. For example, the peak detected at m/z 909.6562 in the negative ion mode was only partially resolved from another peak at approximately m/z 909.635. Interpretation of MS/MS spectra was still needed in this case to identify the major component as C24-OH Sulfatide, ¹³C₃. The orbitrap was therefore valuable as a tool for screening for isobaric peaks, followed by MS/MS for molecular identification. To conclude, ambient methods such as DESI-MS are used to analyze untreated samples that are typically chemically complex. High resolution tissue imaging using the orbitrap, as demonstrated here, is one way to improve the selectivity of ambient ionization methods, a necessity for the analysis of complex mixtures.