MALDI imaging mass spectrometry of lipids by adding ...

4 downloads 7 Views 1MB Size Report
Feb 14, 2011 - Abstract Mass spectrometry imaging of lipids using. MALDI–TOF/TOF mass spectrometers is of growing interest for chemical mapping of ...
Anal Bioanal Chem DOI 10.1007/s00216-011-4814-9

ORIGINAL PAPER

MALDI imaging mass spectrometry of lipids by adding lithium salts to the matrix solution Christopher D. Cerruti & David Touboul & Vincent Guérineau & Vanessa W. Petit & Olivier Laprévote & Alain Brunelle

Received: 21 December 2010 / Revised: 14 February 2011 / Accepted: 14 February 2011 # Springer-Verlag 2011

Abstract Mass spectrometry imaging of lipids using MALDI–TOF/TOF mass spectrometers is of growing interest for chemical mapping of organic compounds at the surface of tissue sections. Many efforts have been devoted to the best matrix choice and deposition technique. Nevertheless, the identification of lipid species desorbed from tissue sections remains problematic. It is now wellknown that protonated, sodium- and potassium-cationized lipids are detected from biological samples, thus complicating the data analysis. A new sample preparation method is proposed, involving the use of lithium salts in the matrix solution in order to simplify the mass spectra with only lithium-cationized molecules instead of a mixture of various cationized species. Five different lithium salts were

tested. Among them, lithium trifluoroacetate and lithium iodide merged the different lipid adducts into one single lithium-cationized species. An optimized sample preparation protocol demonstrated that the lithium trifluoroacetate salt slightly increased desorption of phosphatidylcholines. Mass spectrometry images acquired on rat brain tissue sections by adding lithium trifluoroacetate showed the best results in terms of image contrast. Moreover, more structurally relevant fragments were generated by tandem mass spectrometry when analyzing lithium-cationized species. Keywords Mass spectrometry imaging . MALDI . Lithium salt . Lipid . Rat brain

Introduction Published in the special issue MALDI Imaging on with Guest Editor Olivier Laprévote. Electronic supplementary material The online version of this article (doi:10.1007/s00216-011-4814-9) contains supplementary material, which is available to authorized users. C. D. Cerruti : D. Touboul (*) : V. Guérineau : V. W. Petit : O. Laprévote : A. Brunelle Centre de Recherche de Gif, Institut de Chimie des Substances Naturelles, CNRS, Avenue de la Terrasse, 91198 Gif-sur-Yvette Cedex, France e-mail: [email protected] O. Laprévote Chimie Toxicologie Analytique et Cellulaire, EA4463, Faculté des Sciences Pharmaceutiques et Biologiques, Université Paris Descartes, 4 avenue de l’Observatoire, 75006 Paris, France

Biological imaging allows obtaining in situ information on the physiology and function of organs. There are two broad categories of imaging techniques: that which gives a structural view of a tissue or organ, detailing its histology and its possible pathology, and one that gives a functional view, by measuring its activities. Numerous chemical imaging techniques are available to detect and localize biological compounds, such as lipids, at a tissue or cell level. Lipids are fat-soluble molecules, which are involved in energy storage, cell membrane structure, signaling pathways, cell growth, and death [1]. Among them, phospholipids are of the greatest importance to maintain membrane integrity, to modulate its flexibility, permeability, and to allow anchorage of cell surface proteins. Since all cellular structures are surrounded by a phospholipid bilayer, this class of compounds is very abundant and also directly reflects the cellular state.

C.D. Cerruti et al.

Classical imaging techniques, such as histology using Nile red dye [2] which is a very sensitive fluorescent histochemical staining tool, only enables to localize the total lipid fraction on tissue sections. Immunohistochemistry is of interest for protein mapping but still remains inefficient for the visualization of a specific lipid class. Contrariwise matrix-assisted laser desorption/ionization (MALDI) offers the unique capability to analyze complex mixtures of biological compounds and is not directed a priori to any class of molecule. MALDI consists to mix or to deposit an organic matrix with/on the sample and to irradiate the surface by a UV or IR light generated by a pulsed and focused laser. The matrix absorbs the light at the wavelength of the laser, leading to the soft desorption/ ionization of the intact compounds of interest. The energy deposited for desorption/ionization remains localized allowing the analysis of well-defined areas of the sample. By acquiring mass spectra at regular intervals on a surface, an ion density map, corresponding to the plot of an ion intensity versus (x and y) position, can be reconstructed by an appropriate software [3]. The great advantage of mass spectrometry imaging (MSI) lies in the possibility of reconstructing as many images as detected ions in the spectra acquired in a single run [3–6]. MALDI–mass spectrometry imaging (MALDI–MSI) enables localization of a wide range of biomolecules on tissue sections in a single experiment [4, 7–12] and is a particularly efficient method for the visualization of small endogenous metabolites, especially lipids [13–15]. One of the critical limitations of the spatial resolution of MALDI–MSI is the size of the organic matrix crystals and the analyte migration during the matrix application process. The variation in ionization efficiency in MSI is caused by heterogeneous distribution of organic matrix crystals. A second limitation is the spatial resolution, which is restricted to ∼50 μm in routine analysis, although the best of the state-of-the-art now reaches 5–10 μm, but with very specific conditions [16]. In comparison, cluster TOF-SIMS imaging routinely reaches spatial resolutions in the 0.4–1 μm range [17].

Ideally, during the acquisition of a MALDI spectrum, one single ion peak for each analyte should be obtained. This is usually the case when analyzing protein digests for example, giving mostly protonated species for each peptide. The situation is much more complex with tissues, which are naturally rich in salts, such as of sodium and potassium, leading to unspecific cationization during the MALDI process. Phospholipid ion signals, especially those of phosphatidylcholines, are therefore detected with three different peaks corresponding to protonated [M+H]+, sodium-cationized [M+Na]+ and potassiumcationized [M+K]+ species. Table 1 summarizes the m/z values of some brain glycerophosphatidylcholines (PCs), glycerophosphatidylethanolamines (PEs), and galactosylceramides (GalCers) with various cationizations. For example, the ion at m/z 806.5 can be attributed to a [M+H]+ of PC (38:6) or to a [M+K]+ of PE (38:4) which are difficult to correctly separate with a reflectron time-of-flight and impossible to select independently for tandem mass spectrometry (MS/MS) analysis, whatever the mass analyzer. The aim of the present work is to provide a new sample preparation method allowing the detection of each phosphatidylcholine ion as a single intense ion peak on tissue sections. The original idea is coming from the work of Woods and Setou, who both added lithium salts to the matrix solution [18, 19]. Indeed lithium ion exhibits a high affinity for phospholipids, which is due to the presence of exchangeable hydrogen on the phosphate group. Moreover, the small size of the Li+ cation induces strong ion-dipole interactions and thus tends to form bonds with a strong covalent character. This property has already been widely used by Hsu et al. for the structural analysis of lipids under low collision energy dissociation regime [20–24]. Nevertheless, previous work did not provide an efficient sample preparation for a significant lithium cationization of the targeted molecules. Thus, the present work is an extensive study of five different lithium salts used at two different concentrations in order to optimize the experimental parameters leading to

Table 1 Summary of ion mass-to-charge ratio (m/z) values from glycerophosphatidylcholines (PCs), glycerophosphatidylethanolamines (PEs), and galactocerebroside (GalCers) molecular species: [M+H]+, [M+Na]+, [M+K]+, [M+Li]+, and [M+2Na–H]+ Molecular species PC (diacyl-32:0) PC (diacyl-34:1) PC (diacyl-34:0) PC (diacyl-38:6) PE (diacyl-38:4) PE (diacyl-38:1) d18:1/24:1–GalCer d18:1/h24:0–GalCer

[M+H]+ (m/z)

[M+Na]+ (m/z)

[M+K]+ (m/z)

734.6 760.6 762.6 806.6 768.6 774.6 810.7 828.7

756.6 782.6 784.6 828.6 790.5 796.6 832.7 850.7

772.7 798.5 800.6 844.5 806.5 812.6 848.6 866.7

[M+2Na–H]+ (m/z)

812.5 818.6

[M+Li]+ (m/z) 740.5 766.6 768.6 812.6 774.6 780.6 816.7 834.7

MALDI imaging mass spectrometry of lipids by adding lithium salts

specific lithium-cationized phosphatidylcholines on tissue sections. The MALDI–MSI experiments were performed in order to determinate which lithium salt offers the best image definition and to enlighten possible spatial delocalization. Finally, tandem mass spectrometry (MS/MS) experiments were carried out to confirm the superiority of lithium adducts for structural analysis.

Experimental section Alpha-cyano-4-hydroxycinnamic acid (CHCA), five lithium salts (citrate (LiCitrate), acetate (LiAc), trifluoroacetate (LiTFA), iodide (LiI), chloride (LiCl)), trifluoroacetic acid (TFA), and solvents were purchased from Sigma-Aldrich (Saint-Quentin-Fallavier, France). All experiments made with rat brains were performed in accordance with the protocols approved by the National Commission on animal experimentation and by the recommendations of the European commission DGXI. Male Wistar rats (IBAIC, Orsay, France) between 300 and 420 g were euthanized by an intraperitoneal injection of sodium pentobarbital (>65 mg/kg). The trimmed tissue blocks were immediately frozen in dry ice to prevent crack formation during freezing and stored at -80 °C prior to MS experiments. The frozen intact rat brain was cut at -20 °C in a cryostat (model CM3050-S; Leica Microsystems SA, Nanterre, France). Serial tissue sections of 12 μm thickness were immediately deposited onto a stainless steel plate and controlled before and after matrix deposition with an optical microscope (Olympus BX 51 fitted with ×1.25 to ×50 objectives, Olympus France SAS, Rungis, France) equipped with a Color View Ι camera, monitored by CellB software (Soft Imaging System GmbH, Münster, Germany). Before matrix application, tissue sections were dried under vacuum, at a pressure of a few hectopascals during 30 min. Although the tissue blocks were held by an optimum cutting temperature polymer, they were not embedded into it since any residual polymer might degrade the tissue surface [25]. Matrix solution (α-cyano-4-hydroxycinnamic acid (CHCA)) was prepared at 10 mg mL−1 in acetonitrile/ water/trifluoroacetic acid (70/30/0.1, ν/ν/ν) for control analysis. Methanol, which is a suitable solvent for lipid analysis, was not used here due to the low solubility of lithium salts in this solvent. LiCitrate, LiAc, LiTFA, LiI, and LiCl were tested. Lithium solution at 1 and 2 mg mL−1 concentrations (lithium equivalent) were mixed with the matrix solution for each lithium salt. Rat brain sections were homogeneously covered by the matrix solution using a TM-Sprayer (HTX Technologies, Carrboro, NC, USA) in a single coating step. Experimental parameters were optimized in order to reach the best sensitivity and

reproducibility. This robot is equipped with a heated nozzle making a narrow aerosol of matrix droplets. The sample stage is moved below the thermal spray, allowing a regular matrix coating of a rat brain in less than 2 min. This system is coupled to an isocratic pump which allows a constant flow rate of 240 or 300 μL min−1 of the matrix solution, depending on the lithium salt as shown in Table S1 in the Electronic supplementary material. The spray is finally heated to 120 or 150 °C. The sample plate to receive the matrix is anchored on a stage moving in both x and y directions at a linear velocity of 120 cm min−1. MSI was performed using a 4800 MALDI–time-of-flight (MALDI–TOF)/TOF mass spectrometer from AB-Sciex (Les Ulis, France) equipped with a 200-Hz tripledfrequency Nd/YAG pulsed laser (355 nm) and an electrostatic mirror, leading to a routine mass resolution of about 10,000 in the MS mode. The data were acquired in the positive ion reflectron mode at an accelerating potential of 20 kV and a delayed extraction time of 450 ns. The number of laser shots per pixel was set to 300 and the distance between two adjacent pixels was set to 50 μm, which roughly corresponds to the laser spot diameter. Ion signals with mass-to-charge ratio between m/z 700 and 900 were recorded. External mass calibration was achieved using standard solutions of peptides (Pepmix 5, LaserBio Labs, Sophia Antipolis, France). The images were recorded using 4000 Series Imaging software (www.maldi-msi.org, M. Stoeckli, Novartis Pharma, Basel, Switzerland) and processed using TissueView software (AB-Sciex, Les Ulis, France). For acquisitions of direct on-tissue MS/MS spectra in the positive ion mode, 600 laser shots per pixel were needed, with a collision energy fixed at 2 keV and air pressure in the collision cell of 3×10−6 hPa.

Results and discussion Sample preparation For peptides and proteins analysis, methods for washing the sample surface with organic solvents or buffer solutions are described in the literature [26–29]. For lipids, no preliminary washing step is needed and especially any organic solvent should be avoided in order to prevent removal or delocalization of such compounds. The crucial step during sample preparation is the choice of the matrix and its homogeneous and efficient deposition over the sample surface to acquire good quality chemical images with a good contrast and no delocalization. Despite the promising capability of MALDI–MSI for imaging small metabolites, this technique still has several issues, especially in spatial resolution. One of the critical limitations of the

C.D. Cerruti et al.

spatial resolution of MALDI–MSI is the size of the organic matrix crystal and the analyte migration during the matrix application process. Matrices which crystallize in the form of long crystals will induce displacement of most of the compounds of interest after their deposition on tissue sections and are thus not recommended for MSI experiments. In the positive ion mode, 2,5-dihydroxy benzoic acid and CHCA are usually used for low mass range compound analysis. These two matrices are soluble in water/methanol and water/acetonitrile mixtures, and trifluoroacetic acid can be added for promoting protonation. CHCA seems to be the best option for lipid imaging due to a lower laser power needed for desorption/ionization and a very good sample surface coverage when deposited with a robotic sprayer [21, 30]. A compromise is required between a wet deposition, which is needed for an efficient mixing between the sample compounds and the matrix solution, and a dry deposition, which is needed to avoid delocalization. After many tests in the presence and absence of lithium salts, the best solvent for the matrix solution, which is also compatible with our robotic sprayer, is composed of water/acetonitrile/trifluoroacetic acid (70/30/0.1, v/v/v). The lithium salts are then mixed to the matrix solution at different concentrations of lithium (1 or 2 mg mL−1). It must be noted that lithium iodide and citrate salts led to the clogging of the sprayer nozzle as it solidifies in the capillary block. In each case, the morphology of the crystals on the tissue section was carefully inspected. This crystal morphology, which depends on the analyte/ matrix co-crystallization conditions, is known to strongly influence the spectrum quality in MALDI–MS [31–35]. Characteristic pictures of matrix crystals using 1 mg mL−1 lithium salts are presented in Fig 1. When compared with the “pure” CHCA control sample (Fig. 1a, no lithium salt), the addition of lithium acetate (Fig. 1b) leads to the formation of elongated clusters (∼1 mm long) of small crystals (∼10 μm diameter), whatever the spray temperature. Such morphology is presumed to be responsible of strong delocalization of analytes on the tissue surface. For lithium trifluoroacetate (Fig. 1c), small regular crystals of 10-μm diameter are homogeneously covering the tissue section that looks compatible with a minimal delocalization process. Lithium citrate (Fig. 1d) mixed with CHCA induces the formation of large irregular plaques of about 100 μm diameter which is probably a sign of large compound delocalization. Finally, lithium chloride (Fig. 1e) and lithium iodide (Fig. 1f) both exhibit the same morphology, i.e., a quite complete covering of the section without individual crystals. These observations suggest that lithium trifluoroacetate, iodide, and chloride salts mixed with CHCA are probably good candidates for imaging experiments.

Determination of the lithium salt inducing the highest rate of lithium-cationized phosphatidylcholines The mass spectra acquired from adjacent rat brain sections by MALDI–TOF in the positive ion mode are displayed in Fig. 2. Rat brain cerebellum was chosen as it exhibits two distinct parts, i.e., the white and the gray matters, which show different lipid compositions and can be considered as an easy way to control if delocalization was occurring during the matrix deposition process. The gray matter part was selected as a region of interest to extract the corresponding spectra. The relative proportions of protonated, sodium-, potassium-, and lithium-cationized species were calculated from these spectra using the corresponding peak areas A([M+Ci]+) for the two major phosphatidylcholines PC 32:0 and PC 34:1, according to the following equation:   A ½M þ Ci þ  %¼P  A ½M þ Ci þ With M being the molecule of interest and Ci the cation i (i=H+, Na+, K+, and Li+). With regards to the protonated molecule, sodium and potassium adducts, the formation of lithiated species results in a m/z shift of plus 6 amu and minus 16 and 32 amu, respectively. Comparison of Fig. 2a, b displays a significant m/z shift of 6 amu of PC32:0 and PC34:1 ions which corresponds to the replacement of the protonated species by lithium-cationized ones. Figure 3 shows the relative proportions of adducts in different additions. Lithiated adducts are not observed in the absence of lithium salt (pure CHCA) confirming the fact that there is no detectable lithium in the brain whereas potassium adducts predominate in the spectra (67–73%). Lithium citrate associated with the CHCA matrix forms compact structures requiring high laser fluences for obtaining only weak ion peaks. Thus, this lithium salt gave no usable results. Acetate counterion yields the highest proportion of lithium adducts (80–90%). Nevertheless, the heterogeneous crystallization is not compatible with MALDI–MSI, as described above. For lithium chloride, the percentage of lithium adducts did not exceed 65% and a high rate in protonated species is kept at 1 mg mL−1. Surprisingly at 2 mg mL−1, the relative abundance of potassium adducts is increasing but the lithium-cationized species proportions do not change. As the proportions of the different adducts are not consistent, this lithium salt cannot be retained for the following experiments. With LiTFA, the relative abundance of lithium adducts increases, as expected, with the lithium concentration and reaches 100%. The analysis of the mass spectra shows that LiTFA leads to the highest intensity of [M+Li]+ which is similar to

MALDI imaging mass spectrometry of lipids by adding lithium salts

200 µm

A

B

C

D

E

F

Fig. 1 Pictures of tissue sections sprayed with matrix solutions containing: a no lithium salt, b lithium acetate, c lithium trifluoroacetate, d lithium citrate, e lithium chloride, and f lithium iodide. Lithium concentration was 1 mg mL−1 in all cases

the [M+H]+ intensity of the control condition (Fig. 2a–e). Finally, lithium iodide showed similar results as lithium trifluoroacetate but a lower sensitivity (Fig. 2e versus Fig. 2c). Both last lithium salts are thus favorable for MALDI–MSI of tissue sections. Imaging The next step is to check the image definition by MALDI– MSI to validate the choice of the best lithium salts. A toohigh concentration causes a delocalization of most of the lipid species at the surface of the tissue section, resulting in blurred and distorted ion images (data not shown). The most suitable lithium salt concentration was 1 mg mL−1 for

MALDI–MSI. This concentration will always be used in the following sections. Peak attributions are done with the help of a lipid database, i.e., lipid maps.1 Imaging of cerebellum sections Figure 2 shows the ion images obtained from rat cerebellum tissue section by MALDI–MSI in positive ion mode. The ion image of sagittal rat cerebellum section covered by CHCA presents the intensity values (maximum intensity of a pixel) of 210 for the [PC32:0+H]+ ion species (m/z 734.6)

1

www.lipidmaps.org

C.D. Cerruti et al.

A 4000

798.5

CHCA 772.5

/: 0 - 210

0 700

/

m/z: 734.6

m/z: 760.6

800

750 m/z

400

CHCA+LiAc

G

766.6

/: 0 - 80

100

0 700

798.5

737.5

772.5

200

C 3000

/: 0 - 75

740.6

300 Intensity

/: 0 -150

782.6

756.6

1000

B

F

760.6

2000 734.6

Intensity

3000

/

m/z: 740.6

m/z: 766.6

800

750 m/z

CHCA+LiTFA

H

740.6

2500

766.6

/: 0 - 648

/: 0 - 607

m/z: 740.6

m/z: 766.6

/: 0 - 125

/: 0 - 130

m/z : 740.6

m/z : 766.6

772.5

1000 500

/

0 700

D

800

798.5

1500 737.5

Intensity

2000

800

750 m/z

I

CHCA+LiCl 766.6

760.6

200

740.6

734.6

400

737.5

Intensity

600

/ 0 700

750 m/z

800

E CHCA+LiI

0 700

/: 0 - 130

m/z: 740.6

m/z: 766.6

772.5

750 m/z

798.5

760.6

734.6 737.5

500

/: 0 - 125

740.6

1500

1000

J

766.6

2000

Intensity

Fig. 2 MALDI–MS spectra and ion images acquired in the positive ion mode from the gray matter part of a rat cerebellum tissue section: a CHCA alone, b CHCA plus lithium acetate, c CHCA plus lithium trifluoroacetate, d CHCA plus lithium chloride, and e CHCA plus lithium iodide. m/z values annotated in the spectra correspond to different adducts for phosphatidylcholines PC (diacyl32:0) and PC (diacyl-34:1). f Images with CHCA alone, g images with lithium acetate, h images with lithium trifluoroacetate, i images with lithium chloride, and j images with lithium iodide. Lithium concentration was fixed at 1 mg mL-1. Images of 150×150 pixels with a pixel size of 50 μm. The intensity values (Ι) indicated in each image correspond to the minimum and the maximum intensities in a pixel

/ 800

MALDI imaging mass spectrometry of lipids by adding lithium salts

Fig. 3 Relative intensities of [M+H]+, [M+Na]+, [M+K]+, and [M+Li]+ ions peaks of phosphatidylcholines PC diacyl-32:0 and PC diacyl-34:1, for each lithium salt. a Lithium salts concentrations, 1 mg mL−1. b

Lithium salts concentrations, 2 mg mL−1. Error on the measurement was estimated to be ±5%

and 150 for the [PC34:1+H]+ ion species (m/z 760.6), respectively (Fig. 2f). The lithium acetate mixed with matrix solution forms plaques, which are observable in Fig. 2g and cause delocalization of lipids on the sections. Thus, the results obtained with this salt were not exploitable for imaging. Lithium trifluoroacetate, chloride, and iodide can provide images of excellent quality (Fig. 2h–j). All PC ions are co-located in the lobes and lobules of the cerebellum. As expected, lithium trifluoroacetate gave the best contrast between 0 and 648 in agreement with the highest intensities observed in the mass spectrum. Nevertheless, a salt-added matrix solution, leads to heterogeneous crystallization which results in a slight spot-to-spot variation of signal intensity [36, 37]. The image quality could be enhanced by using a spectrum normalization whose principle is based on the exclusion a number of noise spectra [38], but the software for data processing used in the present experiments did not allow such a procedure.

([PE38:4+K]+ or [PC38:6+H]+), at m/z 812.6 ([PC38:3+H]+ or [PE38:1+K] + ) or ([PE38:4+2Na–H] + ), m/z 816.6 ([PE42:8+H]+), and m/z 834.6 ([PE38:1+Na+K–H]+) was found in the medulla oblongata, in the corpus callosum and in the white matter of cerebellum (Fig. 4e–h). Images recorded under such conditions suggest that phosphatidylcholines are diffusing around the section leading to a very intense halo (Fig. 4a–c). Thus the intensity value of each pixel I does not reflect the real intensity attained in the section only. Figure 4b perfectly illustrates this problem. Different anatomical areas of the brain are poorly differentiated, like the cerebral cortex and the corpus callosum, although I value is very high. The mass spectrum of Fig. 4i shows that the ion peaks at m/z 734.6 ([PC32:0+H]+) and m/z 760.6 ([PC34:1+H]+) are more intense outside of the section. Then CHCA sprayed on a tissue section does not provide optimum conditions for imaging mass spectrometry. In Figs. 5 and 6, ions at m/z 740.6 ([PC32:0+Li]+), m/z 766.6 ([PC34:1+Li]+), m/z 737.6 ([SM18:1+Li]+), and m/z 812.6 ([PC38:3+H]+, [PC38:6+Li]+, [PE38:1+K]+, or [PE38:4+2Na-H]+) are detected in the cerebral cortex, in the hippocampus and in the gray matter of cerebellum. The ions at m/z 806.5 ([PE38:4+K]+ or [PC38:6+H]+), m/z 816.7 ([PE42:8+H]+ or [d18:1/24:1–GalCer+Li]+) and m/z 834.6 ([PE38:1+Na+K–H]+ or [d18:1/h24:0–GalCer+Li]+) are co-localized in the medulla oblongata, in the corpus callosum and in the white matter of cerebellum (Figs. 5e, g, h and 6e, g, h). The spectra extracted from the cerebral cortex or from the corpus callosum are quite different (Figs. 4, 5, and 6). Those shown in Fig. 6 indicate that the

Imaging of whole brain sections On the basis of the above results, LiI and LiTFA were selected for further analysis. The images of several ions acquired under control conditions are shown in Fig. 4, while the images obtained with the addition of LiI and LiTFA are shown in Figs. 5 and 6, respectively. Figure 4a–c shows that ions at m/z 734.6 ([PC32:0+H]+), m/z 760.6 ([PC34:1+H]+), and m/z 731.5 ([SM18:1+H]+) are co-localized in the cerebral cortex and in the gray matter of cerebellum. A co-localization of ions at m/z 806.5

C.D. Cerruti et al.

Fig. 4 MALDI–TOF ion images, mass spectra, and optical image of a sagittal rat brain section in the positive ion mode. CHCA matrix solution. a m/z 734.6 ([PC32:0+H]+), b m/z 760.6 ([PC34:1+H]+), c m/z 731.5 ([SM18:0+H]+), d optical image of the sagittal rat brain section, e m/z 806.5 ([PE38:4+K]+ or [PC38:6+H]+), f m/z 812.6 ([PC38:3+H]+, [PE38:1+K]+, or [PE38:4+2Na-H]+), g m/z 816.6

([PE42:8+H]+), and h m/z 834.6 ([PE38:1+Na+K–H]+). Images of 245×160 pixels with a pixel size of 50 μm. The value of intensity (Ι) indicated in white in each image corresponds to the minimum and the maximum intensities in a pixel. MALDI–MS spectra in the positive ion mode: i acquired from outer edge of the section, j acquired from the cerebral cortex, and k acquired in the corpus callosum (white matter)

addition of LiTFA increases the sensitivity for the detection of PC32:0 and PC34:1 ion species, by a factor of ∼2 if compared with control conditions and by a factor of ∼3 if compared with LiI in the cortex brain (Figs. 4j, 5i, and 6i). The addition of lithium salt can greatly increase the signal of some ions such as the ion at m/z 834.6, whose intensity reaches a value of ∼3,750 after having added LiTFA. The addition of LiTFA salt to the matrix solution provides better contrasts in the cerebellum and also between the cerebral cortex and the corpus callosum. This is therefore the preferred method for imaging compared with LiI salt. Anatomical areas of the brain can be clearly observed and differentiated with a significant gain in image definition such as thalamus and hippocampus when using LiTFA. Table 2 summarizes the results obtained with each lithium salt with respect to the following criterions, easiness of matrix deposition, aspect of the matrix, lithium cationization and image quality.

According to Fig. 3, lithium cationization is probably not total when using a lithium concentration of 1 mg mL−1. This can be clearly observed when mapping ion at m/z 806.5 ([PE38:4+K]+ or [PC38:6+H]+) under control conditions (in absence of lithium salts) and ion at m/z 812.5 ([PC38:3+H]+, [PC38:6+Li]+, [PE38:1+K]+, or [PE38:4+2Na–H]+) using LiI or LiTFA, which are not co-localized. Only a MS/MS analysis would allow discriminating species having very close masses. This will be exemplified for five different ions which are expected to be lithium-cationized. In situ structural characterizations by tandem mass spectrometry and fragmentation processes A MALDI–TOF/TOF mass spectrometer can acquire MS/ MS spectra directly from a tissue section with a good sensitivity [6, 7, 20, 21]. The most difficult step is the careful selection of precursor ions because of the limited

MALDI imaging mass spectrometry of lipids by adding lithium salts

Fig. 5 MALDI–TOF ion images, mass spectra, and optical image of a sagittal rat brain section in the positive ion mode. Matrix solutions CHCA mixed with lithium iodide. a m/z 740.6 ([PC32:0+Li]+), b m/z 766.6 ([PC34:1+Li]+), c m/z 737.5 ([SM18:0+Li]+), d optical image of the sagittal rat brain section, e m/z 806.5 ([PE38:4+K]+ or [PC38:6+H]+), f m/z 812.6 ([PC38:3+H]+, [PC38:6+Li]+, [PE38:1+K]+, or [PE38:4+ 2Na-H]+), g m/z 816.7 ([PE42:8+H]+ or [d18:1/24:1–GalCer+Li]+), h m/z

834.6 ([PE38:1+Na+K–H]+ or [d18:1/h24:0–GalCer+Li]+). Images of 245×160 pixels with a pixel size of 50 μm. The value of intensity (Ι) indicated in white in each image corresponds to the minimum and the maximum intensities in a pixel. MALDI–MS spectra in the positive ion mode acquired: i from the cerebral cortex and j from the corpus callosum

mass resolution of electrostatic gates used for this purpose [39]. Various lipid species have very close molecular masses making them extremely difficult to be selected as single precursor ion. Overlapping isotopic peaks of lipids are frequently observed, leading to mixed MS/MS spectra, which cannot be easily interpretable. A reduction to less than 1 amu for the precursor ion window selection greatly reduces the sensitivity, with the consequence that no MS/MS spectrum can be acquired for low intensity ions. Nevertheless, five different molecular species were successfully identified directly from a rat cerebellum tissue section. The mass-tocharge ratio values of the selected precursor ions were m/z 734.6 ([PC diacyl-32:0+H]+; Fig. 7a), m/z 740.6 ([PC diacyl32:0+Li]+; Fig. 7b), m/z 766.6 ([PC diacyl-34:1+Li]+; Fig. 7c); m/z 768.6 ([PC diacyl-34:0+Li]+; Fig. 7d), m/z 816.7 ([d18:1/24:1–GalCer+Li]+; Fig. 7e), and m/z 834.7 ([d18:1/h24:0–GalCer+Li]+; Fig. 7f), respectively.

Under collision-induced dissociation (CID) conditions, the [M+H]+ ion of glycerophosphatidylcholine lipids yields a prominent fragment ion signal at m/z 184, corresponding to the protonated phosphocholine and a minor signal at m/z 86 characteristic of the choline head group (Fig. 7a). The loss of trimethylamine leads to a fragment ion [M+H-59]+. Other structurally relevant fragment ions are of low abundance thus making difficult the attribution of the complete structure. Thus tandem mass spectra of protonated phosphatidylcholine species are dominated by phosphocholine ion peaks and do not contain peaks that could identify the fatty acid substituents [40, 41]. The spectrum in Fig. 7b shows that the selection of a lithium-cationized species promotes structurally relevant fragmentations confirming the advantage of using lithium for MS/MS analysis of lipids. In Fig. 7b, the product ion spectrum of m/z 740.6 exhibits prominent ion peaks at m/z

C.D. Cerruti et al.

Fig. 6 MALDI–TOF ion images, mass spectra, and optical image of a sagittal rat brain section in the positive ion mode. Matrix solutions CHCA mixed with lithium trifluoroacetate. a m/z 740.6 ([PC32:0+Li]+), b m/z 766.6 ([PC34:1+Li]+), c m/z 737.5 ([SM18:0+Li]+), d optical image of the sagittal rat brain section, e m/z 806.5 ([PE38:4+K]+ or [PC38:6+H]+), f m/z 812.6 ([PC38:3+H]+, [PC38:6+Li]+, [PE38:1+K]+, or [PE38:4+2Na–H]+), g m/z 816.7 ([PE42:8+H]+ or [d18:1/24:1–

GalCer+Li]+), h m/z 834.6 ([PE38:1+Na+K–H]+ or [d18:1/h24:0– GalCer+Li]+). Images of 245×160 pixels with a pixel size of 50 μm. The value of intensity (Ι) indicated in white in each image corresponds to the minimum and the maximum intensities in a pixel. MALDI–MS spectra in the positive ion mode acquired: i from the cerebral cortex and j from the corpus callosum

681.6, 557.6, and 551.6 which are corresponding to the loss of trimethylamine, phosphocholine and lithium salt of phosphocholine, respectively. This indicates that ion at m/z 740.6 is a lithium-cationized PC. This is also confirmed by the presence in the low m/z range of the fragment ion at m/z

86 and 190, which are characteristic of choline and lithium-cationized phosphocholine, respectively. Fatty acid chain lengths are determined by the presence of ion signals at m/z 484.4 and 478.4 corresponding to the loss of palmitic acid C16:0 and lithium salt of palmitic acid.

Table 2 Summary of results obtained with each lithium salt, considering ease of matrix deposition with the robotic sprayer, aspect of matrix deposition, lithium cationization and imaging

Ease of matrix deposition with the robotic sprayer Aspect of matrix deposition Lithium cationization Imaging a

LiCitrate

LiAc

LiCl

Lil

LiTFA

−a Irregular ND −

+ Irregular − −

++ Homogeneous − +

−a Homogeneous + ++

+ Homogeneous ++ +++

Under these conditions, the mixture of matrix and lithium salt precipitated in the capillary nebulizer and clogged it

ND not determinated

MALDI imaging mass spectrometry of lipids by adding lithium salts

A

B 4000

[PC(diacyl-32:0)+H]

+

3500

+

[PC(diacyl-32:0)+Li]

1000

681.6

551.6 557.7

740.6

676.3 734.6

86.1

500

190.1 239.3

2000

1000

484.4

1500

3000

425.4

Intensity

2000

86.1

4000

2500

478.4

184.1

5000

3000

0

0 0

100

200

300

400

500

600

700

0

800

100

200

300

400

m/z

C

D

+

6000

577.6

[PC(diacyl-34:1)+Li]

10000

500

600

700

800

m/z

[PC(diacyl-34:0)+Li]

+

768.6

Intensity

6000

5000

8000

X5

X2 x5

300

400

500

600

700

800

0

100

200

300

1800

400

500

[d18:1/24:1 -GalCer+Li]

F 1200

+

654.4

1600

709.6 +

800

187.1

200

0

451.4

600

400

468.5

398.4 416.4

600

636.5

606.5

800

672.7 624.6

334.3

Intensity

39 1000

187.1

800

1000

1200

200

700

[d18:1/h24:0 -GalCer+Li] 468.3

1400

400

600

m/z

636.5 654.5

200

496.5

100

m/z

Intensity

579.7

0 0

E

585.7

239.3

1000

0

425.4

2000

484.4

86.1

3000

478.4

766.6

707.6

583.6

478.4 504.4 510.4

425.4

239.3

190.2

4000

2000

Intensity

484.4

6000 86.1

Intensity

4000

0 0

100

200

300

400

500

600

700

800

m/z

0

100

200

300

400

500

600

700

800

m/z

Fig. 7 Product ion spectra from the precursor ions: a PC diacyl-32:0 [M+H]+ ion (m/z 734.6), b PC diacyl-32:0 [M+Li]+ ion (m/z 740.6), c PC diacyl-34:1 [M+Li]+ ion (m/z 766.6), d PC diacyl-34:0 [M+Li]+

ion (m/z 768.6), e d18:1/24:1–GalCer [M+Li]+ ion (m/z 816.7), f d18:1/h24:0–GalCer [M+Li]+ ion (m/z 834.7). Collision energy was set at 2 keV, and air was used as collision gas

Finally, the ion at m/z 425.4 is characteristic of a consecutive loss of trimethylamine and palmitic acid and the ion at m/z 239.2 correspond to the acyl ion derived from palmitic acid. Thus, the m/z 740.6 ion can be attributed to a lithiumcationized PC (diacyl-16:0/16:0) species. Figure 7c shows the fragment ion spectrum of the m/z 766.6 precursor ion. Fragment ions at m/z 707.6 ([M+Li59]+), m/z 583.6 [M+Li-183]+, m/z 577.6 ([M+Li-189]+),

m/z 425.4 ([M+Li-59–C17H33CO2H]+), and m/z 239.3 ([C15H31CO]+), already described in Fig. 7b, are detected. The ion at m/z 484 ([M+Li–C 17 H 33CO 2H] + ) could correspond to a neutral loss of oleic acid C18:1. The spectrum presents a higher abundance of the ion at m/z 484 from loss of 18:1 fatty acid substituent than the ion at m/z 504 ([M+Li–C15H31CO2Li]+) and m/z 510 ([M+Li– C15H31CO2H]+) corresponding to the loss of the 16:0 fatty

C.D. Cerruti et al.

acid substituent. Thus the precursor ion at m/z 766.6 can be attributed to [PC (diacyl-16:0/18:1)+Li]+. Figure 7d shows the fragment ion spectrum of the m/z 768.6 precursor ion. Ion signals at m/z 709.6 ([M+Li-59]+), m/z 585.6 ([M+Li-183]+), and m/z 579.6 ([M+Li-189]+) indicates the presence of a lithium-cationized PC, that is confirmed by fragment ion at m/z 86. The fragment ion at m/z 239.2 ion indicates the acylium fragment [C15H31CO]+ which is characteristic of a palmitic acid C16:0. The ions at m/z 484.4 and m/z 478.4 sign the loss of stearic acid C18:0 and its lithium salt, respectively. The ion at m/z 425.4 [M+ Li-59–C17H35CO2H]+ involves first a loss of trimethylamine, then an additional loss of stearic acid C18:0 (C17H35CO2H). The m/z 768.6 ion can thus be attributed to the lithiumcationized PC (diacyl-16:0/18:0) species. The product ion spectrum of the ion at m/z 816.7 which is shown in Fig. 7e yields two prominent fragment ions at m/z 654.3 and at m/z 636.4 corresponding to neutral loss of a hexose moiety ([M+Li–C6H10O5]+ and [M+Li–C6H12O6]+, respectively). The ion at m/z 187 corresponding to a lithiated hexose confirms the osidic moiety of this compound. The ion at m/z 636.4 which is the lithiated oxethane intermediate eliminates HCHO to form ions at m/z 606.5. The m/z 654.3 lithiated ceramide cation also yields the m/z 416.4 ion by loss of the long chain base as an aldehyde [CH3(CH2)12CH= CH–CHO]. This is followed by loss of H2O leading to the ion at m/z 398.4. So the precursor ion at m/z 816.7 can be assigned to a lithium-cationized 1-galactosyl-N-tetracosanoyl-sphingosine (d18:1/24:0–GalCer) [42]. The product ion spectrum of the m/z 834.7 precursor ion is shown in Fig. 7f. The fragmentation of this compound yields to ions at m/z 672.7 ([M+Li–C6H10O5]+), m/z 654.4 ([M+Li– C6H12O6]+), m/z 636.4 ([M+Li–C6H12O6–H2O]+), m/z 624.6 ([M+Li–C6H12O6–CH2O]+), and m/z 187.0 ([C6H12O6+Li]+) which allow identifying a galactose moiety. The cleavage of the amide bond results in the formation of a prominent ion at m/z 468.2 ([M+Li–CO–C22H45CHO]+), which eliminates NH3 to give the m/z 451.7 ion. The fragment ion at m/z 496.7 is specific of the N-hydroxyacyl-sphingenine subclass and further dissociate into the m/z 334.7 ion by loss of the sugar moiety. Thus the precursor ion at m/z 834.7 can be assigned to the lithium-cationized 1-galactosyl-N-α-hydroxytetracosanoylsphingosine (d18:1/h24:0–GalCer) [42]. Unfortunately, for ions at m/z 806.6 and 812.6, MS/MS spectra indicate the presence of mixed different species and did not allow giving a unique structural attribution.

Conclusions Mass spectrometry imaging can be used to determine the location of different molecular lipid species and to elucidate the precise fatty acid composition of biological membranes

in different tissues. In this paper, a protocol for MALDI– MSI detection of phosphatidylcholines and galactosylceramides in positive ion mode by adding lithium salt to the matrix solution was optimized. Five different lithium salts (LiAc, LiTFA, LiCitrate, LiCl, and LiI) were tested at different concentrations. A 2 mg mL−1 concentration of LiTFA leads to total lithium adduct formation for PC (diacyl16:0/16:0) and PC (diacyl-16:0/18:1) while providing a gain in sensitivity. Its use in imaging at a concentration of 1 mg mL−1 does improve the image contrast and does not cause delocalization of the molecules of interest. The presence of alkali metal salts in the matrix solution facilitates the structural identification of lipids by increasing the fragmentation yield during CID experiments, thus making possible the structural elucidation of species from a single peak. Ions at m/z 740.6 [PC 34:0 (diacyl-16:0/16:0)+Li]+, m/z 766.6 [PC 34:0 (diacyl-16:0/18:1)+Li]+, m/z 768.6 [PC 34:0 (diacyl-16:0/ 18:0)+Li]+, m/z 816.7 [d18:1/24:1–GalCer+Li]+, and m/z 834.7 [d18:1/h24:0–GalCer+Li]+ could be unequivocally identified thanks to the presence of lithium ions in the matrix solution. However, MALDI–TOF/TOF mass spectrometers do not have a sufficiently narrow precursor ion selection window to be able to choose only one single ion of interest among all the various lipid species available at the surface of a tissue section. Instrumental work remains to be performed to identify other lipid classes by on-tissue tandem mass spectrometry. Acknowledgements C.D.C. and V.W.P. are indebted to the Institut de Chimie des Substances Naturelles (CNRS) for a Ph.D. research fellowship and a post-doctoral grant, respectively.

References 1. Surette ME, Winkler JD, Fontech AN, Chilton FH (1996) Biochemistry 35:9187–9196 2. Greenspan P, Mayer EP, Fowler SD (1985) J Cell Biol 100:965– 973 3. Stoeckli M, Staab D, Staufenbiel M, Wiederhold KH, Signor L (2002) Anal Biochem 311:33–39 4. Caprioli RM, Farmer TB, Gile J (1997) Anal Chem 69:4751–4760 5. Cornett DS, Reyzer ML, Chaurand P, Caprioli RM (2007) Nat Method 4:828–833 6. Reyzer ML, Caprioli RM (2007) Curr Opin Chem Biol 11:29–35 7. Francese S, Dani FR, Traldi P, Mastrobuoni G, Pieraccini G, Moneti G (2009) Comb Chem High Throughput Screen 12:156– 174 8. Chaurand P, Schwartz SA, Billheimer D, Xu BJ, Crecelius A, Caprioli RM (2004) Anal Chem 76:1145–1155 9. Schwamborn K, Caprioli RM (2010) Nat Rev Cancer 10:639–646 10. Van Remoortere A, Van Zeijl RJ, Van Den Oever N, Franck J, Longuespée R, Wisztorski M, Salzet M, Deelder AM, Fournier I, Mc Donnell LA (2010) J Am Soc Mass Spectrom 21:1922– 1929 11. Benabdellah F, Touboul D, Brunelle A, Laprévote O (2009) Anal Chem 81:5557–5560

MALDI imaging mass spectrometry of lipids by adding lithium salts 12. Nilsson A, Fehniger TE, Gustavsson L, Andersson M, Kenne K, Marko-Varga G, Andrén PE (2010) PLoS ONE 5:e11411 13. Roy S, Touboul D, Brunelle A, Germain DP, Prognon P, Laprévote O, Chaminade P (2006) Ann Pharm Fr 64:328–334 14. Garrett TJ, Prieto-Conaway MC, Kovtoun V, Bui H, Izgarian N, Stafford G, Yost RA (2007) Int J Mass Spectrom 260:166–176 15. Shimma S, Sugiura Y, Hayasaka T, Hoshikawa Y, Noda T, Setou M (2007) J Chromatogr B 855:98–103 16. Guenther S, Römpp A, Kummerb W et al. (2011) Int J Mass Spectrom, doi:10.1016/j.ijms.2010.11.011 17. Brunelle A, Laprévote O (2009) Anal Bioanal Chem 393:31–35 18. Jackson SN, Wang HY, Woods AS (2005) J Am Soc Mass Spectrom 16:2052–2056 19. Sugiura Y, Setou M (2009) Rapid Commun Mass Spectrom 23:326–3278 20. Hsu FF, Bohrer A, Turk J (1998) J Am Soc Mass Spectrom 9:516–526 21. Hsu FF, Turk J (2003) J Am Soc Mass Spectrom 14:352–363 22. Hsu FF, Turk J, Zhang K, Beverley SM (2007) J Am Soc Mass Spectrom 18:1591–1604 23. Hsu FF, Turk J (2008) J Am Soc Mass Spectrom 19:1681–1691 24. Hsu FF, Turk J (2010) J Am Soc Mass Spectrom 21:657–669 25. Schwartz SA, Reyzer ML, Caprioli RM (2003) J Mass Spectrom 38:699–708 26. Lemaire R, Menguellet SA, Stauber J, Marchaudon V, Lucot JP, Collinet P, Farine MO, Vinatier D, Day R, Ducoroy P, Salzet M, Fournier I (2007) J Proteome Res 6:4127–4134

27. Seeley EH, Oppenheimer SR, Mi D, Chaurand P, Caprioli RM (2008) J Am Soc Mass Spectrom 19:1069–1077 28. Lemaire R, Tabet JC, Ducoroy P, Hendra JB, Salzet M, Fournier I (2006) Anal Chem 78:809–819 29. Shimma S, Furuta M, Ichimura K, Yoshida Y, Setou M (2006) J Mass Spectrom Soc Jpn 54:133–140 30. Touboul D, Brunelle A, Laprévote O (2010) Biochimie 93:113–119 31. Benabdellah F, Seyer A, Quinton L, Touboul D, Brunelle A, Laprévote O (2010) Anal Bioanal Chem 396:151–162 32. Sadeghi M, Vertes A (1998) Appl Surf Sci 129:226–234 33. Vaidyanathan S, Winder CL, Wade SC, Kell DB, Goodacre R (2002) Rapid Commun Mass Spectrom 16:1276–1286 34. Onnerfjord P, Ekstrom S, Bergquist J, Nilsson J, Laurell T, Marko-Varga G (1999) Rapid Commun Mass Spectrom 13:315– 322 35. Beavis RC, Bridson JN (1993) J Phys D Appl Phys 26:442 36. Sugiura Y, Shimma S, Setou M (2006) Anal Chem 78:8227–8235 37. Mock KK, Sutton CW, Cottrell JS (1992) Rapid Commun Mass Spectrom 6:233–238 38. Sugiura Y, Konishi Y, Zaima N, Kajihara S, Nakanishi H, Taguchi R, Setou M (2009) J Lipid Res 50:1776–1788 39. Medzihradszky KF, Campbell JM, Baldwin MA, Falick AM, Juhasz P, Vestal ML, Burlingame AL (2000) Anal Chem 72:552– 558 40. Murphy RC, Harrison KA (1994) Mass Spectrom Rev 13:57–75 41. Pulfer M, Murphy RC (2003) Mass Spectrom Rev 22:332–364 42. Hsu FF, Turk J (2001) J Am Soc Mass Spectrom 12:61–79

Suggest Documents