Spatial mapping of phosphatidylcholine sn-positional isomers using CID of divalent metal complexes in imaging mass spectrometry

Abstract

Phosphatidylcholines (PCs) are the main components of cellular membranes. The high degree of structural heterogeneity leads to significant variations in PC functions and complicates structural characterization. For example, the complex mixtures of lipid structures create challenges when analyzing and identifying these compounds directly from tissue in matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry experiments. Phosphatidylcholine (PCs) are preferentially ionized in the positive ion mode in MALDI imaging mass spectrometry. However, low-energy collision induced dissociation (CID) of protonated PCs largely only results in cleavages of the phosphocholine headgroup, with little to no information obtained about the fatty acyl chain identities and positions. Alternatively, metal cationization of lipids is known to generate increased structural information upon CID, but metal coordination has been less studied. Herein, we highlight the use of divalent metal-ligand complexes to produce new ion types for CID analysis in MALDI imaging mass spectrometry. CID of the new [PC + M + ligand]⁺ ion type (where M is a divalent metal) eliminates the headgroup loss fragmentation channel and opens new fragmentation channels at the fatty acyl chain positions. The gas-phase fragmentation behavior of [PC + M + ligand]⁺ ion type is characterized using multiple divalent metals and ligands. The fatty acyl chain product ions are then used to relatively quantify sn-positional isomers. Furthermore, this method is integrated into an imaging mass spectrometry workflow to enable the spatial mapping of PC sn-positional isomers in rat brain and glioblastoma tissues, revealing differential distributions of the sn-positional isomers.

INTRODUCTION

Matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry has been widely used to visualize lipids directly from biological tissues, providing molecular context to tissue biology and pathology. ^{1–4 1–4} During typical imaging mass spectrometry experiments, analytes are ionized directly from tissue and a microprobe raster sampling of the surface results in images of ion intensities across thin tissue sections.^{5, 6} The resulting ion images are visualized for select mass-charge-ratio (m/z) values of analytes of interest. However, accurate identification and mapping of lipids in imaging mass spectrometry is challenging due to the complex chemical environment of biological tissues, which creates in a multitude of isomers at each m/z value that results in composite ion images.^{7, 8} These composite ion images represent imprecise molecular identifications and inaccurate spatial distributions, thus hindering the understanding of localized molecular lipid biology within the tissue microenvironment. Separation of lipid isomers has been achieved using ion mobility coupled to mass spectrometry in some reports, though subsequent interrogation of the separated lipids is necessary to elucidate chemical structures.^{9, 10} Differential distributions of lipid isomers have been measured in a few studies and are indicative of altered metabolism in health and diseases,^11–14 necessitating further improvements in specificity in imaging mass spectrometry to separate and identify these lipid isomers.

Tandem mass spectrometry (MS/MS) has been widely used to improve the separation and identification of analytes in imaging mass spectrometry experiments.^15–17 Collision induced dissociation (CID) has been historically used in imaging mass spectrometry to improve the molecular specificity.^18–20 However, CID analyses of lipids typically produce limited structural information. For example, low-energy CID of protonated phosphatidylcholines (PCs) largely cleaves the phosphocholine headgroup, providing no information on the double bond and fatty acyl tail positions within the lipid. Recently, a variety of gas-phase and condensed-phase reactions have been used in imaging mass spectrometry experiments to differentiate lipid isomers.^{17, 21} For example, the ion/molecule reaction ozone-induced dissociation (OzID) has been shown to resolve PC sn-positional and double bond isomers in MALDI imaging mass spectrometry experiments.^{22, 23}Isomeric distributions of acidic lipids including phosphatidylserines (PSs) and phosphatidylethanolamines (PEs) have been measured using online (i.e., pixel-by-pixel during the imaging experiment) OzID.²⁴ OzID can also be enabled directly on the tissue prior to the imaging experiment through on-tissue ozonization.²⁵ Similarly, on-tissue Paternò–Büchi (PB) reactions have revealed double bond isomer distributions of PCs and PSs.²⁶ PB reactions have also recently been enabled using a reactive MALDI matrix.^{27, 28} This method uses an intrinsically photoreactive Paternò–Büchi compound to react with unsaturation sites during MALDI laser irradiation, enabling the localization of double bonds in PCs. Reactions with singlet oxygen have also been used to reveal spatial distributions of PC double bond isomers in nano-DESI imaging mass spectrometry.²⁹ Alternative ion activation techniques such as ultraviolet photodissociation (UVPD)^{30, 31} and electron induced dissociation (EID)^32–35 have been used to visualize the spatial distributions of PC and fatty acid double bond isomers. Radical directed dissipation (RDD),³⁶ sliver cationization,^{37, 38} and epoxidation with meta-chloroperoxybenzoic acid (m-CPBA)³⁹ have been developed to map the isomeric distributions of lipids. Our group has recently described the use of charge inversion ion/ion reactions to alter ion types in the gas-phase prior to CID in order to reveal isomeric and isobaric distributions of lipids.^{40, 41} For example, a charge inversion reaction between 1,4-phenylenedipropionic acid (PDPA) dianions and PC monocations produced demethylated anions that allow for the assignment of the fatty acyl chain positions upon subsequent CID.^{42, 43} We have also profiled fatty acid double bond isomers using charge inversion ion/ion reactions between fatty acid monoanions and magnesium tris-phenanthroline dications in imaging mass spectrometry analysis of brain and skin tissues.⁴⁴

Each of the MS/MS techniques described above alters either the manner by which energy is imparted into the ion (e.g., UVPD, EID) or the ion type that is subjected to ion activation (e.g., PB reactions, epoxidation, ion/ion reactions). Condensed-phase metal adduction of lipids is a relatively simple approach to changing the ion type. CID of doubly charged lipid-metal ion complexes formed using Fe, Mg, Cu, Ba, Ca, Mn, and Co metals has enabled identification of fatty acyl chain positions,^{45, 46} double bond positions,^{47, 48} double bond orientation,⁴⁹ and methyl branching positions⁵⁰ of lipids in electrospray ionization (ESI) experiments. However, alterations in lipid fragmentation upon metal coordination to other ligands are much less studied. Herein, we report the formation and CID behavior of the singly charged [PC + metal + ligand] ⁺ ion types formed via combinations of positively charged divalent metals and negatively charged ligands in MALDI mass spectrometry. Different metal ions and ligands, including Mg²⁺, Mn²⁺, Ca²⁺, Co²⁺ and Cl⁻, NO₃⁻, CH₃COO⁻ are used to form [PC + metal + ligand]⁺ complexes. CID of these different complexes is described and enables differentiation and relative quantification of PC sn-positional isomers. Application of this approach in MALDI imaging mass spectrometry experiments allows for the visualization of sn-positional isomers in the rat brain tissues.

EXPERIMETNAL

Materials

HPLC-grade water, HPLC-grade acetonitrile, HPLC-grade acetone, manganese nitrate tetrahydrate, cobalt (II) nitrate hexahydrate, calcium nitrate tetrahydrate, and calcium chloride dihydrate were purchased from Fisher Chemical (Waltham, MA, USA). 1,5-diaminonaphthalene (DAN), magnesium chloride anhydrous, magnesium acetate tetrahydrate, magnesium nitrate hexahydrate, manganese chloride tetrahydrate, manganese acetate tetrahydrate, and ammonium acetate were purchased from Sigma-Aldrich (St. Louis, MO, USA). Rat brain tissue was purchased from BioIVT (Westbury, NY, USA) and stored at −80 °C. All lipid standards including isomeric pure PC 16:0/18:1(n-9) and PC 18:1(n-9)/16:0 were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL, USA).

Glioblastoma Tumor Cell Culture

Glioblastoma cells from the established L0 line were cultured and propagated using the neurosphere assay (NSA), following previously published protocols.^{51, 52} Briefly, single cells from L0 were seeded at a density of 5×10⁴ cells/ml in complete human NeuroCult^™ NSA medium, comprising a 9:1 mixture of human NeuroCult^™ NSA Basal Medium and NeuroCult^™ Proliferation Supplement, supplemented with 20 ng/ml epidermal growth factor (EGF), 10 ng/ml basic fibroblast growth factor (bFGF), and 1 μl/ml of 0.2% heparin (2 μg/ml). The cells were maintained in appropriate tissue culture vessels in a 37°C incubator with 5% CO₂. Once the gliomaspheres reached an average diameter of 200 μm, the contents of each flask were transferred to sterile culture tubes and centrifuged at 800 rpm (110 g) for 5 minutes at room temperature. The supernatant was discarded, and the cell pellet was resuspended in 1 ml of 0.05% trypsin-EDTA and incubated at 37°C in a water bath for 2–3 minutes. To quench the trypsin, an equal volume of soybean trypsin inhibitor was added. The cell suspension was then gently pipetted to achieve a single-cell suspension. All tissue culture reagents were obtained from STEMCELL Technologies, Vancouver, BC, Canada.

Intracranial Tumor Implantation

Intracranial xenograft implants of L0 glioblastoma cells were performed in 8-week-old female NOD/SCID mice via stereotactic surgery under isoflurane anesthesia. Mice were obtained from Charles River Laboratories and maintained according to standard murine husbandry protocols. All procedures adhered to the NIH Guide for the Care and Use of Laboratory Animals and were approved by the University of Florida Institutional Animal Care and Use Committee (protocol #201701502). Briefly, 2×10⁵ tumor cells, resuspended in PBS, were injected into the right striatum (coordinates: 2 mm lateral, 2.5 mm dorsal-ventral) following previously published methods.^{51, 53, 54} Animal health was monitored daily post-implantation to ensure normal function. Mice were humanely euthanized upon reaching endpoint criteria, which included >20% body weight loss, a body condition score ≤2, or the development of neurological deficits. Brain tissues were snap-frozen in liquid nitrogen and stored at −80°C for later analysis.

Sample preparation

Synthetic lipid standards solutions were prepared at 1 mg/mL (methanol) and mixed 2:1:1 (v/v/v) ratio with a 20 mM metal solution (water) and a 25 mg/mL DAN matrix solution (acetonitrile). One microliter of the resulting lipid/metal/matrix solutions were manually spotted on an MTP AnchorChip MALDI target plate (Bruker Daltonics, Billerica, MA, USA) and allowed to dry.

Tissues sections were prepared at 10 μm thickness using a Leica CM 3050S cryomicrotome (Leica Biosystems, Buffalo Grove, IL, USA). The section was then thaw-mounted onto conductive indium tin oxide (ITO) coated glass slides (Delta Technologies, Loveland, CO, USA). Washes with chilled (4°C) ammonium acetate (150 mM) were used to remove the endogenous salts.⁵⁵ Briefly, 1 mL of the chilled ammonium acetate solution was deposited on the slide for 10 s and then nitrogen gas was used to remove the excess solution. Manganese acetate tetrahydrate and manganese chloride tetrahydrate salt solutions were prepared at concentrations of 5 mg/mL in 70% acetone. The salt solutions were then sprayed onto the tissue sections using a M5 TM Sprayer (HTX Technologies, LLC, Chapel Hill, NC, USA) with the following the spray conditions: spray temperature of 85°C, plate temperature of 40°C, four passes using a criss-cross pattern, flow rate of 0.1 mL/min, spray velocity of 1,200 mm/min, nitrogen flow of 10 psi, and track spacing of 2.5 mm. Following salt application, the DAN matrix was prepared at a concentration of 10 mg/mL in 70% acetone and was then applied to the tissue section using the TM sprayer under the same conditions.

Mass spectrometry

All mass spectrometry experiments were conducted on a hybrid 7T solariX XR FT-ICR mass spectrometer (Bruker Daltonics, Billerica, MA) in positive ion mode. The MALDI source employs a smartbeam II Nd:YAG laser system (2 kHz, 355 nm). CID mass spectra were acquired from m/z 100 to 2,000 with a 0.4893 s time-domain transient length resulting a mass resolving power of approximately 55,000 at m/z 478. Beam-type CID collision voltages ranging from 22-28 V were used to fragment the lipid-metal ligand-complexes. MALDI experiments were performed with smart walk enabled using a raster size of 200 μm, a 60% global laser power, and 200 laser shots. ESI experiments used a capillary voltage set to 4,500 V, a nebulizer pressure of 3 bar, a dry gas flow rate of 4.0 L/min, a dry gas temperature of 200 °C, and a sample flow rate of 150 μL/h.

Imaging mass spectrometry

Imaging mass spectrometry experiments of rat brain and glioblastoma tissues were performed in positive ion mode. For PC 34:1 in rat brain tissue, ion images were acquired using a pixel spacing of 100 μm in both the x and y dimensions with a laser power of 45% and 500 laser shots. For PC 34:1 in glioblastoma tissue, ion images were acquired using a pixel spacing of 150 μm in both the x and y dimensions with a laser power of 45% and 500 laser shots. For PC 36:1, ion images were acquired using a pixel spacing of 200 μm in both the x and y dimensions with a laser power of 60% and 800 laser shots. CID voltages were set to 28V and 24V for PC 34:1 and PC 36:1, respectively, in rat brain tissue and 28V for PC 34:1 in glioblastoma tissue. CID imaging raw data files were converted to .imzML in flexImaging 5.0 (Bruker Daltonics, Billerica, MA, USA). Isomer ion images were generated and visualized in Python using in-house scripts. Tissue sections were collected after imaging mass spectrometric analysis for hematoxylin and eosin (H&E) staining using a standard protocol. H&E stained tissues were scanned at 10× magnification using an Axio imager M2 Microscope (Carl Zeiss, Jenna, Germany) and visualized using Zen microscopy software (Carl Zeiss, Jenna, Germany).

RESULTS

CID of [PC + metal + ligand]⁺ ion types

Phosphatidylcholines (PCs) are preferentially ionized in the positive ion mode in imaging mass spectrometry experiments. CID analysis of [PC + H]⁺ tends to predominantly give rise to cleavage of the phosphocholine headgroup, though fatty acyl chain fragment ions can be observed in low relative abundances when using high energy CID (Figure 1a). Manipulating the lipid ion types by metal adduction is a simple method to alter the fragmentation behavior upon CID. For example, sodium and potassium-adducted PCs are commonly formed in MALDI and CID of these ion types typically provides more abundant fatty acyl chain fragmentation, though this fragmentation is still far less favorable than cleavage of the phosphocholine headgroup.⁵⁶ Recently, divalent metal adduction has been used in ESI to produce doubly charged lipid-metal ion complexes that allow for detailed structural information upon CID, such as in the analysis of branched chains and fatty acyl chain positions.⁵⁰ For example, CID of ESI formed [PC 16:0/18:1 + Mg]²⁺ (m/z 391.78) generates abundant fatty acyl chain fragment ions (Figure 1b). Fragment ions at m/z 504.34 and 478.33 arise from the neutral loss of the 16:0 fatty acid anion and neutral loss of the 18:0 fatty acid anion along with Mg. Here, we have studied the use of divalent metals in MALDI experiments and have used divalent metal salts to form novel lipid ion types, lipid-divalent metal-ligand complexes of the generic form [PC + metal + ligand]⁺. CID of the complexes reveals new and abundant fatty acyl chain fragmentation channels, confirming the identities and the positions of the PCs. For example, a [PC 16:0/18:1 + Mg + NO₃]⁺ lipid-divalent metal-ligand complex was formed by adding Mg(NO₃)₂•6H₂O into the lipid and matrix solutions. In contrast to CID of protonated PCs that mainly produces a phosphocholine fragment ion (Figure 1a), CID of [PC 16:0/18:1 + Mg + NO₃]⁺ generates two fatty acyl chain fragment ions at m/z 478.33 and 504.34 derived from the neutral loss of the 18:1 or 16:0 fatty acid anion along with the Mg and the ligand. (Figure 1c). No phosphocholine fragment ion is observed, suggesting that CID of this new ion type opens new fragmentation channels and enables facile identification of the fatty acyl chain identities. Notably, the loss of sn-2 fatty acyl chain is favored in [PC 16:0/18:1 + Mg]²⁺ (Figure 1b), while sn-1 fatty acyl chain is more readily lost in the [PC 16:0/18:1 + Mg + NO₃]⁺ ion type (Figure 1c), suggesting that the addition of the anionic ligand changes the fragmentation chemistry. CID of the isomeric [PC 18:1/16:0 + Mg + NO₃]⁺ was also performed (Figure 1d) and shows different relative abundances of the two fragment ions at m/z 478.33 and 504.34, again consistent with more favorable loss of sn-1 fatty acyl chain. CID of the ESI and MALDI formed [PC 16:0/18:1+ Mg + NO₃]⁺ complex is performed and shown to generate the same fatty acyl chain fragment ions (Figure S1).

Figure 1.

CID analyses of (a) [PC 16:0/18:1 + H]⁺ (CID: 24 V) generated via MALDI, (b) [PC 16:0/18:1 + Mg]²⁺ (CID: 11 V) generated via ESI, (c) [PC 16:0/18:1 + Mg + NO₃]⁺ (CID: 22 V) generated via MALDI, and (d) [PC 18:1/16:0+ Mg + NO₃]⁺ (CID: 22 V) generated via MALDI are performed in positive ion mode. Mg metal adduction gives rise to more abundant fatty acyl chain loss, and the addition of the NO₃⁻ ligand changes the preferential loss of the fatty acid. The orange triangle denotes the precursor ion.

The fragmentation behaviors of lipid-Mg-ligand complexes were studied using different ligands. The use of nitrate, chloride, and acetate ligands allows formation of [PC 16:0 /18:1 + Mg + NO₃]⁺, [PC 16:0 /18:1 + Mg + Cl]⁺, and [PC 16:0/18:1 + Mg + acetate]⁺ ion types, respectively. CID of [PC 16:0/18:1 + Mg +NO₃]⁺ and [PC 16:0/18:1 + Mg + Cl]⁺ generates abundant fatty acyl chain fragment ions at m/z 478.33 and 504.34 (Figure 2a and Figure 2b). The m/z 478.33 and 504.34 fragment ions arise from the combined loss of the metal and ligand with either the 18:1 or 16:0 fatty acid anion, respectively. Similar fragmentation behavior is observed for [PC 16:0/18:1 + Mg + acetate]⁺, resulting in fragment ions m/z 478.33, 504.34, 560.32, and 586.34 (Figure 2c). Product ions m/z 560.32 and 586.34 are formed via neutral loss of the 18:1 and 16:0 fatty acids from the precursor ion, respectively. Notably, the neutral loss of acetic acid also produces a [PC 16:0/18:1 + Mn − H]⁺ fragment ion at m/z 782.56. Overall, CID of lipid ion types containing the same metal ion (Mg), but with different ligands Cl⁻ , NO₃⁻, and CH₃COO⁻ gives rise to similar fragmentation pathways arising from most favorable loss of the sn-1 fatty acyl chain. These results are in contrast with CID of [PC 16:0/18:1 + Mg]²⁺, for which the sn-2 fatty acyl chain is cleaved more readily (Figure 1b). These results suggest that the presence of a ligand can significantly alter the preferred fragmentation chemistry.

Figure 2.

CID analysis of [PC 16:0/18:1 + metal + ligand]⁺ ion types formed by MALDI for (a) [PC 16:0/18:1 + Mg + NO₃]⁺, (b) [PC 16:0/18:1 + Mg + Cl]⁺, (c) [PC 16:0/18:1 + Mg + acetate]⁺, (d) [PC 16:0/18:1 + Co(II) + NO₃]⁺, (e) [PC 16:0/18:1 + Mn(II) + NO₃]⁺, and (f) [PC 16:0/18:1 + Ca + NO₃]⁺ gives rise to abundant fatty acyl chain cleavages. The orange triangle denotes the precursor ion. All spectra were collected using a collision energy of 22 V. Note that Figure 2a is reproduced from Figure 1c here to enable a more facile comparison of fragmentation chemistry with different metal and ligand ion types.

The fragmentation behaviors of different lipid-divalent metal-ligand complexes were further studied using different metal ions. Mg, Co, Mn, and Ca were used to generate [PC + Mg + NO₃]⁺, [PC 16:1/18:1 + Co(II) + NO₃]⁺, [PC 16:0/18:1 + Mn(II) + NO₃]⁺, and [PC 16:0/18:1 + Ca + NO₃]⁺ ion types via MALDI. CID of these complexes generates similar fatty acyl chain fragmentation pathways showing a preferential loss of the sn-1 fatty acyl chain. Interestingly, the identity of the metal affects the preference for fatty acyl chain loss. For example, CID of [PC 16:0/18:1 + Ca + NO₃]⁺ (Figure 2f) demonstrates a more favorable loss of the sn-2 fatty acyl chain, while the sn-1 fatty acyl chain is preferentially lost for all other complexes (Figure 2a, 2d, 2e). A similar preference for loss of the sn-2 chain is observed for [PC 16:0/18:1 + Ca + Cl]⁺ , suggesting the identity of the metal plays an important role in this preference (Figure S2). The relative abundances of these fragmentation channels are generally consistent across a range of collision energies (e.g., from 18 V to 30 V for [PC 16:0/18:1 + Mn+ acetate]⁺), though this has not been exhaustively explored (Figure S3). Overall, the use of different metal ions and ligands in complexation with PC 16:0/18:1 results in highly abundant fatty acyl chain fragmentation pathways upon CID.

The CID behavior of the lipid metal complexes is influenced by both the use of different metals and the use of the ligands. The presence of a ligand and the identity of metal ions determines the preferential loss of either the sn-1 or sn-2 fatty acyl chains. Different fragmentation behavior is observed for each of the ion types studied here and is likely related to the size and properties of metal cation and the anionic ligand. For example, the use of a Mg²⁺ metal with Cl⁻ or NO₃⁻ ligands results in selective loss of the metal and ligand with either fatty acid chain (Figure 2a and 2b). However, additional fragmentation pathways (i.e., neutral losses of each fatty acid) are observed with a weaker anionic ligand, such as CH₃COO⁻ (Figure 2c). These additional fragmentation pathways are also observed for complexes containing larger radii metals such as Co²⁺ , Mn²⁺ , and Ca²⁺ (Figure 2d, 2e, and 2f). Some combination of a smaller sized metal cation and/or stronger anionic ligand likely contributes to stronger interaction with lipid, which may result limit the accessible fragmentation chemistry. Proposed fragmentation pathways of the combined fatty acid, metal, and ligand losses and corresponding fragment ion structures suggest that either a six-membered (Scheme S1a) or five-membered (Scheme S1b) ring containing the phosphate group is formed, depending on if the loss involves the sn-1 or sn-2 fatty acyl chain, which is consistent with prior CID studies of metal-cationized lipids.^{45, 50} Conversely, neutral loss of only the fatty acyl chain follows a different mechanism involving the two adjacent chains (Scheme S1c and S1d), which has been previously proposed for the preferential loss of sn-1 fatty acyl chains during CID of lithiated lipids due to a more labile α-hydrogen in the sn-2 fatty acyl chain.⁵⁷ While the mechanisms of these cleavages and preferences are still under study, the resulting MS/MS spectra can aid in PC fatty acyl chain identification. Lipid-divalent metal-ligand ion types are observed in similar abundances to protonated ion types in the same experiment, with [PC 16:0/18:1 + Mn(II) + acetate]⁺ demonstrating the highest ionization efficiency (Figure S4) and thus used for subsequent imaging mass spectrometry experiments (vide infra). The limit of detection of this method was determined to be approximately 5 μM using spotted standards (Figure S5).

Differentiation and relative quantification of sn-isomers

CID of the lipid-divalent metal-ligand complexes can be used to differentiate PC sn-positional isomers, as neutral loss of the sn-1 fatty acyl chain is favored and produces a more abundant fragment ion than neutral loss of the sn-2 chain, which is observed for both neutral loss of the fatty acyl chain as well as neutral loss of the chain anion with the metal and ligand attached. For example, CID of [PC 16:0/18:1 + Mn(II) + Cl]⁺ (Figure 3a) and [PC 18:1/16:0 + Mn(II) + Cl]⁺ (Figure 3b) complexes produces fragment ions resulting from losses of the sn-1 and sn-2 fatty acyl chains in different relative intensity ratios. The loss of sn-1 fatty acyl chain is favored, resulting in higher abundances of m/z 504.34 (neutral loss of 16:0 anion with Mn(II) and Cl) and m/z 593.34 (neutral loss of 16:0) for CID of [PC 16:0/18:1 + Mn(II) + Cl]⁺, while m/z 478.33 (neutral loss of 18:1 anion with Mn(II) and Cl) and m/z 567.23 (neutral loss of 18:1) are more abundant for CID of [PC 18:1/16:0 + Mn(II) + Cl]⁺. The different relative abundances of these diagnostic ions allow for the differentiation of the sn-positional isomers. Similar results are obtained using manganese (II) acetate. CID of [PC 16:0/18:1 + Mn(II) + acetate]⁺ (Figure 3d) and [PC 18:1/16:0 + Mn(II) + acetate]⁺ (Figure 3e) also gives rise to preferential neutral loss of the sn-1 fatty acyl chain. The preferential loss of the sn-1 fatty acyl chain could be due to a more labile α-hydrogen in the sn-2 fatty acyl chain, as the α-hydrogen in the sn-2 fatty acyl chain is involved in the release of sn-1 fatty acyl chain (Scheme S1c).⁵⁷ For the neutral losses containing the acyl chain, metal, and ligand, loss of the sn-1 chain may be favored because formation of the six-membered ring intermediate (Scheme S1a) is more favorable than the five-membered ring intermediate, which is formed during loss of the sn-2 chain (Scheme S1b).⁴⁵ The diagnostic ions resulting from fatty acyl chain loss can be used to perform relative quantification of the sn-isomers and is demonstrated using [PC 16:0_18:1 + Mn(II) + Cl]⁺ and [PC 16:0_18:1 + Mn(II) + acetate]⁺. Good linearity is observed for both ion types (R² = 0.9949 in Figure 3c and R² = 0.9976 in Figure 3f, respectively) by measuring the diagnostic ion intensity ratios as a function of the concentrations of lipid sn-positional isomers PC 16:0/18:1 and PC 18:1/16:0. Similar results are observed when using diagnostic ions from both fatty acyl chain losses as well as fatty acyl chain+metal+ligand loss (Figure S6).

Figure 3.

CID analysis of [PC + metal + ligand]⁺ yields favorable sn-1 fatty acyl chain loss, enabling differentiation of sn-positional isomers PC 16:1/18:1 and PC 18:1/16:0 using (a) [PC 16:0/18:1 + Mn(II) + Cl]⁺ (CID: 22 V) and (b) [PC 18:1/16:0 + Mn(II) + Cl]⁺ (CID: 22 V) ion types. (c) A calibration curve is generated using the intensities of the diagnostic ions allows relative quantification of [PC 16:0/18:1 + Mn(II) + Cl]⁺ and [PC 18:1/16:0 + Mn(II) + Cl]⁺. CID analysis is also performed on (d) [PC 16:0/18:1 + Mn(II) + acetate]⁺ (CID: 28 V) and (e) [PC 18:1/16:0 + Mn(II) + acetate]⁺(CID: 28 V) ion types. (f) A calibration curve generated using the intensities of the diagnostic ions allows relative quantification of [PC 16:0/18:1 + Mn(II) + acetate]⁺ and [PC 18:1/16:0 + Mn(II) + acetate]⁺. Standard deviations are calculated from five replicate measurements. The orange triangle denotes the precursor ion.

Mapping sn-positional isomers in rat brain tissue

CID of lipid-divalent metal-ligand complexes was integrated into an imaging mass spectrometry workflow to visualize the distributions of sn-positional isomers in the rat brain cerebellum. Endogenous salts were removed from the tissue using an ammonium acetate wash and then a divalent metal salt solution was applied to the tissue surface using a robotic sprayer. Although the [PC + Mn + acetate]⁺ ion type produced a greater number of fragment ions compared to the [PC + Mg + NO₃]⁺, the higher ionization efficiency of the [PC + Mn + acetate]⁺ ion type still resulted in more sensitive MS/MS imaging analysis. [PC 34:1 + Mn(II) + acetate]⁺ is readily observed at m/z 873.53 in the granular layer and white matter. CID of this ion results in product ions at m/z 478.33, 504.34, 591.27, and 617.29, identifying the fatty acyl chains as 16:0 and 18:1 (Figure 4a and Figure 4b). Fragment ions m/z 504.34 (16:0) and 478.33 (18:1) are derived from the combined loss of fatty acyl chain along with Mn(II) and acetate ligand. Fragment ions m/z 591.27 and 617.29 result from the loss of the 16:0 and 18:1 fatty acids, respectively. Different intensity ratios of these diagnostic ions indicate the presence of varying levels of PC 16:0/18:1 and PC 18:1/16:0 isomers in the granular layer and white matter. Referencing ratios to the calibration curve (Figure 3f) allows for determination of the relative concentrations of the two isomers. The relative concentration of PC 16:0/18:1 varies in different regions of the rat brain cerebellum and is measured to be 32.9 ± 3.5%, 40.7 ± 2.5%, and 45.7 ± 7.1% in the white matter, molecular layer, and granular layer, respectively (Figure 4c). The relative distributions of sn-positional isomers can also be mapped using diagnostic fragment ions. Using the diagnostic ions m/z 478.33 and 504.34, PC 16:0/18:1 is measured to be relatively more abundant in the white matter (Figure 4g), while PC 18:1/16:0 is more localized to the granular layer (Figure 4f). The combination of all diagnostic ions (m/z 478.33, 504.34, 591.27, and 617.29) was also used for isomer mapping, which gives the same results. (Figure S7) These results are also consistent with previous studies.^{33, 43}

Figure 4.

CID analysis of [PC + Mn(II) + acetate]+ (CID: 28V) for PC 34:1 in (a) the granular layer and (b) white matter in rat brain cerebellum. (c) Relative abundances of the PC 16:0/18:1 and PC 18:1/16:0 isomers are determined using diagnostic fragment ions. The error bars represent the relative standard deviations from five replicate measurements. (d) An H&E stain allows localization of (e) PC 34:1 to structures of the rat brain cerebellum. (f) The relative distribution of PC 16:0/18:1 to PC 18:1/16:0 is calculated using the intensity ratio of m/z 504.34 to the summed intensity of m/z 478.33 and 504.34 and (g) the relative distribution of PC 18:1/16:0 to PC 16:0/18:1 is calculated using the intensity ratio of m/z 478.33 to the summed intensity of m/z 504.34 and 478.33. Pixel size is 100 μm for all the ion images. The orange triangles in panels (a) and (b) denote the precursor ions.

This workflow was also used to study PC 36:1 in rat brain cerebellum (Figure 5a). [PC 36:1 + Mn(II) + Cl]⁺ is readily observed at m/z 877.52. CID of this ion generates fatty acyl chain fragment ions at m/z 478.33, 504.34, 506.36, and 532.37 arising from the loss of the fatty acyl chains along with Mn(II) and Cl, confirming the presence of 16:0, 18:0, 18:1, and 20:1 fatty acids (Figure 5b and Figure 5c). Fragment ions generated from the neutral loss of the fatty acids (m/z 567.23, 593.24, 595.26, and 621.28) are also indicative of the fatty acyl chain identities. This suggests the presence of four isomers: PC 16:0/20:1, PC 20:1/16:0, PC 18:0/18:1, and PC 18:1/18:0. The relative abundances of the diagnostic fragment ions are different in the white matter and molecular layer regions of the cerebellum. For example, the intensity of m/z 532.37 is roughly 50% the intensity of m/z 478.33 in the molecular layer, while the relative abundances of these two ions are almost equal in the white matter, suggesting differences in relative abundances of these isomers in these tissue regions. The relative distributions of the four isomers can be mapped using the diagnostic fragment ions m/z 478.33, 504.34, 506.36, and 532.37 (Figure 5). Opposing relative distributions of PC 16:0/20:1 and PC 20:1/16:0 are observed in the rat brain cerebellum. PC 16:0/20:1 is more abundant in the white matter relative to PC 20:1/16:0 (Figure 5e), while PC 20:1/16:1 is more abundant in the molecular layer relative to PC 16:0/20:1 (Figure 5f). The relative distributions of PC 18:0/18:1(Figure 5g) and PC 18:1/18:0 (Figure 5h) are seemingly homogenous in the rat brain cerebellum, though PC 18:0/18:1 is more abundant. The same results are obtained when using all diagnostic ions for isomer imaging (Figure S8). These results are also consistent with previous studies,^{22, 42} and the different relative abundances of the isomers may reflect different enzyme activities in the rat brain.

Figure 5.

(a) An H&E stain shows structures of the rat brain cerebellum. CID of [PC 36:1 + Mn(II) + Cl]⁺ (CID: 22 V) is performed for ions generated from (b) the molecular layer and (c) the white matter regions of rat brain cerebellum. (d) PC 36:1 in the rat brain cerebellum is composed of four isomers, which are mapped as (e) the abundance of PC 16:0/20:1 relative to PC 20:1/16:0 (diagnostic ion intensity ratio of m/z 532.37 to the sum of m/z 478.33 and 532.37), (f) the abundance of PC 20:1/16:0 relative to PC 16:0/20:1 (diagnostic ion intensity ratio of m/z 478.33 to the sum of m/z 478.33 and 532.37), (g) the abundance of PC 18:0/18:1 relative to PC 18:1/18:0 (diagnostic ion intensity ratio of m/z 504.34 to the sum of m/z 504.34 and 506.36), and (h) the abundance of PC 18:1/18:0 relative to PC 18:0/18:1 (diagnostic ion intensity ratio of m/z 506.36 to the sum of m/z 504.34 and 506.36). Pixel size is 200 μm for all ion images. The orange triangles in panels (a) and (b) denote the precursor ions.

This workflow has also been used to visualize lipid isomer distributions in glioblastoma rat brain tissue (Figure 6a). PC 34:1 is less abundant in the tumor region compared to the non-tumor region (Figure 6b). CID of [PC 34:1 + Mn(II) + acetate]⁺ yields fragment ions m/z 478.33, 504.34, 591.27, and 617.29, confirming the identities of the fatty acyl tails to be 16:0 and 18:1. The m/z 478.33 and 504.34 diagnostic ions can be used to map distributions of sn-positional isomers. PC 16:0/18:1 is relatively more abundant in the non-tumor region (Figure 6c), while PC 18:1/16:0 is relatively more abundant in the tumor region (Figure 6d), suggesting the isomer distributions could be used to differentiate the tumor and nontumor regions of the glioblastoma. These results are in line with OzID experiments that show PC16:0/18:1 to be relatively less localized in medulloblastoma tumor tissue.²² All diagnostic ions indicated by m/z 478.33, 504.34, 591.27, and 617.29 are also used to map the isomer distributions and give the same results (Figure S9) Mapping the relative distributions of the isomers using the intensity ratios of diagnostic ions nicely provides an internal normalization that may increase batch-to-batch and sample-to-sample reproducibility. Overall, isomer-resolved imaging may allow for more reliable differentiation between the tumor and non-tumor regions than using measurement of the sum-composition lipid alone (i.e., MS1 measurement).

Figure 6.

(a) An H&E stain shows localization of (b) PC 34:1 to the tumor and non- tumor regions of glioblastoma brain tissue. CID of [PC 34:1 + Mn(II) + acetate]⁺ generates diagnostic product ions allowing for the isomer-resolved images showing (c) the distribution of PC 16:0/18:1 relative to PC 18:1/16:0 (diagnostic ion intensity ratio of m/z 504.34 to the sum of m/z 478.33 and 504.34) and (d) the distribution of PC 18:1/16:0 relative to PC 16:0/18:1 (diagnostic ion intensity ratio of m/z 478.33 to the sum of m/z 478.33 and 504.34). Pixel size is 150 μm for all ion images.

CONCLUSIONS

This work describes a method to characterize the fatty acyl chain positions of PCs via CID of a new ion type formed using divalent metal salts in MALDI analysis. The [PC + metal + ligand]⁺ ion type is readily formed using a variety of divalent metals and ligands. While formation of these ion types is not as efficient as for protonated and cationized ion types, it is useful for the characterization of PC sn-positional isomers. CID of [PC + metal + ligand]⁺ ion types does not result in phosphocholine headgroup fragmentation, as is typically observed during CID of PC cations. Rather, abundant fragment ions resulting from fatty acyl tail cleavages are observed that allow assignment of the fatty acyl chain positions. These fatty acyl chain product ions can be used to perform relative quantification of sn-positional isomers. PC structural and spatial heterogeneity in rat brain and glioblastoma tissues are also revealed by integrating this method into an imaging mass spectrometry workflow. Differential distributions of lipid isomers are observed in tissues, highlighting the importance of isomer specificity in lipid imaging experiments. The mechanisms and precise roles of the divalent metal and ligand ions in affecting the gas-phase ion chemistry of lipid fragmentation are still unclear and currently under investigation.