Introduction

The cell/tissue lipidome comprises over 10,000 individual species of highly diverse molecules.^[1,2] The identification and detailed structural characterization of lipids have been markedly accelerated by the development of soft ionization techniques for mass spectrometry (MS) combined with chromatography.^[1,2] Recently, MS has advanced into the field of imaging to enable the spatial cataloging of lipids in cellular compartments (reviewed in^[3,4]).

The most commonly employed mass spectrometry imaging (MSI) techniques for lipids are secondary ion mass spectrometry (SIMS), desorption electrospray ionization (DESI), and matrix-assisted laser desorption/ionization (MALDI). These techniques have been successfully applied to depict the spatial distribution of endogenous and exogenous molecules, as well as for revealing correlations of the homeostasis of low molecular mass compounds and proteins with disease occurrence.^[3–6]

In MALDI-MSI, tissue slices are coated with a matrix that is subsequently ablated by absorbing energy from a laser beam. This desorption forms a plume that also contains intact analyte molecules that were in close proximity to the matrix. The analytes either retain pre-formed charges or engage in charge-transfer reactions in the gas phase.^[7,8] This analysis is repeated at discrete locations on the tissue section, resulting in a “molecular fingerprint” that correlates analyte abundance with spatial location.^[3,4] An intrinsic limitation of MALDI-MSI of lipids is the irregular crystallization of matrices on the analyzed tissue surface. Variability in the matrix-to-analyte ratio affects ion intensity,^[9,10] an issue that becomes a limiting factor in MALDI-MSI at high lateral resolution.^[11] Careful optimization of sample preparation and matrix deposition methods are necessary to achieve the best resolution possible with MALDI imaging.^[12–14]

2,5-dihydroxybenzoic acid (DHB) is a widely used matrix for MALDI-MSI of biological molecules, including the mapping of (phospho)lipids.^[3,5] Either sublimation of DHB or dry-coating of DHB crystals through a sieve can give a coating of small crystals over the tissue surface^[15,16]. Spray coating can be optimized to deliver a matrix layer that is fine enough to image individual HeLa cells.^[17] As a general rule, smaller matrix crystals and stronger absorbance by the matrix at the laser wavelength yields a more efficient MALDI process, whereby optimal MALDI-MSI is achieved through maximum contact between analytes and matrix molecules.^[18]

DHB is a weak acid (pKa = 2.95) that facilitates MALDI-MS detection of lipids that are prone to protonation.^[19] For enhanced protonation of analytes, trifluoroacetic acid (TFA; pKa = 0.23) is often used as an additive to DHB. However, recent MALDI-MS studies demonstrated that the combination of DHB/TFA can cause significant hydrolysis of phospholipids.^[20]

DHB exhibits a UV spectrum with a maximum at 355 nm, making it a suitable matrix for analyses with MALDI mass spectrometers with N₂ lasers centered at 337 nm. At the optimal analyte-to-matrix ratio, the UV absorbance maximum of DHB should not deviate significantly from the laser wavelength as this would decrease the energy transfer from the laser to the analytes.^[18] Optimizing the absorbance of DHB by modifying the aromatic ring with various substituents gives an improved energy transfer.^[21] Modifying the physical properties of the matrix will also give improved MALDI-MS performance. Detergents with acid-cleavable alkyl chains have been used to improve the tryptic digestion efficiency of membrane proteins without interfering with MALDI-MS.^[13] Matrices with acid-cleavable alkyl chains resulted in a functional detergent-matrix compound that gave greatly improved signals for cell lysates and membrane proteins and generated the unmodified matrix itself upon sample preparation.^[14] Noticeably, after cleavage of the alkyl chain, the crystallization properties of the resulting matrix were unchanged from that of the non-alkylated matrix. DHB with various length alkyl chains, when added to another matrix, gave improved signals of up to two orders of magnitude for hydrophobic peptides.^[9] The crystallization properties of this mixture of alkylated and non-alkylated matrices were different from that of the non-alkylated matrix alone, and resulted in different distributions of hydrophobic peptides within the sample spot.

In the present study, we have carried out experiments to test whether the introduction of linear hydrocarbon chains in DHB yields matrices with improved analytical characteristics for MALDI-MSI of phospholipids in brain tissue. Targeted synthesis of DHB-C_nH_2n+1 (C₆H₁₃, C₉H₁₉ and C₁₂H₂₅) was carried out following the hypothesis that alkyl derivatives of DHB form smaller crystals, which is a pre-requisite for higher spatial resolution, and that incorporation of the alkyl chains of DHB-C_nH_2n+1 into phospholipid bilayers during matrix deposition on tissue slices will provide better matrix-to-analyte contact.^[22,23]

Materials and Methods

Results

To assess the effects of H⁺ on MALDI-MS of phospholipids, we have performed analyses with DHB solutions of varying acidity (pH = 2 – 8). In this pH range phospholipids do not undergo hydrolysis to any significant extent. However at pH > 4 the carboxylic group of DHB is fully deprotonated and, at 335 nm, its anionic form absorbs approximately 30% less light than the protonated acid (Supplementary Information (SI); figure 1S). Because the buffer potential of most tissues is centered at semi-neutral pH values, the deposition of DHB (after evaporation of the solvent) on tissue slices for MALDI-MSI of lipids may lead to deprotonation of the matrix thereby decreasing analytical sensitivity.

Figure 1 shows the MALDI-MS profiles of an equimolar mixture of 1,2-dioleoylphosphatidylcholine (DOPC; [M + H]⁺,m/z = 786.5; [M + Na]⁺,m/z = 808.4), 1,2-dioleoylphosphatidylethanolamine (DOPE; [M + H]⁺,m/z = 744.5; [M + Na]⁺,m/z = 766.4), 1,1',2,2'-tetraoleoyl cardiolipin (TOCL; [M - H]⁻,m/z = 1456.1; [M - 2H + Na]⁻,m/z = 1479.1) and 1,2-dioleoylphosphatidylserine (DOPS; [M – H]⁻,m/z = 786.6) obtained at 25% laser energy (LE). In both the positive (Fig. 1A) and negative (Fig. 1B) modes of ionization, maximum peak intensities were attained from samples prepared with solutions of DHB at pH 2 (Fig. 2A,D). By varying the LE, peak intensities followed biphasic distributions with maxima at 30% and 45% LE for samples coated with DHB at pH 2 and 4.7, respectively (Fig. 2B,C,E,F). In positive ionization mode at optimal LE, peak intensities decreased up to 50% at pH 4.7 (as compared to pH 2). In the negative mode, no significant changes in sensitivity were observed by varying the pH (Fig. 1, B and E compared to C and F, respectively).

We next conducted comparative MALDI-MS analyses of phospholipids in thin (10 micron) tissue slices from rat brain coated with DHB, DHB-C₆H₁₃ and DHB-C₁₂H₂₅. For each matrix, analytical sensitivity in the negative ion mode was optimized by tuning the laser energy to give the maximum peak intensity of the ion atm/z 806.5 (Fig. 3). Notably, at optimal analytical sensitivity for DHB (45% LE), no desorption/ionization of phospholipids by either DHB-C₆H₁₃ or DHB-C₁₂H₂₅ was detected from tissue. These data may be of relevance to studies by Fukuyama et al., who recently reported that DHB-C₈H₁₇ does not produce MALDI mass spectra of hydrophobic peptides; but the addition of DHB-C₈H₁₇ to a conventional matrix, α-cyano-4-hydroxycinnamic acid, increased the sensitivity of MALDI analysis by 2 orders of magnitude.^[9] In our model, comparable analytical sensitivity for lipids in negative ion mode was attained with DHB, DHB-C₆H₁₃ and DHB-C₁₂H₂₅ at LE of 45%, 60%, and 60%, respectively (Figure 4). LE higher than 65% were not used because of baseline shifts and fragmentation of analytes (data not shown). There was little difference in the spectral quality of lipid profiles obtained directly from tissues slices (Fig. 4A) and purified extracts from these tissues (Fig. 4B). Therefore, endogenous salts and proteins, which are present in tissue slices but not in purified extracts, did not strongly interfere with the analysis. Based on published data, assignments of the observed species are presented inTable 1. The most prominent species detected were sulfatides (ST(d18:1/18:0), ST(d18:1/24:1), and ST(d18:1/24:0) atm/z 806.6, 888.6, and 890.6 respectively) hydroxyl-sulfatides (ST(d18:1/h24:1) and ST(d18:1/h24:0) atm/z 904.6 and 906.6, respectively), and phosphatidylinositol (PI(38:4),m/z 885.5).

To test the hypothesis that alkylated hydrophobic derivatives of DHB may produce smaller crystals, DHB, DHB-C₆H₁₃, and DHB-C₁₂H₂₅ were deposited on brain slices and the structure of the coatings were examined (Fig. 5, panels A, B and C, respectively) using differential interference contrast microscopy (DIC, sub-panels i-iii of each) and polarized light microscopy (sub-panels iv-vi). The total zoomed image in panel ii corresponds to the area shown in the red box in panel i, and the same with the red box in panel ii giving panel iii. A trend of transition from crystals to oily surfaces is apparent in the sequence DHB → DHB-C₆H₁₃ → DHB-C₁₂H₂₅, with DHB-C₆H₁₃ forming fewer (and smaller) visible crystals than DHB (panel A compared to panel B).

Figure 6 presents typical MALDI-MSI ion maps of lipids from brain tissue sections coated with DHB and DHB-C₁₂H₂₅. An optical image of a near-serial tissue section (Figure 6A) shows the locations on opposite hemispheres that were coated with DHB and DHB-C₁₂H₂₅ and subsequently analyzed with MALDI-MSI (black rectangles). Individual MALDI-MS spectra were recorded with a lateral (X-Y) resolution¹ of 50 µm (Figure 6B) in negative ionization mode and at optimal LE. The MALDI ion images withm/z values of 806, 822, 862, 888, 904 and 906 obtained with DHB contained large areas were none of these analytes were detected or were below the sensitivity threshold. In contrast, DHB-C₁₂H₂₅ depicted a more continuous distribution of these analytes. These data indicate that DHB-C₁₂H₂₅ affords higher lateral resolution than DHB.

Discussion

The main finding of this study is that in MALDI-MSI of lipids, DHB-C₁₂H₂₅ is a matrix that affords higher analytical sensitivity/lateral resolution than DHB. Ion maps of select analytes using DHB-C₁₂H₂₅ displayed a relatively continuous distribution. On the other hand, the MALDI ion images of these analytes obtained with DHB contained large areas that were devoid of signal.

MALDI-MSI is a highly-sensitive imaging technique that allows the analysis of the spatial distribution of biomolecules.^[4] Among numerous biomolecular targets, lipids have been the focus of much MALDI-MSI research because of their relative abundance (at least for most lipids) in tissue making them excellent candidates for imaging. In addition, the molecular diversity of lipids allows them to perform essential structural, modulating and signaling functions in cells and tissues making them highly desirable imaging targets.^[31,32] Recent MALDI-MSI studies of lipids led to the development of protocols that afforded higher spatial resolution and addressed the possibility for introduction of quantitative criteria for data analysis.^[10,33,34]

Obtaining better spatial resolution is critical for the analysis of sub-anatomical structures and site-specific metabolic reactions. In general, particle beam methods provide better resolved MS images as compared to MALDI-MSI. Although MALDI-MSI resolution is limited by the laser spot diameter, an oversampling method that allows analysis of areas smaller than the laser spot has been developed by Jurchen et al.^[33] The method is based on the screening of the analyzed surface, with complete ablation of the MALDI matrix at each sample position, following a lateral shift of the laser spot to the next location at a distance less than the diameter of the laser beam.^[33]

In addition to optical limitations, the inherent variability of MALDI signals from lipids within tissue slices is related to the heterogeneity of both the sample surface and the matrix crystallization.^[35] Often matrices crystalize in clusters and illumination of matrix-free areas results in the loss of signal. To circumvent some of these limitations, at least in part, we synthesized improved matrices for lipid analysis based on structural modifications of DHB that do not affect its UV absorbance but change its physical properties.

In a typical spray-deposition MALDI-MSI experiment, tissue sections are sprayed with a solution containing matrix. In turn, evaporation of the solvent leads to crystallization of the matrix on the analyzed surface. Dry-coating of tissue for MALDI-MSI can be achieved by sublimation or sieve-deposition to deliver crystals of matrix to the tissue without solvent. Various other matrix deposition methods include robotic and inkjet spotting, electrospray deposition, and airbrush or capillary spray (ref.^[36] and the references therein). Comparative microphotographs of tissue sections coated with DHB by these methods show that DHB can crystalize in clusters that are separated by uncoated tissue areas with lengths of up to 100 µm.^[36–38] Optimal MALDI-MS spectra are obtained from areas with a high density of matrix crystals,^[9,39] whereby the mass spectral peak intensity, resolution and reproducibility are inversely proportional to the heterogeneity of the matrix coating. The imaging of biological systems that are smaller than the DHB clusters can be achieved with high resolution. For example, Schober et al. detected lipids from a single cell (length, ~ 100 µm) with a resolution of ~ 7 µm.^[17] However, when imaging tissue sections with a length of several centimeters is carried out, the clustering of DHB can affect the intensity of the MS spectra [6,26]. Recently, Hankin et al. have reported that the application of DHB on brain tissue slices by sublimation of the matrix affords a coating with small crystals that do not form clusters.^[15] The latter method is based on the heating of DHB to 120 °C at reduced pressure (0.05 Torr) with concomitant condensation/crystallization of DHB vapors on the analyzed tissue sections. Optimization of the amount of matrix that condenses on the analyzed surface requires careful adjustment of experimental parameters such as pressure, the amount of heat applied to the matrix and the time of sublimation,^[29] with appropriate controls to determine that the analyzed surface has not been perturbed by overheating. The coating of tissue samples with DHB by the sublimation method has allowed MALDI-MSI of lipids with a spatial resolution of approximately 10 µm.^[11]

We have targeted the synthesis of DHB-C_nH_2n+1 (C₆H₁₃, C₉H₁₉ and C₁₂H₂₅) following the rationale that alkyl derivatives of DHB would form smaller crystals than DHB, which is a pre-requisite for higher lateral resolution analyses, and that incorporation of the alkyl chains of DHB-C_nH_2n+1 into phospholipid bilayers (detergent effect) during spray deposition on tissue slices will provide better matrix-to-analyte contact. We hypothesized that DHB-C_nH_2n+1 would form smaller crystals following the empirical observation that the melting points ofO-alkylated phenolic compounds decrease with increases in the length of their alkyl chains. For example, the melting points of 2-hydroxybenzoic (salicylic) acid, 2-methoxybenzoic acid and 2-ethoxybenzoic acid are 158 °C, 101 °C and 19 °C, respectively. The observed decreases in the melting points can be correlated with the decrease in polarity of the corresponding derivatives. Further, assuming that the crystal growth in this example obeys Gibbs' rules of phase equilibria at constant pressure, it is reasonable to speculate that the transition from crystalline to liquid phase will reflect shifts in these equilibria that include the formation of smaller crystals (as opposed to crystal growth). It is noteworthy thatO-alkylation of DHB to DHB-C_nH_2n+1 does not significantly affect the UV spectrum of this matrix. In ethanol, DHB, DHB-C₆H₁₃ and DHB-C₁₂H₂₅ exhibit UV spectra with maximal absorptions at 333 nm, 229 nm, and 229 nm, respectively; for all matrices, ε ≃ 3.2 × 10³ M⁻¹.cm⁻¹).

On rat brain tissue sections, our spray deposition system produced DHB crystals of approximately 20 µm in length that were grouped in clusters spaced by areas of uncoated tissue sections (Fig. 5Aiv). DHB-C_nH_2n+1 exhibited a trend of transition from crystals to an oily layer in the sequence DHB → DHB-C₆H₁₃ → DHB-C₁₂H₂₅, with DHB-C₆H₁₃ forming less (and smaller) crystals than DHB (panel A compared to panel B). Comparison of the data presented inFig. 6A and5C.i., andFig. 2S and 3S (SI), indicates that DHB-C₁₂H₂₅ did not cause any apparent structural changes in the analyzed tissue sections. Both DHB-C₆H₁₃ and DHB-C₁₂H₂₅ required a higher LE than DHB for the detection of peaks with comparable intensities, which correlates with the increased boiling points (b.p.) of theO-alkylated matrices; at 25 °C and 760 mm Hg, the b.p. of DHB, DHB-C₆H₁₃ and DHB-C₁₂H₂₅ are 367.33 °C, 414.03 °C and 490.29 °C (estimations made with ChemDraw Ultra, v. 12). Although the algorithm used for b.p. estimations provides only approximate values,^[40] it denotes the expected trend of increased temperatures for the phase transition of the alkylated matrices. In MALDI experiments, analytes are in contact with matrix molecules, which are then irradiated with high energy laser pulses. This triggers the evaporation of matrix and analyte molecules as ionized molecular species. Vertes et al. proposed that this process proceeds via an initial bulk-desorption of the matrix with concomitant energy transfer to analytes.^[41] It has been further discussed that the mechanical process of desorption activates analyte molecules via transfer of a reaction momentum in the gas phase (“Popcorn” model).^[42] Alternatively, competitive kinetics between sublimation and fragmentation have been considered (“Bottle-neck” model), whereby a “good” matrix has been postulated to possess a low phase-transition temperature.^[41,43] Studies by Price et al., however, have shown that the enthalpy of matrix evaporation/sublimation is not the principle factor in the MALDI process.^[44] In our analysis of lipids with DHB and itsO-alkyl derivatives, we were able to compensate for differences in MALDI sensitivity by increasing the LE. Importantly, DHB-C₁₂H₂₅ was layered exclusively as oil on the tissues sections, which suggests that it provided a more complete, homogenous coating of the analyzed surface. Indeed, comparative MSI analysis of lipids in brain tissues with DHB and DHB-C₁₂H₂₅ indicated that the latter matrix affords better lateral resolution (Fig. 6).

Our results indicate thatO-alkylation of DHB for MALDI-MSI is a promising approach for enhanced spatial resolution. Since hydroxy and amino functions on aromatic ring systems are common functional groups that are the key to the usefulness of several widely used matrices, such as 9-aminoacridine, α-cyano-4-hydroxycinnamic acid, sinapic acid, 2,6-dihydroxyacetophenone, and 1,5-diaminonaphtalene, perhaps O- and N-alkylation may be suitable approaches for modifying these matrices. Currently, the factors that affect the crystallization of these matrices on tissue sections are not well understood and require further study.

Supplementary Material

Acknowledgments

This study was supported in part by grants from the US National Institutes of Health (NS076511 and U19AI068021 (H.B., V.E.K. and D.A.S.), ES021068 (B.R.P. and V.E.K.), NS061817 (H.B.), ES020693 (V.E.K.), U54 GM103529 08 (S.C.W.)), the US National Institute for Occupational Safety and Health (OH008282 to V.E.K.). This project used the UPCI Cancer Biomarkers Facility that is supported in part by Cancer Center Support Grant (CCSG) P30CA047904.

Footnotes

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.