Mass spectrometry imaging with laser-induced postionization

See allHide authors and affiliations

Science  10 Apr 2015:
Vol. 348, Issue 6231, pp. 211-215
DOI: 10.1126/science.aaa1051

Imaging lipid composition

Chemical imaging of cell membranes can be performed with matrix-assisted laser desorption/ionization mass spectrometry (MALDI), but low ionization efficiency often leads to a signal dominated by the main lipid components, such as abundant phosphatidylcholine species. Soltwisch et al. used a tunable laser for post-ionization of neutral species to boost the signal for other membrane components, such as cholesterol and phospho- and glycolipids. Imaging of cells and tissues with these methods allows differentiation based on a more extensive chemical signature.

Science, this issue p. 211


Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) can simultaneously record the lateral distribution of numerous biomolecules in tissue slices, but its sensitivity is restricted by limited ionization. We used a wavelength-tunable postionization laser to initiate secondary MALDI-like ionization processes in the gas phase. In this way, we could increase the ion yields for numerous lipid classes, liposoluble vitamins, and saccharides, imaged in animal and plant tissue with a 5-micrometer-wide laser spot, by up to two orders of magnitude. Critical parameters for initiation of the secondary ionization processes are pressure of the cooling gas in the ion source, laser wavelength, pulse energy, and delay between the two laser pulses. The technology could enable sensitive MALDI-MS imaging with a lateral resolution in the low micrometer range.

Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) is used for the analysis of large nonvolatile biomolecules (1). Typically, the analytes are embedded into crystalline MALDI matrices (often aromatic carboxylic acids). Laser-induced codesorption of both constituents into the gas phase is convoluted with concomitant ionization in primary and secondary ionization processes (24). MALDI-MS imaging (MALDI-MSI) visualizes the lateral distribution of numerous biomolecules or that of administered drugs simultaneously by scanning matrix-coated tissue slices with a finely focused laser beam and recording the ion profiles per irradiated pixel (5, 6). For example, lipid profiles are increasingly used for visualization of malignancies (7, 8) or to retrieve valuable information about the general biochemistry in tissue (9). A lateral resolution below 10 μm can now be achieved (10, 11) that could potentially allow one to visualize the molecular content of single cells, thereby fostering numerous applications in the life sciences (1214).

However, restricted ion yields—less than 1 out of 1000 desorbed molecules is on average ionized (15, 16)—and ion suppression effects (3) can impede obtaining a more comprehensive picture about the tissue composition. For example, lipid mass spectra recorded in the positive ion mode are typically dominated by the signals of phosphatidylcholines (PCs), whereas other phospho- (PLs) and glycolipids (GLs) are often barely detectable, even if these are present in the tissue in high abundance (17). Because of the minute amount of material that is ejected per laser shot, this problem is aggravated in high-resolving imaging (18). To increase the ion yields, single and multiphoton laser ionization of neutral gas phase molecules (19, 20), as well as postionization (PI) by means of an electrospray ionization (ESI) beam (21), have been developed. However, extensive analyte fragmentation limits the use of direct photoionization to small molecules, and ESI requires the use of ambient pressure ion sources. For most analyte classes, the state-of-the art sensitivity achieved by classical MALDI-MSI performed under vacuum is not met with these methods.

Here we introduce a PI strategy, called MALDI-2, that initiates secondary MALDI-like ionization processes in the gas phase. In MALDI-2, the beam of a pulsed ultraviolet (UV) laser intercepts the expanding particle plume in an N2 cooling gas environment (22, 23) (Fig. 1), which contrasts with previous photoionization studies where classical high-vacuum ion sources (p ≤ 10−6 mbar) were implemented (20, 24). An effective diameter of ~5 μm of the primary laser beam (fig. S1) was achieved by beam shaping (23) and by mounting the focusing lens inside the MALDI ion source (Fig. 1) (25).

Fig. 1 MALDI-2 ion source.

Schematic drawing of the modified MALDI ion source of the Synapt G2-S mass spectrometer with primary MALDI and PI laser beams and shielding aperture for increasing the cooling gas pressure in the region of ion generation; 2.0 to 2.5 mbar of N2 were used for the MALDI-2-MSI measurements. The lower panel illustrates the laser pulse triggering sequence.

Two positive ion mode mass spectra that we recorded from mouse cerebellum with and without the PI laser are compared in Fig. 2A and table S1. For this experiment, we chose 2,5-dihydroxybenzoic acid (DHB) as a classical MALDI matrix and prepared the samples using a sublimation and recrystallization protocol, to obtain a particular uniform microcrystalline coating (fig. S1) (25). We observed several classes of membrane lipids that are difficult to image by conventional MALDI-MSI of PC-rich tissue [e.g., cholesterol, phosphatidylethanolamine (PE), plasmalogens (PE-O), phosphatidylserine (PS), and neutral GLs such as galactosylceramide (GalCer)] in the MALDI-2 spectra with [M + H]+ signal intensities that were up to two orders of magnitude higher. The signal intensities of alkali metal adducts increased more moderately by a factor of up to 5, depending on the lipid class, or remained essentially unchanged for PCs. Based on exact mass, we could tentatively assign about two to three times as many lipid species in the mass-to-charge ratio (m/z) range of 670 to 850. The benefit of the high signal intensities that we obtained from a single 5 μm-wide pixel (Fig. 2B) for MS imaging is demonstrated in Fig. 2, C to L, with the example of selected PLs and GalCer analyzed from mouse cerebellum. The distributions of liposoluble vitamins A1 (retinol), D3 (cholecalciferol), E (α-tocopherol), and K2 (menachinon-4) could only be visualized with MALDI-2 (Fig. 2, M to P). These compounds exhibit a high extinction coefficient at 280 nm and are therefore detected both as [M + H]+ and resonantly photoionized M•+ species.

Fig. 2 Mass spectra and MS images recorded from mouse cerebellum in the positive ion mode.

(A) MALDI-2 (red trace) and conventional MALDI mass spectra (blue trace) acquired in the high-resolution mode of the instrument. Each spectrum was accumulated over 144 pixels (two parallel lines, 20 laser shots/pixel); λPI = 280 nm, τ = 9 μs, p = 2.5 mbar. (B) Single pixel spectrum acquired in the resolution mode. (C to L) Conventional MALDI-MS (left) and MALDI-2-MS images (right) of brain lipids and (M to P) of liposoluble vitamins; the molecular identities were corroborated by tandem MS (fig. S2). Images were recorded with a pitch size (pixel-to-pixel distance) of 15 μm. (Q) Hematoxylin and eosin (H&E) stain; the slice was taken approximately from interaural –2.5 to –3 mm. GL: granular layer, ML: molecular layer, WM: white matter. ✦: [cholesterol – H2O + H]+; ∇: [PA + Na]+ (PA: phosphatidic acid) or [PA + K]+; ■: [PC + H]+; □: [PC + Na]+ or [PC + K]+; ●: [PE + H]+; ★: [PE-O + H]+; ✰: [PE-O + Na]+ or [PE-O + K]+; ▲: [PS + H]+; ▸: [PC-O + H]+; +: [GalCer + H]+.

Next we asked if measurements in the negative ion mode would also benefit from the postionization step; typically, positive and negative ion mode measurements provide valuable complementary information about the overall lipid composition in tissue. We obtained up to one to two orders of magnitude higher [M – H] signal intensities for several lipid classes (e.g., for PA, plasmalogens PE-P and PE-O, PE, and PS) if desorbed from a norharmane matrix (Fig. 3 and table S2). This enabled MS imaging with high signal contrast for about five times as many lipid species. Only phosphatidylinositols (PIs; Fig. 3H) and sulfatides (STs; Fig. 3I), compounds that are already detected sensitively by regular MALDI-MSI (23), display a more moderate signal increase or no notable gain (STs).

Fig. 3 Mass spectra and MS images recorded from mouse cerebellum in the negative ion mode.

(A) Mass spectra acquired in the high-resolution mode from 300 adjacent pixels, each; λPI = 260 nm, τ = 10 μs, p = 2.0 mbar (compare fig. S3 for a single pixel spectrum). (B to J) MALDI- and MALDI-2-MS images of selected compounds (compare fig. S4 for tandem MS data); pitch size: 15 μm. (K) H&E stain; the slice was taken approximately from interaural –3.7 to –4.2 mm and comprises also a small part of the medulla (M). ▼: [PA – H]; ●: [PE – H]; ★: [PE-O – H]; : [PE-P - H]; ▲: [PS - H]––; ♠: [PI - H]––; ♣: [ST - H]––.

We also tested a few more tissue types and analyte classes. For example, seminiferous tubule structures of rat testis could be visualized via their triacylglycerol (TG) content (fig. S5). Atypical for conventional MALDI, these polar lipids were detected as protonated molecules. MALDI-2-MSI of plant (apple) tissue produced strongly enhanced signals of mono- and oligosaccharides and of polyphenolic glycosides (fig. S6). In contrast to lipids and glycans, MALDI-2 of peptides has so far not lead to a signal enhancement, possibly because of different ionization mechanisms and plume expansion dynamics (1, 2, 4).

To identify optimal PI conditions, we performed a multiparameter study (Fig. 4). This revealed that the MALDI-2 ionization efficiency depends critically on a few key parameters, namely, the buffer gas pressure p, delay between the laser pulses τ, wavelength of the PI laser λPI, and its pulse energy EPI. Optimal ion yields were found in distinct (matrix-dependent) parameter ranges. In particular, an elevated pressure was required (Fig. 4A and fig. S7). At p = 0.5 mbar, where only a low signal increase was found, the derived mean average velocity of neutral DHB molecules of about 500 ms−1 (Fig. 4C) corresponds well to values that were previously determined for classical high-vacuum MALDI conditions (2). However, at elevated pressures (1.5 and 2.5 mbar), the molecules were effectively slowed down to mean velocities below 100 ms−1. The absence of entrainment in a cooling gas flow and subsequent gas-phase reactions probably account for the inability of previous MALDI-PI studies to produce protonated matrix molecules and higher molecular weight compounds (20).

Fig. 4 Parameters affecting the MALDI-2 ion yields.

(A) Heat map illustrating the effect of the cooling gas pressure p and laser pulse delay τ on the signal intensities of protonated PE molecules. Each data point (black dots) was recorded by applying 600 laser pulses onto approximately 50 positions on DHB-coated homogenized liver tissue. (B) Effect of the PI laser wavelength λ and pulse energy EPI. Solid white line: solution phase mean penetration depth (α−1) of DHB; dashed yellow line: two-photon ionization threshold. (C) MALDI-2 signal intensities of protonated PE as a function of τ for three different gas pressures; average mean velocities are calculated as v = Δz/τ with Δz = 0.5 mm. (D and E) Matrix mass ranges in the positive (DHB matrix) and negative ion mode (norharmane), respectively.

Apart from a small hypsochromic shift, reflecting the difference between solution- and gas-phase absorption spectra (26), the MALDI-2 ion yields of lipids that are nonabsorbing at the PI laser wavelength followed the optical absorption characteristics of the matrix. Particularly high [M + H]+ signals were obtained for wavelengths below the two-photon ionization threshold of DHB of ~310 nm (27) (Fig. 4, B and C). The same behavior was found for the total ion count (TIC) (fig. S7). In contrast, for alkali metal adducts of the lipids, the wavelength characteristics more closely reflected the overall absorption profile of the matrix. Similar to conventional MALDI-MS (28), high signals are here found also above 310 nm (fig. S7D). Possibly, gaseous Na+ and K+ ions were generated by the photodissociation of neutral matrix clusters (29), which gave rise to additional [M + Na]+ and [M + K]+ species.

We hypothesize that the mechanisms underlying ionization by proton transfer in MALDI-2 involve resonant two-photon ionization of the matrix (m) by the PI laser (giving rise to m+• ions and free e), succeeding collisions with neutral matrix molecules (leading to the generation of protonated or deprotonated matrix), and proton transfer to or from neutral analyte molecules (M) in subsequent collisions to yield the observed [M + H]+ or [M – H] products. With regard to the analyte ionization step, similar mechanisms are discussed for conventional MALDI (3). These assumptions are supported by the detection of high abundances of [m – H2O + H]+ ions of the DHB matrix and oligomers thereof, and of [m – H] ions of norharmane and a few oligomers thereof, respectively (Fig. 4, D and E, and table S3). Also, the observed adduct formation between some analytes (e.g., PE) and these matrix-derived ions supports the notion of ample gas-phase reactions (fig. S8). Notably, the soft nature of the ionization is accompanied by a low degree of analyte fragmentation (fig. S9A)

Next to the liposoluble vitamins (fig. S9, C and D), we observed the direct two-photon ionization of analyte molecules, giving rise to additional M•+ and M•– species, also for sialic acid–containing gangliosides. The direct absorption of the laser light by the weakly bound sialic acid residue(s) and succeeding bond cleavage probably led to the high levels of observed asialoganglioside ions (Fig. 3J, fig. S9B, and table S2). So far, we have observed the strong PI effect only with laser spot sizes in the low 10- to sub–10-μm range. This could be caused by the confinement of the particle plume or by an insufficient overlap of the PI laser beam with a more extended particle plume.

Given the high image contrast that was achieved from only 5-μm-wide pixels, we postulate that MALDI-2 will in the future enable MS imaging with even higher lateral resolution in the 1- to 2-μm range. This could provide analytical possibilities that at present are mainly in the domain of optical microscopy techniques. Additionally, the more comprehensive picture that is obtained about the molecular composition in tissue could substantially improve the confidence level of methods that use altered lipid profiles as markers for disease states.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 to S3

References (3039)

References and Notes

  1. Further experimental details can be found in the supplementary online materials.
  2. Acknowledgments: We thank Waters Corp. and GWU Lasertechnik for technical support; F. Spieker and J. Haier for animal tissue; J. Klingauf for scanning electron microscopy images; W. Kramer for Fig. 1; and S. Albrecht, H. Karch, T. Kuhlmann, G. Pohlentz, and J. Y. Yew for helpful discussions. Financial support by the German Science Foundation (grants DR416/8-2 and DR416/9-1 to K.D., SO976/2-1 to J.S., GRK 1409 and MU845/4-2 to J.M.) and the Interdisciplinary Center for Clinical Research (IZKF) Münster (Z03 to K.D.) is gratefully acknowledged. We dedicate this work to the memory of Dr. Franz Hillenkamp in recognition of his lifetime achievements in the field of MALDI mass spectrometry.
View Abstract

Navigate This Article