Imaging mass spectrometry in transmission geometry

Imaging mass spectrometry has been used to provide information on the spatial distribution of lipids,^[1,2] peptides^[3] and proteins,^[4] as well as pharmaceuticals^[5] and metabolites^[6,7] directly from biological tissue sections. Matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry has been successful in investigating protein changes corresponding to several diseases, including Alzheimer’s,^[8] Parkinson’s,^[9,10] and cancer.^[11-13] Secondary ion mass spectrometry (SIMS) has been used to map the distribution of cancer agents^[14] and lipids^[15] with high sensitivity and spatial resolution, although its mass range is limited by extensive molecular fragmentation.^[16] Desorption electrospray ionization (DESI), an ambient ionization method, shows promise in the imaging of living tissue without sample preparation,^[17,18] but its spatial resolution is currently limited.^[19] Common to all surface imaging methods is the ablation from the front side, also referred to as reflection geometry, with the laser beam, cluster beam or liquid stream at an angle to the plane of the surface. Pairing the sensitivity and specificity of mass spectrometry (MS) with imaging techniques has the potential to provide information on the chemical composition of surfaces at the subcellular dimension, but necessitates the development of methods combining high-throughput analysis and high sensitivity with high spatial resolution.

One problem in vacuum MALDI imaging in reflection geometry mode is the loss of mass resolution as well as mass accuracy that accompanies high laser fluence. For this reason, the sample is often ablated near the threshold for ion observation and frequently over 100 laser shots are acquired and their mass spectra summed for each pixel of the image.^[20-22] The time required to obtain an image is dependent on several parameters, including laser repetition rate, the desired spatial resolution of the sample, and the speed of data processing.^[20] As the spatial resolution increases, so does the time required to image the respective surface. Reasonable speed of analysis in reflection geometry requires the use of high repetition lasers, capable of analyzing multiple pixels per second.^[23] Lasers operating in the kHz region significantly reduce analysis time,^[20,24,25] but at a significantly higher laser cost.

An alternative configuration is to align the laser in transmission geometry relative to the sample and mass spectrometer orifice. In transmission geometry, laser irradiation occurs from the back of the sample through a transparent sample holder. The first demonstration of tissue analysis in transmission geometry mode has been reported.^[26] Transmission geometry had previously been demonstrated in both vacuum^[27-30] and atmospheric pressure^[31] (AP) MALDI for standard samples.

Laserspray ionization (LSI) operates at AP in either transmission^[26] or reflection geometry^[32] in the absence of an applied voltage and, as recently shown, the laser is not necessary for ionization.^[33] Because ionization occurs inside the transfer capillary,^[34,35] any means, including laser ablation, of transferring matrix/analyte to the ion entrance orifice produces ions.^[33] Higher temperatures of typically 300°C using 2,5-dihydroxyacetophenone (2,5-DHAP) on Velos LTQ and Orbitrap mass spectrometers have been reported to enhance the abundance of highly charged peptides and protein ions in the positive mode.^[32-37] Transmission geometry allows the sample to be placed very near the ion entrance orifice and here the tissue sample was held approximately 1 mm from the mass spectrometer vacuum entrance thus allowing more efficient sampling of the ablated matrix/analyte clusters. A 337 nm nitrogen laser (Newport Corporation, Irvine, CA, USA; VSL-337ND-S) beam aligned at 180° relative to the entrance of the mass spectrometer and focused using a 102 mm focal length lens (CVI Melles Griot, Albuquerque, NM, USA) was used to ablate the sample.

Because LSI produces electrospray ionization (ESI)-like multiply charged ions, newer fragmentation methods can be employed to characterize materials. For example, nearly complete sequence coverage of ubiquitin, a regulatory protein, was obtained by electron transfer dissociation (ETD) using a single laser shot in transmission geometry.^[35] ETD, in a single LSI acquisition, was also used to sequence and identify directly from mouse brain tissue the doubly charged peptide of m/z 917.5 [molecular weight (MW) 1833 Da] as a myelin basic protein N-terminal fragment.^[37] High MW proteins have also been observed directly from tissue as multiply charged ions using transmission geometry LSI.^[37] Multiply charged ions are also beneficial in imaging when combined with ion mobility spectrometry (IMS)-MS because they promote separation of components in mixtures,^[38] including isomers.^[39]

Another advantage of transmission geometry imaging is that the laser is fired only once at each position because a high laser fluence is used to ablate through the tissue/matrix. This is possible because ionization is at AP, which provides ample collisional cooling to occur before mass analysis.^[26,36,37] Each pixel of the image represents a single laser shot producing a full mass spectrum. This offers the potential for rapid analysis^[36] and imaging using inexpensive nitrogen lasers. A proof of concept experiment produced a crude image of peptides in mouse brain tissue in 30 min acquisition time using LSI in transmission geometry.^[37]

Because the laser travels through the sample in transmission geometry mode, ideally the entire volume of tissue impacted by the laser beam is ablated and ionized in a single pulse. The quality of the obtained image is highly dependent on sample preparation and the thickness of the sample. In the previous transmission geometry images, 2,5-DHAP matrix was applied using the dried-droplet method,^[40] applying nanoliters of solution.^[37] Solvent-based matrix deposition is inappropriate for high spatial resolution imaging, as solvents cause the delocalization of components in the tissue. Solvents also cause problems related to reproducibility, as ‘hot spots’ can arise from uneven matrix application and crystallization. Because each pixel represents one shot, homogenous matrix coverage is required for shot-to-shot reproducibility. Attempts at spray-applying the matrix solution yielded low abundance signals in LSI operated in transmission geometry.^[37]

Solvent-free sample preparation was developed a decade ago,^[41,42] and has recently found new applications related to tissue analysis and imaging.^[26,43-46] The first reported analysis of tissue in transmission geometry using LSI employed both solvent-free and solvent-based sample preparation of the matrix on the tissue.^[26] Tissue imaging was not reported in this study due to limited reproducibility. Another limitation in obtaining high spatial resolution is sensitivity. In a previous report, a strategy was developed in which the glass slides were pre-coated with LSI matrix and a tissue section placed on top of the thin matrix coating.^[37] This enhanced the ion abundance but caused significant larger tissue areas to ablate so that arguably no enhancement of ionization efficiency was achieved.^[37] Large ablated areas are often required to observe sufficient ion current to obtain the necessary reproducible ion abundance for imaging.

Here, we present the first imaging experiment using transmission geometry. This also represents the first report of LSI producing negative ions. In this proof of principle study, the instrument acquisition time for a mouse brain section was approximately 1 h using single laser shot ablations of ~20 μm achieved by solvent-free sample preparation using automated solvent-free matrix deposition on the tissue section.

Mouse brain tissue was obtained from C57 Bl/6 mice, 20 weeks old, that were euthanized with CO₂ gas and transcardially perfused with ice-cold 1× phosphate-buffered saline (PBS, 150 mM NaCl, 100 mM NaH₂PO₄, pH 7.4) for 5 min to remove red blood cells. The brains were frozen at −22°C and sliced into 10 μm sections in sequence using a Leica CM1850 cryostat (Leica Microsystems Inc., Bannockburn, IL, USA). The tissue sections were placed onto prechilled microscopy glass slides that were briefly warmed with a finger from behind to allow sections to relax and attach. Care was taken to avoid water condensation by storing at −20°C and transporting under dry ice the tissue-mounted glass slides in an airtight, desiccant-containing box until use. 2,5-DHAP LSI matrix was pre-ground in a 5 mL glass vial containing 1.3 mm chrome beads for 30 min at 15 Hz in a TissueLyser II ball mill device (QIAGEN, Valencia, CA, USA) similar to previous procedures.^[45] Briefly, the pre-ground matrix was placed in the top compartment of the TissueBox with approximately 30 1.4 mm stainless steel beads. The microscopy slide mounted with mouse brain tissue was separated from the beads and matrix by 5 μm mesh. The TissueBox was placed in the TissueLyser II for 2 min at 25 Hz. Matrix application occurs as the pre-ground matrix is pushed through the mesh by the movement of the TissueBox in the TissueLyser II ball mill device. A 2 min grind time was required to obtain homogenous matrix coverage using 2,5-DHAP. At times greater than 2 min, the layer of applied matrix was too thick to obtain usable MS signals.

LSI-MS and MS/MS experiments were performed on a Thermo Fisher Velos LTQ (Thermo Fisher Scientific, Bremen, Germany) linear ion trap mass spectrometer. As described previously the source housing is removed to allow free access to the ion entrance orifice. The glass microscope slide mounted with mouse brain tissue covered with 2,5-DHAP was attached to a computer-controlled xy-stage (Newmark Systems, Mission Viejo, CA, USA) and moved through the laser beam in transmission geometry mode. The lanes were at 100 μm intervals, and each lane was 10.8 mm long. The instrument was operated in negative ion mode measuring a mass range from m/z 600 to 1000. The laser fluence per pulse was in the range of 0.5–1.0 J cm⁻². Tissue images were created using BioMAP 3.7.5.6 (Novartis Institutes for BioMedical Research, Basel, Switzerland). Thermo XCalibur .RAW files were converted into .IMG files using customized software.

The AP-to-vacuum ion transfer capillary was heated to 450°C. In positive ion mode, 2,5-DHAP reduces the required desolvation temperature to as low as 150°C using solvent-based sample preparation methods.^[39] Solvent-free sample preparation uses higher temperatures, as high as 275°C, to produce LSI ions of peptides with high abundance in the positive ion mode.^[47] As shown in Fig. 1, thermal requirements are notably high in negative mode measurements using solvent-free sample preparation. At temperatures below 400°C, there was a significant decrease in ion abundance and shot-to-shot reproducibility. Furthermore, the key to solvent-free sample preparation appears to be the ability to minimize the sample crystallinity and compactness but provide intimate contact with analyte material.^[48] Using the TissueBox for solvent-free surface coverage, depending on the mesh size used, here 5 μm, produces small matrix particles which have high velocity when using the ball mill at a high frequency. These small projectiles hitting the tissue section may allow sufficient surface penetration so that the spot-to-spot reproducibility is enhanced while keeping the integrity of the tissue relatively undisturbed.

Figure 1.

Transmission geometry negative ion LSI mass spectra of 10 μm thick mouse brain tissue sections using 2,5-DHAP as a matrix and solvent-free sample preparation at (A) 450°C, (B) 400°C, and (C) 350°C ion transfer capillary temperatures. Each spectrum represents a single laser pulse. Mass spectra from similar mouse brain regions are displayed for comparison.

The matrices 2,5-dihydroxybenzoic acid (DHB) and 2,4,6-trihydroxyacetophenone (THAP) were tested, and yielded insufficient negative ion abundance under all experimental conditions evaluated, including variation of temperature and degree of matrix coverage. 2,5-DHAP provided the best sensitivity and reproducibility in negative ion mode measurements and it was thus used exclusively in the following imaging studies.

Many polar phospholipids and sphingolipids are preferentially ionized in negative ion mode, including phosphatidylserines, (PS), phosphatidyinisitols (PI), and sulfatides (ST). Several of the detected lipids (Figs. 1 and 2(A)) are similar to those previously reported in negative ion mode.^[49,50] LSI MS/MS using collision-induced dissociation (CID) with a collision energy of 40 eV and a precursor ion selection window of 1.0 m/z unit identified the lipid at m/z 888.7 as ST (3-O-sulfogalactosylceramide) 18:1,24:1 (Fig. 2(B)), which is present in the myelin sheath.^[51] Differences in intensities and observed lipid species are probably related to solvent-free sample preparation. The observation of different lipids in solvent-free and solvent-based conditions is consistent with previous investigations.^[45,46]

Figure 2.

Transmission geometry negative ion LSI mass spectra of 10 μm thick mouse brain tissue sections using 2,5-DHAP as a matrix and solvent-free sample preparation at 450°C ion transfer capillary temperatures. (A) Full mass spectrum. (B) MS/MS spectrum using CID of m/z 888.7 indicating the chemical composition of phosphatidyl sulfatides (ST) 18:1/24:1. The width of the parent selection window is 1.0. Collision energy of 40 eV was used.

While efforts to reduce the time of analysis for MS imaging have focused on increased laser speed, here an image of mouse brain tissue was obtained in approximately 1 h with a laser operating at approximately 12 Hz. Each row required 33 s of acquisition time, and 93 rows spaced at 100 μm were required to completely image a single section of mouse brain. Laser ablated areas of ca. 20 μm were obtained (Fig. 3). Previous LSI imaging experiments using transmission geometry on delipified mouse brain tissue prepared with 2,5-DHAP and solvent-based sample preparation deposition reported ablated diameters of ca. 15 μm.^[37]

Figure 3.

Optical microscopy of mouse brain tissue after LSI imaging in transmission geometry. Laser ablated dimensions are approximately 20 μm.

Single laser shots were sufficient to differentiate the distribution of lipids appearing at m/z 888.7, 885.6, and 834.5. Using LSI in negative ion mode and the laser aligned in transmission geometry, sufficient ion abundances were obtained from each laser shot to form an image of lipids from mouse brain tissue (Fig. 4), indicating that inexpensive nitrogen lasers may provide a speed of analysis similar to high repetition rate lasers.

Figure 4.

LSI mass spectrum and ion images of [M–H]⁻ lipids from a 10 μm mouse brain tissue section. The tissue section was covered with 2,5-DHAP solvent-free and the acquisition was obtained with the laser aligned in transmission geometry and with ion transfer capillary temperatures of 450°C in the negative ionization mode. For example m/z 888.7, the crescent-shaped figure on the left side is the external capsule/alveolus, which is rich in myelin. The intensely labeled structure in the center corresponds to the thalamus. Other [M-H]⁻ ion images are of m/z 766.59, 790.59, 862.67, 885.59 and 906.67.

Current work includes optimizing sample preparation to further increase ion abundances and shot-to-shot reproducibility. Previous studies^[31] have highlighted the importance of thin matrix coverage for transmission geometry. Application of a thin matrix film that allows the laser to penetrate the complete sample layer while still providing enough matrix coverage for proper ionization of the sample is a challenge that we are addressing. Enhancements are expected with the ability to produce smaller sized and higher velocity crystals. Further, IMS will be incorporated to eliminate washing procedures that delipify tissue and enhance the detection of proteins,^[52] allowing gas-phase separation of lipids and proteins and tissue analysis independent of solvent.^{[38,39,53,54]}

In conclusion, the first example of transmission geometry tissue imaging using solvent-free sample preparation is reported. The laser-ablated area of ~20 μm is similar to that obtained with solvent-based sample preparation;^[37] however, with solvent-free sample preparation, delocalization of the tissue components is eliminated.^[26] The area of ablation reflects the spatial resolution and, combined with the known thickness of the tissue, transmission geometry allows a straightforward determination of the tissue volume that is analyzed, here ~4000 μm³. The ability to achieve the ablation of adjacent holes of ca. 20 μm diameter suggests that transmission geometry capable of subcellular spatial resolution is within reach. More sophisticated laser optics will be needed to achieve this goal. If observed, sensitivity issues at improved spatial resolutions might be overcome, as shown here, by further optimizing the ion transfer capillary temperatures and matrix deposition. Transmission geometry LSI imaging using an inexpensive nitrogen laser allows tissue images to be obtained in a timeframe comparable with high repetition rate (>1 kHz) lasers. Transmission geometry imaging mass spectrometry is potentially useful for many applications including (for example) surface analysis of synthetic materials such as polymers.