Three-Dimensional (3D) Imaging of Lipids in Skin Tissues with Infrared Matrix-assisted Laser Desorption Electrospray Ionization (MALDESI) Mass Spectrometry

Abstract

Three-dimensional (3D) mass spectrometry imaging (MSI) has become a growing frontier as it has the potential to provide a 3D representation of analytes in a label-free, untargeted, and chemically specific manner. The most common 3D MSI is accomplished by the reconstruction of 2D MSI from serial cryosections; however, this presents significant challenges in image alignment and registration. An alternative method would be to sequentially image a sample by consecutive ablation events to create a 3D image. In this study, we describe the use of infrared matrix-assisted laser desorption electrospray ionization (IR-MALDESI) in ablation-based 3D MSI for analyses of lipids within fresh frozen skin tissue. Depth resolution using different laser energy levels was explored with a confocal laser scanning microscope to establish the imaging parameters for skin. The lowest and highest laser energy level resulted in a depth resolution of 7 microns and 18 microns, respectfully. A total of 594 lipids were putatively detected and detailed lipids profiles across different skin layers were revealed in a 56-layer 3D imaging experiment. Correlated with histological information, the skin structure was characterized with differential lipids distributions with a lateral resolution of 50 μm and a z resolution of 7 μm.

Graphical Abstract

INTRODUCTION

Three-dimensional (3D) mass spectrometry imaging (MSI) has emerged as a competitive complementary method to other 3D imaging techniques (e.g., magnetic resonance imaging, X-ray, computerized tomography scan) because it has the potential to reveal the underlying biology by spatially resolved analyses for a diverse range of biomolecules with chemical specificity [1]. 3D MSI has been successfully employed to evaluate pharmaceutical distribution [2], organ structures [3], lipidomic/proteomics examinations [4, 5], and metabolic exchange between organisms[6]. Throughout these investigations, 3D MSI not only retains the merits from 2D MSI that allows simultaneous label-free analyses of hundreds of biomolecules without prior knowledge but also provides contextual information for 2D images, providing a more accurate view of the biological process by adding the depth dimension.

3D MSI methods can be classified into two major approaches, 1) serial-section-based, and 2) ablation-based (i.e., depth-profiling-based), depending on the ionization methods used. 3D MSI is typically performed by reconstruction of 2D MSI of thin serial sections, which has been applied in many ionization techniques including matrix-assisted laser desorption ionization (MALDI)[7], desorption electrospray ionization (DESI) [8] and laser ablation ionization (LDI) [9]. However, imaging of consecutive sections creates both practical and technical challenges such as time-consuming sample preparation which can lead to sample deformation, as well as the need for morphological information for image registration [8][10]. Furthermore, small anatomical egistratioi structure information in between the sections is likely neglected due to the mismatch between the thickness of the section (typically 10-15 μm) and the submicron sampling depth [11]. To circumvent issues arising from serial-sectioning, alternative 3D MSI operated in ablation-based exposed surface as material is removed by a laser or ion beam [12]. This mode has been employed with ablation techniques such as secondary ion mass spectrometry (SIMS) and laser ablation electrospray ionization (LAESI) [13, 14]. Notably, dynamic SIMS operated in this mode allows subcellular 3D imaging by acquiring a series of images interleaved with etching [15, 16]; however, data interpretation has proved challenging due to the low ion fluxes and fragmentation of biomolecules [17]. Additionally, depending on the nature of the primary ion beam used, it would take tens of layers ion beam ablation to sputter away a single cell, thus not practical for tissue imaging [18].

We recently demonstrated an ablation-based 3D MSI approach with infrared matrix-assisted laser desorption electrospray ionization (IR-MALDESI) on an over the counter pharmaceutical using a multiple unit pellet system (MUPS) formulation [19]. Depth resolution was readily adjustable by altering the laser energy level, which allowed removal of sufficient material (depth resolution on the order of microns), suggesting that IR-MALDESI MSI would be a practical tool for 3D tissue examination in terms of the volume to be analyzed in a reasonable amount of time. Over the past decade, IR-MALDSI MSI has evolved as an innovative technique for mapping molecules ranging from lipids, proteins, metabolites to exogenous molecules with minimal preparation, high salt tolerance, and deeper coverage in a wide array of biological specimens including animal tissue sections [20-25]. To fully exploit the 3D potential of IR-MALDESI MSI, the current study applies IR-MALDESI MSI in 3D tissue imaging using skin, a multi-layer modality, as a proof of principle.

Skin is the largest organ in the body with primary functions of moisture retention, protection against pathogens, temperature regulation, and sensation. To enable these functions, skin has a heterogeneous layered structure consisting of epidermis, dermis, and hypodermis. As the outermost skin layer, the epidermis is responsible for maintaining homeostasis and functions as a moisture barrier to fluid loss and protective barrier against the external environment[26]. The most superficial layer of the epidermis is the stratum corneum (SC), which consists of layers of flat corneocytes interconnected by desmosomes and surrounded by intracellular lipids matrix of cholesterol, fatty acids, and ceramides [27]. The dermis, which provides skin elasticity and resilience to physical stress, is dominated by a collagen matrix that supports fewer elastin fibers, and other components such as blood vessels, hair follicles, and skin glands, such as fat-rich sebaceous glands. The hypodermis (i.e., subcutaneous layer) is the deepest and fat-rich skin layer that contains thin septa that attach the skin to underlying deep faschia of muscle and bone. Lipids serve many roles in the skin, providing moisture barrier, energy storage, signal processing, and structural integrity, which depend on their composition and localization. A limited number of MSI anatomical studies that investigated the compositions and distributions of lipids in the skin layers were imaged on the skin cross sections[9, 28-30]. Cross-sectioned tissue makes it possible to examine skin along its entire depth, providing clarity of the species location. However, profiling a cross-section is not enough to show heterogeneous lipids distributions in both xy plane and z-direction, and sample handling can be problematic, especially the potential alteration of lipids localization caused by the embedding procedure[30]. 3D MSI in ablation-based mode enables volumetric molecular profiling as imaging through a full thickness skin tissue from the surface without a need of sectioning or destroying the sample. This method has only been applied to penetration studies of vitamin C[31] and nickel[32] but not to detailed lipid investigations in the skin.

Here we present the demonstration of IR-MALDESI for 3D tissue imaging by mapping diverse lipids in the skin. Both compositional and 3D spatial distributions of lipid species across different skin layers were revealed, and Individual skin layers were chemically distinguished with differential lipids distributions. In this study, a hairless mouse was selected as a model given that hairless skin requires no need for chemical hair removal, which would alter the skin structure and composition. To date, lipids in the skin are characterized in a 3D manner for the first time here, and a method is provided that permits lipids profiling over a wider field of view; the technique could be expanded for general tissue imaging using 3D MSI without cryosectioning.

EXPERIMENTAL

Materials

SKH hairless mice were graciously provided by the Smart group, Department of Biological Science at North Carolina State University. Four pieces of 1 × 1 cm square, full-thickness dorsal skin sections were excised from the back of a hairless mice 3-15 minutes after sacrifice. The skin samples were thaw mounted to tin foil and placed in a closed chamber containing a pre-cooled plastic weigh boat floating on a dry ice bath containing 95% ethanol (Fisher Scientific, Waltham, MA, USA) to allow the samples to be flash-frozen to avoid enzymatic degradation or cracking. The tin foil with frozen sample was then wrapped around a pre-frozen microscope glass slide, which was maintained on dry Ice and then stored in a −80 °C freezer until the MSI measurements were made. Adjacent slices were collected in 10% neutral buffered formalin (Sigma-Aldrich, St. Louis, MO, USA) and processed to obtain histological data to compare the MS image.

IR-MALDESI Imaging Analysis

The IR-MALDESI MSI platform used in these studies has been detailed elsewhere[33]. Briefly, the frozen sample was placed on the Peltier translational stage cooled to −10 °C. One pulse per burst of an infrared laser (IR-Opolette 2731, Opotek, Carlsbad, CA, USA) at a wavelength of 2940 nm was focused and fired at the sample surface at each voxel of a region of interest (ROI) of 10 × 10. The laser energy level was controlled with q-switch delay time which triggers laser pulses and was measured with a laser power meter (Nova 2, Ophir, Jerusalem, Israel). The thermally desorbed material from the laser was then post-ionized with the charged droplets in electrospray plume and the resultant ions were sampled and measured using a QE Plus orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Automatic gain control was disabled, and the injection time was fixed at 25 ms. The mass resolving power was 140,000_fwhm at m/z 200. Mass spectra were collected in positive mode with a lock mass of 371.1012 to provide high mass measurement accuracy (± 2.5 ppm). A solution of 0.2% formic acid (Sigma-Aldrich, St. Louis, MO, USA) in 50/50 methanol/water (LC/MS grade, Fisher Scientific, Waltham, MA, USA) was prepared as the solvent and infused at a flow rate of 2 μL/min.

After IR-MALDESI MSI analysis, the skin samples were measured using a confocal laser scanning microscope (VK-X1100, Keyence, Itasca, IL, USA) to obtain crater profiles and depth determination. Adjacent dorsal back skin was processed for histological evaluation. Samples were stained with hematoxylin and eosin (H&E) using a standard protocol and photomicrographs of 10X, 40X, and 64X magnification were taken for correlation and comparison with the mass spectrometry imaging.

Data Analysis

After MS data collection, data was converted from .RAW to imzML with a free software raw to imzML converter[34]. All MSI datasets were processed with MSiReader software (v1.02) for visualization and analysis[35, 36]. Especially, the 3D plotting function allowed 3D visualization of ion distributions and could be rotated and viewed from any angle.

Lipids from the skin were putatively assigned with mass and spectral accuracy on the level of the molecular formula. Lipids annotations from the MSI analysis were conducted using METASPACE[37] (https://metaspace2020.eu/), searching against the LipidsMaps[38] database (https://www.lipidmaps.org/), with false discovery rate (FDR) controlled at 10%.

RESULTS AND DISCUSSION

This proof-of-principle study aims to map lipid distributions in the skin and characterize skin structure with differential lipids localizations using 3D IR-MALDESI MSI. Figure 1 shows the histologic anatomy of hairless mouse skin in cross section taken from the same mouse that IR-MALDESI 3D MSI was performed on. The skin sample consists of three main layers including epidermis, dermis, and hypodermis with a total thickness of ~500 μm. Other components such as sebaceous glands and fat cells are labeled. A higher magnification image (box inset) of the epidermis depicts its multiple sub-layers, with the uppermost layer being the stratum corneum (SC). As shown in Figure 1, skin structure varies significantly in both the XY plane and with depth, which makes 3D investigations necessary.

Figure 1.

Histology of hairless mouse skin in cross section with identification of the three main skin layers: epidermis, dermis, and hypodermis (total thickness of ~ 500 μm). Other skin components including sebaceous glands and fat cells are labeled. The box image shows the multiple sublayer structure of epidermis where the stratum corneum is the outer most layer. Hematoxlylin and eosin. 10X magnification.

Depth Resolution Determination

Prior to imaging, depth resolution at different laser energies was explored to establish the imaging parameters for skin as previous studies have found that depth resolution varied with different sample types and laser energies[19]. The lowest (0.5 mJ/pulse) and highest (1.2 mJ/pulse) laser energy levels were applied to ablate skin for 5, 10, 15, and 20 layers, and profiles of resultant craters were visualized with a confocal laser scanning microscope (Figure 2 a, c). Linearity between crater depth and layer number (Figure 2 b, d) suggested that the skin sample was ablated at a relatively constant depth, irrespective of cell types and density variation of the different skin layers. The lowest and highest laser energy level resulted in a depth resolution of 7 and 18 μm, respectively. Since the epidermis thickness was approximately 30 μm, the lowest laser energy (i.e., the best depth resolution) was applied for 3D MSI to resolve the stratum corneum and the remainder of the epidermis.

Figure 2.

Depth resolution as a function of laser energy was determined using hairless mouse dorsal skin. Profiles of the craters after ablation for 5, 10, 15, 20 layers at a) 0.5 mJ/pulse and c)1.2 mJ/pulse were examined with optical images, top and side view of laser images using a confocal laser scanning microscope. b) d) Follow up depth measurements show a linear relationship between ablation depth and layer number for both laser energy levels. Depth resolution was determined as 7 and 18 μm at 0.5 mJ/pulse and 1.2 mJ/pulse, respectively.

After the parameters for an acceptable depth resolution was established, a 56 layer-3D IR-MALDESI MSI at a lateral resolution of 50 × 50 μm and depth resolution of 7 μm was performed on a full-thickness skin tissue. To keep the sample surface at the appropriate focal depth of the laser, the sample stage was manually raised up by 50 μm after every 7 imaged layers, which is the best accuracy achieved by the current platform since the smallest tick mark denotes 25 μm in z-direction.

A total of 594 unique ions were putatively annotated as lipids with high mass accuracy (± 2.5 ppm) of the high-resolution accurate mass (HRAM) mass spectrometry (a full list of putative annotations is included in Supplementary Information (ESM) Table S1). Among the 594 lipids, major lipids classes were present such as fatty acyls, glycerolipids, glycerophospholipids, and sphingolipids. The average number of lipid annotations were recorded as a function of estimated sample depth and a v-shaped curve appeared as shown in Figure 3a. This indicates that lipids compositions change as sample depth and lipid variability exists across skin layers, which is expected as cell types change across skin layers. Two representative single acquisition mass spectra (i.e., not averaged) from the 1^st (top) and 56^th layer (bottom) of the 3D skin imaging experiments are displayed in Figure 3b with some annotations labeled. As shown in the figure, significant differences reside in the higher mass range (500-1000 m/z). Spectra from the 1^st layer are featured with ceramide cluster mainly resident between 600-700 m/z, while spectra from the 56^th layer are dominated by phosphatidylcholine (PC) and phosphatidylethanolamine (PE) between 700-800 m/z. Many other lipids such as phosphatidylglycerol (PG), phosphatidylserine (PS), sphingomyelin (SM), diacylglycerol (TAG), and triglyceride (TG) were potentially detected in both spectra, which is comparable to lipids classes found in previous skin imaging experiments using LDI and MALDI[9, 39]. Although further structural information (e.g., location of double bonds) requires fragmentation with tandem mass spectrometry (MS/MS) for isobaric lipid determinations, putative identifications at the molecular formula level are sufficient for the demonstration of 3D imaging without the need for cryosectioning. Those lipids detected, upon further investigation into the literature, are also supported by their presence in mouse skin in previous MSI experiments[9, 39, 40].

Figure 3.

A total of 594 unique ions were putatively identified as lipids with high mass and spectral accuracy from a 3D IR-MALDESI MSI analysis for 56 layers. a) Lipids annotations counts were recorded as a function of sample depth estimated by depth resolution. Annotations decreased with sampling depth for the first 200 μm and then increased for 250-400 μm. The colored boxes shows depth region of different skin layers determined by the histological measurement; “E” represents epidermis, “D” represents dermis, and “H” represents hypodermis; b) Representative single acquisition mass spectra from first layer (1) and last layer (56) of 3D MSI which revealed lipids profiles in the skin, with some annotations labeled.

Various lipids involve in skin physiology and their distribution provides insight into skin structure and function. Specifically, abundance changes in three dimensions can be used to differentiate skin layers due to the varying presence of species in epidermis, dermis, and hypodermis. For example, the ion at m/z 446.2395 has been detected in other skin imaging experiments and was previously assigned to sphingosine 1-phosphate (S1P) in the dermis region[9]. As a bioactive lysophospholipid, S1P plays roles as an intracellular and extracellular message receptor and has been shown to have significant abundance differences between disease and control skin model[9, 41]. In this study, m/z 446.2395 shows noticeable higher ion abundance in the layers 8-33, compared to layers 1-7 and 34-56 (Figure 4a). The average abundance of each layer of 56 was plotted as a function of sampling depth estimated by depth resolution. Two clear interfaces show up between depth at layer 7 and 8, 33 and 34 (Figure 4b), which roughly matches skin regions differentiation with histological staining, suggesting the assignment of imaged layer 1-7 to epidermis, layer 8-33 to dermis, and layer 34-56 to hypodermis. Given the defined depth resolution at 0.5 mJ/pulse, the thickness of epidermis was estimated as 49 μm and the dermis 182 μm. Since the sample was not ablated through completely, the hypodermis thickness was not estimated. This was corroborated using histological preparations which indicated the epidermis was μ50 μm and the dermis was ~200 μm.

Figure 4.

Typical lipid distribution in mouse skin was investigated. a) Distributions of S1P at m/z 446.2395 across 56 imaged layers of skin; b) Average abundance of S1P as a function of sample depth was recorded and average abundance of region assigned to dermis is significantly higher than that of epidermis and hypodermis region; c) A representative mass spectra of cholesterol [M+H⁺-H₂O]⁺ at m/z 369.3516, overlaid with the theoretical isotopic distribution in red dots and d) its distribution across 56 layers in the skin; Other selected IR-MALDESI mass spectrometry images showed spatial distributions of e) ceramide (Cer) at m/z 634.6132 and f) TG at m/z 879.7436.

Epidermal lipids such as cholesterol, wax esters, and TG determine the quality of skin barrier function. [M+H⁺-H₂O]⁺of cholesterol at m/z 369.3516 (isotopic pattern in Figure 4c) is found distributed across all skin layers with strong localization to the epidermis (Figure 4d). The average abundance of each layer of 56 was plotted as a function of layer in ESM Fig. S1a with a rapid decrease in abundance. The average abundance of the first 7 imaged layers was found to be 4-fold higher than the average of layers 8-56, which is in concordance with the known presence of cholesterol in the cell membrane and preferential localization in the epidermis[42]. The abundance change between the 7^th and 8^th imaged layers is not abrupt mainly because the interface of epidermis and dermis reaches at different depths due to varied skin structures at different spots, as each skin layer is not homogenous.

Stratum corneum (SC) is the most superficial sublayer of the epidermis. Besides cholesterol, the other two main lipids in SC namely fatty acids and ceramides have been widely studied due to their pivotal roles in barrier properties. Alternation of those species is related to metabolic abnormality causing diseases such as atopic dermatitis or eczema. Ceramides is especially crucial in preventing water loss due to its lipophilic properties[43]. The heat map of a tentatively identified ceramide in Figure 4e demonstrated its dominance in 1^st image and sparse distribution in the 2^nd imaged layer. A similar distribution of fatty acids (an example is shown in ESM Fig. S1b) suggested the same origin of these species. This indicated that the 1^st and part of the 2^nd imaged layer are most likely corresponding to SC, consistent with the known presence of these intercellular lipids between corneocytes in SC. Hence, the thickness of SC was estimated to be 7-14 μm, which was comparable to 9 μm, the mean thickness of mouse back skin reported in literature[44]. The estimation was not compared to histology because the true thickness of the stratum corneum is not best visualized by histology due to artifact expansion of the SC layers during staining, which evidenced by open basket weave pattern in the image.

The hypodermis is the lowermost section of the skin characterized by a rich abundance of lipids including PC, PE, DG, TG, and SM. Among them, TG is a neutral lipid and is hardly seen in MALDI MSI due to ion depression of phospholipids and interfering matrix signal. TG at m/z 879.7436 has been confidently identified in the skin by Fincher et al with LDI MSI together with MS/MS[9]. The high ion abundance of TG (Figure 4f) and a collection of various other lipids (ESM Fig. S2) confirms that the hypodermis was being represented in layers 34-56. Only part of the hypodermis (~161 μm) was imaged, as the skin was not fully ablated through.

Interfaces between the epidermis, dermis, and hypodermis are clearly visible in the colocalized heatmaps of three selected ions assigned to cholesterol, TG, and S1P (Figure 5). The histological staining image in Figure 1 was correlated to the overlaid image to corroborate the skin layers structure. Even though there are thickness variations at each xy coordinate, those thickness estimations roughly match the histological results, so that the constructed 3D MSI is a reasonably accurate representation of the skin. 3D mappings of three representative ions were constructed in ESM Fig. S3 and their colocalized 3D heat map (Figure 6) provides a better view of skin structure.

Figure 5.

Correlation of MSI images with H&E stained histological image. Overlaid image of three putatively identified lipids (cholesterol at m/z 369.3516 in the green channel, S1P at m/z 446.2395 in the blue channel, TG at m/z 879.7436 in the red channel) shows a clear interface between the 7^th and 8^th layer as well as the 33^rd and 34^th layer. The estimated thickness of the epidermis and dermis is in agreement with the histological data in Figure 1.

Figure 6.

3D three-color overlaid image of cholesterol at m/z 369.3521 (green), S1P at m/z 446.2395 (blue), and TG at m/z 879.7436 (red). The lateral resolution is 50 μm and the depth resolution 7 μm. The magenta color represents the concurrent presence of blue and red color.

We now have evidence that 3D IR-MALDSI MSI is a feasible method in tissue imaging, with micrometer resolution comparable to histological imaging. None of the serial cryosectioning, sample embedding, or chemical matrix is employed, which greatly simplifies the preparation and data interpretations, making 3D MSI a more accessible technique free from tremendous manipulations. It should also be noted that fat tissue underneath the skin section presents a challenge in sample mounting due to poor adhesion to glass slides. Other than laborious solutions such as slides surface modification or using alternative gold plates[45], we thaw mounted skin on a tin foil and wrapped the foil over a frozen glass slide, which provides an alternative of sample mounting methods for skin imaging experiments.

As previously mentioned, all lipids identifications in this experiment are putative using high mass and spectral accuracy, supported by comparison with previously reported studies, which fulfills the demonstration purpose of this study. Exact lipid structural elucidation would be performed with tandem mass spectrometry in future biologically driven studies. Also, further investigations towards skin substructure demand a better spatial resolution for more details. For instance, stratum spinosum containing lamellar granules might be resolved from stratum corneum and gradient lipids distributions as a result of keratinization and cornification could be observed at a higher resolution. Finally, we will implement an autofocus system to IR-MALDESI platform to enable 3D imaging in line with the anatomical appearance of the samples and allow automatic z-stage movement during imaging. Future experiments will not only involve these fundamentals improvements but also biological applications

3D IR-MALDESI generates volumetric molecular imaging without cryosectioning, water removal, or embedding which might change the location of molecules. This allows us to directly access the entire tissue for detailed analysis of molecules such as lipids and metabolites, with clarity of their localization in native state. The method developed in this present proof-of-concept study can be translated to different studies such as studying wound healing, skin cancer, graft rejection, infection. For example, 3D IR-MALDESI allows 3D lipidomics and metabolomics profiling from different skin regions, which will reveal how lipids migrate in the wound site as well as how the cellular environment surrounding the wound site change. This will offer new insights into how skin repairs a wound and may provide better diagnostics and treatment protocols for chronic wounds.

CONCLUSIONS

This was a preliminary study to demonstrate the feasibility of 3D IR-MALDESI MSI with tissue imaging. The suitability of the skin for 3D MSI has been described regarding size and multi-layer characteristics. A total of 594 lipids were tentatively identified, then compositional and 3D spatial distributions of lipids across a nude mouse skin were revealed for 56 layers. The skin structure was successfully resolved with differential lipids distributions, which was comparable to histological images. The method requires further optimization with exact lipids structure identification, resolution improvement, and integration of anatomical information. Nevertheless, 3D IR-MALDESI MSI enables direct volumetric molecular visualization in tissues without cryosectioning and it is foreseeable to investigate lipids and metabolites profiles in the wound healing process, penetration pathways of exogenous compounds, and tumor or cancer margin determinations.