Spatial Neurolipidomics - MALDI Mass Spectrometry Imaging of Lipids in Brain Pathologies

Abstract

Given the complexity of neuronal tissues, understanding neurochemical pathophysiology puts high demands on bioanalytical techniques with respect to specificity and sensitivity. Mass spectrometry imaging (MSI) has evolved to become an important, biochemical imaging technology for spatial biology in biological and translational research. The technique facilitates comprehensive, sensitive elucidation of the spatial distribution patterns of drugs, lipids, peptides, and small proteins in situ. Matrix-assisted laser desorption ionization (MALDI)-based MSI is the dominating modality due to its broad applicability and fair compromise of selectivity, sensitivity price, throughput, and ease of use. This is particularly relevant for analysis of spatial lipid patterns, where no other comparable spatial profiling tools are available. Understanding spatial lipid biology in nervous tissue is therefore a key and emerging application area of MSI research.

The aim of this review is to give a concise guide through the MSI workflow for lipid imaging in neuroendocrine tissue and essential parameters to consider while developing and optimizing MSI assays. Further, a broad overview of key developments and applications of MALDI MSI-based spatial neurolipidomics to map lipid dynamics in neuronal structures and ultimately to further understand neurodegenerative disease pathology.

1. Introduction

With the increase in average life expectancy for the world’s population, the prevalence of age-related diseases, including neurodegenerative diseases such as Alzheimer’s disease (AD) and Parkinson’s disease (PD), is also increasing. Much detail on the pathophysiological mechanisms underlying neurodegenerative disease pathology remains elusive, but mounting data point to a role of neuronal lipid species^1–3. As much as 90% of the brain dry weight consists of lipids though the functional role of brain lipids in distinct neuropathological processes has long been underestimated. Systemically, lipid metabolism has long been known to be essential in multiple relevant metabolic and immune response mechanisms^4–6. However, recent findings have implicated neuronal lipids in neuropathological processes beyond neuroinflammation such as neuronal signaling and amyloidogenic protein aggregation^7–9. For instance, variation in genes encoding lipid transporter proteins mainly APOE^9–12 and lipid-sensing microglial surface receptors (e.g., TREM2)^13–15 have been associated with an increased risk to develop sporadic AD¹⁴. Similarly, mutations in glycolipid-processing enzymes glucocerebrosidase (GBA) are associated with both familial forms PD ³ and lysosomal storage disorders such as Niemann-Pick disease¹⁶, Sandhoff disease and Guillain-Barré syndrome.

A functional understanding of how lipids modulate neuropathological processes is still just beginning to emerge, highlighting the need for further lipid-centric neuroscience research making use of novel tools such as mass spectrometry. Significant advances in our understanding of CNS-related molecular mechanisms in neurodegeneration have been made because of the development of novel, advanced, high-resolution imaging techniques.

Commonly used spatial techniques in molecular biology research include immunohistochemistry for proteins and lipids and in situ hybridization for nucleotides (DNA/RNA) that are interfaced with brightfield and fluorescent microscopy.

The major challenge in biochemical imaging relates to chemical specificity, selectivity, and spatial resolution. This is particularly relevant for lipid detection, as this is not possible using the conventional biological imaging strategies that are used for protein- (IHC) or nucleotide- (PCR, FISH) detection. Spectroscopic methods such as infrared^17–19 or coherent anti-Stokes Raman spectroscopy (CARS)²⁰ are used to for chemical imaging of e.g. lipid classes and their saturation across tissue sections though do not facilitate identification of single lipid species.

A relatively novel molecular imaging technology, mass spectrometry (MS)-based imaging (MSI)^21–24, has, over the last years, made significant inroads into biomedical research and neuroscience research, in particular^25–29

Due to the revolutionizing development of soft ionization techniques such as MALDI³⁰ and electrospray ionization (ESI)³¹, mass spectrometry became the method of choice for protein characterization fueling the development of proteomics³². Further, MS allows fast, sensitive, and specific detection and characterization of intact large biomolecules, including lipids, peptides, and proteins. Though MS-based approaches to tissue extracts facilitate sensitive lipidomic profiling, no accurate spatial information is maintained. Due to the complexity of the CNS, the spatial information of neurochemical distribution patterns is of significant interest to delineate ongoing neuronal mechanisms.

Being a mass spectrometry technique, MSI is characterized by providing high chemical specificity at high sensitivity²⁹. Due to the versatility of the detector range, MSI allows for comprehensive spatial omics profiling of either lipids, metabolites or small proteins in a single tissue imaging experiment. This way typically more than 100, and for dual polarity experiments even >200 different lipid species³³ can be discerned easily at 10 μm resolution.

2. Mass spectrometry imaging

Different MSI modalities have been introduced based on different means of analyte desorption and ionization from the biological sample. The most prominent MSI modalities include laser-based systems such as MALDI MSI as well as less common but complementary tools such as LA-ICP and LAESI. An emerging MSI modality is desorption electrospray ionization (DESI)³⁴. Given the advantage of an atmospheric pressure interface and no requirements for sample preparation, DESI is gaining popularity in research areas spanning from field work to diagnostics and drug development. Surface-enhanced laser desorption/ionization (SALDI) methods offers another LDI based MSI option with no matrix requirement while enabling imaging at high resolution with less noise^{35, 36}. Finally, ion beam-based techniques, i.e., secondary ion mass spectrometry (SIMS)³⁷, are more and more used for biochemical imaging both for lipids but also for proteins.

The principle of MSI is based on sequential acquisition of single mass spectra in a predefined acquisition array across a tissue section. Here the difference between two acquisition events define the pixel resolution. The exact spatial resolution depends on whether a discrete acquisition approach (stepwise moving stage and probe) or a scanning probe and/or a continuously moving stage is used. Aside from this consideration the intensity distribution across the tissue array can be visualized for a given mass signal yielding single ion images (Figure. 1). Different MSI technologies have various strengths and limitations, particularly regarding spatial resolution, molecular information, and mass range, respectively.

Figure 1: Principle of MALDI Mass Spectrometry Imaging.

(A) MALDI MS is performed in situ, where matrix-covered tissue sections are analyzed. Analyte ions are generated upon laser irradiation and transferred into a mass analyzer, e.g., a time of flight. A single mass spectrum is acquired for every xi,yj coordinate of a predefined tissue array defining the image area and pixel distance. (B,C) Single ion images are generated by mapping the intensity of an individual ion signal (m/z; relative intensity) over the whole tissue slide⁸⁷. (C) Example of MALDI MSI of transgenic AD mouse brain showing lipid localization to amyloid plaque pathology. Scalebar: 50μm

MALDI MSI represents a compromise between molecular mass range, sensitivity, and specificity. The technique is based on in situ analyte desorption and ionization upon irradiation-with a UV laser. In MALDI MSI, a crystalline UV radiation-absorbing matrix is sprayed onto the tissue prior to laser irradiation to facilitate the desorption and ionization process³⁰.

The concept of MALD MSI was demonstrated by Bernhard Spengler in 1994²⁴ and Richard Caprioli in 1997²². The main advantage of MALDI over other techniques is its mass range and selectivity, making it a versatile approach for chemical imaging of both low and high molecular weight species ranging from smaller drugs to complex lipids and small proteins. Aside from its versatility, MALDI is further characterized by its fair tradeoff of high resolution, specificity, sensitivity, robustness, and acquisition speed. In terms of resolution, commercial MALDI systems can achieve 5um. Instrumental advancements have allowed for even higher spatial resolution, such as 1.4μm³⁸ with an in house-developed highly focused MALDI source mounted onto an orbitrap system or even 0.6μm using laser post ionization and a transmission mode instrument geometry³⁹. Therefore, MALDI MSI is currently the most common MSI methodology.

Therefore, the present review focuses on MALDI MSI and aims to provide a concise overview of its application for lipid imaging in brain tissue. Here, a step-by-step guide through the experimental setup and workflow is given to enable the reader to replicate MALDI MSI experiments in-house successfully. Different advances and challenges concerning sample preparation, data acquisition, and validation, are discussed. Finally, an overview of MALDI MSI-based lipid applications focusing on neurodegenerative disease pathology is provided.

3. Experimental Considerations

3.1. Sample Preparation for MALDI imaging of lipids

For MSI-based lipid imaging, various target preparation parameters are of critical relevance as they significantly impact final data quality. This particularly concerns signal intensity, reproducibility, and lateral (i.e., spatial) resolution. The sample preparation workflow in MALDI MSI comprises tissue retrieval and storage, tissue sectioning, tissue wash, and matrix application. Tissue handling during recovery (dissection) and slice collection is critical. Similarly, matrix application facilitating sufficient analyte extraction while avoiding lateral diffusion and large crystal formation are cornerstones for each MALDI MSI experiment. These steps must be adjusted for both the type of tissue and, most importantly, the molecular target of interest.

3.1.1. Tissue Handling

For most MSI experiments, including lipid imaging, fresh frozen tissues are used due to the interference of fixation and embedding chemicals with MS analysis. Removal of those polymers is typically achieved through organic washed which consequently would also remove lipids. Hence those samples do not permit the analysis of endogenous lipids due to the extensive washing for polymer removal. Therefore, fresh frozen (FF) tissue samples are preferred for MSI. To ensure appropriate quality of FF tissue, tissue retrieval and storage is essential. Starting with sample retrieval, typically a short post-mortem delay until dissection, collection and processing (freezing) is critical to minimize analyte degradation^{40, 41}. Dissected fresh tissue samples are typically snap-frozen on dry ice or liquid propane or nitrogen and stored at −80°C.

3.1.2. Tissue Sectioning

One of the first and crucial steps in the MALDI MSI workflow is tissue sectioning. The tools and objects should be cooled to the operating temperature of −20°C. Tissues should be transported in a closed container on dry ice and allowed to acclimatize. The cryotome chamber needs to be surface-sterilized with a disinfectant like ethanol. The tissue is mounted to the cryostat chuck by placing a small drop of embedding media (e.g., optimal cutting temperature (OCT) compound), putting the chuck into the cooling chamber of the cryostate and holding the tissue piece onto the mounting medium with ice cold tweezers while the mounting medium freezes. In contrast to cryosectioning protocols used in pathology and histology, tissues cannot be immersed/embedded in OCT for cutting, and contamination with OCT of the cutting surface of the tissue, the cryocollection table or the blade must be avoided at all costs.

The OCT compound should be allowed to solidify at −20°C. The tissue-mounted chuck is then attached to the holder, and sectioning can begin with a thickness of 12 μm. The sections can be transferred onto a glass slide using a brush. Finally, the tissue can be thawed by warming the glass slide from underneath with a finger. It is essential to ensure that the tissue is completely thawed before proceeding with any further staining or analysis. Thawing the tissue too quickly or unevenly can lead to damage and loss of cellular integrity. Therefore, allowing the tissue to thaw at room temperature for sufficient time is recommended before continuing with the desired protocol. It is important to note that the operating temperature and blade condition can affect the outcome, and precautions should be taken to avoid humidity in the chamber. To ensure optimal results, it is recommended to maintain a controlled environment with low humidity during the sectioning process. Additionally, the blade should be regularly held and sharpened to achieve clean and precise sections. Following tissue sectioning the composed sections must be dried in a desiccator for 10min before storage at −80°C to prevent damage by water condensation during freezing⁴².

3.1.3. Tissue Washing and Matrix Application

One of the biggest sources of interference is tissue preparation. This step involves the thawing and washing of the stored cryosections and subsequent matrix application. The initial washings are used to remove contaminants and salts. In the case of lipid imaging this step involves washes with ammonium salts^{43, 44}, while peptide and protein analysis involve organic washes or organic wash accompanied by buffer wash for lipid removal and protein precipitation^{45, 46}. Matrix application can be performed using different approaches of deposition and roughly be divided into three categories: a) nebulizer/sprayer; b) microspotting and c) sublimation.

The application methods are chosen with respect to the analyte targeted, spatial resolution, reproducibility and sensitivity. Desorption and ionization of charged lipid molecules is easily achievable due to their high abundance. Matrix-free, surface-assisted LDI methods have been reported to accomplish this task using functional surfaces such as DIOS⁴⁷, NIMS ³⁶ or Si-nanorods^{48, 49}. These methods are obviously superior with respect to spatial resolution and sample handling. However, they have limitations regarding tissue preparation (sectioning), requiring very thin tissue sections along with sensitivity and reproducibility issues. Consequently, matrix-assisted LDI is more prominent, given the rather straightforward sample preparation, increased sensitivity, range of analytes and robustness.

The most used approach involves matrix-spraying using commercially available robotic nebulizer-assisted pneumatic sprayer systems, such as the TM or M3+ Sprayer (HTX Technologies), SunCollect Sprayer (Sunchrom), or iMLayer (Shimadzu), along with open-source or even manual airbrush systems.

Potential sources of error during matrix application include the matrix solvent being too dry or too wet, which could result in solvent evaporation or analyte diffusion; thus, there is a need to be observant for leaks and clogging during runs. To overcome these challenges, other approaches, such as using a heated spray head (HTX Sprayers), dry coating⁵⁰ or matrix sublimation^51–53, have been introduced allowing to control crystal size^{50, 52}. While sublimation provides significant advantages in terms of crystal size, the earlier approach had limitations with respect to repeatability and extraction efficiency; however, recent developments make matrix sublimation a potent alternative for lipid analysis⁵⁴. .

Different matrices have been proposed for MALDI MSI of various molecular species, including other lipids varying in polarity and hydrophobicity. For lipid analysis, the most used matrices are 2,5-dihydroxy-benzoic acid (DHB)⁵⁵, 1,5 di-amino-naphthalene (1,5-DAN)⁵¹, norharmane⁵⁶, 9-aminoacridine (9-AA)⁵⁷, N-(1-naphthyl) ethylenediamine (NEDC)⁵⁸ and 4-hydroxyl-alpha-cyano-cinnaminic acid (CHCA). The choice of matrix depends on the targeted analyte class, crystal size, mass range, matrix cluster formation.

3.2. MALDI Mass Spectrometry Experimental

3.2.1. MALDI MSI instrumentation

Several MALDI MSI solutions are commercially available. Generally, MALDI MSI instruments can be divided into high vacuum MALDI ToF systems and medium pressure/atmospheric pressure (AP) MALDI setups, where a MALDI ion source is mounted onto an otherwise ESI MS setup.

MALDI ToF systems such as Shimadzu’s Axima, 8030 and 7090 as well as Bruker’s rapifleX/ultrafleXtreme provide very high sensitivity and acquisition speed and have advantages in terms of mass range, e.g., endogenous peptide analysis. Limitations of those setups lie with mass resolution, particularly for compounds over m/z 6000 that are analyzed in linear mode, the systems complexity with a high vacuum source and MS/MS analysis typically relying on post source decay with ion gate-based precursor ion selection. This approach has limitations with respect to precursor ion selection by flight time and, particularly for smaller molecules, and fragmentation efficiency resulting in lower MS/MS sensitivity while consuming sample due to increased laser power. Those challenges have been addressed by including collision chambers or ion traps for CID/CAD fragmentation as found in both Bruker’s and Shimadzus’ MALDI ToF systems, along with MSⁿ capabilities warranted by the iontrap (IT) in the Axima QIT-ToF systems⁵⁹. A limitation of MALDI-ToF systems is mass resolution, which is typically limited to 40000 in reflector mode, and which makes it difficult to discern low molecular weight species from matrix signals. This is overcome in more exotic systems such the Jeol Spiral ToF^{60, 61} or SimulToF⁶², which allow to obtain mass resolution of 5000 (SimulToF) up to 100000 (SpiralToF).

AP or medium-pressure MALDI MS systems bring together the advantages of MALDI MSI in terms of spatial resolution with higher mass resolution (>40000) and more complex mass analyzers including ToF (Shimadzu iMScope⁶³ and Waters Synapt) or FT analyzers such as Bruker’s solariX and scimaX FTICR MS and Thermo MALDI-orbitrap equipped with an intermediate pressure MALDI source^{38, 64}). Further, ion mobility-ToF hybrid analyzers (Bruker TIMS ToF flex, Waters Synapt IMS/cyclic IMS) are available, which are particularly beneficial for resolving isomers in lipid analysis^{19, 65–68}. AP or medium-pressure MALDI MSI systems have limitations in, e.g., terms of sensitivity towards higher molecular weight compounds (peptides and proteins) but offer significant advantages for characterizing and quantifying smaller molecules (lipids) since those systems offer higher mass accuracy, ion mobility and superior MS/MS or MSⁿ capabilities. For MALDI-FTICR and MALDI-orbitrap MSI, some challenges remain with respect to scan speed to the benefit of superior mass resolution.

3.2.2. MALDI MSI Data Acquisition

The parameters for a MALDI imaging experiment are typically specified in an acquisition sequence, specified in the respective MS imaging software. Here, all the acquisition and initial data processing parameters are specified along with the acquisition region, acquisition pattern, and spot-to-spot distance, i.e., spatial resolution. The initial steps of setting up the imaging acquisition experiment involve image registration and teaching for sequence setup, performed in the imaging software provided by the vendor.

For the actual MALDI acquisition method, the optimum number of good-quality spectra that can be obtained from each position must be specified, including the number of laser shots and laser intensity (i.e. fluence). Too high laser fluence can influence the experiment dramatically, leading to signal depletion due to oversampling effects⁶⁹. Here, the spatial resolution and laser focus settings must be considered accordingly, as small spot-to-spot distances at too high laser fluence lead to the described oversampling phenomena. For a typical DHB or DAN matrix for lipid analysis, 5–10 shots at lower energies are needed to allow straightforward acquisition of imaging data at 10μm⁶⁹. When specifying the pixel distance and spatial resolution, a further consideration is the acquisition time. Assuming a MALDI instrument with a 10kHz laser is used, acquiring extensive areas at very high spatial resolution will take several hours. During this time, the matrix might sublimate away, making it impossible to acquire reliable data.

Another critical factor to consider during data acquisition is the external calibration of the instrument before sample analysis to ensure accurate and reliable lipid identification. In lipid MALDI MSI experiments, some of the routine external calibrants used are red phosphorus^70–72, a mixture of standard peptides^{51, 70–76}, or a mix of lipid standards^{77, 78} (Table 1). It is always mandatory to include at least two calibration spots placed in two corners diametrically over the slide, and save the spectra to go back later and recalibrate.

Table 1.

Example of routinely used external calibrants for lipid analysis through MALDI-IMS. 2,5-Dihydroxybenzoic acid (DHB), sinapinic acid (SA), α-cyano-4-hydroxycinnamic acid (CHCA), N-naphthylethylenediamine dihydrochloride (NEDC), 2’,4’,6’-trihydroxyacetophenone monohydrate (THAP), 1,5-diaminonaphthalene (DAN).

Study	External calibrants	Matrix	Tissue	Instrument
Kaya I. et. al.71	Red Phosphorus	Norharmane	Non-human primate coronal brain	MALDI-FTICR (SolariX, Bruker)
Angelini R. et al.78	Mixture of phosphatidylcholine and lysophosphatidylcholine	CHCA	Mouse brain	MALDI TOF/TOF (UltraFlextreme, Bruker)
Liang X. et al.76	Mixture of DHB, SA, CHCA, and peptides.	NEDC	Whole Zebrafish	MALDI-TOF/TOF (RapifleX, Bruker)
Tobias F. et al.72	Red Phosphorus	9AA (lipids) DHB (Cholesterol)	Mouse brain	MALDI TOF/TOF (Sciex)
Kaya I. et al.129	Mixture of standard peptides	DAN	Mouse brain	MALDI TOF/TOF (UltraFlextreme, Bruker)
Mallah K. et al.75	Mixture of standard peptides	DHB	Rat brain	MALDI-LTQ-Orbitrap-XL (Thermo Fisher Scientific)
Prentice BM. et al.73	Mixture of standard peptides for positive mode calibration and matrix clusters for negative mode calibration	DAN	Mouse brain	MALDI TOF/TOF In-house built
O’Rourke MB. et al.70	Red Phosphorus	DHB	Human brain	MALDI TOF/TOF (UltraFlextreme, Bruker)
Jackson SN. et al.77	Mixture of phosphatidylcholines and sphingomyelins	DHA	Rat brain	MALDI-TOF/TOF (Applied Biosystems)
Thomas A. et al.51	Lipid and peptide mixture	DAN	Mouse brain, liver Fish whole body	MALDI TOF/TOF (UltraFlextreme, Bruker)

3.3. Data Analysis

3.3.1. Data Processing

Imaging data acquisition is generally followed by data processing, including baseline subtraction, peak picking, and calibration. On-the-fly MS processing uses a baseline removal algorithm that will not result in negative data (e.g., ConvexHull 0.8 flatness; FlexAnalysis, Bruker Daltonics). Peak picking is performed using sensitive algorithms (Centroid, SNAP) considering peaks with a minimum signal-to-noise (S/N) >3. There are several tools for data analysis for MSI, such as, Flex Analysis (Bruker), Image Reveal (Shimadzu), High Definition Imaging (HDI, Waters), ImageQuest (Thermo Fisher Scientific, Massachusetts, USA) as well as commercial and free-standing, vendor independent options such as Lipostar⁷⁹, SCiLS (Bruker), Mozaic (Spectroswiss, Lausanne, Switzerland), MALDIVision, MSiReader⁸⁰, Cardinal⁸¹, Biomap⁸², msiQuant⁸³ and SpectralAnalysis⁸⁴.

3.3.2. Statistical Analysis

The complexity of high content MALDI MSI data poses a significant challenge for data analysis. High resolution imaging data can result in multiple gigabyte file size that for comparative/multimodal analysis of multiple datasets easily exceeds the computational power of standard systems and highlights the need to balance imaging area, mass range and spatial resolution. A typical MSI data analysis workflow involves three steps: 1) segmentation and region of interest (ROI) identification 2) ROI spectral analysis and finally 3) univariate analysis.

To approach the entire MSI dataset, the first step involves a reduction in dimensionality using multivariate data analysis (MVDA) tools such as principal component analysis (PCA)⁸⁵, maximum autocorrelation factor analysis (MAF)²⁸, hierarchical cluster analysis/bisecting k-means clustering^{85, 86} and orthogonal projections on latent structure analysis (OPLS)^{85, 87}. The aim is to reduce data complexity and capture tissue heterogeneity encoded in covariate patterns of individual chemical species in discrete MVDA factors (e.g., principal components) or clusters, respectively. The goal is then to annotate distinct histological features that are captured into the same MVDA factor and cluster based on their chemical similarity which is reflected in either covariance (PCA) or proximity (HCA). From the corresponding scores and loadings, the variables contributing the most (i.e., mass peak values) to these variances can be deduced, revealing histology-associated chemical changes which allows to annotate the ROI. These ROI can further be annotated based on immunohistochemical staining subsequently performed on the same tissue section⁸⁵. This approach serves also as a means of validation for the MVDA-based image segmentation.

The annotated regions of interest can the further investigated through both supervised and unsupervised MVDA tools such as PCA, HCA, soft independent modelling by class analogy (SIMCA), orthogonal partial least square (OPLS), multiblock OPLS (OnPLS), OPLS discriminant analysis (OPLS-DA) for either classification or discriminative analysis^{85, 87}. Ultimately, univariate tools such as grouped comparisons (e.g., ANOVA, t-statistics, Mann Whitney) or regression modelling (linear and logarithmic) can be used to validate the segmentation and MVDA data as well as verify the histological relevance of the identified chemical localization patterns.

3.4. Lipid Identification

While IMSMSI is a very powerful approach for in situ lipid profiling, there is a need for complementary validation strategies are needed to fully exploit the potential of the technique fully. Like protein and peptide imaging, initial mass peak annotation is based on accurate mass and reference values reported in the literature or online depositories, such as the LIPID MAPS⁸⁸, Metaspace⁸⁸, and SwissLipids Knowledgebase^{89, 90}. A strong advantage in the case of imaging of low molecular weight compounds, such as lipids, is the resolving power of high-end mass analyzers in that mass range, particularly Fourier transform (FT) instruments, but also reflector ToF analyzers. This facilitates compound identification based on low to sub-ppm mass accuracy and the elemental composition deduced by the resolved true isotope pattern. Increased confidence in identification is achieved by follow-up MS/MS or MSⁿ fragmentation analysis and structural characterization⁷⁷. However, regarding MS/MS, in contrast to protein identification, this step is more challenging as lipids compounds, and their fragmentation patterns are far more diverse. This renders the generation of reference databases, such as those available for protein sequences, very challenging. Notable attempts to generate MS/MS spectral libraries have been made to address this issue but are limited due to the vast number of diverse lipid classes and species^{90, 91}.

A further significant challenge in lipid characterization is the identification of lipid isomers. This concerns the identification both structural isomers and stereoisomers⁹². This is one of the most significant challenges, as isomeric lipid species can have entirely different biological functions. Very elegant approaches by in vitro and even in situ derivatization have been proposed^93–96. Most notably, methods using gas phase ozonolysis in lipidomics and, most recently, online within an imaging experiment have been presented^{93, 97}. Similarly, an approach using in situ modification via a Paterno-Büchi reaction have been reported for localizing double bonds in isomeric phosphor- and glycolipids^{94, 96}.

Ion mobility MS (IM-MS) presents another compelling method for identifying various lipid classes and structural isomers⁶⁵. The utilization of matrix-assisted laser desorption/ionization trapped ion mobility spectrometry time-of-flight based mass spectrometry imaging (MALDI TIMS TOF MSI) in lipid identification has been a significant endeavor, as it has enhanced the reliability of correctly identifying lipids by including additional physical factors such as collisional cross-section (CCS) with improved mass accuracy⁶⁷. This further allows identification by comparing with ex situ lipidomic data acquired through LC-IM-MS⁶⁷. Together, these advances indicate that lipid characterization, in vitro and in situ, is becoming increasingly robust and routine.

4. MALDI MSI of neuronal lipids in CNS diseases

4.1. Neurodegenerative Diseases

Lipid analyses are relevant in neuroscience research as they have significant roles as both signaling molecules and structural interactions with membrane proteins. Moreover, it has been established that (membrane) lipids interact with self-aggregating proteins such as amyloidogenic proteins and even promote aggregation of these proteins in a pathological context. This is of relevance as abnormal aggregation and deposition of proteins is a seminal pathological hallmark in many neurodegenerative diseases, including AD, PD, Huntington’s disease, amyotrophic lateral sclerosis, frontotemporal lobe dementia, and Niemann-Pick disease.

MALDI imaging offers the unique opportunity to characterize the protein pathology associated with lipid biochemistry, which in turn provides targets for mechanistic investigations.

In this context, MALDI imaging has been most prominent for delineating amyloid plaque pathology associated with lipid regulations in AD. Specifically, MALDI imaging identified plaque-specific localizations of gangliosides, ceramide, lysophosphatidic acids, and inositol; sulfatides were depleted at plaques^98–100. Moreover, when interfaced with advanced fluorescent microscopy using structure-sensitive luminescent probes, MALDI MSI revealed amyloid polymorphism-specific lipid changes in transgenic AD mice^{101, 102}. Here, ganglioside GM1 correlated with the core structure of senile plaques, while ceramides and phosphoinositols were located in the diffuse periphery¹⁰².

Further development towards multimodal imaging approaches includes bimodal^{101, 102} and trimodal^{103, 104} lipids and peptide imaging from the same tissue section (Figure 3). The low laser desorption energies facilitate this approach, fluences, and shot numbers necessary to obtain satisfactory lipid spectra using 1,5-DAN matrix dry coating protocols or sublimation^{69, 103}. This approach eventually allows correlating in situ lipid changes to associated protein mechanisms identified with MSI, spectroscopy, and conventional microscopy techniques^{101, 105}. In detail, distinct lipids such as ganglioside GM1 were found to correlate with Aβ1–40 within spectroscopy-assigned core structures. In contrast, other lipids, including ceramides and inositols observed in negative mode and lysophosphocholine (LPC) detected in positive mode, correlate with diffuse parts and Aβ1–42, respectively¹⁰⁶. Amyotrophic lateral sclerosis (ALS) is a fatal neurological disorder for which early diagnosis has remained elusive. A study on mouse models of familial ALS demonstrated a notable alteration in docosahexaenoic acid (DHA) distribution pattern containing PCs within the spinal cord. This shift became apparent during the terminal stage of the disease¹⁰⁷. In a separate investigation employing Drosophila melanogaster as a model, alterations in brain phospholipid concentrations throughout the entire disease progression, commencing from the initial stages, were detected, suggesting the possibility of their utility as early biomarkers for the disease¹⁰⁸. Understanding the role of lipids in ALS has remained obscure. Nonetheless, the studies mentioned above have shed light on the substantial involvement of lipids, highlighting the potential of utilizing MSI to identify potential biomarkers for early diagnosis or therapeutic targets for ALS.

Figure 3: MALDI Imaging of lipid species in central nervous system (CNS) tissues in different neurodegenerative diseases.

(A) Multimodal MALDI-MSI revealed amyloid plaque-associated lipids in dual polarity and peptides on a coronal mice brain tissue section of transgenic Alzheimer’s disease mice (tgArcSwe). Ion images of lipids obtained at 10μm spatial resolution: phosphatidylinositols (PI 38:4, m/z 885.6) in negative (green), lysophosphatidylcholines (LPC 16:0, m/z 496.3,) in positive (red) polarities with subsequent amyloid-β (Aβ 1–40, m/z 4257.6) peptide (blue) ion images in the same imaging region¹²⁸. (B) MALDI-MS ion images of sulfatide species (a) SHexCer (41:2), (b) SHexCer (42:2), and (c) SHexCer (42:3) in control (left) and MPTP-lesioned macaque brain tissue, a Parkinson’s disease animal model. The lateral resolution is 150 μm⁷¹. (C) Representative 3D images of lipid in the CNS of a zebrafish model of Niemann-Pick disease 1. The 3D image was prepared by reconstructing 20 consecutive sections of a sample. The lipids shown here are ceramide (Cer 34:1, Cer 37:1), phosphatidylserine (PS 44:11), and phosphatidylethanolamine (PE 40:5). MALDI MSI was obtained in negative ion mode at a spatial resolution of 50 μm⁷⁶.

Lipids are crucial to regeneration and inflammation, making them possible indicators for traumatic brain injury (TBI). High-resolution mass spectrometry revealed similar levels of palmitoylcarnitine within substantial nigra compared to the injury site. Since TBI is a risk factor for PD, the change in carnitines may suggest a pathophysiological relationship between TBI and PD, characterized by dopaminergic neuronal death⁷⁵.

PD pathology can also be mimicked through intracranial injections of neurotoxins that induce cell death of dopaminergic midbrain neurons. Toxins like 6-hydroxy-dopamine (6-OHDA) and 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), are commonly utilized as a neurotoxin in monkeys and mice to simulate PD pathology^{109, 110} as well as for mimic side effects for dopamine replacement therapy²⁶. Here, MALDI-FTICR-MSI in dual polarity of MPTP-treated macaque monkeys found depletion of long-chain hydroxylated sulfatides in motor-related grey matter regions, suggesting myelin sheath disruption around neuron axons, which could negatively affect neural conduction and cause PD⁷¹.

4.2. Lysosomal Storage Disorders

Another neurodegenerative example where MSI can provide valuable insight is Niemann-Pick disease type C1 (NPC1), an autosomal recessive neurodegenerative disorder affecting lysosomal storage and trafficking. It is characterized by a mutation in the lysosomal transmembrane protein NPC1, responsible for most (95%) cases, which leads to dysregulation in the transport and metabolization of cholesterol and glycosphingolipids in the endosome and lysosome, resulting in cellular death. NPC1 patients typically demonstrate progressive neurodegeneration characterized by pronounced impairment of the cerebellum¹⁶. MSI can aid in diagnosing and monitoring the progression of NPC1 disease by accurately measuring lipid levels and their distribution in different brain regions.

In a mouse model of NPC1 study, elevated distributions of GM1, GM2, and GM3 were observed in the cerebellar region. These lipids play crucial roles in cell membrane structure and signaling pathways in the nervous system. The abnormal accumulation of these lipids in the cerebellum suggests dysregulation in lipid metabolism and potentially contributes to the progressive neurodegeneration observed in NPC1 patients. It is worth noting that the authors also disclosed a region-dependent distribution of GM2 and GM3 associated with disease progression⁷². They also observed lower amounts of GM1, HexCer, and Cer in the same region with the same fatty acid backbone (d36:1). This observation prompted them to hypothesize that the elevated levels of GM2 and GM3 were a result of the degradation of GM1. Additionally, a decrease in sulfatide levels at later time points, indicative of demyelination, was seen. They also identified a novel association between ceramide accumulation and the neurodegeneration of Purkinje cells in the cerebellum. The researchers postulated that this alteration may induce cellular death by activating cathepsin D. A marginal increase in unesterified cholesterol levels was seen in the control samples relative to the mutant mice. This observation can be attributed to the relatively lower presence of myelin in NPC1¹¹¹.

It is important to note that imaging MS methods have limitations in accurately quantifying certain lipid species, such as cholesterol and other sterols, because of poor ionization. A study used a new derivatization method and MSI to make the process more sensitive and allow exact quantitation of sterols on sections of the whole mouse brain. Apart from the cerebellum, they also identified a significant reduction in cholesterol in the hypothalamus, midbrain, and pons in the mutant mouse compared to controls⁷⁸. A comprehensive list of MALDI MSI for steroid and specifically cholesterol lipid analysis has been provided in Table 2.

Table 2.

Strategies for steroid lipid analysis through MALDI MSI.

Study	Cholesterol analysis method	Matrix	Tissue	Instrument
Angelini R. et al.78	Enzyme-assisted derivatization for sterol analysis (EADSA) Conversion of 3β-hydroxy group in cholesterol (m/z 386.3543) to 3 β-oxo using cholesterol oxidase. Charge tagging with Girard-P (GP) hydrazine - cholesterol hydrazone (m/z 518.4105) .	CHCA	Mouse brain	MALDI TOF/TOF (UltraFlextreme, Bruker)
Barre FPY. et al.130	MALDI-2, a positionization strategy, was used to enable a second ionization event to increase the yield of ions. Dehydrated cholesterol (m/z 369.3521 [M + H − H2O]+)	DHB	Human cartilage and dog liver	Orbitrap elite MS coupled to an intermediate pressure MALDI/ESI interface
Dufresne M. et al.131	Silver cationized cholesterol [M+Ag]+ (m/z 493–495))	Matrix free	Mouse brain/kidney	MALDI TOF/TOF (UltraFlextreme, Bruker)
Tobias F. et al.72	Unesterified cholesterol (dehydrated ion, m/z 369.3, [M-H2O+H]+)	DHB	Mouse brain	MALDI TOF/TOF (ScieX)
Patti GJ. et al.132	Silver cationized cholesterol [M+Ag]+ (m/z 493) Nanostructure initiator mass spectrometry (NIMS) combined with Silver nitrate (AgNO3)	Matrix free	Mouse brain	MALDI TOF/TOF (Applied Biosystems)
Cobice DF. et al.133 2023	7-ketocholesterol. Charge tagging with Girard-T (GP) hydrazine precoated on ITO slide.	CHCA	Rat adrenal gland/mouse brain	MALDI-FTICR-MS (Bruker SolariX)
Muller L. et al.134	Silver nanoparticles implantation on tissue [M+Ag3]+. Silver adducts of cholesterol and cholesterol ester [M+Ag]+.	Matrix free	Rat brain	MALDI LTQ-XL-Orbitrap (Thermo Fisher Scientific)
Chaurand P. et al.135	Silver-assisted laser desorption ionization mass spectrometry imaging (AgLDI MSI) for cholesterol quantitation (m/z 493.26) [12CHO+107Ag]+	Matrix free	Mouse brain	MALDI TOF/TOF (UltraFlextreme, Bruker)

Here, a pioneering study used three-dimensional MALDI MSI on the whole body of a zebrafish NPC1 model to map lipid dysregulation across different organs. Their 3D MSI displayed elevated levels of ceramide in the brain and increased levels of sphingomyelin species in the spinal cord, where ceramide is both a precursor and degradation product of sphingomyelin, highlighting Cer-SM pathway dysregulation through the CNS⁷⁶ (Figure 3C).

4.3. MALDI MSI of neuronal lipids in brain tumors

Given the pronounced tissue heterogeneity, MSI is a powerful technology to interrogate tumor formation and growth mechanisms and cellular cancer biology in general. The widespread presence and vital role of lipids in the CNS have spurred investigations into their involvement in the analysis of brain cancer. The heightened aerobic glycolysis of tumor cells alters lipid metabolism, rendering tumor cells and their lipid patterns a potential diagnostic and theragnostic in situ target. Consequently, MALDI MSI has been employed in various studies to evaluate the relevance of lipids in distinguishing between different types of brain tumors. In a study graft model of human glioblastoma (GBM) and medulloblastoma (MB) in rat, diminished levels of complex gangliosides, GM1 and GD1, were observed in glioma. Conversely, GM3, a less complex ganglioside, exhibited exclusive upregulation in rat glioma. In turn, GM2 was upregulated in medulloblastoma, facilitating differentiation between these two forms of brain cancer¹¹². Another investigation utilized high-spatial resolution MALDI MSI to scrutinize the metabolic profiles of medulloblastoma and pineoblastoma (PB) tumors. The study identified upregulated glycerophosphoglycerols (PG) and glycerophosphocholines (PC) in MB, while sphingolipids such as the cerebroside hexosylceramide (HexCer (36:2)) as well as ceramide-1-phosphate (CerP (47:2)) were specifically upregulated in PB¹¹³. A study employing 3D MALDI TOF MSI on pediatric brain tumors found that metastasizing tumors exhibited distinct lipid profiles compared to non-metastasizing tumors. A prevalent characteristic in metastasizing tumors was the reduction of lipids with an 18:1 fatty acyl chain configuration. Among the lipids elevated in metastasizing primary tumors were phosphatidylinositol (PI) and the two phosphoinositide phosphates PIP and PIP2. This indicates dysregulation of the PI3K signaling pathway, which is often attributed to alterations in the phosphatase and tensin homolog (PTEN) gene, and has in turn been linked to oncogenesis in humans due to uncontrolled cell growth and proliferation¹¹⁴.

MALDI MSI also contributes to a deeper understanding of the role of lipids in tumor initiation and progression. Here, a study in human glioblastoma multiform (GBM) tissues observed reduced levels of ω−6 fatty acids, including arachidonic acid and adrenic acid, crucial energy sources, in glioma cells compared to peritumoral tissue¹¹⁵. Another study in an GBM animal model, MALDI FTICR-based MSI revealed elevated levels of small-chain fatty acids, such as palmitic acid, linoleic acid, and PI, aligning with the malignant nature of glioma¹¹⁶.

Beyond fatty acids, critical membrane lipids have been identified as dysregulated in mouse models of glioma, particularly phospholipids and sphingolipids. Here, a decrease in common PCs, such as PC 16:0/16:0 and PC 16:0/18:0, was observed in tumor parenchyma and a heterogeneous increase of PC 16:0/16:1, PC 16:0/18:1, and PC 16:0/18:2 in the tumor tissue itself. The surge in de-saturated fatty acids in tumor phospholipids has been suggested to potentially modulate tumorigenic signaling pathways and enhance tumor cell survival. Additionally, MALDI MSI results disclosed a significant reduction in tumoral sphingomyelin (SM d18:1/18:0)¹¹⁷.

Efforts have also been made to integrate the MALDI MSI approach with MRI, a standard in vivo imaging tool for cancer diagnostics. A study combining in vivo MRI data with high-spectral resolution MALDI MSI from a mouse brain model of GBM revealed an upregulated, co-localized distribution of PCs and cholesterol¹¹⁸. Another technological advancement of MALDI MSI involves the integration of stable isotope tracers with MSI to investigate the flux of the metabolic landscape in brains with glioblastoma. In detail, the authors introduced an approach known as Spatial Isotopologue Spectral Analysis (SISA) to assess fatty acid synthesis and elongation flux. SISA unveiled an eight-fold higher difference in palmitate elongation flux in tumor tissue compared to healthy brain tissue, possibly due to the elevated demand for eighteen carbon acyl chains due to their prevalence in the membrane bilayer. This study emphasizes the potential of lipid synthesis as a promising therapeutic target in GBM¹¹⁹.

5. Concluding Remarks

The developments and applications of MS imaging presented here highlight the immense potential of the technique for studying spatial lipid dynamics in complex biological tissues. Recent developments are increasingly overcoming significant challenges associated with sample preparation, sample throughput, and lipid identification. Moreover, standardized protocols^{120, 121} and data formats¹²² along with round robin studies¹²³ and common reporting standards¹²⁴ significantly improve data reproducibility and result accuracy. This is essential to expedite the use of MSI-based applications in clinical settings^{125, 126}.

In summary, MALDI lipid imaging is increasingly becoming a valued technology for molecular histology significantly exceeding standard tools used to study tissue pathology. It has shown great potential for refining molecular disease pathology including neurodegenerative-, neuroinflammatory- and cancer pathologies.

Figure 2: Sample preparation optimization for ganglioside analysis through MALDI-TOF MSI of mouse brain tissue sections.

(A) Optimization by washing tissue sections with ammonium formate, deposition, matrix sublimation, and waiting before analysis increased ganglioside species signal intensity. Different species and fatty acyl chain lengths of gangliosides can be identified, expanding the range of complex lipids that can be studied. The scale bar is 2 mm, and the lateral resolution is 65 μm. ND: not detected. The figure has been adapted from¹²⁷. (B) Application of ozone-induced dissociation (OzID) in resolving double-bond (db) positional isomers. MALDI-CID/OzID fractional images of PC(16:0/18:1), PC(18:0/18:1) and PC(16:0/20:1) in an animal model of medulloblastoma. The images highlight the importance of resolving structural isomers since their distribution is specific, and heterogeneous distributions can be present even for lipids with the same fatty acyl chain, revealing pertinent information regarding disease pathogenesis⁹⁷.