Mass spectrometry imaging to detect lipid biomarkers and disease signatures in cancer

Abstract

Background

Current methods to identify, classify, and predict tumor behavior mostly rely on histology, immunohistochemistry, and molecular determinants. However, better predictive markers are required for tumor diagnosis and evaluation. Due, in part, to recent technological advancements, metabolomics and lipid biomarkers have become a promising area in cancer research. Therefore, there is a necessity for novel and complementary techniques to identify and visualize these molecular markers within tumors and surrounding tissue.

Recent Findings

Since its introduction, mass spectrometry imaging (MSI) has proven to be a powerful tool for mapping analytes in biological tissues. By adding the label‐free specificity of mass spectrometry to the detailed spatial information of traditional histology, hundreds of lipids can be imaged simultaneously within a tumor. MSI provides highly detailed lipid maps for comparing intra‐tumor, tumor margin, and healthy regions to identify biomarkers, patterns of disease, and potential therapeutic targets. In this manuscript, recent advancement in sample preparation and MSI technologies are discussed with special emphasis on cancer lipid research to identify tumor biomarkers.

Conclusion

MSI offers a unique approach for biomolecular characterization of tumor tissues and provides valuable complementary information to histology for lipid biomarker discovery and tumor classification in clinical and research cancer applications.

1. INTRODUCTION

Tumors are highly heterogeneous in composition, including cellular content, genetic, epigenetic, protein, and differentiation_. 1 Despite a large amount of OMICs information available using high‐content genetic, RNA, protein, and functional profiling, the role of lipids and lipid metabolism in oncogenesis has only recently become a focus.2

Lipids are among classes of biomolecules with a high potential value in cancer research and eventually for treatment strategies.3 Lipids have their most widely recognized role in structural biology as key building blocks of the lipid bilayer of cellular and organelle membranes.4 Furthermore, lipids are signaling and metabolic molecules that produce profound effects on immune activation and disease development.5 For mammal cells, the major relevant lipid classes encompass phospholipids (PL), including phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylethanolamine (PE), phosphatidylinositol (PI), phosphatidylglycerol (PG), and phosphatidic acid (PA), along with sphingomyelin (SM), cholesterol, free fatty acids (FA), and di‐ and triglycerides (DAG/TAG).6 It has already been reported that alterations in the lipid composition of tissue are seen in certain diseases, including cancer.7, 8 Differential abundance of certain phospholipids and their enzymatic byproducts was found to be a characteristic for malignant transformations9, 10 and several physiological processes.11 In general, abnormal lipid metabolism is known to be a common feature for cancer cells12, 13, 14 and can already be observed at early stages of the tumor development.15

Due to the lack of methods to examine lipid content, trafficking, and distrbution, several techniques, including mass spectrometry (MS), have become essential for a comprehensive examination of the distribution, functions, and consequences of lipid variation in the context of disease. Mass spectrometry imaging (MSI) is, therefore, a valuable tool for the lipidomic analysis of cancer tissue as it adds the specificity of MS to detailed spatial information. In MSI, not only are intensities of mass to charge (m/z) values recorded but also the respective positions within the sample allowing for the generation of images representing ion intensities at specific tissue localizations. This information is of particular interest when analytes such as lipids are not homogeneously distributed but compartmentalized in tissue. MSI has several advantages over traditional imaging and analytical methods used in clinical diagnostics; most important among these is the ability to visualize thousands of analytes simultaneously and the label‐free nature of the method. Here, we will describe the current and future status of lipid‐based MSI in clinical and nonclinical cancer research.

2. LIMITATIONS OF CURRENT IMAGING AND BIOMARKER DISCOVERY TOOLS

The use of histology, molecular, and imaging data is currently the gold standard for the clinical diagnosis of cancers and identification of prognostic and therapeutic targets.16 Tissue evaluation of stained or labeled slides is traditionally performed manually by highly trained histopathologists. In recent years, computer‐assisted diagnosis (CAD) systems have been implemented to aid histopathologists and clinicians in cancer diagnosis and research, which significantly reduce labor requirements and subjectivity in analysis.17 However, H&E stained images only provide tissue morphology information and may not resolve healthy cells from early‐stage cancerous cells when changes may be purely biochemical in nature.18 Hence, there is a requirement for more specific biochemical markers that can be used to accurately delineate tumor boundaries preventing the potential for tumor re‐emergence and minimizing complications caused by excessive removal of healthy tissue.19

According to the current NIH definition, biomarkers are “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention.”20 Biomarkers may be molecular, histologic, radiographic, or physiologic.21 In clinical laboratories, imaging biomarkers are utilized for disease diagnosis and grading and evaluating therapeutic effect and/or toxicity.22 In research settings, biomarkers are used in pathogenesis and drug discovery and studies to investigate pathophysiological effect and efficacy. To date, they have only gained limited acceptance in clinical decision making,23 and they are typically employed in a complementary manner to histological evaluation. The current state of the art in ex vivo non‐MS–based imaging cancer biomarker research has been discussed in recent excellent reviews.24, 25

Traditional imaging methodologies such as immunohistochemical (IHC) labeling or staining combined with fluorescence microscopy allow for the visualization of tumor biomarkers and tissue structures with high specificity and at high spatial resolution. However, these methods are targeted, requiring prior knowledge about the analyte to be studied, and are limited by the small number of molecules that can be visualized simultaneously.26 Untargeted discovery techniques such as liquid chromatography‐mass spectrometry (LC‐MS) require extraction of the analytes of interest from tissue homogenates and thus lose important information concerning their spatial location within the tissue. In recent years, the metabolome, including small molecules and lipids, has evolved as a source of key potential biomarkers of disease, becoming an area that urgently requires the introduction of new methodologies for their analysis. Due to the high specificity, sensitivity, and increasing spatial capabilities offered, mass spectrometry imaging has rapidly risen to become a premier technology for lipid imaging in biological tissues and cells.27

3. MASS SPECTROMETRY IMAGING

MSI is a powerful technique for visualizing the spatial distribution of biomolecules in situ. A typical MSI experiment is performed by rastering an ionization beam across a tissue surface at defined x,y coordinates, thereby desorbing/ionizing analytes that are subsequently extracted into the mass spectrometer for analysis. Software is used to reconstruct an image from the dataset in which each pixel consists of a mass spectrum. Images can be plotted for any selected peak in the acquired mass spectrum with the relative abundances displayed as a false‐color image across the analyzed region of interest (ROI).28, 29 A simplified outline of the MSI process is shown in Figure 1.

Figure 1.

Overview of MS imaging workflow in cancer research. Biopsy specimens are collected, snap frozen, and sectioned onto compatible slides using a cryostat. MS images are acquired by rastering the ionizing beam across the tissue surface. Ion distribution maps of lipids are reconstructed using software. Multivariate statistical analysis is performed to identify candidate biomarker lipids or lipid signatures of tumors. Histological staining is performed on the same or adjacent tissue section. MS imaging dataset is coregistered with optical image and candidate biomarkers are correlated with histologically defined regions or cell populations. Potential biomarkers have multiple applications in cancer research

The biggest strength of MSI over traditional mass spectrometry approaches is the possibility of combining spatially resolved MS data with microscopic data of the same or serial tissue section, enabling the interpretation of MSI images within their histological context.28

Imaging mass spectrometers are composed of three major components: (a) an ionization source that generates charged gaseous ions from the tissue, (b) a mass analyzer that separates the different types of ions according to their mass‐to‐charge ratio, and (c) an ion detector that reads these ions and records their abundance. The main primary ionization techniques used are described in more detail in the following section. In the mass analyzer, the ions are sorted and resolved based upon their mass to charge ratio before being transmitted to the detector. Multiple types of analyzer are used in MSI setups, and the advantages and disadvantages of each in biomedical research has previously been discussed in detail in excellent reviews by Rubakhin et al30 and Haag.31 A detailed description of MSI detectors is beyond the scope of this review and has been discussed comprehensively by Jungmann and Heeren.32

4. IONIZING MODALITIES

The three most commonly used MSI ion sources are MALDI MS (matrix‐assisted laser desorption/ionization mass spectrometry), DESI‐MS (desorption electrospray ionization mass spectrometry), and SIMS (secondary ion mass spectrometry). A graphical overview of these ionization mechanisms is shown in Figure 2.

Figure 2.

MSI ionization sources predominantly used in lipid‐based tumor imaging. In MALDI, a UV‐absorbing matrix is applied to the tissue surface. Irradiation of the matrix‐coated sample by a UV or IR laser results in the generation of an ion plume (A). In DESI stream of charged droplets are directed at the tissue surface causing desorption and ionization of molecules similar to electrospray ionization (ESI) (B). In SIMS, the tissue surface is bombarded with a focused primary ion beam resulting in sputtering of positive and negative ions and neutral species (C). After ionization, the resulting ions are accelerated into the mass spectrometer

4.1. Matrix‐assisted laser desorption/ionization

MALDI is based on the application of a matrix to the surface of the sample, facilitating co‐crystallization of matrix and analytes. The MALDI matrix must be capable of strongly absorbing the photon energy emitted by the laser pulse to promote efficient desorption and subsequent ionization of analytes.33 Detailed reviews of the mechanisms of MALDI can be found elsewhere.34, 35, 36, 37

Among the several mass spectrometric ionization techniques that can be used to directly analyze tissues, MALDI has led the way in the development of biological and clinical applications for MSI and is, therefore, one of the most commonly used techniques for cancer research.38, 39 By using the appropriate preparation for biological samples, distributional information of a wide range of molecules, eg, lipids,40, 41 metabolites,42, 43 glycans,44, 45 proteins,42, 46 and drugs,47, 48 can be recorded in situ.

Matrix application is certainly one of the main challenges in MALDI MSI as it directly affects the attainable spatial resolution and analytical sensitivity. Consequently, most of the focus for sample preparation has been on the matrix application process. In particular, matrix crystal size and reproducibility of application49, 50, 51 may limit the ability to reach subcellular resolution, ie, imaging at a spatial resolution below 10 μm. The matrix deposition technique has to be carefully considered with respect to the investigated molecules and the spatial resolution desired.52 Matrix preparation must produce a uniform coating of crystals with sufficient amount and density to permit sampling of the surface without having regions that are absent of signal. Most importantly, the localization of the analytes of interest throughout the tissue must be conserved.

MALDI MSI has the advantage of being highly sensitive but can suffer from an abundance of spectral peaks derived from the matrix itself, which may occur in the same mass range as the analyzed lipid.53 Hence, spectra become even more complex. To circumvent the issue, high‐resolution mass analyzers (such as Orbitrap or FT‐ICR) and/or tandem MS (MS/MS) can be applied in MALDI imaging experiments to provide conclusive identification of the lipids in tissue.42, 54, 55 Additionally, MALDI has limitations inherent in the LDI process and the performance of instrumentation. One is the ion suppression effect induced by highly abundant endogenous compounds or salts that hampers direct quantitation from a tissue section, demanding sophisticated quantitation strategies.56, 57 A further limitation is the laser spot size and the trade‐off between image resolution and sensitivity. The smaller the spot size, the less material is desorbed and ionized, which significantly decreases signal intensity.58

4.2. Desorption electrospray ionization

DESI is a complementary ambient ionization technique with a simplified sample preparation workflow that doesn't require the application of matrix to the tissue.59 This ionization method uses electrosprayed charged solvent droplets, which are directed at the tissue surface. The impact of the droplet stream results in the desorption of ions from the surface by electrostatic and pneumatic sources.60 Analytes are ionized by mechanisms similar to electrospray ionization (ESI). The resulting gas‐phase ions are then transferred into the mass spectrometer by use of an ion transfer line.61 The chemistry of the desorption process is highly dependent upon the solvent of choice, and it is very effective for the desorption and ionization of small molecules such as lipids and metabolites.62, 63 Unlike MALDI, DESI requires no advanced sample preparation and occurs under ambient conditions, allowing rapid assessment of molecular information directly from tissue samples.

4.3. Secondary ion mass spectrometry

SIMS ionization utilizes a highly focused primary ion beam directed onto the tissue surface to produce sputtered charged ions. Like DESI, it does not require the addition of an ionization‐enhancing matrix to the tissue surface. However, unlike MALDI or DESI, SIMS imaging of biological tissues is required to be performed under high vacuum conditions.64 SIMS offers the highest spatial resolving capabilities of all the ionization techniques at sub‐100 nm² pixel size.64 Additionally, 3‐D images of cells and tissues can be acquired by profiling at progressive depths through the tissue section.65, 66 As the physical sputtering process is a harsh ionization, excessive in‐source fragmentation of biomolecules can occur resulting in lower ion yield than acquired by alternative ionization methods.67 However, coating of the surface with metal or UV‐absorbing matrix has been shown to enhance ion yield, particularly for larger molecules.68, 69

5. RECENT ADVANCES IN SAMPLE PREPARATION FOR MSI

For a successful MSI experiment, the structural integrity and morphology of the tissue must be maintained from the time of sample collection and prehandling until analysis without delocalization and degradation of the analytes of interest.51 This is of high importance for lipids, which may rapidly undergo oxidation or hydrolysis.70 Thus, the chemical distribution image must resemble the original biological state in situ as accurately as possible. Several methods have been developed to halt metabolite degradation prior to and following tissue collection using high heat or rapid cooling.71, 72, 73 Of these, funnel freezing has been demonstrated to be the most successful for preserving metabolites for molecular imaging.72 The wide variety of sample preparation procedures and potential issues encountered have been discussed in a number of excellent overview articles.51, 74, 75 In brief, tissues from biopsy or autopsy are collected and flash‐frozen, and thin tissue sections are prepared using a cryostat following protocols adapted from traditional microscopy and histology workflows. It is at this stage where sample preparation for the different ionization technologies diverges as MALDI requires the application of a ultraviolet (UV)‐ or infrared (IR)‐absorbing matrix, which can be applied by numerous techniques, including manual spray,76 automated spray,77, 78 dried droplets,79 or sublimation.80, 81, 82

Tissue collection and processing workflows typically used for cancer histology and immunohistochemistry are frequently incompatible with MSI analysis. Formalin fixation without paraffin embedding has been demonstrated to be compatible with downstream lipid MSI,83, 84 but due to their inherent solubility, many lipids will be washed from the tissue during alcohol fixation. Tissue microarrays (TMAs) are rapidly gaining acceptance in clinical cancer workflows due to the ability to assemble multiple small biopsy tissues onto a single histological slide.85 Due to the TMAs being composed of formalin fixed and paraffin embedded tisues, most MS imaging research has been in the compatible field of proteomics.86, 87 Nethertheless, recent publications have shown that some metabolite and lipid species are conserved through the FFPE process and can be visualized by MSI. This potentially opens up a new field for lipid biomarker imaging in TMAs and the possibility of interrogating historical archived FFPE biopsies.88, 89

Matrix selection for lipid analysis has been comprehensively discussed in a recent review by Leopold et al.53 Matrix selection is important as lower molecular weight lipids that have been identified as potential cancer biomarkers, including ceramides,90 acylcarnitines,91 and sterols,92 may potentially be masked by matrix‐related peaks occurring within the same mass range of the spectrum when using instruments with low mass resolving power. The application of matrices suitable for dual polarity imaging, such as norharmane and diaminonaphthalene, has vast potential to increase throughput in cancer lipid biomarker studies as they enable both positive and negative ionizing lipids to be analyzed during the same instrument acquisition.40, 93, 94

The method of matrix application is crucial in enhancing the signal intensity and achieving high spatial resolution. Particularly, matrix crystal size and application consistency49, 50, 51 prove difficult to reach a subcellular resolution, ie, imaging at a spatial resolution below 10 μm. Thus, matrix application strategies providing highly reproducible matrix depositions along with crystals being significantly smaller in diameter than both the laser spot and the single‐cell are mandatory for MALDI MSI below the cellular level. Initially, manual application of the matrix in solution was performed by airspray deposition. To address limitations in throughput and reproducibility, commercial and lab‐built automated spray and droplet‐based instrumentation has been introduced.66, 67 Matrix sublimation, a dry matrix application, first reported by Kim et al,95 traditionally required complicated lab‐built systems. However, it has recently been automated and commercialized, increasing throughput and making it more user‐friendly for clinical cancer applications.96, 97 In this approach, the solid matrix undergoes a direct phase transition to a gas and is then deposited onto the tissue as a homogenous layer. Sublimation generally yields dry matrix crystals at the submicrometer to 3 μm level, allowing an increased spatial resolution down98 to 1.4 μm as well as limited lateral migration of molecules in tissue.52, 99, 100 The narrow size distribution of crystals generated by sublimation enhances reproducibility and S/N ratios in MALDI MS.101, 102 However, due to the lack of solvent assisted extraction, overall ion counts have been shown to be lower than comparable solvent spray methods.98 The ideal matrix application method would provide appropriate analyte extraction without any lateral diffusion and homogenous coverage, as well as produce small crystal sizes. Unfortunately, to date, no universal method exists, and in each experiment matrix deposition must be carefully optimized for the target analyte.

On‐tissue derivatization is a process by which the target analyte (or analyte class) in a tissue section is chemically altered to produce a new compound with more favorable properties for MSI, such as resolving isobaric species, increasing ionization affinity and/or chemical stability during the sample prep and ionization steps.103 Wang et al used this approach to derivatize free fatty acids (FAs) in thyroid cancer tissues enabling FAs to be visualized in the same MSI experiment as phospholipids.104 This enabled different correlations between saturated and unsaturated FFAs and PLs to be detected in specific tumor regions. Kuo et al105 developed an epoxy derivatization approach (called MELDI) combined with DESI‐MSI to identify tumor‐associated change in C═C isomers in a metastatic mouse lung tissue section.

Solvent washing steps typically applied in the preparation of sections for peptide/protein MSI are not appropriate for lipid imaging due to the solubility of lipid markers in the solvent wash phase leading to substantial delocalization. Recently, lipid and metabolite friendly tissue washing protocols using ammonum formate,106 ammonium acetate,107 or saline solutions 108 have been demonstrated to enhance lipid signals in tissue by removing endogenous salts and other suppressants.

While recent developments in sample preparation have significantly improved tissue section quality, analytical reproducibility, and spatial resolution, work still remains before MSI can be integrated routinely into cancer biomarker discovery workflows.

6. RECENT ADVANCES IN MASS SPECTROMETRY TECHNOLOGY

Recent technological advancements in MSI have been focused in several key areas: improvement of analytical sensitivity and specificity; improvement of spatial resolving capabilities; development of high sample throughput; and data handling and statistical analyses. These are discussed in greater detail below.

6.1. Improvement of analytical sensitivity and specificity

Advancements in mass spectrometer sensitivity are essential to keep pace with improvements in mass spectrometry imaging spatial resolving capabilities and the substantially reduced number of ions produced when sampling from much smaller regions. One of the most exciting recent developments within MSI is the incorporation of laser post‐ionization following initial ionization within the mass spectrometer source.109, 110, 111 This is most commonly used for MALDI MSI and is termed MALDI‐2. Using this technique, ion yields for poorly ionizing lipids can be vastly increased, potentially leading to the discovery of cancer biomarkers from lipid classes not routinely detected by MALDI‐MSI. Most recently, laser post‐ionization has been applied to enhance secondary ion production in SIMS by ionizing the neutral species that are cosputtered with secondary ions from the sample surface.112 While the technique has not yet been demonstrated for the analysis of biomolecules in tissues, it has the potential to increase analytical sensitivity, one of the major limitations of SIMS.

A possible issue in the expansion of MSI to research laboratories is the requirement to purchase entire new imaging mass spectrometers. The introduction of a relatively inexpensive dual ion source enabling both MALDI MSI and ESI ionization using Orbitrap mass spectrometers enables MALDI capabilities to be added to a lab containing an existing Orbitrap MS without the requirement to purchase an entire new system or necessitating physical swapping of the source and extended instrument downtime.113 This source has already been utilized to identify lipids associated with medulloblastoma metastasis in a humanized mouse model.40

Historically, the application of time‐of‐flight SIMS (ToF‐SIMS) to lipid imaging has proved challenging due to lower mass‐resolving capabilities than Orbitrap and Fourier transform ion cyclotron resonance (FTICR) analyzer MALDI and DESI systems and the lack of MS/MS capabilities for absolute lipid identification of isobaric species.67 However, recent technological advancements are being used to address these limitations. Independently, Passarelli et al and Smith et al have developed SIMS instruments with high mass resoving capabilities by using Orbitrap or FTICR mass analyzers, respectively, which are highly suited to subcellular lipid imaging.114, 115 In 2016, Fisher et al introduced a novel ToF‐SIMS instrument capable of MS/MS imaging analysis of lipids at submicron spatial resolution.116 The technology was used to identify and visualize fatty acid117 and oxidative tocopherol118 metabolite biomarkers of diseased granulomatous tissue.

6.2. Spatial resolution

The image resolution (ie, the spatial resolution) of an MSI experiment decisively influences the ability to distinguish distributions of components within tissue sections. Herein, the recorded pixel size is key for whether two adjacent regions are perceived as unique and distinct.

Achievable spatial resolution in MSI has its main challenges in sample preparation and instrumental parameters. The diameter of the ionization beam is predominantly the limiting factor, as it determines the ablation area and thus the attainable spatial resolution in an imaging experiment. Several instrumental improvements and other notable advancements in spatial resolution have been reported recently.98, 119, 120 For instance, by adjusting the optics of the ion source along with changes in the geometry on an atmospheric‐pressure MALDI, a spatial resolution of 1.4 μm was reported, allowing for the visualization of lipids on a subcellular level.98 Most recently, MALDI MSI of lipids has been performed at 600 nm pixel size using transmission mode MALDI‐2,111 enabling detailed subcellular imaging of glycolipids that have previously been indicated as potential biomarkers of cancer cells.121

Oversampling80, 122 (decoupling spatial resolution from the ion beam spot size by acquiring spectra from tissue areas smaller than the beam) has recently been demonstrated to produce 5 μm² pixel images using a laser with a spot diameter of 20 μm.123 Yet improving spatial resolution also has its drawbacks. It decreases the area of ablation and consequently decreases signal intensity while increasing data acquisition time and, ultimately, the size of the data file generated.124 Hence, the sensitivity enhancing developments discussed in the previous section gain increased importance.

One limitation of traditional DESI imaging has been the relatively poor spatial resolution offered (100‐200 μm² pixels). However, the recent introduction of nano‐DESI sources has enabled images to be acquired at resolutions approaching single‐cell level (10 μm²).125, 126

Of all the ionization techniques used for MSI, secondary ion mass spectrometry (SIMS) offers the highest spatial resolving capabilities, and its ability to visualize lipids at subcellular resolution makes it a powerful tool for imaging single cancer cells,127 tumor cell spheroids,128 and tissue biopsies.129 However, sensitivity is lower than alternative imaging modalities. Gas cluster primary ion beams (GCIBs) dramatically increase sensitivity in SIMS MSI and have been used to visualize changes in lipid metabolism in human breast cancer tissue.129 As GCIBs utilize clusters containing thousands of particles, the spatial resolution suffers (1‐5 μm) in comparison to traditional SIMS ion beams. Therefore, improving the lateral resolution to sub‐micron levels is an area of highly active research.130

6.3. Sample throughput

Another frequently discussed challenge for MSI is the long analyses times, which may limit the practicality for routine applications. A clinical histopathology laboratory requires high througput for tumor diagnosis and classification, and cases may need to be revisited multiple times.131 Moreover, drug discovery laboratories produce many tissues from a single biomarker animal study.132 Therefore, it is imperative that MSI technologies must evolve to stay current with clinical and medical research needs. The acquisition time and the resulting data load mainly depend on the spatial resolution (ie, point‐to‐point distance), the mass resolution (in case of high‐resolution mass analyzers), and the size of the region of interest (ROI). Advances in laser optics have enabled scan speeds of up to 50 pixels per second,133, 134 compared with three to seven pixels routinely achieved in earlier commercial instruments. Similar to MALDI, significant improvements in the speed of analysis of DESI have been made. While initial DESI imaging studies typically acquired one pixel per second, more recently, scan speeds of up to 30 pixels per second have been achieved.135 The increases in scan speed enable the acquisition of multiple serial sections of tissues, creating 3‐D datasets showing lipid distributions throughout entire organs.136

Throughput becomes even more of a challenge when molecules in the same tissue ionize differently, thus require different polarities for acquisition. This is particularly relevant for lipid analysis as lipids are a diverse analyte class with high structural variability, and, thus, being preferentially detected as either positive or negative ions depending on the headgroup of the different classes. For instance, phosphatidylcholine and sphingomyelin contain positively charged headgroups, whereas phosphatidylethanolamine and ‐inositol are present in permanently neutral or negative charge states. As a result, for comprehensive lipid analyses, MS studies must be performed in both negative and positive ion mode to gather a maximum of information on phospholipid distributions within a single tissue section. Classically, the tissue section is first scanned in one of the ion modes, followed by offsetting the x‐ and y‐raster positions and scanning the tissue in the opposite polarity mode. However, using this strategy demands twice the time for data acquisition. Recently, methods have been employed for lipid imaging using both modes of polarity simultaneously while minimizing analysis time using high‐speed MALDI MSI technology and precise laser control.137 Polarity switching utilizing an infra‐red matrix‐assisted laser desorption ionization source (IR‐MALDESI) identified cholesterol, phosphatidylinositol, and phosphatidylserine lipids as markers of ovarian cancer in a hen model of disease.138

Data handling and storage are another key concern in high‐throughput MSI, particularly when dealing with high spatial and mass resolution data that can reach several hundred gigabytes per tissue section.139 Physical transfer of large imaging datasets between multiple study sites can prove highly cumbersome and time‐consuming. To address this, cloud‐based storage is being developed to enable rapid sharing of datasets between study sites, allowing remote collaborations between mass spectrometrists, histologists, and clinicians in real time.140

6.4. Software and statistical analysis

Basic processing and visualization of mass spectrometry imaging data sets can be performed by a range of commercial and open source software.141, 142, 143, 144, 145 With advances in instrumentation, the complexity of datasets generated in MSI has increased tremendously, demanding advanced statistical methods for reliable interpretation and prospective biomarker discovery. Unsupervised and supervised algorithms are widely employed for data clustering and classification, respectively.146

Unsupervised algorithms do not require any previous knowledge, thus are untargeted approaches and very useful for exploratory data analysis such as in biomarker discovery. In MSI, principal component analysis (PCA) can be applied to reduce the dimensionality of the data set to the most significant m/z distributions.147, 148 Hierarchical cluster analysis (HCA), as another classical unsupervised tool, is used to bin together spatially resolved spectra based on their spectral similarity, allowing for identification of colocalized signals in imaging datasets.

Unsupervised analysis in MSI, such as partial least squares (PLS), has been successfully applied to recognize and distinguish morphologically similar areas in tumor tissue.82, 149 For instance, subareas of gastric cancer in patient tissue sections could be depicted in more detail when compared with histological staining.150 Unsupervised MSI data analysis could also reveal certain subpopulations in sarcomas151 and 3‐D colorectal adenocarcinoma biopsies,152 otherwise being difficult to distinguish with staining methods. The Fournier group applied nontargeted cluster analysis to a MALDI imaging dataset in order to classify anaplastic glioma tissue.46

Supervised algorithms are highly effective in predicting patterns and features of an imaging dataset. Contrary to unsupervised methods, supervised analysis requires a specified set of spectra to be known in advance and thus relies on targeted classification. Machine learning algorithms such as support vector machines (SVM) and random decision forest (RDF) use the predefined set of spectra to determine correlating structures in the data. Supervised algorithms have been successfully utilized to discriminate a variety of cancer types including thyroid cancer, breast cancer, colon cancer, and liver cancer.153, 154 In addition, RDF‐based calculations could identify numerous lipid distributions correlating to the platinum distribution pattern in a malignant pleural mesothelioma (MPM) sample derived from an individual being treated with cisplatin as a cytostatic drug.155 A study using DESI lipid imaging‐derived classifier rapidly classified gliomas into different subtypes and grades.156

Recent software developments and novel workflows are enhancing precise coregistration and correlation of histological and MSI data sets, which is essential for accurate interpretation of lipid biomarker distributions.157, 158, 159 Additionally, methods developed for coregistration of multiple MSI datasets from the same tissue but acquired through complementary imaging modalities may provide substantially more information than offered by a single imaging modality.158, 160, 161, 162

MSI data can be acquired using a variety of instruments, each providing customized tools for acquisition built on proprietary data formats controlled by the manufacturers. However, in 2011, the data format imzML was introduced in order to allow for flexible exchange of imaging data between different instrument vendors and data analysis software.163

Scils is currently the most frequently used commercial processing software for MSI datasets and supports data produced from all mass spectrometer manufacturers following conversion to imzML format.164 The software enables MS and histological image coregisration, multivariate analyses, and spatial segmentation for automated annotation of chemically different tissue regions such as differentiating tumor from healthy tissue.165, 166 Additionally, the software enables construction, visualization, and analysis of 3‐D MALDI imaging data sets as demonstrated to identify lipid markers of medulloblastoma metastasis.40 Further processing software options are emerging to address biomarker discovery needs in cancer research. For example, Veselkov et al recently developed a computational platform for scalable processing of MSI data using machine learning and related pattern recognition for high‐throughput cancer phenotyping studies (BASIS).167 While recent developments are highly encouraging, careful validation and additional advancements in software are required before MSI can be fully integrated into clinical cancer biomarker workflows.

7. ON‐TISSUE LIPID IDENTIFICATION

Accurate identification of lipids in tumor tissue is essential for conclusive determination of tumor biomarkers. This can be challenging by direct MSI due to the lack of chromatographic separation as offered by traditional LC‐MS approaches. In a typical untargeted imaging experiment, putative identifications are assigned based upon m/z values from the initial full scan MSI analysis, which are then confirmed by targeted tandem MS (MS/MS) measurements from the same or adjacent tissue section. The use of high‐resolution mass analyzers such as Orbitrap or Fourier‐transform ion cyclotron resonance (FTICR) systems aids accurate identification of lipid classes and provides basic information regarding the total number of carbons and fatty acids that comprise the fatty acyl chains. However, MS/MS is required to determine the exact fatty acid composition of the lipid. The fatty acid composition of lipids in tumors has been shown to differ substantially from surrounding healthy tissue,168 and fatty acid metabolism has been shown to correlate with malignant phenotypes including metastasis, therapeutic resistance and relapse.169 Therefore, accurate identification of fatty acid composition is an important requirement for MSI studies. Ellis et al have recently introduced a MALDI MSI method enabling automated data‐dependent MS/MS scans to be acquired in parallel during the MSI experiment. Using the technique, they have been able to assign fatty acid compositions to more than 100 lipids identified during the MS image acquisition without increasing the analysis time.170 Numerous databases may be used to assist with lipid identification from both full and MS/MS scan mode data. Examples include LIPID MAPS Structure Database (The LIPID MAPS Lipidomics Gateway, http://www.lipidmaps.org/) and Metabolomics Workbench Metabolite Database (http://www.metabolomicsworkbench.org/), both University of California (San Diego, USA), as well as The Human Metabolome database,171 and METLIN (Scripps Research Institute, LA Jolla, USA).

8. QUANTITATIVE LIPID BIOMARKER IMAGING

When considering the analyte of interest in its biological environment, ion suppression can be understood as a competition for ionization based on secondary reactions.57, 172, 173, 174 The relative concentrations of analyte to MALDI matrix as well as analytes to each other can adversely affect the resulting mass spectra. Hence, if ionization of a certain endogenous compound is thermodynamically favored and/or this substance is highly abundant in tissue, other ions may be masked, or ionization may be depleted to insignificance. Lipids, such as phosphatidylcholine, phosphatidylethanolamine, phosphatidylinositol, and sphingomyelin provide favorable qualities for MS analysis due to their already charged headgroups, which may lead to unfavorable suppression of other lipid class signals if analyzed in parallel.175 This issue can be partially circumvented using phospholipase treatment before matrix application as has been demonstrated to improve analytical sensitivity for cardiolipins.176

Normalization is critically important for the proper interpretation of MSI datasets. Normalization is the process of multiplying a mass spectrum with an intensity‐scaling factor to expand or reduce the range of the intensity axis. It is used to remove instrumental and systemic artifacts that impact signal intensities by projecting spectra of varying intensity onto a common intensity scale.177, 178, 179, 180 Artifacts may originate from matrix application, matrix crystal heterogeneity, or ion source contamination. Among the few different normalization strategies for MSI datasets, total ion current (TIC) normalization is the most commonly implemented algorithm.181 In this approach, all mass spectra are divided by their TIC so that all spectra in a dataset have the same integrated area under the spectrum, based on the assumption that there is a comparable number of signals in each spectrum.177, 182 This limits signal comparison to similar types of tissue.181

Optimal normalization of MSI data is achieved by use of labeled standard/s applied to the tissue surface (MALDI),183, 184, 185 or added to the extraction solvent (DESI). This is the approach most commonly used during targeted MSI of drugs and metabolites.186, 187, 188 For each spectrum or pixel, the signal for the targeted analyte is normalized by the reference standard correcting for compound‐dependent ionization effects. This approach enables targeted imaging of single lipids or lipid classes (as demonstrated for phosphatidylcholine and lysophosphatidylcholine).187 However, lipidome‐wide normalization is still a challenging issue that needs to be addressed.

Direct quantification of phospholipids in tissue by MSI has been achieved by both MALDI and DESI.187, 189 Landgraf et al demonstrated quantitation of seven phosphatidylcholine lipids in nerve tissue by MALDI‐MSI showing good correlation to previously published data.189 Lanekoff et al applied nano‐DESI MSI shotgun lipidomics to quantify 22 phospholipids with reasonable agreement to previously published data.187 However, the feasibility of routine quantification remains to be seen, and to confirm absolute quantitation achieved by MSI, results must be cross‐validated with an established quantification method, such as traditional LC‐MS analysis of tissue homogenates.183, 185, 190

9. APPLICATIONS

In recent years, the role of lipids and the measurement of their spatial distribution have become the subject of increased research activities.191, 192 Here, we focus on applications of MSI to imaging of lipid cancer biomarkers within the last 5 years. MSI has been successfully applied to investigate lipid profiles and specific molecular markers of several types of human cancer including brain,193, 194 breast,129, 195 colon,152, 196 lung,148, 197 kidney,55, 149 prostate,82, 198 thyroid,199, 200 reproductive,201 esophagus,202, 203 lymphatic,204 oral,147, 205 skin,206 and bone.207 Representative imaging data from recent studies are shown in Figure 3, and a concise summary of recent applications is presented in Table 1. These studies have effectively relied on variations in the relative abundances of lipids of a specific lipid class, or a variety of lipid species from many classes, as a means of identifying diagnostic signatures. Some recent examples of MSI applied to tumor lipid biomarker discovery, classification, and pathological interogartion are presented here.

Figure 3.

Examples of MS lipid imaging applied to cancer research. Clockwise from top: Brain—PCA projection (false color heatmap) image of lipid markers resolving glioma (red pixels) from white matter (blue pixels). Reproduced with permission from Jarmusch et al.193 Oral—Visualization of 18 lipid classifiers for resolving tumor (black outline in classifier image, red outline in H&E) from epithelium (grey outline in classifier image, blue outline in H&E) in clinical tongue cancer resection surgery. The heat maps illustrate the probability of being classified as cancer. Reproduced with permission from Bednarczyk et al.148 Colorectal—Chemical reconstruction of tissue regions of interest in colorectal cancer biopsy tissue using multivariate molecular ion patterns. Specific molecular ion patterns resolve colorectal adenocarcinoma (red) from blood vessels (green) and healthy muscle tissue (blue). Reproduced with permission from Veselkov et al.157 Breast—The heterogeneity of human breast tissue biopsies is shown by SIMS imaging. H&E‐stained microscopy image (A) showing cancerous (purple) and stroma (bright) tissue; overlay mass spectrometry image of tissue sections consecutive to the H&E‐stained sections, red (FA(20:4), m/z 303.2, stroma) green (FA(18:1), m/z 281.3, cancerous tissue) (B); PCA scores image generated by principal components analysis of MS image data (C), showing clear differentiation of tumor and stroma. Reproduced with permission from Angerer et al.129 Skin—DESI MSI of basel cell carcinoma biopsy. Tumor areas are outlined in red in the histological image (A); arachidonic acid (m/z 303.5) is enriched in tumor regions as shown in 2‐D DESI image (B). Reproduced with permission from Margulis et al.206 Colon cancer liver metasteses—MALDI‐MSI resolves two distinct types of necrosis, infarct‐like necrosis (ILN) and usual necrosis (UN). PLS classifications of lipid signatures clearly resolves necrotic (purple) from surrounding tissue (A), thin left arrow in ILN, fat right arrow in UN. SM(d18:1/16:0) in red and PC (p‐16:0/18:1) in green are localization markers for ILN and UN respectively (B). Reproduced with permission from Patterson et al₍₂₅₃₎

Table 1.

Summary of recent cancer lipid studies using MSI

Cancer Tissue Type		Investigated Lipid Classes	MSI Technique	Reference
2014	Breast	PL, FA	DESI	248
2014	Breast	PC, FA	MALDI	202
2014	Colon	PL	DESI	157
2014	Colon	Not specified	DESI	157
2014	Colon	PC|PC, SM	MALDI	202, 249
2014	Kidney	PC	MALDI	39
2014	Lung	SL, Cer|PC, FA	MALDI	202, 250
2014	Lungb	PL	MALDI	251
2014	Thyroid	PC, PA, SM|PC, FA	MALDI	202, 252
2014	Gastric	PC, FA	MALDI	202
2014	Gastric	PL|PI, PS, PE, FA	DESI	253, 254
2014	Esophagus	PC, FA	MALDI	202
2014	Gastric	PL	DESI	254
2014	Oral	PC, FA	MALDI	205
2014	Thymus	PL, PG, CL	DESI	255
2014	Prostate	PI	MALDI	82
2014	Breast cancer cells	PA, FA	SIMS	127
2015	Breast	PI, PE, PC, FA	DESI	256
2015	Breastb	PI, PC SM	MALDI	221
2015	Liver	PL	REIMS	217, 257
2015	Brain	PL	DESI	148
2015	Brainb	PA, DAG	MALDI	258
2015	Lung	PC	AFA‐DESI	259
2015	Bone marrow	PC	MALDI	207
2016	Brain	PC, PE, PS, GalCer	DESI	193
2016	Breast cancer cells	PG, PA, SM	MALDI	195
2016	Breasta	PA, PG, SM	MALDI	195
2016	Breast	PL	AFAI	41
2016	Breast	PI, FA	SIMS	260
2016	Breastb	Cer, FA|PI, FA	DESI	261, 262
2016	Breast	PI,FA, SM, PS	SIMS	129
2016	Breast	FA, DAG	SIMS	263
2016	Colon	PI, FA	MALDI	196
2016	Esophagus	PL	DESI	203
2016	Liver	PI, PE, PC, Cer, SM	MALDI	264
2016	Gastric	PC	MALDI	265
2016	Ovary	PA, PS, PG, Cer	DESI	201
2016	Pancreas	FA, PI, PC, PG, PS	DESI	213
2016	Thyroid	CL	DESI	199
2017	Breast	PC, FA|PC, SM	MALDI	266, 267
2017	Kidney	PL, Cer	MALDI	55
2017	Ovary	PC, PG, CL, Cer, FA	DESI	238
2017	Ovary	Cer, PG, PC, CL	DESI	208
2017	Colon	PE, PI, PG, PS	3‐D‐DESI	152
2017	Prostate	PL	DESI	268
2017	Lymphatic	PE, PG, PS, PI, Cer, DAG, FA	DESI	269
2018	Breast	PI, PE, PG, PS	DESI	270
2018	Colon	PE	MALDI	271
2018	Colon	PC, CS, SM, FA	AP‐MALDI/SIMS	272
2018	Kidney	PI, PS	MALDI	273
2018	Brain	PG, SL	MALDI	194
2018	Lung	PL	MALDI	155
2018	Lymphatic	PI, SM, PE, CL	MALDI	166
2018	Thyroid	PC, PA, SM	MALDI	200
2018	Skin	PC, PG, PS, PI, FA	DESI	206
2018	Pancreas	FA, PC, DAG, SM, PE	SIMS	219
2018	Breast	PI, PS, SM	SIMS	129
2019	Lung	PC	MALDI	274
2019	Brain	PA, PE, PS, PI	3‐D‐MALDI	40
2019	Tongue	Multiple	MALDI	147
2019	Kidney	FA	DESI	149
2019	Brain	PI, ST	MALDI	215

Abbreviations: CL, cardiolipins; CS, cholesterols; DAG, diacylglycerols; FA, fatty acids; Cer, ceramides; GalCer, galactosylceramides; LPC, lysophosphatidylcholines; PL, phospholipids; PA, phosphatidic acids; PC, phosphatidylcholines; PE, phosphatidylethanolamines; PG, phosphatidylglycerols; PI, phosphatidylinositols; PS, phosphatidylserines; SL, sphingolipids; SeL, seminolipids; SM, sphingomyelins; ST, sulfatides; DAG, diacylglycerols; TAG, triacylglycerols.

^aCell culture.

^bMurine xenograft model.

9.1. Use of MSI for tumor diagnosis, classification, and grading

Cancer prognosis and clinical treatment decisions crucially depend upon accurate classification of tumor type and grade. Typically, high‐grade tumors are more likely to grow and spread faster than low‐grade tumors. Therefore, molecular markers identifying these tumors at the earliest possible staes of cancer will have high value in clinical cancer research. Multiple MSI approaches have been used to identify and grade a wide range of cancer types as listed in Table 1.

Mao et al41 applied a novel air‐flow assisted ionization MSI method to differentiate breast invasive ductal carcinoma (IDC) and breast ductal carcinoma in situ (DCIS). Tumor subtype was identified and tumor grade was validated. The identified PC, PE, SM, and FA markers showed 100% and 78.6% agreement with the histopathological diagnosis, respectively. This technique has high potential importance as there are no current molecular methods enabling rapid classification in near real time.

Lipid biomarkers of oral cancer were evaluated in comparison to protein biomarkers for their ability to distinguish between tumor and healthy mucosa using MALDI‐MSI and downstream spatial segmentation and PCA analysis.147 Although peptide signatures had a higher weighted accuracy (0.85 v 0.69), both had extremely high precision (0.98 and 0.94, respectively), indicating their suitability as biomarkers of oral malignancies. A significant advantage of the lipid‐based imaging approach is speed as peptide imaging requires time consuming trypsin digestion steps (18 h in this study). MALDI MSI has also been applied to identify phosphatidylglycerols and sphingolipids as the top classifiers to distinguish the two histopathologically similar brain tumor types medulloblastoma (MB) and pineoblastoma (PB), respectively.194

Jones et al39 developed a novel MALDI MSI method to study the distribution of sphingomyelins and ceramides in lung tissue in which the tissues were subjected to enzymatic lipid digestion to release sphingomyelins and ceramides. This strategy was then used to analyze the sphingolipid and ceramide profiles of clinical clear cell renal cell carcinoma biopsies, accurately identifying markers of recurrent disease progressors from nonrecurrent disease individuals.

DESI‐MSI has rapidly gained acceptance for tumor diagnosis and classification studies due to the rapid analysis speed and simple sample preparation.156 Margulis et al applied DESI‐MSI to diagnose early stage baseal cell carcinoma (BCC) from healthy human skin.206 Using the least absolute shrinkage and selection operation (LASSO), 24 molecular lipid and metabolite markers were identified that enabled 94.1% tumor diagnostic accuracy compared with histopathological evaluation in 86 human skin biopsies. Using a similar approach, lipid predictive markers of ovarian high‐grade serous carcinoma (HGSC) aggressiveness were identified (ceramides, glycerophosphoglycerols, cardiolipins, and glycerophosphocholines), enabling their differentiatiation from low grade serous borderline ovarian tumors (BOT) with 93% accuracy.208

Free fatty acids have previously been identified as candidate cancer biomarkers markers in tissue209 and plasma210, 211 by GC/MS and LC/MS. Tamaura et al imaged and identified five potential fatty acid biomarkers of clear cell renal cell carcinoma in 47 patient biopsies using DESI‐MSI and statistical processing by orthogonal projections to latent structures discriminant analysis (OPLS‐DA).149 Of these, oleic acid was shown to corelate with disease progression and shorter duration of progression free survival.

9.2. Use of MSI for intraoperative imaging of tumor margins

Frozen section histology is the traditional technique used to provide tissue diagnosis during surgery.24 This process involves rapid freezing, cutting, staining, and examination of a section of tissue. Alternatively, imprint cytology may be utilized, where cut biopsy surfaces are pressed onto a glass slide, which is then fixed and stained.212 Both techniques require microscopic assessment by a highly skilled pathologist. However, accuracy of histology delineation using traditional approaches can be problematic. For example, delineation of pancreatectomy margins can be variable and subjective, with false‐negative results occurring in up to 20% to 30% of pancreatic adenocarcinoma patients, usually leading to additional surgical intervention.213

Eberlin et al applied DESI‐MSI and downstream LASSO analysis to intraoperative evaluation of surgical resection margins in pancreatic cancer. Using customized training LASSO prediction, they were able to evaluate surgical margins at >98% agreement with histopathology in tissues with complex, mixed histology.213 Interestingly, patients with noncorrelating biopsies in which margins tested positive by MSI but not frozen analysis had significantly lower median survival times. The ability of the MSI method to detect molecular changes at tumor margin in patients who developed early recurrence demonstrates the clear potential role for MSI of lipid biomarkers in clinical cancer surgery workflows.

Most tumor margin studies have been conducted by DESI‐MSI due to the ambient nature of ionization, rapid analysis times, and simple sample preparation protocols. However, recent developments in MALDI‐MSI such as high repetition rate and continuously firing lasers have vastly increased analysis speed.134, 214 Basu et al have presented a rapid MALDI‐MSI method optimized for lipid marker imaging in surgical pathology.215 By depositing tissue sections onto matrix precoated slides and analyzing using a high‐speed RapiFlex mass spectromter with a 10 kHz laser, the entire MSI workflow was condensed to less than 10 minutes. In the proposed workflow, the MS image is colocalized and fused to a H&E‐stained adjacent section image for detailed evaluation by a pathologist or analysis through artificial intelligence guided diagnosis.

Rapid evaporative ionization mass spectrometry imaging (REIMS)216 is an ambient analyticalapproach that ionizes molecules from biological tissue by exposure to an alternating electrical current during electrosurgical dissection. Using the technology, phosphatidylethanolamine markers of tumor tissue were identified, enabling tumor to be resolved from surrounding healthy tissue.217 Although spatial resolution is significantly poorer than alternative MSI technologies (750 μm pixel size), the approach does not require tissues to be frozen and cryosectioned, making it arguably the quickest and simplest intraoperative imaging method.

9.3. Use of MSI for pathogenesis studies

Elucidating molecular markers of cancer pathogenesis is essential for understanding how cancers grow, proliferate, and metastasize. This information forms the basis for the discovery of novel biomarkers and targeted therapies.21

Due to the longer image acquisition times and requirement for bulky, expensive, and high vacuum instrumentation, SIMS is not suitable for intraoperative imaging. However, the primary advantage of SIMS is that it offers unique spatial resolving capabilities, enabling lipid markers to be visualized at subcellular detail. For this reason, SIMS has proven extremely useful for pathogenesis studies where it provides subcellular molecular insights into cancer pathogenesis.127, 129, 218, 219, 220, 221

SIMS has been applied to visualize lipids in individual cancer cell lines. Waki et al used ToF‐SIMS to image lipd distributions in single breast cancer stem cells isolated by fluorescence activated cell sorting (FACS) and identified palmitoleic acid as a biomarker with reduced abundance in tumor cells.127 SIMS lipid analysis of eight human breast cancer cell lines followed by PCA analysis revealed separation of different receptor positive cells based upon fatty acid, cholesterol, and mono‐, di‐, and triacylglycerol profiles.218

In biopsy tissue sections, SIMS enables subcellular imaging of lipid biomarkers providing detailed insight into the tumor microenvironment. ToF‐SIMS was applied to visualize molecular crosstalk between cancer and normal cells in Myc‐induced pancreatic β cell islet tumor microenvironment.219 Using identified lipid markers, cancerous islets were resolved from normal tissue, and the high spatial resolution offered enabled intra tumor heterogeneity to be evaluated. Using the GCIB SIMS instrument described earlier and downstream PCA analysis, Angerer et al identified FA, SM, and PI markers correlating to necrotic tumor, cellular tumor, and stroma microenvironments.129

MALDI‐MSI has been used to investigate hypoxia in tumor biopsy tissues.220, 221

Phosphatidyl markers of hypoxia were identified in a mouse model of human breast cancer. As hypoxia has been suggested as a driver of tumor resistance to radiotherapy or chemotherapy,222 elucidation of its biochemistry may be of high importance for improving therapeutic efficacy.

3‐D MALDI or DESI imaging of tumor biopsies provides a comprehensive overview of lipid distributions throughout the entire tissue.40, 152 Because of the highly heterogeneous nature of tumors, 3‐D imaging approaches may detect lipidomic changes occurring within tumor cells or regions than could be missed in a single 2‐D tissue section.

10. MSI IN MULTIMODAL CANCER IMAGING WORKFLOWS

The combination of different modalities offers a comprehensive analytical tool to answer biological questions that could otherwise not be answered. In the field of laterally resolved chemical analysis, there is a trend towards the combination of two or more technologies, so‐called multimodal imaging approaches, to provide more information than offered by a single modality. While MSI offers high chemical specificity for visualizing and identifying distributions of several molecular species, it lacks the molecular depth that other methods provide and, thus, is typically combined with modalities that complement these features.223, 224, 225, 226, 227, 228 Due to increasing MSI spatial resolving capabilities, accurate cellular‐level coregistration of MSI datasets with histological reference sections is of vital importance for interpreting visual biomarker studies. Recent software developments and semi‐automated workflows are enabling rapid and highly accurate coregistration suitable for use in cancer lipid biomarker discovery.159, 160, 229

Due to the specific ionization properties of the variety of ionization modalities, MSI can be multiplexed with itself to analyze different groups of compounds. In particular, MALDI can be combined with a range of ionization mechanisms such as DESI230 or LA‐ICP‐MS,155, 231, 232 providing a more comprehensive analysis of the sample.

The ability to visualize multi‐omic tumor biomarkers in the same or serial tissue sections has the potential to provide comprehensive information for entire molecular pathways in cancer pathogenesis. The combination of MALDI and DESI in a multimodal imaging approach was introduced by the Cooks group in 2011.230 Sequential imaging of a single human glioma tissue section allowed for correlations between a higher tumor cell density and increased relative abundances of fatty acids as well as a specific calcium‐binding protein.230 Recently, Dewers et al introduced a novel multimodal workflow utilizing lipid biomarkers visualized by MALDI MSI to guide tumor ROI microdissection and subsequent quantitative proteomic analysis.159

In 2018, a single malignant pleural mesothelioma (MPM) section from a patient treated with cisplatin as the cytostatic agent has been analyzed by MALDI and LA‐ICP‐MS,155 providing full distributional information on phospholipids and quantification of drug‐derived platinum.155 By using specific phospholipids as tumor biomarkers, tumor drug penetration could be assessed.

As different imaging platforms often demand distinct sample preparation protocols, analyzing adjacent tissue sections might seem most straight‐forward. However, even two sequential tissue sections cut at some tens of micrometer distance may display variations in analyte distributions. These differences may originate from the sample itself because different cells may have different compositions, or from the sample preparation process where tissue sections are not cut accurately enough and have different thicknesses. Therefore, analyzing the same tissue section using two different methods is highly desirable to exclude intersample variations but also to guarantee appropriate quality after data combination. The latter is an issue because image coregistration is challenging if data are not acquired at the same spatial resolutions or from the same sample, requiring sophisticated data processing methods.233

11. FUTURE DIRECTIONS AND CONCLUSION

Improvements in the speed, sensitivity, and spatial resolution capabilities of MSI are continually expanding the applications to lipidomics research. Coupling MSI to additional separation techniques can significantly enhance both sensitivity and selectivity of analyses. The incorporation of ion‐mobility separation following ionization enables isobaric lipids to be resolved as well as resolving lipids from interfering matrix and endogenous background peaks.234 Advances in tumor spheroid reproducibility have created a viable high‐throughput model for characterizing tumor development and assessing drug therapy approaches.235 When coupled with the specificity of MSI, lipid signatures can be assessed in tumor cell subpopulations,196 and drug penetration into tumor regions of interest may be evaluated.236

Whilst not providing a true image of tissue, ambient ionization technologies such as rapid evaporative ionization mass spectrometry (REIMS) “Iknife,”237 water droplet‐based “MasSpec Pen,”238 and SpiderMass (which uses resonant infrared laser ablation (RIR‐LA) of endogenous water molecules to enhance ionization)239 have the significant advantage of being able to be applied during surgery. Moreover, due to the nondestructive nature of the MasSpec Pen and Spidermass technologies, downstream processing of tissues for histology is possible. These technologies enable surgeons to distinguish tumor from surrounding healthy tissue and accurately determine whether all cancerous tissue is excised during surgery,240, 241, 242 leading to a potential future of seeing mass spectrometers routinely placed in operating suites. Laser capture microdissection (LCM) with online coupling to ESI‐MS or off‐line coupling to LC‐MS/MS has recently been developed for spatial profiling and quantification of lipids in thin tissue sections.243, 244 Again, this is not a true imaging approach, but microscope guided dissection enables absolute quantification of small tissue areas offering valuable complementary data to MSI.

A recently introduced technology, imaging mass cytometry (IMC), applies the specificity of MSI to visualize the localization of metal‐chelated antibodies as an alternative method to classical immunohistochemistry (IHC), fluorescence, or chromogen labels. Metal tags can easily be chelated to existing, validated antibodies and images of metal distributions within the tissue are generated from respective isotopes measured by MS.245, 246 Due to the lack of autofluorescence, this method is accepted to be more sensitive than existing IHC approaches and is capable of simultaneously imaging more than 40 tissue markers.247 It has great potential to provide complementary high‐resolution (1 μm²) protein and enzyme localization to MSI allowing full visualization of lipid metabolic pathways within tumors.

While many significant technical issues affecting MSI of lipids are being addressed by recent advancements, data handling and processing are a crucial area requiring further development before the technique becomes a routine analytical tool in cancer research. Simple and robust online processing tools enabling identification of lipids and classification of tumors type have the potential to bring MS specificity into the hands of surgeons. In basic and translational research, advanced processing algorithms for lipid identification in multimodal MSI studies may allow accurate identification of low abundance tumor lipid markers, potentially leading to the discovery of novel therapeutic targets, which may otherwise be lost amongst the many thousands of peaks detected in a typical high‐resolution mass spectrum. Such applications highlight the promising future of MSI in cancer research. However, potential lipid biomarkers must be accurately identified through use of large and clinically well defined patient cohort studies, and for the immediate future, visual evaluation of histological slides by trained pathologists remains the gold standard in clinical and research cancer studies.

In conclusion, MSI offers a unique approach for biomolecular characterization of tumor tissues. By providing the detailed molecular information lacking in traditional histology approaches, novel lipid biomarkers and therapeutic targets may be discovered, leading to breakthroughs in diagnosis and treatment. We believe MSI will become an increasingly important tool for evaluating lipids in cancer, both in the laboratory and in the operating room.

CONFLICT OF INTEREST

The authors declare no conflicts of interest in connection with this article.

AUTHORS' CONTRIBUTIONS

All authors had full access to the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Conceptualization, M.H., E.E. and B.P.; Resources, E.E. and B.P.; Writing ‐ Original Draft, M.H., E.E. and B.P.; Writing ‐ Review & Editing, E.E. and B.P.; Visualization, M.H. and B.P.; Supervision, E.E. and B.P.; Funding Acquisition, E.E. and B.P.

Holzlechner M, Eugenin E, Prideaux B. Mass spectrometry imaging to detect lipid biomarkers and disease signatures in cancer. Cancer Reports. 2019;2:e1229. 10.1002/cnr2.1229