Analytical challenges of shotgun lipidomics at different resolution of measurements

Abstract

The essence of shotgun lipidomics is to maintain consistency of the chemical environment of lipid samples during mass spectrometry acquisition. This strategy is suitable for large-scale quantitative analysis. This strategy also allows sufficient time to collect data to improve the signal-to-noise ratio. The initial approach of shotgun lipidomics was the electrospray ionization (ESI)-based direct infusion mass spectrometry strategy. With development of mass spectrometry for small molecules, shotgun lipidomics methods have been extended to matrix-assisted laser desorption/ionization mass spectrometry (MALDI MS) and ambient mass spectrometry, including MS imaging methods. Furthermore, the object of analysis has extended from organ and body fluid levels to tissue and cell levels with technological developments. In this article, we summarize the status and technical challenges of shotgun lipidomics at different resolution of measurements from the mass spectrometry perspective.

1. Introduction

The complete set of lipids in tissues or cells is referred to as the lipidome [1]. According to the classification of the LIPID Metabolites and Pathway Strategy (LIPID MAPS) project, lipids can be divided into eight main categories: fatty acyls, glycerolipids, glycerophospholipids, sphingolipids, sterol lipids, prenol lipids, saccharolipids, and polyketides [2, 3]. Each category can be further classified into different lipid classes and subclasses, based on their polarities, charge(s), sizes, linkages to the backbone, and distinct functional groups. The LIPID MAPS structure database currently records 43,616 unique lipid structures (as of 9/13/2019) [4]. Lipid functions are highly related to their structure, concentration, and spatial and temporal distribution. The composition, distribution, and dynamics of the lipidome reflect the status of cellular metabolism [5]. Dysregulation of lipid metabolism is associated in the onset and progression of many diseases, including cardiovascular diseases, neurodegeneration, diabetes, and cancer [5, 6].

Lipidomics is the study of lipids based on qualitative and quantitative analysis of lipidomes in a systematic fashion [7, 8]. Lipidomics analysis is facilitated by decades of biological mass spectrometry advancements [9–14]. Due to the huge difference in lipid properties from water-soluble metabolites, lipidomics is very different from conventional metabolomics in terms of sample preparation, mass spectrometric analysis, and data processing. Due to the chemical nature of lipid molecules, metabolomics and proteomics techniques developed based on aqueous systems are not always suitable for lipidomics analysis. Lipidomic analysis requires non-aqueous phase extraction and ionization, and it considers lipid aggregation under certain conditions, which are unique features of lipid analysis [15].

Shotgun lipidomics analysis occurs under specific experimental conditions with a constant lipid solution concentration [12, 16–18]. In most cases, there is no separation process conducted prior to the mass spectrometric analysis. Thus, shotgun lipidomics can provide sample signals from a uniform chemical environment. Since the chemical environment of the solution do not change during analysis, differences in signal response of homologues are minimized. The signal-to-concentration relationship can be calibrated readily. Accurate absolute quantification after normalization to a common denominator can be achieved with a limited number of internal standards [15, 19]. As a result, high-accuracy and large-scale absolute quantitative analysis is easier to achieve by shotgun lipidomics than by separation-based lipidomics.

According to the essential features of shotgun lipidomics, the scope of shotgun lipidomics has been extended to cover analyses after both direct infusion and desorption (including imaging analysis) [20]. The direct infusion-based electrospray ionization is the earliest ionization strategy used in shotgun lipidomics. This is mainly because, in the early stages of shotgun lipidomics, it was the best approach for direct analysis of lipid extraction in homogeneous solutions. This approach is still the most widely used in quantitative shotgun lipidomics [16, 21]. With the development of MALDI matrices that are specific to small molecules, technical approaches in shotgun lipidomics have been expanded [22–26]. Recently, with the development of ambient ion sources (such as desorption electrospray ionization (DESI), laser ablation electrospray ionization (LAESI), and liquid extraction surface analysis-mass spectrometry (LESA-MS)) and mass spectrometry imaging (MSI) technologies, the scope of shotgun lipidomics has been expanded into direct desorption analysis [27–31]. Currently, analytical methods based on direct desorption or imaging analysis are becoming more important in lipidomics analysis because they provide valuable information on the spatial distribution of lipids. The combined use of these approaches should be very effective to understand lipidomes on different resolution. Figure 1 shows a number of methodological approaches to expand the scope of shotgun lipidomic analysis. Table 1 summarized the significance and purpose of shotgun lipidomics analysis at different levels of measurements for biomedical research.

Figure 1.

Approaches to expand the scope of shotgun lipidomic analysis.

Table 1.

Different measurement levels of shotgun lipidomics for the scenarios and objectives of biomedical research.

	Macro/Organ Level	Micro/Tissue Level	Single-cell Level
Representative samples	Whole organs (e.g., liver, heart)/Body Fluids (e.g., serum, plasma)	Tissue section/Local part of organs (e.g., islet, dorsal root ganglion)/Collection of organelles (e.g., mitochondrial)	Cultured cells/Tissue section
Application scenario	Functional research. The role of lipidome in the development of the disease is understood from the perspective of the overall physiological function of the organs or the whole body.[135, 136]	Corresponds to research at the resolution of histological level or above. It is of great significance for the study of drug functions and release, as well as the mechanism or the development of diseases. [29, 122, 137, 138]	Corresponds to studies at the cellular and molecular levels. It is the cornerstone of basic research.[85, 92, 97]
Technical maturity of biomedical applications	High maturity. There is a need to increase the automation and standardization of quantification. The analysis of the fine structure of lipid molecules is at the forefront of methodological research.	Generally, technical mature in the qualitative analysis. The quantification analysis is not mature enough. Increased sensitivity is the key to solving the problem.	Technically immature. The current research is focused on methodological study. In addition to improving sensitivity, micrometer/nanometer operations and control are also technical bottlenecks.

The ultimate goal of lipidomic analysis is to achieve a full coverage of precise structural analysis, accurate quantification, and a comprehensive understanding of lipid dynamics at all levels. This goal is far from the existing analytical capabilities. We believe this requires a combination of lipidomics based on both efficient separation and direct analysis. Shotgun lipidomics offers unique advantages in both quantitative and spatial distribution analysis.

In this review, we overview the progress in and challenges of shotgun lipidomics analysis at different resolution of measurements, with an emphasis on instrumental analysis and sample processing. We discuss current technical characteristics and bottlenecks, to provide insights that inspire new ideas and drive novel technological advancement.

2. Shotgun lipidomics at spatial levels

2.1. Analysis at macro/organ level

This level is a macroscopic explanation of changes in the whole body or organs. For solid tissue, this level of analysis is for the representative part of the organ or entire organ. Its resolution is usually in the order of a few millimeters or tens of millimeters. For body fluid samples, its analytical difficulty depends on the type of body fluid. Lipid concentrations vary widely among different body fluids. For serum or plasma, approximately lipid content in one hundred microliters corresponds to lipids contents in 10 milligrams of tissue. For sample with a very low lipid content such as cerebrospinal fluid, several times or tens of times of the volume of plasma or serum are required. Analysis of serum and plasma is relatively easier to achieve than tissue samples. The main reason is that the quantity of samples is sufficient, the samples are homogenate, and the lipid content is usually abundant. However, for samples of cerebrospinal fluid, the total amount of samples that can be obtained is limited, and it is hard to quantify low abundance lipid species. At the level of organs and body fluids, quantitative shotgun lipidomics is widely used in mechanism studies of diseases associated with lipid metabolism or in biomarker discovery studies [20, 32–34]. The most commonly measured lipid classes are sphingolipids (SL), glycerophospholipid (GPL), glycerolipids (GL), and non-esterified fatty acid (NEFA). They are involved deeply in the construction of cellular membrane, energy metabolism, transportation, and signal transduction [6]. More than 45 classes of lipids are quantifiable by shotgun lipidomic analysis [5]. In this macrolevel analysis, tens of milligrams of tissue samples are typically required for quantitative analysis. The mass spectrometry strategies or modes that can be used are very diverse, including precursor-ion scan, neutral loss scan, high-resolution mass spectrometry, and multi-dimensional mass spectrometry strategy (MDMS) [11, 35–39]. Absolute quantitative analysis has been easily implemented in a homogeneous direct infusion system. In the case of quantitative analysis using the MDMS method, it is possible to achieve at least one internal standard for each lipid class. The choice of the internal standard needs to avoid interference with endogenous substances, and the spiking content is not less than 10% of the high abundance component. In the calculation, the correction of the abundance of ¹³C isotope with molecular weight changes should be considered. The in-depth review by Wang et al. [15] should be helpful for considering the selection of internal standards. In this level, the overall quantitative accuracy for lipid species that present in modest-to-high abundance can be greater than 90% [40]. For a brief introduction to the basic technical concepts of lipidomics, please refer to a tutorial written by Wang et al. [41].

From the perspective of instrumental analysis, there is no practical difficulty in the quantitative accuracy of the macro/organ level. However, if the sample collection and quality control does not attract enough attention, it will affect the accuracy of the quantification [42]. At the macro/organ level, the minimal sample size for analysis is relatively easy to guarantee. Since human to small animal samples are used, there are still certain requirements for sample consistency to avoid data variation. The difficulty of sample collection is varied among different organs. Samples from the liver and muscle are relatively homogenous and easily collected. However, if only a partial part of the organ is collected, the distribution of fat and blood vessels needs to be considered. For organs with complex and non-homogeneous morphological structures, such as the brain and kidneys, partial excision and collection is more likely to cause data variations. Further, caution should be taken when collecting from small animals, such as mice. The accuracy of the anatomical position and degree of exfoliation of the adherent tissue have a great influence on the sample consistency.

In addition, different biological models should be considered. Usually, sample deviation among patients are very large. If it is a model organism, the wild type should be stable. However, due to the introduction of additional interventions, the level of gene expression is not usually consistent among samples in the experimental group, thus variations in lipid quantification may increase. Therefore, in addition to careful handling of the sampling operation, background information of the sample should be considered to determine the appropriate number of samples for statistical analysis.

There are primarily two analytical challenges at this level: accurate quantification of low-to-modest abundance lipids and determination of fine structural information. The concentration range of lipids in a biological sample varies from amol to nmol/mg of protein (estimated from LIPID MAPS). Ion suppression by high abundant species during ionization or trapping capacity limits the dynamic range of a mass spectrometer. However, the quantitative dynamic range of shotgun lipidomics is only approximately four orders of magnitude. Therefore, using the MDMS strategy, selective ionization, and pre-separation of the lipid extraction by liquid-liquid partitioning or liquid-solid partitioning should be very useful to significantly increase the dynamic range [38, 43]. For example, triglycerides are the predominant component in adipose tissue. Pre-separation of triglycerides from other polar lipids is necessary for shotgun lipidomics analysis of the majority of lipid classes [8]. This challenge is not only due to the full range of lipid concentration distribution, but also because the lipids self are prone to aggregation. To avoid or reduce the adverse effects of aggregation, the concentration of each lipid for infusion should be less than 10 pmol/μl.

Derivatization strategies can be used to significantly improve signal quality in multiple ways [44]. First, derivatization can lead to increases in ionization sensitivity by introducing a functional group that is prone to ionization. Second, derivatization can improve the linear dynamic range of lipid analysis by selectively improving ionization efficiency of low abundance lipid classes. For example, Trimethylsilyl (TMS) derivatization of phosphatidylinositol phosphate [45] and N-(4-aminomethylphenyl)pyridinium (AMPP) derivatization of NEFA [46] greatly enhanced the quantification range. Moreover, derivatization can enhance analytical specificity with tagged moieties. For example, phosphatidylglycerol (PG) and bis(monoacylglycero)phosphate (BMP) are isomers. They can be distinguished by TMS derivatization due to differences in the characteristics of derivatized fragments [47]. In addition to expanding the derivatization coverage of lipid classes, simplification of the derivatization process and methods that do not require purification are important development directions, especially for smaller size analyses. For a more detailed understanding of derivatization methods, please refer to recent review articles written by Hu et al. and Ryan et al. [44, 48].

Comprehensive coverage of fine structural information of lipid molecules is an important research area at this level. Positional determination of unsaturated bonds in lipid molecules and differentiation of cis and trans isomers are currently critical challenges [49]. In diseases such as cancer, the content of isomers differs from the normal state [50]. Therefore, determination of fine structure and quantitative or ratio analysis are required. Current studies on the localization and quantification of carbon-carbon double bonds employ oxidation reactions on double bonds, including ozone induced dissociation (OzID) [51], photochemical method such as Paternò–Büchi reaction [52], and ultraviolet photodissociation (UVPD) [53]. A recently published review article [54] provides detailed information about this topic.

In the high vacuum environment of common mass spectrometry, we do not consider ion collisions because their mean free path is greater than the dimensions of the vacuum cavity. However, after the introduction of collision gas, the mean free path of ions and the average kinetic energy change. These changes are related to the collision cross section of ions. The collision cross section is determined by their molecular structures. Ion mobility mass spectrometry (IMS-MS) is an effective method to distinguish three-dimension conformation such as cis-trans isomers [55]. There are many variants of ion mobility mass spectrometry. These methods combine different collision gas conditions with different electric field modes to produce different separation fields. The ions are separately detected based on differences in their collision cross sections and mass-to-charge ratios. High field asymmetric waveform ion mobility spectrometry (FAIMS) [56], drift tube ion mobility spectrometry (DTIMS) [57], and traveling wave ion mobility spectrometry (TWIMS) [58] are major IMS-MS technologies that have been used for lipidomic analysis. Due to similarities in the common structure of a specific lipid class and small differences in the collision cross section of each lipid species, the separation power of complex mixtures is limited. Nevertheless, the combination of ion mobility mass spectrometry with MALDI, DESI, or other ambient ionization methods for lipid extraction could accurately distinguish high abundance components [59, 60]. A recently published review article [61] can be consulted to learn new methods and ideas for lipid separation and structural elucidation by ion mobility mass spectrometry.

At the macro level, direct infusion analysis is still a dominant approach. Table 2 summarizes the comparison between related technologies and challenges.

Table 2.

Technical characteristics of shotgun lipidomics at different level of measurements.

Scope of Analysis	MS	Identification	Quantification	Derivatization	Fine structure characterization	Major technology trends
Macro/Organ (Of or above 1 mm)	ESI MALDI Ambient MS	>1000 species; > 45 lipid classes	Absolute quantification	No limitation	***** Limited quantitative methods	Automated high-throughput process
Micro/Tissue(20–200 μm)	ESI MALDI Ambient MS	<1000 species; Major species of SL, GPL, GL and NEFA	Relative quantification; Limited Absolute quantification	Limited choices	**	Efficient extraction and derivatization methods
Single-cell/Subcellular(<10 μm;Below mean cell size)	ESI MALDI SIMS LAESI LESA	<100 species; High abundant species of SL, GPL, GL and NEFA	Relative quantification	Rare	*	Improvement of ionization efficiency

2.2. Analysis at micro/tissue level

In this review, we refer to analysis of a sample size of less than 2 mg or a spatial resolution for imaging from 20 to 200 μm as micro- or tissue-level analysis. There are three scenarios for direct infusion analysis at this level. First, some changes occur only in local areas of the organ, and if analyzed at the organ level, this change is likely to be masked in the background. Therefore, the number of available samples for anatomical sampling of these small areas is very limited. Examples are the dorsal root ganglion and islets. For this type of bulk sample, homogenization before lipid extraction is usually performed, followed by a direct infusion strategy.

Second, samples are obtained from isolated primary cells or organelles. For isolated primary cells, considering the loss during isolation and purification, only up to hundreds of thousands of cells can be obtained. For organelle analysis of cultured cells, in theory, this amount of sample is sufficient to ensure quantification; however, the purity of the sample will affect the accuracy and stability of the results. For example, the separation of subcellular organelles is usually carried out by ultracentrifugation, density gradient centrifugation, size exclusion, or immunoaffinity capture. Such methods provide inadequate purity or lipidome integrity for exosome analysis [62]. Recently, the application of novel separation methods, such as flow field-flow fractionation (FIFFF), provided better purity and showed improvements in lipidomic analysis [63]. For cell or organelle samples, the sample can be extracted after sufficient dispersion. Direct infusion analysis is a commonly used strategy.

Third, tissue-level samples are usually obtained from pathological sections or tissue biopsies. For tissue section samples that do not require imaging analysis, the weight of the tissue section is usually around 1 mg, and the strategy of homogenization-extraction and direct infusion analysis is often used; however, analysis from the tissue section requires tissue transfer from a glass slide. This differs from bulk samples in that protein-based quantification becomes more difficult. This is because the tissue must be stripped from the glass slide to quantify protein content. Collection of tissue section samples can be effectively achieved using laser microdissection [64]. In addition to quantification based on protein content, other normalizers (e.g., DNA content, sample weight, or sample volume) can also be used for quantification.

The strategy for direct infusion analysis is characterized by maximum utilization of multiple MS acquisition modes. The major advantages of this approach are high accuracy of targeted or untargeted quantitative analysis. It should be noted that due to the very small amount of sample, such as < 1 mg, conventional extraction methods can cause a large amount of transfer loss. It is possible that global analysis of broad lipid classes is not achievable. Without a simplified extraction strategy, it is possible that there is no signal detectable due to severe sample loss. In this case, a compromise strategy is unavoidable. With a simplified extraction strategy, it is possible to selectively retain some lipid classes for effective detection with a small sample size [64]. For example, a simplified liquid extraction procedure of conventional biphasic lipid extraction such as methyl tert-butyl ether-methanol (MTBE-MeOH) method and butanol-MeOH (BUME) method, or a solid phase extraction (SPE) extraction strategy, can be employed [65–69]. If direct infusion analysis is coupled with monophasic liquid extraction methods (such as MeOH, ethanol, and isopropanol), a signal drop of major lipids caused by the loss of samples during transferring can be reduced [70–73]. Thus, an orthogonal extraction strategy that combines multiple simplified extraction methods, each of which consumes a portion of the sample, and the final combined data can effectively increase the data amount for lipidomic analysis. A recent review provided a good guide for solvent selection in lipid extraction [74]. In addition, the available methods of chemical derivatization in small volume samples are limited. Multiple sample manipulation steps will result in transferring losses, such that selection and development of one-pot derivatization methods, such as Fmoc derivatization of phosphatidylethanolamine (PE) [75] and 2-picolylamine derivatization of fatty acids [76], which do not require separation and purification are necessary.

Microfluidic devices may have the potential to enhance micro level lipid analysis [77]. In this study by Sun et al., the authors integrated solid-phase microextraction materials into microfluidic tubing. The advanced point of this idea is the selective separation of lipid from samples to reduce the suppression effects of other non-lipid molecules. However, current microfluidic devices are mostly developed for operation in an aqueous phase system. The materials used in microfluidic devices are usually not compatible with solvents required for comprehensive lipid extraction, because chloroform is required. Although the fully implemented example has not yet appeared, we believe that development of a slipchip microfluidic device designed with all-glass materials can be used to achieve comprehensive extraction of lipids via microlevel biphasic separation such as Bligh-Dyer extraction [78, 79].

MS imaging analysis at the tissue level is readily available. Desorption electrospray ionization (DESI) and MALDI mass spectrometry imaging are the most common methods. At the modest-to-high spatial resolution level, DESI imaging achieves resolutions between 100 μm and 200 μm. Although the resolution of the DESI method is not high, this method can realize the analysis of the native state of the tissue. This method has been widely used for clinical sample biomarker discovery and interoperation tissue boundary recognition of cancer [27, 28, 80, 81]. If the need for rapid diagnosis is not taken into account, the use of higher resolution methods such as MALDI imaging can provide more detailed information for pathological analysis, making it easier to match the results of histochemistry. In the field of application, the resolution achieved by MALDI imaging is between 20 μm and 100 μm. For example, in a model study of colorectal cancer liver metastasis, the authors screened a matrix that effectively enhanced glycerophospholipid signals in liver tissue, revealing fingerprinting lipid species for tumor metastasis at different spatial resolutions [82]. In the high-resolution mode, small tumor foci at the early-stage of metastasis of the mouse liver were revealed. There are many excellent reviews on mass spectrometry imaging methods that are worth reading [80, 83–86]. Here, we mainly discuss some technical aspects for lipidomic analysis.

The advantage of the DESI-based imaging method is its convenience to obtain essential information needed for pathological diagnosis without additional sample processing [27–29]. This method keeps the sample in its natural state. Inspired by the MDMS strategy [38, 39], which utilizes multiple MS modes and selective ionization, lipid classes can also be selectively ionized by changing the solution composition in DESI. A major drawback of this type of method is the inability to conduct absolute quantification because it lacks a precise way to add internal standards.

The MALDI imaging method enables a higher resolution analysis than DESI. A large number of small molecule matrices are available for lipidomic analysis [87]. In imaging analysis, lipids cannot be physically separated. Therefore, selective ionization is an optimal approach to improve the dynamic range and lipid coverage. Although many matrices have been developed for lipid analysis, there are not sufficient to construct a complete selection list for selective ionization analysis. The current matrix combinations for selective ionization do not compare to the flexibility of direct infusion analysis [22, 24]. For MALDI analysis, discovery of a novel matrix may also promote improvements of ionization efficiency. However, discovery of these matrices, which can significantly improve sensitivity for lipid analysis, is difficult to predict. Nanomaterial matrices are a type of matrix that is expected. However, although the nanomaterial matrix exhibits excellent low background interference properties, their sensitivity to analyzing lipids cannot be compared to organic matrices. Currently, nanomaterial matrices are more advantageous for the analysis of small molecular weight metabolites, but not lipids because the organic matrix (even those specific for metabolites analysis) still has strong or unpredictable interference peaks in the range of molecular weights less than 500 Da.

At modest-to-high resolution levels, MALDI imaging can provide absolute quantitative analysis [88]. An internal standard addition method based on a spotter or inkjet printer can accurately provide a given amount of internal standard at a modest resolution level [89]. As far as we know, there are no example that using a pneumatic spray method for accurate addition of internal standards for high-resolution imaging. However, in the case of correction with optic images, a precisely-controlled pneumatic spray method allows the accurate addition of internal standard. It should be noted that both the internal standard and matrix require a uniform distribution and co-crystallization with lipids in these sections. Therefore, the thickness of the tissue section and deposition time of the matrix need to be optimized. A thickness less than 10 μm is more conducive to quantitative analysis.

At the tissue level of analysis, sampling and extraction of small volume samples, as well as accurate and absolute quantification in mass spectrometry imaging, are currently major challenges. The acquisition of depth information, such as quantification of low abundance lipids and interpretation of structural information, becomes more difficult due to sensitivity or sample size limitations [90, 91]. Therefore, future developments may come by improving ionization efficiency. This is discussed in the following sections. Table 2 provides some comparisons and technical features of related technologies.

2.3. Single-cell level analysis

Single-cell lipidomic analysis is very important for revealing fundamental biological processes [92]. For example, cell division is a highly dynamic process that is rapidly regulated in the cell cycle [93]. This process involves dramatic changes in cell membrane structure and composition. It is known that hundreds of proteins are involved in regulation, synthesis, decomposition, modification, and transport of lipids at complex spatial and temporal levels [6]. However, people understanding of protein and gene expression regulation in this process is far beyond our understanding of lipid changes. In addition, the plasma membrane is the basis for maintaining cellular homeostasis. Analysis of its changes is necessary to reveal the disease mechanism.

Lipids are not synthesized under the guidance of the DNA sequence. Lipid synthesis can be directly derived from conversion of exogenous uptake or occur ab initio [94]. Analysis of the genome at the single-cell level can be performed using amplification and labeling methods to achieve highly sensitive analysis [95]. Proteomics can use the antibody or labeling technology to achieve high-resolution analysis with the laser ablation inductively coupled plasma mass spectrometry imaging [96]. However, the challenge of lipidomics in high-resolution analysis is even greater. There are no methods to amplify, and a labeling method that can be used for signal enhancement is lacking. In addition, due to its chemical properties, many aqueous phase analysis methods are not suitable for lipid analysis. The most important choice to overcome these mismatches and achieve a technological breakthrough is the advancement of mass spectrometry.

Analysis at the single-cell level is divided into two categories: imaging analysis at single-cell or subcellular resolution and sensitivity, which is subject to the sampling area size, mainly through MALDI or secondary ion mass spectrometry (SIMS). This method is characterized by providing important spatial information. As resolution increases, sub-cellular level imaging analysis can be achieved. The other analysis strategy is to sample and analyze single cells directly using infusion ionization.

The cytosol lipids of cells can be extracted by capillary micro sampling [97]. The advantage of this method is that they can be directly sampled from the tissue surface or cultured adherent cells without cell isolation. The disadvantage of this method is that it does not reflect the actual or even major lipid components of a cell. Because the cytoplasm is not uniform, the extracted cytosol is not complete, and the cell membrane components cannot be analyzed. However, this is an effective method when it is necessary to exclude interference from the cell membrane. There have been many studies on single-cell metabolomics and lipidomics methods for extracting cytosol from cells. These methods include different ionization methods [98–100], optimization of sampling and direct infusion volumes [101–103], and use of SPE for selective extraction [104, 105].

Analysis of intact cells can provide more comprehensive information about the cellular lipidome [106, 107]. However, application of this analytical method requires isolation of adherently-grown cells or analysis of suspended cells. It should be noted that the method of analyzing single cells based on droplets and microfluidic systems has matured in single-cell manipulation. However, online direct infusion mass spectrometry is not optimized for lipid analysis because the design of these microfluidic methods is based on aqueous phase analysis. Currently a large number of compromise strategies are the separation of cells using a microfluidic system followed by offline detection using MALDI MS [108, 109].

Currently, relative quantitative analysis of direct infusion single-cell lipidomics is based on a comparison of normalized mass spectral signal intensities [103, 107–111]. This relative comparison cannot be consistent due to the ionization environment, and the precision and reliability of the quantification are affected to some extent. Achieving absolute quantification of single-cell lipidome analysis or quantification via a stable reference is an important direction that needs to be addressed in the future.

MS imaging analysis is an important approach for lipidomic analysis at the single-cell and subcellular levels. For high-resolution MALDI imaging, there are two configurations to perform laser adsorption and ionization: the transmission (Figure 2A) and reflection (Figure 2B, C) modes. Due to the characteristics of the MALDI instrument, the distance from the exit objective lens to the sample is usually a few centimeters long. In the ordinary optical design, it is difficult to focus below 10 μm in the reflective mode. To improve resolution (to decrease the laser spot size), researchers have proposed an improved configuration based on the microscope design, which uses a transmission mode to allow the laser to be incident from the back of the sample to focus on the surface of the sample section [112]. This is a brilliant and simple method that draws directly from proven technologies. This design allows the lens to be placed very close to the sample for single-cell level imaging. This transmission-mode geometry optical design makes it easier to achieve subcellular resolution. However, the transmission mode has high requirements for sample uniformity. Due to the texture of the sample and influences of crystallization of the matrix, the effective intensity of the laser reaching the surface after passing through the sample varies. This inhomogeneity of light intensity affects the ionization efficiency [112].

Figure 2.

Schematic diagram of MALDI MS imaging configuration. A, Transmission mode. B, Reflection mode. C, SMALDI.

Most high-resolution MS imaging systems use the reflection mode. In commercially available products, the resolution for tissue imaging has reached 10 μm. In theory, it reaches the threshold for single-cell resolution. In fact, although the diameter of cells in tissue is mostly in the range of 10–20 μm, an MSI resolution below 5 μm is similar to a single-cell resolution level due to the heterogeneity of cell morphology. To improve the resolution in the reflection mode, the resolution of the 5 μm level can be achieved by expanding the laser spot diameter before focusing, improving the spot shape of the light source, and optimizing the lens-to-sample distance [113]. Due to inherent characteristics of optical systems, efforts to further reduce the spot in the reflective mode require more complex optical designs and are very difficult. It is well known that reducing the distance from the lens to the sample surface can easily reduce the size of the laser spot. To enable the lens to be placed close to the sample for high-resolution imaging while avoiding disadvantages of the transmission mode, an improved reflection mode was created by generating a hollow lens (Figure 2C). In this way, the sample can fly along the lens axil through a hole in the reflection mode, and the laser can focus on the sample surface through the peripheral annular area of the lens. This method is called scanning microprobe matrix-assisted laser desorption ionization (SMALDI) [114, 115]. Recently, the best resolution of mass spectrometry imaging at 1.4 μm using this mode has been reached [116]. At this resolution power, changes in lipids during mitosis can be observed. The autofocus system is an integral part of the system because of the depth of field decreases dramatically. The lens of SMALDI can be very close to the sample to achieve high-resolution imaging; however, the splashed sample can also cause contamination of the lens, affecting the stability for long-term acquisition. In addition, since the central portion of the lens with the best optical performance cannot be utilized, and the shape of the spot (laser energy profile) is not optimized. Yet, overall, this design can well balance the effects of various constraints. By further improving the design of the lens system, it is expected that the resolution will be enhanced to less than 1 μm in the near future.

For the reflection mode, it may be difficult to reduce the spatial resolution to below 400–500 nm for lipid MS imaging. At the 400–500 nm resolution level, the amount of sample that can be utilized is only 1/10 of the current best resolution. To improve the sensitivity, mainly by increasing the ionization efficiency, it would be effective to use a laser with a higher energy density, a shorter pulse, or a method of introducing post-ionization. We are optimistic about achieving resolutions at the near 400–500 nm level, as the major technical obstacles have been resolved in recent years. This is because focusing on this level can be achieved by expanding the laser beam diameter before objective lens focusing [113] and shortening the distance from the objective lens to the sample. The matrix sublimation deposition technique also has no obstacles at this level. The morphology of the tissue section was also stable at atmospheric pressure, and the deformation of the sample was also minimized [117]. The main problem with MALDI at atmospheric pressure is a decrease in sensitivity. The integration of other technologies, such as laser-induced post-ionization technique, can hopefully achieve imaging analysis at the 400–500 nm resolution. Therefore, the SMALDI configuration is likely to be the most accurate MS imaging technology that is able to match optical microscope imaging. It is highly possible that for quantitative lipid imaging, the internal standards can be quantitatively deposited using a pneumatic spray or electrospray-assisted pneumatic spray, and the resolution level around1 μm can be achieved.

In theory, the optical resolution can be greatly improved in the transmission mode, especially under an ultraviolet excitation source, and the Abbe limit can be as low as about 150 nm. However, considering the problems mentioned above, the non-uniformity of the transmitted laser beam must be solved to ensure the stability and reliability of the signal. The technical complexity of achieving the requirements can be very high because crystallization of the matrix can affect laser focusing. In addition, a sophisticated and complex adaptive optics system is required. It is necessary to equip with a more accurate autofocus system, automatic energy compensation system, and short pulse laser. In addition, matrix deposition at a nanometer scale is also a technical problem. Although the matrix sublimation method achieved the best-known minimum crystal size, there are no effective approaches to achieve quantitative sublimation of internal standards. Therefore, it is difficult to perform accurate quantitative MS imaging.

Sensitivity is a central constraint factor for accessing rich information from each pixel. As spatial resolution increases, the amount of limitedly available sample in each pixel leads to insufficient sensitivity. In addition to the quantification challenge, it is increasingly difficult to perform identification by MS/MS in situ.

In the MALDI process, the ionization efficiency is less than 1 out of 1000 desorbed molecules [118]. The sensitivity can be greatly improved by increasing the ionization efficiency. In addition to the development of new matrices and increased the sensitivity of a detector, increasing the energy profile of the laser is a direct way, such as changing the tunable wavelength and improved spot energy distribution. Derivatization is another way to increase the sensitivity. This method works well for targeted analysis. However, derivatization at the single-cell resolution level is much more difficult to implement than the tissue level.

From the perspective of the instrumentation, universal improvement of the ionization efficiency by post-ionization is a fundamental way for single-cell level analysis. An improved post-ionization method (called MALDI-2) based on MALDI imaging was designed by Soltwisch and co-workers (Figure 3A) [119]. A sample plume produced by the first MALDI process was post-ionized by a second laser beam to generate secondary MALDI-like ionization processes in the gas phase. At a 5-μm resolution, the signal increased by two orders of magnitude. A large number of neutral molecules in the gas plume are ionized. The ionization efficiency of low-abundance species, which are suppressed by high-abundance species, is greatly enhanced. This increase in ionization efficiency also increases the dynamic range. The application of laser post-ionization technology in MALDI imaging has proven that this idea is one of the most potential and scalable technologies for single-cell level MS imaging. We expect the combination of SMALDI and MALDI-2 to archive accurate mass spectrometry imaging of lipids at subcellular resolution.

Figure 3.

Schematic diagram of laser-induced post-ionization. A, MALDI MS imaging with laser-induced post-ionization. B, ESI-based direct infusion MS with laser-induced post-ionization.

This laser-induced post-ionization strategy can be used not only for MALDI imaging, but also for ESI-based single-cell analysis (Figure 3B) [99]. If the ionization efficiency is high enough, it can improve sensitivity and reduce the ion suppression effect, which is more conducive for quantitative analysis.

Unlike MALDI and DESI, we believe that the SIMS method is very suitable for providing surface information because there is no extraction process. The gas cluster ion beam (GCIB) technology using the super heavy ion beam in the SIMS can achieve minimal molecular fragments, while the sensitivity is also greatly reduced [120]. If the ion beam energy is increased, the resolution can be increased accordingly; however, the fragmentation becomes severe. So although the idealized resolution can be higher, the current practical resolution is 2 μm [121]. The main development direction of this method is to generate ion beams with a larger cluster size. If a significant breakthrough is made in the development of ion beams, this approach may provide a wealth of meaningful information for lipidomic surface imaging, even including three-dimensional imaging. The advantage of nanoSIMS is that it can achieve a 200-nm resolution [122]; however, valid qualitative information is limited. As such, it is difficult to expect that this technique will have a wide range of applications in lipidomics.

In lipidomic analysis at the single-cell level, direct infusion-based analysis and imaging-based analysis complement each other. At this level of analysis, no matter which method is used, the sensitivity becomes the most important constraint factor. For both electrospray and MALDI ionization processes, there are 2–3 orders of magnitude of signal improvement in ionization efficiency to be achieved. As the performance of ion transmission and detectors increases, detection sensitivity will increase further. Table 2 provides some comparisons and summaries of related technologies and challenges.

3. Perspectives on shotgun lipidomics at temporal levels

The difficulty of analyzing dynamic changes of lipids on a time resolution is far greater than analysis on spatial resolution. In circadian physiology, such as the effects of circadian rhythm changes on metabolism or in stress response studies, such as anesthesia resuscitation or inflammatory response, there have been some studies on lipid changes at different time points, and many interesting discoveries have been revealed [123–128]. These studies are all in hours or days as a time unit. By reducing the sampling interval, temporal resolution can be increased, but the temporal resolution that can be improved is very limited. Stable isotope labeling methods are often used for comparative/quantitative analysis. For cultured cell samples, stable isotope labeling can be used to track the flux of lipid metabolism. Some representative methods, such as X¹³CMS or lipidome isotope labeling of yeast (LILY) [129, 130], have demonstrated that lipid labeling can be efficiently quantified. The stable isotope labeling method can also be used for labeling in small animals. There is little research in applying isotopic labeling for high temporal resolution lipidomics. We believe that lipidomic analysis with higher temporal resolution can be obtained by selecting multiple isotopic labels and applying appropriate time or transient labeling.

Due to the fluidity of the membrane, even if the time interval of sample collection is shortened or stable isotope labeling is used, dynamic information of membrane lipids may change due to a long duration of removal from a physiological environment. Ultra-fast and low-temperature freezing can reduce this effect. The SIMS technology is a surface analysis method that is compatible with cryo-freeze during sample preparation [131]. Therefore, it is possible to study dynamic changes of the cellular lipidome by this means, and simultaneously, high-resolution imaging can be achieved. It is known that PE is usually enriched on the inner leaflet of the cell membrane, whereas in mitotic cytokinesis, PE is distributed on the outer leaflet in the cleave furrow [132]. Similar to such dynamic changes of PE during mitosis, the analysis of time snapshots obtained by SIMS cell surface scanning may reveal many fundamental discoveries that are yet unknown.

Since mass spectrometry requires substantial ‘extraction’ of molecules from the sample, the challenges of using mass spectrometry to analyze dynamic lipidome changes at high temporal resolution far exceeds the challenge in spatial resolution. Optical-based analysis methods have significant advantages in high time-resolved dynamic analysis and real-time analysis in vivo. Fluorescent probe labeling is commonly used in optical imaging methods[133]. The advantage of this type of approach is high sensitivity. The disadvantage is that the structure of the fluorescent probe is inconsistent with the natural structure of the lipid molecule. Since the lipid molecules themselves are relatively small, this type of probe may have a greater negative effect on the results than the protein probe. Therefore, real-time imaging methods based on Raman spectroscopy have been developed in recent years[134]. This method overcomes the shortcomings of fluorescent probes. The insufficiency of Raman imaging is that the sensitivity needs to be improved. Currently, it is usually used to observe specific cells type rich in lipids. Although optical-based methods can achieve high temporal resolution, the specificity is limited. The result often displays the overall functioning of a class of lipids. It is difficult to distinguish specific lipid species.

4. Conclusion

Mass spectrometric analysis of lipids has evolved for nearly 40 years. Shotgun lipidomics has also exceeded 15 years of research history. During the development process, the scope and methods of shotgun lipidomics have undergone profound changes. The scope of shotgun lipidomics has also expanded from massive tissue sample level to micro- or even single-cell levels. The analysis ranges from qualitative and quantitative determinations of lipid molecular species expended to position determinations of unsaturated bonds and isomer characterization and quantification. Data dimensions of the analysis extend from mixtures to spatial distribution of lipid molecules and dynamic change characterization. The expansion of these research areas has been driven by advances in a range of instruments and analytical methods. Quantitative analysis is a major advantage of shotgun lipidomics, including its increased sensitivity as a prerequisite. Increasing ionization efficiency or selective ionization improves sensitivity and the dynamic detection range, which is crucial for shotgun lipidomics.