One of the current major challenges in lipidomics is the difficulty to separate and differentiate isomeric and isobaric species due to the immense structural variability in head groups, acyl chains, number and location of C = C bonds (cis/trans), and regioisomerism (sn-positions), which inhibit the delineation and assignment of their biological roles. The specific functions of the isomer have remained largely unknown due to these challenges [6]. In the case of GSL, the isomer problem is multiplied as many glycans have the same formula (e.g., Glc vs. Gal) as well as there can be distinct linkages in the oligosaccharide chain also with the option for either α- or β-glycosidic bonds, which further complicate GSL analysis [6, 7]. For instance, GlcCer and GalCer have different biological functions with GlcCer required for proper functioning of the epidermis, while GalCer maintains the structure and stability of myelin and the differentiation of oligodendrocytes. GlcCer and GalCer are also accumulated in various lysosomal storage disorders. Accumulation of GlcCer is a typical feature of Gaucher disease, while GalCer is accumulated in Krabbe disease. GlcCer has also been shown to be accumulate in neurodegenerative diseases such as Parkinson’s disease. However, little is still known about their role in tumor progression [8]. In general, a-series gangliosides are known to promote tumor growth while b-series gangliosides may have tumor-suppressive effects. Similarly, antibodies against GD1a and GD1b gangliosides are differentially expressed in various neurological and autoimmune diseases. Differentiation of GSL isomers is therefore crucial because of their distinct biological functions, disease associations, and therapeutic implications [9]. Moreover, the glycosphingolipidome is not only amazingly large, but also expanding with a number of new lipid species. Specifically, GlcCer with less prevalent α linkage rather than β linkage have been recently found [8, 10, 11] together with ceramides either lacking the 1-hydroxyl group [12] or having a fatty acyl attached to the 1-hydroxyl [13] alongside GSL with polyunsaturated very long-chain FA (C26–C36) [14]. Furthermore, humans only synthesize cis (Z) FAs, while trans (E) FAs are not endogenously produced but present in the human body due to dietary intake. They are well known to play an important role in various physiological processes and therefore, the separation of cis/trans isomers is of great interest. The coelution of different lipid subclasses can lead to ion suppression, obscuring the detection of low-abundant lipids [15]. To address these issues, improved separation of lipid subclasses and lipid species together with the ability to distinguish and identify GSL isomers is essential and highly advantageous for the investigation of their physiological role and functions in nature and disease [6, 15].

Several approaches are used to study GSL structures in connection with MS, such as direct infusion (DI), liquid chromatography (LC), supercritical fluid chromatography (SFC), mass spectrometry imaging (MSI), and ion mobility (IM).

DI-MS (also termed shotgun lipidomics) is a technique when lipid extracts are directly introduced into the MS instrument without upfront separation [16]. The molecular characterization of lipid species relies either on the accurate m/z determination in the full scan mode or on the detection of specific fragmentation reactions in MS/MS experiments. In that case, the HRMS instruments are preferred due to their ability to differentiate isomeric and isobaric compounds [5, 17, 18]. Multi-dimensional MS-based shotgun lipidomics (MDMS-SL) allows the separation of many lipid (sub)classes through selective ionization of a certain category of lipids in the ion source (i.e., intra-source separation), even if the lipids are minor [16, 19, 20]. Although electrospray ionization (ESI) and MALDI are by far the most widely used ionization techniques in DI-MS, desorption electrospray ionization (DESI) [21], laser ablation electrospray ionization (LAESI) [22], and matrix-free laser desorption ionization (LDI) [23] have also been successfully applied. The major advantage of shotgun analysis is the reproducibility and relative high-throughput capability, allowing rapid acquisition of the full mass spectrum within seconds while providing similar sensitivity to LC/MS approaches, especially in coupling with nano-ESI [24]. On the contrary, the major drawbacks are the possible carry-over effect and susceptibility to ion suppression due to the presence of other major lipids or polar compounds (e.g., phospholipids, polar metabolites, and salts), which limits the ionization capacity and may even completely suppress the signals of minor or poorly ionizable GSL. Thorough sample preparation is required to ensure the removal of these interfering compounds. More detailed reviews on MS-based shotgun lipidomics can be read elsewhere [25, 26].

Sample preparation for shotgun lipidomics is very straightforward and generally involves simple extraction focusing on efficient total lipid extraction and minimal sample cleanup. The total lipid extract can be obtained commonly by chloroform/methanol-based liquid–liquid extraction such as of that Folch (CHCl3/CH3OH; 2:1, v/v) [27] and Bligh-Dyer (CHCl3/CH3OH; 1:2, v/v) [28]. In these protocols, lipids are partitioned into the lower chloroform phase. Polar solvent, such as methanol, ethanol, or ispropylalcohol, is used to increase the solubility of the lipids in the organic phase [29, 30]. These biphasic systems are able to recover wide range of lipids; however, sialylated and sulfated GSL or neutral GSL with at least four glycan residues mostly partition to the methanol-rich layer. On the contrary, neutral GSL with less than four glycan residues and other less polar lipids remain rather in the chloroform-rich layer. Thus, these methods do not provide effective recovery of the amphiphilic and highly polar GSL, as they generally require more aqueous portion [29, 31]. Over the years, several modifications and alternative strategies to these original protocols have been developed. One of them is the method described by Matysh et al. [17], which utilizes the mixture of methyl tert-butyl ether (MTBE) and methanol in ratio 10:3. This method was specifically developed for shotgun lipidomics of samples with excessive amounts of biological matrices. Furthermore, single-phase butanol-methanol (BUME) extraction system firstly described by Löfgren et al. (n-butanol/CH3OH; 3:1, v/v) [32, 33] and further modified by Alshehry et al. (n-butanol/CH3OH; 1:1, v/v) [34] has been reported to provide a similar yield of lipids compared to traditional Folch and Bligh-Dyer methods. Over the past years, monophasic extractions, also termed as protein precipitation methods, have gained popularity and have been applied to the simultaneous analysis of polar and non-polar lipids and other metabolites. The one-phase extraction methods are generally faster, cheaper, and less complex compared to conventional two-phase partition systems, and eliminate the risk of losses during transfer between phases; however, they do not allow the removal of polar and ionic impurities, leading to an increased risk of matrix effects and ion suppression. Thus, the application of one-phase extractions should be limited to polar lipid classes or should be followed by a sample cleanup using liquid–liquid extraction (LLE) and/or solid-phase extraction (SPE) [5, 35, 36]. The one-phase extraction is usually achieved through simultaneous protein precipitation with a variety of organic solvents including methanol, ethanol, acetonitrile, acetone, isopropanol, n-butanol, as well as their mixtures [35,36,37,38].

LC/MS is a key, well-established, and powerful analytical method used in lipidomics, which allows lipid subclass and/or lipid species separation when coupled to MS. The most frequently used ionization technique in LC/MS-based lipidomics is ESI, which is best suited for a wide range of lipids, including GSL due to several significant advantages including high sensitivity, easy coupling with chromatographic techniques, and structural details based on the use of tandem mass spectrometry (MS/MS) with high mass accuracy. In contrast, APCI and APPI are valuable alternatives for less polar lipids [5, 39]. The initial lipid extraction for the lipidomic analysis using LC–MS methods is similar to those described in the “DI-MS” section, i.e., using organic solvents. In contrast to sample preparation used for DI-MS, the sample preparation for LC–MS often requires more rigorous sample cleanup, including SPE to further purify the lipid extract by removing impurities and concentrating the lipids [30]. If needed, depletion of highly abundant lipids such as glycerolipids and phospholipids can be performed using alkaline hydrolysis [40] or special ZrO2/TiO2-based SPE method can be employed for removal of phospholipids to allow the analysis of low-abundant lipid species [41]. In addition, SPE or open column chromatography can be used to fractionate the lipid extract into subfractions [40].

MSI has become a popular and powerful method perfectly designed for the analysis of solid samples (e.g., tissues) with the ability to simultaneously display both spatial distribution and molecular level information. The most frequently used ionization technique employed in MSI is MALDI, but other ionization techniques could be employed as well, e.g., DESI, LAESI, or secondary ion mass spectrometry (SIMS) [42, 43]. The major advantages of MALDI are minimal sample preparation, high tolerance to salts, and the ability to relatively easily ionize heavily glycosylated GSL, but with lower ionization efficiency compared to ESI [29]. MALDI also has a few limitations, such as the inability to resolve isomers without prior separation and generally high background noise and ion suppression effects due to the formation of matrix clusters [29, 44].

The most widely used MSI technology applied for rapid in situ screening and mapping the spatial distribution of individual lipid species in biological samples is MALDI coupled to time-of-flight analyzers (MALDI-TOF). However, the comprehensive analysis is limited since the MSI is largely based on the qualitative comparison of healthy and diseased samples [45,46,47]. The coupling of MSI with Orbitrap or ion cyclotron resonance has provided deeper insight into the lipidomic complexity of biological samples [48], such as the application of MALDI-Orbitrap using MS/MS spectra to facilitate structural elucidation of even highly complex sulfo-GSL with up to five hexose moieties [49]. MSI techniques generally require minimal sample preparation. In MALDI, the tissue samples are first cryodissected into slices (~ µm), placed on a target surface, co-crystallized and immobilized with a suitable matrix, and then irradiated by laser to produce ions [29]. Common matrices used for GSL analysis include, for example, 2,5-dihydroxybenzoic acid, 1,5-diaminonaphtalene, 4-hydrazinobenzoic acid, 6-aza-2-thiothymine, 6,7-hydroxycoumarin (esculetin), and α-cyano-4-hydroxycinnamic acid [29, 50]. MALDI matrices used in lipidomics were also well-reviewed by Leopold et al. [51].

In recent years, IM has appealed as a suitable technique for the separation of lipid isomers. However, due to the current resolving power limitations, lipid isomers cannot be fully resolved by IM alone in complex mixtures [6], which led to the interfacing of IM with LC/MS with great potential for the separation of lipid isomers together with increased selectivity and sensitivity [52].

Wojcik et al. [6] utilized ultrahigh-resolution IM separation with traveling waves in a serpentine and extended multipass SLIM platform for selected lipid and glycolipid isomers. They achieved partial separation of GlcSph 18:1;O2 vs. GalSph 18:1;O2 and GlcCer 18:1;O2/18:0 vs. GalCer 18:1;O2/18:0, differing only in the identity of glycan, which was achieved after four passes (~ 60 m path). Moreover, the baseline separation of GD1a and GD1b gangliosides, which only differ in the location of sialic acid residues, has been accomplished even with a minimal possible path of 1.25 m (i.e., without using multipass separation). However, the major issue is the limited number of passes due to increasing peak widths with the increasing number of passes, which reduces the detection window and range of mobilities. May et al. [53] also resolved GD1a and GD1b gangliosides in a standard mixture as doubly sodiated species [M + 2Na]2+ along with two pentasaccharide GSL differing in the location and linkage of fucose. Djambazova et al. [54] have reported a partial separation of GD1a and GD1b isomers with 36:1;O2 and 38:1;O2 ceramide in tissue samples using MALDI-TIMS. Xu et al. [55] have shown an effective resolution of GlcCer and GalCer species from human plasma and cerebrospinal fluids using differential mobility spectrometry coupled to LC/ESI–MS. Sample preparation for IM is similar to DI-MS or LC–MS.

In summary, although the separation and characterization of lipid isomers still remain very challenging and only a limited number of lipidomic studies have been carried out to differentiate GSL isomers using IM technologies, recent advances in chromatography and IM through novel instrumental developments have pushed the popularity of IM forward by enhancing its resolution and sensitivity. The coupling of IM with LC/MS is thus likely to become a very valuable tool capable of efficiently separating and reliably distinguishing various GSL isomers. However, further progress is still needed as potential applications of IS are still being discovered [56].

The MS/MS analysis of GSL relies on various dissociation techniques. Each dissociation technique provides a distinct level of structural information since it cleaves bonds at different locations of the molecule. Their combination can provide complementary structural details of GSL [29].

Biological functions of GSL, as well as other lipids, highly depend on their varying expression levels and structural diversity, including carbon–carbon DB locations, cis/trans isomerism, and the sn-position of the fatty acyl chain(s), which complicate the structural elucidation. The MS/MS analysis is essential in the structural elucidation of GSL. The systematic nomenclature of MS/MS fragments for the carbohydrate part of GSL was proposed by Domon and Costello [57] and includes fragments containing the non-reducing end (i.e., A, B, and C) and the reducing end (i.e., X, Y, and Z). Fragments B, C, Y, and Z correspond to glycosidic cleavages that determine the glycan sequence (Fig. 2A), while A and X fragments are cross-ring cleavages allowing the differentiation of linkage positions (Fig. 2B). Since A, B, and C ions do not include ceramide structure, unlike their X, Y, and Z counterparts, their masses are not affected by the lipid moiety. The nomenclature was later updated by Ann and Adams [58] in order to include more detailed ceramide fragments. The major fragments are shown in Fig. 2C, where the NI and NII fragments are diagnostic ions of the long-chain base of the ceramide moiety. It implies that the fragmentation pathways of GSL can be predicted and structural databases can be constructed in silico to identify GSL by matching MS/MS spectra with the database, which is an indisputable advantage in the structural analysis of GSL.

Fragmentation patterns of GSL (adopted from [57, 58, 123]); Hex refers to hexose

Neutral GSL are relatively poorly ionized in the negative ion mode due to their basic (i.e., amino glycan-containing) and acidic counterparts [59]; thus, neutral GSL are commonly analyzed in the positive ion mode, where GSL are better ionized and provide abundant Y/Z-pair ion series indicative of sequence information accompanied by less common B/C-type fragments. A/X-type ions involving C–C bond usually require higher energies [60].

Collision-mediated dissociation is a conventional dissociation technique employed for MS/MS experiments. Both low-energy collision-induced dissociation (CID) and higher-energy collisional dissociation (HCD) provide structural information for elucidating the glycan sequence in the positive ion mode (i.e., sequential cleavage of a monosaccharide unit represented by a series of Z/Y-type ions) or in the negative ion mode (i.e., B- and C-type ions) and ceramide moiety of GSL. An alternative implemented exclusively for ion trap MS is pulsed Q dissociation (PQD), which deposits higher energies on the ions compared to CID and allows the observation of low m/z fragments that are usually excluded from CID, however at the cost of reduced fragmentation efficiency [29, 61, 62].

Incremental Δm/z indicates the loss of a hexose (Δm/z 162) and N-acetlyhexosamine (Δm/z 204) [29]. In addition, the ceramide composition, respectively, sphingoid bases, can be identified from the specific fragment ions in the positive mode (Table 1) [62].