Ganglioside Isomer Analysis Using Ion Polarity Switching Liquid Chromatography-Tandem Mass Spectrometry

Abstract

Gangliosides are ubiquitously present on cell surface. They are more abundantly expressed in nerve cells and tissues and involved in pathology of various diseases. Diversity of molecular structures in the carbohydrate head group, fatty acyl and long chain base increases the complexity of analyzing gangliosides. In this study, an ultrahigh-performance liquid chromatography-tandem mass spectrometry method is developed for analysis of the co-eluting ganglioside isomers, which uses ion polarity switching to integrate glycan head isomer identification, ceramide isomer differentiation, and quantification of ganglioside into one analysis. The method is facilitated with an extensive ganglioside target list by combining the various glycan headgroups, long chain bases and the experimentally determined fatty acyls. Correlation between the retention time of ganglioside and its ceramide total carbon number is experimentally validated and used to predict retention time of ganglioside target list for scheduling the final multiple reaction monitoring method. This method was validated according to the FDA guidelines: 96.5% of gangliosides with good accuracy (80–120%), precision (< 15%) and linearity R² > 0.99. The authenticated gangliosides were quantified from mouse brain by isotope dilution. Overall, 165 gangliosides were quantified using 10 mg mouse brain tissue, including 100 isomers of GM1, GM2, GM3, GD1a, GD1b, GD2, GD3, and GT1b.

Introduction

Gangliosides, a type of glycosphingolipids, are formed by linking a ceramide backbone to a glycan headgroup containing at least one sialic acid moiety. Gangliosides have been ubiquitously found on cell surface, tissues, and more abundantly expressed in the nervous system and are involved in the pathology of various diseases (1–4). In general, the biological function of gangliosides is influenced by both the structures of glycan and the ceramide moieties composed of fatty acyl and the sphingosine long chain base, thus analyzing gangliosides in their intact form is more amenable to understand the important roles they play in different physiological processes (5–8). However, diverse compositions in each building unit of the gangliosides, i.e. glycan, long chain base and fatty acyl can greatly increase the structural complexity and renders it challenging to distinguish the various ganglioside isomers (9–13).

Traditionally, thin-layer chromatography (TLC) coupled with chemical detection or immunochemical binding assays or carbohydrate recognition reagents has been used to differentiate intact gangliosides based on their glycan headgroups (14, 15). However, due to its low resolution, sensitivity, and deficiency of structural information on the ceramide backbone, TLC is inadequate for the identification of various ganglioside molecular species, especially for the trace gangliosides and structurally similar isomers, from complex matrixes.

Mass spectrometry (MS) coupled to different separation techniques have been increasing used for analysis of gangliosides. Gas chromatography−mass spectrometry (GC−MS) analysis involves cleavage of the fatty acyl chain or long chain base from the glycan headgroup; while this can provide accurate information of the individual building units, the whole structural information of the intact ganglioside is lost (16, 17). Under negative ionization mode, matrix-assisted laser desorption/ionization (MALDI) (18–20) or electrospray ionization (ESI) (21–23) has been successfully used in identification of sialic acid-containing glycosphingolipids. However, it is difficult to separate all ganglioside isomers with different glycans or ceramides moieties simultaneously even coupling with the reversed-phase (RP) or hydrophilic interaction chromatography (HILIC) columns (21–24). For example, GD1a and GD1b are not separated using RPLC column (21, 24); although HILIC was introduced to distinguish the isomers with different glycan headgroups, it cannot differentiate isomers with different ceramide moieties (22, 24).

Unique fragment ions have been used for identification and accurate quantification of isomeric metabolites (21, 25, 26). But to date, only a few studies have utilized this strategy for large scale quantification and distinction of isomeric gangliosides (21, 23, 27). On the basis of the characteristic product ions of the glycan headgroup, Ikeda et al. successfully separated and distinguished the regio-isomeric gangliosides like GM1a/GM1b, GD1a/GD1b/GD1c (21). However, the coeluted gangliosides isomeric in the ceramide moiety were indistinguishable without further MS3 fragmentation of the sphingosine base (21). On the other hand, Meng et al. differentiated two types of isomers with different ceramide moieties or glycan headgroups based on their diagnostic fragment ions in positive ionization mode (27).

For comprehensive untargeted profiling of lipids, authentication of identified lipids can be facilitated by the relationship between the total carbon number and the retention time on a reversed phase column (22, 28). Similar strategy was applied by Hu et al. where the authors used a mathematical model to predict the retention times of all theoretically possible acidic glycosphingolipids (23). In current work, an ion polarity switching ultrahigh performance liquid chromatography (UHPLC)-ESI-MS/MS method was developed for large-scale quantification of gangliosides in multiple reaction monitoring (MRM) mode. In one single run, the unique fragment ions generated from the isomeric ceramide moiety of the gangliosides were monitored in positive ion mode and the characteristic product ions used for quantification and differentiation of isomers within different glycan headgroups were acquired in negative ion mode. To shorten the list of potential ganglioside molecular species included in the MRM targets, GC-MS was applied to obtain fatty acyl composition of the gangliosides in the sample (Fig. 1). In addition, several sets of correlation curves were generated using gangliosides standards; the established dependence of retention time (RT) on the total carbon numbers or double bonds of the ceramide backbone was used to predict RTs of gangliosides in the MRM list. In the end, by adding isotope labeled ganglioside standards with known concentration, we absolutely quantified 165 ganglioside species in 10 mg mouse brain with 100 isomers separately quantified.

Figure 1.

Workflow for quantification of ganglioside isomers using UHPLC-MS/MS in MRM mode.

Experimental section

Chemical and Standards

LC-MS or HPLC grade acetonitrile (ACN), methanol (MeOH), water (H₂O), chloroform (TCM), isopropanol (IPA), and formic acid (FA) were obtained from Fisher Scientific (Waltham, MA). MS grade ammonium formate (NH₄COOH) was purchased from Sigma-Aldrich (St. Louis, MO). The gangliosides standards, GM1a, GM2, GM3, GD1a, GD1b, GD2, GD3, and GT1b containing various fatty acyl or long chain bases were purchased from Matreya, LLC (State College, PA). Isotope labeled internal standard (IS) d3-GM3 (d18:1/18:0) was obtained from Cayman Chemical (Ann Arbor, MI).

Standard Solution and Sample Preparation

Stock solutions of ganglioside standards and IS were prepared using TCM/MeOH/ACN (2:1:0.1, v/v/v) in the concentration of 1 μg/μL. All standard solutions used in experiment were diluted using IPA/H₂O/ACN (45:28:27, v/v/v).

Mouse brain tissue in this study was obtained from BioIVT (Westbuty, NY). Total gangliosides were extracted from mouse brain as previously described (29), with minor modification. Briefly, 100 mg mouse brain (BALB/cJ strain) was weighed and to which 750 μL H₂O with 10 ng IS was added and the tissue was homogenized using Precellys Evolution homogenizer (Bertin Technologies) at 6,800 rpm for 30 sec in 2 cycles. Next, to the homogenate 2 mL MeOH and 1 mL TCM was added to extract the total lipids in tissues. After sonicated for 5 min, the mixture was centrifuged at 600 g for 15 min at room temperature. The supernatant was collected and mixed with 650 μL of H₂O to give a final TCM/MeOH/H₂O ratio of 1:2:1.4 (v/v/v). The mixture was centrifuged once again under the same conditions and then the upper layer was collected and dried under a stream of nitrogen gas. The dried extracts were reconstituted using 100 μL IPA/H₂O/ACN (45:28:27, v/v/v) to 1 mg/μL (tissue weight/solvent) and centrifuged at 20,000 g for 10 min at 4 °C before the LC-MS analysis (N=3).

To analyze the fatty acyl composition of gangliosides in commercial ganglioside standards or mouse brain, the fatty acyl chains were cleaved from ceramide backbone of gangliosides using a direct transesterification method with some modifications (30). Briefly, 20 μg ganglioside standards or ganglioside extract from 20 mg mouse brain were dried in borosilicate glass vial and reconstituted in 200 μL methanol/hexane (4:1, v/v), 30 μL acetyl chloride was then slowly added and then the glass vial was sealed tightly and incubated for 1 h at 90 °C with shaking. After glass vial cooled to room temperature, 100 μL of 2 M Na₂CO₃ solution was slowly added to stop the reaction. To extract the generated FAMEs, 100 μL hexane was added and the upper layer was collected for GC-MS analysis.

GC-MS Analysis

An Agilent 7890 GC system (Agilent Technologies) coupled with a Leco Pegasus HT time-of-flight MS (Leco, St. Joseph, MI) was used to analyze the composition of fatty acyls present in gangliosides. A HP-88 column (100 m × 0.25 mm) with a film thickness of 0.2 μm (Agilent Technologies) was utilized for separation of FAMEs. The injection volume of sample was 1 μL using splitless injection mode. The temperature of GC-MS inlet was 250 °C and the oven temperature was kept at 100 °C for 3 min and then increased to 175 °C at rate of 8 °C/min, followed by rate of 3 °C/min up to 240 °C and maintained isothermally for 10 min. The carrier gas was Helium and the flow rate was 2 mL/min. The EI source was set to a potential of 70 eV and the temperature of source and transfer line were 250 °C. Commercial FAMEs standard mixture (Agilent) was used as reference standards.

LC-MS Analysis

A Vanquish UHPLC system (Thermo Scientific) was connected to a Q Exactive HF (Thermo Scientific) mass spectrometer for ganglioside standard analysis to identify all ganglioside species and then coupled with a Quantiva triple quadrupole mass spectrometer (Thermo Scientific) for quantification of all ganglioside standards or extracts from tissue samples.

LC separation was carried out on a Cortecs C18 column (2.6 mm i.d. × 100 mm, 1.6 μm) (Waters). The mobile phase A was ACN/H₂O (60:40, v/v); mobile phase B was IPA/ACN (90:10, v/v); and both of them contained 10 mM NH₄COOH and 0.1% FA. The linear gradient used was as follows: 30% B (0 min), 50% B (1 min), 70% B (7 min), 99% B (13 min), 30% B (13.1 min); and the column was equilibrated for an additional 2 min at 30% B before next injection. The column was maintained at 40 °C and the flow rate was 350 μL/min. The sample injection volume was 10.0 μL.

The following acquisition parameters for the QE HF mass spectrometer were set under both full scan and parallel reaction monitoring (PRM) mode: spray voltage 3.0 kV, vaporizer temperature 400 °C, sheath gas 20 au, auxiliary gas 5 au, sweep gas 1 au, ion transfer capillary temperature 350 °C, S-lens of voltage 50 V, and resolution 120,000. For the HCD fragmentation: maximum injection time 100 ms, automatic gain control (AGC) 1 × 10⁵, and resolution of 15,000. For the quantification of gangliosides using Quantiva in MRM mode with polarity switching, the parameters were: sheath gas 20 Arb, aux gas 7 Arb, sweep gas 1 Arb, ion transfer tube and vaporizer temperature 300 °C. The ion source was operated using heated ESI with ion spray voltage set at 3000 V in negative ion mode and 3500 V in positive mode. Collision energy of each ganglioside standard was optimized via direct infusion. The cycle time was 0.8 sec, the resolution of Q1 and Q3 was 0.7 FWHM (for the ions of d20:1, the Q3 was 0.2 FWHM), the CID gas pressures was 1.5 mTorr.

Method Validation

According to the FDA draft guidelines on validation of bioanalytical methods (31), the linearity, lower limit of quantification (LLOQ), and lower limit of detection (LLOD) of method were evaluated by diluting the standard mixtures of 8 ganglioside standards (GM1a, GM2, GM3, GD1a, GD1b, GD2, GD3, and GT1b) at least 16 times with a 2x dilution factor. IS was added into the dilution solvent at a final concentration of 0.1 ng/μL. The LLOD was defined as the signal to noise ratio (S/N) greater than 3. The LLOQ was defined as the S/N greater than 10 with acceptable accuracy (± 20%) and precision (<20%). The precision and accuracy of method were validated using the same standard mixtures of 8 ganglioside standards at low (LQC), middle (MQC), and high (HQC) concentration covering analytical ranges (N=3).

Data Analysis

GC-MS chromatograms were deconvoluted using ChromaToF (v4.51.6.0) and FAMEs were identified by MS search (v2.0) with an in-house FAME library. Raw data files acquired by LC-MS were analyzed using both the “Transition List Module” in Skyline software (v19.1.0.193) (32) and XCalibur 2.2 (Thermo Scientific). The calibration curves of each analytical standard were constructed by normalizing to the added IS. For the correlation curves between the total carbon number or double bond of ceramide backbone and retention time of gangliosides, high-resolution MS and MS/MS data obtained from QE HF were used to identify all ganglioside species.

Results and Discussion

Fatty acyl analysis to refine the list of potential ganglioside species

Considering the large variety of ganglioside structures in organisms related to the developmental stage and the type of cells or tissues (10, 11, 22, 33), and the limited number of MRM transitions within the defined cycle time, it is necessary to refine the theoretically complied list of gangliosides to those ones most likely exist in the sample. While the composition of glycan headgroup and the ceramide long chain base has less variability, according to the previous literature, the composition of fatty acyls in the ceramide moiety is more flexible in various cells or tissues (16, 34–37). Therefore, knowing the fatty acyl composition of gangliosides would greatly reduce the analytical burden and reduce the large number of targets to be monitored in the MRM method.

We employed a direct transesterification method (30) to get a glimpse of the fatty acyls in the gangliosides and optimized the method to fit with reaction volume of less than 300 μL, using commercial GT1b standard purified from bovine. The reaction conditions optimized including the concentration of acetyl chloride, reaction temperature, and reaction time (Fig. S1A, B, C). According to the intensity of several FAMEs generated from the ganglioside standard, the final condition for cleaving the amido bond between fatty acyl chain and long chain base was fine-tuned to 30 μL of acetyl chloride, 90 °C of reaction temperature, and 1 h of reaction time. Then, we analyzed the composition of fatty acids in GT1b standards using GC-MS and found C18:0 accounted for 84.1% of the total fatty acyls followed by C20:0 with 5.2% (Fig. S1D), which is consistent with the fatty acid composition of GT1b provided by Matreya. To make sure that all fatty acyls are included, intact GT1b standard was analyzed in an untargeted way using LC-MS on a high-resolution mass spectrometer, namely full scan MS data of GT1b standards were acquired under negative ion mode and were processed with mass error < 5 ppm to extract all potential GT1b species and their respective RTs. Then, MS/MS of all potential ganglioside species were acquired using PRM under both negative and positive ion modes. Annotation of the different GT1b molecular species were based on the glycan specific ions in negative ion mode and the long chain base information in positive ion mode (21, 23). Thorough checking at MS and MS/MS levels resulted in a list containing 40 authenticated GT1b species (Table S1). Fatty acyl composition from summing of the fatty acyls in each molecular species matched well with that of the GC-MS data (Fig. S1D), which justifies using GC-MS-based fatty acyl analysis to provide a realistic list of potential ganglioside species, and this result also implies that composition of LC solvent used in the gradient has minimal effect on the ionization efficiency of ganglioside species containing different fatty acyl chain length.

Correlations between RT and total carbon or double-bond number of ceramide backbone within each subclass of gangliosides

To allow more MRM transitions to be monitored in a single LC run while maintaining maximized dwell time and optimized cycle time, the MRM transitions need to be scheduled at a short retention time window. This requires knowing the retention time of each analyte in the transition list (Fig. 1). We analyzed in-depth the eight gangliosides standards (GM1a, GM2–3, GD1a, GD1b, GD2–3, and GT1b) using high resolution LC-MS. By plotting the retention times of more than 500 molecular species with the total carbon number or double bond number of ceramide backbone, we found excellent correlations with coefficients of second degree polynomial regression larger than 0.97 (Table S2), which is consistent with what commonly observed for other lipid classes (22, 28). An example is shown in Fig. 2 for the GM3 species, at the same degree of unsaturation, longer retention time is linearly correlated to the higher total carbon number in the ceramide backbone; conversely, when the total carbon number is the same, retention time is decreasing with the increase of unsaturation. Considering these relationships would be used to predict the RTs of gangliosides in the large MRM transition list, we evaluated their accuracy and precision using ganglioside standard mixtures (N=6) which contained various number of carbon chains and double bonds in the ceramide backbone. 97.0–105.1 % of accuracy and 0.4–2.0 % of precision were obtained, revealing the excellent reproducibility of the RTs of this LC method and the suitability of these correlations to predict RTs of predicted gangliosides in biological samples.

Figure 2.

Correlations between the RPLC retention time and the total number of carbon (A) or double-bond (B) in the ceramide backbone of GM3 gangliosides. X, the number of carbons in the ceramide moiety when unsaturation is fixed; Y, the number of double-bond when total number of carbon is fixed.

Identification and quantification of ganglioside isomers using their unique fragment ions

For more sensitive MRM analysis, we optimized the conditions of ionization and fragmentation of each subclass of gangliosides using their respective standards. In both negative and positive mode, precursors of GM1–3 and GD3 were detected as singly charged ions; and those of GD1a/b, GD2, and GT1b were detected as doubly charged. Considering all gangliosides generate a sialic acid ion (m/z 290.09) as base peak in negative ion mode, it was used as the quantification ion in MRM transition list for all gangliosides. We also manually tuned the collision energy for each ganglioside subclass in negative mode and the CEs were finalized as: 30 eV for GM3, 35 eV for GD1a/b and GM2, 40 eV for GD2–3 and GT1b, and 55 eV for GM1.

However, many isomers with different glycan headgroup or ceremide share almost same physicochemical properties and they are difficult to be totally separated by RPLC column, thus the selectivity of the MRM transitions is critical to distinguish these co-eluting species. To this end, we acquired MS/MS spectra of isomeric ganglioside species in both negative and positive mode to identify isomer-specific diagnostic product ions (Fig. 3 and Fig. 4). For isomers with different ceramide structure, like the GM1 d18:1/20:0 and GM1 d20:1/18:0, the diagnostic product ions were generated by their specific long chain bases in positive ion mode, i.e. m/z 264.2689 for d18:1, and m/z 292.2997 for d20:1 (Fig. 3A), which was also observed in other types of MS instruments (21, 23, 27). GM1 d18:1/20:0 and GM1 d20:1/18:0 co-eluted at 5.27 min, which cannot be distinguished based on their precursor ions in negative mode (Fig. 3B), however, they can be differentiated with these unique fragment ions derived from the specific long chain bases in positive mode (Fig. 3C, D). As a result, the relative amount of these two GM1 isomers was quantified by dividing the total ion intensity of sialic acid (m/z 290.09) obtained in negative ion mode according to the ratio of these two long chain base ions in positive mode.

Figure 3.

Identification and quantification of isomeric gangliosides with different ceramide moieties. (A) MS/MS of GM1 38:1 (d18:1/20:0 or d20:1/18:0) in positive mode. The structure of each fragment ions is labeled in blue. Unique fragment ions for differentiation of isomers with different ceramide moieties are in red (Cer d18:1-H₂O) and green (Cer d20:1-H₂O). (B) Extracted ion chromatogram (EIC) of the precursor ions (MS1) for both isomers in negative mode (left), and of the isomer specific ions in the MS2 scans for GM1 d18:1/20:0 (middle) and GM1 d20:1/18:0 (right).

Figure 4.

Identification and quantification of isomeric gangliosides with different glycan head groups. (A) MS/MS of GD1a d18:1/18:0 in negative ion mode and its unique fragment ion (red color). (B) MS/MS of GD1b d18:1/18:0 in negative ion mode and its unique fragment ion (green color). All fragment ions are detected as singly charged, except for the precursor ion (red diamond, m/z 917.4780/917.4761, [M-2H]^2-). The unique fragment ions for GD1a/b d18:1/18:0 are labeled by dotted box in the molecular structure. (C) EIC and retention time of the precursor ions of both isomers in the MS1 scan in negative ion mode (left), and those of the isomer specific ions in the MS2 scan, GD1a d18:1/18:0 (middle) and GD1b d18:1/18:0 (right).

For isomers with different glycan headgroup (GD1a and GD1b), the specific terminal structures of the glycan can provide diagnostic product ions in negative ion mode (21). In Fig. 4A and 4B, MS/MS of GD1a had the ion of m/z 655.2203 from NeuAc-Gal-GalNAc and GD1b the ion of m/z 581.1821 from NeuAc-NeuAc. Similarly, the partial co-elution of GD1a and GD1b (Fig. 4C) can be quantitatively resolved based on these unique fragment ions (Fig. 4D, E). Using standards, we further optimized the CE for each subclass of gangliosides to improve the limit of detection for all diagnostic ions.

Method validation using ganglioside species in standard mixtures

To further evaluate the feasibility of our whole strategy for efficient quantification of gangliosides, 20 μg commercial GM2 standard mixture purified from human Tay-Sachs cells was hydrolyzed and transesterified under the optimized conditions. The generated FAMEs were analyzed using GC-MS (Fig. S2). Then, these fatty acyls were combined with 9 different long chain bases (d16:0–2; d18:0–2; d20:0–2) to generate an extensive MRM transition list for monitoring 35 molecular species of GM2 species. The MRM transitions were scheduled according to the RTs predicted from the correlations between RT and total carbon numbers in ceramide. Finally, 28 GM2 molecular species containing 8 isomers were quantified using LC-MRM-MS (Table 1). As a further validation, fatty acyl composition results obtained from GC-MS and from LC-MS by analyzing intact GM2 species are similar for the main fatty acids (Fig. S2), demonstrating the good feasibility of this approach.

Table 1.

The abundant GM2 species and their composition in the commercial GM2 standard purified from human Tay-Sachs cells.

Name	Retention Time (min)	Percentage
GM2 d18:1/18:0	4.55	77.77%
GM2 d20:1/18:0	5.42	8.73%
GM2 d20:0/18:0	5.73	4.08%
GM2 d18:1/20:0	5.42	3.81%
GM2 d18:2/18:0	3.89	1.20%
GM2 d18:1/16:0	3.79	0.71%
GM2 d20:1/17:0	5.00	0.63%
GM2 d18:1/24:1	6.20	0.61%
GM2 d16:1/18:0	3.79	0.35%
GM2 d18:2/20:0	4.69	0.33%
GM2 d18:0/20:0	5.73	0.29%
GM2 d18:1/17:0	4.18	0.27%
GM2 d20:2/18:0	4.69	0.19%
GM2 d18:1/22:0	6.27	0.16%
GM2 d18:0/16:0	4.06	0.11%
GM2 d18:1/18:1	3.89	0.10%
GM2 d20:1/20:0	6.27	0.07%
GM2 d20:0/17:0	5.30	0.05%
GM2 d18:2/24:1	5.47	0.05%
GM2 d18:1/22:1	5.37	0.03%
GM2 d20:1/18:1	4.69	0.03%
GM2 d16:2/20:1	4.41	0.02%
GM2 d20:2/17:0	4.28	0.02%
GM2 d18:2/22:0	5.37	0.02%
GM2 d18:0/18:0	4.88	0.01%
GM2 d18:0/17:0	4.46	0.01%
GM2 d18:2/16:0	3.18	0.01%
GM2 d18:2/20:1	5.24	0.01%

Using this approach, we obtained the molecular species exist in 7 other commercial ganglioside standards (GM1a, GM3, GD1a, GD1b, GD2–3, and GT1b) and further evaluated the linearity, sensitivity, precision, and accuracy of the LC-MRM-MS method. Method evaluation was performed according to the FDA guidelines (31). For the 57 highly abundant ganglioside species in these 8 commercial standards (Table 2 & Table S3), 96.5% of them obtained good accuracy (80–120%) and precision (< 15%) by analyzing the QC standards with low, medium, and high concentrations relative to that used for their linearity range. Moreover, all of their linearity R²> 0.99 and LLODs range from 0.012 to 0.039 ng with LLOQs of 0.081–0.269 ng.

Table 2.

Characteristics of method validation based on data from 57 ganglioside species in the ganglioside standards (N=3). (Note: The average value is shown for each ganglioside classes; the detailed information of all 57 ganglioside species is in Table S3)

Class	Sensitivity/ng		Accuracy %			Precision %
	LOD	LOQ	LQC	MQC	HQC	LQC	MQC	HQC
GM1	0.026	0.088	102.3	99.9	102.7	11.0	7.02	7.6
GM2	0.047	0.155	107.8	101.9	105.7	8.0	4.9	7.1
GM3	0.036	0.123	103.4	100.9	101.6	6.8	6.2	8.3
GD1a	0.032	0.106	102.5	102.5	101.6	7.4	9.0	7.0
GD1b	0.034	0.112	107.1	101.6	104.1	7.0	3.2	7.0
GD2	0.047	0.157	103.8	104.4	102.6	4.1	7.8	6.9
GD3	0.031	0.102	101.1	102.8	100.9	7.2	7.1	7.3
GT1b	0.038	0.125	102.8	100.9	99.0	5.7	5.3	7.4

Quantification of ganglioside species in mouse brain

The validated LC-MRM-MS approach was further applied to quantify the ganglioside extracts from mouse brain tissue. GC-MS was first applied to profile the fatty acyl composition from brain ganglioside extract; 17 fatty acids were detected and quantified (Fig. 5A), in which C18:0, C18:1, C22:6, C20:4, and C16:0 were the top five highly abundant fatty acyls, agreeing with the previous reported results (38); then, these fatty acyls were combined with 9 long chain bases that are frequently identified in various organisms and 8 glycan headgroups including GM1a, GM2–3, GD1a, GD1b, GD2–3, and GT1b (21–23). In total, ~500 potential ganglioside species were obtained with unique molecular formula. These species were monitored with scheduled RTs, quantification ions and multiple diagnostic ions to cover all possible ceramide or glycan head group isomers. 165 ganglioside species out of the ~500 target list were detectable and the revised MRM transition list was used to quantify gangliosides from mouse brain. Absolute quantitation was achieved by adding d3-GM3 d18:1/18:0 at known concentration before the extraction step, in the end, 165 gangliosides containing 100 isomers were absolutely quantified with 84% of them had CV of less than 20% in 6 replicates (Table 3, Fig. S3). It is of note that free fatty acids or other sphingolipids could be extracted together with gangliosides using the current extraction buffer (TCM/MeOH/H₂O, 1:2:1.4, v/v/v), such as C20:4 and C22:6 were present in the GC-MS based fatty acyl composition analysis (Fig. 5A). Although this may exaggerate the number and/or distort the relative composition of the fatty acyls in gangliosides as compared to Fig. 5B, GC-MS-based fatty acyl analysis is primarily for identification of a practical list of potential fatty acyls in gangliosides, quantitation is not the intent of this analysis. Besides, authentication using LC-MS with scheduled RTs and signature quantifying and diagnostic fragment ions can remove those artificial gangliosides from the MRM list.

Figure 5.

Individual component analysis of gangliosides extracted from BALB/cJ mouse brain. (A) Fatty acyl composition obtained using direct transesterification and GC-MS analysis. The fatty acyl (B), long chain base (C) and glycan head group (D) composition obtained by summing individual molecular ganglioside species identified using LC-MS.

Table 3.

Concentrations of gangliosides identified from BALB/cJ mouse brain (fmol/mg brain tissue; mean ± SD).

Ceramide	GM1	GM2	GM3	GD1a	GD1b	GD2	GD3	GT1b
d16:0/20:0	109.7 ± 18.9			2133.9 ± 123.5	598.6 ± 50.7			1602.0 ± 103.3
d16:0/20:1	26.5 ± 5.0			298.7 ± 18.3	97.3 ± 8.6			185.4 ± 13.2
d16:1/18:0	38.5 ± 7.9			22.5 ± 5.0	3.6 ± 1.0
d16:1/20:0	7.0 ± 1.5
d18:0/14:0	14.2 ± 1.1
d18:0/16:0	248.1 ± 31.6		15.1 ± 2.3
d18:0/18:0	4952.9 ± 328.1	123.8 ± 13.8	7.5 ± 1.7	7884.0 ± 394.9	2211.2 ± 152.0		95.7 ± 14.3	2462.6 ± 233.8
d18:0/18:1	44.1 ± 8.4			94.9 ± 15.7	30.8 ± 5.0
d18:0/20:0	163.5 ± 15.4			657.2 ± 60.3	347.5 ± 139.1
d18:1/14:0	62.3 ± 7.3
d18:1/16:0	888.2 ± 83.9	47.4 ± 3.2	2.2 ± 0.3	817.7 ± 34.3	131.0 ± 5.9		20.9 ± 1.4	98.9 ± 5.4
d18:1/16:1	29.6 ± 1.1
d18:1/17:0	140.6 ± 27.2			92.2 ± 3.6	27.8 ± 5.1
d18:1/18:0	17535.4 ± 1567.2	580.7 ± 37.1	1031.3 ± 19.9	23790.9 ± 996.0	7747.3 ± 538.3	75.3 ± 5.2	389.9 ± 35.6	7670.8 ± 683.7
d18:1/18:1	147.9 ± 23.2	3.9 ± 0.6	2.7 ± 0.3	81.9 ± 15.0	32.7 ± 4.9			52.9 ± 5.6
d18:1/20:0	176.7 ± 14.4		9.3 ± 1.9	870.3 ± 80.7	540.6 ± 67.5			148.3 ± 27.5
d18:1/20:1	47.0 ± 9.3		1.5 ± 0.4	131.4 ± 11.8	106.5 ± 14.0			90.1 ± 11.5
d18:1/20:2	4.5 ± 0.7
d18:1/22:1	13.4 ± 2.8			20.4 ± 2.1	21.0 ± 3.2
d18:1/24:1	52.4 ± 9.0		16.0 ± 1.1	68.9 ± 11.1	23.3 ± 5.9		8.9 ± 2.1	16.3 ± 3.2
d18:2/16:0	10.4 ± 1.3			11 ± 0.9	6.9 ± 1.1
d18:2/18:0	2141.7 ± 146.3	45.8 ± 6.7	33.0 ± 2.7	2047.5 ± 88.7	830.0 ± 84.5		10.6 ± 1.8	505.9 ± 31.6
d18:2/19:0	14.3 ± 2.5			17.9 ± 1.5	4.5 ± 0.6
d18:2/20:0	65.9 ± 8.9		2.8 ± 0.3	53.1 ± 8.8	43.3 ± 9.6			9.7 ± 1.4
d18:2/20:1	0.8 ± 0.2			5.3 ± 0.7	2.1 ± 0.5
d18:2/24:1	4.5 ± 0.9			24.9 ± 1.8	7.7 ± 1.0
d20:0/16:0	208.1 ± 35.9	80.5 ± 9.2	1.1 ± 0.2				86.4 ± 16.4	775.0 ± 75.6
d20:0/18:0	317.3 ± 52.7			732.2 ± 94.9	387.1 ± 70.3
d20:1/14:0	32.0 ± 7.2							24.4 ± 2.1
d20:1/14:1	0.7 ± 0.1
d20:1/16:0	540.1 ± 91.3	329.8 ± 30.3	99.8 ± 9.0	1165.0 ± 79.3	378.8 ± 24.2		294.8 ± 43.5	1903.4 ± 124.3
d20:1/16:1	72.5 ± 7.5	28.5 ± 5.6	4.4 ± 0.4	113.1 ± 13.4	45.9 ± 9.3			94.8 ± 8.3
d20:1/18:0	1390.5 ± 131.3	94.1 ± 11.6	91.9 ± 7.9	3053.7 ± 563.3	1937.5 ± 575.2			1041.9 ± 148.9
d20:1/18:1	22.8 ± 5.6			12.9 ± 3.5	10.8 ± 4.7			24.8 ± 4.1
d20:1/22:1	2.6 ± 0.4			7.9 ± 1.5	2.7 ± 0.9		8.3 ± 2.1	8.4 ± 1.5
d20:2/16:0	2.8 ± 0.6						8.3 ± 0.8
d20:2/18:0	105.9 ± 9.6		5.5 ± 1.6	43.2 ± 7.3	34.9 ± 8.1			10.0 ± 2.4

Results of the 165 gangliosides in mouse brain were further analyzed to get the distributions of gangliosides according to their respective fatty acyl chains, long chain bases, and glycan head groups. As shown in Fig. 5B, 12 fatty acyls are higher than 0.1% with C18:0, C20:0, and C16:0 accounting for 96.6% of the total fatty acyls in gangliosides which is consistent with previous reports (27). We also found d18:1, d18:0, d20:1, and d20:0 were the predominant long chain bases, similarly as what reported before (36), and the other long chain bases, such as d16:0 and d18:2 were also present with each of them a little over 5% (Fig. 5C). Besides, in agreement with glycan head group distributions of gangliosides obtained from TLC method (39, 40), GD1a was the most dominant in mouse brain tissue and the total content of GD1a, GD1b, GM1, and GT1b accounted for 94.7% (Fig. 5D). Moreover, GD1a d18:1/18:0 was the most abundant ganglioside molecular species in this study, reaching 23.8 ± 1.0 pmol/mg tissue. Comparing with the concentrations of gangliosides in C57B6/J mouse brain reported by Fujiwara et al. (41), it is consistent that GD1a d18:1/18:0 and GM1 d18:1/18:0 are the two most abundant gangliosides followed by GT1b d18:1/18:0 (Table S4). However, the mouse models used in the two studies are different, therefore it is reasonable to see the abundance of different gangliosides varies to a certain extent. On the other hand, GD2 was the least abundant gangliosides class in the BALB/cJ mouse brain in our study, with GD2 d18:1/18:0 as the only GD2 molecular species identified at 75.3 ± 5.2 fmol/mg mouse brain, which nevertheless could be considered as an evidence for the expression of GD2 in mouse brain tissue (42).

Conclusion

The isomeric ganglioside species with variations either in the ceramide or in the glycan head group portion are very challenging to differentiate and quantify, as their properties are so similar that RPLC is incapable of separating them apart. Comprehensive analysis of gangliosides at the resolution of ceramide isomers typically requires two separate LC-MS/MS runs, one in positive mode to differentiate the long chain base and hence fatty acyl, and the other in negative mode to account for the glycan by monitoring the glycan specific ions. To our knowledge, polarity switching MS has rarely been used in ganglioside analysis, which we have proved to be fully capable of identification and accurate quantification of isomeric gangliosides in one LC-MS/MS run. On the other hand, comprehensive profiling of gangliosides requires monitoring sub thousands of gangliosides species when considering all of the possibilities of fatty acyl, long chain base and glycan head, and even larger number of transitions when isomer specific ions are included in the transition list, this renders it technically unfeasible to monitor all theoretically complied list of gangliosides due to limitations in instrument dwell time and cycle time. We used GC-MS to profile the fatty acyls in the gangliosides, which greatly reduced the number of possible ganglioside species that need to be monitored. The reduced list, when combined with scheduled MRM using retention time predicated and validated according to the correlation with the total number of double bond/unsaturation in the ceramides, enables monitoring of ~500 gangliosides and their respective transitions in a single, short gradient LC-MRM-MS analysis. In conjunction with isotope dilution, this thoroughly validated targeted ganglioside assay was able to quantify 165 gangliosides, which contain 100 isomers from 10 mg mouse brain tissue. It is of note that although we profiled fatty acyls as they are the most variable building unit within gangliosides, we used literature reported long chain bases and the 8 ganglioside subclasses (GM1a, GM2, GM3, GD1a, GD1b, GD2, GD3, and GT1b) when constructing the target list. While it serves the purpose well for a well-studied specimen like mouse brain, analysis of the long chain base composition as well as glycan head group very likely will be needed when analyzing a less well-studied specimen or biological species in order to capture all possible ganglioside species in the sample. However, determining the composition of each structural component of gangliosides does not need to be performed on each and every sample in a biological study, in this respect, a pooled sample containing all experimental conditions would be suffice to set up the final LC-MRM-MS method.