Determining β-Monosaccharide Head Group Composition with High-Resolution Cyclic Ion Mobility Separations Coupled to Tandem Mass Spectrometry as a First Step for Unknown Cerebroside Analysis

Abstract

Cerebrosides, a class of biologically important lipids, are comprised of a monosaccharide head group along with their ceramide tail. However, their accurate characterization is challenging because of the isomerism in both the tail, from potential double bond positioning, or in the head from monosaccharide composition and αβ anomericity. In this work, we focused on tackling the identification of the β-monosaccharide head group, as either glucose or galactose, in various cerebroside isomers as well as demonstrating how our methodology could be applied to unknowns found in a porcine extract. To achieve this, we performed collision-induced dissociation prior to cyclic ion mobility separations to generate monosaccharide fragment ions from the starting cerebroside precursor ions. With this pre-cIMS CID approach, we observed that the cIMS separations of the fragment ions were diagnostic of the β-monosaccharide head group composition (i.e., glucose versus galactose), regardless of the ceramide tail length. From there, we demonstrated an example of how this methodology could also be applied to cerebrosides found in a porcine extract and a framework for how this approach could be added to existing workflows in developing collision cross section databases. Overall, we envision that our developed pre-cIMS CID-based approach will be a complementary and orthogonal tool to existing ones in glycolipidomics workflows.

Graphical Abstract

1. Introduction

Cerebrosides are an important subclass of lipids that are responsible for many biological roles in the brain, skin, and nervous system ^1-10. Structurally, they are comprised of a ceramide tail with a monosaccharide head group ^2-4,6,7,11. Additionally, these cerebrosides represent the building blocks for larger gangliosides through extending the carbohydrate head group in either a linear or branched fashion ^2,7,8,12. From a biological perspective, accurate identification of monosaccharide head group composition in cerebrosides is important. ^{1,3,4,9,10,13-17}. However, doing so is largely hindered by their isomerism, which primarily stems from the identity of the hexose monosaccharide at the head group (i.e., glucose or galactose) and its corresponding α/β anomericity. Furthermore, the lack of available authentic analytical standards also prevents the unambiguous annotation of the monosaccharide head group composition in cerebrosides.

Liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) remains the gold standard for lipidomics analyses and has shown utility in cerebroside isomer characterization^{1,10,15,18-26}, but potential challenges related to isomer co-elution or convoluted MS/MS spectra warrants the development of complementary and orthogonal techniques ^{18,20,24,27-29}. An example of this from previous literature is that cerebrosides were characterized as a single species since the carbohydrate head group could not be accurately identified (e.g., an unknown cerebroside is often characterized as a hexosyl ceramide since the head group cannot be accurately identified as either glucose or galactose) ^{1,10,21,25,26}. Thus, the glycolipidomics community as a whole could benefit from the development of new, orthogonal, techniques and methods to complement existing workflows.

Ion mobility spectrometry-mass spectrometry (IMS-MS), where ions are separated in the gas phase based on their size, shape, and charge, has emerged as an orthogonal and complementary analytical technique to condensed-phase separations ^30-39. Technological advances to improve IMS-MS resolution, such as in cyclic IMS ⁴⁰ and structures for lossless ion manipulations ^30,36, have enabled the separation of isomeric species with collision cross section (CCS) differences <1%. Additionally, improved calibration strategies have leveraged these high-resolution IMS-MS platforms to enable the construction of robust and precise CCS databases to aid in molecular identification in omics-based research ^30,39,41. However, these CCS databases rely on authentic standards and thus database matching becomes challenging when an unknown compound is encountered for which no authentic standard exists. Previous work with high-resolution ion mobility separations has demonstrated that various glycolipid isomers can indeed be resolved in ultralong pathlengths ^32,42,43. Extensions of this research focused on using pre-cIMS collision-induced dissociation (CID) to assign sialic acid positioning in ganglioside isomers ⁴³. Interestingly, it was globally observed that shorter pathlengths were required to annotate and resolve fragment ions generated from pre-cIMS CID versus the longer separations needed to resolve isomers at the precursor ion level. Nonetheless, the ability to perform high-resolution IMS-MS separations on both precursor and fragment ions, as well as insights from all-ion fragmentation/mobility-aligned fragmentation (i.e., post-IMS CID), offers flexibility in experimental design as well as complementary information to be obtained from each measurement ^{30-32,34,35,42-44}. Herein, we build upon those previous high-resolution IMS-MS-based studies of glycolipids and present the use of high-resolution cyclic ion mobility spectrometry (cIMS)-based separations coupled with collision-induced dissociation (CID)-based tandem mass spectrometry (MS/MS) performed prior to separations to annotate the β-monosaccharide head group composition as either glucose or galactose in various cerebroside isomers. Initially, we began demonstrating our pre-cIMS CID workflow for monosaccharide head group composition analysis with authentic reference standards. From there, we highlight how it could be used and applied for cerebrosides present in an extract. Lastly, we describe a framework for using this as a potential first step for the future development of robust CCS libraries and thus more accurate cerebroside characterization.

2. Materials and Methods

2.1. Reagents and Sample Preparation

The following were purchased from Avanti Polar lipids (Birmingham, AL USA): C12 galactosyl (β) ceramide (860544), C12 glucosyl (β) ceramide (860543), C12 lactosyl (β) ceramide (860545), C16 galactosyl (β) ceramide (860521), C16 galactosyl (α) ceramide (860431), C16 glucosyl (β) ceramide (860539), C16 lactosyl (β) ceramide (860576), C17 glucosyl (β) ceramide (860569), C17 lactosyl (β) ceramide (860595), C18 galactosyl (β) ceramide (860844), C18 glucosyl (β) ceramide (860547), C18 lactosyl (β) ceramide (860598), C17:0 GA2 (860701), and cerebroside extract, porcine (131303). ESI low concentration tuning mix was purchased from Agilent Technologies (Santa Clara, CA USA).

For cIMS-MS separations, all samples were diluted to 5 μM in LC-MS grade methanol (Fisher Scientific; Pittsburgh, PA USA) and run with direct infusion (5μL/min) negative ion mode electrospray ionization (2.0 kV). The extract was diluted 1000-fold prior to analysis to achieve similar ion counts as those for the standards. For pre-cIMS CID experiments, all samples were run at a concentration of 500 μM with direct infusion (5μL/min) positive ion mode electrospray ionization (2.5 kV). We note that negative ion mode was used for the precursor ions as their singly deprotonated, [M─H]⁻, species while positive ion mode was used for fragment ions generated from pre-cIMS CID as their singly sodiated, [M+Na]⁺, species. This was done because of the low intensity of the singly deprotonated fragment ions in negative ion mode thus limiting the ability to perform high-resolution cIMS separations with sufficient signal-to-noise (see Supporting Information).

2.2. Instrumentation and CCS Measurements

A Waters Select Series Cyclic ion mobility spectrometry-mass spectrometry (cIMS-MS) platform (Wilmslow, UK) was used in all experiments ⁴⁰. Pre-cIMS CID was performed according to our previous work ⁴³. Briefly, precursor ions were m/z selected with the quadrupole and subsequently activated in the trap region by collisions with nitrogen buffer gas. Collision energie were optimized and can be found in respective figure captions. Cyclic ion mobility separations were performed in nitrogen buffer gas at a pressure of 1.74 mBar. Ions can be routed around the 1-meter separation region multiple times to improve resolution and then detected by the time-of-flight mass spectrometer operated in V-mode. Traveling wave conditions were optimized for each separation and can be found in respective figure captions.

Collision cross section (CCS) measurements were done using our previously developed pathlength independent calibration strategy ⁴⁵. This approach uses average ion velocities rather than absolute arrival times enabling us to evaluate and compare CCS values across varying pathlengths, and furthermore does not require calibrants and analytes of interest to be subjected to identical pathlength separations. Agilent tune mix ions were used as our CCS calibrants given that they are robust in nature (i.e., maintain a single IMS peak at any pathlength) and encompass/bracket our desired mobility and m/z ranges of interest for our lipid species. We acknowledge that there is a well-documented class specific bias associated with calibrant selection and resulting CCS values ^30,44,46, but it is difficult in obtaining sufficient cerebroside calibrants especially given their isomeric heterogeneity and potentially unresolvable peaks in low-resolution drift tube IMS measurements. Data was acquired over three separate days and arrival times were corrected for their dead times post-cIMS separation ⁴⁵. We do note that although our CCS calibration strategy is pathlength independent, it is necessary to use identical traveling wave (TW) conditions for calibrants and lipids of interest.

3. Results and Discussion

3.1. Evaluation of Pre-cIMS CID for β-Monosaccharide Head Group Identification in Cerebrosides

Initially, we were interested in determining if the use of pre-cIMS CID and corresponding high-resolution cIMS separations would enable us to assign the composition of the monosaccharide head group in various authentic cerebroside standards. Pre-cIMS CID of the precursor cerebroside ions revealed the presence of the hexose fragment ion for each corresponding isomer at m/z 203.1 (see Supporting Information for a representative tandem mass spectrum from pre-cIMS CID). From there, each of these hexose fragment ions, as their [M + Na]⁺ adducts, were subjected to cIMS separations. Figure 1 highlights the 3 m cIMS separation of sodiated glucose and galactose fragment ions generated from pre-cIMS CID of glucosyl (β) and galactosyl (β) ceramides (d18:1/8:0). We clearly observed that the arrival time distributions for the β-glucose and β-galactose fragment ions were unique from one another (i.e., different arrival times and overall arrival time distributions) after 3 meters of cIMS separation. The Supporting Information contains an arrival time distribution with the galactosyl (α) ceramide (d18:1/8:0), which surprisingly displayed three unique peaks, which we hypothesize is either due to differing sodium cation attachment sites or potential for impurities. We acknowledge that in this work we are highlighting monosaccharide head group composition based on individually run standards and that mixture analyses of both α/β forms may require longer pathlength separations and is the subject for future work. Also, this presented work does not focus on tackling the anomericity (i.e., α/β) bottleneck, and instead exclusively focuses on β ones. When comparing the β-glucose and β-galactose fragment ions, which is the focus of this study, they only required 3 m of separation to be resolved, whereas much longer separation pathlengths were required for cerebroside precursor ions from our previous work ³².

Figure 1.

3 m cIMS separation of the [M + Na]⁺ glucose and galactose fragment ions (m/z 203.1) generated from pre-cIMS CID (75 V) of glucosyl (β) and galactosyl (β) ceramides (d18:1/8:0) with a precursor m/z of 610.5. TW conditions: 550 m/s and 17 V.

Next, we were interested in assessing whether the same fragment ion arrival time distributions could be generated from cerebrosides with longer ceramide tail lengths. Figure 2 highlights the separation of the same fragment ions (i.e., β-glucose and β-galactose) generated from glucosyl (β) and galactosyl (β) ceramides (C18; d18:1/18:0) precursor ions. In comparing the arrival time distributions of the β-glucose and β-galactose fragment ions from Figures 1 and 2, we clearly observed that they matched regardless from what tail length ceramide precursor ion they were generated from (i.e., C8 versus C18). This trend in arrival time distribution matching also held true for β-glucose and β-galactose fragment ions generated from C12 and C16 ceramide tails (see Supporting Information). We believe this clearly demonstrates the utility of pre-cIMS CID coupled with high-resolution cIMS separations to accurately annotate the β-monosaccharide head group composition as either galactose or glucose of cerebrosides irrespective of the ceramide tail length present. Specifically, since β-hexoses are most prevalent as the monosaccharide head group in ceramides^9,20,25,39, our pre-cIMS CID approach coupled with cIMS separations can annotate their identity (i.e., glucose or galactose) irrespective of ceramide tail length (e.g., ranging from C8 to C18 in this study). Again, we acknowledge potential limitations of our approach for complex mixtures of isomeric ceramides, where some front-end separation (e.g., either with LC or IMS) would be required to obtain individual compounds to be subjected to pre-cIMS CID. We also would like to highlight that all fragment ions analyzed in this study were done in positive mode as their singly sodiated adducts, while precursor ions were done in negative mode as their deprotonated species. The rationale behind this is that hexose fragment ions generated from pre-cIMS CID were higher intensity in the MS dimension and resolved better in the cIMS dimension in positive ion mode, while the opposite was true for precursor ions (i.e., they better sensitivity and resolution in negative ion mode), which is consistent with previous observations from ion mobility separations of glycolipid precursor ions ^32,43. Example mass spectra are shown in the Supporting Information, where the hexose fragment ion is too low in intensity to be subjected to cIMS separations in negative ion mode.

Figure 2.

3 m cIMS separation of the [M + Na]⁺ glucose and galactose fragment ions (m/z 203.1) generated from pre-cIMS CID (75 V) of glucosyl (β) and galactosyl (β) ceramides (d18:1/18:0) with a precursor m/z of 750.6. TW conditions: 550 m/s and 17 V.

To demonstrate proof-of-concept feasibility for using our approach for mixture analysis, we subjected a 1:1 mixture of the C16 Glc/Gal(β)Cer isomers to pre-cIMS CID to generate the same β-hexose fragment ions (m/z 203.1) as described above. After a 3 m cIMS-MS separation, both hexose fragment ion isomers could be resolved from one another (see Supporting Information). Interestingly, even with attempted variations in CID voltage, we observed the galactose-containing fragment ion to be in higher intensity than the glucose-containing one (see Supporting Information). This 1:1 mixture of isomers illustrates that detecting β-glucose fragment ions in the presence of more abundant β-galactose ones may be challenging in terms of absolute limits of detection, or more importantly inter-isomer limits of detection. From there, we were interested in determining the opposite – what are the inter-isomer limits of detection for β-galactose fragment ions in the presence of more abundance β-glucose ones. To test this, we performed the same experiment as outlined above but for a 10:1 ratio of C12 Glc(β)Cer versus C12 Gal(β)Cer. Once again, we observed isomeric resolution of the β-hexose fragment ions resulting in two IMS peaks with approximately equal intensity after 3 m of cIMS-MS separation. Lower concentrations resulted in much poorer signal to noise, so we can conservatively estimate that β-galactose fragment ions can be accurately discerned from β-glucose ones at concentrations ~ 10X lower in isomeric mixtures (i.e., at a 1:10 ratio, β-galactose fragment ions could be resolved from β-glucose ones after pre-cIMS CID of precursor ions and a corresponding 3 m cIMS-MS separation). We do note that improved ion accumulation strategies^47,48, varying pre-cIMS CID strategies (e.g., in-source fragmentation), or isomeric resolution at the LC stage could help alleviate some of our observations related to inter-isomer limits of detection as it pertains to fragment ions generated from pre-cIMS CID. In the Supporting Information, we also extended this pre-cIMS CID approach to oligo-glycosyl containing species. Our data, while incomplete in terms of all possible monosaccharides and α/β anomericities, demonstrates that pre-cIMS CID is possible for larger species and could be a relevant addition to glycolipidomics-based workflows.

3.2. Proposed Workflow for Integrating Pre-cIMS CID with CCS Measurements for Cerebroside Analysis from Extracts

With our developed method using pre-cIMS CID and high-resolution cIMS separations to annotate β-monosaccharide head group composition in various authentic analytical cerebroside standards, our next goal was to demonstrate how this approach could be applied toward an extract and thus establish a first step in the overall workflow for analyzing cerebrosides and accurately characterizing their structure. Our proposed workflow for integrating pre-cIMS CID in the analysis of cerebrosides in extracts is as follows. First, we would identify putative tail lengths based on their m/z values (i.e., knowing the number of carbons present and any unsaturation) as well as confirming they are glycosyl-containing ones through MS/MS and the presence of known hexose fragment ions (e.g., m/z 203.1 for singly sodiated species). From there, we would perform pre-cIMS CID and apply it as previously described in the above sections to annotate the head group composition (i.e., glucose versus galactose). Lastly, we could then obtain a cIMS-based CCS value for these cerebrosides using our previously developed pathlength independent calibration strategy ⁴⁵. We acknowledge that other forms of MS/MS, derivatization, chemical modification, or front-end LC separations, would be required to more accurately characterize the ceramide tail. Additionally, more standards, particularly α-monosaccharide-containing ones, are needed to create a complete library.

An example of this proposed workflow is highlighted in Figure 3 where four cerebrosides are putatively annotated based on their ceramide tail length but contain an unknown monosaccharide head group. Through our developed pre-cIMS CID approach, we can identify the β-monosaccharide head group composition present in each of these species. All four cerebrosides shown in Figure 3 were found to be β-galactose-containing ones based on the arrival time matching to a known β-galactose-containing authentic standard (Figure 3, top). From there, we performed high-resolution cIMS separations on each of the singly deprotonated precursor ions, and in conjunction with our CCS calibration approach, we were able to obtain cIMS-based CCS values for both the authentic cerebroside standards as well as the ones observed in our porcine extract (Table 1). Again, we reiterate that mono-glycosyl precursor ions were subjected to cIMS separations in negative ion mode because of better sensitivity and resolution, while hexosyl fragment ions were subjected to cIMS separations in positive ion mode for the same reasons (i.e., especially related to the poor S/N observed for the deprotonated hexosyl fragment ion highlighted in the Supporting Information). We do note that other users applying this methodology will need to tune their traveling wave conditions for resolution as well as CID voltages for sensitivity depending on their instrument platform (e.g., homebuilt SLIM); regardless, similar arrival time distribution trends as well as CCS values within expected calibration error are to be expected ^{30,39,41,44-46}. Additionally, for more complex samples, post-cIMS CID may be necessary to deconvolute isotopic peaks from unsaturated species in a mixture to accurately identify the precursor ion and thus derive a CCS value.

Figure 3.

Monosaccharide head group assignment for cerebrosides in a porcine extract using our pre-cIMS CID methodology for the [M+Na]⁺ ions at m/z values of 750.6 for d36:1, 778.6 for d38:1, 806.7 for d40:1, and 848.7 for d41:1;O3). Identical CID voltages and TW conditions were used as shown in Figures 1 and 2. Hexose fragment ions were analyzed as their [M + Na]⁺ at m/z 203.1.

Table 1.

Cerebroside m/z and CCS values from authentic analytical standards and those from a porcine extract. CCS values are an average of triplicate trials performed on three different days with error bars of one standard deviation. * Denotes cerebrosides from the porcine extract.

Cerebroside	[M─H]−m/z	CCS (Å2)
C8 Glc (β)	586.4	250.1 ± 0.3
C12 Glc (β)	642.5	263.7 ± 0.3
C16 Glc (β)	698.6	275.6 ± 0.6
C17 Glc (β)	874.6	278.7 ± 0.5
C18 Glc (β)	726.6	281.5 ± 0.6
C8 Gal (β)	586.4	245.6 ± 0.5
C12 Gal (β)	642.6	260.0 ± 0.3
C16 Gal (β)	698.6	273.4 ± 0.4
C17 Gal (β)	874.6	276.4 ± 0.5
C18 Gal (β)	726.6	279.4 ± 0.4
*Gal (β) Cer (d36:1)	726.6	278.5 ± 0.4
*Gal (β) Cer (d38:1)	754.6	284.5 ± 0.3
*Gal (β) Cer (d40:1)	782.7	290.4 ± 0.3
*Gal (β) Cer (d42:1; O3)	824.7	296.2 ± 0.4

We would like to acknowledge that Table 1 is not a comprehensive list of all cerebrosides present in an extract or a list of all potential CCS values; doing so is beyond the scope of this current study. Instead, we have presented how pre-cIMS CID could be a useful first step and only one part of an entire glycolipidomics-based workflow for characterizing β-cerebrosides. Furthermore, our goal for this work was to illustrate how pre-cIMS CID with cIMS-MS separations can enable the annotation of monosaccharide head group composition; for this work we have largely ignored the ceramide tail and any isomerism that may be present there. For these reasons, we elected to only showcase four of the most abundant cerebrosides present in this extract to apply our pre-cIMS CID methodology for annotating their monosaccharide head groups. Overall, we envision our described pre-cIMS CID methodology as a first step toward unknown β-cerebroside analysis, especially when coupled with our proposed workflow including high-resolution cIMS-MS separations and corresponding calibration strategies for CCS measurements ^30,39,44,46. Ultimately, we envision our described method for monosaccharide composition analysis using pre-cIMS CID to be integrated with LC separations, various tandem mass spectrometry strategies, chemical modification, derivatization approaches, and CCS measurements to accurately characterize the structure of unknown cerebrosides. Our work described herein is only a small portion, and thus a first step, for a complete workflow.

4. Conclusions

Herein, we have developed a pre-cIMS CID-based method coupled with high-resolution cIMS separations as a first step toward unknown cerebroside analysis. In this work, we focused on characterizing and annotating the β-monosaccharide head group composition in various cerebroside isomers. Initially, we demonstrated that pre-cIMS CID was successful in generating diagnostic arrival time distributions indicative of the monosaccharide head group composition present in various galactose and glucose-containing authentic standard cerebroside isomers with varying ceramide tail lengths. We also demonstrated feasibility of our pre-cIMS CID approach for assessing isomeric mixtures and preliminarily highlighted inter-isomer limits of detection for β-galactose versus β-glucose fragment ions generated from cerebrosides. We are actively pursuing α-monosaccharide-containing cerebrosides. From there, we described a proposed workflow for how this methodology could be applied to assess β-monosaccharide head group composition in unknown cerebrosides present in an extract. We also highlighted how our pre-cIMS CID framework could be integrated with CCS measurements. Future work will focus on building on this proof-of-concept demonstration and apply our approach to various extracts to determine how monosaccharide head group composition changes as a function of various factors (e.g., disease progression). Additionally, we are planning to develop cIMS/cIMS methods to be able to assign monosaccharide head group composition for isomeric mixtures of cerebrosides; we acknowledge cIMS/cIMS could be more challenging as compared to pre-cIMS CID because of the elevated pressure of the cIMS array where ions are activated in cIMS/cIMS. We are also actively working toward developing methods to target the composition of the ceramide tail to enable us to accurately characterize the entire cerebroside structure.

Highlights for:

• Pre-cyclic ion mobility spectrometry collision-induced dissociation enables the unambiguous identification of β-monosaccharide head group composition in cerebrosides

• Collision cross section values were obtained for several unknown cerebrosides in an extract

• β-monosaccharide head group composition could be annotated regardless of ceramide tail length