Fragmentation behavior and gas-phase structures of cationized glycosphingolipids in ozone-induced dissociation mass spectrometry

Abstract

The role of cationization in the fragmentation behavior of glycoconjugates is amply documented in collisional activation techniques but remains less explored in ozone-induced dissociation mass spectrometry (OzID-MS). OzID-MS has been used to elucidate the location of carbon-carbon double bonds in unsaturated lipids. Previously, we demonstrated the structural analysis of unsaturated glycosphingolipids using OzID-MS by mass-selecting the [M+Na]⁺ adduct for fragmentation. In this work, we aimed to examine the effect of different adducts, namely [M+Na]⁺, [M+Li]⁺ and [M+H]⁺ on the OzID-MS fragmentation behavior of a representative unsaturated glycosphingolipid, LacCer d18:1/18:1(9Z). Our data show that [M+H]⁺ primarily undergoes dehydration followed by collision-induced dissociation-like loss of the headgroup, while [M+Li]⁺ and [M+Na]⁺ dissociate at the double bonds albeit with slightly different intensities of the resulting fragments. Using molecular mechanics and theoretical calculations at the semi-empirical level, we report for the first time the gas-phase structure of cationized glycosphingolipids, which helps to rationalize the observed bond cleavage. Our findings highlight that the type of adducts can influence gas-phase ion structure of glycosphingolipids and subsequently affect their fragmentation in OzID-MS. This study contributes to the growing body of knowledge on OzID-MS and gas-phase structures of ionized lipids and the findings have potential to be extended to other more complex glycoconjugates.

Graphical Abstract

INTRODUCTION

Glycosphingolipids are complex lipids composed of a ceramide backbone and a glycan headgroup [1, 2]. Together with glycoproteins, these molecules form the “glycosynapse” in the surface of eukaryotic cells that assume a plethora of vital biological functions such as cell recognition, cell-cell communication, signal transduction, receptor regulation and others [3]. The structural heterogeneity in glycosphingolipids lies in the glycan headgroup and the lipid tail (Fig. 1), both of which dictate the overall biological function of these molecules. This is further evidenced by earlier works that different biological conditions evoke changes not only in the composition of the headgroup but also of the lipid tail [1, 4-6]. As such, analysis of glycosphingolipids at the intact structure level is warranted.

Figure 1. General structure of a glycosphingolipid.

Shown here is LacCer d18:1/18:1(9Z) with glycan headgroup composed of glucose (Glc) and galactose (Gal), and connected to the ceramide backbone. The long chain base shown here is the sphingosine (d18:1) connected to a fatty acyl chain via amide bond. The numbering shown here is used throughout the text to address electronegative atoms.

Generally in tandem mass spectrometry of glycoconjugates, the type of adducts can significantly affect the fragmentation behavior when being selected for fragmentation under collisional activation techniques [7-10]. In this respect, myriad of works have exploited this property to elucidate structure of different molecules. Adducts can remarkably influence not only the energetics of bond cleavage but also the resulting fragmentation mechanism. For example, in oligosaccharides, protonated adducts typically do not yield the more informative cross-ring cleavages, but these fragments can be observed from sodiated adducts [8]. In lipids, lithium has been frequently used in shotgun analysis and for structural elucidation because it provides more informative fragments that are absent in sodium or protonated forms. Glycosphingolipids in particular have been studied previously using collision-induced dissociation (CID) of different adducts, especially lithium, to yield important fragment ions [11-13] that could reveal their intact structure.

To rationalize the data obtained experimentally, it is important to obtain the gas-phase structures of ionized molecules from theoretical calculations. Owing to advancements in computational chemistry and availability of powerful computers, a significant amount of work has been pursued to establish gas-phase structures of simple carbohydrates, peptides, and other small molecules, but to our knowledge, not of glycosphingolipids. Ion mobility spectrometry coupled to mass spectrometry (IMS-MS) [14] is generating huge amount of data that are helpful to characterize the size and shape of molecules, such as glycosphingolipids [15-17], but because the theoretical gas-phase structures of a wide selection of ionized lipids are currently not available, establishing a correlation between experimental data and predicted geometries becomes a challenge.

OzID-MS is a gas-phase fragmentation technique used to pinpoint the location, and in some cases the stereochemistry of carbon-carbon double bonds in unsaturated lipids [18]. This method is based on the highly selective ozone and olefin reaction chemistry [19] and has been implemented in various MS instrument platforms [18, 20, 21]. In these implementations, the typical inert collision gas is replaced by a mixture of ozone (O₃) and oxygen (O₂) such that ozonolysis can take place inside the collision cell. After precursor ion selection, the ion of interest reacts with O₃ in the collision cell and generates Criegee and aldehyde product ions. The mass difference between the precursor ion and product ions can serve as diagnostic masses to pinpoint the location of the double bonds in the molecular ion [18, 21].

It was previously demonstrated that different adducts could be used for OzID-MS which markedly affect product yield but not the resulting fragmentation patterns in phospholipids and fatty acid methyl esters [20, 22]. Previously, we reported the OzID-MS fragmentation of unsaturated glycosphingolipids providing neat spectra of diagnostic OzID-MS ions by selecting the sodiated adduct as precursor ion [23]. In this work, we aimed to study the fragmentation patterns of glycosphingolipids using sodiated, lithiated, and protonated precursors. We observed that the type of adduct clearly influenced the resulting fragmentation pattern. We also carried out theoretical calculations at the semi-empirical level to obtain predicted gas-phase structures of cationized glycosphingolipids that could provide plausible rationalizations of these observations.

2. EXPERIMENTAL PROCEDURES

2.1. Materials

The following standards were purchased from Avanti Polar Lipids (Alabaster, AL): D-lactosyl-ß-1,1' N-oleyol-D-erythro-sphingosine (LacCer d18:1/18:1(9Z), cat. no. 860590), D-galactosyl-ß-1,1' N-oleoyl-D-erythro-sphingosine (GlcCer d18:1/18:1(9Z), cat. no. 860547), N-oleoyl-D-erythro-sphingosine (Cer d18:1/18:1(9Z), cat. no. 860519). Lithium chloride was from Sigma-Aldrich, ammonium formate and formic acid were from Fisher Scientific (Pittsburgh, PA). All solvents were of LC-MS grade and purchased from Fisher Scientific (Pittsburgh, PA).

2.2. Preparation of standard solutions

Standard stock solutions were prepared as 1.0 μg·μL⁻¹ in chloroform/methanol/water (CHCl₃/MeOH/H₂O, 2:1:0.1, v/v) and kept at −20°C until further use. Working solutions for direct infusion experiments were made from these solutions by measuring appropriate aliquots in glass vials followed by drying under a stream of N₂ gas and subsequent reconstitution in acetonitrile/isopropanol/water (ACN/IPA/H₂O, 65/30/10, v/v) to a final concentration of 10 pmol·μL⁻¹. To induce specific adduct formation, ~ 1 mM LiCl and 10 mM NH₄COOH with 0.1% HCOOH was used for [M+Li]⁺ and [M+H]⁺, respectively. For [M+Na]⁺, only the neat ACN/IPA/H₂O solvent system was used.

2.3. Instrument set-up

All experiments described here were carried-out using Synapt™ G2 HDMS instrument (Waters, Manchester, UK) modified to allow the use of O₃ as collision gas as described previously [21]. In this setup, O₃ MEGA integrated ozone delivery system (MKS Inc., Andover, MA) was connected to the trap and transfer regions of the instrument through a three-way valve. The ozone generator was operated following manufacturer’s instructions and produced consistently 6.0-8.0 wt% O₃ in oxygen at a flow rate of 1.0 slm from high purity oxygen (AirProducts, Indianapolis, IN) at 20 psi, and used as collision gas in the mass spectrometer at a flow rate of 2.0 mL/min. The pressure in the trap region of Synapt™ G2 was ~9.9 × 10⁻³ mbar. For safety purposes, the remote switch of the ozone generator was interlocked with an ambient ozone monitoring system (Teledyne, San Diego, CA) and programmed to shut-off production when ambient ozone exceeds safe levels. Excess ozone was destroyed on-line using a destruction catalyst (MKS Inc., Andover, MA) that converts ozone to oxygen.

2.4. OzID-MS experiments

Standards were directly infused at 5.0 μL/min using an automated syringe pump. The ESI source conditions were optimized and the final working conditions were set at the following values: polarity, positive; spray voltage, 3.0 kV; sampling cone, 30 to 50 V; extraction cone, 6 V; source temperature, 100 °C; desolvation temperature, 200 °C; cone gas flow, 50 L/h; desolvation gas flow, 500 L/h. Isolation of precursor ions was carried out in the quadrupole at ~ 1Th isolation width (LM =16, HM = 15). All spectra were acquired for 1.0 min at 0.5 s/scan. The traveling wave in the trap and transfer regions were operated at the following settings as optimized previously [21, 23]: Entrance, 5.0 V; Bias, 2.0; Trap DC, 0.2; Exit, 0; trap wave velocity, 8 m/s; wave height, 0.2 V; transfer wave velocity, 247 m/s; wave height, 0.4 V. Default instrument settings for all the remaining parameters were applied. Under these settings, the reaction time between ozone and ions is estimated to be ~16.65 ms [21]. Instrument was calibrated daily in Resolution mode using sodium formate following manufacturer’s instructions obtaining less than 0.6 ppm (0.5 mDa) RMS residual mass. To compensate for the fluctuations of the ambient conditions during mass measurement, lock mass corresponding to leucine encephalin (m/z 556.2711) was employed. Both the full scan and MS/MS levels were mass-corrected traceable to the reference lock mass using MassLynx v4.1 instrument control software (Waters, Manchester, UK).

2.4. Theoretical calculations

All calculations reported here were performed in SCIGRESS vFJ2.8.1 (Fujitsu, Krakow, Poland) using built-in molecular mechanics and semi-empirical methods. Briefly, molecular structures of non-adducted forms were first geometry optimized using the MM3 force fields and the resulting optimized geometry was used to identify the low-energy conformers via CONFLEX MM3 [24] method. In each instance, at least 800 conformers were generated. Of these, only the twenty lowest predicted structures were used for subsequent calculation of the heat of formation (ΔH^°_f) using MO-G, a semi-empirical method based on Molecular Orbital Package (MOPAC) and employing PM6 parameters [25], both are built-in methods in SCIGRESS. The structure with the lowest heat of formation value was assigned as the most stable conformer of the neutral molecule. This was then used for subsequent calculations to examine preferred localization sites of each adduct.

The preferred localization of adducts was determined by considering all possible sites for adduct attachment, for a total of 14 different sites (cf. Fig. 1). The ion was positioned at about 2.5 Å for each potential site, followed by geometry optimization using MM3 force fields via molecular mechanics method and conformational search using CONLEX MM3 method [24, 26]. Then, the twenty structures with lowest predicted energy were subjected to MO-G PM6 calculations to determine ΔH⁰_f values.

2.5. Nomenclature of glycosphingolipid structures

The shorthand notations used in this manuscript were based on the nomenclatures adapted from the LipidMAPS consortium [27]. For convenience and intuitive presentation of OzID-MS data, the nomenclature for the position of the carbon-carbon double bond was adopted based on the (n-x) system where x corresponds to the position of unsaturation from the terminal −CH₃ of the hydrocarbon chain, for instance, double bonds in GalCer d18:1(4E)/18:1(9Z) are respectively annotated as n-14 and n-9.

3. RESULTS AND DISCUSSION

3.1. Comparison of fragmentation patterns among different cations

Different cations have been previously studied and found to affect gas-phase ozonolysis efficiency in major lipid classes [20, 22] but not in glycosphingolipids. We earlier demonstrated the OzID-MS fragmentation of unsaturated neutral glycosphingolipids [23]. Using sodium adducted species as precursor ions, we have shown that distinct neutral losses in the OzID-MS spectrum pinpoint the location of carbon-carbon unsaturation [23]. In addition, consistent with other works we also reported that sphingosine carbon-carbon double bond is less reactive than the unsaturation in the fatty acyl chain [23]. In this study, we compared the effect of adduct type on the OzID-MS fragmentation of cationized glycosphingolipids.

3.1.1. Lithiated adduct

Lithium has served as a classic cation in the structural elucidation of small molecules including glycosphingolipids by MS [11, 13, 28-31]. This metal promotes charge remote fragmentations that yield informative fragments thus facilitating the structural elucidation of unknown species. Owing to this, [M+Li]⁺ has been widely used for characterization of lipids and glycolipids mainly using collisional activation methods [11, 13]. Using OzID-MS, Thomas et al. showed that the fragmentation of [M+Li]⁺ using PC 16:0/18:1 did not yield distinct fragmentation compared to [M+Na]⁺ [20]. Phospholipids and glycolipids are structurally distinct owing to the head group diversity and stereochemistry on their backbone [1]. Thus, to determine if glycolipids also show similar behavior as phospholipids in OzID-MS, we infused glycolipid standards and subjected the [M+Li]⁺ adducts to OzID-MS fragmentation.

To induce the formation of [M+Li]⁺, the analyte in solution was spiked with ~1 mM LiCl. OzID-MS of [M+Li]⁺ showed the expected diagnostic ions, specifically, the ions at m/z 800.51 and m/z 784.51 corresponding to the Criegee and aldehyde ions of the n-9 double bond in the fatty acyl chain (Scheme 1 and Fig. 2). Also, the ions at m/z 620.29 and m/z 604.28, corresponding to the Criegee and aldehyde ions formed from the sequential cleavage of the n-14 double bond in the long chain base after cleavage of n-9 double bond. Similar to what we reported previously [23], the n-14 sphingosine double bond appeared to be less reactive than the n-9 acyl double bond which could be explained from the gas-phase structure of these ions as discussed later in this report.

Scheme 1.

Overview of the proposed pathways leading to the formation of ozonolysis products observed in the OzID-MS spectrum.

Figure 2. OzID-MS spectra of cationized LacCer d18:1/18:1 (9Z).

The analyte (10 pmol/μL) spiked with ~1 mM LiCl or 10 mM NH₄COOH/0.1% HCOOH to induce the formation of [M+Li]⁺ and [M+H]⁺, respectively. No spiking of Na⁺-containing salt was necessary to induce [M+Na]⁺ formation. Samples were infused using a syringe pump to the ESI source of Synapt G2 HDMS equipped with O₃ gas in the collision cell. Precursor ions corresponding to different adducts were chosen for OzID-MS. Spectra shown are average of 1.0 min acquisition time. Criegee and aldehyde product ions are depicted as (●) and (■), respectively, as shown in Scheme 1. Colors represent different double bond locations. : ozonide stabilized secondary oxidation cleavage product of sphingosine n-14 double bond (see Supplemental Scheme S1), : loss of hexose, : loss of H₂O.

We found that in contrast to previous reports on OzID of phospholipids [20, 22], glycolipids have a distinct fragmentation pattern when ionized as [M+Li]⁺ compared to [M+Na]⁺ (Fig. 2). The [M+Li+O₃]⁺ ion was found to be of higher intensity relative to other fragments within the same spectrum, in contrast to [M+Na+O₃]⁺. Also, when the relative intensities of the Criegee and aldehyde ions were compared with the intensity of the precursor ion, the two adducts showed differential fragmentation efficiency. Specifically, the Criegee and aldehyde ions formed from the cleavage of sphingosine double bond are of higher intensity than those in [M+Na]⁺. As a result, the intensity difference between the fatty acyl and sphingosine OzID characteristic ions is not as drastic as that in [M+Na]⁺. Previous works by others also demonstrated the peculiar behavior of lithiated adducts in CID and high-energy CID [11, 31, 32]. This observation can be ascribed to the relative stability of lithiated over sodiated ozonide, presumably because of the lower ionic radii, but higher electronegativity of Li⁺ than Na⁺, i.e. higher charge density [33].

The most striking difference between [M+Na]⁺ and [M+Li]⁺ was the presence of a peak that differ by 180 Da from that of the ozonide, which appeared at m/z 778.38 and m/z 762.42 for the respective adducts (Supplemental Scheme S1). Using accurate mass measurements, we showed previously that this fragment originates from the secondary oxidation of the sphingosine carbon-carbon double bond, and could not be due to the loss of one of the monosaccharides in the glycan headgroup, as in the case in sodiated adducts [23]. The only major difference is the relative intensity of this fragment between the two metals, specifically, this fragment had a higher relative intensity in [M+Li]⁺ compared to [M+Na]⁺. This suggests that the reactivity of the n-14 sphingosine double bond is higher in lithiated than in sodiated adduct. We speculate that this differential reactivity stems from higher charge density of lithium ion which translates to a relatively more compact structure, providing closer interaction of the charge carrier with the n-14 double bond. This is in agreement with our theoretical calculations described in the succeeding sections. Indeed, distinct behavior between these metal ions has been observed not only in glycolipids, but also OzID-MS of lithium and sodium-adducted fatty acid methyl esters [22] and cholesteryl esters [34]. Taken together, these observations highlight that Li⁺ and Na⁺ ions could affect the efficiency of ozonolysis in the gas-phase.

3.1.2. Protonated adduct

The burgeoning use of protonated adducts in MS stems from the popularity of LC-MS based approaches where often, additives like ammonium formate, ammonium acetate, formic acid or acetic acid are being used. The presence of these additives suppresses the ubiquitous Na⁺ that is typically observed especially for molecules that have high affinity to these ions. In this respect, we studied the OzID-MS fragmentation of protonated LacCer d18:1/18:1(9Z).

We observed substantially distinct OzID-MS spectra for [M+H]⁺ compared to [M+Li]⁺ and [M+Na]⁺. Specifically, we observed persistent neutral loss of H₂O, this dehydration is attributed to the lability of the sphingosine hydroxyl [10] as consistently observed in traditional CID and HCD fragmentation of protonated ceramides and glycosphingolipids [11, 35, 36]. We speculated that the loss of H₂O originated from the sphingosine backbone because infusing Cer d18:1/18:1(9Z) also showed dehydration despite the absence of a glycan headgroup (Supplemental Fig. S1). It is interesting to note that the observed dehydration under OzID conditions was not detected with [M+Li]⁺ or [M+Na]⁺ (Fig. 2 and Supplemental Fig. S1) although it was observed in collisional activation dissociation (CAD)-based fragmentation [31]. This suggests that the lability of the sphingosine hydroxyl in OzID-MS could be attenuated by metal adducts.

In contrast to sodiated and lithiated adducts, where the precursor ion intensity has dropped significantly, the protonated precursor ions remained abundant in the OzID spectrum (Fig. 2). Interestingly, we observed that the ozonide intensity is much weaker compared to that in sodiated and lithiated adducts. Previous OzESI work [37] on sphingomyelin as protonated adduct did not reveal the position of the sphingosine double bond, in contrast, in our work, this double bond is clearly shown, although we failed to detect the Criegee product from this double bond cleavage, probably because of its low stability (Fig. 2). This yields two important implications for [M+H]⁺, first, its reactivity to ozone is lower than that of [M+Li]⁺ and [M+Na]⁺, and second, ozonolysis of the sphingosine double bond in [M+H]⁺ is less pronounced compared to that of the alkali metal adducts.

When protonated adduct was mass-selected for fragmentation, the expected OzID products were observed (Fig. 2C). The ion at m/z 776.50 and m/z 760.50 are the Criegee and aldehyde ions corresponding to neutral losses of 94 and 110 Da, respectively, from the dehydrated precursor ion, diagnostic of the n-9 double bond at the fatty acyl chain [18]. Interestingly, the expected neutral losses of 180 Da and 164 Da from the n-9 cleavage products to locate the n-14 sphingosine double bond were not observed. However, two fragment ions, m/z 436 and m/z 256 that differ by 180 Da were apparent, implying that n-14 oxidative cleavage in [M+H]⁺ adduct occurs differently than that of [M+Na]⁺[23] and [M+Li]⁺.

In addition to the informative OzID fragments corresponding to sphingosine C=C and fatty acyl chain C=C, we observed the formation of a peak that differs by 324 Da from the diagnostic ion for n-9 C=C position in LacCer, and 162 Da for GlcCer (Supplemental Fig. S2), both have a mass of m/z 436. This ion could result from the loss of the sugar headgroup in a CID-like fragmentation mechanism. To confirm this, we subjected ceramide molecule that contains the same long chain base and fatty acyl chain, Cer d18:1/18:1(9Z) to OzID-MS as [M+H]⁺ ions (Supplemental Fig. S1). Indeed, the ion m/z 436 was also observed here, verifying that loss of headgroup, and thus, cleavage of Cer-Glc glycosidic bond has occurred and that the m/z 436 corresponds to the aldehyde ion formed by the cleavage of the n-14 double bond in sphingosine (cf. Scheme 1). The peak at m/z 598 that differ by 162 Da from the aldehyde product of the oxidative cleavage of n-9 double bond (m/z 760) along with the presence of m/z 436 (mass difference of 162 Da from m/z 598) support the potential of identifying the glycan sequence in LacCer d18:1/18:1(9Z).

We previously showed that using [M+Na]⁺, the sphingosine double bond at n-14 with a trans-configuration is relatively less reactive than that of the n-9 double bond in the fatty acyl chain [23]. One of the possible explanations offered in the past was the presence of the hydroxyl group adjacent to the original n-14 double bond along with the possible formation of H-bond between the headgroup and this hydroxyl group as observed using sphingomyelin [20]. In our present results using protonated adduct, the cleavage product of the n-14 double bond remained lower intensity than that of n-9 double bond, despite the elimination of the sphingosine hydroxyl suggesting that the presence of the hydroxyl group was not the sole determinant of the low ozonolysis efficiency at this position. We speculated that the proximity of this unsaturation system to the headgroup has influence on the overall OzID cleavage, this is indeed the case, as we modelled the gas-phase structure of the dehydrated species (Fig. 3).

Figure 3.

Calculated gas-phase structure of dehydrated LacCer d18:1/18:1(9Z). (A) Proposed elimination of H₂O from the sphingosine backbone via charge-remote mechanism. (B) Gas-phase structure of dehydrated LacCer d18:1/18:1(9Z) calculated using MO-G with PM6.

Taken together, these results show that the use of [M+H]⁺ adducts in OzID-MS of glycosphingolipids could potentially reveal the sequence of the glycan headgroup and pinpoint the location of the double bonds.

3.2. Insights on the gas-phase structures of cationized glycosphingolipids

The gas-phase structures of ionized lipids could explain observations in MS, especially in the era of ion mobility spectrometry [38]; however, to our knowledge, few studies are available in this area. We reasoned that in addition to the known effect of adduct size on the resulting fragmentation patterns evident for oligosaccharides [8], the localization of the cations could also play a vital role in deciding the fate of fragmentation. Thus, to partially rationalize the observed OzID-MS behavior, we performed theoretical calculations on LacCer d18:1/18:1(9Z).

In this study, we opted to employ molecular mechanics and semiempirical methods to estimate the gas-phase structures of cationized LacCer d18:1/18:1(9Z). Although ab initio and density functional theory (DFT) are by far considered to provide relatively more accurate predictions of gas-phase structures of ionized species, molecular mechanics (MM) and semi-empirical methods are also suitable and found to be satisfactory by others [26, 39-43]. MM3 has been previously shown to be appropriate for disaccharides analysis [44]. Using MM3, we optimized the geometry of our model molecule in its neutral form and performed a conformational search using CONFLEX method utilizing augmented MM3 force fields [24]. The lowest energy conformers were subjected to further optimization using MO-G by employing PM6 parameters to obtain the values of heat of formation (ΔH⁰_f). Following geometry optimization, arbitrary dihedrals were chosen for conformational search using CONFLEX MM3 method.

To investigate the preferred localization of cations when complexed with the ligand, we employed the lowest energy conformer of the neutral form as starting point (Supplemental Fig. S3). We positioned the cations (Na⁺, Li⁺, H⁺) at ~2.5 Å near each potential cation residence in the LacCer d18:1/18:1(9Z) ligand for a total of 14 different positions and performed identical procedure as mentioned above for the neutral molecule, following an analogous study [45]. Thus, we obtained the low energy structures of cationized LacCer d18:1/18:1(9Z) in the gas-phase using molecular mechanics and semi-empirical calculations as discussed in the subsequent sections.

3.2.1. Energetics of cationization

Cation affinity of ligand L with cation C⁺ is defined as the standard enthalpy change for the following reaction [46]:

$$[\mathrm{LC}{]}^{+}\to {\mathrm{L}}_{(\mathrm{g})}+{{\mathrm{C}}_{(\mathrm{g})}}^{+}\text{Cation affinity}=\Delta H^0>0$$ (Eq. 1)

Cation affinity is an important parameter that serves as a relative measure of the stability of the adduct and becomes very useful when different cations are being compared. The values of cation affinities for simple carbohydrates have been investigated previously[47-49], but those of glycosphingolipids remained relatively less studied [50]. In this present work, we used molecular mechanics and semiempirical methods to estimate cation affinity of LacCer d18:1/18:1(9Z) for three cations, H⁺, Li⁺ and Na⁺. The calculated values LacCer d18:1/18:1(9Z) (Table 1) show that the cation affinities of the ligand for H⁺ and Li⁺ are greater than that of Na⁺ which suggests that formation of [M+H]⁺ and [M+Li]⁺ adduct is more exothermic than [M+Na]⁺. The values shown here are consistent with experimental data by others [8, 33, 49] although the values obtained in this work are slightly higher than those of disaccharides, which are expected due to the additional lipid moiety in glycosphingolipids [47, 51]. The high exothermicity of this process along with the size of the adducts were previously associated with increased metastable decay as well as CID fragmentation of [M+H]⁺ adducted carbohydrates than their metal-cationized congeners [49, 52].

Table 1.

Estimated heat of formation (ΔH⁰_f) and cation affinities of LacCer d18:1/18:1(9Z) calculated using MO-G with PM6. Detailed calculations are shown in Supplemental.

ΔH0f (kJ mol−1)
	Uncomplexed	Complexed	Cation affinity
[M+Na]+	−2,400.86	−2,740.33	+339.47
[M+Li]+	−2,289.75	−2639.25	+349.50
[M+H]+	−1,642.79	−2,445.62	+802.83

3.2.2. Calculated gas-phase structure of [M+Na]⁺

Arbitrarily, we considered an atom to be coordinating with the cation if their distance is less than 2.7 Å [45]. In this respect, in the optimized, lowest energy structure of [M+Na]⁺, we observed that Na⁺ coordinates to glycosidic oxygen (O4, 2.56 Å), (O12, 1.44 Å), and amide oxygen (O2, 1.23 Å) (Fig. 4, for notations, please refer to Fig. 1). These values are within reasonable values as previously calculated using more sophisticated DFT methods for carbohydrates [47, 48, 51]. Of these interactions, the strongest was between Na⁺ ion and the amide oxygen (C=O, O2). In a previous study involving oligosaccharides [53] where one of the sugars involved is N-acetyl neuraminic acid, Na⁺ was also found to coordinate with the carbonyl oxygen.

Figure 4.

Most stable conformer of [M+Na]⁺ of LacCer d18:1/18:1 calculated using MO-G with PM6. Color indicates different atom (Gray: Carbon, White: Hydrogen, Red: Oxygen, Purple: Na⁺, Cyan: Nitrogen).

In this conformation, we found three prominent intramolecular H-bonding interactions between the sphingosine -OH and the glycosidic oxygen (O3), (O8 and O9), and (O11 and O12). These H-bonding interactions contributed to the stability of this structure in the gas-phase.

The overall 3D spatial distribution of functional groups could significantly dictate the overall gas-phase reactivity. Very clearly, the sodium ion appears to be closer to the acyl chain double bond than the sphingosine double bond, presumably because of the size of the headgroup and the trans-configuration of this double bond. Considering that OzID cleavage is charge-induced rather than charge remote [20, 34], the gas-phase structure predicted in our current work offers a plausible explanation to the reduced ozonolysis efficiency of the sphingosine double bond observed previously [20, 23, 54] and in this study.

3.2.3. Calculated gas-phase structure of [M+Li]⁺

In the optimized lowest energy structure of [M+Li]⁺ adduct, Li⁺ ion demonstrated preferential interaction with sphingosine −OH (O1, 1.46 Å), (O11, 2.22 Å), amide oxygen (O2, 1.24 Å), (O12, 2.21 Å), amide nitrogen (N14, 1.48 Å) (Fig. 5, for notations, please refer to Fig. 1). The overall similarity of Li⁺ and Na⁺ is manifested by their stronger interaction with the carbonyl oxygen. A landmark study on ceramide fragmentation also hypothesized that Li⁺ preferably resides proximal to the carbonyl oxygen [31]. Recent works on disaccharides showed similar conformation regardless of the metal identity, but remarkably distinct to that of proton [47], as observed in this current work. We also found three prominent intramolecular H-bonding interactions between O12 and O2, O9 and O8, and O6 and O8. Again, the stability of this gas-phase structure could be ascribed, in part, to these H-bonding interactions.

Figure 5.

Most stable conformer of [M+Li]⁺ of LacCer d18:1/18:1 calculated using MO-G with PM6. Color indicates different atom (Gray: Carbon, White: Hydrogen, Red: Oxygen, Purple: Li⁺, Cyan: Nitrogen).

Consistently, the sphingosine double bond is less reactive towards ozonolysis than the fatty acyl chain double bond. Like what was observed in [M+Na]⁺, the sphingosine double bond in [M+Li]⁺ appeared to have lower ozonolysis efficiency than n-9 fatty acyl chain double bond. This can be explained in a similar manner as above, that the charge-carrier is relatively closer to the n-9 acyl double bond than that of n-14 sphingosine. The observed higher intensities of the fragment ions in the OzID-MS spectrum could be partly explained by the lower ionic radius but higher charge density of Li⁺ ion compared to Na⁺ as evidenced by higher cation affinity of Li⁺ than Na⁺ calculated above (cf. Table 1) and in previous experimental works [11, 13, 28, 31].

3.2.4. Calculated gas-phase structure of [M+H]⁺

The [M+H]⁺ has the most distinct gas-phase structure (Fig. 6), specifically, in the optimized structure of the lowest energy conformer, we found the localization of the proton to oxygen at O7 position (1.46 Å) (for notations, please refer to Fig. 1). This proton is close to the carbonyl oxygen (1.26 Å), sphingosine −OH (1.64 Å), and glucosyl oxygen at O8 (1.43 Å). Further, compared to [M+Na]⁺ and [M+Li]⁺, there are more prevalent intramolecular H-bonding interactions, such as O13 and O1, O7 and O13, O1 and O7, O5 and O8, O4 and O8, and O11 and O12. This seemingly peculiar spatial location of the proton compared to Na⁺ and Li⁺ is analogous to what was observed for saccharides which could be related to the size of the cation [47].

Figure 6.

Most stable conformer of [M+H]⁺ of LacCer d18:1/18:1 calculated using MO-G with PM6. Color indicates different atom (Gray: Carbon, White: Hydrogen, Red: Oxygen, Purple: H⁺, Cyan: Nitrogen).

Overall, comparing the proximity of the acyl double bond to the charge, the cation appears to be closer to the double bond in [M+Li]⁺ and [M+Na]⁺, but reasonably distal in [M+H]⁺. This could explain the maximized H-bonding interactions observed in [M+H]⁺. The larger size of the sodium and lithium compared to proton could have caused the overall more compact structure in [M+H]⁺ over [M+Na]⁺ and [M+Li]⁺, as predicted by LipidCCS web server (http://www.metabolomics-shanghai.org/LipidCCS/) and observed in recent ion mobility experiments of similar GlcCer species [15, 55]. Previous work argued that, instead of remote fragmentation, there is a direct role of charge in the mechanism of OzID of ionized lipids [20, 34, 37]. More recently, a study on the gas-phase structure of 1-deoxysphingosine shows the through-space interaction of charge with the unsaturation site [54]. The gas-phase ion structures we obtained here, clearly support this hypothesis. In this respect, partly, this explains the previous observation that metal adducts provided higher efficiency than protonated adduct.

Seminal works in the past have stressed that protonated species, both in oligosaccharides and in glycolipids [35, 49], could induce substantial fragmentation compared to their metal adduct analogues as observed in MALDI, FAB, and ESI studies. Our results agree with these previous works, and additionally, we provide evidence that not only in collisional activation, but also in OzID-MS that protonated adducts could induce glycosidic bond cleavages.

The intensity of the ozonide ion in [M+H]⁺ being almost completely attenuated compared to [M+Na]⁺ and [M+Li]⁺ (cf. Fig. 2) could not be a consequence of the increased ozonolysis efficiency, but rather, of the CID-like breakdown of the ozonide ion to cleave the glycan headgroup. We reasoned that both protonation and ozonolysis are highly exothermic processes [56], releasing energy upon reaction. This highly energetic process could provide adequate internal energy provoking the cleavage of weak glycosidic linkages. In this respect, it is well-documented [8] that protonated carbohydrates require lower amount of energy to break glycosidic bonds than metal adducts, estimated to be 138.3 kJ mol⁻¹ and 245.6 kJ mol⁻¹, respectively [47] for β anomers. Taken together, these results suggest that the high charge density of the H⁺ ion along with the highly exothermic nature of ozonolysis, could induce glycosidic bond cleavages which are absent in metal-adducted species. This can offer a possibility to determine the location of double bonds, and sequence the glycan headgroup in OzID-MS of intact, unsaturated glycosphingolipids.

CONCLUSION

This study demonstrated the distinct fragmentation behavior of cationized (Na⁺, Li⁺, H⁺) glycosphingolipids in OzID-MS. While similar fragmentation patterns were observed between sodiated and lithiated adducts in terms of the presence of diagnostic OzID ions, relative intensities of these fragments were remarkably different. Protonated adducts gave the most striking difference in the fragmentation pattern as evidenced by the presence of ions resulting from cleavage of the glycosidic bonds. Molecular modeling using molecular mechanics and semiempirical methods show that the size of the cation could influence the gas-phase structures of the cationized glycosphingolipids. This, along with the exothermic nature of cationization, presumably has induced glycosidic cleavages and charge-induced ozonolysis of the carbon-carbon double bonds. This is especially the case for the [M+H]⁺ ion where the high charge density of the proton and its peculiar localization close to glycosidic oxygen evoked the breaking of the glycosidic bonds. Thus, the use of [M+H]⁺ could provide both double bond locations and sequence of the glycan headgroup. Taken together, our work contributes to the current understanding of OzID-MS, gas-phase ion structures of neutral glycosphingolipids, and has potential to be extended to other glycoconjugates.