Introduction

Carbohydrates play vital roles in biology but are often difficult to identify confidently from these samples [1,2,3,4,5]. One of the most challenging aspects of biological carbohydrate research is the need to distinguish between multiple isomeric structures [6,7,8,9,10,11,12,13,14,15,16,17]. This has led to a proliferation of mass spectrometry-based methods aimed at mitigating this problem [3, 6, 8,9,10,11,12, 18,19,20,21,22,23,24,25,26,27,28,29,30]. Complex polysaccharide carbohydrates are formed from simple monosaccharide units in glycosylation reactions. Hexose (C6H12O6) monosaccharides are the main building blocks of complex carbohydrates [5]. In the present article, we investigate the gas-phase fragmentation chemistry of three hexopyranose monosaccharides common in living systems: glucose, galactose, and mannose (Figure 1). Lithium cationization is utilized. These analytes differ only in the stereochemistry of individual hydroxyl groups. However, these simple stereochemical variations can have a profound impact on the chemistry in biological systems [31] and the gas phase.

Figure 1
figure 1

The different monosaccharide systems in this study. (a) Glucose. (b) Galactose. (c) Mannose. The anomeric center (carbon 1) configuration exists as a mixture of the axial (α) and equatorial (β) forms

Full size image

In the present article, we utilize tandem mass spectrometry (MS/MS) [32] and regiospecific isotopic labeling [1,2,3, 16, 21, 23, 24, 33, 34], coupled with theory [1, 2, 34, 35], to elucidate the characteristic fragmentation chemistry of lithiated monosaccharide analytes. The present manuscript is a follow-up to our [1, 2] and others’ [34, 35] recent work on sodiated carbohydrate analytes (specifically requested by reviewers of the earlier paper). While there is certainly a wealth of relevant experimental work on lithiated carbohydrates (for example, [3, 12, 20, 25, 33]) and their fragmentation, there is little/no theoretical data on the specific fragmentation chemistry of these analytes [35, 36]. An analogous situation exists for theoretical data; a substantial number of studies on neutral carbohydrate geometries [37,38,39,40], but data on lithiated forms is lacking. In the present article, we investigate the primary, structurally useful, fragmentation pathways of lithiated glucose, galactose, and mannose: water loss (primarily B1 ion formation) and the cross-ring cleavage reactions producing the 0,2A1 and 0,3A1 ions. We provide evidence for the key gas-phase structures, mechanisms, and energetics underlying these processes.

Experimental

The experimental work was done using an electrospray ionization MaXis plus quadrupole time-of-flight mass spectrometer (Bruker, Billerica, MA). The analytes were diluted to ~ 5 μM with acetonitrile/water/lithium chloride (50/50/0.1%) and then sprayed at a flow rate of 3 μl min−1. Nitrogen was used as both nebulizing and drying gas. The lithiated analytes were selected using the quadrupole followed by activation by collision in the collision cell containing nitrogen. The resulting product ions and remaining precursor ions were dispersed by the time-of-flight mass analyzer. Data was collected as a function of collision energy. Exchange of the analyte hydroxyl protons for deuterons was achieved by dissolving the analytes in deuterium oxide (D2O) for 10 min at room temperature, prior to further dilution in acetonitrile/D2O/LiCl (50/50/0.1%) to a final concentration of ~ 5 μM [3]. Deuterium oxide was purchased from Cambridge Isotopes Laboratories, Inc. (Tewksbury, MA). Lithium chloride, HPLC-grade acetonitrile, and H2O were purchased from Sigma-Aldrich (St. Louis, MO). regioselectively isotopically labeled monosaccharides were purchased from Cambridge Isotopes Laboratories, Inc. (Tewksbury, MA).

Theoretical Methods

Density functional calculations of minima, transition states, product ions, and neutrals were performed with the Gaussian 09 suite of programs [41] at the M06-2X/6-31+G(d,p) level of theory [42, 43]. Multiple conformers of each site of lithiation were examined for each system by scanning the potential energy surface. An initial pool of seed structures was generated using the molecular dynamics engine Fafoom [39, 40] via a genetic algorithm utilizing the MMFF94 force field [44,45,46,47,48]. These structures were sorted based on ring configuration and energy. Once a starting pool has been formed, the genetic algorithm begins with new trial structures generated based on components (i.e., torsion angles and ring configuration) of previous candidates/results. These trials are also subjected to geometry optimization and added to the candidate pool. The neutral structures were geometry optimized at the M06-2X/6-31G(d) level of theory in the Gaussian 09 suite of programs [41]. Following removal of degenerate structures, the optimized neutral candidate structures for each system were then lithiated utilizing a coordinate sensitive script. This process was repeated for all potential sites of lithium attachment. The resulting structures were optimized at the M06-2X/6-31+G(d,p) level of theory. Results of these calculations were then inspected. These structures were ranked based on electronic energy after which the lowest energy, non-degenerate structures were selected for vibrational analysis. Having characterized the low energy minima, multiple transition structures (TSs) were sought. Minima were checked by vibrational analysis (all real frequencies) and TSs were also examined in this manner (one imaginary frequency). The reaction pathway through each particular, energetically competitive TS was determined by intrinsic reaction coordinate (IRC) calculations with up to 18 steps in each direction. The terminating points of these calculations (one on product side, one on reactant side) were then optimized further to determine the exact minima connected by each specific transition structure. Estimates of the lithium affinities of the leaving groups were determined as the difference between the zero-point energy-corrected M06-2X/6-31+G(d,p) total electronic energies (0 K) of the lithiated and neutral form plus Li+ at infinite separation.

Results and Discussion

Experimental Findings

Lithiated glucose, galactose, and mannose analytes, [C6H12O6+7Li]+, populate similar fragmentation pathways (Figure 2; for nomenclature, see Figure S1 [49]). They all produce a water loss peak at m/z 169 which is consistent with the literature on metal-cationized carbohydrate ions [1,2,3, 33, 34, 50, 51]. Lithiated analytes also produce peaks resulting from cross-ring cleavages. Unlike recent work from Chen et al. on sodiated glucose analytes [35], we observe two cross-ring cleavage peaks. Our accurate mass and labeling data supports assignments of 0,2A1, [C4H8O4+7Li]+ at m/z 127 and 0,3A1, [C3H6O3+7Li]+ at m/z 97 in all cases. The key difference between the three analyte populations is manifested in relatively small changes in the relative critical energy required to initiate fragmentation. The lithiated glucose and galactose epimers fragment at lower collision energies than the mannose forms (Figure 2, Figure S2). In addition, the relative abundance of the peaks vary between the systems supporting either differing product dimer-constituents [1, 2] or energetics in each case.

Figure 2
figure 2

Example MS/MS spectra (Ecollisions, lab = 15 eV) of the isomeric lithium-cationized analytes. (a) Glucose. (b) Galactose. (c) Mannose

Full size image

Experimentally, the most facile, useful, reactions for [glucose+7Li]+ are water loss from the anomeric center (B1, m/z 169) and a low abundance cross-ring cleavage 0,2A1 peak. This is followed by another cross-ring cleavage peak, 0,3A1, then consecutive losses of water molecules from the 0,2A1 ion (m/z 109 and 91, Figure 2a, m/z 111 and 91, Figure S2a). We note that direct loss of Li+ also occurs, but this is of no structural benefit.

The lithiated galactose analytes require a similar degree of activation for fragmentation to be experimentally observed. Both the degree of fragmentation as a function of collision energy (reduced relative to glucose) and the nature of the primary fragments differ. For [galactose+7Li]+, the primary fragments are 0,3A1 and water loss (B1) from the anomeric center (Figure 2b, Figure S2b), followed by the 0,2A1 peak at increased collision energies. Similar to the glucose data, the 0,3A1 peak is more prevalent than the 0,2A1 peak at higher collision energies (Figures S3 and S4). Lithiated mannose is the least readily fragmented analyte experimentally (Figure 2c, Figure S2c). Similar to glucose, [mannose+Li]+ produces both the water loss (B1) and the 0,2A1, [C4H8O4+Li]+ peaks. The 0,3A1 ions are increasingly prevalent at higher collision energies (≥ 20 eV). However, unlike for the glucose and galactose congeners, the 0,2A1 ions are most prominent at higher collision energies (Figure S5). In addition, myriad consecutive fragmentation processes are also possible (water losses, C2H4O2 losses, etc.) at higher collision energies along with a substantial decrease in detectable ion signal resulting from loss of Li+ and/or inability to efficiently capture the low m/z products. This general finding though not the exact product distribution also holds for the other monosaccharide analytes.

To distinguish the carbons and the hydrogen atoms contributing to fragment losses, we performed hydrogen-deuterium exchange of the hydroxyl protons forming [C6H7D5O6+7Li]+ precursor ions. These analytes were then subjected to collisional activation (Figure S2). Additional analyses of regiospecifically labeled (13C, 2D) forms of our analytes were also performed. The key findings are the following: (1) support for the 0,2A1 and 0,3A1 peak assignments over the isomeric X0 ion possibilities (Tables S1–S3, Figure S1) and (2) that losses are of D2O and not DOH to furnish the B1 ions at m/z 172, i.e., no loss of C-alpha protons. This is entirely consistent with the prior literature [1,2,3, 33]. Data is provided in Figure S2 and Tables S1–S3 for the interested reader.

Energetics of Lithiated Minima

The lowest energy structures of glucose, galactose, and mannose are shown in Figure 3 and Figure S6. Our calculations indicate that the global minima of lithiated glucose are skew conformations (OS2) [52] in which the Li+ is coordinated to the C3 and C6 hydroxyl oxygens (Figure 3a, Figure S6a). This contradicts the earlier claims of Ni and co-workers who did not locate any skew conformations [34]. The GM structures advocated by those authors have fewer oxygens coordinating the lithium cation and are 6.9 (α) and 17.7 (β) kJ mol−1 less energetically favorable based on our calculations. Skew conformations appear to be characteristic of lithiated systems as these same authors found them to be less competitive for sodiated glucose congeners [35]. Alternate low-energy families of lithiated glucose structures formed are chair conformations (1C4 and 4C1) requiring at least 16 and 17 kJ mol−1 to populate (Figure S7). In contrast, the lowest energy conformation of the lithiated β-galactose anomer is a chair structure (Figure 3b). The [α-galactose+Li]+ analytes are also predicted to form chair conformations (Figure S6b) as are both mannose anomeric forms (Figure 3c, Figure S6c).

Figure 3
figure 3

Global minima of the isomeric lithium-cationized analytes. (a) β-Glucose. (b) β-Galactose. (c) β-Mannose. β-Glucose adopts a skew conformation (OS2) while the galactose and mannose preferentially adopt chair conformations

Full size image

Water Loss Pathways and B1 Ion Formation

The water loss is initiated by proton transfer from one of the hydroxyl groups rather than a Cα proton (Scheme 1, Figure 4). Additional experimental evidence for this proposal is provided by our deuterated hydroxyl MS/MS experiments; loss of D2O to produce the m/z 172 peak holds across all analytes examined in the present study (Figure S2). Our theoretical data predict that the lowest energy pathways to loss of water all include the anomeric oxygen. However, the exact structural specifics of this reaction are predicted to vary as a function of analyte type and anomeric configuration (Scheme 1, Scheme S1, Figure 4, Tables 1, 2, and 3, Tables S1–S3).

Scheme 1
scheme 1

Predicted, lowest energy mechanisms for water loss (B1 ion formation) from the anomeric center of lithiated monosaccharides: (a) α-glucose and α-galactose anomers, (b) β-glucose and β-galactose anomers, (c) β-mannose anomer, and (d) α-mannose anomer

Full size image
Figure 4
figure 4

Transition state structures for water loss from (a) [α-glucose+Li]+, (b) [α-galactose+Li]+, and (c) [α-mannose+Li]+, (d) [β-glucose+Li]+, (e) [β-galactose+Li]+, and (f) [β-mannose+Li]+

Full size image
Table 1 Relative Energies of the Minima, Transition Structures, and Separated Products of Lithiated Glucose (β-D-Glucopyranosyl) Calculated at the M06-2X/6-31+G (d,p) Level of Theory. GM Is the Global Minimum of Potential Energy Surface of [β-D-Glucopyranosyl+Li]+
Full size table
Table 2 Relative Energies of the Minima, Transition Structures, and Separated Products of Lithiated Galactose (β-D-Galactopyranosyl) Calculated at the M06-2X/6-31+G(d,p) Level of Theory. GM Is the Global Minimum of Potential Energy Surface of [β-D-Galactopyranosyl+Li]+
Full size table
Table 3 Relative Energies of the Minima, Transition Structures, and Separated Products of Lithiated Mannose (β-D-Mannopyranosyl) Calculated at the M06-2X/6-31+G(d,p) Level of Theory. GM Is the Global Minimum of Potential Energy Surface of β-D-Mannopyranosyl
Full size table

For all α-anomers, the water loss reaction is initiated by proton transfer to the anomeric hydroxyl group from the C2 hydroxyl (Scheme 1). For the glucose and galactose forms (Scheme 1a, Figure 1a, b), this proton transfer is accompanied by concerted transfer of the Cα-H of C2 to the anomeric center (C1) and cleavage of the glycosidic bond. The net result is formation of a ketone at C2 (B1 ion structures) and loss of water. Formation of the lithiated ketone B1 ion structures from [α-glucose+Li]+ and [α-galactose+Li]+ requires at least 206 or 205 kJ mol−1 through an entropically favorable rate-limiting TS (Figure 4, Tables S4 and S5). [α-Mannose+Li]+, in contrast, proceeds through a zwitterionic oxacarbenium TS and intermediate (Scheme 1d, Figure 4c). The rate-determining TS for the lowest energy lithiated 1,2-anhydromannose B1 ion formation reaction from [α-mannose+Li]+ is substantially more energetically demanding (≥ 234 kJ mol−1, Figure 4c, Table S6). This reaction initially forms a zwitterionic species in which an oxacarbenium functionality is adjacent to the Li+ coordinated hydroxide at carbon 2 (Scheme 1d). Nucleophilic attack of hydroxide into the electropositive carbon 1 then forms the 1,2-anhydromannose B1 ion as the water molecule departs. The alternate 1,2-H shift ketone-forming reaction is initially blocked for [α-mannose+Li]+ by the change in stereochemistry at carbon 2 relative to α-glucose and α-galactose. Consequently, a ketone product is not directly formable. However, it is possible to form the ketone B1 ion from the dimer generated after cleavage of the anomeric C1–OH2+ bond (Scheme 1d, Figure S8). In the dimer, the non-covalently bound water molecule contains the hydroxyl group formerly at the anomeric center. The barrier to the ketone-forming 1,2-H shift reaction within the dimer is lower (227 kJ mol−1, Table S6, Figure S8) than the preceding C1–OH2+ bond cleavage barrier and is entropically favorable (44 J K−1 mol−1). Thus, provided the water molecule is not expelled immediately following C1–OH2+ bond cleavage, the ketone isomer is likely to be competitive. Similar types of rearrangements in post-cleavage dimers have been reported for peptides [53,54,55,56,57].

The β-anomers of lithiated glucose and galactose show distinct water loss pathways from the α-forms. These reactions are initiated from skew structures which facilitate nucleophilic attack by O3 into C1 with concerted transfer of the C3 hydroxyl proton to the anomeric oxygen as the glycosidic bond is cleaved (Scheme 1, Figure 4d, e). Lithiated 1,3-anhydroglucose and 1,3-anhydrogalactose B1 ions are thus generated through comparatively low-energy, but entropically poor, hindered [58,59,60,61] transition structures (Tables 1 and 2; Figure 4d, e). These mechanisms are similar to those described previously by Chen et al. for sodiated glucose [35]. We note that production of a 1,4-anhydrogalactose B1 ion might be expected from lithiated galactose. This possibility was tested but our calculations predict a higher energy barrier. In contrast, the [β-mannose+Li]+ precursors are predicted to expel water from the anomeric center following proton transfer from the C2 hydroxyl group again producing a lithiated 1,2-anhydromannose B1 ion in the process (Figure 4, Scheme 1c). The lowest energy form of this reaction requires at least 195 kJ mol−1 and is sterically hindered (ΔS298K = −4.4 J K−1 mol−1, Table 3). An additional two, energetically more demanding (~ 6–15 kJ mol−1), but entropically more favorable TSs of this type were also located. These structures will become increasingly more competitive as the gas phase in population becomes more energized [1, 58,59,60,61,62].

Our lowest energy calculated transition structures for both the α- and β-glucose analytes differ from those previously proposed [34]. We also located those transition structures [34], as well as many others not highlighted here (including non-anomeric oxygen losses), but these are less competitive (higher relative energies) based on our data.

Cross-Ring Bond Cleavage Transition Structures: the An–Xm Pathways

Experimentally, all three analytes form both 0,2A1 and 0,3A1 ions but with differing onsets and propensities. Our calculations indicate that the mechanisms of formation of the 0,2A1 ion for lithiated glucose and galactose are similar to those previously proposed for larger systems [1,2,3] (Scheme 2). The ring opening occurs simultaneously to proton transfer from the anomeric hydroxyl group to the ring oxygen to form an aldehyde at C1 and a hydroxyl group at C5 from the hemiacetal groups (Scheme 2, Scheme S2). The barriers to ring opening vary with both anomeric configuration and specific monosaccharide. For example, the α-glucose and α-galactose congeners have lower barriers to ring opening than the β-forms, whereas for [mannose+Li]+, this situation is reversed (Figure 5, Tables 1, 2, and 3, Tables S4–S6). Similarly, unlike the larger sodiated systems investigated previously by our group [1, 2], the rate-determining step for 0,2A1 ion formation is not universally the ring-opening TS. This again varies with both anomeric configuration and specific monosaccharide (Figure 5, Tables 1, 2, and 3, and Tables S4–S6). The second, potentially rate-limiting barrier to 0,2A1 ion formation is cleavage of the bond between C2 and C3. Concerted expulsion of 1,2-ethene-diol occurs along with the carbon-carbon bond cleavage and proton transfers (Scheme 2, Scheme S2, Figure S9 and S10), consistent with both the current (Figure S2, Tables S1–S3) and earlier labeling data [1,2,3, 33,34,35]. For the [mannose+Li]+ forms, both the rate-determining TSs require more energy to populate than the glucose forms, consistent with the lower initial abundance of cross-ring cleavage peaks for these analytes (Figure 2c). Additionally, the [mannose+Li]+ forms can expel either a cis or a trans 1,2-ethene-diol with similar barriers (210–214 kJ mol−1), whereas the other hexoses eliminate the cis form preferentially (Table 3, Table S6, Figure S11).

Scheme 2
scheme 2

Mechanism for ring opening followed by 0,2A1 ion formation illustrated for lithiated glucose

Full size image
Figure 5
figure 5

Summarized energetics for 0,2A1 ion formation: (a) [glucose+Li]+, (b) [galactose+Li]+, and (c) [mannose+Li]+. TS1 = ring opening; TS2 = C2–C3 bond cleavage

Full size image

Formation of the 0,3A1 ions is also predicted to begin with ring opening at the hemiacetal (Scheme 3, Figure 6). Direct loss of C3H6O3 from this structure is energetically unfavorable, so instead a further isomerization reaction occurs prior to cleavage of the bond between C3 and C4. The isomerization involves an energetically demanding 1,2-H shift by the Cα-H of C2 to C1 (Figure 6). The anomeric oxygen simultaneously abstracts a proton from the C2 hydroxyl group to leave a ketone at C2. The resulting isomer is the direct precursor for 0,3A1 ion generation. The final covalent bond cleavage stage of this reaction then involves a complex concerted reaction. Transfer of two hydroxyl protons and concerted carbon-carbon bond cleavage (retro-aldol reaction [3]) results in generation of a lithium-bound dimer consisting of 2,3-dihydroxypropanal and (Z)-prop-1-ene-1,2,3-triol. Despite the dimer partners being isomers (C3H6O3), our calculations predict that the 2,3-dihydroxypropanal will dominantly retain the Li+ in agreement with lithium affinity calculations, thereby producing the 0,3A1 peak. This agrees with the loss of C3H3D3O3 (HC(OD)=C(OD)–H2COD, Figure S2, Scheme S3) in our deuterated hydroxyl labeling experiments and the other regiospecific labeling data (Tables S1–S3). For larger sodiated systems, the analogous highly strained 1,2-H shift was found to be the rate-limiting step to 0,3A2 formation [2]. For the lithiated monosaccharides discussed here, this is not uniformly the case. This makes broad statements governing all analyte forms difficult. However, for all lithiated analytes, the ring-open products are entropically favored over the pyranose ring forms. Consequently, these reactions will be increasingly facile once ring opening has been achieved. In most cases, the ring-opening TS is rate limiting enthalpically. Furthermore, even in those cases in which a slightly higher barrier exists after ring opening along the reaction coordinate, the relatively low ΔS298K of the ring-opening TS likely limits [2, 58, 61, 62] the progress of the reaction. Once ring opening is complete and sufficient energy is available for subsequent degradation, the branching ratio between the 0,3A1 and 0,2A1 peaks is a function of the relative entropic favorability of these two processes (and indirectly the stability of the A1 ion products) so ΔG298K is the pertinent measure of the reaction favorability.

Scheme 3
scheme 3

Mechanism for ring opening followed by 1,2-H transfer then 0,3A1 ion formation illustrated for lithiated glucose

Full size image
Figure 6
figure 6

Summarized energetics for 0,3A1 ion formation: (a) [glucose+Li]+, (b) [galactose+Li]+, and (c) [mannose+Li]+. TS1 = ring opening; TS2 = 1,2-H transfer; TS3 = C3–C4 bond cleavage

Full size image

Conclusions

There are broad similarities in the fragmentation chemistry of lithiated glucose, galactose, and mannose, but also structural differences. There are also differences based on anomeric configuration. For example, while all analytes expel a water molecule from the anomeric center at low collision energies, the product ion structure differs between the α- and β-forms for glucose and galactose (lithiated C2 ketones from the α-forms vs. 1,3-anhydrohexose isomers from the β-forms). The dissociation chemistry of both mannose forms is significantly affected by the hydroxyl stereochemistry at carbon 2, which results in production of lithiated 1,2-anhydromannose from both precursor types. Additionally, provided water loss is not instantaneous, the α-anomer can also isomerize to generate a ketone structure at C2 through a concerted 1,2-migration of the of the C2-H to C1. The resulting product is indistinguishable to that formed from [α-glucose+Li]+. All analytes investigated form both 0,3A1 and 0,2A1 ions in mechanisms substantially (though not necessarily solely) limited by the entropically relatively poor ring-opening transition structures. The lowest energy A1 ion-forming mechanisms are consistent with those advocated previously in the literature [2, 3, 33] and our own labeling data.