1 Introduction

Carbohydrates and their derivatives are widespread organic compounds in nature, and they play important roles in biological processes. They contribute to the source of energy for life activities and also interact in the biological functions of cell growth, fertilization, inflammation, and cell–cell interactions [1–4]. The number of possible linear and branched carbohydrate isomers is immense, and has been referred to as the isomer barrier [5]. For instance, there are eight possible hexoses in the pyranose form, which differ in the chirality of the hydroxyl group at C-2, C-3, and C-4. Furthermore, for disaccharides and oligosaccharides, variations in saccharide building blocks and sequence, as well as glycosidic linkage positions and anomericity, give rise to a large number of putative isomers. Scheme 1 illustrates the nomenclature for the example of two glucose saccharide units connected by an α1-4 glycosidic bond.

Scheme 1
scheme 1

Chemical structure of Glcα1-4Glc glycosidic bond in an oligosaccharide

Full size image

Carbohydrates thus represent an extreme case of isomeric heterogeneity, requiring analytical techniques that can address this perplexing but crucial challenge. Nuclear magnetic resonance (NMR) is generally considered to be the gold standard in elucidating carbohydrate structures [6–8]. Mass spectrometry (MS) offers the advantages of lower requirements for analysis time, sample quantity, and purity. A key limitation of mass spectrometry, however, is its inability to distinguish isomers based on mass information alone. Strategies for distinguishing isomers have, thus, focused on distinguishing other physical properties [9], such as the collision cross sections of ions, as measured in ion mobility [10–17] experiments, or their fragmentation patterns in collision-induced dissociation (CID) [18–23], electron-mediated dissociation [24–26], or laser photodissociation [27–36].

Tunable laser photodissociation allows “action” spectra of ions to be recorded [37]. The premise of this approach rests on resonant absorption of photons to induce photodissociation. Laser spectroscopy combines the structural insights from molecular absorbances with insights from the dissociation chemistry and, thus, provides a wealth of information to distinguish isomers. In wavelength-tunable infrared multiple-photon dissociation (IRMPD) of disaccharide isomers, the fragmentation patterns could be considered spectral fingerprints [29, 34].

The charge carrier plays an important role in directing the dissociation chemistry. Protonated oligosaccharides are prone to cleaving exclusively at glycosidic bonds, providing the least structural information [38]. Metal-adducted complexes have been shown to exhibit advantages based on higher ionization efficiencies and more informative fragmentation [14, 20, 39]. For alkali metal adducts, it has been shown that Li+ is bound tightest and causes most abundant cross-ring fragmentation [33, 40], rationalizing the prevalent use of lithiated oligosaccharides in positive-mode oligosaccharides MS analysis [18, 19, 29, 30, 34, 41]. For the negative-mode MS analysis, previous studies have shown that Cl– adducts are highly stable and yield high abundances of fragment ions [42–45]. In addition, deprotonated oligosaccharide precursors can be formed efficiently in electrospray ionization (ESI) in the presence of a basic modulator, such as aqueous ammonium acetate (NH4Ac) or ammonium hydroxide (NH4OH) [46, 47]. Finally, chemical modification of oligosaccharides, such as permethylation or chemical modification of the reducing end, assisted by multiple stages of mass isolation and CID (i.e., MSn), has been demonstrated as a promising avenue to characterize the glycosidic bonds in larger oligosaccharides [23, 48, 49].

Note that in the differentiation of isomers in mixtures, separation techniques prior to structural elucidation by dissociation greatly simplify the analysis. Separations can either be carried out in solution by chromatography [50] or in the gas phase by ion mobility [14]. In some information-rich fragmentation data, it may be possible to identify particular isomers in mixtures based on distinctive features, without the need for separation [51]; however, this often involves asking specific structural questions, as well as having some knowledge of the sample that is being analyzed.

In this work, we implement variable-wavelength IRMPD-MS to characterize the glycosidic bonds in unmodified trisaccharides. In a stepwise approach, the trisaccharides are dissociated to disaccharide units, whose rich wavelength-dependent photodissociation patterns can verify the linkage position and anomericity of the disaccharide unit.

2 Experimental

2.1 Materials

Disaccharides (C12H22O11) and trisaccharides (C18H32O16) comprised of solely D-glucose (Glc) were used, and their nonreducing monosaccharide substructures were in the pyranoside form. Nigerose (Glcα1-3Glc), maltose (Glcα1-4Glc), cellobiose (Glcβ1-4Glc), isomaltose (Glcα1-6Glc), gentiobiose (Glcβ1-6Glc), laminaritriose (Glcβ1-3Glcβ1-3Glc), maltotriose (Glcα1-4Glcα1-4Glc), cellotriose (Glcβ1-4Glcβ1-4Glc), and panose (Glcα1-6Glcα1-4Glc) were purchased from Sigma Aldrich (St. Louis, MO, USA). Laminaribiose (Glcβ1-3Glc), kojibiose (Glcα1-2Glc), and sophorose (Glcβ1-2Glc) were purchased from Carbosynth (Compton, Berkshire, UK). 18O-labeled water (H2 18O) (97 atom % 18O) was also purchased from Sigma Aldrich (St. Louis, MO). All reagents were used without further purification. Saccharide solutions were prepared at a concentration of 0.1 mM in MeOH with adding 1% volume of NH4OH (29.44 % by mass) solution.

2.2 18O-Labeling of Trisaccharides

Laminaribriose (Glcβ1-3Glc) 1.5 mg was dissolved in 100 μL H2 18O at room temperature for 5 to 7 d to label the reducing end with 18O. The solution was then diluted to 0.1 mM with MeOH containing NH4OH prior to mass spectrometric analysis; 1.5 mg of maltotriose (Glcα1-4Glcα1-4Glc) was also labeled using the same procedure.

2.3 Mass Spectrometry and Infrared Multiple Photon Dissociation (IRMPD)

The saccharide samples were ionized in a commercial electrospray ionization (ESI) source (Analytica of Branford, Branford, CT, USA), coupled to a Bruker 47e FTICR (Fourier transform ion cyclotron resonance) mass spectrometer (Bruker Daltonics, Billerica, MA, USA) equipped with a 4.7 T superconducting magnet (Magnex Scientific Ltd., Abington, UK). The negative ions were trapped in an Infinity cell and irradiated with a Lasy-20G tunable continuous wave CO2 laser (Access Laser Co., Marysville, WA, USA), as controlled by a mechanical shutter.

To study wavelength-dependent fragmentation patterns, the disaccharide standards were irradiated with the CO2 laser for 500 ms at a power of 1.5 W. Because of two distinct frequency bands of the CO2 laser emission, the laser was tuned over two separate wavelength bands (9.2–9.7 μm and 10.2–10.7 μm) at 40 specific wavelength channels (see Supplementary Table S-1). Trisaccharides, on the other hand, were broken down stepwise in a two-part experiment. Gas-phase trisaccharide ions were first photodissociated with the CO2 laser for 50 ms at 9.588 μm with a power of 1.5 W, yielding an abundant disaccharide fragment (at m/z 341); as shown in Supplementary Figure S-1, the ion signal for C2 was optimized under those conditions. The output of the CO2 laser was then switched to a variable wavelength (9.2–9.7 μm and 10.2–10.7 μm), and after a 4-s time delay (to allow stabilization of the laser output), the mass-isolated disaccharide fragment ions were further photodissociated at this variable wavelength for 500 ms, under the same conditions as the standard disaccharide experiments. Note that the output wavelengths of a CO2 laser are limited by the rovibrational P and R branches, and that the output wavelength is selected by virtue of a diffraction grating in the laser cavity. The voltages and timing in the experiment were under control of the Predator program, developed by the National High Magnetic Field Laboratory (NHMFL). TTL (transistor–transistor logic) pulses from the Predator program were employed to trigger the mechanical shutter, as well as an in-house LabView software, which controlled the CO2 laser output wavelength and power. In each experiment, a set of 10 mass spectra were collected and averaged.

Data analysis was performed with the aid of another in-house LabView program. The normalized abundance ratios of each mass spectral product fragment were calculated and averaged with three replicate measurements. Normalized abundance ratios were expressed by the following equation:

$$ \mathrm{normalized}\ \mathrm{abundance}\ \mathrm{ratio}=\frac{\mathrm{fragment}\ \mathrm{intensity}}{{\displaystyle \sum \mathrm{fragment}\ \mathrm{intensities}+\mathrm{precursor}\ \mathrm{intensity}}} $$
(1)

The normalized abundance ratios (0.0050–0.8000) of each fragment as a function of infrared wavelength were plotted as heat maps with a color distribution from pure blue (0.0050) to pure red (0.8000) on a logarithmic scale. Ratios in the trace quantity (0–0.0050) were also presented with pure blue.

3 Results and Discussion

3.1 Photodissociation Patterns of Disaccharides

Figure 1 shows the wavelength-variable photodissociation patterns of eight isomeric glucose-containing disaccharides in the form of heat maps. For each disaccharide, the normalized abundance ratios of the observed photodissociation mass channels are given. The abundance ratios of IRMPD fragment ions exhibit considerable differences over the wavelength range of the CO2 laser. Those ions generated by glycosidic cleavage (i.e., m/z 161 and 179) and other high-mass fragments are dominant in the higher wavelength range from 10.2 to 10.7 μm. In contrast, low-mass fragments, including m/z 101, 113, and 115, yield high abundances in the lower wavelength range from 9.2 to 9.7 μm. The higher extent of fragmentation, and likely sequential fragmentation, in the latter range can be explained by strong photon absorbance attributable to C–O stretch vibration in this range.

Figure 1
figure 1

Heat maps of wavelength-dependent infrared photodissociation of isomeric disaccharide standards, showing the normalized abundance ratios of the photofragment mass channels as a function of CO2 laser wavelength. The abundance ratios (0.0050–0.8000) are represented in the color range between pure blue (low) to pure red (high) on a logarithmic scale. Nonobservable photofragment channels are denoted by white

Full size image

Previous studies have demonstrated the use of diagnostic fragment ions that can identify the linkage position of the glycosidic bond for the case of lithiated disaccharides [18, 19, 29, 34]. Table 1 summarizes the appearance of various photodissociation products in the case of deprotonated disaccharides (based on the data shown in Figure 1). For instance, m/z 263 is a unique ion produced from 1-2-linked disaccharides, whereas the ions, m/z 191 and 215 are generated exclusively from dissociation of disaccharides with a 1-3 linkage. The fragments at m/z 221 and 281 are obtained as diagnostic product ions from photodissociation of the two 1-6 linkage anomers of isomaltose (Glcα1-6Glc) and gentiobiose (Glcβ1-6Glc). As a general observation, the larger photofragments from m/z 191 to 323 are found to be diagnostic, whereas the smaller fragments except m/z 149 appear for all isomers.

Table 1 Appearance of Photodissociation Fragment Ions for Isomeric Deprotonated Disaccharides (Diagnostic Fragments Indicated by Shading)
Full size table

Besides specific product ions, diagnostic ratios of key ions over variable wavelengths contain much information in order to differentiate the linkage and anomericity of disaccharide isomers. For the 1-4-linked maltose (Glcα1-4Glc) and cellobiose (Glcβ1-4Glc), which are the only pair with identical product ions, the wavelength-dependent abundance ratios of specific low-mass fragments such as m/z 101 and 113 show some differences at the IR range of 10.2–10.7 μm and provide the means to differentiate them. Also, the normalized abundance ratio of m/z 161 fragment of maltose (Glcα1-4Glc) varies from 0.3 to 0.4 at the range of 10.2–10.4 μm, whereas at the same range the abundance ratio of m/z 161 is about 0.5–0.8 for cellobiose (Glcβ1-4Glc).

3.2 Reproducibility

It is worth noting that the reproducibility of these heat maps depends on a number of experimental parameters, including the alignment of the laser, the laser output power, and the irradiation time. Consistent laser alignment and irradiation time were set up, and output power was kept at 1.50 ± 0.05 W for disaccharide standard experiments. As an example, the day-to-day difference between two wavelength-variable photodissociation experiments for isomaltose (Glcα1-6Glc) is displayed in Supplementary Figure S-2, showing the measurements on two separate days (i.e., measurements 1 and 2). The two measurements are found to be strikingly similar. In order to illustrate the similarities of both measurements, the results are also depicted as a difference heat map, where the numerical results for measurement 2 are subtracted from those in measurement 1. If the patterns matched exactly, the whole figure would be green (i.e., no or minor differences). Indeed, the majority of the difference heat map is green, and only negligible differences are seen, mainly for the m/z 101 product, showing that the experimental conditions can be reproduced faithfully.

In order to better illustrate the fidelity of relative photofragment channel abundances in terms of differentiating isomeric species, selected photofragment channels for maltose and cellobiose are shown in Supplementary Figure S3. The mean and standard deviation values from triplicate measurements indicate that many of the differences in photofragment abundances are reliably reproduced, allowing those isomeric species to be told apart.

3.3 Dissociation of Trisaccharides

When glucose-containing trisaccharides (m/z 503) are irradiated and dissociated by a CO2 laser, of particular interest is the disaccharide fragment at m/z 341 (C12H22O11). To determine whether, in the negative mode, m/z 341 originates from cleavage of either the reducing or nonreducing termini of the trisaccharide precursors (i.e. C2 or Y2 fragments according to Domon and Costello’s systematic nomenclature [52]) (see Scheme 2), the carbonyl oxygen atom of the reducing end was labeled with 18O [21, 53]. With the laser irradiation by the CO2 laser at a fixed wavelength of 9.588 μm, the photodissociation spectra of 18O-labeled [laminaritriose-H]– (m/z 505) and 18O-labeled [maltotriose-H]– (m/z 505) were obtained and are depicted in Figure 2. Both spectra of 18O-labeled [laminaritriose-H]– and [maltotriose-H]– show that exclusively the m/z 341 product ion was observed, whereas there was no evidence for the corresponding fragment at m/z 343. In other words, these control experiments confirmed that the trisaccharides incurred cleavage between the second and third residue, giving rise to the C2 fragment, and not to the isomeric Y2 product ion. As a side note, m/z 161 and 163 are observed as abundant fragments for [maltotriose-H]–, whereas only m/z 161 is seen for [laminaritriose-H]–. The m/z 161 and 163 ions can be rationalized by B1 and Z1, respectively. The Z1 fragment is, in principle, the complementary fragment from C2. It is not clear whether B1 occurs from direct dissociation of the trisaccharide precursor or consecutive dissociation of C2.

Scheme 2
scheme 2

Dissociation of 18O-labeled trisaccharide to C2 and Y2 fragments

Full size image
Figure 2
figure 2

The photodissociation mass spectra for 18O-labeled [laminaritriose-H]– (a), (c) and 18O-labeled [maltotriose-H]– (b), (d)

Full size image

3.4 Photodissociation of Disaccharide Units Derived from Trisaccharides

As the key fragment at m/z 341 (C12H22O11) corresponds to a C2 ion, a more detailed structural probing of C2, in principle, allows one to establish the nature of the glycosidic bond between residues 1 and 2 in the original trisaccharide precursor. For this purpose, C2 ions from various trisaccharides were subjected to variable-wavelength photodissociation, and these results were compared with reference wavelength-dependent photodissociation patterns of disaccharides. Scheme 3 illustrates the strategy of this two-step approach, involving fixed-wavelength irradiation to generate C2 from the trisaccharide precursor, followed by variable-wavelength photodissociation of the C2 ion.

Scheme 3
scheme 3

Stepwise fragmentation for deprotonated trisaccharides

Full size image

Control experiment showed that generation of the C2 fragment was optimized for short irradiation times (i.e., 50 ms) at a moderate power (i.e., 1.5 W). As shown in Supplementary Figure S-1, under these conditions the yield of C2 was maximized at a wavelength of 9.588 μm. In the stepwise dissociation scheme above, the fixed wavelength of the first irradiation was thus set to 9.588 μm, followed by mass isolation of the C2 ion, and a variable wavelength for the second irradiation. For the second irradiation, since the wavelength irradiation time and power are critical factors in determining the dissociation patterns, these parameters had to be kept consistent with the conditions used for the disaccharide standards. It was found that a 4-s time delay between the first and second irradiations minimized fluctuations when switching the laser wavelength.

Figure 3 compares heat maps for wavelength-dependent photodissociation patterns of C2 ions (from four isomeric trisaccharides) with reference photodissociation patterns of disaccharide standards. As expected, heat maps of the C2 ions match most closely those disaccharide standards with the same chemical structures. For instance, C2 from laminaritriose (Glcβ1-3Glcβ1-3Glc) has the same glycosidic linkage and fragmentation pattern as laminaribiose (Glcβ1-3Glc), whereas C2 from panose (Glcα1-6Glcα1-4Glc) is identical to isomaltose (Glcα1-6Glc). Even differences in chirality, such as in the case of maltotriose (Glcα1-4Glcα1-4Glc) and cellotriose (Glcβ1-4Glcβ1-4Glc), are well differentiated in these measurements, as some distinct patterns in the dissociation products of m/z 101, 113, 119, and 161 are reproduced. As an example of how closely the patterns resemble each other, the heat maps for C2 from panose (Glcα1-6Glcα1-4Glc) and the disaccharide isomaltose (Glcα1-6Glc) are subtracted from each other and are represented as a difference heat map (see Supplementary Figure S-4). In this case, the wavelength range from 10.2 to 10.7 μm is largely green, light blue, or yellow and, thus, reflects only minor differences. There are, however, some notable differences in the range from 9.6 to 9.8 μm for the m/z 113 and 131 fragments. These differences are a manifestation of different conditions in the experiment leading to irreproducibility. As a general observation, the wavelength region from 10.2 to 10.7 μm appears to be more reliable and, thus, likely to be more useful in terms of pattern recognition. In the lower wavelength region, the range from 9.3 to 9.4 μm also appears to be well reproducible and yields complementary information.

Figure 3
figure 3

Heat maps of wavelength-dependent infrared photodissociation of C2 fragments derived from trisaccharides (left) and matching disaccharide standards, showing the normalized abundance ratios of the photofragment mass channels as a function of CO2 laser wavelength. The abundance ratios (0.0050–0.8000) are represented in the color range between pure blue (low) to pure red (high) on a logarithmic scale. Nonobservable photofragment channels are denoted by white

Full size image

These measurements also have important repercussions on the chemistry that takes place in sequential fragmentation of saccharides. While nominally, a C2 fragment from a larger oligosaccharide corresponds to the same building block as a reference disaccharide, the present experiments in fact give strong evidence that these structures are chemically equivalent. The dissociation of larger entities into fragments may, in principle, proceed via a number of chemical pathways involving rearrangement reactions and, thus, giving rise to complex product ion structures; complex rearrangement chemistry has been attested in numerous studies on peptide fragmentation pathways [54, 55], which can complicate structural analysis. It appears that such concerns are not merited in the case of oligosaccharide dissociation, where the glycosidic bond and other chemical moieties are conserved upon ion activation [21].

Finally, although this study has confirmed that disaccharide building blocks can be identified based on their photofragmentation patterns, these proof-of-principle experiments were carried out for pure solutions of the analytes in question. The analysis of real-world, complex mixtures requires separation techniques prior to structural probing, as already alluded to in Section 1.

4 Conclusions

In this study, a series of isomeric forms of glucose trisaccharides were ionized in negative ion mode (as deprotonated anions), and were characterized by wavelength-variable infrared photodissociation using a tunable CO2 laser. The main conclusions from this work were:

  • Photodissociation of trisaccharide ions yielded effective glycosidic cleavage between the second and third residues (i.e., C2 fragments), as verified by 18O isotope labeling on the reducing end.

  • C2 photofragment yield was maximized for short (50 ms) irradiation at 9.588 μm.

  • Variable-wavelength photodissociation of these C2 ions resulted in characteristic photodissociation patterns represented here as heat maps. These heat maps were found to be almost identical to those generated in control experiments for corresponding disaccharide ions, indicating a close structural similarity between both.

The stepwise dissociation employed in these measurements allowed a structural elucidation of linkage and anomericity of glycosidic bonds in trisaccharides; this strategy also provides a step towards a reliable and practical tandem MS analysis of oligosaccharides, where their substructures can then be confirmed by variable-wavelength IRMPD, particularly if combined with chemical labeling of the reducing end.