CHAPTER 1: Intramolecular Hydrogen Bonding in Glycans in Aqueous Solution

Researchers in glycan structure determination spend a lot of time chasing their own tails. Because the molecular structures of glycans are more flexible than proteins and DNA, it is sometimes difficult to definitively delineate glycan ″structure″. A perusal of the literature will give one the impression that, for a given glycan in solution, multiple structural possibilities are generally predicted by modeling; however, it is challenging to determine which possibilities are the most reasonable due to ambiguities present in both experimental and modeling data. Thus, solution studies, especially NMR spectroscopy, should aim to provide far more definitive, quantitative parameters to improve model parameterization. Consequently, computer modeling can be more predictive than it is presently. Naturally, this will lead to a more detailed and complete understanding of the forces that dictate glycan structure and conformations. Ultimately, it may even provide a framework to understand and predict chemical reactivity of glycans. One extremely important structural parameter known to contribute to structural stability is hydrogen bonding (Hbonding) in aqueous solution and it is the focus of this chapter.

Hbonding is a vital weak inter/intra-molecular interaction, which is especially important in biological molecules. On one hand, Hbonds can be seen as a putative proton transfer; on the other, Hbonds can stabilize biological structural motifs in proteins, DNA and RNA without proton transfer.

The strength of Hbonds ranges from 1–40 kJ mol⁻¹.¹  Though strong for a non-bonded interaction, it is only a fraction of the strength of a typical covalent bond. Thus, in many biomolecules multiple Hbonds are used to increase the stability of different structural motifs in proteins and nucleic acids.

Hbonding is defined as an interaction between an electronegative atom and a hydrogen atom covalently bonded to an electronegative atom, such as fluorine, oxygen, nitrogen and sometimes carbon. The electrons in the X–H polar covalent bond are delocalized toward the former of the two electronegative atoms (Figure 1.1).

As stated previously, Hbonding is a recurrent structural feature found in proteins and nucleic acids. We were convinced that Hbonds are important in glycans too. Therefore, we sought evidence in support of Hbonding in oligo- and polysaccharides and the methods we developed may be useful in other applications as well. Hbonding in solution has only recently been directly detected in glycans using NMR.^2–7  Details of Hbonding motifs are different in proteins and nucleic acids; therefore, we reasoned that Hbonding in glycans may also differ from any other class of biomolecule.

Hbonds in glycans may differ due to flexibility, or conformational motion in biomolecules, which may cause Hbonds in glycans to be more transient than they are in proteins and nucleic acids. This in turn hampers detection of Hbonds in glycans. Furthermore, the Hbonds expected in glycans are likely to involve multiple OH groups. These are structurally different than those in proteins and nucleic acids where the NH-based Hbond is far more prominent and also dictates structure. Therefore, NMR pulse programs must be specifically tailored for glycan Hbond detection. These considerations drove us to first detect Hbonds in short helical peptides having metastable Hbonds,⁸  and consequently metastable structure, before attempting to detect metastable Hbonds mimicking those in glycans. X-Ray data for Hbonding in oligo- and polysaccharides can be found in Rao et al.⁹  and Jeffrey and Saenger.¹  In this chapter, we limit the discussion to Hbonding in solution.

In oligosaccharides, individual residues are linked between the anomeric carbon (C1) of one residue and virtually any other carbon of the next unit via an oxygen atom. Three-dimensional conformation is defined by the torsions ϕ and ψ (Figure 1.2). The anomeric signal isolation makes these resonances useful as NMR “handles”. Thus, they provide valuable information for sequential assignment. However, because anomeric proton signals commonly overlap with the water resonance, glycan NMR experiments have traditionally been carried out in D₂O as opposed to H₂O. In early glycan NMR studies, pre-saturation was used to remove residual ¹H signals in the HOD peak. Although these procedures enable the observation of anomeric protons, information on all exchangeable protons in the molecule, such as OHs and NHs, is lost through chemical exchange with the water, whose magnetization is saturated. As we will see later, these exchangeable protons could otherwise be used to gather structural data.^10,11  More recently, water suppression in NMR spectra can be achieved by ensuring that water (H₂O, as opposed to D₂O) magnetization remains or ends up on the z-axis and, therefore, with the help of gradients, is not detected. Figure 1.3 shows a comparison of sucrose spectra with no water suppression, presaturation and modern water suppression.

Hbonding from amide NH groups in proteins has been known for many years. In solution, it was usually inferred from H/D exchange in solution NMR spectroscopy or temperature-coefficients of the amide NH chemical shifts. In this experiment, solvent exposed NH groups readily exchange with water.^12,13  Therefore, the temperature coefficient of NH residues' chemical shift under these conditions closely mimics that of water (∼0.09 ppm per °C), unless the NH group is involved in Hbonding. In contrast to solvent exposed NHs, the exchange rate of Hbonded NH groups is slowed. Although the temperature coefficient method is useful, direct evidence of Hbonds would be more desirable.

In 1999 Cordier et al. and Wang et al. reported the direct detection of Hbonds and measurement of ^3hJ_NC' couplings in fully ¹⁵N,¹³C enriched proteins.^2,6  This allowed direct spectral assignments of Hbond donors and acceptors. The NMR spectral correlation is due to N–H⋯O=C orbital overlap of amide NH with the complementary carbonyl oxygen's orbitals. Similar data were reported for nucleic acids by Dingley et al., though not for amide NHs.³  Significantly, this method facilitated the measurement of coupling constants in these Hbonds, which were found to be in the range of 0.2–1.0 Hz. Later, the couplings were shown to correlate to Hbond distances from crystal structures and indicate Hbond strength.⁷  This showed that Hbonding results in orbital overlap and that it could be detected in aqueous solutions of proteins and nucleic acids, even when the J-coupling is small.

Widmalm and co-workers predicted that the structure of the E. coli O142 polysaccharide (PS) should contain a potentially detectable level of Hbonding. These authors based their predictions on experimentally determined NOE's and restrained MD simulations of the pentasaccharide, which is the repeat unit of the O142 PS.¹⁴  The Hbonds were predicted to extend from three NH groups in GalNAc-A, GalNAc-B and GalNAc-C rings to oxygen atoms. However, the Hbond with the highest percentage was predicted to be present in only 20% of the pentasaccharide structures by MD, while the others were predicted to be present in smaller amounts.

Inspired by the results in proteins, Norris et al. attempted to detect similar correlations in the E. coli O142 PS.¹⁵  Unfortunately, direct (through-bond) evidence for the Hbond in this PS was difficult to demonstrate, probably due to exchange broadening combined with a low percentage of amides NH's in the Hbonded state. Nevertheless, these authors reported a measurable temperature dependence of the NH chemical shifts of GalNAc-A, GalNAc-B and GalNAc-E in the O142 PS. This temperature dependence differed from that of H₂O and supported the presence of Hbonding for these three amide NH groups.

Another metric used in the O142 study was the temperature-dependence of the heteronuclear NH J-coupling. In principle, this works like the temperature-dependence of the NH chemical shift. Similar to chemical shifts, coupling constants are averaged as a function of temperature. While the ¹H is localized on the ¹⁵N, the ¹J_NH coupling reflects the average of the coupling of the ¹H to the ¹⁵N while bound to ¹⁵N (∼92 Hz) and the value when it is not bound to ¹⁵N (0 Hz). As the temperature rises and ¹H exchange with water becomes faster, the ¹H lifetime on the ¹⁵N is shorter hence the averaged coupling is reduced. In an Hbonded NH group, the ¹J_NH would be expected to change little as a function of temperature because Hbonding would preclude it from exchanging with water. The ¹J_NH temperature dependence confirmed that GalNAc-A and GalNAc-B are Hbonded for at least part of the time, while GalNAc-C is not, allowing the authors to assign the reduced temperature coefficient displayed by the NH in GlcNAc-E to chemical (conformational) exchange rather than Hbonding.

The ¹J_NH coupling method is more sensitive than chemical shift temperature coefficients to the on–off rate of amide NH's and may therefore be more useful to indirectly deduce Hbonding in glycans. As stated previously, in some cases, the ¹J_NH coupling method may also help in distinguishing Hbonding from conformational motion, both of which modulate the chemical shift from NH exchange with water.

Although Hbonding in the O142 PS was not directly demonstrated, the difficulties were not trivial, as conformational motion, PS heterogeneity and potentially low Hbonded populations likely contributed to the direct detection of HN⋯C=O correlations in the O142 PS. Furthermore, isotopic labels did not result in the types of correlations observed by the groups of Grzesiek, Feigon, Barfield, Torchia and Bax for nucleic acids and proteins.^2,4,5 

However, in 2012, Battistel et al. reported the direct observation of three NH-based Hbonds in a tetramer of α2-8 N-acetylneuraminic (sialic) acid, SiA₄.¹⁶  The studies began in a similar vein to the O142 PS studies, by attempting to detect indirect evidence of NH Hbonding. First, a ¹³C,¹⁵N fully labeled SiA₄ was prepared. Then, good chemical shift dispersion in the NH region was achieved at reduced temperatures (below 15 °C). Subsequently, measurements of NH/ND exchange rates using the SOLEXSY pulse sequence¹⁷  showed that three of the four NH's exchanged with H₂O at a rate constant of approximately 1 s⁻¹, while the remaining NH exchanged with a rate constant of 3 s⁻¹. This strongly suggested that three of the four NHs are Hbonded at −10 °C.

Still, the results did not provide direct evidence for Hbonding in SiA₄. Battistel et al. then used an NMR pulse sequence akin to the HNCO devised by Grzesiek and co-workers,²  but with minor changes to optimize the experiment for the chemical shifts in glycans, because, as stated earlier, they differ in structure from proteins and nucleic acids. At first, no direct correlations were observed because the ¹³C carrier frequency had been set to the C=O region. This suggested that the Hbonded NHs, if present, were not bonded to C=O groups. In hindsight, this may not be surprising: solution Hbonding patterns in glycans are not well established and these patterns may differ from those in proteins and nucleic acids. The ¹³C carrier was then set to multiple different values (100 ppm, 80 ppm and 60 ppm) in order to ensure coverage of the entire ¹³C spectrum. The result was a correlation from the ¹⁵N of the N-acetyl groups in residues I, II and III to the C8 of the same residue (Figure 1.4(A)). Although this correlation peak did not incontrovertibly dictate that Hbonds were present in residues I, II and III of SiA₄, they did strongly support their presence and the notion that the NH's are Hbonded to the glycosidic oxygen atoms (O8) in each of these units. Hbonding from the NH⋯O8 in SiA₄ was further confirmed by an experiment similar to the same HNCO. In this experiment a direct correlation from NH_i → C2_i+1, where i = 2, 3, was detected. Thus Hbonding in α2-8 SiA₄ from the NH → O8 in residues I, II, and III was established at −10 °C.

A simple calculation of the structure with only these Hbonding restraints led to a left-handed helical structure where the helix contains two residues per turn. The helix was thus dubbed a 1₂ helix. Interestingly, the torsional angles in this structure violate the exo-anomeric effect.¹⁸  Were the exo-anomeric effect to play a major role in structure for SiA₄, a four-residue per turn helix with a larger diameter would be expected. While the Hbonds and many of the NOE's supported both structures, the mentioned 1₂ helix and a similar helix and a 1₄ helix, the calculated NOEs from each of the computer models for the 1₂ helix and 1₄ helix showed that over 25 of the experimental NOEs are in disagreement with the 1₄ helix structural model, while no such disagreements were found for the 1₂ helix. Models of the resulting structure extended to 10 residues are shown in Figure 1.4. This helix is in agreement with research published by Yongye et al. who found that SiA₃ yielded NOEs and torsions that, together with calculations, predicted a subpopulation of structures that contain a 1₂ helix.¹⁹  The left-handed helix and inter-residue angles are also consistent with crystallographic results by Schulz et al. where SiA₃ and SiA₅ are found in a helical conformation in the active site.²⁰ 

Interestingly, the obtained helical solution disagrees with crystallographic results of SiA octamers²¹  as well as other crystallographic results of SiA oligomers (see Figure 13 in Nagae et al.). It is not yet clear whether the structural features in the crystallographic studies or those found in solution are more relevant for biological function.

These results have significant implications for three-dimensional glycan structure. First, this Hbonding may help in restricting motion about the glycosidic torsions. Second, even short oligomers can adopt a three-dimensional fold, in this case, it is a helical fold reinforced by NH⋯O Hbonding, where the oxygen is glycosidic. Third, even if not completely rigidified at higher temperatures, this helical conformation is likely to be a conformation that is found as part of an equilibrium mixture of conformers. Finally and most importantly, the Hbonding in SiA₄ may be more general and may aid in explaining some PS structure–function relationships, especially in PS-antibody (Ab) binding.

Amide NH's exchange relatively slowly and can therefore be used as handles in many experiments. However, not all glycans contain NH groups, and consequently Hbonding in such glycans cannot be studied as already discussed. Hbonding in glycans can also occur via the various OH groups in glycans. The oxygen in water has tetrahedral bond angles if the lone pairs are considered. Glycans can be likened to water on a carbon stick. Like the oxygen in water, the carbon atoms in glycans are tetrahedral. This allows the OH groups to easily fit into the water's tetrahedral Hbonding structure. Consequently, the OH groups in glycans are extremely similar to H₂O's OH groups and exchange is more facile than for many other exchangeable hydrogen atoms. In a study of these OH groups, Battistel et al. demonstrated that with appropriate water suppression, Hbonding in glycan OH groups could be directly observed.^11,22,23 

The key to observing these OH Hbonds is a combination of modern water suppression (p3919 WATERGATE) combined with reduced temperatures to slow down the OH exchange rate with H₂O. While OH group resonances could be observed at ambient temperatures, the lines were broad in aqueous solution and consequently, correlations were difficult to observe under these conditions.

Batta and Kover showed that C–H resonances could be correlated to the attached OH resonance and the water resonance could be well suppressed using a WATERGATE pulse sequence element with gradients.^24,25  Triple axis gradients provide significantly enhanced water suppression. In studies, carried out using 1.7 M sucrose at −3 °C, these authors were able to assign the OH resonances using a ¹H–¹H TOCSY sequence. Although these studies broke the barrier for assigning OHs, there are approximately 7 molecules of water for each sucrose molecule dissolved in solution at these high concentrations.²⁶  Ideally, one would like to detect and assign the OH groups at lower concentrations.

Battistel et al. showed that an HSQC–TOCSY (10 ms DIPSI-2 mixing time) improved resolution in the indirect dimension by using ¹³C chemical shifts at 300 mM sucrose in a conventional NMR probe.¹¹  This alleviates spectral overlap in the crowded ¹H spectral regions and thus improves the ability to assign OH groups in glycan NMR spectra. Assigning the OH groups now opens the door to the ability to use the OH groups as handles in NOE experiments as well as in magnetization transfer experiments, however, at the present time, it is not easy to see how the latter could be accomplished.

Battistel et al. then showed that the HSQC–TOCSY with WATERGATE could be extended using a longer mixing time during the TOCSY portion (30 ms DIPSI-2) facilitating magnetization transfer across Hbonds in glycans, when present.²⁷  The sucrose Hbond detected by Battistel et al. with the HSQC–TOCSY using WATERGATE, a through-bond J-coupling experiment. The H1f → C2G and H2g → C1f Hbonds (Figure 1.5) were the ones that had gone undetected for many decades. While present in the neutron structure,²⁸  its existence had not yet been directly detected in solution.^10,29–32  The low signal-to-noise ratio (SNR) for this cross peak suggests one of two scenarios. In the first, line broadening of the OH groups hampers significant magnetization transfer due to relaxation. In the second, a small population of sucrose is Hbonded, leading to a low SNR for this peak. The latter is more consistent with molecular dynamics, thus Battistel et al. ascribed this low SNR to the low population of the Hbonded form. In both scenarios, the Hbonded form may be in rapid exchange with water and would therefore be challenging to detect.

At higher temperatures the OH resonances broaden and become difficult to detect and also to correlate. One can envision that reduced exchange OH rates can still imply the presence of Hbonds, even when they are not directly detectable using HSQC–TOCSY. As mentioned in an earlier section, amide NH exchange rates in disordered proteins were recently published using SOLEXSY.¹⁷  Based on earlier studies of OH and NH exchange at reduced temperatures in mixed solvents,^{10,30,33–39}  we believed that this idea could be applied to glycans to measure OH/OD exchange rates in water without a co-solvent.

In a 1 : 1 mixture of H₂O/D₂O, only 50% of the OH groups are deuterated. This leads to two resonances in the carbon spectrum, one for a ¹³C atom that is attached to an OD group and thus shows a deuterium isotope effect and the other that has the chemical shift of a ¹³C atom that is attached to an OH.^40,41  Although the OH groups are exchanging with the OD groups, the exchange rate constant is slower than the frequency difference between the ¹³C–OH and the ¹³C–OD. The exchange rate can be measured from one-dimensional ¹³C spectra (Figure 1.6).

A variant of the HSQC–TOCSY, called INTOXSY (Figure 1.7(A)), can be used to measure the exchange rate constants.⁴²  In this variant, magnetization is transferred from ¹H to ¹³C, allowed to evolve in ¹³C and then transferred back to ¹H, as in an HSQC. Then, the ¹H magnetization is refocused from anti-phase back to in-phase magnetization. This in-phase ¹H magnetization is then placed on the z-axis and a variable delay time is incorporated for H/D exchange which is followed by a 90° pulse and DIPSI-2 mixing to transfer ¹H magnetization between the C–H and the OH groups. ¹³C atoms bearing an OH now become detectable through the H–C to H–OC coupling. ¹³C atoms bearing an OD group are not detectable. However, as in SOLEXSY, the cross-peak intensities/volumes of ¹³C atoms that, before the variable delay time, bore an OD but after the delay bear an OH are seen to be exponentially increasing in intensity at the end of the variable delay period. The cross-peak intensities/volumes of ¹³C atoms that before the variable delay time bore an OH and now bear an OD are seen to be exponentially decreasing in intensity at the end of the variable delay period. ¹³C atoms that bore an OD and exchanged for an OH also contain a ¹H T₁ component, which can be fit to a bi-exponential function that fits both exponential rates (Figure 1.7(B)).

The INTOXSY NMR experiment described here provides data that can be used to infer Hbonding, even in cases where observation of Hbonds cannot be directly observed or where the correlation is weak. One such example is a recently published study of sialyl Lewis-X (sLe^X). Here OH groups of two glycan residues, Gal{II} and Fuc{III}, which surround a central residue, GlcNAc{II}, form Hbonds with the N-acetyl group, causing the N-acetyl to rapidly flip between the two OH groups, Fuc{III}OH2 and Gal{I}OH2. This facilitates Hbonding between the carbonyl of the N-acetyl group and these two OH groups. Evidence for N-acetyl flipping conformational motion was first found by NMR temperature dependent linewidths. However, the details were revealed by MD simulations and Hbonding was further supported by INTOXSY measurements.⁴²  OH/OD exchange rate constants were <7 s⁻¹, confirming that these OHs are protected from exchange with solvent. OH/OD exchange rates were used in a recent intermolecular complex of Lewis-X hydroxyl groups.⁴¹ 

Presently, it is difficult to estimate the strength of intramolecular Hbonds in glycans. While it is possible to use the populations available from MD simulations, it is widely accepted that the best estimates of ground state energies are accurate to about ±1 kcal mol⁻¹. Given this level of uncertainty, it is difficult to say whether small populations of Hbonded forms predicted by MD simulations are present in the same amounts in solution.

One parameter that would help in this regard is J-coupling (¹H–¹³C and ¹H–¹H) because the magnitude of the J-coupling could lend insight into stability and strength of the Hbonding interaction. O'Leary and co-workers reported some of these J-couplings in model compounds and found that they range from 0.15 to 0.35 in these compounds.^43,44  These small values imply that J-values may be challenging to measure, especially in cases of smaller populations of H-bonded glycans.

As stated in the introduction, C–H Hbonds are controversial for a few reasons. First, it is unknown if they truly contribute to stability of biomolecules. Second, they are weaker than many Hbonds and finally, it is hard to see how they fit into the general definitions of Hbonds. Since the active H atom in the Hbond is attached to a carbon atom, it is not intuitive to predict when these C–H's will be involved in Hbonding.

Nevertheless, such C–H based Hbonds have been reported in crystal structures^1,9  and in solution.⁴⁵  Interestingly, these C–H Hbonds are from Hα in proteins, which can be said to be slightly acidic as they are in the α position to a carbonyl group. However, the carbon atom in a C–H group is not typically thought of as an electronegative group. Furthermore, the hydrogen atoms in these C–H Hbonds are usually not easily exchanged or transferred under biologically relevant solution conditions. It is with this preface that we present studies that identified and demonstrated evidence for C–H Hbonds in Lewis-X (Le^X) based antigens.

In 2013, Zierke et al. published a study of the Le^X trisaccharide conjugated to a protein, which predicted the existence of an Hbond between the H5 of Fuc{II} and the O5 of Gal{III}.⁴⁶  In structural studies, Gal{III} appeared to be situated directly above the H5 of Fuc{II}. This hydrogen displays an unusually deshielded chemical shift and was the basis of the authors’ proposal.

To confirm the proposed Hbond, Zierke et al. covalently attached Le^X to a protein. Presumably, now that the trisaccharide is part of a larger molecule, molecular tumbling would be slowed, which should facilitate NOE measurement in the slow tumbling limit. These measurements provided structural data consistent with an Hbond, but direct NMR evidence of direct orbital overlap (J-coupling) was not reported in this paper. Nevertheless, the NOEs provided validation of the structural proximity of Fuc{II}H5 to Gal{III}'s pyranose oxygen (Figure 1.8).

Interestingly, Pederson and Prestegard reported that Le^X binds to DC–SIGN in two slightly different conformations. One where H5 of Fuc{II} is closer to the pyranose oxygen of Gal{III} and the other when H5 is closer to the glycosidic oxygen between Gal{III} and GlcNAc{I};⁴⁷  both conformations are consistent with NOEs.

Recently, the existence of the CH⋯O Hbond in sLe^X, a Le^X analog, was shown by NMR spectroscopy.⁴⁸  In this report Battistel et al. showed direct through-bond correlations from Fuc{III}H5 to Gal{IV}C1 (these are the same as Fuc{II} and Gal{III}, respectively in the Le^X trisaccharide). Additional correlations from Fuc{III}H5 to GlcNAc{II}C4 and Gal{IV}H1 to Fuc{III}C5 were also observed.

Altogether, the NMR correlations in Battistel et al.'s paper support Zierke's original hypothesis of an Hbond from Fuc{III}H5 to Gal{IV}O5. In addition, Battistel et al.'s report also supports Pederson and Prestegard's NOE based hypothesis of an additional Hbond from Fuc{III}H5 to GlcNAc{II}O4.

The cases discussed so far are representative of the typical situation in glycans where Hbonds involving OHs have relatively small interaction energies, thus representing conformations that are not as populated as others. Despite that disadvantage, Hbonds could be detected by means of through-bond NMR experiments. An unexpected reversal of situation was found when studying the structure adopted by free α(2-8) PSA oligosaccharides in solution.⁴⁹  Because historically the focus was on characterizing the, arguably more important, PSA inter-residue α-linkages, little effort was dedicated into looking at the reducing end, which is mostly found in the β-configuration (90% in the monomer, 97% in the dimer and higher oligomers). Azurmendi et al. report results from MD showing that, thanks to the axial orientation of the anomeric OH, the latter forms a stable Hbond with the carboxylate oxygen of the following residue (red dashed line in Figure 1.9(A)), representing over 90% of the MD population (Figure 1.9(B), bottom panel). Direct evidence was obtained using an HSQMBC.⁵⁰  Additionally, three independent lines of measurements provide the required evidence in support of both, the Hbond and the MD predicted structure for the first two residues at the reducing end of α(2-8) PSA OS: OH exchange rate values, HSQC–ROESY experiments, and residual dipolar couplings (RDCs).

OH exchange rate constants in SiA OS have been recently published.⁵¹  The exchange rate for OH2 in the dimer is 3.4 s⁻¹, about half that of the next low value and an order of magnitude slower than the faster exchanging OHs. The trend is mostly conserved for longer OS of PSA. As mentioned previously, through-bond NMR experiments confirm the Hbond and both MD and the exchange rate indicated that HO2 was Hbonded we also performed ROESY experiments under varied conditions trying to correlate the HO2 with predicted nearby CHs through-space. Figure 1.9(B) shows specific predictions from the MD trajectory, as there is a one to one correspondence between SiA{I}HO2–SiA{II}O1 Hbond condition (bottom panel) and the generally accepted ROE observation limit of 5 Å. In other words, observation of ROESY correlations for the atoms indicated in the top three panels in Figure 1.9(B) would imply the presence of the predicted Hbond. In fact, as reported in Azurmendi et al., the HSQC–ROESY experiments show correlations between all the atoms indicated in Figure 1.9(A). The same report shows that experimental RDC values measured at room temperature are well reproduced for the MD trajectory using SVD, thus supporting the notion that the Hbond is still present at relatively high temperatures as predicted by MD. Finally, the Hbond was confirmed through a long-range coupling measurement.

This Hbond could help explain the unusual behavior of the free Sialic acid oligomers, as suggested by Jennings and co-workers.⁵²  It also may be the cause of the shift in the β : α equilibrium ratio of the reducing ring on going from monomer (90 : 10) to dimer (97 : 3) and higher order oligomers.

Clearly, Hbonding in glycans is an important area of research. Because they can provide additional data, structures derived from data indicating Hbonding are likely to be more precise than those that do not utilize Hbonds. Direct detection of Hbonding can also provide new data for a more complete theory of glycan structure.

The importance of J-coupling in establishing Hbonding should not be underestimated. While direct detection of Hbonding was not in the O142 PS, in the remaining examples given here, a finite long-range J-coupling between atoms in an Hbond and a remote ¹³C was used to directly establish Hbonding in sucrose, SiA₄, SiA₂ and sLe^X.

It is likely that, until regular Hbonding patterns can be established, through bond J-coupling will continue to be the method of choice to establish Hbonding in glycans. Naturally, this raises the issue of the J-coupling values in these Hbonds, which have so far been elusive in aqueous solution. However, the magnitude of the J-coupling may shed light on the strength of Hbonds in glycans. These will be important targets in future studies. O'Leary and co-workers have already begun to pave the way for these studies.^43,44 

As can be seen in the studies presented, the direct detection of Hbonds can occupy an NMR spectrometer for up to one week. Future studies of Hbonding glycans may be facilitated by more sensitive detection, such as the ever-increasing sensitivity of NMR cryoprobe technology and higher magnetic field strength. More importantly, higher magnetic fields may allow the detection and study of Hbonded glycans at higher temperatures due to increased resolution (larger number of Hz in one ppm) at higher magnetic fields.

The studies presented here show that Hbonding glycan can be detected and that these data have important consequences for structure determination in glycans. It is our hope that Hbonding will continue to be a source of new structural data in glycan structure determination.