Glycan OH Exchange Rate Determination in Aqueous Solution

Article pubs.acs.org/JPCB

Glycan OH Exchange Rate Determination in Aqueous Solution: Seeking Evidence for Transient Hydrogen Bonds Marcos D. Battistel, Hugo F. Azurmendi, and Darón I. Freedberg* Laboratory of Bacterial Polysaccharides, Center for Biologics Evaluation and Research, Food and Drug Administration, 10903 New Hampshire Avenue, Silver Spring, Maryland 20903, United States S Supporting Information *

ABSTRACT: Hydrogen bonds (Hbonds) are important stabilizing forces in biomolecules. However, for glycans in aqueous solution, direct NMR detection of Hbonds is elusive because of their transient nature. Here, we present Isotopebased Natural-abundance TOtal correlation eXchange SpectroscopY (INTOXSY), a new 1H−13C heteronuclear single quantum coherence−total correlation spectroscopy based method, to extract OH groups’ exchange rate constants (kex) for molecules in natural 13C abundance and show that OH Hbonds can be inferred from “slower” H/D kex. We evaluate kex measured with INTOXSY in light of those extracted with lineshape analysis. Subsequently, we use a set of common glycans to establish a kex reference basis set and to infer the existence of transient Hbonds involving OH donor groups. Then, we report kex values for a series of mono- and disaccharides, as well as for oligosaccharides sialyl Lewis X and β-cyclodextrin, and compare the results with those from the reference set to extract Hbond information. Finally, we utilize NMR experimental data in conjunction with molecular dynamics simulations to establish donor and acceptor Hbond pairs. Our exchange rate measurements indicate that OH/OD exchange rates, kHD, values kHD. Conversely,

2

I [ppb/Hz]

kcoal [s−1]

50/8.8 162/28.5 217/38.2 151/26.6 74/13.0 161/28.3 146/25.7 35/6.2 172/30.3 148/26.1 77/13.6 158/27.8

19 63 84 59 29 62 57 14 67 66 30 61

LSA [s−1] 33 20 48 18 35 14 15 38 39 43 38 13

± ± ± ± ± ± ± ± ± ± ± ±

18 2 32 8 24 6 6 16 12 10 12 8

INTOXSY [s−1] NA 15 ± NM 13 ± NA 27 ± 22 ± NA 23 ± NM NA 20 ±

10 6 10 4 6

6

a

Errors are reported to 95% confidence intervals; NM, not measureable; NA, not applicable.

from 16 to 196 ppb at 176 MHz. In 50/50 H2O/D2O at −10 °C, the 2Ieq’s for all secondary OH groups fall between 60 and 90 ppb. As expected, tertiary and quaternary carbon atoms, namely, G1, G5, F5, and F2, display a single peak each (Figure 2A and Table 1). Despite not bearing OH groups, differences in chemical shifts were observed for samples in 100% D2O relative to those in 100% H2O (Figure 2B). Hence, the signals corresponding to G1, G5, F5, and F2 are in fact averaged 13C resonances. We further analyzed this observation. We considered the following causes for the observed 13C chemical shift change for 13C with no OH groups: (1) solvation effects, (2) long-range exchange processes (as a result of longrange equilibrium isotope shif ts, long‑rangeIeq’s), (3) exchange processes involving intramolecular Hbonds, and (4) a combination of these events. Even though we are aware of exchange processes taking place in neighboring 13C atoms, it is not possible to discriminate between long-range exchange processes or to attribute the observed effect only to exchange, solvation, or Hbonding. Such effects can also be present in carbon atoms bearing OH groups; however, these minor 13C chemical shift changes (from ∼5 to 12 ppb for sucrose) are masked by the larger direct 2Ieq. We know that the resonances of 13C atoms bearing OH groups reflect OH/OD chemical exchange; therefore, it is likely that long-range exchange effects will be propagated at neighboring sites. Among all of the factors described above that can impact chemical shift and peaks’ line shape, and therefore rates extracted from LSA, only long-range exchange processes can be systematically evaluated. Thus, we explored their impact on kHD determination through LSA (Figure S1). Experimentally, we cannot separate the long‑rangeIeq and 2Ieq effects on kHD. However, their individual contributions can be probed by DNMR simulations and LSA to compare the theoretical and measured kHD values. Briefly, we generated 13C spectra using known 2I, long‑rangeI, and kHD values and then backcalculated kHD by LSA. We found that for 2I = long‑rangeIeq × 10 and long-range OH/OD kHD > 10 s−1, the error induced by exchange processes in neighboring carbon atoms on two-bond kHD determinations is 10 s−1, indicating that for sucrose the contribution of the OH/OD

π

when kHD > 2 (2I ) = 2.2(2I ), the two isotopologues’ resonances coalesce into a single one (Figure 2B). When two resonances are observed, it is possible to calculate the rate at which the peaks would coalesce or a given 2I, using the π approximation kex = 2 (2I ) = 2.2(2I ).54 We refer to this rate constant as the rate at coalescence, kcoal. With this in hand, it is possible to provide a limit on kHD, determined by NMR in equimolar H2O/D2O mixtures. We provide kcoal along with kHD in Table 1. All sucrose 13C resonances showed 2I > 0, ranging 687

DOI: 10.1021/acs.jpcb.6b10594 J. Phys. Chem. B 2017, 121, 683−695

Article

The Journal of Physical Chemistry B

Figure 3. INTOXSY spectra at different exchange times of a 0.3 M sucrose sample, pH 6.5. The spectrum presented in black (τex = 10 ms) shows the correct 1H frequency. Spectra at τex = 100 (purple) and τex = 200 ms (green) were offset in 1H frequency to facilitate comparison. Atom numbers are indicated with Arabic numerals and glucose and fructose rings with G and F, respectively.The inset depicts the fitted buildup/decay profiles obtained from INTOXSY spectra peak volumes as a function of τex for cross-peaks corresponding to the C2 of the glucose ring of sucrose.

for Hbonding by extracting kHD from LSA and INTOXSY methods on sucrose, at different temperatures and concentrations (Table S3). As expected, increasing sucrose concentration from 0.3 to 0.8 M (Tables 1 and S3) improved precision due to SNR increase. However, the increase in concentration appears to have little influence on kHD values in this range, with the exception of F4, for which a 2-fold decrease is observed at 0.8 M. Even though a concentration-dependent intermolecular Hbond has been reported between G4f3 and G3f3,5 the population of dimerized sucrose molecules may be insufficient to affect the rate of exchange within the 95% confidence interval.5 It has been calculated that seven water molecules are sufficient to solvate sucrose.56 Increasing sucrose concentration from 0.3 to 0.8 M decreases the H2O/sucrose molecular ratio by 2.5-fold. This reduction is equivalent to decreasing the number of H2O molecules from ∼25 to ∼10 times predicted to solvate sucrose and may result in dimer formation.56 The fact that immediate H2O layers are responsible for direct exchange with sucrose OH groups and that these layers are not depleted may explain why a reduction in water concentration, for the tested sucrose samples, does not appear to affect the measured kHD. Moreover, our results are consistent with the finding that supramolecular Hbonded networks of sucrose molecules, that could impact kHD, are observed when sucrose concentration exceeds 1.1 M (33% w/w sucrose/water), which is beyond the highest concentration of sucrose that we tested (0.8 M or 24% w/w).57 The exchange rates’ temperature dependence indicates an estimated transition-state activation energy (ΔG⧧, Table S4) of ∼14 kcal/mol for all sucrose OH groups, suggesting that all sucrose OHs are equally solvated and exposed to water

exchange processes occurring on neighboring sites is less than 10% of the kHD determination from LSA. However, as longrange effects are not the only source of variation in chemical shift and peak line-shape, as discussed earlier, one should expect not less than a 10% error in kHD determined from LSA. In summary, LSA studies on sucrose provided a set of reference kHD values and allowed us to establish the required resolution to observe 2Ieq. Hence, INTOXSY experiments were carried out with sufficient resolution aiming to separate the smallest equilibrium isotope shift in sucrose, 60 ppb (∼5 Hz/ point at 176.045 MHz) for F4 (Figure 2A). INTOXSY does not yield 13COD signals at short τex (Figure 3, peaks in black), but these signals become observable when τex is increased (violet and green spectra in Figure 3 are offset in the 1H dimension for comparison). The INTOXSY experiment can yield both kHD and kDH rate constants describing OH/OD and OD/OH exchanges. Because the Kf55 is ∼1 for 50/50 H2O/ D2O solutions, we used kDH = kHD for fitting purposes, as described in Section 2. For F4 and G3, kHD could not be determined with INTOXSY because 2Ieq could only be resolved at the expense of a decreased SNR (Table 1). The kHD values that could be extracted from INTOXSY, by fitting the exchange profiles (Figure 3 inset, Table 1), agree well with the values obtained from LSA (Table 1). Therefore, both INTOXSY and LSA can provide kHD; nevertheless, the method of choice will depend primarily on sample availability and spectral complexity. We previously reported transient Hbonds in sucrose.5 The obtained kHD values for sucrose (Table 1) are homogeneous, consistent with solvent-exposed OH groups, and do not reflect the presence of a persistent Hbond in sucrose. As Hbonding should lead to decreased kHD values, we sought further evidence 688

DOI: 10.1021/acs.jpcb.6b10594 J. Phys. Chem. B 2017, 121, 683−695

Article

The Journal of Physical Chemistry B Table 2. INTOXSY- and LSA-Based kHD (s−1) for Selected Monosaccharidesa INTOXSY (s−1)

LSA (s−1)

rate at coalescence (s−1)

INTOXSY (s−1)

LSA (s−1)

α

rate at coalescence (s−1)

β Glc

1 2 3 4 6

9.8 ± 2.3 9.1 ± 2.6 NM* NM NM

13 18 23 34 21

± ± ± ± ±

18 ± 4 17 ± 2 6±3 NM 7.4 ± 0.8

23 23 13 24 13

± ± ± ± ±

10 17 22 10 4

65 82 49 56 63

1 2 3 4 6

NM 9.8 ± 0.6 27 ± 2 22 ± 3 23 ± 2

>52 13 ± 23 ± 16 ± 16 ±

NM 17 ± 1 NM 22 ± 4 23 ± 2

>48 15 ± 47 ± 13 ± 16 ±

6 10 4 3

48 69 70 56 53

1 2 3 4 6

17 ± 8 10 ± 3 NM 6±3 24 ± 16

11.1 ± 0.7 17 ± 5 10 ± 3 9±3 12.9 ± 0.3

39 60 54 48 44

17 ± 8 NM 21 ± 5 7±1 NM

17 ± 2 >68 16 ± 4 9±1 15 ± 1

39 68 54 50 45

1 2 3 4 6

14 ± 3

15 ± 4

45

NM

20 ± 5

50

15 ± 5 16 ± 1 8±2

29 ± 9 12 ± 5 13 ± 5

68 55 68

NM 9±7 3±2

30 ± 10 15 ± 6 14 ± 6

69 58 75

1 2 3 4 6

12 ± 9 NM 20 ± 7 3±1 NM

15 ± 3 24 ± 8 23 ± 7 6±2 >62

32 86 69 33 62

18 ± 1 NM 15 ± 2 9±0 NM

15 ± 6 24 ± 8 22 ± 7 8±2 >62

36 86 72 26 62

4 8 3 13 4

53 73 50 80 63

6 9 5 3

52 72 72 55 53

Man

Gal

GlcNAc

Fuc

a

Errors are reported to 95% confidence intervals; NM, not measureable due to COH/OD overlap.

First, we will discuss the general findings and later describe specific cases. 3.1. kHD of Glycans at −10 °C. As slowly exchanging OHs yield narrower 1H resonances, for kHD determination on monosaccharides, pH values were set to reduce the line width and thereby maximize OH SNR for peaks in 1D 1H experiments (indicated in Figure S2). As observed for imino protons,58 buffer and salt content also increase kHD and consequently increase the line width and reduce the SNR.59 Therefore, we did not utilize buffers. Because OH/OD exchange is acid/base-catalyzed, changes in pH as small as 0.1 units lead to significant variation in kHD and SNRs. The optimal pH for all cases lies in the ∼5−7 range. pH values lower than ∼6.0 are better for anomeric signals, whereas pH within the range 6.0−7.0 is better for other ring OHs.27−29,59,60 Thus, we adjusted the pH to the value yielding the higher SNR for the whole OH region. Optimizing the pH for each sample enabled us to make sample-to-sample comparison on the basis of overall slowest exchange observed for each sample. LSA and INTOXSY data agree well (Table 2, Figure S2). Most OH groups were found to have kHD ≅ 10−20 s−1, at −10 °C and at their respective optimal pH values (Supporting Information). The anomeric configuration affects the OH exchange rates of only the anomeric OH without noticeable effects on other OHs in the ring (Table 2, Figure S2). Interestingly, when the OH at C4 is axial, the OH hydrogen atom exhibits a kHD below 10 s−1, regardless of anomeric

molecules. This result still contrasts with the fact that inter- and intramolecular Hbonds in sucrose have been reported.5,30 However, analysis of Hbonded forms (G4/G3f3, F1g2, and G2f1) from MD simulations indicates that these conformations represent ∼1% of the total population. It is experimentally challenging to confirm these small populations, even if persistent, because the associated error on exchange rate measurements is ∼10%. Only when the population of Hbonded conformations is significantly larger than the experimental precision for kHD, their effect on the rate will become detectable. Considering this theoretical small population, it is remarkable that Hbonding correlation peaks are observed by NMR. We have now established that OH exchange rates in glycans can be measured by both LSA and INTOXSY. However, inferring Hbonds from exchange rates in glycans or other molecules requires a reference basis set of kHD values to define protection factors and to obtain insight into the role that functional groups, anomeric configuration, linkage position, or OH orientation (axial vs equatorial) play in exchange rate values. To begin building such a database, we assigned the hydroxyl hydrogen atoms and determined the kHD values for the following molecules under the same experimental conditions: (1) the monosaccharides Glc, Gal, Fuc, GlcNAc, and Man; (2) the disaccharides α1,2-, α1,3-, and α1,6mannobiose; and (3) the OS’s sialyl Lewis X (sLeX-5) and βCD (OH assignments and kHD are presented in Table S5). 689

DOI: 10.1021/acs.jpcb.6b10594 J. Phys. Chem. B 2017, 121, 683−695

Article

The Journal of Physical Chemistry B

Figure 4. NMR data displaying H/D exchange rates and Hbonding in βCD. (A) HSQC−TOCSY spectrum of 0.3 M βCD, pH 6.5 with 60 ms isotropic mixing. The asterisk denotes the presence of an impurity. The inset shows INTOXSY buildup and decay profiles for each OH that yielded the rate constant values indicated in the figure. (B) Hydroxyl 1H region of a TOCSY spectrum with 20 ms isotropic mixing, showing cross-peaks that support intramolecular OH3···OH2 Hbonding. (C) βCD structure highlighting Hbonding directionality supported by NMR data.

of the same OH for the β anomer or for the OH2 of RII. However, comparing the kHD values to those of the same OH in the monomer yields a protection factor of 1.1 ± 0.3. Interestingly, RI-OH3 in the α1,6-mannobiose β-anomer exchanges 2-fold slower than for the terminal ring’s OH3 (8 ± 5 vs 18 ± 2 s−1, Table S5). Despite the apparent slower OH kHD described above, no through-Hbond coupling could be detected for these dimannosides. However, the existence of Hbonds in the gas phase has been reported.61 Energetically favored conformations of α1,2-, α1,3-, and α1,6-mannobiose in the gas phase indicate the presence of the following stabilizing Hbonds: OH6′···OH6 for α1,2-; OH2···OH6′ for α1,3-; and OH4···OH6′ for α1,6-mannobioses. In contrast, no persistent Hbonds have been directly detected for these disaccharides in aqueous solution, even when these solutions contain up to 15% acetone.62 Consequently, a reduction in kHD is not expected. As these dimannosides do not appear to have significantly populated intramolecular Hbonded conformations, this disaccharide series can serve as reference for OH exchange rate variation arising from different anomeric configurations, glycosidic linkage positions, and axial versus equatorial orientation in the absence of Hbonding. On the basis of the collected data presented thus far, all OHs exchange at comparable rates. Therefore, we propose that the reference kHD value for fully exposed OHs at −10 °C is ∼18 s−1 and with protection factors smaller than 2. Of course, this value has to be corroborated with further determinations (Table S5). 3.3. βCD. βCD is a cyclic C7 symmetric molecule with 7 α1,4-linked Glc residues. As expected, βCD yields only three hydroxyl 1H signals (Figure 4). Both the HSQC−TOCSY (Figure 4A) and TOCSY experiments (Figure 4B) show clear cross-peaks between the OH3’s and OH2’s, suggesting direct J coupling between these hydroxyl 1H’s. However, molecular symmetry prevents the identification of the correlation as Hbond inter-residue, intraresidue Hbond J coupling, or simply intraresidue vicinal J coupling. INTOXSY exchange profiles clearly show that the OH3’s exchange 4−5 times more slowly than the OH2’s and OH6’s (Figure 4A, inset). This result indicates that OH3’s are

configuration, compared to that for the equatorial counterpart (e.g., Glc vs Gal). This observation agrees with the suggestion that the reduced rate may be due to increased steric hindrance for axial OHs.27,28 The effect is most noticeable for OH4 in Gal and Fuc and, interestingly, less so for OH2 in Man. The upper limit for kHD, kcoal, is shown in Figure S2 (inverted solid pink triangles) and Table 2. However, the kHD measured from 2Ieq cannot be predicted from experimentally determined 2I. Provided that there is an observable 2Ieq,29 a lower kcoal value implies a smaller 2I. Note that for those OHs that did not yield well-resolved 2Ieq, kHD could not be extracted using INTOXSY. The hydroxyl kHD values in monosaccharides provide a frame of reference for expected OH/D kHD values at −10 °C for isolated glycans (Table S5) but do not provide data on how the environment, such as glycosidic linkage, can impact kHD. To this end, we analyzed a series of dimannosides for which the same monosaccharide, Man, is arranged in different configurations. 3.2. Mannobiose Series. At the experimental resolution of 5 Hz/point, the reducing ring for α1,2-, α1,3-, and α1,6mannobiose yielded resolved OH signals for each of the α and β anomers; however, the signals for the nonreducing ring of the α and β anomers could not be differentiated. Thus, a single kex was obtained for the OHs of the nonreducing ring for both anomers (Table S5). In the studied mannobioses, the average kHD for all of the hydroxyl groups in the second ring is ∼18 ± 3 s−1 at −10 °C and pH 6.5. As we measured the monosaccharide Man kHD under the same conditions as for the Man disaccharides, we used the monosaccharide data as reference for computing Man protection factors (p = kreference/ksample) (Table S5). All OH groups for these three disaccharides show protection factors smaller than 2. As in the monosaccharides, the results in the mannobioses imply that the exposed OHs exchange at similar rates, regardless of whether the OHs are in an axial (OH at C2) or equatorial orientation or proximal to a glycosidic linkage (Table S5). Perhaps structural and dynamic factors, inherent to the molecule in question, would modulate OH kinetics. In α1,2-mannobiose, for the OH2 of the α anomer of residue I, the kex is 9 ± 2 s−1, 2-fold slower than that 690

DOI: 10.1021/acs.jpcb.6b10594 J. Phys. Chem. B 2017, 121, 683−695

Article

The Journal of Physical Chemistry B protected from exchange with solvent, implying a donor role in Hbond formation. The HSQC−TOCSY OH3’s ↔ OH2’s correlation peaks indicate that O2 is the Hbond acceptor; however, it does not allow discrimination between intra- and inter-residue Hbonds. On the other hand, the kHD values for OH2 and OH6 indicate that they are barely protected, if at all. We used the crystallographic structure of βCD (PDBID 1DMB)63 as a model to infer Hbonds after adding the missing hydrogen atoms to the structure (Figure 4C). The molecule can be described as a truncated cone with the small side of βCD being stabilized by Hbonding between OH6’s of contiguous Glc residues, whereas the large side is stabilized by OH2/OH3 inter-residue Hbonds. Hbond formation in CDs (α, β, and γ) and the role of cooperativity in structure stability have been discussed in detail as well, on the basis of computational studies.64 Our experimental results are consistent with some of the βCD models reported for OH3’s and OH2’s but do not seem to be compatible for the OH6 case, for which a relatively high Hbond interaction energy was reported (1.1−5.6 kcal/ mol, depending on the βCD model). To address these conflicting results, we conducted very extensive MD simulations (4 × 4.0 μs MD trajectories) and performed a statistical analysis on Hbond formation. We chose a temperature of 300 K because solvated systems and force fields have been thoroughly tested around this temperature, though our experimental H/D exchange measurements were made at lower temperatures. The Hbond statistical analysis results were consistent with the experimental kHD values, yielding an average ∼11% [i]OH3−[i + 1]O2 inter-residue Hbond formation (per residue), with an average H-acceptor distance of 2.8 Å and donor−H-acceptor angle of 161°. Thus, the prediction is that at any one moment there is at least one [i]OH3 → [i + 1]O2 Hbond formed in βCD at ∼27 °C. As the NMR experiments indicate, lowering the temperature further favors this Hbond formation and, consequently, increases its population. For OH6’s, the analysis also predicts a relatively high rate of intraresidue Hbond formation (∼13% between OH6−O5); however, the average donor−H-acceptor angle is ∼108°, leading to a weaker interaction.65−67 Other proposed interresidue Hbonds involving OH6 as donor were present for ∼1% of the time or less, consistent with the weak solvent exchange protection observed by NMR. In summary, βCD is an interesting OH exchange model system because (1) it demonstrates the applicability of INTOXSY to larger OS’s; (2) it shows how INTOXSY can help to infer Hbonds in glycans, establishing a kHD of ∼5 s−1 threshold for hydroxyl groups involved in more persistent Hbonds; and (3) it illustrates the use of MD simulations to provide important insights into these molecular systems, in particular, to discriminate between donor and acceptor atoms when experimentally it may not be possible. For βCD, MD clearly points to inter-residue Hbond formation between [i]OH3 and [i + 1]O2, eliminating the alternative possibilities raised by experimental ambiguities. 3.4. Sialyl Lewis X (sLeX-5). As mentioned above, INTOXSY experiments are time-consuming when sample availability is limited; however, the spectral resolution obtained for OH signals may prove to be an essential advantage with complex molecules. Thus, we were able to probe kHD for OH groups in sLeX-5 using this method (Figure 5). INTOXSY enabled measuring kHD values for 10 out of 14 hydroxyl groups in sLeX-5. For GlcNAc{II}OH6, Gal{IV}OH4, and Sia{V}OH8/OH9, kHD could not be measured, as 2Ieq were unresolved

Figure 5. (A) INTOXSY results for a 0.3 M sLeX-5 sample at pH 6.5 and −10 °C. (B) sLeX-5 chemical structure. Roman and Arabic numerals indicate residue and atom numbers, respectively. Red boxes indicate hydroxyl groups with reduced exchange rates and likely involved in intramolecular Hbonds. Black circles indicate OH groups for which kHD could not be measured due to unresolved 2Ieq.

because kHD > 2.2(2I). For Gal{I}OH4/OH2, Fuc{III}OH2, and Gal{IV}OH2, kHD values indicate that these OHs are shielded from exchange, as they are consistently lower than other sLeX-5 kHD values (Figure 5B, red boxes). However, as HSQC−TOCSY (with 30 ms τm) did not yield Hbond crosspeaks, Hbond donor and acceptor groups could not be directly identified. Nevertheless, INTOXSY data yield the putative Hbond donors through small kHD values when compared to those observed for the Hbonded βCD OH3. We turned to MD simulations to identify the Hbond acceptors. Figure S3 shows the possible donor and acceptor Hbonds and their respective occurrences throughout a 500 ns MD trajectory. In the simulation, Gal{I}OH2 and Fuc{III}OH2, located at opposite sides of the GlcNAc{II} N-acetyl group, form alternating Hbonds with the carbonyl oxygen of the NHAc group in more than 10% of the trajectory frames for both acceptor atoms combined. Moreover, Gal{IV}OH2 forms Hbonds between 20 and 30% of the time with the pyranose oxygen atom of the Sia residue. Interestingly, Sia{V}OH7, belonging to the terminal residue, is predicted to form an intraresidue Hbond with the carbonyl oxygen atom of the Sia{V}NAc group for 30% of the time (Figure S3), protecting the otherwise exposed terminal OH groups from solvent exchange. We could not directly detect OH Hbonds in sLeX-5. Therefore, it is likely that, if they exist in water, they are short-lived or they constitute a minor population at equilibrium. Hbond detection via NMR depends on the magnitude of the through-Hbond coupling constant, the population of Hbonded 691

DOI: 10.1021/acs.jpcb.6b10594 J. Phys. Chem. B 2017, 121, 683−695

Article

The Journal of Physical Chemistry B molecules, and Hbond lifetime. In this context, the absence of HSQC−TOCSY Hbond cross-peaks is not surprising because the Hbonds’ lifetime has to exceed the time period used in the NMR experiment to transfer magnetization from the H(C) to the OH group. For a 1 Hz through-Hbond coupling constant between OH groups, 500 ms is required for a more efficient complete magnetization transfer with TOCSY (1/(2JHH)).68 However, a typical long isotropic mixing experimentally used is of 120 ms, resulting in only a fraction of the magnetization being transferred. Additionally, in this time, assuming a kHD of 20 s−1, OHs could exchange approximately six times with the solvent, further decreasing the detectability. It is noteworthy that COSY-type magnetization transfer is less efficient (1/JHH), thus less suitable for Hbond detection. It has been shown that for structurally restricted 1,3 and 1,4 diols and nonprotic solvent (DMSO) through-Hbond OH/OH coupling values are in the order of ∼0.3 Hz.14 Assuming that the J-coupling values are solvent-independent, if this result is applied to glycans in water, it would predict that OH Hbond detection through 3hJHH-coupling is nearly impossible as multiple H/D exchange cycles will take place during the time required for a small ⟨3hJHH⟩-mediated magnetization transfer. In contrast to scalar coupling, the measured kHD is only a reflection of Hbond lifetime. Although it is less informative than coupling constant values, it can be more readily measured, and with a proper reference Hbond, donor groups and lifetime can be inferred. MD simulations can be used in a complimentary fashion, as shown herein, to provide support for potential Hbond acceptor groups, establishing Hbond pairs that can be used for structural analysis, even when Hbonds are not apparent from direct NMR correlations.

Table 3. LSA vs INTOXSY Summary LSA

INTOXSY

samples concentration dimensionality molecular size

3 >20 mM 1D mono- or disaccharides

experimental time experiments required OH detection resolving 2I for reference resolving 2Ieq type of OH precision data analysis

14−18 h

1 >200 mM 2D tested with pentasaccharides 48−72 h

3

8−10

not required required to minimize error

required not required

recommended 1−3°