Our previous study relied on synthetic analogues of hexasaccharide 1 bearing a single ¹³C-labelled Glc unit in different positions of the chain (e.g., Figure 1B, compounds 2–3) (Delbianco et al., 2018). Additional compounds bearing two ¹³C-labelled Glc units (Figure 1B, compounds 4, 5 and S1) were prepared to provide unambiguous and independent ¹³C-¹H RDC values to allow for the analysis of the global conformational behaviour.

Within a β−Glc residue, all the intra-ring C-H vectors are parallel to each other and should provide the same RDC value. Therefore, in compounds 2-5 only the CH vectors at the methylene moiety could provide additional spatial information. An additional label, with a different orientation than the axial CH vectors, could generate additional spatial information. Given the excellent performance of ¹⁹F for NMR (Linclau et al., 2020), three ¹⁹F-containing hexaglucosides (compound 6, S2, and S3) were targeted to obtain ¹⁹F-¹³C RDCs. The equatorial C-F bond was installed at position C-3, to minimize interference with folding. Two mono-fluorinated compounds (S2 and S3) were used as chemical shift reference, whereas the di-fluorinated analogue 6 was employed in the RDC studies. All target compounds were obtained in good overall yields (13–21%) from BB1-4 using standard AGA conditions (see Supplementary Material).

Several weak alignment media are available for neutral molecules, offering different liquid crystal properties. To evaluate the presence of the helical conformer in solution, we selected the cromoglycate sodium salt (cromolyn) solution in D₂O (Troche-Pesqueira et al., 2014) and the Otting’s medium (Rückert and Otting, 2000): C12E5/hexanol solution in D₂O. The small ionic aromatic cromolyn (a mesogen) creates chromonic phases in which the rigid aromatic moieties self-aggregate in columns in a face-to-face fashion. The aromatic moieties arrange perpendicular to the axis of the formed cylinder. These cylinders in turn are aligned also perpendicular to the magnetic field (Yu and Saupe, 1982). The Otting medium (Rückert and Otting, 2000) generates lamellar L_α crystalline phases, which are formed by series of parallel lipidic bilayers, where the hydrophobic n-alkyl chains aggregate into planar bilayers with the hydrophilic poly (ethylene glycol) headgroups pointing towards the water phase. Within the NMR magnet, the bilayer surfaces orient parallel to the magnetic field direction. This medium has neutral charge, is insensitive to pH and salts, with little or no binding capacity, and can be used at temperatures close to 40°C.

All measurements were performed with each hexasaccharide in a different NMR tube. The splitting in the deuterium lock signal (²H Q splitting) confirmed a similar degree of alignment for each samples. The average obtained splitting in the cromolyn medium was 85 ± 4   Hz , while in E5C12/hexanol was 27 ± 1   ± 1 Hz, allowing for the comparison of the different samples within each medium. RDCs were experimentally determined from the analysis of HSQC NMR spectra at 800 MHz (Figure S1a, see the General Materials and Methods section for details) in two different weak alignment media (Tables 1, 2).

The obtained results are strikingly different for both media. In the chromonic phase (Table 1), the measured RDCs display different signs, depending on the particular residue. Residue G1 always shows RDCs with positive values, while those for residues G2, G3 and G4 are negative. The RDC values within each ring are very similar, as expected from the parallel arrangement of the axial C-H vectors, as can be seen in compound 2 and 3 (residue G2 and G4, respectively). The differences in RDC absolute values for the different residues strongly support the different orientations of the Glc residues (G1, G2, G3, G4) along the hexasaccharide chain in the cromolyn phase. An identical behaviour was observed for compounds 5 and 6, both labelled at residues G1 and G3 either with ¹³C or ¹⁹F. For these two compounds, the magnitude and sign measured for the anomeric C1-H1 RDC is similar. As expected, the RDC values for the equatorial C-F bond in residues G1 and G3 of compound 6 showed a different relative orientation than the axial C-H vectors within the same residue (RDC values with opposite sign). Moreover, the C-F RDC displays different signs for residues G1 and G3, as observed for the corresponding C-H vectors. These data are compatible with a helical shape of the molecule, with the C-H vectors at different Glc moieties pointing to different orientations (Figure 2A gg ₆ conformer).

The data in the C12E5/hexanol solution were drastically different (Table 2). Only positive values were found and the obtained anomeric C1-H1 RDCs for residues G1, G2, and G3 in different molecules were comparable in magnitude, but larger than those for G4. Taken together, these experimental data indicate different relative orientation of the Glc residues in the two media, suggesting a different conformational preference. In this case, the data are compatible with an extended conformation, with all the C-H vectors in parallel disposition (Figure 2B gt ₆ conformer).

The experimental RDCs were fitted to those expected for different conformations of the hexasaccharide by using the MSPIN program, as described in the General Materials and Methods section (Navarro-Vazquez, 2012). Different conformations, helical, linear extended, and intermediate conformations were generated using the Macromodel suite of programs within Maestro 12.7 (Schrödinger, LLC, New York, NY, United States) and the AMBER* forcefield (see experimental section). For the helical conformation, the global minimum (Delbianco et al., 2018), Φ is kept within the exo-anomeric region, ψ adopts the trans conformation, and ω displays the gauche-gauche (gg) orientation. For the linear extended shape, all ω are in the gauche-trans (gt) rotamer (Figure 3). Other intermediate conformers were built with combinations of gg and gt rotamers.

The different data cross-fits are summarized in Table 3. The Cornillescu Quality factor (CQf), calculated from the fitting in MSPIN (see General Materials and Methods section), was used as quality measure. Additionally, the experimental RDC values were compared to those predicted by MSPIN for the canonical individual conformations (Tables S2abc).

The fitting analysis gathered in Table 3 used different data sets. Given the uncertainty provided by the methylene protons, results including or excluding the methylene RDCs were employed for the fitting processes. The results were further divided into two sets including (compounds 2, 3, and 6) or excluding (compounds 2, 3, and 5) the data measured for the fluorinated molecule. The experimental data were considered for fitting to the helix shape (all gg), to the extended geometry (all gt) and to the intermediate structures starting by either the gg or gt rotamers around the β(1–6) linkage (Figure 3). The comparison between the experimental RDCs and those predicted for the individual conformations according to MSPIN are reported in the Supplementary Material.

Data analysis indicates that the results obtained excluding the methylene protons provide better fittings than those including them. The helical conformation (all gg) provides the best fitting (although poor) when only the ¹³C-¹H RDC values are included and the RDC data for the methylene moieties are considered. The relatively high CQf values (0.46 or above) suggest the existence of conformational equilibria around some of these linkages, so that no single rotamer can satisfactorily account for all the observed RDCs. When the ¹³C-¹⁹F RDC values are taken into consideration, the best fitting is obtained for a mixed gg/gt form. In this case, the fitting is considerably improved when the methylene data are neglected. For this set of data, the fitting for the helix shape is also reasonable. The fully extended conformer (all gt) always provides a poor fitting value (0.35 or above), even excluding the methylene protons. In contrast, for the C12E5/hexanol Otting’s medium, the fitting for the helix structure is rather poor in comparison to those obtained for the extended (best fit) or mixed geometries.

Even though it is difficult to provide a quantitative distribution of conformers, it seems that the population of extended conformers in cromolyn is rather low, while the helical conformer should partially contribute to the equilibrium. The existence of gg conformers is favoured in this medium, especially at the region of the reducing end of the hexasaccharide chain. Nevertheless, the gt rotamers should also contribute to the equilibrium, especially for internal linkages. In the C12E5/hexanol solution medium, the contribution of the helical conformer is probably negligible, while the fully extended conformer (i.e., all gt) should be significantly populated. Still, the existence of gg rotamers cannot be disputed. There is not a single structure (conformation), but many of them contributing to the final presentation. This fact also highlights that, in this case, as for other flexible molecules, it is not possible to derive a single 3D structure from the RDC data. This type of approach would generate a virtual structure, with no physical meaning.

The analysis of the RDCs demonstrates the high flexibility of the oligosaccharide. The results clearly show that the alignment medium is not inert and provides interactions with the molecule. The distinct chemical nature of the two employed media generates different interactions with the hexasaccharide, stabilizing different conformational distributions in the two environments. Since these possible intermolecular interactions, including CH-π interactions (Asensio et al., 2013), are quite weak, the observed modulation of the conformational populations reflects the low energy barriers among the contributing conformers.

Automated glycan assembly (AGA) was performed on a home-built synthesizer developed at the Max Planck Institute of Colloids and Interfaces (Eller et al., 2013). All details concerning BB preparation, AGA modules, and post-AGA manipulations can be found in the Supplementary Material.

All spectra were recorded on a Bruker AVANCE III 800 spectrometer operating at a frequency of 800.13 MHz for ¹H, 201.19 MHz for ¹³C, and 122.83 MHz for ²H. One-bond ¹H-¹³C coupling constants were extracted from HSQC spectra acquired without proton decoupling during the acquisition period. These HSQC spectra were acquired with spectral widths of 10 ppm for the direct proton dimension and 10 ppm for the indirect carbon dimension (F1 dimension was aliased) to achieve a time domain matrix of 8K × 512 complex points. These matrixes were apodized in both dimensions with a 90° shifted sinebell, and zero filled to 16 K × 2 K point.

Isotropic samples (D₂O) and C12E5/Hexanol/D₂O samples were shimmed with the TopShim routine, while Chromonic anisotropic samples were heated up to the point where the solution becomes isotropic (²H spectra were recorded to check that the deuterium signal was not a doublet), automatically shimmed using TopShim and then samples were allowed to cool down to the working temperatures.

The coupling constants were extracted from 1D slices of the indirect frequency domain of the 2D spectrum. Each spectrum was duplicated and signals from α- and β-components of the multiplets were shifted relative to each other to reach an ideal overlap of the envelope of the second signal of the multiplet. The magnitude of the shift was the coupling constant value (Kummerloewe et al., 2010).

The residual dipolar couplings (D) were obtained by subtracting the ¹J C-H splitting measured in a F2-coupled HSQC spectrum acquired in isotropic conditions (¹J_CH) to the same coupling obtained in anisotropic conditions (¹T_CH): D 1 CH = T 1 CH ‐ J 1 CH

This then requires the acquisition of two sets of spectra: one in isotropic conditions, and the second in anisotropic media. For the C12E5/Hexanol medium samples, the isotropic values were obtained from a separate sample prepared in D₂O. In the case of the cromolyn samples, both isotropic and anisotropic data were obtained in the same sample, just varying the temperature: 293 K for anisotropic conditions and 308 K for isotropic conditions. Given the spectral resolution, the error in the RDC values was estimated as ca. 1 Hz.

For the MSPIN analysis two files are necessary (Navarro-Vazquez, 2012): a 3D Cartesian coordinates file and a text file with the experimental data which relates every atom pair with their experimental RDC values. Then, six conformers with different ω angle distribution, and their corresponding fluorinated analogues, were generated and energetically minimized using Schrödinger Macromodel software (Figure 2). The 3D Cartesian coordinates of each conformation were correlated with C12E5 and cromolyn experimental RDCs data, taking account all the possible combinations described above.