Docking of 3’SL and 6’SL onto GH33: towards the model of the 3D structure in

To provide a model of the structure in solution, we combined the STD NMR experimental information with molecular docking, using the crystal structure of 2,7-anhydro-Neu5Ac in complex with RgNanH GH33 (PDB ID: 4X4A17_{) as a starting point. First, the docking}

conditions were optimised to reproduce the complex of 2,7-anhydro-Neu5Ac in the GH33 binding as observed in the crystal structure17_.

Figure 5.12. Superimposition of the lowest energy convergent docking solutions for GH33 in

190 | P a g e The coordinate of GH33 from the X-Ray structure with 2,7-anhydro-Neu5Ac (PDB ID: 4X4A) were used to generate the receptor grid; then, 3'SL and 6'SL were docked. Both ligands converged to lowest energy solutions showing the Neu5Ac ring fitting deeply in the catalytic cavity (with glide gscore and glide emodel of circa -5.0 kcal/mol and -55 kcal/mol, respectively, comparable for both ligands). For 3’SL, convergence of the first 10 lowest poses was observed, whereas convergence of the first 6 poses was observed for 6’SL (Figure 5.12). The docking tables are reported in Section A.10 of the Appendix.

The solutions with higher energies presented the ligands flipped by 180°, fitting the glucose in the catalytic cavity, instead. These solutions could safely be excluded based on the experimental STD NMR evidences showing that GH33 does mainly recognise the sialic acid residue (Figure 5.6).

Figure 5.13. (a) 2,7-anhydro-Neu5Ac (in orange) from PDB ID: 4X4A, overlapped with the

Neu5Ac rings of the best docking solutions for 3’SL (cyan) and 6’SL (purple). Only protein residues within 3 Å are shown and labelled. C8 and C9 are pointed by arrows. Protons are omitted for clarity (here and in the following molecular schemes). (b) ΔDEEP-STD for 2,7- anhydro-Neu5Ac87 _{and the Neu5Ac rings of 3’SL and 6’SL, following the same colouring scheme}

as in (a) (adapted from Figure 5.8).

In both 3’SL/GH33 and 6’SL/GH33 models, the Neu5Ac ring was found in a comparable orientation to 2,7-anhydro-Neu5Ac in the crystal structure. Interestingly, the glycerol moieties of the tri-saccharides were further away from the Tyr667 and Trp698 residues

191 | P a g e (resonating at 6.55 ppm) than the polyhydroxy chain of 2,7-anhydro-Neu5Ac (whose hydroxyl on C7 is covalently bound to C2 because of the intramolecular trans - glycosylation). In agreement with the DEEP-STD fingerprint data, H8 and H9s of 6’SL (purple in Figure 5.13) appear to point in the opposite direction relative to the same protons of 3’SL (cyan in Figure 5.13).

It is worth noticing that the main difference between the binding of 3’SL and 6’SL is in the orientation of the lactose moiety. The α2/3 glycosidic linkage directs the galactose of 3’SL to a very efficient π-stacking with Trp698. Figures 5.14a and 5.15a provide two views of 3’SL π-stacking to Trp698 and shows how the galactose region encompassing C3, C4 and C5 point towards the tryptophan side chain (in magenta, resonating around 6.55 ppm, as discussed in Chapter 3). This is in strong agreement with the DEEP-STD data reported with both figures (Figures 5.14a and 5.15a).

Figure 5.14. Lowest energy docking solutions for 3’SL (a) and 6’SL (b) bound to GH33. For the

purposes of comparability, both 3’SL and 6’SL are shown in (b). Top view of the galactose orientation relative to Trp698 (shown in magenta). The ΔDEEP-STD histograms (0.60 ppm/6.55 ppm) of the galactose protons of both ligands (adapted from Figure 5.8) are shown at the bottom. Positions of ligands protons with negative ΔDEEP-STDs are pointed by arrows and labelled to show their close contact to the aromatic residue Trp698.

192 | P a g e The glucose moiety of 3’SL is pointing more towards the solvent, again in accordance with low ΔDEEP-STDs as shown in Figure 5.15a: if the ligand protons are solvent exposed, they will not be strongly affected by differential irradiation frequencies, although the strong positive ΔDEEP-STD observed for the glucose-H4 of 3’SL remains to be investigated.

For 6’SL, the larger flexibility of α2/6 glycosidic linkage makes the galactose ring to sit in an orientation rather perpendicular to the galactose of 3’SL and therefore to Trp698 (Figure 5.15b). This results in galactose H5 and H6s protons pointing towards the tryptophan, while the rest of the galactose protons point to the other side of the binding pocket. This is in agreement with the DEEP-STD data reported below both figures (Figures 5.14b and 5.15b).

Figure 5.15. Lowest energy docking solutions for 3’SL (a) and 6’SL (b) bound to GH33. For the

purposes of comparability, both 3’SL and 6’SL are shown in (b). Side view of Trp698 (shown in magenta) relative to the galactose of 3’SL and the glucose of 6’SL. The ΔDEEP-STD histograms (0.60 ppm/6.55 ppm) of the glucose protons of both ligands (adapted from Figure 5.8) are shown below each figure. Ligands protons with negative ΔDEEP-STDs are pointed by arrows and labelled to show their close contact to the aromatic residue Trp698.

193 | P a g e The distinct bending of 6’SL imposed by its flexible glycosidic linkage brings its glucose ring closer to Trp698 than the adjacent galactose ring: glucose C2 and C6 point towards the tryptophan, as one would expect from the negative DEEP-STD factors observed for H2 and one of the H6s (Figure 5.14b).

To further support that mutation of aspartate to alanine at position 282 did not affect the binding mode of the sialoglycans into the GH33 binding pocket, the mutant was reproduced computationally based on the crystal structure of GH3317 _{by mutating the}

single aspartate 282 residue to an alanine on Maestro Schrodinger (refer to Materials and Methods, Sub-section 5.5.3 for details). From this structure, a new grid was generated, with size and coordinates identical to the WT grid. The docking of 3’SL and 6’SL was repeated in the newly obtained grid, and the resulting poses and clus tering perfectly matched those obtained for the wild type GH33. This result suggests that, whereas the D282A mutation partially inactivates the enzymatic activity, it does not appear to affect the binding mode of 3’SL and 6’SL in the catalytic cleft. This is something important regarding the studies of ligand binding to enzymes: binding in many cases can be considered a separate event from the enzymatic reaction, meaning that the presence or absence of catalytic residues might not impact significantly the binding modes of ligands. This explain why ligands of similar nature can bind in the same catalytic site with very different enzymatic outcomes (reaction or no reaction at all). For example, this situation was also described by NMR regarding the binding of ligands to the human blood group B galactosyltransferase (GTB): UDP-Gal and UDP-Glc sugar nucleotides bind equally well to the enzyme, whereas the reaction only takes place on the first one (the natural substrate).

5.2.5 CORCEMA-ST simulations on 3’SL/GH33 and 6’SL/GH33: validating the 3D structures