Hydrolase–like catalysis and structural resolution of natural products by a metal–organic framework

The exact chemical structure of non–crystallising natural products is still one of the main challenges in Natural Sciences. Despite tremendous advances in total synthesis, the absolute structural determination of a myriad of natural products with very sensitive chemical functionalities remains undone. Here, we show that a metal–organic framework (MOF) with alcohol–containing arms and adsorbed water, enables selective hydrolysis of glycosyl bonds, supramolecular order with the so–formed chiral fragments and absolute determination of the organic structure by single–crystal X–ray crystallography in a single operation. This combined strategy based on a biomimetic, cheap, robust and multigram available solid catalyst opens the door to determine the absolute configuration of ketal compounds regardless degradation sensitiveness, and also to design extremely–mild metal–free solid–catalysed processes without formal acid protons.

X-ray crystallographic data collection and structure refinement. Crystals of 1a@2 and 11a@2 adsorbates were selected and mounted on a MITIGEN holder in Paratone oil, and very quickly placed in a nitrogen stream cooled at 90 K to avoid the possible degradation upon desolvation. Diffraction data were collected on a Bruker-Nonius X8APEXII CCD area detector diffractometer using graphite-monochromated Mo-Kα radiation ( = 0.71073 Å). The data were processed through the SAINT 4 reduction and SADABS 5 multi-scan absorption software. The structure was solved with the SHELXS structure solution program, using the Patterson method. The model was refined with version 2018/3 of SHELXL against F 2 on all data by full-matrix least squares. 6,7 As reported in the main text, even after a single-crystal to single-crystal process, the retained crystallinity of the 3D network of 2, allowed the resolution of the crystal structures of 1a@2 and 11a@2. In particular, in case of 1a@2 it must be underlined that crystals suffered also temperature of 50 °C for 48 hours. For that it is reasonable to expect somewhat mismatches from the routine expected diffraction patterns. Nevertheless, parameters indicating data quality (see also Supplementary Table 1) are as the following for 1a@2 and 11a@2, respectively: resolution, 0.75−0.91 Å; Rint, 6.16%− 8.93%; R(all data), 8.18%−8.13%; Flack parameter, 0.12(2)-0.13 (2); standard uncertainties of C−C bond lengths, 0.0220 Å. Within the limit of X-ray crystallography, we are confident that the structures found are consistent even more bearing in mind the multi-techniques approach that pervades characterizations in the whole paper.
In the refinement of both crystal structures all non-hydrogen atoms were refined anisotropically except some highly dynamically disordered atoms of serine moieties pointing towards pores refined on two positions [O2H and O2H'] and guest molecules, together with lattice water molecules. Some restrains to make the refinement more efficient have been applied, for instance ADP components have been restrained to be similar to other related atoms, using SIMU 0.04 for disordered sections or EADP for group of atoms of the guest molecules expected to have essentially similar ADPs. In particular in 1a@2 the highly diffuse electron density suggests that C7C and C9C sites are disordered and shared with disordered solvent molecules, which were modelled as overlapped positions with EXYZ and EADP and their occupancy factors have been imposed accordingly (see Supplementary Fig. 5 for details). The high disorder detected is likewise at the origin of mismatch between expected and experimental values in some C-C bond lengths.
In 11a@2 atoms sharing the same site (in consequence of statistic disorder C1L with O2' and C8L with O7) have been refined as overlapped positions with EXYZ and EADP and their occupancy factors have been imposed accordingly. For both structures, the hydrogen atoms of the net were set in calculated position and refined isotropically using the riding model. Hydrogen atoms on the guest molecules and for solvent lattice molecules were neither found nor calculated. These molecules are expected to be severely disordered as a direct consequence of their high thermal motion and exhibit also statistic disorder [in 1a@2 O4W and O5W has been refined on two sites as O3W in 11@2]. In general guest molecules are severely disordered, especially for the guest's fragments pointing towards the center of the pores where, undoubtedly, the degrees of freedom, related to diverse possible conformations, increase. It is well known that a crystal structure is the spatial average, representing all molecules, together with all their possible orientations averaged, in the crystal via only one-unit cell. In all cases, as the present one of as synthesized porous materials, where obviously not all unit cells are identical and a variety of orientation are allowed, the description became more challenging. The occupancy of the guests in the pores was found via a free variable and later fixed at the converged occupancy. The use of some C-C and C-O bond lengths restraints of highly disordered atoms for guest molecules during the refinement has been reasonably imposed, as well as related to the expected and severe thermal motion, likely depending on the large size of the huge pores of the frameworks (SADI, DFIX, DANG, SIMU, and DELU).
The guest molecules are also statistically disordered. In 1a@2 the highly diffuse electron density did not allow to model some terminal groups of acetate moieties, suggesting they are affected by very large thermal motion (see Supplementary  quite normal for crystals that suffered a single-crystal to single-crystal process, or from short intermolecular contacts between water molecules and guest molecules or water molecules or guest molecules and the whole network are unavoidable due to the expected severe disorder of both solvent and guest molecules. Reflections that are affected by the beamstop or having (Iobs-Icalc)/σ > 10 were omitted. The comments for the alerts are described in the CIFs using the validation response form (vrf).
A summary of the crystallographic data and structure refinement for 2, 1a@2 and 11a@2 are given in Supplementary Table 1. CCDC Deposition Number are 1985884- 7 The final geometrical calculations on free voids and the graphical manipulations were carried out with PLATON 8,9 implemented in WinGX, 10   was then added and the pH and time were recorded. The hydrolysis was allowed to progress until the pH had dropped to 7.2, corresponding to consumption of approximately 50% of the starting ester. The pH was then adjusted to 8.0 with NaOH 10% and the aqueous solution extracted with 3 x 15 mL of ether. The organic phase was then dried over MgSO4 and filtered, and the solvent was evaporated to provide the unhydrolyzed ester fraction, which may be analyzed for enantiomeric purity without further purification. The aqueous phase was acidified to pH 2 with 1 N H2SO4 than again extracted with 3 x 15 mL. The organic phase was dried over MgSO4 and filtered, and concentrated under vacuum. and the solvent evaporated to provide the hydrolyzed ester fraction as a mixture of both carboxylic acid chiral compounds enriched in the S enantiomer (ca. 80%) respect to the R enantiomer (ca. 20%).

Molecular Recognition
The substrate 11 has been docked into a section of the crystal representing the minimal unit of MOF 2 (32.4 x 32.3 x 49.7 Å 3 ) adopted in the theoretical investigation. AutoDock Vina 14 code has been used in the molecular recognition and 10 output poses have been generated. Box centroid has been determined by a geometric center of the six serine moieties involved into the substrate-binding region and a box of 12.7 Å size for X, Y and 49.7 for Z was used for grid point generation. Each structure has been analyzed and the best docked pose with the lowest binding energy was selected for higher level DFT/MM mechanistic investigations.

QMMM calculations
The reaction mechanism has been investigated with one serine included in the high layer  Supplementary Table 1

Supplementary Figures
Supplementary Fig. 1 Supplementary Fig. 18. Perspective view of a single channel in crystal structure of 11a@2 emphasizing pores filled by guest molecules (grey sticks with the only exception of oxygen atoms, which are depicted as red spheres). The fragment of the 3D networks is also depicted as grey sticks, with the only exception of copper(II), calcium(II) and serine residues oxygen atoms which are represented as green, blue and red sticks. The H-bond interactions are depicted as red dashed lines. Lattice water molecules are not depicted for the sake of clarity except those "locked" being involved in the stabilization of the final serine residue configuration. This unique configuration turns more acid the alcoholic moiety. The crystallographically distinct serine arm, not so blocked, shows statistical disorder (see details of refinement in Crystallographic section).
Supplementary Fig. 19. Details of host-guest interactions involving the chiral fragment 11a packed within pores via strong and medium H-bonds (depicted as red dashed lines) involving hydroxyl groups of glucose moiety and terminal carboxylic groups linked to the net by serine arms and oxygen atoms from oxamate ligand. Atom colour code: All atoms from the fragment of the coordination network are represented as grey sticks, with the only exception of copper(II) (cyan spheres), calcium(II) (blue spheres) and oxygen atoms from serine residues and oxamate-oxygen atoms participating in the intermolecular interactions (red spheres). For the guest molecule, all oxygen atoms are depicted as red spheres whereas all carbon atomswith the exception of C5L (grey sphere)-are depicted as grey sticks. Interacting free water molecules are represented as red spheres.   Supplementary Fig. 30. The reaction mechanisms (left) followed in the hydrolysis of 11 by three water molecules (in the absence of MOF 2) and related PES (right calculated at B3LYP-D3/6-311+G(2d,2p)|UFF//B3LYP/6-31G(d)|UFF level of theory. The mechanism A describes the cleavage of R1 and R2