Discovery of selective monosaccharide receptors via dynamic combinatorial chemistry

The molecular recognition of saccharides by synthetic hosts has become an appealing but elusive task in the last decades. Herein, we combine Dynamic Combinatorial Chemistry (DCC) for the rapid self-assembly and screening of virtual libraries of receptors, with the use of ITC and NMR to validate the hits and molecular modelling to understand the binding mechanisms. We discovered a minimalistic receptor, 1F (N-benzyl-l-phenylalanine), with considerable affinity for fructose (Ka = 1762 M−1) and remarkable selectivity (>50-fold) over other common monosaccharides. The approach accelerates the discovery process of receptors for saccharides.

General procedure for the preparation and analysis of DCC experiments 2 (1 mM) was mixed with A, B, W, F, and D (4 mM each) in 100 mM carbonate buffer pH 10 90%-10% MeCN.Then, the reaction mixture was split into five 1 mL Eppendorf tubes and to each one of them it was added 10 L of a water solution of one sugar (glucose, mannose, galactose, or fructose, 2 mM in reaction mixture); as well as the same volume of distilled water to the fifth tube.The mixtures were left stirring O.N.Then, NaCNBH 3 (20 mM in the reaction mixture) was added to the Eppendorf tubes and they were stirred for 30 min before LC-MS analysis.
The m/z of the (M+H) + ion for each and every one of the library members was searched in the extracted ion chromatogram (EIC) for all the LC-MS runs.
In the case of monofunctionalised library members, there was the possibility of NaCHBH 3 reducing either the imine (2X, figure 2), the unreacted aldehyde, or both groups (2X-OH, Figure 2).LC-MS confirmed the absence of 2X-OH.Both mono-reduced products would be indistinguishable by LC-MS by looking at their m/z as they would have the same mass.However, the product of reduction of starting aldehyde 2 was not detected by LC-MS, suggesting that in the conditions employed, NaCNBH 3 did not reduce the aromatic aldehydes.Therefore, we can conclude that the mono-reduced library members are indeed the secondary amines 2X.
The intensity of the peaks (measured as the number of ions detected on the EIC) for each library member in a templated DCL was divided by the intensity of the same peak in the blank experiment.The greater the value obtained as a result of such operation, the larger the amplification caused by such template to that library member.
As examples, the LC-MS UV and TIC chromatograms, as well as the EIC for the library member 2DD in the blank run as well as in the four templated runs are included here (Figure S1-S4).
The same procedure for the EIC of 2F with fructose is attached too (Figure S5).L-Aspartic acid (D, 652.0 mg, 4.9 mmol) or L-phenylalanine (F, 821.7 mg, 4.9 mmol) was slowly dissolved in a solution of NaOH (0.4 g) in water (5 mL) and methanol (10 mL).Then, benzaldehyde (1, 508 uL, 5.0 mmol) was slowly added to the solution.The reaction mixture was stirred for 1h.NaBH 4 (226.0mg, 6 mmol) was dissolved in methanol (1 mL) and slowly added to the reaction mixture.Reaction left stirring for 1h before acidifying it to pH 5-6 with acetic acid.The solvent was evaporated under reduced pressure.The resulted oils were purified by HPLC to afford 1D (907.2 mg, 83%) and 1F (950.1 mg, 76%) as white solids.HPLC purification method was performed on C18 column and was optimised as follows: A 95%-5% B isocratic for 25 minutes, with a retention time for 1D of 9.1 minutes and for 1F of 20.0 minutes.Being A: water + 0.1% TFA and B: acetonitrile + 0.1% TFA.

S-16
Isothermal titration calorimetry (ITC) binding studies ITC experiments were performed with 2 mM receptor solution (D, F, 1D, 1F, 2DD, 2FF) and 80 mM solution of ligand (D-glucose, D-mannose, D-galactose, D-fructose).100 mM carbonate buffer (pH 10) was employed as solvent.The buffer was degassed prior to solution preparation.The solutions were filtered and degassed prior to use.ITC experiments were performed with the parameters reported in Table S1 by titrating the ligand into the receptor solution.Each experiment consists of 3 titrations and the heat of dilution was measured and subtracted for each experiment.

Molecular modelling studies
All the molecular models were performed in Maestro version 13.6.121,MMshare Version 6.2.121, Release 2023-2, from Schrödinger.Structures for the complexes between 1F and fructose were built with 1F as its monoanionic form (free amine and deprotonated carboxylate) and selected isomers of the saccharide.D-fructose exists as a mixture of five isomers in aqueous solution: three tautomers (linear keto and cyclic five/six-member rings), and two anomers (α/β forms) of each corresponding cyclic hemiacetal.The relative distribution of the different isomers has been thoroughly reported in literature under different experimental conditions, with consistent values of β-D-fructopyranose (65-70%), β-D-fructofuranose (22-25%), α-D-fructofuranose (5-6.5%),α-D-fructopyranose (0.5-3%) and linear ketose (<1%) as the species present in equilibrium in water.i Since we did not observe important changes in the distribution of the species by 1 H NMR upon addition of 1F, we focused the modelling studies on the four cyclic isomers (accounting for >99% species).The complexes were submitted to a conformational searches with the MCMM/LMCS sampling approach that combines Monte Carlo conformational searches with low mode sampling.ii The generated geometries were optimized with the OPLSE4 force field iii in implicit water.iv This protocol produced over 250 local minima for each complex.These minima were energetically ordered and selected representative structures with lowest energy were further optimized by Jaguar package in Maestro suite using DFT calculations at the B3LYP-D3 v level of theory with a 6-31G** basis set.Both gas-phase and C-PCM vi solvation model (water) were used in order to obtain a theoretical estimation of the solvation stabilization energy.The same procedure was also applied to free monoanionic 1F, β-D-fructopyranose, β-D-fructofuranose, α-D-fructofuranose and α-D-fructopyranose to estimate the corresponding binding energies.

1F/β-D-fructopyranose complexes
In this case three representative complexes were identified as depicted in Figure S29 and the relative energies are shown in Table S2.Table S2: Gas-phase and aqueous solution energies, alongside the theoretical solvation stabilization energy for complexes 1, 2 and 3 formed between receptor 1F and saccharide βfructopyranose (as depicted in Figure S29).The differences in energies between complex 1 and 2, as well as in complex 1 and complex 3 are also shown.The three complexes are characterized by stabilizing double H-bonds between the 1Fcarboxylate and the hemiacetal anomeric and C1 hydroxyl groups (dashed lines in Figure S29).The two most stable complexes (complex1 and complex2) are very similar in their geometry.
In both, the 1F receptor shows a V-shaped conformation with the two aromatic rings (Phe side chain and benzyl residue) pointing towards the sugar, thus defining, along with the carboxylate group, a tweezer for the saccharide.
Complex3 is an open structure stabilized exclusively by H-bonds.By looking to the corresponding energies (both in gas phase and in water), remarkable conclusions can be drawn.First, the extra-stabilization of complex1 is mainly due to the effect of water, which allows us to hypothesize that the tweezer conformation is promoted by the corresponding C-H••• interactions, that are favoured in aqueous media.Besides, complex1 shows a more efficient C-H••• contact, explaining the reversed stability in comparison with complex2 in gas phase versus solution.

1F/β-D-fructofuranose complexes
Two complexes were optimized at DFT level (see Figure S30) named as complex1 (left) and complex2 (right).Their structures are shown below and the absolute and relative energies of these two complexes are shown in Table S3.Table S3: Gas-phase and aqueous solution energies, alongside the theoretical solvation stabilization energy for complexes 1 and 2 formed between receptor 1F and saccharide βfructofuranose (as depicted in Figure S30).The differences in energies between complex 1 and 2 are also shown.The results show that complex1 is more stable than complex2 by 2.2 kcal/mol.Also data shows that this extra stabilization sources from the effect of the solvent (compare gas-phase vs. solution energies).The binding in both cases mainly occurs through H-bonding while in complex1 also a sugar C-H••• interaction can be proposed with the side chain of Phe amino acid.This extra-stabilization would explain the solvation effect, since it 'hides' the hydrophobic moiety of the sugar from water.Complex2 is more open and no hydrophobic interactions are present.

Complexes with the minor α-isomers
The corresponding complexes formed between 1F and either α-D-fructofuranose and α-Dfructopyranose were also optimized following the same protocol as previously described.The obtained structures and energies are shown in Figure S31.Table S4: Gas-phase and aqueous solution energies, alongside the theoretical solvation stabilization energy for the complexes 1F with α-fructofuranose and 1F with α-fructopyranose (as depicted in Figure S31).

Comparison of the two major complex species: 1F/β-D-fructopyranose vs. 1F/β-Dfructofuranose
The relative energies for the complexes formed with the two main isomers of fructose in aqueous solution (Table S5) shows that the complex with β-D-fructopyranose is favoured over that with β-D-fructofuranose, explaining why the presence of 1F does not dramatically modify the isomers proportion.
Table S5: Substraction of the values of gas-phase energy, aqueous solution energy and theoretical solvation stabilization energy for the complexes 1F/β-D-fructopyranose minus 1F/β-D-fructofuranose.

Estimation of the binding energies
As a simple estimation of the binding energies for the optimized complexes, we additionally optimized the geometries of the corresponding components, using the same computational protocol.The obtained minima are below in Figure S32.Table S6 contains the corresponding energies for all the components, 1F and the four main isomers of D-fructose.
Table S6: Gas-phase and aqueous solution energies, alongside the theoretical solvation stabilization energy for the four main isomers of D-fructose as well as for 1F free (Figure S32).The binding energies were then estimated as: S-28 For each complex, both in the gas phase and in water.The results are shown in Table S7.
Table S7: binding energies for the complexes formed between 1F and the four main isomers of D-fructose.The analysis of the estimated binding energies confirms the spontaneous formation of the complexes between 1F and the four isomers of fructose.The solvation effect specifically favours the 1F/β-D-fructopyranose complex, which is the most favoured and the most stable one.This result correlates with the experimental observation that addition of 1F over a solution of fructose in water does not change the NMR spectra of fructose, suggesting no variation in isomeric composition upon binding.
We can propose that the flexible structure of 1F allows to dynamically bind all the isomers of the sugar, for which a conventional receptor design would have been difficult.Moreover, in the two most-stable global minima for the two major tautomers, the aromatic rings of the Phe side chain (β-D-fructofuranose) or both side chain and benzyl residue (β-Dfructopyranose) play a key role in the interaction.This would explain the selection of 1F from the dynamic combinatorial mixture.On the other hand, the folded conformation of 1F (specially for β-D-fructopyranose) suggests that the receptor conformationally adapts to the substrate, leading to a compact structure that efficiently de-solvates 1F and, specially, the sugar.De-solvation of host and guest would account for a favourable entropy of binding as experimentally measured by ITC.Finally, the concerted H-bonding of carboxylate of 1F with anomeric OH and C1-OH can only be proposed with the ketose six-member ring, thus suggesting an explanation for the strong binding to fructose vs. the other hexoses, for which such binding motif is not possible.

Figure S1 :
Figure S1: UV (bottom, red).TIC (middle, green) and EIC (top, purple) chromatograms for the LC-MS runs of the untemplated DCC (left) and the DCC templated with D-glucose (right).The m/z shown in the EIC chromatogram is the (M+H) + of molecule 2DD (369.34) and the intensity of the peaks are 8.30e6 and 3.15e7 for non-templated and templated runs, respectively.

Figure
Figure S2: UV (bottom, red).TIC (middle, green) and EIC (top, purple) chromatograms for the LC-MS runs of the untemplated DCC (left) and the DCC templated with D-mannose (right).The m/z shown in the EIC chromatogram is the (M+H) + of molecule 2DD (369.34) and the intensity of the peaks are 8.30e6 and 6.52e7 for non-templated and templated runs, respectively.

Figure
Figure S3: UV (bottom, red).TIC (middle, green) and EIC (top, purple) chromatograms for the LC-MS runs of the untemplated DCC (left) and the DCC templated with D-galactose (right).The m/z shown in the EIC chromatogram is the (M+H) + of molecule 2DD (369.34) and the intensity of the peaks are 8.30e6 and 2.96e7 for non-templated and templated runs, respectively.

Figure S4 :
Figure S4: UV (bottom, red).TIC (middle, green) and EIC (top, purple) chromatograms for the LC-MS runs of the untemplated DCC (left) and the DCC templated with D-fructose (right).The m/z shown in the EIC chromatogram is the (M+H) + of molecule 2DD (369.34) and the intensity of the peaks are 8.30e6 and 7.25e6 for non-templated and templated runs, respectively.

Figure S20 :
Figure S20: ITC titration of D and mannose.

Figure S24 :
Figure S24: ITC titration of F and fructose.

Figure S32 :
Figure S32: DFT optimized geometries for the 1F receptor and fructose isomers.

Table S1 .
ITC parameters for the binding studies with TA nano-LV instrument.