Elucidating the Curtin-Hammett Principle in Glycosylation Reactions: The Decisive Role of Equatorial Glycosyl Triflates

Abstract

The glycosylation reaction represents a crucial and challenging reaction used in oligosaccharide synthesis. Specifically, attaining complete stereocontrol during glycosylation reactions remains challenging. Its complex nature is defined by the highly reactive intermediates that form upon the activation of a glycosyl donor. Low-abundant species may afford the major product via Curtin-Hammett kinetics and have long been proposed to play a major role. Therefore, characterizing these elusive stereodirecting intermediates is key to understanding glycosylation reaction mechanisms. Herein, we applied a combination of (exchange) NMR techniques to establish the equilibration rates of glycosyl triflate reaction intermediates and their ensuing glycosylation reaction kinetics. To this end, we studied the glycosylation reactions of 6,3-mannuronic acid and 6,3-glucuronic acid lactone donors. Using the complete set of reaction kinetics data, we constructed a computational kinetic model that shows that these compounds indeed react according to a Curtin-Hammett scenario. Furthermore, we were able to rationalize the observed stereochemical reaction outcomes using quantum-chemically computed potential energy surfaces for these glycosylation reactions. Hence, this workflow can now be used to obtain a complete reaction kinetics overview to retrieve the reaction pathway(s) that drive product formation.

1

Illustrations of scenarios I–III with the axial triflate as the major species and equatorial triflate as the minor species. In scenario I, product formation originates from the main reaction intermediate observed. The red (dotted) line represents a kinetic quenching (KQ) scenario. In scenario II, product formation derives from both axial and equatorial glycosyl triflates with poor selectivity. In scenario III, the red line represents a Curtin–Hammett-like scenario, where the ratio A/B changes as the reaction progresses. The interconversion rate is faster than the equatorial product-forming step, but it is similar (k _AB,k _BA ≈ k _BD | k _BD ≫ k _AC) or slower (k _BD > k _AB,k _BA > k _AC) than the axial product-forming step, hence still providing axial selectivity. The black line represents a classic Curtin–Hammett scenario where almost exclusively D is formed via B, while the ratio of A/B remains constant over the course of the reaction.

1. Stereodirecting Glycosylation Reaction Intermediates of 6,3-Mannuronic Acid Lactone 1 and 6,3-Glucuronic Acid Lactone 2

2

¹⁹F EXSY NMR experiments of lactones 1 and 2. (A) Illustration of ¹⁹F EXSY NMR experiments on glycosyl triflates. (B) Dissociation rate kinetics of 1 _ax and 2 _ax as a measure of the triflate reactivity at different temperatures. (C) Theoretical unimolecular and bimolecular mechanisms of triflate dissociation in 6,3-uronic acid lactones. (D) Triflate dissociation rate kinetics of lactones 1 and 2 at −30 °C plotted against an increasing triflate concentration.

3

¹H CEST NMR experiments of lactones 1 and 2 overlaid with ¹H NMR spectra. (A) Illustration of ¹H CEST NMR experiments on glycosyl triflates. (B) ¹H CEST profile of activated mannuronic acid lactone 1. (C) ¹H CEST profile of activated glucuronic acid lactone 2. (D) Illustration of ¹⁹F CEST NMR experiments on glycosyl triflates. (E) ¹⁹F CEST profile of activated mannuronic acid lactone 1. (F) ¹⁹F CEST profile of activated glucuronic acid lactone 2.

4

Graphical visualization of reaction kinetic experiments. (A) Key NMR signals tracked over time. (B, C) Schematic representation of the reaction monitoring tube used for the experiment and its mixing mechanism.

5

Reaction kinetics of preactivated donors 1 and 2. Donors were activated using Ph₂SO/Tf₂O/TTBP in CD₂Cl₂. (A) General reaction scheme featuring nucleophilic displacement of the axial or equatorial glycosyl triflate intermediates. (B, C) Concentrations of intermediates (1 _ax , 1 _eq ; 2 _ax , 2 _eq ) and products (22 _eq –25 _eq ; 22 _ax –25 _ax ) vs time, acquired by time-dependent ¹H NMR for 4 different glycosylation reactions (top row). The solid lines represent the best predictions of the kinetic model that was fitted to the data. Next, the kinetic model allowed to predict the rates of the equatorial-axial interconversion (1 _ax /2 _ax → 1 _eq /2 _eq ; 1 _eq /2 _eq → 1 _ax /2 _ax ) as well as the glycosylation reactions (1 _ax /2 _ax → 22 _eq – 25 _eq ; 1 _eq /2 _eq → 22 _ax – 25 _ax ) (bottom row). We note that the concentration of 2 _eq was below the detection limit of the ¹H NMR.

6

(A) Quantum-chemically computed Gibbs free energy surface in DCM (ΔG _DCM) for the reaction of mannuronosyl donor 1 with EtOH (black) and ⁱPrOH (red). (B) Computed product fractions of 1 with glycosyl acceptor 20 or 21 at varying k ₁ (top), as well as the computed TS energy barriers at varying k ₁ (bottom). TS energies are derived from the rate constants found via the fitting of the kinetic model to the data. For the TS energy of the equatorial product-forming step, we used the difference in TS energies (ΔΔG _ax‑eq ^‡) between the respective TSs involved in the formation of the equatorial and axial product as found in the quantum chemical calculations, which was added to the TS energy of the axial product-forming reaction derived from k ₄. (C) Quantum-chemically computed Gibbs free energy surface in DCM (ΔG _DCM) for the reaction of glucuronosyl donor 2 with EtOH (black) and ⁱPrOH (red); (D) computed product fractions of 2 with glycosyl acceptor 20 or 21 at varying k ₁ (top), as well as the computed TS energy barriers at varying k ₁ (bottom). (E) In analogy to the 4 different simulations shown in (B, D), the kinetics are simulated while varying the relative differences in ΔΔG ₁ ^‡ and ΔΔG ₂ ^‡ (left scheme). The 2D heat maps indicate the relative excess in products C and D (i.e., ([C] – [D])/([C] + [D], with +1 indicating only product C formed, and −1 only product D formed). The location pins indicate the values of ΔΔG ₁ ^‡ and ΔΔG ₂ ^‡ that have been determined upon analysis of the experimental kinetic data, complemented by the values of ΔΔG _ax‑eq ^‡ found via the quantum chemical calculations for the reaction of mannuronosyl donor 1 (in analogy to the analysis shown in B).