Conformationally supple glucose monomers enable synthesis of the smallest cyclodextrins

See allHide authors and affiliations

Science  17 May 2019:
Vol. 364, Issue 6441, pp. 674-677
DOI: 10.1126/science.aaw3053

Sweet spot for making oligosaccharides

Sugars pose a challenge for chemists: how to string together functional group–rich building blocks that can adopt multiple conformations. Two papers in this issue used sugar building blocks constrained by a macrocyclic linker to encourage formation of a specific glycosidic linkage (see the Perspective by Pohl). Ikuta et al. used glucose building blocks containing a linker that changes the sugar conformation to synthesize cyclic oligomers with only three or four units. The linker changes the conformation of the glucose monomers, enabling them to come together despite the strain in the final structure. Komura et al. prepared sialic acid building blocks with a linker that allows for selective formation of the α-anomeric linkage with a range of nucleophiles. They synthesized dimers of sialic acid with many different linkages and a pentamer with four α(2,8) linkages. This method enabled chemical synthesis of components of mammalian glycans involved in brain development, cell adhesion, and immune response.

Science, this issue p. 674, p. 677; see also p. 631


Cyclodextrins (CDs) are cyclic oligomers of α-1,4-d-glucopyranoside and are known mainly as hexamers to octamers. The central cavities of CDs can retain small molecules, enabling diverse applications. The smallest members, CD3 and CD4, have ring sizes too small to permit the most stable conformations of glucopyranose and have not been accessible synthetically. In this study, we present methods to chemically synthesize both CD3 and CD4. The main factor in the successful synthesis is the creation of a glucopyranose ring conformationally counterbalanced between equatorial- and axial-rich forms. This suppleness is imparted by a bridge between O-3 and O-6 of glucose, which enables the generation of desirable, albeit deformed, conformers when synthesizing the cyclic trimer and tetramer.

Cyclodextrins (CDs) are cyclic oligomers of d-glucose (1), and CD6 to CD8 (1) (Fig. 1A) are produced enzymatically in bulk (2), making them easily available and nontoxic (35). These properties, along with the ability of CDs to capture small molecules within their hydrophobic central cavities, have permitted diverse applications in industry, medicine, and consumer products. Larger CDs, up to CD35, have been characterized (6). Conversely, the smallest known CD formed through chemical synthesis is CD5 (2) (7). CD3 (3) and CD4 (4) have been discussed without the actual compounds being known. In 1957, French said that space-filling models of 3 to CD-infinity could be constructed when the glucose units had conformational flexibility and that the smallest CD produced by the treatment of glycogen with Bacillus macerans amylase was most likely CD6 (8). In 1970, Sundararajan and Rao reported, on the basis of computational calculations, that CDs having fewer than six glucose molecules could not be cyclized because of steric overlap (9). However, Nakagawa et al. synthesized 2 in 1994 (7). In the following year, Immel et al. indirectly concluded that the difficulty of the synthesis of 3 and 4 is due to the strained glucose units (10). Despite the synthesis of CD-like molecules with smaller rings (1114), the synthesis of 3 and 4 remained an unmet goal. Here, we report the chemical synthesis of these small CDs.

Fig. 1 Structures of CDs and key elements enabled the synthesis of 3 and 4.

(A) Because of the strained glucopyranose rings in 3 and 4, their existence was considered implausible. (B) Conception of the EDB bridge and the α-selective glycosylation attributed to the EDB bridge. The α-selective glycosylation using 5, which has the 3,6-O-o-xylylene bridge, lacked clarity and effectiveness. A desire to improve the reaction led to the 3,6-O-EDB–bridged 8. The glycosylation reaction with 8 proceeded efficiently with α selectivity. (C) Pyranose was made supple by the formation of the 3,6-O-EDB bridge. Because the conformation of the EDB-bridged compounds is not constant, we hesitated to adopt the conventional notation of carbohydrates based on the chair form. For the synthesis of 8 to 12, see SM-9–14. For the determination of each conformation, see SM-8–14. Bn, benzyl; Et, ethyl; MS, molecular sieves; Ph, phenyl; rt, room temperature.

In our synthesis of 3 and 4, one of the decisive factors for success was the adoption of the 3,6-O–EDB [1,1′-(ethane-1,2-diyl)dibenzene-2,2′-bis(methylene)] bridge (Fig. 1B), which was introduced to improve α-selective glycosylation using 5 (15, 16). The bridge in 5 arches over the β face of the pyranose ring and hinders the β face approach of an alcohol, yielding the corresponding product with high α selectivity under kinetic conditions with the use of Cp2ZrCl2 (where Cp is cyclopentadienyl) and AgClO4 in the presence of 4-Å molecular sieves (Suzuki glycosylation) (17). However, the bridge locked the pyranose ring into the 3S1 form, which directed the 2-O-benzyl group axially toward the α face, thus inducing adverse steric hindrance as in 6. The α selectivity and the yield of glucosides 7 therefore decreased with the use of elevated reaction temperatures and more sterically hindered alcohols. We suspected that this issue could be resolved by increasing the steric hindrance at the β face and reducing the overhang of the 2-O-benzyl group on the α face. To satisfy both of these requirements, we planned to modify the pyranose conformation by introducing a longer bridge than the o-xylylene group. With this consideration in mind, we chose an EDB group. The glycosylation reaction using the EDB-bridged glucosyl fluoride 8 afforded the corresponding glucosides with α selectivity even at room temperature [supplementary material sections 2 and 19 to 24 (SM-2, -19–24)]. The reaction proceeded through the corresponding oxocarbenium ion intermediate, as similar α selectivities were observed with 8-α and 8-β (SM-15, -16).

The other key element for synthesizing 3 and 4 was the discovery of the supple pyranose system. The most stable conformation of d-glucopyranose is 4C1. On the other hand, the attachment of a bridge between the two discontiguous oxygen atoms on the pyranose ring produces a bicyclic skeleton in which the newly formed ring modulates the conformation of the pyranose scaffold. A short bridge locks the conformation into a motif with more axial substituents, as seen in 5 and 7 (18) and others (1923). By contrast, when the O-3 and O-6 atoms were bridged by the EDB group, the pyranose conformation was modified by subtle structural alteration (Fig. 1C), revealed by the 1H nuclear magnetic resonance (NMR) coupling constants of 8 to 12 (SM-8–14). Thus, although the difference between the diols 9-α and 9-β relates only to the anomers, the conformations of the pyranose moieties were in 1C4 and 4C1 forms, respectively. In the case of dibenzylated compounds 8 and 10 to 12, where the anomeric substituents were varied, the conformations of the pyranose systems were widely distributed, as displayed on a map of conformations that puts 1C4 and 4C1 forms on both poles (Cremer-Pople-Stoddart coordinates) (24, 25). We propose that the length of the EDB bridge is appropriate to equally balance the innate preference of glucose for the equatorial-rich 4C1 form and the tendency of the bridge to transform the pyranose ring into axial-rich forms. The α selectivity featured in the reaction using 8 and the supple pyranose ring set the stage for the synthesis of strained CDs.

As 3 and 4 repeat α-1,4 linkages, we protected oxygen atoms other than the bonding sites in their synthesis. The EDB group already protected O-3 and O-6 (Fig. 2A, box), so we added protection for O-2 (P2). Full deprotection of 13 yielded the desired product, 4. We obtained the cyclized intermediate 13 by intramolecular glycosylation of a linear tetramer (a) produced by the glycosylation of 15 and 16, alteration of protecting circumstances, and addition of a leaving group. The dimers 15 and 16 were derived from starting monomers 18 and 19. To equip the EDB bridges of 18 and 19, the synthesis began with 1,2,4-orthoacetylglucose (20) (26), the O-3 and O-6 of which are locked in the same direction to ease construction of the bridge. The precursor of 3 was 21, which was produced by intramolecular glycosylation of a linear trimer (b) formed from 15 and 19. The suppleness caused by the EDB bridge enabled the cyclizations leading to 13 and 21, which we suggest are not possible when the glucopyranosyl moieties are in the 4C1 form. Because α selectivity is essential for synthesizing CDs (27, 28), we apply the fluorine atom for the leaving group on the basis of the glycosylation of 8 (Fig. 1B).

Fig. 2 Synthesis of 3 and 4.

The colors in the frame around “G” indicate the corresponding reactions in (A to D). The pyranose conformations of 3, 4, 13, 21, and 25 were determined on the basis of 1H NMR coupling constants (SM-48, -43, -41, -46, and -35, respectively). In the Oak Ridge thermal ellipsoid plot (ORTEP), waters of crystallization were omitted. For a description of protecting groups (Px), see SM-5. Ac, acetyl; DMF, N,N-dimethylformamide; DMP, 2,6-dimethylphenyl; J, coupling constant; Me, methyl; THF, tetrahydrofuran.

Points in the synthesis of the disaccharide 17 (Fig. 2B) (SM-25–33) are as follows. Bisetherification of 20 with 2,2′-bis-(bromomethyl)-dibenzyl (29) furnished the EDB-bridged 23. Keeping the concentration of 20 lower than 10 mM ensured reproducibility. The indium bromide promoted β-specific introduction of the arylthio groups, which, accompanied by the cleavage of the orthoester 23 followed by deacetylation, provided 24 and 9-β, which were converted into 18 and 19, respectively. The glycosylation using 18 and 19 provided the dimer 17 with perfect α selectivity. We attribute the higher selectivity observed in the use of 19 than in the use of 8 (Fig. 1B) to the reduction of steric hindrance at the α face by the alteration of P2, which tends to overhang on the α face, from the benzyl group to the sterically smaller allyl group. We confirmed the α stereochemistry of 17 by transformation to 25, whose pyranoses were 4C1 (SM-3).

We synthesized 4 (Fig. 2C) from dimers 15 and 16, derived from 17 (SM-37–43). The stereochemistry of the formed glycosidic bond in tetramer 14 was undetermined because of the overlapped NMR signals (SM-61). However, the 1H and 13C NMR spectra of cyclized 13 (SM-60) were in agreement with the pattern of a monosaccharide, indicating the unified stereochemistry of the four anomeric positions. These linkages were α because 13 integrated 17, whose 1′-α stereochemistry was confirmed. The reproducibility and yield of the intramolecular glycosylation were poor (6 successes to 17 failures). Removal of the allyl (30) and EDB groups from 13 afforded 4 through 27, demonstrating successful removability of the EDB group by hydrogenolysis. Mass spectrometry of 4 indicated the desired molecular ion peak for the cyclic tetramer (SM-50). 1H NMR spectra revealed that the conformations of the pyranose rings of 13 and 4 were 2H1 and between 4C1 and 2H1 (see 4 in Fig. 1A), respectively (SM-41, -43). The pyranose in these compounds is distorted and flatter than that of larger CDs (3133).

In the synthesis of 3 (Fig. 2D) (SM-44–48), the intramolecular glycosylation was more efficient than that of 4, with the yield of 21 reaching 88%. The 1H and 13C NMR, high-resolution mass spectrometry (SM-49), and x-ray diffraction studies of a single crystal (SM-48) confirmed that 3 was the cyclic trimer. The NMR spectra of 21 and 3 (SM-68, -49) were consistent with equivalent sugar units, and averaged pyranose conformations were E1 in acetone-d6 and 5S1 in D2O (see 3 in Fig. 1A), respectively (SM-46, -48). The x-ray diffraction study of a single crystal of 3 indicated that the conformations of the three pyranoses were different (two with 5S1 and one between 4C1 and OH1) in the crystal lattice (figs. S7 to S9). The interatomic distances among three O-4’s, O-5’s, and H-5’s (figs. S10 to S12) suggest that there is essentially no cavity in the center of the molecule when considering the robes of lone pairs on the O-5’s. The lower field shift of H-5 in the 1H NMR spectrum of 3 than in those of CD6 to CD8 (fig. S5) may be a result of a deshielding effect induced by the O-5’s.

The use of glucose monomers with conformational flexibility, which we term suppleness, allowed for the synthesis of the strained CDs 3 and 4. The creation of suppleness in sugars by the addition of another ring is potentially applicable to the synthesis of other strained compounds or to contexts where function requires flexible structure. The averaged C3 and C4 symmetry observed in NMR spectra of 3 and 4, where multiple stereocenters exist, could be useful in the construction of molecular catalysts or metal-organic frameworks. CDs have general relevance for applications that take advantage of their capacity to hold molecules within the central cavity. We expect that the smaller cavity of 4 will permit selective inclusion of molecules smaller than those accommodated in currently available CDs.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S12

Tables S1 and S2

References (3441)

References and Notes

Acknowledgments: We thank D. Tanaka and Y. Kamakura, Kwansei Gakuin University, for help with the x-ray diffraction study and G. A. Cordell, Natural Products and the University of Florida, for his review of the manuscript before submission. Funding: The research reported here was supported by the MEXT in Japan through the program for the Strategic Research Foundation at Private Universities (S1311046), JSPS KAKENHI (grants JP16H01163 in Middle Molecular Strategy and JP16KT0061), and the Naohiko Fukuoka Memorial Foundation. Author contributions: D.I. and H.Y. conceived the work. D.I., Y.H., S.W., H.S., Y.T., Y.K., K.I., T.H., and S.M. performed the experiments. All members of the team designed the experiments and analyzed the data. S.W. and H.Y. wrote the manuscript. D.I., S.W., Y.T., K.I., T.H., and S.M. assisted in writing and editing the manuscript. Competing interests: Kwansei Gakuin University has filed a provisional patent on this work (application no. JP2018-145108). Data and materials availability: All data are presented in the main text or supplementary materials. The x-ray data and model for 3 are deposited in the Cambridge Crystallographic Data Centre (CCDC) under reference number 1878307.
View Abstract

Stay Connected to Science

Navigate This Article