Recognition properties of receptors consisting of imidazole and indole recognition units towards carbohydrates

Summary Compounds 4 and 5, including both 4(5)-substituted imidazole or 3-substituted indole units as the entities used in nature, and 2-aminopyridine group as a heterocyclic analogue of the asparagine/glutamine primary amide side chain, were prepared and their binding properties towards carbohydrates were studied. The design of these receptors was inspired by the binding motifs observed in the crystal structures of protein–carbohydrate complexes. 1H NMR spectroscopic titrations in competitive and non-competitive media as well as binding studies in two-phase systems, such as dissolution of solid carbohydrates in apolar media, revealed both highly effective recognition of neutral carbohydrates and interesting binding preferences of these acyclic compounds. Compared to the previously described acyclic receptors, compounds 4 and 5 showed significantly increased binding affinity towards β-galactoside. Both receptors display high β- vs. α-anomer binding preferences in the recognition of glycosides. It has been shown that both hydrogen bonding and interactions of the carbohydrate CH units with the aromatic rings of the receptors contribute to the stabilization of the receptor–carbohydrate complexes. The molecular modeling calculations, synthesis and binding properties of 4 and 5 towards selected carbohydrates are described and compared with those of the previously described receptors.


Introduction
Analysis of the binding motifs found in the crystal structures of protein-carbohydrate complexes [1][2][3][4][5] provides much of the inspiration for the design of artificial carbohydrate receptors which use noncovalent interactions for sugar binding [6][7][8][9][10][11][12][13][14][15][16][17][18]. Such receptors provide valuable model systems to study the underlying principles of carbohydrate-based molecular recogni-tion processes and might serve as a basis for the development of new therapeutic agents (for example, anti-infective agents) or saccharide sensors [19][20][21][22][23][24][25][26]. Our previous studies showed that mimicking the binding motifs observed in the crystal structures of protein-carbohydrate complexes by using natural recognition groups or their analogues [27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45] represents an effective strategy for designing carbohydrate receptors. Among other things the crystal structures of protein-carbohydrate complexes revealed that the imidazole and indole groups of His and Trp respectively are able to participate in both hydrogen bonding and stacking interactions with the sugar ring. It should be noted that packing of an aromatic ring of the protein against a sugar is observed in most carbohydrate-binding proteins [1][2][3][4][5]. Such packing arrangements and the hydrogen bonding motifs shown in Figure 1 have inspired the design of receptors 1 and 2 (see Figure 2), including both 4(5)-substituted imidazole or 3-substituted indole units as the entities used in nature, and 2-aminopyridine groups as heterocyclic analogues of the asparagine/glutamine primary amide side chains (in analogy to the binding motif shown in Figure 1a) [31]. The compounds 1 and 2 were established as highly effective receptors for mono-and disaccharides and shown to display remarkable β-vs. α-anomer selectivity in the recognition of glucopyranosides, as well as a binding preference for β-glucopyranoside vs. β-galactopyranoside. It has been shown that both hydrogen bonding and interactions of the carbohydrate CH units with the aromatic rings of the receptors contribute to the stabilization of the receptor-carbohydrate complexes. Compounds 1 and 2 were shown to be more powerful carbohydrate receptors than the symmetrical aminopyridine-based receptor 3. We were interested to see whether compounds 4 and 5 (see Figure 2), which consist of two imidazole or indole groups and one 2-aminopyridine unit, would be more effective with monoand disaccharide substrates. Herein, we describe the synthesis, molecular modeling calculations and the binding properties of the compounds 4 and 5. To compare the binding properties of the new compounds with those of the previously published receptors, octyl β-D-glucopyranoside (6a), methyl β-D-glucopyranoside (6b), octyl α-D-glucopyranoside (7a), methyl α-Dglucopyranoside (7b), octyl β-D-galactopyranoside (8a), methyl β-D-galactopyranoside (8b), methyl α-D-galactopyranoside (9), methyl α-D-mannopyranoside (10) and dodecyl β-D-maltoside (11) were selected as substrates for the binding experiments (see Figure 3). 1 H NMR spectroscopic titrations in competitive and non-competitive media as well as binding studies in twophase systems, such as dissolution of solid carbohydrates in apolar media, revealed highly effective recognition of neutral carbohydrates and interesting binding preferences of these acyclic receptors.

Synthesis of the receptors
The basis for the synthesis of compounds 4 and 5 was 1,3bis(aminomethyl)-5-[(4,6-dimethylpyridin-2-yl)aminomethyl]-2,4,6-triethylbenzene (17). The synthesis of compound 17 is described in reference [27]. The reaction of 17 with the corresponding carbaldehyde, such as 4(5)-imidazole-carbaldehyde (18) [46] or 3-indole-carbaldehyde (19), provided the corresponding imines 20 and 21, which were further reduced with sodium borohydride. The synthesis of receptors 4 and 5 is summarized in Scheme 1. Binding studies in two-phase systems: liquidsolid extractions The dissolution of solid carbohydrates in apolar media provides valuable means of studying carbohydrate recognition by organic-soluble receptors (for examples of receptors which are able to dissolve solid carbohydrates in apolar media, see references [6,27,41,43,[48][49][50]). Extractions of sugars 6b, 7b, 8b, 9 and 10 from the solid state into a CDCl 3 solution of receptor 4 or 5 (1 mM) provided evidence for strong complexation of β-glucoside 6b and β-galactoside 8b. The extraction of solid methyl α-glucoside 7b, α-galactoside 9 and α-mannoside 10 into a CDCl 3 solution of receptor 4 or 5 indicated a weaker binding of these sugars than that of 6b and 8b (see Table 1). The extraction experiments indicated that the imidazole-based receptor 4 is a more powerful carbohydrate receptor than the indole-based compound 5. Receptor 4 was able to dissolve about 1 equiv of β-glucoside 6b and β-galactoside 8b, 0.5 equiv of α-glucoside 7b and about 0.2 equiv of α-galactoside 9. In the case of receptor 5 only about 0.7 equiv of β-glucoside 6b and β-galactoside 8b could be detected in the solution (see Table 1). Regarding 4 and 5, the extractability decreased in the sequence β-glucoside 6b ~ β-galactoside 8b > α-glucoside 7b > α-galactoside 9 > α-mannoside 10 (see Table 1; control experiments were performed in the absence of the receptor). The preference of 4 and 5 for β-vs. α-glucoside (6b vs. 7b) as well as for β-vs. α-galactoside (8b vs. 9) indicated by liquid-solid extractions was further confirmed by 1 H NMR spectroscopic titrations (see below). Compared to the previously studied receptors 1-3, the extraction experiments indicated a significantly higher level of affinity of 4 and 5 towards β-galactoside. It should also be noted that the selectivities observed for 4 and 5 are quite different to those of the recently described phenanthroline/aminopyridine-based receptors 22 and 23 (see Figure 4) [27,29], which show a strong preference for α-glucoside and α-galactoside vs. the β-anomers. Thus, depending on the nature of the recognition units used as building blocks for the acyclic structures, effective carbohydrate receptors with different binding selectivities could be obtained. However, the exact prediction of the binding selectivity still represents an unsolved problem.

Binding studies in homogeneous solution
The interactions of the receptors and carbohydrates were investigated by 1 H NMR spectroscopic titrations in CDCl 3 and DMSO-d 6 /CDCl 3 mixtures. The stoichiometry of the receptor-sugar complexes was determined by mole ratio plots [51,52] and by the curve-fitting analysis of the titration data [53].
The 1 H NMR titration experiments [54] with octyl β-glucoside 6a, α-glucoside 7a, β-galactoside 8a and methyl α-galactoside 9 were carried out by adding increasing amounts of sugar to a solution of receptor 4 or 5. In addition, inverse titrations were performed in which the concentration of the sugar was held constant and that of the receptor was varied. The complexation between receptors 4 or 5 and the monosaccharides was evidenced by several changes in the NMR spectra (for examples, see Table 2 and Figure 5a and Figure 5b). The addition of the monosaccharides 6a, 7a or 8a to a CDCl 3 solution of receptors 4 or 5 caused significant downfield shift of the amine NH A signal (for labeling, see Figure 2), downfield shift and strong broadening of the NH D signal as well as changes of the chemical shifts of the CH 3 F,G , CH 2 B,C,E , pyridine CH and imidazole or indole CH resonances of 4 or 5 (see Table 2). The signal due to the indole NH of 5 shifted downfield by 0.20-0.40 ppm. The complexation-induced chemical shifts of the NH A , indole-NH, CH 2 B , CH 3 F,G and the aromatic CH protons were monitored for the determination of the binding constants, which are summarized in Table 3. Binding studies with β-glucoside 6a and β-galactoside 8a showed the interactions of receptors 4 and 5 with these monosaccharides to be much more favorable than those with the α-anomers 7a and 9.
The curve fitting of the titration data for 4 and β-glucoside 6a suggested the existence of 1:1 and 2:1 receptor-sugar complexes in CDCl 3 solutions with a stronger association constant for 1:1 binding and a weaker association constant for the 2:1 receptor-sugar complex (this model was further supported by the mole ratio plots). The binding constants, however, were too large to be accurately determined by the NMR spectroscopic method (K 11 > 10 5 and K 21 ~ 10 4 M −1 ; see Table 3; for a review discussing the limitations of the NMR method, see ref. [55]). After the addition of 5% DMSO-d 6 the binding constants for 4•6a were determined to be 35000 (K 11 ) and 1000 M −1 (K 12 ). Thus, the affinity of 4 significantly decreases as solvent polarity increases (the addition of dimethyl sulfoxide also caused the change of the binding model; for a discussion on solvent effects in carbohydrate binding by synthetic receptors, see ref. [56]).
The interactions between the β-glucoside 6a and the indolebased receptor 5 in CDCl 3 were shown to be strong but less favorable than those with the receptor 4. The best fit of the titration data was obtained with the "mixed" 1:1 and 1:2 receptor-sugar binding model. The association constants for 5•6a were found to be 45900 (K 11 ) and 730 M −1 (K 12 ).
The interactions between β-glucopyranoside 6a and receptors 4 and 5 were also investigated on the basis of inverse titrations in which the concentration of sugar 6a was held constant and that of receptor 4 or 5 was varied. During the titration of 6a with 4 or 5 the signals due to the OH protons of 6a shifted downfield with strong broadening and became almost indistinguishable from the base line after the addition of only 0.1 equiv of the receptor, indicating important contribution of the OH groups of 6a to the complex formation. Furthermore, the addition of 4 or 5 to a CDCl 3 solution of β-glucoside 6a caused significant upfield shift of the CH signals of 6a, indicating the participation of the sugar CH units in the formation of the CH···π interactions with the aromatic rings of the receptor (for discussions on the importance of carbohydrate-aromatic interactions, see refs. [57][58][59][60][61][62][63]; for examples of CH-π interactions in the crystal structures of the complexes formed between artificial receptors and carbohydrates, see ref. [40]). Among the CH signals, the signal due to the 2-CH proton of 6a showed the largest shift (1.78 and 1.62 ppm for the titration with 4 and 5, respectively). In both cases, 6a•4 and 6a•5, the best fit of the titration data was obtained with the "mixed" 1:1 and 1:2 sugar-receptor binding model. Thus, the inverse titrations fully confirmed the binding model determined through the titrations of 4 or 5 with sugar 6a.
The association constants obtained on the basis of these titrations are identical within the limits of uncertainty to those determined from titrations where the role of receptor and substrate was reversed.
Similar to 4•6a, the best fit of the titration data for receptor 4 and β-galactoside 8a was obtained with the "mixed" 1:1 and 2:1 receptor-sugar binding model. However, the binding constants were again too large to be accurately determined by the NMR spectroscopic method (see Table 3). Studies performed in 5% DMSO-d 6 in CDCl 3 revealed that K 11 = 40700 M −1 and K 12 = 800 M −1 . The titration experiments with β-galactoside 8a clearly showed that receptor 5 is less effective towards this monosaccharide than the imidazole-based receptor 4 but much more effective than the previously described receptors 1-3. The motions of the signals of 5 were consistent with 1:1 and 1:2 receptor-sugar binding and could be analyzed to give association constants of 38000 (K 11 ) and 1100 M −1 (K 12 ). Compared to receptors 1-3 [31,41], receptors 4 and 5 showed a significant-ly higher binding affinity towards the β-galactoside 8a. The differences in the complexation abilities of receptors 1/3 and 4/ 5 towards β-galactoside 8a are clearly visible in the comparison of the chemical shifts of the signals of the four receptors after the addition of β-galactoside 8a (illustrated in parts a-d of Figure 5 for the pyridine CH 3 signals). Our previous studies showed compounds 1-3 to be highly effective receptors for β-maltoside 11 [28,31]. This disaccharide [64] is almost insoluble in CDCl 3 but could be solubilized in this solvent in the presence of the corresponding receptor. Similar solubility behavior of 11, indicating favorable interactions between the binding partners, could be observed in the presence of compounds 4 and 5. Thus, the receptor in CDCl 3 was titrated with a solution of maltoside dissolved in the same receptor solution. The complexation between 4 or 5 and the disaccharide 11 was evidenced by several changes in the NMR spectra (for example, see Table 2 and Figure 6). The saturation occurred after the addition of about 0.7 equiv of 11. Both the curve fitting of the titration data and and the mole ratio plots suggested the existence of 1:1 and 2:1 receptor-sugar complexes in the chloroform solution (with stronger association constant for 1:1 binding and a weaker association constant for 2:1 receptor-sugar complex). In both cases, 4•11 and 5•11, the binding constants in CDCl 3 were too large to be accurately determined by the NMR spectroscopic method (see Table 3). After the addition of DMSO-d 6 a substantial fall in the binding affinity was observed. Studies that were performed with 4 and 11 in 5% DMSO-d 6 in CDCl 3 revealed K 11 = 12000 M -1 and K 21 = 3000 M -1 , those performed with 5 and 11 indicated the formation of complexes with 1:1 receptor-sugar stoichiometry with K 11 = 42000 M -1 .

Molecular modeling
The formation of hydrogen bonds and CH···π interactions between the binding partners was also suggested by molecular modeling calculations. For example, molecular modeling suggested that all OH groups and the ring oxygen atom of the bound β-galactoside 8b in the complex 4•8b are involved in the formation of hydrogen bonds (see Table 4 and Figure 7a and Figure 8). In addition, interactions of sugar C-H units with the central phenyl ring of 4 (see Table 4) were shown to provide additional stabilization of the complex. Furthermore, the molecular modeling calculations indicated that within the 2:1 receptor-sugar complex the two receptor molecules almost completely enclose the sugar, leading to involvement of all sugar hydroxyl groups in interactions with the two receptor molecules (see Table 4 and Figure 7b). The OH groups are involved in the formation of cooperative hydrogen bonds which result from the simultaneous participation of a sugar OH as donor and acceptor of hydrogen bonds. The phenyl units of the both receptors stack on the sugar ring and both sides of the pyranose ring are involved in CH···π interactions (see Table 4 and Figure 7b).

Conclusion
The analysis of the binding motifs which are observed in the crystal structures of protein-carbohydrate complexes has influenced the design of receptors 4 and 5, including two 4(5)substituted imidazole or 3-substituted indole units as well as an   aminopyridine-based recognition group. The compounds 4 and 5 were established as highly effective receptors for neutral carbohydrates and were shown to display a significantly higher level of affinity towards β-galactoside than the previously described acyclic receptors. Both receptors were shown to display high β-vs. α-anomer binding preferences in the recognition of glycosides. The binding properties of 4 and 5 were studied on the base of 1 H NMR spectroscopic titrations in CDCl 3 and DMSO-d 6 /CDCl 3 mixtures as well as binding studies in two-phase systems, such as dissolution of solid carbohydrates in apolar media. The imidazole-based receptor 4 was found to be a more powerful monosaccharide receptor than the indole-based compound 5 and the previously described receptors 1-3. Compared to 1 and 2, incorporating only one imidazole or indole recognition unit, receptor 5 showed increased affinity to β-galactoside but decreased affinity to β-glucoside. The binding affinity of 1-5 towards β-galactoside 8a and β-glucoside 6a increases in the sequence 3 ~ 1 < 2 < 5 < 4 and 3 ~ 5 < 1 ~ 2 < 4, respectively. It is remarkable that the strong enhancement of the binding affinity of 4 and 5 towards β-galactoside was achieved through a relatively simple variation of the receptor structure. In contrast to 4 and 5, the previously described phenanthroline/aminopyridine-based receptors 22 and 23 were shown to display a high binding affinity towards α-galactoside as well as a strong α-vs. β-anomer binding preference. Thus, depending on the nature of the recognition units incorporated into the acyclic receptor structure, effective carbohydrate receptors with different binding preferences can be generated. However, the exact prediction of the binding preference still represents an unsolved problem and remains an important goal for future research.

Experimental section
Analytical TLC was carried out on silica gel 60 F 254 plates employing chloroform/methanol mixtures as the mobile phase.

General procedure for the synthesis of compounds 4 and 5:
To a solution of 4(5)-imidazole-carbaldehyde (18)  The solution was cooled to 0 °C and NaBH 4 (6.80 mmol) was added in portions. The reaction mixture was stirred for 1 h at 0°C and for additionally 6 h at room temperature. The solvent was removed and the residue was taken up in chloroform/water (100 mL, 1:1). The separated organic phase was further washed with water (3×30 mL), dried over MgSO 4 and the solvent was removed. The crude product was purified via column chromatography [CHCl 3 /CH 3 OH (incl. 1% 7 M NH 3 in CH 3 OH), 2:1 or 3:1 v/v].   Hz, 6H), 1.06 (t, J = 7.5 Hz, 3H) ppm; 13