Synthesis and NMR-characterization of three quinamide-based disaccharide mimetics with unusual cyclohexane t wist - conformation

The synthesis of amide-linked disaccharide mimetics has been explored starting with carbohydrate-based amines and a protected quinic acid lactone. Benzyl-2-amino-4,6-O - benzylidene-2-deoxy-α / β -D-glucopyranose ( 12 ) and D-glucamine ( 14 ) were successfully coupled to give the corresponding quinamides ( 13 and 15 ), while the quinoylation of O - acetylated L-fucopyranosyl methylamine ( 7 ) failed. The latter was prepared from per-O -acetyl- L-fucopyranose via the improved multigram scale synthesis of the corresponding per-O -acetyl- L-fucopyranosyl cyanide ( 3 ). Compound 3 was subsequently hydrogenated to yield a mixture of compound 7 and the per-O -acetylated bis-(fucopyranosylmethyl) amine ( 5 ). The vicinal coupling constants in the NMR spectra of all quinamide products revealed considerable flexibility of the cyclohexane ring in solution and substantial contributions by twist -chair conformations.


Introduction
Recently, oligosaccharides were recognized to have functions influencing the entire spectrum of cell activities [1][2][3][4] For example, carbohydrates have been recognized as a completely new cancer prevention target [5][6][7][8] and as possible vaccines 4,[9][10][11][12][13][14][15] Explanations for the striking complexity of naturally occurring carbohydrates have been proposed. 16vailability of metabolically resistant carbohydrate analogs will aid in the study of molecular interactions between carbohydrates and the cell machinery.The synthesis of metabolically stable mimetics of natural carbohydrates as well as oligosaccharides and polyhydroxyl clusters with linkages other than the natural O-glycosidic linkage has attracted considerable interest. 17,18In this paper, we have investigated the possibility for the synthesis of amide-linked Small Cluster Oligosaccharide Mimetics (SCOM) [19][20][21] through nucleophilic opening of a suitably protected lactone from quinic acid by amines to yield disaccharide-like structures.

Results and Discussion
Per-O-acetylation of L-fucose on a multigram-scale has not been reported to the best of our knowledge and was carried out by us to yield an anomeric mixture of 1,2,3,4-tetra-O-acetyl-Lfucopyranose (2a,b, 98 % overall, 15 g starting material) with an α-/β-ratio of 15:1 as estimated by 1 H-NMR (Scheme 1).The observed yield indicated no significant side reactions upon scale up and was consistent with previous reports on small scale. 28,29Lewis acid-promoted direct displacement of the anomeric acetyl group 19,30,31 in 2a,b gave the fucopyranosyl cyanides 3a,b (α/β-ratio <1:20 as estimated by NMR).Scheme 1. Conversion of 2a/b into the fucopyranosyl cyanides 3a/b followed by reduction in the presence of Boc-anhydride. 19fter hydrogenation of compound 3b, 19 the mixture of N-Boc-protected aminomethyl-Cmonosaccharide 4 and N-Boc-protected aminomethyl-C-disaccharide 5 was separated by flash column chromatography on silica gel.The combined fractions of 4 and 5 gave an overall yield of 97 %.The 1 H-NMR spectra of 4 and 5 are shown in Figure 1.The proton assignments and coupling constants were consistent with a 1 C 4 solution-conformation [32][33][34][35] in CDCl 3 for both 4 and 5.A 13 C-DEPT experiment (data not shown) confirmed the presence of one CH 2 -group in 4 and two CH 2 -groups in 5, which supported the monosaccharide structure for the former and the disaccharide structure for the latter.(a) 1 H-NMR spectrum of compound 4 with a 1:3 integration ratio between the fucose-6-CH 3 group and the t-butyl group (solvent CDCl 3 ); b) 1 H-NMR spectrum of compound 5 with a 3:2 integration ratio between the fucose-6-CH 3 group and the t-butyl group (solvent CDCl 3 ).
It has been shown in the literature that methoxide-catalyzed global deprotection of structures similar to 4 suffer from unintended migration of the C-2-acetyl group to the nitrogen atom of the aminomethyl group. 19However, here we were able to achieve selective de-O-acetylation of 4 with a combination of Et 3 N in aq.MeOH at rt., which did not affect the Boc group (NMR data not shown) and prevented O-/N-acetyl migration.Compound 6 was obtained as a syrup and was immediately subjected to sulfuric acid-catalyzed removal of the Boc group.The subsequent neutralization of sulfuric acid was conveniently achieved by BaCO 3 , and the resulting precipitate of BaSO 4 was filtered off.Compound 7 was found to decompose over time.Therefore, the crude product was coupled immediately with lactone 9 whose synthesis from D-(-)-quinic acid 8 is shown in Scheme 2. 36 Compound 9 (91 %) was obtained with identical physical properties to the one described in the literature and with consistent 1 H-NMR data (Table 1).Unfortunately, the coupling reaction between compound 7 and 9 did not yield the intended carboxamide product under a variety of reaction conditions.The quinic acid lactone was partly recovered, whereas the remainder was a complex mixture of at least five unidentified compounds.At the present time, we assume that chemical instability of compound 7 prevented the successful synthesis of the amide.Therefore, we investigated compound 9 as a coupling partner for amines in two model reactions with aniline and benzylamine.Whereas the reaction with benzylamine 10 yielded the reported 21 coupling product 11 in very good overall yield, the coupling between 9 and aniline did not succeed.Resonance-delocalization of the nitrogen lone pair in aniline and concurrent reduced nucleophilicity may be a possible explanation for this result.The benzyl carboxamide 11 was analyzed by NMR spectroscopy (Figure 2).
The solution data for 11 in CDCl 3 indicated a significant twist-chair conformation.In the ring system of 11, H-3 was significantly shielded compared to 9 (Table 1, vide infra) as a result of lactone opening and small vicinal coupling constants between H-3 and both H-2a and H-2b, 5.1 Hz and 3.0 Hz, respectively.This was consistent with the CH-3 bond dissecting the H-2a-C-H-2b bond angle.Proton 4 was assigned to the signal at 4.17 ppm.Besides the two small coupling constants towards H-3 and H-5, the presence of a 4-bond W-coupling of 1.5 Hz towards H-2b was observed.This W-coupling was also present in the signal for H-2b and was confirmed by a weak off-diagonal cross peak in the 1 H-1 H-COSY spectrum (Figure 2b).The signal for H-5 at 4.58 ppm was split by interaction with H-4 ( 3 J 3,4 6.9 Hz) and two small gauche couplings of 2.7 Hz towards H-6a and H-6b.It was concluded, in analogy to H-3, that the CH-5 bond dissected the H-6a-C-H-6b bond angle.Both protons H-3 and H-5 showed NOE cross-peaks to H-2a/H-2b and H-6a/H-6b, respectively (Figure 2b), which cannot be explained by a chair-form and supports a significant contribution by the twist-conformation in CDCl 3 .This conformation might be favorably supported by a hydrogen bond between the OH group at C-3 and the carboxamide (Figure 2a).We also detected hydration between the two OH-groups and residual water in the solvent.This resulted in a significant EXSY 37,38 cross-peak between OH-1 and OH-3 despite the great distance (data not shown).The lactone 9 was also used in the synthesis of three Small Cluster Oligosaccharide Mimetics (SCOM) with sterically more demanding amines.Benzyl-2-amino-4,6-O-benzylidene-2-deoxyα-and β-D-glucopyranoside 12a and 12b 39,40 and D-glucamine 14 were used for coupling with 9.All three amines were successfully coupled, and carboxamides 13a, 13b, and 15, respectively, were obtained in moderate to excellent yields.NMR data for the starting materials (9, 12a and 12b) and for the products (13a and 13b) are shown in Table 1 for comparison.
The 1 H-NMR spectrum of 13a recorded in CDCl 3 is shown in Figure 3a.A change of solvent to DMSO-d 6 resulted in less useful spectra where heavily overlapped signals of diagnostic protons complicated the overall assignment (data not shown).In CDCl 3 , the anomeric configurations for the hexapyranosyl rings A in 13a and 13b were assigned based on the vicinal coupling constants between H1 A and H2 A ( 3 J 1,2 = 3.3 Hz, gauche, α-configuration, and 3 J 1,2 = 8.1 Hz, trans-diaxial, β-configuration, respectively) and did not differ from the starting materials 12a and 12 b (Table 1).In CDCl 3 both OH-groups in ring B were successfully identified based on long-range TOCSY correlations with H-6a/b and H-2a/b in ring B (Figure 3b, left).The signal of OH-3 B at 4.39 ppm was split into a doublet by H-3 B , and the remaining OH-3 in ring A at 3.11 ppm displayed a weak COSY cross-peak to H3 A (data not shown).The set of coupling constants observed in ring B for 13a was found to be similar to that in compound 11.Both signals for H-3 B and H-5 B displayed small vicinal coupling constants towards H-2a/b B and H-6a/b B , respectively.In the molecular model 41 of the twist-conformation, the estimated dihedral angles were [H5-C-C-H6a] +59.9º, [H5-C-C-H6b] -56.6º, [H3-C-C-H2a] +52.9º, and [H3-C-C-H2b] -61.6º.This gauche relationship was in good agreement with the determined small vicinal coupling constants (Table 1).We were able to confirm this assignment by NOESY cross-peaks between H-5 B /H-6a/b B (molecular model: d 5,6a 2.47 Å and d 5,6b 2.46 Å, respectively) and H-3 B /H-2a/b B (molecular model: d 3,2a 2.41 Å and d 3,2b 2.46 Å, respectively) in Figure 3b (center).A weak enhancement was also observed between the OH-group on C-3 and H6a B .The conformation of ring A is defined by the trans-fused 4,6-O-benzylidene protection group, and all observed coupling constants and NOE-enhancements were fully consistent with the 4 C 1 -conformation (Table 1).Inter-ring cross-peaks were observed only among the three OH-groups (Figure 3b, right).We concluded that residual water in CDCl 3 connected the OH-groups in both rings through hydrogen-bonded H 2 O-molecules that can give rise to EXSY cross-peaks through chemical exchange. 37,38The exact geometry of this hydration feature was not further investigated by us and will be subject of future studies.NOESY cross-peaks from the amide proton with H-2 A (d NH,2 2.94 Å) and H-3 A (d NH,3 2.44 Å) were also observed.

ISSN 1551-7004
Page 117 © ARKAT USA, Inc   The 1 H-NMR spectrum for compound 13b is shown in Figure 4a.We observed only a small difference in chemical shift for the anomeric proton (4.71 ppm, 3 J 1,2 8.1 Hz) compared to compound 13a with α-configuration in the sugar moiety (4.85 ppm, 3 J 1,2 3.0 Hz).However, the vicinal coupling constants 3 J 1,2 was indicative of the anomeric configuration.The NMR data obtained for compound 13b suggested similar overall ring conformations in A and B compared to 13a (Figure 4a, Table 1).Interestingly, all three OH-signals were sharp singlets at 2.07, 2.93, and 3.00 ppm with only minor line broadening.Neither 1 H-1 H-COSY nor long-range TOCSY data allowed us to assign the signals unequivocally except for OH-3 A (3.00 ppm) by an NOE cross-peak towards H-2 A (Figure 4b).However, the NOESY spectrum did not show any EXSY cross peaks among the OH-groups despite the presence of H 2 O in CDCl 3 .This implied that 13b was less hydrated and that the OH-groups were possibly stabilized more effectively by intra-molecular hydrogen bonds than 13a.This was supported by observed NOE enhancements between the amide proton and H-1 A and H-2 A and a weak NOE signal towards H-3 A (Figure 4b).This geometry would allow not only the formation of a hydrogen bond between the amide carbonyl oxygen and OH-3 A as well as OH-1 B (1.69 Å and 2.09 Å estimated, respectively) but also between the amide proton and the oxygen of OH-3 B (1.71 Å estimated).
The coupling reaction between 9 and recrystallized D-glucamine yielded the expected carboxamide 15 in low yield.Upon recrystalliztion and charcoal treatment of the crude Dglucamine, the yield of the coupling reaction improved to 62 %.The isolated product had gelatinous consistency and dried into an amorphous powder with uncharacteristic melting point.The NMR spectrum in DMSO-d 6 was found to be consistent with the proposed structure (data not shown).However, extensive overlap especially of signals from the glucamine chain "A" made the full assignment of the structure difficult.In ring B, the proton H-3 B was identified at 3.65 ppm and H-4 B , H-5 B , and H-6b B at 3.98 ppm, 4.51 ppm, and 2.28 ppm, respectively.The three latter showed coupling constants that were similar to the values observed for 13a, and 13b in DMSO-d 6 (data not shown).
Temperature-annealed molecular dynamics at 400 K in vacuum without distance restraints yielded chair-conformations for ring B in compound 11 and in both stereoisomers 13a and 13b.The energy difference between the two possible lowest-energy chair conformations for ring B was small for all compounds (11: 0.4 kcal/mol, 13a and 13b: 0.6 kcal/mol and 1.5 kcal/mol, respectively).In all cases, the amide group and the OH-3 group assumed the axial orientation for favorable hydrogen-bonding in the lowest-energy structure.The observation of a hydrogen bond in the models is a reasonable argument in non-polar solvents such as CDCl 3 .In this particular conformation, H-5 B has a gauche relationship towards H-6b B and a trans-diaxial position towards H-6a B .Proton H-3 B would be gauche to both H-2a B and H-2b B with expected coupling constants between 3-5 Hz.However, the experimental coupling constants for 11, 13a, 13b (Table 1), and 15 (acetone-d 6 , entry not shown: 3 J 5,6a = 3.9 Hz, 3 J 5,6b = 3.9 Hz) indicated that H-5 B is gauche to both protons in position 6 and that the dihedral angles between H-3 B and H-2a,b B are rather ~80º and ~40º, respectively (Table 1, 15 entry not shown: signals submerged).Compound 15 could not be analyzed in CDCl 3 because of low solubility and, unlike 11, 13a, and 13b, the exact conformation of 15 could not be established unequivocally.When we switched to DMSOd 6 , we observed a moderate change in coupling constants for H5 B and H6a,b B in 13a and 13b.Even though the vicinal coupling constants between H-5 B and H-6a B /H-6b B increased from 2.7-3.0Hz in CDCl 3 to 5.1-5.4Hz in DMSO, they remained identical in value consistent with a bisecting geometry between C-H5 B and the H6a B -C-H6b B angle as described above.
Unfortunately, the signals of H-2a B /H-2b B (AB-mixing), H-3 B /H-4 B (submerged) could not be used to establish the overall conformation of ring B in DMSO reliably.Many of the experimental 3 J HH values did not match the calculated values 42,43 (data not shown) neither for any individual conformation nor for an equilibrium of two chair conformers.This also suggests that several solution conformers, including twist-forms, are in rapid equilibrium, and that vicinal coupling We modeled compounds 11, 13a, and 13b in their twist-conformations as shown in Figure 5 to approximate the observed solution data and to match the structure with the observed NOE signals.The lowest-energy twist-chair conformation of 11 was less stable than the two chair conformations by 0.8 kcal/mol and 0.4 kcal/mol, respectively.In the molecular model of 11, the geometry of the amide linkage was defined by two hydrogen bonds, one between the amide carbonyl oxygen and the OH-group at C-1; the other one between the amide NH-group and the OH-group at C-3.The latter can be assumed to possibly contribute to the stabilization of the twist-chair observed by NMR.Similar conformational features were observed in the molecular model of 13a and 13b.In both structures, the amide linkage displayed three stabilizing hydrogen bonds.One between C=O/OH-1 B and one between NH/OH-3 B (see molecular model 11).An additional hydrogen bond resulted from interaction of the amide carbonyl oxygen and the OHgroup at C-3 A in the glucopyranosyl ring.The location of the amide NH-group between H-1 A and H-2 A was supported by an observed Nuclear Overhauser enhancement in compound 13b (Figure 4), whereas the same NOE-signal was absent in 13a.The lowest-energy twist-chair conformation of 13a was less stable by 3. 2.4 kcal/mol and 0.9 kcal/mol, respectively, were calculated (Figure 5).In view of the significant change in free energy for the twist-conformation in 13a, we assume that the observed hydration feature in 13a is crucial in providing conformational stabilization to account for the observed NMR data.
A full conformational analysis of this system is underway to determine the contribution of all solution conformations to the experimentally observed NMR coupling constants.At the present time we assume that the fusion with the isopropylidene ring distorts the neighboring cyclohexane ring, which possibly leads to several rapidly equilibrating twist-conformations.To the best of our knowledge, twist-conformations in the solid state for cyclohexan derivatives with one fused 1,3dioxolane unit have been reported twice 44,45 in the Cambridge Crystallographic Data Bank for small molecules.Cyclohexane twist-conformations in the crystalline state are more frequently found when two 1,3-dioxolane moieties are present or if the cyclohexane ring is part of a tricyclic structure.A twist-conformation in the solid state was reported for the cyanoethylidene derivative of a six-membered carbohydrate tetrahydropyran ring. 20Increased conformational flexibility has been reported for the cyclohexane ring in hydrindanes. 46,47

Experimental Section
General Procedures.All reagents were used without purification unless specified.1-Amino-1deoxy-D-glucitol (D-glucamine) was obtained from Huels America (New Jersey) and was recrystallized from water.Solvents were of HPLC grade.Dichloromethane, isopropyl ether, ethylacetate, and dioxane were purified by simple distillation prior to use.Room temperature NMR spectra were recorded in CDCl 3 unless stated differently on a Varian MERCURY 300 MHz spectrometer with spinning and were referenced to TMS as internal standard.Total Correlation Spectroscopy (TOCSY, Homonuclear Hartmann-Hahn) experiments were carried out non-spinning with a mixing time (duration of the MLEV-16 spin lock) of 50 ms.Data were collected in the 2D-hypercomplex mode (States-Haberkorn method).All ESI mass spectra were recorded on a VARIAN 1200 LC triple-quad mass spectrometer in positive mode.Solutions of c ~ 10 -5 M were used in MeOH/H 2 O (1:1) at 400 µL total volume.The analyte solution was sprayed by continuous infusion from the tip of a capillary with pneumatic assist (N 2 sheath gas) at a flow rate of 10 µL/min.Desolvation of the spray was accomplished at elevated temperature ‫ٯ(‬ API chamber = 50 ºC, ‫ٯ‬ capillary = 120-150 ºC).The instrument was operated at ~ 5*10 -6 torr with a mass window of m/z 0-1500.The detector was set to 1.2-1.5 kV.High-resolution ESI mass spectra were recorded on a double focusing sector instrument (LCmate, JEOL) with polyethylene glycol as internal standard.
Molecular modeling was done on a Power Mac G5 equipped with the Spartan software package (Spartan '02 for Windows, Wavefunction, Inc., Irvine, CA).Initial structures were minimized in the MMFF94 force field (Merck Pharmaceuticals) and were subsequently subjected to searches over the conformational space with collection of the lowest structures within 20 kcal/mol of the global minimum.The lowest energy chair structures and the structures most consistent with NMR data were further optimized by the semi-empirical AM1 method.

Figure 5 .
Figure 5. Molecular models 41 of compounds 11, 13a, and 13b in vacuum.The twistconformations were minimized on the semi-empirical level (AM1-method).H-bonds are shown as dashed lines.Selected experimental NOEs are indicated by arrows.
7 and 4.3 kcal/mol, respectively, and for 13b values of

Table 1 .
1 H-NMR shift values for 9, 11, 12a/b, and 13a/b.J-values in Hz Nitromethane was stirred with Drierite for 24h.Solvents were removed on a Rotovap apparatus (Buechi, Column chromatography was performed with silica gel (J.T. Baker, 40-60 mesh) on a glass column with permanent reservoir (Kimble&Kontes, 50 x 700 and 38 x 500 mm).Thin layer chromatography (TLC, 250 µm SiO 2 , Analtech) was run with indicated solvent systems.The developed plates were sprayed with 10 % aq.H 2 SO 4 , and were charred on a hot plate for visualization.