Synthesis and glycosidase inhibitory activity of new hexa-substituted C8-glycomimetics

Introduction

There is a considerable interest in the design of molecules able to mimic carbohydrates which play critical roles in various biological events such as for example, cell-cell recognition and adhesion, cell growth and differentiation.[1-9] In this context, the goal is to obtain new compounds with improved efficacy, stability and specificity. Thus, a change from an aldopyranoside to a 1-deoxy-iminosugar (Figure 1) decreases the vulnerability of the resulting glycomimetic towards glycosidases, while the core structure and essential network of hydroxyl functionalities for enzyme recognition are retained. An important example is the 1-deoxynojirimycin (DNJ) family, for which DNJ itself is a competitive inhibitor of α-D-glucosidase (K_i = 8–25 μM),[10] while its derivatives miglustat (N-nBu DNJ, Zavesa^®) and miglitol (N-hydroxethyl DNJ, Glyset^® or Diastabol^®) have already found therapeutic applications in Gaucher's disease [11] and type 2 (non-insulino-dependant mellitus) diabetes, [12,13] respectively. In the past decade, works have been targeted to carbasugars originally consisting of six-membered cyclitols, related to valiolamine, [14] voglibose, [15] valienamine, [16] and acarbose. [17] The last two compounds, marketed as Basen^® and Glucor^® or Precose^®, respectively, are also actually used in the treatment of type 2 diabetes. All these compounds can have their amino moiety protonated, and the corresponding ammonium ions mimick the charge of the presumed transition states or intermediates of the enzymatic glycoside hydrolyses [18].

More recently, attention has been increasingly accorded to seven- and eight-membered ring systems [19-30] in order to study the effect of the enhanced flexibility and of the new spatial distribution displayed by these structures on their adaptability in the active site of the enzyme.

As part of a program directed to the synthesis of potential glycosidases inhibitors, [31,32] we focused on the access to new eight-membered carbasugars (Figure 2, A = OH) and related aminocyclitols (A = NHR) from C₂-symmetrical L-ido- or D-manno- cyclooctene, easily available by ring closing metathesis of 1,9-diene derived from L-ido- or D-manno-bis-epoxide [33]. Thus, synthetic potentialities of the newly created cyclic double bond were explored to reach hexa-substituted C8-glycomimetics.

Results and discussion

From the C₂-symmetrical L-ido- or D-manno- cyclooctene, 1 or 2, to obtain the C8 hexa-substitued carbasugars a straightfoward approach seemed to be a dihydroxylation, whereas to obtain the corresponding aminocyclitols it could be an epoxidation followed by the nucleophilic opening of the epoxide moiety by a primary amine or another nitrogen nucleophile. Accordingly (Scheme 1), treatment of the fully O-protected L-ido-cyclooctene 1 with a 5 mol% aqueous solution of osmium(IV) tetroxide [34] in acetone in the presence of N-methylmorpholine oxide and tert-butanol cleanly led to the expected cis-diol 3 in 97% yield. In analogous manner, the D-manno-cyclooctene 2 gave the corresponding cis-diol 4 in 97% yield. In each case, the cis-diol 3 or 4 has been isolated as a single stereoisomer because of the C₂-axis of symmetry displayed by the L-ido or D-manno-cyclooctenes 1 and 2. Then, simultaneous acidic hydrolysis of all the O-protective groups of 3 and 4 furnished the C8 hexa-substitued carbasugars 5 and 6 (80 to 97% overall yield for the two steps).

Now, to reach the corresponding aminocyclitols, we turned to the epoxidation [35] of the cyclooctenes 1 and 2 (Scheme 2). Thus, treatment of 1 and 2 with meta-chloroperbenzoic acid in the presence of sodium hydrogen carbonate afforded the epoxides 7 and 8 in 91–96% yield. As precedently, because of the C₂-axis of symmetry displayed by the L-ido or D-manno-cyclooctenes 1 and 2 the cis-epoxide 7 or 8 has been isolated as a single stereoisomer. However, all attempts to open the epoxide ring involving various nucleophiles, sodium azide, benzylamine, n-butylamine, or serinol in different experimental conditions, protic or aprotic solvent, presence or absence of a Lewis acid catalyst such as ytterbium triflate, revealed unsuccessful, only leading to recover the starting material.

To overcome this difficulty, we turned to a more electrophilic sulfate moiety [36] (Scheme 3). Thus, treatment of the cis-diols 3 and 4 with thionyl chloride in the presence of triethylamine followed by subsequent oxidation with sodium periodate in the presence of ruthenium trichloride gave the cyclic sulfates 9 and 10 in 80–100% yield. Nucleophilic opening of these sulfates by sodium azide in DMF at 80°C, [37] followed by acidic hydrolysis of the resulting acyclic sulfate ester cleanly afforded the corresponding azido-alcohols 11 and 12, isolated as single stereoisomers in excellent yield (95–98%). No other isomer of 11 or 12 was detected by NMR analysis, indicating that the ring-opening reaction is highly regioselective and results in the anti addition of the azido group on the opposite side of the bulky TBDMS group in β-position. However, it has to be pointed out that more hindered nucleophiles, such as primary amines, revealed unable to open the cyclic sulfate 9 or 10.

The absolute configurations of 11 and 12 were established by NMR studies. ¹H signals were assigned (Table 1) using 2D-COSY and 2D-TOCSY experiments starting from hydroxyl group at C8-position.

The determination of all the coupling constants and particularly of the ³J_1,8 was not possible by homodecoupling experiments. Thus, numerical simulation was used for an in depth study of the complex coupling patterns to set the parameters in complete analogy with regard to the experimental spectra. These ³J_1H,1H coupling constants, determined by simulation (Figure 3) of 1D spectra with NMR-SIM started from XWIN-NMR software (Bruker), are gathered in Table 1. These values allowed us to restrain the number of conformations for the eight-membered ring. For example, in compound 11, the large ³J_1H,1H coupling constants found between H1 and H8, H1 and H2 (proR), H8 and H7 (proR), H4 and H5, are in agreement with protons in a pseudo-axial position. On the other hand, the small ³J_1H,1H coupling constants found between H1 and H2 (proS), H8 and H7 (proS) indicate a pseudo-equatorial position of protons H2 (proS) and H7 (proS). NOE measurements and finally Molecular Dynamic calculations using Insight II software (Biosym Technologies, San Diego, CA) allowed to deduce the structure of 11 (Figure 4). Prochiral H2 (proS) and H7 (proS) protons displayed strong NOEs with the TBDMS groups, respectively in positions 3 and 6, indicating they are pointing away from the C8 ring in a pseudo-equatorial position. Hence, the protons H7 (proR), H6, H4 are close together and represent one face (upper) of the C8 ring, while H2 (proR), H5, H3 represent the other face. Then, in compound 11, the strong NOEs H1-H4 and H8-H5 indicate that H1 and H8 are in opposite side in an anti configuration with pseudo-axial positions. A similar approach was applied for the structural determination of 12.

The low reactivity of the epoxide function of 7 could be explained by the steric hindrance of the tert-butyldimethylsilyloxy groups in β-positions as confirmed by the X-ray crystallography (Figure 5). Furthermore, it seems that the C8-carbacycle of the tricyclic system [5-8-3] adopts a twist-boat-chair conformation. Whereas that of the sulfate 9, which crystallizes as a dimer, constituted by a tricyclic system [5-8-5] adopts a more flexible boat-chair conformation, thus allowing its opening by a linear nucleophile (azide anion), but not by a more hindered nucleophile (primary amine).

With the key enantiomerically pure azido-alcohol 11 and 12 in hands, we next turned to the obtention of C8-aminocyclitols. Thus, reduction of the azido group of 11 by dihydrogen in the presence of palladium black in ethyl acetate (Scheme 3) afforded the amino-alcohol 13 which could be submitted to acidic hydrolysis of the O-protective groups to give, after purification by ion-exchange chromatography, the targeted aminocyclitol 15 [20] (95% overall yield from 11). Alternatively, to obtain an analog of voglibose, the amine function of 13 could be alkylated via a reductive amination [39] with a dihydroxyacetone derivative. Thus, treatment of the amine 13 by the commercially available 2,2-dimethyl-1,3-dioxan-5-one in the presence of titanium(IV) tetra-isopropoxide followed by the cyanoborohydride reduction of the imine intermediate gave the expected N-alkylated aminocyclitol 17 (46% overall yield from 11). Then, simultaneous acidic hydrolysis of all protective groups led to the C8-voglibose mimetic 19 after purification by ion-exchange chromatography. The same sequence of reactions was uneventfully applied to the azido-alcohol 12 to afford the aminocyclitols 16 and 20.

The new C8-carbasugars 5 and 6 and C8-aminocyclitols 15, 16, 19 and 20 have been assayed for their inhibitory activity towards 24 commercially available glycosidases [40,41]. They did not inhibit the following enzymes at 1 mM concentration and optimal pH : α-D-glucosidases (maltase) from yeast and rice, β-D-glucosidase from caldocellum saccharolyticum, α-L-fucosidases from bovine epididymis and human placenta, α-D-galactosidases from coffee beans and Escherichia coli, β-D-galactosidases from Escherichia coli, bovine liver, Aspergillus niger and Aspergillus orizae, α-N-acetylgalactosaminidase from chicken liver, β-N-acetylglucosaminidases from Jack bean, bovine epididymis A and bovine epididymis B, α-D-mannosidase from almonds, β-D-mannosidase from Helix pomatia, and β-xylosidase from Aspergillus niger. For other enzymes: α-D-glucosidase from Bacillus stearothermophilus, amyloglucosidase from Aspergillus niger and Rhizopus mold, β-D-glucosidase from almonds, α-L-fucosidase from bovine kidney, and α-D-mannosidase from Jack beans the results are shown in Table 2. Each of these new compounds revealed weak inhibitor of the tested enzymes with a percentage of inhibition not over than 30%. These results show that the enhanced flexibility displayed by C8-glycomimetics does not seem to be correlated with an increase in observed activity. Thus, for example we had previously shown that the corresponding C7-voglibose mimic exhibited interesting activity towards amyloglucosidases from Aspergillus niger and Rhizopus mold (35 and 18 μM respectively, unpublished results). Furthermore, even if data concerning biological activity of C8-glycomimetics are seldom, the reported activities are often weak [24-30].

In summary, utilizing the readily available polyhydroxylated L-ido or D-manno-cyclooctenes, coming from ring closing metathesis of C2-symmetrical 1,9-dienes, we have accomplished the synthesis of a range of new hexa-substituted C8-glycomimetics in enantiopure form. Transformation of the cyclic double bond involved syn-dihydroxylation, then introduction of an azido group by opening of a cyclic sulfate followed by reduction and eventual alkylation of the resulting amine.

According to this strategy and to the nature of the ketones involved in the final reductive amination, various aminocyclitols could be synthesized. Thus, in this study two carbasugars and four aminocyclitols were obtained. Biological evaluation of these compounds towards 24 commercially available glycosidases have been carried out. For these C8-glycomimetics, weak activities were observed, which can probably be explained by a too high conformational flexibility of such structures.

Experimental

See Supporting Information File 1.