New α-galactosidase-inhibiting aminohydroxycyclopentanes

A set of cyclopentanoid α-galactosidase ligands was prepared from a partially protected ω-eno-aldose via a reliable (2 + 3)-cycloaddition protocol with slightly modified conditions. The obtained N-benzylisoxazolidine ring was selectively opened and the configuration of the hydroxymethylgroup was inverted. Consecutive deprotection provided an aminocyclopentane, which was N-alkylated to furnish a set of potential α-galactosidase inhibitors. Their glycosidase inhibitory activities were screened with a panel of standard glycosidases of biological significance.


Introduction
Glycoside hydrolases are a class of abundant and essential enzymes for carbohydrate catabolism, controlled trimming of oligosaccharide motifs in glycoproteins, as well as the release of stored polysaccharides in plants and animals. The Carbohydrate-Active enZYmes database 1,2 (CAZY) currently features well over one hundred and sixty sequence-associated families of glycosidases.
a-D-Galactosides are, in a quite diverse context, an important class of oligosaccharides and carbohydrate conjugates. Their corresponding hydrolases, a-D-galactosidases, are found in glycoside hydrolase (GH) families 4, 27 (containing the Fabry disease related human lysosomal enzyme), 36, 57, 97 and 110 as well as in GH109. GH 4 and GH109 enzymes follow a mechanism requiring NAD + and Mn 2+ that starts with the oxidation at C-3 followed by b-elimination of the aglycone, with subsequent water addition and reduction, 3 while the enzymes of the other ve families follow standard Koshland retaining mechanisms. 4 a-D-Galactosidases have been employed or modulated in several quite diverse applications. For example, a-galactosidases are frequently employed in combination with proteases to remove oligosaccharides from food to improve their nutritional availability and quality. 5,6 Improving the activity of lysosomal agalactosidase through chaperoning is a new approach to the management of Fabry's disease, a hereditary lysosomal disorder resulting from mutations in the GLA gene that result in catalytically compromised mutants of this essential enzyme. [7][8][9][10][11] In a completely different context, a-galactosidases that selectively and efficiently remove the immunodominant a-1,3-galactosyl epitope (Galili antigen) which hampers xenotransplantation from otherwise suitable mammalian sources, have been a focus of organ and tissue transplant medicine. [12][13][14] Similarly, for the purpose of generating red blood cells for readily available universal donor blood, enzymatic conversion of B type red blood cells into type O, by removal of a-1,3-bound galactosides from the cell surface, has been a eld of intense and sustained research. [15][16][17][18][19] Specic inhibitors of a-galactosidases have been developed as chaperones for treatment of the hereditary lysosomal disorder, Fabry's disease. A prime example is 1,5-dideoxy-1,5imino-D-galactitol (DGJ, 1), which enhances the residual activity of various a-galactosidase mutants, ameliorating Fabry related symptoms and providing improved quality of life for afflicted patients. a-Galactosidase inhibitors may also help the discovery, purication and characterization of enzymes for managed degradation of pathologically active a-galactosides.
We have recently been interested in hydroxymethylbranched di-and trihydroxycyclopentylamines such as compound 12, which may be regarded as resulting from ring contraction by formal homolytic extraction of the sugar ring oxygen and bond formation between C-1 and C-5. The new carbacyclic scaffold maintains the stereochemical information of the corresponding parent free sugar or glycosylamine albeit with the characteristic conformational features of vemembered rings. There are only four leading references in which such sugar analogues are shown as proven D- Fig. 1 Carbohydrate-related a-galactosidase inhibitors and inhibitory activities. galactosidase inhibitors. 29,30,34,35 As already outlined, there is only one N-alkylated hydroxymethyltrihydroxycyclopentylamine 12 reported in literature. Yet, this compounds shows sufficient inhibitory activity against agalactosidase (0.43 mM, gbc) but also inhibitory activity against b-galactosidase (1.5 mM, bovine liver) as well as b-glucosidase (0.51 mM, almonds). Aim of this work, is the synthesis of new aminocyclopentanes, as powerful and selective GH27 a-galactosidase inhibitors. These structures were prepared, relying on Vasella's (2 + 3)-cycloaddition approach. 36

Synthesis
Intermediate 21 was synthesised starting with reductive elimination of known iodosugar 13 (ref. 37) with zinc under slightly acidic conditions to give hemiacetal 14. Its treatment with Nbenzylhydroxylamine in CH 2 Cl 2 , provided a mixture of synproduct 15 and the desired known 38,39 anti-congured N-benzylisoxazolidine 16 (experimental details as well as NMR-data had not been reported for this compound) in a ratio of about 1 : 7 (15 : 16). The stereochemical outcome of this cyclisation is highly dependent on the solvent. Jäger and co-workers 39 pointed out that if the reaction is carried out in a polar solvent such as methanol, the main product is compound 15. In contrast, usage of an unpolar solvent such as CH 2 Cl 2 or CHCl 3 mainly provides compound 16. Separation of compounds 15 and 16 was easily achieved by silica gel chromatography. Protection of the free hydroxyl function in compound 16 with chloromethyl methyl ether (MOMCl) in the presence of Hünig's base (i-Pr 2 NEt) yielded fully protected compound 17. "One-pot" hydrogenolytic removal of the N-benzyl group over Pd(OH) 2 /C (20%) in the presence of Boc 2 O (di-tert-butyl-dicarbonate) and Na 2 CO 3 resulted in N-Boc-protected primary alcohol 18. Subsequent Swern or Dess-Martin oxidation furnished the corresponding aldehyde 19. Treatment of aldehyde 19 with pyridine or i-Pr 2 NEt in methanol allowed, via enolization, almost complete isomerisation to epimer 20. This reaction could easily be followed by thin-layer chromatography. The epimerization reaction was then quenched by addition of NaBH 4 at 0 C to obtain desired primary alcohol 21. Unexpectedly, the conditions of this attempted in situ reduction reaction favoured re-isomerization at C-5, now exclusivelyvia intermediate 19providing

Inhibitory activities
Amines 22-24 and 27 were screened for inhibitory activities with a panel of standard glycosidases ( Table 1). As can be seen the compounds act as inhibitors of both aand b-galactosidases, though overall they perform better on a-galactosidases with inhibition constants down to the high nanomolar/low micromolar range, as discussed in more detail below. While the parent molecule with a free amine proved to be a modest inhibitor with inhibition constants in the low to mid micromolar range, alkylation of this nitrogen improved inhibition in each case (with one exception in the case of GCase). Further extension of the alkyl chain generally improved inhibition yet more indeed for the b-glycosidases, all of which belong to the same, structurally related clan GHA, largely parallel increases in inhibition were seen with each addition of carbon atoms. This is consistent with common interactions with closely related structures. 42,43 For the a-galactosidases, the alkylation also generally improved inhibition, but not as predictably as with the b-galactosidases, despite the fact that both, Fabrazyme® and gcb, belong to the same sequence-related family, GH27 (Fig. 3).
Optimal affinity for the human enzyme, Fabrazyme®, was obtained with the n-nonyl substituted derivative 24, with a K i value comparable to that of galacto-isofagomine. This is the compound in current clinical use under the name Migalastat® or Galafold. The presence of the n-nonyl chain in 24 would likely improve its pharmacokinetic prole by keeping the compound in circulation longer, making this a candidate for testing as a chaperone. However, the fact that this compound is also a good inhibitor of other glycosidases tested here, means that its selectivity needs to be further improved.

Conclusions
Following a simple protocol previously outlined by Vasella's and Jäger's groups, 35,36 a series of novel derivatives of 12, bearing chain extensions at the nitrogen at position C-1 have been synthesized. Inhibition constants for these compounds with a range of Dgalactosidases, including human lysosomal a-galactosidase, were then measured. The results of these inhibitory studies revealed that free amine 22 is the worst inhibitor within this series and the activity increases with the introduction of an alkyl chain as seen with compounds 23, 24 as well as 27.
Furthermore, it is both unexpected and interesting that the a-congured compounds developed here are good inhibitors of the b-glycosidases in the rst place, since these have the wrong anomeric conguration for these enzymes. Further work will elucidate the reason for this unexpected nding.

General methods
Optical rotations were measured at 20 C on a Perkin Elmer 341 polarimeter at a wavelength of 589 nm and a path length of 10 cm.
[a] 20 D values are given in 10 À1 degree cm 2 per g. NMR spectra were recorded on a Bruker Ultrashield spectrometer at 300.36 ( 1 H) and 75.53 MHz ( 13 C). CDCl 3 was employed for protected compounds and CD 3 OD or D 2 O for unprotected inhibitors. Carbon and hydrogen numbering in NMR spectra was conducted in analogy to carbohydrate nomenclature and clockwise, starting with the amino bearing carbon as C-1 (Fig. 2). Chemical shis are listed in delta employing residual, nondeuterated solvent as the internal standard. Signals were assigned unambiguously by COSY, HSQC as well as APT analysis. Coupling constants 'J' for protecting groups, alkyl chains as well as the dansyl group (spacer arms) were found in the expected range and are not listed. For intermediate 21, the structure was conrmed by XRD structural analysis: suitable single crystals of compounds were immersed in silicone oil, mounted using a glass ber and frozen in the cold nitrogen stream (100 K). X-Ray diffraction data were collected at low temperature on a Bruker Kappa APEX II diffractometer using Mo K a radiation (l ¼ 0.71073 A) generated by an INCOATEC micro-focus source. The data reduction and absorption correction was performed with the Bruker SMART and Bruker SADABS, respectively. The structures were solved with SHELXT 44 by direct methods and rened with SHELXL 45 by least-square minimization against F 2 using rst isotropic and later anisotropic thermal parameters for all nonhydrogen atoms. Hydrogen atoms were added to the structure models on calculated positions using the riding model. The space group assignments and structural solutions were evaluated using PLATON. 46,47 MALDI-TOF Mass Spectrometry was performed on a Micromass TofSpec 2E Time-of-Flight Mass Spectrometer. Analytical thin-layer chromatography (TLC) was performed on precoated aluminium plates silica gel 60 F254 (E. Merck 5554) and detected with UV light (254 nm). For staining, a solution of vanillin (9 g) in a mixture of H 2 O (950 mL)/ethanol (750 mL)/H 2 SO 4 (120 mL) or ceric ammonium molybdate (100 g ammonium molybdate/ 8 g ceric sulfate in 10% H 2 SO 4 (1 L)) were employed followed by heating on a hotplate. For column chromatography, silica gel 60 (230-400 mesh, E. Merck 9385) or silica gel 60 (Acros Organics, AC 24036) were used. Reaction monitoring was done by TLC.

Kinetic studies
Kinetic studies were performed at room temperature in an appropriate buffer for each enzyme (specic conditions can be found below). All reactions were performed in half-area 96-well- plates (Corning) and monitored with a Synergy H1 plate reader (BioTek). In each experiment, the appropriate concentration of the enzyme was incubated with different concentrations of the inhibitors for 2-5 minutes before initiating the reaction by the addition of the glycoside substrate. The initial rate was then measured by monitoring the increase in, either absorbance at 405 nm (where the aglycone is either p-nitrophenol (pNP) or 2,4dinitrophenol (dNP)) or uorescence at 365/455 nm (where the aglycone is 4-methylumbelliferone (MU)) for up to ve minutes. K i determinations were performed using two different substrate concentrations. For each substrate concentration, a range of three to six inhibitor concentrations was used. Dixon plots (1/v vs. [I]) were constructed to validate the use of competitive inhibition model and to assess the t of the data. The data were then t to a competitive inhibition model using non-linear regression analysis with Grat 7.0.0. (Erithacus Soware).