4-C-Me-DAB and 4-C-Me-LAB – enantiomeric alkyl-branched pyrrolidine iminosugars – are specific and potent α-glucosidase inhibitors; acetone as the sole protecting group

Abstract

The syntheses of 4-C-Me-DAB [1,4-dideoxy-1,4-imino-4-C-methyl-d-arabinitol] from l-erythronolactone and of 4-C-Me-LAB [from d-erythronolactone] require only a single acetonide protecting group. The effect of pH on the NMR spectra of 4-C-Me-DAB [pK_a of the salt around 8.4] is discussed and illustrates the need for care in analysis of both coupling constants and chemical shift. 4-C-Me-DAB (for rat intestinal sucrase K_i 0.89 μM, IC₅₀ 0.41 μM) is a competitive - whereas 4-C-Me-LAB (for rat intestinal sucrase K_i 0.95 μM, IC₅₀ 0.66 μM) is a non-competitive - specific and potent α-glucosidase inhibitor. A rationale for the α-glucosidase inhibition by DAB, LAB, 4-C-Me-DAB, 4-C-Me-LAB, and isoDAB – but not isoLAB – is provided. Both are inhibitors of endoplasmic reticulum (ER) resident α-glucosidase I and II.

This paper describes the synthesis of the enantiomers 4-C-Me-DAB 1D [1,4-dideoxy-1,4-imino-4-C-methyl-d-arabinitol] and 4-C-Me-LAB 1L with only a single acetonide needed as a protecting group, both of which are micromolar inhibitors of some α-glucosidases; in accord with Asano's hypothesis,ⁱ the d-iminosugar 1D is a competitive inhibitor, whereas the enantiomer 1L is a non-competitive inhibitor. Synthetic enantiomers of natural iminosugars are frequently powerful glycosidase inhibitors.ⁱⁱ Natural and synthetic iminosugars comprise a major family of glycosidase inhibitors.ⁱⁱⁱ The introduction of an alkyl substituent into a sugar mimic usually removes any significant glycosidase inhibition,^iv however, introduction of a C6 methyl branch into the piperidine ring of l-swainsonine increases the inhibition of naringinase by an order of magnitude^v in comparison to the parent indolizidine, l-swainsonine.^vi The natural product DAB 3D is a good – but the enantiomer LAB 3L is a more potent and more specific – inhibitor of α-glucosidases.^vii The isomer isoDAB 2D is also a very good inhibitor of α-glucosidases but the enantiomer 2L shows no inhibition of any glycosidase. This paper provides a rationale for isoLAB 2L being the only one of the six simple pyrrolidines 1, 2, and 3 which does not inhibit α-glucosidases; isoLAB 2L is the only one of the sugar mimics which partially rescues the defective F508del-CFTR function in cystic fibrosis.^viii N-Alkylation of monocyclic iminosugars can enhance glycosidase inhibition by several orders of magnitude;^ix such modification of the alkyl-branched parent structures may access a series of new bioactive compounds.

The enantiomeric azidolactones 4L and 4D are key intermediates in the synthesis of 4-C-methyl pyrrolidines 1L and 1D [Scheme 1]. The protected 2-C-methyl arabinono-lactone 5D can be prepared by a Kiliani reaction^x on a protected 1-deoxyribulose^xi derived from d-erythronolactone 6D; 5D is a useful chiron for the preparation of 2′-C-methyl nucleosides^xii and carbon-branched ketoses.^xiii Reaction of 5D with triflic anhydride, followed by treatment with sodium azide, results in nucleophilic displacement at a tertiary center with inversion to give the azidolactone 4D.^xiv The enantiomeric azido-2-C-methyl-l-arabinonolactone 4L was prepared from l-erythronolactone 6L by similar procedures.

Scheme 1. Structural relationship and synthesis of 4-C-methyl branched pyrrolidines.

For the synthesis of 4-C-Me-DAB 1D, it was necessary to invert the configuration at C4 of the azidolactone 4L [Scheme 2]. Hydrolysis of the isopropylidene protecting group in the 1,5-lactone 4L by trifluoroacetic acid in aqueous 1,4-dioxane gave the 1,4-lactone 7L, {m.p. 76-78 °C; [α]_D²⁵ -132.9 (c, 1.06 in CH₃OH)} in 87% yield. Selective esterification of the primary hydroxy group in 7L with tosyl chloride in pyridine afforded the tosylate 8L {m.p. 110-112 °C; [α]_D²¹ -107.7 (c, 1.07 in CH₃CN)} in 57% yield. Reaction of l-ribonolactone 8L with potassium hydroxide in aqueous 1,4-dioxane, followed by treatment with an acidic ion exchange resin, resulted in inversion of configuration at C4 to form the d-lyxonolactone 9D {oil; [α]_D²¹ +48.4 (c, 0.74 in MeOH)} in 87% yield. Esterification of 9D with tosyl chloride in pyridine yielded the primary tosylate 10D {oil; [α]_D²⁰ +13.2 (c, 0.85 in CH₃CN)} (66% yield) which on reduction with sodium borohydride in methanol formed the azidotriol 11D, (oil, 66% yield); it was thus possible to reduce the hindered lactone 10D to the triol 11D without any competing formation of epoxide. Hydrogenation of the unprotected azide 11D in the presence of palladium on carbon in 1,4-dioxane gave the corresponding amine which spontaneously cyclized to form 4-C-Me-DAB 1D, in quantitative yield. The tosylate salt of 1D could be converted into the free base 1D by ion exchange chromatography with Dowex 1-X2 (OH^- form) with water as eluent {free base: oil, [α]_D²⁵ -20.1 (c, 0.57 in H₂O); hydrochloride salt: [α]_D²⁵ -5.22 (c, 1.07 in H₂O)}. The enantiomer 4-C-Me-LAB 1L was prepared by an identical route from d-erythronolactone {free base: oil, [α]_D²⁵ +21.3 (c, 0.71 in H₂O); hydrochloride salt: [α]_D²⁵ +4.10 (c, 1.0 in H₂O)}.^xv It is noteworthy that only one isopropylidene group is used throughout this sequence.

Scheme 2. (i) CF₃COOH:H₂O:1,4-dioxane, 4:1:1, 87% (ii) TsCl, pyridine, 57% (iii) KOH, 1,4-dioxane; then Dowex, 87% (iv) TsCl, pyridine, 66% (v) NaBH₄, MeOH, 66% (vi) H₂, Pd/C, 1,4-dioxane: H₂O (2:1), 100%.

Since the pK_a of the salts of imino sugars is around 7-8, the ¹H NMR spectrum is strongly dependent on the pH. This is illustrated for 4-C-Me-DAB 1D at six different pH values with CH₃CN as the internal standard [Figure 1]. The pK_a of the hydrochloride salt of 1D was determined as 8.38 from the change in the chemical shift of the C-methyl group with pH [Figure 2]. The full resonance assignments of the salt at pH 2.4 and of the free base at pH 11.1 of 1D are given in Tables 1 and 2, respectively.

Figure 1.

¹H NMR (500 MHz) of 4-C-Me-DAB 1D in D₂O at different pHs.

Figure 2.

Chemical shift of the methyl peak in 1D versus pH. The points are the experimental values; the line is the fitted curve, pK_a of 8.38.

Table 1.

NMR [¹H (500 MHz) ¹³C (125 MHz) D₂O] assignments of the HCl salt of 1D, pH 2.4.

Label	1H			13C
pH 2.24	δ (ppm)	mult	3JHH (Hz)	δ (ppm)
C1	3.664 3.202	dd dd	12.7 / 6.8 12.7 / 5.0	47.83
C2	4.401	ddd	6.8 / 5.0 / 5.0	74.26
C3	4.014	d	5.0	77.36
C4	--	--	--	69.48
C5	3.797 3.688	d d	12.3 12.3	63.75
CH3	1.332	s	na	15.12

Table 2.

NMR [¹H (500 MHz) ¹³C (125 MHz) D₂O] assignments of the free base of 1D, pH 11.1.

Label	1H			13C
pH 11.1	δ (ppm)	mult	3JHH (Hz)	δ (ppm)
C1	3.170 2.648	dd dd	12.1 / 7.3 12.1 / 6.2	48.39
C2	4.181	ddd	7.3 / 6.2 / 6.0	77.61
C3	3.801	D	6.0	80.67
C4	--	--	--	63.41
C5	3.485 3.443	D d	11.6 11.6	66.98
CH3	1.010	S	na	17.60

LAB 3L is a more potent and specific inhibitor of α-glucosidases than its naturally occurring enantiomer DAB 3D [Table 3],^xvi whilst isoDAB 2D is a more potent inhibitor of α-glucosidases than DAB 3D and is completely specific; in contrast, isoLAB 2L does not show any inhibition of these enzymes.^viii 4-C-Me-DAB 1D and 4-C-Me-LAB 1L were both found to be specific and potent (μM) inhibitors of α-glucosidases. In particular, these compounds showed strong inhibition of rat intestinal sucrase with IC₅₀ values of 0.41 and 0.66 μM, respectively. The kinetic analysis showed that 4-C-Me-DAB 1D is a competitive inhibitor of rat intestinal sucrase with a K_i value of 0.89 μM, whereas the enantiomer 4-C-Me-LAB 1L is a non-competitive inhibitor (K_i value of 0.95 μM) of the enzyme. Although DAB 3D has been reported to be a good inhibitor of glycogen phosphorylase and has been investigated as a potential therapeutic agent for the treatment of diabetes,^xvii LAB 3L, isoDAB 2D, isoLAB 2L, 4-C-Me-DAB 1D, and 4-C-Me-LAB 1L showed no significant inhibition of this enzyme.

Table 3.

Concentration of iminosugars giving 50% inhibition of various glycosidases and glycogen phosphorylase.

	IC50 (μM)
Enzyme	DAB 3D	isoDAB 2D	4-C-Me-DAB 1D	LAB 3L	isoLAB 2L	4-C-Me-LAB 1L
α-Glucosidase
Rice	250	41	7.1	3.2	NI	5.8
Yeast	0.15	NI	1.9	70	NI	NI
Rat intestinal maltase	55	24	0.74	0.93	NI	2.4
Rat intestinal isomaltase	5.8	20	3.4	0.36	NI	5.1
Rat intestinal sucrase	16	15	0.41	1.0	NI	0.66
β-Glucosidase
Almond	250	NI	NI	NI	NI	NI
Bovine liver	638	NI	NI	NI	NI	NI
Rat intestinal cellobiase	756	NI	NI	NI	NI	NI
α-Galactosidase
Coffee beans	NIa	NI	NI	NI	NI	NI
Human lysosome	NI	NI	NI	NI	NI	NI
β-Galactosidase
Bovine liver	NI	NI	NI	NI	NI	NI
Rat intestinal lactase	323	NI	NI	415	NI	NI
α-Mannosidase
Jack beans	320	NI	NI	NI	NI	NI
β-Mannosidase
Snail	NI	NI	NI	NI	NI	NI
α-l-Rhamnosidase
P. decumbens	NI	NI	NI	NI	NI	NI
α-l-Fucosidase
Bovine epididymis	NI	NI	NI	NI	NI	NI
Trehalase
Rat intestinal trehalase	61	NI	38	75	NI	NI
Glycogen phosphorylase
Rabbit muscle	0.33	NI	NI	NI	NI	NI

^aNI: No inhibition (less than 50% inhibition at 1000 μM).

In order to rationalize the glucosidase inhibition results, it was assumed that these inhibitors mimic the glucoside substrate. The six pyrrolidines were overlayed with glucose to: (i) position the ring nitrogen close to the ring oxygen of glucose so that, in the protonated form, the positive charge can mimic the partial positive charge of the transition state during hydrolysis, and (ii) match as many hydroxy positions as possible, to retain any specific binding interactions in the active sites.^xviii The different glucosidases will recognise the terminal glucose residue of the substrates in different ways. Thus it is not surprising that these inhibitors show slightly different patterns of inhibition for the different glucosidases, however, some general principles can be suggested.

The best match between DAB 3D and glucose placed the C2 and C3 hydroxy (OH) groups of 3D to overlay the C3 and C4 OHs of glucose, respectively [Figure 3(a)]. This positioned the CH₂OH group of 3D close to the C6 OH of glucose and the ring nitrogen of 3D close to the ring oxygen of glucose. In this orientation the inhibitor mimicked three of the glucose OH groups. 4-C-Me-DAB 1D was overlayed with glucose in the same way as 3D, with the inhibitor methyl group in the position of the glucose C5 axial proton [Figure 3(b)]. IsoDAB 2D could be overlayed with glucose in the same way, with the inhibitor CH₂OH group in the position of the glucose C4 axial proton [Figure 3(c)]. This CH₂OH group in 2D might be able to interact with groups which normally bind to the glucose C6 OH, depending on the glucose C5-C6 torsion angle in the bound form. In contrast, the best match between LAB 3L and glucose places the C2 and C3 OHs of 3L overlaying the C2 and C3 OHs of glucose respectively [Figure 3(d)]. This placed the CH₂OH group of 3L close to the C4 OH of glucose and the ring nitrogen of 3L close to the ring oxygen of glucose. Again, in this orientation 3L mimicked three of the glucose OH groups. 4-C-Methyl-LAB 1L was overlayed with glucose in the same way as 3L, with the methyl group of 1L in the position of the glucose C4 axial proton [Figure 3(e)]. If the inactive isoLAB 2L is overlayed in the same manner as 3L, then the CH₂OH group in 2L would be in the position of the glucose C3 axial proton, on the C2OH/C4OH face of glucose. For a terminal α-linked glucose, the rest of the glycan was on the C2OH/C4OH face of the glucose residue (indicated by the large spheres in Figure 3). Alternatively, 2L could be overlayed with the C2 and C3 OHs of 2L overlaying the C3 and C2 OHs of glucose respectively, with the CH₂OH group in 2L in the position of the glucose C2 axial proton, on the opposite face [Figure 3(f)]. In either case, the inhibitor could only mimic two of the four glucose OH groups.

Figure 3.

Overlay of the six inhibitors, (a) DAB 3D, (b) 4-C-Me-DAB 1D, (c) isoDAB 2D, (d) LAB 3L, (e) 4-C-Me-LAB 1L and (f) isoLAB 2L, with an α-glucoside. The glucose residues are in grey, with the oxygens in slightly darker grey, and the inhibitors are in colour. Only the hydroxy protons are shown, for clarity. The large sphere shows the position of the rest of the glycan.

There are several possible explanations for the absence of activity of isoLAB 2L compared to the other five pyrrolidines, including: (i) the inhibitor must mimic at least three of the glucose hydroxy groups for effective binding, or (ii) a hydroxy group near the position of the glucose C4 hydroxy is required for binding, or (iii) an additional axial group at the positions of C2 or C3 in glucose cannot be tolerated, whereas it can at positions C4 or C5. The differential inhibition of different α-glucosidases by the various pyrrolidines showed that the remainder of the substrates can significantly affect the potency of inhibition. Whatever the explanation, it is remarkable to find five such structurally simple pyrrolidines with potent and specific α-glucosidase inhibition.

The effect of 1D and 1L on cellular endoplasmic reticulum (ER) resident α-glucosidase I and II activity in cells was studied using a free oligosaccharide assay.^xix HL-60 cells were treated at concentrations of 1D and 1L as indicated for 24 hours [Figure 4] and were compared with the effect of 1-deoxynojirimycin (DNJ); free oligosaccharides (FOS) were isolated from cells, fluorescently labelled, and followed by separation using NP-HPLC. Peak areas for Glc₁Man₄GlcNAc₁ and Man₄GlcNAc₁ (non-glucosylated control species) were determined and expressed as a ratio. The presence of Glc₁Man₄GlcNAc₁ indicated inhibition of α-glucosidase II; at 500 μM DNJ is five times a more potent inhibitor than 1L and forty times more potent than 1D. The decline in Glc₁Man₄GlcNAc₁ at concentrations of DNJ higher than 500 μM is due to inhibition of α-glucosidase I and therefore production of Glc₃Man₅GlcNAc₁ species at the expense of Glc₁Man₄GlcNAc₁ species. In the presence of 2.5 mM branched imino sugars, cells do not produce any Glc₃Man₅GlcNAc₁ indicating that compounds are either unable to inhibit this enzyme, i.e. are specific α-glucosidase II inhibitors, or most likely, insufficient concentrations have been reached to inhibit the α-glucosidase I step in the pathway. In summary, both 4-C-methyl branched imino sugars of 1D and 1L are weak inhibitors of α-glucosidase II in cells; this is in contrast to the behaviour of the hydroxymethyl branched pyrrolidines 2D and 2L, neither of which showed any inhibition of ER glucosidases.⁸ Both DAB 3D and LAB 3L are presumed inhibitors of processing glucosidases; their N-butyl analogues are weak inhibitors (IC50, 319 μM and 769 μM, respectively) of α-glucosidase I using an in vitro assay and are consequently ineffective at inhibiting glucosidase activity in cellular assays at concentrations of 1 mM or less.^xx

Figure 4.

Glc₁Man₄GlcNAc₁ following α-glucosidase inhibition in FOS assay.

DNJ^xxi [an α-glucosidase inhibitor] and isoLAB 2L⁸ [which has no α-glucosidase inhibition] both show significant rescue of the defective F508del-CFTR function and thus might have potential in the treatment of cystic fibrosis.^xxii Neither of the 4-C-methyl pyrrolidines 1D and 1L had any corrector effect on CFTR function in CF-KM4 cells,^xxiii as assessed by single-cell fluorescence imaging;^xxiv thus, isoLAB 2L – the only one of the DAB analogues that is not an α-glucosidase inhibitor – is the sole pyrrolidine to have chloride channel rescue properties.

In summary, only a single acetonide protecting group is used in the synthesis of the first examples of potent glycosidase inhibitors by C-alkyl branching of the carbon chain of iminosugars. Of the three pairs of enantiomers 1, 2 and 3, five structurally simple pyrrolidines are potent (μM) and specific α-glucosidase inhibitors; isoLAB 2L does not inhibit any glycosidase but is the only one to exhibit significant rescue of the defective F508del-CFTR function. Only DAB 3D shows any inhibition of glycogen phosphorylase.