Large-Scale Synthesis of All Stereoisomers of a 2,3-Unsaturated C-Glycoside Scaffold

Abstract

All stereoisomers of a highly functionalized 2,3-unsaturated C-glycoside can be accessed in 10–100 g quantities from readily available starting materials and reagents in 3–7 steps. These chiral scaffolds contain three stereogenic centers along with orthogonally protected functional groups for downstream reactivity.

Due to their synthetic versatility and high level of stereochemical diversity carbohydrates have served as useful starting points for generating molecular diversity.¹ Carbohydrate-derived glycals in particular have been employed in multiple diversity-oriented synthesis (DOS) pathways.² Of interest to us was the utility of C-alkyl pseudoglycals for developing new build/couple/pair pathways in the context of library development.³ In the present study, we focused on the synthesis of 2,3-unsaturated C-glycosides 1–4 (Figure 1) which incorporate four chemical handles: (1) an ester, (2) an alkene, (3) a primary alcohol and (4) a secondary alcohol/primary amine, thereby, providing a range of options for subsequent modifications and/or functional group pairing reactions.⁴ As part of our design strategy we sought to develop methods for the preparation of all eight stereoisomers of the C-glycoside template to enable the development of stereo/structure-activity relationships.^3b,5 Herein we describe the preparation of C-glycosides 1–4 on large (>50 g) scale.

Figure 1.

Full stereochemical matrix of C-glycosides 1–4

In order to introduce the ester functionality at C-1 we explored a type I Ferrier rearrangement⁶ of tri-O-acetyl-D- and L-glucal to access C-glycosides 1–4. We elected to focus solely on optimizing the large-scale Ferrier reaction for the glucal series with the intention of accessing the galactal-derived material (2) via Mitsunobu inversion of the C-4 allylic alcohol.^7,8 This late stage epimerization strategy is an attractive alternative as it requires optimization of only one Ferrier reaction as well as access to only one unnatural carbohydrate.

Generally, Lewis acids such as TiCl₄,⁹ BF₃·Et₂O,¹⁰ and TMSOTf,^6c are employed in the Ferrier reaction to promote rearrangement to form the active glycosyl intermediate. Once activated, a variety of carbon-based nucleophiles can be utilized,^6d including allyltrimethylsilane,¹¹ trimethylsilylcyanide¹² and various silyl ketene acetals.¹³ We utilized silylketene acetal 5 (1-(tert-butyldimethylsilyloxy)-1-methoxyethene) as a nucleophile for generating the C-glycoside as it offers the advantage of installing an ester side chain without further modifications.

The stereochemical outcome of Lewis acid mediated C-glycosidation reactions are, in general, highly dependent on the conditions employed.^6d For example, it has been reported that in the presence of LiClO₄ in Et₂O, addition of silyl ketene acetal (5) to tri-O-acetyl-D-glucal leads predominantly to the formation of the α-anomer (ratio 3:1, α:β).^13b In contrast, Csuk et al. reported that treatment of tri-O-acetyl-D-glucal with silylketene acetal 5 in the presence of TMSOTf in CH₂Cl₂, affords the β-anomer as the major product (ratio 1:2, α:β).^13c Since we discovered that the α- and β-glycosides could be easily separated by silica gel chromatography and we required access to equal quantities of both anomers, we elected to develop reaction conditions that would achieve a closer to 1:1 α/β ratio.

As shown in Table 1, we evaluated the Ferrier reaction of tri-O-acetyl-D-glucal with 5 under a range of conditions, mainly focused on varying the nature of the Lewis acid and solvent. Use of BF₃·Et₂O as the Lewis acid in CH₂Cl₂ led to the formation of C-glycoside 6 in 45% yield as a 1:2 mixture of anomers favoring the β-anomer (entry 1). Changing the Lewis acid to TMSOTf led to a higher isolated yield (73%) and a more favorable ratio of anomers 1:1.5 (entry 2). Using TMSOTf we next investigated the effect of solvent on reaction selectivity. When employing CH₃CN as a solvent the formation of the α-anomer was slightly favored (α/β ratio = 1.5:1) and a lower yield was obtained (65%, entry 3). The yield of the glycosidation reaction could be improved to 77% on large scale (200 g), using CH₂Cl₂ as a co-solvent (entries 4 and 5), providing a 1.2:1 mixture of anomers. The α/β isomers could be easily separated by silica gel chromatography to provide 42% of the α-anomer (α-D-6) and 35% of the β-anomer (β-D-6).¹⁴ Deacetylation and subsequent selective 6-O-TBDPS protection provided allylic alcohols α-D-1 and β-D-1 in 90% yield over two steps (Scheme 1). This protocol was also applied to tri-O-acetyl-L-glucal¹⁵ on 100-g scale to afford allylic alcohols α–L-1 and β-L-1.¹⁶

Table 1.

Optimization of the Ferrier reaction

entry	Lewis acid	solvent	ratio (α:β)c	yield (%)
1a	BF3·Et2O	CH2Cl2	1:2	45d
2a	TMSOTf	CH2Cl2	1:1.5	73d
3a	TMSOTf	CH3CN	1.5:1	65d
4a	TMSOTf	CH2Cl2/CH3CN(1:1)	1.2:1	72 (39:33)e
5b	TMSOTf	CH2Cl2/CH3CN (1:1)	1.2:1	77 (42:35)e

^a1.2 equiv. Lewis acid, 0.15 M, 2.5 mmol scale.

^b1.2 equiv. Lewis acid, 0.15 M, 735 mmol scale.

^cRatio between α and β stereoisomers determinated by ¹H NMR.

^dIsolated yield for the mixture of α-D-6 and β-D-6.

^eIsolated yield for each anomer (α-D-6:β-D-6)

Scheme 1.

Deacetylation and TBDPS-protection

We next investigated the conversion of α/β-D-1 and α/β-L-1 to the corresponding C-4 epimers via Mitsunobu inversion.⁸ After varying parameters such as phosphine, solvent and temperature, we arrived at a reliable and robust protocol for the Mitsunobu reaction. Focusing initially on the L-series, allylic alcohols α-L-1 and β-L-1 were displaced with p-nitrobenzoic acid in the presence of PBu₃ and DIAD at 0 °C (Scheme 2). Hydrolysis of the resulting benzoates cleanly afforded the enantiopure allylic alcohols α-L-2 and β-L-2 in 68% and 72% yield over two steps respectively. This procedure could then be applied to allylic alcohols α-D-1 and β-D-1 to afford glycosides α-D-2 and β-D-2 thereby completing the full matrix of eight stereosiomers.

Scheme 2.

Mitsunobu inversion of C-4 allylic alcohol

Conversion of allylic alcohols 1 and 2 to amines 3 and 4 was our final task to prepare the desired scaffolds. Introduction of an azido substituent as a precursor to the amine was initially considered using diphenylphosphoryl azide (DPPA)¹⁷ or the zinc complex Zn(N₃)₂Py₂¹⁸ but neither of these options proved useful on multigram scale. In most cases, complex mixtures were obtained and the desired products were isolated in low yield. Successful introduction of the C-4 amine was achieved using phthalimide as a nitrogen nucleophile¹⁹ via a Mitsunobu reaction (Table 2). Essential to reproducible high yields and selectivity was the use of PBu₃, which proved to be more efficient than triphenylphosphine (PPh₃) for the Mitsunobu reaction of glucal-based derivatives α-D-1 and β-D-1 to yield α-D-7 and β-D-7 in 75% and 84% yield respectively (entries 3 and 4). Under these conditions, the formation of undesired S_N2′ byproducts,^19,20 which had been observed during the PPh₃-mediated Mitsunobu reactions, was diminished. By contrast, for the C-4 epimeric glycosides (α–D-2 and β-D-2) PPh₃ proved to be the optimal Mitsunobu reagent affording allylic phthalimides α-D-8 and β-D-8 in 85% and 89% yield (entries 5 and 6).²¹

Table 2.

Introduction of C-4 amine via Mitsunobu reaction

entry	SM	product	PR3	yield (%)a
1	α-D-1	α-D-7	PPh3	62
2	β-D-1	β-D-7	PPh3	54
3	α-D-1	α-D-7	PBu3	75
4	β-D-1	β-D-7	PBu3	84
5	α-D-2	α-D-8	PPh3	85
6	β-D-2	β-D-8	PPh3	89

^aIsolated yield after silica gel chromatography

Lastly, removal of the phthalimide protecting group was achieved via treatment with ethylenediamine²² to afford allylic amines 3 and 4. Application of these conditions to all stereoisomers led to isolation of the desired eight allylic amines in high yield (>90%).

In summary, we have demonstrated that all possible stereochemical permutations of two highly functionalized C-glycosides can be prepared from tri-O-acetyl-D- and L-glucal on large scale. The use of these 16 pyran scaffolds for library synthesis is forthcoming.²³

Experimental Section

General procedure for Ferrier reaction: C-Glycosides (1R,4S,5S)-(+)-α-D-6 and (1S,4S,5R)-(+)-β-D-6

Into a 5-L flask containing CH₂Cl₂ (1.8 L) and CH₃CN (2.0 L) was added tri-O-acetyl-D-glucal (200 g, 735 mmol). The reaction mixture was cooled to −5 °C and silyl ketene acetal 5 (135 mL, 882 mmol, 1.2 equiv) was added. A solution of TMSOTf (153 mL, 845 mmol, 1.1 equiv) in CH₂Cl₂ (200 mL) was added slowly via an addition funnel over 30 min. After complete addition of the Lewis acid, the bath temperature was kept at −5 °C for 2 h. The reaction mixture was poured slowly into sat. aq. NaHCO₃ solution (1.5 L). Excess solvent was removed in vacuo and the crude yellow mixture was extracted with CH₂Cl₂ (3 × 1 L). The combined organic layers were washed with brine and then dried over MgSO₄. Solvent was removed in vacuo and the crude material was purified by silica gel chromatography (hexanes/ EtOAc 95/5 to 70/30) to afford first 73 g (35%) of β-D-6 and then 88 g (42%) of α-D-6 as yellow oils.¹⁴

General procedure for deacetylation/TBDPS protection: Allylic alcohol (1S,4S,5R)-(+)-β-D-1

To a 5-L flask containingβ-D-6 (95 g, 332 mmol) dissolved in 4:1 MeOH:CH₂Cl₂ (3.3 L) was added K₂CO₃ (2.30 g, 16.6 mmol, 0.05 equiv). The reaction mixture was stirred at rt for 100 min and then AcOH (3.80 mL, 66.4 mmol, 0.2 equiv) was added. Solvent was removed in vacuo and the product was filtered through a column of silica gel and eluted with 1 L of 70% hexanes/EtOAc followed by 3 L of straight EtOAc to provide the desired diol, which was isolated as a yellow oil (67 g, 99% yield).

To a 5-L flask containing diol (67 g, 332 mmol) was added CH₂Cl₂ (1.5 L) followed by Et₃N (93 mL, 664 mmol, 2.0 equiv) and DMAP (4.06 g, 33.2 mmol, 0.1 equiv). The reaction mixture was cooled to 0 °C and TBDPS-Cl (90 mL, 349 mmol, 1.05 equiv) was slowly added. The reaction was slowly warmed to rt and stirred for 46 h. The solvent was removed in vacuo and the crude oil was diluted with Et₂O and filtered through Celite to remove the resulting white solid. The mixture was concentrated and the resulting oil was redissolved in Et₂O, washed with sat. NH₄Cl and brine. After drying over MgSO₄, the solvent was removed in vacuo. The yellow oil was then purified by silica gel chromatography (hexanes/EtOAc 95/5 to 70/30) to afford alllylic alcohol β-D-1 as a colorless oil (131 g, 90% yield). [α]_D²⁰ +4.1 (c 1.0, CHCl₃). IR (film) ν 2930, 2858, 1736, 1427, 1361, 1277, 1200, 1111, 1088; ¹H NMR (300 MHz, CDCl₃) δ 7.67 (m, 4H), 7.40 (m, 6H), 5.83 (m, 1H), 5.73 (m, 1H), 4.52 (m, 1H), 4.27 (m, 1H), 3.90 (dd, J = 10.2, 5.1 Hz, 1H), 3.76 (dd, J = 10.2, 7.2 Hz, 1H), 3.64 (s, 3H), 3.50 (m, 1H), 2.45 (m, 1H), 1.04 (s, 9H); ¹³C NMR (75 MHz, CDCl₃) δ 171.2, 135.8, 135.7., 132.9, 132.8, 130.1, 129.7, 129.6, 128.9, 128.0, 127.8, 77.7, 71.3, 66.7, 66.3, 51.9, 40.4, 27.0, 19.3; HRMS (ESI+) calcd for C₂₅H₃₂O₅Si [M + H₂O]⁺: 458.2363, found: 458.2359

General procedure for Mitsunobu Inversion at C-4: Allylic alcohol (1S,4S,5S)-(+)-α-L-2

To an oven-dried 3-neck 1-L round-bottom flask equipped with a 150 mL oven-dried addition funnel and placed under argon, was added a solution of (+)-α-L-1 (29.0 g, 65.8 mmol, 1.0 equiv) and p-nitrobenzoic acid (12.1 g, 72.4 mmol, 1.1 equiv) in dry THF (370 mL). The mixture was degassed with argon (sparge for 30 min) and then cooled to 0 °C. PBu₃ (24.4 mL, 99.0 mmol, 1.5 equiv) was added via syringe to the reaction mixture and a solution of DIAD (19.5 mL, 99.0 mmol, 1.5 equiv) in degassed THF (70 mL) was slowly added via addition funnel over 30 min. After stirring at rt overnight, the solvents were removed in vacuo and the crude product was taken onto the next step without purification.

A 1-L round bottom flask was charged with the crude p-nitrobenzoic ester as a solution in CH₂Cl₂ (130 mL) and MeOH (530 mL) and K₂CO₃ (455 mg, 3.3 mmol, 0.05 equiv) was added. The resulting dark orange mixture was stirred at rt for 2 h and then AcOH (0.57 mL, 9.9 mmol, 0.15 equiv) was added. After evaporation of the solvents, the crude product was dissolved in Et₂O and washed with sat. NH₄Cl aq. solution (2x). The aqueous phase was extracted with Et₂O and the combined organic layers were washed with brine, dried over MgSO₄, filtered and concentrated. The crude product was purified by silica gel chromatography (hexanes to 7:3 hexanes/EtOAc) to afford allylic alcohol α-L-2 (19.8 g, 68% yield, over two steps) as a thick colorless oil. [α]_D²⁰ +78.2 (c 1.5, CHCl₃); IR (film) ν 2930, 2856, 1735, 1428, 1361, 1272, 1200, 1100, 1086; ¹H NMR (300 MHz, CDCl₃) δ 7.67 (m, 4H), 7.39 (m, 6H), 6.08 (ddd, J = 9.9, 5.5, 1.5, 1H), 5.89 (dd, J = 10.1, 3.2 Hz, 1H), 4.73 (m, 1H), 3.96 (dd, J = 8.7, 6.1 Hz, 1H), 3.86 (dd, J = 8.8, 4.4 Hz, 1H), 3.77 (m, 1H), 3.62 (s, 3H), 2.66 (d, J = 15.2, 9.0 Hz, 1H), 2.43 (d, J = 15.3, 5.5 Hz, 1H), 1.85 (d, J = 8.9 Hz, 1H), 1.04 (s, 9H); ¹³C NMR (75 MHz, CDCl₃) δ 171.2, 135.8, 135.8, 133.6, 133.5, 131.8, 130.0, 129.9, 127.9, 127.8, 127.6, 72.4, 70.1, 63.6, 62.2, 52.0, 37.4, 27.0, 19.3; HRMS (ESI+) calcd for C₂₅H₃₂O₅Si [M + H]⁺: 441.2097, found: 441.2100.

General procedure for phthalimide Mitsunobu reaction: Allylic phthalimide (1S,4R,5S)-(−)-β-D-7

To an oven-dried 3-neck 2-L round-bottom flask equipped with a 250-mL addition funnel, was added a solution of (+)-β-D-1 (46.5 g, 106 mmol, 1.0 equiv) and phthalimide (23.3 g, 158 mmol, 1.5 equiv) in dry THF (800 mL). The mixture was degassed with argon (sparge for 60 min) and then cooled to 0 °C. Tributylphosphine (35.0 mL, 137.1 mmol, 1.3 equiv) (or triphenylphosphine, 1.3 equiv) was added, followed by a solution of DIAD (28.7 mL, 137.1 mmol, 1.3 equiv) in degassed THF (200 mL) via addition funnel (over 60 min). After stirring at rt overnight, the solvent was removed in vacuo and the crude product was purified by silica gel chromatography (hexanes to 7:3 hexanes/EtOAc) to afford allylic phthalimide β-D-7 (50.7 g, 84% yield). [α]_D²⁰ −86.2 (c 1.0, CHCl₃). IR (film) ν 2926, 2856, 1716, 1352, 1112; ¹H NMR (500 MHz, CDCl₃) δ 7.75 (dd, J = 5.4, 3.1 Hz, 2H), 7.67 (dd, J = 5.5, 3.1 Hz, 2H), 7.57 (dd, J = 7.9, 1.4 Hz, 2H), 7.49 (dd, J = 7.9, 1.1 Hz, 2H), 7.37 – 7.25 (m, 4H), 7.17 (t, J = 7.5 Hz, 2H), 6.09 (d, J = 10.0 Hz, 1H), 5.78 (ddd, J = 10.0, 5.5, 2.0 Hz, 1H), 4.81 (br d, J = 3.6 Hz, 1H), 4.67 (br t, J = 6.9 Hz, 1H), 3.97 (td, J = 6.0, 3.5 Hz, 1H), 3.67 (s, 3H), 3.65 (dd, J = 10.8, 5.7 Hz, 3H), 3.59 (dd, J = 10.8, 6.5 Hz, 1H), 2.89 (dd, J = 15.8, 7.4 Hz, 1H), 2.71 (dd, J = 15.8, 6.5 Hz, 1H), 0.92 (s, 9H); ¹³C NMR (75 MHz, CDCl₃) δ 171.5, 167.8, 135.5, 135.4, 133.9, 133.8, 133.2, 133.1, 131.8, 129.5, 129.4, 127.5, 127.4, 123.1, 121.4, 75.9, 71.5, 63.6, 51.6, 44.7, 38.8, 26.6, 18.9; HRMS (ESI+) calcd for C₃₃H₃₅NO₆Si [M + Na]⁺: 592.2131, found: 592.2134.

General procedure for phtalimide removal: Allylic amine (1S,4R,5S)-(−)-β-D-3

A 2-neck 2-L flask equipped with a reflux condenser was charged with the allylic phthalimide (25 g, 43.9 mmol, 1.0 equiv) in THF (44 mL) and 1,2-ethylenediamine (8.8 mL, 131.7 mmol, 3.0 equiv) was added at rt via syringe. The reaction mixture was stirred for 2–4 h then MeOH (835 mL) was added. The reaction mixture was stirred at 60 °C for 16 h, at which time a white precipitate had formed and the reaction was deemed complete (LCMS). The solvents were partially evaporated and Et₂O was added to the crude product. After filtration through a pad of Celite, the solvents were partially evaporated and the process was repeated once more to remove any excess of white byproduct. The solution was evaporated to dryness and the crude product was purified by silica gel chromatography (CH₂Cl₂/MeOH 100:0 to 90:10) to afford allylic amine β-D-3 as a pale yellow oil (17.6 g, 91% yield). [α]_D²⁰ minus;44.2 (c 1.2, CHCl₃); IR (film) ν 2930, 2856, 1740, 1428, 1168, 1112; ¹H NMR (500 MHz, CDCl₃) δ 7.66 (d, J = 6.6 Hz, 2H), 7.63 (d, J = 6.5 Hz, 2H), 7.42 – 7.35 (m, 6H), 6.02 (dd, J = 9.8, 5.6 Hz, 1H), 5.69 (d, J = 10.1 Hz, 1H), 4.46 (br t, J = 5.8 Hz, 1H), 3.82 (dd, J = 14.6, 10.0 Hz, 1H), 3.73 – 3.66 (m, 2H), 3.63 (s, 3H), 3.24 (br d, J = 5.2 Hz, 1H), 2.50 (dd, J = 15.4, 7.5 Hz, 1H), 2.43 (dd, J = 15.4, 5.8 Hz, 1H), 1.21 (br s, 2H), 1.04 (s, 9H); ¹³C NMR (125 MHz, CDCl₃) δ 171.0, 135.5, 133.3, 133.2, 130.0, 130.0, 129.7, 129.6, 127.7, 127.6, 77.2, 71.9, 62.7, 51.6, 45.0, 40.0, 26.8, 19.1; HRMS (ESI+) calcd for C₂₅H₃₃NO₄Si [M + H]⁺: 440.2257, found: 440.2252.

Scheme 3.

Phthalimide removal to yield amines 3 and 4