Alkynylation of Pentose Derivatives with Stereochemical Fidelity: Implications for the Regioselectivity of Alkynyl Diol Cycloisomerizations to Cyclic Enol Ethers

Abstract

This work characterizes a previously undetected epimerization in the preparation of alkynyl diols from pentose precursors utilizing the Ohira–Bestmann reagent. Lithium trimethylsilyldiazomethane (Colvin reagent) additions to the d-ribose and d-lyxose-derived benzylidene acetals provide the respective alkynyl diol stereoisomers, without epimerization. Regioselective tungsten-catalyzed cycloisomerizations of the d-ribose- and d-lyxose-derived alkynyl diols yield rigid bicyclic pyranose glycals, confirming the stereochemical fidelity of the Colvin alkynylation process.

Graphical Abstract

Our laboratory previously described the synthesis of seven-membered cyclic enol ethers 2, prepared by tungsten carbonyl-catalyzed cycloisomerizations of alkynyl diols 1 (eq 1).^1,2 The alkynyl diol substrates included

(eq 1)

dioxolane spacers between the terminal alkyne and vicinal diol moieties as a necessary condition for seven-membered ring formation. We recently re-explored this work, aiming to extend its synthetic applications, but uncovered a previously undetected epimerization in the preparation of alkynyl diols closely related to 1.

In our initial report,¹ we claimed that all four diastereomers of 1 gave the corresponding septanose glycals 2 with the assignment of regioselectivity for the seven-membered ring secured by O-acylation of secondary alcohols to the acetate esters 3 (eq 1). This manuscript describes how we identified epimerization of these compounds at carbon 3 (C3) and offers a reliable method for synthesizing alkynyl diols with stereochemical fidelity. We also discuss the implications of substrate stereochemistry on the regioselectivity of tungsten-catalyzed cycloisomerizations of alkynyl diols.

For several of the substrates reported earlier,¹ we prepared the alkynyl diols by the Ohira–Bestmann protocol³ from pentose precursors protected as acetonides. In this extension of our work, we chose benzylidene acetal protection, intending late-stage chemoselective deprotection in the presence of acid-sensitive enol ether and/or glycoside functional groups. Beginning with d-ribofuranose 4 and the known benzylidene acetal 5,⁴ Ohira–Bestmann alkynylation with diazophosphonate 6 gave the alkynyl diol 7 (Scheme 1). As this transformation yielded a single diastereomer, we were initially unaware that the propargylic center at C3 of compound 7 had epimerized.⁵ Tungsten carbonyl-catalyzed cycloisomerization^1,2 of the alkynyl diol 7 produced the septanose glycal 8. In a serendipitous opportunity, we obtained a crystal structure of 8, revealing that the chiral center at C3 was opposite to the configuration of the d-ribofuranose starting material. Thus, compound 8 corresponded to the d-arabino stereochemistry.

Scheme 1.

Synthesis and Crystal Structure of d-Arabinoseptanose Glycal 8, Arising from Epimerization at C3 in the Alkynylation Step

At the same time, we converted d-lyxofuranose 9 into the mixture of benzylidene acetal diastereomers 10a and 10b, followed by Ohira–Bestmann alkynylation, affording alkynyl diols 11a and 11b (Scheme 2). Although the acetal diastereomers 11a and 11b were difficult to separate, tungsten carbonyl-catalyzed cycloisomerization of the mixture gave the chromatographically separable septanose glycals 12a and 12b. Crystallographic analysis of 12a exhibited d-xylo stereochemistry, consistent with epimerization at C3, which we ultimately determined occurred in the alkynylation step.

Scheme 2.

Synthesis of d-Xyloseptanose Glycals 12a and 12b with Epimerization at C3 Confirmed by the Crystal Structure of 12a

Following precedents describing alkynylations of α-chiral aldehydes with lithium trimethylsilyldiazomethane (13) without epimerization,^6,7 we successfully applied this method (Colvin protocol) with benzylidene-protected furanoses 5 and 10a,b (Scheme 3). In contrast with the Ohira–Bestmann protocol, the reaction of 13 with ribose-derived furanose 5 proceeded without epimerization. A mixture of terminal alkyne and alkynyltrimethylsilane was initially produced, but basic methanolysis of the crude product mixture promoted protiodesilylation to provide the alkynyl diol 14. We obtained similar results from the benzylidene-protected lyxofuranose 10a and 10b, producing 15a and 15b. ¹H NMR data for 14 and 15a,b exhibited small but discernible differences in chemical shifts compared with 7 or 11a,b, as expected for diastereomers.

Scheme 3.

Preparations of Alkynyl Diols 14 and 15a,b without Epimerization in the Alkynylation Steps

As the Ohira–Bestmann conditions exposed the furanose substrates 5 and 10 to refluxing methanolic potassium carbonate, followed by slow addition of the diazophosphonate 6, we realized that epimerization of the open-chain aldehyde–diol intermediate (Figure 1, i → ii) likely occurred faster than alkynylation. In hindsight, we recognized that ii was probably thermodynamically favored over diastereomer i, minimizing gauche interactions between the aldehyde and diol substituents on the dioxolane ring. However, the Colvin protocol proceeded at low temperature in THF under aprotic conditions that disfavored epimerization prior to carbon–carbon bond formation.⁸

Figure 1.

Competitive epimerization vs alkynylation depicted for alkynylation of benzylidene-protected d-ribofuranose 5.

In contrast to our initial report, tungsten-catalyzed cycloisomerization of the ribo-alkynyl diol 14 gave the six-membered d-ribopyranose glycal 16 (Scheme 4). Similarly, the lyxo-alkynyl diols 15a,b produced the d-lyxopyranose glycals 17a,b, which were chromatographically separable. In both cases, the seven-membered ring septanose isomers were not observed.

Scheme 4.

Cycloisomerizations of ribo- and lyxo-Alkynyl Diols 14 and 15 Provide the Corresponding Cis-Fused Pyranose Glycals 16 and 17

We sought spectroscopic signatures to distinguish the relative stereochemistry of alkynyl diol diastereomers. Unfortunately, the coupling constants of the propargylic hydrogens (H3) of all alkynyl diol diastereomers were similar in magnitude, exhibiting ³J_H3,H4 coupling constants between 5–8 Hz. However, the alkynyl diol diastereomers did exhibit regular differences in the relative chemical shifts for the hydrogens at C5: the arabino-7 and xylo-11a,b diastereomers bearing a trans-relationship at H3 and H4 (arising from epimerization) had chemical shifts for H5 upfield of 4.0 ppm, whereas the ribo-14 and lyxo-15a,b diastereomers bearing a cis-relationship of H3 and H4 (absence of epimerization) showed chemical shifts for H5 downfield of 4.0 ppm (Figure 2).

Figure 2.

Diagnostic chemical shift trends for the hydrogens at C5 distinguish trans- vs cis-diastereomers at C3 and C4.

For the bicyclic glycals 16 and 17a,b, the coupling constants confirmed the stereochemistry at C3 and C4, due to the rigidity of the bicyclic structures (Figure 3). Specifically, the arabino-8 and xylo-12a,b septanose glycal diastereomers bearing a trans-relationship of H3 and H4 exhibited relatively large ³J_H3,_H4 of 9.7 Hz, consistent with anti-orientations of H3 and H4. In contrast, the ribo-16 and lyxo-17a,b pyranose glycal diastereomers bearing cis-relationships of H3 and H4 (absence of epimerization) displayed smaller ³J_H3,_H4 ranging from 5.8 to 6.8 Hz, in the range expected for gauche orientations of H3 and H4. Consistent with these assignments, the ribo-16 pyranose glycal gave larger ³J_H4,_H5 of 10.0 Hz as expected for trans-diaxial hydrogens, whereas the lyxo-17a–17b pyranose diastereomers showed very small or nonobservable ³J_H4,H5 < 2 Hz.

Figure 3.

Consistent trends in the coupling constants distinguish the cyclic glycal products.

These results demonstrate that only dioxolane-protected alkynyl diols leading to trans-ring fusions produce the septanose glycals by tungsten-catalyzed cycloisomerization. As depicted for the arabino-diastereomer 7 (Figure 4), ring strain disfavors formation of a trans-fused [4.3.0]-bicyclic isomer, allowing the alternative pathway to the seven-membered ring [5.3.0]-bicyclic isomer 8. However, both pathways are now possible from the ribo-diastereomer 14: not surprisingly, the six-membered ring isomer 16 forms faster than the seven-membered ring.

Figure 4.

Regioselectivity of tungsten-catalyzed cycloisomerization depends upon the relative stereochemistry of the C3 and C4 oxygens when protected as cyclic acetals.

These findings require corrections to two of our earlier publications.^1,9