An efficient, palladium-catalyzed, enantioselective synthesis of (2R)-3-butene-1,2-diol and its use in highly selective Heck reactions

Abstract

A robust and scalable procedure for the palladium-catalyzed dynamic kinetic asymmetric transformation of 3,4-epoxy-1-butene into (2R)-3-butene-1,2-diol with water as the cosolvent is reported. Examination of the effects of solvent and temperature led to the identification of conditions that permitted use of 0.025 mol % catalyst, providing (2R)-3-butene-1,2-diol in 84% isolated yield and 85% enantiomeric excess. Subsequent Heck reactions with a diverse range of coupling partners are described and the influence of their electronic nature on maintaining the enantiopurity of the diol is discussed.

Enhanced drug efficacy, more economic use of resources, and, above all else, ever more stringent regulatory policies have all contributed to an increased demand from the pharmaceutical industry for single-enantiomer compounds. This has been a major driving force within the chemical community for the development of novel methods for the asymmetric synthesis of a myriad of chiral building blocks. Single-enantiomer vinylglycols, such as (2R)-3-butene-1,2-diol, (R)-1, are versatile polyfunctional chiral synthons that can be can be manipulated in a highly stereo- and chemoselective manner. The utility of 1, in particular, has been aptly demonstrated in the synthesis of a wide variety of chiral building blocks (1). Furthermore, 1 has been used as an intermediate in the preparation of pharmaceutical agents in a wide variety of therapeutic areas such as HIV protease inhibitors (2), immunosuppression (3), and oncology (4). A number of approaches to the synthesis of enantiomerically enriched 1 based on the use of a chiral auxiliary (5), chiral-pool starting materials (6, 7), and resolution have been reported (8–12). However, these all suffer from factors, such as lengthy synthetic sequences or a maximum theoretical yield of 50%, making them less than ideal approaches from an industrial perspective.

An alternative approach to enantiomerically enriched 1 that proceeded via asymmetric palladium-catalyzed hydrolysis of 3,4-epoxy-1-butene, 2, in which the ligand 3 was superior to ligand 4 (Fig. 1), was reported by Trost and McEachern (13). By taking advantage of the interconversion of the π-allyl intermediates 5 and 6, and the chiral environment provided by the ligand–metal complex, a dynamic system is established that provides high-enantiopurity product from racemic starting material. For the diol 1 to be obtained, cocatalysis with alkylboranes was reported to be necessary, whereas in the absence of such species the cyclic carbonate 7 was the reaction product.

Fig. 1.

Dynamic kinetic asymmetric transformations of 3,4-epoxy-1-butene. Pd₂dba₃·CHCl₃, tris(dibenzylideneacetone)dipalladium(0)·chloroform adduct.

The high synthetic and economic value of single-enantiomer 1 and the potential for its formation in high yield by using a catalytic system led us to evaluate this method for its application on an industrial scale. In this article, we report our early-phase developmental work on this reaction, which led to a highly efficient, scalable, and robust modified route to 1. Along with this development, we also report subsequent selective Heck reactions of (R)-1. The Heck reaction products potentially serve as intermediates in the synthesis of pharmacologically active compounds, such as 2-aminotetralins (14, 15).

Methods and Materials

General. Commercially available chemicals were used without further purification. 3,4-Epoxy-1-butene, 2, was supplied by Eastman Chemical Company. (1S,2S)-(–)-1,2-Diaminocyclohexane-N,N′-bis(2-diphenylphospino-1-naphthoyl), 3, is manufactured by Dowpharma, and research quantities are commercially available from Strem Chemicals (Newburyport, MA). Reactions were carried out under an atmosphere of nitrogen, and the solvents used in palladium-catalyzed reactions were deoxygenated by either sparging with nitrogen or subjecting them to repeated evacuation–nitrogen refill cycles. Reactions were monitored by using GLC or with analytical TLC on silica gel 60 F254 plates (Machery & Nagel) and visualized under UV (254 nm) and/or by staining with either acidic ceric ammonium molybdate or basic potassium permanganate solution. Flash column chromatography was performed on silica gel (35–70 μm). ¹H NMR spectra were recorded on a Bruker WH-400 (400 MHz), and chemical shifts (in ppm) were determined relative to tetramethylsilane in deuterated chloroform (δ = 0 ppm). Coupling constants are reported in Hz. ¹³C{¹H} NMR spectra were recorded on the same Bruker NMR spectrometer (100.6 MHz) and were calibrated with CDCl₃ (δ = 77.0 ppm). Optical rotations were recorded on a Perkin–Elmer 341 polarimeter. Melting points were determined by using a Kohfler hot-stage apparatus and are uncorrected. A typical procedure for the Heck reaction is reported below. Characterization data for 1 and 14–19 are reported separately in Supporting Text, which is published as supporting information on the PNAS web site.

(2R)-3-Butene-1,2-diol, (R)-1. A 5-liter, four-necked jacketed vessel fitted with a mechanical stirrer, thermometer, septum, and gas sparger was charged with water (1,700 ml), dichloromethane (1,090 ml), sodium bicarbonate (240.1 g, 2.86 mol), and tetrabutylammonium chloride (3.97 g, 14.3 mmol). Nitrogen was bubbled through the stirred suspension for 90 min and the sparger was then removed, keeping the system under a nitrogen atmosphere. A 100-ml Schlenk flask fitted with a stirrer bar was charged with Pd₂dba₃·CHCl₃ (185 mg, 0.179 mmol) and (S,S)-3 (310 mg, 0.393 mmol). The flask was evacuated for ≈2 min (≈15 mbar; 1 mbar = 100 Pa) and refilled with nitrogen, and this process was repeated three times. Dichloromethane (50 ml), which was degassed by bubbling nitrogen through it while stirring for 1 h, was then added to the Schlenk flask and the mixture was stirred. The catalyst solution was then added to the 5-liter vessel, followed by 3,4-epoxy-1-butene, 2 (100 g, 1.43 mol). The rapidly stirred mixture was then heated to reflux (internal temperature: 39°C). After 16 h the heating was ceased and GC analysis of the dichloromethane layer indicated >95% consumption of 2 and ≈92% hydrolysis of the cyclic carbonate 7. The aqueous layer was transferred to a separating funnel, washed with dichloromethane (twice with 250 ml), and then concentrated under reduced pressure to ≈500 ml. After careful neutralization to pH 7.2 with 12 M HCl (≈240 ml), the mixture was filtered and the filtrate was extracted with tetrahydrofuran (THF) (five times with 400 ml). The combined THF extracts were concentrated under reduced pressure and the residue was then redissolved in methyl tert-butyl ether (MTBE; 400 ml) and dried (Na₂SO₄). Filtration followed by concentration under reduced pressure provided 1 with 85% enantiomeric excess (ee) as a light yellow liquid (105.1 g, 84%).

(2R,3E)-4-Phenyl-3-butene-1,2-diol (14). A stirred solution of the diol 1 (244 mg, 2.77 mmol) in acetonitrile (2.5 ml) was evacuated (≈15 mbar) and nitrogen refilled. This cycle was repeated three times. Freshly ground potassium carbonate (2.29 g, 16.62 mmol), tri-o-tolylphosphine [P(o-tol)₃; 25.3 mg, 0.083 mmol), palladium acetate (6.2 mg, 0.028 mmol), and bromobenzene (287 μl, 2.77 mmol) were sequentially added. After heating in an oil bath at 95°C for 17 h, the mixture was diluted with MTBE (5 ml) and washed with water (10 ml). The aqueous phase was extracted with MTBE (three times with 10 ml) and the combined organic extracts were washed with brine (10 ml), dried (Na₂SO₄), filtered, and concentrated under reduced pressure. Purification by flash column chromatography (eluant: MTBE/heptane, 9:1, vol/vol) provided 14 with 85% ee as a light tan solid (200 mg, 44%).

Results and Discussion

Initial Evaluation of the Direct Synthesis of (2R)-3-Butene-1,2-diol, (R)-1. Early studies focused on examining the factors that played a key role in determining the enantiopurity of the diol 1 produced by using the boron cocatalyzed protocol. We then planned to assess the reaction by using conditions that would be attractive from an industrial perspective, paying particular attention to the concentration and catalyst loading. Throughout this work we were aware that the formation of oligomers arising from reaction of the diol 1 with residual epoxide 2 would be a key parameter to monitor.

In contrast to the original report, we found that sodium bicarbonate was far more effective than sodium carbonate in producing 1, although the enantiopurity was similar with the two bases (87–88% ee). Increasing the level of triethylborane or the amount of water led to a decrease in the enantioselectivity of the reaction, although the reaction was more sensitive to the former. Increasing the external temperature (30°Cto40°C) led to a much faster reaction without affecting the enantioselectivity, but provided significant oligomeric products. The formation of oligomers remained a consistent problem when the reactions were carried out at the level of catalyst loading described in the original report. However, when the reaction was scaled up and the catalyst loading was decreased to 0.1 mol %, the yield, purity, and enantiopurity of 1 increased. Attempts to further decrease the catalyst loading to 0.05 mol % led to incomplete reaction. Furthermore, as the reaction was scaled up further with 0.1 mol % catalyst, the enantioselectivity decreased and 1 was obtained in only 76% ee. This result, along with susceptibility of the enantioselectivity to subtle changes in the amount of boron cocatalyst and phase transfer catalyst led us to examine a different protocol.

A Modified Synthesis of (2R)-3-Butene-1,2-diol, (R)-1. We reasoned that during formation of the cyclic carbonate 7 sodium hydroxide would be formed as a by-product and that at under sufficiently basic conditions the hydroxide could potentially hydrolyze the cyclic carbonate 7 to provide 1. Furthermore, we anticipated that the immediate hydrolysis product would reside in the aqueous layer of the biphasic reaction media, and exclusion from the other components and products of the reaction would result in minimal oligomer formation. When the conditions reported to provide the cyclic carbonate 7 were used, the level of catalyst loading was decreased from that reported in the original procedure (1 mol % to 0.05 mol %) and the concentration was increased (0.1 M to 1.0 M). The reaction times became extended and 7 was isolated in only 41% yield (Fig. 2). Acidification of the aqueous phase resulted in the isolation of 1 in 38% yield with 87% ee. The diol 1 obtained by using this modified procedure was far cleaner than material prepared via the boron-cocatalyzed route, and we focused our attention on developing this reaction.

Fig. 2.

Initial studies to produce 1 via the cyclic carbonate 7.

From an environmental perspective, dichloromethane is far from the ideal solvent. However, a screen of eight different solvents (THF, MTBE, toluene, ethyl acetate, heptane, acetone, tetrachloroethylene, and α,α,α-trifluorotoluene) indicated that dichloromethane was by far the best solvent for this reaction, with none of the other solvents providing greater than 65% conversion of 2 into 1 and 7. Only THF and MTBE provided a level of enantiopurity similar to that of dichloromethane, but efforts to increase the conversion by using higher temperature were detrimental to the enantioselectivity of the reaction. The amount of phase transfer catalyst had a minimal effect on the outcome of the reaction in terms of the level of hydrolysis of 7 or the enantioselectivity. However, the temperature had a dramatic effect, with essentially complete conversion through to 1 occurring when the external temperature was increased from 30°Cto50°C. The diol 1 was obtained in 63% isolated yield with 86% ee by using 0.05 mol % catalyst. When the reaction was further scaled up to a 100-g input of 2, the diol 1 was obtained in 84% isolated yield with 85% ee by using 0.025 mol % active catalyst (molar substrate/catalyst: 4,000) (Fig. 3). Such a low level of catalyst loading reflects the high activity of the catalyst and is unprecedented in the field of palladium-catalyzed reactions promoted with the ligands 3 and 4 (16). As a consequence of the immediate hydrolysis product being water soluble, the diol 1 was obtained in a very pure state and separation from organic impurities was easily achieved. Methods to upgrade the enantiopurity of 1 have been reported (9).

Fig. 3.

Optimized route to diol 1.

Heck Reactions of 3-Butene-1,2-diol. It is well known that Heck reactions of allylic alcohols 8 and 9 with aryl halides are complicated by competitive formation of the 3-oxoalkylbenzenes 10 and 11, respectively, (path A in Fig. 4) as well as the desired (3-hydroxyalkenyl)benzenes 12 and 13 (path B in Fig. 4) (17). Although conditions have been reported that favor formation of 12 by using stoichiometric AgOAc or Ag₂CO₃ (18), this is not an economically viable solution from an industrial viewpoint. Conditions for the selective formation of 12 and 13 by using iodobenzene have also been reported (19, 20), but both of these methods seem limited to aryl iodides as the coupling partner (21). A more diverse range of aryl bromides are commercially available compared with the corresponding aryl iodides and this availability, in combination with their lower cost, led us to seek conditions which would promote Heck reactions of 1 with aryl bromides.

Fig. 4.

Reaction pathways for Heck reactions of 8 and 9.

Initial studies were carried out with bromobenzene and (±)-1 because of its availability. A number of different palladium sources [Pd(OAc)₂ and Pd₂dba₃·CHCl₃], ligands [PBu₃, P^tBu₃, P(o-tol)₃, 1,2-bis(diphenylphosphino)ethane (dppe), and tris(2,4-di-tert-butylphenyl)phosphite], bases (Na₂CO₃, K₂CO₃, and Cs₂CO₃), and solvents (dimethylformamide, acetonitrile, dioxane, and 1,2-dimethoxyethane) were examined. This search resulted in the combination of Pd(OAc)₂, P(o-tol)₃, and K₂CO₃ in acetonitrile at 90–95°C being identified as selectively providing the allylic diol 14 as a single geometric isomer and regioisomer. No products arising from formal isomerization of the allylic alcohol functionality were observed in the crude reaction products. We investigated the scope of this reaction by examining the reactivity of a number of electronically diverse aryl bromides to obtain the allylic diols 15–19 (Table 1). The conditions used in each case were the same as for those identified for 14 and no further optimization was performed. As well as aryl bromides with electron donating and withdrawing substituents being tolerated, heteroaromatic groups could also be introduced (Table 1, entry 5).

Table 1.

Heck reactions of 1

Entry	ee 1, %	Aryl bromide	Product	Yield, %*	ee, %†
1	87	Bromobenzene	14	44	85
2‡	(±)	Bromobenzene	14	58	(±)
3	(±)	3-Bromoanisole	15	49	(±)
4‡	85	4-Bromoanisole	16	35	66
5	(±)	2-Bromothiophene	17	17	(±)
6	(±)	tert-Butyl 3-bromobenzoate	18	45	(±)
7‡	85	4-Bromoacetophenone	19	29	80

^*Obtained after flash column chromatography and based on the diol 1.

^†Determined by HPLC (see Supporting Text).

^‡Excess aryl bromide used (3 eq).

Whereas the arylated allylic diol 14 was obtained in modest isolated yield, the diol 1 was fully consumed and ¹H NMR analysis of the crude reaction indicated that the reaction was very selective. In an effort to understand the cause of this apparent loss of material, control reactions investigating reactant and product stability were carried out. These reactions clearly established that, whereas the product 14 was stable to the reaction conditions, 1 was rapidly decomposed in the absence of the aryl bromide. Reactions carried out in deuterated solvents did not allow for the elucidation of the decomposition pathway of 1 to be determined. Although the level of catalyst loading or ratio of palladium to ligand had minimal influence on the reaction outcome, the isolated yield of 14 could be increased by using an excess of the aryl bromide (Table 1, entry 2). Subsequent reactions using (R)-1 to obtain 16 and 19 were carried out by using excess aryl bromide. This influence of the amount of aryl bromide used in Heck reactions has been previously reported (20).

One possible mechanism for loss of enantiopurity during the Heck reaction requires β-hydride elimination of the hydrogen at the stereogenic center. The electronic nature of the aromatic substituent may have an effect on the lability of the benzylic hydrogen relative to that at the chiral center, and hence on the degree to which racemization occurs. Indeed, deterioration of enantiopurity was markedly observed only when the coupling partner was an electron-rich aryl bromide (Table 1, entry 4), purportedly because of an enhancement of electron density at the benzylic carbon.

These results clearly demonstrate that aryl bromides undergo highly selective Heck reactions with 1 and that for certain classes of coupling partner, these reactions occur with minimal deterioration of enantiopurity. Further studies are needed to develop this reaction with regard to yield improvement and conservation of enantiopurity.

Conclusion

The palladium-catalyzed dynamic kinetic asymmetric transformation of 3,4-epoxy-1-butene, 2, into (2R)-3-butene-1,2-diol, 1, with water as the cosolvent has been developed into a robust and scalable procedure. The protocol that was used to provide the diol 1 was originally reported to provide the cyclic carbonate 7. However, this original report was based on reactions using small quantities of 2 and relatively large amounts of catalyst and solvent. When the reaction was examined under industrially more favorable conditions at larger scale, hydrolysis of the cyclic carbonate 7 occurred as a consequence of the longer reaction times. Examination of the effect of solvent and temperature led to the identification of conditions that permitted the use of 0.025 mol % catalyst to obtain 1 in 84% isolated yield with 85% ee. The diol 1 produced via the cyclic carbonate 7 was not contaminated with oligomers, which were observed when the original conditions reported to provide 1 were examined. Heck reactions of 1 with a diverse range of coupling partners were explored, and these occurred in a highly selective fashion. The enantiopurity of the Heck reaction product depended on the electronic nature of the coupling partner.