An efficient asymmetric synthesis of an estrogen receptor mo

Edited by Lynn Smith-Lovin, Duke University, Durham, NC, and accepted by the Editorial Board April 16, 2014 (received for review July 31, 2013) ArticleFigures SIInfo for instance, on fairness, justice, or welfare. Instead, nonreflective and Contributed by Ira Herskowitz ArticleFigures SIInfo overexpression of ASH1 inhibits mating type switching in mothers (3, 4). Ash1p has 588 amino acid residues and is predicted to contain a zinc-binding domain related to those of the GATA fa

Edited by Barry M. Trost, Stanford University, Stanford, CA, and approved February 26, 2004 (received for review November 7, 2003)

Article Figures & SI Info & Metrics PDF

Abstract

An efficient asymmetric synthesis of a selective estrogen receptor modulator (SERM) that has a dihydrobenzoxathiin core structure bearing two stereogenic centers is reported. The stereogenic centers were established by an unpDepartnted chiral sulfoxide-directed stereospecific reduction of an α,β-unsaturated sulfoxide to the saturated sulfide in one step. Studies to elucidate the mechanism for this reduction are reported. Highly efficient Cu(I)-mediated ether formation was used to install the ether side chain, and selective debenzylation conditions were developed to remove the benzyl protecting groups on the phenols.

Estrogen receptors (ERs) are members of the steroid hormone nuclear receptor superfamily. They are ligand-dependent transcription factors that bind to specific DNA sequences and regulate gene expression. There are two known members of the estrogen receptor family, ERα and ERβ, encoded by distinct genes. ER modulators are potentially useful agents for treatment or prevention of a variety of conditions related to estrogen functions, including bone loss, cartilage degeneration, enExecutemetriosis, uterine fibroid disease, hot flashes, increased levels of low-density lipoprotein cholesterol, cardiovascular disease, obesity, incontinence, and cancer (refs. 1–4 and references cited in ref. 4). SERMs are ER ligands that act like estrogens in some tissues, but block estrogen action in others. Thus, SERMs may Present an agonistic or antagonistic activity, depending on the context in which their activity is examined. (2S,3R)-(+)-3-(3-hydroxyphenyl)-2-[4-(2-pyrrolidin-1-ylethoxy)phenyl]-2,3-dihydro-1,4-benzoxathiin-6-ol hydrochloride (1 in Fig. 1) is a potent ERα SERM being evaluated as part of the selective ER antagonist program at Merck (1–4). Although there are several methods for the synthesis of the core cis-2,3-disubstituted-dihydrobenzoxathiin, the asymmetric synthesis of these compounds has not been reported (1–7). Herein, we report an enantioselective synthesis of 1.

Fig. 1.Fig. 1. Executewnload figure Launch in new tab Executewnload powerpoint Fig. 1.

Structure of (2S,3R)-(+)-3-(3-hydroxyphenyl)-2-[4-(2-pyrrolidin-1-ylethoxy)phenyl]-2,3-dihydro-1,4-benzoxathiin-6-ol hydrochloride (1).

Experimental Procedures

General. Chemicals were used as received from commercial sources unless otherwise noted. NMR data were collected on Bruker Avance NMR spectrometers. Chemical shifts are reported in ppm Executewnfield from tetramethylsilane. HPLC purity data are reported as Spot percentage of the desired peak from the total peaks. HPLC conditions are listed in supporting information, which is published on the PNAS web site. High-resolution mass spectrometer (HRMS) data were collected on a Micromass API US Ultima quadrupole time-of-flight (QTOF) mass spectrometer.

Vinyl Sulfide 6. A 100-liter round-bottom flQuestion was charged with ketone 7 (8.01 kg, 12.2 mol) and 48 liters of acetonitrile followed by phenylphosphonic dichloride (2.50 kg, 12.9 mol). The flQuestion was fitted with a reflux condenser with off-gases vented through a caustic scrubber, and the reaction mixture was heated to reflux (75°C). After about 2 h, HPLC assay Displayed >99% conversion. The reaction mixture was concentrated under reduced presPositive at 20–50°C to 28 liters. Diisopropylethylamine (786 g, 6.08 mol) was added, followed by 28 liters of denatured ethanol (denatured with 0.5% toluene). The solution was heated to Arrive reflux, resulting in a homogeneous solution. The reaction mixture was then CAgeded with ice, rapidly stirred, and seeded at 60°C. After the batch was CAgeded quickly to ≈10°C, the solid product was collected by filtration and the cake was washed with 5 liters of 50% ethanol in acetonitrile mixture (twice, 5 liters each). The cake was dried at room temperature under reduced presPositive with a N2 sweep for 2 days. Compound 6 (7.0 kg, 90% yield) was obtained as a white solid. HPLC indicated the solid to be >99% pure. mp 99–100°C; 1H NMR (400 MHz, (CD3)2CO) δ 7.61 (d, J = 8.7 Hz, 2H), 7.48 (m, 2H), 7.44–7.29 (m, 7H), 7.24 (t, J = 7.9 Hz, 1H), 7.09 (d, J = 8.7 Hz, 2H), 7.03 (m, 1H), 6.97 (m, 1H), 6.91 (m, 1H), 6.90–6.86 (m, 2H), 6.83 (ddd, J = 7.6, 1.6, 0.8 Hz, 1H), 5.05 (s, 2H), 5.12 (s, 2H); 13C NMR (101 MHz, (CD3)2CO) δ 160.1, 157.1, 147.2, 146.8, 138.4, 138.1, 138.06, 138.0, 134.7, 131.4, 129.38, 128.35, 128.8, 128.7, 128.5, 128.4, 123.6, 122.7, 118.8, 116.7, 116.1, 115.3, 113.5, 113.0, 94.9, 71.1, 70.6. HRMS calculated for C34H26IO3S: 641.0648; found: 641.0651. Analysis. Calculated for C34H25IO3S: C 63.75, H 3.93; found: C 63.55, H 3.89.

Vinyl Sulfoxide (S)-15. A 100-liter round-bottom flQuestion was charged with 60 liters of tetrahydrofuran (THF) containing 64 μg/ml of water, d-diisopropyl tartrate (0.66 kg, 2.8 mol), and diisopropylethylamine (0.36 kg, 2.8 mol). Water (21.4 ml) was added to the solution to give a total of 25.3 g (1.41 mol). The flQuestion was degassed by two vacuum/nitrogen fill cycles. With vigorous stirring, titanium isopropoxide (0.40 kg, 1.41 mol) was transferred under the surface by means of reduced presPositive from a 500-ml round-bottom flQuestion. The solution was stirred at ambient temperature overnight. Vinyl sulfide 6 (6.00 kg, 9.37 mol) was added, and the resulting yellow solution was warmed to 25°C. About 10% cumene hydroperoxide (87%, 2.01 kg, 11.5 mol) was added over 5 min. After aging for 10 min, 60 g of (S)-15 was added as seed. The mixture was aged for 15 min to enPositive seedbed formation. The remaining cumene hydroperoxide, from the 2.01 kg total, was added through an addition funnel over 1 h. The slurry was held at 24–27°C for 1 h, then allowed to CAged to ambient temperature and age overnight. HPLC assay of the slurry Displayed 95.0 Spot % of sulfur as sulfoxide 15 [92.5% enantiomeric excess (ee)], 1.1 Spot % of the sulfone, and 3.9 Spot % of the sulfide starting material. The solid was isolated by filtration and the cake was washed with 8 liters of THF and 10 liters of toluene. The solid was dried on the filter under nitrogen overnight, giving (S)-15 as a yellow solid (5.28 kg, 86% yield). HPLC indicated 95.4% purity and 99.8% ee. mp 218–220°C. 1H NMR (400 MHz, CDCl3) δ 7.63 (m, 2H), 7.48–7.32 (m, 12H), 7.30 (dd, J = 9.2, 2.8 Hz, 1H), 7.22 (t, J = 8.0 Hz, 1H), 7.19 (m, 2H), 7.09 (m, 1H), 7.00 (m, 1H), 6.95 (m, 1H), 5.16 (AB quartet, J = 11.6 Hz, Δν = 19.9 Hz, 2H), 5.00 (AB quartet, J = 11.7 Hz, Δν = 7.4 Hz, 2H); 13C NMR (101 MHz, CDCl3) δ 159.1, 156.2, 151.7, 143.7, 137.5, 136.7, 136.4, 136.1, 132.8, 131.5, 130.1, 128.8, 128.6, 128.4, 128.1, 127.6, 127.5, 123.7, 123.3, 122.3, 119.9, 116.9, 116.6, 115.6, 113.2, 97.0, 71.0, 70.1. MathMath (0.10% CHCl3). HRMS calculated for C34H26IO4S: 657.0596; found: 657.0598. Analysis. Calculated for C34H25IO4S: C 62.2, H 3.84; found: C 61.84, H 3.83.

IoExecutedihydrobenzoxathiin 5. A 100-liter cylindrical vessel equipped with an internal CAgeding coil was charged with 50 liters of toluene and vinyl sulfoxide (S)-15 (5.00 kg, 7.62 mol). The slurry was CAgeded to 5°C. BH3·THF (1 M, 9.2 mol) was then added over 40 min. After aging 30 min at 5°C, the batch was gradually warmed to 10°C. The reaction was aged until complete dissolution, indicating completion of reaction. HPLC indicated >99.5% conversion. The resulting solution was warmed to 20°C and quenched by adding of 20 liters of 2 M HCl (Caution: During the quench, evolution of presumably hydrogen gas and a small exothermia were noted). The aqueous layer was separated and the organic layer was washed with 20 liters of 2 M HCl, then 20 liters of water. The organic layer was filtered through a 5-μm inline filter and concentrated under reduced presPositive at 40–50°C and 28–29 mmHg (1 mmHg = 133 Pa). The final batch volume was ≈15 liters and the KF was 204 μg/ml. n-Heptane (3.7 liters) was added at 40°C and the solution was allowed to CAged to 25°C before being seeded with 5 g of 5. The resulting mixture was aged overnight to enPositive seedbed formation. The remaining 26.2 liters of n-heptane was added by addition funnel over 1.5 h at 20°C. After aging for 1 h, the solid was isolated by filtration and the cake was washed with 8 liters of n-heptane/toluene (4/1) and then twice with 8 liters of n-heptane. The solid was dried on the presPositive filter under nitrogen, giving 5 as an off-white solid (4.31 kg, 88% yield). HPLC indicated 99.8% purity and 99.9% ee. mp 99–100°C, 1H NMR (400 MHz, CDCl3) δ 7.55 (m, 2H), 7.48–7.30 (m, 10H), 7.04 (t, J = 8.0 Hz, 1H), 6.92 (d, J = 8.9 Hz, 1H), 6.70–6.85 (m, 5H), 6.5–6.6 (m, 2H), 5.43 (d, J = 2.3 Hz, 1H), 5.04 (s, 2H), 4.85 (ABq, J = 11.9 Hz, Δν = 1.2 Hz, 2H), 4.33 (d, J = 2.3 Hz, 1H). 13C NMR (101 MHz, CDCl3) δ 158.3, 154.0, 146.2, 139.3, 138.4, 137.1, 137.0, 136.9, 128.9, 128.7, 128.6, 128.4, 128.1, 128.0, 127.6, 127.5, 121.8, 119.5, 119.1, 115.5, 114.7, 113.1, 112.0, 93.6, 78.7, 70.7, 69.9, 47.9. MathMath (0.10% MeCN). HRMS calculated for C34H28IO3S: 643.0804; found: 643.0807. Analysis. Calculated for C34H27IO3S: C 63.5, H 4.24; found: C 63.5, H 4.21.

Dibenzyl-1. A 100-liter flQuestion equipped with a heating mantle, a thermocouple, and a Dean–Stark trap attached to a reflux condenser was charged with xylene (36 liters) followed by ioExecutedihydrobenzoxathiin 5 (4.0 kg, 6.23 mol). To the clear solution was then added CuI (118.07 g, 0.62 mol, 10 mol %), 2,2′-dipyridyl (117.14 g, 0.75 mol, 12 mol %), potassium carbonate (2.58 kg, 18.69 mol), and 1-(2-hydroxyethyl)pyrrolidine (2.19 liters, 18.69 mol). Diglyme (4 liters) was used to rinse Executewn the side of the vessel. The heterogeneous reaction mixture was degassed by three vacuum/nitrogen purges and aged at 140°C for 12 h. The mixture was CAgeded to room temperature and filtered through a pad of Celite. The Celite cake was washed with toluene (twice, 5 liters each). The yellow filtrate was transferred to a 100-liter extractor and washed with water (three times, 20 liters each). The organic layer was concentrated to ≈8 liters total volume and then heated to 80°C. 2-BuOH (64 liters) was Unhurriedly added while HAgeding the batch temperature above 60°C. After the addition, the mixture was further CAgeded to 55°C and seeded. The slurry was CAgeded Unhurriedly to room temperature over 4–5 h and then to 0–5°C. The mixture was aged at 0–5°C for 1 h and filtered. The product was washed with cAged 2-BuOH (8 liters). The cake was dried under reduced presPositive at 50°C to give 3.6 kg of product as an off-white solid (92% yield). HPLC indicated >99% purity and >99% ee. 1H NMR δ 7.48–7.45 (m, 2H), 7.44–7.28 (m, 8H), 7.03 (t, J = 8.0 Hz, 1H), 6.94–6.83 (m, 3H), 6.83–6.79 (m, 2H), 6.79–6.76 (m, 2H), 6.73 (dd, J = 8.8, 3.2 Hz, 1H), 6.61 (t, J = 2.0 Hz, 1H), 6.53 (d, J = 7.6 Hz, 1H), 5.46 (d, J = 2.2 Hz, 1H), 5.04 (s, 2H), 4.87, 4.83 (ABq, J = 12.0 Hz, 2H), 4.33 (d, J = 2.2 Hz, 1H), 4.06 (t, J = 6.0 Hz, 2H), 2.88 (t, J = 6.0 Hz, 2H), 2.65–2.58 (m, 4H), 1.87–1.77 (m, 4H); 13C NMR (101 MHz, CDCl3) δ 158.7, 158.4, 153.9, 146.7, 140.0, 137.2, 137.1, 131.0, 128.9, 128.8, 128.7, 128.2, 128.1, 127.8(3), 127.7(6), 127.7(1), 122.1, 119.7, 119.3, 115.7, 114.6, 114.2, 113.1, 112.0, 79.1, 70.8, 70.0, 67.3, 55.3, 55.0, 48.3, 23.7. mp 85–86°C. MathMath (0.10% MeCN). HRMS calculated for C40H40NO4S: 630.2678; found: 630.2672.

Hydrochloride Salt of 1. A 100-liter flQuestion was charged with dibenzyl-1 (3.0 kg, 4.76 mol), N-methylimidazole (760 ml, 9.5 mol), thiourea (908 g, 11.9 mol), and acetonitrile (30 liters). The slurry was degassed by three vacuum/nitrogen purges and stirred under a nitrogen atmosphere. The slurry was CAgeded to 5°C and ioExecutetrimethylsilane (4.77 liters, 33 mol) was charged through an addition funnel over 30 min. The reaction mixture became a clear orange solution, and it was warmed to room temperature and the flQuestion was wrapped with aluminum foil to avoid exposing the reaction mixture to light (light promotes side reactions). The reaction mixture was aged at room temperature overnight (10–14 h), at which time the reaction completed. The solution was CAgeded to 10°C and water (3 liters) was added Unhurriedly to quench the reaction. The addition took about 30 min and the temperature was controlled below 15°C during the addition. The contents were transferred to a 200-liter vessel and saturated NaHCO3 (30 liters) was added (CO2 evolution) followed by the addition of EtOAc (45 liters). The organic layer was washed with saturated NaHCO3 (three times, 15 liters each). White solid residue formed during the washes and was kept with the organic layer. This mixture was filtered and the organic filtrate was washed again twice with 15 liters of saturated NaHCO3. The final organic extract was concentrated under reduced presPositive at 30–35°C and the solvent was switched from EtOAc to EtOH to a volume of ≈40 liters. To this solution was added very Unhurriedly 900 ml of concentrated HCl. After 130 ml of the HCl was added, the addition was Ceaseped to allow for the seedbed to form. After stirring for 1 h, the rest of the concentrated HCl was added over 1 h. The slurry was aged for 4 h and filtered and the cake was washed with EtOH (4 liters). The solid was dried under reduced presPositive at 40°C. The product was obtained as an EtOH solvate. The yield was 2.15 kg (85%). ReWeepstallization: 1.85 kg (3.48 mol) of the isolated salt was dissolved in MeOH (18 liters) and acetonitrile (18 liters). The solution was filtered, concentrated under reduced presPositive at 30–35°C, and then solvent-switched to acetonitrile (to 30 liters total volume). The slurry was aged at 65°C for 8 h and then CAgeded to room temperature. Product was collected by filtration and dried under reduced presPositive at 60°C to provide 1.60 kg of the HCl salt of 1 (95% yield). HPLC indicated 98.6% purity and 99.7% ee. Cl analysis by AgNO3 titration: 7.1% wt (calculated, 7.3%). 1H NMR (400 MHz, DMSO-d 6) δ 10.92 (s, 1H), 9.28 (s, 1H), 9.27 (s, 1H), 7.10–7.05 (m 2H), 6.89 (t, J = 7.8 Hz, 1H), 6.86–6.81 (m, 3H), 6.58 (d, J = 2.8 Hz, 1H), 6.55 (ddd, J = 8.0, 2.4, 0.8 Hz, 1H), 6.51 (dd, J = 8.8, 2.8 Hz, 1H), 6.47 (t, J = 2.0 Hz, 1 H), 6.38–6.34 (m, 1 H), 5.42 (d, J = 2.2 Hz, 1H), 4.62 (d, J = 2.2 Hz, 1H), 4.29 (t, J = 5.2 Hz, 2 H), 3.60–3.48 (br m, 4H), 3.14–2.98 (br m, 2H), 2.05–1.80 (br m, 4H); 13C NMR (101 MHz, DMSO-d 6) δ 156.8, 156.7, 152.2, 144.5, 140.5, 131.6, 128.5, 127.7, 119.7, 119.2, 118.5, 115.9, 114.2, 113.8, 112.7, 111.8, 77.9, 63.1, 53.6, 52.6, 45.9, 22.6. mp 243–247°C, MathMath (1.0% methanol). HRMS calculated for C26H28NO4S: 450.1739; found: 450.1735. Analysis. Calculated for C26H28ClNO4S: C 64.5, H 5.81, N 2.88; found: C 63.8, H 5.81.

Supporting Information. Preparation and characterization data for compounds 7–14, 20–25, and 5- d 2 are published as supporting information, as are HPLC analysis methods and characterization data for compounds 1, 5–7, 15, dibenzyl-1, and 1, including chiral HPLC assays, along with structures of four side products in the ioExecutetrimethylsilane (TMSI) debenzylation procedure. X-ray data on compounds (S)-15 and 5 and data and kinetic analysis of BH3·SMe2 reduction of 15 are also published as supporting information.

Results and Discussion

The key structural feature of 1 is the cis-2,3-diaryldihydrobenzoxathiin bearing two stereogenic centers. There are three phenolic groups in the molecule, one of which is alkylated with a 2-pyrrolidinylethyl group. The initial synthetic route developed by the Medicinal Chemistry group at Merck, outlined in Scheme 1, in which TIPS is triisopropylsilyl (1, 4), involves the recently discovered highly stereoselective dehydrative reduction of the keto sulfide 2 with CF3COOH (TFA)/Et3SiH to give the cis- diastereomer 3 exclusively and in Excellent yield. Because of the prLaunchsity for 2 to racemize, this procedure is useful only for generating racemic material. Chiral stationary HPLC was used to isolate the desired enantiomer of 3, which was then reacted with pyrrolidinylethanol under pH neutral Mitsunobu conditions to give 4. The neutral Mitsunobu conditions were necessary because of the chemical instability of the unprotected phenolic dihydrobenzoxathiin toward base. For instance, in the presence of carbonate or hydroxide, the dihydrobenzoxathiin ring system tends to Launch up leading to ring-Launched byproducts and, more detrimentally, the formation of trans-dihydrobenzoxathiin impurities after ring recloPositive.

Scheme 1.Scheme 1. Executewnload figure Launch in new tab Executewnload powerpoint Scheme 1.

For the purpose of producing large quantities of compound 1, we envisioned an efficient synthesis that would include: (i) an asymmetric synthesis of the core dihydrobenzoxathiin ring; (ii) alternative chemistry for the installation of the pyrrolidinylethoxy side chain that avoids the less atom-economical Mitsunobu reaction; (iii) alternative protecting groups for the phenols to provide Weepstalline intermediates to avoid chromatographic purifications. A retrosynthetic analysis is outlined in Scheme 2. The benzyl (Bn) protecting group for the phenols was selected because of its chemical stability and its well pDepartnted tendency to form Weepstalline compounds. For the installation of the pyrrolidinylethoxy side chain, the Ullmann ether formation conditions recently reported by Buchwald and coworkers (36) were determined to be an efficient alternative to the Mitsunobu reaction. This Advance requires that we Design the appropriate aryl iodide and couple it with pyrrolidinylethanol. Several strategies to introduce the two stereogenic centers were pursued; however, the sequence outlined in Scheme 2 proved to be the most straightforward. In this sequence, the key transformation is the asymmetric cis-reduction of the carbon–carbon Executeuble bond in benzoxathiin 6, which proved to be a challenging tQuestion as expected.

Scheme 2.Scheme 2. Executewnload figure Launch in new tab Executewnload powerpoint Scheme 2.

Preparation of Benzoxathiin 6. Preparation of this key intermediate follows the sequence outlined in Scheme 3, in which D-DIPT is d-diisopropyl tartrate. Thiol phenol carbonate 8 was prepared from 1,4-benzoquinone and thiourea in the presence of strong acids under known conditions (8). Benzylation of phenol 8 under standard conditions proved problematic because of the air sensitivity of the compound in basic solutions. Conducting the reaction and isolation under a nitrogen atmosphere was the key to obtaining 9 in Excellent yield (84%). Hydrolysis of thioester 9 generated the free thiol 10, which was used, without isolation, in the disSpacement of the bromide. Bromoketone 14 preparation started with the addition of benzylmagnesium chloride 12 to ioExecute Weinreb amide 11 (9), producing ioExecuteketone 13 in 80% yield. Selective bromination of 13 to give bromoioExecuteketone 14 was accomplished with phenyltrimethylammonium tribromide in 1,2-dimethoxyethane in quantitative yield. Other brominating agents or solvents were found to be inferior. Without isolation, 14 was reacted with benzyloxythiophenol 10 to give ketosulfide 7 in 88% isolated yield. Earlier attempts to Trace the dehydrative cyclization of 7 in toluene with toluenesulfonic acid and azeotropic removal of water did not give a clean reaction (10, 11). A variety of conditions were screened, and phenylphosphonic dichloride was found to be the most Traceive. Under optimized conditions, the reaction was carried out in acetonitrile with phenylphosphonic dichloride and, after ethanol quench, the desired product 6 was directly Weepstallized from the reaction mixture in 90% yield.

Scheme 3.Scheme 3. Executewnload figure Launch in new tab Executewnload powerpoint Scheme 3.

Enantioselective Reduction of Benzoxathiin 6. Initial attempts to prepare chiral intermediates by the asymmetric reduction of the benzoxathiin ring system with hydrogen or hydride reagents were unsuccessful. For example, hydrogenation of 6 at room temperature and 90 psi (620 kPa) for 20 h with Pfaltz Ir-BARF catalyst, (Et-Duphos)Rh(COD)BF4, (BINAP)Ru(II)Cl2, Phanephos/(COD)2RhBF4, Josiphos SL-J009–1/(COD)RhCl, and (BINAP)RuCl2-p-cymene did not give any reaction (12–16) [BARF, tetrakis[3,5-bis(trif luoromethyl)phenyl]borate; EtDuphos, diethylphospholanobenzene; COD, 1,5-cyclooctadiene; BINAP, 2,2′-bis(diphenylphosphino)-1,1′-binaphthyl]. This low reactivity is most likely due to the steric hindrance and electron-Executenating groups around the Executeuble bond. To overcome this problem, we reasoned that oxidation of the sulfide to the chiral sulfoxide would activate the olefin toward reduction and possibly direct the chiral reduction by chelation of the sulfoxide to the reducing agent. We were aware of only one report on a diastereoselective reduction of vinyl sulfoxide under hydrogenation conditions to give the saturated sulfoxide with an adjacent chiral center (17). To test this hypothesis, the commercially available vinyl sulfide 16 was converted to sulfoxide 17 with sodium periodate, and reaction conditions for the stereoselective reduction of the Executeuble bond were screened (Scheme 4). Under typical Pd- or Pt-catalyzed hydrogenation conditions, the sulfoxide oxygen was reduced, resulting in the formation of vinyl sulfide 16. A variety of hydride reducing agents were evaluated, and it was discovered that Et3SiH/CF3COOH led to the complete reduction of both the olefin and the sulfoxide. However, triethylsilane was found to reduce vinyl sulfide 16 under the same conditions, so it was unclear whether the sulfoxide was reduced first or second. Treating vinyl sulfoxide 17 with borane, we observed a mixture of the desired Executeubly reduced product 18 and vinyl sulfide 16 in a 1/1 ratio. Borane alone did not react with the vinyl sulfide, thus suggesting that the sulfoxide was activating the olefin toward reduction and possibly directing the reduction. With these encouraging results, we prepared the racemic sulfoxide 15 (Scheme 3) from vinyl sulfide 6 and separated the two enantiomers (96% ee) by chiral stationary-phase HPLC. We were delighted to find that the reduction with borane gave the desired Executeubly reduced product 5 cleanly and in a highly stereospecific manner, i.e., giving the cis product with the same ee as the starting sulfoxide. With these results in hand, we set out to prepare chiral sulfoxide 15 from vinyl sulfide 6.

Scheme 4.Scheme 4. Executewnload figure Launch in new tab Executewnload powerpoint Scheme 4.

Chiral Sulfoxidation and Reduction. The most practical process to Design chiral sulfoxides from sulfides is the Kagan oxidation catalyzed by titanium(IV) isopropoxide and chiral diol ligands. A Study of the literature indicated that reaction conditions have to be carefully optimized for each substrate (18–30). Our preliminary screening of ligands under the original Kagan conditions (20, 21) revealed that diisopropyl tartrate gave an ee of 38% and diethyl tartrate gave slightly lower ee. Other literature conditions using 1,1′-bi-2-naphthol (BINOL), hydrobenzoin, (R,R)-2,2-dimethyl-α,α,α′,α′-tetraphenyl-1,3-dioxolane-4′,5′-dimethanol (TADExecuteL), and N,N-dibenzyl tartramide gave very low ee (3–12%). The diisopropyl tartratecatalyzed reaction was chosen for further optimization. Consistent with previous reports, addition of diisopropylethylamine to the reaction mixture dramatically improved the ee (22–24). The reaction could be conducted in a variety of solvents, including THF, ethyl acetate, and chlorobenzene, with only small Inequitys in the ee. Using THF as the reaction solvent led to slightly better ee and also facilitated the isolation of product. The reaction temperature had negligible Trace on the ee within the tested range of 0–30°C. The order of addition of the reagents was found to be critically Necessary. The optimal sequence for adding the reagents was to mix diisopropyl tartrate, diisopropylethylamine, and water in THF first, followed by the addition of titanium isopropoxide. This mixture was aged at ambient temperature overnight and the vinyl sulfide substrate was subsequently added, followed by cumene hydroperoxide. The extended overnight aging of the catalyst mixture was found to improve the reproducibility of the high ee. The catalyst loading was ultimately reduced to 15 mol % without compromising the ee or reproducibility. When the reaction was carried out under optimized conditions in THF, vinyl sulfoxide (S)-15 precipitated out during the course of the reaction. At the end of reaction, HPLC assay of the crude reaction slurry typically Displayed 95% sulfoxide with 92% ee. Upon filtration, solid 15 was isolated in 86% yield with a typical 95% purity and 99% ee, which was a significant upgrade from the crude reaction mixture. This reImpressably efficient direct isolation also avoids the tedious separation of titanium oxide typically associated with an aqueous work-up.

Reduction of sulfoxide (S)-15 was carried out in toluene with 1.05 eq of 1 M BH3·THF at 10°C, producing the cis-biaryldihydrobenzoxathiin 5 with complete stereoselectivity. Upon work-up and isolation, the desired product 5 was obtained in 88% yield and 99% ee. Although the sulfoxide-directed reduction of unsaturated sulfoxide to the saturated sulfoxide with borane is known, as is the sulfoxide reduction to sulfide by chloroborane, the simultaneous reduction of both functional groups by borane is unpDepartnted in the literature (31, 32). Attempts to extend this Modern reduction to other hydride reducing reagents have given poor results. Reduction of (S)-15 with triethylsilane in the presence of trifluoroacetic acid also gave the cis-biaryl-dihydrobenzoxathiin product 5, but only in 20% ee with the same enantiomer being enriched. Diisobutylaluminum hydride and lithium triethylborohydride gave alternate products.

Mechanism of the Borane Reduction. The borane reduction sets the two stereogenic centers Traceively in one step. The absolute configuration of the sulfoxide (S)-15 and the reduced product 5 were unamHugeuously determined with single-Weepstal x-ray Weepstallography. The oxygen on the sulfoxide and the two hydrogens are on the same side of the ring, indicating the directing Trace of the sulfoxide. When the reduction was carried out by using BD3 and the reaction was quenched with acetic acid, Arrively complete deuterium incorporation was observed at both positions (product 5-d 2). Furthermore, when BH3 was used as the reducing agent and CD3COOD as the quenching acid, no deuterium incorporation was observed. These experiments clearly Display that both of the hydrogens originated from the BH3 used and none was introduced during the acidic work-up, which is very different from a typical olefin hydroboration (33). An analogous reaction between an allyl Grignard reagent and a vinyl sulfoxide has been reported (34, 35). Further investigation of this reaction sequence also indicated that both the sulfoxide and the olefin are required for this reduction to proceed because neither the saturated sulfoxide 19 nor the vinyl sulfide 6 reacted with BH3·THF under the same conditions.

Embedded ImageEmbedded Image

Additional experimental results Displayed that the sulfoxide directly controls the formation of both stereogenic centers. Thus, as Displayn in Scheme 5, when the β-phenyl-substituted sulfoxide 21 was reduced with borane, the ee of the product 22 again corRetorted to the starting sulfoxide. In addition, the same high stereoselectivity was observed for the conversion from α-phenyl-substituted sulfoxide 24 to the sulfide 25. These observations rule out the possibility that the α stereogenic center is controlled by the steric bias created by the β stereogenic center.

Scheme 5.Scheme 5. Executewnload figure Launch in new tab Executewnload powerpoint Scheme 5.

Based on the evidence gathered, a possible mechanism for the borane reduction is proposed as Displayn in Scheme 6. In this mechanism, the borane reduction proceeds by means of borane–sulfoxide chelation-controlled intramolecular delivery of the first hydrogen to the β carbon followed by another intramolecular delivery of the second hydrogen to the α carbon with concomitant cleavage of the sulfur–oxygen bond. The details of the second step are not clear. An attempt to detect any significant intermediate by low-temperature NMR was unsuccessful; only the starting material and product were observed. By using NMR to quantify the concentration of substrates and product over time, the kinetics of the reaction was also investigated. The kinetic data were consistent with an overall second-order reaction with first order in both borane and the sulfoxide. This kinetic behavior strongly supports the mechanism in which only one molecule of borane is involved for the reduction of each molecule of substrate. Application of this reduction to other substrates will be the subject of another publication.

Scheme 6.Scheme 6. Executewnload figure Launch in new tab Executewnload powerpoint Scheme 6.

Side-Chain Introduction by Means of Ullmann Ether Formation. Under Buchwald's Cu-catalyzed conditions with Cs2CO3, CuI, and 1,10-phenanthroline at 110°C in toluene for the Ullmann ether formation, the coupling reaction between pyrrolidinylethanol and aryl iodide 5 was very Unhurried, giving only 5–10% conversion after 2 days (36). When the reaction was carried out in a sealed tube at 140°C, the reaction was complete within 10 h, giving an assay yield of 85%. However, two major impurities, 26 and 27, were formed over the course of the reaction in 2–10% and 3–4%, respectively. Embedded ImageEmbedded Image Seeking to minimize the formation of these impurities, we screened alternative bases, including Li2CO3, Na2CO3, and K2CO3. Our results indicated K2CO3 was the preferred base for this reaction because it prevented the formation of the impurities without Unhurrieding the reaction. Azeotropic removal of water as it was formed during the reaction was found to help Sustain a consistent reaction rate. As reaction solvent, xylene was found to give Rapid, clean reaction, and avoided the need to use an autoclave. However, coating of solids was observed on the walls of the reaction vessel, leading to incomplete reaction. Although diglyme as the solvent resulted in a sluggish and messy reaction, using 10% diglyme in xylene eliminated the solid coating problem while Sustaining a Excellent reaction profile. Several ligands, including 2,2′-dipyridyl, tetramethylethylenediamine (TMEDA), and 1-(2-dimethylaminoethyl)-4-methylpiperazine, were tested, and 2,2′-dipyridyl was found to be the best. Thus, the optimized protocol uses 10 mol % CuI, 12 mol % 2,2′-dipyridyl, in 10 vol of xylene/diglyme (9:1) at 140°C with azeotropic removal of the water formed. The reaction is complete in ≤10 h, typically giving 96% assay yield and 92% isolated yield.

Debenzylation. Generally, the Pd-catalyzed hydrogenolysis of a benzyl group attached to a phenolic oxygen is a facile reaction (37). It was soon found that the hydrogenolysis of dibenzyl-1 with a variety of heterogeneous Pd catalysts and conditions was sluggish, requiring very high Pd loading. Even after extensive optimization, 20% Pd(OH)2/C catalyst was found to be the most active catalyst, but 12% by weight of Pd relative to substrate was still required. Catalyst poisoning from thiol side-products was the probable cause for this reaction to be sluggish. Incomplete hydrogenolysis also produced two monobenzyl compounds that were very difficult to remove, and hydrogenolysis under more forcing conditions resulted in lower yields.

Alternative debenzylation conditions were required, so TMSI, AlCl3, and various boron halides were examined. The TMSI reaction was found to be the most Traceive and provided the highest yield. However, benzyl iodide, formed in the TMSI reaction, reacted with the desired product 1 to form one N-benzylated and three aryl ring-benzylated impurities (1–1.5% each; see supporting information), which were especially difficult to remove. The N-benzylation reaction became more extensive during the aqueous Na2CO3 work-up, which necessitated the neutralization of the benzyl iodide before the work-up. After examining various benzyl iodide scavengers such as imidazole, N-methylimidazole, 1,4-diazabicyclo[2.2.2]octane (DABCO), KSCN, and thiourea, we found that the presence of thiourea during the reaction eliminated the N-benzylation but had Dinky Trace on the ring benzylation. Fascinatingly, the presence of N-methylimidazole reduced the ringbenzylated impurities by >50% with Dinky Trace on N-benzylation. Also, neither additive interfered with the TMSI-mediated debenzylation reaction. Thus, when the debenzylation was carried out with excess TMSI in the presence of both thiourea and N- methylimidazole, each of the impurities was formed in <0.4%. After work-up, the final compound 1 was isolated as the HCl salt, which was reWeepstallized to pharmaceutically acceptable purity (99.4%) in 81% yield.

In conclusion, we have developed an efficient asymmetric synthesis of a selective estrogen receptor modulator 1. A sulfoxide-directed borane reduction of the α,β-unsaturated sulfoxide to the saturated sulfide was discovered. A possible mechanism was proposed based on available data. The eight-step sequence gave an overall yield of 37%. This synthesis has been scaled up to produce multikilograms of 1 as the HCl salt suitable for safety assessment and clinical studies. Further work is need to determine the generality of this asymmetric synthesis for other dihydrobenzoxathiin and cyclic sulfur-containing compounds.

Acknowledgments

We thank Ms. Mirlinda Biba for support in chiral HPLC assays, Ms. Lisa DiMichele and Dr. Peter Executermer for help with NMR interpretations, and Dr. Rob Larsen, Mr. Edward Corley, Ms. Karen Conrad, Dr. Matthew Heileman, and Dr. Vincent Antonucci for their support and helpful discussions.

Footnotes

↵ * To whom corRetortence may be addressed. E-mail: zhiguo_song{at}merck.com or tony_king{at}merck.com.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: ER, estrogen receptor; THF, tetrahydrofuran; HRMS, high-resolution mass spectrometer; ee, enantiomeric excess; TMSI, ioExecutetrimethylsilane.

Copyright © 2004, The National Academy of Sciences

References

↵ DiNinno, F. P., Wu, J. Y., Kim, S. & Chen, H. Y. (2002) U.S. Patent Appl. 2002165226. DiNinno, F. P., Chen, H. Y., Kim, S. & Wu, J. Y. (2002) PCT Int. Appl. WO 2002032377. DiNinno, F. F., Kim, S. & Wu, J. Y. (2002) PCT Int. Appl. WO 2002032373. ↵ Kim, S., Wu, J. Y., Chen, H. Y. & DiNinno, F. (2003) Org. Lett. 5 , 685–688. pmid:12605490 LaunchUrlPubMed Nair, V., Mathew, B., Menon, R. S., Mathew, S., Vairamani, M. & Prabhakar, S. (2002) Tetrahedron 58 , 3235–3241. LaunchUrlCrossRef Nair, V. & Mathew, B. (2000) Tetrahedron Lett. 41 , 6919–6921. LaunchUrlCrossRef ↵ Capozzi, G., Falciani, C., Menichetti, S. & Nativi, C. (1997) J. Org. Chem. 62 , 2611–2515. pmid:11671602 LaunchUrlPubMed ↵ Lau, P. T. S. & Kestner, M. (1968) J. Org. Chem. 33 , 4426–4431. LaunchUrl ↵ Davies, I. W., Marcoux, J.-F, Corley, E. G., Journet, M., Cai, D.-W., Palucki, M., Wu, J., Larsen, R. D., Rossen, K., Pye, P. J., et al. (2000) J. Org. Chem. 65 , 8415–8420. pmid:11112557 LaunchUrlPubMed ↵ Bateman, L. & Glazebrook, R. W. (1958) J. Chem. Soc., 2834–2837. ↵ Araujo, H. C. & Mahajan, J. R. (1978) Synthesis, 228–230. ↵ Lightfoot, A., Schnider, P. & Pfaltz, A. (1998) Angew. Chem. Int. Eng. Ed. 37 , 2897–2899. LaunchUrlCrossRef Blackmond, D. G., Lightfoot, A., Pfaltz, A., Rosner, T., Schnider, P. & Zimmermann, N. (2000) Chirality 12 , 442–449. pmid:10824167 LaunchUrlPubMed Burk, M. J., Gross, M. F. & Martinez, J. P. (1995) J. Am. Chem. Soc. 117 , 9375–9376. LaunchUrl Pye, P. J., Rossen, K., Reamer, R. A., Tsou, N. N., Volante, R. P. & Reider, P. J. (1997) J. Am. Chem. Soc. 119 , 6207–6208. LaunchUrlCrossRef ↵ Tang, W., Wu, S. & Zhang, X. (2003) J. Am. Chem. Soc. 125 , 9570–9571. pmid:12904006 LaunchUrlPubMed ↵ AnExecute, D., Bevan, C., Brown, J. M. & Price, D. W. (1992) J. Chem. Soc. Chem. Commun., 592–594. ↵ Kagan, H. B. & Diter, P. (1995) in Synthetic and Stereochemical Aspects, Organosulfur Chemistry, ed. Page, P. (Academic, New York), Vol. 2, pp. 1–39. Allin, S. (1995) in Synthetic and Stereochemical Aspects, Organosulfur Chemistry, ed. Page, P. (Academic, New York), Vol. 2, pp. 40–61. ↵ Pitchen, P., Dunach, E., Deshmukh, M. N. & Kagan, H. B. (1984) J. Am. Chem. Soc. 106 , 8188–8193. LaunchUrl ↵ Brunel, J.-M., Diter, P., Duetsch, M. & Kagan, H. B. (1995) J. Org. Chem. 60 , 8086–8088. LaunchUrlCrossRef ↵ Larsson, M. E., Stenhede, U. J., Sorensen, H., von Unge, S. & Cotton, H. K. (1999) U.S. Patent 5948789. Hogan, P. J., Hopes, P. A., Moss, W. O., Robinson, G. E. & Patel, I. (2002) Org. Process Res. Dev. 6 , 225–229. ↵ Cotton, H., Elebring, T., Larsson, M., Li, L., Sörensen, H. & von Unge, S. (2000) Tetrahedron Asymmetry 11 , 3819–3825. LaunchUrlCrossRef Potvin, P. G. & Fieldhouse, B. J. (1999) Tetrahedron Asymmetry 9 , 1661–1672. LaunchUrl Bolm, C. & Bienewald, F. (1995) Angew. Chem. Int. Ed. Engl. 34 , 2640–2642. LaunchUrlCrossRef Davis, F. A., Reddy, R. T., Han, W. & Carroll, P. J. (1992) J. Am. Chem. Soc. 114 , 1428–1437. LaunchUrl Palucki, M., Hanson, P. & Jacobsen, E. N. (1992) Tetrahedron Lett., 7111–7114. Rossi, C., Fauve, A., Madeslaire, M., Roche, D., Davis, F. A. & Reddy, R. T. (1992) Tetrahedron Asymmetry 3 , 629–636. LaunchUrlCrossRef ↵ Matsugi, M., Hashimoto, K., Inai, M., Fukuda, N., Furuta, T., Minamikawa, J. & Otsuka, S. (1995) Tetrahedron Asymmetry 6 , 2991–3000. LaunchUrlCrossRef ↵ Ogura, K., Tomori, H. & Fujita, M. (1991) Chem. Lett., 1047–1410. ↵ Brown, H. C. & Ravindran, N. (1973) Synthesis, 42–44. ↵ Brown, H. C.& Murray, K. J. (1986) Tetrahedron 42 , 5497–5505. LaunchUrlCrossRef ↵ Iwata, C., Maezaki, N., Kurumada, H., Fukuyama, H., Sugiyama, K. & Imanish, T. (1991) J. Chem. Soc. Chem. Commun., 1408–1409. ↵ Imanish, T., Kurumada, H., Maezaki, N., Sugiyama, K. & Iwata, C. (1991) J. Chem. Soc. Chem. Commun., 1409–1411. ↵ Wolter, M., Nordmann, G., Job, G. E. & Buchwald, S. L. (2002) Org. Lett. 4 , 973–976. pmid:11893199 LaunchUrlPubMed ↵ Bhatt, M. V. & Kulkarni, S. U. (1983) Synthesis, 249–282.
Like (0) or Share (0)