The remarkable influence of steroid A/B-ring junction on the Wittig olefination reaction of the 11-oxo group: Towards the synthesis of 5α- and 5β-oriented Δ3-isomers of desogestrel

Article abstract of DOI:10.1016/S0039-128X(97)00105-0

The 5α- and 5β-oriented Δ³-double bond isomers 8, 9 of the widely used progestin desogestrel (7) were synthesized. Wittig olefination reaction of the 5α-intermediate 12 showed a dramatically reduced reaction rate compared with the olefination of the 5β-intermediate 13. Computational studies suggest that different energies of the intermediary 1,2- oxaphosphetanes may, at least partially, have been the reason for this phenomenon.

Full text of DOI:10.1016/S0039-128X(97)00105-0

The remarkable influence of steroid
A/B-ring junction on the Wittig
olefination reaction of the 11-oxo
group: Towards the synthesis of
5␣- and 5␤-oriented ⌬3-isomers
of desogestrel
Sven Ring,* Gisela Weber,* Alexander Hillisch,‡ and Sigfrid Schwarz*
*Division of Research and Development, Jenapharm GmbH & Co.KG; and ‡Institute of Molecular
Biotechnology, Jena, Germany
The 5␣- and 5␤-oriented ⌬3-double bond isomers 8, 9 of the widely used progestin desogestrel (7) were
synthesized. Wittig olefination reaction of the 5␣-intermediate 12 showed a dramatically reduced reaction rate
compared with the olefination of the 5␤-intermediate 13. Computational studies suggest that different energies
of the intermediary 1,2-oxaphosphetanes may, at least partially, have been the reason for this phenomenon.
(Steroids 63:21–27, 1998) © 1998 by Elsevier Science Inc.
Keywords: desogestrel isomers; deoxygenation; Wittig olefination; computational studies
markedly by their reaction rates from those of the 5␤-series.
In particular, rate differences were observed when the iso-
mers 12 and 13 were olefinated by the Wittig reaction to
yield the methylene derivatives 14 and 15. Herein we report
our results in this area.
Introduction
Recently, we described a novel total synthetic approach to
desogestrel (7),1 a powerful progestin widely used in oral
contraception.2 The procedure which run on a semi-
technical scale, involved a reductive deoxygenation of the
enone 1 to give compound 4 via the allylic methylether 2 or
the thioketal 3 (Scheme 1). A shortcoming of this reaction
was that, strongly dependent on various reaction parame-
ters, the ⌬3 - double bond isomers 5 and 6 could be formed
as by-products in 3% to more than 15% yield.1 Thus,
contamination, even in traces, of desogestrel (7) by the
isomers 8, 9 must be strictly precluded by highly efficient
purification processes in the course of the synthesis. For the
sake of in-process control and quality assurance of the final
product 7, detection of ⌬3 - double bond isomers with a
definitively established 5␣- or 5␤-configuration became
essential for all steps following the deoxygenation reaction.
These requisites prompted us to separately synthesize the
desogestrel isomers 8 and 9 via the 5␣- and 5␤-oriented
3,4-unsaturated steroids 5 and 6. During our investigations,
we found that selected compounds of the 5␣-series differed
Experimental
General methods
Solvents and chemicals (Merck-Darmstadt, Fluka, Aldrich) were
reagent grade and were used without further purification. Chro-
matography means flash chromatography which was performed on
Kieselgel 60 (Merck-Darmstadt, 0.04–0.063 mm). 1H nmr spectra
were recorded on a Varian 300. Deuteriochloroform was used as
solvent, chemical shifts were reported as ␦ - values in ppm down-
field from tetramethylsilane as internal standard and J were given
in Hz. The spectral data were recorded as ␦ (coupling pattern, J,
proton number). Melting points were measured with a Boetius
equipment. Optical rotations were taken with a Polamat A (Carl
Zeiss Jena), solvent chloroform, c ϭ 1 g/100 mL, t ϭ ϩ20°C.
Sonification was performed by a Bandelin “Sonorex Super” DK
514 B at a frequency of 35 kHz.
Computational studies were performed using the semiempirical
program VAMP 6.03 and the ab initio program Gaussian 94.4 The
calculations were carried out on R8000 and R10000 processors on
parallel SiliconGraphics Challenge and PowerChallenge comput-
ers. Models were built with the program SYBYL.5 All molecules
were optimized with the semiempirical PM36 method in the
Address reprint requests to Sigfrid Schwarz, Division of Research and
Development, Jenapharm GmbH & Co.KG, Otto-Schott-Strasse 15, D-
07745 Jena, Germany. E-mail: Sigfrid.Schwarz@Schering.de
Received May 23, 1997; accepted August 29, 1997.
Steroids 63:21–27, 1998
© 1998 by Elsevier Science Inc. All rights reserved.
655 Avenue of the Americas, New York, NY 10010
0039-128X/98/$19.00
PII S0039-128X(97)00105-0

Papers
18a-Homo-5␣-estr-3-ene-11␣, 17␤-diol (5) and 18a-
homo-5␤-estr-3-ene-11␣, 17␤-diol (6)
Enone 1 (100 g; 328 mmol) was dissolved in acetic acid (7.5 L).
Zinc dust (3 kg) was added and the suspension was ultrasonically
irradiated for 1 h at 25°C. Work-up by extracting the mixture with
toluene followed by rotary evaporation of the solution gave a
mixture of compounds 5 and 6 in a ratio of nearly 1Ϻ2. Extensive
column chromatography (eluent: toluene/ethyl acetate 3Ϻ1 v/v)
followed by crystallization from ethyl acetate yielded the title
compounds 5 (23 g; 24%) and 6 (43 g; 45%) in a 96–99% purity
20
(by gc). 5: mp. 154–159°C. [␣]D ϭ Ϫ45°. 1H nmr: 5.62 (m,
H-3), 5.39 (dd, 9.9, 2.1, H-4), 3.75 (t, 8.3, H-17), 3.69 (td, 10.2,
5.0, H-11), 1.06 (t, 7.5, H-18a). ms: m/z 290.22311 (Mϩ).
C19H30O2 (290.45) calculated C 78.57 H 10.41 found C 78.52 H
10.41. 6: mp. 150–154°C. [␣]D20 ϭ Ϫ23°. 1H nmr: 5.68 (m, H-3),
5.36 (d, 10.0, H-4), 3.75 (t, 8.6, H-17), 3.58 (m, H-11), 1.05 (t, 7.8,
H-18a). ms: m/z 272.21438 (Mϩ-H2O. C19H30O2 (290.45) calcu-
lated C 78.57 H 10.41 found C 78.57 H 10.40.
18a-Homo-5␣-estr-3-ene-11,17-dione (10)
To a solution of dihydroxy steroid 5 (20 g; 69 mmol) in triethyl-
amine (126 mL; 904 mmol) and absolute dimethyl sulfoxide (30
mL; 550 mmol) was added dropwise a suspension of pyridine-
sulfur trioxide complex (60 g; 376 mmol) in absolute dimethyl
sulfoxide (70 mL; 814 mmol) at 25°C. The mixture was stirred for
6 h at 25°C and then diluted with toluene (300 mL) and water (200
mL). The organic phase was separated and the aqueous solution
was twice extracted with toluene (100 mL each). Rotary evapora-
tion of the combined extracts and crystallization of the resulting
product from methanol gave 14 g of the title compound 10 (70%
20
1
yield). 10: mp. 180–184°C. [␣]D ϭ ϩ204°. H nmr: 5.64 (m,
H-3), 5.43 (d, 9.9, H-4), 2.67 (d, 11.9, H-12), 2.51 (dd, 19.3, 9.3,
H-16), 2.28 (d, 11.9, H-12), 2.27 (dd, 19.3, 9.3, H-16), 0.81 (t, 7.5,
H-18a). ms: m/z 286.19409 (Mϩ). C19H26O2 (286.42) calculated
C 79.68 H 9.15 found C 79.67 H 9.17.
Scheme 1 Formation of 3-double bond isomers in the course
of the desogestrel synthesis.
18a-Homo-5␤-estr-3-ene-11,17-dione (11)
VAMP program. The Eigenvector Following optimizer was used
for the energy minimizations. Molecules 12, 13 and the corre-
sponding enolate anions: The rotatable bond of the ethyl substitu-
ent in position 13 of molecules 12 and 13 was varied in 60° steps
to locate the global energy minimum for these structures. All
resulting conformers were semiempirically energy optimized. The
conformers with the lowest energies were subjected to a Hartree-
Fock minimization, applying the 6-31G* basis set. Subsets of
molecules 12 and 13 consisting of rings A, B, and C plus the ethyl
group at position 13 were also energy minimized, to be comparable
to the oxaphosphetane intermediates. The single point energies of
the ab initio optimized molecules 12, 13, and their corresponding
enolate anions in DMSO were calculated using the Self-Consistent
Isodensity Surface Model within the SCRF method.4 1,2-
Oxaphosphetane intermediates 18 and 19: The three rotatable
bonds between the phosphorus atom and the phenyl substituents
were subjected to a potential energy surface scan. Each torsion
angle was varied in steps of 60°. All possible combinations were
generated. The resulting conformations were energy minimized
using the PM3 Hamiltonian. The conformers with the lowest
potential energy were further optimized using the 6-31G* basis set
and the Hartree–Fock method. The ring D and the corresponding
substituents in 18 and 19 were omitted to enable calculations with
larger basis sets. Each energy minimization of the oxaphosphetane
intermediate ABC-ring-subsets 18 and 19 took 18 days of CPU
time on one R10000 processor.
This compound was obtained analogously. Yield: 14 g 11 (70%).
11: mp. 129–135°C. [␣]D20 ϭ ϩ161°. 1H nmr: 5.67 (m, H-3), 5.36
(d, 10.1, H-4), 2.67 (d, 12.1, H-12), 2.28 (d, 12.1, H-12), 0.81 (t,
7.1, H-18a). ms: m/z 286.19241 (Mϩ). C19H26O2 (286.42) calcu-
lated C 79.68 H 9.15 found C 79.83 H 9.28.
17,17-(2,2-Dimethyl)propanedioxy-18a-homo-5␣-
estr-3-en-11-one (12)
A
mixture of dioxo steroid 10 (20 g; 70 mmol), (2,2-
dimethyl)propane-1,3-diol (40 g; 384 mmol), triethylorthoformate
(40 mL; 240 mmol), and p-toluenesulfonic acid monohydrate (1.2
g; 6 mmol) was stirred for 10 h at 40°C. After dilution with toluene
(100 mL), the reaction mixture was subsequently washed with
saturated aqueous sodium hydrogen carbonate solution (100 mL)
and 5 portions of water (50 mL each). The organic phase was dried
and rotary evaporated to give ketal 12 which was purified by
column chromatography (eluent: toluene) and by crystallization
from methanol. Chromatographically recovered dioxo steroid 10
was recycled. Yield in total: 20.3 g 12 (78%). 12: mp. 167–171°C.
20
1
[␣]D ϭ ϩ112°. H nmr: 5.62 (m, H-3), 5.41 (d, 9.8, H-4), 3.63
(d, 11.0, -OCH2C-), 3.36 (q, 10.5, -OCH2C-), 2.78 (d, 11.6, H-12),
2.51 (d, 11.6, H-12), 1.09 (s, -CH3), 1.01 (t, 7.2, H-18a), 0.72 (s,
-CH3). ms: m/z 372.26779 (Mϩ). C24H36O3 (372.55) calculated C
77.38 H 9.74 found C 77.37 H 9.85.
22 Steroids, 1998, vol. 63, January

⌬3-isomers of desogestrel: Ring et al.
water and evaporated to give the title compound 16 which was
purified by chromatography (eluent: toluene: ethyl acetate 10Ϻ1
v/v) and crystallization from methanol. Yield: 10.07 g 16 (66%).
16: mp. 94–98°C. [␣]D ϭ ϩ158°. H nmr: 5.65 (m, H-3), 5.48
(dq, 9.9, 1.9, H-4), 4.89 (s, ϭCH2), 4.74 (s, ϭCH2), 2.57 (d, 12.8,
H-12), 1.84 (d, 12.8, H-12), 0.76 (t, 7.7, H-18a). ms: m/z
284.21384 (Mϩ). C20H28O (284.45) calculated C 84.45 H 9.92
found C 84.50 H 9.94.
17,17-(2,2-Dimethyl)propanedioxy-18a-homo-5␤-
estr-3-en-11-one (13)
20
1
This compound was obtained analogously with the exception that
the reaction required 6 hours for a quantitative transformation into
ketal 13. Yield in total: 20.03 g 13 (77%). 13: mp. 160–164°C.
[␣]D20 ϭ ϩ69°. 1H nmr 5.63 (m, H-3), 5.34 (d, 10.3, H-4), 3.64 (d,
11.0, -OCH2C-), 3.36 (q, 10.5, -OCH2C-), 2.76 (d, 12.1, H-12),
2.51 (d, 12.1, H-12), 1.09 (s, -CH3), 1.01 (t, 7.2, H-18a), 0.72 (s,
-CH3). ms: m/z 372.26730 (Mϩ). C24H36O3 (372.55) calculated C
77.38 H 9.74 found C 77.35 H 9.65.
11-Methylene-18a-homo-5␤-estr-3-en-17-one (17)
This compound was obtained analogously. Yield: 11.6 g 17 (76%).
17: mp. 89–91°C. [␣]D20 ϭ ϩ93°. 1H nmr: 5.66 (m, H-3), 5.38 (d,
9.8, H-4), 4.88 (d, 1.4, ϭCH2), 4.85 (d, 1.4, ϭCH2), 2.56 (d, 12.4,
H-12), 2.40 (dd, 19.4, 7.7, H-16) 1.86 (d, 12.6, H-12), 0.75 (t, 7.7,
H-18a). ms: m/z 284.21460 (Mϩ). C20H28O (284.45) calculated C
84.45 H 9.92 found C 84.65 H 9.91.
17,17-(2,2-Dimethyl)propanedioxy-11-methylene-
18a-homo-5␣-estr-3-ene (14)
Methylene triphenylphosphorane was generated from methyl tri-
phenylphosphonium iodide (166.2 g; 413 mmol) and sodium hy-
dride (12.4 g, 80% in oil; 413 mmol) in absolute dimethylsulfoxide
(300 mL). A solution of ketal 12 (15.38 g; 41 mmol) in toluene
(100 ml) was dropped to the ylide and the resulting mixture was
allowed to react for 108 h at 80°C with ultrasonical irradiation.
After cooling to 25°C, the reaction was quenched by cautious
addition of water (20 mL). The mixture was extracted with cyclo-
hexane (3 times 200 mL). The combined organic phases were dried
and rotary evaporated to give compound 14 which was purified by
column chromatography (eluent: cyclohexane) and by crystalliza-
tion from acetone. Yield: 9.55 g 14 (62.3%). mp. 144–153°C.
11-Methylene-5␣, 17␣-18a-homo-19-nor-pregn-3-en-
20-in-17-ol (8)
The protocol described in ref. 1 for the ethinylation of 11-
methylene-18a-homo-estr-4-en-17-one was used to give title com-
20
1
pound 8 in a 56% yield. 8: mp. 105–108°C. [␣]D ϭ ϩ42°. H
nmr: 5.64 (dq, 9.8, 3.0 H-3), 5.48 (dq, 9.8, 2.2, H-4), 4.95
(s, ϭCH2), 4.68 (s, ϭCH2), 2.62 (s, -CϵCH), 2.61 (d, 11.9, H-12),
2.28 (d, 11.9, H-12), 1.04 (t, 7.2, H-18a). ms: m/z 310.22909
(Mϩ). C22H30O (310.48) calculated C 85.11 H 9.74 found C 85.24
H 9.91.
20
1
[␣]D ϭ ϩ66°. H nmr: 5.63 (m, H-3), 5.47 (dd, 9.6, 2.2, H-4),
4.92 (s, ϭCH2), 4.64 (s, ϭCH2), 3.60 (d, 11.3, -OCH2C-), 3.37 (m,
-OCH2C-), 2.45 (d, 12.6, H-12), 2.32 (d, 12.6, H-12), 1.12 (s,
-CH3), 0.99 (t, 7.2, H-18a), 0.72 (s, -CH3). ms: m/z 370.28720
(Mϩ). C25H38O2 (370.58) calculated C 81.03 H 10.34 found C
81.18 H 10.40.
11-Methylene-5␤, 17␣-18a-homo-19-nor-pregn-3-en-
20-in-17-ol (9)
The protocol described in ref. 1 for the ethinylation of 11-
methylene-18a-homo-estr-4-en-17-one was used to give title com-
pound 9 in a 79% yield. 9: mp. 98–99°C. [␣]D20 ϭ Ϫ31°. 1H nmr:
5.68 (dq, 9.8, 3.2 H-3), 5.38 (d, 9.8, H-4), 4.94 (d, 1.5, ϭCH2),
4.79 (s, ϭCH2), 2.63 (s, -CϵCH), 2.60 (d, 12.4, H-12), 2.27 (d,
12.6, H-12), 1.42 (q, 7.3, H-18), 1.03 (t, 7.3, H-18a). ms: m/z
310.23098 (Mϩ). C22H30O (310.48) calculated C 85.11 H 9.74
found C 85.21 H 9.68.
17,17-(2,2-Dimethyl)propanedioxy-11-methylene-
18a-homo-5␤-estr-3-ene (15)
Compound 13 (24.48 g; 65 mmol) dissolved in toluene (50 mL)
was analogously allowed to react with methylene triphenylphos-
phorane generated from methyl triphenylphosphonium iodide
(84.54 g; 210 mmol) and sodium hydride (5.9 g, 80% in oil; 197
mmol) in absolute dimethylsulfoxide (150 mL). After 1 h the
olefination was complete (tlc). Work-up, column chromatography,
and crystallization gave 15.8 g of the title compound 15 (65%
yield). 15: mp. 142–146°C. [␣]D20 ϭ ϩ9°. 1H nmr: 5.61 (m, H-3),
5.35 (d, 10.2, H-4), 4.91 (s, ϭCH2), 4.75 (s, ϭCH2), 3.60 (d, 11.4,
-OCH2C-), 3.40 (d, 11.4, -OCH2C-), 3,34 (m, -OCH2C-), 2.44 (d,
12.2, H-12), 2.32 (d, 12.2, H-12), 1.12 (s, -CH3), 0.99 (t, 7.3,
H-18a), 0.71 (s, -CH3). ms: m/z 370.28640 (Mϩ). C25H38O2
(370.58) calculated C 81.03 H 10.34 found C 81.07 H 10.29.
Results
Initially, we focused on an effective large-scale preparation
of 5␣-compound 5 and 5␤-compound 6 from the hydroxy
enone 1. Among methods which have been used to prepare
steroid 3-olefins from 4-en-3-oxo steroids are Wolff–Kish-
ner reductions and the reduction with zinc in acetic acid.
Wolff–Kishner reduction of cholest-4-en-3-one, using hy-
drazine with sodium ethoxide in ethanol or with potassium
hydroxide in diethylene glycol reportedly gave a mixture of
5␣- and 5␤-cholest-3-enes appreciably contaminated with
cholest-4-ene. Isolation of the ⌬3-double bond isomers was
accomplished by a laborious bromination, chromatography,
and debromination procedure.7 Due to these disadvantages,
the reduction of compound 1 using zinc in acetic acid8 was
considered to be the method of choice, though a high excess
of zinc is needed which may be a drawback when working
on a larger scale (stirring problems and production of a
highly pyrophorous zinc waste). Nevertheless, high yields
of pure 3-ene steroids have been reported with different
4-en-3-oxo steroids.8,9 To this end, we treated compound 1
with a thirtyfold excess of zinc (by weight) in acetic acid for
11-Methylene-18a-homo-5␣-estr-3-en-17-one (16)
To a solution of (trimethylsilyl)methyl lithium in n-pentane (170
mL; 170 mmol) was added a solution of compound 12 (20 g; 53
mmol) in toluene (120 mL) with stirring at 30°C. The reaction
which was complete after 15 min (tlc), was quenched by addition
of saturated aqueous ammonium chloride solution with cooling.
The organic layer was separated, washed with water until neutral
and evaporated. The resulting 11␣-(trimethylsilyl)methyl-17,17-
(2,2-dimethyl)propanedioxy-18a-homo-5␣-estr-3-en-11␤-ol was
dissolved in acetone, concentrated aqueous hydrochloric acid (1.5
mL) was added, and the mixture was stirred for 1 h at 40°C.
Thereafter, the solution was neutralized with saturated aqueous
sodium hydrogen carbonate solution and evaporated. The residue
obtained was dissolved in toluene, the solution was washed with
Steroids, 1998, vol. 63, January 23

Papers
1 hour at 30°C (Scheme 2). The processing clearly benefited
from sonification.9 After work-up the product was shown by
glc to be a mixture consisting of 5 (30.5%), 6 (64.5%), 4
(3%), and three less polar species (in total 2%; probably
acetates). The mixture was separated by chromatography on
silica gel affording 5 (24% yield) and 6 (45% yield). The
C-5 stereochemistry of both products was established ret-
rospectively, but rigorously, by x-ray diffraction analysis
(vide infra).
Continuing in line with our previously reported deso-
gestrel synthesis,1 compounds 5 and 6 were oxidized by the
Parikh–Doering procedure10 (Scheme 2) to give the di-
ketones 10 and 11 (70% yield each) which were allowed to
react with 2,2-dimethylpropane-1,3-diol, triethylorthofor-
mate and a catalytic amount of p-toluenesulfonic acid.
Within 6 h, the 5␤-compound 11 was transformed quanti-
tatively into the 17-ketal 13 (77% isolated yield), whereas,
under the same reaction conditions, the 5␣-compound 10
showed a merely 65% formation of ketal 12 after 10 h. A
78% isolated yield of ketal 12 was obtained after repeated
ketalization of recycled starting material. A different behav-
iour of the diketones 10 and 11 with respect to the regio-
specificity of ketal formation was not observed.
1 h (65% yield upon isolation and crystallization). However,
using 5␣-ketal 12, a quantitative transformation into the
methylene steroid 14 required 10 equiv. of ylide and a
reaction time of 108 h (61% yield upon isolation and crys-
tallization).
In continuation, the olefinated ketals 14 and 15 obtained
by the Wittig reaction were deprotected using p-
toluenesulfonic acid in acetone to give the ketones 16 and
17 (Scheme 2). More conveniently, the bulk of 16 and 17
was obtained by the Peterson olefination reaction of com-
pounds 12 and 13. As final steps of the title synthesis the
ketones 16 and 17 were subjected to ethinylation by lithium
acetylide/ethylenediamine following the protocol described
earlier1 (Scheme 2). In this way, compounds 8 and 9 were
obtained in a 56% and 79% yield respectively.
The A/B ring junctions belonging to the compounds
described in this paper have been elucidated by x-ray anal-
ysis. For crystallographic reasons the final intermediates 16
and 17 were selected for the measurements. Compound 16
showed a 5␣-configuration (Fig. 1) and compound 17
proved to be 5␤-oriented (Fig. 2). These results retrospec-
tively allowed to define the A/B ring junctions belonging to
series 5, 10, 12, 14, 16, 8 on the one hand and 6, 11, 13, 15,
17, 9 on the other hand.
The following Wittig olefination reaction with oxo ketals
12 and 13 (Scheme 2) was performed in the presence of
ultrasound.1 In spite of this accelerating modification the
reaction rates of 5␣- ketal 12 and 5␤-ketal 13 proved to be
dramatically different. When the 5␤-ketal 13 was treated
with 3.15 equiv. methylene triphenylphosphorane in di-
methylsulfoxide at 80°C, a quantitative transformation into
the corresponding 11-methylene steroid 15 occurred after
Discussion
The reason for the dramatic difference between the 11-oxo
steroids 12 and 13 to undergo olefination with methylene
triphenylphosphorane was not obvious. In principle, the low
reactivity of compound 12 towards the Wittig reagent may
Scheme 2 Formation of title compounds 8, 9 from hydroxy enone 1. i: Zn, AcOH, chromatography; ii: DMSO, NEt3, Py-SO3; iii:
(2,2-dimethyl)propane-1,3-diol, triethylorthoformate, TsOH; iv: Ph3P(ϩ)Me l(Ϫ), NaH, DMSO, PhMe, )))); v: 1. Me3SiCH2-Li, n-pentane, 2.
HCl, acetone, H2O, 40°C; vi: : p-TsOH, acetone; vii: ethine, Li, ethylenediamine, THF.
24 Steroids, 1998, vol. 63, January

⌬3-isomers of desogestrel: Ring et al.
Table 1 Energies of 12, 13, 18, and 19 Calculated with ab initio
and Semiempirical Methods
Compound
12
13
18
19
E [kcal/mol] HF 6-31G*
E [kcal/mol] PM3
Ϫ435957.75
Ϫ435956.33
Ϫ1107419.33
Ϫ1107419.42
Ϫ146.38
Ϫ145.28
Ϫ35.97
Ϫ36.11
philes13,14,15 which, in one analogous case, was suggested to
be due to exclusive enolization of the 11-oxo group. On the
other hand, a structurally comparable compound, showing a
5␤-configuration, was much less prone to enolization using
the same reaction conditions.16 However, since effective
nucleophilic reactions with 11-oxo steroids have also been
shown not to be influenced by their A/B-ring core,17 de-
tailed structural factors involved in more or less smooth
enolization of 11-oxo groups remain undefined as yet. To
investigate the tendency of enolization of compounds 12
and 13 the potential energies of the keto and the enolate
anion forms were calculated and compared. The semiem-
pirically determined starting geometries were minimized
using the ab initio HF 6-31G* method. The single point
energies of the resulting minimum structures were com-
puted in DMSO as solvent.4 The energy difference between
the two keto forms 12 and 13 (Ϫ1.35 kcal/mol in vacuo,
Ϫ1.32 kcal/mol in DMSO) and the enolate anion forms
(Ϫ2.08 kcal/mol in vacuo, Ϫ2.11 kcal/mol in DMSO) was
Ϫ0.73 kcal/mol in vacuo, respectively Ϫ0.79 kcal/mol in
DMSO (see Table 2). The energetic preference of the 5␤-
enolate anion of approximately 0.8 kcal/mol over the 5␣-
enolate anion is considered not to be sufficient to explain
the big difference in the reaction time solely.
A higher energetic burden in the formation of the 5␣-
oriented oxaphosphetane Wittig reaction intermediate ver-
sus the corresponding 5␤-species might be another expla-
nation for the different olefination reaction rates. It was
found by the low-temperature nmr studies of Vedejs et al.
that pentacoordinate 1,2-oxaphosphetanes were in situ in-
termediates in the Wittig olefination reaction which decom-
posed to give the corresponding olefins.18 Reactions at C-11
of estrane and 18a-homo estrane derivatives which were
passing a four-membered transition state were shown to
proceed almost exclusively from the rear side of the mole-
cule.19,1 Hence, in our case, oxaphosphetanes resulting from
methylene triphenylphosphorane attack at the 11-oxo
groups of compounds 12, 13 were suggested to have struc-
tures 18 and 19 (Scheme 2). Ab initio molecular orbital
Figure 1 Crystal structure of 11-methylene-18a-homo-5␣-estr-
3-en-17-one (16) (crystallographic numbering).
have been due to steric shielding of the 11-oxo group from
the ␣- as well as from the ␤-side of the steroid molecule.
However, steric hindrance was considered unlikely, since
ring C of both the compounds 12 and 13 should be equally
accessible from the ␤-side. In the case of an ␣-approach ring
C of compound 13 would be expected to be the more
hindered.11 Indeed, when the oxo steroids 12 and 13 were
subjected to the Peterson olefination reaction12 (Scheme 2),
they reacted equally fast: conversion into the two isomeric
methylene ketones 16 and 17 via the corresponding 11-
(trimethylsilyl)methyl carbinols was complete after 15 min
at 30°C (nearly quantitative yield each). Semiempirical
PM3 calculations also suggested no energetical preference
of the intermediary 5␣- over the 5␤-11-(trimethyl-
silyl)methyl carbinols. The energy difference between these
two reaction intermediates is the same as calculated for 12
and 13 (see Table 1). This indicates that the 11-oxo group of
12 and 13 is comparably accessible for the reaction with
trimethylsilylmethyl-lithium. Thus, we concluded that the
retarded Wittig reaction of compound 12 was not caused by
steric hindrance of the 11-oxo group.
Next, we focused on the preferred ylide mediated 11-oxo
group enolization of compound 12 versus compound 13.
Several unsuccessful attempts have been reported to attack
the 11-oxo group of 5␣-pregnane derivatives with nuleo-
Table 2 Ab initio Calculated Potential Energies of the Keto and
Enolate Anion Forms of Compounds 12 and 13
E [kcal/mol]
in vacuo
E [kcal/mol]
in DMSO
Compound
12
13
Ϫ724388.03
Ϫ724386.68
Ϫ723995.20
Ϫ723993.12
Ϫ724391.52
Ϫ724390.20
Ϫ724025.65
Ϫ724023.54
12 enolate anion
13 enolate anion
Figure 2 Crystal structure of 11-methylene-18a-homo-5␤-estr-
3-en-17-one (17) (crystallographic numbering).
Steroids, 1998, vol. 63, January 25

Papers
calculations were carried out to investigate this explanation
for the differing reaction times. The potential energies for
the educts 12/13 and the intermediates 18/19 in the Wittig
olefination were calculated. The energy difference between
molecules 12/13 (Ϫ1.42 kcal/mol) and the 5␣-/5␤-
intermediates 18/19 (ϩ0.09 kcal/mol) was Ϫ1.51 kcal/mol
(see Table 1). This indicates an energetic preference of 1.51
kcal/mol of the 5␤-oxaphosphetane intermediate over the
corresponding 5␣-intermediate. Steric repulsion between
the axial proton at C-2 of ring A and one of the phosphorus-
bound phenyl group of oxaphosphetane 18 led to a slightly
different and energetically less favorable position of the
four-ring system and the phenyl substituents compared to
the 5␤-fused oxaphosphetane 19 (see Fig. 1). Since the
5␣-oriented 1,2-oxaphosphetane intermediate 18 adopted a
less favorable conformation, a steric hindrance in the for-
mation of the 5␣-intermediate 18 vs. the corresponding
5␤-intermediate 19 seemed to be possible. But care should
be taken with the interpretation of the calculations, since not
the whole potential energy surface, resulting from the three
rotatable bonds at the phenyl substituents, could be calcu-
lated with ab initio methods. Even the smallest 1,2-
oxaphosphetane-intermediate fragments, consisting of at
least 78 atoms (37 non-hydrogen and 41 hydrogen atoms),
are too big to be applicable to ab initio conformational
searches. To ensure a sufficient sampling of conformational
space, instead a semiempirical conformational search was
performed. The energetically most favorable conformations
from this systematic conformational search were further
investigated by ab initio methods. Since ring D and the
corresponding substituents were the same in molecules 18
and 19 and not in direct contact to the phenyl substituents,
these molecule parts were omitted from the ab initio calcu-
lations. These HF/6-31G* calculations were only feasible
by reducing the number of atoms in the system and calcu-
lating a subset of the molecules, consisting of rings A, B, C
and the ethyl group. Neither the conformations nor the
energy differences of the ab initio calculations (Ϫ1.51 kcal/
mol) differed greatly from the PM3 calculations (Ϫ1.24
kcal/mol). This suggests that the PM3 results are approxi-
mately comparable to the ab initio calculations and that the
semiempirical systematic conformational search resulted in
relevant energy minimum structures for the intermediates.
The potential energy values for 12, 13, 18 and 19, as printed
out by Gaussian 944 and VAMP3, are shown in Table 1. The
values listed for the HF 6-31G* method have been calcu-
lated for the ABC-ring subsets. A superposition of the
lowest energy conformers of the ABC-ring subsets of com-
pounds 18 (black) and 19 (grey) is depicted in Fig. 3. The
molecules were fitted at the 10 carbon atoms in the B and C
ring.
Figure 3 Superposition of energy minimum structures of 18
and 19.
12 in comparison to oxo steroid 13. However, whether the
calculated energy difference between molecules 12/13 and
18/19 of 1.51 kcal/mol may be the sole, reasonable expla-
nation for the dramatic reaction rate difference, remains
questionable. Hence, further studies have to be done in this
field.
In conclusion, we have been studying the synthesis of the
5␣- and 5␤-oriented 3-unsaturated analogs 8, 9 of the pro-
gestin desogestrel (7). The Wittig olefination reaction of the
intermediates 12 and 13 showed a dramatically reduced
reaction rate of the 5␣-isomer 12 versus the 5␤-isomer 13.
Computational studies suggest that different energies of the
1,2-oxaphosphetane intermediates may, at least partially,
have been the reason for this phenomenon. The structural
assignment of the 5␣- and the 5␤-oriented compounds was
performed on the basis of x-ray analysis.
Acknowledgments
The x-ray analysis of compounds 16 and 17 by Dr. H. Go¨rls
(Max-Planck-Gesellschaft, Group for CO2 Chemistry at
Friedrich Schiller University, Jena) is gratefully acknowl-
edged. We thank Mrs. S. Ha¨gele and Mrs. U. Zo¨rner for
their help in preparing this manuscript.
As a summing-up, our efforts to find out why the 11-oxo
steroids 12 and 13 showed such dramatic differences in the
Wittig olefination reaction rates, gave an incompletely sat-
isfying answer only. Attempts to get insight into this phe-
nomenon by computational studies suggested that different
enolization or enolate formation rates of the 11-oxo groups
in question were obviously less important. Ab initio molec-
ular orbital calculations showed that an unfavorable geom-
etry of the 1,2-oxaphosphetane intermediate 18 may have
participated in the slower olefination reaction of oxo steroid
References
1.
Schwarz S, Ring S, Weber G, Teichmu¨ller G, Palme H-J, Pfeiffer C,
Undeutsch B, Erhart B, Grawe D (1994). Synthesis of 13-ethyl-11-
methylene-18,19-dinor-17␣-pregn-4-en-20-yn-17-ol (Desogestrel)
and its main metabolite 3-oxo-desogestrel. Tetrahedron 36:10709–
10720.
2.
3.
Hopkins SJ (1982). Desogestrel. Drugs Today 18:361–366.
VAMP 6.0, Oxford Molecular Ltd., Magdalen Centre, Oxford Sci-
ence Park, Sandford-on-Thames, Oxford, OX4 4GA, England.
Gaussian 94 (Revision D.1), Frisch MJ, Trucks GW, Schlegel HB,
4.
26 Steroids, 1998, vol. 63, January

⌬3-isomers of desogestrel: Ring et al.
Gill PMW, Johnson BG, Robb MA, Cheeseman JR, Keith TA,
Petersson GA, Montgomery JA, Raghavachari K, Al-Laham MA,
Zakrzewski VG, Ortiz JV, Foresman JB, Cioslowski J, Stefanov
BB, Nanayakkara A, Challacombe M, Peng CY, Ayala PY, Chen
W, Wong MW, Andres JL, Replogle ES, Gomperts R, Martin RL,
Fox DJ, Binkley JS, Defrees DJ, Baker J, Steward JP, Head-Gordon
M, Gonzales C, Pople JA, Gaussian, Inc. Pittsburgh PA, 1995.
SYBYL, Molecular Modeling System, Version 6.3, (1996), Tripos
Associates, Inc., St. Louis, MO.
11.
12.
13.
14.
15.
Wechter WJ (1959). Stereoselective cis-non-markownikoff hydra-
tion of steroid double bonds. Chem. Ind. (London) 294–295.
Ager DJ (1984). The Peterson reaction. Synthesis 384–398, and
references cited there.
Ringold HJ, Batres E, Zderic JA (1958). Steroids-XC. 11␣-Methyl-
11␤-hydroxy steroids. Tetrahedron 2:164–166.
Fonken GS, Hogg JA (1958). 11-Alkylated steroids. I. The synthesis
of 11-methylhydrocortisone acetate. Tetrahedron 2:365–366.
Elks J (1960). Compounds related to the steroid hormones. Part I.
The preparation and reactions of some 11-methyl-steroids. J Chem
Soc 3333–3340.
Fonken GS, Hogg JA, McIntosh AV Jr (1959). 11-Alkylated ste-
roids. III. Two syntheses of 11-methylhydrocortisone. J Org Chem
24:1600–1604.
Zderic JA, Batres E, Limo´n DCh, Carpio H, Lisci J, Monroy G,
Necoechea E, Ringold HJ (1960). Steroids. CXL. 11-Methyl ste-
roids. J Am Chem Soc 82:3404–3409.
Maryanoff BE, Reitz AB (1989). The Wittig olefination reaction
and modifications involving phosphoryl-stabilized carbanions. Ste-
reochemistry, mechanism, and selected synthetic aspects. Chem Rev
89:863–927 and references cited there.
Van den Broek AJ, van Bokhoven C, Hobbelen PMJ, Leemhuis J
(1975). 11-Alkyliden steroids in the 19-nor series. Recl Trav Chim
Pays-Bas 94:35–39.
5.
6.
7.
Steward JJP (1989). Optimization of parameters for semiempirical
methods. J Comp Chem 10:209–220.
Nickon A, Schwartz N, DiGiorgio JB, Widdowson DA (1965).
Reactivity and geometry in allylic systems. IV. Stereochemical
factors in the photosensitized oxygenation of 5␣- and 5␤-cholest-
3-enes. J Org Chem 30:1711–1717, and references cited there.
McKenna J, Norymberski JK, Stubbs RD (1959). Partial reduction
of steroid hormones and related substances. Part III. The reaction of
␣␤-unsaturated ketones with zinc in acetic acid. J Chem Soc 2502–
2509.
Salvador JAR, Sa´ e Melo ML, Campos Neves AS (1993). Ultra-
sound assisted zinc reactions in synthesis 1. Efficient reduktion of
enones. Tetrahedron Lett 34:357–360.
Parikh JR, Doering WvE (1967). Sulfur trioxide in the oxidation of
alcohols by dimethyl sulfoxide. J Am Chem Soc 89:5505–5507.
16.
17.
18.
8.
9.
19.
10.
Steroids, 1998, vol. 63, January 27