The total synthesis of (±)-ginkgolide B

Article abstract of DOI:10.1021/ja001747s

The total synthesis of the potent PAF antagonist ginkgolide B has been accomplished. The complex architecture of ginkgolide B which includes six rings, eleven stereogenic centers, ten oxygenated carbons, and four contiguous fully substituted carbons is a

Full text of DOI:10.1021/ja001747s

J. Am. Chem. Soc. 2000, 122, 8453-8463
8453
The Total Synthesis of (()-Ginkgolide B
Michael T. Crimmins,* Jennifer M. Pace, Philippe G. Nantermet, Agnes S. Kim-Meade,
James B. Thomas, Scott H. Watterson, and Allan S. Wagman
Contribution from the Venable and Kenan Laboratories of Chemistry,
The UniVersity of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3290
ReceiVed May 19, 2000
Abstract: The total synthesis of the potent PAF antagonist ginkgolide B has been accomplished. The complex
architecture of ginkgolide B which includes six rings, eleven stereogenic centers, ten oxygenated carbons, and
four contiguous fully substituted carbons is a daunting challenge for chemical synthesis. The synthesis of
ginkgolide B was accomplished through a stereoselective intramolecular photocycloaddition of enone 5 to
construct the congested core of the molecule. The photocycloaddition substrate was prepared through technology
for the construction of carboalkoxycyclopentenones previously reported from these laboratories. Regioselective
cyclobutane fragmentation and further functionalization of the photoadduct 4 provided the key pentacyclic
intermediate. Acid-catalyzed rearrangement and epoxide opening were key transformations in the production
of ginkgolide B from the pentacyclic intermediate.
Ginkgo biloba, one of the oldest surviving flora with ancestors
dating to 230 million B.C., flourished during the Jurassic Period
and has been called the “living fossil” by Charles Darwin.1 The
ginkgo tree survives today because of its extraordinary resil-
ience, having endured several planetary mass extinctions of plant
life. Its regenerative powers are supported by a report that a
ginkgo tree sprouted anew from its roots at ground zero
Hiroshima. Extracts of G. biloba have been used as herbal
medicines for approximately 5000 years to treat a variety of
ailments including coughs, asthma, and circulatory disorders.
The traditional Hindu medicine “soma” also contains ginkgo
extracts, and recent clinical studies attest to possible benefits
of ginkgolides in the delay in the onset of dementia.2
Ginkgolides A, B, C, and M (Figure 1), which differ only in
the number and location of hydroxyl groups, were first isolated
as “bitter principles” of the root bark by Furukawa and
co-workers in 1932.3 The structures of the ginkgolides were
first elucidated in 1967 by a series of spectroscopic studies by
Nakanishi.4 The assigned structures were independently con-
Figure 1.
firmed by X-ray crystallography by the Okabe group at Nagoya
University.5 A related C15 compound, bilobalide, was discov-
B is the most potent platelet-activating factor (PAF) antagonist
of the ginkgo extracts with an IC50 value of 0.6 mM.9
ered in 1971,6,7 and in 1987, another member of the ginkgolide
family, ginkgolide J, was isolated and characterized.8 Ginkgolide
Because of its portentous molecular architecture, ginkgolide
B is an intimidating challenge for chemical synthesis. Included
in the ginkgolide skeleton are six rings, eleven stereogenic
centers, ten oxygenated carbons, an unusual tert-butyl group,10
and four contiguous fully substituted carbon atoms. At the
sterically congested core of the molecule are quaternary carbons
C5 and C9, which are jointly embodied in five of the six rings
of the molecule. The stereogenically correct installation of the
C5 and C9 quaternary carbons and the proper orchestration of
functional group manipulation are the most critical issues in
regard to a successful synthetic assault. Synthetic studies on
the ginkgolides have been relatively limited.11 The synthesis of
bilobalide by Corey and co-workers in 1987 was the first report
(1) van Beek, T. A.; Bombardelli, E.; Morazzoni, P.; Peterlongo, F.
Fitoterapia 1998, 69, 195-244. Major, R. T. Science 1967, 157, 1270.
See also: Ginkgo Biloba, A Global Treasure: from Biology to Medicine;
Hori, T., Ed.; Spriger: New York, 1997.
(2) LeBars, P. L.; Katz, M. M.; Berman, N.; Itil, T. M.; Freedman, A.
M. Schatzberg, A. F. J. Am. Med. Assoc. 1997, 278, 1327-1332. Ernst,
E.; Pittler, M. H. Clin. Drug InVest. 1999, 17, 301-308.
(3) Furukawa, S. Sci. Papers Inst. Phys. Chem. Res. Tokyo 1932, 19,
27.
(4) Nakanishi, K. Pure Appl. Chem. 1967, 14, 89-113. Sakabe, N.;
Takada, S.; Okabe, K. J. Chem. Soc., Chem. Commun. 1967, 259-261.
(5) Okabe, K.; Yamada, K.; Yamamura, S.; Takada, S. J. Chem. Soc. C
1967, 2201-2206.
(6) Weinges, K.; Bahr, W. Liebigs Ann. Chem. 1969, 724, 214.
(7) Nakanishi, K.; Habaguchi, K.; Nakadaira, Y.; Woods, M. C.;
Maruyama, M.; Major, R. T.; Alauddin, M.; Patel, A. R.; Weinges, K.;
Bahr, W. J. Am. Chem. Soc. 1971, 93, 3544-3546.
(8) Weinges, K.; Hepp, M.; Jaggy, H. Liebigs Ann. Chem. 1987, 521-
526.
(9) Braquet, P. Drugs Future 1987, 12, 643.
(10) Nakanishi, K.; Habaguchi, K. J. Am. Chem. Soc. 1971, 93, 3546-
3547.
10.1021/ja001747s CCC: $19.00 © 2000 American Chemical Society
Published on Web 08/18/2000

8454 J. Am. Chem. Soc., Vol. 122, No. 35, 2000
Crimmins et al.
Scheme 1
Scheme 2
Scheme 3
of a synthesis of a member of the ginkgolide family.12 The Corey
group disclosed their successful syntheses of ginkgolides B13
and A14 the following year. Corey has also reported on the
synthesis and biological activity of several simpler ginkgolide
analogues.11 Work in our laboratories has also led to the
completion of the syntheses of bilobalide15 and, recently,
ginkgolide B.16 Herein is presented a detailed account of the
investigations which culminated in the total synthesis of
ginkgolide B.
In the original strategic analysis of ginkgolide B (Scheme
1), pentacycle 2 was thought to be a viable intermediate for the
attachment of the C ring lactone. Acetal 2 would be obtained
from bis(methyl acetal) 3, via an acid-catalyzed closure of the
E ring. The bisacetal 3 would ultimately arise from the
regioselective cyclobutane cleavage of tetracycle 4. The key
step in the proposed sequence, a double diastereoselective
intramolecular [2+2] photocycloaddition,17 would provide 4
from enone 5. The photocycloaddition would serve to not only
construct the B ring, but also establish the stereochemistry of
the two quaternary centers at C5 and C9 as well as the C4
stereochemistry. Photosubstrate 5 would be obtained from
acetylenic ester 6 via a zinc homoenolate conjugate addition/
cyclization which was specifically developed for this purpose
during the course of this work.18
The first objective in the synthesis was the development of
a workable synthesis of the acetylenic ester 6. Synthesis of
acetylenic ester 6 began with commercially available 3-fural-
dehyde which was converted to unsaturated ester 7 via a Wittig
reaction with the stabilized ylide, Ph3PdCHCO2Et (Scheme 2).15
Exposure of ester 7 to the higher order tert-butyl cuprate (t-
Bu2CuCNLi2)19 in the presence of TMSCl provided the ester 8
which was directly treated with i-Bu2AlH to provide the
aldehyde 9 in 95% overall yield. Addition of the lithium
acetylide of ethyl propiolate to aldehyde 9 resulted in a 3.3:1
separable mixture of the desired anti diastereomer 10 to the syn
product 11.20
The completion of the assembly of the photosubstrate 5 from
the acetylenic ester 6 was accomplished through a conjugate
addition-cyclization protocol using the Kuwajima-Nakamura
zinc homoenolate21 which had been developed in our laboratory
specifically for this purpose. The zinc homoenolate 12 was
generated via ultrasonic irradiation of [(ethoxycyclopropyl)oxy]-
trimethylsilane22 in the presence of ZnCl2‚OEt2 (Scheme 3).
Treatment of the homoenolate reagent with CuBr‚SMe2 followed
by addition of the acetylenic ester 10 and HMPA to the reaction
mixture resulted in the desired conjugate addition-cyclization
in excellent yield.18 Mechanistically, it is proposed that the zinc/
(11) Corey, E. J.; Gavai, A. V. Tetrahedron Lett. 1988, 29, 3201-3204.
Corey, E. J.; Gavai, A. V. Tetrahedron Lett. 1989, 30, 6959-6962. Corey,
E. J.; Kamiyama, K. Tetrahedron Lett. 1990, 31, 3995-3998. Corey, E. J.;
Rao, K. S. Tetrahedron Lett. 1991, 32, 4623-4626. Weinges, K.; Rummler,
M.; Schick, H.; Schilling, G. Liebigs Ann. Chem. 1993, 287-291.
(12) Corey, E. J.; Su, W. J. Am. Chem. Soc. 1987, 109, 7534-7536.
Corey, E. J.; Su, W. Tetrahedron Lett. 1988, 29, 3423-3426.
(13) Corey, E. J.; Kang, M.; Desai, M. C.; Ghosh, A. K.; Houpis, I. N.
J. Am. Chem. Soc. 1988, 110, 649-651. Corey, E. J. Chem. Soc. ReV. 1988,
17, 111-133. Corey, E. J.; Gavai, A. V. Tetrahedron Lett. 1988, 29, 3201-
3204. Desai, M. C.; Ghosh, A. K.; Kang, M.-C.; Houpis, I. N. In Strategies
and Tactics in Organic Synthesis; Lindberg, T., Ed.; Academic Press: New
York; 1991; pp 89-119.
(14) Corey, E. J.; Ghosh, A. K. Tetrahedron Lett. 1988, 29, 3205-3206.
(15) Crimmins, M. T.; Jung, D. K.; Gray, J. L. J. Am. Chem. Soc. 1992,
112, 5445-5547. Crimmins, M. T.; Gray, J. L.; Jung, D. K. J. Am. Chem.
Soc. 1993, 115, 3146-3155.
(16) Crimmins, M. T.; Pace, J. M.; Nantermet, P. G.; Kim-Meade, A.
S.; Thomas, J. B.; Wagman, A. S.; Watterson, S. H. J. Am. Chem. Soc.
1999, 121, 10249-10250.
(17) Crimmins, M. T.; Reinhold: T. L. Org. React. 1993, 44, 297-588.
Crimmins, M. T. In ComprehensiVe Organic Synthesis; Trost, B. M., Ed.;
Pergamon: Oxford, England, 1991; Vol. 5, pp 123-150. Crimmins, M. T.
Chem. ReV. 1988, 88, 1453-1473. De Keukeleire, D.; He, S.-L. Chem.
ReV. 1993, 93, 359-380. For recent discussions of the mechanism of [2+2]
enone-olefin photocycloadditions see: Schuster, D. I.; Lem, G.; Kaprinidis,
N. A. Chem. ReV. 1993, 93, 3-22. Andrew, D.; Hastings, D. J.; Weedon,
A. C. J. Am. Chem. Soc. 1994, 116, 10870-10882. Maradyn, D. J. Weedon,
A. C. Tetrahedron Lett. 1994, 35, 8107-8110; Maradyn, D. J. Weedon,
A. C. J. Am. Chem. Soc. 1995, 117, 5359-5360.
(18) Crimmins, M. T.; Nantermet, P. G. J. Org. Chem. 1990, 55, 4235-
4236. Crimmins, M. T.; Nantermet, P. G.; Trotter, B. W.; Vallin, I. M.;
Watson, P. S.; McKerlie, L. A.; Reinhold: T. L.; Cheung, A. W. H.; Stetson,
K. A.; Dedopoulou, D.; Gray, J. L. J. Org. Chem. 1993, 58, 1038-1047.
(19) Lipshutz, B. H. Sengupta, S. Org. React. 1992, 41, 135.
(20) Attempts to convert the syn isomer to the anti isomer via the
Mitsunobu inversion led only to Michael addition of the acetylenic ester in
the hydrolysis step.
(21) Nakamura, E.; Kuwajima, I. J. Am. Chem. Soc. 1984, 106, 3368.
Nakamura, E.; Aoki, S.; Sekia, K.; Oshino, H.; Kuwajima, I. J. Am. Chem.
Soc. 1987, 109, 8056.
(22) Salaun, J.; Marguerite, J. Organic Syntheses; Wiley: New York,
1990; Collect. Vol. VII, p 131.

Total Synthesis of (()-Ginkgolide B
J. Am. Chem. Soc., Vol. 122, No. 35, 2000 8455
Scheme 4
Scheme 5
Scheme 6
proximately 2.0 kcal/mol lower in energy than the boatlike
conformation 16. Conformation 17 would lead to the observed
product 15. The enone-furan 18 which includes the tert-butyl
substituent, but lacks the C6 trimethylsilyloxy group, showed
substantially lower selectivity, providing a 1.5:1 ratio of the
diastereomeric photocycloadducts 21:22. Here the two best
conformations for the initial ring closure are calculated to be
separated by only 0.2 kcal/mol, leading to nearly degenerate
conformations 19 and 20. Apparently the gauche interaction
between the tert-butyl group and the C4 position of the furan
raises the energy of the chairlike conformation 19, resulting in
a nonselective photocycloaddition. The implication which results
from these investigations is that the principal controlling factor
for asymmetric induction in the enone-furan 5, required for
the ginkgolide B synthesis, is that the influence of the trimeth-
ylsilyloxy substituent at the C6 center should override the effect
of the tert-butyl group.
On the basis of the results of the enone-furans 14 and 18
with monosubstituted tethers, it was not surprising that irradia-
tion of enone 5 selectively produced diastereomer 23 with none
of the desired diastereomer 24 detected (Scheme 5). However,
when the trimethylsilyl group was removed and the correspond-
ing free hydroxyl was irradiated, a solvent effect on the
diastereoselectivity was observed.25 Specifically, in nonpolar
solvents such as hexanes and benzene, a 1.1:1 mixture of 26:
27 was observed. The stereochemical assignment of the products
was made by exposure of both 26 and 27 to catalytic PPTS in
benzene at reflux (Scheme 6). Hydroxy ester 26 was recovered
unchanged from the reaction, but 27 was rapidly converted to
the bridged lactone 28 in excellent yield. A single-crystal X-ray
of lactone 28 further confirmed the assignment.
copper homoenolate effects a syn addition to the acetylene to
form the C-metalated intermediate which can tautomerize to
the O-metal allenoate. The allenoate can be trapped by trim-
ethylsilyl chloride, which is generated in situ during formation
of the zinc homoenolate. The presence of trimethylsilyl chloride
also results in protection of any free hydroxyls present in the
molecule. The intermediate silyl allenoate can cyclize, with
assistance of the Lewis acidic zinc salts present in the reaction
mixture, to form the desired cyclopentenone 5 in greater than
80% yield.
With substrate 5 in hand, the pivotal synthetic operation, a
double diastereoselective enone-furan [2+2] photocycloaddi-
tion could be investigated.17 Similar photocycloadditions had
been examined on stereochemically simplified systems 14 and
18,23 as shown in Scheme 4, to determine the influence of the
individual substituents at C6 and C8, the two allylic positions
on the tether, on the stereoselectivity of the photocycloaddition.
Irradiation of cyclopentenone 14, which lacked the C8 tert-butyl
group, gave 87% yield of a single diastereomer 15 in which
the methyl ester and silyl ether groups were oriented trans in
the photoadduct. Two reasonable conformations for the first
bond formation in the triplet excited state are shown in Scheme
4. Both position the trimethylsilyloxy group in a pseudoequa-
torial orientation on the forming five-membered ring. The
chairlike conformation 17 has been calculated24 to be ap-
Further investigation of the photocycloaddition of enone 25
in more polar solvents, such as THF and MeOH, was found,
once again, to favor the formation of the undesired product 26.
The rationale for the selectivity trend is found in an examination
of the transition states shown in Figure 2. The two lowest energy
(24) Energy minimizations were performed in MM2 by restricting the
interatomic distance of the enone â-carbon and the alkene carbon to 2.5 Å
and allowing the conformation to be minimized.
(23) Crimmins, M. T.; Thomas, J. B. Tetrahedron Lett. 1989, 30, 5997-
6000.
(25) Crimmins, M. T.; Choy, A. L. J. Am. Chem. Soc. 1997, 119, 10237-
10238.

8456 J. Am. Chem. Soc., Vol. 122, No. 35, 2000
Crimmins et al.
Scheme 7
Scheme 8
hydrogen-bonding effect had been enhanced. Other amides, such
as the diethylamide and the morpholine amide, gave similar,
but slightly diminished selectivities.
While the photocycloadditions of the enone-furans with anti
stereochemistry of the tether substituents had led to only modest
selectivity for the desired cycloadduct, access to the bridged
lactone 28 had been achieved in just seven synthetic steps, and
the material produced through the studies on the anti diastere-
omer served to fuel early investigations on the advanced stages
of the synthesis. Nevertheless, the pivotal nature of the
photocycloaddition in the synthesis required a more highly
selective and efficient cycloaddition. Since the photocycload-
dition of 5 and 19 was not selective for the required tetracycle
24, and the amide photoreactions showed only moderate
selectivities, an alternate route to bridged lactone 28 was
contemplated.27 Specifically, it was postulated that a syn
diastereomer 32 would permit a transition state in the photo-
cycloaddition in which both the trialkylsilyl ether and the tert-
butyl group would be in pseudoequatorial orientations in the
chairlike conformation (Scheme 8). A matched double-diaste-
reoselective effect was anticipated since the individual substit-
uents each favored pseudoequatorial orientations on the basis
of the products observed in the model photocycloadditions. The
photocycloaddition of enone 32 was expected to be highly
diastereoselective for photocycloadduct 33. The requirement that
the C6 hydroxyl would need to be inverted to construct the D
ring lactone was thought to be a minor inconvenience.
A collateral need to improving the selectivity in the photo-
cycloaddition was the supply issue of the photosubstrate. A
revision of the early stages of the synthesis was required to
access ample quantities of photosubstrate 37. Several approaches
were examined in an attempt to improve access to the syn
diastereomer 11. Ultimately, the most practical sequence,
regarding efficient throughput, is illustrated in Scheme 9.
Addition of ethynylmagnesium bromide to aldehyde 8 resulted
in a 97% yield of a 1.2:1 separable mixture of anti acetylenic
alcohol 34 and syn acetylenic alcohol 35. Attempts to improve
the selectivity in favor of the syn diastereomer by changing
solvents and metal counterion and using amine additives did
not significantly alter the selectivity and often led to lower yield
of product.28 However, since the diastereomers were readily
separable by flash chromatography, and the anti diastereomer
34 could be easily converted to the syn diastereomer 35 in near
quantitative yield via a Mitsunobu inversion,29 this strategy was
Figure 2.
conformations (A and B) which would lead to the desired 24
both suffer from severe steric interactions resulting from an axial
oriented tert-butyl or trimethylsilyloxy group. The chairlike
conformation D, which would lead to the undesired product 23,
also has a prohibitive nonbonded interaction involving the tert-
butyl group. The nonbonded destabilization caused by the axial
trimethylsilyloxy group in boatlike conformation C, which
would lead to 23, is minimized by the long O-Si bond and the
possibility for the silyl ether to rotate away from the axial
hydrogen. Therefore, it is conformation C which is postulated
to be the favored transition state, and consequently, 23 is the
favored product in the photocycloaddition of 5.
The altered selectivity observed in the alcohol 25 is thought
to arise from stabilization of the chairlike conformation E which
is similar to conformation B for the trimethylsilyl ether 5. The
added stabilization of a hydrogen bond between the free
hydroxyl and the ester carbonyl as shown may account for the
observed solvent effect.
That a hydrogen-bonding interaction between the secondary
hydroxyl and the ester carbonyl might be responsible for altering
the diastereoselectivity of the photocycloaddition was an
encouraging observation. It was postulated that if the hydrogen-
bonding interaction could be enhanced, the desired selectivity
in the photocycloaddition might be achievable. Since amide
carbonyls are more Lewis basic than ester carbonyls, and
therefore should be more effective hydrogen bond acceptors,
the possibility of using tertiary amides in the photocycloaddition
was investigated, as shown in Scheme 7.26 Indeed, when
dimethylamide 29 was irradiated in hexanes, a 3:1 mixture of
photocycloadducts 30:31 was obtained, indicating that the
(27) All attempts to preform the D ring bridged lactone prior to the
photoaddition were unsuccessful.
(26) The use of secondary amides was also investigated, and although
selectivies in the photocycloaddition were improved (∼5:1 desired:
undesired), yields were low and the mass balance implied competitive
decomposition in the photoreaction.
(28) Carreira, E. M.; Du Bois, J. J. Am. Chem. Soc. 1995, 117, 8106-
8125
(29) Martin, S. F.; Dodge, J. A. Tetrahedron Lett. 1991, 32, 3017-3020.

Total Synthesis of (()-Ginkgolide B
J. Am. Chem. Soc., Vol. 122, No. 35, 2000 8457
Scheme 9
Scheme 10
Figure 3.
photocycloaddition studies on the anti diastereomer 5, in 63%
overall yield from silyl ether 38. Lactone 28 was converted to
enone 43 in 78% yield in one pot via treatment with phenylse-
lenium chloride with catalytic HCl in EtOAc30 followed by in
situ oxidation of the resulting selenide with NaIO4.
With the D ring lactone in place, attention focused on the
regioselective opening of the cyclobutane ring. Originally, it
was anticipated that the cyclobutane opening would be difficult
to preVent after the photocycloaddition, since the cyclobutane
bond between C12 and C4 should be substantially weakened
by the presence of two strong acceptors at C4 and a good donor
at C12 (Figure 3). In fact, the unexpected stability of lactones
28 and 43 necessitated an extensive investigation to ascertain
experimental conditions which would lead to selective cleavage
of the C4-C12 cyclobutane bond. The π overlap of the enol
ether oxygen lone pair with the alkene is believed to diminish
its donor ability, thus preventing cyclobutane rupture.
Initial investigations into the feasibility of the selective
cleavage of the cyclobutane involving the related system 44
were very encouraging23 (Scheme 11). Treatment of cyclobutane
44 with BF3‚OEt2 or HCl in methanol smoothly produced the
bis(methyl acetal) 45. Unfortunately, when this procedure was
implemented on the enone 43, the desired bis(methyl acetal)
47 was not observed. Instead, the furan 46 derived from cleavage
of the C3-C9 cyclobutane bond was obtained. Use of the
saturated ketone 28 also failed to give the desired product under
these conditions. Attempts to execute the cleavage of the
cyclobutane bond prior to introduction of the enone to avoid
the C3-C11 bond rupture were thwarted by the unexpected
formation of the mixed ketal 48. Several factors appear to
contribute to the difference of the behavior of lactones 44 and
28. The vinyl ether oxygen of 28 is a poorer donor than the
ether oxygen of 44, and the gem-dimethyl prevents enolization
of the cyclopentanone in 44, whereas in 28, the keto-enol
tautomerization may decrease the acceptor ability of the
cyclopentanone carbonyl. Most importantly, however, is the
apparent inability of the enol ether of 28 to undergo acid-
adopted for scale-up. Alcohol 35 was then protected as its
triethylsilyl ether, and the corresponding lithium acetylide was
acylated with ethyl chloroformate to provide acetylenic ester
36 in 95% overall yield. Exposure of the acetylenic ester 36 to
the previously described conditions for the conjugate addition-
cyclization reaction resulted in production of cyclopentenone
37 in 82% yield.18 As had been anticipated on the basis of the
analysis of the substituent effects in the model systems and in
the photocycloaddition of enone 5, irradiation of cyclopentenone
37 in hexanes at >350 nm gave a single diastereomeric
photocycloadduct, 38, in quantitative yield. The remarkable
chemical efficiency and stereoselectivity of the photocycload-
dition of enone 37 were instrumental in the advancement of
the synthesis.
With an efficient and scalable synthesis of the photocycload-
duct 38 complete, the next required maneuver was the instal-
lation of the D ring bridged lactone with inversion of the C6
stereogenic center. To this end, the triethylsilyl ether 38 was
cleaved with 5% HF in acetonitrile, and the resulting free alcohol
39 was treated with methanesulfonyl chloride to give mesylate
40 (Scheme 10). The mesylate 40 was dissolved in ethanol
followed by heating of the solution at reflux for 24 h. Water
was then added, and the mixture was heated for an additional
8 h. The addition of water was required since early experiments
indicated the presence of substantial amounts of ortho ester 41
in the crude product mixture. NMR analysis of the initial product
indicated the presence of the desired lactone 28 and the
hydroxyester 42. Subjection of this mixture to pyridinium
p-toluenesulfonic acid (PPTS) in benzene at reflux furnished
the bridged lactone 28, identical to that produced from the earlier
(30) Sharpless, K. B.; Lauer, R. F.; Teranishi, Y. J. Am. Chem. Soc.
1973, 95, 6137-6139.

8458 J. Am. Chem. Soc., Vol. 122, No. 35, 2000
Crimmins et al.
Scheme 11
Scheme 13
Scheme 14
Scheme 12
catalyzed addition of methanol. Formation of the methyl acetal
of enol ether 28 appears to be both kinetically disfavored
because of the highly hindered nature of the vinyl ether and
thermodynamically disfavored due to the steric interaction of
the C8 tert-butyl group and the C10 methylene upon rehybrid-
ization of the C10 carbon to sp3.
More encouraging results were obtained when enol ether 28
was found to undergo electrophilic addition with bromine and
methanol (Scheme 12). Subsequent incorporation of the C1-
C2 double bond produced the enone 49; however, all attempts
to effect rupture of the cyclobutane to afford enone 50 were
unsuccessful.
the vinyl ether occurs syn to the tert-butyl group, indicating
the remarkable steric shielding effected by the cyclopentenone
ring. Subjection of triol 52 to p-TsOH in methanol and
trimethylorthoformate gave 97% of a 5.5:1 mixture of bis(methyl
acetal) 53a and another anomer, 53b, believed to be the C12
diastereomer. The minor diastereomer 53b could be readily
separated and converted to 53a in 92% yield upon exposure to
6 N HCl in acetone. Upon successful cleavage of the cyclobu-
tane, a major milestone in the synthesis had been achieved. Four
of the six rings had been constructed with appropriately
positioned functionality for the installation of the final two rings.
After much experimentation, it was discovered that treatment
of vinyl ether 43 with dimethyldioxirane31 effected selective
epoxidation of the vinyl ether to deliver the epoxide 51, which
could be isolated, but was ordinarily processed in situ (Scheme
13). Addition of p-toluenesulfonic acid and a small amount of
water to the reaction mixture after epoxidation was complete
led to the direct formation of dialdehyde hydrate 52 from vinyl
ether 43 in excellent yield. Apparently, hydrolysis of the epoxide
leads to the hemiacetal which undergoes retroaldol fragmentation
of the cyclobutane, leading to a dialdehyde which forms the
stable hydrate 52. It is important to note that epoxidation of
As noted above, the original strategy required closure of the
E ring at this juncture of the sequence. Before completion of
the central E ring was attempted, oxidation of C4 was required.
Hydroxylation of C4 was accomplished under modified Davis
conditions (Scheme 14).32 Treatment of the â-keto ester 53a
with stoichiometric t-BuLi and catalytic Et2NH (20 mol %)
followed by addition of the Davis oxaziridine gave high yields
of the desired C4 alcohol 54. When stoichiometric amounts of
lithium diethylamide were used, competing oxidation of the
dialkylamine by the oxaziridine necessitated the use of excess
(32) Davis, F. A.; Vishwakarma, L. C.; Billmers, J. M.; Finn, J. J. Org.
Chem. 1984, 49, 3241-3243.
(31) Murray, R. W. Chem. ReV. 1989, 89, 1187-1201.

Total Synthesis of (()-Ginkgolide B
J. Am. Chem. Soc., Vol. 122, No. 35, 2000 8459
Scheme 15
Scheme 16
oxidant, which significantly hampered isolation and purification
of the tertiary alcohol 54.
With diol 54 in hand, only acid-catalyzed closure of the E
ring obstructed the completion of pentacycle 2, the proposed
precursor of ginkgolide B. While the formation of the E ring
acetal from hydroxy acetal 54 appeared uncomplicated, upon
treatment of acetal 54 with p-toluenesulfonic acid in dichlo-
romethane, an unexpected and devastating sequence of events
ensued. The resulting product, hydroxy ketone 56, is believed
to result from the ene-diol rearrangement of intermediate
hydroxy aldehyde 55 which originates from SN2-type E ring
closure with concomitant cleavage of the F ring and expulsion
of methanol. Further support for this proposal emanates from
the observation that acetate 57 results in the production of the
acetoxy aldehyde 58 when exposed to identical reaction
conditions. An attempt was made to reconstitute the F ring from
hydroxy acetal 56 since Corey and co-workers had successfully
prepared acetal 61 from the corresponding aldehyde, albeit under
fairly extreme conditions (see Scheme 15).13
Scheme 17
Exposure of hydroxy ketone 56 (or similar intermediates such
as the C11 carboxylic acid derived from aldehyde 58) to more
forcing conditions (e.g., p-TsOH in benzene at reflux) resulted
only in intramolecular conjugate addition to the C1-C2 enone
to afford the pentacycle 59 in 70% yield. While the Corey
intermediate was similar, the presence of the C3 carbonyl and
the C10 hydroxyl in intermediate 56 served to derail all efforts
to construct the pentacyclic acetal 2.
Two major issues with the original synthetic plan were
indisputable at this point. First, the unwanted ene-diol rear-
rangement, which had precluded closure of the F ring, was a
result of the hydroxyl at C10. Second, the C1-C2 enone was
also interfering with the F ring formation due to its electrophilic
nature and its proximity to the C10 and C11 functional groups.
The former problem was addressed by attempting to remove
the C10 hydroxyl. To obtain reduction product 63, alcohol 53a
was first converted to the corresponding xanthate in 91% yield
upon treatment with CS2 and MeI in the presence of DBU
(Scheme 16).33 An X-ray crystal structure of xanthate 62 served
to confirm the relative stereochemistry shown. When xanthate
62 was treated with tributyltin hydride and AIBN in benzene
at reflux for 10 min, 43% of the desired reduction product 63
was obtained.34 Davis oxidation under unoptimized conditions
gave a 1.6:1 mixture of the desired alcohol 64 and recovered
starting material 63. Exposure of the hydroxy acetal 64 to CSA
in methanol gave a 1:1 mixture of aldehyde 65 and hexacyclic
cage 67, the latter of which presumably arose from intramo-
lecular hemiacetal attack on the enone in 66.
F rings concomitantly without addition at C1, the electronics
of the C1-C2 alkene would have to be modified to reduce its
electrophilicity. In this regard, altering the C3 carbonyl offered
an attractive option: if an intramolecular aldol addition to the
C3 carbonyl could be effected through an enolate attached at
the C4 hydroxyl, not only could the C1-C2 alkene electrophi-
licity be attenuated, but stereoselective addition at the C3
carbonyl could also be accomplished.
The synthetic plan was revised to incorporate the modified
strategy of using an intramolecular aldol reaction to deactivate
the A ring enone (Scheme 17). Specifically, ginkgolide B was
thought to be accessible from lactone 68 via epoxidation of the
C1-C2 alkene, C and E ring formation, and functional
modification of the F ring. Lactone 68 would be formed from
69 via an intramolecular aldol reaction which would not only
remove the enone as a reactive site, but also set the C3
stereochemistry. Bisacetate 69 would be obtained from the
corresponding diol 54.
The requisite diacetate 69 was readily prepared by treatment
of diol 54 with acetic anhydride and triethylamine to give 69
in 48% yield for two steps (hydroxylation and acetylation,
Scheme 18). Subjection of bisacetate 69 to lithium diisopropy-
lamide at -78 °C gave high yields of the desired intramolecular
aldol product 68. A fortuitous discovery was made while to
selectively hydrolyzing the C11 methyl acetal was attempted.
It was determined that treatment of methyl acetal 68 with boron
tribromide in dichloromethane resulted in exclusive formation
Although this approach was temporarily abandoned, it
revealed two important clues to the ultimate synthetic solution.
First, in the absence of the C10 hydroxyl, the E and F rings
could be formed concurrently. Second, to incorporate the E and
(33) Paquette, L. A.; Sauer, D. R.; Cleary, D. G.; Kinsella, M. A.;
Blackwell, C. M.; Anderson, L. G. J. Am. Chem. Soc. 1992, 114, 7375-
7387.
(34) Barton, D. H. R.; McCombie, S. W. J. Chem. Soc., Perkin Trans.
I 1975, 1574-1585.

8460 J. Am. Chem. Soc., Vol. 122, No. 35, 2000
Crimmins et al.
Scheme 18
Scheme 19
subsequent acidification could produce the desired sequence of
events in a single synthetic operation. The product of this
reaction, trilactone 70 (Scheme 17), would be a methylation
(at C14) away from ginkgolide B. Unfortunately, when lactone
75 was treated with aqueous KOH in THF followed by
acidification with HCl, none of the desired product was
observed. Instead, two acids, 76 and 77, were isolated: acid 77
was believed to be the result of attack of the C10 hydroxyl at
C1 of the epoxide, while acid 76 was the result of epoxide
opening at C1 by the C12 methoxy group. A more stepwise
approach was clearly indicated.
With this in mind, allylic alcohol 72 underwent a directed
epoxidation under Sharpless conditions [VO(acac)2, t-BuOOH]36
to give â-epoxide 78 in 95% yield (Scheme 19). It was then
found that treatment of bislactone 78 with methylamine in
methanol resulted in exclusive reaction at the C15 carbonyl to
give 70% of amide 79. It was thought that subjection of the
amide 79 to acidic conditions would result in attack of the amide
carbonyl on the C2 position of the epoxide which, after aqueous
workup, would give the C ring lactone. Furthermore, under the
acidic reaction conditions, the liberated C4 tertiary hydroxyl
could participate in construction of the E ring, giving desired
product 70. However, yet another unexpected result occurred.
Upon exposure of epoxyamide 79 to CSA in dichloromethane,
the protonated epoxide was attacked, not by the amide carbonyl,
but rather by the C12 methoxy group, which appears to be
perfectly oriented with the C1 C-O antibonding orbital of the
epoxide. The ensuing oxonium ion is captured by the C4
hydroxyl, resulting in methyl ether 80.37
of the C11 bromide even in the presence of a large excess of
reagent. The low reactivity of the C12 acetal center is presum-
ably related to its extraordinarily hindered nature because of
its position in the cavity between the A and F rings. Subsequent
treatment of the unstable C11 bromide with activated zinc in
1,2-dimethoxyethane effected reductive elimination to the vinyl
ether 72 (76% yield from 68). The enol ether 72 was subjected
to excess dimethyldioxirane in acetone followed by addition of
water and p-TsOH. The hemiacetal 74 that was produced (78%
yield) arose from initial epoxidation of the electron-rich F ring
vinyl ether with subsequent epoxidation of the C1-C2 alkene
to afford epoxide 73. The stereoselectivity for this first epoxi-
dation is believed to be the result of steric hindrance by the
tert-butyl group, which effectively shields one face of the enol
ether. The R-epoxide, combined with the F ring methoxy group,
then blocks the R-face of the A ring olefin, directing the second
epoxidation to the â-face of the A ring, giving 73. Hydrolysis
of the more reactive C10-C11 epoxide to the hemiacetal 74
ensued. Selective oxidation of the hemiacetal 74 with Br2 and
acetic acid35 provided the R-hydroxy lactone 75 in 64% yield.
An important confluence in the strategy had been reached. All
atoms with the exception of the C14 methyl had been incor-
porated, and each was in the oxidation state identical to the
ultimate objective. Completion of the fully elaborated ginkgolide
skeleton required three events: hydrolysis of the lactone derived
from the intramolecular aldol, formation of the C ring lactone
by attack of the liberated carboxylate (or carboxylic acid) at
C2 of the epoxide, and acid-catalyzed closure of the E ring. An
aggressive approach to complete the synthesis was envisioned.
It was anticipated that basic hydrolysis of lactone 75 and
While the desired outcome was not obtained in the reaction
of 79 with acid, the product 80 did contain some interesting
(36) Sharpless, K. B.; Verhoeven, T. R. Aldrichimica Acta 1979, 12,
63-74.
(37) The structure of 80 is supported by NOE studies.
(35) Corey, E. J.; Kamiyama, K. Tetrahedron Lett. 1990, 31, 3995-
3998.

Total Synthesis of (()-Ginkgolide B
J. Am. Chem. Soc., Vol. 122, No. 35, 2000 8461
Scheme 20
Scheme 21
centers from sp3 to sp2 had altered the steric effects of the
concave face of the molecule. A similar observation was made
during the Corey synthesis in the addition of tert-butyl propi-
onate enolate to the C3 carbonyl. A Sharpless-directed epoxi-
dation36 on the allylic alcohol 84 was attempted, but failed
apparently due to the sensitivity of the ketolactone.
Attention now turned to preventing the ene-diol rearrange-
ment to obtain usable quantities of pentacycle 83 or a similar
species. There were three primary paths by which the ene-
diol rearrangement might be prevented. First, if the C10
hydroxyl of 54 were oxidized to the corresponding ketone or,
second, if the C10 hydroxyl were protected with a base-stable
protecting group, ene-diol rearrangement would not be possible.
This latter solution is complicated by the hindered nature of
the C10 hydroxyl which severely limits the protecting groups
that can be incorporated. The third option to prevent the ene-
diol rearrangement was to remove the C10 hydroxyl altogether.
Ketone 86 (Scheme 21) was treated with NaOH followed by
acidification and esterification of the carboxylic acid with
diazomethane, but ester 87 was the only isolated product. Upon
exposure to acid, alcohol 87 was converted back to 86.
All attempts to protect the C10 hydroxyl as the benzyl ether,
mesylate, or methyl ether at various stages all failed, and once
again efforts focused on excision of the C10 hydroxyl. Deoxy-
genation of alcohol 53a had been previously accomplished in
modest yield by reduction of xanthate 62 with tributyltin hydride
and AIBN in benzene at reflux, but this reaction was somewhat
capricious. An extensive survey of solvents, hydrogen atom
donors, initiators, and reaction temperatures was conducted. The
most consistent results and highest yields (70-75%) were
obtained when the reaction was carried out with tributyltin
hydride and AIBN in benzene at 60 °C (Scheme 22).40 The
reduction product 63 was oxidized under modified Davis
conditions as described earlier and the resulting C4 hydroxyl
converted to the corresponding acetate 88 (Scheme 22). In-
tramolecular aldol addition gave lactone 89 which was then
converted to hemiacetal 90 via treatment with boron tribromide
followed by Ag2CO3. Exposure of 90 to NaOH effected the
base-induced lactone hydrolysis and subsequent cyclization of
the E and F rings. After acidic workup, esterification of the
carboxylic acid with diazomethane, and exposure to methanol,
p-TsOH, and trimethylorthoformate, the methyl acetal 91 was
obtained in good overall yield from 89. A significant milestone
had been achieved: rings A, B, D, E, and F were in place with
suitable functionality to complete the synthesis. Only completion
of the C ring and adjustment of the oxidation states of C10 and
C11 remained. To this end, exposure of acetal 91 to PPTS and
pyridine in chlorobenzene at reflux effected elimination of
methanol to give vinyl ether 92 in 83% yield. Subjection of
allylic alcohol 92 to the Sharpless-directed epoxidation protocol
delivered the epoxide 93 in good yield. Exposure of epoxide
93 to p-TsOH in CH2Cl2 completed the final ring of ginkgolide
B, the C ring, to give lactone 94 (72% for two steps). All that
remained was to methylate the C14 position and oxidize the F
features. Although 80 had the incorrect stereochemistry at C1
and C2 and also contained an unmanageable methyl ether at
C1, the A, B, D, E, and F rings were in place at the same time
for the first time in the synthesis. In an effort to capitalize on
the observed rearrangement of amide 79 to the pentacycle 80,
extensive efforts to incorporate a different protecting group at
C12 other than a methyl ether were undertaken, but none of
the attempts were brought to completion.
Due to the serious complications encountered with undesired
reactivity of the epoxide with the C12 methoxy and the C10
hydroxyl, it became imperative to complete the E ring prior to
incorporation of the C1-C2 epoxide. An attractive option was
to exploit the ene-diol rearrangement which had thwarted early
attempts at E ring formation, but in a slightly different manner.
Bromoacetal 71 was known to be relatively unstable and slowly
hydrolyze to the corresponding hemiacetal 81. The hydrolysis
could be accelerated, as shown in Scheme 20, upon exposure
of the bromoacetal 71 to Fetizon’s reagent.38 Subjection of
hemiacetal 81 to aqueous base followed by an acidic workup
gave a mixture of two carboxylic acids which were immediately
esterified with diazomethane. The primary alcohol 82 was
obtained in 70% yield accompanied by 15% of the hemiacetal
83. In the absence of the C1-C2 epoxide, the base-induced
hydrolysis of the lactone derived from the intramolecular aldol
with subsequent closure of the E ring had been realized in the
form of pentacyclic intermediate 83. The major product,
hydroxyketone 82, is the obvious result of an ene-diol
rearrangement of the intermediate C11 aldehyde. In an effort
to utilize the hydroxyketone 82, it was treated with 2.5 equiv
of the Dess-Martin periodinane39 to provide excellent yields
of the ketolactone 84. Once again, the ultimate goal seemed
imminent. Unfortunately, when the R-ketolactone 84 was
epoxidized with dimethyldioxirane, the epoxide 85 was obtained.
Apparently, a change in hybridization of the C10 and C11
(38) McKillop, A.; Young, D. W. Synthesis 1979, 401. Fe´tizon, M. In
Encyclopedia of Reagents for Organic Synthesis; Paquette, L. A., Ed.; John
Wiley and Sons: New York, 1995; Vol. 6, pp 4448-4454.
(39) Dess, D. B.; Martin, J. C. J. Am. Chem. Soc. 1991, 113, 7277-
7287.
(40) Curran, D. P. Synthesis 1988, 417-439.

8462 J. Am. Chem. Soc., Vol. 122, No. 35, 2000
Crimmins et al.
Scheme 22
Scheme 23
methyl epimer. When the HMPA is added and coordination of
the lithium is minimized, the enolate is free to rotate away from
the F ring, and the new orientation is such that the same,
undesired methyl epimer is obtained. The inability to control
the C14 stereogenicity at this stage was not overly troublesome
since there was ample opportunity to correct this stereocenter
via epimerization at a later stage, especially after the hemispheric
parent ginkgolide skeleton was in place. The C14 methyl group
would be positioned on the more sterically compressed concave
face of the hemisphere and therefore would be expected to
epimerize readily.
In this regard, acetal 97 was converted as before with the
normethyl derivative to hemiacetal 98 followed by base-induced
rearrangement, esterification, and methyl acetal formation. At
the end of this sequence, which was performed in rapid
succession, a 4:1 mixture of methyl epimers 99:100 was
obtained. Apparently, some epimerization had occurred during
the base-induced reorganization of the skeleton. Nonetheless,
the mixture was carried forward whereupon treatment with PPTS
and pyridine in chlorobenzene at reflux gave vinyl ethers 101
and 102. Epoxidation with VO(acac)2 and t-BuOOH resulted
in epoxides 103 and 104. Finally, treatment of the mixture of
103 and 104 with p-TsOH yielded a 4:1 mixture of six-
membered ring, lactone 105 and five-membered ring lactone
106. The undesired C14 methyl epimer 103 had undergone
cyclization at C1 to form the six-membered ring while the
desired C14 methyl epimer 104 cyclized at the C2 position to
form the requisite five-membered ring. As discussed above, in
the undesired C14 methyl epimer 103, the transition state for
ring. The alkylation of 94 with MeI was attempted under a
variety of conditions with no indication of success; instead, only
recovered starting material was observed. Methylation of the
intramolecular aldol product 89 was also attempted, but again,
no alkylation product was observed.
An obvious alternative to methylation at the final stage was
to incorporate the additional carbon prior to the intramolecular
aldol, that is, utilize the propionate instead of the acetate in the
intramolecular aldol reaction. Propionate 96 was obtained from
alcohol 64 via treatment with propionic anhydride and triethy-
lamine (Scheme 23). Intramolecular aldol addition of the
propionate to the A ring carbonyl gave the lactone 97 with the
undesired C14 stereochemistry (Scheme 23). When the enolate
geometry was reversed through addition of HMPA,41 the same
diastereomer was obtained. It is postulated that, in the absence
of HMPA, the Li is coordinated to the enolate as well as the F
ring oxygens, thus orienting the enolate to give the undesired
(41) Ireland, R. E.; Mueller, R. H.; Willard, A. K. J. Am. Chem. Soc.
1976, 98, 2868-2877.

Total Synthesis of (()-Ginkgolide B
J. Am. Chem. Soc., Vol. 122, No. 35, 2000 8463
Scheme 24
of starting lactone 97 and its C14 methyl epimer 107. Prolonged
exposure to base improved the ratio to 1:1, but the yield suffered
due to competitive â-elimination. After the diastereomers were
separated via flash chromatography, lactone 97 could be
resubjected to the NaOMe/MeOH reaction conditions to provide
the required diastereomer 107 in 50% yield after one recycle.
Additional recycles allowed access to higher yields.
With the C14 stereogenic center corrected, the final resolution
to the seemingly endless conundrum had been identified. In the
course of the late stages of the investigation, an additional
important observation was made. It was also discovered that
rather than forming the C11 hemiacetal and effecting a base-
catalyzed cyclization as in the conversion of acetal 97 to 99,
simple subjection of lactone 107 to CSA in methanol at reflux
produced the rearranged acetal 100 in high yield, thus obviating
four synthetic steps. The acid-catalyzed reorganization was only
possible in the absence of the C10 hydroxyl since there was no
possibility for the derailment by the ene-diol rearrangement.
Methyl acetal 100 then underwent elimination with PPTS and
pyridine to provide enol ether 102 in 85% yield. Directed
epoxidation under Sharpless conditions gave the desired epoxide
which, upon in situ exposure to p-TsOH, cyclized to form the
C ring lactone 106. Treament of vinyl ether 106 with dimeth-
yldioxirane resulted in exclusive formation of the epoxide 108
(Scheme 24). The selectivity in this reaction is believed to result
from the fact that the tert-butyl group guards the R-face of the
vinyl ether. Finally, treatment of epoxide 108 with bromine and
acetic acid35 resulted in epoxide opening to the corresponding
hemiacetal, which was followed by an in situ oxidation to
complete the F ring lactone in 52% yield from 106. The
synthesis of ginkgolide B was complete as evidenced by
comparison of the synthetic material to an authentic sample.
The samples provided identical 1H and 13C NMR, infrared
spectroscopic data, and chromatographic properties (TLC). The
final synthetic route required 25 linear steps from 3-furaldehyde.
epoxide opening places the methyl group in the concave,
sterically congested cavity of the molecule. It had been hoped
to exploit this unfavorable interaction to effect epimerization
of the C14 center after lactonization. Unfortunately, this same
steric interference resulted in the cyclization of ester 103 in an
unanticipated manner to form lactone 105. On the positive side,
the cyclization of ester 104, containing the natural configuration
at C14, proceeded to give the desired product 106. It was clear
that epimerization would have to be accomplished prior to the
epoxy lactonization.
Acknowledgment. Financial support of our programs by the
NIH, the NSF, the donors of the Petroleum Research Fund of
the American Chemical Society, and the Randleigh Foundation
is acknowledged with thanks. Thanks are also due to Glaxo-
Wellcome for help with the mass spectral data.
Supporting Information Available: Experimental proce-
dures and spectral data (PDF). This material is available free
of charge via the Internet at http://pubs.acs.org.
Fortunately, it was soon discovered that exposure of lactone
97 to NaOMe in MeOH resulted in a 1.4:1 separable mixture
JA001747S