Stereoselective syntheses of epothilones A and B via nitrile oxide cycloadditions and related studies

Article abstract of DOI:10.1021/jo015791h

The expedient and fully stereocontrolled synthesis of epothilones A and B are described. The routes described make extensive study of nitrile oxide cycloadditions as surrogates for aldol addition reactions and have led to the realization of a highly convergent synthesis based on the Kanemasa hydroxyl-directed nitrile oxide cycloaddition. As well, our synthetic efforts have led to the development of new reaction methodologies and served as the proving ground for several modern methods for asymmetric carbon-carbon bond formation.

Full text of DOI:10.1021/jo015791h

6410
J . Org. Chem. 2001, 66, 6410-6424
Ster eoselective Syn th eses of Ep oth ilon es A a n d B via Nitr ile Oxid e
Cycloa d d ition s a n d Rela ted Stu d ies
J effrey W. Bode and Erick M. Carreira*
Laboratorium fu¨r Organische Chemie, ETH-Zu¨rich, Universita¨tstrasse 16, CH-8092 Zu¨rich/ Switzerland
Received May 24, 2001
The expedient and fully stereocontrolled synthesis of epothilones A and B are described. The routes
described make extensive study of nitrile oxide cycloadditions as surrogates for aldol addition
reactions and have led to the realization of a highly convergent synthesis based on the Kanemasa
hydroxyl-directed nitrile oxide cycloaddition. As well, our synthetic efforts have led to the
development of new reaction methodologies and served as the proving ground for several modern
methods for asymmetric carbon-carbon bond formation.
Ch a r t 1
Ba ck gr ou n d a n d In tr od u ction
The epothilone natural products (1a -f) are structur-
ally unique polyketide macrolides which share with Taxol
the ability to induce tubulin polymerization and associ-
ated cellular effects (Chart 1). This extraordinary biologi-
cally activity has led to the development of anticancer
chemotherapeutics, as evidenced by the phenomenal
success of Taxol as a clinical drug. The inherent limita-
tions of Taxol, in particular the emergence of cancer cell
lines resistant to Taxol, has placed the identification and
study of alternatives at the forefront of chemistry and
biology. It is therefore no surprise that the discovery of
the epothilones and the recognition of their potent
biological activity has prompted interest in the synthesis
and pharmacological study of these exciting molecules.1
The epothilones were first isolated from cultures of
Sorangium cellulosum collected from the banks of the
Zambesi river in South Africa, by Ho¨fle and co-workers
at the Gesellschaft fu¨r Biotechnologische Forschung
(GBF-Braunschweig) in 1987.2 Although the potent
cytotoxicity of these molecules was recognized and pre-
liminary investigations made, they were deemed too toxic
for further study.3 It was not until 1995 that Bollag and
co-workers rediscovered the epothilones using a newly
developed assay to screen for molecules which exhibit
tubulin polymerization activity that significant interest
in the epothilones was generated.4 Although the epo-
thilones were the first nontaxoid molecules for which this
unique mode of action was identified, a number of other
molecules which bind to the same or overlapping binding
site on the microtubules have been subsequently identi-
fied. In addition to the diterpenoids eleutherobin and
sarcodicytin,5 the polyketide natural products discoder-
molide (2)6 and laulimalide (3)7 show similar biological
activity to the epothilones. As well, rhizoxin (4) is a
complex polyketide macrolide known to bind to tubulin
but which acts as an inhibitor of microtubule assembly
8
rather than as a promoter (Chart 2).
Although the gross structure of the epothilones had
been independently deduced by both Ho¨fle and by Bollag,
serious synthetic efforts toward these molecules had to
wait for the disclosure of the relative and absolute
stereochemistry. Initial progress toward the full structure
determination of the epothilones was reported by Georg,9
but the final structure was obtained by X-ray diffraction
analysis of epothilone B by Ho¨fle and co-workers.2a Taken
together, these studies identified the structurally unique
features of the epothilones including their abundance of
functional groups. The epothilones also possess several
uncommon structural features including the C4 gem-
dimethyl group, the thiazole moiety, a cis-epoxide, and
the C9-C11 spacer region which effectively divides these
molecules into two discrete regions of stereochemical and
functional complexity.
As is often the case in natural products synthesis the
unique structure and important biological activity con-
(1) For reviews, see: (a) Mulzer, J . Monatshefte Chemie 2000, 131,
205-238. (b) Nicolaou, K. C.; Roschangar, F.; Vourloumis, D. Angew.
Chem., Int. Ed. Engl. 1998, 37, 2014-2045. (c) Altmann, K. H.; Bold,
G.; Caravatti, G.; End, N.; Florsheimer, A.; Guagnano, V.; O′Reilly,
T.; Wartmann, M. Chemia 2000, 54, 612-621.
(2) (a) Ho¨fle, G.; Bedorf, N.; Steinmetz, H.; Schomburg, D.; Gerth,
K.; Reichenback, H. Angew. Chem., Int. Ed. Eng. 1996, 35, 1567-1569.
(b) Gerth, K.; Bedorf, N.; Ho¨fle, G.; Irschik, H.; Reichenback, H. J .
Antibiot. 1996, 49, 560-563.
(3) Ho¨fle, G.; Bedorf, N.; Gerth, K.; Reichenback, H. (GBF), DE-
4138042, 1993 [Chem. Abstr. 1993, 119, 180598].
(5) The diterpenoids eleutherobin and sarcodicytin natural products
with similar activity have also been recently identified: Lindel, T.;
J ensen, P. R.; Fenical, W.; Long, B. H.; Casazza, A. M.; Carboni, J .
Fairchild, C. R. J . Am. Chem. Soc. 1997, 119, 8744-8745.
(6) (a) ter Haar, E.; Kowalski, R. J .; Hamel, E.; Lin, M. C.; Longley,
R. E.; Gunaskera, S. P.; Rosenkranz, H. S.; Day, B. W. Biochem. 1996,
35, 243-250. (b) Hung, D. T.; Chen, J .; Schreiber, J . Chem. Biol. 1996,
3, 287-293.
(7) Mooberry, S. L.; Tien, G.; Hernandez, A. H.; Plubrukarn, A.;
Davidson, B. S. Cancer Res. 1999, 653.
(8) Takahashi, M.; Iwasaki, S.; Kobayashi, H.; Okuda, S.; Murai,
T. Sato, Y. Biochim. Biophys. Acta 1987, 926, 215-223.
(9) Victory, S. F.; Velde, D. G. V.; J alluri, R. K.; Grunewald, G. L.;
Georg, G. I. Bioorg. Med. Chem. Lett. 1996, 6, 893-898.
(4) Bollag, D. M.; McQueney, P. A.; Zhu, J .; Hensens, O.; Koupal,
L.; Liesch, J .; Goetz, M.; Lazarides; Woods, C. M. Cancer Res. 1995,
55, 5, 2325-2333.
10.1021/jo015791h CCC: $20.00 © 2001 American Chemical Society
Published on Web 08/23/2001

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6411
Ch a r t 2
necessary subunits, their convergent coupling, macro-
cyclization, and stereoselective epoxidation. The second
approach, pioneered by Danishefsky10,22 and continued
by White,23 Schinzer,24 Shibasaki,25 and Panek,26 recog-
nized a disconnection at or near the C12-C13 cis-olefin of
epothilones C and D to achieve a convergent coupling of
two stereochemically advanced fragments. The conver-
gent aldol approach, pioneered by Nicolaou27 and by
Mulzer,28,29 takes advantage of the full aldol retron at
C5-C7 of the epothilone backbone and joins the appropri-
ate fragments at their point of greatest functional and
stereochemical complexity. The various approaches un-
derscore the persistent limitations of modern asymmetric
synthesis that demand a balance between convergency
and efficient stereochemical control. A major current
objective in epothilone syntheses is the development of
a strategy to achieve a concise, convergent, and fully
stereocontrolled approach.
We now report a full account of our studies toward this
goal, culminating in the expedient and stereoselective
syntheses of epothilones A and B.30 Critical to the success
of this endeavor has been the development and applica-
tion of new methodologies and strategies suitable for the
construction of functionally and stereochemically complex
molecules. Our efforts have also focused on the explora-
tion and advancement of the use of nitrile oxides as
surrogates for aldol addition reactions.
spired to make the epothilones attractive targets for total
synthesis. Within six months of the first disclosure of the
relative stereochemistry of the epothilones the Danishef-
sky,10 Nicolaou,11 and Schinzer12 groups each indepen-
dently completed the total synthesis of epothilone A and,
shortly thereafter, epothilone B.13,14 The impressive ef-
forts toward their total synthesis, both by academic
research groups as well as by pharmaceutical compa-
nies,15,16,17 have resulted in a broader understanding of
the role of the functional and stereochemical features
leading to the physiological effects of the epothilones.
These studies, which include both analogue synthesis as
well as elegant degradation studies, have made the
epothilones the subject of intense synthetic scrutiny, and
several promising analogues have been prepared.15,18
The published synthetic routes to the epothilones have
been reviewed.1 In general, the published syntheses fall
into one of three approaches. The macrocyclic ring-closing
metathesis approach has been reported by Nicolaou,11
Schinzer,24a Danishefsky,19 Grieco,20 and Lerner.21 This
strategy takes advantage of the rapid assembly of the
Syn th etic P la n
Our initial synthetic planning of routes for the
epothilones benefited from early reports highlighting the
unique difficulties in controlling the stereochemistry of
the critical polypropionate region. Although we consid-
ered a convergent aldol coupling at C5-C7 to be an
attractive synthetic disconnection, the first generation
syntheses that had been reported documented the dif-
ficulties in stereoselectively accomplishing the necessary
anti-Felkin aldol addition.31 Likewise, although all of the
initial approaches had employed a stereoselective epoxi-
dation reaction as the final step, we believed that early
(19) Meng, D.; Su, D.-S.; Balog, A.; Bertinato, P.; Sorensen, E. J .;
Danishefsky, S. J .; Zheng, Y.-H.; Chou, T.-C.; He, L.; Horwitz, S. B. J .
Am. Chem. Soc. 1997, 119, 2733-2734.
(10) (a) Balog, A.; Meng, D.; Kamenecka, T.; Bertinato, P.; Su, D.-
S.; Sorensen, E. J .; Danishefsky, S. J . Angew. Chem., Int. Ed. Engl.
1996, 35, 2801-2803. (b) Meng, D.; Su, D.-S.; Balog, A.; Bertinato, P.;
Sorensen, E. J .; Danishefsky, S. J .; Zheng, Y. H.; Chou, T. C.; He, L.;
Horwitz, S. B. J . Am. Chem. Soc. 1997, 119, 2733.
(20) May, S. A.; Grieco, P. A. Chem. Commun. 1998, 1597.
(21) (a) Sinha, S. C.; Barbas, C. F., III; Lerner, R. A. Proc. Natl.
Acad. Sci. U.S.A. 1998, 95, 14603. (b) Sinha, S. C.; Sun, J .; Miller, G.
P.; Wartmann, M.; Lerner, R. A. Chem. Eur. J . 2001, 7, 1691-1702.
(22) (a) Balog, A.; Harris, C.; Savin, K.; Zhang, X. G.; Chou, T.-C.;
Danishefsky, S. J . Angew. Chem., Int. Ed. Engl. 1998, 2675, 5-2678.
(b) Harris, C. R.; Kuduk, S. D.; Balog, A.; Savin, K.; Glunz, P. W.;
Danishefsky, S. J . J . Am. Chem. Soc. 1999, 121, 7050-7062.
(23) (a) White, J . D.; Carter, R. G.; Sundermann, K. F. J . Org. Chem.
1999, 64, 684-685. (b) White, J . D.; Sundermann, K. F.; Carter, R. G.
Org. Lett. 1999, 1, 1431-1434.
(11) Yang, Z.; He. Y.; Vourloumis, D.; Vallberg, H.; Nicolaou, K. C.
Angew. Chem., Int. Ed. Engl. 1997, 36, 170-172.
(12) Schinzer, D.; Limberg, A.; Bauer, A.; Bo¨hm, O. M.; Cordes, M.
Angew. Chem., Int. Ed. Engl. 1997, 36, 543-544.
(13) Su, D.-S.; Balog, A.; Meng, D.; Bertinato, P.; Danishefsky, S.
J .; Zheng, Y.-H.; Chou, T.-C.; He, L.; Horwitz, S. B. Angew. Chem.,
Int. Ed. Engl. 1997, 36, 2093-2096.
(14) Nicolaou, K. C.; Pastor, J .; Wissinger, N.; He., Y.; Ninkovic, S.;
Sarabia., F.; Vouloumis, D.; Yang, Z.; Li, T.; Giannakakou, P.; Hamel,
E. Nature 1997, 36, 757-759.
(24) (a) Schinzer, D.; Bauer, A.; Bo¨hm, O. M.; Limberg, A.; Cordes,
M. Chem. Eur. J . 1999, 5, 2483-2491 (b) Schinzer, D.; Bauer, A.
Schieber, J . Chem. Eur. J . 1999, 5, 2492-2500.
(25) (a) Sawanda, D.; Shibasaki, M. Angew. Chem., Int. Ed. Engl.
2000, 39, 209-212. (b) Sawada, D.; Kanai, M.; Shibasaki, M. J . Am.
Chem. Soc. 2000, 122, 10521-10532.
(15) Bristol-Myers-Squibb: (a) J ohnson, J .; Kim, S. H.; Bifano, M.;
DiMarco, J .; Fairchild, C.; Gougoutas, J .; Lee, F. Long, B.; Tokarski,
J .; Vite, G. Org. Lett. 2000, 2, 1573. (b) Borzilleri, R. M.; Zheng, X.;
Schmidt, R. J .; J ohnson, J . A.; Kim, S. H.; DiMarco, J . D.; Fairchild,
C. R.; Gougoutas, J . Z.; Lee, F. Y. F.; Long, B. H.; Vite, G. D. J . Am.
Chem. Soc. 2000, 122, 8890-8897.
(26) Zhu, B.; Panek, J . Org. Lett. 2000, 2, 2575-2578.
(27) (a) Nicolaou, K. C.; He, Y.; Vourloumis, D.; Vallber, H.;
Roschangar, F.; Sarabia, F.; Ninkovic, S.; Yang, Z.; Trujillo, J . I. J .
Am. Chem. Soc. 1997, 119, 7960-7973. (b) Nicolaou, K. C.; Ninkovic,
S.; Sarabia, F.; Vourloumis, D.; He, Y.; Vallberg, H.; Finlay, M. R. V.;
Yang, Z. J . Am. Chem. Soc. 1997, 119, 7974-7991.
(16) Novartis: (a) ref 1c. (b) Altmann, K.-H.; Caravatti, G.; Floer-
sheimer, A.; Wartmann, M. Nicolaou, K. C.; Schinzer, D. Book of
Abstracts, 219th National Meeting of the American Chemical Society,
San Francisco, CA, March 26-30, 2000; American Chemical Soceity:
Washington, DC, 2000; Abstract 287.
(17) Schering AG: Klar, U.; Skuballa, W.; Schwede, W.; Buchmann,
B. Book of Abstracts, 219th National Meeting of the American Chemical
Society, San Francisco, CA, March 26-30, 2000; American Chemical
Soceity: Washington, DC, 2000; Abstract 286.
(28) (a) Mulzer, J . Mantoulidis, A.; O¨ hler, E. Tetrahedron Lett. 1998,
39, 8633. (b) Mulzer, J . Mantoulidis, A.; O¨ hler, E. J . Org. Chem. 2000,
65, 7456-7467.
(29) Martin, H. J .; Drescher, M.; Mulzer, J . Angew. Chem., Int. Ed.
2000, 39, 581-583.
(30) A preliminary report of this work has appeared: Bode, J . W.;
Carreira, E. M. J . Am. Chem. Soc. 2001, 124, 3611-3612.
(18) Harris, C. R.; Danishefsky, S. J . J . Org. Chem. 1999, 64, 8434.

6412 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
Sch em e 1
Sch em e 2
introduction of the epoxide functionally would be advan-
tageous.
In itia l Ap p r oa ch . Closer analysis of the epothilone
structure revealed the potential use of a nitrile oxide
cycloaddition approach to be particularly promising.32 In
analogy to strategies employing aldol couplings, the use
of a nitrile oxide cycloaddition would lead to a discon-
nection of the epothilones at their point of greatest
stereochemical and functional complexity, and a single
reaction step would suffice for the construction of two
bonds, two stereocenters, and two functional groups in
the correct oxidation state (Scheme 1). This was predi-
cated on the known transformation of isoxazolines to
â-hydroxy ketones via N-O bond reduction and imine
hydrolysis. Importantly, the necessary precursors, a cis-
olefin and a nitrile oxide, would be readily accessible and
amenable to concise, convergent synthetic routes.
This use of a strategy based on nitrile oxide cycload-
dition, however, was not without considerable concerns.
Although its successful implementation would undoubt-
edly lead to convergency, issues of regioselectivity,
chemoselectivity, and reactivity of the unactivated cis-
double bond would have to be addressed. Although it is
true that unactivated di- and trisubstituted olefins are
notoriously poor reaction partners in dipolar cycloaddi-
tions, this limitation typically stems from the tendency
of the nitrile oxide to undergo competing decomposition
or dimerization in preference to cycloaddition. The non-
productive decomposition pathways depend on the sterics
associated with the nitrile oxides; hindered nitrile oxide
are known to be ideal reaction partners even with
unactivated olefins. Because we would require a tertiary
nitrile oxide at C5, these precedents offered considerable
promise for the application of this methodology for
epothilone synthesis. The use of cis-olefins also presents
questions of regiochemistry, particularly if they are not
substituted by electronically activating groups. However,
we found promising precedent for steric control of this
reaction in Martin’s report that the reaction of a similarly
substituted cis-olefin and a tertiary nitrile oxide provided
the cycloadduct as a single regioisomer.33
however, an additional site of unsaturation where the
nitrile oxide could potentially react. We anticipated the
trisubstituted olefin to be considerably less reactive than
the requisite cis-olefin, given the reported chemoselective
oxidation of a cis-olefin in the presence of the thiazole
side chain which enabled the late-stage epoxidation of
epothilones C and D to form epothilones A and B.1
However, we felt that we could further influence the
chemoselectivity, regioselectivity, and ultimately stereo-
selectivity of the cycloaddition by effecting a macrocyclic,
ring-closing, nitrile oxide cycloaddition (Scheme 2). Im-
portantly, a number of elegant applications of macrocyclic
nitrile oxide cycloadditions of simpler systems in remark-
ably good yields and with complete regiochemical control
34
further persuaded us to pursue such a route.
These considerations and precedents led to our first
retrosynthetic analysis of the epothilones. In addition to
providing an alternative to the problematic aldol reaction,
we felt the use of an intramolecular nitrile oxide approach
limited superfluous protection group manipulation and
divided the epothilone structure into two accessible
fragments, designated as the C1-C5 fragment (8) and the
C6-C15 fragment (7). With these two subunits as targets,
we initiated our synthetic efforts toward the epothilones.
Although this initial approach proved untenable, the
route employed for the fragment synthesis proved useful
in developing a successful and convergent strategy to the
epothilones. Indeed, the chemistry we developed in the
synthesis of the C6-C15 fragment in the context of the
macrocyclic nitrile oxide cycloaddition approach was
directly applicable to an alternative and fully stereocon-
trolled route. Following the landmark disclosure by
Mulzer that introduction of the C12-C13 epoxide prior to
a convergent aldol reaction resulted in the successful
realization of a highly diastereoselective fragment cou-
pling, we readily adapted our initial route to take
advantage of this development.29 Further refinements in
our synthetic approach ultimately led to the total syn-
thesis of epothilone A (21 steps) and a formal synthesis
of epothilone B (18 steps in total) based on Mulzer’s
stereocontrolled convergent aldol approach (Scheme 3).
These routes, as well as our studies toward the macro-
cyclic nitrile oxide approach, make extensive use of new
reaction methodologies.
In hopes of achieving a convergent approach to the
epothilones, we planned to introduce the thiazole side
chain early in the synthetic route. This presented,
(31) For a more detailed analysis of the challenges in controlling
the stereochemical outcome of this aldol reaction, see Wu, Z.; Zhang,
F.; Danishefsky, S. J . Angew. Chem., Int. Ed. 2000, 24, 4505-4508.
(32) (a) Curran, D. P. J . Am. Chem. Soc. 1983, 104, 4024. (b) Curran,
D. P. Adv. Cycloadditions 1998, 1, 129-189. (c) Kozikowski, A. P. Acc.
Chem. Res. 1984, 17, 410-216.
P r elim in a r y Stu d ies
F r a gm en t Syn th esis: Th e C6-C15 F r a gm en t. Our
initial attempts to achieve a workable synthesis focused
(33) Martin, S. F.; Dupre, B. Tetrahedron Lett. 1983, 24, 1337.

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6413
F igu r e 2.
F igu r e 1.
Sch em e 3
our development of operationally simple, broadly ap-
plicable reaction conditions which enabled the highly
diastereoselective cycloaddition with a wide range of
reaction partners, including aliphatic and C-alkynyl
nitrile oxides.37 However, our attempts to extend this
chemistry to the requisite thiazole-containing oxime 12
were thwarted by the inability to selectively generate the
nitrile oxide precursor. After our initial attempts at
cycloaddition failed to provide any cycloadducts, we
reinvestigated the chlorination (NCS) of oxime 12 and
noted that chlorination occurred selectively on the aro-
matic ring to give 14 (eq 2). Attempts to effect successful
on aldol methodology including both catalytic, asym-
metric aldol reactions as well as chiral reagent-based
approaches. As in the case of the C3-C8 polypropionate
region, we found these methodologies to be limited either
by their stereoselectivity or their practicality, and we
sought to explore less established chemistry to achieve
a novel and expedient synthesis. In this regard, we were
intrigued by the report of a powerful, yet seemingly
overlooked hydroxy-directed nitrile oxide cycloaddition
reaction reported by Kanemasa (eq 1).35 A short series
oxidation with milder halogenating agents, such as tert-
BuOCl or NBS, gave identical results.
As an alternative, we sought a nitrile oxide precursor
which would allow introduction of the sensitive thiazole
at a later stage. In this regard the nitrile oxide derived
from readily prepared oxime 1738 was chosen for the late
stage introduction of the thiazole.39 At the outset, we had
serious concerns with the use of this phosphonate, as it
was expected to be highly base sensitive. However,
hydroximinoyl chloride 18 prepared from oxime 17
underwent diastereoselective nitrile oxide cycloaddition
to give 20 as a single syn diastereomer at the isoxazoline
oxygen (Scheme 4). Oxime 17 proved to be a stable,
versatile nitrile oxide precursor critical to achieving an
efficient, convergent synthesis of C6-C15 epothilone
subunit 21 and became the cornerstone of our epothilone
syntheses (Scheme 5).
of publications documented important effects of Mg(II)
ion on the course of nitrile oxide cycloaddition of benzo-
nitrile oxide with an allylic alkoxide including a signifi-
cant increase in the rate of the cycloaddition and a
stabilization of the reactive nitrile oxide by coordination
of the magnesium ion to the Lewis basic oxygen (Figure
2).36 Furthermore, the allylic alkoxide was proposed to
coordinate to the magnesium and to promote a pseudo-
intramolecular reaction with the magnesium ion acting
as a transient tether. Importantly, with chiral allylic
alcohols, an impressive degree of syn-stereoselectivity
(99:1 syn:anti) in the cycloaddition was observed, and it
is noteworthy that even substituted, unactivated olefins
react under these conditions at 0 °C or below.
The successful implementation of this strategy neces-
sitated convenient access to the appropriate allylic alco-
hols such as 24. Although chiral allylic alcohols are
valuable synthetic building blocks, there exist remark-
ably few methods for their general, enantioselective
synthesis. By far the most common method, kinetic
Although a viable experimental protocol for the pre-
parative use of this reaction was not the focus of Kane-
masa’s studies, his observations provided the basis for
(34) (a) Asaoka, M.; Abe, M.; Takei, H. Bull. Soc. Chem. J pn. 1985,
2145. (b) Asaoka, M.; Abe, M. Chem. Lett. 1982, 215. (c) Confalone, P.
N.; Ko, S. S. Tetrahedron Lett. 1984, 25, 947.
(35) Kanemasa, S.; Nishiuchi, M.; Kamimure, A.; Hori, K. J . Am.
Chem. Soc. 1994, 116, 2324-2339.
(36) (a) Kanemasa, S.; Nishiuchi, M.; Wada, E. Tetrahedron Lett.
1993, 34, 4011. (b) Kanemasa, S.; Okuda, K.; Yamamoto, H.; Kaga, S.
Tetrahedron Lett. 1997, 38, 4095.
(37) Bode, J . W.; Fraefel, N.; Muri, D. Angew. Chem., Int. Ed. 2001,
41, 2082-2085.
(38) For a single step, high yielding preparation of the corresponding
aldehyde from diethylmethyl phosphonate, see Aboujaoude, E. E.;
Collignon, N. Synthesis 1983, 635.
(39) For the use of a related phosphonate-containing nitrile oxide
in a thermal cycloaddition, see: Tsuge, O.; Kanemasa, S.; Suga, H.
Nakagawa, N. Bull. Chem. Soc. J pn. 1987, 60, 2463.

6414 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
Ta ble 1. Op tim iza tion of th e Ep oxid e Op en in g
Sch em e 4
entry
X
solvent
THF
Et2O
base
% 29 % 30 % 31
1
2
3
4
5
6
7
8
9
I
I
I
I
I
I
nBuLi
nBuLi
10
70
trace
30
s
s
1:1 Et2O/hex nBuLi
complex mixture
Et2O
Et2O
Et2O
EtMgBr
NaH
s
30-40
no reaction
no reaction
no reaction
no reaction
no reaction
s
s
KH
OSO3Me Et2O
OSO3Me THF
OSO3Me DME
OSO2CF3 Et2O
nBuLi
nBuLi
nBuLi
nBuLi
10
82
trace
marked improvements (entries 3-6). Reasoning that
changing the counteranion associated with the trialkyl
sulfonium salt from iodide to a nonnucleophilic alterna-
tive would preclude formation of such byproducts, we
conducted the epoxide-opening reaction with commer-
cially available trimethylsulfonium methyl sulfate; how-
ever, the insolubility of this salt rendered the reaction
unworkable (entries 7-9). In contrast, the use of trim-
ethyl sulfonium triflate, prepared by reaction of dimethyl
sulfide and methyl triflate, proved optimal. The desired
allylic alcohol product was reliably obtained in prepara-
tively useful scales in >80% yield with traces of thioether
31 (>1%) as the only detectable side-product (entry 10).
Having developed suitable conditions for the conversion
of a terminal epoxide to the corresponding homologous
allylic alcohol, we were ready to commence with the
synthesis of the necessary fragment 24. The synthesis
began with the convenient alkylation of (R,R)-pseudo-
ephedrine-derived amide 32 with readily available iodide
33 (89%) to give 34 as a single diastereomer, as previ-
ously described by Myers (Scheme 6).45 Reduction of the
chiral amide with the BuLi/NH3‚BH3 afforded alcohol 35
in 99% yield.46 Oxidation of 35 under Ley conditions
(TPAP, NMO, 4 Å MS)47 followed by Wittig olefination
(Ph3EtPI, NaNH2) afforded the cis-olefin 36 in 84% yield
and >95% ee as a 13:1 mixture of cis:trans diastereomers.
The integrity of the methine stereocenter through the
series of reactions was confirmed following desilylation,
Mosher ester formation,48 GC (Hewlett-Packard HP-5
siloxane column) analysis, and comparison to an authen-
tic, racemic sample.49
resolution, reliably provides the desired product in high
enantiomeric excess.40 In our case, however, the presence
of an additional stereocenter in allylic alcohol 24 made
such a strategy unappealing. The leading alternative, the
asymmetric addition of carbanions to an appropriate
aldehyde,41 appeared promising; however, cumbersome
preparations of the chiral vinylorganometallic along with
less than optimal chiral induction rendered such an
approach less than ideal.
Monosubstituted chiral epoxides have recently become
easy to access in high enantiopurity, and we sought a
procedure for the direct conversion of a terminal epoxide
to the corresponding allylic alcohol.42 In this respect, the
single step conversion of epoxides to allylic alcohols
reported by Mioskowski and Falck seemed promising.
This protocol, the addition excess Corey-Chaykovsky
dimethylsulfonium methylide to epoxides and aldehydes,
directly affords chiral allylic alcohols but was unexploited
apart from the original disclosure (eq 3).43,44 Thus, we
were prompted to conduct an investigation of this reac-
tion before applying it to the synthetic strategy en route
to the epothilones. Using commercially available 28 as
our model epoxide, we were initially disappointed to find
that the reported reaction conditions gave the corre-
sponding allylic alcohol 29 in less than 10% yield (Table
1, entry 1). However, we noted that changing the solvent
from the reported THF to Et2O resulted in a dramatic
improvement in consumption of the starting material
(entry 2), albeit the product formed was contaminated
with significant amounts of the corresponding iodohydrin.
Further attempts to influence the product distribution
by variation of reaction solvent and base did not lead to
Terminal epoxide 42 was assembled by reaction of (R)-
glycidol tosylate and Grignard 38 as shown in Scheme
7.50 The organomagnesium reagent was procured by
conversion of silyl ether 36 to bromide 37 (PPh3‚Br2,
(45) (a) Myers, A. G.; Yang, B. H.; Chen, H.; McKinstry, L.; Kopecky,
D. J .; Gleason, J . L. J . Am. Chem. Soc. 1997, 119, 6496-6511. (b)
Myers, A. G.; Yang, B. H.; Chen, H. Org. Synth. 1999, 77, 22.
(46) (a) Myers, A. G.; Yang, B. H.; Chen, H. Org. Synth. 1999, 77,
29. (b) Myers, A. G.; Yang, B. H.; Kopecky, D. Tetrahedron Lett. 1996,
37, 3623.
(47) Griffen, W. P.; Ley, S. V.; Whitcombe, G. P.; White, A. D. Chem
Commun. 1987, 1625.
(48) Dale, J . A.; Mosher, H. S. J . Am. Chem. Soc. 1973, 95, 512.
(49) The preparation of this alcohol has been reported: Reitz, A.
B,; Nortey, S. O.; Maryanoff, B. E.; Liotta, D.; Monaham, R. J . Org.
Chem. 1987, 52, 4191.
(40) For a comprehensive review, see Katsuki, T.; Martin, V. S. Org.
React. 1996, 48, 1-300.
(41) Oppolzer, W.; Radinov, R. N. Tetrahedron Lett. 1988, 29, 5645-
5648.
(42) (a) Ready, J . M.; J acobsen, E. N. J . Am. Chem. Soc. 1999, 121,
1, 6086-6087. (b) Tokunaga, M.; Larrow J . F.; Kakiuchi, F.; J acobsen
E. N. Science 1997, 277, 936-938.
(43) Alcaraz, L.; Harnett, J . J .; Mioskowski, C.; Martel, J . P.; Legall,
T.; Shin, D.-S.; Falck, J . R. Tetrahedron Lett. 1994, 35, 5449.
(44) Corey, E. J .; Chaykovsky, M. J . Am. Chem. Soc. 1965, 87, 7,
1353-1364.
(50) (a) Klunder, J . M.; Onami, T.; Sharpless, K. B. J . Org. Chem.
1989, 54, 1295. (b) Hanson, R. M. Chem. Rev. 1991, 91, 437.

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6415
Sch em e 5
Sch em e 6
Sch em e 7
catalytic ZnBr2),51 Grignard formation (Mg, Et2O), and
copper-catalyzed epoxide opening of commercially avail-
able (R)-tosyl-glycidol (39) to afford hydroxy tosylate 40.52
Although terminal epoxide 42 could be isolated following
the Grignard coupling simply by allowing the reaction
to warm to room temperature, the instability of epoxide
42 prompted us to develop conditions for in situ epoxide
formation and opening to give the chiral allylic alcohol
24. In this regard, treatment of 40 with 5 equiv of 41
and 4.8 equiv of n-BuLi according to our optimized
procedure cleanly afforded allylic alcohol 24 in 85% yield.
With 18 and 24 in hand, we were poised to conduct
the critical hydroxyl-directed nitrile oxide cycloaddition.
Although such cycloadditions had never been documented
on substrates possessing the complexity of these reaction
partners, the dipolar cycloaddition reaction proceeded
smoothly (2 days, rt), providing isoxazoline 23 in 77%
yield (Scheme 8). The secondary alcohol in 23 was
protected as a TBS-ether (tBuMe2SiOTf, iPr2EtN) to
afford 43. Following a survey of bases and reaction
conditions, the use of LiHMDS was identified as optimal
in terms of yield and stereoselectivity for the reaction of
aldehyde 44 and phosphonate 43. Under these conditions,
the formation byproducts by Cannizaro reaction was
minimized, and trans-45 was isolated in 87% yield along
with 9% of the minor cis-diastereomer, which could be
readily separated by chromatography on silica gel.
The preparation of 45 constitutes the completion of the
C6-C15 carbon backbone of fragment 21. However, the
subsequent unmasking of the isoxazoline to furnish the
corresponding hydroxy ketone proved difficult. Although
the reduction of isoxazolines to â-hydroxy-ketones is a
well-studied transformation, surprisingly, it was virtually
unknown for isoxazolines in which CdN is conjugated
with an exocyclic CdC or CtC. Moreover, for the
substrate at hand (45) the presence of both a conjugated
and a nonconjugated olefin provided a particularly chal-
lenging case for chemoselective reduction (Scheme 9).
Torssell and co-workers had examined the reduction
of 3-vinyl-substituted isoxazolines and noted that the
commonly employed reagents, such as Raney-Ni, TiCl3,
and Zn/AcOH, result in concomitant reduction of the
olefin and N-O bond.53 As an alternative, a two-step
protocol was developed involving N-alkylation of the
isoxazoline followed by electrochemical reduction of the
resulting N-alkyl isoxazolinium. Because the compat-
ibility of the reactive thiazole to such conditions was of
(51) Aizpurua, J . M.; Cossio, F. P.; Palomo, C. J . Org. Chem. 1986,
51, 1, 4941-4943.
(52) We are grateful to Diacel coporation for a generous gift of (R)-
tosyl-glycidol.
(53) Isager, P.; Thomsen, I.; Torsell, K. G. B. Acta Chem. Scand.
1990, 44, 806

6416 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
Sch em e 8
Sch em e 9
in moderate yield; however, the desired hydroxy epoxide
could not be isolated following treatment of 48 with
fluoride sources and attempts at subsequent sulfate
hydrolysis.58,59
some concern, we proceed to examine a number of other
reduction protocols. In this respect, Mo(CO)6 in CH3CN
has recently emerged as a useful reagent for N-O bond
cleavage for isoxazoles and isoxazolines.54 This reagent,
indeed, did provide the desired conjugated ketone 46,
albeit in at best 35-40% isolated yield despite extensive
attempts at optimization. A critical problem with this
reaction was the low mass balance following reaction
workup. Following the lead of previous reports that yields
of similar reductions were improved by absorbing the
crude reaction mixture onto silica gel and exposure to
air (1-2 days) led to the isolation of 46 (40-50% yield)
contaminated with saturated ketone 47 (30%) and 20%
recovered starting material (45).55 A mild, selective,
workable solution to the isoxazoline reductive cleavage
to furnish hydroxy ketone 46 eventually emerged from
our studies. Thus smooth cleavage of the isoxazoline N-O
could be effected with 3-4 equiv of SmI2 in THF at 0 °C,
followed by imine hydrolysis with B(OH)3 to give the
desired conjugated ketone (46) in 75% yield.56
We soon decided to focus our attention on the less
reactive, and rarely studied, 1,3-cyclic sulfite which could
be readily prepared in one step from diol 49 and had been
an intermediate in the formation of the cyclic sulfate 48
(Scheme 10). We were pleased to discover that treatment
of the unpurified cyclic sulfite with excess TBAF‚3H2O
in refluxing THF cleanly led to the formation of the
desired epoxide in 80% yield.60 This expedient transfor-
mation from triol to â-hydroxy epoxide effects, in a single
reaction vessel, cyclic sulfite formation, desilylation, ring
closure to form the epoxide, and SO2 extrusion, complet-
ing the synthesis of 21.61,62 The successful formation of
epoxide 21 allowed us to confirm the sense of stereo-
induction observed in the hydroxyl-directed nitrile oxide
1
cycloaddition. Analysis by H NMR spectroscopy of the
coupling constant of the vicinal epoxide protons (J ) 4.2
Hz) was consistent with that of a cis-epoxide.
Following 1,3-syn reduction of ketone 46 under Prasad
conditions (Et3B, MeOH, NaBH4),57 a key chemoselectiv-
ity issue presented itself: differentiating two secondary
hydroxyl groups in anticipation of epoxide formation.
Although considerable work could be relied upon for the
differentiation of secondary alcohols, the fact that the
desired series of reactions would necessitate not only
hydroxyl group differentiation, but also activation, ep-
oxide formation, and deprotection encouraged a novel
solution. An attractive approach would be the formation
of a bis-activated derivative, in which selective C-O
cleavage would lead to the desired epoxide. In this regard,
the obvious candidate, a cyclic sulfate, could be prepared
Th e Nitr ile Oxid e F r a gm en t. The sensitivity of the
thiazole ring toward halogenation presented an obstacle
to the generation of the requisite nitrile oxide precursor
from an oxime; thus, other protocols for the introduction
(58) For a review of the chemistry of cyclic sulfates, see Lohray, B.
B. Synthesis 1992, 1035-1052.
(59) While the suspected byproducts were not isolated, we surmised
that the acidic conditions required to hydrolyze the resulting sulfate
may result in degradation of the newly formed epoxide, perhaps via
intramolecular cyclization.
(60) An epoxide has been implicated as a transient intermediate
following the treatment of related 1,2-cyclic sulfates with TBAF: Ko,
S. Y.; Malik, M. Tetrahedon Lett. 1993, 34, 4675-4678.
(61) The presence of the silyl group is not essential for this
transformation; the corresponding hydroxy-cyclic sulfite undergoes the
same reaction under the identical conditions and can be observed as
an intermediate in this reaction.
(62) In a somewhat related example, a 1-hydroxy,4-mesyloxy,5-
silyloxy-protected triol system undergoes desilylation, epoxide forma-
tion, and 5-exo epoxide opening to afford a substituted tetrahydrofuran
under similar reaction conditions (Hori, K.; Kazuno, H.; Nomura, K.;
Yoshii, E. Tetrahedron Lett. 1993, 34, 2183-2186.). In our case, we do
not observe any tetrahydrofuran formation which would arise from
5-endo opening of the epoxide.
(54) (a) Baraldi, P. G.; Barco, A.; Benetti, S.; Manfredini, S.; Simoni,
D. Synthesis 1987, 276. (b) Cicchi, S.; Goti, A.; Brandi, A.; Guarna, A.;
De Sarlo, F. Tetrahedron Lett. 1990, 31, 3351-3354.
(55) (a) Guarna, A.; Guidi, A.; Goti, A.; Brandi, A.; De Sarlo, F.
Synthesis 1989, 175-178(b) Trost, B. M.; Chupak, L. S.; Lu¨bbers, T.
J . Am. Chem. Soc. 1998, 120, 1732-1740.
(56) Bode, J . W.; Carreira, E. M. Org. Lett. 2001, 3, 1587-1590.
(57) (a) Chen, K.-M.; Gunderson, K. G.; Hardtmann, G. E.; Prasad,
K.; Repic, O.; Shapiro, M. J . Chem. Lett. 1987, 1923-1926.

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6417
Sch em e 10
Sch em e 11
Sch em e 12
nitrile oxide precursor would serve as a stable masking
group for the nitrile oxide. Moreover, it would allow the
generation of the nitrile oxide under very high dilution
conditions, favoring formation of the macrocyclic ring and
disfavoring the two major nitrile oxide decomposition
pathways, both of which are intermolecular: dimeriza-
tion (which would be inconsequential as it would return
starting furoxan), and isocyanate formation. Even more
attractive was that these conditions allowed Curran to
achieve nitrile oxide cycloaddition of similarly substituted
cis-olefins in good yields.67 With this nitrile oxide precur-
sor, a furoxan dimer, as our target, we commenced our
preparation of C1-C5 fragment 51 (Scheme 12).
of this functionality were sought.63 The other general
method for nitrile oxide synthesis, the Mukaiyama-
Hoshino method,64 involves the dehydration of primary
nitroalkanes. Unfortunately, while this method appeared
at first to be viable, it necessitated the preparation of a
neopentyl nitro (Scheme 11). The difficulty in synthesiz-
ing such substrates, as well as concerns about the
dehydration step, encouraged us to seek novel solutions
to the preparation of nitrile oxides in functionally rich
molecules.65
The fact that the desired cycloaddition reaction re-
quired a tertiary nitrile oxide led us to investigate a
protocol reported by Curran in which nitrile oxides are
generated thermally from a furoxan dimer. This reaction
involves cracking the nominal nitrile oxide dimer to
provide the reactive species, a process which Curran had
demonstrated to be viable only for very highly substituted
nitrile oxides (eq 4).66 Further analysis revealed a number
The presence of a gem-dimethyl group in the desired
nitrile oxide 52 presented the ideal opportunity to apply
the enantioselective TiF4-BINOL catalyzed addition of
allyl trimethylsilate to aldehydes (eq 5).68 This reaction
of key advantages to an approach that relied on such a
fragmentation reaction to furnish the reactive nitrile
oxide thermally in situ. The use of a furoxan dimer as a
(65) These and other difficulties in the preparation of nitrile oxides
encountered during our synthesis of the epothilones led to the
development of a novel, general method for nitrile oxide generation
from O-TBDPS hydroxamates: Muri, D.; Bode, J . W.; Carreira, E. M.
Org. Lett. 2000, 2, 539-542.
(63) A model study of the oxime chlorination reinforced our concerns
about this approach to the nitrile oxide:
(64) Mukaiyama, T.; Hoshino, T. J . Am. Chem. Soc. 1960, 82, 5339.

6418 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
Sch em e 13
Sch em e 14
offers several advantages over related processes including
the commercial availability of the ligand, the metal, and
allyl silane; the single step, in situ, preparation of the
active catalyst; the low toxicity of the allyl silane as
compared to the corresponding and often utilized allyl-
stanne reagents; and the ability to run this reaction at
high concentration at 0 °C. This was found to be
particularly effective, in terms of both yield and enantio-
selectivity, for the allylsilylation of sterically demanding
aldehydes such as those required for an epothilone total
synthesis.
Following preliminary results which suggested this
reaction would function well with a suitably functional-
ized aldehyde, we identified aldehyde 57 as an optiminal
substrate in terms of yield, enantioselectivity, and the
possibility to enrichment the enantiopurity by recrystal-
lization. Noteworthy is the ability to perform this reaction
on a preparative scale (40 mmol) using 5 mol % catalyst
and at high substrate concentration (approximately 4 M).
With substantial quantities of 58 available, its elabo-
ration to 52 was subsequently investigated (Scheme 13).
Ozonolytic cleavage of the double bond followed by in situ
protection afforded acetal 59 which was reductively
deprotected (LiAlH4) to furnish diol 60 in 73% overall
yield. Selective oxidation of the primary alcohol (catalytic
TEMPO, NaOCl, catalytic KBr) cleanly afforded hydroxy
aldehyde 61 which was converted directly to its corre-
sponding oxime (62).69
Nitrile oxide formation was readily achieved by oxime
chlorination with tert-BuOCl and deprotonation with
Et3N to provide stable, chromatographically isolatable
nitrile oxide 52 (Scheme 14). In practice, however, the
cleanly formed nitrile oxide was taken directly to the
dimerization step. Refluxing a concentrated solution (1.0
M) of 52 in benzene for 18 h resulted in clean formation
of the UV active, more polar (TLC), unsymmetrical dimer
63.70 Protection of the secondary hydroxyls (TBSOTf,
Hu¨nig’s base) proceeded smoothly to give 64. Selective
acetal hydrolysis in the presences of the TBS ethers was
achieved by the action of FeCl3‚SiO2 in acetone.71 Finally,
Lindgrin oxidation of the dialdehyde to the diacid pro-
ceeded without incident to give 51, completing the
synthesis of the C1-C5 subunit.
In tr a m olecu la r , Rin g-Closin g Nitr ile Oxid e Cy-
cloa d d ition . In preparation for the key step, nitrile oxide
fragment 51 and olefin fragment 21 were coupled under
standard ester forming conditions to provide the unsym-
metrical diester 65 (Scheme 15). Upon heating a dilute
(0.00014 M) solution of dimeric diester 65 in C6H6 at 150
°C for 36 h, a product of nitrile oxide cycloaddition was
(66) Curran, D. P.; Fenk, C. J . J . Am. Chem. Soc. 1985, 107, 7,
6023-6028.
(67) For example, Curran obtained a 59% yield in a related cyclo-
addition of 1.1 equiv of a cis-olefin using a furoxan as the nitrile oxide
precursor. In contrast, the same cycloaddition using a nitro compound
as the nitrile oxide precursor and excess olefin afforded the product in
only 20% yield:
(69) (a) Anelli, P. L., Biffi, C., Montanari, F. Quici, S. J . Org. Chem.
1987, 52, 2559-2562. (b) For a review of TEMPO mediated oxidations,
see de Nooy, A. E. J .; Besemer, A. C.; van Bekkum, H. Synthesis 1996,
1153-1174.
(70) Attempts to dimerize a more sterically hindered nitrile oxide
failed to provide the desired furoxan dimer and the isocyanate was
isolated in good yield:
(68) Gauthier, D. R.; Carreira, E. M. Angew. Chem., Int. Ed. Engl.
1996, 35, 2363.
(71) Kim, K. S.; Song, Y. H.; Lee, B. H.; Hahn, C. S. J . Org. Chem.
1986, 51, 404.

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6419
Sch em e 15
same approach would be amenable to the epothilone A
series. Fortunately, our previously prepared subunit 21
was readily converted to the appropriate aldehyde (70)
and tested in this respect. Following protection of the
secondary alcohol as its TES-ether, we were pleased to
find that the cis-olefin, which had refused to undergo 1,3-
dipolar cycloaddition, could be selectively dihydroxylated
(catalytic OsO4, NMO) and the unpurified diol cleaved
(Pb(OAc)4)to the requisite aldehyde (Scheme 18).
Aldehyde 70 undergoes highly diastereoselective aldol
coupling with 1.5 equiv of 68 at -78 °C (Scheme 19).
Importantly, we found this reaction to be reliable,
consistently affording the desired product in >10:1 dr.
With a solution to the synthesis of the critical polypro-
pionate region in hand, we opted to revisit our epothilone
A synthesis.
Secon d Gen er a tion Syn th eses of Ep oth ilon es A
a n d B. Th e Tota l Syn th esis of Ep oth ilon e A. The
successful total synthesis of epothilone A began with the
preparation of aldehyde 73 from known chiral alcohol 72
(Scheme 20).27 Although our initial report of highly
asymmetric alkyne additions to aldehydes were not
readily extended to unbranched, aliphatic substrates,75a
we have recently developed conditions for the enantiose-
lective addition 3-methylbutyn-2-ol which appeared prom-
ising for the synthesis of the requisite chiral allylic
alcohol.75b In this regard, aldehyde 73 undergoes clean,
highly diastereoselective (>20:1 dr) alkyne addition
under the reported conditions. Following protection of the
unpurified secondary alcohol as a benzoate ester, prop-
argylic ester 75 was obtained in 72% yield. Extrusion of
acetone (K2CO3, catalytic 18-crown-6) and treatment of
the ester with LiAlH4 provided chiral allylic alcohol 77
directly in 81% overall yield from 75.
Sch em e 16
obtained as a single regio- and stereoisomer in 47%
yield.72 Unfortunately, upon closer analysis of both the
crude reaction mixture and the isolated products, we
found that cycloaddition had occurred exclusively on the
more highly substituted olefin (Scheme 16).73,74 This
result prompted us to reevaluate the use of a nitrile oxide
cycloaddition reaction for the macrocyclization step and
subsequently led to the development of a convergent
route that relied on diastereoselective nitrile oxide cy-
cloadditions to assemble the constituent fragments and
lactonization of a seco-acid to furnish the desired mac-
rocycle.
Th e Con ver gen t Ald ol Ap p r oa ch . The disclosure by
Mulzer of a highly stereoselective convergent aldol
coupling in the context of an epothilone B total synthesis
provided an intriguing precendent for a new approach
to the epothilones built upon chemistry developed for the
synthesis of the C6-C15 fragment. Mulzer reported that
while the aldol reaction of aldehyde 67 with a C12-C13
olefin proceeded with only 4:1 dr, the aldol with aldehyde
69 containing a C12-C13 epoxide gave the desired anti-
Felkin product in >10:1 dr and 91% yield which could
be readily transformed to epothilone B (Scheme 17).29
Given the abundance of remote effects in related aldol
reactions, however, we could not be confident that the
Treatment of 77 with the nitrile oxide derived from 18
proceeded cleanly (58% conversion after 48 h, 94%
recovery of 77) to give cycloadduct 78 as a single syn
diastereomer at the isoxazoline oxygen (Scheme 21).
Condensation of 79 with aldehyde 44 under Roush-
Masamune conditions proceeded in good yield,76 but as
a 6:1 mixture of olefin stereoisomers which were not
readily separated by column chromatography. With the
hope that the isomers could be separated following
reduction of the isoxazoline to the â-hydroxy ketone the
mixture was submitted to our recently developed selec-
tive isoxazoline reduction with SmI2. Following the
reaction, analysis of the crude reaction mixture revealed
(74) Attempts to effect an intermolecular nitrile oxide cycloaddition
on the unactivated cis-olefin revealed the pronounced unreactivity of
this double bond. Reaction of either vi or vii with olefin 23 under the
conditions reported by Curran for intermolecular nitrile oxide cycload-
dition quantitatively returned the starting olefin unchanged and the
nitrile oxides were consumed by clean conversion to isocyanate viii.
(72) All of the starting dimer had been consumed and the bulk of
the remaining material was identified as uncyclized nitrile oxide (6
in Scheme 2).
(73) The other potential product, endo-cyclization of the nitrile oxide
onto the trisubstituted to give nine membered ring lactone v was
discounted on the basis of 13C NMR DEPT experiments which clearly
showed that the eight membered ring had been formed.
(75) (a) Franz, D.; Fa¨ssler, R.; Carreira, E. M. J . Am. Chem. Soc.
2000, 122, 1806-1807. (b) Boyall, D.; Lopez, F.; Sasaki, H. Frantz,
D.; Carreira, E. M. Org. Lett. 2000, 2, 4233-4236.
(76) Blanchette M. A.; Choy, W.; Davis, J . T.; Essenfeld, A. P.;
Masamune, S.; Roush, W. R.; Sakai, T. Tetrahedron Lett. 1984, 25,
2183

6420 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
Sch em e 17
Sch em e 18
Sch em e 19
Sch em e 20
a 9:1 mixture of the E and Z olefin stereoisomers, from
which the desired E-isomer was isolated in 75% yield.77
Reduction of the â-hydroxy ketone to the 1,3-syn diol
was followed by our protocol for diol differentiation and
epoxide formation via a cyclic sulfite to afford 84 in 70%
overall yield (Scheme 22). Following bis-O-TES protec-
tion, the primary silyl group was selectively removed and
the unpurified alcohol oxidized to the previously prepared
aldehyde 70 in 84% yield.
that this compound undergoes clean macrolactonization
under modified Yamaguchi conditions, affording 16-
membered macrolide 88 in 74% yield. The C7-OTroc ester
was rapidly removed, even at room temperature, under
conditions similar to those reported by Yamada for Troc
deprotection of a sensitive molecule.79 This procedure
routinely provided the desired product in >80% yield
along with traces of the dichlorocarbonate ester (Scheme
24).80
Following highly diastereoselective coupling of 68 and
70, aldol adduct 71 was protected as its C7-OTroc ester
and the terminal olefin oxidized to the corresponding
aldehyde in good yield (Scheme 23).78 Selective depro-
tection of the C15-OTES and Lindgrin oxidation of the
aldehyde provided seco-acid 87. We were pleased to find
In comparison to the facile deprotection of the C7-OTroc
ester, removal of the remaining C3-OTBS protective
group proved to be challenging. In contrast to the
corresponding transformation of epothilone C, which
lacks the C12-C13 epoxide functionality, deprotection of
epothilone A was hampered by pronounced acid and base
sensitivity. In this regard, reagents such as AcOH, 1:10
aq. HF:CH3CN, CF3COOH (-20 °C in CH2Cl2), HF‚NEt3,
and HF‚pyr (molar rations <2:1 pyr:HF) all resulted in
(77) Upon reduction of a related (Z)-conjugated isoxazoline, we
observed formation of the corresponding (E) product as >25% of the
reaction mixture. It is not clear, however, if this isomerization occurs
during the isoxazoline reduction or upon hydrolysis of the resulting
imine.
(78) The final steps of Mulzer’s epothilone B synthesis were modified
and employed for the completion of epothilone A.
(79) Wakamatsu, K.; Kigoshi, H.; Niyama, K.; Niwa, H.; Yamada,
K. Tetrahedron 1986, 42, 5551-5558.

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6421
Sch em e 21
Sch em e 22
destruction of the sensitive epoxide moiety.81 Likewise,
the highly hindered nature of the silyl ether rendered
numerous reagents for silyl deprotection ineffective for
this transformation, including TBAF-AcOH (THF, 50
°C), TBAF-(2-fluorophenol) (THF, 50 °C), TBAF-HF
(THF, 50 °C), NH4F-HF (DMF:NMP, 45 °C), NH4F
(MeOH, 60 °C), and 0.1 M periodic acid. This difficult
transformation was finally accomplished by the use of a
3:1 mixture of HF‚pyr in pyridine at 40 °C (Scheme 25).82
These conditions resulted in slow but clean deprotection
to afford epothilone A (40% conversion after 7 days)
whose spectral characteristics (1H NMR, 13C NMR, IR,
HRMS, [R]D). were identical in all respects to those
reported for natural material
F or m a l Syn th esis of Ep oth ilon e B. The increasing
importance of epothilone B and related analogues in the
development of new anticancer chemotherapeutics mo-
tivated us to pursue a concise synthesis of this molecule.
An important feature of our synthetic approach to the
epothilones is its ready extension to the synthesis of
related structures. Previously prepared isoxazoline 20
(see Scheme 4) proved to be the ideal starting point for
the epothilone B synthesis as this material could be
readily synthesized in enantiomerically pure form using
commercially available (R)-3-buten-2-ol (19) (Scheme
26).83 Following olefination and oxidation of the second-
ary alcohol, ketone 92 was obtained as highly crystalline
solid which was stereochemically pure following a single
recrystallization (Et2O).
(80) Attempts to eliminate this byproduct by modification of the
reaction reaction solvent (MeOH or EtOH), reaction temperature (rt
or 80 °C) and by careful drying of the reaction components uniformly
failed to change the product distribution. Attempts to deprotected the
dichlorocarbonate ester by resubmission to the reaction condition or
via reduction with SmI2 were equally unsuccessful, and the protected
compound was recovered unchanged. These results, however, were
more curious than consequential as the desired deprotection product
was consistently obtained in >80% yield regardless of the conditions
employed.
(81) Up until this point in the synthesis, the epoxide had proven to
be a robust, trouble-free functional group which at no point exhibited
any notable reactivity, even in the presence of acids (AcOH) and strong
bases (TBAF, THF, reflux).
Using conditions reported by Curran for nucleophilic
additions to similar 2-acyl-∆2-isoxazolines,84 we were
pleased to find that chelation controlled Grignard addi-
tion of 93 to ketone 92 proceeded in good yield and >10:1
(83) We are grateful to Boehringer-Ingelheim for a generous gift of
(82) Mulzer has also reported the use of HF‚pyr to effect the
deprotection of C3-OTBS of epothilone B. Similarly long reaction times
were necessary for this deprotection.
(R)-3-buten-2-ol.
(84) Curran, D. P.; Zhang, J . J . Chem. Soc., Perkin Trans. 1 1991,
2613-2625.

6422 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
Sch em e 23
Sch em e 25
Sch em e 26
Sch em e 24
Sch em e 27
addition to the ketone. The use of the isoxazoline as a
masking group and as a platform for controlling the
stereochemistry of the Grignard addition highlights the
unique advantages enabled by the nitrile oxide approach
to the epothilones.
Following protection of the tertiary alcohol (TESOTf,
Hu¨nig’s Base), the epothilone B synthesis converged with
our previously described route to epothilone A. In this
regard isoxazoline reduction (SmI2, THF), stereoselective
ketone reduction, epoxide formation (via the 1,3-cyclic
sulfite), TES protection, and aldehyde preparation all
proceeded without incident to provide the previously
reported epoxy aldehyde 69 (Scheme 28).
Although Mulzer has described the highly diastereo-
selective aldol coupling of aldehyde 69 and ketone 68,
we have confirmed the highly selective nature of this
reaction which consistently gives the product in good
yield and excellent diastereoselectivity (Scheme 29). The
ability of the remote epoxide moiety to influence the
dr to give the desired syn relationship between the
isoxazoline oxygen and the resulting tertiary alcohol
(Scheme 27).85 To obtain consistently high diastereo-
selectivity, it was essential that the solution of the
Grignard reagent was cooled to -78 °C prior to its
(85) The sense of stereoinduction was confirmed by conversion of
this compound to aldehyde 69 (vida supra) and comparison to the
spectral data of this known compound (see ref 29). We are grateful to
Prof. Mulzer for kindly providing copies of spectral data for 69 and
aldol adduct 100.

Stereoselective Syntheses of Epothilones A and B
J . Org. Chem., Vol. 66, No. 19, 2001 6423
Sch em e 28
Sch em e 29
course of this reaction represents yet another remarkable
long-range effect discovered in the course of synthetic
efforts toward these molecules. It is significant in that it
enables the successful realization of a highly convergent
approach to epothilone B with complete stereocontrol at
all seven chiral centers. Furthermore, the successful
application of this coupling reaction completed the highly
convergent and fully stereocontrolled synthesis of the
epothilone B carbon backbone (11 steps) and constitutes
the expedient formal synthesis of this molecule.
to the development of efficient and scalable approaches
to the epothilones and analogues. The syntheses de-
scribed have proven to be fertile ground for the discovery
and advancement of novel methodologies including: (1)
a new method for nitrile oxide generation,65 (2) a chemose-
lective protocol for isoxazoline reduction,56 and (3) a new
procedure for epoxide formation via 1,3-cyclic sulfites. As
well, we have studied, advanced and debuted a number
of powerful carbon-carbon bond forming reactions in the
context of a complex molecule synthesis, namely (1) the
diastereoselective Kanemasa hydroxyl directed nitrile
oxide cycloaddition, (2) the Falck-Mioskowski allylic
alcohol synthesis, (3) the TiF4-BINOL-catalyzed enan-
tioselective allylsilylation of aldehydes, and (4) the Zn-
(OTf)2-N-Me-ephedrine-mediated asymmetric alkyne ad-
dition. Given the ongoing, intense investigations in
academic and industrial research laboratories on the
chemistry and biology of the epothilones, the results
presented substantively contribute to development of a
Con clu sion s
In summary, we have described the concise, stereo-
controlled total synthesis of epothilone A (21 steps) and
the formal synthesis of epothilone B (18 steps in total)
using hydroxyl-directed nitrile oxide cycloadditions and
Mulzer’s highly diastereoselective convergent aldol cou-
pling. These studies constitute an important contribution

6424 J . Org. Chem., Vol. 66, No. 19, 2001
Bode and Carreira
workable, stereocontrolled synthesis of this important
class of natural products.
Kontaktgruppe fu¨r Forschungsfragen (KGF), and Hoff-
man-La Roche. J .W.B. thanks the National Science
Foundation for a predoctoral fellowship.
Ack n ow led gm en t. We are grateful to Prof. Mulzer
for copies of spectra for 69 and 100 and to Boehringer-
Ingelheim and Diacel Corporation for a generous gifts.
Dr. Don Gautier is gratefully acknowledge for his work
on of the asymmetric allyl silylation reaction. Support
has been provided by generous funds from the ETH,
Su p p or tin g In for m a tion Ava ila ble: Complete experi-
mental procedures, spectral data, and structure correlation for
all relevant compounds (PDF). This material available free of
charge via the Internet at http://pubs.acs.org.
J O015791H