Epothilone B and its derivatives as novel antitumor drugs: Total and partial synthesis and biological evaluation

Article abstract of DOI:10.1007/s007060070099

Microtubule stabilizing natural products, as exemplified by paclitaxel (taxol^R), are being considered as novel drugs against malignant therapy resistent solid tumors. Among these compounds, epothilone B and some of its derivatives have emerged as particularly promising candidates for industrial development. The total and partial syntheses of these compounds are described in detail, and some of the most important recent results on their biological activity are discussed.

Full text of DOI:10.1007/s007060070099

Monatshefte fuÈr Chemie 131, 205±238 (2000)
Invited Review
Epothilone B and its Derivatives as Novel
Antitumor Drugs: Total and Partial Synthesis
and Biological Evaluationa
Johann Mulzer
È
Institut fuÈr Organische Chemie der Universitat Wien, A-1090 Wien, Austria
Summary. Microtubule stabilizing natural products, as exempli®ed by paclitaxel (taxol1), are being
considered as novel drugs against malignant therapy resistent solid tumors. Among these
compounds, epothilone B and some of its derivatives have emerged as particularly promising
candidates for industrial development. The total and partial syntheses of these compounds are
described in detail, and some of the most important recent results on their biological activity are
discussed.
Keywords. Epithilones; Macrolides; Antitumor drugs; Microtubule stabilizing agents; Total synth-
esis; Natural products.
General Background
The chemotherapy of cancer (Fig. 1) strongly depends on the nature of the
individual tumor under consideration. For instance, tumors located in hormone
dependent organs such as the ovaries, breasts, etc. are treated with antihormons,
e.g. anti-estrogens or anti-androgens, which retard the growth of the individual
tissue and thus also that of the tumor enclosed. In the regular case, however, the
drug is intended to kill already existing tumor cells by damaging their DNA (mostly
after intercalation) or by interrupting their mitotic cycle [1].
This aim is most ef®ciently achieved by addressing the so-called microtubles
[2] (Fig. 2). Microtubules are dynamic structures within the cell that play a critical
role in many cellular processes, one of their most important functions being the
formation of the mitotic spindle which controls the movement of the chromatids
throughout the cell division. Microtubules are made up of many individual protein
subunits known as ꢀ- and ꢁ-tubulin (Fig. 2).
In the ®rst step of the formation of microtubules, one ꢀ- and one ꢁ-subunit are
joining to form heterodimers which polymerize to proto®laments. Subsequent
th
Dedicated to Prof. Dr. Gunther Wulff on the occasion of his 65 birthday
a
È

206
J. Mulzer
1. Hormone induced tumors: antihormones (tamoxifen, cyproterone acetate)
2. Cytotoxic drugs (chemicals of low molecular weight)
2.1. Intercalation‡DNA-damage (alkylation: CC-1065, double strand cross linking:
cis-platin, oxidation: anthracyclinone, rupture: calicheamycin)
2.2. Interference with mitotic cycle. Mitotic spindle cannot be formed due to:
2.2.1. Destabilization of microtubules (colchicine, vinca alkaloids)
2.2.2. Stabilization of microtubules (Paclitaxel)
Fig. 1. Chemotherapy of cancer
Fig. 2. Formation of microtubules
aggregation of these proto®laments leads to microtubules which are in a mobile
equilibrium with the smaller fragments. If this equilibrium is disturbed, the mitotic
cycle is interrupted. Alkaloids such as colchicine or vinblastine have been known
for a long time to prevent the aggregation of the proto®laments. With the advent of
paclitaxel (taxol1) (1±1, Scheme 1), however, a new mode of interaction was
discovered. Paclitaxel stabilizes the microtubules by inhibiting their disaggrega-
tion. However, despite its impressive biological pro®le and its wide application
(annual sales of about 1 billion USD), paclitaxel has turned out to be far from ideal
for several reasons, e.g. multidrug resistance, poor bioavailability, and several
serious side effects. However, its mode of action was considered as unique among
all cytostatic drugs, until, quite sensationally, Bollag and coworkers discovered in
1995 [4a] that natural compound called epothilone not only binds to microtubules
in a paclitaxel-like manner, but that it was much more active! Surprisingly,
epothilone was not a new compound at that time; it had been isolated and
structurally elucidated as a secondary metabolite from the soil myxobacterium

Epothilones: Synthesis and Biological Evaluation
207
Scheme 1. Microtubule stabilizing natural products
È
Sorangium cellulosum by Ho¯e and Reichenbach and their coworkers at GBF in
Braunschweig as early as 1987 [5]. Unfortunately, the screening at that time had
been focused on pesticidal activity only. Epothilone, which turned out to be a
mixture of epothilone A und B (2-1 and 2-4, Scheme 2), did show activity towards
certain fungi, but was too toxic for any practical application. Another effect which
was already detected at that time was the high cytotoxicity which, however, was
not pursued any further. Epothilone thus shared the fate of Sleeping Beauty for
about seven years, when, in the aftermath of the sensational success of paclitaxel, a
general test for rapid detection of microtubule stabilizing substances (called
``tubulin assay'') was developed at Upjohn Company in 1994. About 140 000
compounds were screened with respect to their microtubule stabilizing effect, and
they all proved inactive. It was in parallel tests at Merck, Sharp, and Dohme,
however, that the above-mentioned hit [4a] was scored on testing an extract from
Sorangium cellulosum containing epothilone A and B. In the tubulin assay,
epothilone A turned out to be as active as paclitaxel, epothilone B was ®fty times
more active. These results were con®rmed afterwards by a group from the National
Cancer Institute [4b]. Later, several other natural products (1-2 to 1-5) were found
with a similar microtubule stabilizing capability (Scheme 1) [6]. Nevertheless,
however, the main interest continues to be concentrated on the epothilones, in
particular epothilone B because it promises to have some signi®cant advantages

208
J. Mulzer
Scheme 2. Naturally occurring epothilones
over paclitaxel and the other compounds depicted in Scheme 1. First of all,
epothilone B is available in kg-quantities by fermentation. This is particularly
important in considering that compounds 1-2 to 1-5 are all metabolites of marine
organisms and can be isolated from the natural source only with great dif®culty and
in minute quantities. This excludes their practical applicability and limits the
exploration of their biopro®le. Secondly, epothilone B is more soluble in water and
has thus better galenic qualities than paclitaxel. Thirdly, it is still active on cells
showing multidrug resistance. Fourthly, it exhibits similar toxicity towards various
tumor cells (e.g. breast cancer, lung cancer), but acts more rapidly than paclitaxel
does. Fifthly (and in our context most importantly), its molecular architecture is
much simpler. This means that total synthesis of epothilone B and more potent
unnatural derivatives is much more promising and practical than it is in the
paclitaxel series. Meanwhile, the epothilone family comprises altogether six
members (epothilones A-F, 2-1 to 2-6, Scheme 2) which have all been isolated as
metabolites from microbial sources.
On this general background, it was not surprising that soon after the
exceptional biological properties of the epothilones were disclosed a race was
started towards the ®rst total synthesis, which mainly involved the well known US
groups of Danishefsky and Nicolaou and the German team around D. Schinzer. All
three competitors concentrated on the simpler epothilone A ®rst. The ®rst total
synthesis of epothilone A was reported in late 1996 by Danishefsky [7], the other
two groups followed a couple of weeks later [8,9]. The ®rst total synthesis of
epothilone B was completed by Danishefsky in 1997 [10]. Later on, several more
syntheses of epothilone B were published [11].
It has become clear by a variety of biological tests (vide infra) that the epothilone
B series (characterized by the 12-methyl group) contains the more promising
candidates for further clinical development. This means that an ef®cient practical
total synthesis of this class of compounds is essential to gain access to potent
derivatives. Previous reviews [2,3] have dealt with many of the earlier aspects, in
particular those regarding epothilone A and its derivatives. In the meantime great
progress has been made with respect to epothilone B, which is now in the focus of

Epothilones: Synthesis and Biological Evaluation
209
Fig. 3. Crystal structure of epothilone B
general interest. This review will therefore be restricted to the synthesis of the
epothilone B family and the evaluation of their biological properties.
Structure and Conformational Behaviour of Epothilone B
The structure of epothilone B in the solid state [5d] (Fig. 3) has been elucidated by
single crystal diffraction. It reveals the structure of a 16-membered lactone, whose
C1-C4-segment shows a zig-zag antiparallel arrangement including one of the 5-
methyl groups. A second similarly zig-zag oriented segment is found at C7-C12.
These two segments are arranged roughly parallel to each other and are crosslinked
by the C13-C15-and the C5-C6-chains. From this box-like macrolide, three
structural moieties project away and may thus act as anchor groups for the
receptor: the C16-C17-thiazolalkylidene moiety, the 12,13-epoxide, and the 8-
methyl group, and indeed these moieties are responsible for a high biological
activity. Regarding the conformational behaviour in solution so far only epothilone
A has been analyzed in detail [12] by NMR techniques and computational
methods. According to these studies the C1-O15-C15-C14-C13-C12- and the C1-
C5-parts are relatively rigid, whereas the C6-C11-section is ¯exible. Two stable
conformers are found in a ratio of 4:1, the main conformer being very similar to the
solid state conformation.
General Retrosynthetic Considerations
At ®rst sight epothilone B does not look structurally complicated, in particular
when compared to paclitaxel and the other compounds with microtubule stabilizing
effects (Scheme 1). However, some considerations are important in designing a
suitable strategy (Scheme 3).
1. Introduction of the 12,13-epoxide
This is the central issue of the entire synthesis. There are two basic options: 1.1. The
epoxide is generated by epoxidation of the corresponding ole®n (epothilone D). This

210
J. Mulzer
Scheme 3. Basic transformations in epothilone B total synthesis; epoxide formation and ring closure
strategy has to cope with the problem of chemoselectivity (the 16,17-ole®n must not
be affected) and stereoselectivity (a (Z)-double bond has to be generated and to be
epoxidized diastereoselectively from the ꢀ-face). In fact, the solid state structure of
epothilone B indicates that in epothilone D the ꢀ-face should be more accessible that
the ꢁ-face. Additionally, a 12,13-ole®n is a suitable functional entity for connecting
two major fragments (C1-C12 and C13-21) by Wittig type ole®nations.
1.2. A second option is to generate the epoxide from a diol intermediate 3-1 by an
intramolecular SN2-displacement reaction of a 13-mesylate. The regioselectivity of
the mesylation is controlled by the substitution of the OH-functions (13-OH is tertiary
and 12-OH is secondary). The more dif®cult problem, however, is the stereocontrolled
generation of a 12(R),13(R)-diol moiety, such as in intermediate 3-1.
2. Question of ring closure
The most obvious ring closure method is that achieved by macrolactonization. In
view of the highly sensitive allylic and homoallylic 15-OH-function, a carboxyl
group activated lactonization is indicated which proceeds under retention of the
con®guration at C15. Other plausible ring closures would imply ring closing
metathesis (RCM) [13] (12,13-double bond), aldol-type connections (between C2-C3,

Epothilones: Synthesis and Biological Evaluation
211
C3-C4, or C6-C7) or Wurtz type sp3-sp3-coupling (e.g. between C-9-10). So far only
the ring closures A±C shown in Scheme 3 have actually been performed. RCM has
received particular attention, the currently accepted mechanism of which [13] is
outlined in Scheme 4; the crucial intermediates are the metal carbenes 4-3 and 4-6
from which the metallacyclobutanes 4-3 and 4-4 are generated. The driving force to
shift the equilibrium towards the product is the evolution of gaseous ethylene. Two
catalysts (4-7 [14] and 4-8 [15]) have been introduced; the Grubbs catalyst is more
stable than the Schrock catalyst; however, for the formation of trisubstituted ole®ns
such as in epothilone D the much more active Schrock catalyst has to be used.
3. Stereocontrol
Altogether 7 stereogenic centers have to be installed with de®ned con®guration, in
addition to two double bonds. To achieve this there are two basic options [16]: 3.1.
asymmetric induction by chirality transfer from chiral auxiliaries or chiral catalysts
or 3.2. incorporation of stereogenic centers from the chiral carbon pool. The latter
option has the advantage that the absolute con®guration and the optical purity are
guaranteed.
4. Linear or convergent synthesis, number of steps
A convergent approach appears indicated, especially as the disconnections are
obvious (aldol-type additions between C6-C7, C3-C4, and C2-C3. Wittig ole®nations
between C12-C13 and C16-C17, addition of a C14-C13-C12-allylmetal to a C15-
Scheme 4. Mechanism of ring closing metathesis (RCM)

212
J. Mulzer
aldehyde). The fragments should be so chosen that the total number of steps should
not exceed 20±25 (as a rule of thumb, a good synthesis should require no more than
2±3 steps for each stereogenic center and/or functional group in the molecule.
Individual Syntheses
Syntheses Nos. 1 and 2 (Danishefsky)
Danishefsky and his group were the only ones to test all three ring closing processes
shown in Scheme 3. Their ®rst two syntheses [7,10,11b] were based on RCM and
macroaldolization as outlined in Scheme 5.
The macroaldolization was designed to close the C2-C3-bond via intermediate
5-1 by generating a C2-ester enolate in the presence of a nonenolizable sterically
hindered C3-aldehyde.
Intermediate 5-1 is assembled via a C11-C12 Suzuki coupling [17] of the vinyl
iodide 5-2 with organoborane 5-3. On the other hand, the RCM was performed with
a diole®n 5-4 which was prepared by the same aldol-type addition as described
before, however in an intermolecular fashion, using ester 5-5 and aldehyde 5-6. This
makes clear that the molecular fragments used in both approaches are quite similar
and, hence, exchangeable. Thus, 5-3 and 5-6 are prepared by analogous methodology
(Scheme 6).
The common intermediate is 6-7 which is procured via Danishefsky's general
polyketide strategy [18] centered around a hetero-Diels-Alder addition of the chiral
Scheme 5. Overview of Danishefsky's ®rst total syntheses of epothilone B

Epothilones: Synthesis and Biological Evaluation
213
Scheme 6. Danishefsky's hetero-Diels-Alder approach to the C3-C12 fragment
aldehyde 6-1 to diene 6-2. With good stereoselectivity adduct 6-3 is formed
which via routine steps furnishes the C3-C9-precursor 6-7. The metathesis chain
elongation is achieved by organometallic addition to the C9-aldehyde and
subsequent reduction of the 9-hydroxyl function to yield the ole®nic aldehyde 6-8.
The synthesis of fragments 5-2 and 5-5 starts with the aldehyde 7-1 (Scheme 7)
which is subjected to an asymmetric Keck-Umani-Ronchi-allylation [10] to form,
after acylation, 7-2 which is connected with aldehyde 6-8 via an aldoltype addition
to generate a 1:1-C3-epimeric mixture of diole®n 7-3. Functional group adjustment
generates the 5-carbonyl group, and this substrate is subjected to RCM with the
Schrock catalyst 4-8 to furnish a 1:1-mixture of epothilone D and its (E)-isomer
after desilylation. The chemoselective epoxidation of the 12,13-double bond is
achieved by dimethyldioxirane to generate the desired ꢀ-epoxide ( ˆ epothilone B)
with ꢀ/ꢁ-stereoselectivity of > 20:1 (Nicolaou et al. report a 5:1-ꢀ/ꢁ-ratio under
the same conditions [11a]). m-Chloroperbenzoic acid is similarly chemoselective at
15ꢀC, but inferior with respect to stereoselectivity. If applied at room temperature
it also generates the N-oxide (Scheme 32). The epoxidation of epothilone D as the
last step in an epothilone B synthesis has been adopted by all syntheses cited in
Ref. [11].
For the macroaldolization approach (Scheme 8), aldehyde 6-7 is converted to
the enol ether 8-1 via a Wittig reaction. Further functional group manipulation
leads to ole®n 8-2 which is the precursor of the in-situ-organoborane 5-3. The
second component of the Suzuki coupling [17], vinyl iodide 5-2, is prepared (Z)-
selectively from aldehyde 7-2 in three steps. The coupling of 5-2 and 5-3 furnishes
8-4 under retention of the (Z)-ole®n geometry. Deketalization gives the aldehyde 8-

214
J. Mulzer
Scheme 7. Danishefsky's RCM approach to epothilone B
Scheme 8. Danishefsky's macroaldolization approach to epothilone B

Epothilones: Synthesis and Biological Evaluation
215
5, whose macroaldolization furnishes a 1:1-C3-epimeric mixture of macrolide 8-6
which is converted into epothilone D (and eventually epothilone B).
Synthesis No. 3 (Nicolaou)
Very soon after Danishefsky the Nicolaou group was successful with an entirely
different strategy centered around a macrolactonization as the ring forming step
[8,11a]. A C6-C7-aldol-type addition was used as the key CC-connecting operation
of the synthesis. This aldol-type reaction has been ®rst described by our group [23b]
(Scheme 19) and has been applied in a related form both by Nicolaou and Schinzer
[10] in their syntheses of epothilones D and B. In the meantime, all syntheses of
epothilones utilize this aldol-type addition for connecting the C6-C7-bond. Nicolaou's
retrosynthetic planning is shown in Scheme 9.
Macrolactonization is performed on seco acid 9-1, which is assembled by
aldol-type addition of ketone 9-4 to aldehyde 9-2. Quite straightforwardly, 9-4 is
prepared via a Brown allylation [20] of aldehyde 9-5, whereas 9-2 is generated
from aldehyde 9-3 via an (E)-selective Wittig ole®nation. The synthesis of
aldehyde 9-2 is outlined in Scheme 10.
It starts from aldehyde 10-1 which is prepared analogously to Danishefsky's
intermediate 7-2 except for a TBS ether instead of an acetyl as the C15-O-protective
group. The ensuing Wittig reaction has to be performed with the stabilized ylide 10-2
in order to exert stereocontrol on the ole®n formation. With unstabilized ylides
the yield is much lower and (Z/E)-mixtures are obtained. The ester group in 10-3 is
converted into a methyl group, and the terminal ole®nic unit in triole®n 10-4
undergoes a chemoselective hydroboration-iodination sequence to give iodide 10-5
which is subjected to an Enders alkylation [21] with the SAMP-hydrazone 10-6.
Highly diastereoselective CC-elongation is achieved to afford 10-7 which is
converted into the desired aldehyde 9-2 via the nitrile.
The preparation of ketone 9-4 (Scheme 11) starts with a Brown allylation [20]
of the known ketoaldehyde 11-1 to form 11-2 which is converted into 9-4 in three
simple steps. The aldol addition with aldehyde 9-2 leads to a 3:1-mixture of 11-3
and its (6S,7R)-epimer. The major diastereomer is converted into seco acid 11-4
Scheme 9. Overview of Nicolaou's macrolactonization approach to epothilone B

216
J. Mulzer
Scheme 10. Nicolaou's synthesis of key aldehyde 9-2
Scheme 11. Nicolaou's aldol addition-macrolactonization approach to epothilone B

Epothilones: Synthesis and Biological Evaluation
217
Scheme 12. Rationalization of the stereochemical course of the key aldol addition
which is converted into epothilone D by Yamaguchi macrolactonization [22]. The
endgame of this synthesis is notable for the successful differentiation between the
four OTBS protective groups in 11-3. The primary 1-OTBS group is removed ®rst
with camphorsulfonic acid, the O-15-TBS group is removed with tetrabutyl-
ammonium ¯uoride after the generation of the C1-carboxylic acid. On the way
from 11-4 to epothilone D the O-7-TBS group is removed frist with tri¯uoroacetic
acid, whereas the removal of the O-3-TBS group requires prolonged treatment.
The stereoselectivity of the aldol-type addition has traditionally been
interpreted [23a] in terms of the Zimmerman-Traxler transition state model shown
in Scheme 12 which also demonstrates that the facial selectivity, i.e. the ratio of 11-
3 to epi-11-3, strongly depends on the substituents of the keto component, in
particular those at C1 and C3. A chelate has been postulated to explain this.
However, our initial study [23b] (Scheme 13) was performed with the achiral
nonchelating keto component 13-3. In this case a 4:1-ratio of 13-4 and epi-13-4
was obtained. It thus appears that the chelate postulated in Scheme 12 is not so
important; rather, it may be the 6-methyl substituent in the aldehyde component
that is primarily responsible for the facial selectivity.
Synthesis No. 4 (Grieco)
Grieco [11c] (Scheme 14) presented a formal total synthesis of epothilone B based
on Danishefsky's RCM strategy (Scheme 7).

218
J. Mulzer
Scheme 13. Aldol addition by Mulzer-Mantoulidis [23]
Scheme 14. Grieco's formal RCM synthesis of epothilone B
The innovation lies in the preparation of the C3-C13 fragment 14-9 which is
achieved in a straightforward stereoselective manner starting out with the Sharpless
asymmetric epoxidation (SAE) [24] of the known allylic alcohol 14-1 to give 14-2
with unspeci®ed optical purity. After chain elongation to enoate 14-3 an interesting
regio- and stereocontrolled epoxide opening with trimethylaluminum furnishes
adduct 14-4 which after oxidation to the aldehyde is subjected to a Roush
crotylation [25] to give isomer 14-5 as the sole product (matched case in the Roush
procedure!). Acetonide protection of the 1,3-diol and chain elongation furnishes
the unsaturated ketone 14-6 which is transformed into ole®n 14-7. Protective group
manipulations are required to generate aldehyde 14-9 which is subjected to an aldol
addition with Danishefsky's acetate 7-2 to give adduct 14-10 as an epimeric
mixture at C3. Further adjustment of the functional group furnishes diole®n 14-11

Epothilones: Synthesis and Biological Evaluation
219
Scheme 15. Schinzer's synthesis of Nicolaou's macrolactonization key intermediate
which is identical with the intermediate in Danishefsky's synthesis (Scheme 7).
Altogether, Grieco's synthesis of 14-11 requires 23 steps compared to 19 steps in
Danishefsky's route.
Synthesis Nos. 5 and 6 (Schinzer)
Schinzer [11d] described a macrolactonization (Scheme 15) and a RCM (Scheme
16) approach to epothilone B. The crucial CC-connecting step in both approaches
is the familar aldol-type addition of a C6-enolate to a C7-aldehyde. The C1-C6-
component 15-5 is prepared from bromoester 15-1 which is converted into the
ole®nic aldehyde 15-3 via a Reformatsky reaction. The stereogenic center at C3 is
generated via Braun's HYTRA methodology [26] which furnishes hydroxyester
15-4 in high optical purity. Three straightforward steps lead to 15-5. The aldehyde
component 9-2 is identical with that used by Nicolaou and is prepared in close
analogy to Danishefsky's procedure, using an organometallic CC-coupling of the
vinyl iodide 15-6 with the organozinc component 15-7. Aldol coupling proceeds
with an exceptionally high diastereocontrol to furnish Nicolaou's intermediate 11-3
which has already been converted into epothilone B by macrolactonization. In the
RCM variation, 15-5 is connected with aldehyde 16-3 by a highly diastereose-
lective aldol addition to give 16-4 which is concerted into acid 16-5. Esteri®cation
with the known [10] chiral alcohol 16-6 furnishes the Danishefsky-Grieco RCM
intermediate 14-11.
Synthesis No. 7 (Danishefsky)
In their latest approach [11e] the Danishefsky group have also used the aldol-type
connection between C6 and C7 (Scheme 17) in a very direct approach. The ketoester
17-1 serves as an achiral version of the C1-C6-fragment. After conversion into the
enol ether 17-2, an aldol addition is performed with aldehyde 17-3 to give adduct 17-

220
J. Mulzer
Scheme 16. Schinzer's synthesis of Danishefsky's RCM key intermediate
Scheme 17. Danishefsky's aldol addition-macrolactonization approach to epothilone B
4 with acceptable stereoselectivity. 7-OH-protection and chain elongation with vinyl
iodide 15-6 as before furnishes the tricarbonyl compound 17-6 which is subjected to
a chemo- and stereoselective Noyori hydrogenation [27] of the C3-carbonyl to give
17-17 which is O-silylated and cyclized to macrolide 17-8. Deprotection leads to
epothilone D.

Epothilones: Synthesis and Biological Evaluation
221
Synthesis No. 8 (Mulzer 1)
Mulzer's ®rst approach [28] (Schemes 18±22) is based on the alkylation of sulfone
18-8 with allyl iodide 19-9 to form the carbon skeleton of epothilone D. The
preparation of 18-8 starts from the C3±C9-intermediate 13-4 [23] which is
converted into the cyclic acetal 18-1 and oxidized to the C3-aldehyde. Brown
allylation gives 18-2 with 4:1-diastereoselectivity at C3. Readjustment of the
protective groups leads to hydoxy ketone 18-4 which is in equilibrium with the
cyclic acetals 18-5 and 18-6. Tosylation of the primary hydroxyl function in 18-4
proceeds slowly to furnish tosylate 18-7 which is alkylated with the anion of
methylphenyl sulfone to give 18-8. The allyl iodide 19-9 is prepared from (S)-
malic acid (19-1) via the lactone 19-3 and the methyl ketone 19-5 Wittig reaction
with the tributylphosphonium ylide 19-6 furnishes the (E)-ole®n 19-7 selectively.
Swern oxidation, followed by a (Z)-selective Still-Gennari-Horner ole®nation [29]
generates ester 19-8 which is converted into 19-9 eventually. On deprotonation
with butyllithium, 19-8 presumably undergoes an elimination of LiOTBS from the
6,7-positions to form 20-1. After a second deprotonation with butyllithium 20-1
adds 19-9 to form 20-3. A second equivalent of 19-9 is consumed by an SN2
reaction with LiOTBS to form 20-2. This failure obviously results from the 1,5-
arrangement of the carbanion and an acidic 6-H in 20-1. So the plan was changed
to the effect to alkylate 19-9 with the smaller carbonyl free sulfonyl fragment 21-2
which is generated from the TBDPS-protected Roche ester. Alkylation with 19-9
gives 21-3 in good yield. Reductive desulfonation and oxidation of the terminal
primary alcohol furnishes aldehyde 21-4, identical with Nicolaou's aldehyde 9-2
except for the PMB protective group. Aldol addition with the enolate of ketone 9-4
affords adduct 22-1 which is oxidized to carboxylic acid 22-2. However, all
Scheme 18. Mulzer's unsuccessful sulfone alkylation approach

222
J. Mulzer
Scheme 19. Mulzer's synthesis of the key C21-C11-allyliodide fragment
Scheme 20. Unsuccessful Sulfone Coupling
attempts to remove the 15-O-PMP group from 22-2 to get the seco acid failed.
DDQ oxidation, for instance, leads to the 15-ketone 23-3.
Thus, the overall sequence was repeated [11f] with a 15-OTBS instead of the
unsuitable PMB protective group via the same intermediates as described
previously (Schemes 19 and 21, yields in parentheses). Additionally, a novel
synthesis of the C1-C6-fragment 9-4 was developed following a procedure
described by Kiyooka et al. [30] (Scheme 23).

Epothilones: Synthesis and Biological Evaluation
223
Scheme 21. Successful sulfone coupling generates the C21-C7-key aldehyde 21-4
Scheme 22. Endgame fails due to Unsuitable 15-OPMB protecting group
To this end, aldehyde 23-1 is treated with ketene acetal 23-2 in the presence of
borane 23-3, easily prepared in situ from N-Ts-D-valine and diborane. Hydroxy
ester 23-4 is formed in high optical purity. Unfortunately, the addition of ethyl
magnesium bromide to ester 23-5 does lead to ketone 9-4, but larger amounts of the
diethyl alcohol are also produced. Reagent 23-6, however, exclusively furnishes the
methyl ketone 23-7 after methanolysis, which after C-methylation of the enolate
gives 9-4 eventually. With Nicolaou's fragments 9-4 and 9-2 at hand, seco acid 24-
2 is prepared Macrolactonization via Keck's method [31] furnishes macrolide 24-3
which is deprotected and epoxidized to give epothilone B in an unprecedentedly

224
J. Mulzer
Scheme 23. Novel aldol addition approach to the C1-C6-ketone fragment
Scheme 24. Successful sulfone coupling and macrolactonization approach to epothilone B
short and stereoselective sequence. The main advantage over Nicolaou's synthesis
is the concise access to ketone 9-4 and the (E)-selective preparation of aldehyde
9-2. Additionally, the stereogenic centers at C15 and C8 are adapted from the chiral
carbon pool instead of using chiral auxiliaries.
Synthesis No. 9 (White)
J. D. White reported a synthesis of epothilone B [11g] which is based on our
fragment 18-3. The carbon skeleton is constructed by forming a C9-C10-double
bond via a Wittig reaction between the C1-C9-aldehyde 25-2 and the C10-C21-
phosphonium ylide 25-9 (Scheme 25). More speci®cally, compound 18-3 is
oxidized to the carboxylic ester 25-1 and then converted into aldehyde 25-2. The
phosphonium ylide 25-9 is prepared from the Evans oxazolidinone 25-3 which is
converted into 25-5 via a conjugate addition of cuprate 25-4. The introduction of
the 15-hydroxyl function is achieved via enolate hydroxylation with Davis` reagent
[32] to furnish 25-6 after O-silylation. Oxazolidinone removal with mercaptide and

Epothilones: Synthesis and Biological Evaluation
225
Scheme 25. J. D. White's epothilone B synthesis
methyl cuprate addition to the thioester leads to the methyl ketone 25-7 which
is converted into 25-8 via ole®nation. Four additional steps are required to
generate phosphorane 25-9 which is treated with aldehyde 25-2 to give the (Z)-
ole®n 25-10 selectively. Macrolactonization is achieved in low overall yield from
25-10, and after deprotection the super¯uous 9,10-double bond is selectively
hydrogenated with diimide to give epothilone D. White's synthesis is distinguished
by the high stereocontrol exerted on all stereogenic units (double bonds and
stereogenic centers); however, it requires altogether more than 30 steps when
starting from commercially available materials. In a later version [11h] the crucial
CC coupling was performed with alkyne 25-12 and the allylic bromide 25-13 to
give diene-yne ester 25-14 which was selectively hydrogenated over Lindlar's
catalyst to furnish 25-10. In this way the obvious disadvantages of the Wittig
reaction were circumvented.

226
J. Mulzer
Synthesis No. 10 (Mulzer 2)
So far, all syntheses of epothilone B have involved the (more or less)
stereoselective epoxidation of epothilone D in the last step, obviously to avoid
undesired side reactions of the presumably labile 12,13-epoxide. In their second
approach to epothilone B (Scheme 26) the Mulzer group wanted to test this lability
by generating the 12,13-epoxide relatively early in the synthesis and exposing it
deliberately to the hardships of a long sequence [33]. The synthesis starts from
protected lactic ester 26-1 which in a one-pot operation is reduced to the aldehyde
and treated with allylmagnesium bromide to furnish the chelate Cram adduct 26-2
selectively. Oxidation to the aldehyde and addition of isopropenyl magnesium
bromide gives allylic alcohol 26-4 as a 1:1-epimeric mixture which is subjected to
a Claisen-Johnson rearrangement to give ester 26-5. After elaboration of 26-5 to
the methyl ketone 26-6 the double bond is dihydroxylated according to Sharpless'
procedure [34] to give the diol 26-7 which is subjected to a Wittig reaction with 19-
6 to furnish a 3:1-mixture of the lactones 26-8 and 26-9 easily separable by
chromatography. NMR analysis clari®ed that the Wittig reaction has been > 98%
(E)-selective, whereas the dihydroxylation has only furnished a 3:1-diastereomeric
mixture. To avoid the loss of one third of the material, the undesired stereoisomer
26-9 is transformed into the desired one, 26-8, by double inversion at C12 and C13
via the intermediates 27-1 and 27-2. The synthesis is continued with the conversion
of diol 26-8 into the epoxide 27-3. The ester is reduced to the aldehyde which is
Scheme 26. Mulzer's second epothilone B synthesis: The early epoxide approach part I: Claisen
Rearrangement-Dihydroxylation

Epothilones: Synthesis and Biological Evaluation
227
Scheme 27. Chain elongation via Oppolzer's sultam
Scheme 28. Completion of the synthesis: Crucial role of a 7-OTroc-15-OTES protecting group
strategy
subjected to a Horner ole®nation with Oppolzer's phosphonate 27-4 [35] to
generate the enoyl sultam 27-5. 1,4-Hydride addition to the 8,9-double bond and
stereoselective methylation of the resulting 7,8-enolate furnishes 27-6. Next, the
15-OTBS group is exchanged for a OTES group, and this compound is reduced to
the aldehyde 27-7. Aldol addition with ketone 11-2 produces 28-1 with high
stereoselectivity. After transformation of 28-1 into the seco acid 28-2, Yamaguchi-

228
J. Mulzer
macrolactonization gives a satisfactory yield of 28-3 which is smoothly
deprotected to form epothilone B. It thus turns out that the epoxide is rather
stable and tolerates a number of oxidizing, reducing and anionic reagents. The only
epoxide opening process found was the desired conversion of 27-2 into 26-8. The
early introduction of the epoxide has been of advantage in the overall synthesis.
For instance, the formation of thiazol-N-oxides previously observed in the mCPBA
epoxidation of the 12-13-double bond [37] has been avoided as well as the
formation of the (12S,13R)-epoxide which is hard to separate from the correct
stereoisomer. Additionally, the diastereocontrol of the aldol addition with the
epoxyaldehyde is signi®cantly better than it is with the ole®nic aldehyde [11]. The
use of ester 26-5 as an intermediate allows application of the Claisen
rearrangement as an ef®cient chain elongation procedure. Also, it is possible to
introduce additional double bonds after the epoxide has been generated which may
be useful for the preparation of novel derivatives.
In Table 1 the total syntheses of epothilone B reported so far are compared in
terms of (1) total number of the synthetic operations involved, (2) total number of
the steps in the longest linear sequence, (3) overall yield along the longest linear
sequence, and (4) stereocontrol over each stereogenic center. No evaluation,
however, will be tried in this review, as all syntheses are viable and have been
successfully applied and all syntheses certainly have merits and disadvantages.
Table 1. Comparison of the known total syntheses of epothilone B
Synthesis Author
Total
Total ee at ee at ee at ee at ee at
ee at
No.
(year)
Number of
stepsa
yield C-3
(%)
C-6
C-7
C-8
C-12/13 C-15
1
2
Danishefsky 27/21
(1997)
4±6
ca 1
3
40
?
?
?
0/70
70
> 95
> 95
> 97
> 95
100b
100b
70
Danishefsky 30/26
(1997)
34
?
?
?
3
Nicolaou
(1997)
28/24
30/26
34
> 98
20
50
50
> 95
80
84
68
60
>95
>90
> 98
> 95
> 95
80
70
4
Grieco
(1998)
4
> 95
80
0/70
70
5
Schinzer
(1998)
0.6
ca 1
5±7
8
96
6
Schinzer
(1998)
25
96
84
0/70
70
7
Danishefsky 27/17
80
>95
60
68
(1998)
8
Mulzer
(1998)
White
23
95
100b
100b
>98
66
100b
?
9
29
2
80
>95
>90
70
(1998)
Mulzer
(1998)
10
32/26
3±5
>98
>98
80
a
Total number of steps/number of steps in the longest linear sequence; btaken from the chiral carbon
pool

Epothilones: Synthesis and Biological Evaluation
229
Synthesis of Derivatives
Nicolaou made extensive use of his epothilone synthesis [11a] for preparing
derivatives in the hope of improving the biological activity. A host of epothilone A
analogues were produced using combinatorial chemistry especially by applying the
Wittig reaction to construct the C12, C13-ole®nic bond [3]. In the epothilone B
series, however, non-combinatorial methods were applied. For instance, inter-
mediate 10-3 (Scheme 29) served as the starting material for the preparation of
epothilones 29-5 to 29-7, with a halo-methylene unit in position 12 [36].
The synthesis takes advantage of the 12-hydroxymethylene unit which allows
the application of Sharpless' AE reaction [24] to generate the epoxide in 29-4 from
29-3 with high stereoselectivity. The hydroxyl-halogen exchange takes place
without affecting the epoxide. Analogues 30-1 to 30-3 are prepared similarly using
the appropriate ole®nating component (Scheme 30).
The 12,13-trans-epothilones C and D (31-1 and 31-2) are available as side
products from the (E/Z)-12,13-ole®n mixtures generated from unstereoselective
Wittig or RCM reactions in Nicolaou's, Danishefsky's, and Schinzer's total
syntheses [10, 11a-d]. The epoxides 31-3 and 31-4 are obtained, along with the ꢁ-
epoxides, from the (unstereoselective) epoxidation of 31-1 and 31-2 (Scheme 31).
Interesting analogues have also been prepared by partial synthesis from the
È
naturally occuring epothilones. For instance, Ho¯e et al. [37] reported that
epothilone B (2-4) is converted into the N-oxide 32-1 with m-chloroperbenzoic
acid (Scheme 32). On heating with acetic anhydride, 32-1 undergoes a Polonovsky-
Scheme 29. Nicolaou's synthesis of epothilone B derivatives

230
J. Mulzer
Scheme 30. Additional epothilone B derivatives from the Nicolaou group
Scheme 31. Epothilone B derivatives with 12,13-trans-stereochemistry
È
Scheme 32. Ho¯e's partial synthesis of new epothilone derivatives
like rearrangement to form the 21-acetate 32-2 whose saponi®cation leads to
epothilone F (2-6) which has also been detected as a natural metabolite. An
analogous sequence can be performed with epothilone C (2-2). Interestingly, the
oxidation with peracid leads to the N-oxide 32-3 chemoselectively without
touching the 12,13-double bond. Polonovsky-type rearrangements also occur with
anhydrides or sulfonyl chlorides to furnish 21-substituted derivatives, e.g., 32-4.

Epothilones: Synthesis and Biological Evaluation
231
Scheme 33. Partial synthesis of epothilone B lactams
The epothilone lactam analogues (aza-epothilones) 33-3 and 33-4 have been
prepared by an industrial research group (Scheme 33) [38]. They use a palladium
catalyzed SN2-type ring opening of the macrolide [39] at the C15-17-allylic ester
moiety with sodium azide as the nucleophile (Scheme 33). Azide 33-1 is formed
unter retention of con®guration at C15. Reduction of the azide furnishes the amino
acid 33-2 which is cyclized to the lactam 33-3 under carboxyl group activation.
The epoxide can be deoxygenated to the ole®n 33-4 by a tungsten hydride species
generated in situ from tungsten chloride and n-butyl lithium. This deoxygenation
can also be applied to the conversion of epothilone B into epothilone D or of
epothilone A into epothilone C under retention of the ole®n con®guration.
Danishefsky prepared the promising 12,13-benzo-analogue 34-7 of epothilone
B (Scheme 34) [40] essentially along the same route as that one used for epothilone
D (Scheme 17).
Biological Activity
As mentioned above, the initial test for biological activity of the epothilones was
the microtubule aggregation test (`tubulin assay') developed for paclitaxel. At that
time, it was taken for granted that high microtubule aggreation would always go
along with high antitumor activity. Very soon, however, it turned out that ± at least
for paclitaxel analogues ± high activity in terms of microtubule aggreagation is not
always accompanied by high cytostatic activity [41]. However, this could not be
con®rmed for the epothilones, where a high activity in the tubulin assay does
indeed normally lead to high cytotoxicity (Table 2) [42,43].
This is particularly true for epothilones B and D and also the derivatives
30-1, 30-2 and 30-3 [36]. Although there are signi®cant differences with respect
to the individual epothilones they are apparently about as active or even more
active than paclitaxel both in the tubulin assay and in the cytotoxicity tests
(Table 3).

232
J. Mulzer
Scheme 34. Danishefsky's synthesis of a benzo analogue of epothilone D
Table 2. Relative biological activities of epothilones and derivatives and paclitaxel [42,37]
Paclitaxel
15a
14a
2.0cI
2.0cI
50cII
19cII
43cIII
4.2cIII
>100cIV
2.4cIV
Epothilone
A (2-1)
Epothilone
B (2-4)
4.0a
3.3a
25a
39a
0.040cI
2.0cI
25cI
0.035cII
33cII
0.045cIII
3.5cIII
75cIII
0.040cIV
1.5cIV
22cIV
Epothilone
D (2-5)
Epothilone
C (2-2)
>100cII
>100cII
12,13-trans-
Epothilon
D (31-4)
12,13-trans-
Epothilone
B (31-2)
Epothilone
E (2-3)
48cI
75cIII
24cIV
22a
±
3.5cI
30cII
5.5cIII
3.0cIV
>100cI
0.54cI
0.40cI
0.12cI
50cII
20cIII
±
±
±
±
30-1
Highly
activeb
Highly
activeb
Highly
activeb
2.8cII
1.2cII
0.35cII
1.5cIII
2.5cIII
0.14cIII
30-2
30-3
a
EC50 (mg) ˆ quantity required for causing 50% of the tubulin to assemble in polymers; binduction of
c
tubulin assembly; inhibition of human ovarian carcinoma cell growth, IC50 (ng); I ˆ cell line 1A9,
II ˆ cell line 1A9PTX10, III ˆ cell line 1A9PTX22, IV ˆ MDR line A2780AD

Epothilones: Synthesis and Biological Evaluation
233
Table 3. In vitro cytotoxicity of epothilone derivatives and paclitaxel [43]
Compound
Cytotoxicity Ia
Cytotoxicity II
Cytotoxicity III
Paclitaxel
Epothilone A
Epothilone B
Epothilone C
Epothilone D
Epothilone E
Epothilone F
32-1
80
8
12
2
4
1.4
0.2
60
1.4
100
20
1.2
40
24
20
40
10
6
3
1.0
2
0.2
1.5
20
4
32-2
32-3
400
1400
1100
130
200
600
600
1100
32-4
a
Cytotoxicity was determined as IC50 (nM) with the following cell lines: I ˆ mouse ®broblasts
(ATCC CCL 1), II ˆ human cervix carcinoma (DSM ACC 158), III ˆ human lung carcinoma (DSM
ACC 107)
Preliminary SAR results
Using a broad selection of epothilone B analogues, SAR studies have been
performed by Nicolaou [3a] and, more recently by Danishefsky [3b]. Roughly the
molecule may be subdivided into three zones 1±3 (Fig. 4) whose modi®cations
have very different effects on the microtubule stabilization and cytotoxicity. Zone
1, i.e., the polyketide section, is the most intolerant one with respect to structural
modi®cations. For example, inversion of the stereochemistry at C3 or reduction of
the C5 carbonyl resulted in serious arrest of activity. Similarly, analogues with
functionalities at C3, C5, C6, C7, and C8 removed or modi®ed demonstrated
reduced tubulin binding and cytotoxicity. Zone 2 is remarkably tolerant to
modi®cation. For instance, the (Z)-deoxy derivative (epothilone B) and also the
(E)-deoxy derivative show signi®cant activity, and substitution at C12 with ethyl or
propyl and other alkyl residues was well tolerated both in the epoxy and the desoxy
series. Also, the con®guration of C15 appears not crucial for biological activity.
Polar substituents in the 13-appendage enhance the susceptibility to succumb to
MDR. Zone 3 eventually, i.e. the aryl substituents must have an aryl group and the
correct ole®n spacer, but within these limits variations are tolerated.
Fig. 4. Pharmocophoric zones in epothilone B

234
J. Mulzer
Second generation biodata of epothilones;
comparison with other antitumor drugs
An extremely important aspect is the comparison of the epothilones with other
antitumor drugs with respect to multidrug resistance (MDR) [3b, 44] (Table 4). A
comparison of the epothilones, paclitaxel, and two classical antitumor drugs
(actinomycin D and adriamycin) shows that although all these compounds do
exhibit extremely high cytotoxicity the epothilones are signi®cantly superior in
terms of multidrug resistance towards several tumor cell lines. Quite interestingly
epothilone D (2-5) now emerges as the top candidate of all compounds tested so
far, as 2-5 combines high antitumor activity with high multidrug resistance. It is
clearly superior to epothilone B (2-4).
This is in clear contrast to the data of Table 1 where epothilone D shows a
signi®cantly lower cytotoxicity than epothilone B, although the activity in the
tubulin assay is about the same for both compounds. The superiority of epothilone
D over epothilone B is more impressively borne out when discussing the results of
in vivo tests with normal athymic nude mice bearing human mammary
adenocarcinoma MX-1 xenografts (Table 5).
When a daily dose of 0.6 mg/kg of epothilone B was applied to normal nude
mice intraperitonally (i.p.), all mice were dead after seven days. When, however
25 mg/kg of epothilone D was applied i.p., all mice survived. More important,
epothilone B had only a marginal therapeutic effect, whereas epothilone D led to a
drastic reduction of the tumor size, so that one out of six mice was without a
detectable tumor after 35 days. When considering tumor therapy combined with
low toxicity, epothilone D is superior not only to epothilone B, but also to
paclitaxel and other antitumor drugs (Table 4). It has also been demonstrated that,
again in the nude mice test, epothilone D is curative against human tumor
Table 4. Comparison of in vitro growth inhibition potency of epothilone derivatives against various
parent and drug-resistant tumor cell lines [44]
Compound
DC-3F
0.0037
DC-3F/ADX P338/0
P338/Adr SK-N-As SK-N-FI MCF-7
MCF-7/Adr
2-1
0.053
0.0018
0.0010
(5.3 x)
0.012
0.004
0.073
0.021
0.0016
0.023
(1.9 x)
0.010
(25 x)
0.223
(3.1 x)
0.046
(2.2 x)
0.130
(80 x)
0.0030
0.0005
0.032
0.0094
(3.1 x)
0.0027
(5.4 x)
0.144
(14.5 x)
0.017
2-4
0.0006
0.011
0.00029 0.0016
(5.5 x)
(28 x)
0.042
2-2
0.0213
0.0068
0.0029
0.0125
(0.59 x)
0.0042
(0.62 x)
0.326
(3.9 x)
(4.5 x)
0.0071
(2.4 x)
0.150
2-5
0.00097 0.00091
(0.9 x)
0.0029
0.0033
Paclitaxel
0.095
32.0
(338 x)
(111 x)
(46 x)
Actinomycin D 0.00044 0.572
(13000 x)
2.236
(124 x)
0.00015 0.0012
(8 x)
0.00085 0.0119
(14 x)
0.00068 0.00167
(2.5 x)
Adriamycin
0.018
0.0055
2.65
0.077
1.42
0.057
0.216
(482 x)
(18.4 x)
(3.8 x)
Numbers are IC50 (mg); numbers in parentheses are fold of resistance based on the IC50 ratio when compared with
the corresponding parent cell lines except for P388/0 and P388/Adr, and XTT assay was used