Concise asymmetric syntheses of radicicol and monocillin I

Article abstract of DOI:10.1021/ja011364+

Radicicol (1) exhibits potent anticancer properties in vitro, which are likely to be mediated through its high affinity (20 nM) for the molecular chaperone Hsp90. Recently, we reported the results of a synthetic program targeting radicicol (1) and monocil

Full text of DOI:10.1021/ja011364+

J. Am. Chem. Soc. 2001, 123, 10903-10908
10903
Concise Asymmetric Syntheses of Radicicol and Monocillin I
Robert M. Garbaccio, Shawn J. Stachel, Daniel K. Baeschlin, and Samuel J. Danishefsky*
Contribution from The Laboratory for Bioorganic Chemistry, The Sloan-Kettering Institute for Cancer
Research, 1275 York AVenue, New York, New York 10021
ReceiVed June 4, 2001
Abstract: Radicicol (1) exhibits potent anticancer properties in vitro, which are likely to be mediated through
its high affinity (20 nM) for the molecular chaperone Hsp90. Recently, we reported the results of a synthetic
program targeting radicicol (1) and monocillin I (2), highlighted by the application of ring-closing metathesis
to macrolide formation. These efforts resulted in a highly convergent synthesis of radicicol dimethyl ether but
failed in the removal of the two aryl methyl ethers. Simple exchange of these methyl ethers with more labile
functionalities disabled a key esterification in the initial route. Through extended experimentation, a successful
route to both natural products was secured, along with some intriguing results that emphasize the implications
of this design on a broad range of fused benzoaliphatic targets, including analogues of these natural products.
Introduction
Radicicol1,2 (1) and monocillin I2 (2) are resorcylic macrolides
which can both be isolated from Monocillium nordinii2 (Figure
1). While the skeletal structure of radicicol was determined in
1964,3 its relative and absolute stereochemical configuration was
not unambiguously established until 1987.4 The structure of
monocillin I (2) was confirmed by its direct conversion into
radicicol (1).4 Affirmation of these structures was achieved by
their only total synthesis accomplished by Lett and Lampilas.5
Both radicicol (1) and monocillin I (2) exhibit a variety of
antifungal and antibiotic properties not shared by other members
of this class of natural products. Recently, the antitumor
properties of radicicol have come into focus. Its ability to
suppress the transformed phenotype caused by various onco-
genes such as src, ras, and raf has been linked to its high affinity
binding (20 nM) and inhibition of the Hsp90 molecular
chaperone.6 This “anti-chaperone” activity may stimulate deple-
tion of oncogenic proteins and could therefore be of clinical
interest. Importantly, while radicicol displays this and other
activities in vitro, the compound has not yet exhibited in vivo
antitumor activity in animal models,7 though some derivatives
do manifest in vivo efficacy.8 In addition, recent reports on the
neurotrophic ability of geldanamycin (3) and radicicol (1) have
merited attention.9
Figure 1. Structures of radicicol, monocillin I, and geldanamycin.
to bind to Hsp90 (1.2 µM). In addition, a novel class of synthetic
nucleotide analogues were shown to bind to Hsp90 (15-20 µM)
and inhibit the proliferation of cancer cells.11 Hence, the
combination of growing interest in anticancer activity mediated
by Hsp90 inhibition, the high affinity of radicicol (1) for this
chaperone, and the absence of an efficient synthetic route
prompted our total synthesis efforts. Through these efforts, we
hoped to gain access to the radicicol system and thereby gain
the opportunity to address various biological issues starting with
SAR analysis. These efforts could also enable the design of
bifunctional agents by linking radicicol (1) with other ligands,
as has been done with geldanamycin (3),10 thus expanding our
portfolio of selective Hsp90 inhibitors.
Our interest in inhibitors of Hsp90 began with efforts
involving geldanamycin (3),10 an antitumor antibiotic also shown
Synthesis of Radicicol Dimethyl Ether. We focused on a
modular route to monocillin I (2) as well as radicicol (1) that
would be favorable for analogue production. A strategy was
devised utilizing a highly convergent coupling sequence for three
key intermediates (4-6) (Scheme 1).12
(1) Delmotte, P.; Delmotte-Plaque´e, J. Nature 1953, 171, 344.
(2) Ayer, W. A.; Lee, S. P.; Tsuneda, A.; Hiratsuka, Y. Can. J. Microbiol.
1980, 26, 766.
(3) Two independent efforts confirmed the structures of monorden and
radicicol and found them to be identical. McCapra, F.; Scott, A. I.; Delmotte,
P.; Delmotte-Plaque´e, J.; Bhacca, N. S. Tetrahedron Lett. 1964, 869.
Mirrington, R. N.; Ritchie, E.; Shoppee, C. W.; Taylor, W. C.; Sternhell,
S. Tetrahedron Lett. 1964, 365.
(4) Cutler, H. G.; Arrendale, R. F.; Springer, J. P.; Cole, P. D.; Roberts,
R. G.; Hanlin, R. T. Agric. Biol. Chem. 1987, 51, 3331.
(5) Lampilas, M.; Lett, R. Tetrahedron Lett. 1992, 33, 773 and 777.
(6) Roe, S. M.; Prodromou, C.; O’Brien, R.; Ladbury, J. E.; Piper, P.
W.; Pearl, L. H. J. Med. Chem. 1999, 42, 260.
(7) Kwon, H. J.; Yoshida, M.; Nagaoka, R.; Obinata, T.; Beppu, T.;
Horinouti, S. Oncogene 1997, 15, 2625.
(9) Gold, B. G.; Densmore, V.; Shou, W.; Matzuk, M. M.; Gordon, H.
S. J. Pharmacol. Exp. Ther. 1999, 289, 1202. Gold, B. G. Compositions
and Methods for Promoting Nerve Regeneration by Disrupting the Steroid
Receptor Complex and Stimulating MAP Kinase/Kinase Activity. Patent
WO 2001003692, Jan 2001.
(10) Kuduk, S. D.; Zheng, F. F.; Sepp-Lorenzino, L.; Rosen, N.;
Danishefsky, S. J. Biorg. Med. Chem. Lett. 1999, 9, 1233. Kuduk, S. D.;
Harris, C. R.; Zheng, F. F.; Sepp-Lorenzino, L.; Ouerfelli, Q.; Rosen, N.;
Danishefsky, S. J. Bioorg. Med. Chem. Lett. 2000, 10, 4325.
(11) Chiosis, G.; Timaul, M. N.; Lucas, B.; Munster, P. N.; Zhewng, F.
F.; Sepp-Lorenzino, L.; Rosen, N. Chem. Biol. 2001, 8, 289.
(12) Garbaccio, R. M.; Danishefsky, S. J. Org. Lett. 2000, 2, 3127.
(8) Soga, S.; Neckers, L. M.; Schulte, T. W.; Shiotsu, Y.; Akasaka, K.;
Narumi, H.; Agatsuma, T.; Ikuina, Y.; Murakata, C.; Tamaoki, T.; Akinaga,
S. Cancer Res. 1999, 59, 2931.
10.1021/ja011364+ CCC: $20.00 © 2001 American Chemical Society
Published on Web 10/10/2001

10904 J. Am. Chem. Soc., Vol. 123, No. 44, 2001
Garbaccio et al.
Scheme 1
Scheme 3a
The first coupling would entail esterification of an ap-
propriately substituted benzoic acid (4) with an optically pure
secondary alcohol (5) containing all three stereogenic centers
of radicicol. The second coupling requires chemo- and regio-
selective addition of a masked acyl anion equivalent (such as
6) to the benzyl chloride carbon in the presence of a vinyl
epoxide. In the concluding phase, stereospecific ring-closing
metathesis of a diene with a vinyl epoxide would lead to the
desired 14-membered lactone containing the cis-trans diene.
This route would lead to monocillin I (2), which can subse-
quently be converted to radicicol (1) by aromatic chlorination.4
a (a) LiCl, DIPEA (EtO)2P(O)CH2CO2Et, 95%; (b) DIBALH, -20
°C, 96%; (c) (+)-DET, Ti(OiPr)4, TBHP, 90%, >95% ee; (d)
SO3‚pyridine, Et3N, DMSO, 90%; (e) Ph3PCH3Br, NaHMDS, 0 °C,
82%; (f) TBAF, 89%.
epoxide, thus eliminating the Mitsunobu reaction as a method
for subsequent esterification with the aromatic fragment. It will
become evident later that this was a significant limitation. This
route was left unoptimized in favor of a more general and
selective route.
Scheme 2a
Given the low diasteroselectivity of the homoallylic epoxi-
dation and the relative inflexibility of this approach to the
generation of analogous fragments, an alternate sequence was
explored. This route closely follows that described by Waldmann
and co-workers13 (Scheme 3) and initiates with aldehyde (8).
Wadsworth-Horner-Emmons homologation of 8 under Roush-
Masamune conditions14 (LiCl, DIPEA, 95%) followed by
reduction (DIBALH, 96%) yielded trans-allylic alcohol (13).
Sharpless asymmetric epoxidation ((-)-DET, Ti(OiPr)4, TBHP,
90%) gave the desired epoxyalcohol (14) with excellent (>20:
1) selectivity. Epoxyalcohol 14 was then subjected to Parikh-
Doering oxidation (SO3‚pyridine, Et3N, DMSO, 90%)15 and the
resultant aldehyde converted to the vinyl epoxide (15) (PPh3-
CH3Br, NaHMDS, 82%). Removal of the TBDPS group
proceeded smoothly (n-Bu4NF, 89%) to yield the secondary
alcohol (5), which carried all three stereocenters appropriate for
reaching radicicol (1).
a (a) TBDPSCl, imidazole, >95%; (b) DIBALH, -78 °C, 92%; (c)
allyl diethyl phosphonate, 80%, 8:1; (d) TBAF, 85%; (e) VO(acac)2,
CH2Cl2, -30 °C, 2:1.
The second required coupling partner for this design was an
appropriate acyl anion equivalent. Dienyl dithiane (6) was
envisaged to be the optimal choice. This building block could
serve two critical functions. First, it would allow for nucleophilic
introduction of the required diene. Second, it was hoped that
the dithiane function would mask the latent ketone. Previous
work in our laboratory as well as others revealed that the
combination of the free ketone and ester (see asterisks in i, eq
1) before the macrocyclic ring has been established possesses
a high propensity to cyclize to a coumarin (ii). Ring opening
of such a structure such as ii to gain access to substances useful
for macrocyclization can become problematic.
The first and most complex coupling partner needed for this
synthetic plan was the enantiomerically homogeneous vinyl
epoxide 5. A short substrate-directed approach was devised. The
route relies on directed epoxidation of a homoallylic alcohol
(Scheme 2). It was hoped that the orientation of the methyl
group (equatorial versus axial) in the chairlike transition state
would direct the epoxidation to afford the desired vinyl epoxide.
In the event, (R)-3-hydroxybutyric acid methyl ester (7) was
protected (TBDPSCl, imidazole, >95%), and the product was
reduced at low temperature (DIBALH, -78 °C, 92%) to provide
aldehyde (8). Chain extension of this aldehyde with allyl diethyl
phosphonate (n-BuLi, 80%, 8:1 trans:cis) followed by removal
of the TBDPS protecting group (n-Bu4NF, 85%) gave the
required homoallylic alcohol 10. VO(acac)2-catalyzed epoxi-
dation of 10 at 0 °C (TBHP, CH2Cl2, 12 h) afforded desired
epoxide 11 only in a 3:2 ratio. Conduct of the reaction at -30
°C (5 days) gave only minor improvement (2:1). Varying solvent
had little effect on the outcome of this transformation. One
potential shortcoming of this route was that it would require
hydroxyl R stereochemistry in order to produce the all R vinyl
(13) Schlede, U.; Nazare´, M.; Waldmann, H. Tetrahedron Lett. 1998,
39, 1143.
(14) Masamune, S.; Roush, W. R. Tetrahedron Lett. 1984, 25, 2183.
(15) Parikh, V. E., Jr.; Doering, W. J. Am. Chem. Soc. 1967, 89, 5505.

Syntheses of Radicicol and Monocillin I
J. Am. Chem. Soc., Vol. 123, No. 44, 2001 10905
Scheme 4a
lectivity.19 Attempts to run the addition at higher temperatures
or stoichiometry resulted in undesired addition of the dithiane
anion to the vinyl epoxide.
For achieving macrolide formation, we envisaged an unprec-
edented ring closing metathesis of a diene and a vinyl epoxide.
Other resorcylic acid type macrolides have been elegantly
reached by Fu¨rstner and co-workers20 utilizing olefin metathesis,
although not with vinyl epoxides, nor with this degree of
functionality. In our case, the reaction of 21 with commercially
available catalyst (PCy)3Cl2RudCHPh,21 resulted in only trace
amounts of the desired product. In a recent communication,21
Grubbs and co-workers reported the successful intermolecular
cross-metathesis of a vinyl epoxide with a new generation, and
highly active, ruthenium-based olefin metathesis catalyst (22).22
Application of this second generation catalyst to 21 (CH2Cl2,
45 °C) gave the desired Z-E 14-membered lactone (23) in a
gratifying 55% yield. Notably, this transformation was ac-
complished in the presence of two sulfur atoms. Previously,
sulfur-containing substrates were implicated as deactivating
ligands in unsuccessful ring-closing metatheses.23 Thus, the
radicicol-like macrolide 23 was assembled in three sequential
steps.
For completion of the synthesis, there remained only removal
of the dithiane, cleavage of the methyl ethers, and regioselective
aromatic chlorination. Initial efforts to liberate the ketone from
its dithiane protection moiety led to gross decomposition. These
attempts at deprotection (Th(NO3)3, PhI(OCOCF3)2)24,25 were
directed to the generation of a version of a “thionium moiety”
en route to the hydrolytic deprotection. Not surprisingly, the
dienyl epoxide was vulnerable when exposed to these Lewis
acids. To circumvent generation of a thionium intermediate, a
two-step Pummerer-like deprotection26 was enlisted. This
protocol served well in this case, as the dithiane was oxidized
to the monosulfoxide (mCPBA, 0 °C) and the crude monosul-
foxide was exposed to Pummerer-like conditions (Ac2O, Et3N,
H2O, 60 °C) to give the desired ketone (24, 70%), also known
as dimethyl monocillin I. Regiospecific chlorination of the
aromatic ring (Ca(OCl)2, 80%)5 produced radicicol dimethyl
ether (25) without complication. The structure of 25 was
confirmed by comparison of its spectral data with those of
commercial radicicol following methylation (MeI, K2CO3,
acetone, 95%) of both free phenols.
a (a) POCl3, DMF, 75 °C, 93%; (b) NaClO2, 85%; (c) (COCl)2, Et3N,
5, 80%; (d) n-BuLi, 6, 60%; (e) 45 °C, 55%; (f) mCPBA; Et3N, Ac2O,
H2O, 60 °C, 70%; (g) Ca(OCl)2, 80%.
In the event, the allylic dithiane (6), was secured in one step
from commercially available 2,4-hexadienal (sorbaldehyde, 16)
(MgClO4, H2SO4, H2S(CH2)3SH2, 64%, Scheme 3).16
The third coupling partner required for the synthetic plan was
a protected benzoic acid 4. It was synthesized in two steps from
commercially available 3,5-dimethoxybenzyl alcohol (17) (Scheme
4). Formylation and concomitant conversion of the alcohol to
the chloride was effected in excellent yield17 (POCl3, DMF,
93%) to give the desired aldehyde (18). Careful oxidation of
this aldehyde (NaClO2, sulfamic acid, 85%) yielded the desired
benzoic acid (19) with no observed cyclization and minimal18
aromatic ring chlorination (<5%). Here, the methyl groups were
retained for phenolic protection because of their stability. It was
hoped that demethylation could be accomplished at a late stage
in the synthesis.
Unfortunately, we were unable to cleave the two methyl ethers
of 24 required to reach fully synthetic monocillin I (2). This
finding was not wholly unexpected in the light of the highly
sensitive functional core. Representative conditions for ac-
complishing deprotections are depicted below (Table 1). Lewis-
acid promoted deprotections result in the opening of the epoxide
at -78 °C, and only later result in deprotection of the ortho-
With the three coupling partners in hand, assembly anticipat-
ing macrolide formation could then be initiated. Esterification
of the benzoic acid (19) proceeded smoothly via its acid chloride
(COCl2, DMF, Et3N, 5, 80%) to provide the benzoic ester (20).
It is important to note that esterifications using either carbodi-
imide or Mitsunobu-based conditions were unsuccessful because
of intervening cyclization to 2,4-dimethoxy phthalide. These
limitations foreshadowed difficulties we would encounter with
differentially protected benzoic acid systems. Addition of the
lithiated dithiane (n-BuLi, -20 °C) to 20 chemoselectively gave
21 (-78 °C, 60% yield), although only with 4:1 R:γ regiose-
(19) Chemoselective addition of the dithiane at the benzylic center
occurred with predominant R (21) vs γ (21a) regioselectivity (4:1, 60%
yield isolated 21). These regioisomers were separable by HPLC. For a report
on vinyl dithiane regioselectivity and similar reactions, see: Murphy, W.
S.; Wattanasin, S. J. Chem. Soc., Perkin Trans. 1 1980, 2678. Colombo,
L.; Gennari, C.; Santandrea, M.; Narisano, E.; Scolastico, C. J. Chem. Soc.,
Perkin Trans. 1 1980, 136.
(20) Fu¨rstner, A.; Seidel, G.; Kindler, N. Tetrahedron 1999, 55, 8215.
(21) Chatterjee, A. K.; Morgan, J. P.; Scholl, M.; Grubbs, R. H. J. Am.
Chem. Soc. 2000, 122, 3783.
(22) Scholl, M.; Ding, S.; Lee, C. W.; Grubbs, R. H. Org. Lett. 1999, 1,
953.
(23) Armstrong, S. K.; Christie, B. A. Tetrahedron Lett. 1996, 37, 9373.
(24) Ho., T.-L.; Wong, C. M. Can. J. Chem. 1972, 50, 3740.
(25) Stork, G.; Zhao, K. Tetrahedron Lett. 1989, 30, 287.
(26) Kishi, Y.; Fukuyama, T.; Natatsuka, S. J. Am. Chem. Soc. 1973,
95, 6490. Fukuyama, T.; Natatsuka, S.; Kishi, Y. Tetrahedron 1981, 37,
2045. Wu, Z.; Williams, L. J.; Danishefsky, S. J. Angew. Chem. 2000, 39,
3866.
(16) Fang, J.-M.; Liao, L.-F.; Hong, B.-C. J. Org. Chem. 1986, 51, 2828.
(17) Makara, G. M.; Klubek, K.; Anderson, W. K. Synth. Commun. 1996,
26, 1935.
(18) The chlorinated benzoic acid did not participate in esterification
and, thus, easily dropped out of the sequence following purification of 20.

10906 J. Am. Chem. Soc., Vol. 123, No. 44, 2001
Garbaccio et al.
Table 1
1 and 2. It was expected that this cycloaddition should be
selective for the R-double bond. Of course, we recognized that
an alternative aromatic system (30) could arise from reaction
with the â-double bond. However, perhaps naively, we dissected
allene 27 into two potentially competitive dienophile moieties;
one, an R,â-unsaturated ester, and the other, a chloro-olefin.
The sense that acrylates are much more reactive dienophiles
than are vinyl chlorides supported our optimism.
Application of this concept to the synthesis of monocillin I
(2) and radicicol (1) required presentation of the vinyl epoxide
fragment containing all three stereocenters at the stage of the
allene prior to formation of the aromatic system. This subgoal
was accomplished as shown in Scheme 6, using the C-5 epimeric
alcohol (32, see Supporting Information for synthetic details),
which was obtained via the same route alternatively employing
the commercially available (S)-3-hydroxybutyric acid. Mit-
sunobu esterification of acid 3129 with alcohol 32 (DEAD, PPh3,
70%) afforded allenic ester (34) after elimination using Hu¨nig’s
base (70%).
methyl ether. Retreating a step in the synthesis, thiolate-
promoted deprotections were attempted on the dithiane-protected
macrolide (in order to avoid Michael-like degradation pathways).
Again, the methyl ethers were resistant to cleavage, and no
reaction occurred at 80 °C. Higher temperatures led to extensive
decomposition.27
The difficulties with methyl ether deprotection clearly called
for a minimally modified follow-up synthesis with more labile
phenolic protecting groups. Indeed, schemes to this end were
initiated. However, an unexpected difficulty frustrated these
initiatives.
In the critical Diels-Alder cycloaddition between 34 and
diene 28, the desired resorcinylic chloride (35) was obtained
as the major product (50%). Surprisingly, considerable amounts
(15%) of the alternative cycloaddition product 35a (not pictured,
see 30) were observed. Thus, the goal of high convergence
leading to 35 had indeed been realized. However, the modest
yield and regioselectivity of the allene Diels-Alder sequence
were disappointing. Efforts are underway to determine the scope
of this transformation, the origin of this unusual regioselectivity
outcome, and the value of the allene Diels-Alder cycloaddition
concept for the construction of highly functionalized aromatic
systems.
A key enabling reaction in our previous route was the
esterification of benzoic acid 19 with alcohol 5 to produce
benzyl ester 20. Unfortunately, all attempts (including Mit-
sunobu, carbodiimide, and acid chloride) to conduct this reaction
on systems such as 4, where R is more complex than methyl,
were unsuccessful. Such initiatives led, instead, to phthalide
formation (26) (Scheme 5).
Scheme 5
Scheme 6a
An interesting possibility for circumventing this nonachiev-
able esterification of the relevant hindered benzoic acids (cf.,
4) presented itself. This approach anticipated a Diels-Alder
reaction of 1,3 asymmetrically disubstituted allene28 27, where
the ester was preintroduced in the dienophile. (Equation 2).
Hence, the product from the cycloaddition reaction of allene
type 27 with a diene (cf., 28) would directly be 29, en route to
(a) DEAD, PPh3, 70%; (b) i-Pr2NEt, 70%; (c) 50% (4:1).
Given the importance of the radicicol program and our
problems with the allene route, the possibility of direct esteri-
fication of a hindered benzoic acid was revisited (cf., Scheme
5). After much experimentation and fine tuning, it was found
that esterification was possible with the ortho-phenol unpro-
tected. The salicylic acid system 38 was constructed starting
(27) Whether or not the presence of the aromatic chloride could positively
impact these efforts remains uninvestigated.
(28) Diels-Alder reactions of allenic esters have been used to establish
the orsellinic motif (first in 1978 by our laboratory). Danishefsky, S. J.;
Singh, R. K.; Gammill, R. B. J. Org. Chem. 1978, 43, 379. Roush, W. R.;
Murphy, J. J. Org. Chem. 1992, 57, 6622. Fink, M.; Gaier, H.; Gerlach, H.
HelV. Chim. Acta 1982, 65, 2563.
(29) See Supporting Information.

Syntheses of Radicicol and Monocillin I
J. Am. Chem. Soc., Vol. 123, No. 44, 2001 10907
Scheme 7a
Continuing with the synthesis, we were now prepared to
introduce the acyl anion equivalent in preparation for ring-
closing metathesis. Initially, addition of lithiated dithiane 6 to
protected variants of benzyl chloride 39 was discouraging. While
the R:γ selectivity for the addition to the dimethyl ether system
20 (entry 2, Table 3) gave a 4:1 (R:γ) selectivity for the synthesis
of radicicol dimethyl ether, when the same conditions were
applied to the bissilylated variant (entry 3), no selectivity (1:1)
was realized.
An attempt was made to define the parameters which direct
the regiochemical outcome of this reaction. Prior work describes
a “hard-soft” component underlying R:γ selectivity.19 More
specifically, it was observed that hard electrophiles preferably
combined with the harder R-dithiane anion. It should be noted
that from the outset a benzyl chloride was chosen because of
its hardness relative to the corresponding benzyl bromide or
iodide. The experiments in this table clearly show a conforma-
tional (steric) influence on this specific transformation as well.
Steric bulk contained in the ester (entry 1 vs entry 2) favors the
desired R-alkylation product. Sterically demanding protection
at the phenol has the opposite effect and favors γ-alkylation
(entry 2 vs entry 3). It was observed that the presence of the
aromatic chloride (required for radicicol) was especially del-
eterious to the R:γ ratio, giving only the γ-alkylated product
(entry 4). A key experiment showed that having the ortho-phenol
free (as a lithium salt) appeared to increase the amount of the
desired R-product (entry 3 vs entry 5). Following further
experimentation, free ortho-phenol 39 emerged as a remarkably
selective substrate for the dithiane alkylation (entry 6). Addition
of 6 to the lithium salt of benzyl chloride (39) proceeded with
good (6:1) R:γ regioselectivity. Alternative acyl anion equiva-
lents were also explored to circumvent the problem of allylic
dithiane regioselectivity. Notably, the TMS ether of the corre-
sponding cyanohydrin31 gave only R-alkylation selectivity when
applied to simple model systems. Unfortunately, because of the
reduced nucleophilicity of this carbanion, no alkylation product
was produced when applied to the more complex and sterically
hindered system 39.
(a) BBr3, 85%; (b) TBDPSCl, 95%; (c) NaOCl2, 95%.
with 18 (Scheme 7), which was available in one step from
commercially available 17 (Scheme 4). Removal of both methyl
ether groups (36, BBr3, 85%, 25 °C, 24 h) was accomplished
without effect on the benzyl chloride. Selective protection of
the para-phenol (TBDPSCl, imidazole, 95%) provided monophe-
nol benzaldehyde 37. Oxidation to benzoic acid 38 (NaOCl2,
95%) occurred without undesired chlorination or cyclization,
thus completing a short, high yielding route to the aromatic core
of monocillin I (2).
We returned to the Mitsunobu reaction for esterification of
salicylic acid system 38. It was hoped that the hydrogen bonding
of the free phenol with the adjacent carboxylic acid would
suppress phthalide formation during the course of the esterifi-
cation. It was pleasing to find that this was the case. Mitsunobu
esterification did indeed occur in THF, albeit in low yield (Table
2). Reducing the polarity of the solvent by switching from THF
to benzene improved the yield significantly (15% to 40%, entries
1-4),30 but large quantities of phthalide 40 were still recovered.
A survey of tertiary phosphines revealed the suprising observa-
tion that P(fur)3 gave a slow, but highly efficient, esterification
reaction (75%, entry 7) with no competing cyclization. That
the expected inversion of the chiral center did indeed occur was
confirmed by synthesis of the identical product via the allene
Diels-Alder route (vide supra). The combination of a less polar
solvent with a less nucleophilic phosphine served to enhance
the quality of the transformation. The generality of this finding
as it pertains to other difficult systems has yet to be explored.
This solution to the construction of key benzoic ester 39 has a
distinct advantage for analogue synthesis, as it introduces the
most complex fragment, namely the vinyl epoxide, at the latest
stage possible.
Table 3
Table 2
With the dithiane alkylation issue resolved (if imperfectly),
ring closing metathesis was attempted. Protection of the ortho-
(30) Benzoic acids are not particularly good coupling partners in the
Mitsunobu reaction, and solvent can have a dramatic effect. See: Dodge,
J. A.; Trujillo, J. I.; Presnell, M. J. Org. Chem. 1994, 59, 234. Harvey, P.
J.; von Itzstein, M.; Jenkins, I. D. Tetrahedron 1997, 53, 3933.
(31) Allylic TMS cyanohydrins have been noted to give high R:γ ratios.
Hertenstein, U.; Hu¨nig, S.; Oller, M. Chem. Ber. 1980, 113, 3783.

10908 J. Am. Chem. Soc., Vol. 123, No. 44, 2001
Garbaccio et al.
Scheme 8a
Finally, a regioselective chlorination was accomplished using
SO2Cl2 in Et2O (58%), thereby converting monocillin I into
radicicol (1).33 This chlorination represents a significant im-
provement over the literature method (NCS, DMF, 15%) for
this same transformation.4
Conclusions
In summary, our totally synthetic route to radicicol comprises
6 steps from the three fragments (6, 32, 38). Its longest linear
sequence is 14 steps. Most importantly, chemical synthesis can
now provide ample amounts of radicicol and analogues needed
to launch a comprehensive Hsp90 directed drug discovery
program.
While, at the moment, the allene cycloaddition route via
dienophile 34 is not competitive with the route via 38, it too
merits continuing attention, given its potential for reaching
complex aromatic structures in a convergent way. Studies to
this effect are well in progress.
a (a) n-BuLi, -78 °C, 50% (6:1); (b) TBSCl, 88%; (c) 42 °C, 60%;
(d) (i) mCPBA, (ii) Ac2O, Et3N, H2O, 60 °C, (iii) NaHCO3, MeOH,
60%; (e) SO2Cl2, 58%.
Acknowledgment. Postdoctoral fellowship support is ac-
knowledged by R.M.G. (NIH, 1 F32 CA85894-01), S.J.S. (NIH,
1 F32 CA81704), and D.K.B. (Swiss National Science Founda-
tion). The research was supported by the National Institutes of
Health (Grant CA28824). Dr. George Sukenick (NMR Core
Facility, Sloan-Kettering Institute, CA 08748) is acknowledged
for mass spectral analyses. We thank Bill Ayer for supplying
spectral data for natural monocillin I.
phenol of 41 (42, TBSCl, 88%, Scheme 8) was necessary to
facilitate ring closing metathesis. Despite the presence of the
dithiane, and the choice of a vinyl epoxide and a conjugated
diene as coupling partners, the new generation Grubbs catalyst32
again closed to the 14-membered ring 43 in 60% yield. When
this same type of transformation was attempted on free ortho-
phenol 41, both the yield and rate of closure decreased. In
addition, the metathesis reaction of 43 proceeded more rapidly
and in higher yield than that for dimethyl ether substrate 21.
Pummerer-like dispatch of the dithiane, vide supra, followed
by hydrolysis of the resulting phenolic acetates, resulted in
global deprotection and afforded monocillin I (2) in 60% yield.
Supporting Information Available: Detailed descriptions
of experimental procedures and spectral data for all compounds
(pdf). This material is available free of charge via the Internet
at http://pubs.acs.org.
JA011364+
(33) Synthetic radicicol was identical to commercial radicicol (Sigma)
as judged by 1H NMR, 13C NMR, HRMS, IR, UV, and optical rotation.
Spectral data for synthetic monocillin I matched that for natural monocillin
(32) Chatterjee, A. K.; Morgan, J. P.; Scholl, M.; Grubbs, R. H. J. Am.
Chem. Soc. 2000, 122, 3783. Scholl, M.; Ding, S.; Lee, C. W.; Grubbs, R.
H. Org. Lett. 1999, 1, 953.
1
I as judged by H NMR and IR.