Design, synthesis, and biological evaluation of HSP90 inhibitors based on conformational analysis of radicicol and its analogues

Article abstract of DOI:10.1021/ja043101w

The molecular chaperone HSP90 is an attractive target for chemotherapy because its activity is required for the functional maturation of a number of oncogenes. Among the know inhibitors, radicicol, a 14-member macrolide, stands out as the most potent. A molecular dynamics/minimization of radicicol showed that there were three low energy conformers of the macrocycle. The lowest of these is the bioactive conformation observed in the cocrystal structure of radicicol with HSP90. Corresponding conformational analyses of several known analogues gave a good correlation between the bioactivity and the energy of the bioactive conformer, relative to other conformers. Based on this observation, a number of proposed analogues were analyzed for their propensity to adopt the bioactive conformation prior to synthesis. This led to the identification of pochonin D, a recently isolated secondary metabolite of Pochonia chlamydosporia, as a potential inhibitor of HSP90. Pochonin D was synthesized using polymer-bound reagents and shown to be nearly as potent an HSP90 inhibitor as radicicol.

Full text of DOI:10.1021/ja043101w

Published on Web 04/21/2005
Design, Synthesis, and Biological Evaluation of HSP90
Inhibitors Based on Conformational Analysis of Radicicol and
Its Analogues
Emilie Moulin,† Vincent Zoete,‡ Sofia Barluenga,† Martin Karplus,*,‡,§ and
Nicolas Winssinger*,†
Contribution from the Organic and Bioorganic Chemistry Laboratory, Institut de Science et
Inge´nierie Supramoleculaires, UniVersite´ Louis Pasteur, 8 alle´e Gaspard Monge, 67000
Strasbourg, France, Biophysical Chemistry Laboratory, Institut de Science et Inge´nierie
Supramoleculaires, UniVersite´ Louis Pasteur, 8 alle´e Gaspard Monge, 67000 Strasbourg,
France, and Department of Chemistry and Chemical Biology, HarVard UniVersity,
Cambridge, Massachusetts 02138
Received November 16, 2004; E-mail: marci@tammy.harvard.edu; winssinger@isis.u-strasbg.fr
Abstract: The molecular chaperone HSP90 is an attractive target for chemotherapy because its activity is
required for the functional maturation of a number of oncogenes. Among the know inhibitors, radicicol, a
14-member macrolide, stands out as the most potent. A molecular dynamics/minimization of radicicol showed
that there were three low energy conformers of the macrocycle. The lowest of these is the bioactive
conformation observed in the cocrystal structure of radicicol with HSP90. Corresponding conformational
analyses of several known analogues gave a good correlation between the bioactivity and the energy of
the bioactive conformer, relative to other conformers. Based on this observation, a number of proposed
analogues were analyzed for their propensity to adopt the bioactive conformation prior to synthesis. This
led to the identification of pochonin D, a recently isolated secondary metabolite of Pochonia chlamydosporia,
as a potential inhibitor of HSP90. Pochonin D was synthesized using polymer-bound reagents and shown
to be nearly as potent an HSP90 inhibitor as radicicol.
Introduction
The heat shock protein 90 (HSP90) is an ATP-dependent
chaperone whose activity is required for the functional matura-
tion of a number of overexpressed oncogenic proteins that
promote the growth or survival of cancer cells.1-3 The list of
HSP90’s clients include notorious oncogenes such as Bcr-Abl,
Raf1, ErbB2, AKT, and mutated p53. In the absence of this
chaperoning activity, HSP90’s clients are targeted for degrada-
tion by the proteosome. The prospect of stalling a number of
oncogenes by inhibiting a single protein has made HSP90 an
attractive target for chemotherapy.4,5 Two natural products,
radicicol (1, Figure 1) and geldanamycin (2, Figure 1), are
known to be potent inhibitors of HSP90’s chaperone activity.
Cocrystal structures of both of these compounds with HSP90
showed that, while neither compound has structural or topologi-
cal similarity to ATP, they both bind in the ATP-binding pocket
† Organic and Bioorganic Chemistry Laboratory, Institut de Science et
Inge´nierie Supramoleculaires, Universite´ Louis Pasteur.
‡ Biophysical Chemistry Laboratory, Institut de Science et Inge´nierie
Supramoleculaires, Universite´ Louis Pasteur.
Figure 1. Structures of HSP90 inhibitors.
of HSP90; their measured Kd values are 19 nM and 1.6 µM,
respectively.6 Designed inhibitors based on an adenosine scaf-
fold led to the discovery of new leads such as 4 (Figure 1).7-9
More recently, novobiocin was also shown to inhibit HSP90
albeit at millimolar concentration and through a different binding
§ Department of Chemistry and Chemical Biology, Harvard University.
(1) Isaacs, J. S.; Xu, W.; Neckers, L. Cancer Cell 2003, 3, 213-217.
(2) Mosser, D. D.; Morimoto, R. I. Oncogene 2004, 23, 2907-2918.
(3) Prodromou, C.; Pearl, L. H. Curr. Cancer Drug Targets 2003, 3, 301-
323.
(4) Dymock, B. W.; Drysdale, M. J.; McDonald, E.; Workman, P. Expert Opin.
Ther. Pat. 2004, 14, 837-847.
(5) Banerji, U.; Judson, I.; Workman, P. Curr. Cancer Drug Targets 2003, 3,
385-390.
(6) Roe, S. M.; Prodromou, C.; O’Brien, R.; Ladbury, J. E.; Piper, P. W.; Pearl,
L. H. J. Med. Chem. 1999, 42, 260-266.
9
10.1021/ja043101w CCC: $30.25 © 2005 American Chemical Society
J. AM. CHEM. SOC. 2005, 127, 6999-7004
6999

A R T I C L E S
Moulin et al.
Figure 2. Wire-frame representation of radicicol’s three main conformers
and their relative energies (kcal).
site.10-12 Attesting to the potential of HSP90 inhibition for
chemotherapy, a derivative of geldanamycin [17AAG (3), Figure
1] is currently under clinical evaluation.13 While radicicol is
the most potent inhibitor of HSP90, it lacks activity in vivo
presumably due to metabolic instability. It has been shown that
thiols such as DTT can inactivate the action of radicicol.14 This
result has been attributed to a 1,6-Michael addition to the
conjugate diene yielding an inactive product.15 Importantly,
converting the ketone to an oxime affords products that retain
acitivity in vivo by reducing the electrophilicity of the Michael
acceptor.15,16 These compounds were found to be effective
against oncogenic mouse xenograft. Additionally, Danishefsky
and co-workers have reported a series of analogues where the
epoxide has been replaced by a more resistant cyclopropane
and showed that these compounds are active in vitro.17,18 These
results have raised the possibility that radicicol can be modified
to yield a therapeutic agent. However, in the design of modified
inhibitors, one concern has been that many apparently similar
compounds have different affinities. Herein, we report a
molecular dynamics study that addresses this question by
determining the conformational landscape of putative inhibitors
(including radicicol) and use the results to design simplified
analogues based on conformational similarity to radicicol, which
were then synthesized and tested experimentally.
Figure 3. Structures of selected radicicol analogues.
several reported analogues (Figure 3) were analyzed. Each
molecule was simulated by molecular dynamics with the Merck
Molecular Force Field (MMFF94)19-23 in the CHARMM24
program. A dielectric constant of 80 was used to simulate the
effect of solvent in a simple way. The simulations were carried
out at 1000 K during 1 ns and 500 frames were extracted from
the trajectory at 20 ps intervals. The high temperature was used
to ensure that conformational energy barriers were crossed. Each
frame was minimized by 750 steps of the steepest descent (SD)
algorithm in CHARM, and the MMFF energy was calculated.
The resulting 500 conformations were clustered to determine
the main conformations. Starting from the lowest energy
conformation as representative of the first cluster, all conforma-
tions having a root-mean-square deviation (RMSD) lower than
1 Å were grouped into that cluster. The lowest energy conformer
of the remaining conformations was taken as the starting point
for the second cluster, and the process was repeated until all
compounds had been clustered. The RMSD between the lowest
energy representative of each cluster and the bioactive confor-
mation of radicicol in the cocrystal structure6 was then calculated
for all heavy atoms. This analysis led to the identification of
three main conformations: an L-shape conformation, which is
the bioactive conformation of radicicol, an essentially planar
(P-shape) conformation, and an L′-shape conformation that
mainly differs from the L-shape one by the fact that the
macrocycle is positioned on the opposite side of the aromatic
cycle (Figure 2). The calculated L-shape, P-shape, and L′-shape
conformations of the isolated molecules have an average RMSD
from the bioactive radicicol conformation of about 0.6, 1.8, and
2.1 Å, respectively (the bioactive radicicol conformation was
taken from the PDB structure ID 1BGQ).6 Importantly, the
Results and Discussion
Molecular Dynamics Analysis of Radicicol and Potential
Analogues. To evaluate the conformation-activity relationship
of radicicol, its conformational profile (Figure 2) and that of
(7) Vilenchik, M.; Solit, D.; Basso, A.; Huezo, H.; Lucas, B.; He, H.; Rosen,
N.; Spampinato, C.; Modrich, P.; Chiosis, G. Chem. Biol. 2004, 11, 787-
797.
(8) Chiosis, G.; Lucas, B.; Huezo, H.; Solit, D.; Basso, A.; Rosen, N. Curr.
Cancer Drug Targets 2003, 3, 371-376.
(9) Wright, L.; Barril, X.; Dymock, B.; Sheridan, L.; Surgenor, A.; Beswick,
M.; Drysdale, M.; Collier, A.; Massey, A.; Davies, N.; Fink, A.; Fromont,
C.; Aherne, W.; Boxall, K.; Sharp, S.; Workman, P.; Hubbard, R. E. Chem.
Biol. 2004, 11, 775-785.
(10) Marcu, M. G.; Chadli, A.; Bouhouche, I.; Catelli, M.; Neckers, L. M. J.
Biol. Chem. 2000, 275, 37181-37186.
(11) Langer, T.; Schlatter, H.; Fasold, H. Cell Biol. Int. 2002, 26, 653-657.
(12) Yun, B.-G.; Huang, W.; Leach, N.; Hartson, S. D.; Matts, R. L. Biochemistry
2004, 43, 8217-8229.
(13) http://www.clinicaltrials.gov/ct/gui/show/NCT00004241?order)13.
(14) Kwon, H. J.; Yoshida, M.; Fukui, Y.; Horinouchi, S.; Beppu, T. Cancer
Res. 1992, 52, 6926-6930.
(15) Agatsuma, T.; Ogawa, H.; Akasaka, K.; Asai, A.; Yamashita, Y.; Mizukami,
T.; Akinaga, S.; Saitoh, Y. Bioorg. Med. Chem. 2002, 10, 3445-3454.
(16) Ikuina, Y.; Amishiro, N.; Miyata, M.; Narumi, H.; Ogawa, H.; Akiyama,
T.; Shiotsu, Y.; Akinaga, S.; Murakata, C. J. Med. Chem. 2003, 46, 2534-
2541.
(19) Halgren, T. A. J. Comput. Chem. 1996, 17, 616-641.
(20) Halgren, T. A.; Nachbar, R. B. J. Comput. Chem. 1996, 17, 587-615.
(21) Halgren, T. A. J. Comput. Chem. 1996, 17, 553-586.
(22) Halgren, T. A. J. Comput. Chem. 1996, 17, 520-552.
(23) Halgren, T. A. J. Comput. Chem. 1996, 17, 490-519.
(24) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.; Swaminathan,
S.; Karplus, M. J. Comput. Chem. 1983, 4, 187-217.
(17) Yamamoto, K.; Garbaccio, R. M.; Stachel, S. J.; Solit, D. B.; Chiosis, G.;
Rosen, N.; Danishefsky, S. J. Angew. Chem., Int. Ed. 2003, 42, 1280-
1284.
(18) Yang, Z.-Q.; Geng, X.; Solit, D.; Pratilas, C. A.; Rosen, N.; Danishefsky,
S. J. J. Am. Chem. Soc. 2004, 126, 7881-7889.
9
7000 J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005

Design, Synthesis, and Evaluation of HSP90 Inhibitors
A R T I C L E S
Table 1. Relative Energies of Radicicol and Its Derivatives Main
Conformers (Energies Are in kcal/mol Relative to the Lowest
Conformer Energy)
compound
L-shape
L
′
-shap e
P-shape
RMSDa
simil.b
radicicol (1)
0.0
2.7
2.7
0.0
0.0
0.0
0.0
0.3
3.3
0.0
0.0
0.7
3.0
1.7
2.1
0.0
2.4
1.7
1.6
2.1
2.3
1.6
4.2
3.6
0.60
0.60
0.61
0.56
0.62
0.56
0.54
0.62
0.024
0.029
0.025
0.023
0.027
0.023
0.025
0.026
(14S,15S)radicicol (6)
diol-radicicol (7)
monocillin (8)
(17S)-radicicol (9)
cyclopropane (10)
oxime (11)
E-isomer of oxime 11
a RMSD between the L-shape conformation and the bio-active conforma-
tion of radicicol, calculated for common heavy atoms. b Similarity index
defined as the RMSD (see footnote a) divided by the number of atoms
included in the RMSD calculation; the lower is the similarity index, the
greater is the similarity.
Figure 4. Representation of radicicol (wire-frame structure, see Figure 1
for numbering) bound to HSP90 (represented as a surface, blue indicates
positive regions, red indicates negative regions, and + indicates water
molecules) [image generated with Weblab from PDB data ID 1BGQ].
bioactive L-shape is the lowest energy conformer found in the
molecular dynamics analysis. Its energy is 3.3 and 2.4 kcal/
mol lower than that of the L′-shape and P-shape conformers,
respectively (Table 1). This result suggested that the wrong
conformational bias could be involved in the lack of activity of
analogues in which the stereochemistry of specific centers was
altered. Indeed, a corresponding conformational analysis of
compound 617 in which the epoxide stereochemistry was
inverted (Figure 3) showed that, while the macrocycle could
be clustered in the same three shapes, the lowest energy
conformation was no longer the bioactive conformation. Open-
ing of the epoxide to the corresponding diol (7,25,26 Figure 3)
likewise favored the inactive L′ conformation. Removal of the
chlorine atom (monocillin, 8)27,28 lowers the energetic bias for
the bioactive conformation. This result is not surprising
considering that this chlorine can contribute to a large allylic
1,3-strain. Furthermore, it was clear from the crystal structure
that the chlorine atom could fill a hydrophobic pocket. Interest-
ingly, the stereochemistry at the C17 position (9,17 Figure 3)
did not have a significant impact on the conformation. The lack
of activity of this compound is likely to be due to its inability
to fit in the active site rather than to a change in conformational
bias. Compound 1017 wherein the epoxide is replaced by a
cyclopropane has been shown to have comparable activity to
radicicol and indeed has a similar conformational profile.
Likewise, compound 1116 showed a profile similar to radicicol.
The enhanced activity of compound 11 as compared to radicicol
can be rationalized by the fact that the oxime substituent fills a
hydrophobic pocket (see Figure 4). It is interesting to note that
the E-isomer of oxime 11 has a less favorable conformational
bias for the bioactive L-shape than the Z-isomer. This observa-
tion is consistent with experimental data from related oximes
wherein the Z-isomer had a cytotoxicity of 98 versus 282 nM
for the E-isomer.18 The above analysis shows that the epoxide
function influences the relative stability of the main conforma-
tions of radicicol derivatives. In particular, the configuration
of the asymmetric carbons of the epoxide function found in
radicicol favors the bioactive conformation.
Based on the presumed importance of maintaining the
bioactive (L-shape) conformation, we profiled the conformation
space of a series of compounds with modifications of the
epoxide and olefin region (Figure 5). Replacement of the
epoxide for an amide bond (12 and 13) or a bridging cyclo-
pentane (15, Figure 5) led to conformations that disfavored the
L-shape. Removal of the epoxide (17) led to a compound that
strongly disfavors the L-shape. However, replacement of the
epoxide by an olefin (18 and 19) only moderately disfavored
the L-shape conformation. Interestingly, compounds 14 and 16
lacking the R,â-conjugated olefin had an unfavorable bias for
the L-shape. Compound 19 caught our attention as it represents
a significant simplification in terms of chemical synthesis while
only moderately disfavoring the active L-shape conformation
(1.2 kcal). Recently, compound 19 was isolated from the
fermentation broth of Pochonia chlamydosporia and named
pochonin D.29 In the isolation report, the authors did find
pochonin D to be cytotoxic at micromolar concentration, which
could be rationalized by its HSP90 inhibition. Pochonin D was
docked in HSP90 and was found to have a good overlap with
radicicol (Figure 6). Given these computational and experimental
results, a practical and modular synthesis of pochonin D was
sought.
Polymer-Assisted Synthesis of Pochonin D and Analogues
Thereof. The synthetic disconnections of pochonin D (19,
Figure 7) were based on the chemistry that was developed in
the context of our total synthesis of pochonin C.30 We were
particularly interested in developing chemistry amenable to using
polymer-supported reagents31-35 for the purpose of diversity
oriented synthesis. Commercially available Weinreb amide 23
(Scheme 1) was selectively S-alkylated with 3-hydroxythiophe-
nol and 1 equiv of base, and then loaded onto Merrifield resin
in the same pot by adding a second equivalent of K2CO3, the
(29) Hellwig, V.; Mayer-Bartschmid, A.; Mueller, H.; Greif, G.; Kleymann, G.;
Zitzmann, W.; Tichy, H.-V.; Stadler, M. J. Nat. Prod. 2003, 66, 829-837.
(30) Barluenga, S.; Lopez, P.; Moulin, E.; Winssinger, N. Angew. Chem., Int.
Ed. 2004, 43, 3467-3470.
(31) Vickerstaffe, E.; Warrington, B. H.; Ladlow, M.; Ley, S. V. J. Comb. Chem.
2004, 6, 332-339.
(25) Shibata, T.; Oikawa, T.; Kobayashi, T.; Shimazaki, N. Jpn. Kokai Tokkyo
Koho; Sankyo Co., Ltd.: Japan, 1997; 5 pp.
(26) Ikeda, A.; Shinonaga, H.; Fujimoto, N.; Kasai, Y. PCT Int. Appl.; Taisho
Pharmaceutical Co., Ltd.: Japan, 2003; 126 pp.
(27) Ayer, W. A.; Pena-Rodriguez, L. Phytochemistry 1987, 26, 1353-1355.
(28) Danishefsky, S. J.; Garbaccio, R. M.; Baeschlin, D. K.; Stachel, S. J.; Solit,
D.; Shtil, A.; Rosen, N. PCT Int. Appl.; Sloan-Kettering Institute for Cancer
Research: USA, 2002; 135 pp.
(32) Storer, R. I.; Takemoto, T.; Jackson, P. S.; Brown, D. S.; Baxendale, I. R.;
Ley, S. V. Chem.-Eur. J. 2004, 10, 2529-2547.
(33) Bapna, A.; Vickerstaffe, E.; Warrington, B. H.; Ladlow, M.; Fan, T.-P.
D.; Ley, S. V. Org. Biomol. Chem. 2004, 2, 611-620.
(34) Lee, A.-L.; Ley, S. V. Org. Biomol. Chem. 2003, 1, 3957-3966.
(35) Ley, S. V.; Baxendale, I. R.; Bream, R. N.; Jackson, P. S.; Leach, A. G.;
Longbottom, D. A.; Nesi, M.; Scott, J. S.; Storer, R. I.; Taylor, S. J. J.
Chem. Soc., Perkin Trans. 1 2000, 3815-4195.
9
J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005 7001

A R T I C L E S
Moulin et al.
Figure 5. Structure and conformational analysis of designed radicicol analogues.
Scheme 1. Synthesis of Weinreb Amide Fragment 22a
a (a) 3-Mercaptophenol (1.0 equiv), K2CO3 (1.0 equiv), DMF, 23 °C;
after 8 h, K2CO3 (1.7 equiv), Merrifield resin, TBAI (cat), 12 h, 50 °C,
98%; (b) H2O2 (2.0 equiv), (CF3)2CHOH /CH2Cl2 (1/1), 12 h; (c) KOtBu
(1.0 equiv), 1-iodo-5-pentene (1.0 equiv), DMSO, 23 °C, 3 h; (d) toluene,
80 °C, 8 h, 77%. DMF ) N,N-dimethylformamide, DMSO ) dimethyl
sulfoxide, TBAI ) tetrabutylammonium iodide.
Figure 6. Representation of pochonin D (wire-frame structure) docked in
HSP90 superimposed over the structure of radicicol (yellow wire-frame
structure); HSP90 is represented as a surface (see Figure 4 for legend) [image
generated with Weblab from PDB data].
1,4-dioxane gave poor results. Furthermore, the polar nature of
DMSO is unfavorable for the sulfoxide elimination, and the
reaction could be heated to 60 °C without any elimination which
allowed an excess of the alkylating agent to be used. Resus-
pension of the resin in toluene and heating to 80 °C released
the compound by elimination, affording desired fragment 22
with 77% yield and 95% purity (judged by NMR).
Selective Mitsunobu esterification of benzoic acid 20 using
polymer-bound DEAD afforded ester 27 (Scheme 2). The use
of (mClPh)3P was found to be essential to suppress the
competing ether formation with the para-phenol.30,38 Protection
of both phenols with EOM-Cl39 afforded ester 29, which could
be used in the subsequent alkylation without further purification.
All attempts to chlorinate ester 27, 29, or later intermediates
were unsuccessful due to the reactivity of the olefin toward
electrophilic chlorine. The chlorine atom was thus introduced
prior to esterification of acid 20 using HClO generated in situ
by the oxidation of acetaldehyde with NaClO2/sulfamic acid.
Esterification of this product under the same conditions as for
20 afforded compound 28, which was protected with EOM-
Cl to obtain ester 30. As for ester 29, this product could be
used in subsequent reactions without purification. Deprotonation
Figure 7. Retrosynthetic disconnections of pochonin D (19).
resin, and raising the temperature to 50 °C. While preparations
of thiol resins have been reported by direct lithiation of the resin
followed by a quench with elemental sulfur,36 this method was
found more practical as it affords the polymer-bound Weinreb
amide in one step. Oxidation of the thioether 24 to sulfoxide
25 was carried out using H2O2 in HFIP/CH2Cl2.37 This oxidation
procedure was found to be practical and reliable with no
overoxidation to the sulfone in a number of studied cases.
Carrying out the reaction on solid phase also allows for an easy
recycling of the fluorinated solvent. The ketosulfoxide 25 was
then deprotonated with KOtBu, and the resulting enolate was
quenched with 5-iodopentene. The use of DMSO was found to
be critical for the success of this reaction; DMF, tBuOH, and
(36) Farrall, M. J.; Frechet, J. M. J. J. Org. Chem. 1976, 41, 3877-3882.
(37) Ravikumar, K. S.; Begue, J.-P.; Bonnet-Delpon, D. Tetrahedron Lett. 1998,
39, 3141-3144.
(38) Hughes, D. L.; Reamer, R. A.; Bergan, J. J.; Grabowski, E. J. J. J. Am.
Chem. Soc. 1988, 110, 6487-6491.
(39) EOM-Cl: EOM-Cl was used instead of MOM-Cl for availability reasons.
9
7002 J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005

Design, Synthesis, and Evaluation of HSP90 Inhibitors
A R T I C L E S
Scheme 2. Synthesis of Macrocycles 33 and 34a
Table 2. Affinity of Radicicol Analogues for HSP90
IC50 (µM)
radicicol (1)
pochonin D (19)
0.02
0.08
33
34
35
>50
>50
>50
short path of silica could be used to remove all of the catalyst
and its byproducts to afford compound 34 in 60% yield. While
the metathesis reaction carried out on purified triene 32 was
nearly quantitative, it was deemed more practical to carry out
the whole synthetic sequence from compound 20 without
purifications, thus affording the protected pochonin D 34 in 34%
yield for five steps. The EOM groups were removed from both
macrocycles 34 and 33 using sulfonic acid resin in MeOH to
obtain pochonin D (19)43 and its analogue 36 lacking the
chlorine in 90% and 92% yield, respectively. It is interesting
to note that the conjugated olefin proved to have different
reactivity depending on the presence or absence of the chlorine
atom on the aryl ring. Deprotection of compound 33 with HCl
in dioxane led to the conjugate addition of the chlorine ion,
whereas compound 34 proved more resilient to conjugate
addition and could be deprotected with HCl to obtain 19. The
difference in reactivity of the conjugate olefins may be attributed
to the different conformations of the two compounds, which is
due to the 1,3-allylic strain conferred by the bulky chlorine atom.
It is conceivable that the differences in conformation affect the
level of conjugation between the olefin and the adjacent carbonyl
group and thus its susceptibility to Michael addition.
Biological Evaluation of Pochonin D. The compounds were
evaluated for HSP90 affinity in a competition assay with
geldanamycin using a previously described method.45 The results
are shown in Table 2. Pochonin D (19) was found to be a good
ligand for HSP90 with an IC50 of 80 nM as compared to 20
nM for radicicol. This difference is consistent with the calculated
1.2 kcal change in the internal energy of the pochonin D upon
binding. The cocrystal structure of HSP90-radicicol reported
by Pearl et al. showed a tightly bound water molecule making
a hydrogen-bond bridge between the ortho-phenol, the ester,
and Asp79 and a second water molecule making a bridge
between the para-phenol and Leu34.6 Accordingly, compound
34 with alkylated phenols showed no affinity for HSP90. The
importance of the chlorine atom is evident from the comparison
of compounds 19 (80 nM) and 35 (>50 µM). The bulky chlorine
not only comes within van der Waals contact of Phe124 and
fills a hydrophobic void, but it also biases the population of
the bioactive conformer due to allylic strain (see relative energies
of compound 8 vs 1, Table 1).
a (a) NaClO2 (5.0 equiv), NH2SO3H (5.0 equiv), CH3CHO (1.0 equiv),
THF/H2O (5/1), 0 °C, 0.5 h, 92%; (b) PS-DEAD (2.5 equiv, 1.3 mmol
g-1), s-(-)-penten-2-ol (1.0 equiv), P(mClPh)3 (2.0 equiv), CH2Cl2, 23 °C,
i
0.5 h, 68% for 27 and 65% for 28; (c) Pr2EtN (4.0 eqiv), EOMCl (4.0
equiv), TBAI (cat), DMF, 80 °C, 5 h, 95%; (d) LDA (2.0 equiv), THF,
-78 °C, 22 (1.0 equiv), 10 min, Amberlite IRC-50 (20.0 equiv, 10.0 mmol
g-1); (e) Grubbs II (10% mol), toluene (2 mM), 80 °C, 12 h, 40% after
two steps; (f) PS-TsOH (10 equiv, 3.2 mmol g-1), MeOH, 40 °C, 4 h, 90%
for 19; MeOH, 40 °C, 0.5 h, 92% for 35. PS-DEAD ) ethoxycarbonyl-
azocarboxymethyl polystyrene, PS-TsOH ) sulfonic acid resin MP, DMF
) N,N-dimethylformamide, Grubbs II ) ruthenium[1,3-bis-(2,4,6-trimethyl-
phenyl)-2-imidazolidinylidene]dichloro(phenylmethylene)(tricyclohexyl-
phosphine), LDA ) lithium diisopropylamide, EOM ) ethoxymethylene,
TBAI ) tetrabutylammonium iodide, THF ) tetrahydrofurane.
of the toluic esters 29 and 30 followed by addition of Weinreb
amide 22 afforded the desired metathesis precursors 31 and 32
along with 20% undesired product stemming from a 1,4
conjugate addition. Treatment of crude triene 31 and 32 with
the second generation Grubbs catalyst40,41 at 120 °C for 10 min
afforded the desired cyclization product 33 and 34 in excellent
yield albeit as an unseparable 1:4 mixture of cis:trans. Allowing
the reaction to equilibrate under thermodynamic control at 80
°C overnight shifted in both cases the equilibrium to the trans
with >95% selectivity.42 It should be noted that this reaction
could be performed at millimolar concentration without any
detectable amount of dimerization or oligomerization. Impor-
tantly, the 10-membered ring cyclization product was not
observed in either case. Moreover, a simple filtration through a
Conclusion and Future Prospects
The conformational analysis of radicicol and its known
analogues showed that there are three low-energy conformations
and gave a good correlation between their biological activity
and the energetic cost of adopting the bioactive L-shape
conformation. The importance of the decrease in affinity due
(43) Synthetic pochonin D was found to have NMR spectra identical to natural
pochonin D (spectra were kindly provide by Marc Stadler, Bayer Health
Care, Wuppertal, Germany).
(44) Kometani, T.; Kondo, H.; Fujimori, Y. Synthesis 1988, 1005-1007.
(45) Zhou, V.; Han, S.; Brinker, A.; Klock, H.; Caldwell, J.; Gu, X.-j. Anal.
Biochem. 2004, 331, 349-357.
(40) Scholl, M.; Ding, S.; Lee, C. W.; Grubbs, R. H. Org. Lett. 1999, 1, 953-
956.
(41) Chatterjee, A. K.; Morgan, J. P.; Scholl, M.; Grubbs, R. H. J. Am. Chem.
Soc. 2000, 122, 3783-3784.
(42) Lee, C. W.; Grubbs, R. H. Org. Lett. 2000, 2, 2145-2147.
9
J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005 7003

A R T I C L E S
Moulin et al.
to binding a conformer different from the global minimum in
solution has been pointed out previously, although the bioactive
conformation of a flexible ligand often does not correspond to
the global energy minimum of the free ligand; that is, the binding
energy compensates for the reorganization energy and loss of
entropy.46-48 In the present study, the relative conformational
energies were evaluated with a simplified solvent representation.
Thus, the results are only of qualitative significance. Neverthe-
less, the conformational analysis provides a rationale for the
observed changes in activity due to modifications of the epoxide
moiety. Analysis of a series of simplified analogues led to the
identification of pochonin D as having a low energy and high
similarity index to the bioactive conformation. A polymer-
assisted synthesis was developed, which affords pochonin D
from readily available fragments 20, 21, and 22 in six steps
and requires only one chromatographic purification of the final
product. This also represents the first total synthesis of pochonin
D (19), thereby assigning a (+) polarization to (17-R)-pochonin
D. The affinity of pochonin D for HSP90 (80 nM) is close to
that of radicicol (20 nM), the difference being consistent with
the energetic cost of adopting the bioactive conformation. As
the cyclopropane analogues previously reported,18,49 the pocho-
nin D scaffold lacks the sensitive epoxide. In a larger context,
this study clearly illustrates the dramatic impact of small
structural changes on the conformational equilibrium of these
resorcyclic macrolides and may rationalize the diverse biological
activity of structurally similar compounds from this family.
Acknowledgment. A Marie Curie Fellowship to S.B. (Grant
no.: HPMF-CT-2002-01968) and BDI from CNRS and region
d’Alsace to E.M. are gratefully acknowledged. We thank Merck
Bioscience/Novabiochem for samples of resins, as well as
Xiang-ju Gu and Francisco Adrian for their assistance with the
biological screening. This work was partly supported by a grant
from the European Commission (MIRG-CT-2004-505317) and
is part of the COST netword D28. The work done at Harvard
was supported by the National Institutes of Health.
Supporting Information Available: Experimental procedures
utilized and compound characterization including a NMR of
pochonin D. This material is available free of charge via the
Internet at http://pubs.acs.org.
(46) Vieth, M.; Hirst, J. D.; Brooks, C. L., III. J. Comput.-Aided Mol. Des.
1998, 12, 563-572.
(47) Bostrom, J.; Norrby, P.-O.; Liljefors, T. J. Comput.-Aided Mol. Des. 1998,
12, 383-396.
JA043101W
(48) Perola, E.; Charifson, P. S. J. Med. Chem. 2004, 47, 2499-2510.
(49) Geng, X.; Yang, Z.-Q.; Danishefsky, S. J. Synlett 2004, 1325-1333.
9
7004 J. AM. CHEM. SOC. VOL. 127, NO. 19, 2005