Total synthesis of (+)-geldanamycin and (-)-o-quinogeldanamycin: Asymmetric glycolate aldol reactions and biological evaluation

Article abstract of DOI:10.1021/jo034870l

The total synthesis of (+)-geldanamycin (GA), following a linear route, has been completed using a demethylative quinone-forming reaction as the last step. Key steps include the use of two new asymmetric boron glycolate aldol reactions. To set the anti-C11,12 hydroxymethoxy functionality, (S,S)-5,6-bis-4-methoxyphenyldioxanone 8 was used. Methylglycolate derived from norephedrine 5 set the C6,7 methoxyurethane stereochemistry. The quinone formation step using nitric acid gave the non-natural o-quino-GA product 55 10: 1 over geldanamycin. Other known oxidants gave an unusual azaquinone product 49. o-Quino-GA 55 binds Hsp90 with good affinity but is less cytotoxic compared to GA.

Full text of DOI:10.1021/jo034870l

Tota l Syn th esis of (+)-Geld a n a m ycin a n d
(-)-o-Qu in ogeld a n a m ycin : Asym m etr ic Glycola te Ald ol Rea ction s
a n d Biologica l Eva lu a tion
Merritt B. Andrus,* Erik L. Meredith, Erik J . Hicken, Bryon L. Simmons,
Russell R. Glancey, and Wei Ma†
Brigham Young University, Department of Chemistry and Biochemistry, C100 BNSN,
Provo, Utah 84602-5700, and Kosan Biosciences, Inc., 3832 Bay Center Place, Hayward, California 94545
mbandrus@chem.byu.edu
Received J une 21, 2003
The total synthesis of (+)-geldanamycin (GA), following a linear route, has been completed using
a demethylative quinone-forming reaction as the last step. Key steps include the use of two new
asymmetric boron glycolate aldol reactions. To set the anti-C11,12 hydroxymethoxy functionality,
(S,S)-5,6-bis-4-methoxyphenyldioxanone 8 was used. Methylglycolate derived from norephedrine
5 set the C6,7 methoxyurethane stereochemistry. The quinone formation step using nitric acid
gave the non-natural o-quino-GA product 55 10:1 over geldanamycin. Other known oxidants gave
an unusual azaquinone product 49. o-Quino-GA 55 binds Hsp90 with good affinity but is less
cytotoxic compared to GA.
While macbecin I and herbimycin A have received
considerable synthetic attention, including four total
syntheses,1 the closely related ansamycin antitumor
antibiotic geldanamycin (GA) has only recently suc-
cumbed to total synthesis.2 It was isolated (Streptomyces
hygroscopicus var. geldanus) in 1970 by workers at
Upjohn, and the structure was determined by Rinehart
and co-workers shortly thereafter.3 These compounds
possess various biological activities, including great
potential as therapeutic lead compounds for new anti-
cancer agents. Various semisynthetic analogues have
been made and tested.4 Among these, the phase-I clinical
candidate 17-allylamino-GA is the most prominent. Ironi-
cally, geldanamycin, with the least synthetic attention,
was discovered first and is the most potent member of
this class. It shows broad activity with a unique profile
of action within the NCI 60 cell-line panel with an
average ED50 value of 180 nM.5
Lack of initial interest in this target may be attributed
to two factors: One, only recently has its cellular target
been identified, and two, it presents distinct and signifi-
cant synthetic challenges compared to the other ansa-
mycins. Neckers, through the use of a GA-affinity pro-
tocol, demonstrated that geldanamycin binds to the
chaperone heat shock protein 90 (Hsp90).6 Previously, GA
was known to greatly lower cellular levels of various
oncogenic tyrosine kinases, including v-Src, Bcr/Abl, and
ErbB-2,7 proteins that rely on Hsp90 for proper folding
and stability. GA does not effect cellular levels of the
serine/threonine kinases PKA or PKC. In addition, X-ray
crystal structures have recently been reported for the
Hsp90-GA complex.8 GA binds to the ATP binding site
in a C-shaped conformation distinct from its free, solution
conformation.
In light of the recently disclosed biological properties
of GA together with the challenge posed by its structure,
the development of a total synthesis was clearly war-
ranted. In this paper we outline the details of this route,
which establishes the absolute stereochemistry, high-
lights two new glycolate aldol reactions, uncovers the
challenge of p-quinone formation, and provides access to
three new biologically active analogues (Scheme 1).
Unlike the related ansamycins, geldanamycin pos-
sesses a methoxyl at C17 on the quinone that requires
the use of a pentasubstituted benzene precursor in order
† Kosan Biosciences, Inc.
(1) Herbimycin A: (a) Nakata, M.; Osumi, T.; Ueno, A.; Kimura, T.
Tamai, T.; Tatsuda, K. Tetrahedron Lett. 1991, 32, 6015. Macbecin I:
(b) Baker, R.; Castro, J . L. J . Chem. Soc., Perkin Trans. 1990, 47. (c)
Evans, D. A.; Miller, S. J .; Ennis, M. D. J . Org. Chem. 1993, 58, 471.
(d) Panek, J . S.; Xu, F.; Rondon, A. C. J . Am. Chem. Soc. 1998, 120,
4413.
(2) Andrus, M. B.; Meredith, E. L.; Simmons, B. L.; Soma Sekhar,
B. B. V.; Hicken, E. J . Org. Lett. 2002, 4, 3549-3552.
(3) (a) De Boer, C.; Meulman, P. A.; Wnuk, R. J .; Peterson, D. H. J .
Antibiot. 1970, 23, 442. (b) Rinehart, K. L.; Sasaki, K.; Slomp, G.;
Grostic, M. F.; Olson, E. C. J . Am. Chem. Soc. 1970, 92, 7591.
(4) Geldanamycin derivatives: (a) Rinehart, K. L.; McMillan, M. W.;
Witty, T. R.; Tipton, C. D.; Shield, L. S. Li, L. H.; Reusser, F. Bioorg.
Chem. 1977, 6, 353. (b) Schnur, R. C.; Corman, M. L. J . Org. Chem.
1994, 59, 2581. Schnur, R. C.; Corman, M. L.; Gallaschun, R. J .; Cooper,
B. A.; Dee, M. F.; Doty, J . L.; Muzzi, M. L.; Moyer, J . D.; DiOrio, C. I.;
Barbacci, E. G.; Miller, P. E.; O’Brien, A. T.; Morin, M. J .; Foster, B.
A.; Pollack, V. A.; Savage, D. M.; Sloan, D. E.; Pustilnik, L. R.; Moyer,
M. P. J . Med. Chem. 1995, 38, 3806. (c) 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, 1303.
(5) Whitesell, L.; Shifrin, S. D.; Schwab, G.; Neckers, L. M. Cancer.
Res. 1992, 52, 1721.
(6) Whitesell, L.; Mimnaugh, E. G.; De Costa, B.; Myers, C. E.;
Neckers, L. M. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 8324.
(7) Lawrence, D. S.; Niu, J . Pharmacol. Ther. 1998, 77, 81.
(8) (a) Stebbins, C. E.; Russo, A. A.; Schneider, C.; Rosen, N.; Hartl,
F. U.; Pavletich, N. P. Cell, 1997, 89, 239. (b) Roe, S. M.; Prodromou,
R. O.; Ladbury, J . E.; Piper, P. W.; Pearl, L. H. J . Med. Chem. 1999,
42, 260.
10.1021/jo034870l CCC: $25.00 © 2003 American Chemical Society
Published on Web 09/26/2003
8162
J . Org. Chem. 2003, 68, 8162-8169

Total Synthesis of (+)-Geldanamycin and (-)-o-Quinogeldanamycin
SCHEME 1 SCHEME 3
SCHEME 2
cin and herbimycin syntheses. In addition, C11 possesses
a hydroxyl, not a methoxyl. Because of this, the route to
GA needs to address the anti-C11,12 hydroxy-methoxy
functionality in a way that can differentiate between
these two groups. The seco-amino acid 2 reveals the dense
arrangement of functionality. A suitable position for
retrosynthetic cleavage at a midpoint for planning a
convergent approach is problematic. The six stereocenters
and the trisubstituted olefin crowd this region. The C7
TES (triethylsilyl) and C11 TBS (tert-butyldimethyl) silyl
ethers were chosen to allow for the selective introduction
of the urethane group. The aniline moiety would be
masked throughout as an aryl nitro group. In this linear
approach, aldehyde 3 would be used to form the E,Z-diene
carboxylic acid. Enal 4 would undergo an asymmetric
syn-glycolate aldol reaction with the newly reported
norephedrine ester 5 to set the stereochemistry at C6 and
C7.10 While new approaches to the C8,9 olefin and C10-
methyl region were attempted, the previously employed
asymmetric hydroboration route was followed.1a,d Inter-
mediate 6 was made using an asymmetric anti-glycolate
aldol reaction with the newly reported bisaryldioxanone
8 reacting with aldehyde 7.11
Following the lead of Gilman who reported that 1,2,4-
trimethoxybenezene 9 could be selectively metalated at
the 3 position,12 treatment with n-butyllithium, followed
by addition of DMF, lead to the formation of the substi-
tuted benzaldehyde (Scheme 3). Nitric acid in warm
acetic acid then generated nitrobenzene 10. Reduction
and treatment with phosphorus tribromide gave the
stable, crystalline benzyl bromide 11. The C14 methyl
stereocenter was set using an asymmetric Evans alky-
lation13 upon deprotonation of (S)-12 with NHMDS
followed by addition of 11 (1.1 equiv) to give the product
13 in high yield and excellent selectivity. The primary
alcohol was formed with lithium borohydride and Mit-
to ensure proper formation of the desired p-quinone
(Scheme 2). In accord with previous macbecin and
herbimycin studies where 1,4-dimethoxy compounds were
used with success1 and with results from other tri-
methoxybenzene substrates,9 it was anticipated that
oxidative removal of the 1,4-disposed methoxyls, as with
intermediate 1, would generate the needed p-quinone
with high selectivity. Unforeseen conformational and
electronic issues would, only at the very end of route,
reveal the problem of this assumption. The lack of a
hydroxyl at C15, together with the presence of the C17-
methoxyl, prevents the use of an aldol reaction to form
the C14-15 bond, as previously employed in the macbe-
(10) Andrus, M. B.; Soma Sekhar, B. B. V.; Turner, T. M.; Meredith,
E. L. Tetrahedron Lett. 2001, 42, 7197.
(9) (a) Cameron, D. W.; Feutrill, G. I.; Patti, A. F.; Perlmutter, P.;
Sefton, M. A. Aust. J . Chem. 1982, 35, 1501. (b) Witiak, D. T.; Loper,
J . T.; Anathan, S.; Almericao, A. M.; Verhoef, V. L.; Filppi, J . A. J .
Med. Chem. 1989, 32, 1636. (c) Kozuka, T.; Bull. Chem. Soc. J pn. 1982,
55, 2415. (d) Cheng, A. C.; Castagnoli, N. J . Med. Chem. 1989, 32,
1636. (e) Michael, J . P.; Cirillo, P. F.; Denner, L.; Hosken, G.; Howard,
A. S.; Tinkler, O. S. Tetrahedron 1990, 46, 7923. (f) Luly, J . R.;
Rapoport, H. J . Org. Chem. 1981, 46, 2745. (g) Kitahara, Y.; Nakahara,
S.; Shimizu, M.; Yonezawa, T.; Kubo, A. Heterocycles 1993, 36, 1909.
(11) (a) Andrus, M. B.; Soma Sekhar, B. B. V.; Meredith, E. L.;
Dalley, N. K. Org. Lett. 2000, 2, 3035. (b) Andrus, M. B.; Meredith, E.
L.; Soma Sekhar, B. B. V. Org. Lett. 2001, 3, 259. (c) Andrus, M. B.;
Mendenhall, K. G.; Meredith, E. L. Soma Sekhar, B. B. V. Tetrahedron
Lett. 2002, 43, 1789-1792.
(12) Gilman, H.; Thirtle, J . R. J . Am. Chem. Soc. 1944, 66, 858.
(13) Evans, D. A.; Ennis, M. D.; Mathre, D. J . J . Am. Chem. Soc.
1982, 104, 1737.
J . Org. Chem, Vol. 68, No. 21, 2003 8163

Andrus et al.
SCHEME 4
SCHEME 5
sunobu conditions using acetone cyanohydrin14 were
employed to directly give cyanide 14. Two equivalents of
DIBAL used at -78 °C followed by warming and addition
of pH 7 buffer gave the key aldehyde 7 in high yield.
This route was not the initial choice to produce 7. An
asymmetric conjugate addition approach was also pur-
sued by homologating aldehyde 10 with a methoxymethyl
Wittig reagent to give the arylacetaldehyde 15 (Scheme
4).15 Various attempts were made to then access the
unsaturated acyloxazolidinone 17 using phosphonium
and phosphonate derivatives with 15. Finally, it was
found that use of phosphonate 16 reacted with NHMDS
in the presence of lithium chloride gave 17 with good
reactivity and high selectivity.16 Use of DBU in place of
NHMDS gave only a 39% yield with 9:1 selectivity.17
Other base and solvent combinations were also less
effective. While the selectivity of the methylcuprate was
found to be high in this case, typically >19:1, the yields
rarely approached 50%. All variations attempted gave
uniformly low yields.18 It was apparent that under these
conditions, nucleophilic demethylation of the benzene
functionality was occurring leading to decomposition in
this case. The success of the alkylation route ended the
investigation of this route.
ing tin acetal was transformed with tert-butyl bromoac-
etate in the presence of tetra-n-butylammonium iodide
to give dioxanone 8. Following the reported method, 8
was treated with dicyclohexylboron triflate with added
triethylamine to give the boron enolate.11 Aldehyde 7 (1.2
equiv) was then reacted to give the (S,S)-anti-aldol
adduct 21 in 70% yield with 10:1 selectivity. The minor
product was assumed to be the (R,R)-anti product. The
enolate of 8 is locked as the E-isomer and attack of the
aldehyde is constrained to a closed Zimmerman-Traxler
chair arrangement on the si-enolate face away from the
C-5 methoxyphenyl. The alcohol was converted to the
methyl ether and treated with catalytic sodium methox-
ide to give ester 22. The benzyl ether was then cleaved
using ceric ammonium nitrate and the TBS ether was
generated to give the key intermediate 6.
At this point, various routes to install the C8-10
functionality were considered. Ultimately, an asymmetric
hydroboration approach proved successful as pioneered
by Tatsuda with herbimycin A and used recently in
Panek’s approach to macbecin I.1a,d Other approaches
explored at this key juncture will also be described
briefly. To prepare for the hydroboration route, ester 6
was converted to the aldehyde and reacted with trim-
ethylaluminum to give a mixture of alcohols 23 (Scheme
5). Oxidation with Dess-Martin periodinane and treat-
ment with HF gave allyl alcohol 24.21 Excess borane‚THF
at -10 °C was used followed by peroxide to give a 5:1
mixture of syn- and anti-25 diol products (Scheme 6). All
With aldehyde 7 in hand, a scaleable route to the anti-
aldol step was pursued (Scheme 5). On a large scale, (E)-
4,4′-dimethoxystilbene 19 was dihydroxylated under
Sharpless conditions with the components of AD-mix-R
added individually including methylsulfonamide to give
(S,S)-diol 20.19 Batches as large as 10 g were transformed
with reproducible results reacted at 0 °C. Diol 20 was
treated with di-n-butyltin oxide in benzene at reflux
using a Dean-Stark trap to remove water.20 The result-
(14) (a) Wilk, B. Synth. Commun. 1993, 23, 2481. (b) Tsunoda, T.;
Uemoto, K.; Nagino, C.; Kawamura, M.; Kaku, H.; Ito, S. Tetrahedron
Lett. 1999, 40, 7355.
(15) (a) Levine, S. G. J . Am. Chem. Soc. 1958, 80, 6150. (b) Williams,
D. R.; Barner, B. A.; Nishitani, K.; Phillips, J . G. J . Am. Chem. Soc.
1982, 104, 4708.
(16) (a) Broka, C. A.; Ehrler, J . Tetrahedron Lett. 1991, 32, 5907.
(b) Evans, D. A.; Black, W. C. J . Am. Chem. Soc. 1993, 115, 4497.
(17) Blanchette, M. A.; Choy, W.; Davis, J . T.; Essenfeld, A. P.;
Masamune, S.; Roush, W. R.; Sakai, T. Tetrahedron Lett. 1984, 25,
2183.
(18) (a) Hruby, V. J .; J arosinski, M. A.; Li, G. Tetrahedron Lett. 1993,
34, 2561. (b) Williams, D. R.; Kissel, W. S.; Li, J . J . Tetrahedron Lett.
1998, 39, 8593. (c) Wipf, P.; Takahashi, H. J . Chem. Soc., Chem.
Commun. 1996, 2675. (d) Romo, D. R.; Rzasa, R. M.; Shea, H. A. J .
Am. Chem. Soc. 1998, 120, 591. (e) Williams, D. R.; Kissel, W. S. J .
Am. Chem. Soc. 1998, 120, 11198.
(19) Sharpless, K. B.; Amberg, W.; Bennami, Y. L.; Crispino, G. A.;
Hartung, J .; J eong, K.; Kwong, H.; Morikawa, K.; Wang, Z. Xu, D.;
Zhang, X. J . Org. Chem. 1992, 57, 2768.
(20) (a) Burke, S. D.; Sametz, G. M. Org. Lett. 1999, 1, 71. (b) David,
S.; Thieffry, A.; Veyrieres, A. J . Chem. Soc., Perkin Trans. 1 1981, 1796.
(21) Dess, D. B.; Martin, J . C. J . Am. Chem. Soc. 1983, 105, 2487.
8164 J . Org. Chem., Vol. 68, No. 21, 2003

Total Synthesis of (+)-Geldanamycin and (-)-o-Quinogeldanamycin
SCHEME 6
SCHEME 7
SCHEME 8
attempts to improve the selectivity failed. Fortunately,
the yield was high and the minor, undesired anti isomer,
due to a favorable Rf difference, was easily removed by
chromatography. Evans and Burgess have shown that
syn-diols from alkenes of this type can be accessed using
rhodium-catalyzed hydroboration.22 Panek’s macbecin I
approach found that the anti isomer was major under
these conditions.1d While the origin of the asymmetric
induction in this case remains unclear, Houk has pro-
posed a transition-state model for the uncatalyzed reac-
tion.23 Conformations with the hydroxyl adopting an
inside position over the outside position are favored
theoretically, but only slightly by ∼0.1 kcal/mol. In view
of these ambiguities, a structure proof on both isomers
of 25 was performed. The acetonides of the separated
isomers were formed and indicated methine proton
couplings were observed. The syn isomer J ab was 2.5 Hz
consistent with an axial-equatorial disposition and the
anti isomer acetonide gave a diaxial 12.0 Hz value.
Confident in the assignment, the major syn-25 isomer
was protected, with TBS triflate, followed by camphor
sulfonic acid to remove the primary silyl ether and
oxidized to access aldehyde 26. Stabilized Wittig treat-
ment gave the unsaturated ester in quantitative yield
with high E selectivity. DIBAL reduction followed by
Swern oxidation then gave enal 4.
to give the methoxyethoxymethyl (MEM) ether 28.
Oxidation, followed by coupling with ketophosphonate
29,17 previously described,24 gave enone 30 in good yield
with complete E selectivity. Unfortunately, all attempts
to add Mg or Li methylorganometallics gave rapid
decomposition, or no reaction, as with titanium reagents.
The desired alcohol 31, allowing for conversion to the
C8,9 olefin via allylic displacement, was not obtained in
this case.
An alternative to the hydroboration route was also
explored using an asymmetric Sharpless epoxidation
(Scheme 8).25 Adduct 21 was methylated and reduced to
the lactol with DIBAL. Treatment with Wittig reagent
in the presence of LiCl in acetonitrile gave a high yield
of unsaturated ester 32. CAN removal, protection, and
reduction generated the allylic alcohol 33. The standard
conditions were then employed to produce the epoxide
A more convergent approach to GA using an enone
intermediate was explored (Scheme 7). The aldol adduct
21 was converted to the 1,2-diol product 27 and protected
(22) (a) Evans, D. A.; Fu, G. C.; Hoyveda, A. H. J . Am. Chem. Soc.
1992, 114, 6671. (b) Evans, D. A.; Fu, G. C.; Anderson, B. A. J . Am.
Chem. Soc. 1992, 114, 6679. (c) Burgess, K.; Cassidy, J .; Ohlmeyer,
M. J . J . Org. Chem. 1991, 56, 1020. (d) Burgess, K.; van der Donk, W.
A.; J arstfer, M. B.; Ohlmeyer, M. J . J . Am. Chem. Soc. 1991, 113, 6139.
(23) Houk, K. N.; Rondan, N. G.; Wu, Y.-D.; Metz, J . T. Paddon-
Row, M. N. Tetrahedron 1984, 40, 2257.
(24) Andrus, M. B.; Meredith, E. L.; Soma Sekhar, B. B. V. Org.
Lett. 2001, 3, 259.
(25) Sharpless, K. B.; Hanson, R. M.; Klunder, J . M.; Yo, S. Y.;
Masamune, H.; Gao, Y. J . Am. Chem. Soc. 1987, 109, 5765.
J . Org. Chem, Vol. 68, No. 21, 2003 8165

Andrus et al.
SCHEME 9
SCHEME 10
with high selectivity. Trimethylaluminum is well-known
for its ability to open epoxy alcohols at the â-position to
give 1,2-diol products.26 However, all attempts using 34
at the standard temperature of 0 °C gave no reaction,
and at higher temperatures only decomposition was seen.
The γ-TBS ether was changed to smaller protecting
groups, i.e., methyl, without success. No reaction and
decomposition were again observed. The sensitivity of the
substituted benzene substrate was a likely contributor
to these challenges. After these detours, the linear
hydroboration route was followed to completion.
From enal 4 to the lactam stage, the route to geldana-
mycin proceeded without complication. The newly devel-
oped (-)-norephedrine-derived glycolate 5 was reacted
with boron triflate and triethylamine followed by addition
of 4 to give syn adduct 36 in high yield and >20:1
selectivity (Scheme 9).10 Two equivalents of the boron
enolate of 5 were used to ensure complete conversion of
4. The model substrate for this reaction was 2-methyl-
cinnamaldehyde 37. In this case, 5 reacted under similar
conditions with this new auxiliary to give a 9:1 mixture
of syn products. syn-Gylcolate aldol reactions are well-
known using the Evans oxazolidinone auxiliary ap-
proach.27 However, in this case, with substrate 37 reacted
with oxazolidinone 39, a mixture of all four possible 40
isomers, 3:2:2:1, was obtained. From this comparison, the
utility of the norephedrine glycolate is demonstrated for
R-branched enal substrates. Adduct 36 was converted to
the methyl ester 41 using hydroxide followed by trim-
ethylsilyldiazomethane.28 The norephedrine alcohol was
easily recovered by separation from the methyl ester at
this point using chromatography. Attempts to obtain the
alcohol after the carboxylic acid forming step using acid/
base extraction proved futile. Protection as the TES ether
at C7, followed by a half-reduction with DIBAL at -78
°C gave an aldehyde, which was reacted with the Still-
Gennari hexafluorophosphonate in the presence of KH-
MDS and 18-c-6 to generate the (Z)-ester 42.29
Reduction of (Z)-ester 42 initially proved problematic.
When THF was used as solvent, even a great excess of
DIBAL (10 equiv) gave only a <50% yield with the
remainder being recovered 42. Surprisingly, use of ether
with DIBAL with only 2.5 equiv, gave a 91% yield of (Z)-
allyl alcohol (Scheme 10). Use of CH2Cl2 as solvent for
this step resulted in complete reduction and TES ether
cleavage. Oxidation gave the enal, which was then
reacted with allyl ester phosphonate 43 together with
DBU and lithium chloride following the procedure of
Roush.30 The desired E,Z diene ester 44 was obtained in
high yield and selectivity. Previously, the methyl ester
of 44 was employed. Hydrolysis to the carboxylic acid
from this intermediate was very slow and gave multiple
products in that case. The arylnitro group was converted
(26) (a) Suzuki, T.; Saimoto, H.; Tomioka, H.; Oshima, K.; Nozaki,
H. Tetrahedron Lett. 1982, 23, 3597. (b) Roush, W. R.; Adam, M. A.;
Peseckis, S. M. Tetrahedron Lett. 1983, 24, 1377.
(27) (a) Evans, D. A.; Bender, S. L.; Morris, J . J . Am. Chem. Soc.
1988, 110, 2506-2513. (b) Ku, T. W.; Kondrad, K. H.; Gleason, J . G.
J . Org. Chem. 1989, 54, 3487-3489. (c) Andrus, M. B.; Schreiber, S.
L. J . Am. Chem. Soc. 1993, 115, 10420-10421. (d) Evans, D. A.;
Barrow, J . C.; Leighton, J . L.; Robichaud, A. J .; Sefkow, M. J . Am.
Chem. Soc. 1994, 116, 12111-12112. (e) Crimmins, M. T.; Choy, A. L.
J . Org. Chem. 1997, 62, 7548-7549. (f) Crimmins, M. T.; Choy, A. L.
J . Am. Chem. Soc. 1999, 121, 5653-5660. (g) Haight, D.; Birrell, H.
C.; Cantello, B. C. C.; Eggleston, D. S.; Haltiwanger, R. C.; Hindley,
R. M.; Ramaswany, A.; Stevens, N. C. Tetrahedron: Asymmetry 1999,
10, 1353-1367.
(28) Hirai, Y.; Aida, T.; Inoue, S. J . Am. Chem. Soc. 1987, 111, 3062.
(29) Still, W. C.; Gennari, C. Tetrahedron Lett. 1983, 24, 4405.
(30) Roush, W. R.; Sciotti, R. J . J . Am. Chem. Soc. 1998, 120, 7411.
8166 J . Org. Chem., Vol. 68, No. 21, 2003

Total Synthesis of (+)-Geldanamycin and (-)-o-Quinogeldanamycin
SCHEME 11
to the aniline at this point using the reagent described
by Lalancette.31 NaBH2S3 was formed with sodium boro-
hydride reacted with 3 equiv of elemental sulfur as
previously employed by Panek for macbecin I.1d The
aniline from 44 was obtained in 82% yield. Removal of
the allyl ester with palladium tetrakistriphenylphosphine
and morpholine produced the key amino acid 2. Dime-
done, used in place of morpholine for this step, caused
problems at the purification step. Use of BOP-Cl (bis(2-
oxo-3-oxazolidinyl)phosphinic chloride) with diisopropy-
lethylamine in warm toluene (0.001 M) generated the
macrolactam 45 in 76% isolated yield. At this point,
various ansa ring conformations, including rotational
isomers about the amide bond, complicated the NMR
analysis. Changes in solvent and temperature gave only
minimal improvement in the line broadening. Attention
now focused on urethane introduction at C7. Tetra-n-
butylammonium fluoride (TBAF, 2.1 equiv) in THF at 0
°C efficiently removed the TES ether. Previous ansamy-
cin routes have used excess sodium cyanate with TFA to
install the urethane with low yields (∼50%) being typi-
cal.1 We elected to employ trichloroacetyl isocyanate,
following the procedure of Kocovsky.32 The protected
carbamate was initially formed and methanolysis with
potassium carbonate gave the stable C7 urethane in 89%
isolated yield. The standard conditions of TBAF or HF‚
pyridine failed to remove the C11 TBS ether even when
used in excess. Finally, it was found that use of a 10:1
acetonitrile-aqueous HF (48%) mixture gave alcohol 1
in excellent yield. Flexible introduction of the urethane
and removal of protecting groups boded well for explora-
tion of the quinone formation.
SCHEME 12
To access lactam 45 in a more efficient, convergent
manner, a ring-closing metathesis (RCM) strategy was
explored.33 While medium and large rings have found
success with this approach, the combination of a ring of
this size together with a trisubstituted olefin target
possessing allylic substituents has not been reported. In
view of the potential improvement in efficiency and the
ready availability of the intermediates, the RCM route
was attempted. Aldehyde 26 was converted to the ter-
minal olefin-aniline 46 (Scheme 11). The carboxylic acid
47 was made using a series of analogous steps from
isobutenal, including a syn glycolate aldol reaction and
the allylester phosphonate reagent. Coupling under the
BOP-Cl conditions gave the arylamide 48. The Grubbs
imidazolium benzylidene ruthenium catalyst, 10 mol %,
was explored under a variety of conditions, including
various solvents, concentrations, modes of addition, and
additives, to no avail. Even added titanium tetraisopro-
poxide, to disfavor allylic oxygen coordination, in warm
toluene34 failed to give the desired lactam 45.35 The allylic
C7 alcohol, following TES removal, was also explored as
substrate and the diol from TBS removal from 47. A
substrate involving a diphenylsiloxane tether between
the C7 and C11 hydroxyls and intermolecular metathesis
with 46 and 47 also failed.
Formation of the p-quinone from the trimethoxy pre-
cursor 1 is the final issue to be addressed. On the basis
of the other ansamycin routes1 and the various reports
of successful oxidative demethylation reactions of others
using more simple trimethoxybenzenes,9 lactam 1 was
anticipated to give GA directly (Scheme 12). Instead,
using either silver or manganese oxide impregnated with
nitric acid, the unusual aza-quinone variant 49 was
rapidly obtained in 77 and 40% yields, respectively. Other
known oxidants, including CAN in acetonitrile even at
-10 °C, led only to decomposition in this case. The
structure of this unique product 49 was confirmed by
conversion to the corresponding dihydroazaquinone, lac-
tam-phenol 50, by treatment with sodium hydrosulfite.36
The NOE enhancements of 1.54 and 4.1% were observed
(31) (a) Lalancette, J . M.; Freche, A.; Monteaux, R. Can. J . Chem.
1968, 46, 2754. (b) Lalancette, J . M.; Freche, a.; Brindle, J . R.;
Laliberte, M. Synthesis 1972, 526.
(32) Kocovsky, P. Tetrahedron Lett. 1986, 27, 5521.
(33) (a) Trnka, T. M.; Grubbs, R. H. Acc. Chem. Res. 2001, 34, 18-
29. (b) Furstner, A. Angew. Chem., Int. Ed. 2000, 39, 3012-3043.
(34) (a) Furstner, A.; Thiel, O. R.; Blanda, G. Org. Lett. 2000, 2,
3731-3734. (b) Xu, Z.; J ohannes, C. W.; Houri, A. F.; La, D. S.; Cogan,
D. A.; Hofilena, G. E.; Hoveyda, A. H. J . Am. Chem. Soc. 1997, 119,
10302-10316.
(35) A test substrate, o-allylphenyl acrylate, was employed under
these conditions with success to verify the activity of the catalyst.
J . Org. Chem, Vol. 68, No. 21, 2003 8167

Andrus et al.
SCHEME 13
SCHEME 14
between the proton signals indicated. The oxidation was
also explored with the C11 hydroxyl protected as the TBS
ether. In this case, CAN and AgO/HNO3 again produced
the azaquinone as the sole product in only 10 min in good
yield, 70-80%. No trace of the desired quinone was
detected by TLC, NMR, or MS in crude material or in
purified fractions.
To anticipate possible complications, a model tri-
methoxybenzene amide 51, made in a few steps from
aldehyde 15, was explored with various oxidants (Scheme
13). In this case, CAN37 and AgO gave a quantitative
yield of a quinone 52. No trace of aza-quinone product
was detected. Other oxidants including CoF2,38 PhI(OAc)2,
PhI(OTf)2,39 DDQ, and CrO3 gave little or no product.
Initially, through comparison of the NMR data to related
compounds, we were confident that the quinone obtained
in this case was the p-quinone. Only after considerable
effort was a crystal suitable for X-ray analysis obtained
for 52.40 Surprisingly, this unambiguous result confirmed
that the oxidation product in this case was in fact the
o-quinone product and not the p-quinone. With the
nitrogen lone-pair delocalized through carbonyl reso-
nance, oxygen lone-pairs are available to stabilize inter-
mediate radical-cations and control the location of attack
by water leading to quinone formation. For the desired
p-quinone to form, water must attack at the C4 carbon
between the alkyl group and the amide nitrogen. Evi-
dently in this case, water attacks at C2 generating a
carbonyl at that position leading exclusively to o-quinone
formation.
SCHEME 15
Lack of p-quinone formation may be rationalized in the
case of the model system 51, yet the question of aza-
quinone formation with lactam 1 for the synthesis of
geldanamycin remains (Scheme 14). A stereoelectronic
argument is difficult in this case due to the multiple low
energy conformations that both GA and the synthetic
intermediates adopt. A clue may be found in the Hsp90-
bound conformation of GA as shown.8 The amide nitrogen
is found in a twisted s-cis conformation where the lone-
pair on nitrogen is not fully conjugated to the carbonyl
oxygen due to the constraints of the large ring. Instead
of the usual planar dihedral angle of 0°, in this case the
out of plane twisting shows an angle of 12°. This
conformation suggests that the amide nitrogen may
electronically approximate the donation ability of an
aniline type nitrogen with the potential for cation sta-
bilization upon ring oxidation. The less electronegative
nitrogen, compared to oxygen, may then be more able to
donate its lone pair to the ring leading to aza-quinone
formation selectively over either the p- or o-quinone
products. In the case of the acyclic model 51, the amide
is flat and the N-lone pair is not available.
(36) The conditions found in ref 4b were used.
(37) Florke, U.; Frese, P.; Krohn, K. Chem. Eur. J . 2000, 6, 3887.
(38) Nakata, M.; Hashimoto, K.; Takemura, S.; Tomatsu, A. Synlett
1999, 1474.
(39) Kita, Y.; Hashizume, M.; Harayama, Y.; Morioka, H.; Tohma,
H. Tetrahedron Lett. 2001, 42, 6899.
(40) X-ray data: orthorhombic space group P212121, a ) 7.359 Å,
b ) 8.645 Å, c ) 21.12 Å, independent data R1 ) 0.038. Full details of
the synthesis of 51 will be reported elsewhere.
To overcome this electronic conformational bias, a
strategy was devised to further delocalize the nitrogen
lone-pair and favor formation of the desired quinone
(Scheme 15). Following urethane formation with 45,
treatment with excess tert-butyloxycarbonyl anhydride
(BOC2O) in the presence of DMAP gave the N-BOC-
8168 J . Org. Chem., Vol. 68, No. 21, 2003

Total Synthesis of (+)-Geldanamycin and (-)-o-Quinogeldanamycin
TABLE 1. Cytotoxicity a n d Hsp 90 Bin d in g
SCHEME 16
compd
SKBr3 IC50 (nM)
GA
o-GA-55
40
1272
tri-MeO-GA-1
aza-GA-49
>5000
838
to oxidation, rendering the nitrogen lone pair less avail-
able for cation stabilization. As with the model substrate
51, lactam 1 also prefers o-quinone formation. The route
provided ample material for biologically evaluation of
late-stage analogues.
Shown in Table 1 are cytotoxicity results for GA,
o-quino-GA 55, trimethylbenzene 1, and the azaquinone
49. The cell assays were performed with SKBr3 human
cancer cells known to rely on Hsp90 dependent kinase
signaling.6 o-Quino-GA 55 showed comparable binding
to GA, but was far less potent in the cell assay. The
azaquinone 49 showed reasonable, but less potent, activ-
ity in the cell assay. The trimethylbenzene 1 was the least
effective.
The total synthesis of geldanamycin has been ac-
complished for the first time. The linear route involved
41 steps from 1,2,4-trimethoxybenzene. Key reactions
include an asymmetric dioxanone aldol reaction to set
the anti stereochemistry at C11,12 and a norephedrine-
based glycolate aldol step for the syn functionality at
C6,7. Azaquinone was formed under the standard oxida-
tion conditions, instead of the desired p-quinone product
apparently due to ring-size induced amide nitrogen lone-
pair donation. Nitric acid was used to form o-GA and
geldanamycin without formation of the azaquinone prod-
uct. New analogues of geldanamycin were tested for
biological activity. These findings will lead to improved
routes to geldanamycin including selective approaches
to the p-quinone, development of a more convergent,
nonlinear route, and analogues with improved Hsp90
binding.
protected lactam 53 in good yield.41 The urethane was
also BOC protected under these conditions. Treatment
with CAN indeed gave a quinone type product 54 in low
yield from 53. No aza-quinone was observed. The quinone
could not be definitively assigned as either para or ortho
on the basis of the available data at this point. All
conditions attempted in this case for BOC removal, dilute
HCl,42 HF, or the very mild Mg(ClO4)2,43 gave only
decomposition not geldanamycin.
Following a reference by Musgrave, use of nitric acid
alone was attempted for the demethylative oxidation of
1 (Scheme 16).44 A glacial acetic acid solution of 1 was
treated for only 1 min with 70% nitric acid followed by
quenching with sodium bicarbonate. Longer reaction
times in this case led to extensive decomposition. The
more polar, red-orange o-geldanamycin 55 (UV λmax 266,
303, 431 nM) was the major product 10:1 over the less
polar, yellow-orange desired p-quinone geldanamycin,
which was shown to be identical in all respects (NMR,
TLC, UV λmax 260, 311, 422 nM) to an authentic sample.
Multiple 10 mg runs were performed with reproducible
results. The success of nitric acid over the other condi-
tions may be attributed to protonation of the amide prior
Ack n ow led gm en t. We are grateful to the NIH
(GM57275), the American Cancer Society, and BYU for
funding; Bruce J ackson for MS assistance; Prof. Chaitan
Khosla for a sample of authentic geldanamycin; David
C. Myles and Zong-Qiang Tian, Kosan Biosciences, and
Leonard M. Neckers, NCI, NIH, for biological testing.
Su p p or tin g In for m a tion Ava ila ble: Experimental pro-
cedures and characterization for selected compounds and
spectral data. This material is available free of charge via the
Internet at http://pubs.acs.org.
(41) Ragnarrson, U.; Grehn, L. Acc. Chem. Res. 1998, 31, 494.
(42) Boger, D. L.; McKie, J . A.; Nishi, T.; Ogiku, T. J . Am. Chem.
Soc. 1996, 118, 2301.
(43) Stafford, J . A.; Bracken, D. S.; Karanewski, D. S.; Valvano, N.
L. Tetrahedron Lett. 1993, 34, 7873.
(44) Musgrave, O. C. Chem. Rev. 1969, 69, 9.
J O034870L
J . Org. Chem, Vol. 68, No. 21, 2003 8169