Total synthesis of abyssomicin C, atrop-abyssomicin C, and abyssomicin D: Implications for natural origins of atrop-abyssomicin C

Article abstract of DOI:10.1021/ja067083p

In this article, the total syntheses of the antibiotic abyssomicin C (1) and its biologically inactive sibling abyssomicin D (3) are described. A number of unforseen roadblocks in our synthetic plan encouraged innovation, which culminated in the discovery

Full text of DOI:10.1021/ja067083p

Published on Web 12/22/2006
Total Synthesis of Abyssomicin C, Atrop-abyssomicin C, and
Abyssomicin D: Implications for Natural Origins of
Atrop-abyssomicin C
K. C. Nicolaou*,†,‡ and Scott T. Harrison†
Contribution from the Department of Chemistry and The Skaggs Institute for Chemical Biology,
The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, California 92037,
and the Department of Chemistry and Biochemistry, UniVersity of California, San Diego,
9500 Gilman DriVe, La Jolla, California 92093
Received October 6, 2006; E-mail: kcn@scripps.edu
Abstract: In this article, the total syntheses of the antibiotic abyssomicin C (1) and its biologically inactive
sibling abyssomicin D (3) are described. A number of unforseen roadblocks in our synthetic plan encouraged
innovation, which culminated in the discovery of a new Lewis acid-templated Diels-Alder reaction. En
route to abyssomicin C, we prepared and characterized its stable conformational isomer atrop-abyssomicin
C (57), which in the presence of a strong acid underwent an unusual interconversion with the targeted
natural product. Close inspection of the X-ray crystallographic structures of these compounds led to
hypotheses on the mechanism of their interconversion. Attempted reduction of both atropisomers revealed
that atrop-abyssomicin C afforded abyssomicin D much more readily, suggesting that this previously
unknown atropisomer may be synthesized by the host organism and serves as a direct precursor of
abyssomicin D. Finally, to gain insight into the mechanism of antiobiotic activity, several synthetic
intermediates and designed analogues were evaluated for biological activity.
Introduction
In 2004, Su¨ssmuth and Fielder reported the isolation and
characterization of abyssomicin C (1, Figure 1) as part of their
search for natural products possessing antibiotic activity via
inhibition of the p-aminobenzoic acid (pABA) biosynthetic
pathway.1 Of the 201 microbes screened, only the marine
actinomycete strain Verrocosispora AB 18-032, collected from
the depths of the Sea of Japan, met both requirements. Feeding
experiments indicated that abyssomicin C (1) is an inhibitor of
either aminodeoxychorismate synthase or aminodeoxychoris-
mate lyase, enzymes responsible for the conversion of choris-
mate into pABA, making 1 the first known natural inhibitor of
these enzymes. Along with the active component 1, which
inhibited the growth of methicilin-resistant Staphylococcus
aureus (MRSA, MIC ) 4 µg mL-1) and vancomycin-resistant
S. aureus (VRSA, MIC ) 13 µg mL-1), two structurally related
but inactive compounds, abyssomicin B (2) and D (3, Figure
1) were isolated. The presence of an R,â-unsaturated ketone in
1 and its absence in the inactive components (2 and 3) suggested
that this potentially reactive functional group may serve as a
Michael acceptor in both the mechanism of action and the
biogenesis of 2 and 3. Abyssomicin B (2) was proposed to derive
from C (1) via a 1,4-addition of hydroxylamine followed by
Figure 1. Molecular structures of the abyssomicins (1-3) and postulated
biogeneses of 2 and 3.
oxidation, while abyssomicin D (3) was believed to result from
1,4-hydride addition (presumably delivered from NADH),
followed by an intramolecular Michael addition of the resulting
enolate onto the neighboring tetronic ester.
The structural novelty of abyssomicin C combined with its
impressive biological activity spurred considerable interest
among synthetic organic chemists. Its molecular architecture
possesses a number of challenging structural elements that would
need to be appropriately addressed in a total synthesis. These
elements include a strained 11-membered macrocyclic ring,
seven stereogenic centers, a potentially reactive R,â-unsaturated
† The Scripps Research Institute.
‡ University of California, San Diego.
(1) (a) Bister, B.; Bischoff, D.; Strobele, M.; Riedlinger, J.; Reicke, A.; Wolter,
F.; Bull, A. T.; Zahner, H.; Fiedler, H. P.; Su¨ssmuth, R. D. Angew. Chem.,
Int. Ed. 2004, 43, 2574-2576. (b) Riedlinger, J.; et al. J. Antibiot. 2004,
57, 271-279.
9
10.1021/ja067083p CCC: $37.00 © 2007 American Chemical Society
J. AM. CHEM. SOC. 2007, 129, 429-440
429

A R T I C L E S
Nicolaou and Harrison
Figure 3. Key steps in Sorensen’s biomimetic synthesis of abyssomicin
C (1).
Intrigued by the structure and reactivity of abyssomicin C,
we initiated a program directed toward the total synthesis of
this compound. In this article, we describe in detail our efforts
toward abyssomicin C (1) culminating in its total synthesis,11
the identification of a more active atropisomer of 1, the synthesis
of abyssomicin D (3), the development of methods to achieve
these goals, and the synthesis of a number of analogues designed
to probe the mechanism of action of this novel antibiotic.
Retrosynthetic Design. Our approach to the synthesis of
abyssomicin C (1) emphasized convergence with the ultimate
goal of generating analogues from late stage intermediates with
minimal effort (Figure 4). We elected to disassemble the C3-
C8 macrocyclic ring of the natural product by ring-closing
metathesis at C8-C9 and a lithiation/alkylation sequence at C2-
C3, leading back to the oxabicyclo[2.2.2]octane core 11, and
the protected C3-C8 fragment 12 as potential precursors. The
latter fragment was envisaged to arise from routine manipulation
of the enzymatic desymmetrization product 13,12 while the
oxabicyclic core 11 was dissected via the aforementioned
epoxide trapping with the tetronic acid 14. This Dieckmann
condensation product9a,10,13 was predicted to derive from 15,
which was itself anticipated to arise from the Diels-Alder
adduct 16 via standard synthetic transformations.
Figure 2. Selected examples of natural products containing the spiro
tetronic acid cyclohexene structural motif.
ketone, and a novel fused tetronate oxabicyclo[2.2.2]octane core.
The core of abyssomicin C (1) bears a close resemblance to
other spiro tetronic acid2 cyclohexene containing natural
products, including A88696F3 (4, Figure 2), kijanolide4 (5), and
5
quartromicin A1 (6), suggesting that this novel motif may be
derived from intramolecular trapping of a spiro tetronic acid
cyclohexene oxide. To date, all synthetic approaches to this
natural product have centered on this premise, including an
elegant biomimetic total synthesis by Sorensen et al.,6 formal
total syntheses by Snider and Zou7 and Couladouros et al.,8 and
model studies on the oxabicyclo[2.2.2]octane core by Maier et
al.9 and Georgiadis et al.10 In the Sorensen et al. synthesis
(Figure 3), this type of intramolecular epoxide opening was
executed in the final step of the sequence affording the natural
product (1) as a 1:1 mixture with a compound deemed
iso-abyssomicin C, whose conclusive structural assignment was
not published.6
Results and Discussion
(2) For a recent review on the synthesis of tetronic acid containing natural
prodcuts, see: Zografos, A. L.; Georgiadis, D. Synthesis 2006, 3157-3188.
(3) (a) Bonjouklian, R. Tetrahedron Lett. 1995, 36, 332-332. (b) Bonjouklian,
R.; Mynderse, J. S.; Hunt, A. H.; Deeter, J. B. Tetrahedron Lett. 1993, 34,
7857-7860.
(4) Mallams, A. K.; Puar, M. S.; Rossman, R. R.; McPhail, A. T.; Macfarlane,
R. D.; Stephens, R. L. J. Chem. Soc., Perkin Trans. 1 1983, 1497-1534.
(5) (a) Roush, W. R.; Barda, D. A.; Limberakis, C.; Kunz, R. K. Tetrahedron
2002, 58, 6433-6454. (b) Kusumi, T.; Ichikawa, A.; Kakisawa, H.;
Tsunakawa, M.; Konishi, M.; Oki, T. J. Am. Chem. Soc. 1991, 113, 8947-
8948.
Early Studies toward the Oxabicylic Core. Our work on
the abyssomicins began with the preparation of (()-16 (Scheme
1), employing the Lewis acid-templated Diels-Alder reaction
developed by Ward and Abaee.14 Thus, the magnesium salt of
2,4-hexadienol (17) was reacted with n-butanol (to adjust the
(6) Zapf, C. W.; Harrison, B. A.; Drahl, C.; Sorensen, E. J. Angew. Chem.,
Int. Ed. 2005, 44, 6533-6537.
(11) Nicolaou, K. C.; Harrison, S. T. Angew. Chem., Int. Ed. 2006, 45, 3256-
3260.
25
(7) Snider, B. B.; Zou, Y. F. Org. Lett. 2005, 7, 4939-4941.
(8) Couladouros, E. A.; Bouzas, E. A.; Magos, A. D. Tetrahedron 2006, 62,
5272-5279.
(12) Prepared in ca. 90% ee (as determined by [R]D ) +10.5, CDCl3, c )
3.0); see: Lin, G. Q.; Xu, W. C. Bioorg. Med. Chem. 1996, 4, 375-380.
(13) For examples of similar Dieckmann cyclization reactions, see: (a) Roush,
W. R.; Reilly, M. L.; Koyama, K.; Brown, B. B. J. Org. Chem. 1997, 62,
8708-8721. (b) Takeda, K.; Shibata, Y.; Sagawa, Y.; Urahata, M.; Funaki,
K.; Hori, K.; Sasahara, H.; Yoshii, E. J. Org. Chem. 1985, 50, 4673-
4681.
(9) (a) Rath, J. P.; Kinast, S.; Maier, M. E. Org. Lett. 2005, 7, 3089-3092.
(b) Rath, J. P.; Eipert, M.; Kinast, S.; Maier, M. E. Synlett 2005, 314-
318.
(10) Zografos, A. L.; Yiotakis, A.; Georgiadis, D. Org. Lett. 2005, 7, 4515-
4518.
(14) Ward, D. E.; Abaee, M. S. Org. Lett. 2000, 2, 3937-3940.
9
430 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

Abyssomicin C, Atrop-abyssomicin C, and Abyssomicin D
A R T I C L E S
Scheme 1. Synthesis of the Diene (()-23a
a Reagents and conditions: (a) MeMgBr (1.0 equiv), n-BuOH (1.0 equiv),
PhCH3, 0 °C, 30 min; then methyl acrylate (10 equiv), 25 °C, 24 h, 90%.
(b) KHMDS (1.5 equiv), THF, -78 °C, 30 min; then (EtO)3P (2.0 equiv),
O2 (excess), 1 h, 96%. (c) aq 1 M LiOH (2.0 equiv), THF, 25 °C, (not
isolated). (d) NaH (2.0 equiv), PhSH (2.0 equiv), DMF, 0 °C; then 19 (1.0
equiv), 110 °C. (e) (MeO)3CH (15 equiv), H2SO4 (2.0 equiv), MeOH, 65
°C, 5 h, 82% (over two steps). (f) Ammonium molybdate (0.1 equiv), H2O2
(5.0 equiv), EtOH, 0 °C, 10 min. (g) TMSCl (1.2 equiv), Et3N (3.6 equiv),
4-DMAP (0.1 equiv), CH2Cl2, 25 °C, 72% (over two steps). (h) LiHMDS
(1.0 equiv), THF/DMPU (3:1), -78 °C, 5 min; then ICH2Cl (2.0 equiv),
15 min. (i) Na(Hg) (4.0 equiv), Na2HPO4 (7.3 equiv), THF/MeOH (1:1),
-40 °C, 3 h. (j) aq 1 M HCl (0.01 equiv), MeOH, 25 °C, 15 min, 71%
(over three steps). Abbreviations: KHMDS, potassium bis(trimethylsilyl)-
amide; THF, tetrahydrofuran; DMF, N,N-dimethylformamide; TMSCl,
chlorotrimethylsilane; 4-DMAP, 4-(dimethylamino)pyridine; LiHMDS,
lithium bis(trimethylsilyl)amide; DMPU, 1,3-dimethyl-3,4,5,6-tetrahydro-
2(1H)-pyrimidinone.
Figure 4. Retrosynthetic analysis of abyssomicin C (1). RCM: ring-closing
metathesis.
Mg(II) salt aggregation state) and methyl acrylate, which after
cycloaddition via the putative transition state 18 and extrusion
of methanol, afforded the desired lactone (()-16 as a single
diastereomer and in 90% yield. Quenching of the derived
potassium enolate of (()-16 with oxygen and (EtO)3P then
afforded the R-hydroxylation product (()-19 in 96% yield.9a
While hydrolysis of this lactone was successful, attempts to
manipulate the resulting salt (()-20 for the purpose of trans-
forming the C9 alcohol into an olefin were not, instead leading
to substantial amounts of the starting lactone (()-19. To prevent
re-lactonization, (()-19 was opened at the γ-position with the
sodium salt of benzenethiol, and the resulting carboxylate was
methylated. This afforded the thioether (()-21 in 82% yield
for the two-step sequence. Thioether oxidation was achieved
with catalytic ammonium molybdate and hydrogen peroxide,
and the so-obtained crude sulfone was directly protected leading
to the TMS ether (()-22 in a 72% overall yield. This enabled
the execution of a Julia-type olefination15 to install the C9 double
bond as required for the proposed key olefin metathesis step.
Thus, the sulfone was rapidly deprotonated in THF/DMPU (3:
1) at -78 °C and quenched with chloroiodomethane, affording
a 5:1 mixture of chloromethylene sulfones, which was im-
mediately subjected to a sodium amalgam reduction. This
process generated the desired olefin along with partial to full
cleavage of the TMS ether, whose complete deprotection was
achieved by the action of catalytic aqueous HCl in MeOH,
furnishing diene (()-23 in 71% overall yield.
Shortly after the execution of this work, Ward and Souweha
published an enantioselective variant of the above Diels-Alder
reaction (Figure 5) involving the use of a stoichiometrically
generated diene-Zn(II)-BINOL-Mg(II) complex 24.16 Al-
though structural characterization of this complex has not been
achieved, control experiments suggested that Zn(II) serves as a
bridge between BINOL and the diene, while Mg(II) serves as
a Lewis acid to activate incoming dienophiles. Upon reaction
(15) Kocienski, P. Phosphorus Sulfur 1985, 24, 97-127.
(16) Ward, D. E.; Souweha, M. S. Org. Lett. 2005, 7, 3533-3536.
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 431

A R T I C L E S
Nicolaou and Harrison
Scheme 2. Synthesis of Racemic (()-28 and Attempted
Diels-Alder Reaction To Give (()-27a
a Reagents and conditions: (a) DABCO (1.5 equiv), thioanisole (1.5
equiv), n-BuLi (1.5 equiv), THF, 0 °C, 10 min; then 25 °C, 1 h; then 29
(1.0 equiv), -78 °C, 30 min, 65%. (b) MeMgBr (1.5 equiv), PhCH3, 0 °C;
then methyl acrylate (10 equiv), 55 °C, 24 h, 30%. Abbreviations: DABCO,
1,4-diazabicyclo[2.2.2]octane.
chiral center at C9. This functionality was intended to serve
two functions: (a) it would act as a stereocontrol element
enabling the enantioselective synthesis of 23 and (b) its extrusion
via reductive elimination would provide access to the requisite
olefin at C9. Compound 26 was expected to arise from
R-hydroxylation of 27, which was, in turn, expected to be
derived from a diastereoselective Mg(II) templated Diels-Alder
reaction between the chiral diene 28 and methyl acrylate. All
that would then remain was the enantioselective synthesis of
the secondary allylic alcohol 28, a task that we felt modern
synthetic methods could readily accomplish.
Figure 5. Ward’s enantioselective Diels-Alder reaction via a bimetallic
complex.16
Diels-Alder Reaction Development. To test the Diels-
Alder reaction in question, we prepared the racemic diene (()-
28 via alkylation of 2,4-hexadienal (29) with lithiothioanisole17
in 65% yield (Scheme 2). The magnesium salt of (()-28 was
generated and treated with methyl acrylate at room temperature;
however, unlike the smooth, albeit sluggish, reaction seen with
the magnesium salt of 2,4-hexadienol (Scheme 1), the added
bulk of the thiophenylmethylene unit was sufficient to suppress
[4 + 2] cycloaddition. Elevation of the temperature to 55 °C
allowed the desired Diels-Alder reaction to take place after
24 h with complete diastereoselectivity, although in only 30%
yield, with the remainder of the starting material being consumed
to a complex mixture of undesired products (Scheme 2). Also,
unlike the reaction with 2,4-hexadienol, the addition of a
sacrificial alcohol (n-BuOH or i-PrOH) did not enhance the yield
or rate of reaction. Hoping to improve upon this situation, we
screened a number of other Lewis acids (i.e., TiCl4, AlCl3, Zn-
(OTf)2, MgBr2‚OEt2/i-Pr2Net,18 and MeAlCl2) hoping that
modulating the strength of the Lewis acid might suppress
decomposition or increase the rate of cycloaddition, although
these pursuits proved to be fruitless, with the MeMgBr generated
Mg(II)-diene salt being optimal. Temporarily thwarted, we
turned to Ward’s enantioselective variant of this reaction for
inspiration (Figure 5). As the BINOL mediated reaction proceeds
much faster (12 h) than the simple Mg(II) templated reaction
(48 h), we had hoped that we could use a similar complex to
Figure 6. Second generation retrosynthetic analysis of 28.
of this complex with methyl acrylate, an enantio- and diaste-
reoselective Diels-Alder reaction ensues, presumably through
transition state 25, affording lactone (+)-16 in 93% ee and 95%
yield. Interestingly, this reaction proceeds much faster than the
original racemic variant (Scheme 1) despite being carried out
at higher dilution (0.02 vs 0.3 M). This reaction would have
rendered our synthesis of 23 enantioselective; however, we were
dissatisfied with inefficiencies in our existing route, and the use
of stoichiometric BINOL and dimethyl zinc in the first step was
less than desirable. Although we ultimately sought a more
efficient process, the previous sequence (Scheme 1) provided
sufficient quantities of (()-23 to explore subsequent steps in
the synthesis of the natural product’s core.
Improved Approach to 23. In re-evaluating our approach
to 23, we elected to keep several successful features from our
previous work while eliminating unnecessary functional group
manipulations, a theme that led to the more efficient approach
depicted in Figure 6. It was thus envisioned that 23 could derive
from 26, a seemingly more complex molecule with an additional
(17) Corey, E. J.; Seebach, D. J. Org. Chem. 1966, 31, 4097-4099.
(18) Barriault, L.; Thomas, J. D. O.; Cle´ment, R. J. Org. Chem. 2003, 68, 2317-
2323.
9
432 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

Abyssomicin C, Atrop-abyssomicin C, and Abyssomicin D
A R T I C L E S
Table 1. Optimization of the Lewis-Acid Templated Diels-Alder
Reaction between (()-28 and Methyl Acrylate
use of 3 equiv of 33 is less than ideal, the auxiliary could be
removed by washing with aqueous 2 M HCl and recovered (80%
recovery) by careful neutralization with concentrated NH4OH.
We suspect that the requirement for excess auxiliary is a
consequence of the relatively poor affinity between Mg(II) and
amine.
Enantioselective Synthesis of the Oxabicyclic[2.2.2]octane
Core. With a productive and selective Diels-Alder cycloaddtion
in hand, we returned to the enantioselective synthesis of the
tetronate core as shown in Scheme 3. Alkylation of the known
Weinreb amide 3420 with lithiothioanisole17 afforded the ketone
35 in 81% yield. Corey’s Me-(R)-CBS catalyst21 and cat-
echolborane served to reduce this ketone enantioselectively,
affording hydroxy compound 28 in 90% ee and 95% yield. As
expected from a literature precedent,22 and confirmed by the
eventual conversion of 28 to abyssomicin C (1), this reaction
occurs with the diene behaving as the larger of the two ketone
substituents in the putative transition state 36 (Scheme 3).
Subjecting our now enantio-enriched diene to the previously
described Diels-Alder conditions afforded the adduct 27 as a
single diastereomer (80% yield), presumably through the
transition state 37. The desired hydroxy lactone 26 was prepared
from this adduct via R-hydroxylation in 74% yield, setting the
stage for the key Julia-type reductive elimination. Our early
approaches toward this goal employing less direct methods,
specifically oxidation to the corresponding sulfone followed by
reduction with Na(Hg) amalgam15 or Mg(0)/HgCl2,23 were met
with little success. It was then discovered that the direct
reduction of the thioether at C8 proceeds smoothly with a
catalytically generated lithium 4,4′-di-tert-butylbiphenyl radical
anion.24 The resulting carbanion causes an elimination reaction
that fractures the lactone, leading simultaneously to the requisite
olefin at C9 and a carboxlate anion. Methylation of the latter
functionality with MeI in the same reaction vessel proceeded
better than an isolation/methylation sequence leading to the
skipped diene 23 in 99% yield. A vanadium directed epoxidation
of the C11-C12 olefin was then attempted with VO(acac)2 and
t-BuOOH.25 While this delivered the intended epoxide with
complete stereo- and regiocontrol, the reaction was inefficient,
requiring repeated addition of the catalyst to push the reaction
to completion (up to 0.4 equiv). We found that we could
improve the efficiency of this process using rarely employed
a Addition of base was performed at 0 °C before addition of methyl
acrylate (10 equiv) and warming to 55 °C. b Reaction times reflect the time
at which consumption of the diene was complete as measured by 1H NMR
spectroscopy on the crude reaction mixture unless stated otherwise. c Isolated
yield unless stated otherwise. d Reaction performed according to the
conditions summarized in Figure 5. e Reaction stopped before completion
affording 80% of recovered diene. f Trace product detected in the 1H NMR
spectrum of recovered diene.
enhance the rate of cycloaddition. We thus attempted to generate
the diastereomeric mixture of (()-28-Zn(II)-(()-BINOL-
MgBr complexes following the established protocol16 (Table
1), and to our surprise, only trace amounts of (()-27 were
detected; however, decomposition of the diene was slowed,
leading to 80% starting material recovery after 24 h (entry 2).
In a parallel experiment, BINOL was replaced with catechol
(31, entry 3), which led to a substantial improvement (35%
yield), along with complete diene decomposition after 36 h.
Optimistic that the catechol system was catalyzing the intended
cycloaddition, albeit slowly, we investigated the structurally
similar 2-aminophenols. It was reasoned that in this reaction
manifold, the Lewis basic amine would interact with the Mg-
(II) salt of the diene while the Mg(II) bound to the phenol would
serve to activate and orient the incoming dienophile. Using
commercial 2-aminophenol (32, entry 4) with 2 equiv of
MeMgBr led to an immediate increase in reaction turnover,
affording a 49% yield of (()-27 with complete consumption
of (()-28 after 24 h. Auxiliaries possessing alkylated nitrogen
including the commercially available 2-(pyrrolidin-1-yl)phenol
3319 exhibited a slower rate of cycloaddition, along with an
extended lifetime of the diene. Encouraged by these results, the
number of equivalents of 33 and MeMgBr was incremented
(entries 6 and 7) causing the yield of (()-27 to climb up to
80% with 3 equiv of 33 and 4 equiv of MeMgBr. Although the
26
VO(OEt)3 (0.2 equiv) with the same stoichiometric oxidant,
a protocol that afforded the desired epoxide in 93% yield as a
single stereoisomer. Routine acetylation of the free alcohol then
afforded the acetoxy methyl ester 15 (95% yield), setting the
stage for the key Dieckmann condensation reaction. Deproto-
nation of 15 with 2.5 equiv of LiHMDS at -78 °C, followed
by warming to room temperature9a,10,13 and quenching with
saturated aqueous NH4Cl,10 generated the tetronic acid 14, which
was not isolated. Instead, the aqueous/organic mixture was
heated to reflux for 2 h, effecting the intramolecular trapping
(20) Hiyama, T.; Reddy, G. B.; Minami, T.; Hanamoto, T. Bull. Chem. Soc.
Jpn. 1995, 68, 350-363.
(21) Corey, E. J.; Helal, C. J. Angew. Chem., Int. Ed. 1998, 37, 1986-2012.
(22) Cho, B. T.; Choi, O. K.; Kim, D. J. Tetrahedron: Asymmetry 2002, 13,
697-703.
(23) Lee, G. H.; Lee, H. K.; Choi, E. B.; Kim, B. T.; Pak, C. S. Tetrahedron
Lett. 1995, 36, 5607-5608.
(19) (a) 2-(Dimethylamino)phenol was also used with near identical results. (b)
Commercially available from a number of sources; however, we prepared
this compound by reaction of 2-aminophenol (1.0 equiv) with 1,4-
dibromobutane (1.05 equiv) and i-Pr2NEt (3.0 equiv) in toluene (1.3 M) at
65 °C for 12 h (see Supporting Information).
(24) Cohen, T.; Bhupathy, M. Acc. Chem. Res. 1989, 22, 152-161.
(25) Tanaka, S.; Yamamoto, H.; Nozaki, H.; Sharpless, K. B.; Michaels, R. C.;
Cutting, J. D. J. Am. Chem. Soc. 1974, 96, 5254-5255.
(26) Evans, J. M.; Kallmerten, J. Synlett 1992, 269-271.
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 433

A R T I C L E S
Nicolaou and Harrison
Scheme 3. Improved Synthesis of Enantioenriched 23 and Its Conversion to the Tes-Protected Core 11a
a Reagents and conditions: (a) DABCO (1.5 equiv), thioanisole (1.5 equiv), n-BuLi (1.5 equiv), THF, -78 °C, 10 min; then 25 °C, 1 h; then 34 (1.0
equiv), -78 °C, 45 min, 81%. (b) (R)-CBS (0.1 equiv), catecholborane (1.1 equiv), CH2Cl2, -78 °C, 12 h, (95%, 90% ee). (c) 33 (3.0 equiv), MeMgBr (4.0
equiv), PhCH3, 0 °C, 10 min; then methyl acrylate (10 equiv), 55 °C, 12 h, 80%. (d) LiHMDS (1.5 equiv), THF, -78 °C, 30 min; then (EtO)3P (2.0 equiv),
O2 (excess), 1 h, 74%. (e) Li (9.6 equiv), 4,4′-di-tert-butylbiphenyl (0.5 equiv), THF, 0 °C, 1 h; then 26 (1.0 equiv), -48 °C, 3 h; quench with MeOH,
remove excess Li, concentrate; then K2CO3 (1.0 equiv), MeI (10 equiv), DMF, 60 °C, 10 min, 99%. (f) t-BuOOH (3.0 equiv), VO(OEt)3 (0.2 equiv), CH2Cl2,
25 °C, 4 h, 93%. (g) Ac2O (3.0 equiv), 4-DMAP (0.05 equiv), Et3N, 25 °C, 12 h, 95%. (h) LiHMDS (2.5 equiv), THF, -78 °C, 10 min; then 25 °C, 1 h;
then saturated aq NH4Cl, 66 °C, 2 h. (i) TESCl (1.4 equiv), imidazole (2.8 equiv), 4-DMAP (0.05 equiv), DMF, 25 °C, 2 h, 97% (over two steps).
Abbreviations: (R)-CBS, (R)-(+)-2-methyl-CBS-oxazaborolidine.
Scheme 4. Synthesis of C3-C8 Precursor 12a
ether 3927 was prepared using established procedures. Parikh-
Doering oxidation of 39, followed by alkylation with vinyl
magnesium bromide, afforded 40 as a 3:2 mixture of allylic
alcohols that were not separated (74%, over two steps). Three
standard steps, namely, PMB protection (82% yield), TBS ether
cleavage, and another Parikh-Doering oxidation (84%, over
two steps), afforded the mixture of aldehydes 12 that were
carried into the next step without separating each diastereomer.
Fragment Coupling and Attempted Completion of the
Synthesis. With the C3-C8 fragment 12 and the core 11 in
hand, we began investigating the final stages of the synthesis
of abyssomicin C (1, Scheme 5). Treating the protected core
fragment 11 with t-BuLi at -78 °C effected clean lithiation at
C2.28 Quenching of this anion with the mixture of aldehydes
12 afforded a mixture of inseparable diastereomers (61% yield)
that was subjected to deprotection with DDQ, affording another
inseparable mixture 41 in 96% yield (LC/MS characterization).
Treatment of 41 with Grubbs’ second-generation catalyst 4229
in refluxing CH2Cl2 (0.001 M) did not achieve ring-closing
metathesis as desired but instead led to dimeric products
(identified in the crude LC/MS trace). Assuming that the steric
bulk of the TES group prevented macrocyclization, this group
was removed with catalytic aqueous HCl/methanol in 94% yield
affording the mixture of triols 44. Selective oxidation of the
two allylic alcohols was achieved with excess IBX, leading to
10 in 64% yield. Attempted ring-closing metathesis of 10 with
a Reagents and conditions: (a) i-Pr2NEt (3.0 equiv), SO3‚pyridine (3.0
equiv), CH2Cl2/DMSO (2:1), 0 °C, 20 min. (b) Vinyl MgBr (2.0 equiv),
THF, -78 °C, 74% (over two steps). (c) NaH (2.0 equiv), PMBCl (2.0
equiv), tetra-n-butylammonium iodide (0.1 equiv), DMF, 25 °C, 2 h, 82%.
(d) aq 1 M HCl (0.01 equiv), MeOH, 25 °C, 1 h. (e) i-Pr2NEt (3.0 equiv),
SO3‚pyridine (3.0 equiv), CH2Cl2/DMSO (2:1), 0 °C, 20 min, 84% (over
two steps). Abbreviations: DMSO, dimethylsulfoxide; PMBCl, 4-meth-
oxybenzoyl chloride.
of the epoxide exclusively at C12,8,9a,10 leading to 38, containing
the desired tetronate fused oxabicyclo[2.2.2]octane framework.
Chromatographic purification of 38 was met with some decom-
position, so the crude product was immediately protected as its
TES ether 11, providing a near quantitative yield (97%) for the
two-step sequence.
Synthesis of C3-C8 Fragment. To maximize convergence,
we had hoped to attach the macrocyclic portion of the molecule
in a fashion that would require minimal post-coupling manipula-
tions. To this end, we prepared the aldehyde 12 using routine
chemical transformations as shown in Scheme 4. Starting from
the enzymatic desymmetrization product 13,12 the known TBS
(27) Prusov, E.; Rohm, H.; Maier, M. E. Org. Lett. 2006, 8, 1025-1028.
(28) Takeda, K.; Kawanishi, E.; Nakmura, H.; Yoshii, E. Tetrahedron Lett. 1991,
32, 4925-4928.
(29) Scholl, M.; Ding, S.; Lee, C. W.; Grubbs, R. H. Org. Lett. 1999, 1, 953-
956.
9
434 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

Abyssomicin C, Atrop-abyssomicin C, and Abyssomicin D
A R T I C L E S
Scheme 5. Attempted Synthesis of Abyssomicin C (1) via Ring-Closing Metathesisa
a Reagents and conditions: (a) t-BuLi (1.2 equiv), THF, -78 °C, 30 min; then 12 (1.1 equiv), 15 min, 61%. (b) DDQ (3.0 equiv), CH2Cl2: saturated aq
NaHCO3 (10:1), 25 °C, 30 min, 96%. (c) 42 (0.05 equiv), CH2Cl2 (0.001 M), 40 °C, 1 h. (d) aq 1 M HCl (0.01 equiv), MeOH, 25 °C, 1 h, 94%. (e) IBX
(5.0 equiv), DMSO, 25 °C, 64%. (f) 42 (0.1 equiv), 1,4-benzoquinone (0.2 equiv), CH2Cl2 (0.002 M), 40 °C, 1 h, 14%. (g) 42 (0.05 equiv), CH2Cl2 (0.002
M), 40 °C, 1 h, 78%. Abbreviations: DDQ, 2,3-dichloro-5,6-dicyano-1,4-benzoquinone; IBX, o-iodoxyiodobenzene.
catalytic 42 in CH2Cl2 generated a complex mixture of unidenti-
fied byproducts. Conducting the same reaction with 1,4-
benzoquinone, hoping to suppress Ru(III) mediated isomeriza-
tion reactions,30 did allow the isolation of small quantities (14%
yield) of abyssomicin C (1) as an inseparable 2:1 mixture with
10, with the rest of the material being consumed to a mixture
of unidentified byproducts. Whereas 10 proved to be a poor
ring-closing metathesis substrate, the diastereomeric mixture of
triols 44 closed quite effectively under standard conditions,
leading to the ring-closed mixture of triols 45 in 78% yield (LC/
MS characterization). Unfortunately, all attempts to oxidize 45
to 1 with selective oxidants (i.e., IBX and MnO2) afforded only
singly oxidized products, presumably the result of intramolecular
hemiketalization following the first oxidation. Attempts to use
stronger oxidants (i.e., Dess-Martin, PDC, and Swern) were
not successful, leading to oxidation of the unactivated alcohol
at C11. Suspecting that the two additional sp2 centers in 10
(relative to 45) were responsible for its reluctance to cyclize,
we elected to design an improved substrate with the same
hybridization pattern as 45, which was devoid of the intramo-
lecular hemiketalization problem.
OEt2 in 90% yield. Standard deacetylation with K2CO3 in MeOH
then afforded diol 48 in 97% yield. A shorter approach to 48
was also developed via alkylation of the lithio derivative of 11
with lactone 4932 affording a crude lactol, which was thioket-
alized directly to furnish the desired adduct in 76% yield over
the two-step sequence. Oxidation of 48 with IBX occurred
chemoselectively, affording an aldehyde that was immediately
reacted with vinyl magnesium bromide leading to a 2:3 mixture
of allylic alcohols (50) in 65% combined yield. As anticipated,
50 served as an excellent ring-closing metathesis substrate
furnishing a 2:3 mixture of C8-C9 trans-allylic alcohol products
51a and 51b in 85% yield upon exposure to catalyst 42.
As we initiated the ring-closing metathesis strategy, concern
existed that this step could afford atropsiomers at the C8-C9
double bond resulting from hindered rotation of this olefin
through the center of the macrocycle. This form of atropisom-
erism has been observed in trans-olefin containing medium-
sized rings.33 Interestingly, 51a and 51b are each formed as
only the desired atropisomer as evidenced by key NOE
resonances of the purified allylic alcohols (see Figure 7). Our
proposed rationale for this atropselectivity is based on facial
selection in the ring-closing metathesis step. After initiation of
the reaction by the catalyst at the less hindered olefin at C8,
approach of the activated ruthenium carbene species to the C9
olefin could theoretically occur from either the si or the re face
(Figure 7). Approach from the re face leads to the metallocy-
clobutane intermediate 53, which after cycloreversion would
afford the observed products 51a and 51b. The alternate
approach from the si face would lead to metallocyclobutane
54, whose decomposition would afford 55. The failure to
Completion of the Synthesis of Abyssomicin C. We
speculated that blocking one carbonyl group of the cyclization
precursor as a dithioacetal at either C3 or C7 would prevent
this undesired hemiketalization from occurring. The carbonyl
at C3 was selected to undergo this modification, as we
anticipated that this change would be less likely to interfere
with the ring-closing metathesis step. The execution of this
strategy began (Scheme 6) by quenching the lithiated species
of 11 with the known aldehyde 46,31 leading to a mixture of
secondary alcohols (75% yield) that was oxidized with IBX to
afford the ketone 47 (90% yield). The corresponding dithiolane
was prepared by thioketalization with 1,2-ethanedithiol and BF3‚
(32) Prepared according to the sequence reported for ent-49 using (R)-(+)-R-
methylbenzylamine; see: Paquette, L. A.; Boulet, S. L. Synthesis 2002,
888-894.
(33) (a) Pearson, W. H.; Hembre, E. J. J. Org. Chem. 1996, 61, 5546-5556.
(b) Sudau, A.; Munch, W.; Nubbemeyer, U.; Bats, J. W. J. Org. Chem.
2000, 65, 1710-1720. (c) Anderson, E. A.; Davidson, J. E. P.; Harrison,
J. R.; O’Sullivan, P. T.; Burton, J. W.; Collins, I.; Holmes, A. B.
Tetrahedron 2002, 58, 1943-1971.
(30) Hong, S. H.; Sanders, D. P.; Lee, C. W.; Grubbs, R. H. J. Am. Chem. Soc.
2005, 127, 17160-17161.
(31) Ley, S. V.; Dixon, D. J.; Guy, R. T.; Palomero, M. A.; Polara, A.;
Rodriguez, F.; Sheppard, T. D. Org. Biomol. Chem. 2004, 2, 3618-3627.
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 435

A R T I C L E S
Nicolaou and Harrison
Scheme 6. Synthesis of Abyssomicin C Carbocyclic Framework
51a,b via Ring-Closing Metathesisa
a Reagents and conditions: (a) t-BuLi (1.2 equiv), THF, -78 °C, 30
min; then 46 (1.1 equiv), 15 min, 75%. (b) IBX (2.5 equiv), DMSO, 25
°C, 45 min, 90%. (c) BF3‚OEt2 (5.0 equiv), 1,2-ethanedithiol (5.0 equiv),
CH2Cl2, 25 °C, 12 h, 90%. (d) K2CO3 (3.4 equiv), MeOH, 25 °C, 4 h,
97%. (e) t-BuLi (1.2 equiv), THF, -78 °C, 30 min; then 49 (1.1 equiv), 15
min. (f) 1,2-Ethanedithiol (4.0 equiv), TMSOTf (1.0 equiv), -78 °C, 1 h,
76% (over two steps). (g) IBX (2.5 equiv), DMSO, 25 °C, 45 min. (h)
Vinyl MgBr (3.0 equiv), THF, -78 °C, 15 min, 65% (over two steps). (i)
42 (0.05 equiv), CH2Cl2 (0.002 M), 40 °C, 1 h, 85%. Abbreviations:
TMSOTf, trimethylsilyl trifluoromethanesulfonate.
Figure 7. Proposed rationale for the formation of 51a,b and not 55 by the
ring-closing metathesis reaction.
analysis was required to determine that 57 is a diastereomeric
atropisomer of abyssomicin C (1) that we aptly named atrop-
abyssomicin C (see ORTEP drawing, Figure 8 ).
observe 55 suggests that the neighboring tetronic ester provides
sufficient steric bulk as to prevent the latter pathway from
occurring.
Structural Differences. Comparison of the X-ray structures
of abyssomicin C (1)1 and atrop-abyssomicin C (57) reveals
subtle but important differences (Figure 9). The most striking
deviation between the two lies in the R,â-unsaturated ketone
region of the molecule (Figure 9a). In abyssomicin C (1), the
carbonyl adopts a transoid conformation with respect to the C8-
C9 olefin (OdC7-C8dC9, φ ) 144.8°), whereas in atrop-
abyssomicin C (57), the carbonyl adopts a cisoid conformation
(OdC7-C8dC9, φ ) 26.9°). Not surprisingly, the dithiolane
precursor 56 adopts a very similar conformation (Figure 8,
OdC7-C8dC9, φ ) 37.1°). As the enone moiety in 57 exhibits
a higher degree of conjugation between the C7 carbonyl and
the C8-C9 olefinic bond than 1, it was expected that 57 would
be a more reactive Michael acceptor, and further studies appear
to indicate that this hypothesis is correct. Other regions of the
macrocyclic ring exhibit observable structural deviations, es-
pecially around the C19 methyl (C19-C6-C7dO, 1: φ )
-39.0° vs 57: φ ) 91.9°; Figure 9c) and the C3 carbonyl
(OdC3-C2dC16, 1: φ ) 33.8° vs 57: φ ) 56.8°) residues.
Pressing forward in the total synthesis (Scheme 7), the
treatment of 51a and 51b with a DMSO solution of IBX
chemoselectively oxidized the allylic alcohol, leading to 56 (50%
yield). Anticipating that a single deprotection separated 56 from
abyssomicin C (1), this dithioketal intermediate was reacted with
PhI(OTFA)2 in 10:1 CH3CN/H2O to remove the protecting
group moiety. To our surprise, a stable compound (57) was
isolated in 71% yield whose spectroscopic data did not match
that of the previously reported natural product, despite being
very similar to it. While attempting to obtain extensive NMR
data on 57 in unstabilized CDCl3, we fortuitously observed its
slow isomerization into abyssomicin C (1), eventually leading
to a 2:1 equilibrium mixture of the two after 24 h favoring the
targeted natural product (1). Purification of 1 by preparative
TLC afforded a compound whose spectroscopic and physical
properties matched those of abyssomicin C as previously
reported.1,6 Although this meant that our total synthesis of
abyssomicin C (1) was complete, the structure of the synthetic
intermediate 57 remained a mystery. X-ray crystallographic
Closer inspection of the X-ray structures of 1 and 57 suggests
substantial strain within each macrocyclic ring. In both struc-
9
436 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

Abyssomicin C, Atrop-abyssomicin C, and Abyssomicin D
A R T I C L E S
Scheme 7. Synthesis of Abyssomicin C (1) via Atrop-abyssomicin
(57)a
a Reagents and conditions: (a) IBX (2.5 equiv), DMSO, 25 °C, 45 min,
50%. (b) PhI(OTFA)2 (3.0 equiv), CH3CN/H2O (10:1), 25 °C, 10 min, 71%.
(c) Unstabilized CDCl3, 25 °C, 24 h, 67% (100% based on recovered starting
material). Abbreviations: PhI(OTFA)2, bis(trifluoroacetoxy)iodobenzene.
Figure 9. (a) Overlay of the X-ray derived structures of abyssomicin C
(1, blue) and atrop-abyssomicin C (57, yellow). (b) View down the C7-
C8 bond for 1 and 57. (c) View down the C7-C6 bond for 1 and 57.
Figure 8. X-ray based ORTEP drawings of the dithiolane 56 and atrop-
abyssomicin 57. Spheres are drawn at the 50 and 40% probability level,
respectively.
it was surprising that early attempts to convert atrop-abyssomicin
C (57) to abyssomicin C (1) under thermal conditions (i.e.,
refluxing MeOH or toluene at 180 °C, xylenes, and microwave
irradiation)11 were unsuccessful. Thermal equilibration was
eventually achieved by refluxing in 1,2-dichlorobenzene at 180
°C (Table 2, entry 2) for 12 h, affording a 1:1 mixture of isomers
at equilibrium, albeit with substantial decomposition.
Hoping to learn more about the acid-catalyzed interconversion
of 1 and 57, a number of other conditions was screened that
failed to effect isomerization (entries 3-6, Table 2) including
TFA, BF3‚OEt2, CSA, and aqueous 1 M HCl. As expected, the
addition of catalytic ethereal HCl to 57 in a variety of deuterated
organic solvents (entries 7-10) rapidly gave rise to equilibrium
mixtures of 1 and 57 with CDCl3 affording the best ratio (2.5:
1) favoring abyssomicin C (1). Interestingly, subjecting 57 to
the conditions employed in the final step of Sorensen’s synthesis
(p-TsOH, LiCl, CH3CN, 50 °C, 2 h)6 also causes this isomer-
ization to take place, presumably from the in situ generation
of HCl.
tures, the C8-C9 olefin exhibits a deviation from planarity as
the result of twisting, as evidenced by the C10-C9dC8-C7
dihedral angles of 166.9° in 1 and 165.9° in 57.34 Furthermore,
the sp2 center at C2 is substantially pyramidalized in both
compounds, deviating from planarity by roughly 7° in 1 and
11° in 57.
Mechanism of Atropisomer Interconversion. Most litera-
ture examples of room temperature atropisomerism result from
a sterically hindered or blocked rotation, such as that found in
substituted biaryls35 or trans-olefin containing medium-sized
rings,33 although examples of stable conformers in strained
medium-sized rings that lack an obvious steric repulsion have
been observed.36 As no obvious steric obstructions appear to
prevent the rotation of the C6-C7 and C7-C8 σ-bonds in
compounds 1 and 57, short of some minor eclipsing interactions,
(34) Vazquez, S.; Camps, P. Tetrahedron 2005, 61, 5147-5208.
(35) (a) Bringmann, G.; Mortimer, A. J. P.; Keller, P. A.; Gresser, M. J.; Garner,
J.; Breuning, M. Angew. Chem., Int. Ed. 2005, 44, 5384-5427. (b) Lloyd-
Williams, P.; Giralt, E. Chem. Soc. ReV. 2001, 30, 145-157.
(36) (a) Shea, K. J.; Gilman, J. W.; Haffner, C. D.; Dougherty, T. K. J. Am.
Chem. Soc. 1986, 108, 4953-4956. (a) Anastasiou, D.; Campi, E. M.;
Chaouk, H.; Jackson, W. R. Tetrahedron 1992, 48, 7467-7478.
The high thermal barrier to interconversion, the absence of
steric interactions to prevent interconversion, and the substantial
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 437

A R T I C L E S
Nicolaou and Harrison
Table 2. Acid-Catalyzed Isomerization of Atrop-abyssomicin C
(57) and Abyssomicin C (1)
entry
conditionsa
time (h)
ratio (57:1)b
1
2
3
4
5
6
7
8
unstabilized CDCl3
d4-1,2-dichlorobenzene, 180 ˚C
TFA/CH2Cl2 (1:1)
BF3‚OEt2 (1.0 equiv)/CH2Cl2
CSA (1.0 equiv)/CH2Cl2
24
12
24
24
24
24
1
1
1
1
2
1.0:2.0
1.0:1.0
n.r.c
n.r.
n.r.
1 M aq HCl/THF (1:3)
n.r.
1 M HCl in Et2O (0.2 equiv)/d6-THF
1 M HCl in Et2O (0.2 equiv)/CD3CN
1 M HCl in Et2O (0.2 equiv)/CD2Cl2
1 M HCl in Et2O (0.2 equiv)/CDCl3
p-TsOH (1.0 equiv), LiCl
1.6:1.0
1.4:1.0
1.0:2.2
1.0:2.5
2.0:1.0d
9
10
11
(5.0 equiv)/CD3CN, 50 °C
a All reactions were performed at 25 °C on 57 (1 mg) unless indicated
otherwise. b Ratio is at equilibrium and was determined by integration of
the H-7 and H-8 1H NMR (500 MHz) signals. c n.r. ) no reaction as
determined by NMR spectroscopy in MeOD after aqueous workup.
d Equilibrium ratio not determined.
Figure 10. Proposed mechanisms for the acid-catalyzed interconversion
of abyssomicin C (1) and atrop-abyssomicin C (57).
strain in the ground state of each atropisomer of abyssomicin
C begs two hypotheses. The first is that the high thermal barrier
to interconversion is mostly the result of an even more strained
transition state. The second is that the mechanism of the acid-
catalyzed isomerization must operate in a manifold that relieves
this strain.
(1) at the same temperature does not furnish any abyssomicin
D (3) at all, leading instead to a mixture of compounds whose
major component, compound 60, was isolated in 44% yield and
characterized through a combination of NMR and computational
modeling studies (see Supporting Information for details). These
distinct outcomes likely result from the ground-state conforma-
tion of the R,â-unsaturated ketone moiety within these substrates
prior to hydride delivery. Thus, 1,4-reduction of the cisoid enone
in 57 most likely leads to the transient formation of the Z-enolate
61, which after intramolecular Michael addition onto the
tetronate moiety and aqueous workup leads directly to abysso-
micin D (3). On the other hand, the transoid enone of 1 should
result in the formation of the E-enolate 62, whose reaction with
the tetronate core should afford 63, an apparent conformational
isomer of abyssomicin D (3). As this is not the product isolated
from the reaction, tautomerization and epimerization must take
place at C2 and C8, respectively, affording 60. Interestingly,
while attempting to obtain crystals of 60 (for X-ray analysis)
in ethanol, it was discovered that complete isomerization to
abyssomicin D (3) took place. Repeating the experiment with
monitoring by TLC indicated that 60 converted completely to
3 in under 3 h at room temperature, presumably via a series of
proton transfers and bond rotations.
Figure 10 depicts three possible avenues through which this
isomerization could occur. The first possibility (path a) involves
protonation of either atropisomer on the tetronate core by HCl.
This could render the tetronate a sufficiently good leaving group
such that the C11 hydroxy group could re-close onto C12,
generating the postulated biosynthetic precursor 9, which could
be flexible enough so as to allow C6-C7 and C7-C8 bond
rotation before retrapping of the protonated epoxide. One
problem with this hypothesis is the fact that we have not detected
any epoxy compound during our isomerization studies to date,
suggesting that this pathway is not likely operational. Alterna-
tively (path b), protonation of an oxygen atom on the tetronate
core could render C16 electron deficient, allowing it to draw
π-electron density through space from the suitably poised olefin
at C8-C9. This bridging interaction would draw C8 closer to
the tetronate core (see 58, Figure 10), relieving strain. Our last
and preferred hypothesis (path c) is that the core of the molecule
is protonated by the strong acid at C2 affording the oxocarbe-
nium ion 59 (Figure 10). This would convert the already
deformed sp2 center into one that is completely pyramidalized,
relieving strain in the macrocyclic system and allowing the
requisite σ-bond rotation to occur.
The differing reactivity between abyssomicin C (1) and its
atropisomer 57 under 1,4-reduction conditions suggests that
abyssomicin D (3) may arise from reduction of the more reactive
atrop-abyssomicin C (57) and not the originally isolated natural
product (1) as previously hypothesized.1 This pathway is
supported by a kinetics experiment whereby incubation of both
1 and 57 with the NADH analogue 6437 in CDCl3 at 55 °C led
to near complete consumption of 57 and only 30% consumption
Synthesis of Abyssomicin D. To test the proposed biomi-
metic synthesis of abyssomicin D1 (3, Figure 1), both abysso-
micin C (1) and atrop-abyssomicin C (57) were subjected to
reduction with L-Selectride (Scheme 8). Reduction of atrop-
abyssomicin (57) occurs cleanly at -78 °C leading to abysso-
micin D (3, 60% yield). Interestingly, reduction of abyssomicin
1
of 1 after 48 h (Figure 11). Analysis of the crude H NMR
(37) Lavilla, R. J. Chem. Soc., Perkin Trans. 1 2002, 1141-1156.
9
438 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007

Abyssomicin C, Atrop-abyssomicin C, and Abyssomicin D
A R T I C L E S
Scheme 8. 1,4-Reduction of Abyssomicin C (1) and
Atrop-abyssomicin C (57) Leading to Iso-abyssomicin D (60) and
Abyssomicin D (3), Respectivelya
Figure 11. Graph showing the rate of disappearance of abyssomicin C (1)
and atrop-abyssomicin C (57) in their reactions with the NADH analog 64
(5.0 equiv) in CDCl3 (determined by 1H NMR spectrum integration against
2,6-di-tert-butyl-4-methylphenol as an internal standard). (a) Neither
abyssomicin D (3) nor iso-abyssomicin D (60) were detected in the crude
1H NMR spectrum of the product. (b) Abyssomicin D (3) was the only
1
product detected in the crude H NMR spectrum.
Scheme 9. Synthesis of Simplified Analogues 66 and 67 via
Cross-Metathesis and Ac-abyssomicin C (68)a
a Reagents and conditions: (a) L-Selectride (1.2 equiv), THF, -78 °C,
30 min, 60%. (b) L-Selectride (1.2 equiv), THF, -78 °C, 30 min, 44%. (c)
EtOH, 25 °C, 3 h, 100%. Abbreviations: L-selectride, lithium tri-sec-
butylborohydride.
a Reagents and conditions: (a) 42 (0.05 equiv), methyl acrylate (10
equiv), CH2Cl2 (0.1 M), 40 °C, 1 h, 84%. (b) 42 (0.05 equiv), methyl vinyl
ketone (10 equiv), CH2Cl2 (0.1 M), 40 °C, 1 h, 89%. (c) Ac2O (15 equiv),
Et3N, 25 °C, 63%.
spectra following this experiment indicated that 57 had been
converted predominately to abyssomicin D (3), while abysso-
micin C (1) afforded neither previously isolated form of
abyssomicin D (3 or 60). Although these experiments support
the notion that atrop-abyssomicin C (57) may be a product of
biosynthesis, further experimentation with the host organism
would be required for unequivical confirmation of this hypoth-
esis.
Synthesis of Analogues and Biological Evaluation. One of
our initial goals in this program was the synthesis of analogues
for biological and mechanistic evaluations, with a particular
interest in the preparation of analogues involving substitution
along the C3-C8 portion of the macrocyclic ring. It was thus
dissappointing to find that the ring-closing metathesis step
(Scheme 6) is not a general process within this structural series
and that adjusting the length or methyl substitution pattern of
the hydrocarbon chain (C3-C8) prevents efficient macrocy-
clization (data not shown). We instead prepared a number of
simpler analogues based on the core of the natural product as
shown in Scheme 9. Cross metathesis reactions of the core
compound 38 with methyl acrylate or methyl vinyl ketone
afforded two potential Michael acceptor compounds 66 and 67
(in 84 and 89% yield, respectively). Furthermore, the free
hydroxy group of abyssomicin C (1) was acylated under standard
conditions affording 68 in 63% yield to determine whether or
not this functionality is important for activity. These compounds
9
J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007 439

A R T I C L E S
Nicolaou and Harrison
Table 3. Minimal Inhibitory Concentrations (MICs) of Abyssomicin
compound 56, with its R,â-unsaturated ketone intact and in the
more reactive cisoid conformation (see Figure 8), shows activity,
although less so than the two abyssomicin C atropisomers. As
expected, based on the proposed mechanism of action, the allylic
alcohols 51a and 51b are not active (entries 5 and 6). The
simplified core compounds 38, 66, and 67 (entries 7-9) also
lack activity, suggesting that a properly oriented Michael
acceptor is required for activity. The use of a simple nutrient
broth was required in this assay (yeast extract), as conducting
the MIC assays in more complex media (Mueller Hinton II)
resulted in the entire loss of activity, suggesting that the use of
these compounds as bacterialcidal agents in vivo may be
problematic.
C Analogues against Methicillin-Resistant S. aureus (MRSA)a
Conclusion
Described in this full account is a successful total synthesis
of the recently discovered antibiotic abyssomicin C (1) and its
inactive cousin abyssomicin D (3). En route to these natural
products, a number of roadblocks encouraged innovation leading
to the development of a new Lewis acid-templated Diels-Alder
reaction. During this program, atrop-abyssomicin C (57) was
discovered, which demonstrated an unusual form of atropisom-
erism and a novel acid-catalyzed conformational equilibration.
The identification of 57 enabled the hypothesis that this
previously unknown atropisomer is synthesized by the host
organism and that it may serve as a biosynthetic precursor to
abyssomicin D (3). Its eventual isolation from natural sources
will, therefore, not be surprising. Finally, a number of analogues
based on synthetic intermediates were screened for activity
against MRSA, further expanding our understanding of the
structure-activity relationships within the abyssomicin scaffold.
a MIC values were determined with serial dilutions in 96-well plates
against a 1:10 000 dilution in nutrient broth (DIFCO) of a 24 h culture of
methicillin-resistant S. aureus (ATCC 33591) at 35 °C. Reported MIC values
reflect the concentration at which no growth could be visually detected
after 24 h.
Acknowledgment. We thank Dr. D. Thayer for assistance
with the MIC assays, Prof. K. B. Sharpless for advice on
vanadium mediated epoxidations, Prof. E. J. Sorensen for helpful
discussions, and Drs. D. H. Huang and R. Chadha for assistance
with NMR spectroscopic and X-ray crystallographic analyses,
respectively. Financial support for this work was provided by
the Skaggs Institute for Research (predoctoral fellowship to
S.T.H.).
as well as other synthetic intermediates were screened for
activity in a minimal inhibitory concentration (MIC) assay
against methicillin-resistant S. aureus (MRSA), and the results
obtained are provided in Table 3. As anticipated, atrop-
abyssomicin C (57), with its more reactive Michael acceptor,
is a slightly more potent agent than abyssomicin C (1). Acylated
abyssomicin C (68) showed essentially the same activity as the
parent natural product, suggesting that the C11 hydroxy group
is not involved in hydrogen bonding in the active site of the
targeted enzyme, unless the compound serves as a prodrug to 1
through enzymatic acetate hydrolysis. The dithiolane protected
Supporting Information Available: Experimental procedures
and compound characterization, complete ref 1b, and copies of
select 1H and 13C NMR spectra. This material is available free
of charge via the Internet at http://pubs.acs.org.
JA067083P
9
440 J. AM. CHEM. SOC. VOL. 129, NO. 2, 2007