Total synthesis of ustiloxin D and considerations on the origin of selectivity of the asymmetric allylic alkylation

Article abstract of DOI:10.1021/jo048854f

As part of investigations into cell cycle checkpoint inhibitors, an asymmetric synthesis of the antimitotic natural product, ustiloxin D, has been completed. A salen-Al-catalyzed aldol reaction was employed to construct a chiral oxazoline 9 (99% yield, 98

Full text of DOI:10.1021/jo048854f

Total Synthesis of Ustiloxin D and Considerations on the Origin of
Selectivity of the Asymmetric Allylic Alkylation
Andrew M. Sawayama,† Hiroko Tanaka,† and Thomas J. Wandless*,‡
Department of Chemistry, Stanford University, Stanford, California 94305, and
Department of Molecular Pharmacology, Stanford University, Stanford, California 94305
wandless@stanford.edu
Received July 6, 2004
As part of investigations into cell cycle checkpoint inhibitors, an asymmetric synthesis of the
antimitotic natural product, ustiloxin D, has been completed. A salen-Al-catalyzed aldol reaction
was employed to construct a chiral oxazoline 9 (99% yield, 98% ee) that served the dual purpose of
installing the necessary 1,2-amino alcohol functionality as well as providing an efficient synthon
for the requisite methylamino group at C9. The chiral aryl-alkyl ether was assembled using a
Pd-catalyzed asymmetric allylic alkylation that notably delivered a product with stereochemistry
opposite to that predicted by precedent. The linear tetrapeptide was subsequently cyclized to produce
ustiloxin D. The mechanistic origin of the allylic alkylation selectivity was further investigated,
and a working hypothesis for the origin of the observed stereoselectivity has been proposed.
CHART 1. Ustiloxins and Phomopsins
Introduction
Iwasaki and co-workers isolated1 ustiloxins A-D from
the water extract of the fungus Ustilaginoidea virens
growing on rice (Chart 1). This metabolite was reported
to poison domestic animals that consumed the infected
rice plant. In 1992, the structure of ustiloxin A was
determined using a combination of X-ray crystallography,
1H and 13C NMR, and degradation studies. The structures
of ustiloxins B-D were determined by comparison of
their spectral properties and amino acid analyses in
relation to ustiloxin A. Several years before, Edgar
isolated phomopsins A and B as the principle mycotoxin
from cultures of Phomopsis leptostromiformis.2 Of sig-
nificant economic impact to Australia and New Zealand,
sheep, cattle, and horses that consumed lupin infected
with this fungus developed severe liver damage. The
constituent amino acids of phomopsin A were identified
1
using degradation studies followed by extensive H and
13C NMR.3 X-ray crystallographic studies confirmed the
relative configuration of its eight stereocenters in 1986.4
Though the effects of these molecules on livestock are
certainly important, we were more drawn to their mode
of biological activity. Liver specimens of animals with
lupinosis revealed metaphase-arrested cells. Confirming
† Department of Chemistry.
‡ Department of Molecular Pharmacology.
their role as an antimitotic agent, light-scattering experi-
ments have demonstrated that the ustiloxins and pho-
mopsins are potent microtubule depolymerizers both in
vitro and in vivo.5,6 Competitive binding has shown that
this class of molecule binds to the same site as structur-
(1) (a) Koiso, Y.; Natori, S.; Iwasaki, S.; Sato, S.; Sonoda, R.; Fujita,
Y.; Yaegashi, H.; Sato, Z. Tetrahedron Lett. 1992, 33, 4157-4160. (b)
Koiso, Y.; Li, Y.; Iwasaki, S. J. Antibiot. 1994, 47, 765-773. (c) Koiso,
Y.; Morisaki, N.; Yamashita, Y.; Mitsui, Y.; Shirai, R.; Hashimoto, Y.;
Iwasaki, S. J. Antibiot. 1998, 51, 418-422.
(2) Culvenor, C. C. J.; Beck, A. B.; Clarke, M.; Cockrum, P. A.; Edjar,
J. A.; Frahn, J. L.; Jago, M. W.; Lanigan, G. W.; Pane, A. L.; Peterson,
J. E.; Smith, L. W.; White, R. R. Aust. J. Biol. Sci. 1977, 30, 269-277.
(3) Culvenor, C. C. J.; Edgar, J. A.; MacKay, M. F. Tetrahedron 1989,
45, 2351-2372.
(4) MacKay, M. F.; Van Donkelaar, A.; Culvenor, C. C. J. J. Chem.
Soc., Chem. Commun. 1986, 1219-1221.
(5) Luduen˜a, R. F.; Roach, M. C.; Prasad, V.; Banerjee, J.; Koiso,
Y.; Li, Y.; Iwasaki, S. Biochem. Pharmacol. 1994, 47, 1593-1599.
(6) To¨nsing, E. M.; Steyn, P. S.; Osborn, M.; Weber, K. Eur. J. Cell.
Biol. 1984, 35, 156-164.
10.1021/jo048854f CCC: $27.50 © 2004 American Chemical Society
Published on Web 11/11/2004
8810
J. Org. Chem. 2004, 69, 8810-8820

Total Synthesis of Ustiloxin D
ally unrelated rhizoxin and dolastatin 10,7 overlapping
somewhat with the vinca alkaloid binding site. As part
of the larger goal of understanding the mammalian
spindle assembly checkpoint, we are pursuing the syn-
theses of these natural products and structurally related
variants with the goal of dissecting the cell signaling
pathway by which the integrity of microtubules is
monitored.
Structure-activity relationships in microtubule depo-
lymerization assays indicate that the complex phomopsin
A tripeptide side chain is not essential since ustiloxin D
retains most of its antimitotic activity with a simple
glycine side chain (IC50 ) 2.4 vs 6.6 µM).8 Furthermore,
reduction and protection of the glycine acid does not even
change activity by an order of magnitude.8 Hydrogenation
of phomopsin A to octahydrophomopsin A9 had little effect
as well. Other structural differences include the opposite
stereochemistry at the C10 hydroxyl and chlorination of
the hydroxytyrosine arene. These elements are most
likely unnecessary features for microtubule depolymer-
ization. However, the 13-membered macrocyclic motif,
the hydroxytyrosine N-methylation, and the (R)-config-
uration of the C2 ether are preserved throughout each
family member.
FIGURE 1. Ustiloxin D retrosynthesis.
to add to and open this intermediate. Conditions that
weakly activated the acid resulted in no reaction. Though
we had anticipated these issues at the outset of this
strategy, given the myriad peptide coupling agents in
general use, it still appeared to be a tractable problem.
In particular, this failed reaction colored our retrosyn-
thetic analysis of ustiloxin D.
This family of molecules has attracted the interest of
many groups. Semisyntheses of analogues,8,10 pared down
ustiloxin model compounds,11 and the synthesis of the
chiral sulfoxide C12 substituent12 have been reported. Of
particular note, the Joullie´ group completed the first total
synthesis of ustiloxin D13 using a strategy by which the
C2 ether was assembled using nucleophilic aromatic
substitution and the C9/C10 amino alcohol was con-
structed by Sharpless’ asymmetric aminohydroxylation.
Our group has also recently completed an enantioselec-
tive synthesis.14
In pursuing a total synthesis of phomopsin B,15 we
synthesized the complete macrocycle and tripeptide side
chain with the goal of coupling these advanced interme-
diates in a highly convergent manner (Chart 1). Unfor-
tunately, the combination of the weak dehydroproline
nucleophile with the â-branched carbonyl and acylated
amine combined to block this coupling entirely. Condi-
tions that strongly activated the acid resulted in azlac-
tone formation and subsequent epimerization and race-
mization at C3. Furthermore, the side chain was not able
Results and Discussion
Retrosynthesis. During the development of our syn-
thetic strategy for ustiloxin D, our goal was to use
chemistry that would be compatible for application on
the more highly functionalized phomopsin system. Thus,
it was evident from our studies that the macrocycle
needed to be fully formed with the side chain already in
place as shown in Figure 1. We envisioned that such
closure could be brought about by a novel intramolecular
Staudinger ligation16 or a more conventional peptide
coupling strategy. This reduced the problem to linear
tetrapeptide 2 that could be formed from the dipeptide
precursor 3 by sequential coupling of the glycine side
chain and a valine-derived amino acid.
The problem of the tertiary aryl-alkyl ether had
previously been addressed using nucleophilic aromatic
substitution of a homochiral tertiary alcohol onto a
suitably activated arene moiety.13,15d This strategy placed
stringent demands on the electronics of the arene group,
requiring considerable functional group manipulation to
elaborate the natural product. To get around this prob-
lem, we took advantage of the high functional group
tolerance of the Pd asymmetric allylic alkylation.17 New
advances in this methodology led us to believe that this
disconnection might be particularly well-suited to deliver
the desired aryl-alkyl ether from phenol 4.
(7) Lin, Y.; Koiso, Y.; Kobayashi, H.; Hashimoto, Y.; Iwasaki, S.
Biochem. Pharmacol. 1995, 49, 1367-1372.
(8) Morisaki, N.; Mitsui, Y.; Yamashita, Y.; Koiso, Y.; Shirai, R.;
Hashimoto, Y.; Iwasaki, S. J. Antibiot. 1998, 51, 423-427.
(9) Lacey, E.; Edgar, J. A.; Culvenor, C. C. J. Biochem. Pharm. 1987,
36, 2133-2138.
(10) (a) Mutoh, R.; Shirai, R.; Koiso, Y.; Iwasaki, S. Heterocycles
1995, 41, 9-12. (b) Takahashi, M.; Shirai, R.; Koiso, Y.; Iwasaki, S.
Heterocycles 1998, 47, 163-166.
This phenol could in turn be realized by conducting
an aldol reaction between a glycine enolate equivalent
and a suitably protected 3,4-dihydroxybenzaldehyde 5.
Though most work toward these stereocenters has fo-
cused on enantioselective elaboration of an alkene
precursor,13,15c we chose to use an aldol approach for two
(11) La¨ıb, T.; Zhu, J. Synlett 2000, 1363-1365.
(12) Hutton, C. A.; White, J. M. Tetrahedron Lett. 1997, 38, 1643-
1646.
(13) (a) Park, H.; Cao, B.; Jouillie´, M. M. J. Org. Chem. 2001, 66,
7223-7226. (b) Cao, B.; Park, H.; Joullie´, M. M. J. Am. Chem. Soc.
2002, 124, 520-521.
(14) Tanaka, H.; Sawayama, A. M.; Wandless, T. J. J. Am. Chem.
Soc. 2003, 125, 6864-6865.
(15) (a) Stohlmeyer, M. M.; Tanaka, H.; Wandless, T. J. J. Am.
Chem. Soc. 1999, 121, 6100-6101. (b) Woiwode, T. F.; Rose, C.;
Wandless, T. J. J. Org. Chem. 1998, 63, 9594-9596. (c) Woiwode, T.
F.; Wandless, T. J. J. Org. Chem. 1999, 64, 7670-7674. (d) Woiwode,
T. F. Ph.D. Thesis, Stanford University, Stanford, California, 2000;
Chapters 2 and 3.
(16) (a) Nilsson, B. L.; Kiessling, L. L.; Raines, R. T. Org. Lett, 2001,
3, 9-12. (b) Nilsson, B. L.; Kiessling, L. L.; Raines, R. T. Org. Lett.
2000, 2, 1939-1941. (c) Soellner, M. B.; Nilsson, B. L.; Raines, R. T.
J. Org. Chem. 2002, 67, 4993-4996.
(17) Trost, B. M.; Toste, F. D. J. Am. Chem. Soc. 1999, 121, 4545-
4554.
J. Org. Chem, Vol. 69, No. 25, 2004 8811

Sawayama et al.
TABLE 1. Methylation and Reduction of Oxazoline 10
SCHEME 1
reaction conditions
NaBH(OAc)3/EtOH
NaBH(OAc)3/CH2Cl2
NaBH4/EtOH
NaBH4/CH2Cl2
10 (%) 11 (%) 14 (%) 15 (%)
7
4
0
0
6
8
33a
6
42
46
62
55
51
9
29
0
0
4
49
13
21
24
0
NaBH4/CH2Cl2-satd NaHCO3
K-selectride/THF
a This number is a combined yield of 11 and 14.
circumvented by increasing the catalyst loading from 5
to 10 mol % and rigorously removing all traces of water
to deliver 99% yield and 98% ee of cis-oxazoline 9. DBU-
catalyzed isomerization set the correct stereochemistry
for ustiloxin D to provide 10 in 87% yield and 7% recovery
of starting material as a thermodynamic ratio.
reasons. First, our aldol would produce a cis-oxazoline
product, more appropriate for the syn relationship of the
phomopsin system. However, facile base-mediated isomer-
ization is capable of producing the trans-oxazoline, the
desired stereochemistry for the anti configuration found
in the ustiloxins. Thus, this transformation would be
compatible for both molecular scaffolds. Second, the
product oxazoline serves as a masked N-methylated
amino alcohol that could be revealed at the appropriate
time.
Salen-Al Aldol. Our studies toward ustiloxin D were
initiated with the differential protection of the hydroxyl
groups of commercially available 3,4-dihydroxybenz-
aldehyde (Scheme 1). The only conditions that gave
selectivity were to employ 2.5 equiv of strong base to
deprotonate both phenols followed by reaction with 1.2
equiv of acetic anhydride. The acetyl group was allowed
to equilibrate between the two potential phenolates
eventually acylating the more basic C3 hydroxyl of 5. For
the 4-position, we needed a more robust protecting group
that would persist until a final global deprotection. Silyl
ethers were inappropriate because upon deprotection of
the acetate, concomitant migration of a TBS group
between the C3 and C4 hydroxyls was observed. Other
efforts involving stable ethereal linkages amenable to
fluoride deprotection (such as SEM18) were later found
to be incompatible with metal-mediated reactions. The
ethylene glycol substructure likely chelates to the metal
center, attentuating its activity. Though its deprotection
conditions would not be useful in the context of the
phomopsin B synthesis, benzyl protection was finally
settled upon, and after optimization, a 95% yield of 6 was
obtained.
In addition to oxygen and nitrogen installation, the
oxazoline served a dual purpose as a vehicle to function-
alize the N27-nitrogen to an N-methylamine. Using the
combination of methyl trifluoromethanesulfonate fol-
lowed by in situ reduction produced the N-methyloxazo-
lidine as an inseparable mixture of diastereomers.20
Though the methylation step was well precedented,21
many conditions for the reduction were screened to
optimize the yield and are summarized in Table 1. In
addition to the desired N-methylated oxazolidine, we also
observed formation of an overly reduced N,N-dialkyl-
amino alcohol 14 and methylated hydrolysis product 15.
To minimize side products, sodium borohydride was used
as part of a biphasic mixture of saturated sodium
bicarbonate solution and dichloromethane. It is likely
that the transient N-methyloxazolinium ion and sodium
borohydride preferentially dissolve in the aqueous layer
and undergo reaction. The resulting uncharged product
migrates to the organic layer away from the hydride-rich
environment, preventing over-reduction. Control of the
relative rates of reduction vs hydrolysis was used to
minimize production of the benzamide 15. This was
accomplished by using faster reducing agents (NaBH4 vs
NaBH(OAc)3) in more activating solvents (water vs
ethanol). Mild oxalic acid conditions were sufficient to
open the aminal and deliver the desired N-methylamino
alcohol 12 in 80% yield.
Pd π-Allyl Asymmetric Allylic Alkylation. The next
major synthetic hurdle was the asymmetric allylic alky-
lation to form the chiral ether. Work done by the Trost
group had established that Pd π-allyl species undergo
nucleophilic attack at the secondary carbon in preference
to the primary carbon to deliver chiral tertiary centers
over secondary ones.17 This selectivity presumably arises
This set the stage for the first asymmetric reaction,
namely, Evans’ Salen-Al aldol19 between benzaldehyde
6 and the glycine-enolate equivalent 8 (Scheme 1).
Although there was no precedent for this reaction using
an electron-rich aldehyde, the sluggish reactivity was
(20) (a) Gant, T. G.; Meyers, A. I. Tetrahedron 1994, 50, 2297-2360.
(b) Wilson, S, R.; Mao, D. T.; Khatri, H. N. Synth. Commun. 1980, 10,
17-23. (c) Nordin, I. C. J. Heterocycl. Chem. 1966, 3, 531-532. (d)
Barmer, B. A.; Meyers, A. I. J. Am. Chem. Soc. 1984, 106, 1865-1866.
(21) Schmidt, U.; Siegel, W. Tetrahedron Lett. 1987, 28, 2849-2852.
(18) SEM: 2-(trimethylsilyl)ethoxymethyl.
(19) Evans, D. A.; Janey, J. M.; Magomedov, N.; Tedrow, J. S.
Angew. Chem., Int. Ed. 2001, 40, 1884-1888.
8812 J. Org. Chem., Vol. 69, No. 25, 2004

Total Synthesis of Ustiloxin D
TABLE 2. Comparison of Yield and Selectivity of
Asymmetric Allylic Alkylation Using Tertiary Carbonate
16 over 1 h and Primary Carbonate 17 over 10 h
16 delivered unsatisfactory asymmetric induction (10-
20% ee) whereas 17 gave ethers in ∼80% ee.22
With this preliminary data in hand, we tried to take
advantage of the improved enantioselectivity of the
primary carbonate system and improve conversion. We
had reason to believe that the poor yields stemmed from
slow Pd-mediated carbonate ionization to produce the
active π-allyl cation. In the event that 16 was used, an
excess of the electrophile was required (5 equiv) to ensure
that the coupled ether did not compete with the 16 as a
π-allyl precursor. Over the course of the reaction, all 5
equiv of the tertiary carbonate was consumed, with the
generated π-allyl cation predominantly decomposing to
a diene via â-hydride elimination. However, even when
a single equivalent of 17 was reacted using 20 mol % Pd
catalyst loading, no more than 50% consumption was ever
observed. For many systems, olefin coordination of the
π-allyl precursor to the metal center can be rate-limit-
ing.23 In such instances, reducing the sterics and improv-
ing the electronics can alleviate slow turnover. Though
the trisubstituted nature of 17 could not be changed, its
electronics could be adjusted to favor precoordination.
When a ligand such as an olefin forms π bonds with a
metal center, the ligand’s HOMO (π orbital) is thought
to donate its electrons to the metal center’s LUMO (d
orbitals). If ionization ensues, however, a significant
portion of precoordination is dedicated to back-bonding.
In this case, the metal HOMO (d orbital) interacts with
the allylic carbonate LUMO (π* orbital). To improve this
latter HOMO-LUMO match, electron-withdrawing groups
were added to the π-allyl precursor,24 as presented in
Table 3.
To arrive at the required phenol 4, the free amine and
the hydroxyl group of 12 were sequentially protected and
the acetate group was removed using potassium tert-
butoxide in methanol/THF to deliver the free phenol 4
in 62% over three steps. To enhance conversion, catalyst
loading was increased to 10 mol % Pd2(dba)3‚CHCl3 and
50 mol % (R,R) ligand 19 with primary carbonates
reacted for 10 h and tertiary carbonate 16 reacted for 1
h. Leaving groups known to enhance ionization relative
to the methyl carbonate such as trichloroethyl (Troc) or
trifluoromethyl carbonate and phosphates were reacted
with a lack of success similar to the primary methyl
carbonate. On the outside chance that slow nucleophilic
attack was inhibiting reaction progress of the primary
carbonate, we constructed carbonate 34, envisioning that
Pd-mediated ionization would release a proximal pheno-
late better situated to attack and improve conversion.
Unfortunately, we did not observe any change in yield.
Taken together, we concluded that effective coordination
of the trisubstituted olefin was blocked by steric factors
that were impossible to overcome.
from the greater carbocation-like character of the more
highly substituted carbon center following ionization by
the Pd(0) metal. However, we were unsure if we would
be able to make the extension to form quaternary centers.
Such a transformation specifies alkylation at a tertiary
carbon over a primary one via a Pd π-allyl species such
as 18 (Table 2). Model studies were used to evaluate the
viability of this transformation as part of the ustiloxin
D synthetic plan. As 18 can be accessed from ionization
of either methyl carbonates 16 or 17, we sought to
understand if either of these two precursors provided an
advantage over the other.
The results from Table 2 demonstrated that the desired
regiochemical outcome was realized every time using all
manner of π-allyl precursors, Pd(0) sources, and ligands.
More interesting was the divergent conversion and
enantioselection observed using each of the two π-allyl
precursors. In the presence of 1 mol % Pd2(dba)3‚CHCl3
and 5 mol % (R,R) ligand 19 in dichloromethane, phenol
26 and tertiary carbonate 16 reacted to give 94% of the
desired coupled product when the reaction was stopped
immediately after complete phenol consumption (∼1 h).
By contrast, the primary carbonate 17 underwent a much
slower reaction, proceeding to only 44% conversion over
10 h. At this point, if 16 were added to this reaction
vessel, complete and near-instantaneous conversion to
the desired ether was observed. Stereoselectivity was also
affected, as reactions making use of tertiary carbonate
We then turned our attention back to the tertiary
carbonate system, seeking to increase the enantioselec-
tivity. Lower temperatures appeared to stop the catalyst,
as a reaction at -78 °C reduced the yield to 15% of
desired ether 30. Other strategies aimed at increasing
(22) As measured by analytical HPLC using a Chiralpak AD chiral
column.
(23) Trost, B. M.; Crawley, M. L. Chem. Rev. 2003, 103, 2921-2943
and references contained therein.
(24) Trost, B. M.; Toste, F. D. J. Am. Chem. Soc. 2000, 122, 11262-
11263.
J. Org. Chem, Vol. 69, No. 25, 2004 8813

Sawayama et al.
TABLE 3. Asymmetric Allylic Alkylation Using Various
minor compound that was identified as benzamide 39
that was isolated in 18%. We postulate that once the
methyl ester is hydrolyzed it spontaneously decarboxy-
lates to deliver 39. A similar reaction profile was observed
using LiOOH and aqueous K2CO3. The unprecedented
sensitivity to aqueous base of this oxazoline would have
important implications in later transformations (vide
infra). Smaller nucleophiles such as LiAlH4 also reacted
poorly with 10, delivering the alcohol in 26% yield.
Around this time, the Joullie´ group published their
synthesis of ustiloxin D.13b Over the course of this work,
they successfully hydrolyzed a C8 ethyl ester. In contrast
to our system, which had an R-Cbz-protected secondary
amine, the Joullie´ group had before them an R-Cbz-
protected primary amine. In agreement with earlier
work,26 we concluded that LiOH first deprotonates the
free N-H of the carbamate. This prevents epimerization
and elimination via abstraction of the R-proton because
the requisite vicinal dianion is too unstable to form. The
excess base can only saponify the ethyl ester. The
decomposition that was observed in N-methyl carbamate
30 can be attributed to the greater acidity of its R-proton,
as there is no N-H proton to first abstract.
π-Allyl Precursors with Phenol 4
These hydrolysis studies gave us three key points to
consider: (1) The sterics of a Cbz-protected secondary
amine are not compatible with hydrolysis of the C8 ester.
(2) An abstractable proton is required on the N27 amine
to prevent epimerization at C9. (3) The oxazoline is
unstable to aqueous base. Taken together, the best time
to saponify the methyl ester would be immediately after
the oxazoline was opened to the N-methylamino alcohol
and before it would be protected as a benzyl carbamate.
We returned to cis-oxazoline 9 and discovered that in
the presence of anhydrous K2CO3 we could effect both
isomerization and deacetylation in one step to produce
trans-phenol 40 in 95% yield (Scheme 3). Etherification
was accomplished using 5 mol % Pd2(dba)3‚CHCl3, 25 mol
% (R,R) ligand 19, and 3 equiv of tertiary carbonate 16
to afford product 41 in 81% yield and 27% de with 15%
recovered starting material. Though we had evidence
that the primary carbonate 17 produced a more enantio-
selective reaction in model systems (vide supra), we
observed <10% conversion with 17. We did not confirm
the absolute stereochemistry at this juncture and relied
upon the predictive model described by the Trost group
to assign the (2R)-configuration to ether 41a.17 However,
upon completion of the synthesis and comparison with
natural ustiloxin D (vide infra) it became evident that
the palladium coupling had proceeded with opposite
stereoinduction to predominantly give (2S)-ether 41b.
Asymmetric Sharpless Dihydroxylation. Olefin 41
contains structural motifs that had each been studied
individually by the Sharpless group as substrates for
asymmetric dihydroxylation.27 These studies detailed
improved selectivity using the (DHQD)2PYR and
(DHQD)2PHAL ligands,28 so we pursued dihydroxylations
the rate of π-σ-π interconversion17 (vide infra) through
the use of additives such as TBAF, N(Bu)4Cl, N(Bu)4I,
or acetic acid were ineffective in raising enantioselection
in model systems (data not shown). Last, a screen of
various solvents proved to be fruitless, delivering only
lower yields and enantioselectivity across a range of
model systems. Unfortunately, we were unable to deter-
mine the reaction’s enantioselectivity using substrate 4,
as the resulting diastereomers were inseparable by chiral
HPLC.
We next sought to hydrolyze the C8 ester (Scheme 2).
When Cbz-protected secondary amine 30 was subjected
to a mild excess of LiOH, free acid 35 was isolated in
only 13% yield. Most disturbing to our synthetic plan was
the observation of epimerized N-methylamine 36 and
eliminated cinnamate 37 in 19 and 14% yields, respec-
tively. Lithium hydroperoxide, LiBr/DBU, KOtBu/DMSO,
KOtBu/H2O/THF, NaI/pyridine, and TMSI/CH3CN gave
similar myriad products or no reaction at all. In an
attempt to relieve some of the steric issues associated
with the N-methylated carbamate,25 we next considered
preserving the trans-oxazoline until after methyl ester
hydrolysis. As an evaluation of this route, we exposed
methyl ester 10 to LiOH. To our surprise, we only
isolated the deacetylated phenol 40 in 54% yield and a
(26) McDermott, J. R.; Benoiton, N. L. Can. J. Chem. 1973, 51,
2555-2561.
(27) Kolb, H. C.; VanNieuwenhze, M. S.; Sharpless, K. B. Chem.
Rev. 1994, 94, 2483-2547.
(28) (a) Vanhessche, K. P. M.; Sharpless, K. B. J. Org. Chem. 1996,
61, 7978-7979. (b) Wang, Z.-M.; Zhang, X.-L.; Sharpless, K. B.
Tetrahedron Lett. 1993, 34, 2267-2270. (c) Crispino, G. A.; Jeong, K.-
S.; Kolb, H. C.; Wang, Z.-M.; Xu, D.; Sharpless, K. B. J. Org. Chem.
1993, 58, 3785-3786.
(25) Similar difficulties with hydrolysis of an ester with an R-N-
methylated carbamate were observed by: Boger, D. L.; Patane, M. A.;
Zhou, J. J. Am. Chem. Soc. 1994, 116, 8544-8556.
8814 J. Org. Chem., Vol. 69, No. 25, 2004

Total Synthesis of Ustiloxin D
SCHEME 2
SCHEME 3
Since we do not know which set of diastereomers elute
first by HPLC, two interpretations are possible. If pseu-
doenantiomers (2R,3S)-42a and (2S,3R)-42c elute first
and pseudoenantiomers (2R,3R)-42b and (2S,3S)-42d
elute second (case 1), the following is true:
xy + (1 - x)(1 - z)
x(1 - y) + (1 - x)z
peak area 1
peak area 2
)
case 1
xy: mole fraction of (2R,3S)-42a
(1 - x)(1 - z): mole fraction of (2S,3R)-42c
x(1 - y): mole fraction of (2R,3R)-42b
(1 - x)z: mole fraction of (2S,3S)-42d
(1)
On the other hand, if (2R,3S)-42a and (2S,3R)-42c are
the latter eluting isomers, the equation is as follows:
x(1 - y) + (1 - x)z
xy + (1 - x)(1 - z)
peak area 1
peak area 2
)
case 2 (2)
using both ligands. Additionally, to prevent oxazoline
decarboxylation we employed a more weakly basic aque-
ous solution using 2 equiv of NaHCO3/1 equiv of K2CO3
rather than the typically prescribed 3 equiv of K2CO3.
This modification slowed hydrolysis of the intermediate
osmate ester such that increased catalyst loading and
extended reaction times at room temperature were
required to raise conversion to ∼75%.
We needed to solve the equation for the two unknowns
y and z, y being the selectivity of the correct ustiloxin D
(2R,3S)-42a isomer from compound (2R)-41a (minor
product) and z being the selectivity of the C2-epimer of
ustiloxin D (major product). Thus, two data sets are
needed for the determination. The first set of peak areas
was obtained from the analysis of products of dihydroxy-
lation using racemic substrate 41 that was generated by
using a racemic ligand for the π-allyl palladium reaction.
In this case, x ) (1 - x) ) 0.5. The second set of peak
areas was obtained from a dihydroxylation reaction using
substrate 41 that was generated stereoselectively by
using ligand (R,R)-19. In this case, chiral HPLC estab-
lished that x ) 0.365 and (1 - x) ) 0.635. The necessary
data were collected, and y and z values were calculated
from the above equations for both case 1 and case 2 and
compiled in Table 4.
In case 1, all selectivities are greater than 0.5, implying
that the DHQD ligands preferentially form the desired
(3S)-42 compounds, an outcome predicted by the Sharp-
less mnemonic. In case 2, all selectivities are less than
0.5, implying that the reaction is running counter to
expectation. On the basis of the reliability of this mne-
monic and subsequent confirmation by total synthesis,
we believe that our reaction falls into case 1, in which
The dihydroxylation reaction produced a complex
mixture of four diastereomers (two from Pd etherification
and two from dihydroxylation), so evaluation of the
reaction diastereoselectivity was very difficult, particu-
larly because 42a/42c and 42b/42d behaved chromato-
graphically like enantiomers, each pair eluting as a single
peak by reverse-phase C18 analytical HPLC. Because the
dihydroxylation selectivity is a function of both the
catalyst and the substrate, the selectivities for (2R)-41a
and (2S)-41b are different (matched and mismatched
substrates). These details are more cogently illustrated
in Figure 2. The selectivity for the chiral π-allyl pal-
ladium reaction is expressed as the mole fraction of (2R)-
41a, x. The mole fraction of (2S)-41b can be expressed
as (1 - x). In a similar manner, the mole fraction of
(2R,3S)-42a produced from (2R)-41a is expressed as y,
(mole fraction of (2R,3R)-42b is (1 - y)) and the mole
fraction of (2S,3S)-42d produced from (2S)-41b is ex-
pressed as z (ratio of (2S,3R)-42c is (1 - z)).
J. Org. Chem, Vol. 69, No. 25, 2004 8815

Sawayama et al.
SCHEME 4
FIGURE 2. Determination of diastereoselectivies of the
dihydroxylation reaction.
TABLE 4. Selectivities of Dihydroxylation Reaction
Using (DHQD)2PHAL and (DHQD)2PYR
(DHQD)2PHAL
(DHQD)2PYR
area 1 area 2
area 1
area 2
x ) 0.365
x ) 0.5
case 1
54.28
57.25
45.72
42.75
37.08
46.81
62.92
53.19
y (de)
z (de)
y (de)
z (de)
0.5375 (7.5%)
0.8285 (66%)
0.6825 (37%)
0.4625 (-7.5%)
0.3175 (-37%)
0.8923 (78%)
0.1715 (-66%)
0.1077 (-78%)
case 2
for 24 h gave the azide in 56% yield over two steps. It
should be noted that displacements using NaN3 or LiN3
in other solvents were significantly less effective. Because
the potential existed for oxirane-assisted double inversion
at C3, we attempted to effect diastereomer separation
by analytical HPLC. Though this work proved to be
inconclusive, given the stability of trifluoromethane-
sulfonate 45 to SiO2 chromatography and favorable
comparison with natural ustiloxin D upon completion of
the synthesis, we believe that assisted double displace-
ment does not occur to any significant extent.
Next, the oxazoline protecting group was opened to
reveal the N-methylamino alcohol. Methyl triflate fol-
lowed by sodium borohydride provided the intermediate
diastereomeric oxazolidines in 83% yield. This was fol-
lowed by mild acid treatment to unmask the free amine.
As expected, since amine 47 was not protected as a
carbamate, careful reaction with 3 equiv of LiOH at 0
°C selectively hydrolyzed the methyl ester. As soon as
the methyl ester had been fully consumed, 3 equiv of Cbz-
Cl was added to convert the intermediate N-methylamine
to Cbz-protected N-methylamino acid 48 in 71% yield.
Our mechanistic hypothesis as to the kinetics of ester
hydrolysis was borne out as neither epimerization at C9
nor elimination was observed. Furthermore, the glycine
tert-butyl ester was untouched by these relatively gentle
saponification conditions.
pseudoenantiomers (2R,3S)-42a and (2S,3R)-42c elute
first and pseudoenantiomers (2R,3R)-42b and (2S,3S)-
42d elute second. (DHQD)2PYR is therefore the more
selective ligand. Its dihydroxylation on the ustiloxin D
substrate 41a is 66% de, and the selectivity for the C2-
epimer 41b is 78% de, indicating that we were working
with a mismatched pair.
Macrocycle Closure. The secondary alcohol resulting
from dihydroxylation next needed to be converted to an
azide (Scheme 4). Earlier intermediates demonstrated
that first protecting the primary alcohol followed by an
activation/displacement protocol was unsuccessful. It is
likely that the doubly allylic oxygenation renders the
electrophilic center of interest too electron-rich for effec-
tive SN2-displacement (data not shown). To this end, we
chose to oxidize the primary alcohol and couple the
glycine side chain prior to the azide displacement. A two-
step oxidation using TEMPO29/NaClO30 followed by so-
dium chlorite gave the acid in excellent yield (99%), and
BOP31 coupling of the glycine tert-butyl ester proceeded
rapidly to produce amide 44 in 88% yield (Scheme 4).
Despite the strong azide nucleophilicity, preliminary
studies had indicated that the neopentyl center mandated
a highly active trifluoromethanesulfonate to furnish any
substituted product at all (data not shown). Optimized
conditions reacted 3 equiv of triflic anhydride with 44.
Though conversion proceeded slowly over 10 h, exposure
of the purified triflate 45 to anhydrous LiN332 in DMPU33
At this point, we began studies toward closing the
macrocycle. The diverse functionality in 48 demanded
mild conditions, and so we turned to the Staudinger
ligation, a reaction that uses triphenylphosphine to
(29) TEMPO: 2,2,6,6-tetramethyl-1-piperidinyloxy, free radical.
(30) de Nooy, A. E. J.; Besemer, A. C.; van Bekkum, H. Synthesis
1996, 1153-1174.
(31) BOP: benzotriazol-1-yloxy-tris(dimethylamino)-phosphonium
hexafluorophosphate.
(32) Thompson, C. F.; Jamison, T. F.; Jacobsen, E. N. J. Am. Chem.
Soc. 2001, 123, 9974-9983.
(33) DMPU: 1,3-dimethyl-3,4,5,6-tetrahydro-2(H)-pyrimidinone.
8816 J. Org. Chem., Vol. 69, No. 25, 2004

Total Synthesis of Ustiloxin D
SCHEME 5
SCHEME 6
couple an N-terminal azide with a thioester to deliver
an amide bond.16,34 Though we had not found any
examples taking place in an intramolecular context to
produce macrolactams, we felt that this ligation could be
advantageous. The necessary reagent for an intra-
molecular variant of the Staudinger ligation would
require a phosphine and a thiol linked via a methylene
unit. Such a reagent could mediate ring contraction from
a 16-membered thio-lactone to a more strained 13-
membered lactam, in the process taking advantage of
preorganization to introduce ring strain in a stepwise
fashion. Second, since we introduce the C3 nitrogen as
an azide, its direct conversion to an amide would save
steps.
Azide 46 was reduced in 76% yield using PMe3 followed
by quantitative BOP coupling of azido valine39 53 (Scheme
6). The methylation/reduction sequence followed by treat-
ment with oxalic acid produced N-methylamino alcohol
(68%). Elaboration to the azido acid 56 using LiOH and
then CbzCl proceeded as before in 70% yield. Though
Staudinger conditions were once again unable to close
the macrocyle, 56 could be reduced to the amino acid.
When reacted using PyAOP at high dilution in DMF, 57
cyclized in good yield to the desired macrolactam 58. A
strong solvent dependence was observed with no macro-
cycle obtained using dichloromethane.
Earlier results had demonstrated that sufficiently
acidic conditions were capable of eliminating the benzylic
hydroxyl across C9-C10 (data not shown). We were
therefore reluctant to deprotect the tert-butyl ester in this
way. However, our struggles to hydrolyze the hydroxy-
tyrosine methyl ester led us to believe that basic hy-
drolysis of the glycinyl ester would be feasible. This
prediction was realized when hydrogenolysis of benzyl
and Cbz groups were first conducted in 80% yield. Next,
10 equiv of LiOH at room temperature was sufficient to
saponify the tert-butyl ester and give the final product
59 in 85% yield.
The valine thiobenzyl ester was successfully coupled
using BOP reagent in 62% yield (Scheme 5). Compound
49 was added slowly via syringe pump to 20 equiv of
i
linked diphenylphosphinomethane thiol and Pr2NEt at
<1 mM in THF. It was envisioned that the thiol of the
linked compound would first undergo thioester exchange
followed by iminophosphorane formation. Ring contrac-
tion would then give the desired macrocycle. Unfortu-
nately, other than recovered starting material, only a
trace quantity of amine 51 was observed. This indicates
that phosphine-mediated azide reduction was occurring
faster than thioester exchange. This was not unexpected
since it is well-documented that branching at the C-
terminus slows thioester exchange considerably in native
chemical ligations.35 To increase the rate of thioester
exchange, we tried adding (1) benzenethiol,36 a nucleo-
philic acylation catalyst, and (2) a phosphine-thiolate
linker, generated in situ from tBuOK/50. However,
neither of these strategies delivered the desired lactam.
Furthermore, standard peptide coupling using PyAOP,
BOP, HATU,37 or BEP38 on the amino acid derived from
49 did not yield any macrocycle either. Since â-branching
of the hydroxyisoleucine may have been incompatible
with the steric demands of the macrocycle transition
state, we moved our disconnection to the N7-C8 junction.
Though 59 matched the expected mass by MS, its
analytical HPLC retention time differed significantly
from a natural sample of ustiloxin D (Figure S1, see
Supporting Information). Deprotected 59 eluted as a 10:1
mixture with only the minor constituent coeluting with
natural ustiloxin D at 14.15 min. When we compared 1H
NMR in D2O, not only did the spectra not overlay but
(34) Saxon, E.; Bertozzi, C. R. Science 2000, 287, 2007-2010. (b)
Saxon, E.; Armstrong, J. I.; Bertozzi, C. R. Org. Lett. 2000, 2, 2141-
2143.
(35) Hackeng, T. M.; Griffin, J. H.; Dawson, P. E. Proc. Natl. Acad.
Sci. U.S.A. 1999, 96, 10068-10073.
(36) Dawson, P. E.; Churchill, M. J.; Ghadiri, M. R.; Kent, S. B. H.
J. Am. Chem. Soc. 1997, 119, 4325-4329.
(37) HATU: O-(7-azabenzotriazol-1-yl)-N,N,N′,N′-tetramethyluro-
nium hexafluorophosphate.
(39) Lundquist, IV, J. T.; Pelletier, J. C. Org. Lett. 2001, 3, 781-
783.
(38) BEP: 1-ethyl-2-bromopyrdinium tetrafluoroborate.
J. Org. Chem, Vol. 69, No. 25, 2004 8817

Sawayama et al.
SCHEME 7
FIGURE 3. Crystal structure of ustiloxin A.
several key peaks varied by >0.1 ppm.40 The comple-
mentary upfield shifts of the C16 aromatic and C21
methyl protons and downfield shifts of the C22 and C23
ethyl protons prompted us to examine the X-ray crystal
structure of ustiloxin A (Figure 3). Ustiloxin A differs
from ustiloxin D by virtue of a chiral sulfoxide appended
at C12 (Chart 1). Its crystal structure indicated that there
was only 3.4 Å separating the C16 aromatic proton and
the C21 methyl group.1a If an inversion of C2 stereo-
chemistry had occurred, this result could be verified by
NOE studies. Saturating the C16 proton resonance of
natural ustiloxin D led to the observation of an NOE at
the C21 methyl but not at the ethyl C22 or C23. When
the C16 proton resonance of 59 was saturated, the
complementary NOE was observed at the C22 and C23
ethyl group but not at the C21 methyl group. Since our
synthetic strategy did not provide any opportunity to
epimerize at C2, we concluded that the Pd asymmetric
allylic alkylation had proceeded counter to the stereo-
chemical outcome predicted by the model. Such an
outcome is not without precedent.41
SCHEME 8
Synthesis of Ustiloxin D. We returned to phenol 40
(Scheme 7), and using 2 mol % Pd2dba3‚CHCl3 and 6 mol
% enantiomeric (S,S) ligand 60 afforded a 2:1 mixture of
41, with the desired isomer (R)-41a making up the major
product in 71% yield. Dihydroxylation followed by two-
step oxidation of the primary alcohol and BOP coupling
of the glycine tert-butyl ester proceeded in 66% yield over
four steps. Our first synthesis of epi-ustiloxin D 59 had
seen the 2:1 selectivity at C2 changed to 10:1 by the end
of the synthesis. Thus, it appeared as though the dia-
stereomeric epi-C2 stereochemistry was more compatible
with ensuing reactions than was the correct C2 isomer.
As such, we were especially concerned that lower conver-
sions would be observed in the ensuing activation and
displacement of the secondary alcohol using a mixture
containing the correct C2 isomer as the major constitu-
ent. Much to our gratification, no such reduction in yield
occurred. The azide 63 was successfully reduced and
coupled to Boc-Val in 81% yield over two steps (Scheme
8). Because we would not be conducting a Staudinger
ligation to effect macrocyclization, we did not couple azido
valine and instead used a more conventional nitrogen
protection. The N-methylamino alcohol was successfully
unmasked, and the deprotection of both the methyl ester
and the Boc-amine proceeded in moderate yield. PyAOP
macrocyclization proceeded at reduced yield as compared
with the analogous C2-epi compound (56 vs 69%). Two-
step deprotection completed the synthesis and produced
a 2:1 mixture of ustiloxin D 1 and C2-epi ustiloxin D 59.42
(40) As detailed in Supporting Information for ref 14.
(41) (a) Trost, B. M.; Heinemann, C.; Ariza, X.; Weigand, S. J. Am.
Chem. Soc. 1999, 121, 8667-8668. (b) Trost, B. M.; Gunzner, J. J. Am.
Chem. Soc. 2001, 123, 9449-9450. (c) Trost, B. M.; Gunzner, J. L.;
Dirat, O.; Rhee, Y. H. 2002, 124, 10396-10415.
(42) We have attributed the discrepancy in ust D 1:C2-epi ust D 80
using (R,R) 81 and (S,S) 24 ligands to reaction diastereoselectivity for
each of the two isomers.
8818 J. Org. Chem., Vol. 69, No. 25, 2004

Total Synthesis of Ustiloxin D
FIGURE 4. Proposed mechannism for the asymmetric allylic alkylation.
Preparative HPLC separated the isomers, and the major
constituent proved to be identical to a sample of the
natural product by all spectroscopic analyses, including
1H and 13C NMR, HRMS, and specific rotation, as well
as by TLC and HPLC analysis. The synthesis was
completed with a longest linear sequence of 20 steps and
an overall yield of 1.9%.
tion, it would imply that the alkylation is the enantio-
discriminating step. Were the ustiloxin D system to fall
into this kinetic regime, the same enantioselectivity
would be observed regardless of the origin of the Pd
π-allyl, be it (S)-16, (R)-16, 17, or mixtures thereof.
However, using model systems, we observed racemic 16
and 17 to give very different enantioselectivities. This
suggests that the ustiloxin D system does not fall into
this kinetic regime and alkylates according to an alter-
nate mechanism.
At the other extreme, it is possible that ionization of
the π-allyl by Pd is the enantiodiscriminating step. Under
this kinetic regime, alkylation of the phenol occurs much
faster than π-σ-π interconversion of the π-allyl complex.
Thus, the configuration of the π-allyl precursor at ioniza-
tion is the determining factor of the coupled product’s
stereochemistry. In such a case, the identity of the ligand
bound to Pd would be irrelevant, as the mechanism is
essentially proceeding via double SN2 displacement.
Thus, when a racemic mixture of 16 is ionized, a racemic
mixture of π-allyl is formed and a racemic product should
be observed. By contrast, if 17 is ionized to form a single
homochiral π-allyl species, complete enantioselectivity
should be the result. Though not in perfect accord, this
mechanistic model provides a more accurate description
of the observed data, such that ionization of racemic 16
delivers a product of low selectivity (∼20% ee) and
ionization of 17 delivers high selectivity (∼80% ee).
Mechanistic Description of Asymmetric Allylic
Alkylation. An interesting mechanistic question to
address is the difference in absolute configuration and
the degree of selectivity of the Pd asymmetric allylic
alkylation arising from the use of different π-allyl precur-
sors. The chiral space that the (R,R) ligand defines about
Pd can be represented by 68 (Figure 4). Two of the four
phenyl groups make up walls, and the other two make
up flaps and are arranged with C2 symmetry. When
π-allyl precursors such as (S)-16/(R)-16 or 17 are ionized
by Pd(0) and the (R,R) ligand 19, two diastereomeric Pd
π-allyl species are formed, anti-69 and syn-70. These
intermediates can then equilibrate via π-σ-π intercon-
version and alkylate phenol 26 under Curtin-Hammett
conditions.17 Typically, the anti-π-allyl is disfavored
because this orientation places A1,3 strain on the larger
substituent.43 This conformation also places the larger
group closer to the ligand wall and the smaller group
under the flap. In the case of ustiloxin D, however, the
catalytic complex must discriminate between sterically
similar methyl and ethyl groups. For this system, it
happens that the anti conformation is the more reactive
of the two, albeit not by much (2:1 selectivity corresponds
to ∆∆Gq ) 0.4 kcal/mol).
Because conversion was ineffective with the trisubsti-
tuted π-allyl precursor 17 (vide supra), studies were
restricted to the tertiary carbonate 16. However, this
hypothesis suggests that the only element preventing
good selectivity using the tertiary carbonate is its racemic
nature. If a single enantiomer (S)-16 could be prepared,
its subsequent Pd-mediated ionization would be expected
If the rate of π-σ-π interconversion of the π-allyl
complex is significantly faster than the ensuing alkyla-
(43) Trost, B. M.; Lee, C. B. J. Am. Chem. Soc. 2001, 123, 3671-
3686.
J. Org. Chem, Vol. 69, No. 25, 2004 8819

Sawayama et al.
to produce a single enantiomeric ether. We are currently
pursuing this strategy in the context of the synthesis of
phomopsin B.
using a homochiral tertiary carbonate. Our early incor-
poration of the ustiloxin D side chain should be applicable
to the synthesis of phomopsin B and facilitate coupling
of the weakly nucleophilic dehydroproline moiety. Finally,
though we were unsuccessful in our attempts using the
Staudinger ligation to effect macrolactamization of ustilox-
in D, we are still pursuing this strategy in the context of
phomopsin B.
Conclusion
The total synthesis of (-)-ustiloxin D was accomplished
in 20 steps and 1.9% overall yield starting from 3,4-
dihydroxybenzaldehyde. We are applying these studies
to the more complex phomopsin B system. Three catalytic
asymmetric reactions were used to set four of the five
stereocenters. The versatility of the aluminum aldol is
highlighted by its ability to accommodate the opposite
cis-hydroxyl stereochemistries of ustiloxin D and pho-
mopsin B. For ustiloxin D, the aldol product cis-oxazoline
is isomerized to the trans-oxazoline to complete the
synthesis. For our studies toward phomopsin B, the cis-
oxazoline will be preserved. The Pd asymmetric allylic
alkylation proceeded with low selectivity and good yield
using a racemic tertiary π-allyl precursor and proceeded
with high selectivity and poor yield using a primary
carbonate. Improvement of this transformation is ongoing
Acknowledgment. We thank Dr. Yukiko Koiso for
providing a sample of natural ustiloxin D. We also thank
Professors Evans, Trost, and DuBois and members of
their laboratories for helpful advice. This work was
supported by the NSF (CHE-9985214) and a Dreyfus
Foundation Teacher-Scholar Award to T.J.W. and a
NSERC PGS B Award to A.M.S.
Supporting Information Available: Experimental de-
tails and analytical data for all new compounds. This material
is available free of charge via the Internet at http://pubs.acs.org.
JO048854F
8820 J. Org. Chem., Vol. 69, No. 25, 2004