Total synthesis of the antimitotic bicyclic peptide celogentin C

Article abstract of DOI:10.1021/ja909870g

An account of the total synthesis of celogentin C is presented. A right-to-left synthetic approach to this bicyclic octapeptide was unsuccessful due to an inability to elaborate derivatives of the right-hand ring. In the course of these efforts, it was discovered that the mild Braslau modification of the McFadyen-Stevens reaction offers a useful method of reducing recalcitrant esters to aldehydes. A leftto-right synthetic strategy was then examined. The unusual Leu-Trp side-chain cross-link present in the left-hand macrocycle was fashioned via a three-step sequence comprised of an intermolecular Knoevenagel condensation, a radical conjugate addition, and a SmI2-mediated nitro reduction. A subsequent macrolactamization provided the desired ring system. The high yield and concise nature of the left-hand ring synthesis offset the modest diastereoselectivity of the radical conjugate addition. Formation of the Trp-His sidechain linkage characteristic of the right-hand ring was then accomplished by means of an indole-imidazole oxidative coupling. Notably, Pro-OBn was required as an additive in this reaction. Detailed mechanistic investigations indicated that Pro-OBn moderates the concentration of NCS in the reaction mixture, thereby minimizing the production of an undesired dichlorinated byproduct. The natural product was obtained after macrolactamization and deprotection. The chemical shifts of the imidazole hydrogen atoms exhibited significant dependence on temperature, concentration, and pH. Antitumor screening indicated that celogentin C inhibits the growth of some cancer cell lines.

Full text of DOI:10.1021/ja909870g

Published on Web 12/28/2009
Total Synthesis of the Antimitotic Bicyclic Peptide
Celogentin C
Bing Ma, Biplab Banerjee, Dmitry N. Litvinov, Liwen He, and Steven L. Castle*
Department of Chemistry and Biochemistry, Brigham Young UniVersity, ProVo, Utah 84602
Received November 21, 2009; E-mail: scastle@chem.byu.edu
Abstract: An account of the total synthesis of celogentin C is presented. A right-to-left synthetic approach
to this bicyclic octapeptide was unsuccessful due to an inability to elaborate derivatives of the right-hand
ring. In the course of these efforts, it was discovered that the mild Braslau modification of the
McFadyen-Stevens reaction offers a useful method of reducing recalcitrant esters to aldehydes. A left-
to-right synthetic strategy was then examined. The unusual Leu-Trp side-chain cross-link present in the
left-hand macrocycle was fashioned via a three-step sequence comprised of an intermolecular Knoevenagel
condensation, a radical conjugate addition, and a SmI2-mediated nitro reduction. A subsequent macrolac-
tamization provided the desired ring system. The high yield and concise nature of the left-hand ring synthesis
offset the modest diastereoselectivity of the radical conjugate addition. Formation of the Trp-His side-
chain linkage characteristic of the right-hand ring was then accomplished by means of an indole-imidazole
oxidative coupling. Notably, Pro-OBn was required as an additive in this reaction. Detailed mechanistic
investigations indicated that Pro-OBn moderates the concentration of NCS in the reaction mixture, thereby
minimizing the production of an undesired dichlorinated byproduct. The natural product was obtained after
macrolactamization and deprotection. The chemical shifts of the imidazole hydrogen atoms exhibited
significant dependence on temperature, concentration, and pH. Antitumor screening indicated that celogentin
C inhibits the growth of some cancer cell lines.
Introduction
family. The structural assignment of moroidin (5) has been
confirmed by X-ray crystallography.4
In 2001, Kobayashi and co-workers isolated the bicyclic
octapeptide celogentin C (1, Figure 1) from the seeds of the
flowering plant Celosia argentea.1 In addition to 1, the related
natural products celogentin A (2), celogentin B (3), and moroidin
(5) were also discovered during this investigation. Peptide 5
had been previously isolated by Williams and co-workers in
1986 from the leaves of the bush Laportea moroides, which is
native to the Australian rainforest.2 A re-examination of the
seeds of Celosia argentea by Kobayashi and co-workers yielded
six additional bicyclic members of the celogentin family:
celogentin D (4), celogentins E-H (6-9), and celogentin J
(10).3 The celogentin family can be organized into three
structural classes according to variations in the right-hand
macrocycle structure. Thus, the right-hand ring of 1 is a
tetrapeptide that contains a proline, whereas the right-hand rings
of 2-4 are tripeptides that lack this residue. Peptides 5-10
incorporate glycine into their right-hand macrocycles, resulting
in tetrapeptides that presumably adopt a different conformation
than the proline-containing right-hand ring of 1. Two unusual
cross-links, one connecting the leucine ꢀ-carbon with the indole
C-6 of tryptophan and the other joining indole C-2 with the
imidazole N-1 of histidine, are common to the entire celogentin
The seeds of Celosia argentea have been employed as a
traditional Chinese and Japanese herbal medicine for the
treatment of liver and eye diseases.5 An investigation of the
antimitotic activity of the celogentins by the Kobayashi group
revealed that the ability of peptides 1-10 to inhibit tubulin
polymerization was strongly dependent on the structure of the
right-hand ring.1,3,6 Celogentin C (1), the only family member
to possess a proline residue in its right-hand ring, exhibits the
greatest potency in this assay (IC50 0.8 µM), which exceeds
that of the anticancer agent vinblastine (IC50 3.0 µM). In
contrast, the moroidin-type compounds (5-10) are roughly
equipotent with vinblastine (IC50 2.0-4.0 µM), while celogentins
A, B, and D (2-4) are significantly less active (IC50 20-30
µM). Despite the potential of 1 and 5-10 as antitumor agents,
to the best of our knowledge, their toxicity to cancer cells has
not yet been described.
The combination of striking molecular architecture and
significant biological activity has prompted several groups to
address the synthesis of the celogentins. The laboratories of
(4) Suzuki, H.; Morita, H.; Shiro, M.; Kobayashi, J. Tetrahedron 2004,
60, 2489.
(1) Kobayashi, J.; Suzuki, H.; Shimbo, K.; Takeya, K.; Morita, H. J. Org.
Chem. 2001, 66, 6626.
(5) (a) Hase, K.; Kadota, S.; Basnet, P.; Takahashi, T.; Namba, T. Biol.
Pharm. Bull. 1996, 19, 567. (b) Hayakawa, Y.; Fujii, H.; Hase, K.;
Ohnishi, Y.; Sakukawa, R.; Kadota, S.; Namba, T.; Saiki, I. Biol.
Pharm. Bull. 1998, 21, 1154.
(2) (a) Leung, T.-W. C.; Williams, D. H.; Barna, J. C. J.; Foti, S.; Oelrichs,
P. B. Tetrahedron 1986, 42, 3333. (b) Kahn, S. D.; Booth, P. M.;
Waltho, J. P.; Williams, D. H. J. Org. Chem. 1989, 54, 1901.
(3) Suzuki, H.; Morita, H.; Iwasaki, S.; Kobayashi, J. Tetrahedron 2003,
59, 5307.
(6) Morita, H.; Shimbo, K.; Shigemori, H.; Kobayashi, J. Bioorg. Med.
Chem. Lett. 2000, 10, 469.
9
10.1021/ja909870g  2010 American Chemical Society
J. AM. CHEM. SOC. 2010, 132, 1159–1171 1159

A R T I C L E S
Ma et al.
role in the construction of the right-hand ring, and an unusual
additive (Pro-OBn) was critical to the success of this process.
We have conducted further experiments that led us to modify
our initially proposed mechanistic rationale14 for this intriguing
transformation.
Results and Discussion
At the outset of the project, we envisioned developing an
intramolecular Knoevenagel condensation-radical conjugate
addition sequence for installing the key Leu-Trp linkage present
in the left-hand ring of 1 (Scheme 1). Although intramolecular
Knoevenagel condensations are rare, isolated examples exist.15
Accordingly, it was our hope that R-nitroacetamido aldehyde
11 would cyclize to afford R,ꢀ-unsaturated R-nitroamide 12.
Examination of molecular models of 12 indicated that a twist
in the indole ring relative to the plane defined by the double
bond and its substituents would result in shielding of the bottom
face of the alkene by the indole C-7 hydrogen. Therefore, the
conjugate addition of an isopropyl radical16 to 12 should occur
from the top face of the olefin as required. We then planned to
take advantage of the acidic R-nitroamide stereocenter in adduct
13. Simple molecular mechanics minimization of 13 and its
epimer at the nitro-bearing carbon demonstrated that 13 was
ca. 3 kcal/mol lower in energy. Consequently, if epi-13 were
generated in the hydrogen atom abstraction step of the radical
reaction, then treatment of the adduct with a weak base should
form 13 via selective epimerization of the acidic R-nitro
stereocenter.17 Finally, nitro reduction and subsequent coupling
to pyroglutamic acid would deliver 14, the left-hand ring of
celogentin C. Model studies with simple ꢀ-aryl-substituted R,ꢀ-
unsaturated R-nitroamide substrates demonstrated the feasibility
of the radical conjugate addition-nitro reduction sequence.18
Unfortunately, despite numerous attempts, we were unable to
construct an intramolecular Knoevenagel condensation substrate
corresponding to 11. Challenges inherent to installing the nitro
and aldehyde functional groups in the same molecule ultimately
thwarted our efforts.19
Our failure to prepare the left-hand ring of 1 via an
intramolecular Knoevenagel condensation necessitated formula-
tion of a second-generation plan, which is presented in Scheme
2. Prior studies by our group established the viability of an
indole-imidazole oxidative coupling/macrolactamization se-
quence as a means of constructing the right-hand ring of 1.13
These results prompted us to pursue a “right-to-left” synthetic
strategy for the total synthesis. Thus, oxidative coupling of
modified Trp-Pro dipeptide 15 with Arg-His dipeptide 16
would afford heterobiaryl-linked tetrapeptide 17. Deprotection
and macrolactamization would then provide 18, a suitably
functionalized version of the celogentin C right-hand ring. Next,
reduction of the methyl ester would deliver aldehyde 19,
substrate for an intermolecular Knoevenagel condensation.
Adduct 20 would be converted into 1 via a sequence of
transformations including radical conjugate addition, nitro
Figure 1. Celogentins A-H, celogentin J, and moroidin.
Hutton,7 Wandless,8 and Campagne9 have each reported syn-
thetic studies in this area. Notably, the efforts of Moody and
co-workers10 culminated in the construction of stephanotic acid
methyl ester, a derivative of a monocyclic tetrapeptide that is
very similar to the left-hand ring of the celogentins11 (vide infra).
Our group targeted celogentin C for total synthesis, and in 2003
we disclosed the construction of a functionalized tryptophan
derivative suitable for elaboration into 1.12 Later, we prepared
the right-hand ring of 1,13 and recently we achieved the total
synthesis of celogentin C.14 This accomplishment provides the
first synthetic entry into the celogentin family of bicyclic
peptides. Herein, we provide a detailed account of this venture,
including the evolution of our strategy and the discovery of
methodology which should have further applications. An
indole-imidazole oxidative coupling reaction played a central
(7) Yuen, A. K. L.; Jolliffe, K. A.; Hutton, C. A. Aust. J. Chem. 2006,
59, 819.
(8) Grimley, J. S.; Wandless, T. J. Abstracts of Papers, 232nd ACS
National Meeting; San Francisco, CA, Sept. 10-14, 2006; ORGN-
735.
(15) (a) Zhang, Y.; Wada, T.; Sasabe, H. Tetrahedron Lett. 1996, 37, 5909.
(b) Chre´tien, F.; Khaldi, M.; Chapleur, Y. Tetrahedron Lett. 1997,
38, 5977.
(9) Michaux, J.; Retailleau, P.; Campagne, J.-M. Synlett 2008, 1532.
(10) (a) Bentley, D. J.; Slawin, A. M. Z.; Moody, C. J. Org. Lett. 2006, 8,
1975. (b) Bentley, D. J.; Moody, C. J. Org. Biomol. Chem. 2004, 2,
3545. (c) Harrison, J. R.; Moody, C. J. Tetrahedron Lett. 2003, 44,
5189. (d) Comber, M. F.; Moody, C. J. Synthesis 1992, 731.
(11) Yoshikawa, K.; Tao, S.; Arihara, S. J. Nat. Prod. 2000, 63, 540.
(12) Castle, S. L.; Srikanth, G. S. C. Org. Lett. 2003, 5, 3611.
(13) He, L.; Yang, L.; Castle, S. L. Org. Lett. 2006, 8, 1165.
(14) Ma, B.; Litvinov, D. M.; He, L.; Banerjee, B.; Castle, S. L. Angew.
Chem., Int. Ed. 2009, 48, 6104.
(16) For a review of radical conjugate additions, see: Srikanth, G. S. C.;
Castle, S. L. Tetrahedron 2005, 61, 10377.
(17) Although we are unaware of any pKa measurements of R-nitroaceta-
mides, the pKa of methyl nitroacetate in H2O is 5.58. Bernasconi, C. F.;
Pe´rez-Lorenzo, M.; Brown, S. D. J. Org. Chem. 2007, 72, 4416.
(18) Srikanth, G. S. C.; Castle, S. L. Org. Lett. 2004, 6, 449.
(19) Ma, B.; Litvinov, D. N.; Srikanth, G. S. C.; Castle, S. L. Synthesis
2006, 3291.
9
1160 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Antimitotic Bicyclic Peptide Celogentin C
A R T I C L E S
Scheme 1. Intramolecular Knoevenagel-Radical Conjugate Addition Approach to Left-Hand Ring
Scheme 2. Right-to-Left Synthetic Strategy for the Synthesis of 1a
a Pbf ) 2,2,4,6,7-pentamethyldihydrobenzofuran-5-sulfonyl.
reduction, and macrolactamization. The most critical reaction
in the endgame was likely to be the radical conjugate addition.
In the second-generation route, the electron-deficient radical
acceptor would be located in an acyclic portion of the substrate
(cf., 20), in contrast to our prior approach that imbedded the
radical acceptor within a macrocycle (cf., 12, Scheme 1). This
change would presumably reduce the prospects of achieving
satisfactory levels of substrate-directed stereocontrol in the
reaction. Accordingly, we planned to employ a chiral Lewis
acid in the radical conjugate addition.20
Another concern with the new plan involved the reduction
of methyl ester 18 to aldehyde 19. The electron-rich nature of
the indole ring attached to the methyl ester could attenuate its
reactivity toward reducing agents. In fact, the paucity of similar
reductions in the literature21 suggested that conversion of 18
into 19 may not be straightforward. Accordingly, we resolved
to test the indolyl methyl ester reduction on a simpler substrate
than macrocycle 18.
The synthesis commenced with the preparation of C-6
carbomethoxy-substituted tryptophan 25 as outlined in Scheme
3. Previously, screening of various chiral phase-transfer catalysts
in the alkylation of glycinate Schiff base 21 with triethylsilyl
propargyl bromide demonstrated that the trifluorobenzyl-
substituted hydrocinchonidine derivative 22 of Park and Jew22
gave the best results for production of propargylglycine 23.12
Further optimization of this reaction in conjunction with scale-
up efforts resulted in the excellent yield (91%) and ee (94%)
shown in Scheme 3. The alkylation conditions, including the
toluene-chloroform mixed solvent system, are based on those
reported by Park and Jew.22 The relatively labile benzophenone
imine moiety of 23 was then exchanged for a more robust Cbz
group, affording carbamate 24 in 96% yield. Earlier, we had
performed the Larock-type heteroannulation23 of 24 with a
(20) For a review of Lewis acid promoted radical reactions, see: Renaud,
P.; Gerster, M. Angew. Chem., Int. Ed. 1998, 37, 2562.
(22) Jew, S.-s.; Yoo, M.-S.; Jeong, B.-S.; Park, I.-Y.; Park, H.-g. Org. Lett.
2002, 4, 4245.
(21) (a) Moody, C. J.; Shah, P. J. Chem. Soc., Perkin Trans. 1 1989, 2463.
(b) Kim, M.; Vedejs, E. J. Org. Chem. 2004, 69, 7262.
(23) Larock, R. C.; Yum, E. K. J. Am. Chem. Soc. 1991, 113, 6689.
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1161

A R T I C L E S
Ma et al.
Scheme 3. Synthesis of C-6 Carbomethoxy Tryptophan 25
Figure 2. McFadyen-Stevens reduction.
Scheme 4. Modified McFadyen-Stevens Reduction of 25
Me3SnOH26 unexpectedly caused cleavage of the Cbz group.
Clearly, our fears regarding the recalcitrant nature of the C-6
methyl ester were realized.
Table 1. Attempted Reduction/Hydrolysis of 25
As we searched for alternative methods of transforming
methyl ester 25 into aldehyde 26, we were drawn to a report
by Braslau and co-workers of a mild variant of the McFadyen-
Stevens reaction.27 In the classical process, a tosyl-substituted
acyl hydrazine is converted to an aldehyde by treatment with
base at high temperature. Presumably, elimination of the tosyl
group to generate an acyl diazene is followed by loss of N2,
providing the aldehyde (Figure 2).28 In the Braslau modification,
the tosyl group on the acyl hydrazine is replaced with an
o-nitrobenzenesulfonyl (nosyl) moiety. This substitution allows
the elimination to proceed at a lower temperature (ca. 80 °C).
Because acyl hydrazines can be readily prepared from methyl
esters,29 we tested this protocol with methyl ester 25. We were
pleased to discover that exposure of 25 to a 1:2 mixture of
hydrazine hydrate and MeOH at 70 °C provided acyl hydrazine
29 in 90% yield (Scheme 4). If the reaction was conducted in
neat hydrazine hydrate, then the corresponding diacyl hydrazine
derived from tert-butyl ester cleavage was obtained. Nosylation
of 29 was best accomplished in a 1:1 THF-pyridine mixture,
and the elimination-fragmentation of nosyl-substituted acyl
hydrazine 30 proceeded readily at 80 °C in EtOH, delivering
aldehyde 26 in 84% yield. Importantly, chiral HPLC analysis
of 26 demonstrated that no epimerization of the amino acid
R-stereocenter occurred during the three-step sequence.
With a means of accessing the indolyl C-6 aldehyde, we
proceeded to construct the fully functionalized celogentin C
right-hand ring as detailed in Scheme 5. Our previously
developed method for synthesizing the model right-hand ring13
was generally applicable, although some modification was
necessary. First, C-6 carbomethoxy-substituted tryptophan 25
was converted into dipeptide 15 via a three-step process
reagent
result
DIBAL-H
LiAlH4
LiOH
tert-butyl ester reduction
tert-butyl ester reduction
complex mixture
Me3SnOH
Cbz cleavage
hydroxymethyl-substituted o-iodoaniline coupling partner by
utilizing a slight modification of the conditions developed by
Cook and co-workers (Pd(OAc)2, LiCl, Na2CO3).12,24 The
synthesis of C-6 carbomethoxy-substituted tryptophan 25 re-
quired the more electron-deficient 2-iodo-5-(methoxycarbonyl)-
aniline25 as coupling partner. We found that substitution of PdCl2
for Pd(OAc)2 and inclusion of 4 Å MS in the reaction mixture
provided the best yields of 25, which was obtained as a single
detectable regioisomer.
Tryptophan 25 was employed as a model system for
investigating the reduction of methyl ester 18 to aldehyde 19,
and the results of this study are summarized in Table 1.
Surprisingly, attempts to reduce 25 with DIBAL-H or LiAlH4
resulted in exclusive tert-butyl ester reduction, leaving the
methyl ester intact. We then examined selective hydrolysis of
the methyl ester, in hopes of either reducing the carboxylic acid
or converting the acid into a reducible functional group (i.e.,
acid chloride, thioester, or selenoester). Unfortunately, exposure
of 25 to LiOH generated a complex mixture from which the
desired acid 28 could not be identified. Moreover, the use of
(26) (a) Nicolaou, K. C.; Estrada, A. A.; Zak, M.; Lee, S. H.; Safina, B. S.
Angew. Chem., Int. Ed. 2005, 44, 1378. (b) Nicolaou, K. C.; Zak, M.;
Safina, B. S.; Estrada, A. A.; Lee, S. H.; Nevalainen, M. J. Am. Chem.
Soc. 2005, 127, 11176.
(27) Braslau, R.; Anderson, M. O.; Rivera, F.; Jimenez, A.; Haddad, T.;
Axon, J. R. Tetrahedron 2002, 58, 5513.
(24) Ma, C.; Liu, X.; Li, X.; Flippen-Anderson, J.; Yu, S.; Cook, J. M. J.
Org. Chem. 2001, 66, 4525.
(28) (a) McFadyen, J. S.; Stevens, T. S. J. Chem. Soc. 1936, 584. (b) Matin,
S. B.; Craig, J. C.; Chan, R. P. K. J. Org. Chem. 1974, 39, 2285.
(29) (a) Rapoport, H.; Holden, K. G. J. Am. Chem. Soc. 1960, 82, 5510.
(c) Boger, D. L.; Patel, M. J. Org. Chem. 1988, 53, 1405.
(25) (a) Collini, M. D.; Ellingboe, J. W. Tetrahedron Lett. 1997, 38, 7963.
(b) Gabriele, B.; Salerno, G.; Veltri, L.; Costa, M.; Massera, C. Eur.
J. Org. Chem. 2001, 4607.
9
1162 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Antimitotic Bicyclic Peptide Celogentin C
A R T I C L E S
Scheme 5. Synthesis of Right-Hand Ring Aldehyde 37
consisting of simultaneous N-Cbz, tert-butyl ester, and trieth-
ylsilyl deprotection (HBr in HOAc), phthalimide installation
(methyl 2-((succinimidooxy)carbonyl)benzoate),30 and peptide
coupling (Pro-OBn, HOBt, EDCI ·HCl). Oxidative coupling31
of 15 with Arg-His dipeptide 1613 then proceeded readily,
affording indole-imidazole-linked tetrapeptide 17 in 92% yield.
This represents a significant improvement over the analogous
oxidative coupling used to construct the model right-hand ring
(58% yield, 25-30% recovery of each SM).13 In the present
case, optimization of the reaction conditions established that a
slight excess (1.2 equiv) of NCS allowed for complete con-
sumption of the reactants without degrading the product. Our
prior studies established the clear superiority of NCS over NBS,
NIS, and t-BuOCl,13 so alternative oxidants were not examined.
Surprisingly, attempts to cleave the N-Cbz and benzyl ester
moieties from 17 enlisting transfer hydrogenation conditions
(10% Pd/C, HCO2NH4) used with the model system13 failed to
remove the carbamate group. Switching the H2 source to 1,4-
cyclohexadiene yielded the same result, as did employing H2
at both atmospheric and elevated (70 atm) pressures. Fortunately,
PdCl2 in conjunction with triethylsilane32 effected simultaneous
N-Cbz and benzyl ester deprotection, delivering macrolactam-
ization substrate 31 in 76% yield. Cyclization of 31 to afford
32, the C-6 carbomethoxy-substituted celogentin C right-hand
ring, then proceeded smoothly in the presence of HBTU and
HOBt. In preparation for acyl hydrazine installation, the
phthalimide moiety, which prevented the tryptophan N-terminus
from engaging the chloroindolenine intermediate of the oxidative
coupling reaction, was exchanged for a Cbz group. Transforma-
tion of the resulting carbamate 34 into aldehyde 37 was
accomplished via the McFadyen-Stevens-based strategy; how-
ever, some modifications were necessary. Specifically, the
hydrazinolysis step required a 1:1 mixture of hydrazine hydrate
and MeOH instead of the 1:2 mixture used previously with ester
25. Moreover, the reaction mixture was stirred at room tem-
perature and concentrated under high vacuum at 0 °C to prevent
diacyl hydrazine formation. Nosylation of the hydrazine was
best accomplished in a 2:1 THF-pyridine solvent mixture, and
the Braslau-modified McFadyen-Stevens reaction of nosylate
36 provided aldehyde 37 in good (78%) yield.
With the right-hand ring aldehyde in hand, we now required
a suitable nitroacetamide derivative for the key intermolecular
Knoevenagel condensation (see 19 f 20, Scheme 2). Coupling
(30) Casimir, J. R.; Guichard, G.; Briand, J.-P. J. Org. Chem. 2002, 67,
3764.
(31) (a) Bergman, J.; Engqvist, R.; Stålhandske, C.; Wallberg, H. Tetra-
hedron 2003, 59, 1033. (b) Engqvist, R.; Bergman, J. Tetrahedron
2003, 59, 9649. (c) Booker-Milburn, K. I.; Fedouloff, M.; Paknoham,
S. J.; Strachan, J. B.; Melville, J. L.; Voyle, M. Tetrahedron Lett.
2000, 41, 4657.
(32) (a) Birkofer, L.; Bierstein, E.; Ritter, F. Chem. Ber. 1961, 94, 821.
(b) Sakaitani, M.; Ohfune, Y. J. Org. Chem. 1990, 55, 870. (c)
Coleman, R. S.; Shah, J. A. Synthesis 1999, 1399.
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1163

A R T I C L E S
Ma et al.
Scheme 6. Synthesis of Nitroacetamide 42
be interfering with the Knoevenagel condensation. Accordingly,
we resolved to perform this reaction at an earlier stage of the
synthesis. This decision spawned a left-to-right synthetic plan,
which is summarized in Scheme 8. We reasoned that coupling
of nitroacetamide 42 with the relatively simple C-6 formyl
tryptophan 26 would proceed to afford R,ꢀ-unsaturated R-ni-
troamide 44. A chiral Lewis acid-promoted radical conjugate
addition would then provide acyclic peptide 45 after nitro
reduction and coupling to pyroglutamic acid. Macrolactamiza-
tion followed by attachment of the proline residue would
generate cyclic peptide 46, substrate for the indole-imidazole
oxidative coupling reaction with dipeptide 16. This transforma-
tion was valuable in the right-hand ring synthesis (see Scheme
5);13 however, the steric hindrance and conformational con-
straints associated with imbedding the indole moiety in a
macrocycle caused us to view this particular case with some
trepidation. Nonetheless, if we were able to prepare octapeptide
47, our earlier results (cf., Scheme 5) gave us confidence that
macrolactamization between the Pro and Arg residues and
subsequent deprotection would deliver 1 in good yield.
Scheme 7. Failed Knoevenagel Condensation
We had prepared C-6 formyl tryptophan 26 in three steps
from the corresponding methyl ester 25 (Scheme 4), but we
sought a more direct route to this compound. As shown in
Scheme 9, it could be obtained in two steps from C-6
hydroxymethyl-substituted tryptophan 48, a species whose
synthesis was reported previously12 by a route analogous to that
shown for tryptophan 25 in Scheme 3. Removal of the TBS
group from 48 was most easily accomplished by employing an
AcOH:THF:H2O (3:2:2) mixture.40 DDQ-mediated oxidation of
the primary benzylic alcohol41 proceeded smoothly, providing
the desired C-6 formyl tryptophan 26 in 89% yield from 48.
The development of reliable conditions for the union of 26
and nitroacetyl-capped Leu-Val dipeptide 42 is summarized
in Table 2. No reaction between the coupling partners was
observed in the presence of MgBr2 ·OEt242 (entry 1). TiCl4 and
NMM in THF18,34 provided the desired adduct 44, but yields
were low and inconsistent (entry 2). Addition of drying agents
to the reaction mixture failed to improve the results (entry 3).
When the condensation was conducted in Et2O, a low yield of
the product was obtained (27%), and substantial quantities of
starting material were recovered due to poor solubility (entry
4). Although the yield was modest, the reaction was cleaner
than those performed in THF. On the basis of this observation,
we probed the efficacy of mixed solvent systems. Both 1:1 and
2:1 mixtures of THF and Et2O afforded 44 in good yield (entries
5 and 6), with the latter system giving a slightly better result.
Notably, 44 was produced as a single detectable alkene isomer.
The alkene stereochemistry of 44 is assigned on the basis of
the X-ray crystal structure of a simpler model compound that
was produced under similar conditions.43
of Boc-Leu and Val-OBn followed by Boc deprotection
afforded the dipeptide Leu-Val-OBn (39, Scheme 6). Expo-
sure of 39 to Rajappa’s commercially available dithioketene
acetal 4033 in the presence of TsOH then delivered vinyl sulfide
41 as a single isomer of undetermined alkene stereochemistry.
Treatment of 41 with HgCl2 in aqueous MeCN served to
hydrolyze the vinyl sulfide, revealing nitroacetamide 42.
Our first attempt at performing the Knoevenagel condensation
of aldehyde 37 and nitroacetamide 42 enlisted the combination
of TiCl4 and N-methylmorpholine (NMM) in THF,34 a system
that was effective in our synthesis of model radical conjugate
addition substrates.18 Unfortunately, aldehyde 37 was consumed,
but coupling product 43 was not detected (Scheme 7). Numerous
other conditions were screened (4 Å MS, CHCl3, room tem-
perature to reflux;35 PPh3, THF, reflux;36 piperidine, HOAc, 4
Å MS, dioxane, reflux;37 H2O, room temperature to reflux;38
KF-alumina, EtOH, room temperature to reflux39), but no
desired product was formed. This failure to link building blocks
37 and 42 necessitated a redesign of our synthetic strategy.
We hypothesized that the wealth of functional groups present
in aldehyde 37, including numerous Lewis basic sites, might
At this point, we examined the key radical conjugate addition.
Our prior studies with simple substrates established the necessity
of reducing the nitro group immediately after workup to prevent
epimerization of the sensitive R-stereocenter.18,43 To properly
evaluate the stereoselectivity of radical conjugate additions to
44, we required a nitro reduction protocol that was suitable for
(33) (a) Manjunatha, S. G.; Chittari, P.; Rajappa, S. HelV. Chim. Acta 1991,
74, 1071. (b) Manjunatha, S. G.; Reddy, K. V.; Rajappa, S. Tetrahe-
dron Lett. 1990, 31, 1327.
(34) Fornicola, R. S.; Oblinger, E.; Montgomery, J. J. Org. Chem. 1998,
63, 3528.
(35) Sui, Y.; Liu, L.; Zhao, J.-L.; Wang, D.; Chen, Y.-J. Tetrahedron Lett.
2007, 48, 3779.
(40) Paquette, L. A.; Collado, I.; Purdie, M. J. Am. Chem. Soc. 1998, 120,
2553.
(36) Yadav, J. S.; Reddy, B. V. S.; Basak, A. K.; Visali, B.; Narsaiah,
A. V.; Nagaiah, K. Eur. J. Org. Chem. 2004, 546.
(37) Sinha, A. K.; Sharma, A.; Joshi, B. P. Tetrahedron 2007, 63, 960.
(38) Deb, M. L.; Bhuyan, P. J. Tetrahedron Lett. 2005, 46, 6453.
(39) Wang, X.-S.; Zeng, Z.-S.; Shi, D.-Q.; Wei, X.-Y.; Zong, Z.-M. Youji
Huaxue 2005, 25, 1138.
(41) Witiak, D. T.; Loper, J. T.; Ananthan, S.; Almerico, A. M.; Verhoef,
V. L.; Filppi, J. A. J. Med. Chem. 1989, 32, 1636.
(42) Abaee, M. S.; Mojtahedi, M. M.; Zahedi, M. M.; Sharifi, R.; Khavasi,
H. Synthesis 2007, 3339.
(43) He, L.; Srikanth, G. S. C.; Castle, S. L. J. Org. Chem. 2005, 70, 8140.
9
1164 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Antimitotic Bicyclic Peptide Celogentin C
A R T I C L E S
Scheme 8. Left-to-Right Synthetic Strategy
Scheme 9. Streamlined Preparation of C-6 Formyl Tryptophan 26
Table 3. Optimization of Nitro Reduction
Table 2. Knoevenagel Condensation of 26 and 42
entry
nitro reduction conditions
yield (%)
1
2
3
4
5
6
7
8
9
In, 6 N HCl or AcOH, THF-H2O
In, HCO2NH4, MeOH
In, NH4Cl, THF-H2O
Zn, HCO2NH4, MeOH
Zn, HCO2NH4, MeOH-CH2Cl2
Zn, HCO2NH4, THF-H2O
Raney Ni, H2 (1 atm)
NaBH4, NiCl2 · 6H2O, MeOH
SmI2, THF-MeOH 6:1
18
a
<10
a
<10
a
23
13
a
a
10
b
nd
a
<10
90
a The corresponding oxime was also obtained. b No product was
detected.
entry
conditions
yield (%)
a
1
2
3
4
5
6
MgBr2 ·OEt2, Et3N, THF, or CH2Cl2
TiCl4, NMM, THF
TiCl4, NMM, 4 Å MS or MgSO4, THF
TiCl4, NMM, Et2O
TiCl4, NMM, THF-Et2O 1:1
TiCl4, NMM, THF-Et2O 2:1
nr
15-50
15
27
45
b
(HCO2NH4 or NH4Cl) afforded only trace quantities of 51,
65
68
and the corresponding oxime was obtained instead (entries 2
and 3). This observation concerned us, because generation of
an oxime intermediate in the nitro reduction would render futile
any attempts at controlling the R-stereochemistry. Substitution
of Zn for In led to marginally higher yields of 51, but oxime
formation was still problematic (entries 4-6). Neither Raney
a No reaction was observed. b Starting material was recovered (67%
of 26).
a complex substrate. Thus, we reduced the electron-deficient
alkene of 44 with NaBH4 and used the resulting alkyl nitro
compound 50 to develop this reaction (Table 3). We began by
testing In/HCl,44 a reagent system that had proven valuable in
the earlier model studies.43 Unfortunately, both In/HCl and In/
AcOH provided amine 51 in low yield (entry 1), and numerous
byproducts were produced. The use of weaker acids
(44) (a) Lee, J. G.; Choi, K. I.; Koh, H. Y.; Kim, Y.; Kang, Y.; Cho, Y. S.
Synthesis 2001, 81. (b) Singh, V.; Kanojiya, S.; Batra, S. Tetrahedron
2006, 62, 10100.
(45) Gowda, D. C.; Mahesh, B.; Gowda, S. Indian J. Chem., Sect. B 2001,
40B, 75.
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1165

A R T I C L E S
Ma et al.
Table 4. Radical Conjugate Addition to 44
Lewis acid Zn(OTf)2. In this case, a 1:3:2:1 ratio of products
was obtained (entry 5). Interestingly, the reaction with no Lewis
acid present afforded roughly equimolar amounts of 52a-c, with
52d as a minor product (entry 6).
We were disappointed by the performance of the chiral Lewis
acids in the radical conjugate addition, yet intrigued by the result
obtained with Zn(OTf)2, because a major product (52b) was
produced. Repeating this experiment on a preparative scale
returned a very similar product ratio (1.0:2.9:2.0:1.2) in an
excellent 90% yield. Accordingly, the radical conjugate addition
and nitro reduction were each proceeding in g90% yield, and
52b could be obtained in 36% yield from 44. Thus, this process
would provide a viable means of synthesizing 1 as long as the
configuration of 52b matched that of the natural product.
Consequently, we resolved to determine the stereochemistry of
this intermediate.
entry
Lewis acid/ligand
equiv
a:b:c:da
1
2
3
4
5
6
Mg(NTf2)2/53
Mg(NTf2)2/54
Mg(NTf2)2/54
Mg(NTf2)2/55
Zn(OTf)2/none
none/none
2.0
2.0
5.0
2.0
2.0
1:2:1:1
1:2:2:1
2:2:2:1
3:2:2:1
1:3:2:1b
3:4:4:1
In 2006, Moody and co-workers reported the total synthesis
of stephanotic acid methyl ester10a (60, Scheme 10), a derivative
of the natural product stephanotic acid.11 Importantly, 60 and
the left-hand ring of 1 are extremely similar in structure; the
only difference between the two species is the presence of an
isoleucine residue in 60 at the site of the leucine residue in 1.
Because the configurations of 60 and the left-hand ring of 1
are identical, we hoped that a comparison of NMR data would
reveal whether or not the main product of the radical conjugate
addition possessed the requisite stereochemistry for elaboration
into 1. Accordingly, we synthesized the left-hand ring macro-
cycle from isomer 52b as detailed in Scheme 10. The mixture
of compounds 52a-d was partially separable; the two minor
isomers 52a (least polar) and 52d (most polar) could be removed
via chromatography. As a result, the first two steps were
performed on a 1.5:1 mixture of the two most abundant isomers
from the radical conjugate addition. Coupling of amines 52b,c
with pyroglutamic acid provided pentapeptides 45b,c in excel-
lent yield (96%), but the mixture still could not be separated.
However, after tandem Cbz/benzyl ester cleavage, the resulting
peptides 56b and 56c were separable on SiO2. Notably,
isomerically pure 56b was isolated in 31% overall yield from
Knoevenagel condensation product 44, as the excellent yields
of the four intervening steps (radical conjugate addition, nitro
reduction, peptide coupling, hydrogenolysis) compensated for
the low diastereoselectivity of the radical conjugate addition.
In their total synthesis of 60, Moody and co-workers
conducted a macrolactamization at the site corresponding to the
Leu-Val peptide bond in the left-hand ring of 1. This reaction
was plagued by epimerization;10a as a result, we decided to close
our macrocycle at a different site. We were relieved to discover
that cyclization of 56b via Val-Trp amide bond formation was
high-yielding and devoid of epimerization. Removal of the tert-
butyl ester and triethylsilyl groups of macrocycle 57b was then
achieved by enlisting B-bromocatecholborane (BCB).51 Finally,
treatment of the resulting acid 58b with SOCl2 in MeOH
delivered ester 59b. Gratifyingly, the 1H and 13C NMR spectra
of 59b matched quite well with the published spectra of 60.52
This encouraging development gave us confidence that isomer
52b, the major product of the radical conjugate addition, was
of the proper configuration for conversion into 1.
a Calculated from 1H NMR spectra of the mixtures of 52a-d.
b When run on a preparative scale (0.4 mmol), 90% of a 1.0:2.9:2.0:1.2
mixture of 52a-d was obtained.
Ni-H246 nor NaBH4-NiCl2 ·6H2O47 yielded promising results
(entries 7 and 8). At last, we discovered that employing SmI2
in THF-MeOH48 resulted in a facile, high-yielding reduction
(entry 9). Amine 51 was isolated as a mixture of diastereomers
due to the lack of selectivity in the 1,4-reduction of 44.
With the availability of a high-yielding nitro reduction
protocol, we investigated the radical conjugate addition to 44.
Previous studies conducted with ꢀ-aryl-substituted R,ꢀ-unsatur-
ated R-nitroamides revealed that the chiral Lewis acid generated
by complexation of Mg(NTf2)2 and DBFOX/Ph49 (54, Table 4)
produced the adducts with high ee’s but low dr’s.43 Subse-
quently, we synthesized the second-generation ligands DBFOX/
Nap (53) and DBFOX/Bn (55), which promoted the radical
conjugate addition with slightly higher dr’s.50 On the basis of
these findings, we tested ligands 53-55 in the conjugate addition
of isopropyl radical to 44, and the results are collected in Table
4. Unfortunately, each reaction was characterized by low levels
of diastereoselectivity, and all four possible products 52a-d
(arranged in order of increasing polarity) were produced.
Because of the multitude of Lewis basic sites in radical acceptor
44, we conducted most of the reactions with 2 equiv of Lewis
acid to ensure that sufficient complexation to the R-nitroamide
moiety would occur. An increase in the amount of chiral Lewis
acid to 5 equiv had a small impact on the product ratio, but did
not yield a synthetically useful result (entry 3). In fact, the best
selectivity for a single product, albeit modest, was achieved by
employing substrate-directed stereocontrol with the achiral
(46) Boger, D. L.; Borzilleri, R. M.; Nukui, S.; Beresis, R. T. J. Org. Chem.
1997, 62, 4721.
(47) Handa, S.; Gnanadesikan, V.; Matsunaga, S.; Shibasaki, M. J. Am.
Chem. Soc. 2007, 129, 4900.
(48) (a) Kende, A. S.; Mendoza, J. S. Tetrahedron Lett. 1991, 32, 1699.
(b) Sturgess, M. A.; Yarberry, D. J. Tetrahedron Lett. 1993, 34, 4743.
(49) (a) Kanemasa, S.; Oderaotoshi, Y.; Yamamoto, H.; Tanaka, J.; Wada,
E.; Curran, D. P. J. Org. Chem. 1997, 62, 6454. (b) Iserloh, U.; Curran,
D. P.; Kanemasa, S. Tetrahedron: Asymmetry 1999, 10, 2417. (c)
Iserloh, U.; Oderaotoshi, Y.; Kanemasa, S.; Curran, D. P. Org. Synth.
2003, 80, 46.
We also transformed compounds 52a, 52c, and 52d, the minor
isomers from the radical conjugate addition, into the corre-
sponding cyclic peptides. Isomers 52a and 52d could be isolated
(50) Banerjee, B.; Capps, S. G.; Kang, J.; Robinson, J. W.; Castle, S. L. J.
Org. Chem. 2008, 73, 8973.
(51) Boeckman, R. K., Jr.; Potenza, J. C. Tetrahedron Lett. 1985, 26, 1411.
(52) See the Supporting Information for details.
9
1166 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Antimitotic Bicyclic Peptide Celogentin C
A R T I C L E S
Scheme 10. Synthesis of Macrocycle 59b and Comparison to Stephanotic Acid Methyl Ester
Scheme 11. Synthesis of Other Diastereomeric Macrocycles
in pure form, so these compounds were individually processed
into peptides 56a and 56d with yields similar to those obtained
previously (Scheme 11; cf., Scheme 10). In contrast to the
smooth cyclization of 56b, macrolactamizations of 56a, 56c
(obtained previously, see Scheme 10), and 56d under identical
conditions were low-yielding (20-52%). As a result of the poor
macrolactamization yields, we are unable to rule out the
possibility that epimerization of the activated carboxylate group
may have produced epimers of 57a, 57c, and 57d at the Val
R-carbon. However, comparisons of spectral data to those
published by Moody10a (vide infra) suggest that 57d contains
an L-Val rather than a D-Val residue and is not derived from
epimerization during the peptide coupling.
low (unoptimized) yield of the esterification prevented us from
isolating methyl ester 59a. Fortunately, we did obtain sufficient
quantities of 59c and 59d to compare their 1H NMR spectra to
those of 59b, 60, and diast-60, a diastereomer of 60 constructed
by Moody and co-workers.10a These comparisons were informa-
tive, as the spectrum of 59d matched closely with the spectrum
of diast-60.52 Thus, we believe that our “d” series of compounds
possesses the same (2R,3S)-ꢀ-substituted leucine residue as
Moody’s diast-60. Moreover, the spectrum of 59c did not match
that of either 60 or diast-60.52 The distinct nature of the H
1
NMR spectra of macrocycles 59b-d and the clear similarities
between 59d and diast-60 strengthened our belief that the most
abundant “b” series of compounds was of configuration identical
to that of 1.
The stage was now set for annulation of the right-hand ring
of celogentin C onto the left-hand ring. Coupling of acid 58b
As with 57b, treatment of 57a, 57c, and 57d with BCB
afforded carboxylic acids 58a, 58c, and 58d. However, the
scarcity of the least abundant isomer 58a combined with the
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1167

A R T I C L E S
Ma et al.
Scheme 12. Construction of Octapeptide 47 via Oxidative Coupling
Scheme 13. Preliminary Mechanistic Rationale for Oxidative
to Pro-OBn provided 46, substrate for the crucial oxidative
coupling reaction (Scheme 12). Disappointingly, hexapeptide
46 did not react with Arg-His dipeptide 16 under the conditions
optimized for the coupling of Trp-Pro dipeptide 15 with 16
(see Scheme 5). Monitoring of the attempted coupling by mass
spectrometry revealed that 46 was reacting with NCS to form
a dichlorinated species instead of the required chloroindolenine,
which incorporates only one chlorine atom. Additionally,
comparison of a 1H NMR spectrum of the crude reaction mixture
to one of 46 showed clear differences in the signals emanating
from the proline residue. On the basis of these data, we
hypothesized that the chloroindolenine intermediate was un-
dergoing a second chlorination, perhaps on the Trp-Pro tertiary
amide. The dichlorinated species was unreactive, as increasing
the temperature, time, or concentration of the reaction did not
promote coupling with 16.
Remarkably, a serendipitous discovery provided the solution
to this problem. When a sample of 46 that was contaminated
with unreacted Pro-OBn from the prior peptide coupling was
subjected to the oxidative coupling conditions, the desired
octapeptide was produced. Further experimentation confirmed
the beneficial effect of Pro-OBn on the reaction. The optimized
protocol involved stirring a solution of 46, Pro-OBn (2 equiv),
NCS (3 equiv), and 1,4-dimethylpiperazine in CH2Cl2 at ambient
temperature for 6 h, then adding dipeptide 16 (5 equiv), and
stirring for an additional 24 h. An excess of the imidazole
component was required to ensure a reasonable reaction rate,
but the presence of unreacted 16 complicated the product
isolation. We found it most convenient to subject the crude
oxidative coupling product to transfer hydrogenation, which
cleaved both the Cbz and the benzyl ester protecting groups.
The deprotected octapeptide 47 could then be isolated in 64%
yield.
Coupling
sites (probably the indole C-3 atom and the Trp-Pro amide)
with approximately equal rates. However, in the presence of
the additive, the dichlorinated species could transfer the
undesired chlorine atom to the nitrogen atom of Pro-OBn,
generating the requisite monochlorinated species (i.e., chlor-
oindolenine) along with N-chloroamine A. The chloroindolenine
could then react with dipeptide 16 to form the coupled product,
while elimination of HCl from A promoted by either 1,4-
dimethylpiperazine, excess 16, or Pro-OBn would produce
imine B. Consequently, the unwanted chlorine atom of the
dichlorinated intermediate would be sequestered as a hydro-
chloride salt of the base. In this scenario, both the nucleophilic
nitrogen atom and the acidic R-hydrogen atom of Pro-OBn
are essential structural features that enable its function as a
scavenger of electrophilic chlorine.
The precious nature of hexapeptide 46 limited our ability to
perform detailed mechanistic investigations with this compound.
N-Phthaloyl Trp-Pro dipeptide 61, which was originally used
to develop the oxidative coupling conditions,13 was readily
available in our laboratory. In contrast to 46, 61 underwent clean
oxidative couplings with imidazole nucleophiles in the presence
of 1 equiv of NCS and in the absence of Pro-OBn.13 In these
reactions, we did not observe dichlorinated intermediates.
Apparently, the rate of the desired chlorination of 61 was quite
rapid, so the undesired chlorination was not competitive. Thus,
we had concerns regarding the suitability of 61 for probing the
role of Pro-OBn in the oxidative coupling. Fortunately,
dichlorination of 61 could be accomplished by employing excess
NCS. We then embarked upon the mechanistic investigation
summarized in Table 5 with the viability of 61 as a test substrate
established.
This intriguing result prompted us to investigate the role of
Pro-OBn in the oxidative coupling. Mass spectrometric analysis
of reactions conducted with added Pro-OBn indicated the
presence of four noteworthy species: a monochlorinated adduct
of 46 (presumably the desired chloroindolenine), a dichlorinated
adduct of 46 (presumably the unreactive byproduct), a monochlo-
rinated adduct of Pro-OBn, and the product of HCl elimination
from this adduct. On the basis of this information, we formulated
the preliminary mechanistic hypothesis summarized in Scheme
13. As in the reactions conducted without added Pro-OBn,
hexapeptide 46 may be chlorinated by NCS at two different
9
1168 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Antimitotic Bicyclic Peptide Celogentin C
A R T I C L E S
Table 5. Investigation of Oxidative Coupling with Dipeptides 61 and 62
entry
dipeptide
equiv NCS
additive (equiv)
producta
1
2
3
4
5
6
7
8
9
10
61
61
61
62
61
61
61
61
61
61
3
2
3
3
3
3
3
3
3
3
Pro-OBn (2)
Pro-OBn (2)
Pro-OBn (2)b
Pro-OBn (2)
Pro-OMe (2)
pyrrolidine (2)
none
pyrrolidine (2)
Pro-OBn (2)
Pro-OMe (2)
C
rec. SM (major), C (minor)
D + 2Cl (major), C + Cl (minor)
E (major), D (minor)
C
C (major), C + Cl (minor)
C (47%)
C (41%)
C (68%)
C (79%)
a Values in parentheses refer to isolated yields of product. b Added 3 h after addition of NCS and 1,4-dimethylpiperazine.
Oxidative couplings of 61 and its Cbz-protected congener
62 were conducted with excess imidazole (8 equiv) in CD2Cl2.
The deuterated solvent was chosen to enable reaction monitoring
of Pro-OBn adduct D. Thus, double protection of the Trp amine
is required for successful indole-imidazole oxidative couplings
with acyclic substrates.
1
by H NMR. However, we found that more instructive data
In an effort to determine the structural requirements of the
additive, the reaction was conducted with Pro-OMe (entry 5)
and pyrrolidine (entry 6) in place of Pro-OBn. In the former
case, imidazole adduct C was produced, demonstrating that the
benzyl ester of Pro-OBn is not essential. Interestingly, C was
observed along with its chlorinated adduct in the latter reaction,
suggesting that pyrrolidine may be a viable, albeit less effective,
additive in the oxidative coupling. These results prompted us
to obtain a quantitative measure of the effectiveness of various
additives. Accordingly, the preparative-scale reactions listed in
entries 7-10 were conducted. Surprisingly, adduct C was
isolated in 47% yield in the absence of an additive (entry 7).
Thus, with the simple substrate 61, an additive is not required
for production of the desired oxidative coupling product even
when excess NCS is employed. Compound C was actually
obtained in a slightly lower yield (41%) in the presence of
pyrrolidine (entry 8), indicating that a cyclic secondary amine
alone does not confer a beneficial effect on the reaction. In
contrast, reactions conducted with Pro-OBn (entry 9) and
Pro-OMe (entry 10) delivered C in 68% and 79% yield,
respectively. Therefore, both Pro-OBn and Pro-OMe are
useful additives in the oxidative coupling, with Pro-OMe being
somewhat more effective.
could be obtained by mass spectrometry (ESI-MS). Nonetheless,
for consistency, CD2Cl2 was employed in all of the reactions
listed in Table 5. Use of the conditions developed for oxidative
coupling of 46 with 16 resulted in successful coupling of 61
and imidazole, providing adduct C (entry 1). Reducing the
amount of NCS to 2 equiv gave predominantly starting material,
demonstrating that an excess of NCS relative to Pro-OBn is
required for oxidative coupling to occur (entry 2).
Our first indication that the hypothesis outlined in Scheme
13 was incorrect came when we waited 3 h after the start of
the reaction to add Pro-OBn (entry 3). If the additive was
indeed scavenging chlorine from the dichlorinated species, then
its inclusion at the beginning of the reaction should not be
essential. However, after addition of imidazole, C was not
observed. A chlorinated version of C was detected by ESI-MS,
but the major component of the reaction mixture was a
dichlorinated adduct of D, in which Pro-OBn was functioning
as a nucleophile rather than a chlorine scavenger. This instructive
experiment revealed several important facts about the oxidative
coupling reaction. First, in contrast to the situation with
macrocyclic hexapeptide 46, overchlorinated versions of dipep-
tide 61 are viable electrophiles. Second, Pro-OBn must be
added to the mixture at the outset to exert a beneficial effect on
the reaction; contrary to our earlier hypothesis, this additive is
unable to remove chlorine atoms from dichlorinated species.
Third, if Pro-OBn is added to the mixture after the NCS has
been consumed by 61, then it simply adds to the chloroindo-
lenine intermediate, forming D.
In our early studies of the oxidative coupling, we employed
the phthalimide-protected substrate 61 to prevent the Trp amine
from cyclizing onto the chloroindolenine. Although the chlor-
oindolenine derived from 46 was capable of cyclization via
nucleophilic attack of the Trp-Val amide nitrogen, such a
process was not observed. Presumably, the conformational
constraints of the macrocycle of 46 prevent this undesired event
from occurring. To determine if acyclic substrates possessing a
monoprotected Trp amine could couple with imidazole, we
subjected N-Cbz congener 62 to the standard conditions (entry
4). Tricyclic product E was obtained, along with a minor amount
The data in Table 5, especially the results shown in entry 3,
prompted us to propose a new mechanism for the role of
Pro-OBn in the oxidative coupling of 46 and 16. A hypothesis
that is consistent with our observations is presented in Scheme
14. It is likely that chloroindolenine F is produced by the
reaction of hexapeptide 46 with 1 equiv of NCS (step i). This
intermediate can react further with a second equiv of NCS,
forming a dichlorinated species (step ii). Although we do not
possess definitive data regarding the structure of the dichlori-
1
nated species, H NMR spectra of crude reaction mixtures
suggest that the proline residue is involved in the second
chlorination. In the absence of Pro-OBn, the dichlorinated
species predominates, and upon addition of dipeptide 16, the
desired product 47 is not generated. When Pro-OBn is present
at the outset of the process, it can react with NCS directly to
form N-chloroamine A (step iii). As in our prior proposal,
elimination of HCl from A to give imine B sequesters an
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1169

A R T I C L E S
Ma et al.
Scheme 14. Revised Mechanistic Rationale for Oxidative Coupling
Scheme 15. Completion of Total Synthesis of 1
electrophilic chlorine atom as an HCl salt. Apparently, the rates
of steps i, ii, and iii are balanced such that chloroindolenine F
predominates over the dichlorinated species when 3 equiv of
NCS and 2 equiv of Pro-OBn are employed. Finally, coupling
product 47 evolves from F upon addition of dipeptide 16. Thus,
the role of Pro-OBn is to modulate the concentration of NCS
such that step ii is minimized and the amount of F is maximized.
harmony with prior observations regarding Pbf deprotections,53
no alkylation of the Trp indole was observed. Moreover,
chromatographic purification of 1 was unnecessary as long as
its precursor 63 was rigorously purified.
1
Virtually all of the signals in the H NMR spectrum of
synthetic 1 precisely matched the published data for celogentin
C.1 However, the chemical shift of the imidazole H2 atom
differed significantly, and a smaller deviation was exhibited by
the imidazole H5 atom. We first considered the possibility that
we had synthesized an unnatural atropisomer of celogentin C.
However, diagnostic NOE correlations (indole NH/imidazole
H2 and Trp ꢀ-H/imidazole H5) indicated that our synthetic
compound was identical to the natural product in its orientation
about the heterobiaryl axis.54 Upon further investigation, we
discovered that the chemical shifts of imidazole H2 and H5 were
dependent on both temperature and concentration. The chemical
shift of H2 decreased from 9.22 to 8.92 ppm as the temperature
of a DMSO-d6 solution of 1 was increased from 16 to 44 °C
(Table 6). A less dramatic decrease was observed for H5.
Interestingly, dilution of the solution by a factor of 20 caused
a dramatic decrease in chemical shift (9.16 to 8.04 ppm for
To obtain further evidence in favor of this hypothesis, a
solution of Pro-OBn in CDCl3 was treated with NCS. This
resulted in consumption of the starting material and formation
of N-chloroamine A according to ESI-MS and 1H NMR analysis.
Next, addition of 1,4-dimethylpiperazine to the solution con-
verted A into imine B. In addition to supporting our mechanistic
proposal, this experiment highlights the critical roles of both
the nucleophilic amine and the acidic R-hydrogen of Pro-OBn.
Indeed, it is remarkable that such an exquisitely effective
additive was discovered in serendipitous fashion.
At this stage, only two steps (macrolactamization and
deprotection) were required to complete the total synthesis of
celogentin C. In earlier model studies, formation of the isolated
right-hand ring proved quite facile in the presence of HOBt and
HBTU.13 We were relieved to find that these conditions induced
smooth cyclization of 47, affording bicyclic octapeptide 63 in
83% yield (Scheme 15). Cleavage of both the Pbf and the tert-
butyl ester protecting groups was then accomplished by the
action of TFA, providing 1 in 88% yield. Importantly, and in
(53) Fields, C. G.; Fields, G. B. Tetrahedron Lett. 1993, 34, 6661.
(54) Regarding the NOE correlations observed with natural celogentin C,
the text of ref 1 is in error. The correct data can be located in Table
3 and on the NOESY spectrum given in the Supporting Information.
This discrepancy caused us previously to erroneously conclude that
our model right-hand ring of celogentin C was the unnatural atropi-
somer (see ref 13).
9
1170 J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010

Antimitotic Bicyclic Peptide Celogentin C
A R T I C L E S
Table 6. Temperature Dependence of Chemical Shifts of
Imidazole Hydrogens in 1
growth), and the MDA-MB-468 breast cancer cell line (34%
growth). Accordingly, the selective anticancer properties of
celogentin C may be a subject worthy of future study.
temp (°C)
H2 (ppm)
H5 (ppm)
nata
16
18
20
22
27
37
44
9.41
9.22
9.20
9.18
9.16
9.11
9.03
8.92
7.79
7.73
7.73
7.72
7.72
7.71
7.69
7.66
Conclusions
We have achieved the total synthesis of celogentin C. Early
efforts, which focused on a right-to-left approach to formation
of the bicyclic framework of the natural product, were ultimately
unsuccessful. Nevertheless, in the course of exploring this route,
we discovered that the Braslau modification to the McFadyen-
Stevens reaction27,28 provided an effective solution to the
reduction of an extremely unreactive methyl ester. It is likely
that this heretofore underutilized transformation will find
application in the reduction of other recalcitrant ester groups.
Ultimately, success was achieved after recourse to a left-to-
right annulation strategy. The Leu-Trp side-chain cross-link,
which is imbedded in the left-hand macrocycle, was constructed
via a three-step protocol consisting of an intermolecular
Knoevenagel condensation, a radical conjugate addition, and a
nitro group reduction. The low diastereoselectivity obtained from
the radical conjugate addition was largely offset by the excellent
yield and concise nature of this sequence. The key Trp-His
side-chain linkage was formed by an indole-imidazole oxidative
coupling reaction that joined the intact left-hand ring with an
Arg-His dipeptide. Interestingly, the success of this transforma-
tion depended on the use of Pro-OBn as an additive. Further
experiments were supportive of a mechanistic hypothesis in
which Pro-OBn reacts directly with NCS to form an N-
chloroamine that collapses to an imine upon elimination of HCl.
This process serves to moderate the concentration of NCS so
that the desired chloroindolenine intermediate is favored at the
expense of the unproductive dichlorinated species. The ability
of proline esters to scavenge and sequester electrophilic chlorine
atoms could prove to be useful in other synthetic endeavors.
Celogentin C was obtained in three straightforward steps after
the oxidative coupling reaction, and sufficient quantities of 1
were produced to permit evaluation of its anticancer activity.
In the course of comparing spectral data of synthetic 1 to the
reported data for the natural product, we discovered that the
chemical shifts of the imidazole hydrogen atoms were dependent
on temperature, concentration, and pH. This behavior is presum-
ably an indication of the participation of the imidazole moiety
of 1 in intermolecular hydrogen bonding and/or acid-base
chemistry. Our total synthesis of 1 is the first of a constituent
of the celogentin/moroidin family of bicyclic peptides, and we
believe that our route can be readily adapted to enable the
preparation of other members of this interesting class of
antimitotic compounds.
a Data from natural sample of 1 (ref 1).
Table 7. Concentration Dependence of Chemical Shifts of
Imidazole Hydrogens in 1
factora
H2 (ppm)
H5 (ppm)
natb
1.0
0.50
0.33
0.050
9.41
9.16
9.05
9.04
8.04
7.79
7.72
7.68
7.68
7.40
a Concentration of original solution (ca. 6 mg of 1 in 0.5 mL of
DMSO-d6) was set to 1.0. Subsequent dilutions are reported as a
fraction of the original solution. b Data from natural sample of 1 (ref 1).
H2, 7.72 to 7.40 ppm for H5), as shown in Table 7. It is likely
that the imidazole N3 atom of 1 participates in intermolecular
hydrogen bonding and/or acid-base chemistry, and the tem-
perature and concentration dependence of these phenomena are
reflected in the chemical shifts of the neighboring H2 and H5
atoms. These observations gave us some comfort regarding the
discrepancies between the NMR data for synthetic 1 and the
data collected on the natural sample. Unfortunately, the reported
chemical shift values for natural 1 were outside the ranges
manifested by synthetic 1. On the basis of the observed trends,
it appeared that a colder or more concentrated sample of
synthetic 1 would exhibit H2 and H5 chemical shifts equal to
or greater than those recorded from natural 1. However, the
high freezing point of DMSO and the limited amount of
synthetic 1 on hand both precluded us from verifying this
hypothesis.
We reasoned that the imidazole signals in the 1H NMR would
also exhibit pH dependence. Indeed, addition of a small quantity
of TFA (ca. 2 µL) to a solution of 1 in DMSO-d6 at roughly
the original concentration given in Table 7 caused a large
downfield shift in both signals (9.53 ppm for H2, 7.83 ppm for
H5). As a result, the range of chemical shifts observed for
imidazole H2 and H5 in synthetic 1 at various temperatures,
concentrations, and pH values (9.53-8.04 ppm for H2, 7.83-7.40
ppm for H5) now encompassed the reported chemical shifts of
these hydrogen atoms in natural 1 (9.41 ppm for H2, 7.79 ppm
for H5). This fact greatly increased our confidence that we had
in fact synthesized celogentin C. The identity of our synthetic
sample was then confirmed by HPLC coinjection with a sample
of natural 1.
The potent antimitotic activity of celogentin C prompted us
to synthesize sufficient quantities of 1 for screening against the
National Cancer Institute’s panel of 60 cancer cell lines. In a
single-dose assay conducted at 10 µM concentration of 1, the
cell lines exhibited a mean growth percent of 86%. While the
overall result was disappointing, 1 demonstrated significant
tumor growth inhibition in four cancer cell lines: the SR
leukemia cell line (35% growth), the MDA-MB-435 melanoma
cell line (23% growth), the HS 578T breast cancer cell line (30%
Acknowledgment. We thank Brigham Young University (Gradu-
ate Research Fellowship to B.M., startup funding to S.L.C.) and
the National Institutes of Health (GM70483) for financial support.
We also thank Professor Hiroshi Morita of Hoshi University for
providing an analytical sample of celogentin C, Joshua W. Robinson
for assistance in preparing compound 52, and the Developmental
Therapeutics Program of the National Cancer Institute for perform-
ing the 60-cell assay.
Supporting Information Available: Experimental procedures,
characterization data, and NMR spectra for all new compounds.
This material is available free of charge via the Internet at http://
pubs.acs.org.
JA909870G
9
J. AM. CHEM. SOC. VOL. 132, NO. 3, 2010 1171