A library of spirooxindoles based on a stereoselective three-component coupling reaction

Article abstract of DOI:10.1021/ja045089d

A collection of structurally complex and chemically diverse small molecules is a useful tool to explore cell circuitry. In this article, we report the split-pool synthesis of more than 3000 spirooxindoles on high capacity macrobeads. The key reaction to a

Full text of DOI:10.1021/ja045089d

Published on Web 11/19/2004
A Library of Spirooxindoles Based on a Stereoselective
Three-Component Coupling Reaction
Michael M.-C. Lo, Christopher S. Neumann, Satoshi Nagayama,
Ethan O. Perlstein, and Stuart L. Schreiber*
Contribution from the Program in Chemical Biology, Broad Institute of HarVard and MIT, and
the Howard Hughes Medical Institute, Department of Chemistry and Chemical Biology,
HarVard UniVersity, Cambridge, Massachusetts 02138
Received August 15, 2004; E-mail: stuart_schreiber@harvard.edu
Abstract: A collection of structurally complex and chemically diverse small molecules is a useful tool to
explore cell circuitry. In this article, we report the split-pool synthesis of more than 3000 spirooxindoles on
high capacity macrobeads. The key reaction to assemble the spirooxindole core stereoselectively is a Lewis
acid variant of the Williams’ three-component coupling. After formation, the skeleton was elaborated using
Sonogashira couplings, amide forming reactions, and N-acylations of γ-lactams. The final library was
analyzed by sampling individual macrobeads and by using binomial confidence limits. It was determined
that at least 82% of the library compounds should have better than 80% purity. To demonstrate the utility
of our discovery process, a high-throughput chemical genetic modifier screen was performed using stock
solutions of the resultant products. A number of positives were identified as enhancers of the cellular actions
of latrunculin B, an actin polymerization inhibitor. Through resynthesis, we confirmed one of the positives
and demonstrated that, in yeast cells, it has an EC50 in the sub-micromolar range.
from graph theory5 and an Internet-accessible analysis environ-
ment (ChemBank),6 we hope not only to have chemistry inform
Introduction
A goal of chemical genetics is to find small molecules that
modulate the individual functions of gene products with high
potency and high specificity.1 Nature has provided effective
examples via natural products, which in turn have stimulated
the development of target-oriented synthesis (TOS). However,
synthetic chemists have shown that Nature has no monopoly
on such molecules. An unproven but intuitive hypothesis is that
synthetic compounds that embody features characteristic of
natural products may prove equally suited as modulators:
rigidity, from either covalent bonding or hydrogen bonding to
reduce the conformational flexibility of the molecule; stereo-
chemistry, to fit into the chiral active sites of proteins; and
multiple hydrophilic and hydrophobic groups, to provide the
enthalphic driving force for protein binding.2
Our laboratory has been interested in exploring whether
diversity-oriented synthesis (DOS), especially using a “one-
support, one-stock solution” format,3 might be useful in testing
the above hypothesis. This approach has already led to the
discovery of many small molecules having novel biological
properties.4 By combining this approach with analysis tools, e.g.,
biology but also to have biology inform chemistry. Chemists
can gain insight into underlying property-activity relationships
from the systematic collection and analysis of matrices of data
involving different small molecules, assay measurements, and
cell states.7
DOS aims to synthesize compounds whose diversity results
from variations in skeletons and stereochemistry. In addition,
products having functionalities that enable follow-up chemistry
to be performed effectively and systematically, for example,
using appending processes, are highly valued. The latter point
follows from the expectation that DOS compounds will be
screened in assays and that bioactive compounds identified in
this manner will be modified in ways that facilitate their use as
cellular probes.8 The selection of reactions to be incorporated
in DOS pathways is thus critical to the value of the resultant
library as a tool to investigate biology. Complexity-generating
reactions are appealing, because molecules embodying the
aforementioned features of natural products can be assembled
(4) (a) Pelish, H. E.; Westwood, N. J.; Feng, Y.; Kirchhausen, T.; Shair, M.
D. J. Am. Chem. Soc. 2001, 123, 6740-6741. (b) Spring, D. R.; Krishnan,
S.; Blackwell, H. E.; Schreiber, S. L. J. Am. Chem. Soc. 2002, 124, 1354-
1363. (c) Kuruvilla, F. G.; Shamji, A. F.; Sternson, S. M.; Hergenrother,
P. J.; Schreiber, S. L. Nature 2002, 416, 653-657. (d) Wong, J. C.; Hong,
R.; Schreiber, S. L. J. Am. Chem. Soc. 2003, 125, 5586-5587. (e) Koehler,
A. N.; Shamji, A. F.; Schreiber, S. L. J. Am. Chem. Soc. 2003, 125, 8420-
8421.
(1) (a) Schreiber, S. L. Science 2000, 287, 1964-1969. (b) Schreiber, S. L.
Chem. Eng. News 2003, 81 (9), 51-61.
(2) For discussions of libraries inspired by natural products, see: (a) Hall, D.
G.; Manku, S.; Wang, F. J. Comb. Chem. 2001, 3, 125-150. (b) Nicolaou,
K. C.; Pfefferkorn, J. A. Biopolymers 2001, 60, 171-193. (c) Breinbauer,
R.; Vetter, I. R.; Waldmann, H. Angew. Chem., Int. Ed. 2002, 41, 2878-
2890. (d) Boldi, A. M. Curr. Opin. Chem. Biol. 2004, 8, 281-286.
(3) (a) Blackwell, H. B.; Pe´rez, L.; Stavenger, R. A.; Tallarico, J. A.; Cope-
Eatough, E.; Foley, M. A.; Schreiber, S. L. Chem. Biol. 2001, 8, 1167-
1182. (b) Clemons, P. A.; Koehler, A. N.; Wagner, B. K.; Sprigings, T.
G.; Spring, D. R.; King, R. W.; Schreiber, S. L.; Foley, M. A. Chem. Biol.
2001, 8, 1183-1195.
(5) Haggarty, S. J.; Clemons, P. A.; Schreiber, S. L. J. Am. Chem. Soc. 2003,
125, 10543-10545.
(6) (a) ChemBank. http://chembank.med.harvard.edu (accessed August 2004).
(b) Strausberg, R. L.; Schreiber, S. L. Science 2003, 300, 294-295.
(7) Kim, Y.-K.; Arai, M. A.; Arai, T.; Lamenzo, J. O.; Patterson, N.; Clemons,
P. A.; Schreiber, S. L. J. Am. Chem. Soc., in press.
(8) Burke, M. D.; Schreiber, S. L. Angew. Chem., Int. Ed. 2004, 43, 46-58.
9
10.1021/ja045089d CCC: $27.50 © 2004 American Chemical Society
J. AM. CHEM. SOC. 2004, 126, 16077-16086
16077

A R T I C L E S
Lo et al.
Figure 1. Outline of the Williams’ synthesis of (-)-spirotryprostatin B and structures of two additional diketopiperazines from Aspergillus fumigatus.
from simple building blocks. Multicomponent coupling reactions
are illustrative;9 however, stereoselective versions are not yet
routine.10
In the original conception of this synthetic pathway, we were
attracted to several features of the Williams 3-CR. We recog-
nized the theoretical possibility of varying the stereochemistry
of cycloadducts in a partially systematic manner by exploiting
the stereospecific nature of [2 + 3]-cycloaddition reactions (cf.,
E vs Z dipolarophiles) and the influence of auxiliaries or
catalysts, although we anticipated that control over endo-exo
selectivity might be formidable. We envisioned skeletal diversity
arising by the use of alternative dipolarophiles and by the
removal of the auxiliary, yielding an amino acid that we
perceived as a substrate for subsequent skeleton-determining
reactions. Here, we report several advances that address the
initial challenges necessary to realize these theoretical advan-
tages. The results, involving a pilot split-pool synthesis16 of more
than 3000 single-skeleton spirooxindole products, should enable
more extensive stereochemical and skeletal diversification in
future studies. Numerous technical and tactical advances were
developed that are prerequisites for future studies, including a
Lewis acid mediated variant of the Williams 3-CR that proceeds
on macrobeads,17 and biological assays showing that synthetic
compounds having structural features in common with com-
pounds in Figure 1 are indeed useful as probes. These advances
enabled the synthesis of compounds that derive structural
diversification through appending processes.
We were therefore inspired by the report of Williams and
co-workers in which a stereoselective three-component reaction
(3-CR) was used as the key step in the TOS of the natural
product (-)-spirotryprostatin B.11 A spirocyclic oxindole-
pyrrolidine core was constructed with the simultaneous creation
of three bonds and four stereogenic centers in one single
chemical step. The spiro fusion between the 3-position of the
pyrrolidine ring and the 3′-position of the oxindole ring
distinguishes spirotryprostatins12 from structurally related trypros-
tatins13 and cyclotryprostatins (Figure 1).14 These tryptophan-
derived natural products were all isolated from the fermentation
broth of Aspergillus fumigatus by Osada.15
(9) For reviews, see: (a) Weber, L.; Illgen, K.; Almstetter, M. Synlett 1999,
366-374. (b) Bienayme´, H.; Hulme, C.; Oddon, G.; Schmitt, P. Chem.s
Eur. J. 2000, 6, 3321-3329. (c) Do¨mling, A.; Ugi, I. Angew. Chem., Int.
Ed. 2000, 39, 3168-3210. (d) von Wangelin, A. J.; Neumann, H.; Go¨rdes,
D.; Klaus, S.; Stru¨bing, D.; Beller, M. Chem.sEur. J. 2003, 9, 4286-
4294.
(10) Some examples include: (a) Mannich reaction: List, B.; Pojarliev, P.; Biller,
W. T.; Martin, H. J. J. Am. Chem. Soc. 2002, 124, 827-833. Notz, W.;
Tanaka, F.; Watanabe, S.; Chowdari, N. S.; Turner, J. M.; Thayumanavan,
R.; Barbas, C. F., III. J. Org. Chem. 2003, 68, 9624-9634. (b) Tandem
aza[4 + 2]-cycloaddition/allylboration: Toure´, B. B.; Hoveyda, H. R.;
Tailor, J.; Ulaczyk-Lesanko, A.; Hall, D. G. Chem.sEur. J. 2003, 9, 466-
474. (c) Addition of organoboronic acids to in situ generated imines:
Petasis, N. A.; Zavialov, I. A. J. Am. Chem. Soc. 1997, 119, 445-446. (d)
1,3-Dipolar cycloaddition: Fokas, D.; Ryan, W. J.; Casebier, D. S.; Coffen,
D. L. Tetrahedron Lett. 1998, 39, 2235-2238. (e) Passerini reaction: Frey,
R.; Galbraith, S. G.; Guelfi, S.; Lamberth, C.; Zeller, M. Synlett 2003,
1536-1538. (f) Propargylamine synthesis: Gommermann, N.; Koradin,
C.; Polborn, K.; Knochel, P. Angew. Chem., Int. Ed. 2003, 42, 5763-
5766. (g) aza-Baylis-Hillman reaction: Balan, D.; Adolfsson, H. Tetra-
hedron Lett. 2003, 44, 2521-2524.
(11) (a) Sebahar, P. R.; Williams, R. M. J. Am. Chem. Soc. 2000, 122, 5666-
5667. (b) Sebahar, P. R.; Osada, H.; Usui, T.; Williams, R. M. Tetrahedron
2002, 58, 6311-6322.
(12) (a) Cui, C.-B.; Kakeya, H.; Osada, H. J. Antibiot. 1996, 49, 832-835. (b)
Cui, C.-B.; Kakeya, H.; Osada, H. Tetrahedron 1996, 52, 12651-12666.
(13) (a) Cui, C.-B.; Kakeya, H.; Okada, G.; Onose, R.; Ubukata, M.; Takahashi,
I.; Isono, K.; Osada, H. J. Antibiot. 1995, 48, 1382-1384. (b) Cui, C.-B.;
Kakeya, H.; Okada, G.; Onose, R.; Osada, H. J. Antibiot. 1996, 49, 527-
533. (c) Cui, C.-B.; Kakeya, H.; Osada, H. J. Antibiot. 1996, 49, 534-
540.
(14) Cui, C.-B.; Kakeya, H.; Osada, H. Tetrahedron 1997, 53, 59-72.
(15) The intriguing scaffold and antimitotic activities of spirotryprostatins have
attracted the attention of synthetic chemists. For total syntheses of
spirotryprostatin B, see: (a) von Nussbaum, F.; Danishefsky, S. J. Angew.
Chem., Int. Ed. 2000, 39, 2175-2178. (b) Reference 11a. (c) Wang, H.;
Ganesan, A. J. Org. Chem. 2000, 65, 4685-4693. (d) Overman, L. E.;
Rosen, M. D. Angew. Chem., Int. Ed. 2000, 39, 4596-4599. (e) Bagul, T.
D.; Lakshmaiah, G.; Kawabata, T.; Fuji, K. Org. Lett. 2002, 4, 249-251.
(f) Meyers, C.; Carreira, E. M. Angew. Chem., Int. Ed. 2003, 42, 694-
696.
Results and Discussion
In this initial use of the title reaction, spirooxindoles were
elaborated using building blocks having diverse properties and
orthogonal chemical reactivities. The spirooxindole skeleton is
assembled in the first step, through a highly diastereoselective
3-CR using macrobead-supported aldehydes, either enantiomer
of the Williams’ chiral auxiliary 1,18 and isatin-derived di-
polarophiles bearing the allyl ester (2 and 3).
Optimization of the 3-CR. We were unable to achieve the
3-CR on macrobeads using the conditions shown to be suc-
(16) (a) Furka, A.; Sebestye´n, F.; Asgedom, M.; Dibo´, G. Int. J. Pept. Protein
Res. 1991, 37, 487-493. (b) Lam, K. S.; Salmon, S. E.; Hersh, E. M.;
Hruby, V. J.; Kazmierski, W. M.; Knapp, R. J. Nature 1991, 354, 82-84.
(c) Houghten, R. A.; Clemencia, P.; Blondelle, S. E.; Appel, J. R.; Dooley,
C. T.; Cuervo, J. H. Nature 1991, 354, 84-86. (d) Bunin, B. A.; Ellman,
J. A. J. Am. Chem. Soc. 1992, 114, 10997-10998.
(17) Tallarico, J. A.; Depew, K. M.; Pelish, H. E.; Westwood, N. J.; Lindsley,
C. W.; Shair, M. D.; Schreiber, S. L.; Foley, M. A. J. Comb. Chem. 2001,
3, 312-318.
(18) Both enantiomers of Boc-protected Williams’ chiral auxiliary 1 are
commercially available.
9
16078 J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004

Library of Spirooxindoles
A R T I C L E S
Table 1. Scope of the Lewis Acid Mediated 3-CR: Variation of the Aldehydea
1
a Conversion determined relative to residual aldehyde. d.r. determined by H NMR.
cessful in solution (MS 3 Å, toluene). We therefore examined
various mild Lewis acids to promote the reaction. Most of these
were found to be effective accelerants for the reaction, furnishing
the 3-CR product in high purity and excellent diastereoselec-
tivity.19 To minimize the cleavage of the loaded aldehyde from
the silicon linker by adventitious acid, a base was added to the
reaction as a buffer. Among the amine bases examined, pyridine
was found both to prevent cleavage and to promote the
reaction.20 Finally, molecular sieves were replaced with methyl
orthoformate as the dehydrating agent for operational simplicity.
With the new reaction conditions (eq 1), aromatic aldehydes
of various substitution patterns (Table 1, entries 1-8) are
excellent substrates, affording good levels of diastereoselectivity
(g88:12) and conversion (g89%).21 R,â-Unsaturated aldehydes
(entries 9 and 10) and heteroaromatic aldehydes (entries 11 and
12) offer lower levels of diastereoselectivity, so we decided to
exclude them from the library synthesis.22
dipolarophile in the 3-CR. The major obstacle, however, is the
challenging synthesis of the Z-dipolarophiles. The E-dipolaro-
philes have been reported to isomerize into the Z-isomers with
AlCl3,23 but, in our hands, the reaction is capricious and does
not work for the majority of the substrates that we have
examined. As a result, Z-dipolarophiles were not included in
the subsequent library synthesis. We note that this shortcoming
represents an important challenge for subsequent research. The
stereospecificity of the key cycloaddition reaction offers the
possibility of at least partial stereochemical diversification of
the resulting products. Exploiting this possibility and imparting
further stereochemical diversification through, for example,
alteration of the relative face selectivity in the reaction, constitute
challenging but important goals of future research.
Library Synthesis. I. Reaction Survey and Building Block
Selection. Split-pool synthesis is the most effective method to
generate a desired library in the fewest chemical operations,
but it involves solid-phase synthesis and hence introduces a
different set of criteria for evaluation. While solid-phase
synthesis reactions can be driven to completion by excess
reagents, the desired products are immobilized and cannot be
separated from unreacted starting material or purified from
We planned to increase the stereochemical diversity of the
products further by using either diastereomer (E/Z) of the
(19) Lewis acids examined that are capable of promoting the reaction include:
LiOTf, Mg(OTf)2, Mg(ClO4)2, Sc(OTf)3, Y(OTf)3, La(OTf)3, Yb(OTf)3,
Cu(OTf)2, AgOTf, Zn(OTf)2, Sn(OTf)2, In(OTf)3, and Bi(OTf)3. Strong
Lewis acids cleave the silicon linker on the macrobead and, thus, were not
examined.
(20) Other amine bases that were examined include 2,6-lutidine, 2,6-di-tert-
butyl-4-methylpyridine, diisopropylethylamine, and 2,2′-bipyridine.
(21) The stereochemistry of the product was assigned using NOE analyses, and
it is consistent with Williams’ results reported in ref 11b.
(22) Our protocol also works with other dipolarophiles bearing electron-
withdrawing substituents such as ketones, nitriles, and amides.
(23) Faita, G.; Mella, M.; Righetti, P.; Tacconi, G. Tetrahedron 1994, 50,
10955-10962.
9
J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004 16079

A R T I C L E S
Lo et al.
Scheme 1. Test Results of Alkyne Building Blocks for Sonogashira Coupling
immobilized byproducts. Hence, the reactions used should
convert the immobilized substrates into products with high
purity.
ladium-based protocol with thiosalicylic acid did not succeed
with spirooxindoles bearing the allyl ester, we discovered that,
during Sonogashira coupling, the allyl ester was cleaved
concomitant with the carbon-carbon bond formation. If the
alkyne is omitted from the reaction conditions altogether, the
aromatic iodide is left intact while the allyl ester is cleaved to
the acid.
Relatively high levels of palladium and copper were used in
the reaction to ensure reliable coupling.27 Although the products
are immobilized on macrobeads and the excess reagents can in
theory be washed away, substantial amounts of metal can still
be retained. To obtain reliable cellular screening results, it is
desirable to keep the levels of adventitious metal to a minimum.
The thiourea moiety has been reported to be effective for
scavenging several late-transition elements, presumably due to
its soft nitrogen and sulfur atoms.28 Glyoxal bis(thiosemicar-
bazone), which contains two thiourea subunits, was found to
be a good scavenger for our purpose. Color serves as a useful
indicator for the completion of metal removal, as the light yellow
solution turns deep red on exposure to copper and deep yellow
to palladium (Figure 2).29 Hence, after reaction, the macrobeads
were washed iteratively with a DMF solution of the scavenger
until the solution retained its native light yellow color.
Reactions that are general, are insensitive to minor alterations
of reaction conditions, and have broad functional-group compat-
ibility are preferred in DOS. Reactions are also evaluated based
on their ability to modify the initial chemical properties of the
starting material, e.g., the number of rotatable bonds (cyclization
reactions), or the value of cLogP (replacing a hydrophobic group
with a hydrophilic one).24 Unfortunately, the two goals are often
conflicting, so the challenge of pathway development is to select
reactions that strike a balance.25
The spirooxindole scaffold is densely packed with function-
alities. We therefore sought orthogonal methods to derivatize
the products without use of protecting groups. Three different
approaches were explored and demonstrated to be practical: (a)
palladium-catalyzed carbon-carbon bond formation with con-
comitant allyl ester cleavage; (b) amide formation between the
resultant carboxylic acid with amines; and (c) N-acylation of
γ-lactams with electrophiles.
Several palladium-catalyzed coupling reactions were studied,
and while Suzuki and Stille reactions proceed to some extent,
Sonogashira coupling with aromatic iodide-bearing spirooxin-
doles was found to be the most reliable (eq 2). We surveyed
the scope of the alkynes and classified their reactivities as
outlined in Scheme 1. The classification is only approximate,
as the reactivity of some alkynes is sensitive to minor structural
changes in the spirooxindole.
(26) We have installed other esters (e.g., methyl, ethyl, tert-butyl, benzyl,
p-methoxybenzyl) in the 3-CR product and explored their corresponding
deprotection methods, but none show the desired reactivity.
(27) We could have conceivably used the more potent palladium catalysts that
are based on bulky phosphines/carbenes to lower the catalyst loading. (For
a review, see: Littke, A. F.; Fu, G. C. Angew. Chem., Int. Ed. 2002, 41,
4176-4211.) However, in the library synthesis, electrophoric tags that
contain multiple aryl chlorides are used for encoding, and it is quite likely
that these tags would be activated as well.
The cleavage of the ester group proved to be difficult,
presumably due to steric hindrance.26 While a standard pal-
(28) A case study was presented by Silicycle, a company that markets silica-
bound metal scavengers. http://www.silicycle.com/html/english/products/
product_line.php?cat_id)15 (accessed August 2004).
(29) Amberlite resins have been coated with glyoxal bis(thiosemicarbazone) to
extract metal ions such as Hg(II), Pd(II), Cu(II), Cd(II), and Pb(II) from
aqueous solutions. See: Hoshi, S.; Fujisawa, H.; Nakamura, K.; Nakata,
S.; Uto, M.; Akatsuka, K. Talanta 1994, 41, 503-507.
(24) cLogP is defined as the calculated log P value of a compound, which is
the logarithm of its partition coefficient between n-octanol and water [log-
(coctanol/cwater)]. It is a measure of the hydrophilicity of the compound.
(25) For discussions on the factors governing the selection of reactions in a
split-pool library, see: Marx, M. A.; Grillot, A.-L.; Louer, C. T.; Beaver,
K. A.; Bartlett, P. A. J. Am. Chem. Soc. 1997, 119, 6153-6167.
9
16080 J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004

Library of Spirooxindoles
A R T I C L E S
Figure 3. Acylating reagents for which reaction conditions were optimized.
Next, various methods to effect amide coupling were
explored. Standard coupling reagents PyBOP31/(i-Pr)2NEt are
consistently effective over a broad range of amines. Most of
the amines that present difficulties contain R-heteroatoms or
are sterically demanding (Scheme 3).
Acid chlorides, chloroformates, and isocyanates were screened
for the electrophilic capping of the γ-lactam nitrogen.32 We
identified three electrophiles from a panel of isocyanates and
chloroformates during a preliminary screen of reactivity (Figure
3). The optimal reaction conditions for N-acylation depend on
the electrophile used. By experimenting with changes in solvent,
amine base, or addition of DMAP33 as catalyst, we found
conditions that provide good conversions for individual acylating
reagents. Difficulty in identifying more general conditions
suggests that N-acylation can be sensitive toward changes in
substrate structure, an expected result given the increasing
diversity of library members.
Figure 2. Glyoxal bis(thiosemicarbazone) as colorimetric indicator for
copper and palladium.
To establish the efficiency of glyoxal bis(thiosemicarbazone)
as a metal scavenger quantitatively, we conducted a Sonogashira
coupling to yield a product with several ligating heteroatoms
that are likely to extract some degree of metal from the reaction
mixture (Scheme 2). After the reaction, the macrobeads were
then separated into two portions, one of which was washed with
the scavenger solution. Compared with the control experiment
in which the beads were washed with the same sequence of
solvents without the scavenger, the copper content of the cleaved
product was lowered by 98% to 26 ppm and the palladium
content by 87% to 330 ppm.30
Library Synthesis. II. Preparation. Once the above three
reaction pathways were validated with a range of building
blocks, a library synthesis was planned according to Scheme 4.
Scheme 2. Effectiveness of Scavenger in Reducing Metal Contents from Sonogashira Coupling
Scheme 3. Test Results of Amine Building Blocks for Amide Formation
9
J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004 16081

A R T I C L E S
Lo et al.
Scheme 4. Library Scheme and Final Building Block Selectiona
a The TBDMS groups in black are removed upon cleaving the products from macrobeads.
Starting from macrobead-supported aldehydes,34 a Lewis acid
mediated, stereoselective coupling was performed with 4
aldehydes, 2 morpholinones, and 2 dipolarophiles to yield 16
spirooxindole cores (4 × 2 × 2). The sequence of reaction for
the eight cores containing the aryl iodide was as follows: (a)
Sonogashira coupling with concomitant ester cleavage (skip +
8 alkynes); (b) amidation (skip + 11 amines); and (c) N-
acylation (skip + 3 N-acylating reagents). This sequence
generated 8 × (1 + 8 × 12) × 4 ) 3104 members.35 The
sequence of reaction for the other eight cores without the aryl
iodide was as follows: (a) ester cleavage (skip + deprotection);
(b) amidation (skip + 11 amines); and (c) N-acylation (skip +
3 N-acylating reagents). This sequence generated another 8 ×
(1 + 1 × 12) × 4 ) 416 members, so the complete library
contains 3520 theoretical members.
substitution) and on calculations to diversify the cLogP values
of the library members. The analysis was performed in a linear
fashion, but ultimately we hope to use software tools in
development that allow the selection of building blocks based
on their ability to diversify the desired chemical properties of
all library members. To verify that the building blocks selected
were viable in the synthesis pathway, thirteen library members,
with each building block represented at least once, were
randomly selected for synthesis on the macrobead support and
full characterization.
Split-pool synthesis allows us to maximize operational
efficiency. To ensure that each library member is equally
represented, the macrobeads are split and pooled in different
proportions. Hence, we devised a flowchart to streamline the
process (Scheme 5). Each yellow rectangle represents a reaction
vessel with the identity of the building block. In parentheses
are shown the percentage of the pool that should go into the
vessel and the number of intact beads expected after the
reaction.36 Each green diamond denotes a pool, and the number
shown is the projected total number of beads. The flowchart
not only helps the logistics of moving macrobeads through the
library realization process but also can be useful to describe
some late changes to the process itself.
Building blocks were selected based on intuition to maximize
diversity (e.g., functional groups, heterocycles, degree of
(30) Analyzed by ICP-MS (West Coast Analytical Labs, CA).
(31) (benzotriazol-1-yloxy)tripyrrolidinophosphonium hexafluorophosphate.
(32) In one model study, a spirooxindole was treated with 3-methoxybenzoyl
chloride to give the acylated product in high yield. Although the product
can be cleaved from macrobeads and characterized, it appears prone to
hydrolysis upon prolonged storage as a 10 mM solution in DMSO. Hence,
we excluded acid chlorides from further studies.
(33) N,N-(Dimethylamino)pyridine.
(34) The aldehydes were loaded on 500-600 µm polystyrene beads at a level
of 150-200 nmol per bead.
(35) After the “skip” step in Sonogashira coupling, the allyl ester remained intact.
Thus, amidation is not possible, and only one library member was generated.
(36) Based on the nature of the reaction, a percent loss was assigned to estimate
bead loss or breakage during the reaction.
9
16082 J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004

Library of Spirooxindoles
A R T I C L E S
Scheme 5. Flowchart Guiding the Split-Pool Synthesis of the Library
The compatibility of the pathway with the chemical encoding
procedure used to record the chemical history of the beads was
next determined.37 One combination of the three building blocks
was selected, and the synthesis was performed with encoding
at each step. Successful decoding directly from the bead after
compound cleavage demonstrated that the operating chemical
encoding procedure could be used with this library synthesis.38
Library Synthesis. III. Realization. With preparation work
completed, we were ready to realize the library. The four
aromatic aldehydes A1, A2, A3, and A5 were loaded onto the
macrobeads according to the reported procedure.3b Since no
purification or isolation of intermediates is possible during a
solid-phase split-pool synthesis, we tested each chemical reaction
for high conversion and purity before the beads were transferred
to the next step. Three beads were sampled from each reaction
vessel and the compounds cleaved from them. The crude
The following example from the Sonogashira coupling of
spirooxindoles to 1-propargyl-1H-benzotriazole illustrates our
analysis (Figure 4). Using the above analytical protocol, we
determined that the transformation was incomplete in the first
run, so the beads were resubmitted to the same reaction
conditions until satisfactory conversion was accomplished. In
addition to the progress of the reaction, we monitored the beads
for effective tagging as well. After every encoding step, three
beads were again retrieved and the compounds cleaved from
them. Both the bead and the crude product were analyzed for
proper tagging, and it was determined that compound decoding
is more reliable than bead decoding for identifying library
members.38a
During library realization, we discovered that there were
encoding problems associated with building blocks C3 and C12.
We decided to keep the beads from these two reactions away
from the pool and to perform the final N-acylation on them in
a parallel fashion. We can reflect this change in the flowchart
by disconnecting the corresponding reaction vessels from the
pool step and connecting each of them to two parallel
processes.40 The change to the numbers in the flowchart is
minimal, as only the numbers of beads entering the last pool
and exiting the final N-acylation require modification. As a result
of this change, the number of library members was reduced to
3232.
1
products thus obtained were then analyzed by LC/MS and H
NMR.39 Each library member contains a pyrrolidine moiety that
ensures the detection of the molecular ion (M+1) with elec-
trospray ionization (ESI) detection in LC/MS. In addition, the
high loading of the macrobeads (100-200 nmol/bead) allowed
us to acquire a diagnostic NMR spectrum of material from a
single macrobead within 15 min.
(37) For accounts of the development of chemical-encoding strategy, see: (a)
Ohlmeyer, M. H. J.; Swanson, R. N.; Dillard, L. W.; Reader, J. C.; Asouline,
G.; Kobayashi, R.; Wigler, M.; Still, W. C. Proc. Natl. Acad. Sci. U.S.A.
1993, 90, 10922-10926. (b) Nestler, H. P.; Bartlett, P. A.; Still, W. C. J.
Org. Chem. 1994, 59, 4723-4724.
Library Synthesis. IV. Postsynthesis Quality Control. We
used a sampling approach to estimate the quality of the library.
(38) For details regarding the operation of chemical encoding, see: (a) Blackwell,
H. E.; Pe´rez, L.; Schreiber, S. L. Angew. Chem., Int. Ed. 2001, 40, 3421-
3425. (b) Reference 3a.
(39) The frequency of 1H NMR analysis depends on the reliability of the
reactions involved and the nature of the products.
(40) The flowchart depicting the realized library is provided in the Supporting
Information.
9
J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004 16083

A R T I C L E S
Lo et al.
Figure 4. (a) 1H NMR of the crude product from one macrobead. The bold arrows indicate signals corresponding to the product. The regular arrows
indicate signals corresponding to the deprotected intermediate. (PhMe2Si)2O was used as an internal standard for determining the loading level. (b) LC/MS
of the corresponding sample. One peak was shown in the UV trace, but the peak actually consists of the product (MW ) 745) and the deprotected intermediate
(MW ) 715). There was a 0.6 min shift between the UV trace and the MS trace.
Dolle has reported a similar approach to the same problem.41
The sample was recorded “pure” if the UV trace at λ ) 254
nm suggests a purity better than 80%, and the predominant
compound has a molecular mass that matches one of the possible
combinations. Within a sample of size n, the number of pure
samples should follow a hypergeometric distribution, which can
be approximated by a binomial distribution.42 The confidence
intervals for the population purity p can be readily determined
from the number of pure samples and the sample size.43 Since
we are interested only in the lower limit, the one-sided lower
limit at the 95% confidence interval is determined by setting R
) 0.10.
Figure 5 shows the purity analysis of the samples. Since the
pools that are in parallel (C3/D0, C3/D1, C12/D0, C12/D1)
are expected to contribute only about one-tenth of the number
of compounds from the main pools, we normalized them so
that each of these pools contributes only one datum point. Hence
the total number of samples analyzed was 44, and the adjusted
number of pure compounds was between 40 and 41.44 The lower
limit of purity for the whole library thus calculated is ap-
proximately 82% (one-sided 95% confidence interval), satisfying
our target figure of 80%.45
As the final step prior to moving library members to a
collection maintained at the screening center sponsored by the
Initiative for Chemical Genetics (ICG), the macrobeads were
arrayed into 384-well plates. Compounds were cleaved with
HF•pyridine and eluted from the solid phase in an automated
fashion, a process referred to as formatting.3bAltogether 20 plates
of compounds were formatted. A final quality control was
performed with the formatted compounds, which revealed that
the purity of the library was lowered. This observation was
expected, since the formatting process involves a number of
automated liquid-handling steps and solution transfers. Using
a less stringent criterion, 45 out of 59 wells had purity better
than 70% and were recorded as pure. The lower limit of purity
thus determined was 65%. By removing the plate that contains
the building block combination C3/D1, we improved the lower
limit to 70%. The remaining 19 plates of library compounds
are now stored as part of the ICG DOS collection, available for
screening by the scientific community.46 The library was also
printed on glass slides, enabling its use in small-molecule
microarray assays.47
(44) For pools C3/D0, C3/D1, C12/D0, and C12/D1, the number of pure
compounds coming out from each of them can have a value of either 0 or
1. Based on the analysis on the macrobeads from these pools, we assign a
value of 1 for C3/D0 (p1 ) 0.875), a value of 0 for C3/D1 (p2 ) 0), and
a value of 1 for C12/D0 (p3 ) 1.0). Since pool C12/D1 gives a p4 value
of 0.6, we cannot assign a definite value of either 0 or 1. We therefore
report the total number of pure compounds as a range and calculate the
lower limit in each case.
(45) The average of the two lower limits from 40 pure compounds (80.4%) and
41 pure compounds (83.3%).
(46) For a listing of libraries available for screening at the ICG, see http://
iccb.med.harvard.edu/chemistry/compound_libraries/libraries_available.htm
(accessed August 2004).
(41) Dolle, R. E.; Guo, J.; O’Brien, L.; Jin, Y.; Piznik, M.; Bowman, K. J.; Li,
W.; Egan, W. J.; Cavallaro, C. L.; Roughton, A. L.; Zhao, Q.; Reader, J.
C.; Orlowski, M.; Jacob-Samuel, B.; DiIanni Carroll, C. J. Comb. Chem.
2000, 716-731.
(42) The multiplicative finite population correction factor for the confidence
limit is given as x((N - n)/(N - 1)), where N is the number of beads in
the whole library and n is the number of beads in the sample. The factor
is greater than 0.99 and hence approximated by 1.
(43) Intercooled Stata 8.0 for Macintosh; Stata Corporation: College Station,
TX, 2003.
9
16084 J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004

Library of Spirooxindoles
A R T I C L E S
Figure 5. Purity analysis of the library before formatting.
Figure 6. Plot of cLogP versus molecular weight of the final library members.
Preliminary Results from Cell-Based Screening. For a
small-molecule library to be broadly useful in chemical genetics,
members must be able to cross cell membranes and to bind
specifically to protein targets. An examination of two key
chemical properties of our realized library reveals that the
members tend to be more lipophilic than those deemed
“druglike” and have molecular weights more reminiscent of
natural products than druglike compounds (Figure 6).48 To
demonstrate the value of these compounds as effective probes,
we used a chemical genetic modifier screen to search for
bioactive compounds.49 This type of screen identifies compounds
that enhance or suppress cellular phenotypes, for example, those
induced by a small molecule with a known mechanism of action.
An assay was developed to identify enhancers of the growth
arrest induced by latrunculin B,50 a natural product that
(49) For a general review on screening, see: (a) Stockwell, B. R. Nat. ReV.
Genet. 2000, 1, 116-125. For examples of chemical genetic modifier
screens, see: (b) Koeller, K. M.; Haggarty, S. J.; Perkins, B. D.; Leykin,
I.; Wong, J. C.; Kao, J. M.-C.; Schreiber, S. L. Chem. Biol. 2003, 10, 397-
410. (c) Butcher, R. A.; Schreiber, S. L. Chem. Biol. 2003, 10, 521-531.
(d) Haggarty, S. J.; Koeller, K. M.; Kau, T. R.; Silver, P. A.; Roberge, M.;
Schreiber, S. L. Chem. Biol. 2003, 10, 1267-1279. (e) Huang, J.; Zhu H.;
Haggarty, S. J.; Spring, D. R.; Snyder, M.; Schreiber, S. L. Submitted for
publication.
(47) McPherson, O. M.; Koehler, A. K. Broad Institute, Cambridge, MA.
Unpublished work.
(48) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. AdV. Drug
DeliVery ReV. 1997, 23, 3-25.
9
J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004 16085

A R T I C L E S
Lo et al.
limit of 30 µM, yeast growth is unaffected. This experiment
shows that 9 does not show the same phenotype as latrunculin
B and that it is synthetic lethal with latrunculin B. Second, we
chose a member of the library 7i, which has a median molecular
weight and cLogP, to serve as a negative control. Yeast cells
treated with 7i alone or with a combination of 7i and latrunculin
B show no inhibition of growth relative to untreated yeast,
confirming that the activity of 9 is not a general property of
the spirooxindole core. Further studies of 9 and the elucidation
of its mechanism are underway. For the moment, the experi-
ments confirm that small molecules emanating from the
synthetic pathway described herein are capable of yielding novel
probes of cell circuitry, in this case of the actin regulatory
network.
Figure 7. Spirooxindoles screened in follow-up assays for latrunculin B
enhancement.
sequesters monomeric actin and prevents the formation of actin
microfilaments.51 Latrunculin B has been a valuable tool in
elucidating the roles of the actin cytoskeleton in both yeast and
mammalian cells.52 Small molecules that perturb this system
should provide further insight into its regulation.
Conclusion
We have developed a stereoselective, Lewis acid mediated
variant of the Williams 3-CR that is suitable for macrobead-
supported aldehydes. The diversity of these structurally complex
and stereochemically diverse products can be further increased
by efficient Sonogashira coupling, amide formation, and N-
acylation of γ-lactams. A library of 3232 theoretical spiro-
oxindoles was synthesized, and through sampling of individual
macrobeads, it was determined that at least 82% of the library
compounds have greater than 80% purity. To demonstrate our
overall discovery process, a cell-based screen was performed
with library members to identify enhancers of latrunculin B,
an actin polymerization inhibitor. Through resynthesis, one of
the positives was confirmed and found to have an EC50 in the
sub-micromolar range. One of several important goals that
emerged from these studies is the use of computed properties
of virtual products to guide the selection of subunits and
appendages, so that the actual products are distributed across
regions of chemical space of greatest promise for interrogation.
The initial screen was performed in 384-well plates using
wild-type yeast growing in a nutrient-rich medium. Latrunculin
B was added to a final concentration of 7 µM, followed by pin
transfer of library stock solutions to a final concentration of
approximately 33 µM. Ten formatted stock plates were used
for the initial screen, resulting in 3520 individual assays. The
assay plates were allowed to stand at room temperature, and
yeast growth was quantitated over a span of 24 to 96 h. If the
small molecule enhances the effect of latrunculin B, the growth
of the yeast in the assay well will be retarded. By comparing
growth in the assay wells with that in the control wells lacking
library members, 36 compounds were scored as enhancers.
Thirty-three were successfully decoded, revealing that these
positives represented 19 unique chemical structures.53
A crude structure-activity relationship can be deduced from
these initial positives. Two building blocks, aldehyde A3 (28/
33) and the (5R,6S)-morpholinone E1 (28/33), dominate the
positives. All positives were derived from the dipolarophile F2,
which is not substituted at the 5-position of the indole ring,
even though these members should constitute only 12% of the
library. These results hint that the positives might be acting
through a common target or mechanism. Based on these
observations and the relative strengths of the positives, com-
pound 9 was chosen for resynthesis and follow-up assays (Figure
7). For these assays, the amount of latrunculin B was reduced
to 1.0 µM, which is 12.5% of the experimentally determined
EC50 value.54 At this concentration, latrunculin B alone has no
observable effect on yeast growth. When 9 was added to this
same medium, the inhibitory effect of latrunculin B is signifi-
cantly enhanced. The EC50 value55 for 9 was determined to be
550 ( 50 nM.
Acknowledgment. We thank Robert M. Williams for a
generous gift of Boc-protected 1 and for sharing unpublished
results with us. We thank Paul Clemons for generating computed
properties of the library. We thank Leticia Castro, Jennifer
Raggio, Kerry Pierce, and James Roger for their expert
assistance in library decoding, formatting, and quality control.
We thank Justin Klekota and Newton Y.-L. Wai for discussions
concerning sampling statistics. We acknowledge NIH/NCI Grant
2P01 CA78048-06A1 for support of this research and thank
the NCI for sponsorship of the ICG. S.N. and C.S.N. were
supported by JSPS and NSF fellowships, respectively. S.L.S.
is an Investigator with the Howard Hughes Medical Institute at
Harvard University.
Two additional experiments were performed as controls. First,
when yeast cells are treated with 9 alone, up to the solubility
Supporting Information Available: Experimental procedure
for library synthesis, characterization data of 3-CR products and
demonstration compounds, assay protocols, structures of assay
positives. This material is available free of charge via the
Internet at http://pubs.acs.org. The SDF file containing the
product structure and 6 computed chemical properties for each
member of the realized library can be downloaded at http://
people.med.harvard.edu/∼pclemons/DOS/props/props.htm.
(50) Detailed procedures for the high-throughput assay and corresponding follow-
up assays are available in the Supporting Information.
(51) Spector, I.; Shochet, N. R.; Kashman, Y.; Groweiss, A. Science 1983, 219,
493-495.
(52) Peterson, J. R.; Mitchison, T. J. Chem. Biol. 2002, 9, 1275-1285.
(53) See Supporting Information for the structures of these positives.
(54) Effective concentration that produces 50% of the maximum possible
response for the small molecule.
(55) Defined as the concentration required to restore latrunculin B to show the
same phenotype at its EC50 value 8 µM.
JA045089D
9
16086 J. AM. CHEM. SOC. VOL. 126, NO. 49, 2004