Synthesis of a 10,000-membered library of molecules resembling carpanone and discovery of vesicular traffic inhibitors

Article abstract of DOI:10.1021/ja056338g

Split-and-pool synthesis of a 10,000-membered library of molecules resembling the natural product carpanone has been achieved. The synthesis features development of solid-phase multicomponent reactions between nitrogen nucleophiles, enones, and hydroxylamines, and a solid-phase application of the Huisgen cycloaddition affording substituted triazoles. The synthesis was performed in high-capacity (500 μm) polystyrene beads using a one bead-one stock solution strategy that enabled phenotypic screens of the resulting library. Using whole-cell fluorescence imaging, we discovered a series of molecules from the carpanone-based library that inhibit exocytosis from the Golgi apparatus. The most potent member of this series has an IC₅₀ of 14 μM. We also report structure-activity relationships for the molecules exhibiting this interesting phenotype. These inhibitors of exocytosis may be useful reagents for the study of vesicular traffic.

Full text of DOI:10.1021/ja056338g

Published on Web 04/01/2006
Synthesis of a 10,000-Membered Library of Molecules
Resembling Carpanone and Discovery of Vesicular Traffic
Inhibitors
Brian C. Goess,† Rami N. Hannoush, Lawrence K. Chan, Tomas Kirchhausen,* and
Matthew D. Shair*
Contribution from the Department of Chemistry & Chemical Biology, HarVard UniVersity,
Cambridge, Massachusetts 02138, HarVard Institute of Chemistry & Cell Biology, HarVard
Medical School, Boston, Massachusetts 02115, and the Department of Cell Biology and the CBR
Institute for Biomedical Research, HarVard Medical School, 200 Longwood AVenue,
Boston, Massachusetts 02115
Received September 14, 2005; E-mail: shair@chemistry.harvard.edu
Abstract: Split-and-pool synthesis of a 10,000-membered library of molecules resembling the natural product
carpanone has been achieved. The synthesis features development of solid-phase multicomponent reactions
between nitrogen nucleophiles, enones, and hydroxylamines, and a solid-phase application of the Huisgen
cycloaddition affording substituted triazoles. The synthesis was performed in high-capacity (500 µm)
polystyrene beads using a one bead-one stock solution strategy that enabled phenotypic screens of the
resulting library. Using whole-cell fluorescence imaging, we discovered a series of molecules from the
carpanone-based library that inhibit exocytosis from the Golgi apparatus. The most potent member of this
series has an IC50 of 14 µM. We also report structure-activity relationships for the molecules exhibiting
this interesting phenotype. These inhibitors of exocytosis may be useful reagents for the study of vesicular
traffic.
Introduction
tures contribute most to overall activity. However, one major
difference between molecules derived from DOS and those
Natural products and their derivatives serve as medicines and
as probes of complex biological processes. For example,
rapamycin is being used to elucidate the role of mTOR
(mammalian target of rapamycin),1 and brefeldin A,2 a unique
inhibitor of the GTPase Arf1, is a valuable reagent for studying
vesicular traffic (Chart 1). Many medicines are themselves
natural products (erythromycin,3 taxol,4 and lovastatin5) or are
derived from natural products (aspirin).
Recently, there has been an effort to determine whether
molecules generated through diversity-oriented synthesis (DOS)
might serve, as natural products do, as useful probes of
biological processes or as lead structures for medicines.6
Advantages of molecules derived from DOS over those derived
from natural sources include the following: (1) they can be
generated in a format more amenable to high-throughput
biological screens; and (2) the structures of compounds entering
screens can be controlled, allowing planned experimentation to
determine which structures are most active and which substruc-
derived from natural sources is that natural products may be
exposed to the pressures of evolution (mutation, selection, and
amplification), leading to molecules that provide some advantage
to the producing organisms.7 This means that certain natural
products are more likely to be, for example, high affinity binders
to their protein targets than are molecules generated via DOS.
We have decided to construct DOS libraries around core
structures that are based directly on natural product structures.8
Our hypothesis is that library members with such core structures
will have a higher intrinsic ability to bind proteins than would
library members with core structures that are not based directly
on natural product structures. Furthermore, natural products offer
the opportunity to use biomimetic synthesis as a strategy for
rapid assembly of their complex structures.9 It is important to
distinguish our strategy, using natural product-based libraries
to discover molecules with new biological functions, from the
existing strategy of using natural product-based libraries to
improve the properties of natural products.8
† Present address: Furman University, Greenville, SC.
(1) Mita, M. M.; Mita, A.; Rowinsky, E. K. Cancer Biol. Ther. 2003, 2, S169-
S177.
(2) Klausner, R. D.; Donaldson, J. G.; Lippincott-Schwartz, J. J. Cell Biol.
1992, 116, 1071-1080.
(3) Pathak, A. N.; Srivastava, A. J. Sci. Ind. Res. India 1998, 57, 862-872.
(4) Shigemori, H.; Kobayashi, J. J. Nat. Prod. 2004, 67, 245-256.
(5) Alberts, A. W. Am. J. Cardiol. 1988, 62, 10J-15J.
(6) (a) Webb, T. R. Curr. Opin. Drug DiscoVery 2005, 8, 303-308. (b) Spring,
D. R. Org. Biomol. Chem. 2003, 1, 3867-3870. (c) Schreiber, S. L. Science
2000, 287, 1964-1969.
(7) (a) Piggott, A. M.; Karuso, P. Comb. Chem. High Throughput Screening
2004, 7, 607-630. (b) For an alternative theory on why nature produces
diverse natural products, see: Firn, R. D.; Jones, C. G. Nat. Prod. Rep.
2003, 20, 382-391.
(8) For a review on the integration of natural products in library synthesis,
see: Paululat, T.; Tang, Y.-Q.; Grabley, S.; Thiericke, R. Chim. Oggi 1999,
17, 52-56 and references therein.
(9) de la Torre, M. C.; Sierra, M. A. Angew. Chem., Int. Ed. 2003, 43, 160-
181.
9
10.1021/ja056338g CCC: $33.50 © 2006 American Chemical Society
J. AM. CHEM. SOC. 2006, 128, 5391-5403
5391

A R T I C L E S
Goess et al.
Chart 1. Natural Products and Their Derivatives Used as Medicines or to Probe Biology
Scheme 1
Scheme 2
We demonstrated the successful application of this strategy
in our recent discovery of secramine. We developed a biomi-
metic diversity-oriented synthesis of galanthamine-like mol-
ecules, screened these molecules in a fluorescence-based
phenotypic screen, and discovered secramine, a cell-permeable
inhibitor of vesicular traffic from the Golgi.10 Here we report
the development, split-pool synthesis, and phenotypic screening
of a 10,000-member encoded library of molecules based on the
biomimetic synthesis of carpanone first reported by Chapman.11
Our synthesis features three-component conjugate addition
reactions that may prove useful in other DOS projects. We
screened the library in a fluorescence-based phenotypic assay
and identified members of the library that inhibit vesicular
traffic, and the results of this screen are described. The molecules
discovered in this study may be useful reagents for studying
vesicular traffic and the proteins involved in vesicular traffic.
group that could be transformed through split-and-pool synthesis
in myriad ways;13 and (5) its rigid, cup-like shape should enable
diastereoselective reactions from its convex face.14
Recently, we reported a solid-phase, biomimetic synthesis
of carpanone-like molecules via the heterodimerization of
electronically differentiated o-hydroxystyrenes (Scheme 2).15
Electron-rich o-hydroxystyrene 1, immobilized in polystyrene
resin, was treated with electron-deficient o-hydroxystyrene 2
in solution, and, upon addition of iodobenzene diacetate (PhI-
(OAc)2), an oxidative coupling followed by an inverse-electron
demand Diels-Alder reaction occurred to yield resin-bound
heterodimer 3. The ability to produce a carpanone-like het-
Results and Discussion
Carpanone has a number of features that make it useful as a
scaffold for split-and-pool library synthesis (Scheme 1): (1)
carpanone possesses significant complexity (six rings and five
contiguous stereocenters); (2) it is generated from easily
prepared precursors in an efficient, single-step transformation11
that could potentially be adapted to the solid phase;12 (3) its
methyl substituents offer a handle for attachment to the solid
phase; (4) carpanone contains an enone, a versatile functional
(12) For a solid-phase synthesis of carpanone itself which appeared after our
initial report,15 see: Baxendale, I. R.; Lee, A.-L.; Ley, S. V. J. Chem. Soc.,
Perkin Trans. 1 2002, 1850-1857.
(13) Though technically a dienone, one of the dienone olefins in the scaffold is
not in conjugation with the carbonyl, rendering it unreactive as a conjugate
acceptor (vide infra).
(14) nOe experiments on the conjugate addition adduct of thiophenol and a
representative library core synthesized in our previous communication15
confirmed that conjugate addition occurs diastereoselectively from the
convex face of the enone.
(15) Lindsley, C. W.; Chan, L. K.; Goess, B. C.; Joseph, R.; Shair, M. D. J.
Am. Chem. Soc. 2000, 122, 422-423.
(10) Pelish, H. E.; Westwood, N. J.; Feng, Y.; Kirchhausen, T.; Shair, M. D. J.
Am. Chem. Soc. 2001, 123, 6740-6741.
(11) Chapman, O. L.; Engel, M. R.; Springer, J. P.; Clardy, J. C. J. Am. Chem.
Soc. 1971, 93, 6696-6698.
9
5392 J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006

Synthesis of Carpanone-like Molecules
A R T I C L E S
Scheme 3
erodimer in the solid phase was significant not only because it
demonstrated that the key carpanone dimerization reaction could
be performed in resin, but it also transformed the dimerization
reaction into a diversity-generating step. Thus, many scaffolds
could be produced from the same resin-bound o-hydroxystyrene
simply by varying the structure of the o-hydroxystyrene
introduced in solution.
A screen of different linkers (R1) and spacers (R2) was
conducted to optimize the yield of heterodimer 3 relative to 4,
the undesired product of intra-resin homodimerization of 1. The
optimal combination, depicted in Scheme 2, proved general.
Five unique electron-deficient o-hydroxystyrenes were found
to couple with 1 to yield resin-bound heterodimers that were
isolated in 77-81% yield after cleavage from the resin. Analyses
of the reaction products indicated that the only significant
byproduct of these reactions was undesired homodimer 4, which
was produced in 10-15% yield.
The presence of homodimer byproduct, however, complicated
the use of this reaction as the basis for a split-and-pool library
synthesis. At maximal resin loading levels of 1 (1.1 mmol/g
resin), formation of homodimer 4 effectively suppressed het-
erodimerization. At a lower loading level of 1 (0.15 mmol/g
resin), heterodimer 3 was favored over 4 by the ratio of 5.3:1,
but this loading level was considered to be too low to support
multiple biological screens using the ICCB16 standard protocols
for library formatting and analysis.17 Furthermore, the unwanted
homodimer present in each resin bead following oxidative
coupling would be retained throughout the split-and-pool
synthesis and would contaminate the stock solutions of final
library compounds.
With these considerations in mind, we elected to use a
homodimer similar to 4 as a scaffold for our library synthesis.
This strategy offers several benefits over a heterodimeric library.
A scaffold formed in a solid-phase homodimerization would
not suffer from the impurity problems that plagued our
heterodimeric scaffolds. Furthermore, loading levels could be
significantly increased over those used for heterodimerizations
since there would be no need to spatially separate the resin-
bound substrates in order to minimize intra-bead homodimer-
ization.18 Finally, the diethylsilyl linker and amide spacer, which
were developed empirically solely to enhance selectivity for
heterodimerization, could be replaced with a more robust
diisopropylsilyl linker19 and an all-carbon spacer, thus simplify-
ing resin preparation. The primary drawback to this approach
is that, unlike the analogous situation with heterodimeric
scaffolds, the dimerization reaction itself could not be used as
a diversification step to yield multiple scaffolds20 since there is
only one possible homodimer that can be formed from a resin-
bound monomer. We envisioned that the consequences of this
modification could be minimized if allylated phenol 5 were used
as the monomer (Scheme 3). After formation of homodimer 6,
a selective deallylation would reveal a free phenol (7) which
would provide a handle for diversification, thereby recapturing
some of the diversification potential at this position. We also
hoped to utilize the remaining allyl group as a component of
reactions designed to alter the scaffold structure of 7 and yield
multiple new scaffolds.
The synthesis of resin-bound racemic homodimer is depicted
in Scheme 4. Hydroquinone was monoallylated to yield 8, which
was formylated under basic conditions to produce aldehyde 9.21
An E-selective Wittig reaction22 gave acid 10, which was
reduced to alcohol 11 with lithium aluminum hydride. Direct
resin loading of 11 gave a small amount of undesired attachment
through the phenol hydroxyl, so the phenol was selectively
protected as the acetate (12) prior to resin loading.23 Protode-
silylation of the resin19 with triflic acid, addition of 12, and
deprotection of the resulting resin-bound methyl ester with
piperidine yielded the loaded resin 13 (loading level ) 1.05
mmol/g resin).
Our attempts to prepare resin-bound homodimer 18 directly
via PhI(OAc)2-mediated24 dimerization of silyl-protected 14
(step g) with 13 were not successful. The attempted dimerization
proceeded cleanly;25 however, the mass recovery of cleaved
homodimer 16 was only 50% based on the desired incorporation
of one molecule of 14 per molecule of 13. The most likely
explanation for this occurrence is that exclusive intra-bead
homodimerization occurred producing cross-linked 15. Indeed,
treatment of 13 with PhI(OAc)2 in the absence of 14 gave a
quantitative yield of homodimer 16 after resin cleavage,
presumably via 15.
Intra-bead homodimerization was unacceptable because it
reduced the loading level of the resin by 50%. To avoid this
complication, we loaded onto the resin a mixture of molecules
17a and 17b. 17a and 17b were prepared by a solution-phase
homodimerization of 11 to produce 16 (step j) and subsequent
monosilylation. Loading took place successfully to yield the
(16) Institute for Chemistry and Cell Biology, Harvard University, Cambridge,
MA.
(17) The standard resin used by the ICCB is a derivative of 500 µm polystyrene
resin with a diisopropyl-p-methoxyphenylsilyl terminus. See: (a) Sternson,
S. M.; Louca, J. B.; Wong, J. C.; Schreiber, S. L. J. Am. Chem. Soc. 2001,
123, 1740-1747. (b) Spring, D. R.; Krishnan, S.; Blackwell, H. E.;
Schreiber, S. L. J. Am. Chem. Soc. 2002, 124, 1354-1363. (c) Blackwell,
H. E.; Pe´rez, L.; Stavenger, R. A.; Tallarico, J. A.; Eatough, E. C.; Foley,
M. A.; Schreiber, S. L. Chem. Biol. 2001, 8, 1167-1182. (d) See ref 19.
(18) For an analysis of site-site interactions, see: Hodge, P. Chem. Soc. ReV.
1997, 26, 417-424.
(21) Baker, R.; Castro, J. L. J. Chem. Soc., Perkin Trans. 1 1990, 47-65.
(22) Suzuki, Y.; Takahashi, H. Chem. Pharm. Bull. 1983, 31, 1751-1753.
(23) Torrini, I.; Zecchini, G. P.; Agrosi, F.; Paradisi, M. P. J. Heterocycl. Chem.
1986, 23, 1459-1463.
(24) For a discussion of the reactivity of hypervalent iodine reagents, see:
Varvoglis, A. HyperValent Iodine in Organic Synthesis; Elsevier: London,
1996.
(19) 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.
1
(25) As judged by the crude H NMR of product 16, which was cleaved from
(20) Burke, M. D.; Schreiber, S. L. Angew. Chem., Int. Ed. 2004, 43, 46-58.
the resin with HF‚pyridine.
9
J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006 5393

A R T I C L E S
Scheme 4 a
Goess et al.
a Conditions: (a) (i) LiH, DMF, 0 °C, (ii) allyl bromide, 36%; (b) NaOH, CHCl3, H2O, reflux, 56%; (c) LiN(TMS)2, 4-carboxybutyltriphenylphosphonium
bromide, THF, 0 °C, 65%; (d) LAH, THF, -78 °C, 75%; (e) NaOH, 1-acetyl-1H-1,2,3-triazolo[4,5-b]-pyridine, THF, 84%; (f) (i) 5% TfOH, diisopropyl-
p-methoxyphenylsilyl resin, CH2Cl2, (ii) 12, (iii) piperidine, THF; (g) TIPSCl, imidazole, DMF, 85%; (h) PhI(OAc)2, F6-IPA, 14, CH2Cl2; (i) (i) HF‚pyridine,
THF, (ii) TMSOCH3, 50%; (j) PhI(OAc)2, CH2Cl2, 0 °C, 60%; (k) TIPSCl, imidazole, DMF, 46%; (l) (i) 5% TfOH, diisopropyl-p-methoxyphenylsilyl resin,
CH2Cl2, (ii) 17a/b; (m) (i) HF‚pyridine, THF, (ii) TMSOCH3, 90%.
Scheme 5
corresponding mixture of resin-bound isomers 18a and 18b.26
With resin-bound scaffold 18 in hand, we began to investigate
reactions that would lead to a diverse library of molecules.
On the Mechanism of PhI(OAc)2-Promoted Oxidative
Dimerization. The Chapman biomimetic synthesis of carpanone
used stoichiometric PdCl2 as the oxidant to effect dimerization
of o-hydroxystyrene 19 (Scheme 5(a)).11 According to Chap-
man’s proposal, the initial step involves formation of bisphenoxy
Pd(II) intermediate 20. Oxidation of 20 and concomitant
reduction of Pd(II) to Pd(0) generates 21, which then undergoes
an endo-selective, intramolecular, inverse-electron demand [4
+ 2] cycloaddition affording carpanone. It was suggested that
the dimerization of intermediate 20 may be radical-mediated,
in which case the high diastereoselectivity observed in the
formation of 21 is difficult to explain.
We found that PhI(OAc)2 is a superior reagent for this
transformation based on our investigations with similar o-
hydroxystyrenes in solution-phase and solid-phase syntheses
(Scheme 4). We propose a different mechanism for the PhI-
(OAc)2-mediated oxidative dimerization that accounts for the
high diastereoselectivity and suggest that the reaction reported
by Chapman with PdCl2 may proceed via similar intermediates
(Scheme 5(b)). Our proposed mechanism involves activation
of 19 by PhI(OAc)2 to yield 22 followed by oxidative addition
of a second molecule of 19 generating 23. This mechanism of
oxidation of phenols is well-precedented.27 Intermediate 23 is
(26) As deduced by 1H NMR and mass recovery of 16, obtained by HF‚pyridine-
mediated cleavage of 18a and 18b from the resin. All subsequent studies
showed that the mixture of 18a and 18b behaved as if it were one isomer.
Furthermore, the distinction between 18a and 18b is transient. Following
cleavage from the resin, both 18a and 18b yield the same product (16).
(27) (a) Zhdankin, V. V.; Stang, P. J. Chem. ReV. 2002, 102, 2523-2584. (b)
See ref 24.
now poised for an intramolecular, inverse-electron demand,
endo-selective [4 + 2] cycloaddition to afford carpanone. A
feature of this mechanism is that the observed diastereoselec-
tivity is derived from the endo-selectivity of a cycloaddition
rather than from a radical-mediated coupling.
9
5394 J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006

Synthesis of Carpanone-like Molecules
A R T I C L E S
Scheme 6 a
Scheme 7 a
a Conditions: (a) Pd(OAc)2, Ag2CO3, PPh3, CH3CN, 60 °C, dark, 96%;
(b) (dihydroimidazolylidene)RuCl2(PCy3)dCHPh, benzene, 55 °C, 74%;
(c) HCl, CH3OH, 0 °C, 74%.
a Conditions: (a) PhI(OAc)2, CH2Cl2, 0 °C 55%; (b) Pd(PPh3)4, sodium
2-ethylhexanoate, CH2Cl2, 40%; (c) Cs2CO3, CH3I, DMF, 0 °C, 85%; (d)
CeCl3‚7H2O, NaBH4, CH3OH, 0 °C, 95%; (e) acrylic acid, Et3N, EDC,
DMAP, CH2Cl2, 0 °C, 74%.
noted.17b A screen of alternate reduction methods also failed.32
These results highlight the difficulties associated with translation
of solution-phase chemistry to the solid phase.
We had also sought to increase the skeletal diversity of the
library by removing the allyl group of 29 and cleaving the
resulting hemiketal to the corresponding ketone and phenol, two
functional groups that could be further diversified. Although
the allyl group could be removed by treatment with Pd(PPh3)4
and PhSiH3, the resulting hemiketal resisted cleavage under both
acidic and basic conditions.
Solid-Phase Reaction Development. Given the difficulties
associated with translating our solution-phase chemistry to the
solid phase, we elected to perform all subsequent reaction
development in the solid phase. Our attention first turned to
diversity-generating reactions involving the enone of 18 (Scheme
8). Treatment of 18 with methylamine and O-ethylhydroxyl-
amine resulted in the formation of oxime 33 as a mixture of
oxime geometrical isomers (Path A).33 Such multicomponent
reactions (MCRs) are advantageous as they incorporate building
blocks more rapidly than the corresponding stepwise processes,
and they minimize the potential for resin bead breakage by
minimizing the number of reactions to which each resin bead
is exposed. Other combinations of amines and hydroxylamines
that participated in this MCR are presented in Chart 2. Building
blocks were validated by performing each solid-phase test
reaction on a scale sufficient to provide approximately 5 mg of
crude product following cleavage from the resin. An 1H NMR
of each crude product was then taken to provide a rough
assessment of product purity, and only building blocks that gave
>90% yield of the desired product were incorporated into the
library.34
Solution-Phase Reaction Development. Understanding that
a library composed of molecules that each possess the same
scaffold as 18 is less diverse than one with multiple scaffolds,
we sought to increase the skeletal diversity of the library.20 Initial
studies were performed on model system 29 in solution, which
allowed convenient monitoring of reaction progress by thin-
layer chromatography (Scheme 6). Monomer 24, synthesized
using chemistry analogous to that employed in the synthesis of
10 (see Supporting Information), was exposed to PhI(OAc)2,
yielding 25. Enone 25 was selectively deallylated28 to yield
phenol 26, which was then methylated to yield 27. Methyl ether
27 was then reduced diastereoselectively under Luche condi-
tions29 to generate allylic alcohol 28, which was esterified to
yield acrylic ester 29.
Ester 29 was processed along a branching pathway to produce
two distinct scaffolds depending on which transition-metal-
mediated reaction was employed (Scheme 7). Treatment of 29
with palladium acetate, silver carbonate, and triphenylphosphine
initiated a rearrangement via a π-allyl palladium intermediate
to yield 30, a molecule with one additional ring.30 Alternately,
treatment with the second-generation Grubbs metathesis cata-
lyst31 yielded 31, a molecule with two new rings formed at the
expense of one ring from 29. Furthermore, treatment of 31 with
acid resulted in an additional scaffold modification to yield furan
phenol 32, a molecule that lacks two of the four rings originally
present in 29. From a single precursor (29), three new scaffolds
were created utilizing three different reactions. Unfortunately,
despite considerable effort, all attempts to effect the Luche
reduction of enone 18 in the solid phase failed, which effectively
precluded application of these scaffold diversification reactions
in the solid phase. Possible reasons for this failure include the
deleterious effect that polar solvents have on the swelling
properties of the hydrophobic polystyrene resin and ineffective
diffusion of the metal or hydride source into the resin.
Occasional difficulties with reagent diffusion through these
macrobeads, especially with metallic reagents, have been
Primary amines were used exclusively in our conjugate
additions since a MCR with a primary amine would generate a
secondary amine that could be further functionalized. For
example, addition of chloroformates to resin-bound 33 yielded
carbamate products of which 34 is a representative example
(32) L-selectride, K-selectride, LAH, DIBAL-H, 9-BBN, and BH3 each led to
a complex mixture of products including products of over-reduction and
allyl group cleavage.
(28) Jeffrey, P. D.; McCombie, S. W. J. Org. Chem. 1982, 47, 587-590.
(29) Luche, J. L.; Rodriguez-Hahn, L.; Crabbe, P. J. Chem. Soc., Chem.
Commun. 1978, 601-602.
(30) Oppolzer, W. Angew. Chem., Int. Ed. Engl. 1989, 28, 38-52.
(31) Grubbs, R. H. Tetrahedron 2004, 60, 7117-7140.
(33) A mixture of oxime geometrical isomers was generally observed, in most
cases with a ratio >4:1, presumably in favor of the E isomer where the
oxime substituent is oriented away from the carpanone core.
1
(34) The crude H NMR spectra for all numbered compounds produced in the
solid phase are included in the Supporting Information.
9
J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006 5395

A R T I C L E S
Scheme 8 a
Goess et al.
a Conditions: (a) o-EtONH2‚HCl, Et3N, CH3NH2, p-TsOH, CH3OH, CH2Cl2, 72 h; (b) methyl chloroformate, 2,6-lutidine, THF, 5 h; (c) p-nitrophenyl
chloroformate, 2,6-lutidine, THF, 5 h; (d) piperonylamine, THF, DMSO, 60 °C, 48 h; (e) PhSH, Et3N, THF, 18 h; (f) o-BnONH2‚HCl, Et3N, p-TsOH,
CH3OH, CH2Cl2, 12 h; (g) o-BnONH2‚HCl, Et3N, TMSN3, AcOH, CH3OH, CH2Cl2, 12 h; (h) for 40, dimethylacetylene dicarboxylate, THF, 60 °C, 8 h; for
41, phenylacetylene, CuI, DIPEA, THF, 60 °C, 12 h.
(Scheme 8). Furthermore, conversion of 33 to p-nitrophenyl-
carbamate 35 allowed, after subsequent heating with a primary
amine, access to ureas, such as 36 (Path B).35 Amides and
sulfonamides were also produced by reaction of 33 with the
corresponding acyl and sulfonyl chlorides (see Chart 2 and
Supporting Information). During subsequent preparation of a
small test library prior to synthesis of the full 10,000-membered
library, we observed that the secondary amines bearing the
largest substituents frequently did not undergo complete conver-
sion to the corresponding sulfonamides and ureas, so these
combinations of diversification reactions were excluded from
the final library plan.
We then sought to extend this MCR to sulfur nucleophiles;
however, treatment of 18 with aryl thiols in place of amines
under the conditions developed for production of 33 led to highly
variable results.36 This reaction was not general enough to be
included in the library synthesis. However, the 1,4-addition of
aryl thiols to enones was successfully accomplished in a previous
library,10 so we attempted to apply this reaction to our system.
Treatment of 18 with benzenethiol under basic conditions
cleanly produced thioether 37 (Path C). In a second step, oxime
38 was successfully prepared using an adaptation of the
conditions developed for the synthesis of 33. Every hydroxyl-
amine that proved successful for oxime formation in the amine
MCR was also successful in the thiol series. Aliphatic thiols
could be incorporated if n-BuLi and 2,6-lutidine were used in
place of triethylamine. Successful thiol building blocks are
presented in Chart 2.
Azides have been described as one of the functional groups
that reliably participates in a number of solution-phase organic
reactions.37 We sought to incorporate the azide functional group
into our library in hopes that its well-established solution-phase
reactivities would prove equally reliable in our solid-phase
synthesis. Given the success of using primary amines as
nucleophiles in a MCR with enone 18 (Path A), we investigated
whether azide anion could also act as a nucleophile in the MCR
in hopes of taking advantage of, in subsequent steps, the
reactivity of azides toward alkynes.38 The desired MCR was
successfully achieved (Scheme 8, Path D) to give azide 39.
(37) Kolb, H. C.; Finn, M. G.; Sharpless, K. B. Angew. Chem., Int. Ed. 2001,
40, 2004-2021.
(38) (a) Rostovtsev, V. V.; Green, L. G.; Fokin, V. V.; Sharpless, K. B. Angew.
Chem., Int. Ed. 2002, 41, 2596-2599. (b) Breinbauer, R.; Kohn, M.
ChemBioChem 2003, 4, 1147-1149.
(35) Hutchins, S. M.; Chapman, K. T. Tetrahedron Lett. 1995, 36, 2583-2586.
(36) Depending on the thiol, the outcomes included no reaction, 1,4-thiol
incorporation only, oxime formation only, and the desired MCR. The results
could not be correlated with the structure of the thiol.
9
5396 J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006

Synthesis of Carpanone-like Molecules
A R T I C L E S
Chart 2. Compilation of Building Blocks Used in the Library Synthesis
Every hydroxylamine that proved successful for oxime forma-
tion in the amine MCR was also successful in the azide MCR.
We were then pleased to discover that the thermal cycloaddition
reaction of azide 39 with dimethylacetylene dicarboxylate
yielded triazole 40 and that the copper-catalyzed cycloaddition
reaction of 39 with phenylacetylene yielded 41.39 Successful
acetylene building blocks are presented in Chart 2.
With the enone functional group now sufficiently diversified,
the phenol allyl group was selectively cleaved in the presence
of the ketal allyl group using Pd(PPh3)4 and PhSiH3. Repre-
sentative examples of this reaction are presented in Scheme 9.
With careful control of reaction time, a selective deprotection
could be achieved for all library members except those contain-
ing urea and triazole substructures. In molecules containing these
two substructures, the ketal allyl group was partially removed
as well and could be completely removed after an extended
reaction time.40 To avoid potential complications arising from
having two free alcohols present within any library compound,
final library members containing urea and triazole substructures
were segregated following global deallylation and were not
further diversified.
Reaction development on the free phenols produced from the
selective deprotection reactions was complicated by the growing
size of the potential library. Using all combinations of reactions
and building blocks discussed to this point, over 700 unique
(40) We speculate that the increased coordinating capability of highly nitroge-
nated triazoles and ureas might deliver palladium to the allyl group nearest
these functional groups, thus promoting undesired cleavage of the ketal
allyl group.
(39) For recent insights into the mechanism of this copper-mediated reaction,
see: Himo, F.; Lovell, T.; Hilgraf, R.; Rostovtsev, V. V.; Noodleman, L.;
Sharpless, K. B.; Fokin, V. V. J. Am. Chem. Soc. 2005, 127, 210-216.
9
J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006 5397

A R T I C L E S
Scheme 9 a
Goess et al.
using a “one bead-one stock solution” strategy.17 The library
was prepared in triplicate, using approximately 30,000 macro-
beads (polystyrene, 500 µm, 100 nmol per bead capacity) at
the outset to account for such factors as occasional bead
breakage, removal of a small quantity of beads after each step
for quality control analysis, and the inherent inability of the
split-and-pool method to synthesize 100% of the theoretical
number of compounds.43 After each step, a fraction of the resin
was reserved so that each intermediate in the synthesis would
be represented in the final library of compounds. To facilitate
structure determination of library compounds, we used an
adaptation of the encoding strategy developed by Still. This
strategy involves Rh(II)-catalyzed carbene insertion of diazo-
ketones bearing unique combinations of halogenated aryl groups
(“tags”) into the polystyrene resin following each reaction.44
These tags are oxidatively removed from each resin bead after
compound release and are analyzed by GC. The peaks in the
GC trace indicate the specific tags used to encode the bead’s
reaction history, thus revealing the identity of the compound
attached to that bead.
a Conditions: (a) Pd(PPh3)4, PhSiH3, CH2Cl2, dark, 6 h.
Scheme 10 a
The initial quantity of scaffold-loaded resin (18) was split
into three portions. One portion was divided into 36 batches.
Each batch was tagged with a unique combination of tags and
was treated with one of 36 combinations of six amines and six
hydroxylamines to yield 36 compounds (Scheme 11, Path A).
Rigorous quality control was performed on all 36 batches. A
small amount of each compound was cleaved (HF‚pyridine) and
analyzed by LC-MS. All 36 batches showed >90% purity, and
the GC trace from each bead showed all expected peaks
corresponding to the proper tag combinations for each reaction.
These compounds were pooled, and a portion was removed for
inclusion in the final library. The remainder was split into
batches and tagged prior to treatment with one of 19 acid
chlorides, sulfonyl chlorides, and chloroformates (Paths B and
C). The 24 compounds bearing large amine groups at the R1
position were found to be incompatible with the larger members
of the chloroformate, acid chloride, and sulfonyl chloride
families of building blocks, so these compounds were reacted
with a subset of these building blocks bearing smaller R3 groups
(Path B, see Supporting Information for full details). The 12
compounds bearing small amine groups at the R1 position were
reacted with all 19 building blocks (Path C). A subset of 12 of
these products bearing a p-nitrophenyl chloroformate moiety
was further converted into ureas through treatment with one of
five primary amines (Path D). An additional 576 compounds
were produced overall in this sequence (Paths B, C, and D).
GC analysis on a sample from each batch of resin confirmed
the presence of proper tag combinations.
a Conditions: (a) R3-ArOH, Cu(OAc)2, Et3N, 4 Å m.s., CH2Cl2, 48 h;
(b) PPh3, DIAD, R4-OH, CH2Cl2, dark, 48 h; (c) R5-NCO, Et3N, CH2Cl2,
48 h.
compounds could be generated in a split-and-pool synthesis.
Clearly, further reaction development and building block testing
could be performed efficiently on only a small subset of these
700 compounds. The inability to optimize late-stage reactions
for all members of a split-and-pool synthesis is one drawback
of this synthesis technique. Nonetheless, we felt that careful
optimization of reaction conditions and building blocks for a
small subset of potential library compounds would produce
reactions that could be extended to all library members.
Combined phenols 44 were diversified using three distinct
reactions: (1) copper-catalyzed biaryl coupling with phenols
(44 f 45, Path A, Scheme 10),41 (2) Mitsunobu substitution
with aliphatic alcohols (44 f 46, Path B),42 and (3) carbamate
formation with isocyanates (44 f 47, Path C). The building
blocks that were successfully employed in these reactions are
presented in Chart 2.
A second portion of the initial quantity of scaffold-loaded
resin (18) was split into nine batches and tagged, and each batch
was treated with one of nine thiols (Scheme 12, Path E). Quality
control on each batch indicated that one of the nine products
was not generated with >90% purity, so this batch was removed
from the library. The remaining batches were pooled, a portion
was removed for inclusion in the final library, and the remainder
Split-and-Pool Library Synthesis. Utilizing the above
diversity-generating reactions and building blocks, a final library
of 10,102 molecules was prepared via split-and-pool synthesis
(43) Due to the statistical improbability of splitting any mixture of solid-phase
resin beads into identical heterogeneous batches, a certain percentage of
the theoretical number of compounds in any split-and-pool library will not
be made. This percentage is minimized when multiple copies of the library
are prepared. Three copies of the carpanone library were prepared.
(44) (a) Nestler, H. P.; Bartlett, P. A.; Still, W. C. J. Org. Chem. 1994, 59,
4723-4724. (b) See ref 17c.
(41) For general reviews, see: (a) Whiting, D. A. Oxidative Coupling of Phenols
and Phenol Ethers. In ComprehensiVe Organic Synthesis; Trost, B. M.,
Fleming, I., Pattenden, G., Eds.; Pergamon: Oxford, 1991; Vol. 3, pp 659-
703. (b) Quideau, S.; Feldman, K. S. Tetrahedron 2001, 57, entire issue.
(42) Mitsunobu, O. Synthesis 1981, 1, 1-28.
9
5398 J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006

Synthesis of Carpanone-like Molecules
A R T I C L E S
Scheme 11
Scheme 12
Scheme 13
was split into seven new batches and tagged. Six batches were
treated with one of six hydroxylamines to yield 48 oximes (Path
F). The seventh batch was left unreacted to ensure the presence
of ketones in the final library. GC analysis on a sample from
each batch of resin confirmed the presence of proper tag
combinations.
A third portion of the initial quantity of scaffold-loaded resin
(18) was split into six batches and tagged, and each batch was
treated with trimethylsilyl azide and one of six hydroxylamines
(Scheme 13, Path G). Quality control analyses indicated that
all six products were formed with >90% purity and confirmed
the presence of proper tag combinations. The six batches were
pooled, split into 13 new batches, tagged, and treated with one
of 13 alkynes (Path H). Again, GC analysis on a sample from
each batch of beads confirmed the presence of proper tag
combinations. A total of 716 unique compounds had been
produced at this stage (Paths A-H).
(Scheme 14). A small portion was removed to ensure the
presence of bisallylated structures in the final library. The
remainder was tagged and subjected to deallylation conditions
(Path I). A small portion was then removed to ensure the
presence of free phenols in the final library. The remaining resin
was split into 15 batches and reacted with one of 15 alcohols,
aryl boronic acids, and isocyanates (Path J). In lieu of tagging,
these batches were not pooled at the end of the synthesis. Each
batch was arrayed separately during library formatting,
which allowed the identity of the final reaction performed on
individual beads to be deduced from the location of the bead in
the array.45
Library Formatting and Quality Control. The 30,000
individual macrobeads were then arrayed into 384-well micro-
titer plates such that only one bead was present in each well.
The plates were subjected to an automated cleavage process17
in which each well was treated with 20 µL of an HF‚pyridine
At this stage, all batches of beads, with the exception of
batches bearing urea and triazole substructures, were pooled
(45) This process is known as position encoding.
9
J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006 5399

A R T I C L E S
Scheme 14
Goess et al.
solution (5% HF‚pyridine, 5% pyridine in THF) to remove the
compound from the resin followed by treatment with 20 µL of
methoxytrimethylsilane to quench any excess HF. After evapo-
ration of volatiles, the beads were washed with three successive
30 µL portions of DMF, and the combined eluates were
transferred into fresh 384-well plates. The stock solutions
contained in these “mother plates” were transferred to multiple
sets of “daughter plates” for use in cell-based assays, compound
archives, and quality control analyses. The encoding tags on
the beads were then oxidatively cleaved and subjected to GC
analysis for structure identification.
Quality control was performed on the final library by
analyzing three randomly selected compound solutions from
each of the forty-eight 384-well daughter plates containing the
final library members. In all, 146 compound solutions were
analyzed by LC-MS, which accounts for approximately 1.5%
of the total number of compounds in the library. The results of
these analyses are detailed in the Supporting Information and
are summarized here; 84% of the compounds were successfully
identified by matching the structure inferred from GC tag
decoding to the mass spectrum of the major peak in the LC;
12% of the compounds analyzed had a structure that could be
deduced from GC tag decoding but that could not be correlated
to the major peak in the LC-MS; 4% of the compounds
analyzed had structures that could not be deduced from GC tag
decoding.
Cell-Based Phenotypic Screen. The secretory pathway
involves the transport of properly folded proteins from the
endoplasmic reticulum (ER) through the Golgi apparatus to the
plasma membrane. The study of individual components of this
pathway would be aided if specific small-molecule regulators
of each of its many steps were available. A number of small
molecules have been shown to affect the secretory pathway.
For example, the natural product brefeldin A specifically and
reversibly blocks protein transport out of the ER and causes
tubulation of the Golgi membranes via inhibition of the GTPase
Arf1.46 Secramine, a molecule derived from a recent DOS library
prepared in our laboratory,10 blocks protein transport from the
Golgi apparatus to the plasma membrane, an effect comple-
mentary to that of brefeldin A.2 Other molecules, such as
norrisolide,47 nocodazole,48 and illimaquinone,49 have been
shown to induce fragmentation of the Golgi apparatus. Each of
these molecules has proven useful in studying vesicular traffic
since vesicular traffic is a rapid process that is best perturbed
by fast-acting small molecules.
Through a combination of diversity-oriented synthesis and
high-throughput phenotypic screens, our laboratory intends to
discover additional small molecules that perturb the proper
function of the secretory pathway in order to provide additional
tools for probing this pathway. Utilizing an image-based
phenotypic screen that we developed previously,50 we screened
the carpanone-based library for molecules that perturb exocytic
traffic of membrane proteins from the ER to the plasma
membrane. The screen utilizes the temperature-sensitive ve-
sicular stomatitis virus glycoprotein fused to green fluorescent
protein (VSVGts-GFP). At the nonpermissive temperature of
40 °C, VSVGts-GFP accumulates in the ER. On cooling to 32
°C, VSVGts-GFP synchronously passes from the ER to the
plasma membrane via the Golgi apparatus.51 We screened 8,685
members of the carpanone library in duplicate at ∼22 µM. Each
compound was added to BSC1 cells (monkey kidney epithelial
cells) expressing VSVGts-GFP at 40 °C in 384-well plates.
After 1 hour, the cells were shifted to 32 °C for 3 hours, fixed,
and imaged using an automated fluorescence microscope.
Thirteen compounds were visually identified for perturbing the
intracellular distribution of VSVGts-GFP. The inhibitory activ-
ity of these compounds could be characterized by two distinct
phenotypes (Figure 1A): (1) Golgi exit block, and (2) Golgi
fragmentation. The remaining compounds either had no effect
on trafficking or showed toxic effects exemplified by cell
rounding and/or reduced cell count. In this report, we focus on
the class of compounds that blocks VSVGts-GFP exit from
the Golgi (Figure 1B).
We selected the most potent inhibitors from this class (“hits”)
for resynthesis and further investigation. In addition to preparing
a large quantity (approximately 10 mg) of each hit via solid-
phase synthesis, we prepared samples of intermediates along
each synthesis pathway so that preliminary structure-activity
relationships could be established. Following cleavage from the
resin, each sample was purified using standard flash chroma-
tography prior to biological evaluation.
All members of this class have a secondary amine or
carbamate at R2 and an oxime at R1 (Figure 1C). CLL-19 is the
most potent and displays an IC50 of ∼14 µM. In general,
substitution of an o-fluorobenzyl group for an allyloxy or
hydroxyl group at R3 led to a decrease in inhibitory activity
(Figure 1E). Furthermore, replacement of the secondary amine
(46) (a) Fujiwara, T.; Oda, K.; Yokota, S.; Takatsuki, A.; Ikehara, Y. J. Biol.
Chem. 1988, 263, 18545-18552. (b) See ref 2.
(47) Brady, T. P.; Wallace, E. K.; Kim, S. H.; Guizzunti, G.; Malhotra, V.;
Theodorakis, E. A. Bioorg. Med. Chem. Lett. 2004, 14, 5035-5039.
(48) Rogalski, A. A.; Bergman, J. E.; Singer, S. J. J. Cell Biol. 1984, 99, 1101-
1109.
(50) Feng, Y.; Yu, S.; Lasell, T. K. R.; Jadhav, A. P.; Macia, E.; Chardin, P.;
Melancon, P.; Roth, M.; Mitchison, T.; Kirchhausen, T. Proc. Natl. Acad.
Sci. U.S.A. 2003, 100, 6469-6474.
(51) (a) Presley, J. F.; Cole, N. B.; Schroer, T. A.; Hirschberg, K.; Zaal, K. J.
M.; Lippincott-Schwartz, J. Nature 1997, 389, 81-85. (b) Scales, S. J.;
Pepperkok, R.; Kreis, T. E. Cell 1997, 90, 1137-1148.
(49) Takizawa, P. A.; Yucel, J. K.; Veit, B.; Faulkner, D. J.; Deernick, T.; Soto,
G.; Ellisman, M.; Malhotra, V. Cell 1993, 73, 1079-1090.
9
5400 J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006

Synthesis of Carpanone-like Molecules
A R T I C L E S
Figure 1. High-throughput screening of a library of molecules resembling carpanone for inhibitors of VSVGts-GFP traffic. (A) Statistics of the hits that
scored in the original screen: nine compounds were identified that block VSVGts-GFP exit from the Golgi, and four compounds were identified that cause
Golgi fragmentation. Also, many compounds exhibited nonspecific phenotypes that were mixtures of ER block and Golgi block. (B) Schematic illustrating
the phenotype for the class of compounds that inhibits VSVGts-GFP exit from the Golgi. (C) Representative examples of the hits from the original screen
and their derivatives that were resynthesized. (D) Representative dose response profiles for two of the most potent hits from the library. Also shown is the
inactive carpanone natural product. (E) Structure-activity relationships (SAR) for the various hits at R3. Variations in this series did not significantly perturb
the potency of the compounds, with the hydroxy and allyloxy groups showing the highest potencies compared to that of the o-fluorobenzyloxy group. (F)
SAR of hits varying at position R2. In this series, potency decreases when the secondary amine is replaced with an ethyl carbamate or a benzyloxy ethyl
carbamate. The methyl oxime was least potent regardless of the identity of the R2 group.
at R2 with an ethyl carbamate reduced the inhibitory activity
among all members of this class (compare CLL-23 to CLL-19,
and Figure 1E). Benzyloxyethyl carbamate substitution further
compromised the inhibitory activity compared to the secondary
amine (Figure 1F). Interestingly, carpanone did not perturb
VSVGts-GFP trafficking at 300 µM. Furthermore, the library
core (16) did not affect trafficking and is severely toxic to cells
at concentrations >33 µM (Figure 1C).
We chose to further investigate CLL-19, which inhibits
exocytosis by blocking exit of VSVGts-GFP from the Golgi
with an IC50 of ∼14 µM. To investigate its specificity for
membrane traffic inhibition, we tested the effect of CLL-19 on
other cellular properties. CLL-19 did not inhibit the uptake of
fluorescently labeled cholera toxin from the cell surface to
perinuclear regions (Figure 2A). It also did not inhibit the uptake
of the fluorescently labeled endocytic receptor ligand transferrin
9
J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006 5401

A R T I C L E S
Goess et al.
Figure 2. Hit CLL-19, identified from a screen of a library of molecules resembling carpanone, is a specific inhibitor of VSVGts-GFP traffic. (A) Endocytosis
experiments illustrating that CLL-19 (33 µM) does not have an effect on the uptake of fluorescently labeled cholera toxin (CT-594) or transferrin. Also
shown are immunofluorescence experiments demonstrating that CLL-19 has no effect on the steady-state dynamics of actin or tubulin. (B) Immunofluorescence
experiments probing the effect of CLL-19 on various cellular organelles. Three-hour treatment of 33 µM CLL-19 does not affect the intracellular distribution
of the ER exit sites marker sec13p, and the cis-Golgi marker GM130, suggesting that the architectural integrity of the Golgi is maintained. The KDEL
receptor, which recycles between the ER and the Golgi, appears slightly condensed on treatment with CLL-19 compared to the control, while clathrin
distribution appears normal.
from the plasma membrane to endosomes; however, it did alter
the intracellular distribution of the perinuclear transferrin-
containing endosomes (Figure 2A). Moreover, CLL-19 did not
affect the cellular cytoskeleton as evidenced by the integrity of
both actin and tubulin networks following treatment with CLL-
19. Thus, CLL-19 did not induce general defects in clathrin-
dependent and clathrin-independent intracellular membrane
trafficking or in the organization of the cellular cytoskeleton
and is, therefore, a specific inhibitor of VSVGts-GFP traffick-
ing.
To further characterize the activity of CLL-19, and to better
understand its inhibitory effects on vesicular traffic, we exam-
ined whether CLL-19 altered the steady-state distribution of a
series of different intracellular organelle markers along the
exocytic and endocytic pathways (Figure 2B). Three hour
treatment of BSC1 cells with 33 µM CLL-19 did not affect the
distribution of the sec13p marker specific for the ER exit sites.
CLL-19 also did not affect the localization of the Golgi matrix
protein GM130 (Figure 2B), suggesting that the integrity of the
Golgi apparatus is maintained. The KDEL receptor (KDELr),
which recycles between the Golgi and ER, appeared slightly
more condensed upon treatment with CLL-19, while the
perinuclear signal of the coat protein clathrin was normal. These
results suggest that the Golgi apparatus remains intact upon
treatment with CLL-19, although VSVGts-GFP export from
the Golgi is abrogated.
We next investigated the effect of CLL-19 on cell cycle
progression. We incubated CLL-19 with BSC-1 cells, followed
by fixing and staining for DNA and phosphoHistone 3 antibody
(a marker for cells in the mitosis phase). After a 6 hour
9
5402 J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006

Synthesis of Carpanone-like Molecules
A R T I C L E S
Figure 3. Effect of CLL-19 on cellular division and cell cycle progression. (A) Mitotic index experiments measuring the percent of cells in mitosis after
6 h treatment with 33 µM CLL-19 or 33 µM nocodazole. (B) Flow cytometry experiments showing the decrease in the percent of cells in the G2/M phase
and their increase in the S phase following 6 h treatment with CLL-19. Nocodazole is shown as a control.
treatment, we found that the mitotic index, measured by
percentage of cells in mitosis, decreased from 7.3 to 0.8% on
treatment with 33 µM CLL-19 (Figure 3A). As expected, the
anti-mitotic control compound nocodazole increased the mitotic
index to 16.7% (i.e., more cells were present in mitosis),
indicating that cells were not able to exit the mitotic phase. The
results of cell cycle analysis by flow cytometry allowed a better
understanding of drug effects on cell cycle progression. CLL-
19-treated cell cultures (33 µM) became enriched in S phase
(up 12.2%) after 6 hours. CLL-19-treated cell cultures also
showed a decrease in number of cells in the G2/M phase, in
agreement with the mitotic index assays. These results strongly
suggest that CLL-19 has an effect on growth that may be linked
to partial blockage of dividing cells in S phase, hence slowing
their rate of entry into mitosis (Figure 3B; see drop in number
of cells in the G2/M phase). Further detailed investigations into
the biology of CLL-19 and other hits from the carpanone library
with distinct phenotypes are ongoing and will be reported in
due course.
of the library has identified compounds that perturb the secretory
pathway by inhibiting the exit of proteins from the Golgi
apparatus. This study further underscores how libraries based
upon natural products combined with phenotypic screens can
be useful in identifying molecules with new functions.
Acknowledgment. Financial support from an NSF CAREER
award to M.D.S. and from NIH (P01 GM62566-05) to T.K. is
acknowledged. In support of the ICCB, grants from the NCI,
Merck Research Labs, and Merck KgaA are acknowledged. B.G.
acknowledges an NDSEG graduate research fellowship. R.H.
acknowledges an NSERC postdoctoral research fellowship. We
thank John Tallarico, Jennifer Raggio, and Leticia Castro for
expert supervision of library formatting and quality control;
Ramiro Massol, Werner Boll, and Yan Feng for experimental
assistance and thoughtful discussions; and Yoshiki Tanaka for
resynthesis of two key compounds.
Supporting Information Available: Complete experimental
procedures for the preparation of compounds 8-18 and 24-
47 with characterization data, and detailed procedures for library
Conclusions
1
synthesis and biological screens. H NMR, 13C NMR, and IR
A 10,000-membered library of carpanone-like molecules has
been prepared in a split-and-pool solid-phase synthesis that relied
on adaptations of well-established solution-phase methodologies,
application of known solid-phase methodologies, and develop-
ment of a key multicomponent reaction between nitrogen
nucleophiles, enones, and hydroxylamines. A phenotypic screen
spectra and LR/HRMS data for compounds 8-18, 24-41, CLL-
8, and CLL-19. GC and LC-MS quality control spectra for 8
final library members (PDF). This material is available free of
charge via the Internet at http://pubs.acs.org.
JA056338G
9
J. AM. CHEM. SOC. VOL. 128, NO. 16, 2006 5403