Stereoselective synthesis of over two million compounds having structural features both reminiscent of natural products and compatible with miniaturized cell-based assays [25]

Full text of DOI:10.1021/ja981746o

J. Am. Chem. Soc. 1998, 120, 8565-8566
Stereoselective Synthesis of over Two Million
8565
Compounds Having Structural Features Both
Reminiscent of Natural Products and Compatible
with Miniaturized Cell-Based Assays
Derek S. Tan, Michael A. Foley, Matthew D. Shair, and
Stuart L. Schreiber*
Department of Chemistry & Chemical Biology
HarVard Institute of Chemistry & Cell Biology
Howard Hughes Medical Institute
HarVard UniVersity, 12 Oxford St.
Cambridge, Massachusetts 02138
ReceiVed May 19, 1998
Understanding the cellular function of a protein usually requires
a means to alter the function. This is commonly done by mutating
the gene encoding the protein (genetic approach). It can also be
done by binding the protein directly with a small molecule ligand
(chemical genetic approach). Such ligands, primarily natural
products or synthetic variants of them, can either inactivate1 or
activate2 protein function. For the chemical genetic approach to
have its maximal impact, efficient methods of ligand discovery
will be required to provide, in the limit, a small molecule partner
for every gene product.
Figure 1. Synthesis of tetracycles and nitrones. a: X ) H, b: X ) 2-I,
c: X ) 3-I, d: X ) 4-I, e: X ) 4-CF3, f: X ) 3,4-(OMe)2.
Our laboratory has recently described miniaturized cell culture
assays for screening large numbers of small molecule ligands,
potentially on a genome-wide basis.3 However, despite its success
in natural products synthesis, synthetic chemistry has not yet been
applied to the synthesis of vast numbers of compounds with struc-
tures both reminiscent of natural products and compatible with
miniaturized assays. These features of a synthesis will likely be
required in order to discover nonnatural compounds having the
binding affinities and specificities characteristic of natural prod-
ucts.
The synthetic strategy we have undertaken is to develop highly
efficient multistep syntheses of natural product-like compounds
that include several coupling steps and to use split-pool tech-
niques4 at these steps in order to generate diverse outcomes. This
technique requires that one of the coupling substrates be im-
mobilized on a solid support. Our syntheses are further con-
strained by the water-compatible photolabile supports required
when carrying out large numbers (>105) of miniaturized “nano-
droplet” cell culture assays aimed at identifying cell permeable
ligands.3 This presents a considerable synthetic challenge since
the structural elements that provide these properties (e.g., poly-
(ethylene glycol), nitrobenzyl moieties) are incompatible with
numerous reagents used in conventional synthetic chemistry.
With these considerations in mind, we first converted shikimic
acid, 1, into both enantiomers of epoxycyclohexenol carboxylic
acid 2,5 which were then coupled to a photocleavable linker on
solid support by standard methods (Figure 1).6 Treatment of the
resin-bound epoxycyclohexenol, 3, with various nitrone carboxylic
acids,7 7a-f and 9b-d, under esterification conditions yielded
tetracyclic compounds 4a-f and 5b-d with complete regio- and
Figure 2. Compounds derived from templates 4 and 5.
stereoselectivity, presumably via tandem acylation/1,3-dipolar
cycloaddition.8 No intermediate nitrone ester or carboxy isox-
azolidine structures were observed. The iodoaryl tetracycles
4b-d and 5b-d were selected for further study since the support-
bound aryl iodides could be modified with commercially available
reagents, avoiding the need for solution-phase syntheses of
numerous nitrone acids.
Tetracycles 4b-d and 5b-d are rigid, densely functionalized
compounds that can undergo further reactions to introduce a
variety of functional groups around the central octahydrobenzisox-
azole structure, notably, without the use of protecting groups. The
iodoaryl groups can serve as substrates for palladium cross-cou-
pling reactions. The electrophilic lactone and epoxide can react
with nucleophiles while simultaneously unmasking alcohols for
subsequent reactions. Furthermore, reductive N-O bond cleav-
age would provide two additional handles for functionalization.
We developed several such reactions (Figure 2) with product pur-
ity ranging from ∼50 to g98% following photocleavage.9 Since
purification is not practical in a large split-pool synthesis, only
reactions generating products in g90% purity were acceptable.
The most promising reactions were studied in detail to define
the scope, limitations, and optimal conditions for each (Figure
3). Iodobenzyl tetracycles 4b-d were selected as scaffolds since
(1) See http://www-schreiber.chem.harvard.edu on the Web.
(2) Crabtree, G. R.; Schreiber, S. L. Trends in Biochem. Sci. 1996, 21,
418-422.
(3) (a) Borchardt, A.; Liberles, S. D.; Biggar, S. R.; Crabtree, G. R.;
Schreiber, S. L. Chem. Biol. 1997, 4, 961-968. (b) You, A. J.; Jackman, R.
J.; Whitesides, G. M.; Schreiber, S. L. Chem. Biol. 1997, 4, 969-975.
(4) (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.
(5) (a) Wood, H. B.; Ganem, B. J. Am. Chem. Soc. 1990, 112, 8907-
8909. (b) McGowan, D. A.; Berchtold, G. A. J. Org. Chem. 1981, 46, 2381-
2383. (c) Mitsunobu, O. Synthesis 1981, 1-28.
(7) Keirs, D.; Overton, K. Heterocycles 1989, 28, 841-848.
(8) Tamura, O.; Okabe, T.; Yamaguchi, T.; Gotanda, K.; Noe, K.;
Sakamoto, M. Tetrahedron 1995, 51, 107-118.
(6) The resin structure shown represents Tentagel S NH2, a poly(ethylene
glycol)-polystyrene copolymer, loaded with a photocleavable 3-amino-3-o-
nitrophenylpropionic acid linker. (Brown, B. B.; Wagner, D. S.; Geysen, H.
M. Mol. DiV. 1995, 1, 4-12.).
(9) Photocleavage products were analyzed by 1H NMR, FAB-MS, HPLC,
and TLC.
S0002-7863(98)01746-6 CCC: $15.00 © 1998 American Chemical Society
Published on Web 08/08/1998

8566 J. Am. Chem. Soc., Vol. 120, No. 33, 1998
Communications to the Editor
Figure 4. Building blocks used in library synthesis.
Fourth, during the encoded library synthesis below,14 the ef-
fectiveness of carbene tagging was verified by analyzing beads
from every pool at each step of the synthesis.
Figure 3. Synthesis of 24 demonstration compounds and a 2.18 million
compound library.
A large encoded library was constructed by split-pool synthesis
beginning with attachment of two spacers, ꢀ-aminocaproic acid
and glycine, to the photolinker on resin. A third portion of the
resin was left without a spacer. The resin was pooled, divided
into two portions, and one enantiomer of epoxycyclohexenol
carboxylic acid 2 was coupled to each pool. After the resin was
pooled and divided into three portions, iodobenzyl nitrone acids
7b-d were coupled, resulting in a total of 18 tetracyclic scaffolds.
The synthesis was completed by reaction with 30 terminal alkynes,
then with 62 primary amines, and finally with 62 carboxylic acids
(Figures 3 and 4) with an additional skip codon at each step.15
This six-step reaction sequence resulted in a collection of
compounds calculated to contain 2.18 million distinct, spatially
separated, and encoded chemical entities. These synthetic
compounds are rigid, stereochemically defined, and structurally
diverse, characteristics common to many natural products.
Furthermore, the synthesis allows the controlled release of
compounds from the individual 90 µ supports into nanodroplets
containing engineered cells, features critical to the miniaturized
cell-based assays now being used to screen this library for cell
permeable, protein-binding ligands.3 Encouragingly, we have
already found that several members of this library activate a
reporter gene in mink lung cells.16 We note, however, that this
synthesis relies in part on simple acylation chemistry. Syntheses
benefiting from a greater range of mild and selective synthetic
methods may facilitate the routine discovery of compounds with
protein-binding properties rivaling those of natural products.
their palladium-mediated cross-coupling reactions were more
efficient than those involving iodophenyl tetracycles 5b-d. A
carefully controlled alkyne coupling reaction was used to convert
the aryl iodides (10) into aryl alkynes (11).10 Cycloreversion was
a persistent problem in uncatalyzed lactone aminolysis; however,
2-hydroxypyridine mediated the efficient reaction of primary
amines with the γ-butyrolactone to afford γ-hydroxyamides (12).11
After testing numerous acylating conditions, we found that the
unmasked secondary alcohol could be capped reliably with DIPC/
DMAP-activated carboxylic acids to give γ-acyloxy amides (13).12
Split-pool synthesis provides the theoretical means to synthesize
the large numbers of molecules likely required for chemical
genetic screens, where molecules replace mutations. However,
such syntheses present enormous analytical challenges. We have
developed a four-step protocol in order to provide maximum
confidence that a complex split-pool synthesis of encoded
molecules yields the anticipated products in high purity and
efficiency. First, to demonstrate the suitability of the entire
reaction sequence for library synthesis, we synthesized and fully
characterized 24 compounds,13 10a-f through 13a-f, with the
latter providing photocleavage products in acceptable 80-90%
purity. Second, potential building blocks at each step were tested
by reaction with a single selected substrate. Thus, 50 alkynes
were tested by reaction with 10a, 87 amines with 11a, and 98
acids with 12a. Of these, 23 alkynes, 54 amines, and 44 acids
reacted with g90% conversion and purity (LC-MS). These
building blocks, along with a limited number of less optimal
candidates (generally g70% conversion), were selected for
inclusion in library synthesis. Third, to verify effective synthesis
in a split-pool format, we generated a small test library with
building blocks carefully selected such that the products within
each final pool would have unique masses, allowing analysis by
LC-MS. Indeed, all of the expected 456 masses were detected.
Acknowledgment. We thank Dr. A. Tyler, J. Athanasopoulos, and
J. Lynch for expert mass spectral support.
Supporting Information Available: Acknowledgments, additional
discussion, complete experimental procedures, and analytical data for 2,
3, 6b-d, 7b-d, 10-13a-f, building block testing, test library synthesis,
large library synthesis, binary encoding (77 pages, print/PDF). See any
current masthead page for ordering information and Web access
instructions.
(10) (a) Sonogashira, K.; Tohda, Y.; Hagihara, N. Tetrahedron Lett. 1975,
4467-4470. (b) Young, J. K.; Nelson, J. C.; More, J. S. J. Am. Chem. Soc.
1994, 116, 10841-10842. (c) Collini, M. D.; Ellingboe, J. W. Tetrahedron
Lett. 1997, 38, 7963-7966. (d) Odingo, J.; Sharpe, B. A.; Oare, D., Presented
at the 213th National Meeting of the American Chemical Society, San
Francisco, CA, April, 1997; Paper ORGN 574.
JA981746O
(14) Binary encoding was performed by a modified version of the literature
procedure (see Supporting Information): Nestler, H. P.; Bartlett, P. A.; Still,
W. C. J. Org. Chem. 1994, 59, 4723-4724.
(15) Combs, A. P.; Kapoor, T. M.; Feng, S.; Chen, J. K.; Daude´-Snow, L.
F.; Schreiber, S. L. J. Am. Chem. Soc. 1996, 118, 287-288.
(16) Compounds with the general structure 13 in Figure 3, with no spacer,
R1 ) carbon-based substituents attached to an alkyne substituted at the meta
position (see S46 in Supporting Information), R2 ) 3,4-dimethoxybenzyl, and
R3 ) methoxymethyl, were found to activate the reporter plasmid p3TPLux
in a stably transfected mink lung cell line, both independently and synergisti-
cally with TGF-â1 (B. R. Stockwell and S. L. S., unpublished results and
Curr. Biol. 1998, 8, 761-770.).
(11) (a) Rony, P. R. J. Am. Chem. Soc. 1969, 91, 6090-6096. (b)
Openshaw, H. T.; Whittaker, N. J. Chem. Soc. C 1969, 89-91.
(12) (a) Atuegbu, A.; Maclean, D.; Nguyen, C.; Gordon, E. M.; Jacobs, J.
W. Bioorg. Med. Chem. 1996, 4, 1097-1106. (b) Hydroxyl groups introduced
during the alkyne coupling step were also acylated efficiently under these
conditions.
(13) 1H NMR with complete peak assignments by extensive homonuclear
decoupling and/or DQF-COSY, TOCSY, and NOESY experiments, FAB-
MS, HRMS, HPLC, TLC.