Synthesis of 7200 small molecules based on a substructural analysis of the histone deacetylase inhibitors trichostatin and trapoxin.

Article abstract of DOI:10.1021/ol016915f

Seventy-two hundred potential inhibitors of the histone deacetylase (HDAC) enzyme family, based on a 1,3-dioxane diversity structure, were synthesized on polystyrene macrobeads. The compounds were arrayed for biological assays in a "one bead-one stock sol

Full text of DOI:10.1021/ol016915f

ORGANIC
LETTERS
2001
Vol. 3, No. 26
4239-4242
Synthesis of 7200 Small Molecules
Based on a Substructural Analysis of
the Histone Deacetylase Inhibitors
Trichostatin and Trapoxin
Scott M. Sternson, Jason C. Wong, Christina M. Grozinger, and
Stuart L. Schreiber*
Howard Hughes Medical Institute (HHMI), Institute of Chemistry and Cell Biology
(ICCB), Department of Chemistry & Chemical Biology, HarVard UniVersity,
Cambridge, Massachusetts 02138
sls@slsiris.harVard.edu
Received October 16, 2001
ABSTRACT
Seventy-two hundred potential inhibitors of the histone deacetylase (HDAC) enzyme family, based on a 1,3-dioxane diversity structure, were
synthesized on polystyrene macrobeads. The compounds were arrayed for biological assays in a “one bead-one stock solution” format.
Metal-chelating functional groups were used to direct the 1,3-dioxanes to HDAC enzymes, which are zinc hydrolases. Representative structures
from this library were tested for inhibitory activity and the 1,3-dioxane structure was shown to be compatible with HDAC inhibition.
Post-translational modification of proteins through acetylation
and deacetylation of lysine residues has a critical role in
regulating their cellular functions.1 While small molecule
probes for specific protein kinases and phosphatases exist,
probes of histone acetyl transferases and deacetylases are
limited as a result of their lack of selectivity. Here we report
a step toward our goal of identifying selective small molecule
inhibitors of histone deacetylases (HDACs) by preparing
thousands of such compounds using the one bead-one stock
solution format2 for synthesis.
participate in cellular pathways that control cell shape and
differentiation, and an HDAC inhibitor has been shown
effective in treating an otherwise recalcitrant cancer.4 Nine
human HDACs have been characterized5 and two inferred;6
these members fall into two related classes (class I and II).
No small molecules are known that selectively target either
the two classes or individual members of this family.7
(2) (a) Sternson, S. M.; Louca, J. B.; Wong, J. C.; Schreiber, S. L. J.
Am. Chem. Soc. 2001, 123, 1740-1747. (b) Blackwell, H. E.; Pe´rez, L.;
Stavenger, R. A.; Tallarico, J.; Eatough, E.; Foley, M. A.; Schreiber, S. L.,
Chem. Biol. In press. (c) Clemons, P.; Koehler, A.; Wagner, B.; Sprigings,
T.; Spring, D.; King, R.; Schreiber, S.; Foley, M. A., Chem. Biol. In press.
(3) Hassig, C.; Schreiber, S. L. Curr. Opin. Chem. Biol. 1997, 1, 300-
308.
HDACs are zinc hydrolases that modulate gene expression
through deacetylation of the N-acetyl lysine residues of
histone proteins and other transcriptional regulators.3 HDACs
(4) Warrell, Jr., R. P.; He, L.-Z.; Richon, V.; Calleja, E.; Pandolfi, P. P.
J. Natl. Cancer Inst. 1998, 90, 1621-1625.
(1) Kouzarides, T. EMBO J. 2000, 19, 1176-1179.
10.1021/ol016915f CCC: $20.00 © 2001 American Chemical Society
Published on Web 11/30/2001

The natural products trapoxin8 (TPX) and trichostatin A9
(TSA) (Figure 1) are potent inhibitors and useful probes of
acids at the entrance of the N-acetyl lysine binding channel.
The cap and the metal-binding functionality are connected
by a linker, often a 5-6 atom hydrocarbon chain.11 Synthetic
molecules incorporating these substructural elements are
likely to inhibit HDAC enzymes.
Comparison of amino acid sequences around the active
site was used to infer structural differences between indi-
vidual HDAC family members that could be exploited in
the design of selective inhibitors. Most of the amino acids
that contact TSA in the HDLP structure are conserved across
all HDACs. However, this conservation diverges for amino
acids at the solvent-exposed rim of the channel, indicating
that this is a selectivity-determining region (Figure 2). The
Figure 1. HDAC inhibitors trichostatin (TSA) and trapoxin (TPX).
Analysis of these natural products and other HDAC inhibitors reveal
a substructural organization consisting of a metal-binding functional
group, a linker region, and a cap region.
HDACs, but their lack of selectivity among family members
limits their ability to dissect the functions of individual
members. The absence of atomic resolution structures of
human HDACs complicates a structure-based solution to this
problem; therefore, we have undertaken a screening-based
approach. On the basis of sequence alignments of HDACs
and structural analyses of natural HDAC inhibitors (Figure
1), including TPX and TSA, we conceived a synthetic
pathway leading to 7200 1,3-dioxanes, all biased toward
HDAC inhibition. Representative compounds that resulted
from this pathway were shown to inhibit two different human
HDACs; the varied structures of the dioxanes and their
tendency to inhibit HDACs suggest that suitable screening
methods may identify the sought after, specific inhibitors.
A structural rationale for HDAC inhibition is suggested
from the X-ray crystal structure of TSA-bound HDAC-like
protein (HDLP), an HDAC ortholog from the thermophilic
bacterium Aquifex aeolicus.10 In this structure, the hydrox-
amic acid of TSA penetrates a narrow, hydrophobic channel
and chelates a buried zinc ion. The substructural organization
of most HDAC inhibitors can be rationalized in light of the
HDLP structure. Inhibitors typically possess metal-binding
functionality and a cap substructure that interacts with amino
Figure 2. Sequence comparison of residues on the rim of the
N-acetyl lysine binding channel. Amino acids in HDLP that contact
TSA are boxed in gray. The numbering is based on the HDLP
sequence.
most significant sequence differences are observed between
class I and class II HDACs. The sequence diversity in the
rim of the N-acetyl lysine binding channel suggests that
selective inhibitors may be identified from collections of
compounds having varied cap groups, since these groups
would be expected to interact with the rim residues. It is
also of note that an Arg265Pro single nucleotide polymor-
phism has been recently identified in the rim region of
HDAC3,12 providing the structural rationale for polymorph-
specific design of HDAC inhibitors. By synthesizing mol-
ecules that possess diversity elements targeted toward regions
predicted to be structurally divergent, discovery of selective
inhibitors may be possible.
Our synthetic plan (Scheme 1) generates diversity in the
cap region of the small molecules by using the split-pool
synthesis technique. The chain length for the hydrocarbon
linker ranges from 3 to 6 methylene groups so that the
orientation of the cap relative to the enzyme channel is
varied.13 Literature precedence11 and confirmatory HDAC
(5) (a) Taunton, J.; Hassig, C. A.; Schreiber, S. L. Science 1996, 272,
408-411. (b) Yang, W. M.; Yao, Y.; Sun, J.; Davie, J. R.; Seto, E. J. Biol.
Chem. 1997, 272, 28001-28007. (c) Grozinger, C.; Hassig, C.; Schreiber,
S. L. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 4868-4873. (d) Kao, H.;
Downes, M., Ordentlich, P.; Evans, R. M. Genes DeV. 2000, 14, 55-66.
(e) Hu, E.; Chen, Z.; Fredrickson, T.; Zhu, Y.; Kirkpatrick, R.; Zhang, G.;
Johanson, K.; Sung, C.; Liu, R.; Winkler, J. J. Biol. Chem. 2000, 275,
15254-15264. (f) Zhou, X.; Marks, P. A.; Rifkind, R. A.; Richon, V. M.
Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 10572-10577.
(6) Venter, J. C. et al. Science 2001, 291, 1304-1351.
(7) Ortholog-selective HDAC inhibitors have been reported: (a) Meinke,
P. T.; Colletti, S. L.; Doss, G.; Myers, R. W.; Gurnett, A. M.; Dulski, P.
M.; Darkin-Rattray, S. J.; Allocco, J. J.; Galuska, S.; Schmatz, D. M.;
Wyvratt, M. J.; Fisher, M. H. J. Med. Chem. 2000, 14, 4919-4922. (b)
Meinke, P. T.; Liberator, P. Curr. Med. Chem. 2001, 8, 211-235.
(8) Kijima, M.; Yoshida, M.; Sugita, K.; Horinouchi, S.; Beppu, T. J.
Biol. Chem. 1993, 268, 22429-22435.
(11) Jung, M.; Brosch, G.; Ko¨lle, D.; Scherf, H.; Gerha¨user, C.; Loidl,
P. J. Med. Chem. 1999, 42, 4669-4679.
(12) Wolfsberg, T.; McEntyre, J.; Schuler, G. Nature 2001, 409, 824-
826.
(13) Unsaturated linkers would presumably reduce the entropic penalty
of binding this narrow channel. However, because the structural features
required for selective inhibition are not known, the simplest set of linkers
was chosen for use in this initial screening library. Unsaturated linkers will
be tested during optimization of selective inhibitors.
(9) Tsuji, N.; Kobayashi, M.; Nagashima, K.; Wakisaka, Y.; Koizumi,
K. J. Antibiot. 1976, 29, 1-6.
(10) Finnin, M. S.; Donigian, J. R.; Cohen, A.; Richon, V. M.; Rifkind,
R. A.; Marks, P.; Breslow, R.; Pavletich, N. P. Nature 1999, 401, 188-
193.
4240
Org. Lett., Vol. 3, No. 26, 2001

styrene support (1). The resin was pooled, split, encoded for
the subsequent reactions with 50 combinations of six
diazoketones, and reacted with 50 nucleophile building
blocks to generate 100 1,3-diols (2) in high purity (Figure
3). The solid-supported 1,3-diols were pooled and split into
Scheme 1
Figure 3. Percentage of synthetic intermediates that are >90%
pure and >70% pure (LC). Letters refer to reactions labeled in
Scheme 1.
six portions that were reacted with Fmoc-amino dimethyl-
acetal building blocks under HCl catalysis to form 600 Fmoc-
amino-1,3-dioxanes (3). The resin was tagged with six
combinations of three diazoketones to encode the ketalization
reactions. To encode the subsequent reactions, the resin was
pooled and split into four portions and reacted with four
combinations of three diazoketones. After Fmoc removal,
these pools were reacted with TESCl to protect free hy-
droxyls incorporated from the nucleophile building blocks.
The amino-1,3-dioxane resin was reacted with four diacid
building blocks: pyridine-activated glutaric anhydride or
PyBOP-activated monophthalimidomethylester diacids.15 Treat-
ment with hydrazine generated 2400 carboxyamides (4). One-
third of these carboxyamide beads was set aside for screening
experiments. The high purity of these reaction products
indicates that the phthalimidomethylester is well-suited for
carboxylic acid protection in solid-phase organic synthesis,
where ester hydrolysis can be difficult as a result of the poor
aqueous swelling properties of polystyrene resins. The
remaining carboxyamides were split into two portions. One
portion was reacted with diisopropylcarbodiimide (DIC) and
phenylenediamine to generate 2400 o-aminoanilides (6).
Reaction of the remaining portion of resin 4 with O-2-
methoxypropanehydroxylamine in the presence of PyBOP
generated 2400 protected hydroxamic acids (5). The 2-meth-
oxypropane protecting group was essential for this reaction
as O-TBDMS protection gave impure reaction products, and
O-allyl and O-THP protecting groups were not sufficiently
labile to be removed under conditions compatible with every
synthesized compound. Treatment of resin 5 with PPTS
generated 2400 hydroxamic acids (7).
inhibition studies with a model compound led us to avoid
shorter linkers (see below). Three metal-binding function-
alities previously demonstrated to inhibit HDAC1 and likely
adopting different binding orientations were used: carboxylic
acid, o-aminoanilide, and hydroxamic acid. The 1,3-dioxane
is a rigid core that can be synthesized stereoselectively and
with enormous structural diversity.2a
To simplify structure determination of synthetic products,
we used an adaptation2b,c of the encoding strategy reported
by Still and co-workers14 entailing covalent [Rh(OCOCPh3)2]2-
mediated attachment of electrophoric diazoketone tags to
individual, high capacity polystyrene synthesis beads. These
tags can be removed oxidatively after the synthesis and
analyzed by GC.
Two portions of a silane-derivatized polystyrene resin2a
were tagged with two diazoketones, and then enantiomeric
γ,δ-epoxy alcohols were attached to afford modified poly-
The purity of the reaction products at each synthetic step
was determined by LC-MS analysis of the crude material
cleaved from single beads (Figure 3). Consistent with model
(14) Ohmeyer, 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.
(15) Nefkens, G. H.; Tesser, G.; Nivard, R. J. Recueil 1963, 82, 941-
953.
Org. Lett., Vol. 3, No. 26, 2001
4241

molecule microarrays,17 in vitro histone deacetylase inhibition
assays, and cytoblot assays18 using anti-acetyl lysine anti-
bodies. To demonstrate the ability of the synthesized
compounds to inhibit HDAC, we synthesized two compounds
(8 and 9) representative of the molecules in the library and
measured IC50s against HDAC1 and HDAC6 (Table 1).
Calculated IC50s of ∼1 µM were similar to the IC50s of
the substructure 10, indicating that the 1,3-dioxane portion
of the molecules is not detrimental to HDAC inhibition. In
contrast, a compound (11) with a hydrocarbon linker shorter
than those used in the library synthesis weakly inhibited
HDAC1 and HDAC6 (IC50s > 50 µM). A full analysis of
the inhibitory activity of all individual library members
against HDAC1, HDAC4, and HDAC6 is underway.
Table 1. HDAC-Inhibitory Activity (IC50 ( SE, µM)
Acknowledgment. This research was supported by the
NIH (GM38627). S.M.S., J.C.W., and C.M.G. were sup-
ported by NSF predoctoral fellowships, and S.M.S. was
supported by a Roche fellowship. The ICCB is supported
by the NCI, NIGMS, Merck, Merck KGaA, and the Keck
foundation. We acknowledge Dr. Edward Wintner for initial
work toward the 1,3-dioxane scaffold. S.L.S. is an Investiga-
tor at the Howard Hughes Medical Institute at the Department
of Chemistry and Chemical Biology, Harvard University.
compound
HDAC1
HDAC6
8
9
10
1.2 ( 0.5
1.7 ( 1.2
1.5 ( 0.2
0.9 ( 0.2
1.1 ( 0.1
0.38 ( 0.04
reactions,16 the purity of reactions A-C was high. However,
the purity of the o-aminoanilides and hydroxamic acids, while
acceptable, was less than in model studies.16 These results
illustrate the difficulty of optimizing the final steps of a split-
pool synthesis when many synthetic intermediates are
required to react uniformly. Regardless of purity, however,
for every reaction product analyzed (50 out of 50), GC
analysis of the electrophoric tags allowed their structures to
be inferred. In each instance, the mass of the structure
inferred was consistent with the LC-MS data.
The 7200 polystyrene beads were arrayed and cleaved in
the one bead-one stock solution format2 to generate 7200
stock solutions. We are now developing high throughput
assays to test each stock solution for inhibitory activity
against individual HDACs and against cellular HDAC
activity. These include protein-binding assays using small
Note Added after ASAP: This Letter was released ASAP
on 11/30/01 with an error in the structure of compound 4 in
Scheme 1. The print and final Web version (12/05/01) are
correct.
Supporting Information Available: Building blocks
used, experimental procedures, and LC-MS characterization.
This material is available free of charge via the Internet at
http://pubs.acs.org.
OL016915F
(17) MacBeath, G.; Koehler, A. N.; Schreiber, S. L. J. Am. Chem. Soc.
1999, 121, 7967-7968.
(18) Stockwell, B. R.; Haggarty, S. J.; Schreiber, S. L. Chem. Biol. 2000,
7, 275-286.
(16) See Supporting Information for LC traces from model reactions.
4242
Org. Lett., Vol. 3, No. 26, 2001