Rational design and simple chemistry yield a superior, neuroprotective HDAC6 inhibitor, tubastatin A

Article abstract of DOI:10.1021/ja102758v

Structure-based drug design combined with homology modeling techniques were used to develop potent inhibitors of HDAC6 that display superior selectivity for the HDAC6 isozyme compared to other inhibitors. These inhibitors can be assembled in a few synthetic steps, and thus are readily scaled up for in vivo studies. An optimized compound from this series, designated Tubastatin A, was tested in primary cortical neuron cultures in which it was found to induce elevated levels of acetylated α-tubulin, but not histone, consistent with its HDAC6 selectivity. Tubastatin A also conferred dose-dependent protection in primary cortical neuron cultures against glutathione depletion-induced oxidative stress. Importantly, when given alone at all concentrations tested, this hydroxamate-containing HDAC6-selective compound displayed no neuronal toxicity, thus, forecasting the potential application of this agent and its analogues to neurodegenerative conditions.

Full text of DOI:10.1021/ja102758v

Published on Web 07/09/2010
Rational Design and Simple Chemistry Yield a Superior,
Neuroprotective HDAC6 Inhibitor, Tubastatin A
Kyle V. Butler,† Jay Kalin,† Camille Brochier,‡ Guilio Vistoli,§ Brett Langley,‡ and
Alan P. Kozikowski*,†
Drug DiscoVery Program, Department of Medicinal Chemistry and Pharmacognosy, UniVersity
of Illinois at Chicago, 833 South Wood, Chicago, Illinois 60612, Burke Medical Research
Institute, 785 Mamaroneck AVenue, White Plains, New York 10605, and Dipartimento di Scienze
Farmaceutiche “Pietro Pratesi”, UniVersita` degli Studi di Milano, Via Mangiagalli 25,
I-20133 Milan, Italy
Received April 2, 2010; E-mail: kozikowa@uic.edu
Abstract: Structure-based drug design combined with homology modeling techniques were used to develop
potent inhibitors of HDAC6 that display superior selectivity for the HDAC6 isozyme compared to other
inhibitors. These inhibitors can be assembled in a few synthetic steps, and thus are readily scaled up for
in vivo studies. An optimized compound from this series, designated Tubastatin A, was tested in primary
cortical neuron cultures in which it was found to induce elevated levels of acetylated R-tubulin, but not
histone, consistent with its HDAC6 selectivity. Tubastatin A also conferred dose-dependent protection in
primary cortical neuron cultures against glutathione depletion-induced oxidative stress. Importantly, when
given alone at all concentrations tested, this hydroxamate-containing HDAC6-selective compound displayed
no neuronal toxicity, thus, forecasting the potential application of this agent and its analogues to
neurodegenerative conditions.
agents.6 The potential toxicities associated with the inhibition
of certain isozymes may lead to additional difficulties for the
Introduction
Protein function can be regulated by the enzymatic addition
and removal of acetyl groups at specific lysine residues. Lysine
acetylation is mediated by two classes of enzymes with opposing
functions: histone acetyltransferase (HAT) and histone deacety-
lase (HDAC), which catalyze the addition and removal of acetyl
groups, respectively.1 The domain of this regulatory mechanism
is vast: mass spectrometry profiling identified 3600 sites on 1750
proteins subject to acetylation.2 HDAC inhibitors (HDACI) have
been aggressively pursued as therapies for cancer and CNS
disorders, and two inhibitors, Vorinostat and Romidepsin, have
been FDA approved for treatment of cutaneous T-cell lym-
phoma.3 HDACIs act on 11 zinc-dependent HDAC isozymes;
their classification and properties have been reviewed elsewhere.4,5
These enzymes are divided into four groups: class I (HDACs
1, 2, 3, 8), class IIa (HDACs 4, 5, 7, 9), class IIb (HDACs 6,
10), and class IV (HDAC11). Most HDACI so far identified
primarily inhibit the class I enzymes, producing an antiprolif-
erative phenotype which is useful for oncology applications,
but unwarranted for the many nononcology applications of these
clinical development of pan-HDAC inhibitors.7-9 Because the
network of cellular effects mediated by acetylation is so vast
and because inhibition of some isozymes may lead to undesir-
able side effects, isozyme selective inhibitors may hold greater
therapeutic promise than their nonselective counterparts.10
HDAC6 has emerged as an attractive target for drug develop-
ment and research.11,12 A diverse set of substrates have been
identified for this enzyme, including R-tubulin, HSP90, perox-
iredoxins, and nuclear histones.13-15 Presently, HDAC6 inhibi-
tion is believed to offer potential therapies for autoimmunity,
cancer, and many neurodegenerative conditions.9,16-18 Selective
inhibition of HDAC6 by small molecule or genetic tools has
(6) Glaser, K. B.; Li, J.; Staver, M. J.; Wei, R. Q.; Albert, D. H.; Davidsen,
S. K. Biochem. Biophys. Res. Commun. 2003, 310, 529–36.
(7) Witt, O.; Deubzer, H. E.; Milde, T.; Oehme, I. Cancer Lett. 2009,
277, 8–21.
(8) Drummond, D. C.; Noble, C. O.; Kirpotin, D. B.; Guo, Z.; Scott, G. K.;
Benz, C. C. Annu. ReV. Pharmacol. Toxicol. 2005, 45, 495–528.
(9) Minucci, S.; Pelicci, P. G. Nat. ReV. Cancer. 2006, 6, 38–51.
(10) Thomas, E. A. Mol. Neurobiol. 2009, 40, 33–45.
(11) Grozinger, C. M.; Hassig, C. A.; Schreiber, S. L. Proc. Natl. Acad.
Sci. U.S.A. 1999, 96, 4868–73.
† University of Illinois at Chicago.
‡ Burke Medical Research Institute.
(12) Boyault, C.; Sadoul, K.; Pabion, M.; Khochbin, S. Oncogene 2007,
26, 5468–76.
§ Universita` degli Studi di Milano.
(1) Xu, W. S.; Parmigiani, R. B.; Marks, P. A. Oncogene 2007, 26, 5541–
52.
(13) Hubbert, C.; Guardiola, A.; Shao, R.; Kawaguchi, Y.; Ito, A.; Nixon,
A.; Yoshida, M.; Wang, X. F.; Yao, T. P. Nature 2002, 417, 455–8.
(14) Kovacs, J. J.; Murphy, P. J.; Gaillard, S.; Zhao, X.; Wu, J. T.; Nicchitta,
C. V.; Yoshida, M.; Toft, D. O.; Pratt, W. B.; Yao, T. P. Mol. Cell
2005, 18, 601–7.
(2) Choudhary, C.; Kumar, C.; Gnad, F.; Nielsen, M. L.; Rehman, M.;
Walther, T. C.; Olsen, J. V.; Mann, M. Science 2009, 325, 834–40.
(3) Kelly, W. K.; Marks, P. A. Nat. Clin. Pract. Oncol. 2005, 2, 150–7.
(4) Dokmanovic, M.; Clarke, C.; Marks, P. A. Mol. Cancer Res. 2007, 5,
981–9.
(15) Wang, Z.; Zang, C.; Cui, K.; Schones, D. E.; Barski, A.; Peng, W.;
Zhao, K. Cell 2009, 138, 1019–31.
(5) de Ruijter, A. J.; van Gennip, A. H.; Caron, H. N.; Kemp, S.; van
Kuilenburg, A. B. Biochem. J. 2003, 370, 737–49.
(16) Wang, L.; de Zoeten, E. F.; Greene, M. I.; Hancock, W. W. Nat. ReV.
Drug DiscoVery 2009, 8, 969–81.
9
10842 J. AM. CHEM. SOC. 2010, 132, 10842–10846
10.1021/ja102758v  2010 American Chemical Society

Tubastatin A, a Superior, Neuroprotective HDAC6 Inhibitor
A R T I C L E S
been demonstrated to promote survival and regrowth of neurons
following injury, offering the possibility for pharmacological
intervention in both CNS injury and neurodegenerative condi-
tions.19 Unlike other histone deacetylases, inhibition of HDAC6
does not appear to be associated with any serious toxicity,
making it an excellent drug target.7 Tubacin, an HDAC6
selective inhibitor, was identified in 2003 by combinatorial
chemistry methods.20 The use of Tubacin in models of disease
has helped to validate, in part, HDAC6 as a drug target, but its
non-drug-like structure, high lipophilicity (ClogP ) 6.36
(KOWWIN)), and tedious synthesis conspire to make it more
useful as a research tool than a drug.21 Other compounds have
been reported to have modest preference for HDAC6.22-24
Encouraged by the possible use of HDAC6 inhibitors as
neuroprotective agents, we initiated a drug design campaign to
identify highly selective and drug-like inhibitors of this enzyme.
We now show how rational drug design was used to generate
an HDAC6 inhibitor with a drug-like structure, simple synthesis,
and superior target selectivity.
Results and Discussion
Homology Modeling. We chose to study selectivity by
comparing HDAC6 against HDAC1, the latter being an
important regulator of cell proliferation and a key oncology
target. Their comparison is most useful, as these two enzymes
have diverse phylogeny and are members of separate deacetylase
classes. Lacking crystal structures for both subtypes, we
generated reliable models for these isozymes by employing
homology techniques.
Figure 1. Top-down view of active site for homology models of HDAC1
(top) and HDAC6 (bottom). Distances between boundaries of the catalytic
channel rim are shown with orange arrows. Letters A-D denote four
boundary areas surrounding the channel rim.
HDAC1 and HDAC6 homology models were generated by
exploiting multiple resolved HDAC crystal structures as tem-
plates, followed by multiple-threading alignments, as imple-
mented in the I-TASSER approach.25 I-TASSER is an auto-
mated bioinformatics tool for predicting protein structure from
amino acid sequence. The catalytic sites of both models were
established by extracting zinc and chelating residues from the
human HDAC8 structure in complex with trichostatin A (PDB:
3FOR) and inserting them into the generated models. Analysis
of the two modeled catalytic pockets revealed that, while the
active site is highly conserved, the dimensions of the catalytic
channel rim differ greatly between the two isozymes. Figure 1
shows four regions, A-D, which represent boundaries of the
catalytic channel rim. Region A corresponds to P32 in HDAC1
and P501 in HDAC6; region B corresponds to L271 and Y204
in HDAC1 and L749 and F679 in HDAC6; region C corre-
sponds to D99 and F205 in HDAC1 and D567 and F680 in
HDAC6; region D corresponds to Q26, G27, M30, and K31 in
HDAC1 and S498, H500, E502, and V503 in HDAC6. Variation
at the D region produces a significantly wider channel rim in
the homology model of HDAC6. The structural manifestation
of this variation is that the measured D-B distance in the energy
minimized conformation is 12.5 Å in HDAC1, compared to 17.5
Å in HDAC6. Following the canonical HDACI structure of
aromatic cap group-linker-hydroxamic acid, we sought to
design compounds that target this area of structural diversity
with a cap group that is both large and rigid enough to
comfortably occupy the rim region of HDAC6, but not HDAC1.
After investigating various polycyclic cap group frameworks,
the carbazole system was found to fit these criteria. When
attached to a zinc-chelating hydroxamic acid by an appropriate
linker, the carbazole nicely fits the D region of HDAC6, whereas
it clashes with the B and D regions in HDAC1 (Figure 3
Supporting Information).
Design of Tricyclic HDACI and Enzyme Inhibition Data.
While lined with conserved residues, the modeled catalytic
channels of HDAC1 and HDAC6 also differ in shape. The
HDAC6 channel appears wider and shallower, suggesting that
replacement of the traditional alkyl chain linker with bulkier
and shorter aromatic moieties might further enhance HDAC6
selectivity. A series of carbazole hydroxamic acids with alkyl
and alkylaryl linker groups were synthesized (Supporting
Information). Inhibitor selectivity was monitored by comparing
IC50 values at HDAC6 against HDAC1 by assays using purified
human HDAC protein. All of the carbazole-based compounds
preferentially inhibited HDAC6 (Table 1). Ease of synthesis
was of paramount importance to our drug design, and most of
these inhibitors can be made in two or three steps. The simple
syntheses allowed for rapid generation of structural analogues.
(17) Dompierre, J. P.; Godin, J. D.; Charrin, B. C.; Cordelieres, F. P.; King,
S. J.; Humbert, S.; Saudou, F. J. Neurosci. 2007, 27, 3571–83.
(18) Kazantsev, A. G.; Thompson, L. M. Nat. ReV. Drug DiscoVery 2008,
7, 854–68.
(19) Rivieccio, M. A.; Brochier, C.; Willis, D. E.; Walker, B. A.;
D’Annibale, M. A.; McLaughlin, K.; Siddiq, A.; Kozikowski, A. P.;
Jaffrey, S. R.; Twiss, J. L.; Ratan, R. R.; Langley, B. Proc. Natl. Acad.
Sci. U.S.A. 2009, 106, 19599–604.
(20) Wong, J. C.; Hong, R.; Schreiber, S. L. J. Am. Chem. Soc. 2003, 125,
5586–7.
(21) Haggarty, S. J.; Koeller, K. M.; Wong, J. C.; Grozinger, C. M.;
Schreiber, S. L. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 4389–94.
(22) Schafer, S.; Saunders, L.; Schlimme, S.; Valkov, V.; Wagner, J. M.;
Kratz, F.; Sippl, W.; Verdin, E.; Jung, M. ChemMedChem 2009, 4,
283–90.
(23) Itoh, Y.; Suzuki, T.; Kouketsu, A.; Suzuki, N.; Maeda, S.; Yoshida,
M.; Nakagawa, H.; Miyata, N. J. Med. Chem. 2007, 50, 5425–38.
(24) Manku, S.; Allan, M.; Nguyen, N.; Ajamian, A.; Rodrigue, J.; Therrien,
E.; Wang, J.; Guo, T.; Rahil, J.; Petschner, A. J.; Nicolescu, A.;
Lefebvre, S.; Li, Z.; Fournel, M.; Besterman, J. M.; Deziel, R.;
Wahhab, A. Bioorg. Med. Chem. Lett. 2009, 19, 1866–70.
(25) Roy, A.; Kucukural, A.; Zhang, Y. Nat. Protoc. 2010, 5, 725–38.
9
J. AM. CHEM. SOC. VOL. 132, NO. 31, 2010 10843

A R T I C L E S
Butler et al.
Table 1. HDAC Inhibition Data for 1-7 and Other Referenced
ity, while decreasing the chain length caused a loss of potency
and selectivity. The optimal linker was the p-tolyl group of 5,
which had an IC50 value of 19 nM at HDAC6 with ap-
proximately 574-fold selectivity over HDAC1. For 5, the angle
between the rotational axis of the tolyl linker and the C(sp3)-N
bond connecting linker and cap is fixed at ∼113°, forcing the
carbazole to lie against the catalytic channel rim, stressing the
reported differences between the two rim regions.
Inhibitorsa
We next examined how modifications to the tricyclic group
influenced activity. As the carbazoles are too lipophilic to be
good drugs, we reduced lipophilicity by disrupting planarity in
the tricyclic ring system and introducing a tertiary amine.
Analogues of 1 (compounds 2-4) were less potent than the
parent, but all preferentially inhibited HDAC6. Analogues of 5
displayed enhanced potency and selectivity. The tetrahydro-γ-
carboline analogue, 6, was superior to the parent in HDAC6
potency and selectivity, with an IC50 of 15 nM at HDAC6 and
over 1000-fold selectivity against HDAC1. The tetrahydro-ꢀ-
carboline, 7, was even better, with an IC50 of 1.4 nM at HDAC6
and approximately 3700-fold selectivity against HDAC1. We
note that modifications to the tricyclic group did not significantly
enhance inhibitory activity when employed in conjunction with
the n-pentyl linker, but greatly enhanced activity and selectivity
when used with the tolyl linker, where the bent conformation
forces tighter interactions between the tricycle and the catalytic
channel rim, giving a greater response to structural changes
made to the tricycle. The enhanced selectivity of the carboline
derivatives may also derive from the presence of the N-methyl
group, as this substituent will further expand the dimensions of
the cap group, further favoring interactions with HDAC6.
The tricyclic inhibitors were compared with other compounds
reported to be highly selective for HDAC6. Tubacin was found
to potently inhibit HDAC6, with an IC50 value of 4 nM and
approximately 350-fold selectivity over HDAC1. The optimized
inhibitors, 6 and 7, are far more selective for HDAC6 than any
other compounds reported in literature. These agents are also
very druggable: ClogP ) 2.41 (KOWWIN) and tPSA ) 57 for
both 6 and 7; AlogPs water solubility ) 45.2 mg/L for 6 and
43.7 mg/L for 7. The tertiary amine group can be used to form
pharmaceutically useful salts, thus, facilitating compound solu-
bilization. Furthermore, their facile, three-step syntheses allow
easy scale-up for in ViVo studies. These compounds confirm
our structure based design approach, as less than 30 compounds
were made to arrive at these findings.
To further validate this compound series, we chose to profile
HDACI 6, hereafter designated as Tubastatin A, against all 11
isozymes, to investigate its ability to induce R-tubulin acetylation
in cells, as well as to profile its neuroprotective action in a cell
model of oxidative stress. This compound was chosen for further
profiling after it was found to have favorable in Vitro ADME
properties. Thus, both Tubastatin A and Tubacin were tested at
all 11 HDAC isoforms (Table 2). Tubastatin A was substantially
more selective than Tubacin at all isozymes except HDAC8
and maintained over 1000-fold selectivity against all isoforms
excluding HDAC8, where it had approximately 57-fold selectiv-
ity. The moderate activity of Tubastatin A at HDAC8 may be
the product of a known conformational change that occurs upon
binding to HDAC8, which dilates the catalytic pocket, to better
accommodate the bulky tricyclic group.27
a Data are shown as IC50 values in µM ( standard deviation. Values
are the mean of two experiments, except TSA, which is a mean of 9
experiments. Compounds were tested in duplicate in a 10-dose IC50
mode with 3-fold serial dilution starting from 30 µM solutions. IC50
values were extracted by curve-fitting the dose/response slopes. TSA
was used as an internal standard. b This isoxazole-based inhibitor
(ISOX) was previously reported by our group to have a low picomolar
IC50 at HDAC6.26 When this compound was tested in the present assays,
a more reasonable value was observed. After investigating the source of
this discrepancy, we found that lack of a detergent (Triton X100) in the
original assay, which was performed by a different company, caused the
anomalously high activity. Screening was performed by Reaction
Biology (Malvern, PA).
Compound 1, featuring carbazole in conjunction with an
n-pentyl linker, was synthesized first and was found to have an
HDAC6 IC50 value of 62 nM with approximately 226-fold
selectivity for HDAC6 over HDAC1. Analogues of 1 with
shorter or longer alkyl chain linkers were synthesized, and it
was found that increasing the alkyl chain length from 5
methylene groups gave increased potency with loss of selectiv-
Selectivity of Tubastatin A was analyzed in the more complex
cellular environment by comparing R-tubulin hyperacetylation,
(26) Kozikowski, A. P.; Tapadar, S.; Luchini, D. N.; Kim, K. H.; Billadeau,
D. D. J. Med. Chem. 2008, 51, 4370–3.
9
10844 J. AM. CHEM. SOC. VOL. 132, NO. 31, 2010

Tubastatin A, a Superior, Neuroprotective HDAC6 Inhibitor
A R T I C L E S
Table 2. Enzyme Inhibition Data for Tubacin and Tubastatin A at
all 11 HDAC Isozymesa
Tubacin
IC50 (µM) ( SD
Tubastatin A
IC50 (µM) ( SD
HDAC1
HDAC2
HDAC3
HDAC4
HDAC5
HDAC6
HDAC7
HDAC8
HDAC9
HDAC10
HDAC11
1.40 ( 0.24
6.27 ( 0.29
1.27 ( 0.16
17.3 ( 2.1
3.35 ( 0.03
0.004 ( 0.001
9.7 ( 1.8
1.27 ( 0.16
4.31 ( 0.34
3.71 ( 0.16
3.79 ( 0.10
16.4 ( 2.6
>30
>30
>30
>30
0.015 ( 0.001
>30
Figure 3. HCA oxidative stress assay: neurons were treated with Tubastatin
A, with or without addition of HCA. Viability was assessed after 24 h by
MTT assay. Yellow bars, Tubastatin A alone; blue bars, Tubastatin A and
HCA. Asterisk (*) indicates significant increase in survival relative to HCA-
treated control, P < 0.01, by two-way ANOVA followed by Bonferroni
post-tests.
0.854 ( 0.040
>30
>30
>30
a Values are the mean of two experiments. Data are shown as IC50
values in µM ( standard deviation. Compounds were tested in duplicate
in 10-dose IC50 mode with 3-fold serial dilution starting from 30 µM
solutions. IC50 values were extracted by curve-fitting the dose/response
slopes.
(Figure 3). This result compares well with results reported by
Parmigiani et al., showing that Tubacin induces R-tubulin
acetylation at 5 µM and protects prostate cancer (LNCaP) cells
from hydrogen peroxide-induced death at 8 µM via peroxire-
doxin acetylation.29 Importantly, when tested alone at all of the
concentrations shown, Tubastatin A exhibited no toxicity,
indicating that neurotoxicity is likely a product of class I HDAC
inhibition, and not a property inherent to hydroxamic acids.
Tubastatin A is the first neuroprotective hydroxamic acid-based
HDACI that we have found, that does not cause neuronal death
when tested alone in the HCA model. These results provide
further evidence that the inhibition of HDAC6 may provide a
means for achieving therapeutic intervention in neurodegen-
erative conditions.
In summation, this work presents compounds that exhibit
outstanding selectivity and potency for the inhibition of HDAC6.
Rational drug design principles combined with homology
modeling have made it possible to create these novel chemical
entities which can be assembled in multigram quantities in a
few days time. The in Vitro enzyme biology is consistent with
cellular activity, and multiple in ViVo studies, which will be the
subject of future reports, demonstrate the superior biological
profile of Tubastatin A in various models of inflammation.
Figure 2. Comparison of histone and R-tubulin hyperacetylation for TSA,
Tubastatin A, and Tubacin.
corresponding to HDAC6 inhibition, with histone hyperacety-
lation, which corresponds to class I HDAC inhibition (Figure
2). Both Tubastatin A and Tubacin preferentially induced
R-tubulin hyperacetylation at 2.5 µM. Slight induction of histone
hyperacetylation was seen for Tubastatin A at 10 µM, which
may reflect inhibition of the limited histone deacetylase activity
of HDAC6 or slight inhibition of class I HDACs at these higher
concentrations.
Experimental Details
Homology and Molecular Modeling. Human HDAC sequences
were obtained from the UniProt database (www.uniprot.org).
Multiple sequence alignments of HDAC isozymes were carried out
using CLUSTALW and BLOSUM matrices for alignment scoring
(www.clustal.org). The HDAC1 and HDAC6 homology models
were generated by exploiting resolved crystal structures as tem-
plates, followed by multiple-threading alignments, as implemented
in the I-TASSER approach (zhanglab.ccmb.med.umich.edu/
I-TASSER/). For HDAC6, modeling focused on the sole second
catalytic subunit (CDII, Gly482-Gly801), as it is the major
functional domain of HDAC6. Following construction of the
models, hydrogen atoms were checked using VEGA (www.
vegazz.net). To remain compatible with physiological pH values,
the side chains of Arg, Lys, Glu, and Asp were ionized, while His
and Cys residues were considered neutral by default. The complete
Analysis of Tubastatin A in a Model of Neurodegeneration.
Previously, it has been shown that inhibition of HDAC6 protects
against neuronal degeneration and stimulates neurite outgrowth
in dorsal root ganglion neurons, thus, suggesting a means for
achieving therapeutic intervention in CNS diseases.19 Tubastatin
A was examined in a model of oxidative stress induced by
homocysteic acid (HCA). This model leads to depletion of
glutathione, the cells major intracellular antioxidant. HDAC6
inhibition rescues neuronal death in this model, possibly by
causing hyperacetylation of peroxiredoxins. Importantly, Tub-
astatin A did not prevent the depletion of glutathione in this
model (as measured by GSH-GloTM Glutathione Assay;
Promega; data not shown). In previous work, we reported that
nonselective, hydroxamic acid HDACIs displayed considerable
toxicity to the primary cortical neurons.28 In HCA-induced
neurodegeneration assays, TSA was moderately neuroprotective
at 0.5 µM, although protection declined at higher concentrations
due to dose-dependent neurotoxicity. Tubastatin A displayed
dose-dependent protection against HCA-induced neuronal cell
death starting at 5 µM with near complete protection at 10 µM
(27) Somoza, J. R.; Skene, R. J.; Katz, B. A.; Mol, C.; Ho, J. D.; Jennings,
A. J.; Luong, C.; Arvai, A.; Buggy, J. J.; Chi, E.; Tang, J.; Sang,
B. C.; Verner, E.; Wynands, R.; Leahy, E. M.; Dougan, D. R.; Snell,
G.; Navre, M.; Knuth, M. W.; Swanson, R. V.; McRee, D. E.; Tari,
L. W. Structure 2004, 12, 1325–34.
(28) Kozikowski, A. P.; Chen, Y.; Gaysin, A.; Chen, B.; D’Annibale, M. A.;
Suto, C. M.; Langley, B. C. J. Med. Chem. 2007, 50, 3054–61.
(29) Parmigiani, R. B.; Xu, W. S.; Venta-Perez, G.; Erdjument-Bromage,
H.; Yaneva, M.; Tempst, P.; Marks, P. A. Proc. Natl. Acad. Sci. U.S.A.
2008, 105, 9633–8.
9
J. AM. CHEM. SOC. VOL. 132, NO. 31, 2010 10845

A R T I C L E S
Butler et al.
models were carefully checked to avoid unphysical occurrences
such as cis peptide bonds, wrong configurations, or colliding side
chains. A preliminary energy minimization was performed on the
models to avoid high-energy interactions, in which the backbone
atoms were fixed to preserve the predicted folding.
To correctly arrange zinc at the catalytic site, metal ion and
chelating residues were extracted from the resolved structure of
human HDAC8 complexed with trichostatin A (PDB: 3FOR) and
inserted into the generated models. This was accomplished by
superimposing the CR carbon atoms of the zinc-coordinating
residues in PDB:3FOR over the corresponding atoms of HDAC1
and HDAC6. All atoms except the zinc ion were then deleted.
Finally, an energy minimization was performed by fixing the
backbone and zinc ion, completing the homology modeling
procedure.
The optimized models were exploited for docking simulations.
The conformational behavior of simulated compounds was inves-
tigated by a MonteCarlo procedure (as implemented in the VEGA
suite of programs which generated 1000 conformers by randomly
rotating the rotors). All geometries obtained were stored and
optimized to avoid high-energy rotamers. The 1000 conformers
were clustered by similarity to discard redundancies; in this analysis,
two geometries were considered nonredundant when they differed
by more than 60° in at least one torsion angle. For each derivative,
the lowest energy structure was then submitted to docking simula-
tions. All minimizations were carried out on a 16 CPU Tyan-VX50
system using Namd2.6 with CHARMM force field and Gasteiger’s
atomic charges.
27 000. The crossover rate was increased to 0.8, and the number
of individuals in each population to 150. All other parameters were
left at the AutoDock default settings.
Neuroprotection Assays. Primary cortical neuron cultures were
obtained from the cerebral cortex of fetal Sprague-Dawley rats
(embryonic day 17) as described previously. All experiments were
initiated 24 h after plating. Under these conditions, the cells are
not susceptible to glutamate-mediated excitotoxicity. For cytotox-
icity studies, cells were rinsed with warm PBS and then placed in
minimum essential medium (Invitrogen) containing 5.5 g/L glucose,
10% fetal calf serum, 2 mM L-glutamine, and 100 µM cystine.
Oxidative stress was induced by the addition of the glutamate
analogue homocysteate (HCA; 5 mM) to the media. HCA was
diluted from 100-fold concentrated solutions that were adjusted to
pH 7.5. In combination with HCA, neurons were treated with either
TSA or Tubastatin A at the indicated concentrations. Viability was
assessed after 24 h by MTT assay (3-(4,5-dimethylthiazol-2-yl)-
2,5-diphenyltetrazolium bromide) method.
Acknowledgment. This work was supported by the National
Institutes of Health [R01 AG022941] (A.P.K.). K.V.B. is supported
by an American Chemical Society division of medicinal chemistry
predoctoral fellowship. We also thank the International Rett
Syndrome Foundation for their support allowing the scale-up and
ADMET studies on Tubastatin A and the laboratory of Stuart
Schreiber at Harvard University, funded by the NIH’s Initiative
for Chemical Genetics, for providing a sample of Tubacin.
Docking calculations were carried out on the HDAC1 and
HDAC6 models by AutoDock4.0. The grid boxes were set to
include all residues within a 15 Å radius sphere around the catalytic
Zn ions, thus, comprising the entire catalytic cavities. In both cases,
the resolution of the grid was 60 × 60 × 60 points with a grid
spacing of 0.450 Å. Each substrate was docked into the grid with
the Lamarckian algorithm as implemented in AutoDock. The ligand
bonds were allowed to freely rotate during docking The genetic-
based algorithm ran 20 simulations per substrate with 2 000 000
energy evaluations and a maximum number of generations of
Supporting Information Available: Experimental details for
synthesis of all compounds shown, synthesis of additional
compounds and their enzyme inhibition data, structures of
referenced compounds, procedures for enzyme inhibition assays,
additional homology modeling figures, TSA neuroprotection
data, and Western blot analysis procedures. This material is
available free of charge via the Internet at http://pubs.acs.org.
JA102758V
9
10846 J. AM. CHEM. SOC. VOL. 132, NO. 31, 2010