Structure-guided optimization of small molecules inhibiting human immunodeficiency virus 1 Tat association with the human coactivator p300/CREB binding protein-associated factor

Article abstract of DOI:10.1021/jm070014g

Human immunodeficiency virus 1 (HIV-1) trans-activator Tat recruits the human transcriptional coactivator PCAF (p300/CREB binding protein-associated factor) to facilitate transcription of the integrated HIV-1 provirus. We report here structure-based lead optimization of small-molecule inhibitors that block selectively Tat and PCAF association in cells. Our lead optimization was guided by grand-canonical ensemble simulation of the receptor/lead complex that leads to definition of chemical modifications with improved lead affinity through displacing weakly bound water molecules at the ligand-receptor interface.

Full text of DOI:10.1021/jm070014g

J. Med. Chem. 2007, 50, 2285-2288
2285
Structure-Guided Optimization of Small
Molecules Inhibiting Human Immunodeficiency
Virus 1 Tat Association with the Human
Coactivator p300/CREB Binding
Protein-Associated Factor
Figure 1. Structures of chemical ligands for the PCAF BRD.
improve their potency for cell-based functional validation of
this novel anti-HIV therapy target.
Chongfeng Pan,† Mihaly Mezei,‡ Shiraz Mujtaba,‡
Michaela Muller,‡ Lei Zeng,‡ Jiaming Li,†
Zhiyong Wang,*,† and Ming-Ming Zhou*,‡
Lead optimization often requires not only design and synthesis
of appropriate chemical analogs but also estimation of the
resulting effects of chemical modifications on target binding.
Successful execution of the latter could reduce the burden of
synthesis of a large number of chemical analogs for lead
improvement. Toward this end, we report in this study a protocol
for identifying sites for chemical modifications in a lead that
would lead to an enhanced affinity to a receptor by defining
water molecules weakly bound at the lead-receptor interface.
This strategy involves three steps: (1) identifying water
molecules trapped around a receptor-bound lead; (2) defining
solvation sites near the ligand with low desolvation penalty;
and (3) estimating effects of adding a new chemical group
occupying solvation sites on receptor binding. Such effects are
of the net change in free energy of binding of a ligand between
energy gain in receptor binding upon chemical modifications
and energy penalty for ligand desolvation at the point of the
attachment.
The estimation of such effects are realized by Monte Carlo
simulations of a lead-receptor complex in the grand-canonical
(T, V, µ) ensemble (GCE),8,9 in which a system’s temperature
and volume but not the number of molecules are kept constant.
Besides conventional rotations and translations, the GCE
simulation is also performed with periodical insertions and
deletions of water molecules. This ensures adequate sampling
of isolated cavities that exist between the receptor and the ligand
or inside the receptor, which are otherwise difficult to treat by
simulations in canonical, microcanonical, or isothermal-isobaric
ensembles.10-13 The “cavity biased” variant attempts insertion
of a molecule only if a cavity of an appropriate radius is
found.9,11 Insertion and deletion attempts are accepted with
probabilities Pi and Pd, respectively
Hefei National Laboratory for Physical Science at Microscale and
Department of Chemistry, UniVersity of Science and Technology of
China, Hefei, Anhui 230026, People’s Republic of China, and
Department of Structural and Chemical Biology, Mount Sinai
School of Medicine, New York UniVersity, One GustaVe L. LeVy
Place, New York, New York 10029-6574
ReceiVed January 5, 2007
Abstract: Human immunodeficiency virus 1 (HIV-1) trans-activator
Tat recruits the human transcriptional coactivator PCAF (p300/CREB
binding protein-associated factor) to facilitate transcription of the
integrated HIV-1 provirus. We report here structure-based lead
optimization of small-molecule inhibitors that block selectively Tat and
PCAF association in cells. Our lead optimization was guided by grand-
canonical ensemble simulation of the receptor/lead complex that leads
to definition of chemical modifications with improved lead affinity
through displacing weakly bound water molecules at the ligand-
receptor interface.
The fundamental importance of the human immunodeficiency
virus (HIVa) life cycle in the progression of the AIDS disease
in infected subjects is well recognized.1,2 Because of the
development of resistance to the available anti-HIV drugs due
to mutations in the targeted viral proteins,3 continued viral
production by chronically infected cells leads to HIV-based
immune dysfunction and other disorders, including cancer.
Recent studies suggest that new therapeutic strategies that target
different steps in the viral life cycle could complement the
existing anti-HIV therapies.4 It has been shown that transcrip-
tional activation of the integrated HIV provirus requires a
molecular interaction between lysine 50-acetylated HIV-1 trans-
activator Tat and the bromodomain (BRD) of the human
transcriptional co-activator PCAF (p300/CREB binding protein-
associated factor),5,6 and blocking this interaction leads to a
major reduction in Tat-mediated viral transcription.6 These
findings suggest that the Tat/PCAF recruitment could serve as
a new therapeutic target for intervening HIV-1 replication. To
validate this potential novel anti-HIV therapy target, we seek
to develop small-molecule chemical inhibitors that selectively
block this molecular interaction, which is essential for viral
production. Toward this goal, we have identified from structure-
based NMR screen N1-aryl-propane-1,3-diamine compounds (1
and 2, Figure 1) for the PCAF BRD binding to HIV-1 Tat.7
However, these initial lead compounds require optimization to
Pi ) min{1, PcNav exp[B + (E(rN+1) - E(rN))/kT]/(N + 1)}
cav
N-1
Pd ) min{1, N exp[-B + (E(rN) - E(rN-1))/kT]/P
}
where E(rN) is the potential energy of the system of N molecules
at configuration rN, Pcav is the probability of finding a cavity of
N
specific size, k is the Boltzmann constant, and T is the absolute
temperature. The parameter B controls the final density of the
system. To match experimental conditions, B is tuned to a value
that results in the experimental water density in the outer layers
of the simulation cell, using a recently developed automatic
tuning method.14 The water sites around a ligand are calculated
by extracting water within 4 Å of a ligand.15 At each site,
orientation of a water molecule is an average of those contribut-
ing to that site. Overall, this strategy provides a sensitive probe
of local chemical potential in the vicinity of the ligand.
The simulation defines generic solvation sites,15 mean posi-
tions of water molecules around the ligand, with fractional
occupancy calculated from the number and position of water at
the corresponding sites. Water molecules found trapped at the
* To whom correspondence should be addressed. Phone: 212-659-
8652 (M.-M.Z.); +86-551-3603185 (Z.Y.W.) Fax: 212-849-2456 (M.-
M.Z.); +86-551-3631760 (Z.Y.W.). E-mail: ming-ming.zhou@mssm.edu
(M.-M.Z.); zwang3@ustc.edu.cn (Z.Y.W.).
† University of Science and Technology of China.
‡ Mount Sinai School of Medicine.
a Abbreviations: BRD, bromodomain; CBP, CREB-binding protein;
GCE, grand-canonical ensemble; HIV, human immunodeficiency; NMR,
nuclear magnetic resonance; PCAF, p300/CBP associated factor.
10.1021/jm070014g CCC: $37.00 © 2007 American Chemical Society
Published on Web 04/20/2007

2286 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 10
Letters
ligand/receptor interface are indicative of sites on the ligand
for modifications. Because trapping a water molecule is en-
tropically unfavorable, displacing it by a chemical group may
improve lead affinity to a receptor. During simulation, sur-
roundings of a ligand are examined for sites with high
probability of successful water deletions. Water sitting at such
sites may be displaced by a new functional group tethered to
the ligand without incurring significant desolvation penalty. Such
spots are thus deemed as good candidate sites for chemical
modifications in lead optimization.
Change in binding affinity of a ligand, ∆∆Gbind, can in
principle be calculated as a thermocyle that involves free-energy
simulations on the ligand in free and receptor-bound states.16
However, such simulations require high precision because
∆∆Gbind is typically a small value between two simulated
quantities. To circumvent this problem, we estimated solvation
free energy of a ligand by defining the extent it destabilizes the
surrounding water. Once a region of low desolvation penalty is
identified in the GCE simulation, an atom X is selected on the
ligand in that region, replacement groups are designed, and
effects are evaluated.
We tested this approach on a chemical ligand N1-(4-methyl-
2-nitro-phenyl)-propane-1,3-diamine (2) bound to the BRD of
the co-activator PCAF.5 We selected the most representative
structure of the 20 final NMR structures5 for this study, as
judged by the lowest root-mean-square deviations. We used the
program Simulaid17 to orient and fit the complex in the smallest
enclosing rectangular box to minimize the number of water
molecules needed for the simulation. This box was enlarged in
all three directions to ensure a water layer of a 15 Å thickness.
Figure 2. (A), (B) Ribbon and surface representations of the PCAF
BRD/2 complex structure showing mean positions of trapped water
molecules (W1-W5) identified in the GCE simulation. The deletion
sites are indicated as spheres in B.
Table 1. Number of Water Deletions in the Vicinity of the Ligand
Rc
region
2
3
4
5
6
7
8
9
10
216 149 575 538 418 551 397 400 626
32 20 131 43 272 32 26 37 212
3.5 NP1 in
3.5 NP1 out
We performed GCE simulation of the PCAF BRD/2 complex
using the program MMC17 and the cavity-biased technique,9
with insertions only made into voids. We extended preferential
sampling to insertions and deletions,18 thereby ensuring a
thorough sampling of regions in the vicinity of the ligand. The
simulation on the PCAF BRD/2 complex led to the identification
of five water molecules, three of which are trapped between
the protein and the ligand with significant propensity for deletion
replacement (W3-W5, Figure 2A,B). Notably, one such water
molecule (W3) was found in the acetyl group binding pocket
that is filled by the acetylated lysine 50 (K50ac) of Tat in the
PCAF BRD/Tat-K50ac peptide complex but unoccupied when
bound to the ligand 2. Linking of an acetyl moiety to 2 with an
aliphatic carbon chain has been shown to displace that water
molecule in the acetyl binding pocket, which results in a major
enhancement in binding affinity to the protein (unpublished
results, Z.Y. Wang and M.-M. Zhou).
3.5 NP1 total 248 169 706 581 690 583 423 437 838
3.5 CZ
3.5 CZ
3.5 CZ
in
out
109 87
31 15
139 158 225 298 139 242 221
129 11 266 14 25 33 196
total 140 102 368 169 491 312 164 275 417
4.0 NP1 in
4.0 NP1 out
290 198 771 634 586 675 574 508 608
36 32 159 65 347 73 52 44 275
4.0 NP1 total 326 230 930 701 933 848 626 552 973
4.0 CZ
4.0 CZ
4.0 CZ
in
out
90
31
99
35
365 367 326 414 219 324 403
143 18 335 17 37 37 258
total 221 134 508 385 661 431 256 361 661
The number of successful water deletions in the vicinity of
each ligand derivative is shown in Table 1. For each compound,
we calculated corresponding deletion sites within 3.5 and 4.0
Å of the ligand and for sites between the ligand and the protein
(labeled as “in”), as well as for sites on the solvent exposed
site of the protein (labeled as “out”). Furthermore, sites were
counted in the proximity of the entire ligand (labeled as “NP1”)
or only near the chemical groups involved in the derivatization,
up to and including the anchoring carbon in the aniline ring
(labeled as “CZ”).
Because attachment of heavy atoms to a chemical ligand
likely results in an increase of its surface area, comparison of
deletion counts was done in individual groups. The salient
features that we concluded from GCE calculations with these
hypothetical chemical derivatives are as follows: (i) the -CF3
derivative 3 has better solvation than ligand 1. The latter lacks
the para-methyl group; (ii) of the derivatives that contain -CH2-
OH, -CH2NH2, -CH2SH, and -CH2Cl (4-7) at the para-
methyl group, 5 shows consistently lower water deletion counts
at both 3.5 and 4.0 Å distances for the entire molecule as well
as near the derivation site; and (iii) the compounds with an extra
methyl display solvation strength in the order of -CHClCH3
(8) > -CHBrCH3 (9) . -CHICH3 (10), of which the iodine
derivative 10 is markedly less well solvated than the former
Besides the protein/ligand interface, the para-methyl group
on the aniline ring of 2 was shown to be associated with water
prone for deletion replacement (Figure 2B). Due to its solvent
exposure nature, we reasoned that derivatization at this methyl
group would likely replace the weakly bound water without
incurring major desolvation penalty. To identify appropriate
functional groups for new chemical analogs, we then performed
the GCE simulation for several hypothetical derivatives that
contain the following functional groups at the para-methyl
position in 2: -CF3, -CH2OH, -CH2NH2, -CH2SH, -CH2-
Cl, -CHClCH3, -CHBrCH3, and -CHICH3 (compounds
3-10, as listed in Figure 1 and Table 1). The simulations were
carried out using corresponding structure models that were built
on the template of the PCAF BRD/2 structure. During the
calculations, the torsion angles in the different derivatives and
the orientation of the methyl group were sampled, while the
rest of the structure was held rigid.

Letters
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 10 2287
demonstrating the high selectivity of these PCAF BRD-specific
ligands over other structurally similar proteins. Collectively,
these data confirm our water deletion replacement predictions
by the GCE simulations and support the notion that appropriate
modifications of a lead that displaces its surrounding weakly
trapped water when bound to a receptor can result in an
enhancement of lead binding to the receptor. Furthermore, these
studies highlight the potential of the markedly improved lead 8
as a tool in facilitating our further validation of PCAF BRD/
HIV Tat association as a target for new anti-HIV therapy.
In summary, our study reported here shows that identification
of displaceable bound water molecules near the receptor-bound
ligand that are determined based on the open-ensemble simula-
tion may offer a valuable strategy in guiding chemical modifica-
tions of lead optimization in small-molecule drug design.
Figure 3. Blocking of PCAF BRD/Tat-K50ac binding by 2 or 8. (A)
In vitro inhibition of 2 or 8 for GST-PCAF BRD binding to a
biotinylated Tat-K50ac peptide, as assessed by anti-GST Western blot.
(B) HIV LTR luciferase assay assessing efficacy of 2 or 8 in inhibiting
Tat-mediated HIV transcription in the cell.
Acknowledgment. We thank the National Institutes of
Health (to M.M.Z. and S.M.), the GlaxoSmithKline and the
Campbell Foundations (to M.M.Z.), and National Science
Foundation of China (#30572234 to Z.Y.W.) for financial
support of this work.
two. Notably, the GCE calculated deletion counts between the
entire ligand and the proximity of the para-methyl group are
in accord for most of the compounds, with the exception of 6
(Table 1). In the latter case, the -CH2SH group is better solvated
at the protein/ligand interface than at the entire ligand, but this
gain was countered by its worsened solvation on the solvent-
exposed side.
Supporting Information Available: Experimental details and
procedures. This material is available free of charge via the Internet
at http://pubs.acs.org.
References
(1) Wei, X.; Ghosh, S. K.; Taylor, M. E.; Johnson, V. A.; Emini, E. A.;
Deutsch, P.; Lifson, J. D.; Bonhoeffer, S.; Nowak, M. A.; Hahn, B.
H. Viral dynamics in human immunodeficiency virus type 1 infection.
Nature 1995, 373, 117-122.
(2) Ho, D. D.; Neumann, A. U.; Perelson, A. S.; Chen, W.; Leonard, J.
M.; Markowitz, M. Rapid turnover of plasma virions and CD4
lymphocytes in HIV-1 infection. Nature 1995, 373, 123-126.
(3) Ho, D. D.; Zhang, L. HIV-1 rebound after anti-retroviral therapy.
Nat. Med. 2000, 6 (7), 736-737.
(4) Jeang, K.-T.; Xiao, H.; Rich, E. A. Multifaceted activities of the
HIV-1 transactivator of transcription, Tat. J. Biol. Chem. 1999, 274,
28837-28840.
(5) Mujtaba, S.; He, Y.; Zeng, L.; Farooq, A.; Carlson, J. E.; Ott, M.;
Verdin, E.; Zhou, M. M. Structural basis of lysine-acetylated HIV-1
Tat recognition by PCAF bromodomain. Mol. Cell 2002, 9 (3), 575-
586.
(6) Dorr, A.; Kiermer, V.; Pedal, A.; Rackwitz, H. R.; Henklein, P.;
Schubert, U.; Zhou, M. M.; Verdin, E.; Ott, M. Transcriptional
synergy between Tat and PCAF is dependent on the binding of
acetylated Tat to the PCAF bromodomain. EMBO J. 2002, 21 (11),
2715-2723.
(7) Zeng, L.; Li, J.; Muller, M.; Yan, S.; Mujtaba, S.; Pan, C.; Wang,
Z.; Zhou, M. M. Selective small molecules blocking HIV-1 Tat and
coactivator PCAF association. J. Am. Chem. Soc. 2005, 127 (8),
2376-2377.
(8) Adams, D. J. Grand canonical ensemble Monte Carlo for a Lennard-
Jones fluid. Mol. Phys. 1975, 29, 307-311.
(9) Mezei, M. Grand-canonical ensemble Monte Carlo study of dense
liquid Lennard-Jones, soft spheres and water. Mol. Phys. 1987, 61
(3), 565-582.
(10) Tieleman, D. P.; Berendsen, H. J. A molecular dynamics study of
the pores formed by Escherichia coli OmpF porin in a fully hydrated
palmitoyloleoylphosphatidylcholine bilayer. Biophys. J. 1998, 74 (6),
2786-2801.
(11) Resat, H.; Mezei, M. Grand canonical Monte Carlo simulation of
water positions in crystal hydrates. J. Am. Chem. Soc. 1994, 116,
7451-7452.
To validate the predications of chemical derivations from the
GCE simulations, we chemically synthesized several 2 analogs,
that is, 3, 4, 5, and 8, which were predicted in the aforemen-
tioned GCE simulations to have improved binding affinity for
the PCAF BRD. The corresponding synthesis routes and
conditions for these chemical derivatives are described in detail
in Support Information. We then assessed the ability of these
newly synthesized chemical analogs in blocking PCAF BRD
binding to a Tat-K50ac peptide in a competition assay. In this
assay, inhibition of a ligand against PCAF BRD binding to the
biotinylated Tat-K50ac peptide that was bound to streptavidin
agarose beads was measured in a concentration-dependent
manner. Of these 2 analogs, 3, 4, and 5 showed a modest
increase in the inhibition against PCAF BRD/Tat binding as
compared to 2 (data not shown), whereas the new ligand 8,
N-(3-amino-propyl)-4-(1-chloroethyl)-2-nitrobenzene amine, ex-
hibited an estimated 15-fold improvement of the inhibitory
activity in vitro (Figure 3A).
We further showed that 8 exhibited EC50 of ∼1 µM (as
compared to an estimated EC50 of 10 µM for 2) in its cellular
efficacy on inhibiting Tat-mediated transcription of the viral
promoter in a HIV-1 LTR-luciferase reporter gene assay (Figure
3B). Moreover, our study showed that the PCAF BRD ligands
are selective over other members of the structurally conserved
BRD family,19 including the BRD from the human transcrip-
tional co-activator CBP (CREB binding protein). It has recently
been shown that CBP via its BRD can bind specifically to the
human tumor suppressor p53 at the C-terminal acetylated lysine
382 (K382ac)20 and that this CBP BRD/p53-K382ac interaction
is responsible for p53 acetylation-dependent co-activator recruit-
ment in response to DNA damage or external stress signals, a
step essential for p53-induced transcriptional activation of
cyclin-dependent kinase inhibitor p21 in G1 cell cycle arrest.20,21
While transcriptional activity of p53 can be modulated by CBP
BRD-specific small molecules,22,23 neither 8 nor 2 showed any
effect at 5 µM on p53-induced p21 transcription in response to
DNA damage upon UV treatment (data not shown), thereby
(12) Verdonk, M. L.; Chessari, G.; Cole, J. C.; Hartshorn, M. J.; Murray,
C. W.; Nissink, J. W.; Taylor, R. D.; Taylor, R. Modeling water
molecules in protein-ligand docking using GOLD. J. Med. Chem.
2005, 48 (20), 6504-6515.
(13) Li, Z.; Lazaridis, T. The effect of water displacement on binding
thermodynamics: concanavalin A. J. Phys. Chem. B 2005, 109 (1),
662-670.
(14) Speidel, J. A.; Banfelder, J. R.; Mezei, M. Automatic control of
solvent density in grand canonical ensemble Monte Carlo simulations.
J. Chem. Theory Comput. 2006, 2, 1429-1434.
(15) Mezei, M.; Beveridge, D. L. Generic solvation sites in a crystal. J.
Comput. Chem. 1984, 6, 523-527.

2288 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 10
Letters
(16) Tempe, B. L.; McCammon, J. A. Ligand-receptor interactions. J.
Comput. Chem. 1984, 8, 281-283.
(21) Barlev, N. A.; Liu, L.; Chehab, N. H.; Mansfield, K.; Harris, K. G.;
Halazonetis, T. D.; Berger, S. L. Acetylation of p53 activates
transcription through recruitment of coactivators/histone acetyltrans-
ferases. Mol. Cell 2001, 8, 1243-1254.
(22) Sachchidanand; Resnick-Silverman, L.; Yan, S.; Mutjaba, S.; Liu,
W. J.; Zeng, L.; Manfredi, J. J.; Zhou, M. M. Target structure-based
discovery of small molecules that block human p53 and CREB
binding protein association. Chem. Biol. 2006, 13 (1), 81-90.
(23) Mujtaba, S.; Zeng, L.; Zhou, M. M. Modulating molecular functions
of p53 with small molecules. Cell Cycle 2006, 5 (22), 2575-2578.
(17) Mezei, M. URL: http://inka.mssm.edu/∼mezei.
(18) Owicki, J. C. Computer Modeling of Matter; ACS Symposium Series;
American Chemical Society: Washington, DC, 1978; Vol. 86.
(19) Zeng, L.; Zhou, M. M. Bromodomain: An acetyl-lysine binding
domain. FEBS Lett. 2002, 513 (1), 124-128.
(20) Mujtaba, S.; He, Y.; Zeng, L.; Yan, S.; Plotnikova, O.; Sachchidan-
and; Sanchez, R.; Zeleznik-Le, N. J.; Ronai, Z.; Zhou, M. M.
Structural mechanism of the bromodomain of the coactivator CBP
in p53 transcriptional activation. Mol. Cell 2004, 13 (2), 251-263.
JM070014G