Biophysical probes reveal a "compromise" Nature of the Methyl-lysine Binding Pocket in L3MBTL1

Article abstract of DOI:10.1021/ja110432e

Histone lysine methylation (Kme) encodes essential information modulating many biological processes including gene expression and transcriptional regulation. However, the atomic-level recognition mechanisms of methylated histones by their respective adaptor proteins are still elusive. For instance, it is unclear how L3MBTL1, a methyl-lysine histone code reader, recognizes equally well both mono- and dimethyl marks but ignores unmodified and trimethylated lysine residues. We made use of molecular dynamics (MD) and free energy perturbation (FEP) techniques in order to investigate the energetics and dynamics of the methyl-lysine recognition. Isothermal titration calorimetry (ITC) was employed to experimentally validate the computational findings. Both computational and experimental methods were applied to a set of designed "biophysical" probes that mimic the shape of a single lysine residue and reproduce the binding affinities of cognate histone peptides. Our results suggest that, besides forming favorable interactions, the L3MBTL1 binding pocket energetically penalizes both methylation states and has most probably evolved as a "compromise" that nonoptimally fits to both mono- and dimethyl-lysine marks.(Figure Presented)

Full text of DOI:10.1021/ja110432e

ARTICLE
pubs.acs.org/JACS
Biophysical Probes Reveal a “Compromise” Nature of the
Methyl-lysine Binding Pocket in L3MBTL1
Cen Gao, J. Martin Herold, Dmitri Kireev,* Tim Wigle, Jacqueline L. Norris, and Stephen Frye
Center for Integrative Chemical Biology and Drug Discovery, Division of Medicinal Chemistry and Natural Products, UNC Eshelman
School of Pharmacy, University of North Carolina, Chapel Hill, North Carolina 27599-7363, United States
S Supporting Information
b
ABSTRACT: Histone lysine methylation (Kme) encodes
essential information modulating many biological processes
including gene expression and transcriptional regulation. How-
ever, the atomic-level recognition mechanisms of methylated
histones by their respective adaptor proteins are still elusive. For
instance, it is unclear how L3MBTL1, a methyl-lysine histone
code reader, recognizes equally well both mono- and dimethyl
marks but ignores unmodified and trimethylated lysine residues.
We made use of molecular dynamics (MD) and free energy
perturbation (FEP) techniques in order to investigate the energetics and dynamics of the methyl-lysine recognition. Isothermal
titration calorimetry (ITC) was employed to experimentally validate the computational findings. Both computational and
experimental methods were applied to a set of designed “biophysical” probes that mimic the shape of a single lysine residue and
reproduce the binding affinities of cognate histone peptides. Our results suggest that, besides forming favorable interactions, the
L3MBTL1 binding pocket energetically penalizes both methylation states and has most probably evolved as a “compromise” that
nonoptimally fits to both mono- and dimethyl-lysine marks.
’ INTRODUCTION
require a detailed understanding of how lysine methylation states
are recognized by their respective adaptor proteins. More
specifically, it is unclear how adding a single methyl group to
Kme0 or removing it from Kme3 results in a huge gain in affinity
to a “reader” protein, whereas adding a methyl to Kme1 does not
affect the affinity at all.
Histone lysine-methylation plays a key role in transcriptional
regulation.1 Abnormal methylation patterns may lead to various
pathologies including cancer.2,3 At the molecular level, methyla-
tion marks act via recruitment of protein complexes that mediate
transcriptional activation or repression.4 MBT repeats form the
smallest and the least studied class of “chromatin readers”, i.e.,
protein modules that bind to methyl-lysine marks on histone
tails. Functionally, these proteins localize to chromatin and
regulate transcription by a currently unknown molecular
mechanism.5 For instance, L3MBTL1, a protein that features
three MBT domains, recognizes mono- and dimethyl-lysine
marks on H1.4K26, H3K4, H3K9, H3K27, and H4K20 in vitro
and associates with heterochromatin protein 1 (HP1γ) and the
retinoblastoma protein (Rb).6 These interactions result in a
transcriptionally nonpermissive chromatin structure in vitro
and the negative regulation of multiple genes through the
E2F/Rb oncogenic pathway.7,8 Remarkably, despite the com-
plexity of the L3MBTL1 interaction network, it has been
demonstrated that transcriptional repression by L3MBTL1 relies
upon a single methylation mark (H4K20me1).9 Moreover, the
lower methylation state specificity of L3MBTL1 is exclusively
mediated by the second MBT domain.9
Free energy perturbation (FEP),10 coupled to molecular
dynamics (MD) simulations, is well suited to study energetics
and dynamics at the atomic level. Moreover, valuable additional
information, such as electrostatic or van der Waals (vdW)
contributions to the free energy of binding can also be extracted
from computational simulations. Obtaining this valuable infor-
mation requires a judicious choice of a study system. Ideally, the
potential energetic changes should be confined to the interaction
between the methylated lysine side chain and the “reading”
pocket of L3MBTL1. Although straightforward in silico, this
approach may appear less biologically relevant than previously
reported experimental studies6,7,11 involving 10ꢀ15 residue
histone fragments. However, a substantial body of evidence
demonstrates that such a reductionist hypothesis (that biological
function relies upon a localized pocket-residue interaction) is
applicable in the context of Kme recognition by L3MBTL1. For
instance, it has been shown that the MBT-Kme recognition
occurs in a histone sequence independent manner.6,7,11 More-
over, available X-ray structures of MBT-containing proteins
Although the biological implications of the various histone
methylation states are subject to intense investigations,5 the
atomic-scale energetics and dynamics of Kme recognition by
respective histone code readers remain elusive. Yet, exogenous
modulation of the histone code using small molecule agents will
Received: November 19, 2010
Published: March 23, 2011
r
2011 American Chemical Society
5357
dx.doi.org/10.1021/ja110432e J. Am. Chem. Soc. 2011, 133, 5357–5362
|

Journal of the American Chemical Society
ARTICLE
Table 1. Six Designed Biophysical Probes, Their Binding Affinities to L3MBTL1, and Respective Computed Energetic
Contributionsa
Exp
cf2
FEP
cf2
FEP
cfi
N
ID
R
Kd (μM)
ΔΔG
ΔΔG
isomorph
R0
Cme0
ΔGΔ
1
2
3
4
5
6
UNC587
UNC588
UNC589
UNC590
UNC592
UNC591
Nme0
>10000
104 ( 10
73 ( 19
>10000
91 ( 11
21 ( 2
>2.7
0.00
4.73
10
20
30
40
50
60
ꢀ1.28
0.55
1.95
0.57
3.21
4.70
Nme1
0.00
ꢀ0.30
1.81
Cme1
Nme2
ꢀ0.21
>2.7
Cme2
Nme3
Cme3
N(me)et
pyrrolidine
ꢀ0.08
ꢀ0.94
ꢀ0.65
ꢀ1.82
C(me)et
cyclopentane
a Dissociation constants Kd are measured by ITC and are the average of two independent measurements. Free energy values are expressed in kcal/mol.
Binding free energies calculated by FEP (ΔΔGFcfEP2 are given relative to the experimental binding free energy of compound 2 (ꢀ5.38 kcal/mol). The R0
FEP
cfi
column denotes the neutral isomorphs of compounds 1ꢀ6. ΔΔG is the binding free energy difference between a compound and its neutral isomorph.
Compounds 1 and 4 did not demonstrate any detectable binding in our ITC experiments (see Supplementary ITC data in Supporting Information).
demonstrate a “cavity-insertion” recognition mode, with the
modified lysine side-chain fully buried within the MBT binding
cage,4 while the rest of the histone chain is exposed to the solvent
and forms very low-impact interactions with the reader protein.
Finally, the reductionist hypothesis is also supported by cell-
based data. More specifically, it has been demonstrated that
repressive function of L3MBTL1 relies solely on its second MBT
repeat and does not need its interaction partners (i.e., Rb and
HP1γ).9 From the computational perspective, the use of a
monopeptide-like probe whose interactions would be confined
to the Kme binding pocket is also highly preferred because it
makes it possible to avoid any parasitic interactions that might
occur between L3MBTL1 and a polypeptide probe. Moreover,
the smaller probes will have significant phase space overlaps,
which will result in a better convergence of FEP calculations.12
Here, in line with the above rationale, we investigate the atom-
level energetics of MBT-Kme recognition in a system where
L3MBTL1 interacts with small molecules that mimic a single
lysine residue. Six compounds that we refer to as biophysical
probes were synthesized. The compounds 1ꢀ4 (Table 1; Sup-
porting Information) demonstrate the same affinity profile
toward L3MBTL1 as cognate peptides do. To gain a deeper
understanding of the L3MBTL1 binding pocket architecture, we
have also synthesized and studied two exogenous alkylated lysine
analogues, compounds 5 (N(me)et) and 6 (pyrrolidine). In
order to corroborate computational findings with experimentally
observable data, we measured the affinity of the designed
biophysical probes to L3MBTL1 by means of isothermal titration
calorimetry (ITC).
compounds are consistent with the trend observed for histone
peptides, which supports their use as biophysical probes to study
MBT-Kme recognition. In order to make sure that our probes do
actually bind to the second MBT domain of L3MBTL1, we have
also measured binding of compounds 2, 3, 5, and 6 to the D355A
L3MBTL1 mutant. This mutation has previously been demon-
strated to efficiently switch off any binding of mono- or dimethy-
lated histones6 because of the absence of the main binding
anchor Asp355. Our ITC experiments with the mutant did not
show any measurable binding (Supplementary ITC Figures).
This strongly supports the hypothesis that the studied com-
pounds bind to the lysine binding pocket of the second MBT
domain.
Our experimental affinity data confirm the intriguing struc-
tureꢀactivity relationships (SAR), where the same chemical
modification, i.e. adding or removal of a single methyl group,
in a similar context, may result either in no change in affinity or in
its full loss. More specifically, it is unclear how adding a single
methyl group to Nme0 or removing it from Nme3 results in a
huge gain in affinity to a “reader” protein but does not affect the
affinity when the methyl is added to Nme1. Consequently, we
performed a series of FEP and MD simulations (see Experi-
mental Section and Supplementary Scheme 1a) in order to
provide an atomic-scale structural rationale for these “atypical”
SAR. The computed free energies obtained demonstrate a strong
correlation with ITC results, which justifies the use of FEP for the
energetic analysis of Kme recognition.
The same computational protocol was also used to determine
relative weight of polar and nonpolar contributions to the
binding affinity and to ascertain “preferred” interaction modes
for each compound. To this end, we made use of virtual probes
that represent neutral isomorphs of compounds 1ꢀ6 obtained
by replacing their amino nitrogens with carbon atoms. As
previously demonstrated,13ꢀ15 the binding free energy difference
between an ionizable compound and its respective nonpolar
isomorph (ΔΔGFcfEPi) can be exclusively attributed to polar
interactions, i.e., hydrogen bonding, cationꢀπ, and long-range
ionic interactions (see Experimental Section and Supplementary
Scheme. 1b). Furthermore, the nonpolar contribution to the
difference in affinities of two compounds can be expressed as the
’ RESULTS AND DISCUSSION
At the first step, we studied binding of all six synthesized
compounds to a fragment of L3MBTL1 consisting of three MBT
domains (residues 200ꢀ522) by means of ITC. The results
(Table 1, Supplementary Figures) suggest that the probes, with
exception of compounds 1 (Nme0) and 4 (Nme3), demonstrate
a dose-dependent interaction with affinities in the order pyrro-
lidine > Nme2 ≈ N(me)et ≈ Nme1 . Nme3 ≈ Nme0. The
relative order and absolute binding free energies of the studied
5358
dx.doi.org/10.1021/ja110432e |J. Am. Chem. Soc. 2011, 133, 5357–5362

Journal of the American Chemical Society
ARTICLE
Scheme 1. Free Energy Decomposition for Compounds 1ꢀ6
(as Annotated in Table 1)a
a P/NP refer, respectively, to polar and non-polar contribution to
ΔΔGFEP
.
difference in affinities of the two respective nonpolar isomorphs.
Here, we performed six additional FEP calculations to compute
relative binding affinities of neutral isomorphs 10ꢀ60 to
L3MBTL1 (see Table 1) and used these affinities in the further
energetic and structural analyses.
We first compared data obtained for compounds 2 (Nme1)
and 3 (Nme2). Their respective affinities to L3MBTL1 were
similar in both ITC and FEP experiments. However, binding free
energies of their respective neutral isomorphs 20 and 30 are
significantly different (Scheme 1, Supplementary Scheme 2).
The neutral isomorph 20 binds more tightly than 30 by 1.10 kcal/
mol due to its more efficient nonpolar interactions. Additionally,
we made use of conventional MD simulations to provide a
structural interpretation for the MBT-Nme recognition. Ten
thousand structural snapshots from 20 ns simulations were
clustered and analyzed in order to determine the most repre-
sentative bound ligand conformations. As shown in Figure 1a, the
bound state of compound 2 (Nme1) may be represented by 5
conformations, varying significantly in positions of both the
amino nitrogen and the methyl group. However, despite its high
mobility within the binding pocket, the only methyl group of
compound 2 manages to keep optimal distances of ca. 3.5 Å to
each of three aromatic side chains, Phe379, Trp382 and Tyr386
(see Experimental Section and Figure 2). In contrast, the bound
compound 3 (Nme2) is represented by three quite distinct
orientations of its methyl groups (Figure 1b). The distance
distribution in Figure 2 shows that only one methyl group can
keep an optimal distance to the aromatic side chains, while the
other can significantly diverge beyond the range of optimal vdW
interactions. One of representative conformers of compound 3 is
identical to the X-ray conformation of Kme2 with L3MBTL111
and two others differ only in methyl group orientations while
keeping the amino nitrogen in a position of optimal electrostatic
interaction with Asp355.
Therefore, both structurally and energetically, the identical
binding affinities of 2 (Nme1) and 3 (Nme2) are achieved
through quite different recognition mechanisms (as depicted in
Figure 3). Most remarkably, both methylation forms are non-
optimal MBT binders. For instance, compound 2 (Nme1)
benefits from more efficient vdW interactions and a higher
bound-state mobility, leading to a favorable entropic contribu-
tion. However, it is penalized by suboptimal electrostatic inter-
actions because its positive charge is delocalized between two
hydrogens of which only one can efficiently interact with the
carboxyl group of Asp355. Alternatively, the compound 3
(Nme2) is penalized by less favorable nonpolar interactions
but benefits from a much stronger electrostatic contribution due
to its highly localized positive charge. Consequently, it appears
Figure 1. Representative conformers of compounds 2, 3, and 6 derived
from a 20 ns MD simulation in the Kme binding site of L3MBTL1: (a)
compound 2 (Nme1) demonstrates a high conformational diversity, (b)
compound 3 (Kme2) features 3 distinct orientations of its methyl
groups, (c) the pyrrolidine group of 6 is fixed in the cage allowing only
for the carbon chain flexibility, (d) superposition of the most frequent
orientations of Nme1, Nme2, and pyrrolidine heads.
that the lysine pocket of L3MBTL1 that equally nonoptimally
binds both cognate ligands ( Kme1 and Kme2) has emerged as an
evolutionary
“compromise”
between
two
possible
pocket designs, each of which would tightly bind either Kme1
or Kme2. Energetically, this “compromise” pocket equally pena-
lizes both methylation states, although through different inter-
action forces.
The proposed “compromise” pocket hypothesis is also com-
patible with the extremely weak affinity of compounds 1 (Nme0)
and 4 (Nme3). Indeed, compound 1 (Nme0) lacks all favorable
vdW interactions, from which its closest analogue 2 (Nme1)
benefits. This results in a nonpolar free energy loss of 2.90 kcal/
mol. Furthermore, compound 1 (Nme0) is also penalized by a
weaker electrostatic contribution because its positive charge is
delocalized between three hydrogen atoms, resulting in a loss of
an additional 1.83 kcal/mol. As to the low affinity of compound 4
(Nme3), it has been previously hypothesized that steric repul-
sion is the major force preventing MBT domains from binding
the trimethylated lysine.4 Our decomposition analysis suggests
that while the nonpolar term (0.73 kcal/mol) does penalizes
compound 4, the polar term (1.38 kcal/mol) is even more
important for the decline of its affinity (relative to 3). The above
is in agreement with the observation that a negatively charged
residue is essential for a functional MBT domain, but not
mandatory for a Kme3 binding domain.4 For example, the
Kme3 binding pocket of BPTF, a PHD finger-containing protein,
does not contain any acidic residues; interestingly, a Y17E
5359
dx.doi.org/10.1021/ja110432e |J. Am. Chem. Soc. 2011, 133, 5357–5362

Journal of the American Chemical Society
ARTICLE
Figure 2. Distance distributions between the methyl groups of biophysical probes and aromatic side chains of the binding pocket of L3MBTL1. The
only methyl group of compound 2 (Nme1, red curve) is most of the time in the distance range corresponding to the optimal vdW interactions. Only one
of two methyl groups of compound 3 (Nme2, blue curves) manages to keep an energetically favorable distance to the aromatic side chains.
(pyrrolidine), whose ability to bind MBT domains was established
in our previous studies.16,17 ITC data show that compound 6 is the
most potent of all alkylation states we have examined and binds
tighter than 3 (Nme2) by 0.66 kcal/mol, while the potency of
compound 5 (N(me)et) is comparable to that of 3 (Nme2). Our
free energy decomposition analysis suggests that the calculated gain
in affinity of 6 (pyrrolidine) over 3 (Nme2) (1.51 kcal/mol) is due
to a huge electrostatic contribution of 2.75 kcal/mol. Compound 6
is, however, penalized for its two extra carbons by a nonpolar
penalty of 1.23 kcal/mol compared to 3. On the atomic level, the
significant electrostatics-related advantage of 6 is hard to explain
considering a relatively modest change in partial charge as com-
pared to 3 (Nme2). To further examine this interaction, we
analyzed the conventional MD trajectory to provide a structural
rationale for the apparent electrostatics reward. The analysis
demonstrates that the pyrrolidine ring, slightly withdrawn from
the pocket, has a very low mobility in its bound state and is
constrained to optimally interact with Asp355 (Figure 1d). More-
over, the pyrrolidine ring is parallel to the aromatic side chain of
Trp382 and forms an aromaticꢀaliphatic ring stacking motif,
commonly seen in the Cambridge Structure Database18 and in
some proteins featuring an aromatic binding cage.19
Figure 3. Qualitative recognition model of various lysine methylation
states by MBT domains. The Nme0 state demonstrates weak interac-
tions to all binding site residues, whereas Nme3 is lacking a critical
interaction with Asp355. The Nme1 and Nme2 states display the same
binding affinity but quite different interaction patterns.
To conclude, in this study we proposed that the L3MBTL1
lysine binding pocket has evolved as a “compromise” that
features equally nonoptimal fits to both mono and demethylation
states. This finding may help in understanding the evolutionary
development of the methylation machinery and also provide
guidance to the development of more efficient exogenous
modulators of the histone code.
mutation in the binding pocket led to BPTF preferentially
recognizing Kme2 over Kme3.6
Finally, in line with the finding that the native methylation state
binding is imperfect by design, one might also infer that a tighter
MBT binder is possible. To test this hypothesis, we have synthe-
sizedandstudiedtwoalkylatedlysineanalogues5(N(me)et) and6
5360
dx.doi.org/10.1021/ja110432e |J. Am. Chem. Soc. 2011, 133, 5357–5362

Journal of the American Chemical Society
ARTICLE
’ EXPERIMENTAL SECTION
transformation from compounds 5 (N(me)et) to 6 (pyrrolidine), simula-
tions were performed on a 23 window basis, with three additional windows
in the beginning of the simulation and three in the end.
Synthesis. Detailed synthetic procedures and characterization are
described in the Supporting Information.
For each system, two independent calculations with different random
seeds were performed. Results from both simulations were combined
and subjected to the Bennett acceptance ratio (BAR) analysis33 in order
to compute free energy differences between every two adjacent simula-
tion windows. The block bootstrapping method was used to estimate the
statistical uncertainties.34 In this study, 20 block bootstrapping trials
were conducted, and the standard errors over these trials are reported in
Supplementary Table 2.
Free Energy Decomposition. The binding site of the second
MBT domain of L3MBTL1 features an aromatic cage and an aspartate
residue. Hence, the interaction with cationic lysine-like ligands may
involve ionic, hydrogen bonding, cationꢀπ, and vdW interactions. MD/
FEP simulations make it possible to determine the energetic compo-
nents of the ligandꢀprotein binding. More specifically, virtual nonpolar
isomorphs of the actual ligands have proved a useful tool to decompose
the relative free energy of binding ΔΔGifj into electrostatic and
nonpolar contributions.14,35 The electrostatic term mainly includes
hydrogen bonding, cationꢀπ, and long-range ionic interactions, while
the nonpolar contribution is often referred to as a charge-independent
term, including van der Waals interactions and the hydrophobic effect.
Here we performed six FEP calculations to compute the relative binding
affinities between our polar compounds 1ꢀ6 and their respective
nonpolar isomorphs 10ꢀ60 (e.g., the nonpolar isomorph for Nme3 is
Cme3). These calculations completed the thermodynamic cycle shown
in Supplementary Scheme 1b. Because the nonpolar compounds have an
identical size and shape to their corresponding polar counterparts, the
relative binding free energy between each two neutral isomorphs
Protein and Ligand Structure Preparation for MD and FEP.
We used the second MBT domain (residues 366 to 424) of the X-ray
structure of L3MBTL1 at 2.05 Å resolution (PDB entry 2RJF)11 as a
starting point for all MD/FEP simulations. The protein structure was
prepared using Protein Preparation Wizard from the Maestro software
suite.20 Hydrogen atoms were added to the structure, and the whole
system was minimized using the OPLS force field.21 Three-dimensional
structures of small-molecule ligands were generated with Ligprep,20 and
multiple low energy conformations were saved. To obtain initial binding
poses, all ligands were docked to the protein using Glide’s Standard
Precision (SP) mode.22 These binding poses were visually inspected and
the most reasonable structure was selected as the initial bound structure
for dynamic simulation. No acceptable docking pose was found for
compound 4 (Nme3). Its initial bound structure was obtained by
superimposing it to compound 3 (Nme2), followed by a local mini-
mization. All simulation systems were solvated in an orthorhombic
TIP3P water box with a minimum of 10 Å between the box boundary
and any solute atom.
Molecular Dynamics. Molecular dynamics was employed to
sample the system ensembles during FEP calculation. AMBER99SB
force field23 was used to describe the receptor. This force field has been
proven to efficiently reproduce the cationꢀπ interaction,24 which is
crucial for our model system. It is also recently applied to investigate the
methyl-lysine dependent binding of H3 histone tails to the HP1
chromodomain.14 General Amber Force Field (GAFF)25 was used to
model ligand molecules. Atomic partial charges were derived by
restrained electrostatic potential (RESP)26 method, using geometries
optimized at HF/6-31G* level in Gaussian 03.27
0
0
(ΔΔGi fj in Supplementary Scheme 1b) can be used to estimate the
nonpolar effect during the binding of the charged compounds i and j. On
All MD simulations were conducted using Desmond v2.2.28,29
Particle mesh Ewald (PME) summation was used to treat long-term
interactions.30 The SHAKE algorithm31 was used to restrain all bonds
involving hydrogen atoms. The system was energy-minimized and then
gradually heated to 300 K during 0.5 ns. An additional 1.0 ns simulation
at 300 K was performed to further equilibrate the system. All production
runs were performed with the NPT ensemble. System temperature and
pressure were regulated by Langevin thermostat and barostat.32 The
integration time steps were 1, 1, and 3 fs for, respectively, the bonded
term, the near term (short-range electrostatic and van der Waals
interactions), and the far term (long-range electrostatic interactions).
Free Energy Perturbation. Binding free energy differences
(ΔΔGifj) for two similar ligands i and j that bind to the same protein
were calculated by means of the FEP method according to the thermo-
dynamic cycle shown in Supplementary Scheme 1a. Relative binding
affinities ΔΔGifj were computed by calculating the free energy differ-
ence of two ligands in solution and in protein, where
0
0
the other hand, the difference between ΔΔGi fj and the relative binding
free energy of the polar compounds ΔΔGifj can be considered as the
0
0
electrostatic contribution (ΔΔGifj ꢀ ΔΔGi fj in Supplementary
Scheme 1b). The decomposed free energies are summarized in Scheme 1
and Supplementary Scheme 2.
Representative MD Conformations. In order to provide struc-
tural insights into the Kme recognition mechanism, we selected repre-
sentative ligandꢀprotein conformations from MD simulations. To this
end, we performed conventional 20 ns MD simulations for each
proteinꢀligand complex. System coordinates were saved every 2 ps,
yielding 10,000 snapshots per MD run. The snapshots were then
clustered based on the distances from the ligand’s heavy atoms to the
aromatic residues that constitute the active site. Clustering was per-
formed using Pipeline Pilot v7.5.36 The final representative structures
were obtained by a visual inspection of all cluster centers.
General Procedure for Isothermal Titration Calorimetry
(ITC) Experiments. For the ITC measurements, L3MBTL1 was
extensively dialyzed into ITC buffer (20 mM Tris-HCl, pH 8, 25 mM
NaCl and 2 mM β-mercaptoethanol). Subsequently, the concentration
was spectrometrically established; the extinction coefficient ε for
L3MBTL1 is 90870. The ITC experiments were performed at 25 °C,
using an AutoITC200 microcalorimeter (MicroCal Inc., USA). Experi-
ments were performed by injecting 1.5 μL of a 1 mM solution of the
compounds into a 200 μL sample cell containing 50 μM or 100 μM
L3MBTL1; all weakly binding compounds 2, 3, and 5 were titrated to a
100 μM protein solution, and compounds 1, 4, and 6 were titrated to
50 μM protein. A total of 26 injections were performed with a spacing of
180 s and a reference power of 8 μcal/s. Compounds were dissolved in
ITC buffer at 10 mM and were diluted to 1 mM. A control experiment
for each compound was also performed, and the heat of dilution was
measured by titrating each compound into the buffer alone. The heat of
dilution generated by the compounds was subtracted, and the binding
ΔΔGi f j ¼ ΔGj ꢀ ΔGi ¼ ΔGicofmpjlex ꢀ ΔGsolution
i f j
Alchemical mutations were performed on a dual topology basis,
comprising a series of intermediate states. To speed up the convergence,
soft-core potential was used to scale down the vdW potential. For a
simulation window k, the hybrid Hamiltonian was the sum of Len-
nardꢀJones, electrostatic, and bonded interaction terms:
Hðx, λkÞ ¼ HLJðx, λk,vdwA, λk,vdwBÞ þ Helecðx, λk,elecA, λk,elecB
þ Hbondedðx, λk,bondA, λk,bondB
Þ
Þ
where λk is a coupling parameter between the initial and final Hamiltonians.
Most FEP simulations involved a total of 17 windows, with an 8.0 ns
production run in each window. The λk values used in our calculations are
listed in Supplementary Table 1. To ensure a smooth alchemical
5361
dx.doi.org/10.1021/ja110432e |J. Am. Chem. Soc. 2011, 133, 5357–5362

Journal of the American Chemical Society
ARTICLE
isotherms were plotted and analyzed using Origin Software (MicroCal
Inc., USA). The ITC measurements were fit to a one-site binding model.
(18) Paul, P. Cryst. Eng. 2002, 5, 3–8.
(19) Hopkins, A. L.; Ren, J.; Milton, J.; Hazen, R. J.; Chan, J. H.;
Stuart, D. I.; Stammers, D. K. J. Med. Chem. 2004, 47, 5912–5922.
(20) Maestro Suite; Schrodinger LLC: New York, 2009.
(21) Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J. J. Am. Chem.
Soc. 1996, 118, 11225–11236.
(22) Halgren, T. A.; Murphy, R. B.; Friesner, R. A.; Beard, H. S.;
Frye, L. L.; Pollard, W. T.; Banks, J. L. J. Med. Chem. 2004, 47, 1750–9.
(23) Kini, R. M.; Evans, H. J. J. Biomol. Struct. Dyn. 1991, 9, 475–88.
(24) Felder, C.; Jiang, H.-L.; Zhu, W.-L.; Chen, K.-X.; Silman, I.;
Botti, S. A.; Sussman, J. L. J. Phys. Chem. A 2001, 105, 1326–1333.
(25) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case,
D. A. J. Comput. Chem. 2004, 25, 1157–74.
’ ASSOCIATED CONTENT
S
Supporting Information. Supplementary schemes; sup-
b
plementary tables; binding curves for ITC experiments; experi-
mental procedures for all new compounds. This material is
available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
Ph:(919) 843-8457. Fax:(919) 843-8465. E-mail: dmitri.kireev@
unc.edu.
(26) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Kollmann, P. A. J. Am.
Chem. Soc. 1993, 115, 9620–9631.
(27) Frisch, M. J. et al. Gaussian 03, Revision C.02; Gaussian, Inc.:
Wallingford, CT, 2004.
(28) Desmond Molecular Dynamics System, version 2.2; D. E. Shaw
Research: New York, 2009.
’ ACKNOWLEDGMENT
(29) Bowers, K. J.; Chow, E., Xu, H., Dror, R. O., Eastwood, M. P.,
Gregersen, B. A., Klepeis, J. L., Kolossvꢀary, I., Moraes, M. A., Sacerdoti,
F. D., Salmon, J. K., Shan, Y., Shaw, D. E. Proceedings of the ACM/IEEE
Conference on Supercomputing (SC06); IEEE: Piscataway, NJ, 2006; p 43.
(30) Essmann, U.; Perera, L.; Berkowitz, M.; Darden, T.; Lee, H.;
Pedersen, L. J. Chem. Phys. 1995, 103, 8577–8593.
(31) Andersen, H. J. Comp. Phy.s 1983, 52, 24–34.
(32) Pastor, R. W.; Brooks, B. R.; Szabo, A. Mol. Phys. 1988,
65, 1409–1419.
We thank Structural Genomics Consortium, Toronto for
providing protein constructs. This research was supported by
startup funds for SVF provided by the Carolina Partnership and
by the grants RC1-GM090732-01 from the National Institutes of
Health and the Innovation Award from the University Cancer
Research Fund. We also thank Dr. Harry Stern (University of
Rochester) for helpful discussions.
(33) Bennett, C. H. J. Comp. Phys. 1976, 22, 245–268.
(34) Boyce, S. E.; Mobley, D. L.; Rocklin, G. J.; Graves, A. P.; Dill,
K. A.; Shoichet, B. K. J. Mol. Biol. 2009, 394, 747–763.
(35) Gallivan, J. P.; Dougherty, D. A. Proc. Natl. Acad. Sci. U.S.A.
1999, 96, 9459–9464.
’ REFERENCES
(1) Zhang, Y.; Reinberg, D. Genes Dev. 2001, 15, 2343–2360.
(2) Rivenbark, A. G.; Coleman, W. B.; Stahl, B. D. FASEB J. 2009,
23, 38.1.
(36) Pipeline Pilot, ver. 7.5; Accelrys Software Inc.: San Diego, 2009.
(3) Kondo, Y.; Shen, L.; Issa, J.-P. J. Mol. Cell. Biol. 2003,
23, 206–215.
(4) Taverna, S. D.; Li, H.; Ruthenburg, A. J.; Allis, C. D.; Patel, D. J.
Nat. Struct. Mol. Biol. 2007, 14, 1025–1040.
(5) Bonasio, R.; Lecona, E.; Reinberg, D. Semin. Cell Dev. Biol. 2009,
21, 221–230.
(6) Li, H.; Fischle, W.; Wang, W.; Duncan, E. M.; Liang, L.;
Murakami-Ishibe, S.; Allis, C. D.; Patel, D. J. Mol. Cell 2007,
28, 677–691.
(7) Trojer, P.; Li, G.; Sims, R. J., 3rd; Vaquero, A.; Kalakonda, N.;
Boccuni, P.; Lee, D.; Erdjument-Bromage, H.; Tempst, P.; Nimer, S. D.;
Wang, Y. H.; Reinberg, D. Cell 2007, 129, 915–928.
(8) Nevins, J. R. Hum. Mol. Genet. 2001, 10, 699–703.
(9) Kalakonda, N.; Fischle, W.; Boccuni, P.; Gurvich, N.; Hoya-
Arias, R.; Zhao, X.; Miyata, Y.; Macgrogan, D.; Zhang, J.; Sims, J. K.;
Rice, J. C.; Nimer, S. D. Oncogene 2008, 27, 4293–4304.
(10) Chipot, C.; Pohorille, A. Free Energy Calculations: Theory and
Applications in Chemistry and Biology, study ed.; Springer: New York,
2007.
(11) Min, J.; Allali-Hassani, A.; Nady, N.; Qi, C.; Ouyang, H.; Liu, Y.;
MacKenzie, F.; Vedadi, M.; Arrowsmith, C. H. Nat. Struct. Mol. Biol.
2007, 14, 1229–1230.
(12) Pohorille, A.; Jarzynski, C.; Chipot, C. J. Phys. Chem. B 2011,
114, 10235–10253.
(13) Ma, J. C.; Dougherty, D. A. Chem. Rev. 1997, 97, 1303–1324.
(14) Lu, Z.; Lai, J.; Zhang, Y. J. Am. Chem. Soc. 2009,
131, 14928–14931.
(15) Hughes, R. M.; Wiggins, K. R.; Khorasanizadeh, S.; Waters,
M. L. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 11184–11188.
(16) Kireev, D.; Wigle, T. J.; Norris-Drouin, J.; Herold, J. M.; Janzen,
W. P.; Frye, S. V. J. Med. Chem. 2010, 53, 7625–7631.
(17) Herold, J. M.; Wigle, T. J.; Norris, J. L.; Lam, R.; Korboukh,
V. K.; Gao, C.; Ingerman, L. A.; Kireev, D. B.; Senisterra, G.; Vedadi, M.;
Tripathy, A.; Brown, P. J.; Arrowsmith, C. H.; Jin, J.; Janzen, W. P.; Frye,
S. V. J. Med. Chem. 2011, in press.
5362
dx.doi.org/10.1021/ja110432e |J. Am. Chem. Soc. 2011, 133, 5357–5362