Content uploaded by Wai Keat Yam
Author content
All content in this area was uploaded by Wai Keat Yam on Dec 30, 2014
Content may be subject to copyright.
Renement of a Low-resolution Crystal
Structure to Better Understand Erythromycin
Interactions on Large Ribosomal Subunit
http://www.jbsdonline.com
Abstract
Macrolides are a group of diverse class of naturally occurring and synthetic antibiotics made
of macrocyclic-lactone ring carrying one or more sugar moieties linked to various atoms
of the lactone ring. These macrolides selectively bind to a single high afnity site on the
prokaryotic 50S ribosomal subunit, making them highly effective towards a wide range of
bacterial pathogens. The understanding of binding between macrolides and ribosome serves
a good basis in elucidating how they work at the molecular level and these ndings would be
important in rational drug design. Here, we report renement of reconstructed PDB structure
of erythromycin-ribosome system using molecular dynamics (MD) simulation. Interesting
ndings were observed in this renement stage that could improve the understanding of the
binding of erythromycin A (ERYA) onto the 50S subunit. The results showed ERYA was
highly hydrated and water molecules were found to be important in bridging hydrogen bond at
the binding pocket during the simulation time. ERYA binding to ribosome was also strength-
ened by hydrogen bond network and hydrophobic interactions between the antibiotic and the
ribosome. Our MD simulation also demonstrated direct interaction of ERYA with Domains
II, V and with C1773 (U1782EC), a residue in Domain IV that has yet been described of its
role in ERYA binding. It is hoped that this renement will serve as a starting model for a
further enhancement of our understanding towards the binding of ERYA to ribosome.
Key words: Molecular dynamics simulation; Renement; Erythromycin A; 23S rRNA;
Large ribosomal subunit; and Water-bridged hydrogen bond.
Introduction
The macrolides are clinically well-established and highly prescribed antibiotics,
ranging from the erythromycins to the newer analogues such as ketolides deriva-
tized in numerous ways to improve their pharmacological properties (1). These
antibiotics which consist of a 12- to 16- membered lactone ring, to which one or
more sugar substituents are attached, mainly target exclusively at 50S prokaryotic
ribosomal subunit. Biochemical and genetic data have revealed that the macrolides
interacted with the loop of helix 35 in Domain II of the bacterial 23S ribosomal
RNA (rRNA) and with the peptidyl transferase center (PTC) in Domain V (2-7).
The X-ray structure (8) also conrmed similar location for the binding site for mac-
rolides (erythromycin, clarithromycin, and roxithromycin), chloramphenicol, and
clindamycin where the binding site was found to be near to the PTC, specically to
nucleotides in the Domain V of 23S rRNA. The crystallographic study also showed
that the binding of the antibiotics has no signicant interactions with nearby ribo-
somal proteins and did not cause major conformational change to the PTC (8).
Erythromycin A (ERYA), produced from Saccaropolyspora erythraea is the rst
macrolide antibiotic introduced in 1952 (9). It is often used to treat patients who are
Journal of Biomolecular Structure &
Dynamics, ISSN 0739-1102
Volume 26, Issue Number 1, (2008)
©Adenine Press (2008)
Habibah A. Wahab
1,2,*
Wai Keat Yam
1,2
Mohd-Razip Samian
3
Nazalan Najimudin
3
1
Pharmaceutical Design and Simulation
(PhDS) Laboratory
2
School of Pharmaceutical Sciences
3
School of Biological Sciences
Universiti Sains Malaysia
11800 Minden, Pulau Pinang, Malaysia
131
*
Phone: 604 653 2212
Fax: 604 657 0017
Email: habibahw@usm.my
132
Wahab et al.
allergic to penicillin. Despite the discovery and the use of ERYA since half a cen-
tury ago, it is still one of the most important antimicrobial agent as it is effective to-
wards many common bacterial pathogens and some non-typical pathogens (10-12).
ERYA has a 14-membered lactone ring (L), attached with a desosamine sugar (D)
and a cladinose sugar (C) extended from C-5 and C-3 position of the lactone ring,
respectively (Figure 1). ERYA showed high specicity and afnity for the bacterial
50S subunit with K
d
~ 10
-8
M (3, 7, 13-16). However, the actual binding mode of
ERYA or its inhibition mode to the bioactivity of ribosome is merely understood to
the most general idea. A comprehensive understanding of the drug binding sites is
therefore needed to understand the mechanisms of drug action, and hence enabling
a rational approach for antibiotic development.
Ribosomes are macromolecules that are responsible for the production of protein
chains. They are complex structures made of two subunits; the large and small
subunits, and both subunits comprised enormous amount of rRNA and ribosomal
protein. Both subunits have different features in strengthen the transition between
various catalytic sites that leads to efcient protein biosynthesis process. The small
subunit is responsible for the formation of initiation complex and contains messen-
ger decoding site. On the other hand, the large subunit contains PTC that catalyzes
peptide bond formation and provides exit tunnel for nascent polypeptide chain to
leave ribosome. Due to its prominent role in this important bioprocess and the fact
that prokaryotic ribosome differ signicantly from the eukaryotic counterpart, ribo-
some has become favourite target for many natural or synthetic antibiotics to inhibit
the bacterial protein biosynthesis process, thus suppressing the bacterial growth.
The availability of X-ray structures of ribosome (8, 17-28) has opened up the possi-
bility of performing computational calculation like molecular dynamics (MD) sim-
ulations for further understanding of the structure and functions of ribosome as well
to design novel inhibitors. Yet, seven years after the rst ribosome structure being
elucidated at the atomic level using X-ray crystallography, very few MD simula-
tions on ribosome have been carried out as compared to other biological molecular
system such as protein, nucleic acid, lipid bilayer, and so forth. One possible ex-
planation for this is the enormous structure of ribosome could easily yield a system
containing massive number of atoms that is prohibitive to be simulated on a typi-
cal workstation. Nevertheless, MD simulation of 70S ribosome was successfully
done by Trylska et al. (29) in 2005. In their work, the coarse-grained method was
adapted as an approach to shrink the simulation size to a reasonably size, feasible
enough to be simulated to half a microsecond. These ndings have important im-
plication for understanding ribosome’s movement in translocation of tRNA during
the translation process. The use of coarse-grained model has reduced the number
of atoms in a simulation signicantly, however it also gives less information and re-
parameterization of force eld might bring less reliability in simulation accuracy.
Sanbonmatsu et al. (30) overcome the coarse-grained limitation in understanding
the translation machinery by producing the rst all-atom targeted MD simulation of
the accommodation of tRNA into the 70S ribosome in explicit solvent. This large
scale all-atom simulation yielded 2.6 million atoms and some 10
6
computer hours
were used to simulate seven systems of 2 ns time frame in order to understand
the critical step of identifying cognate tRNA selection in the decoding stage. The
simulations have enlightened the understanding of tRNA movement into ribosome
during decoding stages in atomic level. However, these MD studies (29, 30) in-
vestigated the role of ribosome in protein biosynthesis process and so far no MD
simulation is used to study ribosome’s function as a drug target.
Investigation on the ribosomal subunit crystal structures in the Protein Data Bank
(PDB) revealed majority of them are lacking structural information due to the
low percentage completeness of the structure. In the case of the large ribosomal
subunit, some rRNA and ribosomal proteins were not able to be captured due to
Figure 1: Schematic representation of erythromycin A.
It consists of a 14-membered lactone ring, a desosamine
sugar and a cladinose sugar.
133
Erythromycin-
ribosome Complexes
high density and their complicated structural organization in this macromolecule.
The coarse-grained model allows for much longer simulation than the simulation
with all-atom resolution. The latter has an advantage of high accuracy due to
specic interactions and is more suitable for drug design. Based on this, we did
a reconstruction of the missing residues in the crystal structure and then followed
by renement stages using MD simulation. In the process, we were able to obtain
some valuable information from this short simulation with regards to the binding
of ERYA to the large ribosomal subunit.
Here, we report the results of MD simulation of ribosome-ERYA complex, with
the aim to elucidate the mechanism of action of this macrolide on the 50S ribo-
somal subunit. To date, this is the rst kind of MD simulation on such system and
we hoped the results could provide both qualitative and quantitative information
to explain the molecular basis of the antibacterial action at the molecular level.
This simulation has enabled us to investigate water molecule’s involvement and
their contribution in mediating the binding of ERYA with the binding pocket. Our
study is also consistent with the experimental results (8) in terms of the existence
of hydrogen bond network in the ERYA binding at the large subunit; and also the
non-involvement of ribosomal proteins in the ERYA binding. In addition, we were
able to identify nucleotides from domains II and IV that were involved in the opti-
mum binding of ERYA to the large subunit.
Methods
Construction of Simulation System
The starting structure was based on a 3.50 Å resolution X-ray structure of the 50S
large ribosomal subunit, taken from the PDB: PDB code, 1JZY (8). This structure
contained one chain of 23S rRNA and three chains of ribosomal proteins namely L4,
L22, and L32, in complex with ERYA including two magnesium ions. This crystal
structure is 79.5% complete and 106 nucleotides were found missing from rRNA in
domains I, II, V, VI, and 14 missing amino acid residues from L4 and L32 protein
chains. The missing side chains for the ribosomal proteins were built based on the
carbon alpha atoms found in the PDB structure using LEAP module of AMBER 8
package (31). Missing RNA residues were constructed using the following proce-
dure: (I) Missing residues were identied and their corresponding contiguous resi-
dues were extracted to Hyperchem7.5 (Hypercube Inc.). (II) The missing residues
were built according to their sequence using Hyperchem7.5. (III) These newly con-
structed part from the Hyperchem7.5 were then inserted into the starting structure to
form a complete structure of the large ribosomal subunit (It is worth noted here that
all of these missing residues were not located in the binding pocket). Subsequently,
the structure was rened for a total of 10,000 steps each of Steepest Descent (SD)
and Conjugate Gradient (CG) minimizations to relieve possible steric clashes and
overlaps of side chains. The structure was examined at every 5000
th
step of minimi-
zation, in terms of its energy, RMS values and atoms that were overlapped or made
unfavorable contacts. The nal structure was taken when all of the above criteria
were satised and used as the starting structure for subsequent MD simulation runs.
The geometries of ERYA were fully optimized and their electrostatic potentials were
computed, both at the B3LYP level with the 6-31G(d,p) basis set using GAUSSIAN
03 (32) program and their partial charges were obtained by restrained electrostatic
potential (RESP) using ANTECHAMBER (33). The Amber 99 force eld (34) was
used to describe the molecular mechanics of rRNA and ribosomal proteins, while
the general amber force eld (GAFF) (35) was used to describe ERYA. Hydrogen
atoms for the entire complex were added explicitly using LEAP. The complex
(50S ribosomal subunit and ERYA) was immersed in a TIP3P (36) waterbox of
edge lengths of 233.736 Å × 270.195 Å × 198.736 Å containing 296,254 water
molecules. The complex was found to be largely anionic due to the presence of
134
Wahab et al.
phosphates in the RNA. Therefore, to ensure the neutrality of this system, potas-
sium and sodium counterions were added almost equally to the system and they
were placed by LEAP at the most negative position of the complex. The details of
the simulation system including number of counterions and number of atoms for
each molecule are shown in Table I. The parameters of magnesium, sodium, and
potassium ions were taken from the standard AMBER database.
Minimizations and Molecular Dynamics Simulations
Minimizations and molecular dynamics (MD) simulations were carried out using
the SANDER module of AMBER 8. Prior to MD equilibration runs, the system
was subjected to a total of 4,500 steps of SD and 10,500 steps of CG minimizers on
three different minimization stages where the complex was xed on place with a
positional restraint force of 300 and 150 kcal mol
-1
Å
-2
, respectively, in the rst and
second stages of minimization, and the entire system was permitted to move freely
after that. The relaxed structure was then subjected to heating stages, each with 20
ps from 0 to 150K then to 300K with positional restraints of 150 kcal mol
-1
Å
-2
on
the complex. It was then followed by 20 ps of fully unrestrained equilibration at
constant temperature of 300K, controlled by the Langevin thermostat with collision
frequency of 1.0 ps
-1
(37). A 40 ps MD simulation was done using the canonical
ensemble before switching over to the isobaric-isothermal ensemble. Pressure of
the system was regulated at 1 bar with isotropic position scaling of 1 ps pressure
relaxation time. SHAKE (38) method was applied throughout the MD simulation
to allow the integration of force equation at 2 fs and Particle Mesh Ewald (PME)
(39) was turned on for proper treatment of long range electrostatic interactions and
the non-bonded cutoff was set to 8.0 Å. The translational and rotational around the
center of mass were removed every 1000 steps, and the non-bonded pair list was
updated every 25 steps. Trajectories were saved every 0.1 ps during the simulation
for later analyses. Approximately 700 hours were used on a 16-CPU Linux cluster
upon the completion of the current 1.3 ns of MD simulation.
Table I
MD system setup details.
Simulation system 23S rRNA Ribosomal proteins Ligand / Ions Water molecule Total
ERYA+ large
ribosomal subunit
2880 residues
(92,922 atoms)
L4: 205 residues (3148 atoms)
L22: 134 residues (2214 atoms)
L32: 60 residues (958 atoms)
ERYA
(118 atoms)
2 Mg
1444 Na
+
1400 K
+
296,254990,968 atoms
Figure 2: Thermodynamics properties of (A) potential
energy (B) simulation box volume for the 1.3 ns simula-
tion showed equilibration from 500 ps onwards. The
rst 100 ps that are intended for heating are omitted.
135
Erythromycin-
ribosome Complexes
Analysis
Trajectories were analyzed using PTRAJ from AMBER 8, while snapshots from
the trajectories were visually examined and illustrated using VMD (40). Hydrogen
bond analysis was calculated using PTRAJ and represented in the schematic form
generated by Ligplot (41). Hydrophobic interaction was computed using HBPLUS
(42), where HBPLUS generates a list of non-bonded interactions by computing
all possible positions for heavy atoms that are less than a specied distance apart.
Those interaction/forces that were exerted 3.90 Å or less for paired carbon-carbon
were then extracted. The distribution of water molecules in the hydration shell of
ERYA and radial distribution function (RDF) was calculated using PTRAJ. Analy-
sis of MD trajectories was focused on its production stage (500-1300 ps) while
thermodynamics properties were monitored throughout the simulation.
Results and Discussion
Stability of the Trajectories
Stability of the trajectory was demonstrated by thermodynamics properties versus
simulation time as shown in Figure 2A-B. In general, the plots were stable through-
out the simulation, but sharp increase could be observed in the beginning of the simu-
lation (data not shown) due to heating stages. Consequently, the raise of temperature
and release of positional restraints have increased the potential energy of the system
and nally, causing the system to reach equilibration from 500 -1300 ps. Figure 2B
showed simulation box volume versus simulation time. It can be seen that the simu-
lation box reached an average value of 1.05 × 10
7
Å
3
and it did not change much dur-
ing the production stages, which also signied the system has reached equilibration.
Convergence and stability of the simulation could be further reected from its de-
viation from a reference structure using the root mean square deviation (RMSD).
The mass-weighted RMSD of the erythromycin binding pocket (taken as 15 Å from
the center of mass of ERYA, as shown in Figure 3) over 1.3 ns of MD simulation
was calculated with reference to the initial structure. The RMSD plots (Figure 4A-
C) showed structure reached a stable state only after 500 ps where average displace-
ment of the binding site was 3.38 ± 0.14 Å. For the ligand, it only has an average
displacement of 1.57 ± 0.09 Å from the initial structure during the production stage.
The low RMSD values showed ERYA was well maintained in the pocket through-
out the 1.3 ns simulation. The RMSD of ribosomal proteins (Figure 4C) inside the
pocket was also calculated based on their backbone with respect to the initial struc-
ture and it was found to have an average value of 2.80 ± 0.19 Å.
To have better insight on the exibility of each nucleotide residue in the binding
pocket, a root mean square uctuation (RMSF) analysis was evaluated on the ribo-
Figure 3: Line and tube representation of rRNA in 50S
ribosomal subunit (left). Red arrow showed location of
ERYA binding pocket (taken as 15 Å from mass center
of ERYA) with yellow surface representation (right).
The backbone of the pocket is indicated by light blue
tube and ERYA is represented in dark blue licorice.
136
Wahab et al.
some backbone with reference to the average structure. Several key residues in-
volved in the interaction of ERYA with the binding pocket, including C765, C1773,
A2041, A2042, A2045, A2418, A2482, G2484, U2564, U2588, and U2589 [here
and throughout the entire manuscript, nucleotides and amino acids are numbered
according to D. radiodurans, unless stated in parentheses with E. coli (EC) num-
bers or H. marismortui (HM) numbers] showed low exibility as indicated by their
relatively smaller RMSF values (Table II). These important residues were mainly
located in Domain II and V (except for C1773 in Domain IV) and formed the bind-
ing site for ERYA. They also appeared to be the residues that were responsible for
the binding of ERYA, either through H-bonds, hydrophobic interactions, etc. These
interactions indeed restricted the motion of these residues resulting in low uctua-
tion and less mobility compared to other residues that were not at the vicinity of
the binding pocket. Some residues (those of not in the binding pocket) were found
to have high uctuation (> 10 Å) as shown in Table II. Visual inspection showed
these residues were located at the peripheral and exterior of the large ribosomal
subunit and therefore experienced high uctuations as the simulation progressed.
We have also measured their distances to ascertain that these regions indeed were
located very distant from the binding pocket (ranging from 80-150 Å). Therefore,
it is expected that these residues (which uctuated remarkably) did not exert any
signicant and direct effects to the binding of ERYA in the binding pocket.
Solvent Effects and Water-bridged Hydrogen Bond in the Binding Pocket
The characteristics of solvation effects and hydration shell are difcult to be stud-
ied by experimental studies due to the high mobility of water molecules (43, 44).
Therefore, computational simulation such as MD simulation on the other hand, are
often used to study details on water networks within the molecule, their dynamic
movements, and contribution in receptor-ligand binding (45).
In our study, the distribution of water molecules in ERYA binding pocket was inves-
tigated using radial distribution function (RDF), also known as the pair distribution
function or g(ij, r). It was calculated by dividing the space around the two atoms
(i, j) into spacing-bins at intervals of 0.05 Å, ranging between 0 to 10 Å. The water
molecules that were found in each shell were counted and averaged over congu-
rations that were generated by the 500-1300 ps window of MD simulation. The
average number was then divided by the volume of each shell to obtain the average
density of water as a function of the distance r from a reference atom in the ligand.
Figure 4: Root mean square deviation (RMSD) plots
for (A), backbone of erythromycin binding pocket (B)
ERYA, and (C) backbone of ribosomal proteins in the
binding pocket, with reference to the initial structure. The
rst 100 ps that are intended for heating are omitted.
137
Erythromycin-
ribosome Complexes
In this study, RDF was calculated based on selected acceptor atom in ERYA (see
below) to an oxygen atom of water molecule (Table III). In order to have a clearer
insight of the distribution of water molecules around the acceptor atoms, the esti-
mated number of water molecules that were found within the rst solvation shell
(spherical radius of <3 Å) and second solvation shell (spherical radius between 3-5
Å) were also counted and averaged in the 500-1300 ps window as shown in Table
IV. Figure 5 showed the general pattern for RDF plot that was obtained for oxygen
of water (Ow) and O6 (4ʹʹ-OH group of cladinose). RDF plots of other calculated
ligand atom also showed similar pattern (not shown here) and their rst peak of the
plot together with their intensity values are also recorded in Table III.
Table II
Root mean square fluctuation (RMSF) and types of interactions involved for residues found 4 Å
or less to ERYA in the binding pocket. Selected reconstructed residues that were distant from the
binding pocket are labeled with #.
Residue
Number
Domain
RMSF
(Å)
Types of interactions
predicted in this work
Types of interactions predicted
in crystal structure (8)
U727#II14.35 --
G728#II14.46 --
A729#II12.35 --
C759 II 0.42 HP -
C765 II 0.86 H-bond-
A1059# II 12.10- -
C1090# II 16.05- -
C1120# II 18.41- -
C1506# III17.27 --
A1511# III11.86 --
U1521# III12.72 --
C1773IV0.52H-bond and HP -
A2041V 0.44 -H-bond
A2042V 0.61 HP H-bond and HP
A2045V 0.70 HP H-bond
C2046V 0.80 HP -
G2286# V10.29 --
U2298# V10.49 --
A2418V 0.78 H-bond-
A2482V 0.47 H-bond and HP -
G2484V 0.52 HP H-bond and HP
U2564V 0.47 H-bond and HP -
C2565V 0.43 HP -
U2588V 0.57 H-bondH-bond and HP
C2589V 0.63 HP HP
U2590V 0.52 HP HP
Hydrophobic (HP) interaction was calculated using HBPLUS (42), accounted for interactions by
paired carbon-carbon that were those of 3.90 Å away.
Table III
Radial distribution function (RDF) calculated for
selected ligand atoms in ERYA.
RDF
First Solvation
Peak (Å)
RDF Intensity
at First Peak
Ow-O13 2.68 0.13
Ow-O62.781.28
Ow-O10 2.78 0.25
Ow-O52.780.62
Ow-O11 2.83 0.24
Ow-O82.880.23
Ow-O12 2.98 0.22
Ow-N 3.43 0.32
Hw-O61.780.60
Table IV
Estimated number of water molecules from ERYA and selected acceptor atoms of ERYA.
ERYA NO5O6O8O10 O11O12 O13
Simulation
Time (ps)
1
st
2
nd
1
st
2
nd
1
st
2
nd
1
st
2
nd
1
st
2
nd
1
st
2
nd
1
st
2
nd
1
st
2
nd
1
st
2
nd
501-600613 0602030201000000
601-700716 0704130301000000
701-800817 0714150402000001
801-900618 0614141402000001
901-1000 6180304250502000002
1001-11007 24 0304150501010013
1101-120010280305270602020013
1201-130010310415150602140114
60 165039 332937 135013 1701414
1
st
signified first solvation shell with spherical radius of <3 Å from acceptor atom and 2
nd
is second solvation shell with spherical radius
between 3-5 Å.
138
Wahab et al.
RDF of Ow and O6 showed rst and highest sharp peak at 2.78 Å with
the coordination number of 1 when integrated up to the rst minimum at
3.48 Å. This RDF prole also showed the highest RDF intensity, indicat-
ed most water molecules clustered around O6 (refer to Table IV) as com-
pared to the other acceptor atoms. RDF of Ow-O13 (oxygen of 12-OH of
lactone ring) and Ow-O10 (oxygen of 6-OH of lactone ring) showed the
rst peaks at 2.68 and 2.78 Å, respectively. The low intensity indicated
less number of water molecules surrounded these acceptor atoms. This
is also agreed with the low number of water molecules found in the rst
and second hydration shells (Table IV). Comparing to the other RDF
proles (with the exception of Ow-O6), Ow-O5 has the highest RDF
intensity values and more water molecules (totaling 35) at its rst and
second solvation shells. In contrast, Ow-O11 has only a total of 8 water
g(r)
Distance (Angstrom)
Figure 5: Radial distribution function (RDF) between
oxygen of water molecule (Ow) and O6 of ERYA.
Desosamine
Lactone
1.70
1.88
O6
Cladinose
2.03
U2564
O5
2.61
2.83
1.86
2.98
2.37
G2562
1.92
WAT79259
C2589
C 2046
A 2045
A 2482
U 2564
O1P
P
O2P
A 2042
ERYA
C 2565
U 2590
C 2589
G 2484
C 759
C 1773
U 2588
Figure 6: Water-bridged H-bonds at the binding site between O5
and O6 of ERYA and nucleotides G2562, U2564, and C2589. H-
bond distances (in Å) are also shown.
Figure 7: Ligplot (41) representation of average structure of
500-1300 ps window, revealing hydrogen bonds (green dotted
lines) and hydrophobic contacts (red spoked arcs pointing towards
ERYA) between ERYA and bases C765, U2588, A2045, A2482,
G2484, C2565, C2589, and U2590.
139
Erythromycin-
ribosome Complexes
molecules in the rst and second solvation shells. The RDF plots for Ow-O8 (2ʹ-
OH of desosamine sugar) and Ow-O12 (11-OH of lactone ring) showed their rst
peaks at 2.88 and 2.98 Å, respectively. The average number of water molecules
was found to be the lowest for O12 (only 1 in the second solvation shell). In the
case of N atom, the rst peak occurred only after 3 Å and 39 water molecules were
found in the second solvation shell (no water molecule found in the rst solvation
shell) indicating that this atom was less hydrated than the other oxygen atom in
the sugar moiety.
To conrm whether O6 is making any H-bonding with water molecules, RDF and
H-bond analysis were also performed for Hw (hydrogen atom of water molecule)
and O6 (Table III). Hw-O6 plot showed the rst peak formed at 1.78 Å. H-bond
analysis (Table V) also showed that 86.12% of occupancy when water molecules
acted as H-bond acceptor to O6, with the average length of 2.84 Å and angle of
160.88º (see below Hydrogen Bond Analysis and Hydrophobic Interactions for
cutoffs used). On the other hand, when water molecules acted as H-bond donor
16
14
12
10
8
6
4
2
0
Distance (Angstrom)
200 400 600 800 1000 1200
Time (ps)
2ʹOH@ERYA-N1@A2041
2ʹOH@ERYA-N6@A2041
2ʹOH@ERYA-N6@A2042
6ʹOH@ERYA-N6@A2045
11-OH@ERYA-O4@U2588
12-OH@ERYA-O4@U2588
Figure 8: Distance for selected H-bonds suggested
by Schlünzen and co-workers (8) as seen in this 1.3 ns
simulation. Majority H-bonds had its distance for >3.50
Å when approaching equilibration and production stag-
es. As a result, H-bonds between them were no longer
formed and this result did not show agreement with the
pattern found in crystal structure.
Figure 9: (A) (Left) Distance of ribosomal proteins
(in Å) L4, L22, and L32 from mass center of ERYA is
shown here. New cartoon representation of backbone
indicated three nearest ribosomal protein: L4 (yellow),
L22 (cyan), and L32 (purple), together with ERYA in
licorice representation. Other nucleotide residues are
omitted for clarity. (Right) Time dependent distances
between the backbone of selected amino acid from L4,
L22, and L32 to the mass center of ERYA. (B) (Left)
Time dependent distance for magnesium ions from
nearest atoms of ERYA. (Right) Figure showed two
magnesium ions (MG3281 and MG3282), licorice rep-
resentation of ERYA (all H atoms were omitted) and
nucleotides A806, C2420, C2421, C2431, and U2485.
Both magnesium ions are found far from ERYA; how-
ever, they are mainly coordinated to the phosphate ox-
ygen of these nucleotides. Other nucleotide residues
were omitted for clarity.
24
22
20
18
16
14
Distance (Angstrom)
0 200 400 600 800 1000 1200
Time (ps)
14
12
10
8
6
4
Distance (Angstrom)
0 200 400 600 800 1000 1200
Time (ps)
L4
L22
L32
ERY-A
18.37
21.55
14.72
R109-S113 of L22
M1-K8 of L32
Y59-A67 of L4
MG3281-O11@ERY-A
MG3282-C28@ERY-A
A806
MG3281
10.11
21.55
7.88
C2421
C2420
O11
Lactone
Cladinose
Desosamine
C28
MG3282
C2431
U2485
A
B
140
Wahab et al.
to O6, occupancy of 52.25% was found with 3 Å and 145.99º, respectively. It is
interesting to note that O5 also formed H-bond with water molecule with high
occupancy of 66.62% (length of 2.97 Å and angle of 144.05º).These observations
showed strong and persistent H-bonds formed between water molecules and cla-
dinose sugar, thus maintaining the cladinose sugar at the binding pocket. These
H-bonds were also found at appropriate geometry to bridge H-bonds between cla-
dinose sugar and nucleotides in the binding site.
It is interesting to note that there were many water molecules clustered in between
U2564 (U2585EC) and O6. However, when H-bond cutoff were applied to classify
them, water-bridged H-bonds were found to occur only in 30% of the simulation
time. These water-bridged H-bonds were formed between O6 and mainly with N3
and H3 of U2564. However these water-bridged H-bonds were found to be not per-
manent, as when O6 and U2564 came closer together, water molecules moved aside
to make way for direct H-bond. It is also worth noting here that the water molecule
that bridged H-bonds between O6 and U2564 was not always the same molecule.
Another water-bridged H-bond was found between O5 and G2562 (G2583EC) and
C2589 (C2610EC). Unlike the former H-bond, water molecule that bridged H-
bond between O5 and G2562 was found to be the same molecule (WAT79259) from
600 ps onwards. A snapshot of these water-bridged H-bonds is shown in Figure
6. In contrast to this nding, Mao and co-worker (46) in their Structure Activity
Relationship (SAR) studies previously showed that O6 was not able to mediate any
H-bonds due to its dispensability for macrolide binding. The crystal structure (8)
also showed that no H-bonds were mediated from this sugar. The analysis of our
trajectory gives, however, some indication that this sugar moiety might involved in
the binding of ERYA to ribosomal subunit through water molecules.
We also observed that there were many water molecules at the binding site and in the
rst solvation shell of ERYA, but not all of them were involved directly in bridging
ERYA with the binding pocket. These water molecules might possibly contributed
to the overall stabilization of ERYA binding pocket by holding them in the right
position via interconnecting network of H-bonds. Some crystallography and MD
simulation studies have previously shown the importance of water and water-bridged
H-bond in many receptor-ligand system such as protein-ligand (47-50), DNA-ligand
(51-53), RNA-ligand (54-57), etc.; therefore, we believe this nding is important and
should be taken into consideration in the understanding of their binding interaction.
Hydrogen Bond Analysis and Hydrophobic Interactions
Hydrogen bond (H-bond) is one of the main interaction for the binding of ligand to the
binding pocket and there were at least six H-bonds observed from the X-ray structure
(8) that kept ERYA in the pocket. H-bonds were assumed to be present if distance of
H-bond donor-acceptor was <3.5 Å and the angle formed between donor-H-acceptor
was >120º. The classication of H-bond occupancies used are as follows: persistent
H-bond if occupancy is >60%, medium H-bond if occupancies are 30% to 60% and
weak H-bond if <30% (58). Due to the enormous amount of H-bonds found during
this simulation, only H-bond with >20% of occupancy are discussed here.
It was found that C1773 (U1782EC), U2588 (U2609EC), C765 (A752EC), A2418
Table V
H-bonds formed between water molecules and ERYA in the 500-1300ps window.
H-bond Acceptor H-bond Donor
Occupancy
(%)
Average
Distance (Å)
Average
Angle (°)
Solvent
H10···O6@ERYA
86.122.84160.88
O5@ERYASolvent 66.622.96144.05
O6@ERYASolvent 52.253.00145.99
H-bond distance and angle cutoff used in this analysis are 3.50 Å and 120º, respectively.
Occupancy <50% are not shown here.
141
Erythromycin-
ribosome Complexes
(A2439EC), and A2482 (A2503EC) played an important role in the binding of
ERYA. C1773 formed at least ve different H-bonds with O12 and O13 of the
lactone ring for ERYA throughout the 1.3 ns simulation (Table VI). The H-bond
between N3 of the pyrimidine base of C1773 and hydrogen of O13 of ERYA were
found to occupy 50.85% of the total simulation with an average distance of 2.93
Å and mean angle of 145.25º. This H-bond was found with medium occupancy
during the equilibration stage but became persistent until the end of simulation.
C1773 also formed medium H-bond contacts with hydrogen of O12 and O13 of
ERYA through its N3 and O2 at the pyrimidine base, with the percentage occupancy
of 31.00% (average distance and angle of 3.11 Å and 152.32º, respectively) and
32.46% (3.14 Å and 149.21º) correspondingly. There were also H-bonds formed
between hydrogen of O12 of ERYA with O2@C1773 and between oxygen O13 to
N4 of C1773. The dynamic features of these two bonds were correlated with the
percentage occupancy of 28.77% and 24.85%, with average bond length and angle
of 2.99 Å and 139.94º, 3.16 Å and 136.31º, respectively. These two H-bonds were
seen with low occupancy in the early stage of MD but strengthened during MD
production stage before it weakened in the nal stage of simulation.
U2588 formed a medium H-bond at O4 of its pyrimidine base with hydrogen of
O12 of ERYA, with mean distance of 2.80 Å and angle 142.97º. This H-bond
was found to be persistent in early stage of MD, but was deformed during 700 ps
to 1000 ps. However, it was reformed after 1000 ps and stayed on moderately as
the production time evolved. This H-bond was 44.38% occupied during the 1.3 ns
simulation, indicated its signicance in keeping ERYA at its pocket. C765 formed
H-bond with oxygen of O13 at ERYA via N4. This H-bond has an occupancy per-
centage of 34.69% and with average bond distance of 3.07 Å and a mean angle of
137.10º. This bond interacted strongly and persistently even from the start of the
simulation but became weaker after 500 ps. Other residues in Domain V that were
found to be involved in H-bonds interaction included A2418 and A2482. A2418
formed a weak H-bond at N6 of its purine ring with oxygen of O10 of lactone ring
with the occupancy of 22.46%, average distances and angle of 3.21 Å and 138.27º,
respectively. H-bond between N6 of A2482 and oxygen O9 of desosamine sugar
was found in the early stage of the simulation but was weakened after 500ps. The
average distance and angle formed was 3.18 Å and 145.08º, respectively.
Hydrophobic interaction was computed using HBPLUS, where the interactions/
forces exerted were those of 3.90 Å away for paired carbon-carbon and it is il-
lustrated in Ligplot representation. Figure 7 showed H-bonds and hydrophobic
interaction found in the average structure with its interacting nucleotide residues.
There were a total of four H-bonds found in the average structure, namely H-
bonds between oxygen O12 of ERYA and O4@U2588 with distance of 3.26 Å,
nitrogen N3 of U2564 and O6@ERYA with distance of 2.96 Å, oxygen O12@
ERYA and O2@C1773 of 3.02 Å and lastly, oxygen O13@ERYA and N3@C1773
Table VI
H-bonds formed between ERYA and binding pocket during the 1.3 ns MD simulation.
H-bond Acceptor H-bond Donor
Occupancy
(%)
Average
Distance (Å)
Average
Angle (°)
Predicted in
crystal structure
N3@C1773
H31···O13@ERYA
50.852.93145.25-
O4@U2588
H36···O12@ERYA
44.382.78142.97+
O13@ERYA
H41···N4@C765
34.693.07137.10-
O2@C1773
H31···O13@ERYA
32.463.14149.21-
N3@C1773
H36···O12@ERYA
31.003.11152.32-
O2@C1773
H36···O12@ERYA
28.772.99139.94-
O13@ERYA
H41···N4@C1773
24.853.16136.31-
O10@ERYA
H61···N6@A2418
22.463.21138.27-
O9@ERYA
H61···N6@A2482
20.623.18145.08-
H-bond distance and angle cutoff used in this analysis are 3.50 Å and 120º, respectively. Occupancy <20% are not
shown here. +, predicted H-bond; -, H-bond not predicted.
142
Wahab et al.
with 2.90 Å. There were also 20 hydrophobic interactions found between carbon
atoms in ERYA with carbon atoms of interacting residues in the average structure.
Of all the 20 hydrophobic interactions, some of these interacting residues made
edge to face π-π interaction to ERYA. For example, the hydrophobic interactions
between C30 of ERYA to C6@G2484, and C32 of ERYA to C2 and C6 of A2042
showed lactone ring of ERYA was located perpendicular to the base sugar of
these purine bases, forming the T-shape π-π interaction. Comparing to the crystal
structure, we observed six H-bonds and 11 hydrophobic interactions between the
ERYA and the nucleotides in the binding pocket.
The crystal structure (8) showed all the six H-bonds between erythromycin and its
binding pocket were mediated through the reactive group of desosamine sugar and
lactone ring of the ERYA. Our MD simulation did not showed consistencies with the
observed pattern in crystal structure as the observed H-bonds were mainly mediated
from O12 and O13 of lactone ring. H-bonds that were found in the crystal structure
were H-bonds between 2ʹOH group of desosamine sugar with N1 and N6 of A2041
and to N6 of A2042, 6-OH group of lactone ring with N6@A2045, 11-OH group of
lactone ring with O4@U2588 and 12-OH group of lactone ring with O4 at U2588. In
our simulation, all of these H-bonds (except H-bond between O12 and O4@U2588)
were found to have bonds length >3.50 Å in most of the simulation time and none
of them were able to form permanent H-bonds (Figure 8). The importance of the
H-bonds between nucleotides A2041 (A2058EC) and A2042 (A2059EC) and mac-
rolides were previously shown by crystallography studies (8, 59, 60). Schlünzen and
co-workers found three H-bonds that connect erythromycin to A2041 and A2042. Tu
and co-workers (59) found a H-bond between 2ʹOH of its desosamine sugar with the
N1 of the mutated G2099A,HM (A2058EC) from the crystal structure of Haloarcula
marismortui large subunit with MLS
B
K antibiotics (including erythromycin). Hansen
and co-workers (60) on the other hand found in the crystal structure of H.marismortui
50S subunit complexed with the 15- and 16-membered macrolides, that the myca-
minose of 16-membered macrolide (or desosamine sugar from 15-membered mac-
rolide) formed a H-bond from its 2ʹOH to G2099HM (A2058EC). Although these
H-bonds were known to be important in the overall binding of macrolides, we did
not found these H-bonds in our simulation. The only interaction that we were able to
associate with A2041 and A2042 is the H-bond between 2ʹOH of desosamine sugar
and N1@A2041 that was found in the very early stage of the simulation. However,
as simulation time evolved, this H-bond slowly disappearing during the equilibration
stage and it was totally distorted shortly after this stage (Figure 8).
One of the possible explanation for these observations could be the orientation of
the system was different from the orientation found in the crystal structure. This
might caused by minimizations done prior the start of this simulation and therefore,
causing their interactions to be varied in accordance with their chemical nature.
Other explanations might be the criteria used to dene H-bonds differ from the one
used in the crystal structure and it should also be noted that the classication of
these H-bonds were based on their occupancies (total time of existence divided with
total simulation time) and therefore those with low occupancies were not discussed
and shown here. In a static system, like in a crystal structure, only one snapshot was
able to be captured while on the other hand; a dynamic system like the MD simula-
tion offered a trajectory of many snapshots capturing many different conformations,
thus giving a different pattern of H-bond network when comparing with the crystal
structure. Nevertheless, the H-bond pattern that showed here has provided another
perspective of H-bond pattern as explicit analysis of H-bond properties enabled us
to look at the occupancies and assignment of each H-bond that was formed.
Involvement of Ribosomal Protein and Magnesium Ion
Figure 9A showed distances between the center of mass for the backbone of se-
lected nearest amino acid residues to ERYA’s center of mass. These plots demon-
143
Erythromycin-
ribosome Complexes
strated that the nearest ribosomal protein chain, specically the Met1 to Lys8 of
L32, were about approximately 15 Å away from ERYA at the start of the simula-
tion, and uctuated around the mean value of 15.50 Å until the end of the simula-
tion. The second nearest protein chain, the ribosomal protein chain of L4 (from
residue Tyr59 to Ala67) was ~17 Å distance away from the ligand at the start of
the trajectory. It oscillated ~16.50 Å during production stage and the distance
was maintained at ~16 Å in the last few hundred ps. The most distant protein
chain from ERYA, L22 as demonstrated by its nearest amino acid to the ligand,
namely Arg109 to Ser113 were ~19 Å away at the start of the MD simulation.
The distance increased to ~22 Å during production stage, and uctuated around
1-2 Å after 500 ps until the end of simulation. On the average, these three nearest
ribosomal protein chains, namely the L4, L22, and L32 were found to be at least
15 Å away from the ERYA, as seen in this MD simulation and also in the crystal
structure (8). These nding indicates that the ribosomal proteins were too distant
for any signicant chemical interactions to occur, and hence supporting the claim
that erythromycin binding site is made of only rRNA and ribosomal proteins have
no direct implication to the binding of ERYA (8).
Figure 9B showed distances of the two magnesium ions that were found in crystal
structure from ERYA. Two magnesium ions (MG3281 and MG3282) were found
to be located distant from ERYA. On average, the nearest atom from ERYA to
MG3281 was found to be 10.96 Å from O11 of the lactone ring of ERYA and on
the other hand, MG3282 was 7.03 Å from C28 of desosamine sugar of erythromy-
cin. Based on these distances, both magnesium ions were far for any signicant
direct interaction to the binding of erythromycin. Both magnesium ions were found
near to the phosphate oxygen of A806, C2420, and C2421 (for MG3281); C2431
and U2485 (for MG3282) although no restraints were applied in this simulation
to maintain this interaction. Based on these, we believe magnesium ions were not
directly involved in the binding of ERYA; instead they might be involved in main-
taining the stability of the binding pocket.
Further Discussion
Our MD simulation showed direct interaction of ERYA to Domain II of the 23S
rRNA. An H-bond with occupancy of 34.69% was found between N4 of C765 and
hydrogen of 12-OH group of the lactone ring of ERYA. Nucleotide C759 formed
hydrophobic interaction with ERYA in most of the simulation time. These observa-
tions therefore agree with the protection effect found on hairpin 35 of Domain II in
footprinting experiments (2, 3, 7) and supported the speculation that the Domain II
has direct and signicant interactions with ERYA. Nucleotides A2041 (A2058EC),
A2042 (A2059EC), and G2484 (G2505EC) were previously demonstrated to be
important, as it was shown to be strongly protected by erythromycin from chemical
modication (2, 3, 7, 14, 61). Although our simulation was not able to prove the in-
volvement of these residues in H-bond analysis (as mentioned above), hydrophobic
analysis showed these nucleotide exerts its hydrophobic interaction to ERYA by π-π
interaction between its purine bases and desosamine sugar of ERYA.
Our MD simulation also showed the signicance of residue C1773 (U1782EC) to
the binding of ERYA. C1773 is located in Domain IV, a Domain that has never
discussed before to have any direct interaction with the binding of erythromycin.
However, C1773 was previously associated to the binding of quinolyallyl group of
ketolide ABT-773 (62) through hydrophobic interaction and other than that, it was
never reported to have contribution in the binding of other macrolides to 23S rRNA.
Visual inspection showed that C1773 was folded in a proximal distance to C765
(from Domain II) and the exposure of this residue to ERYA became apparent in the
equilibration stage. H-bond analysis also revealed that C1773 formed at least ve
different H-bonds with ERYA at different locations and yet, most H-bonds occurred
for >20% of the simulation time. This nucleotide residue might be as important as
144
Wahab et al.
other residues that interacts with ERYA, considering it is a cytosine in Deinococ-
cus radiodurans (prokaryotic ribosome) and is a uracil in Haloarcula marismortui
(eukaryotic ribosome), where the difference in nucleotide sequences between a pro-
karyote and eukaryote might potentially affect the drug afnity and selectivity of
ERYA to the ribosome. With this, we believed that this residue might have some
important implications in the binding of ERYA and further research should be done
to investigate the importance of this residue.
Conclusions
In the process of rening a low-resolution structure, enormous amount of valu-
able information were obtained in understanding ribosome’s role as a drug target.
Based on our study, it was found that the drug binding site was composed only by
rRNA and no ribosomal proteins were found to be involved directly as they were
too distant from the binding pocket such that direct interaction is irrelevant. Similar
observation was seen for the putative magnesium ions as discussed in (8) and it was
found that the two ions, were also far from the drug. Therefore, they would also not
be able to be involved directly in the overall binding of ERYA-ribosome.
Our study also showed the importance of water molecules as ERYA was highly hy-
drated during the course of the simulation. Water molecules were found to form H-
bonds with ERYA with high occupancies and interestingly, water-bridged H-bonds
were found in maintaining the strong interaction of ERYA to the binding pocket.
Water molecules at the binding site have played important role by forming multiple
H-bonds, bridging ERYA with binding site and possibly stabilizing the entire drug-
binding site interactions. This study should also be useful in understanding hydra-
tion sites at ERYA binding site and would be applicable in methods for de novo
drug designing. Besides that, H-bond analysis and hydrophobic interactions also
contributed to the overall ERYA binding. Although some of H-bond formed did
not show consistencies with the experimental results, in general H-bond still exist
between ERYA and their contribution is conclusive.
We believe this simulation served as the stepping stone of studying more rigor-
ously into the detail mechanism of action of the ERYA to 50S large ribosomal
subunit that currently is not able to be offered by the experimental studies. A lon-
ger simulation (which is now in progress) would ensure a higher reliability, and
insight obtained from it would denitely be needed to assist in the development
of safer macrolide antibiotics in the future.
Acknowledgements
This work was supported by the Ministry of Science, Technology and Innovation
(MOSTI) grant (grant number: 304/PFARMASI/640038/K105). The generous
supply of computational time by MIMOS Berhad for this research is gratefully
acknowledged. We thank National Biotechnology Network for providing our lab
with basic computational facilities. We also thank Prof. Janez Mavri for criti-
cally reading this manuscript.
References and Footnotes
1.
2.
3.
4.
5.
6.
7.
8.
9.
S. Omura. Macrolide Antibiotics: Chemistry, Biology, and Practice. Academic Press, Am-
sterdam, Boston (2002).
L. Xiong, S. Shah, P. Mauvais, and A. S. Mankin. Mol Microbiol 31, 633-639 (1999).
L. H. Hansen, P. Mauvais, and S. Douthwaite. Mol Microbiol 31, 623-631 (1999).
B. Weisblum. Antimicrob Agents Chemother 39, 797-805 (1995).
B. Vester and S. Douthwaite. Antimicrob Agents Chemother 45, 1-12 (2001).
L. Xiong, Y. Korkhin, and A. S. Mankin. Antimicrob Agents Chemother 49, 281-288 (2005).
S. Douthwaite, L. H. Hansen, and P. Mauvais. Mol Microbiol 36, 183-193 (2000).
F. Schlunzen, R. Zarivach, J. Harms, A. Bashan, A. Tocilj, R. Albrecht, A. Yonath, and F.
Franceschi. Nature 413, 814-821 (2001).
J. M. McGuire, R. L. Bunch, R. C. Anderson, H. E. Boaz, E. H. Flynn, H. M. Powell, and J.
W. Smith. Schweiz Med Wochenschr 82, 1064-1065 (1952).
145
Erythromycin-
ribosome Complexes
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
S. Pal. Tetrahedron 62, 3171-3200 (2006).
A. M. Nilius and Z. Ma. Curr Opin Pharmacol 2, 493-500 (2002).
J. C. Gasc, S. G. d’Ambrieres, A. Lutz, and J. F. Chantot. J Antibiot (Tokyo) 44, 313-
330 (1991).
S. Pestka and R. A. Lemahieu. Antimicrob Agents Chemother 6, 479-488 (1974).
S. Douthwaite and C. Aagaard. J Mol Biol 232, 725-731 (1993).
G. P. Dinos, S. R. Connell, K. H. Nierhaus, and D. L. Kalpaxis. Mol Pharmacol 63, 617-
623 (2003).
M. Lovmar, T. Tenson, and M. Ehrenberg. J Biol Chem 279, 53506-53515 (2004).
N. Ban, B. Freeborn, P. Nissen, P. Penczek, R. A. Grassucci, R. Sweet, J. Frank, P. B. Moore,
and T. A. Steitz. Cell 93, 1105-1115 (1998).
N. Ban, P. Nissen, J. Hansen, P. B. Moore, and T. A. Steitz. Science 289, 905-920 (2000).
F. Schluenzen, A. Tocilj, R. Zarivach, J. Harms, M. Gluehmann, D. Janell, A. Bashan, H.
Bartels, I. Agmon, F. Franceschi, and A. Yonath. Cell 102, 615-623 (2000).
A. Vila-Sanjurjo, W. K. Ridgeway, V. Seymaner, W. Zhang, S. Santoso, K. Yu, and J. H. Cate.
Proc Natl Acad Sci USA 100, 8682-8687 (2003).
B. T. Wimberly, D. E. Brodersen, W. M. Clemons, Jr., R. J. Morgan-Warren, A. P. Carter, C.
Vonrhein, T. Hartsch, and V. Ramakrishnan. Nature 407, 327-339 (2000).
D. E. Brodersen, W. M. Clemons, Jr., A. P. Carter, R. J. Morgan-Warren, B. T. Wimberly, and
V. Ramakrishnan. Cell 103, 1143-1154 (2000).
J. Harms, F. Schluenzen, R. Zarivach, A. Bashan, S. Gat, I. Agmon, H. Bartels, F. Franceschi,
and A. Yonath. Cell 107, 679-688 (2001).
P. Nissen, J. Hansen, N. Ban, P. B. Moore, and T. A. Steitz. Science 289, 920-930 (2000).
J. M. Ogle, D. E. Brodersen, W. M. Clemons, Jr., M. J. Tarry, A. P. Carter, and V. Ramakrish-
nan. Science 292, 897-902 (2001).
T. M. Schmeing, A. C. Seila, J. L. Hansen, B. Freeborn, J. K. Soukup, S. A. Scaringe, S. A.
Strobel, P. B. Moore, and T. A. Steitz. Nat Struct Biol 9, 225-230 (2002).
A. Yonath, J. Mussig, B. Tesche, S. Lorenz, V. Erdmann, and H. G. Wittmann. Biochem Int
1, 428-435 (1980).
M. M. Yusupov, G. Z. Yusupova, A. Baucom, K. Lieberman, T. N. Earnest, J. H. Cate, and H.
F. Noller. Science 292, 883-896 (2001).
J. Trylska, V. Tozzini, and J. A. McCammon. Biophys J 89, 1455-1463 (2005).
K. Y. Sanbonmatsu, S. Joseph, and C. S. Tung. Proc Natl Acad Sci USA 102, 15854-
15859 (2005).
D. A. Case, T. A. Darden, T. E. I. Cheatham, C. L. Simmerling, J. Wang, R. E. Duke, R.
Luo, K. M. Merz, B. Wang, D. A. Pearlman, M. Crowley, S. Brozell, V. Tsui, H. Gohlke, J.
Mongan, V. Hornak, G. Cui, P. Beroza, C. Schafmeister, J. W. Caldwell, W. S. Ross, and P.
A. Kollman. AMBER 8. 2004: University of California, San Francisco.
M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman,
J. A. J. Montgomery, T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Lyengar,
J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H.
Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima,
Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V.
Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R.
Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K. Morokuma, G. A. Voth, P. Salvador, J.
J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. train, O. Farkas, D. K.
Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S.
Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L.
Martin, D. J. Fox, T. Keith, M. A. Al-Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe,
P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, and J. A. Pople. Gaussian 03,
Revision C.02. 2004: Gaussian, Inc., Wallingford CT.
J. M. Wang, W. Wang, and P. A. Kollman. Abstr Pap Am Chem S 222, U403-U403 (2001).
W. D. Cornell, P. Cieplak, C. I. Bayly, I. R. Gould, K. M. J. Merz, D. M. Ferguson, D. C.
Spellmeyer, T. Fox, C. J. W. and P. A. Kollman. J Am Chem Soc 117, 5179-5197 (1995).
J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, and D. A. Case. J Comput Chem 25,
1157-1174 (2004).
W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L. Klein. J Chem
Phys 79, 926-935 (1983).
R. W. Pastor, B. R. Brooks, and S. A. Mol Phys 65, 1409-1419 (1988).
Ryckaert J. P., G. Ciccotti, and H. J. C. Berendsen. J Comput Phys 23, 327-341 (1977).
T. A. Darden, D. M. York, and L. G. Pedersen. J Chem Phys 98, 10089-10092 (1993).
W. Humphrey, A. Dalke, and K. Schulten. J Mol Graph 14, 33-38, 27-38 (1996).
A. C. Wallace, R. A. Laskowski, and J. M. Thornton. Protein Eng 8, 127-134 (1995).
I. K. McDonald and J. M. Thornton. J Mol Biol 238, 777-793 (1994).
K. Kulinska, T. Kulinski, A. Lyubartsev, A. Laaksonen, and R. W. Adamiak. Comput Chem
24, 451-457 (2000).
P. Aufnger and E. Westhof. J Mol Biol 300, 1113-1131 (2000).
P. Cozzini, M. Fornabaio, A. Marabotti, D. J. Abraham, G. E. Kellogg, and A. Mozzarelli.
Curr Med Chem 11, 3093-3118 (2004).
J. C. Mao and M. Putterman. J Mol Biol 44, 347-361 (1969).
146
Wahab et al.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
G. Settanni, A. Cattaneo, and P. Carloni. Biophys J 84, 2282-2292 (2003).
C. S. Poornima and P. M. Dean. J Comput Aided Mol Des 9, 500-512 (1995).
Y. Tie, P. I. Boross, Y. F. Wang, L. Gaddis, F. Liu, X. Chen, J. Tozser, R. W. Harrison, and I.
T. Weber. FEBS J 272, 5265-5277 (2005).
K. Wittayanarakul, O. Aruksakunwong, P. Sompornpisut, V. Sanghiran-Lee, V. Parasuk, S.
Pinitglang, and S. Hannongbua. J Chem Inf Model 45, 300-308 (2005).
B. Nguyen, D. Hamelberg, C. Bailly, P. Colson, J. Stanek, R. Brun, S. Neidle, and W. D.
Wilson. Biophys J 86, 1028-1041 (2004).
B. Wellenzohn, M. J. Loferer, M. Trieb, C. Rauch, R. H. Winger, E. Mayer, and K. R. Liedl.
J Am Chem Soc 125, 1088-1095 (2003).
S. A. Shaikh, S. R. Ahmed, and B. Jayaram. Arch Biochem Biophys 429, 81-99 (2004).
Q. Vicens and E. Westhof. Biopolymers 70, 42-57 (2003).
Q. Vicens and E. Westhof. Structure 9, 647-658 (2001).
Q. Vicens and E. Westhof. Chem Biol 9, 747-755 (2002).
Q. Vicens and E. Westhof. J Mol Biol 326, 1175-1188 (2003).
S. J. Wodak, D. v. Belle, and M. Prevost. In Computer Modelling in Molecular Biology.
VCH, 62-102 (1995).
D. Tu, G. Blaha, P. B. Moore, and T. A. Steitz. Cell 121, 257-270 (2005).
J. L. Hansen, J. A. Ippolito, N. Ban, P. Nissen, P. B. Moore, and T. A. Steitz. Mol Cell 10,
117-128 (2002).
D. Moazed and H. F. Noller. Biochimie 69, 879-884 (1987).
F. Schlunzen, J. M. Harms, F. Franceschi, H. A. S. Hansen, H. Bartels, R. Zarivach, and A.
Yonath. Structure 11, 329-338 (2003).
Date Received: October 8, 2007
Communicated by the Editor Thomas E Cheatham
