A R T I C L E S
Celik et al.
Table 1. hSERT Homology Models Built Using Alignments A, B,
and C. Structural and Energetic Data for the 15 Models Are Listed
Ramachandran coordinates are optimized during the homology model-
ing procedure, it is obvious that the models with lowest DOPE score
and objective function within one sequence alignment will have the
best possible Ramachandran plots. In this way, the different alignments
can be evaluated by comparing the computed Ramachandran scores;
the better the alignment, the more favorable Ramachandran plots are
expected. The volume of the binding cavity was measured by the
Molegro Virtual Docker suite of programs, using the solvent accessible
surface method with a maximum of 25 cavities per model.45 Sodium
ions were included by manual placement in the same relative place as
they are located in the LeuTAa structure. A chloride ion was manually
inserted between residues Tyr121, Ser336, Asn368, and Ser372 in a
few models (please see Supporting Information). The four residues and
the Cl- ion were minimized in MacroModel version 9.146 with the
OPLS force field.
homology Ramachandran
DOPE
distance
GlideScore
volume of
model
plot (%)a
(kJ/mol)
D98(O
δ)−
Na1 (Å)b (kcal/mol)c binding site (Å3)
A-1
A-2
A-3
A-4
A-5
B-1
B-2
B-3
B-4
B-5
C-1
C-2
C-3
C-4
C-5
94.9
94.6
94.7
95.7
94.7
93.8
93.5
94.0
94.6
94.4
95.3
94.4
95.3
94.9
96.4
-78607
-78374
-78040
-78360
-78627
-77388
-76433
-78284
-77728
-77268
-79159
-78962
-79844
-79205
-79703
4.1
90.1
101.9
84.0
128.0
100.9
46.6
67.6
36.9
67.6
61.4
3.3
3.5
3.3
3.3
-6.6
3.2
4.2
62.4
114.2
115.2
68.0
3.3
3.3
3.0
2.3. Ligand Modeling. 5-HT was drawn in Maestro46 as charged
on the primary amino group and minimized with the OPLS 2005 force
field as implemented in MacroModel 9.146 for 10 000 steps of conjugate
gradient iterations or until convergence, according to default settings.
-6.7
143.9
a Percentage of residues with psi- and phi-values in the “favored” or
“most favored regions” of the Ramachandran plot. b Only distances <5.0
Å are included. c GlideScore computed for docking of 5-HT in a rigid protein
in Glide 4.0 using the SP scoring function.
2.4. Docking. Initial docking of 5-HT in the 15 homology models
of hSERT was performed with Glide47,48 version 4.0 using standard
procedures.46 This docking, with a rigid protein and a flexible ligand,
was generally performed with the default standard precision (SP) scoring
function in Glide;47 however, the Glide extra precision (XP) scoring
function49 was tested for two models to evaluate if the special extra
terms in this scoring function is of importance for this particular
protein-ligand system. The binding site was defined from two residues
(Asp98 and Ile172) which make up the two ends of the binding cavity
parallel to the membrane plane.
and the short C-terminal were included. Three alignments of the target
sequence of hSERT to the template, LeuTAa, are evaluated. The first
alignment, A, is based on an automatic alignment created in MOD-
ELLER (version 8.1),40 which uses the align2d methodology.41 Ad-
ditional manual adjustments were performed to eliminate unwanted gaps
in the produced alignment for hSERT and to incorporate knowledge
from available experimental data placing Asp9819 and Ala169,20 in the
binding site by iteratively evaluating the DOPE (Discrete Optimized
Protein Energy) scores42 after model building. Ile172 has been shown
to be important for binding of inhibitors to hSERT,20,21 and its sequence
proximity to Ala169 also places this residue in the binding cavity. Ile172
corresponds to Val104 in LeuTAa which is indeed placed in the
hydrophobic part of the binding cavity in LeuTAa.12 No further attempts
were done to refine the extra- and intracellular loops or the C-terminal,
since they are distant from the transmembrane region and the putative
ligand binding cavity (more than 20 Å), which is the focus of the present
study, and thus are probably not affecting the binding of the substrate.
The second alignment, B, was published with the LeuTAa structure12
and is used for generating the second set of homology models of
hSERT. The third alignment, C, stems from a study by Beuming et
al.25 where experimental data was included for all known proteins in
the NNS family to propose a detailed alignment.
2.2. Model Building. Five homology models, 1-5, were built for
each of the three alignments, A-C, using MODELLER in stand-alone
mode and default settings.40 The produced 15 homology models, A1-
5, B1-5, and C1-5, were evaluated by visual inspection (with respect
to reproducing the protein backbone in the 12 TM helices, especially
the kinks in TM1 and TM6) and by quantitative examinations through
the Objective Function, DOPE scores, the volume of the binding site,
and sterical features through Ramachandran plots (Table 1). The
MODELLER Objective Function describes how well the model fits
with all the input structural data,43 and the DOPE-score42 evaluates,
per residue basis, the quality of the model against the template. The
objective function and DOPE scores are energetic measurements and
should thus both be as low as possible.42,43 Stereochemical features of
the obtained models were examined from Ramachandran plots generated
in VMD 1.8.4 using the ramaplot plugin, version 1.0.44 As the
2.5. Induced Fit Docking. To fully explore the concerns about
flexible amino acid side chains when performing molecular docking
simulations into homology models, we decided to include protein
flexibility in the docking protocol using one homology model of hSERT
from each alignment. Models of hSERT that allowed docking of 5-HT
into a rigid protein were chosen for alignments A and C, whereas for
B, where no poses were generated during normal docking, the homology
model with the best energy scores and an interaction between Na1 and
Asp98 was selected for further evaluation, resulting in selection of
homology models A-4, B-3, and C-5 for further calculations. The IFD
protocol37,46 from Schro¨dinger Inc. that combines Glide 4.046 and Prime
1.546 was employed for docking of 5-HT into hSERT allowing for
protein flexibility.37,50 The IFD workflow consists of three steps. First,
the ligand is docked flexibly into a rigid protein with a soft Van der
Waals potential, thereby allowing for some steric clash. During this
step it is possible to make point mutations of highly flexible residues
to alanine, to artificially create more room and secure that at least a
few poses of the protein-ligand complex are generated; this was not
necessary for introducing 5-HT into the hSERT binding site. In the
second step during an IFD workflow the protein is optimized, using
Prime, within 5.0 Å of the ligand poses from the first step, and if any
residues were mutated in the first step, they are reintroduced. The final
step is a redocking of the ligand into the relaxed protein binding cavity,
where normal Van der Waals terms are used. Either the SP or XP
scoring function can be applied in the last step. The final scoring from
an IFD calculation is the computed GlideScore (SP or XP) reflecting
the interaction between the protein and the ligand. Another reported
(45) Thomsen, R.; Christensen, M. H. J. Med. Chem. 2006, 49, 3315-3321.
(46) Schro¨dinger, LLC, 2006.
(47) Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.;
Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.;
Shaw, D. E.; Francis, P.; Shenkin, P. S. J. Med. Chem. 2004, 47, 1739-
1749.
(40) Sali, A.; Blundell, T. L. J. Mol. Biol. 1993, 234, 779-815.
(41) Sanchez, R.; Sali, A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 13597-13602.
(42) Eramian, D.; Shen, M.; Devos, D.; Melo, F.; Sali, A.; Marti-Renom, M.
A. Protein Sci. 2006, 15, 1653-1666.
(48) 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, 1750-1759.
(49) Friesner, R. A.; Murphy, R. B.; Repasky, M. P.; Frye, L. L.; Greenwood,
J. R.; Halgren, T. A.; Sanschagrin, P. C.; Mainz, D. T. J. Med. Chem.
2006, 49, 6177-6196.
(43) Eswar, N.; John, B.; Mirkovic, N.; Fiser, A.; Ilyin, V. A.; Pieper, U.; Stuart,
A. C.; Marti-Renom, M. A.; Madhusudhan, M. S.; Yerkovich, B.; Sali, A.
Nucl. Acids Res. 2003, 31, 3375-3380.
(44) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graph. 1996, 14, 33-38.
(50) Teague, S. J. Nat. ReV. Drug DiscoVery 2003, 2, 527-541.
9
3856 J. AM. CHEM. SOC. VOL. 130, NO. 12, 2008