Novel insights into GPCR-Peptide interactions: Mutations in extracellular loop 1, ligand backbone methylations and molecular modeling of neurotensin receptor 1

Article abstract of DOI:10.1016/j.bmc.2008.08.051

Investigating prototypical interactions between NT(8-13) and the human neurotensin receptor 1 (hNTR1), we created a receptor-ligand model that was validated by site-directed mutagenesis and structure-activity relationship studies. Stabilization of the extracellular loop 1 (EL1) by π-stacking clusters proved to be important for agonist binding when substitution of six conserved amino acids by alanine resulted in an agonist specific loss of maximal binding capacity. In agreement with our modeling studies, EL1 seems to adopt a clamp-type border area controlling the shape of the binding site crevice. Employing chemically manipulated peptide analogs as molecular probes, the impact of backbone modifications on receptor-ligand interaction, especially the influence on ligand conformation, was examined in binding studies and explained by in silico analysis.

Full text of DOI:10.1016/j.bmc.2008.08.051

Bioorganic & Medicinal Chemistry 16 (2008) 9359–9368
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Novel insights into GPCR—Peptide interactions: Mutations in extracellular
loop 1, ligand backbone methylations and molecular modeling of
neurotensin receptor 1
*
Steffen Härterich, Susanne Koschatzky, Jürgen Einsiedel, Peter Gmeiner
Department of Chemistry and Pharmacy, Emil Fischer Center, Friedrich Alexander University, Schuhstrasse 19, D-91056 Erlangen, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Investigating prototypical interactions between NT(8–13) and the human neurotensin receptor 1
(hNTR1), we created a receptor–ligand model that was validated by site-directed mutagenesis and struc-
Received 8 May 2008
Accepted 22 August 2008
Available online 27 August 2008
ture–activity relationship studies. Stabilization of the extracellular loop 1 (EL1) by
p-stacking clusters
proved to be important for agonist binding when substitution of six conserved amino acids by alanine
resulted in an agonist specific loss of maximal binding capacity. In agreement with our modeling studies,
EL1 seems to adopt a clamp-type border area controlling the shape of the binding site crevice. Employing
chemically manipulated peptide analogs as molecular probes, the impact of backbone modifications on
receptor–ligand interaction, especially the influence on ligand conformation, was examined in binding
studies and explained by in silico analysis.
Dedicated to Professor H.-D. Stachel on the
occasion of his 80th birthday.
Keywords:
Neurotensin receptor
Neuropeptide receptor interactions
Molecular modeling
Ó 2008 Elsevier Ltd. All rights reserved.
Extracellular loop 1
Radioligand binding
Site-directed mutagenesis
Peptide backbone modifications
1. Introduction
indicate any site-specific binding for these residues (for detailed
information, see Table S2 in Supplementary material). Recently,
Neurotensin (NT) is a neuropeptide that is largely distributed in
the central nervous system and some regions of the periphery.
When administered into the CNS, neurotensin appears to inhibit
dopamine function and produce antipsychotic activity.1 These ef-
fects are associated with the G protein-coupled neurotensin recep-
tor 1 (NTR1).2 As a consequence, metabolically stable NTR1 agonists
that are able to cross the blood–brain barrier may have great poten-
tial for the treatment of schizophrenia.1,3,4 Structure–activity rela-
tionship (SAR) studies including a gradual truncation of the
the conformation of NT(8–13) bound to NTR1 was determined by
solid state NMR spectroscopy.6 Taking advantage of these results,
lactam-bridged NT(8–13) mimetics constraining the backbone
dihedral angle of Pro10 were developed.7–9 Previous binding site
models of NTR1 offered valuable insights into key interactions
and basic structural elements for species specifity.10,11 To further
elucidate structural and functional properties of peptidergic GPCRs,
in general and to advance a recently published pharmacophore
model for NTR1 agonists,12 we attempted to combine structure–
activity relationship studies, molecular modeling and site-directed
mutagenesis. Exploiting NTR1 as a prototypical peptide GPCR, our
investigations especially aimed to identify conserved structural
and functional motifs within the extracellular loop region and thus,
influence the design of new ligands for NTR1.
peptide sequence and a D-amino acid scan identified the C-terminal
hexapeptide NT(8–13) (H-Arg8-Arg9-Pro10-Tyr11-Ile12-Leu13) as the
necessary and sufficient portion for both binding affinity and ligand
efficacy.5 These findings could be corroborated when we showed
that formal shuffling of the amino acids 2–3, 4–5, and 6–7 did not
Abbreviations: TM, transmembrane helix; EL, extracellular loop; SCR, structur-
ally conserved region; GPCR, G protein-coupled receptor; NT, neurotensin; NT(8–
13), peptide containing residues 8–13 of neurotensin; hNTR1, human neurotensin
receptor 1; RMSD, root mean square deviation; MD, molecular dynamics; DPPC, 1,2-
dipalmitoyl-sn-glycero-3-phosphocholine; AUC, area under the curve; SAR, struc-
ture–activity relationship.
2. Results
2.1. Protein sequence alignment
The crystal structure of rhodopsin is widely used as a template
in GPCR homology modeling and at the time this model was cre-
* Corresponding author. Tel.: +49 9131 8529383; fax: +49 9131 8522585.
E-mail address: gmeiner@pharmazie.uni-erlangen.de (P. Gmeiner).
0968-0896/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2008.08.051

9360
S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
ated this was the only GPCR whose crystal structure had been
solved.13–15 Our homology modeling concept was based on a pro-
tein sequence alignment which helps to identify structurally con-
served regions (SCRs) present in both bovine rhodopsin and the
human neurotensin receptor 1 (hNTR1). Rhodopsin and hNTR1
share only 24% sequence identity and 56% similarity within the
transmembrane helix (TM) bundle. Because of the low overall
identity, the alignment process was improved by the addition of
409 sequences which share the same conserved patterns, that is,
other family A GPCRs.16
2.2. Membrane environment
A membrane patch, consisting of 180 1,2-dipalmitoyl-sn-glyce-
ro-3-phosphocholine (DPPC) lipids, was equilibrated over 3 ns of
MD simulation. As a validation criterion for our membrane model,
we monitored the order parameter SCD. During the later stages of
the equilibration, SCD closely approached experimental values17
showing an area per lipid head group of 0.66 nm2 (experiment:
0.62 nm2) 18 and the expected distribution of water and lipid head
group densities. After integration of the receptor into the equili-
brated membrane patch,19 polar residues pointing outwards at
the extra- and intracellular side of the TMs were used as additional
markers for solvent exposition and helped to validate the position
of the receptor model within the membrane.
Figure 2. Membrane snorkeling. Representation of the final docked structure,
showing the orientation of Asp138 and Arg142 towards the hydrophilic head group
of two DPPC molecules as indicated by two arrows. Parts of TM2, TM3 and EL1 are
represented as colored ribbons (TM2 in blue, TM3 in green, and EL1 in cyan). The
aromatic cluster formed by the conserved amino acids that were mutated in this
study is also shown.
2.3. Molecular dynamics simulation
During the ensuing MD simulation, the general structure of the
TM bundle remained stable. As expected, the highest conforma-
tional flexibility could be observed in the loop regions, especially
in intracellular loop 3, which, however, refolded and finally
adopted a stable conformation. After 5500 ps of simulation time,
a stable root mean square deviation (RMSD) of the SCRs of about
0.23 nm (Fig. 1) and stable potential and kinetic energies indicated
that the receptor model had reached equilibrium state. Extracellu-
lar loop 3 (EL3), which is expected to be of special importance for
ligand binding2 gave an acceptable RMSD value of 0.32 nm (Fig. 1).
Structures in the ensuing timeframe (5500–6500 ps) were em-
ployed for the calculation of the averaged receptor model used
for docking.
described by Hawtin et al. for the V1a vasopressin receptor,20 these
residues fulfill the function of a molecular anchor that fixes the
upper region of TM3 to the membrane (Fig. 2).
2.4. Docking
To guide and validate the docking procedure, interacting key
residues derived from mutagenesis studies and solid state NMR
experiments were identified,10,21 receiving distance restraints be-
tween ligand and receptor side chains. Data from further solid state
NMR experiments guided dihedral angle restraints and the starting
conformation of NT(8–13).6 A docking position between the extra-
cellular ends of TM6, TM7, and EL2 was defined for the C-terminal
During the timeframe used for the averaged model, amino acids
Asp138 and Arg142 formed ionic interactions with the polar head
groups of two phospholipid molecules in the vicinity of the extra-
cellular end of TM3 which were indicated by mean distances be-
tween 0.44 and 0.48 nm. In accordance with recent observations
carboxyl function, the adjacent C and the b-positioned CH2 moiety
a
at a place where the carboxyl group was able to interact with
Arg322. In the first ensemble of structures generated with the
MODELLER program package,22 Met207 at the junction between
TM4 and EL2 was too far away from the ligand binding site to form
the presumed hydrophobic interactions with Ile12 and Leu13 of
NT(8–13). Secondary structure consensus prediction of EL2, which
was modeled in analogy to bovine rhodopsin, keeping the con-
served disulfide bond between Cys141 (TM3) and Cys224 (EL2)
and the hairpin structure, did not indicate a preferred folding.
Therefore, we chose to remodel the region between Met203 and
Asp215. A gap was introduced in the alignment file used for MOD-
ELLER, which resulted in loop structures which only had to fulfill a
distance restraint between Met207 and Ile12, Leu13 of the ligand.
This procedure led to an ensemble of diverse loop conformations
and allowed Met207 to take part in the formation of a hydrophobic
pocket for the side chain of Ile12.
For further refinement of receptor–ligand interactions, side
chain atoms of the residues in EL3 were omitted in the input file.
This procedure retained the backbone conformation derived from
MD but removed the bias of chi angle restraints, thus resulting in
higher side chain flexibility during the modeling process. In order
to allow some kind of annealing process between receptor and li-
gand, side chain atoms of NT(8–13), except for Pro10, were also
Figure 1. RMSD of the receptor model during equilibration. RMSD of the backbone
of EL3 (black) and of the backbone of the SCRs (gray) is shown as a function of MD
simulation time. The timeframe between 5500 and 6500 ps, showing stable values,
is used for the generation of the equilibrated receptor model applied for docking.

S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
9361
omitted. The best model was reintegrated into the membrane envi-
ronment and subjected to MD simulation. To detect favorable
ensembles of ligand–receptor interactions within the resulting tra-
jectory, two-dimensional root mean square deviation (2D-RMSD)
plots were generated using the GROMACS routine g_rms.23,24 RMSD
was calculated on the ligand’s backbone atoms and on the back-
bone atoms of the key residues Met207, Arg322, Phe326, Trp334,
Tyr339, and Tyr342 which are supposed to take part in ligand bind-
ing.10 The resulting plot was used to identify five homologous clus-
ters within the trajectory, marking stable receptor–ligand
interaction and stages of conformational evolution. As expected,
the largest changes could be observed during the first stages of
the restraintless MD simulation, finally leading to the most stable
conformation found in the last cluster. Coordinate-averaged struc-
tures of each cluster were calculated. Then, RMSD values between
this structure and every snapshot in the cluster were calculated in
order to identify the cluster’s single structure with the lowest
RMSD. This timeframe was used as a representative for the cluster
and for energy minimization, followed by evaluation by Drug-
between À60° to À80°, being more likely for a proline residue than
the proposed À147° of the NMR based structure.6,26 Deviations ob-
served for psi 10 and phi 11 antagonized each other. The fluctua-
tion of these angles was especially pronounced between cluster 3
and cluster 5, changing from 137° and À143° in cluster 3 to 86°
and À90° in cluster 5 for psi 10 and phi 11, respectively (Table 1).
Comparison of the recently solved crystal structure of a beta
adrenoceptor (PDB code 2RH1) with our hNTR1 model indicates
an RMSD value of 0.33 nm for the SCR backbone.27 The value be-
tween our model and rhodopsin is 0.27 nm. The overall shape of
the model shows a larger kink in TM1 and a lesser kink in TM5
as compared to rhodopsin and thus, showing similarity to the beta
adrenoceptor.
2.5. Binding studies
Extracellular loop 1 (EL1) contains a high number of amino
acids conserved among NTR1 and NTR2 species variants.2 Never-
theless, in the described model, we could not detect any direct
interactions between NT(8–13) and EL1, although it is positioned
directly opposite of the binding site at EL3. To assess the impor-
tance of EL1 and prove our model’s explanatory power, we con-
structed single point mutations of the most conserved aromatic
amino acids and one proline in EL1. Interestingly, mutants
F127A, H131A, P133A, and F136A showed no change in affinity
for NT(8–13) and only a minor decrease in affinity for the antago-
nist SR 48692 when compared to the wild-type receptor. However,
the amount of available binding sites was strongly reduced for the
agonist NT(8–13) (Fig. 4). On the other hand, Bmax values for the
antagonist SR 48692 remained unchanged, showing that the reduc-
tion in agonist binding sites was due to the mutation and not
caused by reduced receptor expression. The binding site of SR
48692 is supposed to be situated between the TMs, partly overlap-
ping with the binding site of NT.2 Therefore, it should be less af-
fected by mutations in the extracellular loop region and reflect
correct folding and correct arrangement of the TM bundle. Com-
parison of the ratio between Bmax values of NT(8–13) and SR
48692 for the wild-type and mutant receptors showed that the
Score ,
CSD 25 indicating cluster 5 as the best docked structure. During
the simulation, NT(8–13) moved away from the TM bundle to-
wards the core of EL3. This movement still allowed an ionic inter-
action between Arg322 at the ligand’s C-terminal end and provided
a close contact of the Tyr11 side chain with an aromatic pocket at
the receptor formed by Trp374, Tyr339, Tyr342, and His218
(Fig. 3A). The N-terminal end of NT(8–13) moved upwards into a
position near the outer regions of EL3, allowing Arg9 to interact
with Trp334. Before the last cluster was generated, the guanidi-
nium group of Arg8 settled above Trp334, where it interacted with
the backbone oxygen of Gln333 and the side chain of Thr335
(Fig. 3A and B). Additionally, a large cavity was observed around
the ligand’s C-terminal end which was only partly filled by the
Leu13 side chain.
Analysis of the dihedral angles of NT(8–13) and subsequent
comparison with the backbone conformation obtained from NMR
investigations6 showed agreement for psi 9, 11, and 12 and for
phi 12 (Table 1). Deviations were found for Arg8 psi and Arg9
phi angles. The calculated phi angle of Pro10 was in a narrow range
Figure 3. Best docked structure. Representation of cluster 5 with NT(8–13) docked into hNTR1. Not all residues are displayed. Residues of the ligand and of the presumed
binding site are coated in blue and pink, respectively. (A) Binding site at EL3, showing the aromatic pocket (H218, Trp334, Tyr339, and Tyr342) for Tyr11, the ionic interaction
of the ligand’s C-terminal end with the side chain of Arg322 and the interaction of Arg8 with Gln333 and Thr335 above EL3. (B) Putting special emphasis on ligand backbone
interaction. N –H of Arg9 is oriented towards the solvent. Ile329 and the backbone oxygen of Arg9 are in the vicinity of N –H of Tyr11. The side chain of Ile12 is next to its
a
a
own N –H position. N –H of Leu13 forms an H-bond with the backbone oxygen of Cys327.
a
a

9362
S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
Table 1
Backbone dihedral angles of NT(8–13) observed in the clusters
Angle
Lit.a
Cluster 1
Cluster 2
Cluster 3
Cluster 4
Cluster 5
R8 psi
R9 phi
R9 psi
P10 phi
P10 psi
Y11 phi
Y11 psi
I12 phi
I12 psi
L13 phi
À34 15
À134 31
136 25
À147 16
146 15
À121 22
131 19
120 16
À131 16
103 12
À72 11
119 31
À111 31
134 14
À115 15
150 11
À129 11
124 16
À131 16
112 9.2
À69 8.8
111 23
À109 24
132 13
À109 15
151 12
À139 15
122 14
À130 16
99 10
123 19
À128 21
117 13
À70 10
101 26
À104 27
147 12
À126 13
157 14
À144 14
110 15
53 17
113 10
À73 8.8
86 18
À90 21
139 15
À125 15
155 14
À142 14
À68 11
137 25
À143 22
138 15
À119 15
156 14
À145 14
À114
8
134 24
À104 16
Data represents mean values and standard deviation as observed for NT(8–13) in the docked MD simulation.
a
6
See Luca et al.
of binding (Table 3), indicating a loss of backbone interaction or
conformational torsion. Even more, introduction of N-Me-Tyr11 re-
sulted in only poor affinity to the active conformation of hNTR1 la-
beled with [3H]NT(8–13) and a Ki that was almost 10,000-fold
below the Kd value (Table 3). Our ligand–receptor model indicated
that this difference is the effect of intramolecular clashes of the
introduced N-methyl substituent with the carbonyl oxygen of the
N-acylproline unit (Fig. 3B). Alternatively, repulsive receptor–li-
gand interactions with Tyr328, Ile329, and Ser330 could be respon-
sible for the poor radioligand displacement. Especially the side
chain of Ile329 was in close vicinity of Tyr11-N and seemed to
a
take part in receptor–ligand interaction (Fig. 3B). Thus, we con-
structed the I329G single and YIS328-330AAA triple mutants.
However, the affinity of both NT(8–13) and the congener 2 was
unaffected by these mutations (Table 3). The binding affinity of
the adamantane amino acid substituted derivative [Aac13]NT(8–
13) (see experimental), was comparable to NT.
Figure 4. Competition binding curves. Graph showing the data obtained in
homologous competition binding assays with [3H]NT(8–13) at receptors mutated
at EL1. Values from the competition binding assays were normalized for each
receptor mutant and multiplied with the corresponding Bmax SR 48692/NT(8–13)
ratio, showing the reduction in specific receptor binding of NT(8–13).
Table 3
Effect of backbone methylation on ligand affinity and efficacy
amount of agonist binding sites was reduced to less than 10% for
P133A and F136A. No specific binding could be detected for
NT(8–13) at the W129A and W134A mutants. On the other hand,
ligand affinity of SR 48692 remained almost unchanged when com-
pared to the wild-type (Table 2).
To break down the binding energy to specific contacts and, thus,
the influence of backbone-induced H-bonds and conformational
restraints, a microwave assisted solid-phase peptide synthesis
protocol was applied to yield the backbone methylated NT(8–13)
analogs [N-Me-Arg9]NT(8–13) (1), [N-Me-Tyr11]NT(8–13) (2),7 [N-
Me-Ile12]NT(8–13) (3), and [N-Me-Leu13]NT(8–13) (4). Receptor
binding studies of the NT(8–13) surrogate 1 showed that substitu-
tion at Arg9 did not significantly influence ligand affinity (Ki
0.51 nM). This indicates that the disposition might be at a sterically
less demanding entrance of the binding pocket. On the other hand,
exchange of Ile12 and Leu13 led to a more than 200-fold reduction
Compound
[3H]NT(8–13)
Rel. Ca2+ signal
( SEM)
displacement ( SEM)
NT
NT(8–13)
2.1 nM ( 12%)
0.63 nM ( 11%)
0.51 nM ( 5.7%)
100% ( 16%)
385% ( 24%)
362% ( 17%)
À154% ( 52%)
96% ( 8%)
[N-Me-Arg9]NT(8–13)
[N-Me-Tyr11]NT(8–13)
[N-Me-Ile12]NT(8–13)
[N-Me-Leu13]NT(8–13)
[Aac13]NT(8–13)
5.1 lM ( 21%)
0.16
lM ( 19%)
0.19
lM ( 17%)
238% ( 9%)
180% ( 11%)
2.4 nM ( 8.7%)
Receptor
Kd NT(8–13)
Ki compound 2
Wild-type
I329G
YIS328-330AAA
0.63 nM ( 11%)
0.47nM ( 23%)
0.80 nM ( 36%)
5.1
4.1
7.1
l
l
l
M ( 21%)
M ( 19%)
M ( 39%)
Data obtained from competition binding assays with [3H]NT(8–13) (2–8 experi-
ments in triplicate). Relative calcium signal refers to calcium levels observed at
100 nM relative to NT.
Table 2
Radioligand binding studies at the EL1 mutants
Receptor
Kd NT(8–13)
Bmax (pmol/mg)
Kd SR 48692
Bmax (pmol/mg)
Bmax SR 48692/NT(8–13) (%)
Wild-type
F127A
W129A
H131A
P133A
0.63 nM (0.49–0.80)
0.74 nM (0.60–0.94)
Not detectable
0.64 nM (0.51–0.84)
0.64 nM (0.39–1.1)
Not detectable
11
4.8
—
3.5
0.47
—
8.9 nM (7.5–11)
19 nM (12–29)
27 nM (19–38)
8.4 nM (6.9–10)
9.2 nM (7.1–12)
23 nM (12–44)
33 nM (17–62)
15
18
9.4
16
9.3
10
10
100
36
—
30
7
W134A
F136A
—
9
0.46 nM (0.36–0.64)
0.71
Data are derived from normalized curves of 4–6 experiments done in triplicate (values in brackets are 95% confidence intervals). Bmax ratio percentages are calculated relative
to the wild-type receptor ratio.
Data obtained from homologous competition binding assays with [3H]NT(8–13) and [3H]SR 48692, respectively.

S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
9363
2.6. Ligand potency
For Arg9, Tyr11, Ile12, and Leu13 both maps, representing methyl-
ated and non-methylated model peptides, had low energy phi and
psi angles corresponding to those found in the trajectory. Backbone
dihedral angles of Pro10 next to the methylated residue regarding
[N-Me-Tyr11]NT(8–13) (2), were found in a region which partly
showed an increase in conformational energy. Because of proline
being a particular amino acid known to differ in its conformation
from most other amino acids,26 we also calculated conformational
energy plots for N-Ac-Pro-Xxx-Ala-NMe where Xxx = Tyr and
(NMe)Tyr. As expected, the resulting maps showed additionally
confined allowed regions, but gave similar results for the observed
angles. The highly correlated changes between psi of Pro10 and phi
of Tyr11, observed in the trajectory, inspired us to also create con-
formational energy maps for the combination of these two angles,
including the conformation observed in cluster 3. The results indi-
cated that the angles of cluster 5 were at an energetically more
In order to evaluate the effect of backbone methylation on Gq
transactivation, the N-methyl substituted peptide analogs 1–4
were tested for their ability to increase the intracellular calcium
concentration. Comparison of calcium release at concentrations
of 100 nM indicated, that 1 was almost as effective as NT(8–13)
whereas 3 and 4 showed reduced activation properties (Table 3).
These results were in agreement with data from receptor binding
studies. [N-Me-Tyr11]NT(8–13), however, reduced the calcium sig-
nal to À154% which is comparable to neurotensin when the abso-
lute values are compared. To further investigate this effect, the
experiments were performed at different concentrations revealing
a dose-dependent inverse agonist effect of [N-Me-Tyr11]NT(8–13)
(2) (Fig. 5). This effect is specifically mediated by the hNTR1 as cal-
cium levels in untransfected HEK cells did not differ from control
upon administration of 2.
favorable region (À1.1 kcal molÀ1
) than those of cluster 3
(+1.03 kcal molÀ1) (Fig. 6). Methylation at tyrosine reversed the sit-
uation. Now the conformation observed in cluster 3 was at the rim
of the energetically most favorable region (7.85 kcal molÀ1) and
the conformation of cluster 5 was less favorable (12.7 kcal molÀ1).
2.7. Conformational energy calculations
To examine the effect of backbone methylation on the confor-
mation of NT(8–13), force field based energy maps were created,
based on a capped alanine tripeptide as truncated model peptide.
3. Discussion
The neurotensin receptor NTR1 is a prototypical peptidergic
GPCR that plays an important role in the discovery of antipsychotic
drugs. Aiming to better understand the overall shape and plasticity
of the binding site crevice and to identify crucial interactions, this
study benefits from a tight interplay of site-directed mutagenesis,
in silico investigations and SAR studies.12
To make sure that the binding site of NT(8–13) is not placed
within the TM bundle, like in aminergic GPCRs or opioid receptors,
we created a preliminary model based on the
l opioid receptor
model published by Mosberg and co-workers28 Docking of NT(8–
13) into this structure according to opioid ligands revealed that
the central part of the binding pocket was lined by Arg148,
Asp149, Tyr153, Tyr319, Phe35, and Tyr354. These residues have
already been examined by mutagenesis studies and no direct influ-
ence on NT binding could be detected.29 For a great number of pep-
tidergic GPCRs, ligand binding sites are suggested to be located
mainly at the extracellular ends of TM6, TM7 and the connecting
EL3, which is likely to be exposed to solvent.30 Since vacuum MD
simulations can lead to severe distortions of loop structures,31
Figure 5. Concentration dependent calcium release. AUC of calcium fluorescence
signals induced by hNTR1 activation at different concentrations of NT(8–13), NT and
[N-Me-Tyr11]NT(8–13). Control values were derived by the addition of pure buffer.
Figure 6. Conformational energy maps. Data obtained by GRIDSEARCH, sampling backbone phi and psi angles of N-Ac-Pro-Xxx-Ala-NMe where Xxx = Tyr and (NMe)Tyr.
Angles found in clusters 3 and 5 are indicated in the plots. The combination of proline-psi and tyrosine-phi angles is depicted, showing an increase in conformational energy
upon methylation.

9364
S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
we chose an explicit water/membrane environment for the
equilibration of our rhodopsin-based homology model, which
adopted a stable conformation after moderate MD simulation time.
Instead of creating a single heuristic best model, we took advan-
tage of an efficient docking procedure, building and scoring an
ensemble of different models.32 Since we expected the structurally
nonconserved extracellular loop 3 to form a significant part of the
binding site of NTR1, we chose to dock our ligand into a model, in
which the carefully constructed EL3 had already been intensely
equilibrated. MD simulation of the ligand–receptor complex al-
lowed ligand and receptor to adapt to each other and form energet-
ically favorable interactions, finally leading to cluster 5 with the
best DrugScore result and lowest RMSD of the binding site. The
obvious improvement during the first quarter of the MD simula-
tion was caused by the described movement of NT(8–13) towards
the aromatic pocket at EL3. Peptide backbone dihedral angles
agreed with experimentally derived values for psi 9, 11, and 12
and for phi 12. Deviations, observed at psi 8 and phi 9, were prob-
ably due to the solvent exposed position of Arg8. High flexibility
existed in this residue’s side chain until it finally settled at the de-
scribed site above EL3. The correlated shifts in psi 10 and phi 11
were due to flexibility in the backbone orientation while the side
chains of the two residues showed only minor fluctuations.
Figure 7. Functional role of EL1. Representation of the region around EL1 as an
abstract model showing the presumed function. Asp138 and Arg142 anchor TM3 to
the membrane, aromatic residues in EL1 form a rigid
disulfide bond connects TM3 and EL2.
p-stacking cluster and the
Besides the above discussed primary binding site that proved to
form a tunnel between extracellular loop 3, TM6 and TM7, extra-
cellular loop 1 was expected to build a crucial border area. Interest-
ingly, a DXXCR motif at the interface between TM3 and EL1 is
conserved among peptidergic GPCRs and it was shown for the
V1A vasopressin receptor that this motif is important for receptor
activation.20 NTR1 mutants R142G and D138A were reported ear-
lier to lead to a complete loss of ligand binding but no explanation
could be provided.2 As observed in our MD simulation, these two
residues ‘snorkeled’ in the membrane/water interface region,33
thereby anchoring the position of TM3 at the membrane. Mutagen-
esis of the EL1 residues revealed a reduction of Bmax values in com-
bination with no change in binding affinity of the agonist
[3H]NT(8–13). This data points to a loss of high affinity binding
sites which are needed for receptor activation. A comparable im-
pact was observed at some residues in EL3 which are not involved
in ligand binding but receptor activation.34 Furthermore, at the C5a
receptor the conserved WXFG motif at the C-terminal end of EL1,
which is also present in the hNTR1, was reported to be important
for receptor activation.35 Our homology model predicted aromatic
agreement with the model, revealing that the amide function was
outward directed, lacking attractive interactions with the receptor
protein. On the other hand, our model indicated a crucial H-bond
between the carbonyl oxygen of Cys327 and
(Fig. 3B), explaining the moderate loss in affinity for [N-Me-
Leu13]NT(8–13) (Ki = 0.19
M). The decrease in affinity observed
N
a
of Leu13
l
for [N-Me-Ile12]NT(8–13) is obviously due to intramolecular steric
interactions with the CH2CH3 group of the side chain (Fig. 3B). Sup-
porting evidence is found in the literature when it was shown that
exchange of the NH function by a CH2 group did not alter binding
affinity.36 Special susceptibility was expected from backbone N-
methylation at the Tyr11 position. We could not detect a hydrogen
bond that could be lost and the triple mutant YIS328-330AAA indi-
cated that additional space did not improve binding affinity of [N-
Me-Tyr11]NT(8–13). However, intramolecular clashes were de-
tected between the N-methyl function and the carbonyl oxygen
of Arg9. Thus, we could deduce that [N-Me-Tyr11]NT(8–13) was
unable to adopt a secondary structure similar to the bioactive con-
formation of NT(8–13) evolved from the molecular dynamics sim-
ulation. In fact, an almost 10,000-fold reduction of [3H]NT(8–13)
displacement was observed for [N-Me-Tyr11]NT(8–13).
p-stacking between Phe127, Phe136, Trp129, and Trp134. This re-
sults in a rigid EL1, forming a kind of clamp, which keeps TM3 and
TM2 together. In combination with the membrane anchoring ob-
served in our MD simulation, EL1 seems to stabilize the extracellu-
lar regions of TM2, TM3 and parts of EL2 which is connected to
Comparison of the conformational energies of backbone phi/psi
angles shows that methylation has little effect in regard to Arg9,
Leu13, and Ile12. Methylation at Tyr11, however, considerably
influences psi of Pro10 and phi of Tyr11. After methylation, the
conformation observed in cluster 5 of our molecular dynamics sim-
ulation is energetically less favored. On the other hand, the inter-
TM3 by a disulfide bond (Fig. 7). Perturbation of the
p p interac-
À
tions in this region by means of site-directed mutagenesis obvi-
ously interferes with receptor activation and strongly reduces
ligand binding.
mediately generated cluster
3 of our trajectory seemed to
represent a structural family that could be in good agreement with
the conformational requirements for [N-Me-Tyr11]NT(8–13) (2).
Since cluster 3 is energetically less favored than cluster 5, the con-
formational restraint inflicted by N-methylation in position 11
might hinder the induced fit of the ligand and prevent agonist
binding as well as activation. This is in agreement with data from
our functional calcium assay, clearly revealing inverse agonist
properties for the peptide analog 2. As a functional switch, back-
bone conformation at the Pro10/Tyr11 region might be of funda-
mental importance for receptor activation when methylation at
this position induces a change in the preferred binding mode. This
would explain why the compound is a strong inverse agonist at
nanomolar concentrations but shows only micromolar affinity
when competing with the agonist NT(8–13).
According to our receptor model, the binding site is primarily
forming a tunnel between extracellular loop 3, TM6 and TM7 har-
boring major parts of the residues 11, 12, and 13 whereas residues
8–10 are located at the outer surface of the loop. The cavity, which
is only partly filled by Leu13, could easily accommodate a sterically
highly demanding adamantane residue of the respective unnatural
amino acid (Fig. 8). This observation could be verified by chemical
synthesis and subsequent ligand binding experiments revealing an
affinity of [Aac13]NT(8–13) that is comparable to neurotensin
(Ki = 2.4 nM) (Table 3). Backbone N-methylations were chosen as
a further instrument to characterize the shape of the binding pock-
et and the bioactive conformation of NT(8–13). Thus, methylation
of N at Arg9 leads to the peptide analog [N-Me-Arg9]NT(8–13)
a
that showed binding properties similar to NT(8–13) which was in

S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
9365
Figure 8. Docked NT(8–13) and [Aac13]NT(8–13). The receptor is represented as solvent accessible surface. (A) NT(8–13) is shown in cyan, the sterically demanding
adamantyl substituted Leu13 is shown in pink within the binding site tunnel. (B) NT(8–13) is shown as CPK representation within the binding site.
5.2. Receptor preparation
4. Conclusion
HEK-293 cells, transiently transfected with wild-type and mu-
tant receptor cDNA by the CaHPO4 method,40 were cultured in
150-mm Petri plates containing 20 mL of MEM alpha medium sup-
plemented with 10% (v/v) fetal bovine serum, 100 U/mL penicillin
The benefit of GPCR homology modeling based on the crystal
structure of bovine rhodopsin remains controversial.14,37 In
addition, we had to omit the long N-terminal and C-terminal
ends in order to prevent unwanted bias. Therefore, producing
and implementing experimental results obtained from chemical
synthesis and site-directed mutagenesis is essential to overcome
the above mentioned limitations. Recently the crystal structure
of the beta adrenoreceptor has been solved and future will tell
whether models based on the adrenoceptor crystal structure
will be more accurate than those based on the rhodopsin
crystal.
Our model correctly predicted that EL1 does not take part in li-
gand binding and we were able to deduce a hypothesis for the ef-
fect of mutations in EL1, based on published results, radioligand
binding studies and homology modeling. This data showed that
EL1 forms an integral part in receptor function and signal transmis-
sion. Furthermore we identified [N-Me-Tyr11]NT(8–13) as an in-
verse agonist and we derived an explanation for the dramatic
loss of binding affinity of this compound when compared to back-
bone methylation at Arg9, Ile12, or Leu13.
G, 100 lg/mL streptomycin, and 2 mM L-glutamine.
HEK-293 cells were harvested 48 h after transfection. Cells were
removed by aspiration of the medium, followed by a wash with
phosphate buffered saline, which was discarded, resuspension in
10 mL of harvest buffer (10 mM Tris–HCl, 0.5 mM EDTA, 5.5 mM
KCl, and 140 mM NaCl, pH 7.4), scraping the cells with a rubber
spatula into a centrifuge tube, and collection of cells by centrifuga-
tion at 220g for 8 min. The cellular pellet was resuspended in 5 mL
of homogenate buffer (50 mM Tris–HCl, 5 mM EDTA, 1.5 mM CaCl2,
5 mM MgCl2, 5 mM KCl, and 120 mM NaCl, pH 7.4) and stored at
À80 °C. After thawing, the cells were diluted in homogenate buffer,
homogenized using a Polytron (20,000 rpm, 5 Â 5 s each in an ice
bath), and spun at 50,000g for 15 min. The membrane pellet was
resuspended in storage buffer (50 mM Tris–HCl, 1 mM EDTA,
5 mM MgCl2 0.1 mM dithiothreitol, 100
lg/mL bacitracin, and
5
lg/mL soybean trypsin inhibitor, pH 7.4), homogenized with a
Potter-Elvehjem homogenizer, and stored in small aliquots at
À80 °C. Protein concentration of the membrane preparation was
estimated by the method of Lowry et al.41 using bovine serum
albumin as a standard.
NT agonists and antagonists were described as potential drugs
in the treatment of schizophrenia and Parkinson’s disease,
respectively.1,38 In combination with a recently published phar-
macophore hypothesis based on SAR studies,12 our results will
help to guide the development of new ligands for NTR1 and facil-
itate a better understanding of the functionality of peptidergic
GPCRs.
5.3. Radioligand binding studies
The radioligand [3H]NT(8–13) was synthesized in collaboration
with Amersham/GE Healthcare UK Limited and obtained with a
specific activity of 136 Ci/mmol. The neurotensin receptor binding
assay followed a previously published protocol7 and was adapted
to 96-well plates, employing 0.2 nM [3H]NT(8–13) or 0.8 nM
[3H]SR 48692 (specific activity 81 Ci/mmol; Amersham Biosciences
UK Limited), respectively. Unlabeled SR 48692 was a generous gift
from the Chemical Synthesis and Drug Supply Program of the Na-
tional Institute of Mental Health. Ki values were derived from the
corresponding EC50 data.42
5. Experimental
5.1. Site-directed mutagenesis
The hNTR1 cDNA, subcloned into a pcDNA3.1(+) eukaryotic
expression vector, was purchased from the UMR cDNA Resource
Center. Oligonucleotidic primers were purchased from Bio-
mers.net. Site-directed mutagenesis was performed by polymerase
chain reaction (PCR) using oligodeoxynucleotides bearing the de-
sired mutation.39 Fidelity of PCR amplification and introduction
of mutations in the receptor cDNAs was confirmed by sequencing
with the ABI sequencer system (ABI Systems, Weiterstadt, Ger-
many) at the laboratory of C.-M. Becker (Department of Biochem-
5.4. Functional calcium assay
Fluorescence measurement of intracellular calcium release was
carried out applying a Victor3V multilabel counter (Perkin Elmer,
Freiburg, Germany) with a 485/510 nm filter set, following an
adapted protocol published by Lee et al.43 and Kassack et al.44 In
istry,
FAU
Erlangen,
Germany)
using
appropriate
oligodeoxynucleotidic primers.

9366
S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
brief, HEK-293 cells were harvested 48 h after transient transfec-
tion with wild-type hNTR1 cDNA and kept at 37 °C/5% CO2 for
15 min. After centrifugation, the cell pellet was washed once with
calcium free Krebs-Hepes buffer (118 mM NaCl, 4.7 mM KCl,
HCO2H in 18 min, System 2 (S2) if not noted more specifically:
3–40% CH3CN in H2O + 0.1% HCO2H in 26 min.
5.5.2. H-Arg-N-Me-Arg-Pro-Tyr-Ile-Leu-OH ([N-Me-Arg9]NT(8–13))
(1)
1.2 mM MgSO4, 1.2 mM KH2PO4, 4.2 mM NaHCO3, 117 mM
D-glu-
cose, and 10 mM Hepes, pH 7.4) and incubated with 2 M Fluo-4
l
The synthesis was performed starting from commercially avail-
able Fmoc-Leu-Wang resin. Fmoc-N-Me-Arg(Mtr)-OH and Fmoc-
Arg(Pbf)-OH were each coupled using a double coupling, first acyl-
ation: amino acid/HATU/DIPEA 5 equiv/5 equiv /5 equiv second
acylation: BTC protocol as described for peptide 4, both cycles un-
der microwave irradiation. Purification: gradient: 5–22% A in
26.9 min, then isocratic elution, tR: 28.7 min, ESI-MS: [M+H]+;
Calcd: 831.5, found: 831.5. Purity: S1: >99% (tR: 12.7 min); S2:
>99% (tR: 14.1 min).
AM (Invitrogen, Karlsruhe, Germany) and 2.5 mM probenecid (Sig-
ma–Aldrich, Steinheim, Germany) for 1 h in the dark at 37 °C/5%
CO2. Cells were washed two times and split in 96-well plates with
105 cells in 80
lL Krebs-Hepes buffer per well. Plates were kept at
37 °C/5% CO2 for 30 min. Baseline fluorescence was measured for
48 s. The fluorescence signals of agonist stimulated alterations in
intracellular calcium levels were collected at 8 s intervals for
200 s. Data were analyzed using PRISM (GraphPad Software, San
Diego, CA). For quantification of ligand efficacy, the area under
the curve (AUC) was calculated from the fluorescence signals of
À16 s to +48 s, the test compound being added at 0 s. The average
value of the fluorescence signals from 56 to 200 s was used for
baseline correction.
5.5.3. H-Arg-Arg-Pro-Tyr-N-Me-Ile-Leu-OH ([N-Me-Ile12]NT(8–13))
(3)
The synthesis was performed starting from commercially avail-
able Fmoc-Leu-Wang resin (loading 0.8 mmol/g). PyBOP-coupling
of Fmoc-N-Me-Ile-OH was repeated 2Â and the following Fmoc-
Tyr-OH was coupled with BTC as described for peptide 4 (single
coupling without microwave). Purification: gradient: 5–26% A in
26.9 min, tR: 23.8 min, ESI-MS: [M+H]+ Calcd: 831.5, found:
831.6. Purity: S1: 99% (tR: 13.8 min); S2: 98% (tR: 14.5 min).
5.5. Peptide synthesis
General method for the microwave assisted Fmoc deprotection/
acylation procedure: the peptide synthesis was achieved by man-
ual microwave assisted solid-phase techniques starting from the
respective resin. Amino acids were incorporated as their commer-
cially available derivatives in the following order: Fmoc-N-Me-Leu-
OH, Fmoc-Ile-OH, or Fmoc-N-Me-Ile-OH, Fmoc-Tyr(tBu)-OH, Fmoc-
Pro-OH, Fmoc-Arg(Pbf)-OH, or Fmoc-N-Me-Arg(Mtr)-OH, Fmoc-
Arg(Pbf)-OH. If not specified more precisely, elongation of the pep-
tide chain was done by repetitive cycles of Fmoc deprotection
applying 20% piperidine in DMF (microwave irradiation: 5 Â 5 s,
100 W), followed by 5 washings with DMF and subsequent peptide
coupling employing 5 equiv of the corresponding Fmoc-amino
acid/PyBOP/diisopropylethylamine and 7.5 equiv HOBt, dissolved
in a minimum amount of DMF (irradiation: 15 Â 10 s, 50 W). In be-
tween each irradiation step, cooling of the reaction mixture to a
temperature of À10 °C was achieved by sufficient agitation in an
ethanol-ice bath. Coupling of the amino acid next to the N-methyl-
ated derivative was performed with BTC/2,6-lutidine in dioxan.
After the last acylation step the N-terminal Fmoc-residue was
deprotected, the resin was rinsed ten times with CH2Cl2 and dried
in vacuo. The cleavage from the resin was performed using a mix-
ture of TFA/phenol/H2O/triisopropylsilane 88:5:5:2 for 2 h, in case
of peptide 1 for 6 h. After evaporation of the solvent and precipita-
tion in tert-butylmethylether, the crude peptides were purified
using preparative RP-HPLC and subsequent lyophilization. Peptide
purity and identity were assessed by analytical HPLC. The proce-
dures yielded the N-methylated peptides [N-Me-Arg9]NT(8–13)
(1), [N-Me-Ile12]NT(8–13) (3), [N-Me-Leu13]NT(8–13) (4), and the
NT(8–13) derivative [Aac13]NT(8–13) (5) with 2-aminoadaman-
tane-2-carboxylic acid in position 13.
5.5.4. H-Arg-Arg-Pro-Tyr-Ile-N-Me-Leu-OH ([N-Me-Leu13]NT(8–13))
(4)
Commercially available 2-chlorotritylchloride resin (loading:
1.9 mmol/g) was stirred in a solution of Fmoc-N-Me-Leu-OH (1
equiv)/ diisopropylethylamine (DIPEA, 4 equiv) in CH2Cl2. After
2 h the resin was washed 3Â with CH2Cl2/CH3OH/DIPEA (17:2:1),
3Â CH2Cl2, 2Â DMF, 2Â CH2Cl2 and dried in vacuo. After removing
the Fmoc group, Fmoc-Ile-OH (5 equiv) was activated with BTC45
(1.67 equiv) and 2,6-lutidine (12.50 equiv) in dioxane and added
to the resin, that was suspended in a solution of DIPEA (10 equiv)
in dioxane.
BTC-coupling was repeated using microwave irradiation and
peptide synthesis was finished according to the protocol described
above. Purification: gradient: 5–30% A in 26.9 min, tR: 24.2 min,
ESI-MS: [M+H]+; Calcd: 831.5, found: 831.6. Purity: S1: 98% (tR:
15.3 min); S2: 98% (tR: 16.0 min).
5.5.5. H-Arg-Arg-Pro-Tyr-Ile-Aac-OH ([Aac13]NT(8–13)) (5)
At first, Fmoc-2-aminoadamantane-2-carboxylic acid (Fmoc-
Aac) was synthesized. 2-Aminoadamantyl-2-carboxylic acid
hydrochloride (509 mg, 2.20 mmol) was suspended in CH2Cl2
(20 mL) and subsequently treated with DIPEA (1.13 mL,
6.60 mmol) and TMSCl (280 lL, 2.20 mmol). The mixture was stir-
red at room temperature for 20 min, and then a solution of FmocCl
(568 mg, 2.20 mmol) in CH2Cl2 (5 mL) was added dropwise. After
stirring for 2 h, aqueous HCl (2 N, 25 mL) was added and the organ-
ic layer was separated. After a further extraction of the aqueous
layer with CH2Cl2 (25 mL) the combined organic layers were dried
with MgSO4, concentrated and the residue was purified by flash
5.5.1. Purification of the peptides
Preparative RP-HPLC (Agilent 1100 preparative series) was per-
formed applying the following conditions (if not noted more spe-
column
chromatography
(CH2Cl2/methanol,
gradient
99:1 ? 97:3), furnishing 586 mg (64%) of Fmoc-Aac as a colorless
solid.
cifically): column: Zorbax Eclipse XDB-C8, 21.2 Â 50 mm, 5
lm
IR (neat) 3319 (NH), 3066, 2912, 2860 (CH), 1728, 1714,
1666 cmÀ1 (C@O); 1H NMR (CD3OD, 600 MHz)
d 1.60–1.64
particles, eluent: 0.1% TFA in acetonitrile (A) and 0.1% TFA in H2O
(B) applying a linear gradient starting from 5% A in 95% B, flow
rate: 18 mL/min. Peptide purity and identity were assessed by ana-
lytical HPLC (Agilent 1100 preparative series, column: Zorbax
(m, 2H, adamantane), 1.74–1.78 (m, 6H, adamantane), 2.03–2.11
(m, 4H, adamantane), 2.53 (br s, 2H, adamantane), 4.21 (br t, 1H,
J = 6.7 Hz, fluorene CH), 4.31 (bd, 1H, J = 6.7 Hz, OCH2), 7.29 (ddd,
2H, J = 7.5, 7.5, 1.1 Hz, aryl CH), 7.37 (ddd, 2H, J = 7.5, 7.5, 1.2 Hz,
aryl CH), 7.66 (bd, 2H, J = 7.5 Hz, aryl CH), 7.78 (bd, 2H, J = 7.5 Hz,
aryl CH); 13C NMR (CDCl3, 150 MHz) d 26.5, 26.7, 29.7, 32.5, 33.8,
37.5, 47.3 (6Â adamantane C, fluorene CH), 63.7, 67.0 (adamantane
Eclipse XDB-C8 analytical column, 4.6 Â 150 mm, 5
lm, flow rate:
0.5 mL/min, detection wavelength: 220 nm) coupled to a Bruker
Esquire 2000 mass detector equipped with an ESI-trap. System 1
(S1) if not noted more specifically: 10–55% CH3OH in H2O + 0.1%

S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
9367
C, OCH2), 120.0, 125.0, 127.1, 127.8, 141.4, 143.7 (6Â aryl C), 155.9
5.7. Construction of the homology model
(carbamate C@O).
To a suspension of commercially available Wang functionalized
PS resin (100 mg, 0.11 mmol) in THF (0.5 mL) was added a solution
of Fmoc-Aac (138 mg, 0.330 mmol) and PPh3 (86.6 mg,
0.330 mmol) in THF (2 mL). After addition of a solution of di-tert-
butylazodicarboxylate (76.0 mg, 0.330 mmol) in THF (approx.
1 mL), the resin was gently agitated under nitrogen for 14 h. Sub-
sequently, the resin was separated and washed thoroughly with
THF (2Â), DMF (2Â), methanol (2Â), and CH2Cl2 (4Â). Subse-
The initial homology modeling process was conducted accord-
ing to a protocol successfully used for the D2 dopamine receptor
family,15 based on the crystal structure of bovine rhodopsin (PDB
entry 1F88)13 and applying the SYBYL6.9 programm package (Tri-
pos Inc., St. Louis, Missouri). Construction of the protein loops
was additionally influenced by a secondary structure consensus
prediction.47 Extracellular loop (EL) 2 was modeled in analogy to
bovine rhodopsin, keeping the conserved disulfide bond between
Cys141 (TM3) and Cys224 (EL2) and the hairpin structure. Second-
ary structure consensus prediction did not indicate a preferred
folding of this loop.
quently capping employing acetic acid anhydride (32
lL, 10 equiv),
2,6-collidine (45 L, 10 equiv) and DMAP (0.4 mg, 0.1 equiv) in
l
DMF was performed (microwave irradiation: 5 Â 10 s, 50 W). After
cleavage of the Fmoc protecting group, a double coupling of Fmoc-
Ile-OH (134 mg, 5 equiv) using HATU (144 mg, 5 equiv) and DIPEA
5.8. Membrane environment
(130
The peptide was then elongated as described above. Purification
particles, flow:
21.2 mL/min) gradient: 5–28% A in 32 min, then isocratic elution,
tR: 30.1 min, ESI-MS: [M+H]+; Calcd: 881.5, found: 881.5. Purity:
S1: 99% (tR: 18.2 min); S2: 99% (tR: 18.6 min).
lL, 10 equiv) in DMF (least volume possible) was carried out.
The DPPC lipids were described using a previously developed
topology file (Tieleman, see http://moose.bio.ucalgary.ca) and an
equilibrated membrane patch consisting of 128 DPPC molecules.18
An initial membrane, consisting of 180 DPPC molecules, was equil-
ibrated with the GROMACS program package23,24 under periodic
(Zorbax 300SB-C18, 21.2 Â 250 mm,
7 lm
boundary conditions, applying the GROMACS force field.
A
5.5.6. Synthesis of the formally2–3/4–5 and2–3/4–5/6–7 shuffled
NT derivatives
0.9 nm cutoff was applied to both van der Waals and Coulomb
interactions, which were completed using the particle-mesh Ewald
summation.48 Berendsen temperature bath coupling of 0.1 ps at
300 K and anisotropic pressure coupling at 1 bar in x, y, and z direc-
tions with a coupling constant of 1.0 ps were applied. All bonds
were constrained by the LINCS algorithm.49 This simulation setup
was used to equilibrate the membrane template over a total simu-
lation time of 3 ns.
The respective amino acids were introduced as the following
derivatives via the acylation/deprotection protocol described
above: Fmoc-Ala-OH, Fmoc-Lys(Boc)-OH, Fmoc-Asn(Trt)-OH,
Fmoc-Glu(OtBu)-OH, and p-Glu-OH ((S)2-pyrrolidone-5-carboxylic
acid), in case of a positive ninhydrine test (except for proline) pep-
tide coupling was repeated.
Integration of the hNTR1 model into the membrane was accom-
plished by aligning the receptor within the membrane and remov-
ing all lipid and solvent molecules closer than 0.07 nm to the
receptor with the VMD programm package.50 Its position along
the z-axis, being perpendicular to the membrane, was adjusted
according to investigations by Baldwin et al.19 The size of the sim-
ulation box was increased along the z-axis and water was added to
the empty space. Chloride and sodium ions were added to neutral-
ize the charge of the receptor and to create an isotonic environ-
ment (0.3 osmol).
5.5.7. pGlu-Tyr-Leu-Asn-Glu-Lys-Pro-Arg-Arg-Pro-Tyr-Ile-Leu-OH
(2<->3/4<->5NT)
Purification (Zorbax Eclipse XDB-C8, 21.2 Â 150 mm,
5 lm
particles, eluent: 0.1% HCO2H in acetonitrile (A) and 0.1% HCO2H
in H2O (B) applying a linear gradient 5–17% A in 95–83% B in
15 min, then isocratic elution, flow rate: 10 mL/min): tR:
19.0 min, ESI-MS: [M+H]+; Calcd: 1672.9, found: 1673.2. Purity
(column/flow rate see above): System 1 (10–55% CH3OH in
H2O + 0.1% HCO2H in 18 min): >99% (tR: 16.5 min); System 2
(3–55% CH3CN in H2O + 0.1% HCO2H in 26 min): 99% (tR:
14.2 min).
5.9. Molecular dynamics simulation
5.5.8. pGlu-Tyr-Leu-Asn-Glu-Pro-Lys-Arg-Arg-Pro-Tyr-Ile-Leu-OH
(2<->3/4<->5/6<->7NT)
After energy minimization, water and membrane were equili-
brated over 1 ns of simulation time. Strong positional restraints
of 1000 kJ molÀ1 nmÀ2 on all atoms of the protein ensured that
no artificial distortions were introduced during this process. The
parameters used were the same as described above for the pure
membrane patch. After this step, MD simulation continued with-
out restraints upon the protein, until equilibrium was reached, de-
tected by stable root mean square deviation and stable kinetic and
potential energy.
A coordinate averaged receptor structure of the timeframe be-
tween 5500 and 6500 ps was calculated. In order to relax confor-
mational strain caused by the averaging process, the resulting
structure was subjected to energy minimization.
Purification (Zorbax Eclipse XDB-C8, 21.2 Â 150 mm,
5 lm
particles, eluent: 0.1% HCO2H in acetonitrile (A) and 0.1% HCO2H
in H2O (B) applying a linear gradient 5–18% A in 95–82% B in
18 min, then isocratic elution, flow rate: 10 mL/min): tR:
19.8 min, ESI-MS: [M+H]+; Calcd: 1672.9, found: 1673.2. Purity
(column/flow rate see above): System 1 (10–55% CH3OH in
H2O + 0.1% HCO2H in 18 min): >99% (tR: 16.3 min); System 2
(3–55% CH3CN in H2O + 0.1% HCO2H in 26 min): >99% (tR:
14.1 min).
5.6. Alignment
The alignment process followed
a
procedure successfully
5.10. Docking
applied in homology modeling of the D2 dopamine receptor
family.15 The amino acid sequence of hNTR1 and bovine rho-
dopsin were retrieved from the SWISS-PROT/TrEMBL database
A docking position between the extracellular ends of TM6, TM7,
and EL2 was defined for the C-terminal carboxyl function and the
(http://www.expasy.org).46
Structurally
(SCRs) forming the transmembrane helices (TMs) were identi-
fied after 409 aligned sequences of members of family
conserved
regions
adjacent C and Cb atoms of NT(8–13), allowing the carboxyl group
a
to interact with Arg322. Mutagenesis data was introduced by
means of distance restraints between ligand and receptor, taking
special advantage of a correlation between receptor mutants and
ligand derivatives which allowed mapping of mutant effects to
A
GPCRs (obtained from http://www.gpcr.org/7tm) were added
to the alignment.16

9368
S. Härterich et al. / Bioorg. Med. Chem. 16 (2008) 9359–9368
specific residues of NT(8–13).10 Experimental data available on
NT(8–13) backbone conformation by means of solid state NMR
spectroscopy6 was introduced by means of dihedral angle re-
straints and the construction of a ligand template backbone struc-
ture, omitting side chain atoms, except for Pro10. The
MODELLER7v7 program package22 was used to construct 100 mod-
els, applying our averaged receptor model, the ligand template
structure and described restraints. Models were scored by Drug-
ScoreONLINE (http://www.ag-klebe.de), applying the empirically de-
7. Bittermann, H.; Einsiedel, J.; Hübner, H.; Gmeiner, P. J. Med. Chem. 2004, 47,
5587.
8. Bittermann, H.; Böckler, F.; Einsiedel, J.; Gmeiner, P. Chem. Eur. J. 2006, 12,
6315.
9. Hoffmann, T.; Lanig, H.; Waibel, R.; Gmeiner, P. Angew. Chem., Int. Ed. 2001, 40,
3361.
10. Barroso, S.; Richard, F.; Nicolas-Ethève, D.; Reversat, J.-L.; Bernassau, J.-M.;
Kitabgi, P.; Labbé-Jullié, C. J. Biol. Chem. 2000, 275, 328.
11. Pang, Y.-P.; Cusack, B.; Groshan, K.; Richelson, E. J. Biol. Chem. 1996, 271, 15060.
12. Einsiedel, J.; Hübner, H.; Hervet, M.; Härterich, S.; Koschatzky, S.; Gmeiner, P.
Bioorg. Med. Chem. Lett. 2008, 18, 2013.
13. Palczewski, K.; Kumasaka, T.; Hori, T.; Behnke, C. A.; Motoshima, H.; Fox, B. A.;
Trong, I. L.; Teller, D. C.; Okada, T.; Stenkamp, R. E.; Yamamoto, M.; Miyano, M.
Science 2000, 289, 739.
rived scoring function DrugScore .
CSD 25 The model yielding the best
score was used as a new template in an iterative docking process.
After minimization in vacuum, the final model was reintro-
duced into the membrane simulation box. During the first 250 ps
of MD simulation only phospholipids and water molecules were al-
lowed to move, while the receptor was fixed by strong positional
restraints (1000 kJ molÀ1 nmÀ2). After removing water seeping into
the cleft between membrane and receptor, the simulation contin-
ued for another 250 ps. Now only the backbone was kept fixed
while the side chains were allowed to move freely. Subsequently, the
restraints were reduced to 100 kJ molÀ1 nmÀ2 and 10 kJ molÀ1 nmÀ2
over 250 ps, each. After this initial equilibration of 1 ns, restraintless
MD simulation continued for another 4 ns.
14. Oliveira, L.; Hulsen, T.; Lutje Hulsik, D.; Paiva, A. C. M.; Vriend, G. FEBS Lett.
2004, 564, 269.
15. Boeckler, F.; Lanig, H.; Gmeiner, P. J. Med. Chem. 2005, 48, 694.
16. Oliveira, L.; Paiva, A. C. M.; Vriend, G. J. Comput. Aided Mol. Des. 1993, 7, 649.
17. Seelig, A.; Seelig, J. Biochemistry 1974, 13, 4839.
18. Tieleman, D. P.; Berendsen, H. J. C. J. Chem. Phys. 1996, 105, 4871.
19. Baldwin, J. M.; Schertler, G. F. X.; Unger, V. M. J. Mol. Biol. 1997, 272, 144.
20. Hawtin, S. R.; Simms, J.; Conner, M.; Lawson, Z.; Parslow, R. A.; Trim, J.;
Sheppard, A.; Wheatley, M. J. Biol. Chem. 2006, 281, 38478.
21. Williamson, P. T. F.; Bains, S.; Chung, C.; Cooke, R.; Watts, A. FEBS Lett. 2002,
518, 111.
22. Šali, A.; Blundell, T. L. J. Mol. Biol. 1993, 234, 779.
23. Berendsen, H. J. C.; van der Spoel, D.; van Drunen, R. Comput. Phys. Commun.
1995, 91, 43.
24. Lindahl, E.; Hess, B.; van der Spoel, D. J. Mol. Model. 2001, 7, 306.
25. Velec, H. F. G.; Gohlke, H.; Klebe, G. J. Med. Chem. 2005, 48, 6296.
26. MacArthur, M. W.; Thornton, J. M. J. Mol. Biol. 1991, 218, 379.
27. Cherezov, V.; Rosenbaum, D. M.; Hanson, M. A.; Rasmussen, S. G. F.; Thian, F. S.;
Kobilka, T. S.; Choi, H.-J.; Kuhn, P.; Weis, W. I.; Kobilka, B. K.; Stevens, R. C.
Science 2007, 318, 1258.
5.11. Conformational energy maps
Conformational energy calculations were performed on the N-
methylated and unmethylated tripeptide N-Ac-Ala-Xxx-Ala-NMe
where Xxx = Ala and (NMe)Ala, applying a modified method pub-
lished by Vogen et al.,51 applying the GRIDSEARCH procedure of
the Sybyl software package (Tripos Inc., St. Louis, Missouri). Sam-
pling of N-Ac-Pro-Xxx-Ala-NMe where Xxx = Tyr and (NMe)Tyr
was also conducted. Each map contained 5184 data points.
28. Fowler, C. B.; Pogozheva, I. D.; LeVine, H., III; Mosberg, H. I. Biochemistry 2004,
43, 8700.
29. Labbé-Jullié, C.; Barroso, S.; Nicolas-Etève, D.; Reversat, J.-L.; Botto, J.-M.;
Mazella, J.; Bernassau, J.-M.; Kitabgi, P. J. Biol. Chem. 1998, 273, 16351.
30. Lawson, Z.; Wheatley, M. Biochem. Soc. Trans. 2004, 32, 1048.
31. Mehler, E. L.; Periole, X.; Hassan, S. A.; Weinstein, H. J. Comput. Aided Mol. Des.
2002, 16, 841.
32. Sherman, W.; Day, T.; Jacobson, M. P.; Friesner, R. A.; Farid, R. J. Med. Chem.
2006, 49, 534.
33. Monné, M.; Nilsson, I.; Johansson, M.; Elmhed, N.; von Heijne, G. J. Mol. Biol.
1998, 284, 1177.
Acknowledgments
34. Richard, F.; Barroso, S.; Nicolas-Ethève, D.; Kitabgi, P.; Labbé-Jullié, C. Eur. J.
Pharmacol. 2001, 433, 63.
35. Klco, J. M.; Nikiforovich, G. V.; Baranski, T. J. J. Biol. Chem. 2006, 281,
12010.
36. González-Muñiz, R.; García-López, M. T.; Gómez-Monterrey, I.; Herranz, R.;
Jimeno, M. L.; Suárez-Gea, M. L.; Johansen, N. L.; Thøgersen, H.; Suzdak, P. J.
Med. Chem. 1995, 38, 1015.
37. Archer, E.; Maigret, B.; Escrieut, C.; Pradayrol, L.; Fourmy, D. TiPS 2003, 24, 36.
38. Antonelli, T.; Fuxe, K.; Tomasisni, M. C.; Mazzoni, E.; Agnati, L. F.; Tanganelli, S.;
Ferraro, L. Prog. Neurobiol. 2007, 83, 92.
39. Breitinger, H. C.; Villmann, C.; Becker, K.; Becker, C.-M. J. Biol. Chem. 2001, 276,
29657.
40. Graham, F. L.; van der Eb, A. J. Virology 1973, 52, 456.
41. Lowry, O. H.; Rosebrough, N. J.; Farr, A. L.; Randall, R. J. J. Biol. Chem. 1951, 193,
265.
42. Cheng, Y. C.; Prusoff, W. H. Biochem. Pharmacol. 1973, 22, 3099.
43. Lee, S. P.; So, C. H.; Rashid, A. J.; Varghese, G.; Cheng, R.; Lança, A. J.; O’Dowd, B.
F.; George, S. R. J. Biol. Chem. 2004, 279, 35671.
We thank C.-M. Becker, Carmen Villmann and Jana Oertel (Insti-
tute of Biochemistry, FAU), Frank Boeckler (MRC, Cambridge), Har-
ald Hübner, and Holger Bittermann for technology transfer and
helpful discussions. We also acknowledge the use of the GPCR data
base and the UMR cDNA Resource Center (http://www.cdna.org).
This work was supported by the Deutsche Forschungsgemeinschaft
(DFG Gm13/7).
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2008.08.051.
44. Kassack, M. U.; Höfgen, B.; Lehmann, J.; Eckstein, N.; Quillan, J. M.; Sadee, W. J.
Biomol. Screening 2002, 7, 233.
References and notes
45. Falb, E.; Yechezkel, T.; Salitra, Y.; Gilon, C. J. Pept. Res. 1999, 53, 507.
46. O’Donovan, C.; Martin, M. J.; Gattiker, A.; Gasteiger, E.; Bairoch, A.; Apweiler, R.
Brief. Bioinform. 2002, 3, 275.
1. Boules, M.; Fredrickson, P.; Richelson, E. Peptides 2006, 27, 2523.
2. Dobner, P. R. CMLS 2005, 62, 1946.
3. Wustrow, D. J.; Davis, M. D.; Akunne, H. C.; Corbin, A. E.; Wiley, J. N.; Wise, L. D.;
Heffner, T. G. Bioorg. Med. Chem. Lett. 1995, 5, 997.
4. Hadden, M. K.; Orwig, K. S.; Kokko, K. P.; Mazella, J.; Dix, T. A.
Neuropharmacology 2005, 49, 1149.
47. Combet, C.; Blanchet, C.; Geourjon, C.; Deléage, G. TIBS 2000, 25, 147.
48. Darden, T.; York, D.; Pederson, L. J. Chem. Phys. 1993, 98, 10089.
49. Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M. J. Comput. Chem. 1997,
18, 1463.
50. Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics Modell. 1996, 14, 33.
51. Vogen, S. M.; Paczkowski, N. J.; Kirnasky, L.; Short, A.; Whitmore, J. B.;
Sherman, S. A.; Taylor, S. M.; Sanderson, S. D. Int. Immunopharmacol.
2001, 1, 2151.
5. St-Pierre, S.; Lalonde, J.-M.; Gendreau, M.; Quirion, R.; Regoli, D.; Riouxle, F. J.
Med. Chem. 1981, 24, 370.
6. Luca, S.; White, J. F.; Sohal, A. K.; Filippov, D. V.; van Broom, J. H.; Grisshammer,
R.; Baldus, M. PNAS 2003, 100, 10706.