Molecular basis for selective serotonin reuptake inhibition by the antidepressant agent fluoxetine (Prozac)

Article abstract of DOI:10.1124/mol.113.091249

Inhibitors of the serotonin transporter (SERT) are widely used antidepressant agents, but the structural mechanism for inhibitory activity and selectivity over the closely related norepinephrine transporter (NET) is not well understood. Here we use a combination of chemical, biological, and computational methods to decipher the molecular basis for high-affinity recognition in SERT and selectivity over NET for the prototypical antidepressant drug fluoxetine (Prozac; Eli Lilly, Indianapolis, IN). We show that fluoxetine binds within the central substrate site of human SERT, in agreement with recent X-ray crystal structures of LeuBAT, an engineered monoamine-like version of the bacterial amino acid transporter LeuT. However, the binding orientation of fluoxetine is reversed in our experimentally supported model comparedwith the LeuBAT structures, emphasizing the need for careful experimental verification when extrapolating findings from crystal structures of bacterial transporters to human relatives. We find that the selectivity of fluoxetine and nisoxetine, a NET selective structural congener of fluoxetine, is controlled by residues in different regions of the transporters, indicating a complex mechanism for selective recognition of structurally similar compounds in SERT and NET. Our findings add important new information on the molecular basis for SERT/NET selectivity of antidepressants, and provide the first assessment of the potential of LeuBAT as a model system for antidepressant binding in human transporters, which is essential for future structure-based drug development of antidepressant drugs with fine-tuned transporter selectivity. Copyright

Full text of DOI:10.1124/mol.113.091249

Supplemental material to this article can be found at:
http://molpharm.aspetjournals.org/content/suppl/2014/02/10/mol.113.091249.DC1.html
1521-0111/85/5/703–714$25.00
MOLECULAR PHARMACOLOGY
http://dx.doi.org/10.1124/mol.113.091249
Mol Pharmacol 85:703–714, May 2014
Copyright ª 2014 by The American Society for Pharmacology and Experimental Therapeutics
Molecular Basis for Selective Serotonin Reuptake Inhibition by
the Antidepressant Agent Fluoxetine (Prozac) s
Jacob Andersen, Nicolai Stuhr-Hansen,1 Linda Grønborg Zachariassen, Heidi Koldsø,2
Birgit Schiøtt, Kristian Strømgaard, and Anders S. Kristensen
Department of Drug Design and Pharmacology, University of Copenhagen, Copenhagen, Denmark (J.A., N.S.-H., L.G.Z., K.S.,
A.S.K.); and Center for Insoluble Structures (inSPIN) and Interdisciplinary Nanoscience Center (iNANO), Department of
Chemistry, Aarhus University, Aarhus, Denmark (H.K., B.S.)
Received December 20, 2013; accepted February 10, 2014
ABSTRACT
Inhibitors of the serotonin transporter (SERT) are widely used LeuBAT structures, emphasizing the need for careful experimental
antidepressant agents, but the structural mechanism for inhibitory verification when extrapolating findings from crystal structures of
activity and selectivity over the closely related norepinephrine bacterial transporters to human relatives. We find that the se-
transporter (NET) is not well understood. Here we use a com- lectivity of fluoxetine and nisoxetine, a NET selective structural
bination of chemical, biological, and computational methods to congener of fluoxetine, is controlled by residues in different regions
decipher the molecular basis for high-affinity recognition in SERT of the transporters, indicating a complex mechanism for selective
and selectivity over NET for the prototypical antidepressant drug recognition of structurally similar compounds in SERT and NET.
fluoxetine (Prozac; Eli Lilly, Indianapolis, IN). We show that Our findings add important new information on the molecular basis
fluoxetine binds within the central substrate site of human SERT, for SERT/NET selectivity of antidepressants, and provide the first
in agreement with recent X-ray crystal structures of LeuBAT, an assessment of the potential of LeuBAT as a model system for
engineered monoamine-like version of the bacterial amino acid antidepressant binding in human transporters, which is essential
transporter LeuT. However, the binding orientation of fluoxetine is for future structure-based drug development of antidepressant
reversed in our experimentally supported model compared with the drugs with fine-tuned transporter selectivity.
Although compounds targeting SERT and NET have had
Introduction
important clinical significance for several decades, the molec-
The early recognition of the serotonin (5-hydroxytryptamine)
ular details underlying binding of antidepressants to these
transporter (SERT) and the norepinephrine transporter (NET)
transporters are not clearly understood. Structural informa-
as important targets for antidepressant drugs fostered extensive
tion of SERT and NET is still lacking, but X-ray crystal
drug discovery efforts dedicated to the design and synthesis of
structures of the bacterial homolog LeuT (Yamashita et al.,
compounds selectively targeting SERT and/or NET (Kristensen
2005; Krishnamurthy and Gouaux, 2012) have proved to be
et al., 2011). In 1986, fluoxetine (Prozac; Eli Lilly, Indianapolis,
excellent structural templates for its mammalian counterparts
IN) was approved as one of the first selective serotonin reuptake
inhibitors (SSRIs) for the treatment of depression, and it has
since become widely acknowledged as a breakthrough drug for
depression (Wong et al., 1995, 2005). Unlike the tricyclic
antidepressants (TCAs), fluoxetine and other SSRI drugs are
highly selective for SERT, and today the SSRIs remain among
the most widely prescribed antidepressant drugs (Waitekus and
Kirkpatrick, 2004; Bauer et al., 2008).
and facilitated identification of the location and structure of
ligand-binding sites in human transporters (Beuming et al.,
2008; Celik et al., 2008; Andersen et al., 2009, 2010; Kaufmann
et al., 2009; Koldsø et al., 2010, 2013a; Sinning et al., 2010;
Plenge et al., 2012; Severinsen et al., 2012, 2013). Structures of
LeuT have also provided direct insight into the binding
mechanism of antidepressants. SSRIs and TCAs bind LeuT
with low affinity to a site (denoted S2) located in an ex-
tracellular facing vestibule (Fig. 1) (Singh et al., 2007; Zhou
et al., 2007, 2009), leading to the proposal that antidepressant
drugs also bind to the S2 site in human transporters (Zhou
et al., 2007, 2009). In contrast, recent structures of LeuBAT, an
engineered version of LeuT with residues from SERT inserted
into the central substrate site (denoted S1), and the dopamine
transporter (DAT) from Drosophila melanogaster, displayed
high-affinity binding of antidepressants within the central S1
This work was supported by the Lundbeck Foundation, the Danish National
Research Foundation (DNRF59), and the Danish Council for Independent
Research, Natural Sciences.
1Current affiliation: Department of Chemistry, University of Copenhagen,
Copenhagen, Denmark.
2Current affiliation: Department of Biochemistry, University of Oxford,
Oxford, United Kingdom.
dx.doi.org/10.1124/mol.113.091249.
s
This article has supplemental material available at molpharm.aspetjournals.
org.
ABBREVIATIONS: b-CIT, (2)-2b-carbomethoxy-3b-(4-iodophenyl)tropane; DAT, dopamine transporter; IFD, induced-fit docking; NET, norepi-
nephrine transporter; RMSD, root-mean-square deviation; SAR, structure-activity relationship; SERT, serotonin transporter; SSRI, selective
serotonin reuptake inhibitor; TCA, tricyclic antidepressant; TM, transmembrane helix; WT, wild-type.
703

704
Andersen et al.
2013). Surprisingly few biochemical studies have addressed the
location of the fluoxetine binding site in human SERT, but it
has been found that binding of fluoxetine is chloride-dependent
and highly sensitive toward mutation of Ile172 to Met (Henry
et al., 2006; Walline et al., 2008; Tavoulari et al., 2009;
Sørensen et al., 2012). As Ile172 and the chloride binding site
are both located within the central part of SERT, these
observations are consistent with the S1 binding mode found
in LeuBAT. However, both effects have been proposed to be
allosterically induced (Walline et al., 2008; Tavoulari et al.,
2009), and do thus not unequivocally pinpoint the location of
the fluoxetine binding site. In contrast, nisoxetine, a structural
congener of fluoxetine with selectivity for NET, was recently
proposed to bind in the S2 site of NET (Wang et al., 2012).
Given the structural similarity between nisoxetine and fluox-
etine (Fig. 1) it is tempting to believe that they share the
same binding site in SERT and NET, respectively. Here we
have used a combination of chemical, biological, and compu-
tational approaches to decipher the molecular basis for binding
of fluoxetine in SERT and selectivity over NET. Our study finds
that fluoxetine binds within the S1 site of SERT and allows
for the first assessment of LeuBAT as a model system for
directly revealing the binding mode of antidepressants in
human transporters.
Materials and Methods
Synthesis. A complete description of the synthesis and full
characterization of fluoxetine, nisoxetine, and analogs 7–11 are found
in Supplemental Methods.
Molecular Biology. As expression vectors, pcDNA3.1 and pCI-
IRES-neo–containing human SERT and NET, respectively, were used
(Kristensen et al., 2004; Andersen et al., 2011). Generations of point-
mutants in SERT and NET were performed by site-directed mutagen-
esis using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA).
Multiple mutants were generated by introducing one or more
mutations into existing mutants using site-directed mutagenesis. The
mutations were verified by DNA sequencing of the entire gene (GATC
Biotech, Constance, Germany). Synthetic cDNA encoding the two 15-
fold mutants SERT-(NET S1) and NET-(SERT S1) was purchased from
GeneArt (Thermo Fisher Scientific, Waltham, MA) and subcloned into
the pCI-IRES-neo expression vector as detailed previously (Andersen
et al., 2011).
Transport Assays and Radioligand Binding Experiments.
[3H]5-Hydroxytryptamine and [3H]dopamine uptake measurements in
COS-7 cells expressing wild-type (WT) and mutant forms of SERT and
NET, and binding of [125I]-labeled b-CIT [(2)-2b-carbomethoxy-3b-(4-
iodophenyl)tropane] to membranes of COS-7 cells expressing WT and
mutant forms of SERT and NET were performed essentially as described
(Andersen et al., 2011; Sørensen et al., 2012). Detailed descriptions
of the functional uptake assay and the radioligand binding assay
are provided in Supplemental Methods.
Fig. 1. (A and B) Chemical structure and pharmacological characterization
of the SSRI fluoxetine (A) and the NET selective congener nisoxetine (B).
The binding affinities of fluoxetine and nisoxetine toward SERT (filled gray
circles) and NET (filled blue circles) were determined in an [125I]b-CIT
competition binding assay. Data points represent mean 6 S.E.M. from
triplicate determinations. (C) Left: cross-sectional illustration of the crystal
structure of LeuT in complex with the substrate leucine (C-atoms shown in
yellow) in the central S1 site and fluoxetine (FLX) (C-atoms shown in green)
in the vestibular S2 site (PDB ID 3GWW). Right: cross-sectional illustration
of the crystal structure of LeuBAT in complex with fluoxetine (C-atoms
shown in magenta) in the central S1 site (PDB ID 4MM8).
site (Fig. 1) (Penmatsa et al., 2013; Wang et al., 2013).
Combined with biochemical studies showing that most SSRIs
inhibit SERT in a competitive manner (Graham et al., 1989;
Koe et al., 1990; Apparsundaram et al., 2008), and several
residues located within the S1 site of SERT and NET have been
Computational Methods. The R- and S-enantiomers of flu-
shown to be important for binding of SSRIs (Barker et al., 1999; oxetine were docked into homology models of human SERT and
NET, which were generated and validated as described previously
(Koldsø et al., 2013b). To allow for protein flexibility during docking
calculations, the induced-fit docking (IFD) workflow within the
Schrödinger software suite (Schrödinger LLC, 2011; Sherman et al.,
2006) was used. To address the binding of fluoxetine both in the central
S1 site and the S2 site in the extracellular facing vestibule, two different
types of IFD calculations were set up for each transporter. R- and
S-fluoxetine were docked into the transporters utilizing the endogenous
substrate bound to the S1 site as the binding site definition, with the
default settings for the size of the binding site (a box measuring 26 Â 26 Â
Henry et al., 2006; Mason et al., 2007; Walline et al., 2008;
Andersen et al., 2009, 2010; Koldsø et al., 2010; Sørensen et al.,
2012), LeuBAT and Drosophila DAT seem to represent im-
proved structural frameworks for studying the molecular
pharmacology of human transporters compared with LeuT.
Fluoxetine has been cocrystallized together with both LeuT
and LeuBAT, and these studies have provided ambiguous
insight into the binding mechanism of this important SSRI
drug. Whereas fluoxetine binds to the S2 site in LeuT, it binds
to the S1 site in LeuBAT (Fig. 1) (Zhou et al., 2009; Wang et al., 26 Å3 as the outer boundary for ligand and an inner box of 10 Â 10 Â 10 Å3,

The Fluoxetine Binding Site in Human SERT
705
which should include the center of the ligand). These IFDs are
termed small. In separate calculations, IFDs were performed where the
binding site was defined from residues within the S1 site (SERT: Asp98
and Ile172; NET: Asp75 and Val148) in addition to S2 site residues
(SERT: Arg104 and Glu493; NET: Arg81 and Asp473). In these IFDs,
which are termed large, the inner box dimensions were increased to
20 Â 20 Â 20 Å3. Additionally, Trp80 of NET was mutated to Ala during
the initial docking stage, since this residue was blocking the S2 site. The
aligning residue in SERT (Trp103) is pointing away from the S2 site in
the homology model and accordingly there was no need for mutation of
this residue during initial docking. The maximum number of output
structures was set to 20, and the recovered binding poses were ranked
according to their XP GScores and Emodel scores. The XP GScore is an
empirical scoring function that accounts for the interaction energy
between the ligand and the protein and approximates the ligand
binding free energy, while the Emodel score is a combination of the XP
GScore, the nonbonded interactions, and the internal strain of the
ligand (Friesner et al., 2004). The docking results have been divided into
structural clusters based on heteroatom root-mean-square deviation
(RMSD) , 2 Å in each docking. The structural clusters identified in each
docking setup were further divided into global clusters. The global
clusters were defined based on an RMSD , 2.3 Å between the
heteroatoms of fluoxetine. A model of fluoxetine binding in SERT (a
representative binding mode from the dominating SERT-Cluster 1) is
provided as Supplemental Material.
SERT and NET (Owens et al., 1997; Andersen et al., 2009).
To investigate the role of the aminoethyl chain of fluoxetine
for activity and selectivity toward SERT and NET, we also
prepared analogs of fluoxetine with modifications around this
chain (Table 1) (see Supplemental Methods).
The binding affinities of the synthesized compounds were
determined by displacing binding of the [125I]-labeled cocaine
analog b-CIT to human SERT and NET transiently expressed
in membranes from COS-7 cells. As expected, fluoxetine
had a high affinity toward SERT and selectivity over NET
(7 versus 887 nM), whereas nisoxetine had a high affinity
toward NET and selectivity over SERT (4 versus 167 nM) (Fig. 1;
Table 1). Hence, as recognized from early structure-activity
relationship (SAR) studies, ortho substitution of the phenoxy
ring appears to confer selectivity toward NET, whereas
substituents in the para position induce selectivity toward
SERT (Wong et al., 1975). However, the analog with an
unsubstituted phenoxy ring (compound 8) displayed compa-
rable affinity toward NET as nisoxetine (4 versus 12 nM)
whereas the analog with both ortho and para substitution
(compound 9) had similar binding affinity toward SERT as
fluoxetine (7 versus 4 nM) (Table 1), showing that the ortho
substituent is a minor determinant for selective binding in
NET over SERT. In contrast, the CF3 substituent is essential
for high-affinity binding in SERT and greatly reduces binding
in NET (Table 1), thus showing that this substituent is the
Results
Structure-Activity Relationship Study. Fluoxetine main determinant for the distinct selectivity profiles observed
and nisoxetine share a substituted 3-aryloxy-3-phenylpro- for fluoxetine and nisoxetine. In agreement with previous
pan-1-amine skeleton, and are distinguished by their sub- SAR studies of fluoxetine (Wong et al., 1975; Horng and Wong,
stitution on the phenoxy ring only. Where fluoxetine has 1976), we found that addition of an N-methyl group to
a CF3 group in the para position, nisoxetine has a methoxy fluoxetine (as in compound 7) reduced affinity toward SERT
group in the ortho position (Fig. 1). To delineate the role of (7 versus 37 nM) but increased the selectivity for SERT over
these two diverging structural elements for activity at SERT NET (127- versus 286-fold). Extending the aminoalkyl chain
and NET, we designed and prepared derivatives of fluoxetine of fluoxetine with one methylene group (as in compound 10)
and nisoxetine with different substituents on the phenoxy had a minor effect on the affinity for SERT (7 versus 12 nM)
ring (Table 1) (see Supplemental Methods). Furthermore, the but increased the selectivity over NET (127- versus 279-fold).
length and methyl substitution patterns of the aminoalkyl The N,N-dimethyl analog of homofluoxetine (compound 11)
chain of SSRIs and TCAs have previously been shown to be had low activity toward both SERT and NET. The length and
important determinants for activity and selectivity toward substitution pattern of the aminoalkyl chain on fluoxetine are
TABLE 1
Binding affinities of fluoxetine and nisoxetine derivatives at SERT and NET
The binding affinities at human SERT and NET were determined in an [125I]b-CIT competition binding assay, and the
selectivity ratio was calculated as Ki(NET WT)/Ki(SERT WT).
SERT/NET
Selectivity
Compound
R1
R2
R3
n
SERT WT
NET WT
nM
Fluoxetine (1)
7 6 2
3 6 1
887 6 115
1324 6 246
572 6 67
4 6 1
10,563 6 1131
12 6 2
125 6 22
3352 6 593
7958 6 1196
127
389
108
0.02
286
0.08
45
S-Fluoxetine (S-1)
H
CF3
H
H
1
R-Fluoxetine (R-1)
Nisoxetine (2)
N-Methylfluoxetine (7)
Des-CF3-fluoxetine (8)
2-OCH3-fluoxetine (9)
Homofluoxetine (10)
N-Methylhomofluoxetine (11)
5 6 1
OCH3
H
H
OCH3 CF3
H
H
H
1
1
1
1
2
2
167 6 31
37 6 4
157 6 29
3 6 1
12 6 1
992 6 34
CF3 CH3
H
H
H
H
CF3
CF3 CH3
279
8

706
Andersen et al.
thus important determinants for binding and SERT/NET
selectivity; this has also been observed for other SSRIs
and TCAs (Owens et al., 1997; Andersen et al., 2009).
Fluoxetine is a racemate consisting of a 1:1 mixture of R- and
S-enantiomers and, unlike other antidepressant drugs that are
highly enantioselective, the two enantiomers of fluoxetine have
similar binding affinities for SERT (Wong et al., 1985). We
determined the binding affinities of the R- and S-enantiomers
of fluoxetine at SERT and NET and found the two enantiomers
to be 389- and 108-fold selective toward SERT over NET,
respectively, thereby showing that the stereochemistry of
fluoxetine is an important determinant for SERT/NET selec-
tivity (Table 1). In summary, our SAR analysis demonstrates
that minor modifications of the chemical scaffold of fluoxetine
can improve the affinity toward SERT and increase the se-
lectivity over NET, whereas none of the compounds tested
had greater affinity for NET or improved selectivity over
SERT compared with nisoxetine.
Molecular Docking. To create models of possible binding
modes of fluoxetine in SERT and NET, we performed IFD
calculations of R- and S-fluoxetine into homology models of
human SERT and NET. The SERT and NET homology models
were constructed using LeuT as template (Koldsø et al., 2013b).
Previous LeuT-based models of TCA binding in human SERT
(Sinning et al., 2010) have proved to be very predictive when
compared with the recent structure of the eukaryotic Drosoph-
ila DAT in complex with the TCA nortriptyline (Penmatsa
et al., 2013) (Supplemental Fig. 1). Specifically, we observed
the aromatic lid (Tyr176/Phe335) to be broken in our IFD
calculations of fluoxetine in human SERT, showing that the
outward-occluded LeuT structure can be used as template for
human SERT and still provide an inhibitor-bound transporter
model in an outward-open conformation. Furthermore, the
overall structure of Drosophila DAT is very similar to that of
LeuT, emphasizing that the LeuT-fold is conserved from
prokaryotic to eukaryotic transporters, and together these
findings substantiate the continued use of LeuT as a structural
template for human transporters in the study of drug binding.
Initially, IFD calculations were confined to the S1 binding site
of SERT and NET (denoted small IFDs). In separate runs, the
entire S1/S2 region was included in the docking calculations
(denoted large IFDs). The docking results have been divided
into clusters based on heteroatom RMSD , 2 Å, and further Fig. 2. (A and B) Global binding clusters obtained from IFD simulations
of R-fluoxetine (C-atoms shown in yellow) and S-fluoxetine (C-atoms
mapped into global clusters describing common binding modes
by comparison of all clusters obtained from both small and
shown in orange) into homology models of SERT (A) and NET (B). The
two dominating binding clusters (SERT-Cluster 1 and NET-Cluster 1)
large IFDs of both enantiomers of fluoxetine (Fig. 2; Table 2).
The global clusters were defined based on RMSD , 2.3 Å
between the heteroatoms of fluoxetine.
Overall, we observed three global clusters of fluoxetine
binding in SERT. In the most prevalent binding mode
identified in SERT-Cluster 1 (containing 89% of the poses
obtained) (Table 2), fluoxetine is located almost entirely within
the central S1 site except for the CF3-substituted phenyl ring,
represent 89 and 67% of all R- and S-fluoxetine poses obtained from IFD
simulations into SERT and NET, respectively (Table 2). Note that the
R- and S-enantiomers are completely overlapping in the two dominating
binding clusters, whereas only a single enantiomer is found in the minor
binding clusters (SERT-Cluster 2–3 and NET-Cluster 2–3). Selected
residues in proximity of the proposed binding clusters are shown as stick
representations, and the sodium and chloride ions are shown as magenta
and green spheres, respectively. The stippled lines indicate the curvature
of the S1 and S2 sites.
which protrudes out toward the S2 site (Fig. 2A). In agreement (SERT-Cluster 2 and SERT-Cluster 3; 9 and 2%, respectively,
with the similar binding affinities of R- and S-fluoxetine for of the poses obtained) (Table 2). In these two clusters,
SERT (Table 1), the XP GScores are also very similar for the fluoxetine is exclusively located in the S2 site and the amine
two enantiomers. Additionally, a significant overlap between of the ligand is located close to Glu493 from TM10 in both
the two enantiomers in SERT-Cluster 1 is seen, where the clusters (Fig. 2A).
amine of fluoxetine is anchored between Tyr95 and Asp98 on
From IFD calculations in NET, we also found three global
transmembrane helix (TM) 1, and the unsubstituted phenyl clusters of fluoxetine binding. In the dominating binding
ring is located close to Ile168, Ile172, and Phe341 on TM3. Two mode (NET-Cluster 1; 63% of the poses obtained) (Table 2),
minor global clusters of fluoxetine binding were also obtained fluoxetine is located in the S1 site below an aromatic lid

The Fluoxetine Binding Site in Human SERT
707
TABLE 2
Results from SERT and NET docking calculations
The IFD setup is mentioned (either Small or Large). It is indicated if the cluster is located within the S1 or S2 site in addition to the cluster name
for that setup. The number of poses in each cluster is listed with respect to the total number of poses. The average XP Gscore is listed with the
standard deviation indicated in brackets. Also, the average Emodel score is listed with the standard deviation in brackets, and last it is listed in
which global cluster the setup cluster belongs to.
Protein
IFD
Compound
S1 or S2 Site
Cluster
#Pose/#Total Poses Avg. XP GScore Avg. Emodel
Global Cluster
kcal/mol
hSERT Small R-Fluoxetine
S-Fluoxetine
S1
S1
S2
S1-C1
20/20
0/20
211.7 (0.5)
211.5 (0.7)
211.0 (0.4)
274.0 (9.5) SERT-Cluster 1
271.0 (8.1) SERT-Cluster 1
265.2 (3.2) SERT-Cluster 2
Outliers
S1-C1
Outliers
19/19
0/19
Large R-Fluoxetine
S-Fluoxetine
S2-C1
Outliers
4/4
0/4
S1
S2
S1-C1
S2-C1
Outliers
1/2
1/2
0/2
212.3 (-)
211.0 (-)
281.4 (-)
261.8 (-)
SERT-Cluster 1
SERT-Cluster 3
hNET
Small R-Fluoxetine
S-Fluoxetine
S1
S1-C1
16/17
1/17
210.6 (1.3)
269.8 (5.8) NET-Cluster 1
Outliers
S1
S1
S1-C1
S1-C2
Outliers
12/19
6/19
1/19
211.4 (0.6)
29.9 (0.3)
269.1 (5.0) NET-Cluster 1
255.3 (5.6) NET-Cluster 2
Large R-Fluoxetine
S-Fluoxetine
S1
S2
S1-C2
2/5
2/5
1/5
210.6 (0.5)
210.9 (0.2)
253.0 (5.8) NET-Cluster 2
266.8 (1.8) NET-Cluster 3
S2-C1
Outliers
S1
S1
S1-C1
S1-C2
Outliers
3/8
4/8
1/8
211.7 (0.0)
210.0 (0.2)
274.8 (0.7) NET-Cluster 1
257.0 (2.4) NET-Cluster 2
(Tyr152/Phe317) (Fig. 2B). The amine of fluoxetine is co- As all five residues are located within the S1 site of SERT, these
ordinated by Asp75, the CF3-substituted phenyl ring is results suggest that SERT-Cluster 1 represents the bioactive
located just above Phe72, and the unsubstituted phenyl binding conformation of fluoxetine in SERT. This is in accordance
ring is located close to Ile144, Val148, and Phe323. There is with IFD calculations that also indicated fluoxetine to have the
a significant overlap between the two enantiomers of fluoxe- tightest binding in this cluster (Table 2), most significantly
tine in NET-Cluster 1, which is in agreement with the revealed in the Emodel scores. In contrast, mutations of residues
comparable binding affinities of the enantiomers toward NET within the S2 site generally induced ,3-fold shifts in fluoxetine
(Table 1). Two minor global clusters were also obtained from Ki (Fig. 3; Supplemental Table 1), thus speaking against the
IFDs in NET (NET-Cluster 2 and NET-Cluster 3; 24 and 4%, binding modes predicted in SERT-Cluster 2 and SERT-Cluster 3.
respectively, of the poses obtained) (Table 2). In NET-Cluster Notably, an ionic interaction between the amine of fluoxetine
2, fluoxetine is binding exclusively in the S1 site with the and the negatively charged side-chain of the S2 residue Glu493
amine and unsubstituted phenyl ring located at similar on TM10 is predicted in the two minor binding clusters (Fig. 2A).
positions as found in NET-Cluster 1, and in NET-Cluster 3 However, removing the negatively charged side-chain by the
fluoxetine is binding in the S2 site in a similar pose as found in E493A mutation had no significant effect on the potency of
SERT-Cluster 3 (Fig. 2). NET-Cluster 1 and NET-Cluster 2 fluoxetine (Fig. 3; Supplemental Table 1). Furthermore, muta-
are not similar to any of the binding modes observed for tions of six hydrophobic residues within the S2 site (Trp103,
fluoxetine within the S1 site of SERT.
Ile179, Trp182, Tyr232, Val236, and Val489) that seem to be
Experimental Validation of Suggested Binding Modes important for the overall shape of the extracellular vestibule of
of Fluoxetine in SERT. To distinguish between the three SERT, generally only led to small shifts (,3-fold) in fluoxetine Ki
obtained clusters of fluoxetine binding in SERT, we performed (Fig. 3; Supplemental Table 1), substantiating that SERT-
a mutational analysis of residues within 6 Å of the predicted Cluster 2 and SERT-Cluster 3 do not represent the bioactive
binding modes to determine their role in fluoxetine binding. In binding conformation of fluoxetine in SERT. Interestingly, TM10
total, 59 point mutations across 27 different positions in the S1 residues have previously been suggested to have an important
and S2 sites of SERT were included in the study. Nine mutants role for inhibitor binding within the S1 site of Drosophila and
rendered the transporter nonfunctional and were not studied human DAT (Bisgaard et al., 2011; Penmatsa et al., 2013). Here
further (Supplemental Table 1). The inhibitory potency (Ki) of we show that mutation of residues in TM10 of SERT (Ala486,
fluoxetine was determined at each of the 50 functional point- Val489, Lys490, and Glu493) induce ,3-fold changes in the
mutants across 24 different positions (Fig. 3; Supplemental potency of fluoxetine (Fig. 4; Supplemental Table 1), suggesting
Table 1). At five positions (Tyr95, Asp98, Ile168, Ile172, that TM10 residues in DAT have a more important role in
Asn177), point mutations induced .10-fold shifts in the Ki value inhibitor binding than TM10 residues in SERT. The amino group
for fluoxetine (ranging from 11- to 79-fold), suggesting these of fluoxetine was found to have a key role for high-affinity bind-
residues are key determinants for fluoxetine binding in SERT. ing in SERT (Table 1). According to data from SERT-Cluster 1,

708
Andersen et al.
Fig. 3. (A) Graphical summary of the fold-change in fluoxetine potency (shown on x-axis) induced by point mutations at various positions in the S1 and
S2 sites and in the S1/S2 interface (shown on y-axis). The fold-change is calculated as Ki(WT SERT)/Ki(mutant) or Ki(mutant)/Ki(WT SERT) for
mutations increasing or decreasing the potency of fluoxetine, respectively. The gray shaded region indicates ,3-fold change in fluoxetine potency. Open
circles specify that the mutations do not significantly affect the Ki value for fluoxetine compared with WT SERT, whereas gray and blue shading of data
points specify a significant change (P , 0.01; Student’s t test). Mutations producing .10-fold changes in Ki and their corresponding positions are
highlighted in blue. Fluoxetine Ki, substrate Km, and functional activity of the mutations are shown in Supplemental Table 1. (B–E) Close-up views of
a representative binding pose of S-fluoxetine (C-atoms shown in orange) in SERT from the dominating SERT-Cluster 1. Selected binding site residues
are shown as stick representations. Positions where mutation induced .10-fold change in fluoxetine Ki are highlighted in blue. Sodium and chloride ions
are shown as magenta and green spheres, respectively.
the amine forms direct interactions with Tyr95 and Asp98 on loss of potency for fluoxetine, respectively. In contrast, the
TM1, similar to what has previously been observed for N177E mutation did not significantly affect fluoxetine Ki
escitalopram (Koldsø et al., 2010). Accordingly, the D98E (Fig. 3; Supplemental Table 1), indicating that a side-chain
mutation induced a 12-fold loss of potency for fluoxetine, and bearing a carbonyl group in this position is important for
removal of the aromatic ring of Tyr95 (Y95A and Y95V) induced recognition of fluoxetine. In SERT-Cluster 1, Asn177 is located
a 40-fold loss of potency for fluoxetine (Fig. 3; Supplemental .6.5 Å away from fluoxetine but interacts with Thr439
Table 1). Interestingly, when substituting Tyr95 for Trp (Y95W) through an H-bond (Fig. 3). Hence, the N177A and N177S
we found a significant 8-fold gain of potency (Fig. 3; Supplemen- mutations might affect fluoxetine Ki in an indirect manner by
tal Table 1). Since Trp is a better cation–p interaction partner modulating the overall shape of the S1 pocket by disruption of
than Tyr (Gallivan and Dougherty, 1999), the gain of potency the H-bond between TM3 and TM8. The CF3 substituent of
induced by Y95W suggests a cation–p interaction between the fluoxetine was found to be a key determinant for obtaining
amine of fluoxetine and the aromatic side-chain of Tyr95. The high-affinity binding in SERT (Table 1). In SERT-Cluster 1,
I168F mutation induced an 11-fold gain of potency for fluoxetine the CF3-substituted phenyl ring is located in a hydrophobic
(Fig. 3; Supplemental Table 1), which is likely induced by pocket with a direct interaction between the CF3 group and the
aromatic interactions between the inserted Phe and the backbone of Gly100 and aromatic p–p stacking interactions
unsubstituted phenyl ring of fluoxetine in the lower part of with Tyr176 (Fig. 3). Backbone interactions are notoriously
the S1 site (Figs. 2 and 3). Previously, the I172M mutation was difficult to address by conventional mutagenesis, and we have
shown to decrease fluoxetine potency (Henry et al., 2006; previously shown that SERT is very sensitive to mutation of
Walline et al., 2008; Thompson et al., 2011; Sørensen et al., Tyr176 (Andersen et al., 2010). Only the conservative Y176F
2012). Here, we show that mutation of Ile172 to Ala, Gln, and mutation has so far been found to be functionally tolerated,
Met induced 6- to 79-fold loss of potency for fluoxetine, and we and it had no significant effect on fluoxetine Ki (Fig. 3;
corroborate earlier findings that Ile172 plays a key role in re- Supplemental Table 1). Therefore, as an alternative approach
cognition of fluoxetine (Fig. 3; Supplemental Table 1). Muta- to probe specific interactions between the CF3-substituted
tion of Asn177 on TM3 to Ala or Ser induced 7- and 25-fold phenyl ring of fluoxetine and SERT, we tested nisoxetine,

The Fluoxetine Binding Site in Human SERT
709
Fig. 4. (A and B) Topology diagram of SERT (A) and NET (B) illustrating the identity, TM location, and numbering of the 15 nonconserved SERT/NET
residues within 6 Å of the putative S1 site, and a graphical representation of selected multiple SERT and NET mutants (see also Supplemental Fig. 2;
Supplemental Tables 3, and 4). SERT mutants are shown on a gray background with mutations indicated in blue (A), and NET mutants are shown on
a blue background with mutations indicated in gray (B). (C–F) Inhibitory potency of fluoxetine and nisoxetine at single point-mutants in SERT (C) and
NET (D) and at multiple mutants in SERT (E) and NET (F). The inhibitory potency of fluoxetine and nisoxetine was determined in a functional uptake
inhibition assay, and data represent mean 6 S.E.M. from at least three independent experiments each performed in triplicate (Supplemental Tables 3
and 4). The stippled lines indicate the potency of the inhibitors at WT transporters. Asterisks denote significantly different Ki values compared with WT
transporters (P , 0.01; Student’s t test).
des-CF3-fluoxetine (compound 8), and 2-OCH3-fluoxetine (com- summary, our mutational analysis establishes the S1 site as the
pound 9) at selected S1 mutations that induce significant primary binding site for fluoxetine in SERT and identifies
changes in fluoxetine Ki (Supplemental Table 2). Nisoxetine, SERT-Cluster 1 as the most likely binding mode model.
compound 8, and compound 9 have different aromatic sub-
Molecular Determinants for SERT/NET Selectivity.
stituents than fluoxetine, and we thus envisioned that if the We next sought to identify specific residues within SERT and
substituted phenyl ring of fluoxetine interacts with one of the NET that determine the distinct selectivity of fluoxetine and
mutated residues, the analogs would be differentially affected by nisoxetine. We have shown that nonconserved SERT/NET
the mutation compared with fluoxetine. However, the potency of residues within the S1 site dictate the selectivity for citalopram
fluoxetine and the three analogs were generally affected to the (Andersen et al., 2011). Accordingly, we hypothesized that
same level across the tested mutations (Supplemental Table 2), the molecular determinants for fluoxetine and nisoxetine
thereby indirectly substantiating that the CF3-substituted selectivity are also found among these residues. We have
phenyl ring is located in the hydrophobic pocket, potentially previously mutated each of the 15 nonconserved SERT/NET
engaging in backbone and aromatic stacking interactions residues found within 6 Å of the S1 site to the aligning
that are difficult to address by site-directed mutagenesis. In residue in the other transporter, resulting in 15 individual

710
Andersen et al.
SERT-to-NET mutations in SERT and 15 individual NET-to- NET. Specifically, the 4-fold N9 mutant (NET-A77G-M424L-
SERT mutations in NET (Fig. 4; Supplemental Fig. 2) A426G-T453C) induced the largest effect and rendered NET
(Andersen et al., 2011). To systematically probe for the 14-fold more sensitive toward fluoxetine than NET WT (2993
individual contribution of these nonconserved S1 residues versus 217 nM). Hence, nonconserved residues within the S1
for the selectivity of fluoxetine and nisoxetine, we determined site of SERT and NET are key determinants for the selectivity
the inhibitory potency of the two inhibitors at the 15 SERT-to- of fluoxetine, which are supportive of our proposed binding
NET and the 15 NET-to-SERT mutations (Fig. 4; Supple- mode of fluoxetine (Fig. 2).
mental Tables 3 and 4). For six of the SERT-to-NET
Interchanging Binding Sites between SERT and
mutations in SERT, the potency of fluoxetine was signifi- NET. Interchanging nonconserved residues within the S1
cantly decreased (3- to 7-fold). None of the SERT mutants site of SERT and NET modulated the potency of fluoxetine
induced an increase in the potency of nisoxetine (Fig. 4; and nisoxetine. However, the selectivity was not fully re-
Supplemental Table 3). In contrast, 12 of the 15 NET-to-SERT versed by any of the tested mutants. We therefore generated
mutants in NET increased fluoxetine potency (3- to 11-fold), a series of mutant constructs in which all nonconserved
whereas only three mutants decreased the potency of residues in the S1 site and/or all nonconserved residues in the
nisoxetine (2- to 3-fold) (Fig. 4; Supplemental Table 4). Hence, S2 site were simultaneously interchanged between SERT and
as observed for citalopram (Andersen et al., 2011), inhibitor NET, and thereby in principle transplanting these binding
selectivity can be modulated, but is not controlled by a single sites from SERT into NET and vice versa. Hereby, three
nonconserved residue within the S1 site. Next, we combined SERT constructs containing NET S1: SERT-(NET S1), NET
the single point-mutants into a set of 30 multiple SERT-to- S2: SERT-(NET S2), and NET S1/S2: SERT-(NET S1S2), in
NET mutants in SERT (designated S1–S30) and 17 multiple addition to two NET constructs containing SERT S1: NET-
NET-to-SERT mutants in NET (designated N1–N17) (Fig. 4; (SERT S1) and SERT S2: NET-(SERT S2) were created (Fig.
Supplemental Tables 3 and 4). The design of these mutants 5; Supplemental Fig. 2; Supplemental Table 5). Additionally,
was initially directed by results from single point mutants we created two constructs in which all nonconserved residues
and later combined with results from multiple mutants and within 6 Å of the predicted binding mode of fluoxetine (SERT-
with the binding poses obtained from IFD calculations. Cluster 1) were interchanged—SERT-(NET S1S2i) and NET-
Functional uptake activity was retained for 29 of the 30 (SERT S1S2i), respectively. For a detailed description of
multiple SERT mutants and for 10 of the 17 multiple NET mutant constructs, see Supplemental Fig. 2 and Supplemen-
mutants (Supplemental Tables 3 and 4). The nonfunctional tal Table 5.
mutants were not studied further. For fluoxetine, a signifi-
Initially, we performed saturation binding analysis on
cant loss of potency (3- to 11-fold) was observed for eight of membrane preparations from COS-7 cells expressing WT
the multiple SERT mutants (Fig. 4; Supplemental Table 3). and mutant transporters and found that all SERT constructs
Mutations within the Ile172/Ala173/Ser174 motif on TM3 bind [125I]b-CIT, whereas only NET WT and NET-(SERT S2)
was included in the five multiple mutants that displayed showed specific [125I]b-CIT binding (Fig. 5; Supplemental
the largest loss of potency, suggesting that this motif is Table 5). Consistent with saturation binding analyses,
an important determinant for the selectivity of fluoxetine. confocal imaging of green fluorescent protein–tagged variants
Interestingly, combining mutations in the Ile172/Ala172/ of WT and mutant transporters showed that the two NET
Ser174 motif with A441G and L443M (as in S27, S28, and constructs that do not bind [125I]b-CIT—NET-(SERT S1) and
S29) induced a significant gain of potency for fluoxetine (Fig. NET-(SERT S1S2i)—were primarily retained within intra-
4; Supplemental Table 3), indicating that A441G and L443M cellular compartments (Supplemental Fig. 3). Next, we de-
hold a positive role for binding of the SSRI. For nisoxetine, termined the binding affinities of fluoxetine and nisoxetine
seven of the multiple SERT mutants induced a significant at the S1/S2 constructs that bind [125I]b-CIT in a competition
gain of potency (10- to 24-fold). Combining mutations within binding assay. For fluoxetine, insertion of NET S1 into SERT
the Ile172/Ala173/Ser174 motif with A441G and L443M on induced a 15-fold decrease in binding affinity (7 versus 102
TM8 induced the largest gain of nisoxetine potency, indicating nM), whereas insertion of the NET S2 site did not affect
that residues on TM3 and TM8 are cooperative determinants binding of fluoxetine (Fig. 5; Supplemental Table 5). In-
for binding of nisoxetine in SERT. The 3-fold S29 mutant terestingly, the decreased binding affinity of fluoxetine in
(SERT-S174F-A441G-L443M) had the largest effect, and SERT-(NET S1) was reversed by simultaneous insertion of S2
rendered SERT 24-fold more sensitive to nisoxetine than site (Fig. 5; Supplemental Table 5). The binding affinity of
SERT WT (60 versus 1422 nM), showing that key determi- nisoxetine was increased by 6-fold after inserting the S1 site
nants for nisoxetine selectivity are located within the S1 site from NET into SERT (167 versus 29 nM), whereas insertion of
of SERT.
the S2 site had no significant effect on nisoxetine. Insertion of
Determination of the potency of nisoxetine at the 10 func- both the S1 and S2 sites from NET into SERT improved the
tional multiple NET mutants surprisingly showed that all binding affinity of nisoxetine to a similar level as observed
retained WT potency of the NET selective ligand (Fig. 4; when the S1 site was inserted alone (15 versus 29 nM),
Supplemental Table 4). These data strongly suggest that, in showing that residues located in the S1 site of SERT are key
contrast to SERT, nonconserved S1 residues do not define the determinants for the selectivity of nisoxetine.
inhibitory potency of nisoxetine in NET. In contrast, all
Similar detailed analysis was not possible for NET, since
multiple NET-to-SERT mutants induced a significant gain only the NET-(SERT S2) construct could bind [125I]b-CIT.
of fluoxetine potency (2- to 14-fold) (Fig. 4; Supplemental Interestingly, the binding affinity of nisoxetine was decreased
Table 4). Mutations in TM1 (F72Y and A77G), TM8 (M424L almost to the same level as observed in SERT WT by inserting
and A426G), and TM9 (Thr453) were found to be most SERT S2 into NET (167 versus 104 nM compared with 4 nM
important for improving inhibitory potency of fluoxetine in at NET WT). Together with our initial analysis, which showed

The Fluoxetine Binding Site in Human SERT
711
Fig. 5. (A) Location of nonconserved SERT/NET residues within 6 Å of the S1 site (left), S2 site (middle), and the putative fluoxetine binding site (right)
shown on a homology model of SERT (see Supplemental Fig. 2 and Supplemental Table 5 for specific residues). (B) Saturation binding curves for
[
125I]b-CIT binding to COS-7 membranes expressing WT and mutant forms of SERT (left) and NET (right) where all nonconserved residues within 6 Å of
putative binding regions have been interchanged (for Kd and Bmax values, see Supplemental Table 5). Data points represent mean 6 S.E.M. from
duplicate determinations. (C and D) The binding affinities of fluoxetine and nisoxetine were determined in an [125I]b-CIT competition binding assay at
WT and mutant forms of SERT (C) and NET (D), and data represent mean 6 S.E.M. from at least three independent experiments each performed in
duplicate (Supplemental Table 5). The stippled lines indicate the binding affinities of fluoxetine and nisoxetine at WT transporters. Asterisks denote
significantly different Ki values compared with WT transporters (P , 0.01; Student’s t test).
nonconserved S1 residues in NET to have a minor role in the 2010; Severinsen et al., 2013). Very recently, Drosophila DAT
selectivity of nisoxetine, this result suggested that non- and LeuBAT were crystallized in complex with antidepres-
conserved residues in the S2 site of NET are key determinants sants (Penmatsa et al., 2013; Wang et al., 2013). These X-ray
for the selectivity of nisoxetine. In summary, our mutational crystal structures offer a new platform for understanding
analysis of nonconserved SERT/NET residues supports the ligand interactions that have obvious potential to further
view that the selectivity of fluoxetine for SERT over NET is push the field toward more reliable and realistic models of
largely determined by nonconserved residues within the S1 antidepressant binding in human transporters. However,
site of both SERT and NET. This interpretation is fully in both Drosophila DAT and LeuBAT are inactive in transport,
accordance with our proposed IFD models of fluoxetine and their pharmacological profiles seem to be a hybrid of the
binding within the S1 site in the two transporters. In contrast, human SERT, NET, and DAT. Thus, to assess the potential of
we found that the selectivity of nisoxetine for NET over SERT LeuBAT and Drosophila DAT for studying the molecular
is controlled by nonconserved residues in different regions of pharmacology of human transporters, it is critically impor-
SERT (S1 residues) and NET (S2 residues). Differentiation tant to establish similarities and discrepancies between
between direct and indirect effects in mutagenesis studies is Drosophila DAT and LeuBAT and their human relatives.
inherently difficult. Mutations can induce long-range alloste- For this purpose, comparison of our present fluoxetine model
ric effects that perturb a distinct binding site or induce a shift with the X-ray crystal structure of LeuBAT in complex with
in the conformational equilibrium of the transporter that fluoxetine therefore provides an excellent first opportunity for
changes the temporal accessibility to the binding site. Thus, assessment of the potential of LeuBAT as a model system for
the mutational analyses do not allow us to conclusively specify SSRI binding in human transporters. First and foremost, our
the location of the nisoxetine binding site in SERT or NET, proposed binding model of fluoxetine in SERT (SERT-Cluster
but they suggest that there is a complex mechanism for 1) is in agreement with the LeuBAT structure (Wang et al.,
selective recognition of inhibitors in SERT and NET.
2013) by showing that the inhibitor binds within the S1 site
(Figs. 2 and 3). The observed S1 binding modes in LeuBAT
and our SERT model correlate very well with our experimen-
tal validation, and can explain two key findings from the
Discussion
The models of fluoxetine binding in human SERT and NET mutational analysis. Firstly, fluoxetine potency is largely
produced in the present study are based on X-ray crystal affected by mutations in the S1 site (up to 79-fold loss-
structures of LeuT and have been constructed using well- of-potency), whereas mutations in the S2 site generally
established procedures that have been implemented for induce ,3-fold changes in fluoxetine Ki (Fig. 3; Supplemen-
modeling of other important monoamine transporter inhib- tal Table 1). Secondly, our mutational analysis suggests
itors (Andersen et al., 2010; Koldsø et al., 2010; Sinning et al., a cation–p interaction between Tyr95 and the amino group

712
Andersen et al.
of fluoxetine, which is in agreement with our binding model
showing that the amino group of fluoxetine is anchored
between Tyr95 and Asp98 in the S1 site of SERT. A similar
interaction is also observed in LeuBAT, where the amino
group of fluoxetine is coordinated by the aligning residues
(Tyr21 and Asp24, respectively) within the S1 site (Wang
et al., 2013). Together with experimentally validated models
showing other SERT inhibitors to also be anchored within
the S1 site (Andersen et al., 2010; Koldsø et al., 2010;
Sinning et al., 2010; Combs et al., 2011; Severinsen et al.,
2013), this is in contrast to previous structural studies of
LeuT in complex with antidepressants, that proposed SSRIs
and TCAs to bind in the S2 site (Zhou et al., 2007, 2009).
Hence, although crystal structures of LeuT have provided
seminal improvements for our understanding of the over-
all structure and function of SERT and NET, our results
emphasize that LeuBAT is an improved model system
compared with LeuT for understanding the mechanism of
drug binding in human transporters.
However, we do observe differences between fluoxetine
binding in our model compared with the LeuBAT structure.
Specifically, although the amine of fluoxetine is anchored
similarly between the Tyr and Asp residues, the orientation of
the two aromatic moieties of the inhibitor is reversed in our
model compared with the LeuBAT structure (Fig. 6). While
the CF3-substituted phenyl ring binds between TM3 and TM8
in the LeuBAT structure, it is protruding up toward the S2 fluoxetine (C-atoms shown in orange) in SERT from the dominating
site in our model. Notably, neither from IFDs in SERT nor in
NET did we identify any poses of fluoxetine in an orientation
Fig. 6. (A and B) Superimposition of a representative binding pose of
SERT-Cluster 1 (same binding pose as shown in Fig. 2, B–E) and the
binding mode of fluoxetine (C-atoms shown in green) in LeuT (PDB ID
3GWW) (A) and fluoxetine (C-atoms shown in magenta) in LeuBAT (B).
similar to the one in the LeuBAT structure (Fig. 2).
Furthermore, observations from our mutational analysis of
SERT are difficult to reconcile with the orientation of the two
aromatic moieties of fluoxetine as found in LeuBAT. Specif-
ically, the CF3 group of fluoxetine is sandwiched in a groove
between TM3 and TM8 in LeuBAT (Wang et al., 2013). In
contrast, in our SERT model we observed an H-bond between
the carbonyl group on the Asn177 side-chain and the hydroxyl
group of the Thr439 side-chain, which is not conserved in
LeuBAT. This H-bond constrains flexibility of TM3 and TM8
The overlays are shown on the SERT model. Selected residues in
proximity of the proposed binding clusters are shown as stick representa-
tions. Positions where mutations induce .10-fold changes in fluoxetine
potency are highlighted in dark blue. (C) Superimposition of a represen-
tative binding pose of fluoxetine (C-atoms shown in orange) in SERT
(shown in gray) and the binding mode of fluoxetine (C-atoms shown in
magenta) in LeuBAT (shown in blue). Selected residues in proximity of the
proposed binding clusters are shown as stick representations. Numbering
and identity of LeuBAT residues are shown in brackets. H-bond between
SERT residues Asn177 and Thr439 is indicated by stippled lines.
and does not allow a similar binding mode of the CF3 group of inhibitor (Fig. 6), making it less likely that mutation at this
fluoxetine in SERT (Fig. 6). Accordingly, we found that site will have a pronounced effect on fluoxetine potency.
mutations of Asn177 and Thr439 induce a marked loss of Overall, even though fluoxetine shares the same binding site
potency for fluoxetine (Figs. 3 and 4), likely due to disruption in human SERT and LeuBAT, our experimentally supported
of the H-bond and thereby the overall shape of the binding site binding model of fluoxetine in SERT suggest that the SSRI
rather than affecting direct interactions with the inhibitor. has distinct binding modes in human and bacterial trans-
Also, if fluoxetine adopts the same binding mode in SERT as porters, emphasizing the continued need for careful experi-
found in LeuBAT, Asn177 would be pointing directly toward mental validation when extrapolating findings from LeuBAT
the CF3 group of the SSRI (Fig. 6). Thus, introduction of to human transporters.
a negative charge into this subsite would likely cause an
Our study provides novel insight into the molecular deter-
electrostatic repulsion with the electronegative CF3 group, minants for selective nisoxetine binding in NET by showing
and thereby induce a loss of potency for fluoxetine. Notably, that nonconserved residues within the S2 site are important
the N177E mutation had no effect on fluoxetine potency (Fig. for high-affinity nisoxetine binding in NET (Figs. 4 and 5).
3), and since the H-bond to Thr439 can be preserved in the Mutation of residues outside the S1 site in NET has previously
N177E mutant, this provides further support for our proposed been found to affect binding of nisoxetine (Paczkowski et al.,
binding model. Additionally, mutation of Ile179 in SERT has 2007; Wenge and Bönisch, 2013). In addition, cocaine-like
previously been shown to induce a marked loss of potency for compounds, which are believed to bind in the S1 site of
fluoxetine (Zhou et al., 2009). This is in good agreement with monoamine transporters (Beuming et al., 2008), display
our binding model, in which Ile179 is located within 3.5 Å noncompetitive binding with nisoxetine in NET (Zhen et al.,
from fluoxetine and points directly toward the subsite where 2012). Together, these observations indicate that nisoxetine
the CF3 group of fluoxetine binds (Fig. 6). In contrast, if binds outside the S1 site, and are thus supportive of a recent
fluoxetine adopts a similar binding mode in SERT as found in model suggesting that nisoxetine binds to the S2 site in NET
LeuBAT, Ile179 would be located .6.5 Å away from the (Wang et al., 2012). However, nisoxetine is also affected by

The Fluoxetine Binding Site in Human SERT
713
serotonin transporter: importance of Ser-438 in recognition of citalopram and tri-
cyclic antidepressants. J Biol Chem 284:10276–10284.
Apparsundaram S, Stockdale DJ, Henningsen RA, Milla ME, and Martin RS (2008)
Antidepressants targeting the serotonin reuptake transporter act via a competitive
mechanism. J Pharmacol Exp Ther 327:982–990.
Barker EL, Moore KR, Rakhshan F, and Blakely RD (1999) Transmembrane domain
I contributes to the permeation pathway for serotonin and ions in the serotonin
transporter. J Neurosci 19:4705–4717.
Bauer M, Monz BU, Montejo AL, Quail D, Dantchev N, Demyttenaere K, Garcia-
Cebrian A, Grassi L, Perahia DGS, and Reed C et al. (2008) Prescribing patterns of
antidepressants in europe: Results from the Factors Influencing Depression End-
points Research (FINDER) study. Eur Psychiatry 23:66–73.
Beuming T, Kniazeff J, Bergmann ML, Shi L, Gracia L, Raniszewska K, Newman
AH, Javitch JA, Weinstein H, and Gether U et al. (2008) The binding sites for
cocaine and dopamine in the dopamine transporter overlap. Nat Neurosci 11:
780–789.
Bisgaard H, Larsen MAB, Mazier S, Beuming T, Newman AH, Weinstein H, Shi L,
Loland CJ, and Gether U (2011) The binding sites for benztropines and dopamine
in the dopamine transporter overlap. Neuropharmacology 60:182–190.
Celik L, Sinning S, Severinsen K, Hansen CG, Møller MS, Bols M, Wiborg O,
and Schiøtt B (2008) Binding of serotonin to the human serotonin transporter.
Molecular modeling and experimental validation. J Am Chem Soc 130:3853–3865.
Combs S, Kaufmann K, Field JR, Blakely RD, and Meiler J (2011) Y95 and E444
interaction required for high-affinity S-citalopram binding in the human serotonin
transporter. ACS Chem Neurosci 2:75–81.
Friesner RA, Banks JL, Murphy RB, Halgren TA, Klicic JJ, Mainz DT, Repasky MP,
Knoll EH, Shelley M, and Perry JK et al. (2004) Glide: a new approach for rapid,
accurate docking and scoring. 1. Method and assessment of docking accuracy.
J Med Chem 47:1739–1749.
Gallivan JP and Dougherty DA (1999) Cation-pi interactions in structural biology.
Proc Natl Acad Sci USA 96:9459–9464.
mutations of S1 residues in NET (Mason et al., 2007;
Sørensen et al., 2012), and the recent structures of Drosophila
DAT and LeuBAT showed a common inhibitor binding site to
be located within the central S1 site (Penmatsa et al., 2013;
Wang et al., 2013). Taken together, it seems most likely that
the high-affinity binding site for nisoxetine is located within
the S1 site in NET, and that the selectivity is determined by
nonconserved residues lining the S2 site that nisoxetine needs
to permeate to reach the central S1 site. In contrast, we find
that the molecular determinants that underlie the lower
potency of nisoxetine in SERT are primarily located among
nonconserved residues within the S1 site of this transporter
(Figs. 4 and 5). This is in agreement with previous findings
for other S1 residues in SERT that have been shown to be
important for recognition of nisoxetine in SERT (Walline
et al., 2008; Sørensen et al., 2012). Hence, in contrast to
fluoxetine, where S1 residues in both SERT and NET control
binding and selectivity (Figs. 3 and 4), selective binding of
nisoxetine is controlled by residues in separate regions of the
two transporters. Interestingly, the same pattern has also
been found for the SSRI escitalopram and the structurally
closely related NET selective inhibitor talopram (Andersen
et al., 2011). Thus, the finding that the selectivity of seemingly
closely related inhibitors is controlled by residues located in
different regions of two closely related transporters suggests
a complexity of the molecular pharmacology of monoamine
transporters that warrants further studies.
In summary, our findings add important new information
on the molecular basis for SERT/NET selectivity of anti-
depressants and provide the first assessment of the potential
of LeuBAT as a model system for antidepressant binding to
human transporters. Along with a growing number of other
LeuT-derived models of inhibitor binding, we can now begin to
understand the differences and similarities among the in-
hibitory mechanisms of antidepressants in a structural context.
This is essential for establishing a useful framework for
structure-based drug development of future monoamine trans-
porter drugs with fine-tuned transporter selectivity.
Graham D, Esnaud H, Habert E, and Langer SZ (1989) A common binding site for
tricyclic and nontricyclic 5-hydroxytryptamine uptake inhibitors at the substrate
recognition site of the neuronal sodium-dependent 5-hydroxytryptamine trans-
porter. Biochem Pharmacol 38:3819–3826.
Henry LK, Field JR, Adkins EM, Parnas ML, Vaughan RA, Zou MF, Newman AH,
and Blakely RD (2006) Tyr-95 and Ile-172 in transmembrane segments 1 and 3 of
human serotonin transporters interact to establish high affinity recognition of
antidepressants. J Biol Chem 281:2012–2023.
Horng JS and Wong DT (1976) Effects of serotonin uptake inhibitor, Lilly 110140, on
transport of serotonin in rat and human blood platelets. Biochem Pharmacol 25:
865–867.
Kaufmann KW, Dawson ES, Henry LK, Field JR, Blakely RD, and Meiler J (2009)
Structural determinants of species-selective substrate recognition in human and
Drosophila serotonin transporters revealed through computational docking stud-
ies. Proteins 74:630–642.
Koe BK, Lebel LA, and Welch WM (1990) [3H] sertraline binding to rat brain
membranes. Psychopharmacology (Berl) 100:470–476.
Koldsø H, Autzen HE, Grouleff J, and Schiøtt B (2013a) Ligand induced conforma-
tional changes of the human serotonin transporter revealed by molecular dynamics
simulations. PLoS ONE 8:e63635.
Koldsø H, Christiansen AB, Sinning S, and Schiøtt B (2013b) Comparative modeling
of the human monoamine transporters: similarities in substrate binding. ACS
Chem Neurosci 4:295–309.
Koldsø H, Severinsen K, Tran TT, Celik L, Jensen HH, Wiborg O, Schiøtt B,
and Sinning S (2010) The two enantiomers of citalopram bind to the human se-
rotonin transporter in reversed orientations. J Am Chem Soc 132:1311–1322.
Krishnamurthy H and Gouaux E (2012) X-ray structures of LeuT in substrate-free
outward-open and apo inward-open states. Nature 481:469–474.
Kristensen AS, Andersen J, Jørgensen TN, Sørensen L, Eriksen J, Loland CJ,
Strømgaard K, and Gether U (2011) SLC6 neurotransmitter transporters: struc-
ture, function, and regulation. Pharmacol Rev 63:585–640.
Kristensen AS, Larsen MB, Johnsen LB, and Wiborg O (2004) Mutational scanning of
the human serotonin transporter reveals fast translocating serotonin transporter
mutants. Eur J Neurosci 19:1513–1523.
Mason JN, Deecher DC, Richmond RL, Stack G, Mahaney PE, Trybulski E, Winneker
RC, and Blakely RD (2007) Desvenlafaxine succinate identifies novel antagonist
binding determinants in the human norepinephrine transporter. J Pharmacol Exp
Ther 323:720–729.
Owens MJ, Morgan WN, Plott SJ, and Nemeroff CB (1997) Neurotransmitter re-
ceptor and transporter binding profile of antidepressants and their metabolites.
J Pharmacol Exp Ther 283:1305–1322.
Acknowledgments
The authors thank Shahrokh Padrah for technical assistance with
synthesis of fluoxetine analogues and Birgitta Kegel for performing
elemental analysis of the synthesized compounds.
Authorship Contributions
Participated in research design: Andersen, Koldsø, Schiøtt,
Strømgaard, Kristensen.
Conducted experiments: Andersen, Stuhr-Hansen, Zachariassen,
Koldsø.
Performed data analysis: Andersen, Koldsø, Schiøtt, Strømgaard,
Kristensen.
Wrote or contributed to the writing of the manuscript: Andersen,
Koldsø, Schiøtt, Strømgaard, Kristensen.
Paczkowski FA, Sharpe IA, Dutertre S, and Lewis RJ (2007) x-Conotoxin and tricyclic
antidepressant interactions at the norepinephrine transporter define
a new
transporter model. J Biol Chem 282:17837–17844.
Penmatsa A, Wang KH, and Gouaux E (2013) X-ray structure of dopamine trans-
porter elucidates antidepressant mechanism. Nature 503:85–90.
Plenge P, Shi L, Beuming T, Te J, Newman AH, Weinstein H, Gether U, and Loland
CJ (2012) Steric hindrance mutagenesis in the conserved extracellular vestibule
impedes allosteric binding of antidepressants to the serotonin transporter. J Biol
Chem 287:39316–39326.
References
Andersen J, Olsen L, Hansen KB, Taboureau O, Jørgensen FS, Jørgensen AM, Bang-
Andersen B, Egebjerg J, Strømgaard K, and Kristensen AS (2010) Mutational
mapping and modeling of the binding site for (S)-citalopram in the human sero-
tonin transporter. J Biol Chem 285:2051–2063.
Schrödinger LLC (2011) Induced Fit Docking Protocol, Glide Version 5.7 and Prime
Version 3.0, Schrödinger LLC, New York.
Severinsen K, Koldsø H, Thorup KA, Schjøth-Eskesen C, Møller PT, Wiborg O,
Jensen HH, Sinning S, and Schiøtt B (2013) Binding of mazindol and analogs to the
human serotonin and dopamine transporters. Mol Pharmacol 85:208–217.
Severinsen K, Kraft JF, Koldsø H, Vinberg KA, Rothman RB, Partilla JS, Wiborg
O, Blough B, Schiøtt B, and Sinning S (2012) Binding of the amphetamine-like
1-phenyl-piperazine to monoamine transporters. ACS Chem Neurosci 3:
693–705.
Andersen J, Stuhr-Hansen N, Zachariassen L, Toubro S, Hansen SMR, Eildal
JNN, Bond AD, Bøgesø KP, Bang-Andersen B, and Kristensen AS et al. (2011)
Molecular determinants for selective recognition of antidepressants in the hu-
man serotonin and norepinephrine transporters. Proc Natl Acad Sci USA 108:
12137–12142.
Andersen J, Taboureau O, Hansen KB, Olsen L, Egebjerg J, Strømgaard K,
and Kristensen AS (2009) Location of the antidepressant binding site in the

714
Andersen et al.
Sherman W, Day T, Jacobson MP, Friesner RA, and Farid R (2006) Novel procedure
for modeling ligand/receptor induced fit effects. J Med Chem 49:534–553.
Singh SK, Yamashita A, and Gouaux E (2007) Antidepressant binding site in a bac-
terial homologue of neurotransmitter transporters. Nature 448:952–956.
Sinning S, Musgaard M, Jensen M, Severinsen K, Celik L, Koldsø H, Meyer T, Bols
M, Jensen HH, and Schiøtt B et al. (2010) Binding and orientation of tricyclic
antidepressants within the central substrate site of the human serotonin trans-
porter. J Biol Chem 285:8363–8374.
Sørensen L, Andersen J, Thomsen M, Hansen SMR, Zhao XB, Sandelin A,
Strømgaard K, and Kristensen AS (2012) Interaction of antidepressants with the
serotonin and norepinephrine transporters: mutational studies of the S1 substrate
binding pocket. J Biol Chem 287:43694–43707.
Tavoulari S, Forrest LR, and Rudnick G (2009) Fluoxetine (Prozac) binding to sero-
tonin transporter is modulated by chloride and conformational changes. J Neurosci
29:9635–9643.
Thompson BJ, Jessen T, Henry LK, Field JR, Gamble KL, Gresch PJ, Carneiro AM,
Horton RE, Chisnell PJ, and Belova Y et al. (2011) Transgenic elimination of high-
affinity antidepressant and cocaine sensitivity in the presynaptic serotonin
transporter. Proc Natl Acad Sci USA 108:3785–3790.
Waitekus AB and Kirkpatrick P (2004) Duloxetine hydrochloride. Nat Rev Drug
Discov 3:907–908.
Walline CC, Nichols DE, Carroll FI, and Barker EL (2008) Comparative molecular
field analysis using selectivity fields reveals residues in the third transmembrane
helix of the serotonin transporter associated with substrate and antagonist rec-
ognition. J Pharmacol Exp Ther 325:791–800.
Wang CIA, Shaikh NH, Ramu S, and Lewis RJ (2012) A second extracellular site is
required for norepinephrine transport by the human norepinephrine transporter.
Mol Pharmacol 82:898–909.
Wang H, Goehring A, Wang KH, Penmatsa A, Ressler R, and Gouaux E (2013)
Structural basis for action by diverse antidepressants on biogenic amine trans-
porters. Nature 503:141–145.
Wenge B and Bönisch H (2013) The role of cysteines and histidins of the norepi-
nephrine transporter. Neurochem Res 38:1303–1314.
Wong DT, Bymaster FP, and Engleman EA (1995) Prozac (fluoxetine, Lilly 110140),
the first selective serotonin uptake inhibitor and an antidepressant drug: twenty
years since its first publication. Life Sci 57:411–441.
Wong DT, Bymaster FP, Horng JS, and Molloy BB (1975) A new selective
inhibitor for uptake of serotonin into synaptosomes of rat brain: 3-(p-
trifluoromethylphenoxy). N-methyl-3-phenylpropylamine. J Pharmacol Exp
Ther 193:804–811.
Wong DT, Bymaster FP, Reid LR, Fuller RW, and Perry KW (1985) Inhibition of
serotonin uptake by optical isomers of fluoxetine. Drug Dev Res 6:397–403.
Wong DT, Perry KW, and Bymaster FP (2005) Case history: the discovery of fluox-
etine hydrochloride (Prozac). Nat Rev Drug Discov 4:764–774.
Yamashita A, Singh SK, Kawate T, Jin Y, and Gouaux E (2005) Crystal structure of
a bacterial homologue of Na1/Cl—dependent neurotransmitter transporters. Na-
ture 437:215–223.
Zhen J, Ali S, Dutta AK, and Reith ME (2012) Characterization of [³H]CFT binding to
the norepinephrine transporter suggests that binding of CFT and nisoxetine is not
mutually exclusive. J Neurosci Methods 203:19–27.
Zhou Z, Zhen J, Karpowich NK, Goetz RM, Law CJ, Reith MEA, and Wang DN (2007)
LeuT-desipramine structure reveals how antidepressants block neurotransmitter
reuptake. Science 317:1390–1393.
Zhou Z, Zhen J, Karpowich NK, Law CJ, Reith MEA, and Wang DN (2009) Antide-
pressant specificity of serotonin transporter suggested by three LeuT-SSRI struc-
tures. Nat Struct Mol Biol 16:652–657.
Address correspondence to: Jacob Andersen, Department of Drug Design
and Pharmacology, University of Copenhagen, Universitetsparken 2, DK-2100
Copenhagen, Denmark. E-mail: jaa@sund.ku.dk