Molecular Basis for Selective Serotonin Reuptake ...

1521-0111/85/5/703–714$25.00 MOLECULAR PHARMACOLOGY Copyright ª 2014 by The American Society for Pharmacology and Experimental Therapeutics

http://dx.doi.org/10.1124/mol.113.091249 Mol Pharmacol 85:703–714, May 2014

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 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 compared with the

Introduction The early recognition of the serotonin (5-hydroxytryptamine) transporter (SERT) and the norepinephrine transporter (NET) as important targets for antidepressant drugs fostered extensive drug discovery efforts dedicated to the design and synthesis of compounds selectively targeting SERT and/or NET (Kristensen et al., 2011). In 1986, fluoxetine (Prozac; Eli Lilly, Indianapolis, 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). This work was supported by the Lundbeck Foundation, the Danish National Research Foundation (DNRF59), and the Danish Council for Independent Research, Natural Sciences. 1 Current affiliation: Department of Chemistry, University of Copenhagen, Copenhagen, Denmark. 2 Current 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.

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.

Although compounds targeting SERT and NET have had important clinical significance for several decades, the molecular details underlying binding of antidepressants to these transporters are not clearly understood. Structural information of SERT and NET is still lacking, but X-ray crystal structures of the bacterial homolog LeuT (Yamashita et al., 2005; Krishnamurthy and Gouaux, 2012) have proved to be excellent structural templates for its mammalian counterparts 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 extracellular 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

ABBREVIATIONS: b-CIT, (2)-2b-carbomethoxy-3b-(4-iodophenyl)tropane; DAT, dopamine transporter; IFD, induced-fit docking; NET, norepinephrine 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 fluoxetine (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 computational 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

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 shown to be important for binding of SSRIs (Barker et al., 1999; 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 improved 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.,

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 pCIIRES-neo–containing human SERT and NET, respectively, were used (Kristensen et al., 2004; Andersen et al., 2011). Generations of pointmutants in SERT and NET were performed by site-directed mutagenesis 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 15fold 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-(4iodophenyl)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. Computational Methods. The R- and S-enantiomers of fluoxetine 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 26 26 Å3 as the outer boundary for ligand and an inner box of 10 10 10 Å3,

The Fluoxetine Binding Site in Human SERT

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 comparable 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 main determinant for the distinct selectivity profiles observed for fluoxetine and nisoxetine. In agreement with previous SAR studies of fluoxetine (Wong et al., 1975; Horng and Wong, 1976), we found that addition of an N-methyl group to fluoxetine (as in compound 7) reduced affinity toward SERT (7 versus 37 nM) but increased the selectivity for SERT over NET (127- versus 286-fold). Extending the aminoalkyl chain of fluoxetine with one methylene group (as in compound 10) had a minor effect on the affinity for SERT (7 versus 12 nM) but increased the selectivity over NET (127- versus 279-fold). The N,N-dimethyl analog of homofluoxetine (compound 11) had low activity toward both SERT and NET. The length and substitution pattern of the aminoalkyl chain on fluoxetine are

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 20 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.

Results Structure-Activity Relationship Study. Fluoxetine and nisoxetine share a substituted 3-aryloxy-3-phenylpropan-1-amine skeleton, and are distinguished by their substitution on the phenoxy ring only. Where fluoxetine has a CF3 group in the para position, nisoxetine has a methoxy group in the ortho position (Fig. 1). To delineate the role of these two diverging structural elements for activity at SERT and NET, we designed and prepared derivatives of fluoxetine and nisoxetine with different substituents on the phenoxy ring (Table 1) (see Supplemental Methods). Furthermore, the length and methyl substitution patterns of the aminoalkyl chain of SSRIs and TCAs have previously been shown to be important determinants for activity and selectivity toward

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).

Compound

R1

R2

R3

n

SERT WT

NET WT

SERT/NET Selectivity

nM

Fluoxetine (1) S-Fluoxetine (S-1) R-Fluoxetine (R-1) Nisoxetine (2) N-Methylfluoxetine (7) Des-CF3-fluoxetine (8) 2-OCH3-fluoxetine (9) Homofluoxetine (10) N-Methylhomofluoxetine (11)

705

H

CF3

H

1

OCH3 H H OCH3 H H

H CF3 H CF3 CF3 CF3

H CH3 H H H CH3

1 1 1 1 2 2

7 3 5 167 37 157 3 12 992

6 6 6 6 6 6 6 6 6

2 1 1 31 4 29 1 1 34

887 1324 572 4 10,563 12 125 3352 7958

6 6 6 6 6 6 6 6 6

115 246 67 1 1131 2 22 593 1196

127 389 108 0.02 286 0.08 45 279 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 selectivity (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 selectivity 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 Drosophila 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 mapped into global clusters describing common binding modes by comparison of all clusters obtained from both small and 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, which protrudes out toward the S2 site (Fig. 2A). In agreement with the similar binding affinities of R- and S-fluoxetine for SERT (Table 1), the XP GScores are also very similar for the two enantiomers. Additionally, a significant overlap between the two enantiomers in SERT-Cluster 1 is seen, where the amine of fluoxetine is anchored between Tyr95 and Asp98 on transmembrane helix (TM) 1, and the unsubstituted phenyl ring is located close to Ile168, Ile172, and Phe341 on TM3. Two minor global clusters of fluoxetine binding were also obtained

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 shown in orange) into homology models of SERT (A) and NET (B). The two dominating binding clusters (SERT-Cluster 1 and NET-Cluster 1) 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.

(SERT-Cluster 2 and SERT-Cluster 3; 9 and 2%, respectively, of the poses obtained) (Table 2). In these two clusters, fluoxetine is exclusively located in the S2 site and the amine of the ligand is located close to Glu493 from TM10 in both clusters (Fig. 2A). From IFD calculations in NET, we also found three global clusters of fluoxetine binding. In the dominating binding mode (NET-Cluster 1; 63% of the poses obtained) (Table 2), fluoxetine is located in the S1 site below an aromatic lid


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

hSERT

Small

R-Fluoxetine

S1

S1-C1 Outliers

20/20 0/20

211.7 (0.5)

274.0 (9.5)

SERT-Cluster 1

S-Fluoxetine

S1

S1-C1 Outliers

19/19 0/19

211.5 (0.7)

271.0 (8.1)

SERT-Cluster 1

R-Fluoxetine

S2

S2-C1 Outliers

4/4 0/4

211.0 (0.4)

265.2 (3.2)

SERT-Cluster 2

S-Fluoxetine

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

R-Fluoxetine

S1

S1-C1 Outliers

16/17 1/17

210.6 (1.3)

269.8 (5.8)

NET-Cluster 1

S-Fluoxetine

S1 S1


12/19 6/19 1/19

211.4 (0.6) 29.9 (0.3)

269.1 (5.0) 255.3 (5.6)

NET-Cluster 1 NET-Cluster 2

R-Fluoxetine

S1 S2


2/5 2/5 1/5

210.6 (0.5) 210.9 (0.2)

253.0 (5.8) 266.8 (1.8)


S-Fluoxetine

S1 S1


3/8 4/8 1/8

211.7 (0.0) 210.0 (0.2)

274.8 (0.7) 257.0 (2.4)


kcal/mol

Large

hNET

Small

Large

(Tyr152/Phe317) (Fig. 2B). The amine of fluoxetine is coordinated by Asp75, the CF3-substituted phenyl ring is located just above Phe72, and the unsubstituted phenyl ring is located close to Ile144, Val148, and Phe323. There is a significant overlap between the two enantiomers of fluoxetine in NET-Cluster 1, which is in agreement with the comparable binding affinities of the enantiomers toward NET (Table 1). Two minor global clusters were also obtained from IFDs in NET (NET-Cluster 2 and NET-Cluster 3; 24 and 4%, respectively, of the poses obtained) (Table 2). In NET-Cluster 2, fluoxetine is binding exclusively in the S1 site with the amine and unsubstituted phenyl ring located at similar positions as found in NET-Cluster 1, and in NET-Cluster 3 fluoxetine is binding in the S2 site in a similar pose as found in SERT-Cluster 3 (Fig. 2). NET-Cluster 1 and NET-Cluster 2 are not similar to any of the binding modes observed for fluoxetine within the S1 site of SERT. Experimental Validation of Suggested Binding Modes of Fluoxetine in SERT. To distinguish between the three obtained clusters of fluoxetine binding in SERT, we performed a mutational analysis of residues within 6 Å of the predicted binding modes to determine their role in fluoxetine binding. In total, 59 point mutations across 27 different positions in the S1 and S2 sites of SERT were included in the study. Nine mutants rendered the transporter nonfunctional and were not studied further (Supplemental Table 1). The inhibitory potency (Ki) of fluoxetine was determined at each of the 50 functional pointmutants across 24 different positions (Fig. 3; Supplemental Table 1). At five positions (Tyr95, Asp98, Ile168, Ile172, Asn177), point mutations induced .10-fold shifts in the Ki value for fluoxetine (ranging from 11- to 79-fold), suggesting these residues are key determinants for fluoxetine binding in SERT.

As all five residues are located within the S1 site of SERT, these results suggest that SERT-Cluster 1 represents the bioactive binding conformation of fluoxetine in SERT. This is in accordance with IFD calculations that also indicated fluoxetine to have the tightest binding in this cluster (Table 2), most significantly revealed in the Emodel scores. In contrast, mutations of residues within the S2 site generally induced ,3-fold shifts in fluoxetine Ki (Fig. 3; Supplemental Table 1), thus speaking against the binding modes predicted in SERT-Cluster 2 and SERT-Cluster 3. Notably, an ionic interaction between the amine of fluoxetine and the negatively charged side-chain of the S2 residue Glu493 on TM10 is predicted in the two minor binding clusters (Fig. 2A). However, removing the negatively charged side-chain by the E493A mutation had no significant effect on the potency of fluoxetine (Fig. 3; Supplemental Table 1). Furthermore, mutations of six hydrophobic residues within the S2 site (Trp103, Ile179, Trp182, Tyr232, Val236, and Val489) that seem to be important for the overall shape of the extracellular vestibule of SERT, generally only led to small shifts (,3-fold) in fluoxetine Ki (Fig. 3; Supplemental Table 1), substantiating that SERTCluster 2 and SERT-Cluster 3 do not represent the bioactive binding conformation of fluoxetine in SERT. Interestingly, TM10 residues have previously been suggested to have an important role for inhibitor binding within the S1 site of Drosophila and human DAT (Bisgaard et al., 2011; Penmatsa et al., 2013). Here we show that mutation of residues in TM10 of SERT (Ala486, Val489, Lys490, and Glu493) induce ,3-fold changes in the potency of fluoxetine (Fig. 4; Supplemental Table 1), suggesting that TM10 residues in DAT have a more important role in inhibitor binding than TM10 residues in SERT. The amino group of fluoxetine was found to have a key role for high-affinity binding 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 TM1, similar to what has previously been observed for escitalopram (Koldsø et al., 2010). Accordingly, the D98E mutation induced a 12-fold loss of potency for fluoxetine, and removal of the aromatic ring of Tyr95 (Y95A and Y95V) induced a 40-fold loss of potency for fluoxetine (Fig. 3; Supplemental Table 1). Interestingly, when substituting Tyr95 for Trp (Y95W) we found a significant 8-fold gain of potency (Fig. 3; Supplemental Table 1). Since Trp is a better cation–p interaction partner than Tyr (Gallivan and Dougherty, 1999), the gain of potency induced by Y95W suggests a cation–p interaction between the amine of fluoxetine and the aromatic side-chain of Tyr95. The I168F mutation induced an 11-fold gain of potency for fluoxetine (Fig. 3; Supplemental Table 1), which is likely induced by aromatic interactions between the inserted Phe and the unsubstituted phenyl ring of fluoxetine in the lower part of the S1 site (Figs. 2 and 3). Previously, the I172M mutation was shown to decrease fluoxetine potency (Henry et al., 2006; Walline et al., 2008; Thompson et al., 2011; Sørensen et al., 2012). Here, we show that mutation of Ile172 to Ala, Gln, and Met induced 6- to 79-fold loss of potency for fluoxetine, and we corroborate earlier findings that Ile172 plays a key role in recognition of fluoxetine (Fig. 3; Supplemental Table 1). Mutation of Asn177 on TM3 to Ala or Ser induced 7- and 25-fold

loss of potency for fluoxetine, respectively. In contrast, the N177E mutation did not significantly affect fluoxetine Ki (Fig. 3; Supplemental Table 1), indicating that a side-chain bearing a carbonyl group in this position is important for recognition of fluoxetine. In SERT-Cluster 1, Asn177 is located .6.5 Å away from fluoxetine but interacts with Thr439 through an H-bond (Fig. 3). Hence, the N177A and N177S mutations might affect fluoxetine Ki in an indirect manner by modulating the overall shape of the S1 pocket by disruption of the H-bond between TM3 and TM8. The CF3 substituent of fluoxetine was found to be a key determinant for obtaining high-affinity binding in SERT (Table 1). In SERT-Cluster 1, the CF3-substituted phenyl ring is located in a hydrophobic pocket with a direct interaction between the CF3 group and the backbone of Gly100 and aromatic p–p stacking interactions with Tyr176 (Fig. 3). Backbone interactions are notoriously difficult to address by conventional mutagenesis, and we have previously shown that SERT is very sensitive to mutation of Tyr176 (Andersen et al., 2010). Only the conservative Y176F mutation has so far been found to be functionally tolerated, and it had no significant effect on fluoxetine Ki (Fig. 3; Supplemental Table 1). Therefore, as an alternative approach to probe specific interactions between the CF3-substituted phenyl ring of fluoxetine and SERT, we tested nisoxetine,


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 (compound 9) at selected S1 mutations that induce significant changes in fluoxetine Ki (Supplemental Table 2). Nisoxetine, compound 8, and compound 9 have different aromatic substituents than fluoxetine, and we thus envisioned that if the substituted phenyl ring of fluoxetine interacts with one of the mutated residues, the analogs would be differentially affected by the mutation compared with fluoxetine. However, the potency of fluoxetine and the three analogs were generally affected to the same level across the tested mutations (Supplemental Table 2), thereby indirectly substantiating that the CF3-substituted phenyl ring is located in the hydrophobic pocket, potentially engaging in backbone and aromatic stacking interactions that are difficult to address by site-directed mutagenesis. In

summary, our mutational analysis establishes the S1 site as the primary binding site for fluoxetine in SERT and identifies SERT-Cluster 1 as the most likely binding mode model. Molecular Determinants for SERT/NET Selectivity. We next sought to identify specific residues within SERT and NET that determine the distinct selectivity of fluoxetine and nisoxetine. We have shown that nonconserved SERT/NET residues within the S1 site dictate the selectivity for citalopram (Andersen et al., 2011). Accordingly, we hypothesized that the molecular determinants for fluoxetine and nisoxetine selectivity are also found among these residues. We have previously mutated each of the 15 nonconserved SERT/NET residues found within 6 Å of the S1 site to the aligning residue in the other transporter, resulting in 15 individual

710

Andersen et al.

SERT-to-NET mutations in SERT and 15 individual NET-toSERT mutations in NET (Fig. 4; Supplemental Fig. 2) (Andersen et al., 2011). To systematically probe for the individual contribution of these nonconserved S1 residues for the selectivity of fluoxetine and nisoxetine, we determined the inhibitory potency of the two inhibitors at the 15 SERT-toNET and the 15 NET-to-SERT mutations (Fig. 4; Supplemental Tables 3 and 4). For six of the SERT-to-NET mutations in SERT, the potency of fluoxetine was significantly decreased (3- to 7-fold). None of the SERT mutants induced an increase in the potency of nisoxetine (Fig. 4; Supplemental Table 3). In contrast, 12 of the 15 NET-to-SERT mutants in NET increased fluoxetine potency (3- to 11-fold), whereas only three mutants decreased the potency of nisoxetine (2- to 3-fold) (Fig. 4; Supplemental Table 4). Hence, as observed for citalopram (Andersen et al., 2011), inhibitor selectivity can be modulated, but is not controlled by a single nonconserved residue within the S1 site. Next, we combined the single point-mutants into a set of 30 multiple SERT-toNET mutants in SERT (designated S1–S30) and 17 multiple NET-to-SERT mutants in NET (designated N1–N17) (Fig. 4; Supplemental Tables 3 and 4). The design of these mutants was initially directed by results from single point mutants and later combined with results from multiple mutants and with the binding poses obtained from IFD calculations. Functional uptake activity was retained for 29 of the 30 multiple SERT mutants and for 10 of the 17 multiple NET mutants (Supplemental Tables 3 and 4). The nonfunctional mutants were not studied further. For fluoxetine, a significant loss of potency (3- to 11-fold) was observed for eight of the multiple SERT mutants (Fig. 4; Supplemental Table 3). Mutations within the Ile172/Ala173/Ser174 motif on TM3 was included in the five multiple mutants that displayed the largest loss of potency, suggesting that this motif is an important determinant for the selectivity of fluoxetine. Interestingly, combining mutations in the Ile172/Ala172/ Ser174 motif with A441G and L443M (as in S27, S28, and S29) induced a significant gain of potency for fluoxetine (Fig. 4; Supplemental Table 3), indicating that A441G and L443M hold a positive role for binding of the SSRI. For nisoxetine, seven of the multiple SERT mutants induced a significant gain of potency (10- to 24-fold). Combining mutations within the Ile172/Ala173/Ser174 motif with A441G and L443M on TM8 induced the largest gain of nisoxetine potency, indicating that residues on TM3 and TM8 are cooperative determinants for binding of nisoxetine in SERT. The 3-fold S29 mutant (SERT-S174F-A441G-L443M) had the largest effect, and rendered SERT 24-fold more sensitive to nisoxetine than SERT WT (60 versus 1422 nM), showing that key determinants for nisoxetine selectivity are located within the S1 site of SERT. Determination of the potency of nisoxetine at the 10 functional multiple NET mutants surprisingly showed that all retained WT potency of the NET selective ligand (Fig. 4; Supplemental Table 4). These data strongly suggest that, in contrast to SERT, nonconserved S1 residues do not define the inhibitory potency of nisoxetine in NET. In contrast, all multiple NET-to-SERT mutants induced a significant gain of fluoxetine potency (2- to 14-fold) (Fig. 4; Supplemental Table 4). Mutations in TM1 (F72Y and A77G), TM8 (M424L and A426G), and TM9 (Thr453) were found to be most important for improving inhibitory potency of fluoxetine in

NET. Specifically, the 4-fold N9 mutant (NET-A77G-M424LA426G-T453C) induced the largest effect and rendered NET 14-fold more sensitive toward fluoxetine than NET WT (2993 versus 217 nM). Hence, nonconserved residues within the S1 site of SERT and NET are key determinants for the selectivity of fluoxetine, which are supportive of our proposed binding mode of fluoxetine (Fig. 2). Interchanging Binding Sites between SERT and NET. Interchanging nonconserved residues within the S1 site of SERT and NET modulated the potency of fluoxetine and nisoxetine. However, the selectivity was not fully reversed by any of the tested mutants. We therefore generated a series of mutant constructs in which all nonconserved residues in the S1 site and/or all nonconserved residues in the S2 site were simultaneously interchanged between SERT and NET, and thereby in principle transplanting these binding sites from SERT into NET and vice versa. Hereby, three SERT constructs containing NET S1: SERT-(NET S1), NET S2: SERT-(NET S2), and NET S1/S2: SERT-(NET S1S2), in addition to two NET constructs containing SERT S1: NET(SERT S1) and SERT S2: NET-(SERT S2) were created (Fig. 5; Supplemental Fig. 2; Supplemental Table 5). Additionally, we created two constructs in which all nonconserved residues within 6 Å of the predicted binding mode of fluoxetine (SERTCluster 1) were interchanged—SERT-(NET S1S2i) and NET(SERT S1S2i), respectively. For a detailed description of mutant constructs, see Supplemental Fig. 2 and Supplemental Table 5. Initially, we performed saturation binding analysis on membrane preparations from COS-7 cells expressing WT and mutant transporters and found that all SERT constructs bind [125I]b-CIT, whereas only NET WT and NET-(SERT S2) showed specific [125I]b-CIT binding (Fig. 5; Supplemental Table 5). Consistent with saturation binding analyses, confocal imaging of green fluorescent protein–tagged variants of WT and mutant transporters showed that the two NET constructs that do not bind [125I]b-CIT—NET-(SERT S1) and NET-(SERT S1S2i)—were primarily retained within intracellular compartments (Supplemental Fig. 3). Next, we determined the binding affinities of fluoxetine and nisoxetine at the S1/S2 constructs that bind [125I]b-CIT in a competition binding assay. For fluoxetine, insertion of NET S1 into SERT induced a 15-fold decrease in binding affinity (7 versus 102 nM), whereas insertion of the NET S2 site did not affect binding of fluoxetine (Fig. 5; Supplemental Table 5). Interestingly, the decreased binding affinity of fluoxetine in SERT-(NET S1) was reversed by simultaneous insertion of S2 site (Fig. 5; Supplemental Table 5). The binding affinity of nisoxetine was increased by 6-fold after inserting the S1 site from NET into SERT (167 versus 29 nM), whereas insertion of the S2 site had no significant effect on nisoxetine. Insertion of both the S1 and S2 sites from NET into SERT improved the binding affinity of nisoxetine to a similar level as observed when the S1 site was inserted alone (15 versus 29 nM), showing that residues located in the S1 site of SERT are key determinants for the selectivity of nisoxetine. Similar detailed analysis was not possible for NET, since only the NET-(SERT S2) construct could bind [125I]b-CIT. Interestingly, the binding affinity of nisoxetine was decreased almost to the same level as observed in SERT WT by inserting SERT S2 into NET (167 versus 104 nM compared with 4 nM at NET WT). Together with our initial analysis, which showed


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 selectivity of nisoxetine, this result suggested that nonconserved residues in the S2 site of NET are key determinants for the selectivity of nisoxetine. In summary, our mutational analysis of nonconserved SERT/NET residues supports the view that the selectivity of fluoxetine for SERT over NET is largely determined by nonconserved residues within the S1 site of both SERT and NET. This interpretation is fully in accordance with our proposed IFD models of fluoxetine binding within the S1 site in the two transporters. In contrast, we found that the selectivity of nisoxetine for NET over SERT is controlled by nonconserved residues in different regions of SERT (S1 residues) and NET (S2 residues). Differentiation between direct and indirect effects in mutagenesis studies is inherently difficult. Mutations can induce long-range allosteric effects that perturb a distinct binding site or induce a shift in the conformational equilibrium of the transporter that changes the temporal accessibility to the binding site. Thus, the mutational analyses do not allow us to conclusively specify the location of the nisoxetine binding site in SERT or NET, but they suggest that there is a complex mechanism for selective recognition of inhibitors in SERT and NET.

Discussion The models of fluoxetine binding in human SERT and NET produced in the present study are based on X-ray crystal structures of LeuT and have been constructed using wellestablished procedures that have been implemented for modeling of other important monoamine transporter inhibitors (Andersen et al., 2010; Koldsø et al., 2010; Sinning et al.,

2010; Severinsen et al., 2013). Very recently, Drosophila DAT and LeuBAT were crystallized in complex with antidepressants (Penmatsa et al., 2013; Wang et al., 2013). These X-ray crystal structures offer a new platform for understanding ligand interactions that have obvious potential to further push the field toward more reliable and realistic models of antidepressant binding in human transporters. However, both Drosophila DAT and LeuBAT are inactive in transport, and their pharmacological profiles seem to be a hybrid of the human SERT, NET, and DAT. Thus, to assess the potential of LeuBAT and Drosophila DAT for studying the molecular pharmacology of human transporters, it is critically important to establish similarities and discrepancies between Drosophila DAT and LeuBAT and their human relatives. For this purpose, comparison of our present fluoxetine model with the X-ray crystal structure of LeuBAT in complex with fluoxetine therefore provides an excellent first opportunity for assessment of the potential of LeuBAT as a model system for SSRI binding in human transporters. First and foremost, our proposed binding model of fluoxetine in SERT (SERT-Cluster 1) is in agreement with the LeuBAT structure (Wang et al., 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 experimental validation, and can explain two key findings from the mutational analysis. Firstly, fluoxetine potency is largely affected by mutations in the S1 site (up to 79-fold lossof-potency), whereas mutations in the S2 site generally induce ,3-fold changes in fluoxetine Ki (Fig. 3; Supplemental Table 1). Secondly, our mutational analysis suggests 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 overall 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 site in our model. Notably, neither from IFDs in SERT nor in NET did we identify any poses of fluoxetine in an orientation 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. Specifically, 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 and does not allow a similar binding mode of the CF3 group of fluoxetine in SERT (Fig. 6). Accordingly, we found that mutations of Asn177 and Thr439 induce a marked loss of potency for fluoxetine (Figs. 3 and 4), likely due to disruption of the H-bond and thereby the overall shape of the binding site rather than affecting direct interactions with the inhibitor. Also, if fluoxetine adopts the same binding mode in SERT as found in LeuBAT, Asn177 would be pointing directly toward the CF3 group of the SSRI (Fig. 6). Thus, introduction of a negative charge into this subsite would likely cause an electrostatic repulsion with the electronegative CF3 group, and thereby induce a loss of potency for fluoxetine. Notably, the N177E mutation had no effect on fluoxetine potency (Fig. 3), and since the H-bond to Thr439 can be preserved in the N177E mutant, this provides further support for our proposed binding model. Additionally, mutation of Ile179 in SERT has previously been shown to induce a marked loss of potency for fluoxetine (Zhou et al., 2009). This is in good agreement with our binding model, in which Ile179 is located within 3.5 Å from fluoxetine and points directly toward the subsite where the CF3 group of fluoxetine binds (Fig. 6). In contrast, if fluoxetine adopts a similar binding mode in SERT as found in LeuBAT, Ile179 would be located .6.5 Å away from the

Fig. 6. (A and B) Superimposition of a representative binding pose of fluoxetine (C-atoms shown in orange) in SERT from the dominating 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). The overlays are shown on the SERT model. Selected residues in proximity of the proposed binding clusters are shown as stick representations. Positions where mutations induce .10-fold changes in fluoxetine potency are highlighted in dark blue. (C) Superimposition of a representative 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.

inhibitor (Fig. 6), making it less likely that mutation at this site will have a pronounced effect on fluoxetine potency. Overall, even though fluoxetine shares the same binding site in human SERT and LeuBAT, our experimentally supported binding model of fluoxetine in SERT suggest that the SSRI has distinct binding modes in human and bacterial transporters, emphasizing the continued need for careful experimental validation when extrapolating findings from LeuBAT to human transporters. Our study provides novel insight into the molecular determinants for selective nisoxetine binding in NET by showing that nonconserved residues within the S2 site are important for high-affinity nisoxetine binding in NET (Figs. 4 and 5). Mutation of residues outside the S1 site in NET has previously been found to affect binding of nisoxetine (Paczkowski et al., 2007; Wenge and Bönisch, 2013). In addition, cocaine-like compounds, which are believed to bind in the S1 site of monoamine transporters (Beuming et al., 2008), display noncompetitive binding with nisoxetine in NET (Zhen et al., 2012). Together, these observations indicate that nisoxetine binds outside the S1 site, and are thus supportive of a recent model suggesting that nisoxetine binds to the S2 site in NET (Wang et al., 2012). However, nisoxetine is also affected by


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 antidepressants 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 inhibitory mechanisms of antidepressants in a structural context. This is essential for establishing a useful framework for structure-based drug development of future monoamine transporter drugs with fine-tuned transporter selectivity. 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. References Andersen J, Olsen L, Hansen KB, Taboureau O, Jørgensen FS, Jørgensen AM, BangAndersen B, Egebjerg J, Strømgaard K, and Kristensen AS (2010) Mutational mapping and modeling of the binding site for (S)-citalopram in the human serotonin transporter. J Biol Chem 285:2051–2063. 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 human 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

713

serotonin transporter: importance of Ser-438 in recognition of citalopram and tricyclic 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, GarciaCebrian A, Grassi L, Perahia DGS, and Reed C et al. (2008) Prescribing patterns of antidepressants in europe: Results from the Factors Influencing Depression Endpoints 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. 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 transporter. 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 studies. 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 conformational 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 serotonin 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: structure, 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 receptor and transporter binding profile of antidepressants and their metabolites. J Pharmacol Exp Ther 283:1305–1322. 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 transporter 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. 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.

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 bacterial 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 transporter. 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 serotonin 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 highaffinity 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 recognition. 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 transporters. Nature 503:141–145.

Wenge B and Bönisch H (2013) The role of cysteines and histidins of the norepinephrine 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-(ptrifluoromethylphenoxy). 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 fluoxetine 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. Nature 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) Antidepressant specificity of serotonin transporter suggested by three LeuT-SSRI structures. 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: [email protected]

Molecular Pharmacology

Supplemental Material

The molecular basis for selective serotonin re-uptake inhibition by the antidepressant agent fluoxetine (Prozac) Jacob Andersen, Nicolai Stuhr-Hansen1, Linda Grønborg Zachariassen, Heidi Koldsø2, Birgit Schiøtt, Kristian Strømgaard & Anders S. Kristensen

Department of Drug Design and Pharmacology, University of Copenhagen, Universitetsparken 2, DK-2100 Copenhagen, Denmark (JA, NSH, LGZ, KS, ASK); and Center for Insoluble Structures (inSPIN) and Interdisciplinary Nanoscience

Center (iNANO), Department

of

Chemistry, Aarhus University,

Langelandsgade 140, DK-8000 Aarhus C, Denmark (HK, BS). 1

Current affiliation: Department of Chemistry, University of Copenhagen, Copenhagen, Denmark.

2

Current affiliation: Department of Biochemistry, University of Oxford, Oxford, United Kingdom.

Correspondence should be addressed to Jacob Andersen ([email protected]).

Supplemental Figures ................................................................................................................................. S2-S4 Supplemental Tables .................................................................................................................................. S5-S9 Supplemental Methods ........................................................................................................................... S10-S17 Supplemental References .............................................................................................................................. S18

S1

Drosophila DAT/nortriptyline (crystal structure) SERT/imipramine (LeuT-based homology model) TM3

TM1 Y124 Y176

D46 D98

V120 I172 Nortriptyline Imipramine A117 A169

F43 Y95

Supplemental Figure 1. Overlay of the crystal structure of Drosophila DAT in complex with the TCA nortriptyline (shown in wheat) (PDB ID 4M48) with a LeuT-based model of SERT in complex with the TCA imipramine (shown in teal) (Sinning et al., 2010). Selected residues in proximity of the inhibitor binding sites are shown as stick representations. The overlay illustrates that LeuT-based models are very predictive of inhibitor binding in eukaryotic transporters, and thus verifies the use of LeuT as structural template for human transporters.

S2

hSERT hNET

1 METTPLNSQKQLSACEDGEDCQENGVLQKVVPTPGDKVESGQISNGYSAVPSPGAGDDTRHSIPATTTTLVAELH 1 -----------------------MLLARMNPQVQPENNGADTGPEQPLRARKTAELLVVKERNGVQCLLAPRDGD

hSERT hNET

ddddddddddddd --------TM1------------------TM2---------76 QGERETWGKKVDFLLSVIGYAVDLGNVWRFPYICYQNGGGAFLLPYTIMAIFGGIPLFYMELALGQYHRNGCISI 53 AQPRETWGKKIDFLLSVVGFAVDLANVWRFPYLCYKNGGGAFLIPYTLFLIIAGMPLFYMELALGQYNREGAATV asdfsdfsdfsd dfs●●●●●●●●

ddddd -----------------TM3---------------hSERT 151 WRKICPIFKGIGYAICIIAFYIASYYNTIMAWALYYLISSFTDQLPWTSCKNSWNTGNCTNYFSED------NIT hNET 128 W-KICPFFKGVGYAVILIALYVGFYYNVIIAWSLYYLFSSFTLNLPWTDCGHTWNSPNCTDPKLLNGSVLGNHTK dsfddddddddkkk●●k●●●●●●●●●● ddddddddddddddddddddddddddddddddd --------TM4-----------------TM5----hSERT 220 WTLHSTSPAEEFYTRHVLQIHRSKGLQDLGGISWQLALCIMLIFTVIYFSIWKGVKTSGKVVWVTATFPYIILSV hNET 202 YSKYKFTPAAEFYERGVLHLHESSGIHDIGLPQWQLLLCLMVVVIVLYFSLWKGVKTSGKVVWITATLPYFVLFV ------ dddddddddddddddd -------------TM6------------------hSERT 295 LLVRGATLPGAWRGVLFYLKPNWQKLLETGVWIDAAAQIFFSLGPGFGVLLAFASYNKFNNNCYQDALVTSVVNC hNET 277 LLVHGVTLPGASNGINAYLHIDFYRLKEATVWIDAATQIFFSLGAGFGVLIAFASYNKFDNNCYRDALLTSSINC ddddddddddddddddddddddddddddddddddddd*●●●●●●dd●●●●d●●dddddddddddddddddddd● --TM7------- ddddddddddddddddddddddddddddddddddddddd ------------TM8------hSERT 370 MTSFVSGFVIFTVLGYMAEMRNEDVSEVAKDAGPSLLFITYAEAIANMPASTFFAIIFFLMLITLGLDSTFAGLE hNET 352 ITSFVSGFAIFSILGYMAHEHKVNIEDVA-TEGAGLVFILYPEAISTLSGSTFWAVVFFVMLLALGLDSSMGGME dddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd●●d●●●d●●●● -------------TM9-----------------TM10----------hSERT 445 GVITAVLDEFPHVWAKRRERFVLAVVITCFFGSLVTLTFGGAYVVKLLEEYATGPAVLTVALIEAVAVSWFYGIT hNET 426 AVITGLADDFQVLKR-HRKLFTFGVTFSTFLLALFCITKGGIYVLTLLDTFAAGTSILFAVLMEAIGVSWFYGVD ●ddddddddddddddddddddddddddd●dddddddddddddddddddddd●● ● Dddddddddddddddd ----------TM11-----------------TM12-----hSERT 520 QFCRDVKEMLGFSPGWFWRICWVAISPLFLLFIICSFLMSPPQLRLFQYNYPYWSIILGYCIGTSSFICIPTYIA hNET 500 RFSNDIQQMMGFRPGLYWRLCWKFVSPAFLLFVVVVSIINFKPLTYDDYIFPPWANWVGWGIALSSMVLVPIYVI hSERT 595 YRLIITPGTFKERIIKSITPETPTEIPCGDIRLNAV------hNET 575 YKFLSTQGSLWERLAYGITPENEHHLVAQRDIRQFQLQHWLAI

Residues within 6Å of the putative S1 site Residues within 6Å of the putative S2 site Residues within 6Å of proposed fluoxetine binding site

Supplemental Figure 2. Amino acid sequence alignment of hSERT and hNET based on a previously published alignment of LeuT and SLC6 transporters.(Beuming et al., 2006) Conserved residues are shown as white text on grey background, similar residues are shown as white text on black background and nonconserved residues are shown as black text on white background. Overall, 44% of the residues are conserved and 22% of the residues are similar in hSERT and hNET. The putative TM regions are depicted on top of the alignment. The filled black circles indicate residues that are located within 6Å of the putative S1 binding site [32 out of 47 residues (68%) are conserved], blue stars indicate residues that are located within 6Å of the putative S2 binding site [11 out of 24 residues (46%) are conserved] and green triangles indicate residues that are located within 6Å of the proposed binding site for fluoxetine in hSERT [28 out of 42 residues (67%) are conserved]. Non-conserved residues within these regions are highlighted in red boxes. The S1 site was defined as the proposed escitalopram binding site in hSERT (Andersen et al., 2010) and the S2 site was defined as the equivalent position in hSERT as the binding site for imipramine found in LeuT (PDB ID: 2Q72). The proposed fluoxetine binding site was defined from a representative pose of fluoxetine binding in SERT from the dominating SERT-Cluster 1 (see Fig. 2 and Fig. 3). S3

SERT(NET S1)

SERTSERTSERT(NET S2) (NET S1S2) (NET S1S2i)

NET WT

NETNETNET(SERT S1) (SERT S2) (SERT S1S2i)

Overlay

CellMask Deep Red

GFP

SERT WT

Supplemental Figure 3. Expression and cellular distribution of GFP-tagged versions of WT and mutant forms of SERT and NET where non-conserved SERT/NET residues within 6Å of putative binding regions have been interchanged (see also Supplemental Figure 1 and Supplemental Table 5). WT and mutant forms of GFP-tagged SERT and NET were transiently transfected into COS7 cells and the plasma membrane was visualized with CellMask Deep Red plasma membrane stain. Confocal laser scanning microscopy images show that all SERT mutants and NET-(SERT S2) displayed surface expression patterns comparable to WT transporters, whereas NET-(SERT S1) and NET-(SERT S1S2i) showed loss of cell-surface expression.

S4

Supplemental Table 1. Impact of mutations on [3H]5HT uptake kinetics and fluoxetine potency Site

Mutant

Km µM

WT S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1/S2 S1/S2 S1/S2 S1/S2 S1/S2 S1/S2 S1/S2 S1/S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 S2

Y95Aa Y95Va Y95W D98Ea I168A I168F A169Da I172Aa I172Ma I172Qa N177Aa N177Ea N177Sa F341Aa F341Ia F341La F341Ya V343Na V343Sa S438Ta R104A Y175Fa Y175V Y176Fa F335Ea F335Ha F335Ya F335W a W103A W103E I179C I179H W182A W182E Y232A Y232E Y232F V236A V236E V236F G338D G338F D400Fa D400Ka D400La A401E G402H P403A P403E P403F A486E A486I V489Fa K490Da K490Ta K490Fa E493A E493F E493K

0.98 ± 0.09 0.22 ± 0.04 2.35 ± 0.31 2.16 ± 0.25 0.22 ± 0.06 0.93 ± 0.11 1.15 ± 0.07 0.62 ± 0.06 1.08 ± 0.34 0.77 ± 0.05 3.12 ± 0.72 7.65 ± 0.42 1.55 ± 0.14 4.25 ± 0.12 0.77 ± 0.18 0.40 ± 0.11 0.54 ± 0.1 0.08 ± 0.02 0.24 ± 0.01 0.43 ± 0.304 0.14 ± 0.02 N.F. 0.24 ± 0.08 3.70 ± 0.61 0.88 ± 0.04 0.29 ± 0.09 0.68 ± 0.18 0.62 ± 0.2 0.06 ± 0.01 0.96 ± 0.09 N.F. 0.53 ± 0.11 N.F. 2.22 ± 0.26 5.66 ± 3.79 0.19 ± 0.02 0.48 ± 0.08 1.33 ± 0.18 1.29 ± 0.05 N.F. 0.42 ± 0.05 N.F. N.F. 0.18 ± 0.03 0.12 ± 0.03 0.60 ± 0.30 0.55 ± 0.13 N.F. 1.49 ± 0.08 0.30 ± 0.04 0.67 ± 0.14 0.51 ± 0.09 0.46 ± 0.06 0.63 ± 0.43 0.16 ± 0.02 0.27 ± 0.09 0.17 ± 0.06 1.50 ± 0.11 N.F. N.F.

Uptake activity n

% of WT

30 3 4 4 3 5 5 3 5 4 3 3 3 3 3 3 3 3 3 3 3

100 45 ± 4 7 ± 1 28 ± 6 65 ± 8 59 ± 4 41 ± 4 89 ± 2 63 ± 5 86 ± 3 21 ± 3 42 ± 2 76 ± 1 63 ± 1 51 ± 7 44 ± 2 52 ± 2 57 ± 8 20 ± 1 87 ± 2 20 ± 4

3 5 3 3 4 3 3 5

54 49 86 7 13 73 63 15

3

19 ± 1

5 5 5 5 5 5

86 112 39 48 69 67

± ± ± ± ± ± ± ±

± ± ± ± ± ±

Fluoxetine K i nM

5 6 5 1 2 7 11 2

110 5,129 4,629 13 1,300 25 10 29 663 8,786 2,002 811 124 2,708 125 16 29 524 24 39 364

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

8 243* 1,283* 2* 190* 5* 1* 7* 64* 708* 261* 209* 38 125* 24 2* 5* 96* 7* 6* 60*

35 3 4 8 6 4 4 3 4 7 4 4 4 3 4 6 6 4 6 3 4

-46.3 -41.8 8.4 -11.7 4.4 11.4 3.8 -6.0 -79.4 -18.1 -7.3 -1.1 -24.5 -1.1 7.0 3.8 -4.7 4.6 2.9 -3.3

34 201 94 45 100 148 342 17

± ± ± ± ± ± ± ±

3* 45* 15 5* 14 77 64* 5*

6 4 3 3 4 4 5 3

3.2 -1.8 1.2 2.5 1.1 -1.3 -3.1 6.3

3

3.0

3 3 3 4 3 3

-1.8 1.1 1.4 1.3 -1.4 1.1

37 ± 4*

6 12 4 5 7 7

Fold-change in K i-value n

194 97 82 82 160 97

± ± ± ± ± ±

40* 35 9 16 22 16

5

20 ± 2

31 ± 9*

4

3.6

3 3 3 3

37 21 23 21

± ± ± ±

4 3 3 2

67 64 94 55

± ± ± ±

34 17 48 4

4 4 4 3

1.6 1.7 1.2 2.0

5 5 5 5 5 3 4 4 3 5

85 17 41 59 60 110 61 109 95 55

± ± ± ± ± ± ± ± ± ±

9 2 5 7 8 9 4 4 5 8

84 21 53 86 173 189 77 153 232 99

± ± ± ± ± ± ± ± ± ±

29 8* 11* 13 29* 71* 36 44 69* 14

4 3 4 6 6 5 4 5 5 4

1.3 5.2 2.1 1.3 -1.6 -1.7 1.4 -1.4 -2.1 1.1

Uptake activity, substrate Km values and fluoxetine Ki values were determined in a [3H]5HT uptake assay. Results are presented as mean ± SEM from n indipendent experiments each performed in triplicate. The fold-changes in Ki values were calculated as Ki(WT)/Ki(mutant) or -Ki(mutant)/Ki(WT) for mutants increasing or decreasing, respectively, the potency of fluoxetine. Asterisks denote significantly different Ki value compared to WT (p < 0.01; Student's t-test). N.F. = non-functional. a Values for Km and uptake activity are taken from (Andersen et al., 2009, 2010, 2011) and are included for comparison.

S5

Supplemental Table 2. Inhibitory potency of fluoxetine and analogues at selected S1 mutants in SERT Fluoxetine (1) nM

SERT WT

110 ± 8

R -Fluoxetine (R -1)

S -Fluoxetine (S -1)

nM

n

nM

n

35

388 ± 38

3

332 ± 43

3

3

4,983 ± 692*

3

Y95A

5,129 ± 243*

3

7,105 ± 277*

Y95W

13 ± 2*

8

N.D.

D98E

1,300 ± 190*

6

2,722 ± 457*

3

687 ± 62*

I172M

8,786 ± 708*

7

10,799 ± 1,148*

3

Y175F

34 ± 3*

6

55 ± 10*

N177S

2,708 ± 125*

3

N- Methylfluoxetine (7) nM

SERT WT

2,587 ± 343

Y95A

16,963 ± 1,673*

Y95W

N.D.

D98E

2,766 ± 297

Nisoxetine (2)

n

n

3

356 ± 88

3

3

3,821 ± 558*

3

12,904 ± 1,495*

3

210,739 ± 36,687*

5

3

68 ± 6*

3

255 ± 19

3

3

2,806 ± 176*

3

17,023 ± 4,456*

3

Des-CF3 -fluoxetine (8) nM

n

61

15,144 ± 2,059*

N.D.

4,082 ± 759*

nM

1,422 ± 162

2-OCH3-fluoxetine (9)

n

nM

n

Homofluoxetine (10) nM

n

7

3,793 ± 501

7

44 ± 4

21

1,026 ± 127

11

4

6,134 ± 1,069*

4

3,307 ± 291*

5

4,615 ± 599*

3

54 ± 6

4

42 ± 8*

4

N.D. 4

7,257 ± 1,069*

4

179 ± 34*

8

4,401 ± 487*

3

I172M

26,210 ± 3,315*

7

20,273 ± 3,571*

7

5,695 ± 569*

10

25,113 ± 1,881*

3

Y175F

446 ± 61*

8

583 ± 81*

4

15 ± 4*

4

350 ± 101*

4

N177S

15,801 ± 1,638*

8

6,040 ± 1,501

4

567 ± 123*

5

13,787 ± 3,071*

3

3

Inhibitor Ki values were determined in a [ H]5HT uptake assay. Results are presented as mean ± SEM from n independent experiments each performed in triplicate. Asterisks denote significantly different inhibitor Ki value compared to WT (p < 0.05; Student's t-test). N.D. = not determined. Ki values for fluoxetine are taken from Supplemental Table 1, and are included for comparison.

S6

Supplemental Table 3. Impact of mutations of non-conserved residues in the S1 site of SERT on uptake kinetics and inhibitor potency Km

Uptake activity

Fluoxetine K i

µM

% of WT

nM

SERT WT

0.98 ± 0.09

100

110 ± 8

1,422 ± 162

Y95F

0.35 ± 0.09

85 ± 5

387 ± 9*

3,447 ± 697*

G100A

0.57 ± 0.08

60 ± 3

75 ± 21

296 ± 88

I165V

0.73 ± 0.08

82 ± 2

261 ± 55

1,396 ± 304

C166I

0.87 ± 0.16

97 ± 2

315 ± 98*

1,678 ± 469

F170L

0.65 ± 0.08

83 ± 1

323 ± 69*

981 ± 204

I172V

0.73 ± 0.20

36 ± 3

162 ± 29

657 ± 97

A173G

0.26 ± 0.07

53 ± 11

189 ± 11

752 ± 25

S174F

0.63 ± 0.05

70 ± 3

274 ± 90

T439S

0.75 ± 0.10

78 ± 6

822 ± 63*

A441G

0.37 ± 0.09

72 ± 2

L443M

0.49 ± 0.11

58 ± 5

231 ± 32

486 ± 79

G445A

0.73 ± 0.17

85 ± 3

295 ± 78*

1,662 ± 319

C473T

0.87 ± 0.24

74 ± 7

533 ± 63*

1,541 ± 171

T497A

1.03 ± 0.41

71 ± 5

138 ± 25

V501I

2.87 ± 1.06

90 ± 3

128 ± 35

325 ± 114

Y95A-G100A

0.67 ± 0.08

51 ± 3

183 ± 19

2,575 ± 359

S2

I172V-A173G-S174F

1.89 ± 0.50

78 ± 1

1,226 ± 220*

4,186 ± 237*

S3

G100A-I172V-A173G-S174F

1.49 ± 0.09

21 ± 3

708 ± 166*

1,272 ± 166

S4

G100A-L443M-G445A

1.50 ± 0.28

29 ± 5

324 ± 90*

1,472 ± 533

S5

L443M-G445A

2.62 ± 0.23

71 ± 4

240 ± 58

1,430 ± 302

S6

Y95A-G100A-L443M-G445A

1.12 ± 0.22

16 ± 4

111 ± 27

S7

Y95A-G100A-I172V-A173G-S174F

0.85 ± 0.12

36 ± 5

544 ± 147*

S8

I172V-A173G-S174F-L443M-G445A

2.76 ± 0.38

57 ± 4

1,160 ± 227*

S9

L443M-G445A-C473T

2.21 ± 0.34

90 ± 3

249 ± 82

766 ± 138

S10

Y95A-G100A-L443M-G445A-C473T

1.07 ± 0.05

54 ± 8

85 ± 16

1,332 ± 260

S11

Y95A-G100A-I172V-A173G-S174F-L443M-G445A-C473T

1.82 ± 0.41

31 ± 1

416 ± 17*

412 ± 212

S12

G100A-I172V

0.36 ± 0.11

19 ± 3

48 ± 11*

602 ± 229

S13

G100A-A173G

0.35 ± 0.12

13 ± 3

55 ± 7*

326 ± 78

S14

G100A-S174F

1.09 ± 0.33

31 ± 3

63 ± 18*

S15

I172V-L443M

0.41 ± 0.13

13 ± 2

65 ± 8

S16

A173G-L443M

3.02 ± 0.94

18 ± 1

111 ± 29

75 ± 17*

S17

S174F-L443M

1.43 ± 0.38

53 ± 6

150 ± 22

132 ± 24*

S18

G100A-I172M-L443M

0.52 ± 0.20

15 ± 1

85 ± 45

561 ± 20

S19

G100A-A173G-L443M

0.77 ± 0.30

65 ± 4

79 ± 20

348 ± 114

S20

G100A-S174F-L443M

0.72 ± 0.27

47 ± 4

107 ± 22

876 ± 175

S21

G100A-I172M-A173G-S174F-L443M

1.21 ± 0.44

21 ± 2

208 ± 73

361 ± 57

S22

I172M-A173G-S174F-L443M

0.65 ± 0.29

39 ± 3

337 ± 51*

453 ± 109

S23

G100A-L443M

1.66 ± 0.97

10 ± 2

110 ± 73

169 ± 65

S24

A173G-S174F-L443M

0.70 ± 0.17

43 ± 5

129 ± 40

140 ± 31*

S25

G100A-A173G-S174F-L443M

3.70 ± 2.20

6 ± 1

55 ± 11

71 ± 20*

S26

G100A-A173G-S174F-A441G-L443M

N.F.

S27

A173G-A441G-L443M

0.30 ± 0.16

56 ± 9

26 ± 5*

173 ± 64

S28

A173G-S174F-A441G-L443M

0.50 ± 0.14

40 ± 8

44 ± 11*

127 ± 23*

S29

S174F-A441G-L443M

2.32 ± 0.17

6 ± 1

20 ± 8*

60 ± 18*

S30

I172V-A173G-S174F-T439S

0.97 ± 0.48

33 ± 4

586 ± 200*

651 ± 200

Mutant

S1

Residues mutated

61 ± 20*

Nisoxetine K i nM

488 ± 103 1,659 ± 66 544 ± 172

524 ± 143

933 ± 269 3,110 ± 410* 2,425 ± 334

85 ± 15* 193 ± 41

Uptake activity, substrate Km values and Ki values for fluoxetine and nisoxetine were determined in a [3H]5HT uptake assay. Results are presented as mean ± SEM from 3 - 61 independent experiments each performed in triplicate. Asterisks denote significantly different inhibitor Ki value compared to WT (p < 0.05; Student's t-test). N.F. = nonfunctional. Values for substrate Km and uptake activity for the 15 single mutants and multiple mutants S1-S10 are from (Andersen et al., 2011), and are included for comparison.

S7

Supplemental Table 4. Impact of mutations of non-conserved residues in the S1 site of NET on uptake kinetics and inhibitor potency Km

Uptake activity

Fluoxetine K i

µM

% of WT

nM

NET WT

0.57 ± 0.09

100

2,993 ± 266

6 ± 1

F72Y

0.26 ± 0.16

28 ± 3

1,230 ± 213

1 ± 0.3

A77G

0.30 ± 0.04

46 ± 4

317 ± 97*

2 ± 1

V141I

0.47 ± 0.06

10 ± 1

318 ± 76*

1 ± 0.1

I142C

0.14 ± 0.01

52 ± 2

985 ± 507*

4 ± 1

L146F

0.11 ± 0.03

52 ± 3

683 ± 247*

2 ± 0.3

V148I

0.35 ± 0.05

18 ± 3

276 ± 114*

4 ± 2

G149A

0.39 ± 0.21

29 ± 4

689 ± 193*

19 ± 6*

F150S

0.37 ± 0.16

10 ± 2

3,515 ± 413

S420T

0.29 ± 0.08

16 ± 4

582 ± 93*

6 ± 2

G422A

0.08 ± 0.02

65 ± 9

442 ± 52*

2 ± 1

M424L

0.25 ± 0.16

43 ± 8

390 ± 33*

1 ± 0.4

A426G

0.07 ± 0.01

51 ± 3

627 ± 102*

3 ± 1

T453C

0.41 ± 0.15

23 ± 3

484 ± 83*

3 ± 1

T477A

1.08 ± 0.40

65 ± 5

2,363 ± 418

9 ± 1

I481V

0.84 ± 0.36

83 ± 3

1,077 ± 155*

13 ± 2*

26 ± 2

746 ± 92*

3 ± 1

Mutant

Residues mutated

Nisoxetine K i nM

20 ± 6*

N1

F72Y-A77G

5.71 ± 1.25

N2

V148I-G149A-F150S

N.F.

N3

A77G-M424L-A426G

0.79 ± 0.24

60 ± 2

399 ± 83*

4 ± 1

N4

M424L-A426G

0.56 ± 0.26

68 ± 3

754 ± 161*

3 ± 0

N5

F72Y-A77G-M424L-A426G

1.50 ± 0.44

28 ± 8

412 ± 77*

6 ± 4

N6

M424L-A426G-T453C

0.60 ± 0.26

39 ± 5

299 ± 101*

4 ± 1

N7

F72Y-A77G-M424L-A426G-T453C

8.06 ± 1.98

35 ± 6

308 ± 112*

4 ± 0

N8

F72Y-M424L-A426G

0.21 ± 0.08

61 ± 4

1,324 ± 296*

2 ± 0

N9

A77G-M424L-A426G-T453C

0.78 ± 0.27

22 ± 7

217 ± 30*

4 ± 1

N10

V148I-S420T

N.F.

N11

V141I-V148I-S420T

N.F.

N12

V141I-T453C

N.F.

N13

V148I-S420T-T453C

N.F.

N14

V141I-S420T-T453C

N.F.

N15

V141I-T453C

0.87 ± 0.38

12 ± 1

448 ± 25*

11 ± 3

N16

A77G-V141I

1.18 ± 0.07

7 ± 3

N17

G149A-F150S-M424L

N.F.

367 ± 24*

4 ± 1

Uptake activity, substrate Km values and Ki values for fluoxetine and nisoxetine were determined in a [3H]DA uptake assay. Results are presented as mean ± SEM from 3 - 48 independent experiments each performed in triplicate. Asterisks denote significantly different inhibitor Ki value compared to WT (p < 0.05; Student's t-test). N.F. = nonfunctional. Values for substrate Km and uptake activity for the 15 single mutants and multiple mutants N1-N8 are from (Andersen et al., 2011), and are included for comparison.

S8

Supplemental Table 5. Binding affinity of fluoxetine and nisoxetine Table S5. Binding affinity of fluoxetine and nisoxetine [125I]β-CIT a

Mutant

Fluoxetine K i

Nisoxetine K i

K d, nM

B max , pmol/mg

n

nM

n

nM

n

SERT WT

6±2

6±3

5

7±2

7

167 ± 31

7

SERT-(NET S1)b

8±3

4 ± 0.2

4

102 ± 16*

4

29 ± 4*

4

SERT-(NET S2)c

3±1

4

3 ± 0.1

3

73 ± 8

3

SERT-(NET S1S2)d

13 ± 1

SERT-(NET S1S2i)e

3±1 13 ± 0.2

NET WT

11 ± 2 2 ± 0.1

4

10 ± 4

4

15 ± 7*

4

11 ± 3

4

65 ± 5*

3

95 ± 17

3

10 ± 6

3

887 ± 115

5

4±1

5

4

7,831 ± 2,644*

3

NET-(SERT S1)f

no specific binding

NET-SERT S2)g

2 ± 0.1

NET-(SERT S1S2i)h

no specific binding

3±1

104 ± 4*

3

The binding affinities of fluoxetine and nisoxetine were determined in a [125I]β-CIT compeition binding assay. Results are presented as mean ± SEM from n independent experiments each performed in dublicate. Asterisks denote significantly different inhibitor Ki value compared to WT (p < 0.05; Student's t-test). a Kd and Bmax values determined from a saturation binding assay. b hSERT-Y95F-G100A-I165V-C166I-F170L-I172V-A173G-S174F-T439S-A441G-L443M-G445A-C473T-T497AV501I. c hSERT-G100A-I108L-Q111K-D400T-A401E-P403A-S404G-L406V-A486I-V489L-K490T-E493D-T497A. d hSERT-Y95F-G100A-I108L-Q111K-I165V-C166I-F170L-I172V-A173G-S174F-D400T-A401E-P403A-S404GL406V-T439S-A441G-L443M-G445A-C473T-A486I-V489L-K490T-E493D-T497A-V501I. e hSERT-Y95F-G100A-F170L-I172V-A173G-S174F-T439S-F440M-A441G-L443M-G445A-K490T-E493D-T497A. f hNET-F72Y-A77G-V141I-I142C-L146F-V148I-G149A-F150S-S420T-G422A-M424L-A426G-T453C-A477TI481V. g hNET-A77G-L85I-K88Q-T381D-E382A-A384P-G385S-V387L-I466A-L469V-T470K-D473E-A477T. h hNET-F72Y-A77G-L146F-V148I-G149A-F150S-S420T-M421F-G422A-M424L-A426G-T470K-D473E-A477T.

S9

SUPPLEMENTAL METHODS Materials. Dulbecco’s modified Eagle’s medium (DMEM), fetal bovine serum (FBS), trypsin, and penicillin-streptomycin was purchased from Invitrogen. All DNA restriction enzymes and T4 DNA ligase enzyme were from New England Biolabs. Cell culture dishes were from Sarstedt AG & Co, and white 96well plates were from Nunc. [3H]5HT (27–28 Ci/mmol), [3H]DA (91–139 Ci/mmol), [125I]β-CIT (2200 Ci/mmol), MicroScint-0 and -20 scintillation mixtures, and 96-well glass fibre filter plates were obtained from PerkinElmer Life Sciences. Polyethyleneimine, R- and S-fluoxetine (HCl salts), 5-HT (creatine sulfate complex) and DA (HCl salt) were purchased from Sigma-Aldrich, and β-CIT were purchased from ABX.

Cell culturing and expression of SERT and NET. Cell culturing and transfection was performed as described.(Andersen et al., 2011) Briefly, COS7 cells were cultured in DMEM media with 10% FBS, 100U/mL pencillin and 100µg/mL streptomycin at 37°C in a humidified 5% CO2 environment. Cells were transfected using TransIT LT-1 DNA transfection reagent by following the protocol supplied by the manufacturer. Before transfection, confluent cells growing in monolayer were resuspended in DMEM at a concentration of 1.3x106 cells pr. mL. Per 96-well plate, 7µg of DNA and 13µL of transfection reagent were mixed in 0.4mL of DMEM and incubated at 20°C for 20 min. Subsequently this mixture was added to the cell suspension and, immediately afterward, the cells were dispensed into white 96-well plates at 50% confluence.

Uptake assays. Uptake assays were performed 40-48 h after transfection. Cells were washed twice with PBSCM buffer (in mM: NaCl, 137; KCl, 2.7; Na2HPO4, 4.3; KH2PO4, 1.4; CaCl2, 0.1; MgCl2, 0.5) on an ELx50 automated microplate strip washer (BioTek Instruments) and left with PBSCM for 20 min prior to uptake experiments. In all NET assays, the PBSCM buffer was supplemented with 1 mM L-ascorbic acid. In inhibition studies (IC50 determinations), cells were preincubated in 40µL of PBSCM per well with 0 or increasing concentrations of inhibitor for 30 min before addition of 10µL of PBSCM containing [3H]5HT (SERT assays) or [3H]DA (NET assays) giving a final substrate concentration of 50nM. Uptake was allowed to proceed at 20°C for 10 min before uptake was terminated as described below. For determination of KMvalues, cells were incubated with increasing concentration of cold substrate (5-HT in SERT assays and DA in NET assays) in PBSCM and 50 nM [3H]5-HT (SERT assays) or 50 nM [3H]DA (NET assays) at 20°C for 10 min. Uptake was terminated by washing three times with PBSCM (ELx50 automated microplate strip washer). The amount of accumulated radiolabelled substrate was determined by solubilising cells in scintillant cocktail (MicroScint-20) with counting of plates in a Packard TopCounter. Non-specific uptake was determined as uptake in non-transfected cells and total uptake was determined in the presence of PBSCM buffer alone. Assays were carried out in triplicate and repeated at least three times.

Cell membrane preparation and radioligand binding assay. Cell membranes of COS-7 cells were prepared as previously described.(Andersen et al., 2009) For saturation binding studies, increasing concentrations of [125I]β-CIT diluted 1:20 with unlabelled β-CIT in PBSCM and 5-30 μg total membrane protein per sample were combined in 96-well plates and total volume adjusted to 200 μL PBSCM per sample. Binding proceeded for 2 h on ice with gentle rocking. Subsequently, membranes were transferred to S10

96-well glass fibre filter plates (Unifilter C, PerkinElmer) preincubated with 0.1% polyethyleneimine (30 µL pr well) using a Packard Bell cell harvester and washed four times with water. Non-specific binding was determined in parallel at membranes from non-transfected COS7 cells. Filter plates were dried and soaked in scintillant followed by counting in a Packard Topcounter. Saturation binding assays were carried out in duplicate and repeated at least three times. For competition binding assays, 5-30 μg total membrane protein was incubated with 0.25 nM [125I]β-CIT in the presence of increasing concentrations of inhibitor in PBSCM using the same protocol as for saturation binding experiments. Competition binding assays were carried out in duplicate and repeated at least three times.

Data analysis. All data analysis was performed using Prism 6 software (GraphPad Inc, San Diego, CA, USA), and carried out as previously described.(Andersen et al., 2009) The inhibitory potency (Ki) for inhibitors were calculated from IC50 values determined in functional uptake assays using the equation: Ki = IC50/(1+([L]/KM)), where [L] is the concentration of [3H]5-HT (SERT assays) or [3H]DA (NET assays) and KM is the Michaelis-Menten constant for the substrate. The binding affinity (Ki) for inhibitors were calculated from IC50 values determined in competition binding assays using the equation: Ki = IC50/(1+([L]/KD)), where [L] is the concentration of [125I]β-CIT and KD is the dissociation constant for β-CIT. Ki values were compared using Student's t-test unless otherwise indicated.

S11

Synthesis of fluoxetine derivatives General procedures. All reagents were obtained from commercial sources and used without any further purification. Proton (1H) NMR and carbon (13C) NMR spectra were recorded on a Varian Mercury Plus (300 MHz) or a Bruker Avance (400 MHz) instrument. Chemical shifts (δ) are reported relative to tetramethylsilane (TMS) as an internal standard. The following abbreviations are used for the proton spectra multiplicities: s, singlet; br s, broad singlet; d, doublet; dd, double doublet, triplet; q, quartet; sep, septet m, multiplet. Coupling constants (J) are reported in hertz (Hz). Microanalyses were performed at the Microanalytical Laboratory at the Department of Chemistry, University of Copenhagen, Denmark. Gas chromatography-mass spectrometry (GC-MS) analysis was performed on a Shimadzu GCMS-QP5050A instrument.

Scheme S1. The synthesis of fluoxetine analogues was initiated from (3-chloropropyl)benzene (3) and (4chlorobutyl)benzene (4). The chloroalkylbenzenes were brominated at the benzylic position using Nbromosuccinimide in the presence of a catalytic amount of benzoyl peroxide in chlorobenzene affording (1bromo-3-chloropropyl)benzene (5) and (1-bromo-4-chlorobutyl)benzene (6), respectively (Silvestri et al., 2004). Phenolate, 2-methoxy-phenolate, 4-trifluoromethyl-phenolate and 2-methoxy-4-(trifluoromethyl)phenolate were subsequently generated by reacting the corresponding phenols with sodium hydride and then reacted with 5 and 6 affording (3-chloro-1-aryloxypropyl)benzenes and (4-chloro-1-aryloxybutyl)benzenes, respectively. The favorable chemoselectivity of the phenolates towards the benzylic bromine over the terminal chlorine gave the desired products in 70-80% yield with purities above 85% (determined by GCMS). The crude products were used directly in reactions with N-methyl- or N,N-dimethylamine affording fluoxetine (1), nisoxetine (2) and five analogues (7-11). Reagents and conditions: (a) NBS, (PhCO2)2, PhCl, 85 ºC, 4 h; (b) ArONa, DMF, room temperature, 12 h; (c) N-Methyl- or N,N-dimethylamine, Bu4NI, K2CO3, MeOH, 80 ºC, 24h (sealed tube). Isolated yields are shown in brackets.

(1-Bromo-3-chloropropyl)benzene (5) (Silvestri et al., 2004).

A slurry of 1-chloro-3-phenylpropane (3) (14.3 mL, 100 mmol), N-bromosuccinimide (16.0 g, 90 mmol), and dibenzoyl peroxide (30 mg) in chlorobenzene (40 mL) was heated at 85 ˚C for 4 hours and then allowed to slowly cool at room temperature for one hour. Precipitated succinimide was filtered off and chlorobenzene S12

was removed on a rotary evaporator. Distillation at 84-85 ˚C/0.3 mbar afforded the benzylbrominated product 5 (14.3 g, 68%) as a colorless liquid. Anal. Calcd for C9H10BrCl: C, 46.29; H, 4.32. Found: C, 46.52; H, 4.39. 1H NMR (400 MHz, CDCl3):  2.40-2.52 (m, 1H), 2.64-2.75 (m, 1H), 3.52-3.60 (m, 1H), 3.66-3.75 (m, 1H), 5.18-5.23 (m, 1H), 7.27-7.42 (m, 5H). 13C NMR (100 MHz, CDCl3):  42.20, 42.85, 51.55, 127.46, 128.82, 129.00, 141.00. GC-MS: EI (m/z, relative intensity): 234 (M+, 1), 153 (47), 117 (16), 104 (10), 91 (100).

(1-Bromo-4-chlorobutyl)benzene (6).

A slurry of 1-chloro-4-phenylpropane (4) (3.37 g, 20 mmol), N-bromosuccinimide (3.56 g, 20 mmol), and dibenzoyl peroxide (10 mg) in chlorobenzene (10 mL) was heated at 85 ˚C for 4 hours and then allowed to slowly cool at room temperature for one hour. Precipitated succinimide was filtered off and chlorobenzene was removed on a rotary evaporator. The product was separated by flash chromatography on silica gel 60F (20 g) by means of heptane. Evaporation of the solvent followed by bulb-to-bulb distillation (110 ˚C, 0.3 mbar) afforded the benzylbrominated product 6 (1.87 g, 76%) as a colorless liquid. Anal. Calcd for C10H12BrCl: C, 48.52; H, 4.89. Found: C, 48.38; H, 4.97. 1H NMR (300 MHz, CDCl3):  1.73-1.85 (m, 1H), 1.92-2.08 (m, 1H), 2.23-2.46 (m, 2H), 3.53 (t, J = 6.3 Hz, 2H), 4.93 (t, J = 7.4 Hz, 1H), 7.26-7.39 (m, 5H). 13

C NMR (75 MHz, CDCl3):  31.30, 37.30, 44.24, 54.57, 125.89, 127.19, 128.80, 141.67. GC-MS: EI (m/z,

relative intensity): 248 (M+, 1), 167 (26), 131 (15), 117 (8), 104 (10), 91 (100).

2-Methoxy-4-(trifluoromethyl)phenol (12).

A solution of 2-methoxy-4-(trifluoromethyl)benzaldehyde (2.04 g, 10 mmol) and 3-chloroperbenzoic acid (70%, 2.47 g, 15 mmol) in dichloromethane (30 mL) was stirred for 12 hours at room temperature. At this stage GC-MS showed full conversion into the corresponding formyl ester and a white slurry had formed. A solution of p-toluenesulfonic acid monohydrate (1.90 g, 10 mmol) in methanol (30 mL) was added and the resulting colorless solution was stirred at room temperature for 2 hours in order to obtain complete transesterification. A saturated aqueous sodium thiosulfate solution (10 mL) was added and after evaporation on celite separation was performed by flash chromatography on silica gel 60F (20 g) by means of EtOAc/heptane (1:2). Evaporation of the solvent and drying in a vacuum oven (1 mbar, 40 ˚C, 2 hours) gave 12 (1.69 g, 88%) as a colorless liquid. Anal. Calcd for C8H7F3O2: C, 50.01; H, 3.67. Found: C, 49.69; H, 3.74. 1H NMR (300 MHz, CDCl3):  3.95 (s, 3H), 5.94 (br-s, 1H), 6.98 (d, J = 8.0 Hz, 1H), 7.07 (s, 1H), 7.18 (d, J = 8.0 Hz, 1H).

C NMR (75 MHz, CDCl3):  56.24, 107.66, 114.40, 119.06, 122.48, 126.13,

13

146.42, 148.40. GC-MS: EI (m/z, relative intensity): 192 (M+, 79), 177 (100), 149 (39).

S13

(3-chloro-1-aryloxypropyl)benzenes or (4-chloro-1-aryloxybutyl)benzenes.

To a solution of the appropriate phenol (1.1 mmol) in dry DMF (5 mL) cooled in an ice-bath under nitrogen was added sodium hydride (60% in mineral oil, 44 mg, 1.1 mmol) in one portion. The resulting slurry was treated in an ultrasound bath at room temperature for 30 minutes under nitrogen to give a clear yellow solution.

Upon

cooling

in

an

ice

bath

(1-bromo-3-chloropropyl)benzene

5

or

(1-bromo-4-

chlorobutyl)benzene 6 (1 mmol) was added in a dropwise fashion during a ten minute period and stirring at room temperature was maintained for 12 hours. The resulting reaction mixtures were poured into ice (150 g) and extracted with Et2O/heptane (1:2, 3 x 25 mL). The pooled extracts were washed with sodium hydroxide (4M, 2 x 10 mL), dried with magnesium sulfate, and concentrated in vacuo to afford (3-chloro-1aryloxypropyl)benzenes or (4-chloro-1-aryloxybutyl)benzenes in 70 to 80 % yields as colorless liquids in purities above 85% with only negligible amounts of bis-substituted by-products according to GC-MS. These materials were used directly for synthesis of 1, 2 and 7-11.

N-Methyl or N,N-dimethyl 3-phenyl-3-aryloxy-1-propylamines or 4-phenyl-4-aryloxy-1-butylamines (1, 2, 7-11).

General procedure: A slurry of the appropriate (3-chloro-1-aryloxypropyl)benzene or (4-chloro-1aryloxybutyl)benzene (approx. 0.5 mmol), tetrabutylammonium iodide (0.037 g, 0.1 mmol) and potassium carbonate (0.27 g, 2 mmol) in a solution of N-methylamine or N,N-dimethylamine (2.0M in methanol, 4 mL, 8 mmol) was heated in a sealed tube for 24 hours. After cooling in an ice bath the white slurry was poured into water (30 mL) and extracted with ethyl acetate (3 x 25 mL). The combined organic phases were pooled, dried over sodium sulphate, filtered, and evaporated. The product was purified by flash chromatography through a short plug of silica by means of triethylamine-ethyl acetate (1:4) followed by evaporation and bulb-to-bulb distillation. For the synthesis of compound 10 (N-methyl-4-phenyl-4-(4-(trifluoromethyl)phenoxy)-1-butylamine), a modified Hindsberg alkylation procedure was also applied (see below).

N-Methyl-3-phenyl-3-(4-(trifluoromethyl)phenoxy)-1-propylamine (1) (Molloy and Schmiegel).

Bulb-to-bulb distillation (150 ˚C, 0.8 mbar) gave the amine 1 (0.10 g, 32 % from 5) as a colorless liquid. Anal. Calcd for C17H18F3NO: C, 66.01; H, 5.87; N, 4.53. Found: C, 66.22; H, 5.66; N, 4.38. 1H NMR (300 S14

MHz, CDCl3):  1.40 (br-s, 1H), 1.97-2.04 (m, 1H), 2.16-2.23 (m, 1H), 2.42 (s, 3H), 2.73 (t, J = 7.2 Hz, 2H), 5.26-5.31 (m, 1H), 6.88 (d, J = 8.5 Hz, 2H), 7.23-7.32 (m, 5H), 7.40 (d, J = 8.5 Hz, 2H). 13C NMR (75 MHz, CDCl3):  36.72, 38.96, 48.43, 78.72, 115.77, 123.12, 125.79, 126.11, 126.75, 127.83, 128.79, 141.05, 160.51. GC-MS: EI (m/z, relative intensity): 309 (M+, 43), 251 (8), 232 (5), 205 (4), 183 (15), 59 (100).

N-Methyl-3-(2-methoxyphenoxy)-3-phenyl-1-propylamine (2) (Srebnik et al., 1988).

Bulb-to-bulb distillation (170 ˚C, 0.5 mbar) gave the amine 2 (0.11 g, 41 % from 5) as a colorless liquid. Anal. Calcd for C17H21NO2: C, 75.25; H, 7.80; N, 5.16. Found: C, 75.34; H, 5.24; N, 5.27. 1H NMR (300 MHz, CDCl3):  1.84 (br-s, 1H), 2.01-2.06 (m, 1H), 2.21-2.29 (m, 1H), 2.43 (s, 3H), 2.75-2.79 (m, 2H), 3.89 (s, 3H), 5.16-5.20 (m, 1H), 6.65-6.69 (m, 2H), 6.82-6.86 (m, 2H), 7.21-7.36 (m, 5H). 13C NMR (75 MHz, CDCl3):  36.68, 38.70, 48.90, 56.07, 80.63, 111.96, 116.26, 120.66, 121.41, 125.96, 127.49, 128.49, 141.96, 147.58, 149.96. GC-MS: EI (m/z, relative intensity): 271 (M+, 2), 167 (36), 148 (66), 124 (100).

N,N-Dimethyl-3-phenyl-3-(4-(trifluoromethyl)phenoxy)-1-propylamine (7) (Molloy and Schmiegel).

Bulb-to-bulb distillation (160 ˚C, 0.5 mbar) gave the amine 7 (0.092 g, 28% from 5) as a colorless liquid. Anal. Calcd for C18H20F3NO: C, 66.86; H, 6.23; N, 4.33. Found: C, 66.88; H, 6.31; N, 4.34. 1H NMR (300 MHz, CDCl3):  1.95-2.02 (m, 1H), 2.14-2.21 (m, 1H), 2.23 (s, 6H), 2.37-2.44 (m, 2H), 5.24-5.26 (m, 1H), 6.88 (d, J = 8.8 Hz, 2H), 7.23-7.33 (m, 5H), 7.40 (d, J = 8.8 Hz, 2H). 13C NMR (75 MHz, CDCl3):  37.01, 45.72, 55.91, 78.63, 115.84, 122.93, 125.90, 126.75, 127.33, 127.85, 128.79, 141.16, 160.64. GC-MS: EI (m/z, relative intensity): 323 (M+, 2), 178 (6), 115 (4), 104 (4), 91 (6), 73 (11), 58 (100).

N-Methyl-3-phenoxy-3-phenyl-1-propylamine (8) (Castelli and Vailati).

Bulb-to-bulb distillation (130 ˚C, 0.2 mbar) gave the amine 8 (0.083 g, 34 % from 5) as a colorless liquid. Anal. Calcd for C16H19NO: C, 79.63; H, 7.94; N, 5.80. Found: C, 79.56; H, 7.96; N, 5.71. 1H NMR (300 MHz, CDCl3):  1.65 (br-s, 1H), 1.97-2.06 (m, 1H), 2.12-2.22 (m, 1H), 2.42 (s, 3H), 2.75 (t, J = 7.0 Hz, 2H), 5.21-5.26 (m, 1H), 6.83-6.88 (m, 3H), 7.15-7.30 (m, 3H), 7.32-7.37 (m, 4H).

13

C NMR (75 MHz,

CDCl3):  36.64, 38.96, 48.63, 78.56, 116.01, 120.82, 125.96, 127.60, 128.72, 129.42, 142.04, 158.23. GCMS: EI (m/z, relative intensity): 241 (M+, 17), 209 (2), 148 (94), 45 (100). S15

N-Methyl-3-(2-methoxy-4-(trifluoromethyl)phenoxy)-3-phenyl-1-propylamine (9).

Bulb-to-bulb distillation (160 ˚C, 0.5 mbar) gave the amine 9 (0.12 g, 35 % from 5) as a colorless liquid. Anal. Calcd for C18H20F3NO2: C, 63.71; H, 5.94; N, 4.13. Found: C, 63.76; H, 6.04; N, 4.32. 1H NMR (300 MHz, CDCl3):  1.90 (br-s, 1H), 2.02-2.08 (m, 1H), 2.21-2.31 (m, 1H), 2.43 (s, 3H), 2.73-2.78 (m, 2H), 3.91 (s, 3H), 5.25-5.29 (m, 1H), 6.70 (d, J = 8.3 Hz, 1H), 6.96 (d, J = 8.3 Hz, 1H), 7.02 (s, 1H), 7.22-7.33 (m, 5H).

C NMR (75 MHz, CDCl3): 36.69, 38.71, 48.69, 56.26, 80.41, 108.56, 114.78, 118.10, 123.81,

13

125.81, 127.84, 128.63, 128.73, 141.11, 149.77, 150.91. GC-MS: EI (m/z, relative intensity): 235 (M+, 20), 192 (74), 177 (72), 148 (100).

N-Methyl-4-phenyl-4-(4-(trifluoromethyl)phenoxy)-1-butylamine (10) (Seeman).

Bulb-to-bulb distillation (180 ˚C, 0.5 mbar) gave the amine 10 (0.14 g, 43 % from 6) as a colorless liquid.

Alternatively, compound 10 was also synthesized by using a modified Hindsberg alkylation procedure which gave the desired compound in higher yields compared to the Finkelstein conditions used above: To a solution of N-methyl-2-nitrobenzenesulfonamide (0.23 g, 0.6 mmol) and tetrabutylammonium iodide (0.037 g, 0.1 mmol) in DMF (2 mL) was added potassium tert-butoxide (1M in THF, 0.6 mL, 0.6 mmol) at 0 ˚C. After stirring for 5 minutes at room temperature 1-(4-chloro-1-phenylbutoxy)-4-(trifluoromethyl)benzene (0.16 g, 0.5 mmol) was added and heating at 90 ˚C was maintained for 24 hours under nitrogen. 1,8Diazabicycloundec-7-ene (DBU, 0.30 mL, 1.2 mmol) and thiophenol (0.21 mL, 2 mmol) were added and stirring was maintained at room temperature for one hour. The orange reaction mixtuxe was poured into brine (30 mL) and extracted with ethyl acetate (3 x 20 mL). The pooled organic extracts were washed twice with brine (10 mL), dried over sodium sulfate, filtered and evaporated. Separation on silica by means of triethylamine-ethyl acetate (1:3), evaporation and bulb-to-bulb distillation (180 ˚C, 0.5 mbar) gave 10 (0.11 g, 68 %) as a light-yellow oil. Anal. Calcd for C18H20F3NO: C, 66.86; H, 6.23; N, 4.33. Found: C, 66.85; H, 6.45; N, 4.26. 1H NMR (400 MHz, CDCl3):  1.42-1.73 (m, 2H), 1.74-1.93 (m, 2H), 2.26 (s, 3H), 2.46 (t, J = 7.0 Hz, 2H), 5.01 (q, J = 6.3 Hz, 1H), 6.74 (d, J = 8.5 Hz, 2H), 7.06-7.17 (m, 5H), 7.26 (d, J = 8.5 Hz, 2H). 13

C NMR (100 MHz, CDCl3):  26.05, 36.14, 36.71, 51.73, 80.26, 115.61, 122.45, 125.70, 126.56, 126.64,

127.62, 128.56, 140.98, 160.50. GC-MS: EI (m/z, relative intensity): 323 (M+, 3), 178 (6), 162 (100).

S16

N,N-Dimethyl-4-phenyl-4-(4-(trifluoromethyl)phenoxy)-1-butylamine (11).

Bulb-to-bulb distillation (180 ˚C, 0.5 mbar) gave the amine 11 (0.13 g, 39 % from 6) as a colorless liquid. Anal. Calcd for C19H22F3NO: C, 67.64; H, 6.57; N, 4.15. Found: C, 67.54; H, 6.77; N, 12.33. 1H NMR (300 MHz, CDCl3):  1.59-1.74 (m, 2H), 1.88-2.06 (m, 2H), 2.21 (s, 6H), 2.30 (t, J = 7.2 Hz, 2H), 5.17 (q, J = 6.4 Hz, 1H), 6.89 (d, J = 8.3 Hz, 2H), 7.24-7.29 (m, 2H), 7.31-7.34 (m, 3H), 7.42 (d, J = 8.3 Hz, 2H). 13C NMR (75 MHz, CDCl3):  24.04, 36.63, 45.60, 59.50, 80.40, 115.74, 122.64, 125.84, 126.69, 126.74, 127.75, 128.69, 141.14, 160.65. GC-MS: EI (m/z, relative intensity): 337 (M+, 1), 176 (3), 162 (8), 143 (5), 91 (6), 58 (100).

S17

SUPPLEMENTAL 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 serotonin transporter. J Biol Chem 285(3): 2051-2063. Andersen J, Stuhr-Hansen N, Zachariassen L, Toubro S, Hansen SMR, Eildal JNN, Bond AD, Bøgesø KP, Bang-Andersen B, Kristensen AS and Strømgaard K (2011) Molecular determinants for selective recognition of antidepressants in the human serotonin and norepinephrine transporters. Proc Natl Acad Sci USA 108(29): 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 Serotonin Transporter: Importance of Ser-438 in Recognition of Citalopram and Tricyclic Antidepressants. J Biol Chem 284(15): 10276-10284. Beuming T, Shi L, Javitch JA and Weinstein H (2006) A comprehensive structure-based alignment of prokaryotic and eukaryotic neurotransmitter/Na+ symporters (NSS) aids in the use of the LeuT structure to probe NSS structure and function. Mol Pharmacol 70(5): 1630-1642. Castelli E and Vailati A (2006) An isolated atomoxetine impurity, processes for the preparation of atomoxetine impurities and their use as reference standards. WO Patent 2006004979. Molloy BB and Schmiegel KK (1977) 3-Aryloxy-3-phenylpropylamines and acid additions salts thereof, useful as psychotropic agents, particularly as anti-depressants. US Patent 4018895. Molloy BB and Schmiegel KK (1986) Treatment of obesity with aryloxyphenylpropylamines. US Patent 4626549. Seeman P (2007) Preparation of (phenyl)phenoxybutanamine derivatives as inhibitors of serotonin and dopamine reuptake for treatment of CNS disorders. WO Patent 2007095756. Silvestri R, Artico M, La Regina G, Di Pasquali A, De Martino G, D'Auria FD, Nencioni L and Palamara AT (2004) Imidazole Analogues of Fluoxetine, a Novel Class of Anti-Candida Agents. J Med Chem 47(16): 3924-3926. Sinning S, Musgaard M, Jensen M, Severinsen K, Celik L, Koldsø H, Meyer T, Bols M, Jensen HH, Schiøtt B and Wiborg O (2010) Binding and orientation of tricyclic antidepressants within the central substrate site of the human serotonin transporter. J Biol Chem 285(11): 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(52): 43694-43707. Srebnik M, Ramachandran PV and Brown HC (1988) J Org Chem 53: 2916-2920.

S18