Letters pubs.acs.org/acschemicalbiology
Mapping the Binding Site for Escitalopram and Paroxetine in the Human Serotonin Transporter Using Genetically Encoded PhotoCross-Linkers Hafsteinn Rannversson,† Jacob Andersen,† Benny Bang-Andersen,‡ and Kristian Strømgaard*,† †
Center for Biopharmaceuticals, Department of Drug Design and Pharmacology, University of Copenhagen, Jagtvej 162, 2100 Copenhagen Ø, Denmark ‡ Lundbeck Research, H. Lundbeck A/S, Ottiliavej 9, 2500 Valby, Denmark S Supporting Information *
ABSTRACT: In spite of the important role of the human serotonin transporter (hSERT) in depression treatment, the molecular details of how antidepressant drugs bind are still not completely understood, in particular those related to potential high- and low-affinity binding sites in hSERT. Here, we utilize amber codon suppression in hSERT to encode the photocross-linking unnatural amino acid p-azido-L-phenylalanine into the suggested high- and low-affinity binding sites. We then employ UVinduced cross-linking with azF to map the binding site of escitalopram and paroxetine, two prototypical selective serotonin reuptake inhibitors (SSRIs). We find that the two antidepressant drugs exclusively cross-link to azF incorporated at the high-affinity binding site of hSERT, while cross-linking is not observed at the low-affinity binding site. Combined with previous homology models and recent structural data on hSERT, our results provide important information to understand the molecular details of these clinical relevant binding sites.
T
structures of hSERT11 and dDAT7−10 supporting this proposition. Whereas the S2 site was initially described as a secondary substrate binding site in LeuT,19 it was later suggested to harbor a low-affinity allosteric inhibitor binding site in hSERT.20 The latter is consistent with the recent X-ray crystal structure of hSERT, which showed simultaneous binding of two escitalopram molecules within both the S1 and S2 site (Figure 1A).11 The structure thus illuminates how the binding at the S2 site can decrease dissociation from the central S1 site, as has been observed in hSERT binding assays.21 Incorporating light-sensitive unnatural amino acids sitespecifically into proteins provides an attractive approach to examine biological function as well as molecular mechanisms with good spatial and temporal resolution. Specifically, photocross-linking amino acids, such as p-azido-L-phenylalanine (azF, Figure S1), have been employed in studies of integral membrane protein drug targets, such as neurotransmitter26,27 and G protein-coupled receptors,22−25 to investigate receptor structure and function and to indentify ligand binding sites. We have recently employed the amber codon suppression technology to introduce photo-cross-linking amino acids into hSERT.28 Here, we apply the methodology to introduce azF into the high- and low-affinity binding sites of hSERT and
he magnitude and duration of serotonergic neurotransmission is controlled by the serotonin transporter (SERT). SERT is an integral membrane-bound protein that facilitates the Na+-, K+- and Cl−-dependent reuptake of serotonin (5-hydroxytryptamine; 5-HT) against its concentration gradient.1 Compounds influencing SERT transport activity influences several neurophysiological processes, and inhibitors of SERT have been widely used for treating various psychiatric diseases, such as anxiety and major depressive disorder.2 The first selective 5-HT reuptake inhibitors (SSRIs) were approved for the treatment of depression in the 1980s, and they remain among the most widely used antidepressants in current clinical use.3 SERT belongs to the solute carrier 6 family of transporters, which also include the transporters for the neurotransmitters dopamine, norepinephrine, γ-aminobutyric acid (GABA), and glycine.1 The first structural insight into this family of transporters was provided by X-ray crystal structures of the bacterial amino acid transporters LeuT and MhsT,4−6 and the Drosophila melanogaster dopamine transporter (dDAT).7−10 Importantly, human SERT (hSERT) was recently crystallized in complex with the two prototypical SSRIs, escitalopram and paroxetine (Figure 1B).11 The transporter structures have shown two distinct binding sites; the central substrate binding (S1) site and a vestibular (S2) site, which is believed to be located at the extracellular side of the S1 site (Figure 1A). Extensive computational and mutational studies, based on homology models of hSERT, have proposed that antidepressants bind with high-affinity within the S1 site,12−18 with recent © XXXX American Chemical Society
Received: April 22, 2017 Accepted: September 14, 2017 Published: September 14, 2017 A
DOI: 10.1021/acschembio.7b00338 ACS Chem. Biol. XXXX, XXX, XXX−XXX
Letters
ACS Chemical Biology
Figure 1. Introduction of photo-cross-linking UAAs into hSERT. (A) X-ray crystal structure of hSERT with two escitalopram molecules (blue spheres) bound (PDB ID 5I73). The locations of the S1 and S2 binding sites are illustrated with dashed lines. TM10−12 have been removed for clarity. (B) Chemical structures of escitalopram and paroxetine. (C) Close-up view of residues within the S1 and S2 binding sites that were mutated to azF (orange sticks).
perform UV-induced photo-cross-linking experiments to address binding of the two SSRIs escitalopram and paroxetine. We have recently established that hSERT is highly amenable to introduction of the photo-cross-linking amino acid azF.28 In brief, this is achieved by cotransfecting HEK293T cells with plasmids carrying the hSERT genes with an amber codon at a preferred position together with an orthogonal pair of transfer RNA (tRNA) and aminoacyl tRNA synthetase (aaRS) for azF. The cells are then cultured in media containing azF ensuring incorporation of the photo-cross-linking amino acid at the desired position, and subsequently the translational efficiency and fidelity is determined in radioligand binding or flux assays.28 In our previous study, we examined five azF mutants in the S1 binding site (Tyr95, Ile168, Tyr175, Phe335, and Phe341) and five mutants in the S2 binding site (Trp103, Tyr107, Trp182, Tyr487, and Lys490). In order to get a more comprehensive coverage of the putative binding sites in hSERT, we decided to include a considerable number of additional putative cross-linking mutants. Selection of the additional mutants was based on three-dimensional orientation and distance from the escitalopram and paroxetine binding sites observed in the recent X-ray crystal structures of hSERT (Figure 1A and C).11 Specifically, all residues within 10 Å of the bound molecules in the crystallized state were assessed, as well as in all other predicted orientations. Residues with side chains facing away from the bound molecule and residues with protein structure residing between the residue and the bound molecule were excluded. In addition to the 10 mutants listed above, we included 13 new azF mutants (Asp98, Val102, Ala169, Ile172, Ala173, Tyr176, Asn177, Tyr232, Asn368, Asp400, Ser438, Thr497, and Val501; Figure 1C). The residues are primarily located in the core transmembrane (TM) domains (TMs 1, 3,
6, 8, and 10) that form the two putative binding sites, thus providing a substantially broader mapping of these binding sites. We introduced azF into the 23 positions in hSERT and functionally characterized the resulting mutants. First, to examine how the transport kinetics were affected by the incorporation of azF, cells expressing hSERT azF mutants were incubated with increasing concentrations of the substrate (5HT) to determine the maximum rate of transport (Vmax) and the substrate concentration needed to achieve half-maximum rate of transport (Km; Table S1). This demonstrated that eight mutants (Asp98, Ile172, Ala173, Asn368, Ser438, Thr497, and Val501) did not transport 5-HT. For a number of these mutants, this was not surprising, as we and others have previously demonstrated the key functional importance of, for example, Asp98, Ile172, and Ser438.12−14,29−31 These mutations either inhibit protein folding and surface expression or rigorously decrease the ability of hSERT to transport 5-HT. In the latter case, the structural integrity of the inhibitor binding site may well be preserved, and information regarding the interaction of ligands with the binding site can still be achieved through binding studies.13,32 Importantly, the remaining 15 mutants all maintained sufficient uptake activity for functional measurements (Table S1). Introduction of azF into 13 of the 15 positions led to a decreased Vmax value, which is likely caused by a reduced level of surface expression, while the Km values were generally in the same range as WT hSERT (Table S1). Importantly, all 15 of these mutants maintained sufficient uptake activity for functional measurements. Next, we determined the inhibitory potency of escitalopram and paroxetine on the 15 functionally active hSERT crosslinking mutants (Table S2). The inhibitory effect of escitalopram was generally not strongly affected by the B
DOI: 10.1021/acschembio.7b00338 ACS Chem. Biol. XXXX, XXX, XXX−XXX
Letters
ACS Chemical Biology
Figure 2. Cross-linking of escitalopram and paroxetine in the high-affinity binding site. (A, C) Results from photo-cross-linking experiments between hSERT azF mutants and 100 nM [3H]escitalopram (A) or [3H]paroxetine (C). The bars represent the radioactivity signal expressed as counts per minute (c.p.m.), and results are presented as mean ± s.e.m. from three to 12 independent experiments. Asterisks (*) denote significantly different cross-linking signal compared to hSERT WT (p < 0.05; Student’s t test). (B, D) X-ray crystal structure of hSERT with escitalopram (B; PDB ID 5I73) and paroxetine (D; PDB ID 5I6X) bound in the S1 binding site. Residues shown to engage in photo-cross-linking (Tyr95, Phe341, Val501) are highlighted (orange sticks).
treated we did not observe cross-linking and second, that the cross-linking signal was abolished when outcompeting the radioligand with an unlabeled inhibitor (Figure S1). All three cross-linking residues, Tyr95, Phe341, and Val501, are located in the central S1 binding site of hSERT (Figures 1 and 2). Importantly, a positive cross-linking signal is possible only if there is close proximity between the engineered azF residue and the ligand in the presence of UV light, and our results thus unambiguously verify that both drugs bind to the S1 binding site with high affinity. Interestingly, the two residues that crosslink to [3H]escitalopram, Y95azF and F341azF, have previously been shown to cross-link to [3H]imipramine and [3H]vortioxetine, respectively, thus suggesting related binding modes of these three different drugs. Moreover, the crosslinking of [3H]escitalopram to Y95azF and F341azF is in very good agreement with both homology models and the X-ray crystal structure of the escitalopram-hSERT complex.11,15,16 In contrast, when comparing the binding modes of paroxetine in the X-ray crystal structure11 and in a recent docking study,18 an apparent difference is observed in the orientation of the fluorophenyl group of paroxetine. Specifically, in the crystal structure the fluorophenyl group is in close proximity to the V501 residue, while it is oriented deeper into the S1 site in the model and is situated close to F341. Due to the inherent distance constraint for cross-linking formation (∼3 Å), our cross-linking results demonstrate that paroxetine binds in close proximity to V501. This could indicate that the binding mode of paroxetine observed in the crystal structure is correct, due to the close proximity of paroxetine and V501. The cross-linking of [3H]paroxetine to V501azF required some additional investigation, as the hSERT V501azF mutant was not functionally active (Table S1). To examine the V501azF mutant further, we performed binding experiments using [125I]-labeled (−)-2β-carbomethoxy-3β-(4-iodophenyl)tropane ([125I]RTI55) on WT hSERT and the V501azF mutant (Figure S3). V501azF bound [125I]RTI55 with an affinity similar to WT hSERT (Kd = 1.4 ± 0.3 nM for WT hSERT and Kd = 1.1 ± 0.4 nM for V501azF), and in a [125I]RTI55 competition binding experiment we found that paroxetine had
hSERT mutants, as 13 mutants induced less than 10-fold change in the potency of escitalopram. One mutant, A169azF, was significantly more sensitive to escitalopram (19-fold increased potency), whereas another mutant, Y95azF, induced 240-fold loss of inhibition for escitalopram compared to WT hSERT. These data support previous mutational and modeling studies, demonstrating that Tyr95 serves a critical role for escitalopram binding in hSERT.15−17 This is also consistent with the recent hSERT X-ray crystal structure showing that Tyr95 is located beneath the amino group of escitalopram forming a cation−π interaction, and it might also form a hydrogen bond with the oxygen of escitalopram.11 For paroxetine, 14 of 15 mutants induced less than 10-fold change in paroxetine potency, with only one mutant, D400azF, being 10-fold more sensitive to paroxetine compared to WT hSERT. Thus, in general, paroxetine was less affected by azF mutants compared to escitalopram. Next, we examined if the two antidepressants, escitalopram and paroxetine, could undergo UV-induced cross-linking to the 23 azF mutants located in the putative binding sites of hSERT. Such a cross-link would provide unequivocal confirmation that the drug binds in close proximity (
