Allosteric modulation of serotonin and dopamine transporters: New insights from computations and experiments

Abstract

Human monoamine transporters (MATs) are critical to regulating monoaminergic neurotransmission by translocating their substrates from the synaptic space back into the presynaptic neurons. As such, their primary substrate binding site S1 has been targeted by a wide range of compounds for treating neuropsychiatric and neurodegenerative disorders including depression, ADHD, neuropathic pain, and anxiety disorders. We present here a comparative study of the structural dynamics and ligand-binding properties of two MATs, dopamine transporter (DAT) and serotonin transporter (SERT), with focus on the allosteric modulation of their transport function by drugs or substrates that consistently bind a secondary site S2, proposed to serve as an allosteric site. Our systematic analysis of the conformational space and dynamics of a dataset of 50 structures resolved for DAT and SERT in the presence of one or more ligands/drugs reveals the specific residues playing a consistent role in coordinating the small molecules bound to subsites S2–I and S2-II within S2, such as R476 and Y481 in dDAT and E494, P561, and F556 in hSERT. Further analysis reveals how DAT and SERT differ in their two principal modes of structural changes, PC1 and PC2. Notably, PC1 underlies the transition between outward- and inward-facing states of the transporters as well as their gating; whereas PC2 supports the rearrangements of TM helices near the S2 site. Finally, the examination of cross-correlations between structural elements lining the respective sites S1 and S2 point to the crucial role of coupled motions between TM6a and TM10. In particular, we note the involvement of hSERT residues F335 and G338, and E493-E494-T497 belonging to these two respective helices, in establishing the allosteric communication between S1 and S2. These results help understand the molecular basis of the action of drugs that bind to the S2 site of DAT or SERT. They also provide a basis for designing allosteric modulators that may provide better control of specific interactions and cellular pathways, rather than indiscriminately inhibiting the transporter by targeting its orthosteric site.

Highlights

• •Human monoamine transporters (MATs) are critical to regulating monoaminergic neurotransmission.
• •We presentstudies of the structural dynamics and ligand-binding properties of two MATs, with focus on the allosteric modulation.
• •DAT and SERT differ in their two principal modes of structural changes.
• •Structural elements lining the S1 and S2 sites point to the crucial role of coupled motions between TM6a and TM10.
• •We provide a basis for designing allosteric modulators with therapeutic potential.

1. Introduction

Monoamine transporters (MATs) are Na⁺/Cl⁻ dependent transporters and part of the solute carrier 6 (SLC6) family. They are composed of three types of neurotransmitter transporters: dopamine transporter (DAT), norepinephrine transporter (NET), and serotonin transporter (SERT) (Kristensen et al., 2011; Rudnick et al., 2014). They mediate the reuptake of their substrate, dopamine (DA), norepinephrine (NE), or serotonin (or 5-hydroxytryptamine, 5-HT), from the synaptic or extrasynaptic space into the intracellular presynaptic compartment, and consequently play a critical role in regulating monoamine homeostasis by terminating monoamine signaling and recycling the monoamines.

Maintenance of the physiological levels of neurotransmitters at the synapse through the efficient function of MATs is essential to preventing various neurological disorders or neurodegenerative diseases. For example, excess dopamine signaling is critically involved in most actions of drugs of abuse (Volkow et al., 2017); whereas its depletion underlies Parkinson's disease (Aguilar et al., 2021). The monoamine transport requires the translocation of the substrate against its concentration gradient, and hence it is accompanied by the thermodynamically driven co-transport of Na⁺ ions down their electrochemical gradient, hence the qualification of MATs as secondary transporters.

Since MATs control the amplitude and duration of monoaminergic neurosignaling, they have been and are being pursued as important therapeutic targets for the treatment of several neurological and mental health disorders (Aggarwal and Mortensen, 2023; Howell and Negus, 2014). Tricyclic antidepressants (TCAs), the selective inhibitors of SERT (SSRIs), and serotonin and norepinephrine reuptake inhibitors (SNRIs) are commonly prescribed for depression, anxiety, and panic disorders and primarily inhibit SERT and NET (Fakhoury, 2016; Hamon and Blier, 2013; Sorensen et al., 2022). Bupropion, a DAT inhibitor, is an antidepressant and smoking cessation aid (Dwoskin et al., 2006; Mortensen and Amara, 2006). The DAT inhibitor, methylphenidate, is used for treating attention-deficit hyperactivity disorder (ADHD) (Faraone, 2018; Heal et al., 2009). Compounds that act as non-selective substrates of MATs and promote reverse transport of monoamines are also in clinical use. These include the racemic mixture of amphetamine isomers (Adderall) that is prescribed to treat ADHD (Faraone, 2018; Heal et al., 2009). In addition, MATs are the primary targets of action of several psychostimulants and recreational drugs of abuse such as cocaine, methamphetamine, 3,4-methylenedioxy–methamphetamine (“ecstasy” or MDMA), and cathinones (or “bath salts”) (Docherty and Alsufyani, 2021; Glennon, 2014; Sora et al., 2009). These substances either block or reverse the transport of monoamines leading to increased dopaminergic neurotransmission and, consequently, increased stimulatory and rewarding effects which further exacerbate addictive behavior (George et al., 2012; Koob and Volkow, 2016). All the compounds described above have in common that they interact with the orthosteric substrate binding site S1 in MATs, competing directly with the monoamine substrates.

On the other hand, allosteric effects in MATs have been observed through the binding of different small molecules to secondary binding sites. This study provides a comprehensive analysis of known allosteric effects elicited upon binding of small molecules to secondary binding sites, observed in experiments and computations conducted for SERTs and DATs. We carried out a systematic integrated analysis of all structures resolved for SERT and DAT, currently available in the Protein Data Bank (PDB), which shows the existence of various functional conformations and ligand binding pockets that have long been proposed to be involved in allosteric regulation based on computational studies of MATs (see reviews (Cheng and Bahar, 2019; Grouleff et al., 2015; Manepalli et al., 2012)). Additionally, structural evidence of an allosteric binding site and its potential allosteric modulation, which was elusive in dDAT structures, is presented based on structural studies of human SERTs. We have also identified the residues that closely (within 4.5 Å atom-atom distance) interact with substrates and ligands, revealing their crucial role in coordinating substrate and ligand binding events between the primary (orthosteric) and secondary (allosteric) binding sites of either SERT or DAT.

For a comprehensive assessment of possible mechanisms of allosteric couplings intrinsically accessible to DAT and SERT, we further examined the space of conformations and transitions accessible to these transporters. Several studies in the last decade examined the dynamics of transporters using molecular dynamics simulations, which provide insights into full-atomic interactions and various stages of the transport function (Koldsø et al., 2011, 2013; Hellsberg et al., 2019; Sinning et al., 2010; Thomas et al., 2012; Cheng and Bahar, 2015; Gradisch et al., 2022). Here we take a different approach, mainly focusing on cooperative changes in structure which are typical of allosteric events. We make inferences based on ensemble analysis of extensive experimental (structural) data, and on elastic network model (ENM) analyses of representative structures, using the ProDy interface (Zhang et al., 2021). Excellent agreement between the information conveyed by experiments and computations supports the utility of ENMs for efficient assessment of the space of conformers accessible to this family of transporters. Notably, the transmembrane helices TM6 (TM6a part) and TM10 contribute significantly to the allosteric action induced upon binding ligands to the secondary binding site. Residues within these helical regions, which line the S2 site, exhibit strong dynamic couplings to the orthosteric (primary binding) site, as revealed by the inter-residue cross-correlations computed for equilibrium motions using the Gaussian Network Model (GNM) (Haliloglu et al., 1997; Rader et al., 2005). These collective findings provide valuable insights into the allosteric mechanisms controlling the functional dynamics of SERTs and DATs, paving the way for new potential strategies for modulating their function and for formulating therapeutic interventions.

2. Structural data reveal two sites S1 and S2 (adaptively spanning subsites S2–I and –II) for neurotransmitter and/or ligand binding

Recent significant discoveries have facilitated our structure-based understandings of how MATs work and interact with endogenous and exogenous ligands. Many studies have been pivotal in guiding further studies to elucidate the molecular mechanisms of substrate transport as well as those of antidepressant and pychostimulant binding, starting from the elucidation of crystal structures of the homologous bacterial (Aquifex aeolicus) transporter LeuT (Singh et al., 2007; Yamashita et al., 2005; Zhou et al., 2007), followed by those of the Drosophila melanogaster dopamine transporter (dDAT) (Penmatsa et al., 2013, 2015), and human serotonin transporter (hSERT) (Coleman and Gouaux, 2018; Coleman et al., 2016, 2019; Plenge et al., 2021; Yang and Gouaux, 2021), which share the LeuT fold. This fold consists of twelve transmembrane (TM) domains (TMDs) or TM helices, ten of which are organized into two topologically inverted repeats (TM1-TM5 and TM6-TM10) (Yamashita et al., 2005). At the center of the two repeats is the primary site S1 for substrate binding. Nearby residues transiently bind co-transported ions. These occupy hinge-like regions disrupting the helices TM1 and TM6 at approximately their mid points. Finally, MATs have large N- and C-termini that extend into the cytoplasm.

Many dDAT and hSERT structures have been co-crystallized with their substrates (Penmatsa et al., 2015; Yang and Gouaux, 2021), NET-specific reuptake inhibitors (NRIs) (Penmatsa et al., 2015; Pidathala et al., 2021), SSRIs (Coleman and Gouaux, 2018; Coleman et al., 2016, 2019, 2020), TCAs (Penmatsa et al., 2013; Plenge et al., 2021), SNRIs (Pidathala et al., 2021), amphetamine (Penmatsa et al., 2015), cocaine (Penmatsa et al., 2015), GRIs (Joseph et al., 2022), and other small molecules (Coleman et al., 2019; Singh et al., 2023) (Table S1). In all these studies, the ligands are bound to the primary substrate binding site, S1. Fig. 1A illustrates the binding of sertraline, an SSRI, to SERT at site S1, together with two sodium ions (blue spheres) and one chloride ion (magenta sphere). On the other hand, further structural and computational studies as well as biochemical evidence have suggested the presence of additional ligand-binding sites within the MATs. These observations opened the intriguing possibility that multiple low-affinity binding sites are present in regions distinct from the direct translocation pathway, which could serve as allosteric sites for modulating the conformational dynamics of the transporter thereby affecting its function.

Fig. 1.

Cryo-EM structures resolved for SERT demonstrate two major sites for binding ligands. All panels display side (upper) and top (lower) views of the PDB structures as labeled between the two diagrams. (A) SERT in outward-facing (OF) state, complexed with the SSRI sertraline (red) bound to primary site S1 (Coleman and Gouaux, 2018). (B) SERT complexed with two serotonins, 5-HT (cyan), bound to S1 and secondary/allosteric site S2 (subsite I, shortly designated as S2–I) in the occluded conformation closed to both extracellular (EC) and intracellular (IC) regions (Yang and Gouaux, 2021). (C) Human SERT, resolved in OF state, complexed with S-citalopram (green) at S1 and Br-citalopram (yellow) at S2-II (Coleman et al., 2016). (D) Wild-type hSERT (in OF state) complexed with vilazodone (dark blue) at S2–I and imipramine (light pink) at S1 (Plenge et al., 2021). The green circles highlight the site S1; red and yellow highlight the allosteric subsites S2–I and S2-II, respectively. (E) Close-up view of S2-binding sites for serotonin (left), Br-citalopram (middle) and Vilazodone (right).

Early molecular dynamics (MD) simulations using the crystal structure resolved for LeuT indeed revealed that, along with a leucine molecule bound to the primary binding site, another leucine could simultaneously occupy a secondary site relatively closer to the extracellular (EC) surface (Shi et al., 2008). This second site (designated as S2) was located near the so-called EC gate (composed of F253, R30 and D404) (Shi et al., 2008), the opening/closing of which would provide/prevent access to the primary site S1 from the EC medium. As such, small-molecule binding to S2 was proposed to “allosterically” trigger a conformational change that stabilizes the closure of the gate in the currently resolved outward-facing (OF) state of the transporter. This would promote the transition of the transporter from substrate-bound occluded to inward-facing (IF) open (IFo) conformation thus enabling the release of the substrate and Na⁺ ions to the intracellular (IC) side. Simultaneous binding of two alanine molecules to S1 and S2 site was also observed in later MD simulations of LeuT (Cheng and Bahar, 2013), together with a coupling between the motions of the alanines bound to the two sites, in support of an allosteric effect on S1 induced upon substrate-binding to S2. While the presence of a secondary site S2 (as a high-affinity or low-affinity site for substrate binding) has been recognized in LeuT and other members of the neurotransmitter sodium symporter (NSS) family (Nyola et al., 2010), like MhST (Quick et al., 2018), its presence has also been questioned through other experimental methods (Erlendsson et al., 2017). Furthermore, its significance in facilitating substrate transport remains relatively uncertain (Krishnamurthy and Gouaux, 2012; Wang et al., 2012; Wang and Gouaux, 2012). Finally, whether this mechanism also exists and/or is required in the mammalian MATs has eluded unambiguous determination (Yang and Gouaux, 2021).

The structure resolved by Coleman et al. (2016) for the OF state of hSERT in the presence of two (S)-citaloprams (Fig. 1C), an antidepressant provided structural evidence of the propensity of a MAT to simultaneously bind two ligands. In addition, this also provided mechanistic and structural support for earlier biochemical experiments that are described below demonstrating allosteric effects on SERT (Plenge and Mellerup, 1985; Wennogle and Meyerson, 1982). One of the citaloprams occupied the orthosteric site S1, analogous to the primary site seen in LeuT and dDAT structures, whereas the other resided at a position in the EC vestibule that was somewhat distinct from the S2 site in LeuT. However, in the same study, the authors also reported the density of a different ligand resembling maltose, part of the detergent molecule, that resided in the same allosteric site. This discovery provided the stimulus for more studies on the validity of the allosteric site, its substrate specificity as well as its function in allosteric modulation. A more recent study (Yang and Gouaux, 2021) indeed confirmed the simultaneous binding of two serotonins to SERT (Fig. 1B). This structure showed that the 2nd serotonin was bound to a pocket partly overlapping with the allosteric site found by Coleman et al. (2016). The site remained occupied by serotonin throughout the transport cycle (Yang and Gouaux, 2021), and it was therefore proposed to not directly couple to the translocation process. Furthermore, it was noted that the second serotonin bound deeply into the structural “cradle” composed by TM10 and TM12 (Yang and Gouaux, 2021); and thus, its binding would not occlude access to the central orthosteric site but would, on the other hand, induce conformational changes that could affect the interactions of serotonin within the site S1.

The same group recently reported additional conclusive structural support for the existence of the allosteric site S2 in SERT, upon resolving a cryo-EM structure in the presence of the antidepressant vilazodone bound to that allosteric site (Plenge et al., 2021). Furthermore, vilazodone was shown to allosterically prevent the dissociation of ligands from the orthosteric site. In this structure (Fig. 1D), vilazodone occupies an extensive space in the EC vestibule encompassing the sites where both the second citalopram and serotonin were found to bind in the above mentioned studies (Coleman et al., 2016; Yang and Gouaux, 2021), suggesting that the site S2 is not as precisely defined as the primary site S1 but does exist and impacts the transport function. It essentially forms an extended binding pocket that adapts to different types of ligands.

The adaptability of site S2 to accommodate different ligands is in line with the conformational predisposition of allosteric sites to prompt a cooperative change in conformation upon binding an allosteric modulator. In this case, a variety of small molecules have been observed to adopt different binding poses at/near the site S2 of DAT and SERT, highlighting the necessity for a unified nomenclature that distinguishes between subsites.

Based on examination of current experimental and simulated data for DAT and SERT conformations, we group the allosteric binding sites into two subsites, S2–I and S2-II. Representative members of S2–I and S2-II include serotonin bound hSERT (Yang and Gouaux, 2021) and (S)-citalopram bound hSERT (Coleman et al., 2016) illustrated in Fig. 1 panels B and C, respectively. The differentiation between S2–I and S2-II can be visualized based on the residues that coordinate the ligand, as illustrated in Fig. 1E. While S2–I and S2-II share two coordinating residues (F556 and E494) (as can be viewed in Supplementary Figs. S1 and S2), S2–I is relatively closer to TM12 with Y579 playing an important role in its coordination, whereas S2-II is closer to the center of the TM helical bundle, interacting with EC gating residues R104 and E493. In the case of vilazodone bound to the allosteric site, the ligand spans both S2–I and S2-II regions: it tends to penetrate closer to TM12.

In summary, these significant structural findings have provided evidence for the existence of an extended allosteric binding site S2, different from the primary substrate binding site S1. S2 is composed of two subsites S2–I and S2-II. Several small molecules (substrates and inhibitors) bind to this site. Some of these small molecules have been observed to engage in modulating MAT function in an allosteric manner.

3. Evidence of functional impact made by allosteric modulators

In addition to structural support for MAT sites that can allosterically modulate the transport activity, earlier functional and behavioral studies have pointed to the allosteric modulation of MATs. The idea of a functional allosteric site in SERT has been around for almost four decades as biochemical dissociation studies demonstrated that some antidepressants and serotonin itself had allosteric effects by slowing the dissociation of a radiolabeled ligand bound to the orthosteric site (Plenge and Mellerup, 1985; Wennogle and Meyerson, 1982). Mutagenesis approaches and MD simulations further explored this idea and identified key residues (Plenge et al., 2012) that constitute the allosteric site which closely resembles the S2 site in LeuT (see details in review (Cheng and Bahar, 2019)). Mutations at these residues decreased the allosteric potency of ligands bound to the secondary site, but did not affect the affinity to bind the primary site (Plenge et al., 2012).

These earlier studies have now led to the identification of compounds with therapeutic potential demonstrating similar allosteric effects on ligand dissociation but with significantly higher affinities for the allosteric site over the orthosteric site (Plenge et al., 2020; Salomon et al., 2023). In addition to the observations on allosterically modulating ligand dissociation from the orthosteric site, allosteric effects have been observed in studies of MATs including effects on the affinity of ligands that bind S1, non-competitive inhibition of substrate translocation, and effects on behaviors in animal models (Niello et al., 2020).

Our group identified a potential allosteric pocket (Kortagere et al., 2013; Mortensen and Kortagere, 2015) in the EC vestibule of SERT using a combination of MD simulations and comparative genomics analyses. This allosteric pocket, conserved across all MATs, overlaps significantly with the site S2 occupied by the second serotonin and citalopram in the above-described structures (Fig. 1). Virtual screening of compound libraries yielded several ligands that engage with this site, and our biochemical assays showed that engaging this site would result in allosteric effects such as enhanced transport function at low substrate concentrations, potentiated MDMA induced reverse transport, and the stabilization of an OF conformation of SERT (Kortagere et al., 2013).

More recent application of the same integrated experimental and computational protocol to DAT led to the discovery of KM822, a non-competitive allosteric ligand specific to DAT (Aggarwal et al., 2019). Biochemical assays confirmed the interaction of this compound with the allosteric pocket and functional assays demonstrated that it acts as a non-competitive inhibitor of dopamine uptake and decreases the affinity of cocaine for DAT. In an animal model, we demonstrated that the compound did not affect the locomotion of planarians but inhibited cocaine-induced hyperlocomotion. In a follow up study our group in collaboration with the Blakely group also demonstrated that KM822 attenuates an amphetamine-induced behavior in C. elegans roundworms and does not induce this behavior by itself (Refai et al., 2022). Furthermore, we found that this did not occur in an animal with DAT genetically removed, supporting that this behavioral effect is elicited through the specific allosteric modulation of DAT function.

By combining the above mentioned biochemical-substituted cysteine scanning accessibility experiments with molecular structure-based modeling, we further characterized the EC vestibular site in DAT where KM822 binds (Fig. 2A) (Aggarwal et al., 2019, 2021). The binding site for KM822 aligns with the site involved in the allosteric effects of antidepressants described above (Plenge et al., 2012) and broadly overlaps structurally with the site S2-II. hDAT residues coordinating KM822 include W84, R85, F217, D476, H477, and Y548 (thick sticks in Fig. 2A). Notably, the dopaminergic psychostimulant sydnocarb, which is another non-competitive inhibitor of DAT, was observed in MD simulations to sample both subsites S2–I and –II, exhibiting the intriguing capability of translocating between the two subsites (Fig. 2B) (Aggarwal et al., 2021). Fig. 2C shows that among the two binding sites visited by sydnocarb during the course of simulations, one closely approximates the S2–I site occupied by 5-HT (Yang and Gouaux, 2021) (shown in Fig. 1B), and the other closely overlaps with the extended S2-I-II site selected by vilazodone (Plenge et al., 2021) (shown in Fig. 1D). Fig. 2D provides a view of the spatial locations of these experimentally and computationally detected and confirmed sites, with the help of color-coded space-filling representation. Fig. 2E shows the sequence alignment between hDAT, hSERT and hNET near the residues coordinating this site.

Fig. 2.

Multiple binding poses of substrates and/or inhibitors to hDAT from simulations and experiments. (A) MD simulations of hDAT OFo conformer (orange), with explicit membrane (green) and in 0.15M NaCl solution (not shown) identified the binding site for KM822 (pink spheres) at the EC vestibule. Residues in thick sticks are KM822 binding residues, captured in 100 ns-long MD simulations (Aggarwal et al., 2021). (B) Binding poses sampled by the allosteric modulator sydnocarb obtained in multiple MD runs. The panel displays the superposition of two snapshots from different runs where sydnocarb is stabilized at the respective subsites S2–I and S2-II (light orange and yellow green). KM822 binding site resembles the Site S2-II of sydnocarb. (C) Comparison of the binding pose of sydnocarb to the S2–I site of hDAT as predicted by MD simulations, with the binding poses of vilazodone (light blue) and imipramine (light green) bound to SERT resolved (Plenge et al., 2021) by cryo-EM. (D) Binding sites S1, S2–I, and S2-II are shown in green, red and blue surface, respectively. For details, see refs (Aggarwal et al., 2021; Plenge et al., 2021). (E) Sequence alignment near the allosteric binding sites shown in D. Residue numbers are displayed for hDAT.

Other studies have similarly reported allosteric modulation of MATs. A peptide from a predatory marine cone snail, χ-MrIA, is a non-competitive and selective inhibitor of NET and was modeled to bind a large domain filling up the extracellular vestibule of the transporter (Paczkowski et al., 2007). The Rothman group has identified compounds that show unconventional pharmacology and suggested that they allosterically modulate both DAT and SERT. Observed effects of these compounds include modulation of amphetamine-elicited reverse transport, attenuation of orthosteric ligand dissociation, and distinct effects on transport versus reverse transport activity (Pariser et al., 2008; Rothman et al., 2009). Some of these compounds have been developed further and demonstrated to interfere with the interaction of DAT with the HIV-derived transactivator of transcription (TAT) peptide (Sun et al., 2017; Zhu et al., 2022). In some recent studies, lignan glycosides from an East Asian shrub used to treat depression were found to modulate MATs in an allosteric non-competitive manner. Computational modeling suggested the lignan glycosides would bind to a site different from the orthosteric site but related to S2 in LeuT (Huang et al., 2022; Liu et al., 2022). Finally, the zinc ion that is present in the brain has been shown to specifically modulate DAT function but not NET and SERT functions (Meinild et al., 2004; Scholze et al., 2002). The site of interaction was identified to lie outside the orthosteric site, and the zinc ion was found to stabilize an outward-facing conformation (Loland et al., 1999). It was furthermore determined that zinc displays an intriguing biphasic effect on DAT activity as it stimulates transport activities at low concentrations and inhibits incompletely at higher concentrations (Li et al., 2015).

These findings further support the existence of an (extended) allosteric site (with multiple subsites) within the EC vestibule whose targeting would impact the transport function. The studies further demonstrate the ability of the substrate and ligands to translocate between them, and/or to simultaneously have two substrates or ligands occupying them. The binding of small molecules to these sub-sites could potentially allow for the allosteric modulation of DAT activity, opening novel possibilities for therapeutic interventions.

4. Identification of SERT- and DAT-specific residues coordinating the ligands bound to sites S1 and S2

The functional and behavioral studies described above point to the significance of exploiting the allosteric sites of MATs for developing novel strategies for modulating neurotransmission. Toward this goal, we performed a comparative analysis of the coordination geometry of substrates/ligands bound to MATs at the primary and secondary sites using exhaustive data from the PDB, mainly 50+ DAT and SERT structures resolved to date and listed in Table S1. The table provides information on the type of transporter, its conformational state (OF/IF) and the conformation of its EC or IC gates (open or closed, designated as o or c), and the identity of ligands (substrate and non-substrate). The hSERT EC gates include 2 pairs of gating residues, R104 in TM1B and E493 in TM 10 form EC1, while Y176 in TM3 and F335 in TM6A form EC2. Likewise, for dDAT, R52 in TM1B and D475 in TM10 form EC1, while Y124 in TM3 and F319 in TM6A form EC2. Concerning hSERT IC gates, R79 in TM1A and D452 in TM8 form IC1, while K85 in TM1A and D360 in TM7 form IC2. Similarly, in dDAT, R27 in TM1A and D435 in TM8 form IC1, while W30 in TM1A and Y331 in TM6B form IC2. The distance between these residues pairs indicates the presence of a salt bridge formed between them, which serves as one of the factors considered when assessing whether the structures are in open or closed conformations while being in the OF or IF states (see more in Table S1a).

The dDAT and hSERT residues that coordinate the primary and secondary binding sites were identified based on their atom-atom distance (≤4.5 Å) from the ligand (Table S2). Key residues that appeared to be consistently in close interaction with the substrates in the majority of dDAT and hSERT structures are illustrated in Fig. 3. The figure shows the superposition of 21 OF dDAT (panel A) and 19 OF hSERT (panel B) structures.

Fig. 3.

A few key residues consistently coordinate the substrates/ligands bound to the S1 and S2 sites of dDAT and hSERT. (A) Superposition of 21 dDAT OF structures with substrates bound to S1 (cyan) and S2 (pink) sites. Small molecules (substrates, drugs, or small molecules) bound to S1, shown in spheres, are D-amphetamine, nortriptyline, nisoxetine, reboxetine, methamphetamine, cocaine, L-norepinephrine, milnacipran, duloxetine, and (1S,2S)-tramadol; and those at S2 site include ethylene glycol, diethylene glycol diethyl ether, and SKF89976A (see Table S1). Note that S2 is occupied by organic solvents or other molecules such as monosaccharides or lipids, rather than DAT-specific substrates or inhibitors. (B) Superposition of 19 hSERT structures with substrates bound to S1 (cyan), S2–I (yellow) and S2-II (pink). Residues that interact with substrates are shown in green sticks on the left diagram. Phe319 and Phe335 are EC2 gating residues in dDAT and hSERT, respectively. Small molecules at the S1 site are Br-citalopram, sertraline, fluvoxamine, paroxetine, Br-paroxetine, I-paroxetine, 4-(4-fluorophenyl)-2-piperazin-1-yl-1,3-thiazole, serotonin; those at S2–I are dodecyl-β-D-maltoside, serotonin, and vilazodone; and those at S2-II, dodecane, (S)-citalopram, and Br-citalopram.

At the primary site S1, all dDAT residues that coordinate the substrate (F43, D46, V120, Y124, F319, F325, S421, and G425) sequentially and structurally align with their counterparts in hSERT (Y95, D98, I172, Y176, F335, F341, S438, and G442). While the identity of the amino acid may differ, their physicochemical properties and spatial positions and orientations are closely retained. This supports the notion that the primary binding site is conserved across MAT family members and plays a critical role in binding the substrate - dopamine, serotonin, or norepinephrine in DAT, SERT and NET, respectively. Evidently, substitutions at these key residues lining the orthosteric site would have functional consequences.

The allosteric site, on the other hand, presents more variations, as could also be inferred by the adaptable shape of S2 pointed out above. Furthermore, the lack of dDAT structures resolved in the presence of S2-bound substrate or ligands limits our ability to make comparisons between SERT and DAT. Aside from the compound SKF89976A bound to S2 site of GABA transporter-like dDAT structure (PDB ID: 7WLW) which carries GABA transporter (GAT) mimetic mutations (Joseph et al., 2022), all dDAT whose S2 site is occupied by a ligand correspond to non-specific binding of an organic solvent molecule (e.g., diethylene glycol diethyl ether) or a monosaccharide (e.g. beta-D-glucose). While this shows the avidity of that site to bind a ligand, it does not lend itself to making inferences on specific residues that coordinate the ligand. Nevertheless, it is worth noting that SKF89976A resides in a pocket that closely resembles the site S2-II of hSERT and the residues coordinating SKF89976A in GAT-like dDAT (R476, Y481, and G538) structurally align with those (E494, P499, and F556) at the S2-II site of hSERT.

In the following, we present the evaluation of MAT allosteric binding sites based exclusively on SERT structures experimentally resolved in the presence of S2-bound substrates or ligands (Fig. 3B). Notably, in 6 out of 11 such structures, the S2 site is occupied by a detergent (dodecyl-β-D-maltoside), and in one case, by dodecane (Table S1). Yet, SERT structures present a sufficiently robust dataset to make inferences on residues contributing to the S2–I and S2-II subsites. As previously discussed, the S2–I and S2-II subsites are distinguished by the position of the ligand relative to TM12, and the extent of penetration of ligand into the gap between TM10 and TM11, and EL6 (Fig. 2D). If a ligand crosses this gap (yellow spheres in Fig. 3B), it belongs to S2–I subsite; otherwise, it belongs to S2-II (salmon spheres). A key difference between S2–I and –II is at the positioning of F556. In S2–I, the phenyl group of F556 points out perpendicularly with respect to TM11 helix (with the dihedral angle χ₁ varying in the range −90° to 90° (Fig. S2). This permits the ligand to penetrate the gap between TM11 and TM10. In contrast, the phenyl group of F556 in S2-II is almost parallel to TM11 with a dihedral angle of approximately 174° (Fig. S2). This trans-conformation of F556 prevents the substrate from entering the space between TM10 and TM11. However, it is important to note that both subsites I and II are coordinated by the same key residues F556 and E494, despite their different sidechain orientations. In the case of subsite II, these residues are complemented by P561.

The classification and criteria of these two distinct subsites concurs with the binding poses of the lead compound Lu AF88273 for SERT allosteric inhibitors that were recently proposed (Salomon et al., 2023). The lead compound Lu AF88273 was shown, through docking, MD simulations as well as mutagenesis experiments, to adopt two distinct binding poses IPC (S2–I) and CPI (S2-II) at the S2 site with high affinity, where the F556 residue would accommodate the binding with different dihedral angles similarly to what is observed in Fig. S2. We note that a prior docking simulations of hSERT and MDMA, a psychotherapeutic agent, also determined two possible binding poses for MDMA, termed “5-HT-like” (S2–I) and “escitalopram-like” (S2-II) (Islas and Scior, 2022).

The identification and comparison of critical residues that coordinate substrates or drugs bound to hSERT or dDAT provides new insights into the design and refinement of transporter-specific allosteric binding site. Clearly, specificity is harder to achieve by targeting the primary site that is evolutionarily conserved; in contrast, the allosteric site presents more variability depending on the specific transporter. The residues identified here are within sufficiently close contact distance (<4.5 Å) from the substrate or ligands to play an important role in selectively binding these small molecules.

5. Ensemble analysis of DAT and SERT structures reveals the principal changes in structure that distinguish the two transporters

To gain an overall understanding of the conformational space sampled by MAT family members DAT and SERT, we performed an ensemble analysis of all the 50 structures experimentally resolved to date for dDATs and hSERTs. The properties of these structures are listed in Table S2. Ensemble analysis is broadly employed to study MD trajectories. In our case, we analyzed the ensemble of experimentally resolved structures to determine the principal modes of structural variations or their relationships to theoretical predictions. To this aim, first we carried out a principal component analysis (PCA) of the ensemble of structures to determine the two principal components PC1 and PC2 of structural variations. Then, we examined the distribution of dDAT and hSERT structures in the conformational subspace spanned by these first two PCs, shown in Fig. 4A. The structures form clusters indicative of subpopulations of dDAT and hSERT structures that exhibit similar conformational (OF, IF, or occluded) features. Notably, the two types of MATs, SERT and DAT, also separate in two halves of the conformational map (divided by the diagonal dashed line).

Fig. 4.

Principal structures changes observed in resolved DAT and SERT structures. (A) Projection of the resolved structures onto the conformational subspace spanned by principal components PC1 and PC2 derived from experimental data. Each structure is shown by a color-coded dot. The structures are grouped into five main subgroups, enclosed in dotted circles, as labeled. The diagonal dashed line separates hSERT and dDAT structures. PC1 describes the passage from OF state to occluded and then to IF state. PC2 refers to orientational changes in selected helices (TM11-12, TM4-5, TM9-10) accompanied by rearrangements in the EC loop regions. (B) Movements of residues along PC1 (left; green arrows) and PC2 (right; red arrows) illustrated for OF hSERT (PDB ID: 5I6X (Coleman et al., 2016),). Top and bottom diagrams display the structure from front and back views. See also the Movies S1 and S2for the animations of the collective movements occurring along the two PCs.

Examination of the organization of structures along the abscissa shows that changes in structure defined by PC1 refer to the transition of the transporters from OF state (with EC gates in open (OFo) or closed (OFc) conformations) to the occluded state (approximately in the middle of PC1), and then to the IC state. In other words, PC1 accounts for the transition of the transporter all the way from its OF open state to IF state, via the occluded state (see also Movie S1). Panel B (left) describes the major structural changes occurring along this mode (hSERT OF diagrams viewed from front (top) and back (bottom)). The movements of TM1b, TM6a, TM11, and TM12 facilitate the closing of EC vestibule, between OF and occluded states. Subsequently, the movement of TM1a together with the partial contributions from TM5, TM4, and TM9 open the IC vestibule as the protein transitions from occluded to IF state (Fig. 4B, Movie S1). The principal changes observed in the resolved SERT conformers resemble those observed in detailed simulations of DAT transition between OF and IF states (Cheng and Bahar, 2015).

As to PC2, careful examination of the motions along this mode (see Movie S2) points to helical displacements indicated by the arrows in Fig. 4B (right). We observe therein selected tilting of helices in TM10b, TM11, and TM12 (see Movie S2). The conformational representations of these TM helices may be instrumental in accommodating ligand binding to S2. PC2 further enables cooperative changes at the cytoplasmic end of the helices (Fig. 4B, bottom), thus providing additional conformational flexibility to accommodate the release of the substrates. Notably, PC2 also separates dDAT OF from hSERT OF, highlighting the difference between them.

Table S2 lists the structural properties as well as the substrate/ligands bound to each of the examined structures. Structural properties include the state (open or closed) of the two EC gates (EC1 and EC2) and IC gates (IC1 and IC2) as well as the dihedral angle of the phenylalanines near the EC2 gate (F335 in hSERT and F319 in dDAT). These geometric criteria distinguish the OFo and OFc conformers, but these two subsets of conformers almost overlap when projected onto PC1-PC2 subspace (see Fig. 4A) as their differences mostly originate from different side chain rotational states at the EC gates. The accompanying minimal changes in α-carbon coordinates are not visible in this subspace of cooperative changes in the overall architecture. In contrast, hSERT OFo structures are relatively spread out in the conformational space, pointing to the higher conformational flexibility of hSERT in the OF state, compared to dDAT in the OF state.

Overall, this mapping of existing structures onto the principal subspace of conformations spanned by PC1 and PC2 highlights (i) the structural differences between SERT and DAT (even though both share the LeuT fold). (ii) elucidates the major structural elements whose concerted rearrangements enable the transition of the transporters between IF, occluded, and OF states along the transport cycle. Furthermore, the analysis shows that (iii) PC1 plays a key role in enabling the transition between OF and IF states, evidenced by the cooperative movements of structural elements surrounding the S1 site and the EC and IC gates; whereas (iv) PC2 allows for the rearrangements of TM11 and TM12, which have been pointed out to play an important role in accommodating the ligands bound to S2.

6. Hinge regions modulating the global movements coincide with substrate/ligand binding sites

To gain a deeper understanding of the cooperative conformational motions accessible to these transporters under physiological conditions, and the way substrate- or ligand-binding affects (or controls) these motions, and thereby the transport function, we performed a Gaussian Network Model (GNM) analysis of the intrinsic dynamics of four representative structures of SERT and DAT, as presented in Fig. 5, Fig. 6. First, we verified that the residue fluctuation profile predicted by the GNM is consistent with that inferred from experiments (Fig. 5A) despite the fact that the experimental dataset does not necessarily represent a complete sampling of the conformational space. Further comparison of GNM results with those from a 3D-dependent elastic network model (ENM), and the anisotropic network model (ANM), supported the robustness of theoretical results (Fig. 5B).

Fig. 5.

Fluctuation profiles of residues from theory and experiments. Root-mean-square fluctuations (RMSFs) of residues are plotted against residue index in both panels. (A) Comparison of experimental RMSF profile, obtained from the PCA (PC1-20) of the ensemble of structurally resolved 50 hSERT and dDAT structures (blue curve); and RMSF profile, obtained from GNM analysis (GNM modes 1–20; orange curve) using a representative structure (hSERT in OFo state; PDB ID: 5I6X)). (B) Comparison of the fluctuation profile of residues in OFo hSERT predicted by the GNM (modes 1–20 – orange curve) and ANM (modes 1–20 – green curve). The horizontal bar along the abscissa describes the secondary structure (black represents the TMs; and orange, the IC or EC loops).

Fig. 6.

GNM analysis of hSERT, dDAT, and hDAT dynamics. (A) Distribution of residue movements along the most cooperative mode of motion predicted by the GNM (mode 1) for IF hSERT (PDB ID: 6DZZ), OFo hSERT (PDB ID: 5I6X), OFo dDAT (PDB ID: 4M48) and OFo hDAT (homology model from the previous study (Cheng et al., 2015)). Residues intersecting the x-axis (dotted lines, movement = 0), which serve as hinges and thus control the collective motions of the overall transporters, are labeled. Hinge residues experimentally or computationally observed to bind substrates or drugs are highlighted in green (see also Fig. S3). The eigenvalues represent the normalized distributions of movements along the normal coordinates, hence the absence of labels along the y-axis. (B) Color-coded ribbon diagrams where regions exhibiting maximal movements along in mode 1 are shown in red and blue (opposite directions), and those exhibiting minimal motions (acting as hinges/anchors) are in green (in line with the colored bar along the right ordinate in panel A). Hinge residues are shown in spheres and labeled. (C) Sequence alignment of selected hinge residues between hSERT, dDAT, and hDAT highlighting the conserved positions (shown in semi-transparent pink boxes). The complete alignment can be found in Fig. S4.

Next, we proceed to a more detailed analysis of the most cooperative movements accessible to these MATs and the identification of critical sites mediating them. The GNM permits us to dissect the overall conformational dynamics accessible under equilibrium conditions into a spectrum of normal modes. The lowest frequency mode (also called GNM mode 1) provides information on the relative size and direction of residue movements in the most cooperative mode of motion uniquely defined for each structure. The curves in Fig. 6A display the GNM-predicted residue displacements along the most cooperative mode axis for hSERT (IF and OF), dDAT (OF) and hDAT (OF) (from top to bottom), and the color-coded ribbon diagrams in panel B illustrate the structural elements undergoing positive (red) and negative (blue) displacement along the global mode axis. Residues serving as anchors or hinge centers during these movements (located at the crossover between positive and negative displacements) are labeled in panel A. By virtue of their critical position at the interface between oppositely moving substructures, those residues play a critical role in mediating the collective movements relevant to the transition of the transporters during the transport cycle.

The global hinge sites determined by the GNM (Bahar et al., 1997) have been pointed out in previous work to coordinate the binding of ligands or implicated in serving mechanical roles in protein functions (reviewed in (Bahar et al., 2015)). Consistent with prior work, several hinge-residues highlighted in green (and shown in space-filling representation in panel B) coincide with residues observed in experimentally resolved structures to coordinate the substrate or small molecules bound to sites S1 or S2. We note among them dDAT F43 and F319 highlighted in Fig. 3A to consistently coordinate the small molecule bound to S1, or hSERT D98 and P561 that consistently coordinate the ligands at S1 and S2, respectively (Fig. 3B). Likewise, among the residues that coordinate KM822 bound to S2-II in hDAT (Fig. 2A), several (D79, Y156, F320) are distinguished among the hinge sites in the global mode of hDAT (Fig. 6A, bottom).

Thus, the binding sites for small molecules (substrates or antidepressants) include hinge or anchor sites that effectively trigger or alter cooperative responses to promote or inhibit the intrinsically accessible motions of the transporter. Furthermore, the hinge sites either coincide with, or closely neighbor, the EC gate residues (Y176 in hSERT, Y124 and F319 in dDAT or Y156 in hDAT, all of whom also coordinate the substrate when bound to S1). Finally, certain hinge sites are sequentially close to the S1-site coordinating residues (e.g., hSERT L436 and F440 closely neighbor S438 and G442, respectively; dDAT D420 is adjacent to S421) suggesting that their proximity may ensure efficient communication between mechanically and chemically functional sites.

As expected, many hinge sites are conserved (highlighted in pink in Fig. 6C) among MAT members, consistent with their critical role in modulating cooperative movements relevant to function. They would not thus tolerate mutations (Haliloglu and Bahar, 2015), and/or if mutated, might impact significantly the protein dynamics. Interestingly, a quadruple DAT mutant, where four hinge site residues in the TM4 and TM9 were mutated to their SERT counterparts (I248F, V249T, L458F, and L459G), displayed altered dopamine transport kinetics and reduced effects of ligand-induced DAT oligomerization (Sorkina et al., 2021). The sequence information in Fig. 6C shows which hinge residues are conserved or varied among hSERT, hDAT and dDAT. The sequential heterogeneity at some of these sites provides valuable information for selecting target sites for potential drugs or allosteric modulators that can selectively affect the specific transporter.

7. Inter-residue cross-correlations analysis demonstrates that the allosteric sites communicate with the primary binding site

In order to evaluate the existence (and strength) of communication, if any, between the sites S1 and S2 such that binding of substrate/drug to S2 would affect S1, we computed the cross-correlations between the fluctuations of residues lining those sites. To this aim, we generated the cross-correlation map for hSERT containing a vilazodone molecule at its S2 site (PDB ID: 7LWD (Plenge et al., 2021)). The resulting heat map based on GNM modes 1–10 is presented in Fig. 7A. The map displays a detailed pattern of positively correlated (red, moving concertedly in the same direction), and anticorrelated (blue, moving concertedly in opposite direction) pairs of residues, as indicated by the color bar on the right. The orientational correlations vary in the range [−1, 1], as they represent the correlation cosines between the directions of motions, and both limits represent strong couplings, while uncorrelated pairs of residues yield cross-correlations close to zero.

Fig. 7.

Cross-correlations between residues lining the sites S1 and S2. (A) Residues cross-correlations computed from GNM analysis mode 1–10 are plotted as a heatmap, evaluated for a representative hSERT OFo structure with substrate bound to S1 and S2–I (PDB ID: 7LWD). Red and blue regions refer to pairs exhibiting the strongest correlations (positive and negative, respectively). Top panel shows the average cross-correlation between each residue and all other residues. Peaks highlight residues that are strongly coupled to others. (B) Cross-correlations between residue pairs that coordinate the substrates bound to S1 (ordinate), and S2–I, and S2-II (abscissa). Top and side curves show the average cross-correlations with the other pocket-lining residues for respective residues coordinating S2 (top panel) and S1 (side panel), respectively. Residue pairs with high average cross-correlations are more influential in establishing the allosteric coupling between S1 and S2.

To aid in the identification of couplings between S1 and S2 binding sites, critical residues that coordinate each binding site are extracted from the map and plotted against each other in the cross-correlation map presented in Fig. 7B. Dark blue and dark red regions in the map show the residue pairs that are subject to the strongest pairwise couplings. We note that residues F335 and G338 from S1 site exhibit very strong positive correlations with D328, A331, Q332 and F335 on site S2 (see the labels along the upper abscissa). Additionally, site S1 T497 is strongly positively correlated with E493, E494, Y495, P499, L502, and Y579; and S1 I172 is strongly coupled to S2 E493. The residues that are strongly coupled are also physically located in proximity, giving rise to different coupling in helices movement. These couplings, at a local scale, may directly contribute to the respective function at that region (Fig. S5). Hence, based on GNM analysis, TM6a (from residues D328 to F341) and TM10 (from residues E493 to L502) are crucial for the coupling between orthosteric and allosteric binding sites. Likewise, S1 residues A169, I172, and T497 exhibit strong anticorrelations with S2 R104, showing that in this case there is strong communication inducing opposite direction fluctuations between those residues.

When considering the average cross-correlation values contributed by each residue in relation to others, some striking points are observed. Average cross-correlations were calculated by calculating for each residue its average cross-correlation absolute value with all other residues (Fig. 7A top panel). These peaks therein point to residues that are most effective in transmitting allosteric signals, due to their high correlations to others. Zooming in on the substrate binding sites (Fig. 7B top panel for S2 and side panel for S1), we note that S2 residues E493, E494, Y495 are the most effective in transmitting allosteric signals. Whereas, in S1, T497 and F335 contribute the most. This is consistent with another study where the authors using MD simulations as well as mutagenesis experiments found that T497 and F335 played an important role in transmitting allosteric signals between S1 and S2 where the residue E494 could form a salt bridge with ligands bound (Plenge et al., 2020). Furthermore, other studies of hSERT allosteric binding have shown supportive findings where SERT ligands engaged in close interactions with T497, Q332, R104, F335, F556, D328, and E494 (Islas and Scior, 2022; Xue et al., 2022).

Overall, these results reveal the existence of strong cross-correlations between selected pairs of residues that line the S1 and S2 pockets. Among them, five (F335, E493, E494, Y495 and T497) are distinguished as those imparting the highest cross-correlation between S1 and S2, revealing their important role in enabling allosteric communication.

8. Conclusion

Significant efforts have been deployed to identify and characterize allosteric binding sites in hSERT and dDAT. The structural data on hSERT in the presence of substrate/ligand bound to a secondary site, pointed to the existence of a new site partitioned into two partially overlapping subsites S2–I and S2-II, where S2–I penetrates the gap between EL6, TM11 and TM10 (Cheng and Bahar, 2019; Coleman et al., 2016; Plenge et al., 2012). Computational data from MD simulations also suggested that such two subsites that bind substrates or ligands also exist in hDAT, as illustrated by the structural model generated for KM822 bound hDAT (Aggarwal et al., 2021). Even though experimental structural data were unavailable for dopamine transporter, the structural and dynamic similarities between SERT and DAT supported the idea that S2–I and S2-II sites also exist in dDAT.

The present work focused on these allosteric sites and their functional significance in the light of the involvement of these sites in the conformational dynamics and cross-correlations within the transporter. We presented the key residues that facilitate substrate-binding to SERT through categorizing the residues in proximity (4.5 Å) to the substrates in orthosteric and allosteric binding sites. These substrate-coordinating residues and their sequential counterpart among the MAT family members can serve as target sites for further studies of the allosteric behavior of MATs experimentally. We further carried out an ensemble analysis (PCA) of all experimentally resolved structures in that family, to provide a clear view of the landscape of conformations and transitions they are subject to. It is noteworthy that the structural variability deduced from the PCA of experimental data for 50 structures shows strong agreement with the conformational motions predicted by elastic network model (ENM) analysis of only one representative structure, underscoring the significance of ENM analysis, especially in the presence of sparse structural data, for assessing the space of conformers and their transitions. Particularly, the GNM analysis revealed the highly conserved hinge regions operating in SERT and DAT, which may serve as hot spots for ligand/drug binding, as alterations (or small molecule binding) to those mechanically critical sites can impact the global dynamics and thereby function of the transporter.

Finally, to gain a deeper understanding of the role of S2-site coordinating residues in potentially transmitting allosteric signals to the substrate binding site S1, we examined the cross-correlations between residues coordinating the ligands bound to these two sites, using a representative hSERT structure. The cross-correlations between the equilibrium fluctuations of these residues revealed that TM6A and TM10 helices were most involved in the coupling between S1 and S2 binding sites; and in particular, five residues, F335 and T497 in S1 as well as E493, E494, and Y495 in S2 were distinguished by their efficient communication/signaling role between the two sites. Future functional research on these reported residues may yield further insight into how allosteric modulation is effectuated and how we can engage such mechanisms to design allosteric modulators with therapeutic potential.

9. Methods

All analyses described in this section were performed using the ProDy package in Python (Zhang et al., 2021).

Ensemble analysis (or PCA) of the dataset of experimentally resolved DAT and SERT structures. The ensemble of structures was constructed from 50 hSERT and dDAT wildtype as well as mutant structures (with or without ligands bound), experimentally resolved and deposited in the PDB (Table S1). First, each of 49 hSERT and dDAT structures was sequentially or structurally aligned (when sequence alignment fails) using trivial alignment based on residue number or the combinatorial extension alignment (CE align) algorithm against the selected reference chain, the SERT OFo structure with (PDB ID: 5I6X). In each structure, the coordinates of α-carbons were used for structural alignments. The structural coordinates of the transporters in the ensemble were then iteratively superposed until convergence (using an ΔRMSD threshold of 0.0001 Å). The mean coordinate was calculated and set as the new reference coordinate in every iteration. The residue number was adopted from that of the first reference protein (PDB ID: 5I6X) and used as the standard for comparison between SERT and DAT. The sequence alignment can be found in Fig. S4.

Principal components analysis (PCA) to determine the principal modes of structural changes was performed similarly to the procedure previously described (Bakan and Bahar, 2009). A covariance matrix (3N × 3N, where N is number of residues) was constructed for the ensemble and decomposed into PCs. Coordinates of different structures were then projected onto the conformational subspace spanned by PC1 and PC2 (Fig. 4A).

The square fluctuations of each residue along the respective PC as well as the sum of the square fluctuations along multiple PCs are calculated from the respective projections of the deformations onto the PCs. The summation over all PCs represents the mean-square fluctuations (MSFs) of SERT and DAT based on experimental data. The accuracy of the MSF profile (as a function of residue index) may be limited by the number of structures in the ensemble as well as their distribution in the conformational space.

Elastic Network Model (ENM) predictions. GNM and ANM analyses (Rader et al., 2005) were performed by constructing Kirchhoff matrix (N x N) and Hessian matrix (3N × 3N), respectively, for each SERT or dDAT structure or hDAT homology model (Cheng et al., 2015) based on α-carbons coordinates. Similarly to PCA, these matrices were decomposed into ANM/GNM normal modes and the corresponding residue fluctuations were calculated and compared to each other (Fig. 5). GNM analysis was also used to predict the hinge sites by plotting the individual mode shapes (N-dimensional eigenvectors of the Kirchhoff matrix) associated with each mode as a function of residue number and locating the crossovers between positive and negative displacements (Fig. 6). The cross-correlation between residues were also evaluated by calculating the GNM covariance matrix based on 10 modes (Fig. 7).

CRediT authorship contribution statement

Hoang Nguyen: and. Mary Hongying Cheng: and. Ji Young Lee: conducted the computational experiments and all authors contributed to the analysis and interpretation of the data. All authors contributed to writing and editing, All authors approved the final version of the manuscript. Ole Valente Mortensen: conceptualized the content. Ivet Bahar: conceptualized the content.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Ole Valente Mortensen has patent #9616065 issued to Assignee. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.