Composition of buffered solution (ND96) (in mM): NaCl 96, KCl 2, CaCl₂ 1.8, MgCl₂ 1, HEPES 5, pH 7.6. Composition of NDE solution: ND96 plus 2.5 mM pyruvate and 50 μg/ml Gentamycin sulphate. Composition of external control buffer (ND98) (in mM): NaCl 98, MgCl₂ 1, and CaCl₂ 1.8 with or without 0.01% DMSO. In tetramethylammonium (TMA)-chloride “zero sodium” buffer (TMA98), equimolar TMACl replaces NaCl. In Li⁺ buffer, equimolar LiCl replaces NaCl. The final pH was adjusted using respective hydroxides (NaOH or TMAOH or LiOH) to 7.6 for all external solutions. Substrates used were Dopamine (DA) (Calbiochem - Sigma, Milan, Italy), Lithocholic acid (LCA) (Sigma), and Obeticholic acid (OCA) (Adipogen, Switzerland). LCA or OCA powder was dissolved in DMSO at 50 mM and 100 mM, respectively.

Oocytes were obtained from adult Xenopus laevis females. Animals were anaesthetised in 0.1% (w/v) MS222 (tricaine methanesulfonate; Sigma) solution in tap water. Abdomens were sterilized with antiseptic agent (Povidone-iodine 0.8%), laparotomy was performed, and portions of the ovary were collected. The oocytes were treated with 0.5 mg/ml collagenase (Sigma Type IA) in ND96 calcium-free for at least 30 min at 18°C. Healthy and fully-grown oocytes were selected and stored at 18°C in NDE solution (Bossi et al., 2007). The day after the removal, the oocytes were injected with cRNA using a manual microinjection system (Drummond Scientific Company, Broomall, PA). Injected concentrations were 12.5 ng/50 nl for the mouse dopamine transporter (mDAT), 2 ng/50 nl for human Takeda G protein-coupled receptor (hTGR5).

mDAT cDNA in pcDNA3.1 was kindly gifted from Prof. Dr. Harald Sitte of Medical University of Vienna. The cDNA mDAT was amplified with forward and reverse primers containing SmaI and EcoRI restriction sites, respectively (5′-GAC​TCCC​GGGACC​CAT​GAG​TAA​AAG​CAA​ATG-3′; 5′-GCA​TGAA​TTCTTA​CAG​CAA​CAG​CCA​ATG​GCG​C-3′). The amplified coding sequence was then subcloned into the pGHJ vector after double digestion with SmaI e EcoRI restriction enzymes (Promega). hTGR5 gene was in pCMV6-Entry (GPBAR1 Human cDNA ORF Clone, NM_001077191; Origene Technologies, Inc., Rockville, MD, United States). The two plasmids were linearized with SalI (mDAT) and with NdeI (hTGR5), in vitro capped, and transcribed using T7 RNA polymerase. Enzymes were supplied by Promega Italia. The oocytes were incubated at 18°C for 2–3 days prior to electrophysiological experiments. The experimental protocol was approved locally by the Committee of the “Organismo Preposto al Benessere degli Animali” of the University of Insubria (OPBA-permit #02_15) and nationally by Ministero della Salute (permit nr. 1011/2015).

Electrophysiological studies were performed using the Two-Electrode Voltage Clamp (TEVC) technique (Oocyte Clamp OC-725; Warner Instruments, Hamden, CT, United States). Controlling software was WinWCP version 4.4.6 (J. Dempster, University of Strathclyde, United Kingdom) or Clampex (Molecular Devices, Sunnyvale, CA, United States, www.moleculardevices.com). Borosilicate microelectrodes, with a tip resistance of 0.5–4 MΩ, were filled with 3 M KCl. Bath electrodes were connected to the experimental oocyte chamber via agar bridges (3% agar in 3 M KCl). The holding potential was kept at −60 mV for all the experiments. The minimal concentration of OCA inducing inward current was evaluated by applying concentration of OCA from 1 pM to 100 µM. Each dose was tested on one single oocyte, exposed before to 30 µM DA. The current in the presence of OCA is the maximal current measured at the peak. For the I/V relationships, a step protocol from −160 mV to +40 mV with 20 mV of increment was applied for 100 ms. The mean of the transport-associated currents plotted in the scatter diagrams and in the I/V relationships were determined by subtracting the current recorded in the ND98 buffer from the current recorded in the presence of DA or OCA. In experiments using TMA, subtraction was performed with the current in TMA98 buffer. To chelate intracellular calcium, oocytes were injected with 50 nl of an intracellular solution containing 13 mM EGTA 30 min prior to electrophysiological recording. Intracellular solution had the following composition (in mM): KCl 130, NaCl 4, MgCl₂ 1.6, HEPES 10, glucose 5, pH 7.6. The dose-response experiments in Figure 4 were performed exposing the same oocytes first to all the concentrations of DA alone, then to the same concentrations plus 10 µM OCA. Data analysis was performed using Clampfit 10.2 software (Molecular Devices, Sunnyvale, CA, United States, www.moleculardevices.com); OriginPro 8.0 (OriginLab Corp., Northampton, MA, United States, www.originlab.com) and GraphPad Prism (www.graphpad.com/scientific-software/prism) were used for statistical analysis and figure preparation.

The mDAT (residues R56 to N596; UniProt ID O55192) in the outward-facing open (OFo), occluded, and inward-facing open (IFo) states were generated using SWISS-MODEL (Waterhouse et al., 2018) based on the structures resolved for OFo Drosophila DAT (dDAT) (PDB: 4M48) (Penmatsa et al., 2013), occluded hSERT (PDB: 6DZV) and IFo hSERT (PDB: 6DZZ) (Coleman et al., 2019). Homology models for hDAT in these conformational states were taken from previous work (Cheng et al., 2018; Aggarwal et al., 2021).

OCA 3D structures were downloaded from the ZINC database (Irwin and Shoichet, 2005) (ZINC14164617) and DrugBank (Wishart et al., 2006) (DB05990). The net charge of OCA is indicated to be −1 in the ZINC database, and zero in DrugBank. Docking simulations were performed using both electronic states, designated as OCA(−) and OCA(n) for the negatively-charged and neutral OCAs, respectively. The binding sites and binding poses on both mDAT and hDAT, in different conformational states, were assessed using docking simulation software AutoDock4 (Morris et al., 2009) and AutoDock Vina (Trott and Olson, 2010). Autodock Vina simulations were carried out using a grid with dimensions set to 84 × 58 × 86 Å³ and center at (−1.22, 1.16, −6.26 Å) for each conformer and each transporter. This grid box encapsulated the entire structure of the transporter. The number of runs (exhaustiveness parameter in Autodock Vina) was set to 50 and the algorithm returned 20 binding modes of interest for each conformer. AutoDock4 simulations were performed following previously published protocols (Cheng et al., 2015; Aggarwal et al., 2021). Briefly, Lamarckian genetic algorithm with default parameters was employed with the maximal number of energy evaluations set to 2.5 × 10⁷. The binding energy was estimated from the weighted average of multiple binding poses at a given site observed in 100 independent runs.

To begin to investigate the regulatory effects of OCA on DAT electrical activity, Xenopus laevis oocytes were tested by two-electrode voltage clamp (TEVC) (Figure 1A). Perfusion of DA onto the oocytes expressing the mouse DAT (mDAT) generated an inward current (−12.5 nA ± 0.63) (Figures 1B,C, left), confirming the expression and functionality of the transporter. On the same oocytes, following DA washout, perfusion of 10 µM OCA also generated an inward transport current (−8.28 nA ± 0.6). This membrane conductance exhibited a lower amplitude and inactivated rapidly (Figures 1B,C, left). The OCA-induced transient inward current was also elicited when OCA was perfused prior to DA exposure (data not shown).

Next, we set out to understand whether the OCA current was mediated by endogenous targets of OCA. One of the main targets of OCA, the G-Protein Coupled Receptor TGR5, is shown to be only minimally expressed in Xenopus laevis oocytes as reported in the platform Xenbase (Yanai et al., 2011; Session et al., 2016; James-Zorn et al., 2018). As shown in Figure 1B, no response from the perfusion of OCA was obtained in control oocytes (i.e., not expressing any heterologous protein). Thus, an activation of endogenous targets seems unlikely. As previously reported in literature, oocytes can express functional human TGR5 (hTGR5) (Lieu et al., 2014) as demonstrated by modulation of various membrane proteins upon TGR5 activation. We expressed hTGR5 in oocytes and measured OCA-induced currents. No currents were observed in hTGR5 expressing oocytes upon OCA perfusion (Figure 1B, right and Figure 1D). Next, we tested oocytes co-expressing mDAT and hTGR5. The currents recorded from these cells in response to DA (−12.88 nA ± 0.9) and OCA (−9.37 nA ± 0.82) alone (Figure 1B, center and Figure 1C, right) were comparable to the ones recorded in oocytes expressing mDAT alone (Figure 1D). This data suggests that the inward current induced by OCA is directly mediated by mDAT.

Dose-response experiments determined that 10 µM OCA was a saturating concentration. The EC50 was not calculated for this interaction because the current amplitude was consistent in response to over an eight order of magnitude range of OCA concentrations, as low as 1 pM and as high as 100 µM OCA (Figure 1E). The concentration of 10 µM was chosen throughout this study because it approximates the concentration of OCA in the plasma after oral administration and is sufficient to activate physiological targets of OCA in vitro (Roda et al., 2014).

To better characterize the OCA-induced current, we performed a voltage step protocol and obtained I/V curves for the transporter with DA and OCA (Figure 1F). As expected, DA-induced currents increased at negative potential without reverting. At positive voltages, the I/V curve for DA shows the distinctive inverted U shape, previously reported by Sonders and coworkers (Sonders et al., 1997), which is a hallmark of blockage of the tonic Na⁺-leak current. OCA displays a similar curve shape. However, it shows an increase in conductance at negative potentials and a lack of inversion at positive potentials (Figure 1F). These characteristics point towards OCA opening a conductance in mDAT.

Several members of the solute carrier 6 (SLC6) family of transporters exhibit basal leak currents (Lester et al., 1994; Bossi et al., 1999; Andrini et al., 2008) that are augmented in the presence of Li⁺ ions. The Li⁺-leak current can be utilized to highlight the effect of molecules that interact with transporters and modify their electrical activity in the absence of Na⁺. DAT shows a significant Li⁺-leak current that is partially blocked upon DA perfusion (Giros et al., 1992; Sonders et al., 1997). Therefore, we used Li⁺ to investigate whether OCA, similarly to DA, binds mDAT in the absence of Na⁺. Oocytes were perfused first either with DA or OCA in Na⁺ bathing buffer (Figures 2A,B). Switching to Li⁺ bathing buffer induced a large inward current (−92.41 nA ± 7.45) (Figures 2A,B). As expected, the addition of DA to the Li⁺ buffer partially blocked the Li⁺-leak current (Figures 2A,B). This inhibition occurred in two phases; a rapid transient component (current at the peak: −25.72 nA ± 4.07) followed by a steady-state condition (−38.97 nA ± 5.11). After DA removal, the Li⁺-leak current returned to initial values (Figure 2A). Interestingly, perfusion of OCA also partially blocked the Li⁺-leak current (Figures 2A,B). As with DA, this inhibition displayed two phases; a transient rapid component (current at the peak: −66.86 nA ± 6.13) and a steady-state component (−82.42 nA ± 7.69). Inhibition of the Li⁺-leak current is a strong indication of direct binding of OCA to the transporter in the absence of Na⁺.

In addition to a coupled mechanism, Na⁺ also permeates through DAT in the absence of DA. This generates a leak current that can be uncovered when Na⁺ is substituted by non-permeant cations such as choline or TMA⁺ (Sonders et al., 1997). To better understand the effect of OCA on membrane conductance and the relevance of Na⁺ in the OCA-induced inward current, experiments were repeated with TMA⁺ as the cation substituting for Na⁺ in the bathing buffer. TMA⁺ blocked the Na⁺-leak current (15.12 nA ± 1.71) (Figures 2C,D). As expected, in the presence of TMA⁺, DA elicited only a fast-transient inward current (−1.98 nA ± 0.37), confirming that Na⁺ is necessary for mDAT-mediated DA currents. OCA behaved similarly to DA: the inward current was still present but greatly reduced in amplitude (−1.83 nA ± 0.38) (Figures 2C,D). The substitution of TMA⁺ for Na⁺ greatly affected the amplitude of the OCA-induced currents, pointing to a major role for the permeation of Na⁺ ions in causing the OCA-induced currents recorded in physiological solution.

OCA has been shown to induce IC Ca²⁺ fluctuations (Hao et al., 2017). Thus, it is possible that the OCA-induced inward current could be generated by the activation of chloride conductance due to an increase in IC Ca²⁺ concentrations. To investigate this possibility, experiments were conducted in oocytes expressing mDAT and injected with the Ca²⁺-chelating agent EGTA (Figure 3A). The presence of EGTA did not alter mDAT-mediated currents elicited by DA or OCA (Compare Figure 3B with Figures 1B,C). Maximal OCA-induced inward currents were also not altered by the presence or absence of EGTA (Figure 3C). Together, these data strongly suggest that IC Ca²⁺ does not regulate OCA-induced currents and further suggest that effects of OCA are mediated by direct interaction with mDAT.

Data thus far indicate that OCA interacts directly with mDAT in the absence of DA. We next investigated whether OCA regulates DAT-mediated DA-induced currents. Currents generated from increasing concentrations of DA were unaltered by the presence of 10 µM OCA even at low DA concentration (Figures 4A,B). Mean of maximal currents were fitted to a Hill equation (Figure 4C) and the results reported in Figure 4D. These data indicated that the presence of OCA does not significantly affect either the affinity or the maximal transport currents.

To determine whether the OCA-induced mDAT-mediated current is specific to OCA, or is a common phenotype induced by bile acids, we investigated the effect of the natural bile acid, lithocholic acid (LCA). OCA and LCA share the same sterol-based structure with different R groups at positions 5 and 6 of the B ring (Figure 5A). Specifically, ethyl and hydroxyl groups present in OCA are substituted by hydrogen in LCA. Similar to OCA, perfusion of 10 µM LCA onto mDAT-expressing oocytes induced a transient inward current (−10.25 nA ± 2.06) (Figures 5B,C). Direct comparison of OCA and LCA revealed no significant differences in their ability to induce DAT-mediated currents (Figure 5C). These data suggest that the ability of OCA to promote DAT-mediated currents is shared by another BA.

To identify potential sites for binding of OCA onto mDAT and hDAT, docking simulations were performed, using AutoDock4 (Morris et al., 2009) and AutoDock Vina (Trott and Olson, 2010). To account for alternative protonated and deprotonated states of OCA, both the neutral (OCA(n)) and negatively-charged (OCA(−)) forms (Figure 6D) were used in the simulations. Computations were performed for the outward-facing open (OFo), occluded (OCC), and inward-facing open (IFo) states of both mDAT and hDAT. The computations revealed that OCA selected similar binding poses, with comparable binding affinities, for either transporter when the same conformational state was targeted regardless of the species. Therefore, representative results for hDAT are presented unless otherwise stated.

Figure 6A illustrates the top-ranking distinct binding poses and sites of both OCA forms, observed in docking simulations, onto hDAT in the OFo and IFo states in the absence of DA. Panel B shows the comparison to the DA-bound structures resolved for dDAT (Wang et al., 2015) in the presence of OCA, and panel C to ibogaine-bound hSERT (Coleman et al., 2019). Notably, in the case of the OFo conformer, the most favorable binding sites for both the neutral and negatively charged forms of OCA are within the extracellular (EC) vestibule but the exact locations are determined by the protonation state of the ligand. In the absence of DA, OCA(−) occupies the S2 site proposed to allosterically modulate transport (Cheng and Bahar, 2019), whereas OCA(n) binds an extended region spanning between the primary (S1) and secondary (S2) sites (Figure 6A, top panels). The binding pose of OCA(−) thus differs from that of the DA bound to OFo DAT (Figure 6B) where the residue equivalent to hDAT D79 (D46 in dDAT) plays a major role in coordinating the binding of the amine group through a salt bridge formation (Penmatsa et al., 2013; Wang et al., 2015; Cheng and Bahar, 2019). Notably, when DA is present, the top binding poses of OCA do not block the binding of substrate DA (Figure 6B, left panel), such that DA is able to bind in the proximity of OCA. This is consistent with the non-competitive binding of OCA revealed in TEVC experiments (Figure 4).

For the IFo conformer, no high affinity binding is observed to the EC vestibule. For OCA(n), the top binding site is in proximity to D421 within the IC vestibule (Figure 6A, bottom right panel). OCA(−) preferentially binds near the IC entrance, in the proximity to R445, and no high affinity binding is observed within the vestibule. Notably, docking simulations also revealed binding instances within the transmembrane region, including the site known to be occupied by cholesterol near TM1a (Penmatsa et al., 2013; Cheng and Bahar, 2019), or a site close to residues R443/H444 (Figure 6A, middle left panel) reported previously as a PIP₂ binding site (Belovich et al., 2019). However, these sites could potentially be obstructed by lipid molecules, which are not included in docking simulations, and hence will not be further elaborated.

Unexpectedly, in the case of the occluded DAT conformer (Figure 6C), both charged and neutral OCAs (OCA(n/−)) are predicted to bind to the S1 site, with almost identical affinity, and the binding pose closely resembles that resolved for ibogaine-bound hSERT. Of note, the binding pocket in the occluded state may be much larger than anticipated, with possible involvement of TM9 (Figure 6C, right panel).

Taken together, these docking simulations suggest that the binding sites and affinities depend on the net charge carried by OCA and on the conformational state of the transporter. In the OFo state, the binding site for OCA(n) closely neighbors, but does not overlap with, the S1 site resolved for DA (Wang et al., 2015); whereas, in the absence of DA, the binding site for OCA(−) overlaps with the broadly-defined S2 site (Figure 6A). In the IFo state, on the other hand, OCA binds near D421 or R445, depending on its protonation state. Notably, in the occluded state of the transporter, OCA appears to select a binding site similar to that resolved for ibogaine-bound to hSERT (Coleman et al., 2019) regardless of its charge. However, caution is needed when interpreting docking-predicted binding sites, especially in the occluded state. The S1 site suggested in the occluded DAT conformer may not be accessible to OCA(−) in practice, even though it forms the pocket with the highest affinity, due to the fact that OCA(−) does not bind to the S1 site in the OFo conformer.