Gating and anion selectivity are reciprocally regulated in TMEM16A (ANO1)

A variety of essential physiological functions, including blood pressure, electrolyte and hormone secretion, gastrointestinal rhythmicity, pain, neuronal and muscle excitability, and olfactory transduction, depend on the activation of Ca²⁺-dependent chloride (Cl⁻) channels TMEM16A and TMEM16B (Hartzell et al., 2005; Duran et al., 2010; Berg et al., 2012; Pedemonte and Galietta, 2014). Both channels are homodimer proteins harboring an independent anion permeation pathway in each monomer (Paulino et al., 2017; Dang et al., 2017; Sheridan et al., 2011). TMEM16A and TMEM16B are gated by the direct, voltage-dependent binding of intracellular Ca²⁺ ions to high-affinity binding sites (Arreola et al., 1996; Xiao et al., 2011; Contreras-Vite et al., 2016; Qu and Hartzell, 2000; Terashima et al., 2013). Three Ca²⁺ binding sites (sites 1–3) have been projected in each subunit (Yu et al., 2012; Tien et al., 2014; Le and Yang, 2020). Among them, site 1 is the best characterized both structurally and functionally. It is formed by four glutamates from transmembrane helices (TM) 6, 7, and 8, one aspartate from TM8, and one asparagine from TM6, which together coordinate the binding of two Ca²⁺ ions (Paulino et al., 2017; Dang et al., 2017).

Fig. 1, based on the Ca²⁺-bound structure of TMEM16A PDB accession no. 5OYB (Paulino et al., 2017), shows the TM3–TM8 (gray ribbons) surrounding the pore (teal mesh) in a single subunit. Site 1 locates toward the cytoplasmic side, near a wide water-filled cavity accessible to the cytoplasm where it faces the permeation pathway. Site 2 has a low affinity for Ca²⁺ (within the micromolar range) and is occupied by this cation in a voltage-independent manner (Jeng et al., 2016; Lim et al., 2016). A third Ca²⁺ binding site is formed by one glutamate and one lysine on the TM2 and two aspartates on the TM10 (Le and Yang, 2020). Interestingly, it has been suggested that site 3 can allosterically regulate site 1 (Le and Yang, 2020).

The pore is formed by two large vestibules located at the entrances connected by a narrow neck, thinner than the hydrated Cl⁻ ion size (Fig. 1). At the narrow neck region, an inner hydrophobic gate formed by a triad of isoleucine residues (I641, I550, and I551 in TMEM16A ac) maintains the pore closed (Lam et al., 2021). In addition, the apo Ca²⁺ pocket of site 1 generates a strong negative electrostatic potential known as the electrostatic gate that hampers anion conductance (Lam and Dutzler, 2018). When the intracellular Ca²⁺ increases, the site 1 electrostatic potential and depolarizing voltages assist Ca²⁺ binding to site 1 (Fig. 1). The immediate consequence of Ca²⁺ binding is the plummeting of the electrostatic barrier, also referred to as “electrostatic gate opening.” A subsequent effect of Ca²⁺ binding is the hinge movement of the cytosolic half of TM6 around glycine 644 (G644). The move of TM6 marks the opening of the steric hydrophobic gate. Thus, the concerted action of electrostatic and steric gating triggered by Ca²⁺ binding is required to open the pore and enable Cl⁻ permeation (Lam and Dutzler, 2018; Paulino et al., 2017). Additionally, upon disruption of the hydrophobic gate or the hinge movement of TM6 by mutation, the voltage can induce Cl⁻ permeation without intracellular Ca²⁺ (Lam et al., 2021).

The juxtaposition of the anion permeation pathway and site 1 would induce an interaction of the permeant anion with the gating elements that may influence Ca²⁺ binding, a critical step for TMEM16A activation. In agreement, early observations have shown that anions more permeable than Cl⁻ accelerate the onset and slow the offset of the Ca²⁺-activated Cl⁻ currents in rat parotid acinar cells, thus affecting the voltage dependence of endogenous TMEM16A (Perez-Cornejo et al., 2004; Xiao et al., 2011). In TMEM16B, such anions aided the voltage dependency and increased the apparent Ca²⁺ sensitivity (Betto et al., 2014). Furthermore, applying thiocyanate to the cytoplasmic side of membrane patches excised from Xenopus laevis oocytes increased the apparent Ca²⁺ sensitivity of the endogenous TMEM16A (Wozniak et al., 2018; Qu and Hartzell, 2000). Thus, permeant anions interact with TMEM16A gating. However, during molecular interactions, the charge distribution of each participant is affected by the presence of others (Murray and Politzer, 2017). Therefore, the gating elements may be expected to affect anion permeation. In agreement, structural studies of TMEM16A revealed that Cl⁻ ions pass in front of site 1 during permeation, and site 1 alters the channel’s anion conductivity by generating a negative electrostatic potential (Fig. 1; Lam and Dutzler, 2018). Together, these observations suggested that permeant anions and the gating machinery of TMEM16 interact reciprocally. This idea implies that permeant anions affect Ca²⁺ and voltage sensitivities, key gating properties, whereas the gating elements alter permeability and anion selectivity.

Here, we used inside-out patches and whole-cell recordings to determine the effects of permeant anions on the apparent Ca²⁺ sensitivity and voltage dependence of WT TMEM16A. In addition, we applied site-directed mutagenesis to investigate whether the gating elements (site 1, TM6, and the steric hydrophobic gate) regulate anion permeability and selectivity. Our results indicate that the Ca²⁺ sensitivity is influenced by the presence of anions in the pore; conversely, the gating elements control anion selectivity. Furthermore, channels lost their anion selectivity and sensitivity to thiocyanate mole fractions (MFs) after we mutated key residues at several gating elements. We suggest that the activation and anion selectivity of TMEM16A depends on a reciprocal regulation of permeation and gating.

We used human embryonic kidney 293 (HEK 293) cells to express WT and mutant channels. HEK 293 cells were cultured in DMEM (Gibco) supplemented with 10% FBS and 0.1% gentamicin at 37°C in a 95% O₂/5% CO₂ atmosphere. WT mouse TMEM16A (ac) and mutant DNAs cloned in the pIRESII-EGFP vector (Clontech) were transfected (1 µg/µl cDNA) using the Polyfect transfection reagent (Qiagen), according to the manufacturer’s instructions. Successfully transfected cells were used 12 h after transfection. Cells were seeded on coverslips at low density for whole-cell recording. Cells stably transfected with TMEM16A or transiently transfected with mutants were plated onto poly-L-lysine–coated coverslips for inside-out recordings. On the day of the experiment, we submerged one coverslip in a recording chamber flooded with the standard external solution and placed it on the stage of an inverted microscope (Nikon).

We introduced site-directed mutations using the Quick-change kit (Agilent) and verified them by sequencing.

We recorded macroscopic currents at room temperature (21–23°C) from whole cells or inside-out patches using the patch-clamp technique as we previously reported (Cruz-Rangel et al., 2017). Cells expressing the channel of interest were selected by their GFP fluorescence using an inverted microscope (Nikon) equipped with UV illumination. We fabricated borosilicate patch pipettes using the P-97 electrode puller (Sutter Instruments). When filled with the standard pipette solution, the pipette resistances were 3–5 MΩ for the whole cell or 1–2 MΩ for inside-out patches. Currents from inside-out patches held at +80 mV were induced by rapid application to the cytosolic side of solutions containing increasing [Ca²⁺]. The stimulation protocol for whole-cell recordings consisted of voltage steps from −100 to +200 mV delivered every 7 s from a holding potential of −30 mV, followed by a repolarization potential to −100 mV. Alternatively, we recorded whole-cell currents using a ramp protocol that varied the membrane potential between −80 and +80 mV. Data were acquired using an Axon Digidata 1550, an Axopatch 200B amplifier, and the pClamp10 software (Molecular Devices). Currents were filtered at 5 kHz and digitized at 10 kHz. For inside-out recordings using patch pipettes filled with solutions containing foreign anions, we embedded the Ag/AgCl recording electrode in a 1-M KCl agar guard. Recording chamber solutions were applied using a homemade gravity perfusion system and grounded using a 3-M KCl agar-bridge connected to an Ag/AgCl reference electrode.

We obtained all the recordings using extracellular solutions made hypertonic with D-mannitol, with their pH adjusted to 7.3 to inhibit the activation of endogenous volume-sensitive Cl⁻ and proton-activated channels (Hernández-Carballo et al., 2010; Ullrich et al., 2019). The standard extracellular solution contained (in mM) 139 TEA-Cl, 20 HEPES, 0.5 CaCl₂, and 110 D-mannitol. We adjusted the pH to 7.3 with TEAOH or NaOH. The solution’s tonicity was 380–400 mOsm/kg measured by the vapor pressure point method with a VAPRO (Wescor). We replaced the Cl⁻ content in the standard extracellular solution with the desired foreign anion (SCN⁻, I⁻, NO₃⁻, or Br⁻) for anion selectivity sequence determination. External solutions with varying SCN⁻ MF were prepared by mixing the appropriate volumes of solutions containing 100% Cl⁻ and 100% SCN⁻. The total anion concentration of the external solutions was 140 mM. The intracellular solutions with different [Ca²⁺] contained (in mM) 35.1 TEA-Cl, 25 EGTA-TEA, and 2.43 CaCl₂, for [Ca²⁺]_i = 0.1 µM; 30 TEA-Cl, 25 EGTA-TEA, and 5.24 CaCl₂ for [Ca²⁺]_i = 0.2 µM; 30 TEA-Cl, 25.24 EGTA-TEA, and 5.24 CaCl₂ for [Ca²⁺]_i = 0.25 µM; 22.48 TEA-Cl, 25 EGTA-TEA, and 8.76 CaCl₂ for [Ca²⁺]_i = 0.5 µM; 27.4 TEA-Cl, 25 HEDTA-TEA, and 6.3 CaCl₂ for [Ca²⁺]_i = 1.3 µM; 11.8 TEA-Cl, 25 HEDTA-TEA, and 14.1 CaCl₂ for [Ca²⁺]_i = 5 µM; and 2 TEA-Cl, 25 HEDTA-TEA, and 19 CaCl₂ for [Ca²⁺]_i = 12 µM. In addition, these solutions contained 50 mM HEPES and 85 mM D-mannitol. We adjusted the pH to 7.3 with TEA-OH and the tonicity to 290–300 mOsm/kg. The free [Ca²⁺] was calculated using Maxchelator (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/). The standard intracellular solution contained [Ca²⁺]_i = 0.2 µM. The [Cl⁻] in these solutions was 40 mM. All chemicals were purchased from Sigma-Aldrich or Merck.

The apo (PDB accession no. 5OYG) and holo (PDB accession no. 5OYB) structures of mouse TMEM16A isoform ac (Paulino et al., 2017) were embedded in a POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) membrane and solvated with 140 mM NaCl. These ensembles were then equilibrated for 30 ns using Gromacs 5.1. I-tasser (https://zhanglab.ccmb.med.umich.edu/I-TASSER/) was used to model the protein gaps. By solving the linearized Poisson–Boltzmann equation in CHARMM-GUI PBED solver, the average electrostatic potential along the pore was calculated on a 255 × 197 × 273-ų (0.5-Å grid spacing), with CHARMM36 forcefield. We generated the in silico mutants with CHARMM-GUI tools. The pore in the structure was located using CAVER 3.0 (https://caver.cz/index.php?sid=100) and visualized with PyMol. We calculated the average electrostatic potential on the transversal area determined by the pore radius. For this calculation, the thickness of the membrane was 35 Å. The dielectric constants were set to 2 for the protein, 2 for the membrane, 30 for the membrane headgroups, and 80 for solvent containing 140 mM salt concentration (Jo et al., 2008a; Im et al., 1998; Jo et al., 2008b; Lam and Dutzler, 2018).

Pooled data are presented as mean ± SEM of n (number of independent experiments). We performed statistical analysis of mean values using OriginLab. Significant differences between means were determined using Student’s t test or one-way ANOVA followed by Tukey’s correction with P = 0.01 (indicated by ** in the figures).

Fig. S1 A shows the magnitude of the background currents recorded from HEK 293 cells transfected with an empty vector. Cells were bathed in Cl⁻- (black) and SCN⁻-containing (blue) solutions. Fig. S1 B shows that 0.5 µM Ani9, a TMEM16A blocker, inhibited the ion currents through the 5Q mutant channel. Fig. S2 shows the relationship between P_X/P_Cl ratios and anion hydration energies in different channels. In the WT pore, the P_X/P_Cl ratios increase linearly (olive symbols and line) as the hydration energy decreases, a relationship that is lost in the mutant channels.

To establish whether foreign anions regulate the apparent Ca²⁺ sensitivity of WT TMEM16A, we determined the Ca²⁺ EC₅₀ using recordings from inside-out patches held at +80 mV as previously reported in TMEM16B (Betto et al., 2014). First, we excised patches using a pipette filled with a solution containing Cl⁻, SCN⁻, F⁻, or NO₃⁻, and exposed the cytosolic side to increasing [Ca²⁺]_i. Fig. 2 A shows examples of the macroscopic currents induced by the indicated [Ca²⁺]_i while the extracellular side of the membrane patch was exposed to Cl⁻ (black), SCN⁻ (blue), F⁻ (green), or NO₃⁻ (red). As the [Ca²⁺]_i increased, the magnitude of the current flowing through WT channels increased until it saturated at 5–12 µM regardless of the external anion. Fig. 2 B shows the resulting dose–response curves to [Ca²⁺]_i; the curves for SCN⁻ (blue) and NO₃⁻ (red) are shifted toward lower [Ca²⁺] values compared with the curves for Cl⁻ (black) and F⁻ (green) that are on higher [Ca²⁺] values. The curves were analyzed using Eq. 1 (continuous lines in Fig. 2 B) to obtain the EC₅₀ and n_H coefficient values. Next, we plotted these parameters in Fig. 2 C as a function of the permeability ratio for the tested anion. As the permeability ratio increased, the EC₅₀ value (black) decreased from 0.75 µM in the presence of F⁻ to 0.3 µM in SCN⁻, whereas the Hill coefficient stayed 2.5–3 (light blue).

As more than one permeant anion can simultaneously be present in the TMEM16A pore, the zero-current potential and the anion conductance can display anomalous mole fraction (AMF) behavior (Qu and Hartzell, 2000; Xiao et al., 2011). We hypothesized that more than one permeating anion could influence the site 1 Ca²⁺ affinity during the permeation process. If so, the Ca²⁺ affinity would display AMF behavior because Ca²⁺ ligation also occurs within the pore. We tested this idea by determining the Ca²⁺ EC₅₀ at different SCN⁻ MF on the extracellular side. We chose SCN⁻ because it is approximately fourfold more permeable than Cl⁻ and induced small currents at +80 mV in mock-transfected cells (Fig. S1 A, upper panel). Fig. 3 A shows Ca²⁺-induced macroscopic currents recorded from four independent inside-out patches excised with a pipette containing 0.25 (olive), 0.5 (orange), 0.75 (green), or 1 (blue) SCN⁻ MF. The [Ca²⁺]_i used to activate the currents are indicated at the top of the records. Fig. 3 B shows the resulting dose–response curves to Ca²⁺ under different external SCN⁻ MF. Compared with the 0 SCN⁻ MF condition, the curves shifted to lower [Ca²⁺] values as the SCN⁻ MF increased to 0.75. However, the curve with SCN⁻ MF of 1 was similar to that obtained with 0.5 MF. The EC₅₀ and n_H parameters were calculated by fitting the dose–response curves to the Hill equation (continuous lines in Fig. 3 B). The calculated parameters are plotted as a function of the SCN⁻ MF in Fig. 3 C (black). The EC₅₀ decreased as the SCN⁻ MF increased, reaching a minimum value when the SCN⁻ MF was 0.75. In contrast, the Hill coefficient remained constant at ∼2.5. Although at +80 mV, the TMEM16A EC₅₀ did not show a typical AMF behavior, at +100 mV (Fig. 3 C, gray symbols), we saw a minimum of ∼0.5 and 0.75, indicative of abnormal behavior of Ca²⁺ sensitivity.

Given that the rates of association and dissociation of Ca²⁺ determine the apparent Ca²⁺ sensitivity, a property increased by highly permeant anions, we expected to see changes in on- or off-rates of the Ca²⁺-induced currents in the presence of different anions or SCN⁻ MF. As we reported previously, it is possible to estimate these rates from inside-out patch recordings (Xiao et al., 2011). We calculated the apparent binding and unbinding rates of Ca²⁺ by fitting a single exponential function to the activation and deactivation segments of the TMEM16A current elicited by 5 µM Ca²⁺ applied to patches held at +80 mV. This [Ca²⁺] is ∼10-fold the Ca²⁺ EC₅₀ (Fig. 2 B); therefore, changes in the Ca²⁺ sensitivity would be readily apparent. Fig. 4 A shows examples of macroscopic currents recorded in the presence of the F⁻, Cl⁻, NO₃⁻, and SCN⁻ on the extracellular side. The red lines are the monoexponential fits to calculate τ_on and τ_off. With extracellular NO₃⁻ or SCN⁻, the off-rate was slowed by 2.4- and 2.9-fold, respectively, without altering the on-rate (Fig. 4 A, right side). Similar recordings are shown in Fig. 4 B. In this case, we exposed the extracellular side to SCN⁻ MF = 0.25, 0.5, and 0.75. Increasing the SCN⁻ MF slowed down the off-rate nearly twofold without changing the on-rate (Fig. 4 B, right side). We obtained the same trend with data collected with 1.3 µM Ca²⁺. On the right side, the average on and off time constant values are plotted as open circles in Fig. 4, A and B. These results further support the idea that permeant anions increase the apparent Ca²⁺ affinity of TMEM16A by slowing down the unbinding of Ca²⁺. Thus, these data show that permeant anions probably regulate the apparent Ca²⁺ sensitivity of TMEM16A by stabilizing bound Ca²⁺.

Provided that high [Ca²⁺]_i facilitates the voltage dependence of TMEM16A (Xiao et al., 2011) and the apparent Ca²⁺ affinity displayed AMF behavior with SCN⁻ MF, we anticipated alterations in the voltage dependence by the presence of anion MFs as previously demonstrated in CLC-0 and CLC-1 voltage-gated Cl⁻ channels (Pusch et al., 1995; Rychkov et al., 2001). We measured Cl⁻ currents from cells bathed in solutions containing different SCN⁻ MF to test this idea. Fig. 5 A shows an example of currents obtained in Cl⁻ (black) and SCN⁻ (red) recorded in the voltage range of −100 and +200 mV. The presence of SCN⁻ resulted in currents of greater magnitude and tail currents slower than those recorded in the presence of Cl⁻. The G/G_{max_app, 0.5 µM Ca}²⁺ calculated under different SCN⁻ MF was plotted as a function of voltage in Fig. 5 B. The curves seem to steadily shift to lower voltage values as the SCN⁻ MF increases. The total V_{0.5_app} shift was +83 mV (from +126 ± 7.8 mV with SCN⁻ MF = 0 to +43 ± 8 mV with SCN⁻ MF = 1). However, the V_{0.5_app} values show a nonlinear behavior (Fig. 5 C) with little or no change in the slope (Fig. 5 D). In addition, the reversal potential shifts plotted against the SCN⁻ MF (Fig. 5 E) cannot be explained by a constant P_SCN/P_Cl value (P_SCN/P_Cl = 4.7 in red and = 5.5 in blue). Instead, P_SCN/P_Cl varies as the SCN⁻ MF changes (Fig. 5 F). This behavior is compatible with AMF behavior, as previously reported (Qu and Hartzell, 2000; Xiao et al., 2011). Thus, in addition to increasing the apparent Ca²⁺ affinity, SCN⁻ MF applied extracellularly strongly facilitated the TMEM16A voltage dependence.

Because the permeant anions stabilized the open conformation by decreasing the off-rate of Ca²⁺-induced currents, which facilitates the gating process of CaCC (Perez-Cornejo et al., 2004), we anticipated that the gating machinery would reciprocally regulate anion selectivity. As described in the Introduction, the gating machinery of TMEM16A comprises electrostatic (acidic residues of site 1) and steric (TM6 and hydrophobic gate) elements. Therefore, we performed a mutagenesis analysis of these elements and determined the mutant channels’ anion permeability and selectivity sequence to test further the idea of reciprocal regulation between permeation and gating.

To infer a possible role of site 1 in anion selection, we produced mutant TMEM16A channels E702Q, E705Q, E702Q/E705Q, and E654Q/E702Q/E705Q/E734Q/D738Q (5Q), determined their anion selectivity using whole-cell recordings, and compared it to WT TMEM16A channels recorded in the presence of 0.2 µM Ca²⁺. These mutations decreased the Ca²⁺ affinity of site 1 (Tien et al., 2014; Yu et al., 2012); thus, we recorded them with 12 µM intracellular Ca²⁺. Fig. 6 A displays I-V relations obtained from cells dialyzed with 0.2 mM Ca²⁺ that express WT channels (upper left) or with 12 µM Ca²⁺ that express E702Q (upper right), E705Q (middle left), and E702Q/E705Q (middle right) mutant channels. We bathed cells in 140 mM Cl⁻, Br⁻, NO₃⁻, I⁻, or SCN⁻ solutions and determined the reversal potential from the I-V relations under each ionic condition. We observed that currents increased with all foreign anions for the WT channel, and reversal potentials changed between −62 and −28 mV when external Cl⁻ was replaced by the indicated anions (Fig. 6 A, upper left). We used the changes of zero-current reversal potentials and the Goldman–Hodgkin–Katz Equation (Eq. 3) to calculate permeability ratios (P_X/P_Cl) and determine the anion selectivity sequences for each channel (Table 1). The anion permeability ratios are plotted as gray bars in Fig. 6 B, and the characteristic lyotropic selectivity sequence SCN⁻ > I⁻ > NO₃⁻ > Br⁻ > Cl⁻ is listed in the right column of Table 1. The reversal potentials of the currents generated by the E702Q mutant in the presence of Cl⁻, Br⁻, NO₃⁻, I⁻, or SCN⁻ were −22, −33, −31, −41, and −37 mV, respectively. Wider variation in the corresponding values was observed for E705Q (−26, −40, −46, −50, and −33 mV) but not for the double mutant E702Q/E705Q (−20, −25, −26, −31, and −23 mV). The calculated ionic permeability ratios (Table 1) are compared in Fig. 6 B. In general, permeability ratios for NO₃⁻, I⁻, Br⁻, and SCN⁻ were smaller in the mutants than in the WT TMEM16A recorded in the presence of 0.2 µM Ca²⁺ (Fig. 6 B, gray bars).

Because E702Q and E702/E705Q mutations make the pore nearly unselective to anions, we decided to test the role of the whole acidic residues of site 1 in anion selection by mutating them. For this purpose, we tested the 5Q mutant in the absence and presence of Ca²⁺. In the absence of intracellular Ca²⁺, 5Q activity produces small currents at negative potentials (Fig. 6 A, bottom left), displaying an outwardly rectifying open phenotype (Segura-Covarrubias et al., 2020), thus compromising the determination of the reversal potential. To show that 5Q mediated these currents, we recorded I-V relationships in the presence of Cl⁻ and SCN⁻ from the same cells exposed to 10 µM DCPIB (4-(2-butyl-6,7-dichlor-2-cyclopentylindan-1-on-5-yl)), an inhibitor of volume-regulated anion channels. Then, we added 0.5 µM ANI9, an inhibitor of TMEM16A, in the presence of DCPIB. ANI9 inhibited the current recorded in the presence of SCN⁻ without changing the reversal potential, indicating that 5Q activation generates the small currents. Fig. S1 B shows an example of these I-V relations.

Fig. 6 A (bottom left) shows I-V relations obtained in the presence of Cl⁻, Br⁻, NO₃⁻, I⁻, or SCN⁻ from cells dialyzed with solutions without Ca²⁺. Interestingly, the current reversal potentials and anion selectivity sequence of 5Q in the absence of Ca²⁺ were similar to the WT channel in the presence of 0.2 µM Ca²⁺ (Fig. 6 B, lime green bars; Table 1). This result revealed that the negative charges at the Ca²⁺-binding site 1 have to be mostly neutralized to allow the channel to discriminate among anions. Then, we determined the anion selectivity in the presence of 12 µM free Ca²⁺ in the cytosolic side. Unexpectedly, the rectification was lost, the current at negative potentials increased, and the zero-current potential values between tested anions were less significant (Fig. 6 A, bottom right). Accordingly, the corresponding P_X/P_Cl ratios were <1.5 for all foreign anions (Fig. 6 B, sea green), indicating that 5Q in the presence of 12 µM Ca²⁺ lost anion selectivity. The anion selectivity sequence of this mutant in the presence of Ca²⁺ was I⁻ > NO₃⁻ = SCN⁻ = Br⁻ > Cl⁻ (Table 1).

The alterations in anion selectivity produced by single or double Glu (E) to Gln (Q) substitutions were surprising, provided that this procedure presumably only neutralizes the negative charge, similar to titration with Ca²⁺. As we cancel the charges by a mutation in site 1, the electrostatic potential of the pore should drop until it is completely balanced, as was previously shown using Ca²⁺ in the G644P mutant (Lam and Dutzler, 2018). Therefore, we assumed that the site-directed mutagenesis gradually opened the so-called “electrostatic gate,” which is known to hamper anion conduction (Lam and Dutzler, 2018). To find out whether this idea was correct, we calculated the electrostatic potential profiles along the pore for WT, I641A, E702Q, E705Q, E702Q/E705Q, and 5Q channels. To mimic the experimental conditions used to determine the anion selectivity sequences, we performed the calculations in the absence and presence of one or two Ca²⁺ ions in the Ca²⁺ pocket of E702Q, E705Q, and E702Q/E705Q mutant pore structures and without Ca²⁺ in the 5Q. We used the 5OYB structure as a template for replacing the Ca²⁺ pocket residues in silico. In Fig. 7 A, we compare the electrostatic potential profiles obtained.

Overall, electrostatic potentials inverted around the middle of the pore and did not change in the extracellular side. However, compared with WT, we saw significant changes at site 1 and the hydrophobic gate in single and double mutant channels. In the absence of Ca²⁺ (continuous lines), the electrostatic potential became more negative at the site 1 Ca²⁺ pocket region of E702Q, E705Q, and E702Q-E705Q channels. Moreover, in the presence of one Ca²⁺ ion placed near the Ca²⁺ pocket (dotted lines), the potential profiles in these mutants became more positive at the site 1 region but were more negative than the values from the WT profile with two Ca²⁺ ions (black dots). On the other hand, E702/E705 with two Ca²⁺ ions randomly placed near the Ca²⁺ pocket produces a more positive potential profile (dotted blue line; by ∼10 kcal/mol). The excess of positive charges due to the presence of the two Ca²⁺ ions could explain this effect. Finally, if the binding of two Ca²⁺ ions balances the electrostatic potential, then 5Q without Ca²⁺ and WT with two Ca²⁺ ions in the pocket should have nearly equal electrostatic potential profiles. Indeed, the electrostatic potential profile of the 5Q pore (continuous red line) was almost the same as that of the WT pore with Ca²⁺ (black dotted line).

This analysis shows that mutating negatively charged residues and increasing Ca²⁺ occupancy in the Ca²⁺ pocket decreases the electrostatic potential in this region. Whether this electrostatic potential drop takes part in the loss of anion selectivity is uncertain. However, it is worth noting that as the electrostatic potential gradually decreases, the channels lose their anion selectivity. Additionally, the electrostatic potentials of WT 0.2 µM Ca²⁺ and 5Q 0 Ca²⁺ are both closer to 0 (neutralized) at site 1 (Fig. 7 B, continuous red line and black dotted line). Functionally, these are the channels that stay selective. This result supports the idea that the electrostatic potential participates in anion selection.

We also compared the electrostatic potential profiles of WT and I641A channels in the absence and presence of Ca²⁺ using the Ca²⁺-free structure (PDB accession no. 5OYG) of WT TMEM16A (Paulino et al., 2017). The resulting profiles were similar to those previously reported by Lam and Dutzler, (2018). The potential polarity of both pores inverts in the presence of Ca²⁺, just as in the Ca²⁺-bound structure 5OYB (Fig. 7 A). In the absence and presence of Ca²⁺ (continuous versus dotted lines, respectively, in Fig. 7 B), the magnitude of the negative and positive potentials was the same in both pores (black versus pink continuous and dotted lines). Note the substantial reduction of the −35 kcal/mol potential around site 1 by adding Ca²⁺. Also, the potential is inverted around the hydrophobic gate by the presence of Ca²⁺. However, the total change is minor and, as shown below, the I641A mutation changes anion selectivity, suggesting that the hydrophobic gate sterically contributes to anion selectivity.

In addition to electrostatic, the steric elements in the pore also play a role in controlling the Cl⁻ conductance (Lam et al., 2021). For example, the structural rearrangements of TM6, after Ca²⁺ binding, control the opening of the hydrophobic gate of TMEM16A channels (Paulino et al., 2017), which participates in anion conduction. Therefore, the TM6 may sterically compromise Cl⁻ permeation and anion selectivity by partially occluding the apo conformation pore. Consequently, we assessed whether TM6 could be involved in the selectivity process. To this end, we determined the anion selectivity sequence of the single-mutant G644P. This mutation hinders the TM6 hinge flexibility, and the cytosolic half of α-6 presumably adopts an open conformation (Lam and Dutzler, 2018). Additionally, this mutation renders a channel with increased Ca²⁺ sensitivity that can be activated by voltage without intracellular Ca²⁺ (Paulino et al., 2017). Fig. 8 A shows I-V relations around the reversal potentials under different anionic conditions on the extracellular side. Notably, the currents reverted in the narrow voltage range of −35 and −27 mV. This result shows that the G644P mutant channel lost anion selectivity, as corroborated by calculating the permeability ratios for Br⁻, NO₃⁻, I⁻, and SCN⁻ (Fig. 8 B) and its anion selectivity sequence (Table 1).

The other steric element of TMEM16A gating is the hydrophobic gate, whose opening is coupled to α-6 movement. The I641 is a critical residue for the integrity of the hydrophobic gate (Lam et al., 2021); therefore, we mutated this residue to Ala (A) and determined the anion selectivity to evaluate the role of the gate on the selectivity process. Because the I641A channel has high Ca²⁺ sensitivity and is activated by voltage in the absence of intracellular Ca²⁺, it allowed us to determine anion selectivity in the absence and presence of 0.2 µM intracellular Ca²⁺ and compare results with those for WT TMEM16A. Fig. 8 C shows I-V relationships around the reversal potential when Cl⁻, Br⁻, NO₃⁻, I⁻, or SCN⁻ was present in the extracellular side. The reversal potentials of the currents generated by cells that expressed the I641A mutant ranged from −29 to −20 mV (upper panel of Fig. 8 C) and from −34 to −20 mV in the absence of intracellular Ca²⁺ (lower panel of Fig. 8 C). Fig. 8 D shows that mutating I641 sharply reduced the permeability ratios and changed the anion selectivity sequence from SCN⁻ > I⁻ > NO₃⁻ > Br⁻ > Cl⁻ in WT TMEM16A (gray bars) to I⁻ > Br⁻ ≥ SCN⁻ > NO₃⁻ > Cl⁻ in the I641A mutant (cyan bars) in the presence of 0.2 μM Ca²⁺ and to I⁻ ≥ SCN⁻ > Br⁻ > NO₃⁻ > Cl⁻ in I641A in absence of Ca²⁺ (pink bars; Table 1).

To further support the role of the gating elements in anion selectivity, we obtained the permeability ratios in the absence of Ca²⁺ of channels with site 1 and hydrophobic gate (5Q/I641A) or site 1 and TM6 (5Q/G644P) mutated (Fig. 9). As we expected, the shifts in reversal potentials induced by anion replacements in the 5Q/I641A and 5Q/G644P channels (Fig. 9 A) were significantly reduced compared with 5Q in the absence of Ca²⁺ (Fig. 6 A). However, these reversal potentials were closer to G644P or I641A (Fig. 8). The resulting P_X/P_Cl values (Fig. 9 B) and the anion selectivity sequence (Table 1) indicate that both mutants are almost anion unselective.

The 5Q selectivity sequence in the absence of Ca²⁺ is probably like WT, because its gating machinery remained unaltered. Even if site 1 charges are neutralized, if the gating process is changed by mutating either the hydrophobic gate (5Q/I641A) or TM6 movement (G644P-5Q), the channel will become unselective. This comparison highlights the importance of the hydrophobic gate and TM6.

The last column in Table 1 lists anion selectivity sequences for the channels we studied. Mutations at the hydrophobic gate, TM6, and site 1 changed the anion selectivity sequence. For example, channels with a mutated hydrophobic gate preferred I⁻ over other ions and Br⁻ over NO₃⁻, while channels with mutations in site 1 selected I⁻ and NO₃⁻ over SCN⁻. Taken together, the results of the mutagenesis analysis show that the electrostatic and steric gating elements of TMEM16A control the anion selectivity.

Because SCN⁻ MF facilitated the voltage dependence of WT channels that have their anion selection mechanism intact (Fig. 5), we sought to test whether voltage dependence facilitation is altered in channels made nonselective by mutating their hydrophobic gate (I641A) or site 1 (E702Q-E705Q). We determined the voltage dependence at different external SCN⁻ MF from cells dialyzed with 0 and 0.2 (I641A) or 12 (E702Q/E705Q) µM Ca²⁺. Fig. 10, C and E, shows examples of whole-cell currents recorded from two different cells expressing I641A (A) and E702Q/E705Q (C) channels recorded in the presence of Ca²⁺. The cells were initially in a Cl⁻-containing solution (black) and then in an SCN⁻-containing solution (red). Although SCN⁻ increased both mutants’ current, just as in WT channels (see Fig. 5), the effects of SCN⁻ MF on I641A and E702Q/E705Q differed from WT. For example, the zero-current potentials did not significantly hyperpolarize by increasing SCN⁻ MF (Fig. 10 G). Consequently, the change in P_SCN/P_Cl (Fig. 10 H) was negligible in both channels, indicating that channels whose hydrophobic gate or Ca²⁺ pocket was mutated were insensitive to SCN⁻ MF. Importantly, by increasing the SCN⁻ MF, the G/G_{max_app, 0.2 µM Ca}²⁺-versus-V curves of I641A (Fig. 10 D) shifted toward negative potentials, just like WT channels. However, the V_{0.5_app} shifts were more significant in the I641A mutant than in WT (when SCN⁻ MF = 1.00, V_{0.5_app} was −28 in I641A and +45 mV in WT) with no change in the slope (Fig. 10, I and J, gray spheres). In sharp contrast, the SCN⁻ MF negligibly altered the G/G_{max_app, 12 µM Ca}²⁺-versus-V curves (Fig. 10 F), V_{0.5_app} (Fig. 10 I), and slope (Fig. 10 J) values in the E702Q/E705Q channel (orange spheres). The complete substitution of Cl⁻ by SCN⁻ in this mutant shifted the G/G_{max_app, 12 µM Ca}²⁺ curve −30 mV on average and the zero-current potentials by −11 mV. By contrast, the same substitution in the WT channel moved these parameters by −83 and −39 mV, respectively.

A salient result from these experiments is the strong sensitivity of I641A V_{0.5_app} on SCN⁻ MF. This result was obtained from experiments performed in the presence of intracellular Ca²⁺; thus, we presume that Ca²⁺ still binds to site 1 because it is unaltered in the I641A mutant. Moreover, this mutant displays higher sensitivity for Ca²⁺ than WT (Peters et al., 2018). Thus, we suppose that permeant SCN⁻ would stabilize the Ca²⁺ bound to site 1, as previously shown in WT channels (Fig. 4), hence increasing the voltage sensitivity. To evaluate this idea, we repeated an SCN⁻ MF experiment recording the I641A mutant without intracellular Ca²⁺. Fig. 10 A shows that the current is enhanced by replacing the extracellular Cl⁻ with SCN⁻. However, the G/G_{max_app, 0 µM Ca}²⁺-versus-V curves (Fig. 10 B) and their V_{0.5_app} (Fig. 10 I, lime symbols) were not strongly affected by the SCN⁻ MF. Also, the change in reversal potential was less than −20 mV after the total substitution of Cl⁻ with SCN⁻ (Fig. 10 G). P_SCN/P_Cl ratios were ∼2 with different SCN⁻ MF (Fig. 10 H). Interestingly, the slope remained at ∼0.03 mV⁻¹ (Fig. 10 J), a value slightly higher than the one for WT channels (0.02 mV⁻¹).

In conclusion, these data demonstrate that mutations at the hydrophobic gate or site 1 not only hinder anion selectivity but also alter the ability of the channel to change its voltage dependence when highly permeant anions are present.

This work describes that permeable anions increase the Ca²⁺ sensitivity of TMEM16A as their permeability ratio increases, presumably by stabilizing Ca²⁺ binding to site 1 (an essential step to start the gating process). Conversely, mutations at the Ca²⁺ binding site 1, TM6 hinge, and hydrophobic gate turned TMEM16A unselective, thus revealing the essential role of the gating elements as structural determinants of anion selectivity. Overall, our data point to a reciprocal interaction between permeant anions and gating.

A critical aspect for understanding the permeant anions’ role in TMEM16A gating, and probably in TMEM16B too, is that site 1 is next to the permeation pathway, thus enabling a two-way interaction (Lam and Dutzler, 2018). On the one hand, permeant anions in the pore or near the Ca²⁺ pocket stabilize Ca²⁺ ligation, thus enhancing the apparent Ca²⁺ sensitivity. On the other hand, the gating elements of TMEM16A control anion selectivity. This mutually affecting interaction within the pore is in agreement with present and previous results showing that (1) anions more permeable than Cl⁻ as well as SCN⁻ MF increase the apparent Ca²⁺ sensitivity and slow the off-rate of Ca²⁺-induced currents; (2) the apparent Ca²⁺ sensitivity shows AMF behavior; (3) mutations at the gating elements make TMEM16A nonselective and alter the SCN⁻-dependent regulation of voltage dependence; (4) Cl⁻ conductance is facilitated by neutralizing site 1 electrostatic potential with Ca²⁺ (Lam and Dutzler, 2018); (5) P_I/P_Cl and P_SCN/P_Cl decrease in the WT channel upon increase of the intracellular Ca²⁺ concentration (Peters et al., 2015); (6) the apparent Ca²⁺ affinity of TMEM16A increases by neutralizing or reversing the charge of K588 and K645 adjacent to site 1 (Lam and Dutzler, 2018); (7) the on and off of the Ca²⁺-activated Cl⁻ currents from parotid acinar cells are accelerated and slowed, respectively, by permeant anions (Perez-Cornejo et al., 2004); (8) intracellular SCN⁻ increases the apparent Ca²⁺ sensitivity of Ca²⁺-activated Cl⁻ currents in Xenopus oocytes (Qu and Hartzell, 2000); (9) anions with higher permeability than Cl⁻ increase the apparent Ca²⁺ sensitivity of TMEM16B (Betto et al., 2014); and (10) dynamic anion selectivity is observed in TMEM16B exposed to Ca²⁺ jumps (Sagheddu et al., 2010). These results strongly suggest that permeation and Ca²⁺ ligation are coupled and highlight the critical role of permeant anions in TMEM16A activation. Alternatively, anions could hamper the conformational transition of α-6 following Ca²⁺ binding or the closing of the hydrophobic gate instead of directly interacting with the Ca²⁺-sensing machinery, thus slowing channel closure.

Other channels have a gating mechanism that shows dependence on the permeant ion. For example, in voltage-gated Cl⁻ channels CLC-0 and CLC-1, a Cl⁻ ion could be responsible for opening a glutamate gate protruding the permeation pathway (Pusch et al., 1995; Chen and Miller, 1996; Chen, 2003; Pusch, 1996; Rychkov et al., 2001). Furthermore, we showed that the interaction between the hyperpolarization-propelled intracellular Cl⁻ and the pore gate triggers CLC-2 gating (De Jesús-Pérez et al., 2016, 2021; Sánchez-Rodríguez et al., 2010, 2012). In TMEM16B, highly permeable anions facilitated the voltage-gated opening and increased the apparent Ca²⁺ sensitivity (Betto et al., 2014). Furthermore, we reported that the slow gating mode of TMEM16A and the process by which TMEM16A reaches full conduction depend on Cl⁻ ions (Contreras-Vite et al., 2016; Cruz-Rangel et al., 2017). Thus, the previous data and the present work indicate that in TMEM16A and TMEM16B Cl⁻ channels, the permeant anion plays an essential role in gating.

It was previously established that the extracellular entrance and the narrow neck of the pore (Fig. 1) take part in anion selection in TMEM16A. Alanine scanning of the extracellular residues R511, K599, R617, and R784 located in α-3, α-5, α-5–α-6 linker, and α-9, respectively, showed that the P_I/P_Cl and P_SCN/P_Cl increased without altering the apparent Ca²⁺ sensitivity, but, in the same mutants, P_SCN/P_Cl increased by raising the [Ca²⁺]_i from 400 nM to 1 mM (Peters et al., 2015). Furthermore, mutation of two lysines, K584 and K645 in TMEM16A (a; K588 and K649 in TMEM16A [ac], respectively) located at the narrow neck modulated channel rectification and ion permeation under extremely high Ca²⁺ concentrations to neutralize the electrostatic potential generated by the Ca²⁺ pocket (Reyes et al., 2015; Lam and Dutzler, 2018). In addition, varying the electrostatic potential of the pore changes the anion conductance of TMEM16A (Nguyen et al., 2019; Lim et al., 2016; Jeng et al., 2016). Here, we report that anion selectivity is controlled by the gating elements of TMEM16A: the Ca²⁺ binding pocket of site 1, TM6, and the hydrophobic gate. Thus, the anion selectivity process by TMEM16A occurs at five structural domains along the pore: one located on the extracellular side formed by four basic residues, and four inside the pore, including (1) the hydrophobic gate, (2) two lysines at the narrow neck, (3) site 1, and (4) the cytosolic side of the TM6 (Fig. 1). Thus, anion selection in TMEM16A occurs along the narrow section of the pore, a process similar to that described in BEST1, another Ca²⁺-activated Cl⁻ channel (Vaisey et al., 2016).

The anion selectivity mechanism of TMEM16A is poorly understood. The consensus anion selectivity sequence determined from the reversal potentials of macroscopic currents recorded from different native cells expressing endogenous CaCC or cells expressing cloned TMEM16A is SCN⁻ > I⁻ > NO₃⁻ > Br⁻ > Cl⁻ (Evans and Marty, 1986; Large and Wang, 1996; Yang et al., 2008; Schroeder et al., 2008; Adomaviciene et al., 2013; Ni et al., 2014; Qu and Hartzell, 2000; Perez-Cornejo et al., 2004; Berg et al., 2012). This lyotropic sequence has been reported in different anion channels (Large and Wang, 1996) and suggests a selectivity mechanism for a pore that contains weak field strength binding sites (Wright and Diamond, 1977). Our data indicate that neutralizing the negative charges at site 1 is necessary for TMEM16A channels to present their characteristic lyotropic selective sequence. However, it seems unlikely that these charges are part of the binding sites. Unlike K⁺ channels (Gouaux and Mackinnon, 2005), TMEM16A structures do not reveal Cl⁻ ions within the permeation pathway (Paulino et al., 2017), thus making it challenging to find the pore domains involved in ion binding and their physical-chemistry properties. Molecular dynamics simulations show that Cl⁻ paused at the two regions of the pore during the permeation (Jia and Chen, 2021). However, no coordination between Cl⁻ and pore residues was determined. Instead, the simulation suggests that the hydrophobic portion of the neck generates energy barriers for Cl⁻ permeation that are lower for SCN⁻ due to its lower hydration energy. More structural data are needed to determine the Cl⁻ binding sites within the TMEM16A pore and its relation to Cl⁻ permeation.

The poor anion selectivity and the linear relationship between permeability ratios and anion hydration energies suggest that the anion dehydration energy determines a weak interaction between the permeant anion and putative binding sites within the WT TMEM16A pore (Dawson et al., 1999; Perez-Cornejo et al., 2004; Qu and Hartzell, 2000). Interestingly, molecular dynamics simulations showed that a single Cl⁻ ion permeates the entire TMEM16A pore surrounded by four to five water molecules at the narrow neck and six to seven water molecules at the vestibules (Jia and Chen, 2021). This observation would suggest that Cl⁻ permeates the pore in the hydrated state. Furthermore, we found that channels with their hydrophobic gate, TM6, and Ca²⁺ pocket mutated lost the linear relationship between permeability ratios and anion hydration energy (Fig. S2). This observation indicates that the tested anions experience almost equal energy barriers during the permeation process through mutant pores. Our finding that channels with a mutated Ca²⁺ pocket display similar electrostatic potential profiles supports this conclusion.

Interestingly, SCN⁻ and I⁻, the most permeable anions through TMEM16A, can be oxidized to produce antimicrobial compounds in secretory glands and the extracellular fluid layer of the airways. This oxidation process is caused by the hydrogen peroxide produced by dual oxidase 1/thyroid oxidase 1 and dual oxidase 2/thyroid oxidase 2 (Geiszt et al., 2003). In normal conditions, it is estimated that the airway secretions could have ≤0.4 mM SCN⁻ (Fragoso et al., 2004), which dual oxidases can use as a source for manufacturing bactericides. This high concentration of extracellular SCN⁻ is partly maintained by the elevated SCN⁻ permeability of the CFTR Cl⁻ channel (Moskwa et al., 2007). However, in cystic fibrosis patients, this antibacterial system is inoperative (Moskwa et al., 2007). Therefore, TMEM16A, which is expressed along with CFTR in airway epithelial cells, could serve as a CFTR surrogate to maintain the extracellular SCN⁻ concentration and sustain the microbicidal function (Pedemonte et al., 2007). However, the role of TMEM16A in this defense mechanism is unclear. The permeability of TMEM16A to SCN⁻ and I⁻ and the production of antimicrobial substances would be compromised if residues controlling the hydrophobic and site 1 are mutated or by pharmacological interventions targeting the hydrophobic gate (Dinsdale et al., 2021). However, mutations of the four extracellular basic residues can upregulate the permeability to SCN⁻ and I⁻ (Peters et al., 2015). Therefore, additional experiments aimed at testing the role of TMEM16A in this defense system are required.

In conclusion, we present evidence that anion selectivity occurs along the pore of TMEM16A. Also, we show that reciprocal interaction between permeant anions and the gating elements alter the Ca²⁺ sensitivity, voltage dependence, and the anion selectivity of TMEM16A. This type of regulation likely impacts the mean open time of the channels, ultimately tuning anion flux during physiological responses. In addition, these data add to the growing body of evidence supporting the idea that permeant anions play an active role in ion channel gating.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Joseph A. Mindell served as editor.

The authors thank C.Y. Hernandez-Carballo for technical assistance.

The work was supported by grant FORDECYT-PRONACES (Ciencia de Frontera 2019) 1308052 2020 from CONACyT, México. J.J. De Jesús-Pérez was supported by a Graduate Student Fellowship #234820 and Postdoctoral Fellowship #711128 from CONACyT, México. O. Posadas was supported by a Graduate Student Fellowship # 933084 from CONACyT, México. G. Segura-Covarrubias was supported by a Graduate Student Fellowship # 297721 from CONACyT, México.

The authors declare no competing financial interests.

Author contributions: All experiments described in this manuscript were performed at the Institute of Physics of the Universidad Autonoma de San Luis Potosi. J.J. De Jesús-Pérez: Design of the work, acquisition, analysis, and interpretation of data, and revising the manuscript critically for important intellectual content. A.E. López-Romero: Acquisition, analysis, and interpretation of data, and revising the manuscript critically. O. Posadas: Acquisition, analysis, and interpretation of data, and revising the manuscript critically. G. Segura-Covarrubias: Acquisition, analysis, and interpretation of data, and revising the manuscript critically. I. Aréchiga-Figueroa: Acquisition, analysis, and interpretation of data, revising the manuscript critically for important intellectual content, and securing funds to support the project. B. Gutiérrez-Medina: Interpretation of data and revising the manuscript critically for important intellectual content. P. Pérez-Cornejo: Design of the work, interpretation of data, revising the manuscript critically for important intellectual content, and securing funds to support the project. J. Arreola: Conception and design of the work, analysis and interpretation of data, drafting the work and revising it critically for important intellectual content, and securing funds to support the project.