An Additional Ca²⁺ Binding Site Allosterically Controls TMEM16A Activation

SUMMARY

Calcium (Ca²⁺) is the primary stimulus for transmembrane protein 16 (TMEM16) Ca²⁺-activated chloride channels and phospholipid scramblases, which regulate important physiological processes ranging from smooth muscle contraction to blood coagulation and tumor progression. Binding of intracellular Ca²⁺ to two highly conserved orthosteric binding sites in transmembrane helices (TMs) 6–8 efficiently opens the permeation pathway formed by TMs 3–7. Recent structures of TMEM16K and TMEM16F scramblases revealed an additional Ca²⁺ binding site between TM2 and TM10, whose functional relevance remains unknown. Here, we report that Ca²⁺ binds with high affinity to the equivalent third Ca²⁺ site in TMEM16A to enhance channel activation. Our cadmium (Cd²⁺) metal bridging experiments reveal that the third Ca²⁺ site’s conformational states can profoundly influence TMEM16A’s opening. Our study thus confirms the existence of a third Ca²⁺ site in TMEM16A, defines its functional importance in channel gating, and provides insight into a long-range allosteric gating mechanism of TMEM16 channels and scramblases.

Graphical Abstract

In Brief

Le and Yang present electrophysiological evidence for a third Ca²⁺ binding site in the Ca²⁺-activated chloride channel TMEM16A. They show that Ca²⁺ binds with high affinity to this site to facilitate channel opening, providing insight into an allosteric gating mechanism in TMEM16 ion channels and scramblases.

INTRODUCTION

Transmembrane protein 16A (TMEM16A) belongs to the TMEM16 family whose members include both Ca²⁺-activated chloride channels (CaCCs) (Caputo et al., 2008; Schroeder et al., 2008; Yang et al., 2008) and dual functional Ca²⁺-activated non-selective ion channels and phospholipid scramblases (Bushell et al., 2019; Le et al., 2019b; Malvezzi et al., 2013; Suzuki et al., 2010, Suzuki et al., 2013; Yang et al., 2012). The CaCC TMEM16A plays an important role in regulating smooth muscle contraction, gut motility, fluid secretion, nociception, and anxiety (Cho et al., 2020; Scudieri et al., 2012; Whitlock and Hartzell, 2017). Abnormal expression of TMEM16A is intimately associated with numerous types of cancers, such as gastrointestinal stromal tumors; head and neck squamous carcinomas; and breast, pancreatic, and prostate cancer, thus highlighting its potential as a biomarker and promising anticancer therapeutic target (Crottès and Jan, 2019).

As a polymodal sensor, the gating landscape of TMEM16A is shaped by the synergistic action of Ca²⁺ binding and membrane depolarization, both of which promote channel opening (Falzone et al., 2018; Whitlock and Hartzell, 2017). TMEM16A’s channel activity is also positively regulated by the phosphatidylinositol (4,5)-bisphosphate (PIP₂) (De Jesús-Pérez et al., 2018; Le et al., 2019a; Ta et al., 2017; Tembo et al., 2019; Yu et al., 2019). Extensive structural and functional studies have advanced our mechanistic understanding of TMEM16A’s Cl⁻ permeation, Ca²⁺-dependent gating, and its regulation by lipids. Assembled as a homodimer, each TMEM16A subunit contains 10 membrane-spanning segments with its own independent pore (Figure 1A) (Dang et al., 2017; Jeng et al., 2016; Lim et al., 2016; Paulino et al., 2017a; Paulino et al., 2017b). Transmembrane helix (TM)3–TM8 form the pore-gate domain that is essential for ion permeation, Ca²⁺ sensing, and PIP₂ regulation, whereas TMs 1, 2, 9, and 10 form the supporting domain that establishes the dimer interface (Le et al., 2019a). The hourglass-shaped permeation pathway is composed of residues from TM3 to TM7, many of which have been implicated in controlling anion selectivity (Dang et al., 2017; Peters et al., 2015; Yang et al., 2012). Binding of intracellular Ca²⁺ to five highly conserved acidic residues and an asparagine from TM6–TM8 (Figure 1A) (Brunner et al., 2014; Paulino et al., 2017a; Tien et al., 2014; Yu et al., 2012) not only induces a prominent conformational rearrangement of TM6 to help open the steric gate but also relieves a second, electrostatic gate imposed by the Ca²⁺-binding acidic residues, subsequently allowing chloride conduction (Dang et al., 2017; Lam and Dutzler, 2018; Paulino et al., 2017a; Peters et al., 2018).

Figure 1. An Additional Ca²⁺ Binding Site in TMEM16 Ion Channels and Scramblases.

(A) Structure of the Ca²⁺-bound mouse TMEM16A homodimer (PDB: 5OYB). The anion permeation pathway in one monomer is shown as a yellow surface. TMs 2 and 10 are indicated in magenta and cyan, respectively; TMs 6, 7, and 8 are indicated in light blue; and Ca²⁺ ions are indicated as red spheres. The second monomer is indicated as orange ribbons. Insets, close-up views of the orthosteric Ca²⁺ binding sites (blue box) and the putative third Ca²⁺ site (red box). Out: extracellular; in: intracellular.

(B and C) Cartoon representation of TMEM16A showing the proposed modular design as viewed from the side (B) or top (C) (Le et al., 2019a).

(D) Protein sequence alignment of TMEM16 proteins at the third Ca²⁺ binding site in TM2 and TM10.

(E and F) The Ca²⁺-bound structures of mouse TMEM16F (PDB: 6QP6) (E) and human TMEM16K (PDB: 6R7X) (F). Insets indicate their respective third Ca²⁺ sites in TM2 and TM10.

Intriguingly, recent structural studies of TMEM16F and TMEM16K scramblases have revealed yet another Ca²⁺ binding site in TM2 and TM10 of the supporting domain (Figures 1D–1F) (Alvadia et al., 2019; Bushell et al., 2019). We refer to this additional Ca²⁺ site as the “third Ca²⁺ site” to distinguish it from the primary “orthosteric sites” in TM6–TM8. Despite not being resolved initially, it was later suggested that the Ca²⁺-bound TMEM16A structure (Paulino et al., 2017a) also harbors an electron density at the equivalent site, presumably deriving from a bound Ca²⁺ ion (Alvadia et al., 2019). This assumption is supported by the fact that the Ca²⁺-coordinating residues at the third site are highly conserved among TMEM16 proteins (Figure 1D). More importantly, the distant location of this Ca²⁺ site from the pore-gate domain (Figures 1A–1C) also suggests that TMEM16 proteins may undergo a long-range allosteric modulation, a regulatory mechanism common in many ion channels and membrane receptor proteins that harbor multiple topologically distinct, yet tightly coupled, sensors (Changeux and Christopoulos, 2016). Nevertheless, lack of functional evidence raises questions on the existence and functional relevance of this Ca²⁺ site in regulating TMEM16 proteins. Interestingly, a missense mutation of a third Ca²⁺-coordinating residue, D615N, in TMEM16K was identified in a spinocerebellar ataxia type 10 (SCAR10) patient with unknown patho-physiology (Figures 1D–1F) (Balreira et al., 2014). Therefore, understanding the functional role of this third Ca²⁺ binding site not only informs an allosteric gating mechanism of TMEM16 proteins but also provides potential insight into the molecular patho-physiology of TMEM16K variant in SCAR10.

To define the functional existence and the molecular basis for channel-gating modulation by the third Ca²⁺ binding site, we studied the TMEM16A CaCC using electrophysiology, mutagenesis, and Cd²⁺-mediated metal bridging. Here, we show that the third site in TM2 and TM10 binds intracellular Ca²⁺ with high affinity. Notably, mutating the Ca²⁺-binding residues at this site not only significantly reduces its Ca²⁺ sensing but also directly affects the channel’s activation gate. Using Cd²⁺-mediated metal bridging of cysteines engineered to the third Ca²⁺ site, we further demonstrate that the conformational states of TM2 and TM10 can strongly influence TMEM16A’s channel opening in a manner that is independent of the orthosteric Ca²⁺ sites. Our functional analyses thus unveil an unknown regulatory mechanism in which channel activation is facilitated by Ca²⁺ binding to the TM2–TM10 site. We anticipate that our present findings could stimulate future mechanistic investigations on the allosteric regulation and the development of allosteric modulators for TMEM16 ion channels and phospholipid scramblases.

RESULTS

Mutation of the Putative Third Ca²⁺ Site Residues Alters TMEM16A’s Ca²⁺ Sensing

Given the structural and sequence conservation between mammalian TMEM16 proteins at the putative third Ca²⁺ binding site (Figures 1A and 1D–1F), we examined whether the equivalent site in TMEM16A contributes to channel gating by measuring the Ca²⁺-dependent activation of the alanine mutations of E425 and K428 from TM2 and of D879 and D884 from TM10 (Figures 2A–2E). Paradoxically, the K428A and D884A mutations markedly reduce, while E425A and D879A enhance, the apparent Ca²⁺ sensitivity of TMEM16A (Figures 2F and 2I; Table S1). Similar phenotypes were also observed in the charge-reversing mutations of these residues (Figure S1; Table S1). Although they do not tell us whether the third Ca²⁺ site residues indeed bind Ca²⁺, these results offer clear evidence to support that this putative Ca²⁺ site in TM2 and TM10 significantly impacts the Ca²⁺-dependent activation of TMEM16A.

Figure 2. Mutations at the Putative Third Ca²⁺ Site Alter TMEM16A’s Ca²⁺ Sensitivity.

(A–E) Representative recordings showing Ca²⁺-dependent activation of TMEM16A wild type (WT) (A), E425A (B), K428A (C), D879A (D), and D884A (E). Ca²⁺ concentrations are shown as numbers above outward current deflections (in micromolar). The membrane was held at +60 mV. Red dashed line indicates 0-pA baseline.

(F) The Ca²⁺ dose-response curves of TMEM16A WT and mutants.

(G and H) Representative recording showing Ca²⁺-dependent activation of TMEM16A D884N (G) and its dose-response curve (H).

(I) Quantifications of the half-maximal activation concentration of Ca²⁺ (Ca²⁺ EC₅₀) values of TMEM16A WT and mutants (see also Table S1).

Numbers in parentheses denote the number of individual recordings. One-way ANOVA: p values are 0.0244, <0.0001, 0.0027, <0.0001, and <0.0001 for E425A, K428A, D879A, D884A, and D884N, respectively. Data are represented as means ± SEM.

TMEM16A’s D884 residue is absolutely conserved among mammalian TMEM16 proteins and corresponds to TMEM16K D615, whose D615N variant was identified in a patient with autosomal recessive SCAR10 (Figures 1A, 1D, and 1F) (Balreira et al., 2014; Bushell et al., 2019). To gain some insight into the mutational effect of TMEM16K D615N, we measured the Ca²⁺ sensitivity of the equivalent TMEM16A D884N mutant (Figure 2G). Similar to D884A, the D884N mutation also significantly reduces TMEM16A Ca²⁺ sensitivity, albeit to a lesser extent (Figures 2H and 2I; Table S1). Although this observation implies a possibility that D615N also impairs TMEM16K’s function, the previous study by Bushell et al. (2019) found that this mutant displayed no obvious defects in its lipid scrambling under both Ca²⁺-free and saturating Ca²⁺ conditions using reconstituted liposomes. We speculated that the mutational effect of TMEM16K D615N may require a more rigorous assessment of its Ca²⁺ dose-dependent lipid scrambling, especially considering the relatively lower sensitivity/resolution of lipid scrambling assays compared to that of electrophysiology as well as the less disruptive Asp-to-Asn substitution.

Taken together, our single point mutation characterizations of TMEM16A reveal that disrupting the putative third Ca²⁺ binding site in TM2 and TM10 alters channel activation.

Disrupting the Third Ca²⁺ Site Residues Alters Channel Activation independent of Ca²⁺

Considering the pronounced effects from the alanine and charge-reversing mutations of the third Ca²⁺ site on channel activation, we speculated that such perturbations could either affect the orthosteric-site-mediated, Ca²⁺-dependent gating or modulate the channel’s activation gate independent of Ca²⁺. To test whether disrupting the third Ca²⁺ site residues would remotely impact TMEM16A’s activation gate in the pore-gate domain independent of Ca²⁺ binding, we utilized the gain-of-function mutant Q645A (Peters et al., 2018), which allows us to measure TMEM16A’s channel activation in the complete absence of Ca²⁺ (Figure 3A). To accurately quantify their voltage-dependent activation, we normalized the measured conductance in the absence of Ca²⁺ to the maximal conductance (G_max), which was obtained by exposing the excised membrane patches to saturating 100 μM Ca²⁺ (Figure 3A). Interestingly, alanine mutations of the putative third Ca²⁺ site residues in the Q645A background markedly shift the conductance-voltage (G-V) curves in the absence of Ca²⁺ (Figures 4B–4F). Whereas E425A and D879A enhance channel activation and shift the G-V curve toward more a negative voltage range, K428A and D884A shift it toward more positive voltages. Furthermore, the charge-reversing mutants E425K and D879K also promote, whereas K428E and D884K reduce, TMEM16A’s channel activation in the absence of Ca²⁺ (Figure S2). The G/G_max-V curves of these mutant channels did not go beyond 0.5 (Figure 3F), which prevents us from extrapolating the half-maximal activation voltage (V_1/2) values for further comparisons. Nonetheless, the mutational effects of the third Ca²⁺ site residues on TMEM16A’s Ca²⁺-independent activation strikingly mirror their respective effects on the channel’s Ca²⁺-dependent activation seen in the wild-type (WT) background (Figure 2; Figure S1). Because the orthosteric Ca²⁺ sites in TM6–TM8 account for the measured Ca²⁺-dependent currents in the WT background (Figure 2), the observed alterations of TMEM16A’s Ca²⁺ sensitivity from the third site mutations should arise from their Ca²⁺-independent allosteric effects on channel gating. Therefore, the paradoxical effects of the third Ca²⁺ site mutants on WT TMEM16A’s Ca²⁺ sensitivity (Figures 2A–2F) likely reflect a net result of the Ca²⁺-dependent activation by the orthosteric Ca²⁺ sites and the Ca²⁺-independent allosteric effects from perturbation of the third Ca²⁺ site on the channel’s activation gate.

Figure 3. Mutation of the Third Ca²⁺ Site Residues Directly Affects Channel Gating independent of Ca²⁺.

(A–E) Representative voltage-current (I-V) measurements of TMEM16A Q645A (A, or QA background construct), E425A(QA) (B), K428A(QA) (C), D879A(QA) (D), and D884A(QA) (E) The same membrane patch was sequentially recorded in 0 μM Ca²⁺ (left recording) and then in 100 μM Ca²⁺ to obtain maximum current (right recording). Voltage protocols are indicated in (A) (right).

(F) Ca²⁺-independent conductance-voltage (G-V) curves of QA mutants. Numbers in parentheses denote the number of individual recordings. Data are represented as mean ± SEM.

Figure 4. Single Alanine Mutations of the Third Ca²⁺ Site Residues Reduce Ca²⁺ Binding.

(A) Schematic diagram showing the experimental design. E701K and D734R mutations were introduced to the Q645A background to abolish Ca²⁺ binding to the orthosteric sites. The Q645A mutation enables channel activation in absence of Ca²⁺ binding to the orthosteric sites.

(B–F) Representative Ca²⁺-activated currents of the Q645A/E701K/D734R background (B, QA/EK/DR), or QA/EK/DR carrying an additional alanine mutation: E425A (C), K428A (D), D879A (E), and D884A (F). Ca²⁺ concentrations are shown as numbers above outward current deflections (in micromolar). The membrane was held at +60 mV. Red dashed line indicates 0-pA baseline.

(G) Quantifications of the Ca²⁺-activated currents of TMEM16A mutants in response to different Ca²⁺ concentrations. bg, QA/EK/DR background; non-trans., non-transfected.

(H and I) Normalized Ca²⁺ dose-response curves of QA/EK/DR background, K428A(QA/EK/DR), D879A (QA/EK/DR), and D884A(QA/EK/DR) (H) and their estimated Ca²⁺ EC₅₀ (I). The Ca²⁺-sensitive currents were normalized to the maximum Ca²⁺-sensitive current.

One-way ANOVA: p values are 0.0002 for K428A(QA/EK/DR) and <0.0001 for E425A(QA/EK/DR), D879A(QA/EK/DR), and D884A(QA/EK/DR). Numbers in parentheses denote the number of individual recordings. Data are represented as means ± SEM.

The Putative Third Ca²⁺ Site in TM2 and TM10 Binds Ca²⁺

Because neutralizing the orthosteric Ca²⁺ binding residues severely diminishes Ca²⁺-dependent activation in TMEM16A (Tien et al., 2014; Yu et al., 2012), it is thus intriguing that neutralizing E425A and D879A mutants at the third Ca²⁺ site enhance, whereas K428A and D884A decrease, the apparent Ca²⁺ sensitivity of TMEM16A (Figure 2I; Table S1). However, the fact that these mutants are sufficient to sensitize or attenuate channel activation in the absence of Ca²⁺ binding is suggestive of their potential pleiotropic effects on not only the channel’s activation gate (Figure 3; Figure S2) but also Ca²⁺ binding to this third site. We reasoned that the mutational effects on channel activation from disrupting the third Ca²⁺ site residues could be confounded by the intact orthosteric Ca²⁺ sites. Thus, to specifically investigate the Ca²⁺ binding properties of the third site, we utilized the E701K and D734R mutations, two charge-reversing mutations of the Ca²⁺-coordinating E701 and D734 on TM7 and TM8, respectively (Figure 1A), previously shown to abolish Ca²⁺ binding to the orthosteric Ca²⁺ sites (Tien et al., 2014; Yu et al., 2012) (Figure 4A). We also introduced the Q645A mutation on TM6 to allow the channel to adopt a partially released steric gate that can be opened via membrane depolarization (Lam and Dutzler, 2018; Peters et al., 2018). Remarkably, the resulting triple mutant Q645A/E701K/D734R (henceforth referred to as QA/EK/DR) channels display robust Ca²⁺ sensitivity, as Ca²⁺ application strongly potentiates the basal current mediated by Q645A (Figure 4B).

Because the E701K and D734R mutations abolish Ca²⁺ binding to the orthosteric sites, we attributed the observed Ca²⁺ sensing capability of QA/EK/DR to the third Ca²⁺ site in TM2–TM10 due to the following lines of evidence. First, by analyzing the Ca²⁺-sensitive currents, which were obtained by subtracting the basal current from each Ca²⁺-elicited steady-state current (Figure 4B), we found that the third Ca²⁺ site has an estimated Ca²⁺ half-maximal activation concentration (EC₅₀) of ~320 nM ± 10 nM (Figures 4G–4I; Table S1), which is well within physiological range and is thus unlikely to be mediated by the already impaired orthosteric sites. To rule out any potential confounding effects from Ca²⁺-dependent rundown on our Ca²⁺ EC₅₀ analysis, we performed rundown-corrected measurements of the QA/EK/DR mutant and obtained a comparable Ca²⁺ EC₅₀ value of ~360 nM ± 20 nM (Figures S3A–S3C; Table S1). Second, consistent with its location closer to the cytosolic side, the Ca²⁺ sensitivity of the third site shows no apparent voltage dependence (Figures S3D–S3F; Table S1), which is distinct from the highly voltage-dependent Ca²⁺ sensitivity of the orthosteric Ca²⁺ binding sites, owing to their location deep into the membrane field (Dang et al., 2017; Paulino et al., 2017a; Xiao et al., 2011). Finally, we designed single alanine mutations of the third Ca²⁺ site—E425A, K428A, D879A and D884A—in the context of the QA/EK/DR background, examined their Ca²⁺ sensitivity, and found that they substantially reduce the Ca²⁺-dependent currents of the QA/EK/DR channels (Figures 4C–4F). The reduced Ca²⁺-dependent currents are not due to decreased channel expression, as these quadruple mutants display robust channel activation in the absence of Ca²⁺ (Figures S4A–S4F, S4M, and S4N). Notably, alanine mutations of the three acidic residues, E425, D879, and D884, most dramatically attenuate the third site’s Ca²⁺ sensing, whereas that of the basic K428 residue causes a less pronounced reduction, consistent with the potential structural role of K428 in stabilizing the third Ca²⁺ site (Figures 4G–4I). Our observation that E425A and D884A strongly reduce the Ca²⁺ sensitivity of the third site is particularly fascinating in light of the recent structural studies showing that the equivalent E395-D859 and E259-D615 residues in TMEM16F and TMEM16K, respectively, directly coordinate the bound third Ca²⁺ ion via their side-chain carboxylates (Alvadia et al., 2019; Bushell et al., 2019). Taken together, our characterizations of the third Ca²⁺ site on the QA/EK/DR background demonstrate that Ca²⁺ binds to this site with high affinity to promote channel opening.

Charged Residues at the Third Ca²⁺ Site Play Differential Roles in Ca²⁺ Binding

To further explore the functional roles of the third Ca²⁺ site residues, we carried out Ca²⁺-dependent current measurements of all six combinations of double mutants at the third Ca²⁺ site on the QA/EK/DR background. Provided the strongly reduced Ca²⁺ sensitivity by single alanine mutations of E425, D879, and D884 (Figures 4G–4I), we reasoned that double-mutant variants comprising two acidic residues at the third site should abolish its Ca²⁺ binding. In agreement with this hypothesis, we found that the acidic double mutants E425A/D879A, E425A/D884A, and D879A/D884A completely lose their Ca²⁺ sensing capability (Figures 5A, 5B, and 5E). The loss of Ca²⁺ sensitivity by the basic/acidic E425A/K428A double mutant further highlights the importance of E425 at the third Ca²⁺ site (Figure 5E). It is worth noting that the lack of Ca²⁺ sensing by these mutants is not due to impaired membrane trafficking, as evidenced by their robust membrane expression (Figure S4Q) and voltage-dependent channel activation in the absence of Ca²⁺ (Figures S4G–S4L, S4O, and S4P). By contrast, the basic/acidic double mutants K428A/D879A and K428A/D884A exhibit an apparent Ca²⁺ sensitivity comparable to that of D879A and D884A, respectively (Figures 5C–5F; Table S1). The observation that neutralizing two acidic residues completely extinguishes its Ca²⁺ sensing further strengthens our conclusion that the third site binds Ca²⁺ to facilitate channel opening.

Figure 5. Differential Roles of the Third Ca²⁺ Site Residues in Ca²⁺ Binding.

(A–D) Representative recordings of the QA/EK/DR background carrying an additional double mutation at the third Ca²⁺ site: E425A/D879A (A), E425A/D884A (B), K428A/D879A (C), and K428A/D884A (D). Ca²⁺ concentrations are shown as numbers above outward current deflections (in micromolar). The membrane was held at +60 mV. Red dashed line indicates 0-pA baseline.

(E and F) Quantifications of the Ca²⁺-activated currents from double mutants at the third Ca²⁺ binding site on the QA/EK/DR background (E) and their estimated Ca²⁺ EC₅₀ (F). Structure of the third Ca²⁺ site indicates the locations of the binding residues.

(G and I) Normalized Ca²⁺ dose-response curves from mutations of K428-D879 and K428-D884 pairs on the QA/EK/DR background.

(H and J) Double-mutant cycle analyses of K428-D879 and K428-D884 pairs. The free energy changes (ΔG_(s)) were calculated to obtain the interaction energy: ΔΔG_int = ΔG₍₄₎ ΔG₍₁₎ = ΔG₍₃₎ ΔG₍₂₎. Background refers to the QA/EK/DR construct.

One-way ANOVA: in (F), p values are 0.0013 for K428A versus K428A/D879A, <0.0001 for K428A versus K428A/D884A, 0.9996 for D879A versus K428A/D879A, and 0.0001 for D884A versus K428A/D884A; in (H) and (J), p values are all <0.0001. Numbers in parentheses denote the number of individual recordings. Data are represented as means ± SEM.

The ability of K428A/D879A and K428A/D884A double mutants to retain some Ca²⁺ sensitivity (Figures 5C–5E; Table S1) allowed us to conduct double-mutant cycle analysis (Hidalgo and MacKinnon, 1995) to study the potential coupling between these charged residues (Figures 5F, 5G, and 5I). Interestingly, our analyses reveal that K428 is energetically coupled to D879 and to D884 with a significant energy of interaction (ΔΔG_int) of −1.07 ± 0.03 kcal · mol⁻¹ and −0.98 ± 0.02 kcal · mol⁻¹, respectively (Figures 5H and 5J). This suggests that, although K428 is unlikely to directly coordinate Ca²⁺, its energetically favorable interactions with D879 and D884 may serve an important structural role in stabilizing these first-shell acidic residues for Ca²⁺ binding.

Taken together, in excellent agreement with previous structural studies (Alvadia et al., 2019; Bushell et al., 2019; Paulino et al., 2017a), our functional analyses not only validate that Ca²⁺, indeed, binds to the third site in TM2 and TM10 but also unveil the distinct roles of its charged residues in mediating Ca²⁺ binding and regulating channel activation.

TM2-TM10 Interaction Can Strongly Influence TMEM16A’s Channel Gating

The multifaceted role of the third site in modulating TMEM16A’s Ca²⁺-dependent and Ca²⁺-independent channel activation prompted us to speculate that the conformation of TM2 and TM10 at this region may be key to allosterically controlling channel gating. To test this hypothesis, we performed double-cysteine substitutions to the third Ca²⁺ site and sought to assess whether the transition metal ion Cd²⁺ could form a metal bridge between TM2 and TM10 and how such restricting their conformation would impact channel activation. The readily reversible formation of Cd²⁺ metal bridge permits temporal probing of the effects from crosslinking TM2 and TM10 on channel activity (Linsdell, 2015). By screening different TM2–TM10 interhelical pairs of cysteines, we found that application of Cd²⁺ (0.5, 1, or 10 μM) largely abolishes channel activity of the double mutant E425C/D879C, but not WT or the single-cysteine mutants E425C and D879C, even under saturating 100 μM Ca²⁺ (Figures 6A–6E). The partial inhibition seen in WT channels is likely caused by Cd²⁺ competing with Ca²⁺ for binding to the orthosteric Ca²⁺ sites, as this inhibition is strongly attenuated by elevating intracellular Ca²⁺ and is rapidly reversed upon Cd²⁺ removal (Figures 6A and 6F; Figure S5). In stark contrast, Cd²⁺-mediated inhibition of E425C/D879C is much longer lasting, as evidenced by the slow current recovery rate (Figure 6F), which is indicative of a high-affinity Cd²⁺-E425C/D879C metal bridge. In fact, we found that Cd²⁺ also potently inhibits the double mutant E425H/D879C (Figures S6A, S6D and S6E), in accordance with the ability of Cd²⁺ to bridge a histidine and a cysteine (Linsdell, 2015). Cd²⁺ elicits no obvious effects on other TM2–TM10 interhelical double-cysteine mutants, including E425C/D884C and K428C/D879C (Figures S6B–S6E), presumably owing to a lack of optimal distance and/or orientation for Cd²⁺ metal bridging. The K428C/D884C double mutant shows Ca²⁺-elicited currents that are too small for analysis (Figure S6F).

Figure 6. Cysteine-Mediated Stabilization of TM2-TM10 via Cd²⁺ Metal Bridging Inhibits TMEM16A Channel Activation.

(A–D) Representative recordings indicating the effects of 1 μM Cd²⁺ on channel opening of TMEM16A WT (A), E425C/D879C (B), E425C (C), and D879C (D). The membrane was held at +60 mV. Red dashed line indicates 0-pA baseline.

(E) Quantifications of the dose-dependent Cd²⁺-mediated inhibition of TMEM16A WT, E425C/D879C, E425C, and D879C. n = 4–25 recordings for each concentration of Cd²⁺.

(F) Current recovery time constant (off-rate) of wild type (WT) and mutants after application of 1 μM Cd²⁺ in the presence of 100 μM Ca²⁺. One-way ANOVA: p values are <0.0001, 0.1208, and 0.8519 for E425C/D879C, E425C, and D879C, respectively. Numbers in parentheses denote the number of individual recordings. Data are represented as means ± SEM.

We also examined whether the inhibitory effect from chemically restraining TM2 and TM10 is dependent upon Ca²⁺ binding to orthosteric sites. By again exploiting the QA/EK/DR background construct to interrogate the effects of Cd²⁺-mediated bridging in the absence of Ca²⁺, we found that 10 μM Cd²⁺ markedly inhibits the voltage-dependent channel activity of E425C/D879C but not the Cd²⁺-insensitive E428C/D884C mutant (Figure S7). Taken together, these results reveal that altering the conformation of TM2–TM10 at the third Ca²⁺ site by bridging the substituted cysteines E425C and D879C leads to profound channel inhibition in a manner that is independent of Ca²⁺ binding to the orthosteric sites.

DISCUSSION

Here, we present functional evidence to establish that, in addition to the two orthosteric sites in the pore-gate domain, TMEM16A also possesses an additional Ca²⁺ binding site comprising E425 and K428 in TM2 and D879 and D884 in TM10 (Figure 7A). Our data not only consolidate recent structural findings of a third Ca²⁺ binding site in TMEM16F and TMEM16K scramblases (Alvadia et al., 2019; Bushell et al., 2019) but also functionally validate that the unassigned electron density observed in TM2–TM10 of the liganded TMEM16A structure, indeed, represents the third bound Ca²⁺ ion (Paulino et al., 2017a).

Figure 7. Proposed Mechanism of Allosteric Modulation via the Third Ca²⁺ Site.

(A) A proposed allosteric gating mechanism shown in a TMEM16A topological model. Ca²⁺ binding to the orthosteric sites and the third Ca²⁺ site can activate and modulate TMEM16A activation gate through short-range and long-range allosteric coupling, respectively. Movement of TM10’ is speculated based on the structural study of TMEM16K (Bushell et al., 2019).

(B) A simplified Ca²⁺-dependent gating scheme of TMEM16A describing the relationships between the activation gate, the orthosteric Ca²⁺ binding sites, and the third Ca²⁺ site. Blue arrows represent allosteric coupling among the Ca²⁺ sensors and the activation gate.

Allosteric Modulation of TMEM16A’s Gating via an Additional High-Affinity Ca²⁺ Site

In TMEM16 proteins, the pore-gate domain accommodates both the two orthosteric Ca²⁺ sites and the activation gate, thereby allowing Ca²⁺ binding to efficiently open the activation gate (Figures 7A and 7B). This structural arrangement differs from that of the Ca²⁺-activated large conductance potassium BK channels, whose two Ca²⁺ sensors are distantly located within disparate parts of the cytosolic gating ring and independently control activation via long-range allosteric coupling mechanisms (Hite et al., 2017; Horrigan, 2012; Horrigan and Aldrich, 2002; Tao et al., 2017; Yang et al., 2015; Zhou et al., 2017). Therefore, the presence of an additional Ca²⁺ site that is distant from the pore-gate domain raises the question of whether TMEM16 proteins may also utilize a long-range allosteric mechanism to regulate ion/lipid permeation gating.

To address this question and to understand the relationship between the third Ca²⁺ site and the orthosteric Ca²⁺ sites, we used three different TMEM16A backgrounds: WT, Q645A, and QA/EK/DR. First, our Ca²⁺-dependent measurements on the WT background revealed the involvement of the third Ca²⁺ site residues in channel activation (Figure 2). Second, the Q645A background allowed us to demonstrate that mutation of the third site residues alters channel activation independent of Ca²⁺ binding to the orthosteric sites, which is in support of a long-range allosteric coupling between this additional Ca²⁺ site and the activation gate. Finally, because simultaneous Ca²⁺ binding to the orthosteric sites hinders assessment of the third site’s Ca²⁺ sensitivity, we generated the QA/EK/DR construct and used it to reliably measure the Ca²⁺ sensitivity of the third site. This approach allowed us to reveal that the third Ca²⁺ site possesses a high affinity for Ca²⁺ with an apparent dissociation constant (K_D) of ~320 nM, which is well suited for sensing physiological Ca²⁺ concentrations (Figure 4B; Table S1). Our comprehensive mutagenesis experiments, in conjunction with double-mutant cycle analyses, further support that the three acidic residues E425, D879, and D884 directly participate in Ca²⁺ binding, whereas K428 serves a stabilizing structural role (Figures 4 and 5). Interestingly, in addition to impairing Ca²⁺ binding, disruption of the third Ca²⁺ site also differentially alters channel gating in a Ca²⁺-independent manner (Figure 3; Figure S2), consistent with a pleiotropic role of these residues in both mediating Ca²⁺ binding and regulating the channel’s gate.

Conformationally Restricting TM2–TM10 at the Third Site Inhibits Channel Activation

Our observation that chemically bridging E425C and D879C inhibits TMEM16A activation is both striking and counterintuitive, considering that Ca²⁺ binding to the third site promotes channel opening (Figure 4B). However, this is unsurprising, as we believe that Cd²⁺-mediated metal bridging likely elicits distinct effects from Ca²⁺ binding to an intact third site based on the following reasons. First, in the intact site, the Ca²⁺ ion is coordinated by the acidic residues E425, D879, and D884 (Figures 4 and 5) (Alvadia et al., 2019; Bushell et al., 2019), while in the double cysteine-substituted mutant, Cd²⁺ should be sandwiched between E425C and D879C. Second, given the shorter side chain of cysteine compared with those of aspartate and glutamate, the conformation resulting from Cd²⁺-E425C/D879C complex formation should, therefore, be distinct from that induced by Ca²⁺ binding to the intact site. Finally, Cd²⁺ coordination with E425C and D879C is more stable than Ca²⁺ within the third-site acidic residues, as signified by the extremely slow off-rate of Cd²⁺ that is characteristic of a cadmium-cysteine metal bridge (Figure 6) (Linsdell, 2015). Although precisely how such stabilization of TM2 and TM10 prohibits channel activation requires additional studies, our data further underscore the importance of this site in channel gating.

Potential Role of TM10’ in Allosteric Modulation by the Third Ca²⁺ Site

Our functional evidence suggests that Ca²⁺ binding to the third site can lead to long-range allosteric effects on TMEM16A activation. However, it is still unclear how Ca²⁺ binding energy at the third site is propagated to the pore-gate domain. We speculate that TM10’, a short α helix following TM10, may play an important role in mediating this allosteric coupling based on the following observations (Figure 7A). First, it was shown that replacing or truncating the C-terminal region following TM10’ after the third Ca²⁺ site at L904 (Dang et al., 2017) or K895 (Scudieri et al., 2016) markedly altered the Ca²⁺ sensitivity of TMEM16A. Second, using chemical crosslinking experiments, Scudieri et al. (2016) suggested that TM10’ may regulate channel gating via its inter-subunit coupling to the TM2–TM3 loop, which was previously implicated in TMEM16A’s voltage-dependent channel activation (Ferrera et al., 2009; Xiao et al., 2011). Finally, the recent structural study of TMEM16K suggested that TM10’ forms an inter-subunit contact with the TM2–TM3 loop (mostly β9–β10) and undergoes a pronounced swinging movement during Ca²⁺-dependent activation; this enables rearrangement of the TM2–TM3 loop and, subsequently, the N-terminal cytosolic domain (NCD), all of which culminate in the opening of TMEM16K’s lipid pathway (Bushell et al., 2019). On the basis of these structural and functional observations, we suggest that Ca²⁺ binding to the third site in TM2–TM10 likely induces a conformational change of TM10’, perhaps altering its inter-subunit interaction with the TM2–TM3 loop and, subsequently, the NCD to modulate channel opening (Figure 7A). The lack of conformational changes at the third site upon Ca²⁺ binding in the TMEM16A and TMEM16F structures likely reflects their desensitized/rundown states after Ca²⁺-dependent activation (Alvadia et al., 2019; Bushell et al., 2019; Dang et al., 2017; Feng et al., 2019; Le et al., 2019a; Ye et al., 2018). Future studies are required to fully elaborate the molecular basis of this long-range allosteric coupling pathway between the third Ca²⁺ site and the activation gate (Figure 7B).

In summary, our work builds upon previous structural studies and provides the physical basis underlying a long-range regulatory gating mechanism by which Ca²⁺ binding to the TM2–TM10 site enhances channel opening. Our present findings also suggest that structurally conserved TMEM16 ion channels and scramblases may exploit a similar mode of allosteric regulation by Ca²⁺. We believe that the additional Ca²⁺ site endows TMEM16 proteins with enhanced regulatory mechanisms, increasing their complexity necessary for diverse Ca²⁺ signaling functions in various tissues. Mechanistic insight into this Ca²⁺ binding site not only facilitates our understanding of the patho-physiology of TMEM16 mutations but may also assist future therapeutic design of allosteric modulators for the treatment of TMEM16-related diseases.

STAR★METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Huanghe Yang (huanghe.yang@duke.edu).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

This study did not generate unique datasets/code.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

Adherent human embryonic kidney 293T (HEK293T) cells, which were authenticated and tested negative for mycoplasma, were obtained from Duke Cell Culture Facility (ATCC CRL-11268). HEK293T cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, GIBCO) supplemented with 10% fetal bovine serum (FBS, Sigma), 1% penicillin/streptomycin (GIBCO) in an incubator supplied with 5% CO₂ and kept at 37°C. HEK293T cells were passaged every 2–3 days and were used within 15 passages.

METHOD DETAILS

Cloning and Cell Culture

Mouse TMEM16A (mTMEM16A, NCBI accession code NP_001229278) tagged with a C-terminal EGFP (in an N1 vector) was used for all experiments. This mTMEM16A(‘a’) isoform lacks the EAVK segment (‘c’) in the first intracellular loop. Individual point mutations of TMEM16A were generated using QuikChange site-directed mutagenesis kit (Agilent), and multiple mutations were generated sequentially. All PCR primers were purchased from IDT DNA Technologies (see Table S2 the complete list of primers used for mutagenesis). Substituted amino acid sequences were validated by Sanger sequencing (Genewiz). Glass coverslips were pre-coated with poly-L-lysine (Sigma) and laminin (Sigma) for at least 1–2 h in a 24-well plate (Eppendorf) and were seeded with freshly dissociated HEK293T cells. HEK293T cells were transiently transfected in the following day using X-tremeGENE 9 DNA transfection reagent (Sigma). Each well was typically transfected with 200 to 250 ng of total DNA plasmid to achieve sufficient expression for current recordings.

Electrophysiology

Electrophysiology recordings were performed at room temperature within 24 to 36 h after transfection. Glass electrodes were made by pulling borosilicate capillaries (Sutter Instruments) and fire-polished with a microforge (Narishige). Electrodes typically had a resistance of 2–3 MΩ when filled with recording solutions. Excised inside-out membrane patches were obtained by rapidly pulling gigaohm (GΩ) cell-attached seals from HEK293T cells expressing fluorescently tagged TMEM16A channels.

External (pipette) solution contains 140 mM NaCl, 10 mM HEPES, 2 mM MgCl₂, and pH 7.3 (adjusted with NaOH). Bath solution contains 140 mM NaCl, 10 mM HEPES, 5 mM EGTA, and pH 7.4 (adjusted with NaOH). Intracellular (perfusion) solutions with less than 100 μM Ca²⁺ contain were made by adding CaCl₂ to a solution containing 140 mM NaCl, 10 mM HEPES, 5 mM EGTA, and the amount of CaCl₂ added was calculated using WEBMAXC (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/) to achieve desired free Ca²⁺. The free Ca²⁺ concentrations were further corrected using the ratiometric Ca²⁺ dye Fura-2 (ATT Bioquest) by fitting their absorbances against a standard Ca²⁺ curve generated from a Ca²⁺ calibration kit (Biotum). For solution with 100 μM or 500 μM Ca²⁺, CaCl₂ was directly added to a solution containing 140 mM NaCl, 10 mM HEPES, pH 7.3.

Application of intracellular solutions to excised inside-out patches was carried out using a pressurized perfusion apparatus (ALA-VM8, ALA Scientific Instruments) in which the glass electrodes containing the excised patches were placed directly in front of the perfusion outlet. For measurements of the Ca²⁺-dependent dose-response behavior of TMEM16A WT and mutant channels with intact (WT or Q645A backgrounds) or disrupted (QA/EK/DR background) orthosteric Ca²⁺ sites, excised inside-out patches were held at +60 mV and TMEM16A-mediated outward currents (Cl⁻ influx) were elicited by sequential application of EGTA-buffered internal solutions containing increasing Ca²⁺ concentrations. Once the Ca²⁺-elicited currents reached a steady state, a Ca²⁺-free solution (140 mM NaCl, 10 mM HEPES, 5 mM EGTA, pH 7.3) was immediately applied to allow the channels to return the closed state (WT background channels) or the basal activity level (QA/EK/DR background channels). To ensure that current rundown does not impact our Ca²⁺-dependent measurements and quantifications, we also performed Ca²⁺-dependent recordings of the QA/EK/DR mutant using a rundown-corrected protocol akin to one reported previously (Lim et al., 2016; Ni et al., 2014). In this rundown-corrected protocol, after application of each test Ca²⁺ concentration (0.255, 0.387, 0.94, 2.26, 6 and 100 μM), saturating 500 μM Ca²⁺ was applied to obtain the reference peak current, which was used for normalizing the test Ca²⁺ current responses.

For current-voltage (I-V) recordings of the Q645A background construct-containing TMEM16A mutant channels in the absence of Ca²⁺, the excised membrane patches were held at 60 mV, and test voltage steps from 120 to +140 mV were applied to elicit channel opening in the absence of 0 Ca²⁺. The same patch was immediately exposed to 100 μM Ca²⁺ and subjected to an I-V protocol in which the membrane was held at 0 mV and stepped to +140 mV and 60 mV. The steady-state currents measured during test voltage steps were used for subsequent analysis.

For cadmium (Cd²⁺) metal bridging experiments, the desired Cd²⁺ concentrations were directly added to a solution containing 140 mM NaCl, 10 mM HEPES, 100 μM Ca²⁺, pH 7.4 using 1000× Cd²⁺ stock solutions. 1000× Cd²⁺ stock solutions (0.5, 1, 10 mM) were made from 10× serial dilutions from a 100 mM Cd²⁺ stock solution using ultrapure water (Invitrogen). Excised inside-out patches were held at +60 mV using a gap-free protocol and TMEM16A channels were activated with the control 100 μM or 500 μM Ca²⁺ solution for 2–3 s to obtain the reference peak current before being switched to a Cd²⁺-containing solution to assess current inhibition. Application of Cd²⁺-containing solution usually took 4–5 s until the appearance of the steady-state currents after which the control 100 μM or 500 μM Ca²⁺ solution was re-applied (for ~30 s) to allow current recovery from Cd²⁺-mediated inhibition.

For experiments assessing the effects of Cd²⁺ metal bridge formation in the absence of Ca²⁺, E425C/D879C(QA/EK/DR) and E425C/D879C(QA/EK/DR) channels were first activated in the absence of Ca²⁺ and immediately exposed to solution containing 10 μM Cd²⁺ (140 mM NaCl, 10 HEPES, pH 7.4) using an I-V protocol. The membrane was held at 60 mV and the channels were activated via voltage steps from 120 to +140 mV. The E425C/D884C(QA/EK/DR) was used as a negative control for its lack of Cd²⁺ sensitivity as seen in the WT background (see Figures S6B, S6D, and S6E).

All electrophysiology recordings were low-pass filtered at 5 kHz (Axopatch 200B) and digitally sampled at 10 kHz (Axon Digidata 1550A) and digitized by Clampex 10 (Molecular Devices).

Confocal microscopy

HEK293T cells were co-transfected with EGFP-tagged mouse TMEM16A mutant channels (mTMEM16A-EGFP) and OFPSpark®-tagged human phospholipid scramblase 1 (hPLSCR1-OFPSpark®, Sino Biological) at a 2:1 DNA ratio. hPLSCR1 is a single-pass transmembrane protein that favors plasma membrane expression (Sun et al., 2002). The fluorescently tagged hPLSCR1 was used as a plasma membrane marker in this study. Cells were imaged approximately 24–30 h after transfection. Fluorescent images were acquired using a Zeiss 780 inverted confocal microscope at 60x magnification.

QUANTIFICATION AND STATISTICAL ANALYSIS

Electrophysiological data analysis was done in Clampfit, Microsoft Excel and MATLAB (MathWorks). Fluorescence image analysis was done in ImageJ (https://imagej.nih.gov/ij/).

Ca²⁺-dependent channel activation

For WT and mutant channels with intact orthosteric Ca²⁺ sites, the Ca²⁺-elicited steady-state currents were normalized to the peak current (see Figure 2). For TMEM16A mutant channels with impaired orthosteric Ca²⁺ sites (QA/EK/DR containing constructs), the basal current right after each Ca²⁺ application was subtracted from the steady-state Ca²⁺-elicited current to avoid the issue with changing basal current, and these Ca²⁺-sensitive current components (see Figures 4B, 4G, and 5E) were then normalized to the peak Ca²⁺-sensitive current (see Figures 4H, 5G, and 5I). For rundown-corrected Ca²⁺-dependent measurements of the QA/EK/DR mutant (Figures S3A–S3C), the Ca²⁺-sensitive current of each test Ca²⁺ concentration was normalized to its respective reference peak current elicited by 500 μM Ca²⁺, and these reference peak-normalized values of the test Ca²⁺ concentrations were used for constructing the dose-response curves. In all cases, the sigmoidal Ca²⁺ dose-response curves were similarly constructed by fitting the normalized Ca²⁺-elicited currents using a non-linear regression curve fit with the equation:

$$\frac{1}{I_{\mathit{\text{max}}}}=\frac{1}{1+{\left(\frac{\left[E{C}_{50}\right]}{\left[{\text{Ca}}^{2+}\right]}\right)}^H}$$ (Equation 1)

where I/I_max is the normalized Ca²⁺-elicited current, [Ca²⁺] is the applied free Ca²⁺ concentration, H is the Hill coefficient, and EC₅₀ is the estimated concentration of Ca²⁺ at which I/I_max is 0.5. The EC₅₀ values from individual recordings are extracted for statistical analyses.

Cd²⁺ metal bridging experiments

The extent of Cd²⁺-mediated inhibition of TMEM16A in the presence of 100 μM Ca²⁺ was determined as:

$$\mathit{\text{Current}}\mathit{\text{ratio}}(\%)=\frac{I_{C{d}^{2+}}}{I_{\mathit{\text{peak}}}}\times 100$$ (Equation 2)

where I_peak and I_Cd2+ denote steady-state currents measured in the presence of 100 μM or 500 μM Ca²⁺ before and during Cd²⁺ application, respectively.

To assess the extent of current recovery following Cd²⁺ treatment, we determined the apparent off-rate of Cd²⁺ (Figure 6F; Figure S6E). As in the case of Cd²⁺-insensitive channels such as WT or the single cysteine mutants E425C, D879C, the rapid current recovery of inhibition following Cd²⁺ washout mostly represents the rate of displacement of Cd²⁺ from the orthosteric Ca²⁺ binding sites (Figures 6A and 6E; Figure S5). On the other hand, in the Cd²⁺-sensitive double mutant E425C/D879C, the tight association between Cd²⁺ and the substituted cysteines E425C and D879C mostly accounts for the much slower recovery rate (Figures 6B and 6E; Figure S5). As a result, we used the time constant (τ) of current recovery after application of 1 μM Cd²⁺ to quantitatively assess the effects of Cd²⁺ crosslinking by fitting the recovering current with a standard exponential fit with the equation:

$$I_{(t)}=I_0+A\left(1-e^{\frac{-t}{\tau }}\right)$$ (Equation 3)

where I_(t) represents the current at time point t, I₀ is the initial current amplitude, t is the time constant.

Current-voltage (I-V) recordings

For I-V measurements of Q645A mutant channels, we used the steady state currents obtained during the depolarization steps for analysis due to the rapid deactivation at hyperpolarization and the intrinsically low open probability of the Q645A mutant in the absence of Ca²⁺ binding to the orthosteric sites (Lam and Dutzler, 2018; Peters et al., 2018). The calculated conductance from each test voltage step measured in the absence of Ca²⁺ (first recording) was normalized to the peak conductance measured at +140 mV in the presence of 100 μM Ca²⁺ (second recording). The resulting normalized G/G_max-V curves were fit with a Boltzmann equation:

$$\frac{\text{G}}{{\text{G}}_{\mathit{\text{max}}}}=\frac{1}{1+{\text{e}}^{\frac{-zF\left(Vm-V_{1/2}\right)}{RT}}}$$ (Equation 4)

where G/G_max is the normalized channel conductance, z is the number of equivalent gating charges, V_m is the applied membrane voltage, V_1/2 is the estimated voltage at which G/G_max is 0.5, F is the Faraday’s constant, R is the gas constant, and T is the absolute temperature. Because of the low open probability of the Q645A-containing mutants in 0 Ca²⁺, their G/G_max-V curves did not go beyond 0.5 (Figure 3F). We therefore did not extract the V_1/2 values for further comparisons.

Double mutant cycle analysis

To assess the potential energetic coupling between the charged residues at the third Ca²⁺ site, we extracted the Ca²⁺ EC₅₀ values from our Ca²⁺-dependent dose-response measurements of single and double mutations of the third Ca²⁺ site in the QA/EK/DR background. Because we were able to reliably measure the Ca²⁺ dose-response curves of K428A/D879A(QA/EK/DR) and K428A/D884A(QA/EK/DR), the K428-D879 and K428-D884 residue pairs were further analyzed using double mutant cycle analysis by calculating the free energy change associated with a perturbation (Hidalgo and MacKinnon, 1995). For each pair of residues, namely X and Y, the free energy change associated with their perturbations via single or double mutations was calculated using the following equations:

$$\Delta G_{XY>X^{\prime }Y}=RT*In\left(\frac{E{C}_{50\left(X^{\prime }Y\right))}}{E{C}_{50(XY)}}\right)(\text{or}\Delta {\text{G}}_{(1)})$$ (Equation 5)

$$\Delta G_{X{Y}^{\prime }>X^{\prime }{Y}^{\prime }}=RT*In\left(\frac{E{C}_{50\left(X^{\prime }{Y}^{\prime }\right)}}{E{C}_{50\left(X{Y}^{\prime }\right)}}\right)(\text{or}\Delta {\text{G}}_{(4)})$$ (Equation 6)

$$\Delta G_{XY>X{Y}^{\prime }}=RT*In\left(\frac{E{C}_{50\left(X{Y}^{\prime }\right)}}{E{C}_{50(XY)}}\right)(\text{or}\Delta {\text{G}}_{(2)})$$ (Equation 7)

$$\Delta G_{X^{\prime }Y>X^{\prime }{Y}^{\prime }}=RT*In\left(\frac{E{C}_{50\left(X^{\prime }{Y}^{\prime }\right)}}{E{C}_{50\left(X^{\prime }Y\right)}}\right)(\text{or}\Delta {\text{G}}_{(3)})$$ (Equation 8)

where XY denotes intact residues on the background; X’Y, XY’, X’Y’ denote mutations at X, Y, and both, respectively; EC₅₀ is the estimated half-maximal activation concentration of Ca²⁺. Note that the background refers the QA/EK/DR construct on which single and double mutations of the third site residue pairs are made. In principle, ΔG₍₁₎ and ΔG₍₄₎ are related and represent the free energy change as a result of mutation of X (or X’) on the background channel (ΔG₍₁₎) or on the Y’ background (ΔG₍₄₎). Similarly, ΔG₍₂₎ and ΔG₍₃₎ represent the free change associated with Y’ mutations on background and X’ background, respectively. If X and Y residues are energetically coupled, the free energy change associated with the double mutations–X’ mutation on Y’ background (ΔG₍₄₎) or Y’ mutation on X’ background (ΔG₍₃₎)–would differ from that associated with the single mutations X’ and Y’, or ΔG₍₁₎ and ΔG₍₂₎, respectively. Please also refer to Figures 5G–5J for the schematic diagrams of our double mutant cycle analyses. To obtain the standard error for the interaction energy (ΔΔG_int), the individually calculated free energy change (ΔG_(s)) values were used as:

$$\Delta \Delta G_{int}=\Delta G_{(4)}-\Delta G_{(1)}=\Delta G_{(3)}-\Delta G_{(2)}$$ (Equation 9)

The calculated ΔΔG_int values were subsequently confirmed using the equation:

$$\Delta \Delta G_{int}=RT*In\left(\frac{E{C}_{50\left(X^{\prime }{Y}^{\prime }\right)}\cdot E{C}_{50(bg)}}{E{C}_{50\left(X^{\prime }Y\right)}\cdot E{C}_{50\left(X{Y}^{\prime }\right)}}\right)$$ (Equation 10)

Structural and sequence analysis

PDB coordinate files were obtained from the Protein Data Bank (https://www.rcsb.org/). Protein structure figures were prepared in Pymol (Schrödinger, Inc.). Protein sequence alignment was done in UniProt (https://www.uniprot.org/align/).

Statistical analysis

Statistical analyses were performed in Prism software (GraphPad). Two-tailed Student’s t test was used for comparisons between two groups (paired or unpaired), and one-way ANOVA was used for comparisons between multiple data groups. Comparisons with p values < 0.05 are deemed to be statistically significant. Data in summary bar graphs are represented as mean ± standard deviation (SD) or mean ± standard error of the mean (s.e.m.), and each data point represents an independent recording. *, **, *** and **** represent statistical significance corresponding to p value < 0.05, < 0.01, < 0.001, < 0.0001, respectively.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
Escherichia coli (MAX Efficiency DH5α competent cells)	Invitrogen	Cat#18258012
Chemicals, Peptides, and Recombinant Proteins
Opti-MEM Reduced Serum Medium	GIBCO	Cat#11058021
Poly-L-lysine (PLL)	Sigma-Aldrich	Cat#P2636
Laminin	Sigma-Aldrich	Cat#L-2020
X-tremeGENE 9 DNA Transfection Reagent	Sigma-Aldrich	Cat#6365787001
Fura-2 pentapotassium salt	ATT Bioquest	Cat#21025
Critical Commercial Assays
Calcium Calibration Buffer Kit	Biotum	Cat#59100
Experimental Models: Cell Lines
Human Embryonic Kidney (HEK293T) cells	ATCC	RRID: CVCL_0045
Oligonucleotides
Custom-designed primers for mutagenesis of mTMEM16A constructs (see Table S2)	IDT-DNA Technologies	N/A
Recombinant DNA
TMEM16A-pEGFP-N1 from Mus musculus	(Tien et al., 2014)	N/A
mTMEM16A E425A	This paper	N/A
mTMEM16A K428A	This paper	N/A
mTMEM16A D879A	This paper	N/A
mTMEM16A D884A	This paper	N/A
mTMEM16A E425K	This paper	N/A
mTMEM16A K428E	This paper	N/A
mTMEM16A D879K	This paper	N/A
mTMEM16A D884K	This paper	N/A
mTMEM16A Q645A	(Peters et al., 2018)	N/A
mTMEM16A E425A/Q645A	This paper	N/A
mTMEM16A K428A/Q645A	This paper	N/A
mTMEM16A D879A/Q645A	This paper	N/A
mTMEM16A D884A/Q645A	This paper	N/A
mTMEM16A E425K/Q645A	This paper	N/A
mTMEM16A K428E/Q645A	This paper	N/A
mTMEM16A D879K/Q645A	This paper	N/A
mTMEM16A D884K/Q645A	This paper	N/A
mTMEM16A Q645A/E701K/D734R	This paper	N/A
mTMEM16A E425A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A K428A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A D879A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A D884A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A E425A/K428A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A E425A/D879A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A E425A/D884A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A K428A/D879A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A K428A/D884A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A D879A/D884A/Q645A/E701K/D734R	This paper	N/A
mTMEM16A E425C	This paper	N/A
mTMEM16A K428C	This paper	N/A
mTMEM16A D879C	This paper	N/A
mTMEM16A D884C	This paper	N/A
mTMEM16A E425C/D879C	This paper	N/A
mTMEM16A E425C/D884C	This paper	N/A
mTMEM16A K428C/D879C	This paper	N/A
mTMEM16A K428C/D884C	This paper	N/A
mTMEM16A E425H/D879C	This paper	N/A
pCMV3-hPLSCR1-OFPSpark®	Sino Biological	Cat#HG16877-ACR
Software and Algorithms
PyMOL	Schrödinger	https://pymol.org/2/
Axon PClamp 10 software	Molecular Devices	https://www.moleculardevices.com/
MATLAB	Mathworks	https://www.mathworks.com/products/matlab.html
Excel	Microsoft Office	N/A
Prism 8	GraphPad	https://www.graphpad.com/scientific-software/prism/
Illustrator	Adobe	https://www.adobe.com/products/illustrator.html
WEBMAXC (metal chelation prediction software)	Chris Patton, Stanford University	https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/
Uniprot Sequence Alignment	UniProt	https://www.uniprot.org/align/

Highlights.

• TMEM16A possesses a third high-affinity Ca²⁺ binding site between TM2 and TM10

• Binding of Ca²⁺ to the third Ca²⁺ site allosterically promotes channel opening

• Neutralizing coordinating residues abolishes Ca²⁺ binding to the third site

• An engineered Cd²⁺ metal bridge at the third Ca²⁺ site inhibits channel activation