Voltage-Dependent Protonation of the Calcium Pocket Enable Activation of the Calcium-Activated Chloride Channel Anoctamin-1 (TMEM16A)

Anoctamin-1 (ANO1 or TMEM16A) is a homo-dimeric Ca2+-activated Cl− channel responsible for essential physiological processes. Each monomer harbours a pore and a Ca2+-binding pocket; the voltage-dependent binding of two intracellular Ca2+ ions to the pocket gates the pore. However, in the absence of intracellular Ca2+ voltage activates TMEM16A by an unknown mechanism. Here we show voltage-activated anion currents that are outwardly rectifying, time-independent with fast or absent tail currents that are inhibited by tannic and anthracene-9-carboxylic acids. Since intracellular protons compete with Ca2+ for binding sites in the pocket, we hypothesized that voltage-dependent titration of these sites would induce gating. Indeed intracellular acidification enabled activation of TMEM16A by voltage-dependent protonation, which enhanced the open probability of the channel. Mutating Glu/Asp residues in the Ca2+-binding pocket to glutamine (to resemble a permanent protonated Glu) yielded channels that were easier to activate at physiological pH. Notably, the response of these mutants to intracellular acidification was diminished and became voltage-independent. Thus, voltage-dependent protonation of glutamate/aspartate residues (Glu/Asp) located in the Ca2+-binding pocket underlines TMEM16A activation in the absence of intracellular Ca2+.


Voltage-dependent protonation of TMEM16A enables activation. Under physiological [Ca 2+ ] i
and [H + ] i , TMEM16A is gated by Vm-dependent binding of intracellular Ca 2+ to the Ca 2+ -binding pocket [22][23][24][25] . However, intracellular acidification in the presence of Ca 2+ inhibited Ca 2+ -activated Cl − currents in salivary acinar cells and HEK-293 cells expressing TMEM16A by competing for high-affinity binding sites in the Ca 2+ -binding pocket 32,33 . Based on this observation we hypothesized that positive Vm could drive intracellular H + into the Ca 2+ -binding pocket to protonate Glu/Asp residues, open the channel and thus generate I Cl,Vm . To test this idea, we recorded I Cl,Vm from TMEM16A-expressing HEK-293 cells dialyzed with an internal solution containing 25.24 mM EGTA/0 Ca 2+ . The pH of this solution was adjusted to 8.0, 7.3, 6.0, 5.0, and 4.0 thus changing the [H + ] i by four-orders of magnitude. To ensure that pH i remained constant during our recordings, we increased the buffer capacity of the solutions by adjusting the pH with 50 mM of bicine (8.0), HEPES (7.3), MES (6.0, 5.0, and 4.0), citric acid (5.0), or tartaric acid (4.0). No difference in channel activation was observed between data obtained with different buffers at the same pH. Figure 2A shows I Cl,Vm recorded at −100, −40, +20, +60, +120 and +160 mV using the protocol shown in Fig. 1G. Each row shows I Cl,Vm recorded at the indicated pH (left column), traces are representative of 5 independent experiments. As pH i decreased from 8.0 (top) to 4.0 (bottom), the magnitude of I Cl,Vm increased. Under acidic conditions I Cl,Vm activated and deactivated very rapidly. At +160 mV, the time constant of activation was 0.24 ± 0.02 ms at pH i = 4 (n = 5). Tail currents were nearly absent but at pH i 4.0 a small and fast inward current became evident (inset shows a magnification of the tail current). The corresponding I Cl,Vm -Vm relationships are shown on the right column of Fig. 2A. At −100 mV a reduction of pH i from 6 to 4.0 produced a 5-fold increase in I Cl,Vm (−1.5 ± 0.4 to −7.5 ± 2.7 pA/pF, n = 5. Lastly, no currents were recorded from HEK-293 cells transfected with the empty vector or dialyzed with pH i 4.0 containing EGTA or BAPTA (Supplementary Fig. S2). Hence, the potentiation of I Cl,Vm under acidic conditions resulted from activation of TMEM16A. Unfortunately, the effect of intracellular H + on the Vm-dependent activation of I Cl,Vm cannot be deduced from the analysis of macroscopic conductance (G = I Cl,Vm /Vm-Vr) vs Vm at different [H + ] i for two reasons. First, the reversal potentials (Vr) measurements at pH i ≥ 6.0 were unreliable due to I Cl,Vm rectification (at pH i 5 and 4, reversal potentials were −34.4 ± 2.3 and −31.4 ± 0.8 mV, respectively, closer to the expected reversal potential for Cl − ). Second, the instantaneous current-voltage relationship after a depolarization to +160 mV was not linear (Supplementary Fig. S3, blue) indicating that I Cl,Vm does not follow Ohm´s law and the tail current magnitudes at −100 mV (orange) were the same for steps between −100 to +160 mV as if the open probability was Vm-independent.
To advance the hypothesis that Vm activation of TMEM16A is due to protonation we must show that the equilibrium constant of protonation (K) is Vm-dependent. Figure 2B show I Cl,Vm titration curves at different Vm. As the [H + ] i increase the magnitude of I Cl,Vm is increased. Unfortunately, only at +160 mV and pH i 4.0, we obtained a hint of saturation. Experiments at pH i 3.0 were not possible because the cells died quickly. Thus, to determine the The voltage protocol used to activate TMEM16A consisted of a holding potential of −30 mV, 250 or 500 ms steps between −100 to +160 mV in 20 mV increments, and unless otherwise indicated, a repolarization Vm of −100 mV. equilibrium constant of protonation (K) at each Vm, we first fitted the data collected at +160 mV to a Hill equation (Eq. 1) to obtain an I Cl,Vm maximum. Then we used this value in the Hill equation to fit all the curves. This way, our fitting procedure had only two free parameters, K and Hill coefficient (N). Continuous lines in Fig. 2B and their corresponding R 2 values show that this fitting procedure is a good quantitative description of the experimental data at all Vm. Figure 2C displays the resulting pKa (= −log 10 K) values as a function of Vm. At positive Vm, the pKa value increased indicating that less H + are required to generate 50% of I Cl,Vm . This Vm dependence of pKa resembles the Vm dependence of EC 50 for intracellular Ca 2+ 22 . The relation pKa -Vm was fit with Eq. 2 to www.nature.com/scientificreports www.nature.com/scientificreports/ calculate K 0 (equilibrium constant of protonation at 0 mV and δ, the fraction of electrical field sensed by H + going from the intracellular side to the extracellular side). A pK 0 value of 0.97 ± 0.07 and a δ value of 1.39 ± 0.04 were estimated this way. We conducted similar experiments with inside-out patches; unfortunately, most patches were unstable at acidic conditions. Figure 2C (blue symbols) shows averaged data obtained from two super-patches ( Supplementary Fig. S4) that withstood exposure to pH i 8.0, 6.0, 5.0, and 4.0. The pKa obtained from inside-out patches has a similar Vm-dependence as the pKa obtained from whole cell recordings. In this case, the average pK 0 and δ values were 1.77 and 0.79, respectively. Thus, taken together our results support the hypothesis that Vm gates TMEM16A due to intracellular protonation.
Voltage-dependent protonation of glutamate and aspartate residues located in the ca 2+ -binding pocket increase the open probability of TMEM16A. To identify Glu and Asp residues within the Ca 2+ -binding pocket that are targets of intracellular H + we mutated in an incremental fashion 4 Glu and 1 Asp residues. These residues were mutated into Gln to resemble a permanent protonated Glu side chain. We reasoned that if H + neutralizes the COO − group of Glu and Asp residues to activate wild type TMEM16A (WT) then Vm would activate Gln mutants at pH i 7.3 without the need for acidification. Figure 3A shows I Cl,Vm www.nature.com/scientificreports www.nature.com/scientificreports/ recordings at −100, −40, +20, +60, +120 and +160 mV from five independent cells expressing (from top to bottom) WT, E702Q/E705Q (2 M), E702Q/E705Q/E734Q (3 M), E702Q/E705Q/E734Q/D738Q (4 M), and E654Q/ E702Q/E705Q/E734Q/D738Q (5 M) channels. All cells were dialyzed with 25.24 mM EGTA/0 Ca 2+ and pH i 7.3. As the number of mutations accumulates, the magnitude of I Cl,Vm increased mirroring the effect of intracellular acidification. I Cl,Vm onset was fast, had no time dependence and no tail currents at −100 mV. Very fast tail currents were observed only with the quintuple mutant (see magnification in the lower-left panel). The corresponding I Cl,Vm -Vm relationships (Fig. 3A, right column) displayed outward rectification just like WT channels but the rectification changed without showing a particular pattern. Although the magnitude of I Cl,Vm increased at all Vm, at +160 mV I Cl,Vm appeared to saturate as the number of mutations (or Gln) in the Ca 2+ -binding pocket increased ( Fig. 3B). At pH i 7.3, a strong depolarization (+160 mV) induced a 6.5-fold increase in I Cl,Vm in 5 M channels compared to WT.
The above data show that mutating the acidic residues of the Ca 2+ -binding pocket into Gln allowed activation of TMEM16A at pH i 7.3 in the absence of intracellular Ca 2+ . If these residues are the main source of Vm sensitivity under acidic conditions, then increasing the [H + ] i should have little or no additional effect on I Cl,Vm , and the Vm dependence of pKa should be abolished. To test these predictions, we recorded I Cl,Vm from cells dialyzed with an intracellular solution with pH i 4. Figure   www.nature.com/scientificreports www.nature.com/scientificreports/ Despite this limitation and contrary to the result obtained at pH i 7.3 (Fig. 3B), Vm activation of I Cl,Vm at pH i 4.0 was strongly weakened but not eliminated. Depolarizing to +160 mV induced a 1.4-fold increase in I Cl,Vm generated by 5 M channels compared to WT at pH i 4.0. Thus, mutating 4 Glu and 1 Asp located in the Ca 2+ -binding pocket of TMEM16A reduced but did not completely abolish the ability of intracellular H + to increase I Cl,Vm .
To investigate if the Vm dependence of pKa was eliminated by mutating the acidic residues we performed concentration-response experiments in 5 M channels. Figure 4C shows titration curves constructed using the I Cl,Vm recorded between +60 to +160 mV from the whole cell (left) or inside-out patches (right). Unlike WT channels (Fig. 2B), the 5 M mutant channels display nearly overlapping titration curves at different Vm indicating less Vm dependence. This behaviour was more evident for inside-out patch data possibly because we were able to test all [H + ] i s in each patch. Curves were fit with the Hill equation to calculate pKa. Figure 4D (purple = whole cell; green = inside-out patches) shows pKa values against Vm. pKa values fall in a range comprising 4.7 to 6 (note a smaller range for inside-out patch data), indicating that residues other than Glu and Asp are being protonated. These pKa -Vm curves were fitted with Eq. 2 to estimate pK 0 and Vm sensitivity. pK 0 /δ values were 4.69 ± 0.08/0.13 ± 0.04 (inside out) and 3.91 ± 0.14/0.70 ± 0.07 (whole cell). Thus, compared to WT (replotted in orange) protonation in 5 M channels is less Vm-dependent. This loss of Vm dependence is also illustrated by the fact that this channel is partially open at a low [H + ] i . Thus, mutating the acidic residues decreased the Vm dependence of TMEM16A titration lending support to the idea that the Ca 2+ -binding pocket is the target of intracellular H + . In addition, the activity of TMEM16A-5M channels was enhanced under acidic conditions by a weakly Vm-dependent protonation mechanism.
To further understand the mechanism by which intracellular H + activates TMEM16A, we investigated whether protonation of the Ca 2+ -binding pocket increases the open probability. To this end, we recorded I Cl,Vm -Vm curves from inside-out patches consecutively exposed to pH i 7.3 and 4.0. We used a ramp protocol that changed the Vm between −100 to +200 mV in 635 ms. pH i 4.0 was chosen because under this acidic condition 64% of the time the Glu side chain will be in the protonated state, assuming the pKa value of Glu is 4.25. Figure  7.3 for 5 M channels was Vm-independent (orange), albeit acidification produced a 3.5-fold increase in the open probability in the −100 to +200 mV range. Therefore, protonation is enough to grant Vm-dependent activation to TMEM16A channels in the absence of intracellular Ca 2+ . Also, the titration data indicates that TMEM16A activity is enhanced by Vm-independent protonation.
The data described show that the activation of TMEM16A in the absence of intracellular Ca 2+ is enhanced by mutating the acidic residues of the Ca 2+ -binding pocket. We took advantage of the strong activation of 4 M and 5 M channels at pH i 7.3 to record tail currents to analyse the effect of acidification on voltage-dependent activation. We recorded tail currents at +100 mV using a P/8 protocol in cells expressing WT, 4 M and 5 M channels dialyzed with  Fig. 6C. By comparing this result with that obtained at pH i 4.0 (closed blue symbols) we can conclude that intracellular protons shifted the Vm activation by more than −100 mV.

Discussion
TMEM16A gating in the absence of intracellular Ca 2+ can be prompted by strong depolarizations, an elevation in temperature, mutating residues Ile637 and Gln645 in the sixth transmembrane segment (Ile641 and Gln649 in our clone) and deleting EAVK segment in the first intracellular loop 23,[36][37][38] . To explain these results, an intrinsic Vm sensitivity in TMEM16A channels has been proposed 37 . However, in TMEM16A an obvious Vm-sensing domain is lacking. Since intracellular acidification reduces TMEM16A activity due to competition between H + and Ca 2+ for the Ca 2+ -binding pocket 32,33 , we hypothesized that TMEM16A could be gated by Vm-dependent protonation of Glu and Asp residues within the pocket. This is indeed what we found. Our data do not rule out the presence of voltage-sensing domains, instead reveals an unexpected source of Vm dependence, namely Vm dependent protonation. As we acidified the cytosolic side of TMEM16A a large outward I Cl,Vm that lacked time dependence was activated by depolarizations in the absence of intracellular Ca 2+ . The equilibrium constant of protonation and the apparent open probability both increased in a Vm-dependent manner. These effects were reduced after mutating Glu and Asp residues from the Ca 2+ -binding pocket, confirming that H + titrate these residues. Nevertheless, in TMEM16A-5M a channel with all the residues mutated, intracellular acidification enhanced channel activity although the effect was Vm-independent. The estimated pKa of this secondary activation is between 4.5 and 6 suggesting protonation of His and probably Asp residues located in the cytosolic side. Together, our data is consistent with a mechanism depicted by the Scheme in Fig. 7.
We propose that intracellular H + ions interact in a Vm-independent manner (K 1 = β 1 /α 1 ) with acidic residues (shown in red) located in the cytosolic side of TMEM16A (shown in blue light embedded in a grey membrane; modified from 5OYB 25 ) outside the electrical field. A depolarizing stimulus will push H + into the electrical field where they can interact with the acidic residues of the Ca 2+ -binding pocket (K 2 = β 2 /α 2 ). Once these residues are protonated (shown in navy), the channels reach the conductive state through a Vm dependent transition (K 3 = β 3 / α 3 ) and generate I Cl,Vm .
The effects of intracellular H + observed on TMEM16A are comparable to those reported on human SLO1 BK potassium channel 39 . Both channels are activated in the absence of intracellular Ca 2+ by intracellular acidification www.nature.com/scientificreports www.nature.com/scientificreports/ and the targets are Ca 2+ sensing residues. Interestingly, intracellular H + target two His residues, as well as one Asp residue located within the RCK1 domain of BK channels; mutating the His residues, abolished the activating effect of H + . In agreement with this, the equilibrium constant of BK protonation has a value of about 6.5 at +100 mV and displays shallow Vm dependence 39 . Here intracellular H + activates TMEM16A by Vm-dependent and independent mechanisms. The Vm-independent pKa has a value of about 5.0 at +100 mV, which may suggest titration of Glu, Cys or His residues 40 just like in BK channels.
Full activation of TMEM16A by intracellular Ca 2+ is achieved by neutralization of the electrostatic potential generated by the acidic residues of the Ca 2+ -binding pocket and the subsequent movement of TM6 towards TM8 30 . This process seems to be assisted by phosphatidylinositol 4,5 bisphosphate [41][42][43] . Neutralization of the electrostatic potential by Ca 2+ is illustrated in TMEM16A Gly644Pro mutant channels where intracellular Ca 2+ can abolish the strong outward rectification displayed by these channels 30 . In the "constitutively protonated" TMEM16A 5 M channel the electrostatic potential has been neutralized, however, the channel was still activated by Vm and showed outward rectification. This implies that neutralizing the electrostatic potential is not sufficient to abolish rectification. This idea is supported by the strong outward rectification observed under intracellular acidic conditions, which should partially or abolish the electrostatic potential. Alternatively, TM6 may remain bound to TM4 in TMEM16A 5 M thus inducing rectification.
The present work together with a previous report from our group 44 shows that TMEM16A is the target of extraand intracellular H + . In both cases, protonation of acidic residues enhanced the open probability of the channel albeit extracellular H + do so independently of Vm and Ca 2+ . What would be the physiological consequences of TMEM16A regulation by protons? Although we cannot answer this question yet, we envision that this regulatory process would be important to both TMEM16 channels and scramblases since the residues targeted by H + are present in these proteins. Cancer cells overexpress TMEM16A channels and experience a large pH gradient 20,21,45,46 , conditions that facilitate cell migration and cancer progression. In these cells, the extracellular side is acidic whereas the cytosol is alkaline, and this favours TMEM16A activity. A key salient property of this regulation is that increasing the [H + ] e or the [H + ] i increases the current size without changing the fast kinetics. The activation of TMEM16A by Vm in the absence of intracellular Ca 2+ occurred in less than 1.0 ms, well within the time scale of the electrical activity of excitable cells. Rapid activation of TMEM16A can regulate the electrical activity by inducing membrane depolarization or by accelerating action potential repolarization; this has been shown in neurons from dorsal root ganglia, cholinergic neurons of the medial habenula and muscle cells 36,47 . Thus, activation of TMEM16A enables neurons to respond to thermal stimulus, control anxiety-related behaviour 36,47 and increase the frequency of action potentials in skeletal muscle cells of zebrafish 48 . A more direct physiological role for TMEM16A regulation by protons is suggested by its simultaneous activation with H + ATPase in the apical membrane of proximal tubules of mouse kidney 49 . In this scenario, a parallel Cl − flux via TMEM16A would serve as a counter ion for H + transport by the V-ATPase. Also, it is interesting to notice that the dimeric channels TMEM16A and CLC-0 are both activated by intracellular to protons 50 . In conclusion, we propose that intracellular H + endow TMEM16A with Vm gating in the absence of intracellular Ca 2+ by a mechanism that includes Vm-independent titration of cytosolic residues and Vm-dependent titration of acidic residues located in the Ca 2+ -binding pocket.  . Schematic representation of the proposed mechanism of TMEM16A activation by voltageindependent and voltage-dependent protonation. We advocate that voltage activation of TMEM16A in the absence of intracellular Ca 2+ by protonation proceeds in three steps. The first and second are voltageindependent and voltage-dependent protonation steps, respectively. Channel opening is achieved by a voltagedependent transition occurring during the last step. α 1 , α 2 , and α 3 are forward rate constants whereas β 1 , β 2 and β 3 are backward rate constants. TMEM16A is depicted in blue embedded in a grey membrane. Red dots are intracellular un-protonated acidic residues that are potential targets of intracellular protons. Residues outside the electrical field are protonated in a voltage-independent manner (navy). Inside the black rectangle are the acidic residues of the Ca 2+ pocket (one subunit) that are protonated in a voltage-dependent manner.

Methods
1 µg/µl cDNA using Polyfect transfection reagent (QIAGEN), according to the manufacturer's instructions. Cells were used 12 h after transfection. For whole cell recordings, we seeded cells at low density whereas for inside-out recordings stably transfected cells with TMEM16A or transiently transfected mutants were plated onto poly-llysine coated coverslips. chloride current recordings by patch clamp. Vm-activated macroscopic chloride (Cl − ) currents (I Cl,Vm ) were recorded at room temperature (21-23 °C) from whole cells or inside-out patches expressing wild type (WT) or mutant TMEM16A channels using the patch clamp technique as we previously reported 44,51 . We selected EGFP fluorescent HEK-293 cells using an inverted microscope equipped with UV illumination. Borosilicate patch pipettes were fabricated using P-97 electrode puller (Sutter Instruments CO.). The electrode resistance was 3-5 MΩ for whole cell or 1-2 MΩ for inside-out patches. The stimulation protocol consisted of Vm steps from −100 to +160 mV delivered every 7 s from a holding potential of −30 mV, followed by a repolarization potential to −100 or +100 mV. I Cl,Vm from inside-out patches were recorded using a 635 ms Vm ramp (−100 and +200 mV). Data were acquired using an Axopatch 200B amplifier and the pClamp10 software (Molecular Devices). I Cl,Vm was filtered at 5 kHz and digitized at 10 kHz while the bath was grounded using 3 M KCl agar-bridge connected to an Ag/AgCl reference electrode. Solutions were applied using a home-made gravity perfusion system.  50 where I max and I min are the maximum and minimum response, IC 50 is the concentration of inhibitor ([B]) needed to obtain half I max − I min inhibition; it also represents the equilibrium constant of protonation K, and N is the Hill coefficient. For the Vm dependence of K, we fitted titration curves at different Vm, the corresponding K was converted to pK (−log K) and plotted against Vm. The curve was fit with Eq. 2 53-55 : where pK 0 is the effective pH i needed to obtain half-activation when Vm = 0 mV, R is the gas constant, T is absolute temperature, F is Faraday's constant, z is the charge, δ is the electrical distance from the inside. To obtain the V 0.5 value of TMEM16A-WT, the current-voltage curve was adjusted to the following equation: where Vm is the clamping voltage, Vr is the reversal potential, V 0.5 is the voltage at which the 50% of channels were activated, G max is the estimated maximum conductance and dx is the Vm sensitivity. We estimate the liquid junction potentials using the Clampex routine of pClamp and used them to correct the reversal potential values.