Current-direction/amplitude-dependent single channel gating kinetics of mouse pannexin 1 channel: a new concept for gating kinetics

The detailed single-channel gating kinetics of mouse pannexin 1 (mPanx1) remains unknown, although mPanx1 is reported to be a voltage-activated anion-selective channel. We investigated characteristics of single-channel conductances and opening and closing rates of mPanx1 using patch-clamp techniques. The unitary current of mPanx1 shows outward rectification with single-channel conductances of ~20 pS for inward currents and ~80 pS for outward currents. The channel open time for outward currents (Cl− influx) increases linearly as the amplitude of single channel currents increases, while the open time for inward currents (Cl− efflux) is constant irrespective of changes in the current amplitude, as if the direction and amplitude of the unitary current regulates the open time. This is supported by further observations that replacement of extracellular Cl− with gluconate− diminishes the inward tail current (Cl− efflux) at a membrane potential of −100 mV due to the lowered outward current (gluconate− influx) at membrane potential of 100 mV. These results suggest that the direction and rate of charge-carrier movement regulate the open time of mPanx1, and that the previously reported voltage-dependence of Panx1 channel gating is not directly mediated by the membrane potential but rather by the direction and amplitude of currents through the channel.

The gating mechanism of Panx1 channels is of great interest. The whole-cell currents of Panx1 are augmented by membrane depolarization 8,26,40 and the open probability of single mPanx1 channel also increases in a voltage-dependent manner 40 , implying that Panx1 gating is regulated by membrane potential. Because of its voltage-dependent gating, one may expect that the transmembrane domains of mPanx1 contain charged amino acids as a voltage sensor. However, in fact, its transmembrane domains lack such a canonical voltage sensor 41 . Thus, the mechanism of the voltage-dependent gating of Panx1 remains unknown.
In this study, we examined the conductance and gating properties of single mPanx1 channels expressed in HEK293 cells using the patch-clamp technique. Our observations indicate that the open time of the channel for outward currents increases linearly as the amplitude of single channel currents increases, while the open time of the channel for inward currents is constant irrespective of changes in the amplitude of single channel current. These results suggest that the direction and amplitude of currents regulates the open time (closing rate) of mPanx1 channels, and that the previously reported voltage-dependence of Panx1 channel gating is not directly due to the membrane potential but rather due to the direction and amplitude of the single channel current driven by membrane potential. This concept is supported by our observation of reduced inward tail currents (Cl − efflux) caused by diminution of outward currents after replacement of extracellular Cl − with gluconate − (to which Panx1 channels are less permeable than Cl − ). The present study proposes a novel concept that the direction and magnitude of ion flux through the channel pore regulates the gating. These data have been partly reported in abstract form 42 .

Materials and Methods
Cell culture and transfection. Human  Single-channel current recordings. All experiments were performed in the cell-attached configuration of the patch-clamp technique at room temperature (23~25 °C). The pipette solutions contained (in mM) 140 NaCl (140 NaF, 140 NaNO 3 , 140 NaBr, 140 NaSCN or 140 Na-gluconate), 2 KCl, 1 CaCl 2 , 1 MgCl 2 and 10 HEPES (pH 7.4 adjusted with NaOH). The bath solution contained 140 NaCl, 2 KCl, 1 CaCl 2 , 1 MgCl 2 and 10 HEPES (pH 7.4 adjusted with NaOH). Borosilicate glass pipettes (Drummond Scientific Co., Broomall, PA, USA) were pulled using a pipette puller (PP-830, Narishige, Tokyo, Japan) and polished with a microforge (MF-9, Narishige, Tokyo, Japan) to a diameter corresponding to a pipette resistance within a 7.2~9.2 MΩ range in the NaCl pipette solution. Currents were amplified with an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA). Data were acquired at 5 kHz with a Digidata 1322A interface using pCLAMP 10 acquisition software (Molecular Devices, Sunnyvale, CA, USA), and stored in a personal computer. The single channel current traces were low-pass filtered at 1 kHz through a Bessel filter. Single-channel mPanx1 channel currents occur in bursts ( Fig. 1B and C) with fast open-closed kinetics within bursts (Fig. 1D). In the present study, we determined the unitary conductance from the slope of the current-voltage relationship and the single channel gating kinetics in cell-attached patches by analyzing the single-channel currents within bursts.
Open probability (P o ) of a single channel. Single-channel currents were analyzed using pCLAMP 10 software (Molecular Devices, Sunnyvale, CA, USA). We determined the number of active channels per patch membrane (N) as reported in our previous study 43 . We considered "N" as the true number of active channels per a patch membrane with a probability <0.001 that the patch contained more than "N" channels 43 . The open probability (P o ) was calculated from the cell-attached patch-clamp recordings in a patch membrane using the following equation.
where T oi is the total time of a channel staying at the open state "i" (i means the number of channels simultaneously opened: e.g., "i = 2" means that only 2 channels simultaneously opened) during the total recording time period (T R ) within bursts.
Closed and open times of a single channel. To determine mean closed and open times of a single channel, we created histograms for closed and open dwell times within bursts obtained from single channel recordings where the patch membrane was considered to contain only one active channel with a probability <0.001 that the patch membrane contained more than one active channel as described in our previous study 43  Opening and closing rates of a single channel. The opening and closing rates were calculated from the single-channel recordings in a patch membrane using the following equations. Whole cell current recordings. Whole cell currents were obtained with 3.0~4.5 MΩ borosilicate glass pipettes from Neuro2a cells transiently transfected with mPanx1Δ371. Cells were continuously perfused (~3 mL/min) at room temperature (23~25 °C) with a bath solution (pH 7.4) containing (in mM) 140 NaCl, 5 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 glucose and 10 HEPES. To substitute extracellular Cl − with gluconate − , NaCl was replaced with equimolar Na-gluconate. The intracellular solution contained 140 CsCl, 5 NaCl, 5 MgCl 2 , 5 EGTA and 10 HEPES (pH 7.4). mPanx1Δ371-expressing cells were identified by GFP fluorescence. Whole-cell currents were recorded with an Axopatch 1D amplifier and a Digidata 1321A interface using pCLAMP 8 acquisition software (Molecular Devices, Sunnyvale, CA, USA). The current data were low-pass filtered at 500 Hz with a four-pole Bessel filter and digitized at 2 kHz. Cells were held at −100 mV and ramp voltage pulses (−100 mV to +100 mV over 500 ms; +100 mV for 125 ms) were applied at 5 s intervals.

Statistical analysis.
All values are shown as mean ± SEM. The Student's t-test was used for statistical analysis as appropriate. P value < 0.05 was considered as a statistically significant difference.

Results
Single-channel current and unitary conductance of mPanx1 channel. The single channel activity of mPanx1 was examined during application of negative and positive pipette potentials in cell-attached membrane patches in HEK293 cells. Figure 2 shows representative single-channel current traces of mPanx1 channels in the presence of various anions (Cl − , F − , NO 3 − , Br − , SCN − and gluconate − ) in the pipette solution at −Vp of ±80 mV (−Vp is the displacement of the patch potential from the resting intracellular potential, e.g. −Vp = +80 mV means that the patch membrane was clamped at +80 mV more positive than the resting membrane potential, while −Vp = −80 mV means that the patch membrane was 80 mV more negative than the resting membrane potential; −Vp of 0 mV means that the patch membrane was clamped at the resting membrane potential). The single-channel conductance (~80 pS) for outward Cl − currents is consistent with previous reports 26,40 . Such channels were not observed in mock-transfected HEK293 cells.
Apparent voltage dependent gating of mPanx1. Representative single-channel currents of mPanx1 recorded at −Vp of ±40, ±60 and ±80 mV obtained from cell-attached patches with the pipette and bath solutions containing Cl − are shown in Fig. 4A. P o increased with membrane depolarization (Fig. 4B). The plots were fitted with Boltzmann equation to obtain the apparent half maximal activation voltage (V 1/2 ) and the number of gating charges. The apparent V 1/2 of the channel in the presence of Cl − was at −Vp of 34.3 ± 5.8 mV (i.e. 34.3 ± 5.8 mV more positive potential from the resting potential; n = 8). The apparent number of gating charges was 0.89 ± 0.07 (n = 8). Thus, it appears that mPanx1 gating is dependent on membrane voltage despite a lack of canonical voltage sensors, suggesting that there are as-yet-unknown mechanisms in the regulation of mPanx1 gating. Current-direction/amplitude-dependency of opening and closing rates of mPanx1 channel at various voltages. Figure 5A   fitted with a single exponential. The opening and closing rates at various voltages are shown in Fig. 5C and D. The opening rate monotonically increased within the voltage range tested (−Vp = −80 to +120 mV) ( Fig. 5C-a). However, interestingly the slope of the opening rate changes was steeper in a voltage range where inward currents were elicited (−Vp = −80 to −40 mV: the closed column in Fig. 5C-b) than in a voltage range where outward currents were elicited (−Vp = +40 to +120 mV: the open column in Fig. 5C-b). By contrast, as the membrane voltage increased, the closing rate decreased under conditions with outward currents (−Vp = +40 to +120 mV: Fig. 5D-a and the open column in Fig. 5D-b), whereas it was unchanged under conditions with inward currents (−Vp = −80 to −40 mV: Fig. 5D-a and the closed column in Fig. 5D-b). Thus, the slopes of both the opening and closing rates abruptly changed with the reversal potential as the boundary, indicating the current-direction-dependent single-channel kinetics of mPanx1 channels. Figure 6A and B show the correlation between open time and "|i|"; the absolute value of the single channel current. No significant correlation was seen in Cl − efflux (inward current; hyperpolarized) conditions (r = 0.13) (Fig. 6A), whereas a strong positive correlation was observed in Cl − influx (outward current; depolarized) conditions (r = 0.99) (Fig. 6B). The slope under Cl − influx (outward current; depolarized) conditions (1.09 ± 0.11 ms/ pA, n = 5) was significantly higher than that under Cl − efflux (inward current; hyperpolarized) conditions (0.02 ± 0.02 ms/pA, n = 5) (Fig. 6C). These results suggest that the mean open time (closing rate) depends on the magnitude of charge carrier flux through the channel (|i|) under Cl − influx (outward current; depolarized) conditions, and that 'the direction of ionic flux' through the channel plays a key role in the regulation of channel gating: i.e., the single channel closing kinetics of mPanx1 channel shows current-direction/amplitude-dependency.
The gating of mPanx1 channel is regulated by the amplitude of unitary outward currents. In the single channel analyses of cell-attached patch clamp recordings, we found that the mean open time of mPanx1 channels under Cl − influx (outward current; depolarized) conditions appears to be linearly correlated with the amplitude of unitary outward currents (|i out |) (Fig. 6B). However, a correlation also exists between the mean open time and the membrane potential, because different |i out | were achieved by changing the membrane potential. To determine which parameter, |i out | or membrane potential, is actually correlated with the mean open time of mPanx1, we employed tail current analysis using the whole-cell patch clamp configuration. HEK293 and Neuro2a cells were transiently transfected with wild-type mPanx1 and membrane currents were recorded during voltage ramps from −100 to +100 mV from a holding potential of −100 mV. Consistent with the previous findings 25 , expression of mPanx1 generated currents in both cell lines (427.4 ± 118.2 and 391.5 ± 75.0 pA at +100 mV in HEK293 and Neuro2a cells, n = 9 and 5, respectively), which were too small for analysis of tail currents. We therefore created a mutant of mPanx1 (mPanx1Δ371) that lacks the C-terminal auto-inhibitory region (residues 371-426) 25 . Expression of mPanx1Δ371 generated large enough outwardly-rectifying currents (2037 ± 330 pA at +100 mV, n = 4, in Fig. 7A and B) for tail current analysis: clearly identifiable tail currents at −100 mV (529.0 ± 111.9 pA, ▴ in Fig. 7B) in Neuro2a cells. Outward currents at +100 mV (I out ) were taken at the end of the pulse (⦁) and the tail currents (I tail ) were calculated by subtracting the holding currents at −100 mV from the peak inward currents observed immediately after repolarizing from +100 mV to −100 mV (▴). This suggests that mPanx1Δ371 channels retain the P o dependence of either |i out | or membrane potential.
In the voltage pulse protocol used here, the outward current at +100 mV (I out ) reflects the rate of influx of the major extracellular anion, but the tail current (I tail ) is carried by Cl − efflux from the intracellular space and should not depend upon extracellular anions. Its amplitude is linearly related to the product of: 1) the P o of the channel at the end of the depolarization pulse (+100 mV), and 2) the driving force of the Cl − efflux at −100 mV, which is   . The slope value in depolarization (1.09 ± 0.11 ms/pA, n = 5) was significantly larger than that in hyperpolarization (0.02 ± 0.02 ms/pA, n = 5) (C). Asterisk indicates significantly difference between inward and outward currents (***p < 0.001). Data are shown as mean ± SEM (n = 5).
SCienTiFiC REpoRTS | 7: 10512 | DOI:10.1038/s41598-017-10921-x determined by the reversal potential (E rev ). As expected, replacement of extracellular Cl − with gluconate − diminished |i out |, but, surprisingly, reversibly reduced the amplitude of I tail as well as I out (Fig. 7A-C). Because of the lower permeability of gluconate − to the channel than Cl − (Fig. 3), this anion substitution causes a positive shift in the reversal potential (E rev ) as observed in Fig. 7A, thereby increasing the driving force for I tail by 37% (considering approximate relative permeability of Cl − : gluconate − = 1.0: 0.186 (Fig. 3C), the calculated E rev in extracellular Cl − or gluconate − is 1.1 or 38.5 mV, respectively). Nevertheless, I tail recorded in extracellular gluconate − was 48.4% smaller than that in extracellular Cl − (Fig. 7C), indicating that the P o of the channel at +100 mV is 62.3% lower in gluconate − than in Cl − . These data clearly demonstrate that the P o of mPanx1 is regulated by the magnitude of |i out | independently of the membrane potential, strongly supporting the concept of the current-dependent gating of mPanx1.

Discussion
In the present study, we investigated the gating properties of mPanx1 channels expressed in HEK293 and Neuro2a cells at the single-channel level by the cell-attached and whole patch-clamp techniques and found that their kinetics are drastically changed by voltage stimuli around the reversal potential (E rev ). This voltage-dependence of the kinetics of mPanx1 channels may be due to electrostatic interactions between the surface potential and extracellular loops and/or the cytoplasmic N-and C-termini that have many charged residues and may interact with the gate in a voltage-dependent manner. Although the channel gating kinetics appears to depend on voltage (membrane potential), our results indicate that they are instead regulated in a manner that depends upon the direction and amplitude of currents through the channel. This result suggests that channel gating is not regulated directly by voltage, but is rather directly regulated by the direction and magnitude of charge carrier flux (current) through the channel. Because currents through the channel can occur in either direction, mPanx1 channels demonstrate apparent voltage-dependence.
A model of the channel gating is shown in Fig. 8A, in which we assume that the gate contains a charged residue(s) and is located within the pore but not in the transmembrane region (i.e., in extracellular or intracellular portion of the pore). As shown in Fig. 8A-a, if we assume a 'negatively' charged gate at the 'cytoplasmic region' in the pore, the channel gate located on the right side in the channel pore (a blue circle with -in Fig. 8A-a, left panel) shows the closed state, while the channel gate located on the left side in the channel pore (a red circle with -in Fig. 8A-a, right panel) shows the open state. Figure 8A-b indicates the transition of the channel from the open state to the closed state. Under Cl − efflux (inward current: hyperpolarized) conditions ( Fig. 8A-b, left panel), the channel gate closes spontaneously (the channel gate moves from the red circle location (open sate) to the blue circle location (closed state) as indicated with a blue arrow in Fig. 8A-b, left panel) without any Cl − efflux (inward) -current-induced force affecting the channel gating movement. In contrast, under Cl − influx (outward current: depolarized) conditions ( Fig. 8A-b, right panel), the force (a red arrow in Fig. 8A-b, right panel) induced by the Cl − influx (outward current) keeps the channel gate open in a manner dependent on the magnitude of the Cl − influx (outward current) against the spontaneous movement of the channel gate to the closed state (a blue arrow in Fig. 8A-b, right panel). Thus, the length of the channel open time increases as the amplitude of single channel outward current increases. The closed time (opening rate) may be controlled by interactions between "gating charges" and membrane potential.
The behavior of Panx1 is reminiscent of certain types of K + channels, in which the magnitude of current and the rate of gating depend upon the direction of ion flux [45][46][47][48][49] . All of these channels have multiple binding sites and multiple asymmetric energy barriers to ion flux. The channel is modeled by a chain of saturable binding sites that can be occupied by a number of ions. The ions move in single file to one of its immediate neighboring vacant sites with absolute rates determined through Eyring equation by the barrier heights between sites, the membrane potential, and the ion-ion repulsion. The Panx1 channel appears to follow this paradigm. For the Panx1 channel, we can approximate the properties of the channel if the pore of the channel has a strong binding site (deep energy well) near the cytosolic surface with an adjacent high energy barrier to ion flux. In addition, the channel would need to have multiple (at least 2) weak binding sites and small barriers near the extracellular end of the channel pore (see Fig. 8B in detail).
The voltage-sensing mechanisms of voltage-gated channels are of general interest. In many channels, voltage gating is regulated by a voltage sensor domain within the transmembrane domain, for example the S4 domain The mPanx1 channel is represented as a series of multiple anionic binding sites S1 -S3 and multiple energetic barriers to ion movement, B1 -B4. Barrier B2 is relatively high and binding site S1 is relatively deep. (b) The situation for outward Cl − movement (inward current). Cl − enters the channel from the cytosol inside the cell and binds to the strong binding site, S1. After binding, Cl − will usually return to the cytosol but occasionally will acquire enough energy to cross the adjacent barrier B2 and continue across the multiple small barriers and weak binding sites to the exterior of the cell. Crossing barrier B2 would be associated with channel gating. Thus, since the rate limiting step is crossing barrier B2, the Cl − efflux (inward current) would be small, and the gate is quickly closed. (c) The situation for anion influx (outward current). Cl − enters the outer mouth of the channel easily and multiple ions (Cl − ) rapidly occupy all of the weak binding sites, S3 and S2. The buildup of multiple anionic (Cl − ) charges in the outer part of the channel provides additional coulombic energy to reduce energy barrier B2 (dotted line: the difference between the levels of barrier B2 and binding site S2 reduces to the dotted line from the solid line due to the buildup of S2 level by multiple anionic (Cl − ) charges) and allow ion (Cl − ) movement to the strong binding site S1 near the cytosolic surface, but coulombic force also provides energy to move anions (Cl − ) out of this well into the cytosol (dotted line). Thus, Cl − influx (outward current) is large, and the gate is kept continuously open depending on the Cl − influx (outward current).
SCienTiFiC REpoRTS | 7: 10512 | DOI:10.1038/s41598-017-10921-x of the super-family of voltage-gated K + , Na + and Ca 2+ channels 50,51 . However, Panx1 lacks an obvious voltage sensor domain. Other ion channels and receptors are voltage-dependent despite the absence of a voltage sensor. For example, nicotinic ACh receptors, P2X 2 receptors 52, 53 and some metabotropic glutamate receptors, including mGluR1a and mGluR3 54 , have ligand binding sites at either the extracellular or intracellular sides with ligand binding causing conformational changes that produce an apparent voltage dependence [52][53][54] . Replacement of extracellular Cl − with other anions, including MeSO 3 − , Br − , I − and SCN − , modulates the gating properties of ASIC1a recorded in rat hippocampal neurons with the whole-cell patch-clamp technique 27 . ASIC1a channels consist of three subunits with a trimeric extracellular domain containing three "anion-binding sites" that are highly conserved among ASIC isoforms 55 . Cl − regulates ENaC gating 36,56 .
In contrast, our results here demonstrate that mPanx1 channels show apparent voltage-dependent gating kinetics (Fig. 4A), but the voltage dependence is not mediated directly by membrane potential but rather by the direction and amplitude of currents (anion flux). Thus, mPanx1 may not require a voltage sensor to gate. However, mPanx1 has charged residues located at the cytoplasmic side of the channel. It is plausible that some of these charged residues are located within the electric field across the plasma membrane and act as voltage sensors, causing conformational changes that lead to modulation of channel gating. Alternatively, the present study supports a novel model in which the direction and amplitude of currents through the channel influence channel gating, providing voltage dependence mediated by interactions of ions in the permeation with the channel gating machinery.
In summary, the present study suggests that the direction and magnitude of charge flow regulate the 'voltage-dependent' single-channel gating kinetics of mPanx1 channel, thus providing a novel model in which single channel gating kinetics are dependent upon the direction and amplitude of currents through the channel.