ISOLATION AND CHARACTERIZATION OF A HIGH AFFINITY PEPTIDE INHIBITOR OF CLC-2 CHLORIDE CHANNELS

The ClC protein family includes voltage-gated chloride channels and chloride/proton exchangers. In eukaryotes, ClC proteins regulate membrane potential of excitable cells, contribute to epithelial transport, and aid in lysosomal acidification. Although structure/function studies of ClC proteins have been aided greatly by the available crystal structures of a bacterial ClC chloride/proton exchanger, the availability of useful pharmacological tools, such as peptide toxin inhibitors, has lagged far behind that of their cation channel counterparts. Here we report the isolation, from Leiurus quinquestriatus hebraeus venom, of a peptide toxin inhibitor of the ClC-2 chloride channel. This toxin, GaTx2, inhibits ClC-2 channels with a voltage-dependent apparent K(D) of approximately 20 pm, making it the highest affinity inhibitor of any chloride channel. GaTx2 slows ClC-2 activation by increasing the latency to first opening by nearly 8-fold but is unable to inhibit open channels, suggesting that this toxin inhibits channel activation gating. Finally, GaTx2 specifically inhibits ClC-2 channels, showing no inhibitory effect on a battery of other major classes of chloride channels and voltage-gated potassium channels. GaTx2 is the first peptide toxin inhibitor of any ClC protein. The high affinity and specificity displayed by this toxin will make it a very powerful pharmacological tool to probe ClC-2 structure/function.

GaTx2: a high affinity peptide inhibitor of ClC-2 Thompson (1,4), and participation in lysosomal acidification (ClC-5, ClC-6) (2). Defects in the genes encoding ClC proteins are linked to a number of diseases including myotonia, epilepsy, Dent's disease, and Bartter's syndrome (1)(2)(3). It has been suggested recently that ClC-2 may play a role in constipation-associated irritable bowel disease as well as in atherosclerosis (5,6). Most ClC channels show localized tissue expression; ClC-1 for example is expressed solely in skeletal muscle, while ClC-Ka/b is localized to the kidney. ClC-2, on the other hand, is expressed nearly ubiquitously, suggesting that this channel plays an important, yet largely undefined, physiological role (1,2).
ClC proteins are structurally unrelated to cation channels, with the functional unit being a homodimer (1). ClC channels display two equidistant conductance levels for a single channel opening. In 2002, the crystal structure of a bacterial ClC protein from S. typhimurium was solved, revealing a very complicated membrane topology consisting of 18 α-helical units per subunit in the homodimer, only some of which fully traverse the membrane (7). Examination of the crystal structure revealed no obvious pore, such as is evident in K + channel structures, even though bound Clions were present near the proposed selectivity filter (7,8). Shortly after the crystal structure was solved it was shown that the bacterial ClC protein was actually a Cl -/H + exchanger, and not a channel (9). Comparison of the amino acid sequence of the bacterial ClC protein with that of the eukaryotic ClC channels ClC-0, -1, and -2 revealed only 22%, 16%, and 19% overall identity, respectively (data not shown). The divergence is largely in the cytoplasmic domains, which are absent in bacterial ClC proteins; sequence identity is much higher in the transmembrane domains.
Single-channel gating in ClC proteins is complicated, involving both fast and slow gating processes which are thought to involve separate regions of the protein (1). Fast gating controls the opening and closing of both protopores independently, operating on the millisecond timescale or faster. Through examination of the crystal structure, and subsequent electrophysiological analysis, the fast gating process was revealed to involve a conserved glutamate residue deep within each pore (10). This acidic residue lies near a Clbinding site and moves slightly to open the pathway in response to changes in membrane voltage and subsequent changes in occupancy of that site, thus providing the link between permeation and gating observed in ClC channels (4). In contrast, slow gating controls both pores simultaneously, operating on the hundreds of milliseconds to seconds timescale.
Unlike fast-gating, the regions of the ClC protein involved in slow gating are still unknown, despite the availability of the bacterial ClC crystal structure. It is believed that the dimer interface contributes to slow gating, as well as the long cytoplasmic Cterminal domain, an isolated version of which was recently crystallized (11)(12)(13). However, the conformational changes involved in the fast and slow gating processes are still largely unknown. Also, in both ClC-1 and -2, fast and slow gating are linked through an undetermined mechanism (14,15).
Despite the availability of the bacterial ClC protein crystal structure, our understanding of gating mechanisms and structural rearrangements of ClC proteins has lagged behind that of their cation channel counterparts. This is due in large part to a lack of useful pharmacological agents, such as peptide toxins, that may be used as tools. Toxins from venomous animals such as scorpions, snakes, and cone snails have been used for a number of years to define the permeation pathways and gating processes of cation channels (16). However, no peptide toxins have been isolated that inhibit a ClC channel, and only one toxin has been isolated that inhibits any Clchannel of known molecular identity (17). We recently showed that venom from the scorpion L. quinquestriatus hebraeus contains a peptide component that inhibits the ClC-2 chloride channel (18). Here, we report the isolation of this peptide toxin, its proteomic properties, and primary characteristics of the biophysical mechanism of inhibition.

Experimental Procedures Oocyte and cRNA Preparation
Xenopus oocytes were isolated as previously described (18) and incubated at 18 o C in a modified Liebovitz's L-15 medium (pH 7.5) by guest on March 24, 2020 http://www.jbc.org/ Downloaded from with a cocktail of antibiotics (gentamicin, penicillin, streptomycin) and HEPES. Oocytes were injected with cRNAs encoding ClC-1, -2, -3, -4, CFTR, GABA C , Shaker B-IR, or K V 1.2. Injection details may be found in Supplemental Methods. Data were collected 2 -5 days post injection. Methods for animal handling are in accordance with the NIH guidelines and the protocol was approved by the Animal Use and Care Committees of the Georgia Institute of Technology and Emory University.

Venom Preparation and Toxin Purification
Venom from L. quinquestriatus hebraeus was obtained from Latoxan (France), and prepared as described in Supplemental Information or ref. (18). Reversed-phase HPLC (RP-HPLC) was performed on partially-fractionated scorpion venom (Lqh pf-venom), or on venom fractions. For toxin purification, venom was prepared at a concentration of 5 mg/mL in solution that contained (in mM): 150 NMDG-Cl, 5 MgCl 2 , 10 TES, and 10 Tris·EGTA, pH 7.5. Subsequent fractions were resuspended in ND96 bath solution for the electrophysiological bio-assay (see Supplemental Methods). All venom and fraction concentrations are stated as equivalent to the dried weight of whole venom.

Electrophysiology
Standard two-electrode voltage-clamp (TEVC) techniques were used to study whole cell ClC-2 currents expressed in Xenopus oocytes. Currents were acquired using a GeneClamp 500B amplifier and pClamp 8.4 software (Axon Instruments / Molecular Devices Union City, CA). Data were acquired at 2 kHz and filtered at a corner frequency of 500 Hz. Bath solutions and voltage protocols for TEVC experiments with ClC channels were as described in Supplemental Methods or ref. (18). Inside-out multi-channel patch, inside-out and outside-out macropatch, and single-channel patch clamp experiments were performed as previously described (17,18), using an Axopatch 200B amplifer (Axon). For all recordings the dashed line represents the zerocurrent level. Further details of electrophysiology methods are provided in Supplemental Methods.
In order to prevent loss of material, all solutions containing toxin were prepared immediately prior to experiments from a 20 μM stock. All solution lines and recording chambers were coated with bovine serum albumin to prevent loss of material due to sticking to plastic or glass.

Proteomic Analysis
The active toxin was analyzed via MALDI-MS at the Georgia Institute of Technology Bioanalytical Mass Spectrometry Facility. Edman degradation was performed at the Emory University Microchemical and Proteomics Facility. Sequence analysis of the natural and modified (reduced and alkylated) peptides was performed using Applied Biosystems model Procise-cLC automated protein sequencer (Applied Biosystems 491cLC CLC capillary protein sequencing system; Foster City, CA) using the manufacturer's cycles with slight modifications. Prior to sequencing, natural and modified samples were de-salted and purified via RP-HPLC.

Homology Modeling
The homology model of the GaTx2 toxin was created using the Modeller 8v2 program, using Neurotoxin P01 as the template structure (19). The GaTx2 homology model was then minimized via a 5 ps, 2500 step simulation with NAMDv2 utilizing the charmm22 force field. Prior to minimization, disulfide bridges were patched in order to ensure that the disulfide bonds remained intact during energy minimization.

GaTx2 Synthesis
Methods for peptide synthesis, purification, and folding have been previously described (20,21). Solid phase synthesis was performed by Fmoc chemistry, using the HBTU/HOBT/DIPEA method on an Applied Biosystems 431A synthesizer. Oxidative cyclization of the crude linear peptide was performed under equilibrating conditions. The cyclized peptide was then purified from the reaction mixture via RP-HPLC and characterized by MALDI-MS and analytical HPLC.
GaTx2: a high affinity peptide inhibitor of ClC-2 Thompson et al.,4 Statistics Data are expressed as mean + SEM for n obervations. Differences were determined to be significant when p< 0.05, using paired and unpaired Student's t-tests.

Isolation and Proteomic Characterization of the Active Toxin
We recently showed that a peptide component of scorpion venom is capable of specifically inhibiting the ClC-2 chloride channel (18). In order to isolate the active species from venom, we used RP-HPLC to separate the components of Lqh pf-venom, which contains only components smaller than 10 kDa. We tested each fraction for inhibition of ClC-2 channels expressed in Xenopus oocytes using TEVC. With initial separation performed using a C3 column, we observed that the fraction collected from 0 -10 minutes (0 -10% acetonitrile), Fraction A, retained activity similar to Lqh pf-venom (Supplemental Information (SI) Fig. 1A, 45.7 + 6.0% inhibition, fraction concentration 0.1 mg/mL equivalent, n = 3, p = 0.01).
The brief retention time of the active component on the C3 column suggested that the toxin is very hydrophilic; therefore, further separation of components and isolation of the active toxin were accomplished by RP-HPLC using a C18 column, as summarized in SI Figure  1B. Fractions were collected every 5 minutes and tested using the same protocol as in the prior round of fractionation. Activity was spread over a period of 15 minutes of elution covering a range of 7 -15% acetonitrile. The active fractions showed activity similar to that of venom and fraction A (SI Fig. 1B, 45.2 + 5.0% inhibition, fraction concentration 0.1 mg/mL equivalent, n = 10, p = 0.01). These fractions were pooled, and subjected to a final round of RP-HPLC using a C18 column (Fig. 1A).
In the final stage of isolation, the active peak eluted at ~5% acetonitrile (Fig. 1A). This peak fully recapitulated the activity of Lqh pfvenom (64.2 + 5.3% inhibition at 0.1 mg/mL equivalent, n = 4, p = 0.025). A conservative estimate of toxin concentration was obtained from Edman degradation of the isolated peptide. These results suggested that this dilution of venom (0.1 mg/mL equivalent) contained 10 nM isolated toxin. A dose response curve constructed using native toxin suggested a K D ~ 50 pM and a Hill coefficient of ~ 1 (V M = -160 mV) (SI Fig. 1C). MALDI-MS analysis showed that the active peak had a mass of 3.2 kDa (Fig.  1B), well within the range expected for scorpion toxin inhibitors.
The primary structure of the peptide corresponding to the active peak was determined using N-terminal protein degradation and amino acid analysis (see Supplemental Methods). The toxin, which we named "Gating modifier of anion channels 2" (GaTx2), had a protein sequence of 1 VSCEDCPDHCSTQKARAKCDNDKCVCE PI 29 (Fig. 1C). MALDI-MS analysis of a reduced form of the toxin exhibited a mass shift of 348.2 Da (from 3191.29 to 3539.46 Da) upon reduction by DTT and modification of free cysteine residues by iodoacetimide, confirming that the six cysteine residues participate in three disulfide bonds in the folded toxin (data not shown). Comparison of this sequence to other known toxins revealed that GaTx2 bore 100% identity to a previously identified toxin called Leiuropeptide II (LPII). This toxin was initially isolated in 1997, although no target was identified; it was initially postulated that this toxin was a K + channel inhibitor based on disulfide bridge arrangement, and spacing of critical lysine residues (22). Since it actually inhibits a Clchannel, we have renamed this toxin to avoid confusion. Of the toxins shown in Figure 1C only four have been shown to be active against ion channels: neurotoxin P01, which inhibits apamin sensitive K + channels (23), scyllatoxin, which inhibits Ca 2+ -activated K + (K Ca ) channels (24), and maurotoxin and OdK1, both of which inhibit K V 1.2 (25,26).
The NMR structure of GaTx2 has been solved previously (22); however, the coordinates are not publically available. Therefore, we created a homology model of GaTx2 based on the NMR structure of neurotoxin P01 (19), which bears 93% sequence identity to GaTx2, in order to make structural comparisons of GaTx2 to other known scorpion toxins (Fig. 1D  middle). This model predicts that GaTx2 is by guest on March 24, 2020 http://www.jbc.org/ Downloaded from composed of 2 β-strands and one α-helix, which are connected via 3 disulfide bonds. The basic fold and disulfide connectivity are consistent with the published NMR structure of LPII (22). This basic fold is highly conserved among numerous toxins isolated from various scorpion species (27). We compared the tertiary structure of GaTx2 to both scyllatoxin, a K + channel inhibitor, and chlorotoxin, a toxin isolated from a related scorpion that is thought to inhibit an unidentified Clchannel from rat brain vesicles when reconstituted into lipid bilayers (24,28). GaTx2 is noticeably more compact than both scyllatoxin (Fig. 1D left) and chlorotoxin (Fig.  1D right). Although the basic fold is similar between all three of these toxins, the tertiary structure itself is not enough to impart inhibitory activity against ClC-2, as chlorotoxin is unable to inhibit ClC-2 channels (18). The primary and secondary structures of GaTx2 are also very different from GaTx1, a recently isolated peptide inhibitor active against the CFTR Clchannel, which also does not inhibit ClC-2 (17).

Inhibition of ClC-2 by Synthetic GaTx2
In order to determine if the component that was isolated from venom is truly responsible for the observed inhibitory activity, we produced GaTx2 via solid-phase chemical synthesis, as has been done for many other peptide inhibitors (20,21). The folded synthetic toxin shows the same mass as the purified native toxin (SI Fig. 2A right). Activity of the synthetic form of GaTx2 was tested initially via TEVC using the same voltage protocols as used during the isolation of the toxin. TEVC experiments showed that 10 nM synthetic GaTx2 inhibited steady-state ClC-2 mediated currents by 61.4 + 8.0 % at V M = -160 mV (n = 3) (SI Fig. 2B, C). This activity was almost exactly the same as that observed for the native toxin at the same concentration (64.2 + 5.3, n = 4) (SI Fig. 2C).

Inhibition of ClC-2 by Synthetic GaTx2 is Similar to That of Lqh pf-venom
We previously showed that Lqh pfvenom inhibited ClC-2 currents across a range of voltages, for both the quasi steady-state activation currents and the tail currents elicited by a voltage step to +40 mV (18). In order to determine if GaTx2 showed similar behavior, we tested the effect of 10 nM GaTx2 across a range of voltages from -160 to +60 mV. Similar to the activity of Lqh pf-venom, GaTx2 strongly inhibited inward ClC-2 currents, and showed no effect at membrane potentials that do not activate the channel ( Fig. 2A), drastically altering the current voltage relationship. There is also strong inhibition of tail currents at +40 mV from test potentials that strongly activate the channel (Fig. 2B), which could induce a shift in the V 1/2 of channel activation to more hyperpolarizing potentials. However, due to the fact that ClC-2 never reaches full activation, even at strongly hyperpolarizing potentials, it is not possible to determine the effect of GaTx2 on the conductance/voltage relationship for ClC-2 by calculating a true V 1/2 .
Lqh pf-venom also inhibited ClC-2 currents in a strongly voltage-dependent manner, with greater inhibition at more depolarizing potentials (18). In order to determine if synthetic GaTx2 showed similar voltage-dependence of inhibition, we examined the fractional current remaining across a range of membrane potentials from -160 mV to -60 mV. Currents were inhibited by 62.5 + 2.6% at -160 mV (n = 3), but at -60 mV currents were inhibited by 86.6 + 2.3% (n = 3, p = 0.006) (Fig. 3A), consistent with the behavior observed for Lqh pf-venom (18). Also consistent with the activity of Lqh pfvenom is a lack of voltage dependent inhibition of tail currents elicted at +40 mV (n = 3) (Fig.  3B). These results, therefore, suggest that GaTx2 is indeed the active component of Lqh pf-venom and inhibits via a similar mechanism.

GaTx2 is a Highly Selective ClC-2 Inhibitor
In order to determine if GaTx2 could inhibit any other ion transport proteins, we tested the toxin on a number of related and unrelated ion channels and transporters, including members of the ClC family, other major chloride channels, and voltage-dependent K + channels. We previously reported that Lqh pf-venom could not inhibit ClC-0 or ClC-1 (18). To confirm this we asked whether synthetic GaTx2 could inhibit ClC-1. We compared currents in the presence of toxin to those after washout of toxin in order to account for channel rundown. In the presence of 10 nM GaTx2, ClC-1 tail currents at V M = -120 mV from pulses to +60 mV were 1.03 + 0.01 fold larger than currents after washout of toxin, implying no inhibition of ClC-1 (Fig. 4A). Application of 10 nM GaTx2 to ClC-3 likewise did not alter currents at V M = +80 mV (Fig. 4B, 0.90 + 0.05 fold, n = 3, p = 0.18). GaTx2 also was not capable of inhibiting ClC-4 mediated Cltransport measured at V M = +80 mV (10 nM GaTx2, 0.97 + 0.02 fold, n = 3, p = 0.42) (Fig.  4C), indicating that GaTx2 does not inhibit a ClC protein that functions as a transporter. We also know from previous experiments that Lqh pf-venom does not inhibit ClC-0 (18). Thus, GaTx2 is not capable of inhibiting currents mediated by any other ClC channel or transporter on which we have tested it.
We recently identified and characterized GaTx1, a peptide inhibitor of currents mediated by CFTR chloride channels (17). CFTR was not inhibited by the venom fraction that contains GaTx2 (29). However, in order to corroborate these results we tested the effect of synthetic GaTx2 on CFTR. We first tested GaTx2 on CFTR currents by application of toxin to the extracellular face of the channels using TEVC. Channels were activated by 200 μM IBMX + 25 μM db-cAMP. After steady-state activation was achieved, 20 nM GaTx2 was applied for 5 minutes. No change in current was observed during the application of toxin (Fig. 4D, 0.95 + 0.02 fold, n = 2), consistent with the notion that Lqh pf-venom could not inhibit CFTR currents from the extracellular face. Due to the fact that GaTx1 only inhibited CFTR from the intracellular face (17), we applied 60 nM GaTx2 to the intracellular face of the channel using multi-channel patches. For these experiments we used Flag-cut-ΔR-CFTR, a mutant CFTR lacking the R-domain, rendering it independent of phosphorylation by PKA and insensitive to dephosphorylation-mediated rundown. Channels were activated with 0.2 mM ATP, then reactivated with 0.2 mM ATP + 60 nM GaTx2. No inhibition of channel currents was observed (Fig. 4E, 0.91 + 0.13 fold, n = 10, p=0.74). Therefore, GaTx2 also is not capable of inhibiting CFTR from either the extracellular or intracellular face.
We next asked whether GaTx2 is capable of inhibiting currents from ligand-gated chloride channels formed by homomeric GABA C receptors. GABA C currents were measured by TEVC both in the absence and presence of GaTx2 (Fig. 4F). Comparison of current in the presence of 10 μM GABA to currents in the presence of 10 μM GABA plus 10 nM GaTx2 showed no change (13 + 9.4 % increase, n = 3, p = 0.32). GABA-induced Clcurrents also did not significantly increase upon washout of toxin (14 + 19% increase, p = 0.74). This suggests that GaTx2 is not capable of inhibiting GABA C receptors.
Xenopus oocytes contain a large number of endogenous calcium-dependent chloride channels (Cl Ca ). Application of 10 mM Ca 2+ to the face of an excised inside-out patch rapidly activates these channels, which then rundown over the course of the experiment. Therefore, control records and records in the presence of GaTx2 were obtained from different patches. In order to assess toxin activity, we measured current decay over the course of three pulsed exposures to high [Ca 2+ ]; toxin was applied between the 1 st and 2 nd Ca 2+ pulses. If GaTx2 inhibited channel currents, we expected to observe a greater degree of apparent rundown in toxin-containing patches versus control patches. However, the decay between pulses was the same with and without 10 nM GaTx2 (Fig. 4G, 25.67 + 3.65% (control, n = 5), vs. 34.11 + 5.26, (+ GaTx2, n=6), p = 0.24), suggesting that GaTx2 is not capable of inhibiting these channels from the intracellular side. We were unable to test the effect of GaTx2 on Cl Ca when applied to the extracellular face because receptor-mediated activation of these channels exhibits irreversible desensitization after one exposure to receptor ligand (not shown).
Finally, because the toxin bearing the sequence of GaTx2 was originally predicted to be a K + channel inhibitor, we sought to determine if this toxin could inhibit the voltagegated Drosophila Shaker K + channel B variant with inactivation-removed (ShB-IR), by exposing oocytes expressing ShB-IR to 20 nM GaTx2, a concentration that strongly inhibits ClC-2 currents (Fig. 4H). We observed no inhibition of ShB-IR currents, indicating that GaTx2 does not interact with this channel (1.00 + 0.007 fold, n = 2). We similarly tested the effect of GaTx2 on the mammalian Shaker equivalent Kv1.2, and observed no change in K + by guest on March 24, 2020 http://www.jbc.org/ Downloaded from currents ( Fig. 4I, 0.99 + 0.03 fold, n = 4, p=0.55). It is possible that GaTx2 may inhibit these channels at a much higher concentration, but it had no effect at a concentration that is saturating for inhibition of ClC-2 mediated currents (see below). Thus, we have shown that GaTx2 is a specific ClC-2 inhibitor, being unable to inhibit other members of the ClC family, CFTR, GABA C , Cl Ca , or major voltagedependent K + channels.

Kinetics of Inhibition of ClC-2 by GaTx2
In order to gain a better understanding of the mechanism of inhibition of ClC-2 by GaTx2, we measured the rate of both onset of and recovery from inhibition of ClC-2 currents by direct application of GaTx2 to the extracellular face of the channel using TEVC; solution exchange in these experiments was complete within ~10 seconds. In order to allow sufficient channel activation, and maximal deactivation between pulses, channels were activated every 15 seconds with a 1 second voltage pulse to V M = -100 mV, followed by a tail pulse to +40 mV. As a kinetic control, inhibition of ClC-2 currents by 1 mM Zn 2+ showed that both inhibition and washout occurred very rapidly, with a time constant for recovery of ~38 seconds (SI Fig. 3). The data obtained for both onset of inhibition and recovery from inhibition were fit with a single exponential function from the beginning of either onset or recovery from inhibition until the process had reached steady-state. Timedependent currents were inhibited by ~95% by 1 mM Zn 2+ , as expected (30). These experiments were then repeated using GaTx2 over a broad concentration range. Data obtained in the presence of 100 pM GaTx2 showed a fast onset of inhibition and a very slow recovery from inhibition (Fig. 5A, B). For low concentrations of toxin, steady-state inhibition was achieved within five minutes. Using this protocol, we were able to construct a dose-response curve for steady-state inhibition of ClC-2 time-dependent currents by synthetic GaTx2 at V M = -100 mV, which gave an apparent K D of 22 + 10 pM, and a Hill coefficient of ~1 (Fig. 5C), consistent with the high affinity interaction observed for native GaTx2 (SI Fig. 1C, 2C). Maximal inhibition of ~50% at this voltage likely arises from strong voltage-dependence as well as protocol dependence of inhibition of ClC-2; this is also consistent with the behavior of Lqh pf-venom (18). The voltage-dependence of inhibition likely reflects the voltage-dependence of channel gating, although it is possible that the toxin may also bind in an intrinsically voltage-dependent manner. If GaTx2 does, indeed, function as a gating modifier, as suggested for Lqh pf-venom (18) and by results described below, its activity also should be state-dependent. Therefore, the incomplete inhibition observed using this voltage protocol may reflect incomplete inactivation of ClC-2 during the voltage protocol. We also observed that the pseudo-first order on-rate for GaTx2 increased as a function of [GaTx2], while the off-rate was independent of toxin concentration (Fig. 5D), as expected. The very slow off-rate (k off = 0.0034 s -1 ) is consistent with an intimate interaction between toxin and channel, and is consistent with the offrates of other very high affinity toxins (31). While the on-rate is clearly concentrationdependent, we are unable to measure the forward rate constant accurately for rates faster than 0.1 s -1 . This is due to a combination of the slow solution exchange rates and the slow voltage-induced activation and inactivation of ClC-2 channels requiring a voltage clamp protocol that pulses no more frequently than every 15 seconds. Thus, we cannot determine whether the onset of inhibition is a linear function of [GaTx2], and therefore represents a simple bimolecular interaction.

Inhibition of ClC-2 by GaTx2 is Voltage Dependent
In order to confirm the activity of synthetic GaTx2, we recorded from inside-out multi-channel patches, where the pipette was backfilled with varying concentrations of synthetic toxin. The pipette was backfilled in such a way as to allow ~10 minutes of control recording (at V M = -100 mV or -60 mV), followed by ~10 minutes of recording in the presence of GaTx2. We then calculated the average window current from five separate fourminute windows in both control and experimental conditions (17). When no toxin was backfilled into the pipette, we observed no change in average window current, suggesting that there was no rundown over the course of the experiment (Fig. 6A). However, when 2 nM synthetic GaTx2 was backfilled into the pipette, average window currents were drastically reduced at the end of the experiment (Fig. 6A,  bottom; 80.4 + 2.0% decrease). We repeated these experiments with varying concentrations of GaTx2 to obtain a dose-response curve, which provided K D = 12 + 5 pM and a Hill coefficient of ~1 for steady-state inhibition at -100 mV using this multichannel patch protocol (Fig. 6B). This value is not significantly different from the K D measured using TEVC at the same membrane potential (Fig. 5C), suggesting that the pipette backfill technique was sufficient to accomplish steady-state inhibition of the channel. The small difference in magnitude of maximal inhibition using the TEVC and multichannel patch protocols likely arises from differences in the Clgradient (no gradient for multichannel patches vs. ~3:1 out:in gradient for TEVC).
Our previous experiments with Lqh-pf venom indicated that inhibition of ClC-2 was voltage-dependent, with improved inhibitory efficacy at less hyperpolarizing potentials (18), suggesting that the affinity of GaTx2 is voltagedependent. In order to further characterize this phenomenon, we compared the degree of inhibition of ClC-2 current by 2 pM synthetic GaTx2 at V M = -100 and -60 mV using multichannel patches (Fig. 6C). At V M = -100 mV, 2 pM toxin did not inhibit ClC-2 significantly (4.0 + 9.9 % inhibition, n = 11 windows); however, at V M = -60 mV, ClC-2 currents were inhibited 41.4 + 4.1 % (n = 17 windows) by 2 pM toxin, confirming the notion that inhibition of ClC-2 by GaTx2 is voltage-dependent.

GaTx2 Slows ClC-2 Channel Opening
Toxins isolated from animal venoms may inhibit channel activity either by blocking the channel pore or by modifying channel gating via altering the voltage dependence of activation, thus causing a change in opening or closing rates at a given membrane potential (32). The activity of Lqh pf-venom suggested that the active component may act as a gating modifier due to the significant slowing of the activation kinetics at -160 mV in the presence of venom (18). We tested the effect of 10 nM GaTx2 on the activation kinetics of ClC-2 in order to determine if the activity of the isolated toxin was similar to that of Lqh pf-venom. Activation of ClC-2 macroscopic currents is best described by the sum of two exponential functions. In the presence of 10 nM GaTx2, we observed a 2.45 + 0.18 fold (n=3, p = 0.001) increase in the timeconstant for the fast component of channel activation, with no change in the slow time constant at low toxin concentrations (1.10 + 0.06 fold increase, n = 3, p = 0.12) (Fig. 7A, B). This activity is very similar to that observed for Lqh pf-venom, which showed a slowing of both the fast and slow components of channel activation at very high venom concentrations, although the change in the slow component of activation was less pronounced (18). This further suggests that GaTx2 inhibits ClC-2 channel activity via the same mechanism as Lqh pf-venom.
In order to examine the mechanism of inhibition in greater detail, we studied the effect of GaTx2 on single ClC-2 channels. To determine whether GaTx2 alters the single channel conductance of ClC-2, which may be indicative of pore block, we created all-points amplitude histograms from segments of records with only 1-2 open channels, in the absence and presence of GaTx2 (records not shown). In the control portion of the experiment, 2 channels were active simultaneously, resulting in four conductance levels and no periods with zero channel activity (Fig. 8A left). In the presence of 2 nM GaTx2, channel activity was drastically reduced such that only one active channel was observed, with the majority of the time spent in the closed state (Fig. 8A right). Also, in the presence of a saturating concentration of GaTx2 (2 nM), the single protopore amplitude at -100 mV was unchanged from control (0.25 + 0.01 pA vs. 0.26 + 0.01 pA, n = 3, p = 0.45) (Fig.  8A), as measured from the difference between levels O1 and O2 from the amplitude histograms created from 2-minute sections of record in the absence and presence of toxin. This suggests that GaTx2 does not induce partial conductances, although this does not fully discount the possibility of inhibition via a pore block mechanism.
We next performed patch clamp experiments with either one or two channels in the patch and 20 pM GaTx2 backfilled into the by guest on March 24, 2020 http://www.jbc.org/ Downloaded from pipette ( Fig. 8B -D). Measuring channel activity as NP O /N, we observed that NP O /N was reduced by 59.5 + 8% (Fig. 8D, n = 2) at -100 mV, which is consistent with the reduction of window current observed in multichannel patches for this concentration of toxin (Fig. 6B). Some toxins act via a pore block mechanism by inducing long closed states within a channel burst (33). However, comparison of expanded recordings of bursts in the presence and absence of GaTx2 (Fig. 8C) showed no toxin-induced intraburst closures that would be consistent with a pore block mechanism, although it is of course impossible to know whether the toxin was bound to the channels from which these currents arose, due to fact that toxin does not alter conductance. This also suggests that once the channel opens in the presence of toxin, GaTx2 exerts no effect on fast gating, which controls the fast transitions between conductance levels observed within a channel burst for ClC channels.
Prior experiments with Lqh pf-venom led us to propose that ClC-2 is inhibited by this toxin via modification of channel gating; channel activation was slowed in the presence of venom (18). This was also found to be the case for synthetic GaTx2 (Fig. 7).
The slow activation of ClC-2 channels evident in macroscopic recordings, such as shown in Figure 7, reflects the slow voltage-induced activation of both the slow-gates and the fastgates. The long interbust closures, as seen in Figure 8B, are likely due to closure of the slow gate, which controls both protopores simultaneously. However, for ClC-2 both fast and slow gating appear to be coupled (15), thus, an inhibitor that affects either fast or slow gates could lead to prolongation of the apparent intraburst closed state. We reasoned that the effect of GaTx2 on single-channel gating may be apparent as a change in the latency to first opening of ClC-2 channels, measured as the time from the voltage step until the beginning of the first double-barrelled open burst. For ClC channels. In the presence of 20 pM GaTx2, we observed a 7.67 + 0.25 fold increase (from 13 + 5 s to 107 + 40 s) (Fig. 8B,E; n = 2) in the latency to first opening upon stepping from V M = 0 to V M = -100 mV, which is consistent with a modification of channel gating. The effect of GaTx2 on individual ClC-2 microscopic opening and closing rate constants under steady-state conditions cannot be determined from these experiments, but warrants further attention.

GaTx2 is not an Open Channel Blocker
Thus far, our experiments have suggested that GaTx2 acts to slow channel opening via modification of channel gating. To provide further evidence for this mechanism, we tested the effect of GaTx2 on channels that had reached steady-state activation using outside-out macropatches. We reasoned that if GaTx2 modifies channel gating, application of toxin to channels in the open state should not lead to a reduction in current amplitude, or current amplitude would be reduced very slowly. If the toxin inhibited via an open channel pore block mechanism, then currents should be reduced even if toxin was applied when channels are open. To test this notion, we activated ClC-2 channels in outside-out macropatches with a pulse to -100 mV, and allowed current to reach steady-state (Fig. 9). After steady-state activation had been achieved, usually after several minutes, 2 nM GaTx2 was applied for at least 1 minute using a fast perfusion system with a solution exchange time constant of ~20 ms (18). Application of toxin did not inhibit open channel currents (1.02 + 0.02 fold change, n = 5, p = 0.15) (Fig. 9A, C), although application of toxin before steady state was reached led to substantial inhibition under these conditions (Fig. 9A inset). In contrast, application of 1 mM DPC, a known pore blocker, for 30 seconds to open channels resulted in 44.9 + 3.0 % inhibition (n = 9, p < 0.01) (Fig. 9B, D)

Discussion
We showed previously that Lqh pfvenom contained a peptide component that inhibited the ClC-2 chloride channel by slowing channel opening (18). Here we report the successful isolation of the first peptide inhibitor of a ClC protein, GaTx2, and initial characterization of the mechanism of action. GaTx2 is a 3.2 kDa peptide with three disulfide bonds holding together a secondary structure composed of one α-helix and two β-strands. Although the sequence of GaTx2 has been described previously (22), this is the first identification of its molecular target. The target for this toxin originally was predicted to be a K + channel based solely on sequence alignments with other known or putative K + channel toxins. Therefore, this is also the first toxin that has been shown to be misclassified, raising the possibility that a number of other toxins capable of inhibiting anion-selective channels have already been sequenced.
The crystal structure of a bacterial ClC protein has been very useful for directing experiments aimed at gaining further understanding of the pore structures and fast gating mechanisms of ClC channel proteins (10,35). Although the crystal structure has not been as useful for understanding slow gating, some progress has been made through mutagenesis studies (11,13,36,37). The new crystal structures of the isolated C-terminal domain from ClC-0 will likely be very helpful in this regard (12). However, many of the conformational changes associated with channel gating are still unknown, and are unlikely to be elucidated strictly by studying the available structures, which do not capture the dynamic nature of these proteins. The GaTx2 toxin will enable the biophysical study of these gating processes in greater detail.
The apparent dissociation constant of GaTx2 for ClC-2 at -100 mV, determined using multiple approaches including TEVC, which allowed direct application of toxin to the cell, and toxin application via a diffusion mechanism using pipette backfill, is ~ 15 pM (Fig. 5C, 6B). Consistent with this high affinity interaction, recovery of ClC-2 currents from inhibition is very slow, with a time constant of >300s. With apparent K d in the low pM range, GaTx2 inhibits ClC-2 with higher affinity than any other available drug and, in fact, is the best inhibitor of any chloride channel. The previously best available ClC inhibitor, a pentameric DIDS hydrolysis product, inhibits ClC-Ka with a K 1/2 of 0.5 μM while GaTx1 inhibits CFTR with a K D of 25 nM at -100 mV (17,38).
Interestingly, inhibition of ClC-2 by GaTx2 seems to be protocol dependent. Application of toxin to channels that have reached steady state activation did not reduce currents. However, significant inhibition was observed when channels were allowed to close in the presence of toxin, as observed for both TEVC recording and inside-out multi-channel patches. Therefore, GaTx2 does not mediate a simple open channel block mechanism such as that seen for some peptide toxins. Our TEVC and single channel data suggest that GaTx2 inhibits ClC-2 activity by slowing channel activation. Given that GaTx2 does not inhibit open ClC-2 channels, it is likely that the toxin preferentially binds to the closed state of the channel, and prevents channel opening. The possibility that inhibition of ClC-2 currents by GaTx2 may not be mediated by a simple bimolecular interaction is very intriguing. The seemingly biphasic appearance of both the doseresponse relationships (SI Fig. 1C; Fig. 5C, 6B), and recovery from inhibition (Fig. 5B), suggest that GaTx2 may bind to two or more sites that differ in affinity. ClC proteins are both structurally complex and regulated by very complicated gating mechanisms that are linked to permeation. In addition, fast and slow gating in ClC-2 are linked. Therefore, the idea that the interaction between GaTx2 and ClC-2 is not a simple one is not unreasonable, and warrants further investigation.
Inhibition by GaTx2 is also voltagedependent, with stronger inhibition at more physiological membrane potentials. This voltage-dependent inhibition also is consistent with a gating modification mechanism. Hanatoxin, a gating modifier of K V 2.1 channels, inhibits with a K D of 42 nM at 0 mV. However, at +50 mV, hanatoxin inhibits the K V 2.1 channels by only 40% at concentrations as high as 5 μM (39). This phenomenon is similar to the voltage-dependent inhibition reported here for GaTx2. It is at present mechanistically unclear by guest on March 24, 2020 http://www.jbc.org/ Downloaded from why maximal inhibition is ~60% for TEVC, while 80% inhibition can be achieved using patch clamp techniques. It is possible that the differences in recording conditions, such as chloride concentrations, may affect the inhibitory activity of GaTx2. Such details regarding the mechanism of inhibition still need to be addressed. The voltage-dependence and protocol dependence suggest that the mechanism by which GaTx2 inhibits ClC-2 is complex.
The physiological role of ClC-2 is still largely undefined. It is thought that ClC-2 may play a role in vascular smooth muscle cells and may be expressed on the apical membrane of epithelial cells along with CFTR, although this is still controversial (40). GaTx2 will be useful in determining the role of ClC-2 in these cells and may aid in determining the membrane localization of ClC-2 in specific cell types. Also, mutations in ClC-2 have been implicated in epilepsy (3), while under-activity of WT ClC-2 has been implicated in constipation-associated inflammatory bowel disease (5). Therefore, GaTx2 may serve as a lead compound for peptidomimetic drugs that target ClC-2.
GaTx2 is protected by a pending U.S. provisional patent, held by Georgia Tech Research Corporation. The authors declare no competing interests.     Representative TEVC experiment in which WT-CFTR channels were activated by a mixture of IBMX and db-cAMP. Channel activity increased rapidly, then decayed and settled to steady state level. Channels were then exposed to 20 nM extracellular GaTx2 for 5 minutes. Summary data presented in the bottom panel show that GaTx2 cannot inhibit CFTR from the extracellular side (n =2). (E) Inside-out multichannel patch recording of Flag-cut-ΔR-CFTR, V M = -80 mV. Channels were activated by application to the intracellular face of CFTR of either 0.2 mM ATP or 0.2 mM ATP + 60 nM GaTx2. Window currents for multiple 3-minute windows were analyzed in the absence and presence of GaTx2, for multiple patches. Summary data in the bottom panel show that GaTx2 cannot inhibit CFTR from the intracellular side (n = 10 windows). (F) TEVC recording of oocytes expressing GABA C channels (rho subunit) activated by 10 μM GABA (V M = -60 mV). Channels were exposed to 20 nM GaTx2 for 5 minutes then activated by GABA again. GaTx2 was then removed, while the oocyte was still exposed to GABA. The bottom panel shows that GaTx2 did not inhibit GABA C currents (n = 3), and GABA-activated current did not increase when GaTx2 was removed. (G) Inside-out macropatch recording of endogenous Xenopus Ca 2+ -dependent Clchannels, V M = -50 mV. Channels were activated by multiple applications of 10 mM Ca 2+ , and the amount of rundown was quantified. In some patches (lower traces), toxin was applied between the first and second application of Ca 2+ . The degree of rundown between the 1 st and 2 nd application of Ca 2+ did not differ between control patches and patches exposed to toxin (control, n = 5; GaTx2, n = 6). TEVC recording of Shaker ShB-IR K + channels exposed to 20 nM GaTx2 (H), and K V 1.2 channels exposed to 10 nM GaTx2 (I). Channels were activated by a voltage pulse to 0 mV from a holding potential of V M = -90 mV. Summary data show that GaTx2 was not capable of inhibiting Shaker ShB-IR K + channels (n = 2), or K V 1.2 channels (n = 4). Onset of inhibition occurred rapidly (n = 5 for this concentration), while recovery from inhibition was very slow, taking >7 minutes to return to baseline. (C) Dose-response curve for inhibition of ClC-2 currents by synthetic GaTx2 at V M = -100 mV. Each point represents data from 3 -5 TEVC experiments, fit with a three-parameter Hill equation. (D) Plot of first order rate constants for the onset of inhibition (open squares, n = 3 -5), and recovery from inhibition (filled circles, n = 2 -3). The dashed line represents an estimate of the detection limit of our experimental system, constrained by both the solution exchange rate and the low frequency of voltage pulses required to allow channels to return to the closed state. Note that some error bars are smaller than the symbol.    GaTx2 cannot inhibit open ClC-2 channels. ClC-2 channels were studied using outside-out macropatches activated by a voltage step to V M = -100 mV. Channel activity was allowed to reach steadystate. Patches were then exposed to either 2 nM GaTx2 (A), or 1 mM DPC (B). Summary data shown in (C) and (D) indicate that while GaTx2 was unable to inhibit ClC-2 steady-state currents from open channels (n = 5) (C), DPC was able to inhibit very well (n = 9) (D). The inset in (A), on the other hand, shows that if GaTx2 was applied to channels during the activation phase in this protocol, then currents were inhibited.