Trapping of ivermectin by a pentameric ligand-gated ion channel upon open-to-closed isomerization

Ivermectin (IVM) is a broad-spectrum anthelmintic drug used to treat human parasitic diseases like river blindness and lymphatic filariasis. By activating invertebrate pentameric glutamate-gated chloride channels (GluCl receptors; GluClRs), IVM induces sustained chloride influx and long-lasting membrane hyperpolarization that inhibit neural excitation in nematodes. Although IVM activates the C. elegans heteromeric GluClα/β receptor, it cannot activate a homomeric receptor composed of the C. elegans GluClβ subunits. To understand this incapability, we generated a homopentameric α7-GluClβ chimeric receptor that consists of an extracellular ligand-binding domain of an α7 nicotinic acetylcholine receptor known to be potentiated by IVM, and a chloride-selective channel domain assembled from GluClβ subunits. Application of IVM prior to acetylcholine inhibited the responses of the chimeric α7-GluClβR. Adding IVM to activated α7-GluClβRs, considerably accelerated the decline of ACh-elicited currents and stabilized the receptors in a non-conducting state. Determination of IVM association and dissociation rate constants and recovery experiments suggest that, following initial IVM binding to open α7-GluClβRs, the drug induces a conformational change and locks the ion channel in a closed state for a long duration. We further found that IVM also inhibits the activation by glutamate of a homomeric receptor assembled from the C. elegans full-length GluClβ subunits.


Results
Structural and basic functional properties of the chimeric α7-GluClβ receptor. In this study, we have used a chimeric Cys-loop receptor that consists of five identical subunits. Each subunit is a chimera generated by fusing the N-terminal extracellular sequence of the neuronal α 7 nicotinic acetylcholine receptor (α 7-nAChR) subunit to the C-terminal transmembrane and intracellular sequence of the C. elegans GluClβ subunit (GLC-2) (Supplementary Fig. S1) 80,81 . This chimeric α 7-GluClβ Cys-loop receptor robustly responds to ACh that binds to the extracellularly facing ligand-binding domain (LigBD) to open a Cl − -selective ion channel pore 80,81 . Figure 1A and B, presents a homology model of the chimeric α 7-GluClβ R. Figure 1C depicts a 3-D structure of IVM. Figure 2 shows representative ACh-evoked currents and a dose-response curve indicating an ACh-EC 50 value of 46 ± 14 μ M and a Hill coefficient of 2.55 ± 0.3. These values are very close to previous determinations 81 . Application of IVM prior to ACh inhibits the responses of the α7-GluClβR. IVM can activate the C. elegans homomeric GluClα and heteromeric GluClα /β receptors expressed either in Xenopus oocytes [4][5][6][7][8]61,82 , HEK cells 78 or CHO cells 79 , but it cannot activate the C. elegans homomeric GluClβ R expressed in these cells 4,7,8,76,78,79 . Since, in the GluClα cryst R, IVM binds at the periphery of the ion-channel pore domain 77 , we reasoned that IVM would not activate the chimeric α 7-GluClβ R whose ion-channel pore domain is that of the GluClβ R. On the other hand, previous studies performed with the α 7-nAChR showed that IVM potentiated the ACh-elicited currents and decreased the ACh-EC 50 by 20 fold 54,55 . Hence, we were wondering whether IVM would be able to potentiate the response of the chimeric α 7-GluClβ R to ACh. Our curiosity emerged since, in the GluClα cryst R, IVM forms a few van der Waals interactions with the M2-M3 loop 77 that is located at the interface between the LigBD and the pore domain and interacts with entities of the LigBD to gate Cys-loop receptors. For example, in the GABA A R 83-85 , GlyR 86,87 , 5HT3 A R [88][89][90] , muscle nAChR [91][92][93] , neuronal α 7-nAChR [94][95][96] , as well as in the chimeric α 7-GlyR 97 and α 7-5HT3 A R 98 . Figure 3 shows robust responses to ACh alone (I P1 ) or to ACh applied immediately after pre-application of 0.01% DMSO (the content in the IVM-containing solution) (I P2 ). The DMSO-containing solution had no significant effect on the current amplitude (I P2 /I P1 of 95 ± 1.8%; P > 0.05, N = 42, paired, two-tailed Student's t-tests). In contrast, pre-application of 5 μ M IVM for 10 sec reduced the ACh-elicited response to 16 ± 5.2% of the response obtained immediately after the DMSO pre-application (percent I P3 /I P2 ; P < 0.001, N = 8) (Fig. 3). Notably, the initial responses to ACh could not be recovered even after a long-term wash (* > 2 min) (e.g., Fig. 3, I P4 ), and their magnitude was 4.4 ± 1.3% of the responses obtained after the DMSO pre-application (percent I P4 /I P2 ; P < 0.001, N = 5).
Since the ACh-evoked response could not be recovered after exposure of the cell to IVM, we had to perform the above application protocol in different cells for each IVM concentration separately. By that, we obtained the inhibition graph shown in the inset of Fig. 3. The IVM-IC 50 was found to be 156 nM, and the Hill coefficient ∼ 3. The results in Fig. 3 may indicate that IVM binds to the resting (closed) state and cannot dissociate from the receptor during the application of ACh. We however cannot exclude the possibility that the immediate application of ACh after the application of IVM stabilized the receptor in a pre-open state for a long duration. We therefore changed the application protocol and introduced intermediate washes between the IVM and ACh applications. Clearly, neither 5-nor 10-sec-long intermediate washes could prevent the strong inhibitory effect of IVM pre-application (I P3 and I P3 /I P2 in Fig. 4A and B, respectively). Furthermore, a long-term wash introduced after the first recovery trial (Fig. 4A, asterisk), was followed by even a greater inhibitory effect (I P4 and I P4 /I P2 in Fig. 4A and B, respectively).
Since the GluClα cryst R indicates that IVM is accommodated at the protein-lipid interface 77 , we examined the effect of the membrane voltage on the IVM inhibitory effect. Similar effects have been observed at − 60 mV and − 20 mV (Fig. 4B), suggesting that the electric field of the membrane does not influence IVM association with the chimeric α 7-GluClβ R.
IVM accelerates the decline of ACh-elicited macroscopic currents in cells expressing α7-GluClβ receptors. Application of 350 μ M ACh (saturating concentration) for 5.5 sec on cells expressing the chimeric α 7-CluClβ R, resulted in currents that rapidly reached a peak and then declined to 31 ± 8.6% (N = 5) of the peak amplitude (Fig. 5A, black trace). We took advantage of the slow and partial current decline observed here, and examined the capability of IVM to associate with the α 7-GluClβ Rs after activation, during the phase of current decline. To this end, immediately after reaching to the ACh-evoked peak, we added 5 μ M IVM (dissolved in 0.01% DMSO) to the ACh-containing solution. The red colored current trace in Fig. 5A clearly indicates that IVM accelerates the current decline, and reduces the amplitude of the steady-state current. The grey current trace (Fig. 5A) corresponds to a control showing that 0.01% DMSO does not substantially change the response to ACh alone. Statistical analysis of the time constant of current decline (τ d ) indicates that DMSO alone had no significant effect on τ d , neither at − 60 mV nor at − 20 mV (Fig. 5B). In contrast, IVM reduced the τ d values by more than two folds at both membrane voltages, − 60 mV and − 20 mV, to a statistically similar extent (Fig. 5B). These results reinforce our above suggestion that the electric field of the membrane does not influence IVM association with the chimeric α 7-GluClβ R.
How does IVM decrease Cl − ion flow through the chimeric α7-GluClβR? The α 7-nAChR is a fast desensitizing receptor whose current response rapidly declines in the presence of saturating ACh concentrations due to channel closing 96,[98][99][100][101][102][103][104][105][106][107][108][109][110] . Compared with the α 7-nAChR, in the case of the chimeric α 7-GluClβ R studied here, the current declines much slower under saturating ACh concentration (e.g., Fig. 5A, black trace). Therefore, Figure 3. Inhibitory effect of IVM pre-application on the response of the α7-GluClβR to ACh. Following the first ACh application, which gave peak I P1 , the cell was washed with a physiological solution for 40 sec. Then, as a control, 0.01% DMSO was applied for 10 sec prior to the second ACh application that gave peak I P2 . Next, following a 40-sec-long wash, 5 μ M IVM (in the presence of 0.01% DMSO) was applied for 10 sec before its replacement by ACh (giving peak I P3 ). Finally, following a 2-min-long wash (*), ACh was applied again (giving peak I P4 ). In all applications, 50 μ M ACh was applied for 0.5 seconds. Inset, a curve corresponding to the inhibition of ACh (50 μ M)-elicited currents by increasing concentrations of IVM. Data were extracted from experiments performed as in panel A. Since the currents could not be recovered following an exposure to IVM, each cell was exposed only to one concentration of IVM. Then, to provide the normalized ACh response, the peak current recorded in response to ACh immediately after IVM application (i.e., I P3 ) was divided by the peak current recorded in response to ACh immediately after DMSO application (i.e., I P2 ). The curve was fitted to the averaged data points with a non-linear regression using the Hill equation 2 (Eq. 2; r 2 = 0.996). Error bars correspond to SEM. Total of 41 cells were used to obtain the averaged points. Recordings were performed at − 60 mV.
Scientific RepoRts | 7:42481 | DOI: 10.1038/srep42481 over a relatively long-term recording of inward currents through the Cl − -selective α 7-GluClβ R, the driving force acting on Cl − ions may decrease due to the outflow of Cl − ions and the decrease in their intracellular local concentration near the channels during the recording time window. A decrease in intracellular Cl − ion concentration in the vicinity of the channels has already been suggested for slowly-desensitizing GABA and Gly Cys-loop receptors 111 . To distinguish between desensitization (channel closing) and a decrease in the driving force acting on Cl − ions, we activated the receptors with 30 μ M ACh, and after the current reached to its peak, we performed voltage ramps throughout the current-declining phase, as has been performed previously 111 . Figure 6A presents an example for such an experiment, where the output currents of the voltage ramps are obtained in the absence of IVM. Superimposition of the voltage-ramps' output currents shows a moderate decrease in slopes (Fig. 6A, right). The current-voltage relations (I/V curves) extracted from the output currents (Fig. 6C) indicate that, during the application of ACh alone (in the absence of IVM), the reversal potential (E rev ) decreases from − 5.3 mV to − 28.9 mV. The black squares in Fig. 6E correspond to the E rev values determined for each voltage ramp when ACh ; and a fourth peak (I P4 ) after further long-term wash (* > 100 sec). In all cases, 50 μ M ACh was applied for 0.5 seconds. DMSO (0.01%) and IVM (5 μ M; containing 0.01% DMSO) were applied for 4 seconds. Washes performed immediately after the first and second ACh applications were 40-sec long (not indicated by a bar above the trace). Intermediate washes (IW) were applied for 5 seconds. Recording of this specific trace was performed at − 60 mV. (B) Histograms of data extracted from experiments performed as in A, which include a 4-sec-long pre-application of IVM (5 μ M), followed by a 5-or 10-sec-long intermediate wash (IW) and an ACh (50 μ M) application (for 0.5 sec), at − 60 or − 20 mV membrane voltages. Each current peak (I P3 or I P4 ) was normalized to the current peak recorded after the application of 0.01% DMSO (I P2 ) in the same cell (Norm. ACh Resp. stands for normalized acetylcholine response). Each cell was exposed to IVM only once, because the currents could not be recovered even after a long-term wash. The bars correspond to the averaged normalized currents with error bars corresponding to SEM of N cells, as indicated within the figure. The time of the intermediate wash (IW) is indicated at the bottom. Paired, two-tailed Student's t-tests indicate that ***P < 0.0005 and ** 0.0005 < P < 0.005. Two-way analysis of variance (ANOVA) between the I P3 /I P2 ratios obtained following 5-sec-long or 10-sec-long intermediate washes, at − 60 mV or − 20 mV, indicates that there is no difference between the extent of inhibition observed under these four conditions (P > 0.05). The same two-way ANOVA probability (P > 0.05) was obtained for the I P4 /I P2 ratios.
Scientific RepoRts | 7:42481 | DOI: 10.1038/srep42481 is applied alone. The leftward (negative) shift in the E rev values observed in the absence of IVM (∆ E rev = 23.6 mV; Fig. 6C and E, black squares) reflects a change in the Nernst potential for Cl − ions. By using the experimental E rev values and the Nernst equation (Eq. 4), we calculated the intracellular concentrations of Cl − ions ([Cl] i ) and found that it is 122.7 mM at the beginning and 48.9 mM at the end of the current-decline phase (r1 and r13 in Fig. 6A, respectively). Therefore, when measuring inward Cl − currents over 9.6-sec-long applications of ACh alone (e.g., Fig. 6A), a significant component of the current decline can be attributed to a decrease in the electrochemical driving force (V DF ) acting on Cl − ions (V DF = V m − E rev ). V DF probably decreases because Cl − ions flow out of the cell, and [Cl] i in the vicinity of the channels decreases with no capability to recover within the time window of the recording.
We applied the same experimental protocol on the representative cell shown in Fig. 6A, but in the presence of IVM that was added immediately after the current reached to its peak (Fig. 6B). In this experiment, we used 4-μ M-IVM solution that contained 0.008% DMSO. Superimposition of the voltage ramps' output currents recorded in the presence of IVM shows a sharper decrease in slopes (Fig. 6B, right) than in the absence of IVM (Fig. 6A, right). The I/V relations indicate that, within the recording time in the presence of IVM, the reversal potential (E rev ) decreases from − 7.9 mV to − 18.9 mV (Fig. 6D). The leftward (negative) shift in the reversal potential observed in the presence of IVM (∆ E rev = 11 mV; Fig. 6D and E, red circles) reflects a much smaller change in the Nernst potential for Cl − ions, as compared with the shift observed in the absence of IVM. The I/V relations allowed us to determine the change in the chloride conductance throughout the 9.6-sec application time. For that, we have calculated the chloride chord conductance at − 65 mV (Eq. 5), which turned to be virtually the same as the chloride chord conductance at + 15 mV. Figure 6F shows exponential curves fitted to the chord conductance points (Eq. 6) that were calculated from the currents shown in Fig. 6A and B. Clearly, in the presence of 4 μ M IVM, the chloride conductance declines much faster (red curve) than in the absence of IVM (black curve) (Fig. 6F). Using the experimental E rev values, we calculated the [Cl] i and the V DF for each voltage  ramp output current at − 65 mV, as has been performed above for the ACh application in the absence of IVM. While the changes in [Cl] i and V DF were slower and weaker in the presence of IVM (Fig. 6H) than in the absence of IVM (Fig. 6G), the decline of the conductance and current was much faster and stronger in the presence of IVM (Fig. 6H) than in the absence of IVM (Fig. 6G). Moreover, the current decline and the conductance decline in the presence of IVM overlap. Hence, we suggest that the conductance decline observed in the presence of IVM reflects structural changes in the ion conductance pathway, which eventually prevent Cl − ions to cross the ion-channel pore.
To determine the kinetics of conductance decline at lower IVM concentrations we had to use different cells since we could not reproduce the initial robust response (the one shown in Fig. 6A) after the exposure of the cell to IVM, even following a long-term wash (70 seconds) ( Supplementary Fig. S2). Figure 6I and J shows the conductance decline in representative cells under IVM concentrations lower than 4 μ M. The entire analysis described in this section was applied to 26 cells; and the averaged normalized conductance decline, under three different IVM concentrations, is presented in Fig. 6K-M.
We then took advantage of the capability of IVM to accelerate the current decline in order to determine the binding affinity of IVM. To this end, the voltage ramps' output currents obtained in the presence of IVM were subtracted from their counterpart currents obtained in the absence of IVM. Then, we calculated the chloride conductance based on the resulting subtracted currents at each ramp. This chloride conductance is hereafter termed G IVM . In each experimented cell, we fitted exponential curves to the first 5-6 conductance points, where there was no substantial difference between the ∆ E rev values in the absence or presence of IVM. These exponential fits for the G IVM points were obtained at various IVM concentrations, and they actually provide the time constant of the IVM-dependent Cl − -conductance decay (τ IVM ). Figure 6N shows that the averaged reciprocal τ IVM values increase linearly with the increase in IVM concentrations.
IVM does not readily dissociate from the receptor ( Supplementary Fig. S2) and the number of possible intermediate IVM-bound states is not known. Hence, the simplest possible kinetic model that could describe the acceleration of current decline in the presence of IVM would be one in which an ACh-activated conducting receptor (R* ACh ) closes a gate in the ion-channel pore upon IVM binding (IVM·R ACh ), as follows: Figure 6N shows a curve fitted to the 1/τ IVM points by a linear regression using equation 7. The slope of this curve corresponds to the IVM association rate constant (k forward, is the extrapolated intercept of the linear curve with the y-axis in Fig. 6N. The K d (k b /k f ) for IVM binding to an activated α 7-GluClβ chimeric receptor would therefore be 1.4 × 10 −6 M. This K d value is nine times higher than the IC 50 value observed when IVM is applied prior to activation (156 nM; Fig. 3).

IVM becomes trapped during the receptor open-to-closed isomerization. The residency time for
a drug bound to its target is provided by 1/k b 112 . The results of the previous section indicate that, for IVM, the residency time is ~2.5 sec, which means that IVM should have been removed within a wash of a few seconds. However, as said, the initial robust ACh-elicited current response (e.g., Fig. 6A) could not be reproduced following cell exposure to IVM, even after a wash lasting 70 seconds (Supplementary Fig. S2). To better understand the effect of IVM, we applied ACh (alone) for 9.6 seconds and then re-applied the same ACh concentration (alone) several times, each for 350 millisecond. Figure 7A and C (black bars) shows that when re-applications were performed 5, 10, 20 and 40 seconds after the end of the initial (long-term) ACh-application, the current amplitudes did not fully recover. The unrecovered current portion may indicate that the [Cl] i near the channels and the V DF at these time points have not been restored to initial values. We however do not exclude the possibility that part of the current is missing due to acute desensitization 113 , which may result in reversible stabilization of the receptor in a deeply desensitized (closed) state -as has been suggested for nAChRs in midbrain dopamine neurons 114 and recombinant α 4β 2 neuronal nicotinic AChRs 115 (reviewed by Giniatullin et al. 116 ). Figure 7A and C (black bars) also shows that it takes at least 70 seconds for the macroscopic current amplitude to recover. This recovery excludes the possibility of functional rundown since there was no permanent inactivation as described, for example, in recombinant α 3β 2 117 , α 4β 2 118 and α 7 119 nAChRs. When IVM was combined with ACh after the current reached to its peak in the long-term application, the current amplitude could not be recovered even after 100 sec of wash ( Fig. 7B and C, red bars).
To examine if IVM can be removed from the receptor by longer wash periods, we extended the washing time. Figure 8A shows a control experiment, where ACh was applied alone for 9.6 seconds followed by two additional short-term ACh applications (350 milliseconds), with two intermediate washes each lasting 3 minutes. As can be seen in Fig. 8A, the initial (long-term) ACh application resulted in a response that reached a peak (I P1 ) and declined with one exponential time course (τ d 1, Fig. 8C). The current remaining at the end of the initial ACh application was 35 ± 4% of the initial peak (percent I S /I P1 in Fig. 8D, black bars). Figure 8A and D (black bars) also shows that the initial current amplitude (I P1 ) was completely recovered (I P3 /I P1 ). In contrast, when IVM was added to ACh immediately after the current reached to its peak (Fig. 8B, leftmost response), the current declined with double exponential time course. One of the two time constants (τ d 2) is significantly smaller than the time constant measured in the presence of ACh alone ( Fig. 8C; P < 0.0001, unpaired, two-tailed Student's t-test). In addition, the steady-state current (I S ) was found to be 6% ± 0.01% of the initial amplitude, significantly smaller than in ACh alone (Fig. 8D, percent I S /I P1 , red bar vs black bar). Furthermore, very small responses could be seen upon re-application of ACh alone after 3 and 6 minutes of wash (Fig. 8B, inset; Fig. 8D, I P2 /I P1 and I P3 /I P1 in red). Since, in the control experiments, the current could be recovered almost completely after a 3-min wash (Fig. 8A, I P2 ; Fig. 8D, I P2 /I P1 in black), we conclude that the [Cl] i in the vicinity of the channels and V DF largely recovered within the 3-min wash. When IVM is added, the recovery of the [Cl] i and V DF is expected to take place even sooner, because compared with the control, the current declines much faster to I S and I S /I P1 is much lower (Fig. 8A  vs B; Fig. 8C and D , I S /I P1 , black vs red bars). Hence, the incapability of ACh to elicit currents after exposure to IVM (Fig. 8B, I P2 and I P3 ; Fig. 8D, I P2 /I P1 and I P3 /I P1 in red) strongly indicates that IVM stabilizes, for a long duration, a channel state having an obstructed ion-conduction pathway.

Does IVM accumulate in the CHO cell plasma membrane during the time of its application?
The capability of a fluorescent ivermectin probe to partition into biological membranes and diffuse laterally in the lipid bilayer was previously demonstrated following its incubation for 30 minutes with muscle vesicle membranes prepared from the parasite Ascaris sum 120 . The lipophilic properties of IVM were also demonstrated by incubating Times indicated above the short bars were measured from the end of the initial, long-term ACh application (that gave I P1 and I S ). (B) Representative currents for the same experiment as in A, but with IVM (0.5 μ M) added after peak I P1 . Currents in A and B were measured at − 60 mV. (C) Normalized ACh response (y-axis) calculated by the current peak ratios that are indicated below the x-axis. Current peak ratios were extracted from experiments like those shown in A and B. Black bars correspond to results obtained with ACh alone, as in panel A (N = 8). Red bars correspond to ACh-elicited currents where IVM was added as in panel B (N = 5). Error bars correspond to SEM. Asterisks correspond to unpaired, two-tailed Student's t-tests with 0.001 < *P < 0.05, 0.0001 < **P < 0.001 and ***P < 0.0001. (D) Fractional removal of IVM from the aqueous phase of a 1-ml solutions containing 140,000 CHO cells and 4 μ M IVM. Removal of IVM was determined following the indicated incubation times by measuring the absorbance of supernatants at 245 nm wavelength (e.g., Supplementary Fig. S3; Supplementary Materials and Methods). Note that, no IVM removal is indicated at the top of the y-axis. Mean ± SEM of seven independent experiments are presented. Paired, two-tailed Student's t-tests indicate that for "ns" (not significant) P > 0.2, and ***P < 0.0001.

Figure 8. The inhibition of ACh-elicited currents by addition of IVM is irreversible. (A)
Representative currents elicited by a long-term application (9.6 sec) of ACh followed by two short-term (350 ms) applications of ACh, with intermediate long-term washes, each lasting 3 minutes. (B) Representative currents for the same experiment as in A, but with the exception of IVM (0.5 μ M) addition after the ACh-elicited current reached to its peak (I P1 ). Inset, magnification of the indicated segment of the current trace. In both panels A and B, black bars above the current traces indicate ACh or IVM applications. (C) Kinetics of current decline under longterm application (9.6 sec) of 30 μ M ACh, in the absence or presence of 0.5 μ M IVM (performed as described in A and B). In both cases, the application of ACh alone or the co-application of ACh with IVM, the kinetics of current decline was measured from 0.75 sec after the ACh application started until the end of the application. %A stands for the amplitude percentage. Error bars correspond to SEM of N = 10 cells per each condition (ACh alone or ACh + IVM). (D) Normalized ACh response (y-axis) calculated by the current peak ratios that are indicated below the x-axis. Current ratios were extracted from experiments performed as shown in A and B. Black bars correspond to experiments where all applications contained only ACh (e.g., panel A; N = 10). Red bars correspond to experiments where IVM was added during the initial, long-term application of ACh (e.g., panel B; N = 10). Error bars correspond to SEM. Unpaired, two-tailed Student's t-tests indicate *** P < 0.0001. IVM with artificial large unilamellar vesicles (LUVs; 6 mM) for 5 minutes, and determining the depletion of IVM from the aqueous solution following a further 20-minutes long ultracentrifugation (50 or 70 percent depletion, depending on the LUVs' lipid composition) 58 . Hence, one may argue that the irreversible inhibition of the chimeric α 7-GluClβ R by IVM is due to the accumulation of IVM in the cell membrane and the consequent increase of the effective concentration of the drug around the receptors. To the best of our knowledge, no data is available regarding the partitioning of ivermectin into the plasma membrane of CHO cells in a period as short as the time of IVM application in the current study (no longer than 10 seconds).
To assess the IVM partitioning into the CHO cell plasma membrane during a short time period, we followed the procedure of Silberberg et al. 58 with a few modifications accounting for our experimental conditions. To this end, we incubated 140,000 CHO cells (correspond to 3.2 μ M plasma membrane lipids 121 ) with 4 μ M IVM for either 0, 15 or 300 seconds, pelleted the cells by a brief centrifugation (30 seconds), and analyzed the extent of IVM removal from the aqueous solution (elaborated in the Supplementary Materials and Methods). Supplementary Fig. S3 shows examples of absorption spectra of supernatant obtained after the CHO cells' sedimentation. These spectra indicate that IVM dissolved in NES containing 0.02% DMSO has an ultraviolet absorption maximum at 245 nm, a wavelength used by others to detect IVM in other solvents 58,122 . The maximum absorbance values were used as a measure for the fraction of IVM removal during the incubation times. Figure 7D indicates that following 5 minutes of incubation, ∼ 24% of IVM was removed from the aqueous solution by the CHO cells. This is less than in artificial LUVs 58 , probably because of the differences in the: membrane lipid composition; IVM hydration conditions; duration of the post-incubation centrifugation; and lipid content (lower in our case as mentioned above). Importantly, within 15 seconds of incubation, no statistically significant amount of IVM was removed from the aqueous solution by the CHO cells ( Fig. 7D; e.g. Supplemental Fig. S3). Although these experiments are not sufficiently sensitive to provide the IVM concentration in the CHO cell membrane during the electrophysiological experiments, they indicate that the accumulation of IVM in the membrane is not substantial during the longest IVM applications we employed (10 seconds) (further discussed below).

Can IVM inhibit homomeric receptors assembled from the C. elegans GluClβ subunit?
The results presented so far show that IVM strongly inhibits the chimeric α 7-GluClβ R. Since IVM putatively binds in the ion-channel pore domain and this region of the chimeric α 7-GluClβ R is contributed by the C. elegans GluClβ subunit, an inevitable question emerged: would IVM inhibit homomeric receptors assembled from the C. elegans GluClβ (GLC-2) subunit ? To answer this question, we attempted to express GluClβ R homomers in CHO cells, although in a previous study such functional homomers were rarely expressed in CHO cells and provided small responses to 10 mM Glu 79 . Currently, we have used a more powerful transfection reagent (Supplementary Materials and Methods) and challenged the cells with 100 mM Glu. Four of 46 cells gave reliable currents of hundreds of picoamperes at + 60 mV (e.g., Fig. 9A and B). These responses were reminiscent of the behavior displayed by homomeric GluClβ Rs expressed in Xenopus oocytes 4,123 . That is, (i) the responses desensitized (e.g., Fig. 9A) as previously shown under a saturating Glu concentration (see Fig. 7A in Etter et al. 123 ), and (ii) the responses at + 60 mV were larger than the responses obtained at − 60 mV, with I + 60mV /I −60mV = 2.8 ± 0.2 (mean ± SEM) (e.g., Fig. 9A, I P1 /I P0 ). Accordingly, the I/V curve rectifies outwardly (e.g., Fig. 9B), albeit with less steepness than that observed previously in Xenopus oocytes 4 . This difference in the I/V-curve steepness is partially because we have used approximately equimolar extra-and intracellular Cl − ion concentrations (150.8 mM and 138 mM, respectively), whereas the extracellular and calculated intracellular Clion concentrations used in the experiments with Xenopus oocytes were 122.6 mM and ~33 mM, respectively 4 . Notably, the responses of the other 42 cells that we patched were very small (10.8 ± 2.6 pA at + 60 mV and 31.1 ± 5.6 pA at − 60 mV), providing I + 60mV /I −60mV = 0.26 ± 0.06; and therefore these responses were considered as nonspecific and irrelevant to the homomeric GluClβ R. Figure 9A and C, also shows that pre-application of 2 μ M IVM for 10 seconds leads to the inhibition of ∼ 67% of the response of the homomeric GluClβ Rs to 100 mM Glu (I P2 /I P1 ). Further application of 100 mM Glu after a wash of 25 seconds provided peak I P3 , which revealed that most of the initial response disappeared ( Fig. 9A and C, I P3 /I P1 ). Note that, in spite of the efficient perfusion system, which provides a time constant of activation at the millisecond range (33.1 ± 3.7 ms), the Glu-elicited currents slowly returned to baseline upon the application of wash. This slow current decline is probably due to the very high concentration of the applied Glu (100 mM), which is ~260 times the EC 50 determined in Xenopus oocytes (380 μ M) for the homomeric GluClβ R 4 .

Discussion
Cys-loop receptor chimeras have long been used to shed light on the interplay between the ligand-binding domain and the ion-channel pore domain 55,80,81,97,98,[124][125][126][127][128][129] . Here we used a highly expressible, homomeric α 7-GluClβ chimeric receptor in an attempt to understand why IVM does not activate the C. elegans homomeric GluClβ R although it can activate the C. elegans heteromeric GluClα /β R. The rationale behind our decision to work with a chimera having the extracellular domain of the α 7-nAChR was based on the assumption that the LigBD-pore domain interface might affect IVM action. This assumption emerged because of the following reasons. (i) IVM strongly potentiates ACh-elicited currents in the α 7-nAChR 54 . (ii) In the GluClα cryst R, the disaccharide and lactone moieties of IVM form van der Waals interactions with the M2-M3 loop 77 that is located at the LigBD-pore domain interface. (iii) The M2-M3 loop of Cys-loop receptors interacts with the β 1β 2, Cys and β 8β 9 loops of the ligBD as well as with the pre-M1 segment to couple agonist binding to channel gating 50,83-98,130-136 . (iv) Incorporation of the GluClβ subunit in GluClR assemblies increases the sensitivity of the receptor to IVM 78 , which suggests that IVM interacts with the GluClβ subunit. We therefore expected that IVM would, at least, potentiate the response of the chimeric α 7-GluClβ R to ACh. To our great surprise, IVM was found here to strongly inhibit the chimeric α 7-GluClβ R.
Interestingly, IVM negatively modulates a few other Cys-loop receptors 55,137,138 . Yet, little is known about the mechanism of the IVM inhibitory phenomenon. Here, based on experiments where IVM was applied prior to ACh, we suggest that IVM can bind to the resting (closed) state of the α 7-GluClβ R and stabilize this conformation for a long duration. However, we do not exclude the possibility that the stabilized conformation corresponds to a desensitized-like state, since unliganded as well as liganded Cys-loop receptors can interconvert from the resting state to a desensitized state directly without opening; for instance, the neuronal α 4β 2 nAChR 115 .
IVM binds in the periphery of the ion-channel pore domain of the GluClα cryst R 77 , and it was likewise suggested to bind in a homologous peripheral site in the GlyR 137 . Assuming that IVM also binds in the periphery of the ion-channel pore domain of the chimeric α 7-GluClβ R, then the possibility of competitive inhibition by IVM is excluded. Uncompetitive inhibition, which requires receptor-activator complex formation prior to binding of the inhibitor 139 , is also excluded because IVM can bind already before ACh association with the α 7-GluClβ R (e.g., Fig. 4). Since non-competitive inhibition can occur with or without the presence of the activator, IVM most likely acts as a non-competitive inhibitor of the chimeric α 7-GluClβ R (e.g. , Figs 4 and 5). Therefore, the IC 50 determined in Fig. 3 (156 nM) actually corresponds to the inhibition constant (K i ) 139 .
The response of the chimeric α 7-GluClβ R to ACh declines slowly after reaching to its peak. This decline could be attributed, at least in part, to a loss of electrochemical driving force for Cl − ions, as discussed in the "Results". However, one cannot exclude the possibility that a fraction of this decline corresponds to channels' closing in the presence of the agonist due to desensitization. When IVM was added to an already activated receptor population, immediately after the response reached to its peak, the decline of the current became faster and the extent of current loss at the end of the co-application period was larger. As discussed in detail in the "Results" (Fig. 6), the faster and stronger decline in the Cl − conductance and the lesser decrease in [Cl] i and V DF during the co-application of ACh with IVM, as compared with the application of ACh alone, strongly suggest that, in the presence of IVM, the receptor adopts a conformation that hinders chloride ion flow. The exponential fits to the specific IVM-dependent Cl − -conductance (G IVM ) decay (Fig. 6) allowed us to obtain the affinity of the α 7-GluClβ R for IVM when the drug is applied after activation (K d = 1.4 μ M; Fig. 6N). When IVM is applied prior to the activation by ACh, the affinity of the receptor for IVM is nine times higher (K i = 156 nM; Fig. 3). Hence, we suggest that IVM can bind to two distinct receptor states. One state is possibly the resting state and the other state is likely to be an open state of the α 7-GluClβ R. We further suggest that IVM binding to the open state induces a conformational change that results in pore closure. In respect to IVM association with the open state, the aforementioned kinetic model should therefore be revised as follows.
where k f and k b are the rate constants described above in the first kinetic model, and α and β are the rates of receptor interconversion between the IVM-bound open state (IVM⋅ R* ACh ) and the IVM-bound non-conducting (closed) state (IVM⋅ R C ACh ). We could not determine the receptor interconversion rates, but it is reasonable to assume that if α ≫ β, then the IVM-bound closed conformation would be stable for a long duration.
Most recently, the three dimensional structure of a homomeric α 1 GlyR from the zebrafish (termed GlyR EM ) was elucidated by electron cryo-microscopy (ECM) in conformations that correspond to the receptor complexed with either glycine alone or with both glycine and IVM 140 . These ECM structures indicate that the ion-channel pore of a receptor occupied by both glycine and IVM is narrower than that of a receptor occupied by glycine alone. Yet, the most constricted portion of the doubly occupied (glycine/IVM) GlyR EM has a diameter of 5 Å, which is sufficiently wide to allow Cl − ions to pass 140 . We speculate that the association of IVM with ACh-occupied open α 7-GluClβ R leads to constriction of the ion-channel pore, but unlike the case of the doubly occupied GlyR EM , here the constriction of the α 7-GluClβ R is not sufficiently wide to allow Cl − ions to flow across the ion-channel pore.
The capability of ACh-elicited currents to recover after the first response to long-term application of ACh alone ( Fig. 7A and C, black bars), clearly indicates that no change in the size of the receptor population has occurred. This conclusion excludes the possibility that functional rundown accompanied with permanent inactivation took place. Hence, the gradual recovery seen in Fig. 7A could indicate that reversible desensitization or/ and a decrease in V DF underlie the lack of current when ACh is reapplied alone 5 to 70 seconds after the end of the initial long-term ACh application (see also Fig. 7C, black bars). We therefore conclude that the small responses to ACh that were obtained after the co-application of ACh with IVM ( Fig. 7B and C, red bars) likely correspond to a receptor population that is 'stuck' in the above-mentioned channel-constricted conformation.
When IVM is added after the ACh-elicited current reached to its peak, it most likely associates with open receptors. The calculation of the time constant of G IVM decay (Fig. 6N) allowed us to determine the residency time of IVM bound to the α 7-GluClβ R (1/k b = ~2.5 sec). This value means that a few seconds after removal of water-soluble IVM from around the cell by a wash, the inhibitory effect should have been weakened due to IVM dissociation and lateral diffusion in the membrane. However, the responses were not even partially recovered after the first exposure of ACh-bound open α 7-GluClβ Rs to IVM (Figs 7 and 8), indicating that there might be an additional binding step that possibly causes trapping of IVM in the pore periphery. Yet, one cannot exclude with absolute certainty the possibility that the irreversibility of the inhibition could be due to a steady, high membranous IVM concentration encountered by the receptors, rather than an irreversible trapping of IVM by the receptors. This possibility arises because, within time, IVM tends to accumulate in lipid bilayers due to its lipophilic properties, and it cannot be removed from the lipidic phase under physiological conditions. Previous studies involving IVM accumulation in artificial lipid bilayers or biological membranes 58,120 are relevant to relatively much longer incubation times than used in the current study (detailed in the "Results"). Indeed, we found that no substantial accumulation of IVM in the membrane of CHO cells is apparent during a 15-sec-long incubation with each other (Fig. 7D and Supplementary Fig. S3). The longest time of exposure to IVM in the electrophysiological experiments performed here was 10 seconds. The perfusion exchange in our system removes soluble small molecules from around the cell within less than a second, as can be judged by the time of the current's return to the baseline when replacing ACh-containing solution by a wash (407 ± 34 ms, in the experiments represented by Figs 7A and 8A). This means that, after 10 seconds of IVM application, it probably takes less than 1 second to remove all the water-soluble IVM molecules from around the cell, as well. At this time point, no further IVM is available for binding by the receptors, except for those membrane-embedded IVM molecules that, as said above, do not likely reflect a substantial accumulation.
Noteworthy, the effect of IVM might take place due to IVM partitioning into the membrane at the site of its action, or by IVM molecules that partition into the membrane far from their targets and approach their binding sites by lateral movement within the lipid bilayer. Lateral movement of a fluorescently labeled IVM probe in biological membranes was previously determined to be slow, with a lateral diffusion coefficient (D L ) of 14⋅ 10 −10 cm 2 /s (see page 554 in Martin and Kusel 120 , under "The lateral mobility of the ivermectin probe"). In the six experiments represented by Fig. 5A (e.g., red trace), the time between the start of ACh or IVM applications and the onset of their effects was 67 ± 7 or 132 ± 24 milliseconds, respectively. Taking into account the differences in accessibility and complexity between the ACh and IVM binding sites, the two-fold slower onset of effect observed for IVM is reasonable if IVM acts at the site of its partitioning. In this context, it is intriguing to point out that in the GluClα cryst receptor the disaccharide and lactone moieties of IVM interact with the M2-M3 loop 77 that is putatively situated above the lipid leaflet in all Cys-loop receptors. So, partitioning of IVM into the membrane at the site of action perhaps involves a recognition step that occurs outside of the membrane (still have to be studied).
Collectively, our IVM partitioning experiments and the arguments raised above suggest that the concentration of IVM in the perfused solution approximates the concentration of IVM in the membrane at the end of the 10-sec-long IVM application. Hence, our calculation of k b based on the concentration of IVM in the perfused solution is a reasonable approximation. Still, one may argue that since the washes do not extract IVM from the membrane, the irreversibility may reflect an exchange between receptor-bound and membrane-resident IVM molecules. The counter argument is based on Fig. 6K, which indicates that 740 milliseconds after the application of 0.5 μ M IVM, there is already a statistically significant IVM-dependent decrease in the chloride conductance by 20 ± 2% (P = 0.0005, paired, two-tailed Student's t-test). This IVM concentration (0.5 μ M) does not exert the maximal inhibitory effect according to the declining rate of the chloride conductance (Fig. 6). Yet, we did not observe even a partial recovery of the initial response following IVM removal from the external solution and the subsequent ACh applications (Figs 7B and 8B); which likely indicates that IVM is trapped in its binding site(s) for a long duration.
In conclusion, the results presented here suggest that IVM can bind to both, the resting and active states of the α 7-GluClβ R. After initial binding of the drug to the active (open) state, it accelerates pore constriction to a degree that impedes Cl − ion flow through the ion-channel pore. We suggest that at the time the pore becomes constricted in a non-conductive conformation, IVM is likely trapped between the transmembrane segments for a long duration and it reciprocally locks the pore-lining segments. As such, gating motions necessary for re-opening of the ion-channel pore are prevented.
So far, it was not known why homomeric receptors assembled from the C. elegans GluClβ (GLC-2) subunit could not be activated by IVM 4,7,8,76 . The current study provides an answer to this question by demonstrating that IVM actually inhibits the responses of the C. elegans homomeric GluClβ R to Glu (Fig. 9). IVM likely binds in the periphery of the ion-channel pore domain, between adjacent subunits of the homomeric GluClβ R, akin to the case of the GluClα cryst R. If so, IVM is accommodated by the homomeric GluClβ R at β /β intersubunit interfaces. In a previous study, we inferred that the subunits of the C. elegans heteromeric GluClα (GLC-1)/β (GLC-2) receptor are arranged in an anticlockwise β -α -β -α -α fashion, as viewed from the extracellular milieu 79 . This subunit arrangement does not have β /β intersubunit interfaces. Hence, we further hypothesize that in the IVM-activatable heteromeric GluClα /β R, the subunit arrangement precludes potential restraints that could have emerged by IVM binding at such β /β interfaces. where I is the current response, I max is the maximal current response, EC 50 is the agonist effective concentration that elicits 50% of maximal current response, [ACh] is the concentration of acetylcholine, and n H is the Hill coefficient.

Materials and Methods
In the case of inhibition of the current response, curves were fitted to the data points by a nonlinear regression using the Hill equation 2, where I is the current response, I max is the maximal current response, IC 50 is the IVM concentration that inhibits 50% of the current response, [IVM] is the concentration of ivermectin, and n H is the Hill coefficient. The declining phase of the currents was fitted with an exponential time course to give the time constant of current decline by using equation 3, t P S / S or its bi-exponential variant where appropriate, using data that relates to ∼ 90% of the period between the peak of the current and the termination of a typical 5.5-or 9.6-sec-long ACh (or ACh + IVM) application (i.e., the fitting was performed from the point of IVM addition). I is the current, I P is the peak of the current, I S is the steady state current, t is the time, and τ is the time constant of current decline.
To calculate the changes in the intracellular concentration of Cl − ions ([Cl] i ) in the voltage-ramp experiments, the Nernst equation (4) was used as follows: [Cl] Cl rev o i where E Cl is the Nernst (equilibrium) potential for Cl − ions; E rev is the measured reversal potential; R and F are the gas and Faraday constants, respectively; T is the absolute temperature (298.15 K); z is the valance of a Cl − ion (− 1); [Cl] o is the extracellular concentration of Cl − ions; and [Cl] i is the intracellular concentration of Cl − ions. Chord conductance (G) was calculated by using equation 5 as follows: where G is the chloride conductance, I corresponds to the current measured at the respective membrane voltage (V m ), and E rev is the reversal potential. The relationship between the conductance values and time was plotted for each cell, and the time constant of conductance decay was determined by fitting an exponential curve to the data points using equation 6 as follows: Scientific RepoRts | 7:42481 | DOI: 10.1038/srep42481 t Cl Cl Cl / Cl or its bi-exponential variant where appropriate. G Cl (t) is the chloride conductance at a given time; G Cl (max) and G Cl (plat) are the chloride conductance values at the maximum and the plateau, respectively; t is the time; and τ is the time constant of the chloride-conductance decay to the plateau. To calculate the IVM association (k f ) and dissociation (k b ) rate constants, a curve was fitted to the 1/τ IVM values by a linear regression using equation 7 as follows: IVM f b where τ IVM is the time constant of the IVM-dependent Cl − -conductance decay, and [IVM] is the concentration of IVM.