An Engineered Glutamate-gated Chloride (GluCl) Channel for Sensitive, Consistent Neuronal Silencing by Ivermectin*

Background: Ivermectin (IVM) silences activity in neurons expressing glutamate-gated chloride channels (GluCl). Results: Rational point mutations in GluCl robustly increase IVM-induced conductance and reduce the variability of spike suppression in cultured neurons. Conclusion: The newly engineered receptor improves heteromeric receptor expression and sensitivity to IVM. Significance: GluClv2.0 is an improved tool for IVM-induced silencing. A modified invertebrate glutamate-gated Cl− channel (GluCl αβ) was previously employed to allow pharmacologically induced silencing of electrical activity in CNS neurons upon exposure to the anthelmintic drug ivermectin (IVM). Usefulness of the previous receptor was limited by 1) the high concentration of IVM necessary to elicit a consistent silencing phenotype, raising concern about potential side effects, and 2) the variable extent of neuronal spike suppression, due to variations in the co-expression levels of the fluorescent protein-tagged α and β subunits. To address these issues, mutant receptors generated via rational protein engineering strategies were examined for improvement. Introduction of a gain-of-function mutation (L9′F) in the second transmembrane domain of the α subunit appears to facilitate β subunit incorporation and substantially increase heteromeric GluCl αβ sensitivity to IVM. Removal of an arginine-based endoplasmic reticulum retention motif (RSR mutated to AAA) from the intracellular loop of the β subunit further promotes heteromeric expression at the plasma membrane possibly by preventing endoplasmic reticulum-associated degradation of the β subunit rather than simply reducing endoplasmic reticulum retention. A monomeric XFP (mXFP) mutation that prevents fluorescent protein dimerization complements the mutant channel effects. Expression of the newly engineered GluCl opt α-mXFP L9′F + opt β-mXFP Y182F RSR_AAA receptor in dissociated neuronal cultures markedly increases conductance and reduces variability in spike suppression at 1 nm IVM. This receptor, named “GluClv2.0,” is an improved tool for IVM-induced silencing.

range to raise concern about potential side effects because IVM is known to activate (or potentiate) other ligand-gated ion channels in the CNS, although it does so much less potently than GluCl (6 -9). In addition, the silencing phenotype displayed a bimodal distribution that correlated with both the extent of viral infection (i.e. the volume of striatum expressing GluCl ␣␤) and the extent of spike suppression (i.e. individual neurons showing either full, partial, or no inhibition), which further correlated with the co-expression levels of ␣ and ␤ subunits.
GluClv1.0 was thus a candidate for rational protein optimization that would 1) alleviate the concern of off-target side effects and 2) avoid suboptimal spike inhibition. Toward aim 1, we screened heteromeric GluCl ␣␤ for increased sensitivity to IVM by introducing channel gating mutations in the second transmembrane domain of the ␣ subunit. To employ IVM effectively as an orthogonal silencing tool, however, we had to reinstate glutamate insensitivity (by reintroduction of the ␤Y182F mutation) to this hypersensitive mutant receptor. For aim 2, we suspected that inadequate surface expression of the ␤ subunit was responsible for the variability in spike suppression. To improve heteromeric receptor expression at the plasma membrane, putative trafficking determinants of the ␤ subunit were mutated. An additional mutation designed to prevent fluorescent protein dimerization appeared to enhance heteromeric subunit assembly. Altogether, in vitro silencing experiments unequivocally show that the resulting mutant receptor is a more sensitive and less variable GluCl/IVM silencing tool.
Cell Culture-Human embryonic kidney 293 (HEK293) cells were purchased from ATCC (catalog no. CRL-1573). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM; catalog no. 11965, Invitrogen) supplemented with 10% FBS (catalog no. 26140, Invitrogen), 100 units/ml penicillin, 100 g/ml streptomycin (catalog no. 15140, Invitrogen), and 1 mM sodium pyruvate (catalog no. 11360, Invitrogen) and maintained at 37°C and 5% CO 2 in a humidified incubator. Cells were passaged when confluent at a subcultivation ratio of 1:5 or 1:10 every 3-4 days. For FlexStation assays, HEK293 cells were plated at 20,000 cells/well, with a plating volume of 100 l/well, in a black-sided/clear-bottomed 96-well imaging plate (catalog no. 353219, BD Falcon). For transfection, 16 g of total DNA in 750 l of DMEM was mixed with 30 l of ExpressFect in 750 l DMEM, preincubated for 20 min, and then added at 15 l/well to cells containing 100 l of fresh culture medium. Transfection mixes were removed from cultures following a 4 -6-h incubation period at 37°C/5% CO 2 and replaced with fresh culture medium. For total internal reflection fluorescence (TIRF) imaging, HEK293 cells were plated on 35-mm glass bottom culture dishes (catalog no. P35G-1.5-10-C, MatTEK) at 50,000 cells/dish. Cells were transfected with 1 g of DNA in 100 l of DMEM plus 4 l of ExpressFect (catalog no. E2650, Denville Scientific) in 100 l of DMEM that was preincubated for 20 min before adding to culture dishes containing 2 ml of fresh culture medium. The transfection mix was removed after 4 -6 h and replaced with fresh culture medium. For all experiments, HEK293 cells were transfected 24 h after plating and assayed 48 h after transfection.
Hippocampal neurons were obtained from day 18 Wistar rat embryos (10) and plated at a density of 40,000 cells/dish on 35-mm glass bottom culture dishes coated with poly-DL-lysine (catalog no. P9011, Sigma). Neurons were cultured in Neurobasal medium (catalog no. 21103-049, Invitrogen) containing 2% B27 (catalog no. 17504-044, Invitrogen) and 0.5 mM Glutamax (catalog no. 35050, Invitrogen). Medium was supplemented with 5% equine serum (catalog no. SH30074, Hyclone) during plating. Cultures were maintained at 37°C and 5% CO 2 in a humidified incubator with a 50% medium exchange once per week. For imaging experiments, neurons were treated with 1 M cytosine arabinoside (catalog no. C1768, Sigma) on culture day 10, followed by a 100% medium change the next day. Cultures used for electrophysiological experiments were not treated with cytosine arabinoside. Neurons were transiently transfected after 13-14 days in culture and assayed 24 h later. Transfections were prepared using 4 g of plasmid DNA with 20 g of Nupherin-neuron (catalog no. SE-225, BIOMOL) and gen) diluted separately in 400 l of Neurobasal without phenol red (catalog no. 12348-017, Invitrogen). The diluted reagents were preincubated separately at room temperature for 15 min and then combined and incubated for another 45 min. An 800-l volume of conditioned medium was then removed from the neuronal culture dish and replaced with the 800-l transfection mix. After incubating cultures for 1 h at 37°C/5% CO 2 , an 800-l volume was removed from the dish and replaced with the original 800 l of conditioned medium.
Fluorescent Membrane Potential Measurements-Fluorescent membrane potential assays were performed using the FlexStation 3 multimode benchtop microplate reader and the BLUE formulation kit (catalog no. R8042, Molecular Devices). Dye loading buffer was prepared according to the instructions of the manufacturer. To perform the assay, culture medium was removed from the cells and replaced with 50 l of DMEM. Cells were then loaded with 50 l of blue dye loading buffer and incubated for 40 min at 37°C/5% CO 2 . Fluorescence was measured at ambient temperature using the "Low PMT" setting of the FlexStation 3 operated by SoftMax Pro Data Acquisition and Analysis Software. Excitation and emission wavelengths were set at 530 and 565 nm, respectively. The emission cut-off was 550 nm. Other FlexStation parameters included a pipette dispensing height of 230 l for a 300-l volume well, an initial well volume of 100 l, a transfer volume of 50 l (therefore, drug concentrations were prepared three times), and a transfer rate setting of 2, corresponding to ϳ31 l/s. Ivermectin concentrations for the FlexStation assay were prepared as 1:10 serial dilutions of 10, 5, and 2 M using a 1ϫ commercial stock of Hanks' balanced salt solution (without phenol red; catalog no. 14025, Invitrogen) with 20 mM HEPES, pH 7.4, containing 0.1% DMSO.
Electrophysiology-We used an Axopatch 200A amplifier equipped with a CV201 headstage and Digidata 1200 series interface operated by Clampex version 9.2 software (Axon Instruments) to obtain electrophysiological recordings from neurons in whole-cell current clamp mode. Stepwise square current pulses (Ϫ100 to 250 pA, 25-pA increments) were used to generate action potentials. Data were sampled at 50 kHz and low pass-filtered at 5 kHz. Neurons were superfused (or incubated) with artificial cerebrospinal fluid composed of 110 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl 2 , 0.8 mM MgCl 2 , 10 mM D-glucose, 10 mM HEPES, pH 7.4, 230 mosM. Patch pipettes were pulled from borosilicate glass and had resistances of 7-12 megaohms when filled with the internal solution (100 mM potassium gluconate, 0.1 mM CaCl 2 , 5 mM MgCl 2 , 1.1 mM EGTA, 10 mM HEPES, 3 mM Mg-ATP, 0.3 mM GTP, 3 mM phosphocreatine, pH 7.2, 215 mosM. IVM was dissolved in artificial cerebrospinal fluid containing 0.1% DMSO and preincubated with cultures at 37°C/5% CO 2 for 15 min. All recordings were obtained at ambient temperature. Data were analyzed using Clampfit version 9.2 software. Resting membrane potential was measured in the absence of injected current and corrected for the liquid junction potential at the electrode tip. To minimize the confounding effects of seal rupture during the transition to whole-cell configuration, cells with resting potentials of ϾϪ45 mV or seal resistances of Ͻ100 megaohms were omitted from analysis. The steady-state voltage response was plotted against the amount of current injected to measure a voltage-current (V-I) relationship. Total input resistance of the cell was determined from the slope of the linear portion of this relation, according to Ohm's law, V ϭ IR. Input conductance was calculated as the inverse of resistance (G ϭ 1/r ϭ I/V). Evoked spikes were counted and plotted against the amount of injected current.
Immunofluorescent Labeling-Live neurons were immunolabeled without fixation using a published protocol (11). A V5 epitope tag including a linker sequence followed by a His 6 tag already encoded in the pcDNA3.1 vector was added to the C terminus of GluCl ␣ and ␤ subunits (-CASKGNSADIQHSGGRSSL-EGPRFEGKPIPNPLLGLDSTRTGHHHHHH) by mutation of the stop codon (see above). The His 6 tag was not utilized. Surface receptors were labeled with primary mouse monoclonal anti-V5 antibody (1:200; catalog no. R960-25, Invitrogen) followed by a conjugated secondary Alexa Fluor 555 donkey antimouse antibody (1:400; catalog no. A-31570, Invitrogen). Antibodies were diluted into warm artificial cerebrospinal fluid and applied sequentially, incubating each for 30 min at 37°C/5% CO 2 with appropriate rinse steps. Live cultures were imaged immediately after immunostaining.
Imaging-Living HEK293 cells and cultured neurons were imaged at 37°C in a stage-mounted culture dish incubator (Warner Instruments). Transiently transfected HEK293 cells were imaged by TIRF microscopy, which enabled visualization of fluorescent receptors expressed in the plasma membrane and nearby intracellular vesicles (within 200 nm of the cellcoverslip interface). Prior to imaging, cell culture medium was replaced with phenol red-free CO 2 -independent Leibovitz L-15 medium (catalog no. 21083-027, Invitrogen). TIRF images were obtained using an inverted microscope (Olympus IX81) with a 100ϫ/1.45 numerical aperture Plan Apochromat oil objective. A stepping motor (T-cube stepper motor, Thorlabs) was used to control the position of the fiber optic and TIRF evanescent field illumination. A 488-nm laser was used to excite monomeric YFP fluorescence. Images were acquired using MetaMorph Premier software (Molecular Devices) at 16-bit resolution over 512 ϫ 512 pixels and captured using a backilluminated EMCCD camera (iXON DU-897) supported by ANDOR iQ2 software (Andor Technology).
Transiently transfected hippocampal neuron cultures were imaged using a laser-scanning confocal microscope (Nikon Eclipse C1si) equipped with a 63ϫ/1.4 numerical aperture VC Plan Apochromat oil objective. Monomeric YFP fluorescence was acquired with 514-nm laser excitation. Alexa Fluor 555 and pDsRED2 (catalog no. 632409, Clontech) fluorescence was acquired with 561-nm laser excitation. Images were collected as z-stacks at a step size of 1.0 m with 16-bit resolution over 512 ϫ 512 pixels and a dwell time of 6.72 s.
For confocal image analysis, the two different fluorescent signal intensities were correlated on a pixel-by-pixel basis using the Colocalization Colormap ImageJ plug-in (Adam Gorlewicz; available at the ImageJ web site). This software is based on a previously published algorithm (12). The correlation of a pair of pixels was calculated by the normalized mean deviation product (nMDP) as follows, where the intensity (I) for a given pixel in one image (a or b) is normalized to the average intensity and maximum intensity for that image. The nMDP values for each pixel range from Ϫ1 to 1 and are visualized on a color scale (see Figs Statistics-Pooled data are shown as means Ϯ S.E. Box plots represent the mean, median, 25th, and 75th percentiles. Statistical significance (p Ͻ 0.05) was determined by one-way analysis of variance on ranks using multiple pairwise comparison.

Gain-of-function Gating Mutation Increases Receptor
Sensitivity to IVM-GluCl is a member of the Cys-loop superfamily of ligand-gated ion channels in which five individual subunits form homomeric or heteromeric receptor subtypes with varying functional properties. A number of studies on both cationselective (nicotinic acetylcholine and serotonin 5-HT 3 ) and anion-selective (GABA and glycine) Cys-loop receptors have demonstrated that mutation of a highly conserved pore-lining residue, leucine 9Ј, to one of several other amino acids can dramatically increase agonist sensitivity through effects on channel gating (13)(14)(15)(16)(17)(18)(19)(20)(21). This gain-of-function effect is evident in the dose-response relation, in which channel activation occurs with lower concentrations of agonist.
To test GluCl receptors for increased IVM sensitivity, the highly conserved leucine 9Ј residue in the M2 domain of the ␣ subunit was mutated to one of seven other residues: L9ЈI, L9ЈF, L9ЈV, L9ЈA, L9ЈG, L9ЈS, or L9ЈT. Heteromeric GluCl ␣␤ wildtype (WT-XFP) and leucine 9Ј mutant channels (also XFPtagged) were expressed in HEK293 cells and screened for receptor activation using a fluorescence-based assay with a membrane potential-sensitive dye. All mutants, except for L9ЈF, displayed reduced signals for IVM activation compared with WT-XFP (Fig. 1A). Weak agonist-induced signals were the result of elevated base-line fluorescence (data not shown), suggesting an increased level of background conductance, which has been reported for some leucine 9Ј mutations in other Cysloop receptors (13,20,(22)(23)(24)(25). The L9ЈF mutant retained a maximal response comparable with that of WT-XFP. Fitted IVM activation parameters for the normalized mean signal show a biphasic concentration-dependent relation for the L9ЈF mutant with an unambiguous increase in IVM sensitivity throughout the unsaturated concentration range compared with WT-XFP ( Fig. 1B and Table 1). As a result, L9ЈF was designated as the best candidate mutation for increasing the sensitivity and preserving the amplitude of the ivermectin response.
The membrane potential-sensitive dye assay unexpectedly revealed that HEK293 cells expressing GluCl ␣ homomers were indeed responsive to IVM (Fig. 1C), despite the failure of previous electrophysiological studies to detect IVM responses from ␣ homomers expressed in mammalian cell lines (1,2). GluCl ␣ homomers containing the L9ЈF mutation were also responsive but did not show increased sensitivity to IVM ( Table  1), implying that the gain-of-function effect of L9ЈF on IVM sensitivity requires co-assembly with the ␤ subunit. GluCl ␤ homomers were not responsive to applications of IVM.
Mutation of a Putative ER Retention Motif Enhances the High Sensitivity IVM Response-In the cell, multimeric membrane proteins are synthesized, matured, and assembled in the endoplasmic reticulum (ER) (26,27). Proteins that are misfolded, unassembled, or improperly assembled are retained in the ER and ultimately targeted for ER-associated degradation (28,29). This mechanism ensures that only properly assembled receptors are transported to the cell surface.
The amino acid sequence of a plasma membrane protein often contains specific signaling motifs to facilitate retention or export of the protein from the ER. The best characterized ER exit signals include DXE (30,31), LXX(L/M)E (32), and (I/L)XM (33). These motifs exist on the cytosolic loops of a variety of membrane-associated proteins, where they are recognized by The IVM concentration-response relation for the L9ЈF mutant was best fit by the sum of two Hill equations. C, normalized IVM concentration-response relations for WT-XFP and L9ЈF GluCl ␣ homomers. Fitted parameters for IVM concentration-response relations are given in Table 1. Error bars, S.E. the primary cargo selection protein (Sec24) of the COPIIcoated vesicles for transport from ER to Golgi. Well characterized ER retention signals include the classical C-terminal motifs, KDEL and KKXX (34,35), and the cytosolic argininebased signal, RXR (36). The RXR motif occurs in potassium channels (36), G-protein-coupled receptors (37), voltage-dependent Ca 2ϩ channels (38), ionotropic glutamate receptors (39 -41), and ionotropic Cys-loop receptors (42)(43)(44). Proteins containing the RXR motif are inferred to be retained in the ER (or maintained by COPI retrieval) until the retention signal is masked by multimeric assembly with other receptor subunits (45)(46)(47). Fig. 2A notes putative ER signaling motifs in the amino acid sequences of GluCl ␣ and ␤, in the large M3-M4 intracellular loop of each subunit. The ␣ subunit contains a single putative ER export motif (LNLLE) immediately following the fluorescent fusion protein (XFP tags were inserted at restriction sites (3)). The ␤ subunit contains two putative ER export motifs (LEM and DAE) and two putative ER retention motifs (RSR and RRR). The presence of putative retention motifs in ␤ and their absence in ␣ is consistent with the functional expression observed for these subunits in mammalian systems. For example, in the non-mammalian Xenopus laevis oocyte expression system, both ␣ and ␤ subunits form functional homomeric channels, of which ␣ homomers are activated by IVM and ␤ homomers are activated by glutamate (note that ␣␤ heteromers can be activated by either glutamate or IVM) (48). In HEK293 cells, however, ␣ homomer activation can be observed, but no agonist-induced response has successfully been detected for ␤ homomers using either electrophysiology or fluorescencebased membrane potential assays (data not shown). This suggests that one or both of the retention motifs in the ␤ subunit may prevent/limit its export to the plasma membrane in the absence of ␣ subunits. Removal of ER retention signals could potentially allow for more uniform receptor expression and consistent neuronal silencing. To investigate this possibility, the putative ER retention motifs of the ␤ subunit were mutated to alanine residues: GluCl ␤-YFP R318A/S319A/R320A for an "RSR_AAA" mutant and GluCl ␤-YFP R329A/R330A/R331A for an "RRR_AAA" mutant.
The mutant ␤ subunits were co-expressed with ␣ to produce heteromeric receptors in HEK293 cells and tested for IVM activation using the fluorescent membrane potential assay. The RSR_AAA and RRR_AAA mutant receptors as well as a double mutant receptor, RSR_AAA/RRR_AAA, produced IVM concentration-response relations similar to that of the WT-XFP receptor ( Fig. 2B and Table 1). However, the RSR_AAA mutant displayed a distinct biphasic relation, suggesting the presence of an additional receptor population with high IVM sensitivity, presumably one with greater ␤ subunit incorporation. The addition of the L9ЈF gain-of-function mutation to the RSR_AAA mutant receptor further enhanced the high sensitivity component of the IVM concentration-response relation. A comparison of peak amplitudes in response to 10 nM IVM indicated that the L9ЈF ϩ RSR_AAA mutant receptor significantly increased receptor activation using lower concentrations of IVM (Fig. 2C). No IVM response was observed when mutant ␤ subunits were expressed the in absence of the ␣ subunit, implying that the removal of these putative retention motifs was not sufficient for surface expression of functional ␤ homomers.
Addition of the Glutamate-insensitive Mutation Eliminates High IVM Sensitivity-To use this L9ЈF ϩ RSR_AAA hyper-IVM-sensitive GluCl mutant receptor for pharmacogenetic silencing, sensitivity to the endogenous neurotransmitter glutamate had to be eliminated. Astonishingly, reintroduction of the glutamate binding site mutation Y182F in the ␤ subunit abolished the high sensitivity component of the biphasic IVM concentration-response relation (Fig. 3A). The addition of the Y182F mutation to the singly mutated L9ЈF and RSR_AAA mutant receptors showed a similar effect (Fig. 3, B and C). Despite the loss of high IVM sensitivity, the Y182F mutation did suppress the glutamate sensitivity of the receptors (data not shown).
A Mutation That Prevents XFP Dimerization Restores High IVM Sensitivity-Fluorescent proteins tend to dimerize at high concentrations (49), but dimerization can also occur when XFPs are spatially restricted, such as when they are fused to membrane proteins (50). We suspected that XFP dimerization might interfere with IVM sensitivity because heteromeric WT-XFP receptors with YFP and CFP tags on either or both sub- GluClv2.0, an Improved Neuronal Silencer JULY 19, 2013 • VOLUME 288 • NUMBER 29 units all displayed IVM concentration-response relations that were right-shifted relative to the untagged WT receptor (Fig.  4A). The full data set for XFP-tagged and untagged subunits revealed that rightward shifts occurred only when the fluorescent protein was present in the ␣ subunit (Fig. 4B). In further observations, the IVM concentration-response relation of the ␣-XFP homomer was also right-shifted relative to untagged ␣ homomers (data not shown), again suggesting a loss of IVM sensitivity mediated by the fluorescent protein insertion. Strictly monomeric XFPs (mXFPs) can be obtained by mutating the Ala-206 residue to Lys, a long, positively charged side chain that disrupts the hydrophobic dimer interface (51). Incorporating this mutation into the fluorescently tagged GluCl constructs prevented the rightward shift and even added a high sensitivity component to the IVM concentration-response relation of WT-mXFP (Fig. 4C). Including mXFP mutations in the heteromeric L9ЈF receptor produced a biphasic concentration-response relation more pronounced than any other previously obtained (Fig. 4D). More importantly, introducing mYFP into the initially favored L9ЈF ϩ RSR_AAA receptor restored the high IVM sensitivity component that was lost upon the addition of the glutamate-insensitive Y182F mutation. This GluCl ␣-mXFP L9ЈF ϩ ␤-mXFP Y182F RSR_AAA mutant receptor was more sensitive to IVM than GluClv1.0 by ϳ2 orders of magnitude ( Fig. 4E and Table 1) while preserving insensitivity to physiologically relevant concentrations of glutamate (Յ1 mM; Fig. 4, F-H) (52)(53)(54).
Biphasic Concentration-Response Relation Reflects Mixed Receptor Subunit Stoichiometries-In pentameric Cys-loop receptors, multicomponent concentration-response relations typically result when several subunit stoichiometries with differing agonist sensitivities are present in the receptor population (55)(56)(57). Variation in subunit stoichiometry may account for the biphasic IVM concentration-response relations observed for certain mutant GluCl subunit combinations. The ␣ subunit L9ЈF mutation adds a high sensitivity component to The ␣ subunit contains a single putative ER export motif (LXX(L/M)E). The ␤ subunit contains two possible ER export motifs ((I/L)XM and DXE) and two potential ER retention motifs (RXR). B, normalized IVM concentration-response relations for heteromeric receptors containing RSR_AAA, RRR_AAA, and the double RSR_AAA/ RRR_AAA mutations compared with WT-XFP. The RSR_AAA mutant produces a biphasic concentration-response relation. The addition of an L9ЈF mutation further enhances the high sensitivity component. C, the L9ЈF ϩ RSR_AAA mutant receptor significantly increases the peak amplitude of the 10 nM IVM response relative to WT-XFP. IVM activation was measured using a fluorescent membrane potential assay. Fitted parameters for IVM concentration-response relations are given in Table 1. Error bars, S.E.

GluClv2.0, an Improved Neuronal Silencer
the IVM concentration-response relation only when co-expressed with the ␤ subunit. Mutation of an ER retention sequence in the ␤ subunit, which presumably increases the probability of ␤ incorporation into the receptor, has a similar effect. Thus, the high IVM sensitivity component probably manifests from a shift in the ratio of ␣/␤ subunits in the assembled pentamer to include more ␤. To test this hypothesis, different ratios of ␣ and ␤ subunit DNA (1:1, 4:1, and 1:4) were transfected into HEK293 cells, and the IVM concentration-response relations were again measured using the fluorescent membrane potential assay. Consistent with our previous results, a 1:1 ratio produced a biphasic concentration-response relation. Biasing the ratio in favor of ␣ (4:1) produced a decrease in IVM sensitivity, whereas biasing the ratio in favor of ␤ (1:4), produced no further increase in IVM sensitivity (Fig. 5). Therefore, incorporation of the ␤ subunit confers increased receptor sensitivity to IVM, but transfection of equal amounts of ␣ and ␤ subunit DNA appears to be optimal.
The fluorescent protein tags permit direct visualization and localization of GluCl subunits expressed in a cell. Images of HEK293 cells obtained by TIRF microscopy show that GluCl ␣␤ heteromers and ␣ homomers are expressed at the plasma membrane, but ␤ homomers are not (Fig. 6, A and B). Specifically, the ␣-mYFP subunit shows plasma membrane fluorescence when transfected alone or cotransfected with untagged ␤. The ␤-mYFP subunit, on the other hand, shows plasma membrane fluorescence only when cotransfected with untagged ␣. Transfection of the ␤-mYFP subunit alone displays complete ER retention, as evidenced by the reticulated pattern of fluorescence and the lack of fluorescent filopodia at the periphery. GluCl ␤-mYFP subunits bearing the RSR_AAA, RRR_AAA, or the double RSR_AAA/RSR_AAA mutations display fluorescence patterns similar to the ␤-mYFP homomer (Fig. 6C). Thus, disruption of the putative ER retention motifs in the ␤ subunit is not sufficient to allow plasma membrane expression of ␤ homomers.
RSR Mutation Increases ␤ Subunit Expression at the Plasma Membrane and the ER-The results above suggest that the ER retention mutation RSR_AAA enhances incorporation of the ␤ subunit into heteropentameric GluCl receptors expressed on the cell surface. Although HEK293 cells were convenient for screening receptor function, primary neuronal cultures provide a more appropriate environment for investigating patterns of subunit expression. Preliminary confocal images of fluorescently tagged subunits transfected into embryonic rat hippocampal neurons displayed a deficient neuronal expression pattern for GluCl ␤ homomers with minimal extension into the processes and comparatively few fluorescent neurons per imaging dish. GluCl ␣ homomers and ␣␤ heteromers, on the other hand, exhibit extensive fluorescent projections with no gross discernible differences (data not shown). Distinguishing between receptors retained in the ER and those expressed at the plasma membrane required colocalization markers.
Receptors expressed at the plasma membrane were targeted by live cell immunofluorescent surface staining. A V5 epitope tag was added to the extracellular C terminus of both ␣ and ␤ and subunits, which were then assayed for channel function in HEK293 cells to verify that pentameric assembly was not disrupted (data not shown). Cultured rat hippocampal neurons were then transfected with fluorescently tagged ␣ and/or ␤ subunits with one of the subunits containing the V5 epitope (e.g. ␣-mYFPV5 ϩ ␤-mYFP). V5-tagged subunits expressed at the cell surface were detected by immunostaining with an anti-V5 primary antibody followed by a fluorescent Alexa Fluor 555conjugated secondary antibody. Immediately after staining, the entire neuron was scanned for red and yellow fluorescence using confocal microscopy. Yellow fluorescence contributed by both subunits (mYFP) reflected total protein expression and was used to locate transfected neurons within the dish. Red fluorescence from live cell immunostaining (Alexa Fluor 555) emitted only from labeled subunits expressed at the cell surface. The extent of red and yellow fluorescent colocalization, therefore, served as a relative measure of surface expression, minimizing the effects of nonspecific antibody staining in our analysis. The fluorescence intensity correlation for each pixel was calculated by the nMDP and visualized on a color scale (Fig. 7A). Colocalization is evidenced by all positive nMDP values (0 Ͻ x Յ 1; shown in hot colors) and occurs only for receptors  Fig. 2). B and C, the Y182F mutation similarly affected the concentration-response relations of the individual L9ЈF and RSR_AAA mutant receptors. IVM activation was measured using a fluorescent membrane potential assay. Fitted parameters for IVM concentrationresponse relations are given in Table 1. Error bars, S.E. JULY 19, 2013 • VOLUME 288 • NUMBER 29 expressed at the surface. Hence, greater nMDP values indicate a greater amount of GluCl expression at the plasma membrane.

GluClv2.0, an Improved Neuronal Silencer
The overall average of positive nMDP values provides a relative representation of total surface expression levels of receptor (note that the sum of all positive nMDP values is distorted by the size and number of cells imaged). Heteromeric WT receptors show the same level of total surface expression regardless of whether the V5 tag was on the ␣ or ␤ subunit (Fig.   7B). Homomeric receptors (transfected 1:1 with V5-tagged and untagged subunits for comparison with heteromers) corroborated the previous HEK293 cell observations that ␣ homomers are expressed at the plasma membrane of neurons at levels comparable with ␣␤ heteromers but that ␤ homomers are not. Heteromeric receptors bearing the RSR_AAA retention mutation in the ␤ subunit showed the same total surface expression levels as WT heteromers and ␣ homomers, whereas those with  Fig. 3A). This newly engineered receptor (␣-mXFP L9ЈF ϩ ␤-mXFP Y182F RSR_AAA) is ϳ100-fold more sensitive to IVM than GluClv1.0. Fitted parameters for IVM concentration-response relations are given in Table 1. Examining the number of maximally correlated pixels (nMDP ϭ 1) within an image more directly represents surface expression levels of ␣ versus ␤ subunits. On average, the heteromeric WT, RRR_AAA, and RSR_AAA/RRR_AAA mutant receptors showed a greater number of maximally correlated pixels when the V5 tag was on the ␣ subunit than on the ␤ subunit, suggesting a greater number of ␣ subunits present in the assembled pentamer (Fig. 7C). Heteromeric RSR_AAA mutant receptors, on the other hand, showed that maximal colocalization occurred more often for V5-tagged ␤ than V5-tagged ␣, consistent with a shift in subunit stoichiometry. Altogether, the immunofluorescent results suggest that mutation of the putative RSR retention motif in the ␤ subunit does not increase the total number of receptors trafficked to the plasma membrane but does increase the number of ␤ subunits incorporated into assembled surface receptors.
To determine the relative amounts of WT and mutated ␤ subunits remaining in the ER, colocalization analysis was per-  GluCl ␣ homomers are also expressed at the plasma membrane. B, a reticulate pattern of fluorescence and the lack of hairlike filopodia at the periphery suggest that ␤ homomers are exclusively retained in the ER. C, mutation of the putative ER retention motifs in the ␤ subunit does not alter this pattern of fluorescence and therefore does not enable plasma membrane expression of ␤ homomers.

GluClv2.0, an Improved Neuronal Silencer
formed on neurons cotransfected with non-fluorescent ␣ subunits and the various mYFP-tagged ␤ subunits together with the fluorescent ER marker dsRED (Fig. 8A). Once again, the RSR_AAA mutant differed significantly from WT and the RRR_AAA and RSR_AAA/RRR_AAA mutants, but rather than showing reduced ER localization, as would be presumed by relief of retention, the RSR_AAA mutant showed greater expression levels in the ER compared with the other ␤ subunits (Fig. 8B). Because newly synthesized, improperly folded, or unassembled Cys-loop subunits remaining in the ER are known to be rapidly degraded (58 -62), mutation of an ER retention motif could conceivably influence subunit degradation. These results may suggest that the RSR_AAA mutant ␤ subunit is not being degraded as quickly, leaving more available for heteromeric assembly with ␣ subunits.
An Improved Neuronal Silencing Tool-Rational mutation has led to a newly engineered GluCl receptor with increased sensitivity to IVM and improved expression of the ␤ subunit.
To determine whether this GluCl ␣-mXFP L9ЈF ϩ ␤-mXFP Y182F RSR_AAA receptor is indeed an improved neuronal silencing tool over GluClv1.0, IVM-induced conductance and spike inhibition were measured from dissociated hippocampal neurons expressing these receptors (identified by fluorescence microscopy). Neurons were incubated with 0, 1, or 20 nM IVM for 15 min, rinsed with IVM-free saline, and then recorded in whole-cell current clamp mode for V-I relations using square current injection steps of Ϫ100 to 250 pA in 25-pA increments (Fig. 9A). The newly engineered receptor showed a significant increase in conductance (measured from the linear portion of the V-I relation; Fig. 9C) and a reduction in the mean evoked spike count (Fig. 9B) for both 1 and 20 nM IVM compared with GluClv1.0. Not only was spike suppression enhanced, the variability in spike suppression was also significantly reduced (Fig.  9D). Neither the presence of IVM nor receptor expression itself affected the resting membrane potential, confirming the integrity of induced spike inhibition (Fig. 9E). Thus, neuronal silencing achieved through IVM activation of the newly modified GluCl is indeed more sensitive and less variable than for GluClv1.0.

DISCUSSION
Previous in vitro and in vivo studies report that co-expression of both ␣ and ␤ subunits is required for GluCl-mediated neuronal silencing (1,2). Variability in GluCl subunit expression (particularly that of the ␤ subunit) appeared to be responsible for whether or not an individual neuron was inhibited by IVM. Our results show that functional, IVM-sensitive ␣ homomers are actually expressed at the plasma membrane, but inclusion of the ␤ subunit increases GluCl sensitivity to IVM. Therefore, variation in the proportion of heteromeric and homomeric receptors within individual neurons may account for the previously reported variation in IVM-induced spike suppression. We improved on the previous GluClv1.0 construct, introducing three new amino acid modifications. These modifications still do not provide a homogeneous receptor population, but, nevertheless, they generate a receptor population that is significantly more sensitive to IVM and that exhibits reduced variability in spike suppression. Therefore, this revised GluCl receptor, dubbed "GluClv2.0," should allow lower doses of IVM to be administered for in vivo silencing, alleviating concerns of side effects and suboptimal inhibition. itive nMDP values for a given receptor is representative of its total surface expression. GluCl ␣ homomers are expressed at the plasma membrane, and ␤ homomers are not. For heteromers, the same level surface expression was obtained regardless of whether the V5 tag was on the ␣ or ␤ subunit. Receptors bearing putative ER retention motif mutations do not increase total receptor surface expression compared with WT. C, the average number of pixels with nMDP values of unity indicates that more maximally correlated pixels occur with V5-tagged ␣ than with V5-tagged ␤ when WT, RRR_AAA, and RSR_AAA/RRR_AAA receptors are expressed as heteromers. In contrast, the RSR_AAA mutant receptor shows more maximal colocalization with V5-tagged ␤ than with V5-tagged ␣, suggesting a shift in surface receptor subunit stoichiometry. Error bars, S.E.

GluClv2.0, an Improved Neuronal Silencer
Mechanisms of the Optimized Receptor-Concentration-response relations of GluCl ␣ homomers and ␣␤ heteromers bearing an L9ЈF mutation in the ␣ subunit show that co-assembly with the ␤ subunit significantly increases IVM sensitivity. The effects of biasing the stoichiometric ratio of transfected ␣ and ␤ subunits not only support this conclusion, they suggest that the L9ЈF mutation prefers or possibly promotes ␤ subunit incorporation because biasing for ␣ subunit expression with a 4:1 ␣/␤ subunit ratio still yields a biphasic IVM concentrationresponse relation for this mutant. Even the low sensitivity component of this relation is more sensitive (EC 50 ϭ 150 nM) than that of L9ЈF homomers (EC 50 ϭ 450 nM). Cys-loop receptor assembly (as described by one theory) involves an initial subunit dimerization step, followed by the dimerization of dimers and subsequent inclusion of a fifth subunit to form the final pentamer (63). The L9ЈF mutation may either promote ␣-␤ dimerization or hinder ␣-␣ dimerization, favoring the formation of a predominantly heteromeric ␣␤ receptor population that includes more ␤ subunits per assembled receptor.
According to conjecture, the arginine-based ER retention motif of the ␤ subunit requires masking by co-assembly with the ␣ subunit to exit the ER. This mechanism may explain the absence of ␤ homomers at the plasma membrane. Our data show that removal of the retention motif (RSR_AAA) increases the amount of ␤ exported to the plasma membrane in the presence of ␣. However, the RSR_AAA mutation also elevates ␤ subunit levels in the ER, indicating that it does not simply enhance ␤ subunit surface expression by reducing ER retention. Moreover, simply mutating the ER retention motif is not sufficient to allow surface expression of ␤ homomers, suggesting that an additional unknown quality control mechanism is involved in ␤ subunit retention. One possible explanation is that the RSR_AAA mutation impedes ER-associated degradation of the ␤ subunit. It could do so either directly, prolonging its availability for ␣-␤ dimerization, or indirectly, by facilitating stable ␣-␤ dimer formation. A similar stabilizing effect within the ER appears to underlie nicotine-induced up-regulation of ␣4␤2 nicotinic acetylcholine receptors to the plasma membrane (44). We do not know the extent to which the availability of ␤ subunits in the ER impacts the propensity of ␣ homomer formation. Limiting ␤ subunit degradation may keep ␣ homomer expression to a minimum.
The reason why reintroduction of the glutamate-insensitive mutation eliminated increased IVM sensitivity is less clear. The Y182F binding site mutation may reduce protein expression/ stability of the ␤ subunit or somehow inhibit ␣-␤ dimer formation. We cannot exclude the possibility that this mutation affects IVM activation of GluCl by interfering with the coupling between agonist binding and gating (64). However, restoration of the high sensitivity component of the IVM concentrationresponse relation by introducing the mYFP mutation suggests that it is more likely to be a consequence of altered subunit stoichiometry. The glutamate binding site is positioned at subunit interfaces where structural modifications could potentially disrupt heteromeric assembly. In this instance, the Y182F mutation would be preventing efficient pentameric incorporation of the ␤ subunit, counteracting the benefits of the L9ЈF and RSR_AAA mutations.
Application of GluClv2.0-Since the initial proof-of-concept study, GluClv1.0 has been used in conjunction with channelrhodopsin-2-mediated activation to define an inhibitory microcircuit within the amygdala involved in mouse fear conditioning (65) and to identify a hypothalamic locus responsible for male mouse aggression and its close neuroanatomical relationship to mating circuits (66). In the former study, an intersectional approach was used to restrict GluCl expression to PKC-␦-containing GABAergic neurons in the central amygdala. Selective localization was achieved by targeted transgenic expression of GluCl ␣-CFP in all PKC-␦ ϩ neurons, followed by stereotaxic injection of an adeno-associated vector encoding GluCl ␤-YFP Y182F. Although GluCl/ IVM-mediated silencing of PKC-␦ ϩ neurons yielded a significant enhancement of conditional freezing, the behavioral result was again confounded by a bimodal phenotype. Histological data revealed considerable variation in expression of the virally injected ␤ subunit, reminiscent of the striatal proof-of-concept study. Interestingly, control animals transgenically expressing ␣ alone were not affected by treatment with 10 mg/kg IVM, despite our present results indicating that ␣ homomers are indeed expressed at the plasma membrane. Most likely, ␣ homomer activation elicits insufficient Cl Ϫ current to achieve neuronal silencing.
GluClv2.0 maintains the requirement of both ␣ and ␤ subunits for silencing. The pharmacokinetic properties of the silencing tool have not been altered. Another variant of IVM-induced silencing has been achieved by increasing the sensitivity of a mammalian homomeric glycine receptor (67), although this tool remains to be verified in vivo. In addition, an alternative Cys-loop orthogonal pharmacological silencing tool (based on chimeric nicotinic acetylcholine-glycine receptors of mammalian origins) capable of activating and inactivating on shorter time scales has recently been constructed (68). However, the cognate synthetic nicotinic agonist has demonstrated weak to moderate binding of other endogenous nicotinic acetylcho-line receptors, and the absence of potential co-assembly of chimeric subunits with endogenous nicotinic acetylcholine receptors has not yet been verified. Even faster time-resolved neuronal silencing can be achieved using optogenetic techniques (69,70), but this method is invasive and is not suitable for silencing large anatomical regions or diffuse signaling networks. The duration of light-induced manipulation is further limited by heat generation, which may alter neuronal activity or damage cell health (71). Thus, orthogonal phar- FIGURE 9. An optimized neuronal silencing tool. Action potential firing in response to depolarizing current pulses (Ϫ100 to 250 pA, 25-pA increments) was obtained from cultured hippocampal neurons in whole-cell current clamp mode following a 15-min preincubation with 0, 1, or 20 nM IVM. A, linear range of the V-I relations from neuronal cultures transfected with the revised construct (␣-mYFP L9ЈF ϩ ␤-mYFP Y182F RSR_AAA) was compared with those of GluClv1.0 (␣-YFP ϩ ␤-YFP Y182F) and non-transfected neurons. The revised receptor exhibits lower input resistance (as indicated by the slope) at 1 and 20 nM IVM. B, the revised receptor also reduced mean evoked spike counts at 1 and 20 nM IVM. Insets, voltage response to current injections for individual neurons expressing the original (top) and revised (bottom) receptors. C, the revised receptor induced a significant increase in conductance for both 1 and 20 nM IVM compared with GluClv1.0 (*, p Ͻ 0.01; **, p Ͻ 0.001). D, variability in evoked spike suppression, noted by the S.D. (error bars) of the spike count along the abscissa, is also significantly reduced with the revised receptor. E, neither construct expression nor the presence of IVM alone significantly altered the resting membrane potential (V mrest ). macogenetic silencing using IVM activation of Cl Ϫ -permeable pentameric Cys-loop receptors remains a viable option for long term neuronal inhibition (such as that required in learning paradigms) and for assessing modulatory as opposed to regulatory roles in neuronal circuitry.