Extrasynaptic δ‐GABAA receptors are high‐affinity muscimol receptors

Abstract Muscimol, the major psychoactive ingredient in the mushroom Amanita muscaria, has been regarded as a universal non‐selective GABA‐site agonist. Deletion of the GABAA receptor (GABAAR) δ subunit in mice (δKO) leads to a drastic reduction in high‐affinity muscimol binding in brain sections and to a lower behavioral sensitivity to muscimol than their wild type counterparts. Here, we use forebrain and cerebellar brain homogenates from WT and δKO mice to show that deletion of the δ subunit leads to a > 50% loss of high‐affinity 5 nM [3H]muscimol‐binding sites despite the relatively low abundance of δ‐containing GABAARs (δ‐GABAAR) in the brain. By subtracting residual high‐affinity binding in δKO mice and measuring the slow association and dissociation rates we show that native δ‐GABAARs in WT mice exhibit high‐affinity [3H]muscimol‐binding sites (KD ~1.6 nM on α4βδ receptors in the forebrain and ~1 nM on α6βδ receptors in the cerebellum at 22°C). Co‐expression of the δ subunit with α6 and β2 or β3 in recombinant (HEK 293) expression leads to the appearance of a slowly dissociating [3H]muscimol component. In addition, we compared muscimol currents in recombinant α4β3δ and α4β3 receptors and show that δ subunit co‐expression leads to highly muscimol‐sensitive currents with an estimated EC 50 of around 1–2 nM and slow deactivation kinetics. These data indicate that δ subunit incorporation leads to a dramatic increase in GABAAR muscimol sensitivity. We conclude that biochemical and behavioral low‐dose muscimol selectivity for δ‐subunit‐containing receptors is a result of low nanomolar‐binding affinity on δ‐GABAARs.

The functional and pharmacological characteristics of extrasynaptic d-GABA A Rs are quite different from classical c2-GABA A Rs. d-GABA A Rs have much higher affinity for GABA, are insensitive to classical benzodiazepines, show high sensitivity to neurosteroids and Zn 2+ (Semyanov et al. 2004;Mortensen and Smart 2006;St orustovu and Ebert 2006) and d-GABA A R-mediated tonic currents show high sensitivity to ethanol (Hanchar et al. 2005;Fleming et al. 2007). Recombinantly expressed d-GABA A Rs show increased maximal currents with the 5,6,7tetrahydroisoxazolo[5,pyridin-3-ol (THIP, also known as gaboxadol) when compared to GABA. This is likely because of GABA being a partial agonist on these receptors (Bianchi and Macdonald 2003) and d subunit incorporation dramatically increases their THIP sensitivity (Meera et al. 2011). This is consistent with the finding that GABA A R d subunit knock-out (dKO) mice lose low dose THIP effects on tonic currents in neurons in brain slices and behavioral sensitivity to low doses of THIP (Boehm et al. 2006;Chandra et al. 2006). Similar to THIP, low-dose muscimol behavioral effects are also dependent on the presence of a6, a4 and d subunits, with both a4KO and dKO mice much less sensitive to behavioral muscimol effects, whereas ectopic over-expression of a6 in mice resulted in increased behavioral muscimol sensitivity (Chandra et al. 2010). This clearly indicates that a4/6bd-GABA A Rs are critical for low-dose behavioral effects of the GABA analogs THIP and muscimol.
Muscimol, the principal psychoactive constituent of Amanita muscaria and related species of mushroom (Krogsgaard-Larsen et al. 1981), is produced from ibotenic acid by decarboxylation (Fig. 1a) and has been considered as a general GABA A agonist that activates all GABA A R subtypes (Krogsgaard-Larsen et al. 1979;DeFeudis 1980), including specialized rho-GABA receptors (Ogurusu et al. 1999). However, muscimol shows GABA A R selectivity with exceptionally high affinity to d-GABA A Rs (Quirk et al. 1995). In addition, when measured with autoradiography, dKO mice lose high-affinity (6 nM) [ 3 H]muscimol binding in the forebrain sections, with drastically reduced binding in the cerebellum (Mihalek et al. 1999, Fig. 1b), indicating that under these experimental conditions muscimol shows much higher affinity for d-GABA A Rs when compared to abundant c-GABA A Rs.
In this study, we investigated high-affinity (5 nM) [ 3 H]muscimol binding in wild-type (WT) and dKO mouse brains and to several abc and abd-type recombinant GABA A Rs by measuring binding and unbinding kinetics. Subtraction of residual high-affinity (5 nM) [ 3 H]muscimol binding that is seen on abundant GABA A R subtypes in dKO mice from binding in WT membranes allowed us to isolate a native d-GABA A R component. This isolated component showed very slow muscimol dissociation rate with an apparent K D (calculated from k on and k off rates) for muscimol of 1.6 nM for a4bd receptors in the forebrain and around 1 nM for a6bd receptors in the cerebellum. Recombinant a4b3d receptors expressed in oocytes revealed a biphasic response to muscimol with the high-muscimol affinity (slowly deactivating/dissociating) component showing an approximate EC 50 of around 1-2 nM.
We conclude that muscimol is a high-affinity ligand for both native and recombinant d-GABA A Rs, providing the molecular basis for the biochemical and behavioral selectivity of muscimol actions on a4/6bd GABA A Rs (Chandra et al. 2010).

Animals
Wild-type (C57BL/6J, WT; RRID: IMSR JAX:000664), and GABA A R d subunit knockout (C57BL/6J, dKO; RRID: MGI:3 639693) mice (age 3-12 months, both sexes, University of California at Los Angeles) were used for the studies. However, dKO mice were originally generated by the Homanics lab (Mihalek et al. 1999), using ES cell injection into C57BL/6J blastocysts and backcrossed for at least 10 generations with C57BL/6J mice (Jackson Laboratories, stock No. 000664). The mice weighed 19-32 g and they were housed in 12 : 12 h light:dark cycle in static plastic cages in groups of 2-4 mice having ad libitum access to Rodent Lab Chow #5001 food and filtered tap water. The animals were killed by decapitation, their brains were removed, the cerebellum was separated with a scalpel from the rest of the brain (i.e., forebrain and midbrain, but loosely referred to here as forebrain), frozen on dry ice and stored at À70°C.
All procedures were in accordance with protocols approved by the University of California at Los Angeles (UCLA) Chancellor's Animal Research Committee (Animal Welfare Approval number: A3196-01). Preparation of brain membranes WT and dKO forebrain and cerebellar membranes were prepared using a modification of the method of Squires and Saederup (2000) essentially as described by Uusi-Oukari et al. (2014). Rat forebrain along with midbrain region was homogenized into 10 mM Tris-HCl, pH 8.0 buffer containing 2 mM EDTA, using an Ultra-Turrax T25 (Janke & Kunkel IKA labortechnik) for 20 s at 9500 rpm. The homogenates were centrifuged at 20 000 g for 10 min at +4°C and the resulting pellets were washed three times by resuspension and re-centrifugation with 10 mM Tris-HCl, pH 8.5 buffer containing 0.2 M NaCl, and 5 mM EDTA. The resulting pellets were then suspended in ice-cold H 2 O and centrifuged. The pellets were again washed three times with Tris, pH 8.5/NaCl/EDTA as described above. The resulting pellets were finally suspended in assay buffer consisting of 50 mM Tris-base, pH adjusted to 7.4 with citric acid, and frozen at À70°C. Before a binding experiment, the suspension was thawed, centrifuged, and suspended in assay buffer.

Recombinant GABA A receptor expression in HEK293 cells and Xenopus oocytes
Human embryonic kidney (HEK) 293 cells (ATCC Cat# CRL-1573; RRID: CVCL_0045; a commonly misidentified cell line by ICLAC; last authenticated by STR DNA profiling in December 2018) were transfected with rat cDNAs (a1, L08490; a6, L08495; b2, X15467; b3, X15468; c2S, L08497; d, L08496) in pRK5 plasmids under the control of the cytomegalovirus promoter (Uusi-Oukari et al. 2000), using the calcium phosphate precipitation method essentially as described (L€ uddens and Korpi 1997). The plasmids were used in 1 : 1 and 1 : 1 : 1 ratios for transfections containing 2 [(a1 or a6) + (b2 or b3)] or 3 [(a1 or a6) + (b2 or b3) + (c2S or d)] different subunits, respectively (5 lg of each plasmid DNA for a 10 cm plate). Mock transfection was done using 5 lg pRK5 plasmid backbone. The cells were harvested 48 h after transfection. Culture medium was removed and the cells were detached from the plates by pipetting in ice-cold assay buffer containing 2 mM EDTA. The cells were homogenized (Ultra Turrax, 20 s at 9500 rpm), the homogenates centrifuged at 20 000 g for 10 min at +4°C, and washed once with the same buffer. The homogenates were finally suspended in assay buffer (1 mL/plate) and either used directly to binding assays or stored frozen at À70°C.
Human a4, a6, b3, and d cDNA clones for oocyte expression were made by PCR amplification of the coding region (NcoI site introduced with the amplifying 5 0 oligonucleotide at the ATG initiation codon) and a HindIII (or SpeI) containing oligo) and cloning it into a NcoI-HindIII (or SpeI) cut vector backbone derived from pEGFP-N1 (Addgene 6085-1). The entire transcribed region was confirmed by sequencing to ensure that protein sequences conform to consensus sequences found in the RefSeq database (https://www.ncbi.nlm.gov/RefSeq). Plasmids were linearized with NotI (New England Biolabs) and cRNA was transcribed, using T7 RNA polymerase (Ambion, mMESSAGE mMACHINE T7 Fig. 1 (a) Structures of the muscimol precursor ibotenic acid, GABA and the GABA A R agonists muscimol and THIP. The backbone of GABA is shown in bold to illustrate that muscimol and THIP are conformationally restricted GABA analogs. (b) [ 3 H]Muscimol (6 nM) autoradiography in brain sections comparing wild-type (WT) with a6 knockout (a6KO) and delta knockout (dKO) mouse lines. This shows that high-affinity muscimol binding in the forebrain is d subunit dependent, wheras in the cerebellum it is a6 subunit dependent. Transcription Kit, Ambion Austin TX, USA, Cat. No. 1344). Xenopus laevis (Nasco, product number LM00531) oocytes were prepared from oocyte lobes shared by Dr. Olcese (UCLA, Anesthesiology). The oocytes were injected with 2 ng of each a6 and b3 subunit cRNA alone or together with 10 ng d cRNA. Currents were measured 5-10 days after injection by two electrode voltage clamp using an Axoclamp 2B amplifier and pCLAMP software. Drug solutions were applied in ND96 (in mM 96 NaCl, 2 KCl, 1.8 CaCl 2 , 1 MgCl 2 , 5 HEPES pH 7.4) by gravity perfusion with bath exchange time of about 2 s. Muscimol was prepared as an aqueous 100 mM stock solution.

Measurement of [ 3 H]muscimol binding kinetics
The binding of [ 3 H]muscimol (5 nM) was measured in assay buffer at 22°C in a total volume of 300 lL. Triplicate technical replicates of mouse forebrain (190-215 lg protein), cerebellar (180-210 lg protein) or HEK cell (92-132 lg protein) membranes for each time point were incubated with shaking for various times (15 s-15 min) to measure association of the binding. Non-specific binding was determined in the presence of 100 lM GABA. The incubation was terminated by filtration of the samples with a Brandel Cell Harvester (model M-24, Gaithersburg, MD, USA) onto Whatman GF/B filters (Whatman International Ltd., Maidstone, UK). The samples were rinsed twice with 4-5 mL of ice-cold assay buffer. Filtration and rinsing steps took a total time of~15 s. Air-dried filters were immersed in 3 mL of Optiphase HiSafe 3 scintillation fluid (Wallac, Turku, Finland) and radioactivity determined in a Wallac model 1410 liquid scintillation counter (Wallac, Turku, Finland). The maximal binding disintegrations per minute (DPM) values (at 15 min in association) for recombinant studies with 5 nM [ 3 H]muscimol were between 700 and 2500 DPMs of specific binding (background subtracted). In native membranes, the maximal DPM values were between 2500 and 3000 for WTs and 1300-1500 for dKOs. Mock transfection with pRK5 plasmid did not produce any specific binding over the background.
To measure dissociation of [ 3 H]muscimol binding, triplicate technical replicates of each sample of mouse brain or HEK cell membranes for each time point were first pre-incubated at 22°C in a total volume of 300 lL for 15 min with 5 nM [ 3 H]muscimol in the absence and presence of 100 lM GABA. The dissociation was then started by adding 100 lL of 400 lM or 100 lM (non-specific binding) cold GABA to the incubation mixtures to reach a final 100 lM GABA concentration in all tubes. The tubes were mixed and incubations at 22°C were terminated at various time points (30 s -30 min) as described above. Dissociation of [ 3 H]muscimol from recombinant receptors in HEK cell membranes was also measured at 0-4°C (on ice) to evaluate how fast [ 3 H]muscimol dissociates from receptors while washing the filter with ice-cold assay buffer during filtration.
Saturation analysis of [ 3 H]muscimol to WT and dKO mouse forebrain and cerebellar membranes was performed essentially as described by Uusi-Oukari and Korpi (1989). Triplicate samples of the membranes were incubated in assay buffer with concentration series of hot [ 3 H]muscimol (0.1-30 nM) at 0 to 4°C for 30 min in the absence and presence of 100 lM GABA determining the non-specific binding. The incubations were terminated as described above.
The hypothetical values for binding of [ 3 H]muscimol to d-GABA A Rs in WT animals, 'native d-GABA A Rs', were calculated by subtracting the specific dKO binding values (binding to c2-GABA A Rs) from the corresponding WT values at each time point: native d-GABA A Rs = WT-dKO. Because of the lack of low-affinity binding and the relatively small number of time points in our assays, the binding curves fitted better in 'one binding site' model. However, varying fast and slow dissociation components are obvious in the graphs (see Figs 3 and 4).

Protein measurement
In all ligand-binding studies, protein concentrations of membranes were determined with the Bio-Rad Coomassie blue dye-based protein assay kit (Hercules, CA, USA) according to manufacturer's instructions.

Data analysis
Association and dissociation curves for estimation of association and dissociation rate constants, and saturation binding for estimation of B max and K D values were analyzed with Graph Pad Prism 7 software (Graph Pad, San Diego, CA, USA). Statistical significances between the groups were analyzed using unpaired t-test and one-way ANOVA followed by Tukey's post hoc test Graph Pad (Graph Pad Prism 7). p-values of < 0.05 were considered significant. In this study, no sample calculation, assessment of data outliers and data normality were performed, and experiments were done unblinded.

Results
The majority of high-affinity (5 nM) [ 3 H]muscimol binding is because of binding to low abundance d-GABA A R To evaluate the contribution of d-GABA A Rs to high-affinity muscimol binding, we measured the time course of 5 nM [ 3 H]muscimol binding to forebrain and cerebellar membranes from both wild-type and dKO mice. Deletion of the d subunit led to > 50% reduction of 5 nM [ 3 H]muscimol binding at 22°C to both forebrain and cerebellar membranes when compared to WT mice (Fig. 2). This finding is remarkable, considering that the proportion of d-GABA A Rs in the mammalian fore/midbrain is only up to 10%, depending on the exact brain region (Whiting 2003;H€ ortnagl et al. 2013). In the cerebellum, the fraction of d-GABARs is close to 30% (Tretter et al. 2001;P€ oltl et al. 2003), but this is accompanied by a relatively high muscimol affinity of cerebellar a6bc2 receptors (see Fig. 1b, Mihalek et al. 1999;M€ akel€ a et al. 1997). The increased muscimol binding by these a6bc2 receptors likely explains why the percent reduction in high-affinity muscimol binding in dKO cerebellum is about the same as in the forebrain despite the much higher abundance of d-GABA A Rs in the cerebellum. Total 5 nM [ 3 H] muscimol binding (fmol/mg membrane protein) was around four times higher in the cerebellum when compared to forebrain both in WT as well as in the dKO mice (Fig. 2), which is consistent with a much higher dexpression in the cerebellum and also a slightly higher muscimol affinity of a6bd GABA A Rs (see below). Binding of 5 nM [ 3 H]muscimol to non-d-GABA A Rs in dKO forebrain was around 100 fmoles and about 300 fmoles (per mg membrane protein) in the cerebellum (see Fig. 2). Considering that generally binding to brain membranes is about ten times higher (1-2 pmol/mg membrane protein (Sieghart et al. 1987;Kontturi et al. 2011) for benzodiazepine ligands (with only one binding site, vs. two for muscimol), the amount of [ 3 H]muscimol binding suggests that in the forebrain only a rather small fraction (~5%) of non-d-GABA A Rs were occupied by muscimol under our binding conditions (Table 1).

High-affinity [ 3 H]muscimol binding to d receptors is because of changes in binding kinetics, particularly very slow dissociation kinetics
To better illustrate high-affinity muscimol binding kinetics to d-GABA A Rs, we subtracted binding from non-d-GABA A Rs in dKO mice from binding in WT mice and also normalized the level of 5 nM [ 3 H]muscimol binding to 100% at 15 min when the maximal binding was achieved (Figs 2 and 3). Both in the cerebellum and in forebrain, high-affinity muscimol association was faster to the small fraction of high-affinity non-d-GABA A Rs (mostly c2-GABA A Rs) when compared to d-GABA A Rs, which was surprising since faster muscimol association would contribute to higher muscimol affinity in d-GABA A Rs. This slower muscimol association to d-GABA A Rs is reflected in higher forebrain and cerebellar Fig. 2 Majority of high-affinity muscimol binding is d subunit-dependent. Association of [ 3 H]muscimol binding to forebrain (n = 8 independent experiments in both mouse lines using individual forebrains in each experiment) and cerebellar (n = 3 independent experiments using pools of 3 individual cerebella from the mouse line in each pool) membranes of WT and dKO mice (mean AE SEM). The experiments were performed in triplicate technical replicates. Forebrain and cerebellar membranes were incubated with 5 nM [ 3 H]muscimol alone and in the presence of 100 lM GABA to determine non-specific binding. The incubations were terminated at various time points by filtration onto GF/B filters. The values are expressed as fmol/mg protein (left panels) and as % of binding at 15 min (right panels). Binding to d-GABA A Rs (WT-dKO) was calculated by subtracting binding to non-d-GABA A RS in dKO mice from binding to WT membranes. association rate constants (k on ) of [ 3 H]muscimol binding to dKOs than to WT mouse membranes (Table 2 We also looked at muscimol dissociation in WT and dKO cerebella and forebrains by evaluating high-affinity (5 nM) [ 3 H]muscimol unbinding for up to 30 min. A comparison of muscimol dissociation between WT and dKO animals shows that almost all of the slow muscimol dissociation is because of d-GABA A Rs, with only a minor component present in both the forebrain and cerebellum of dKO animals, which is because of the high-affinity muscimol binding to non-d-GABA A Rs (Fig. 3).
d-GABA A Rs muscimol association (k on ) and dissociation rates (k off ) determine muscimol K D values in the low nM range After subtraction of binding to non-d-GABA A Rs in dKO mice from binding to total GABA A Rs in WT mice, we were able to determine a K D value based on the equation K D = k off /k on . The calculated K D value for d-GABA A Rs in the fore(mid)brain (predominantly a4bd) is 1.6 nM, and the K D for d-GABA A Rs in the cerebellum (a6bd) is 1.1 nM. Therefore, under our binding conditions [5 nM [ 3 H]muscimol and 22°C], the majority of d-receptors both in forebrain and cerebellum should be occupied at equilibrium.
We also determined dissociation rate constants of the highmuscimol affinity component in dKOs and WT brains, although the majority of non-d-GABA A Rs have low affinity and are therefore not occupied at 5 nM [ 3 H]muscimol. Dissociation rate constants k off of [ 3 H]muscimol binding were higher in dKOs than in WTs in both forebrain (p < 0.001) and cerebellar membranes (p < 0.05) (  of the calculated native d-GABA A Rs in both forebrain and cerebellum were smaller than those of dKO indicating slower [ 3 H]muscimol dissociation from d-GABA A Rs than from c2-GABA A Rs (Table 2, Fig. 3). Both forebrain and cerebellar k off values were also lower in calculated native d-GABA A Rs than in WT mice.
Association and dissociation-binding kinetics of 5 nM [ 3 H]muscimol to recombinant GABA A R subtypes Measurements in native brain tissues have the advantage that we can measure native receptors. The disadvantages is that the fraction of d receptors is variable (up to 10% of a4bd receptors in the fore/midbrain depending on brain region, up to 30% of a6bd receptors in the cerebellum). In addition, because of the low-muscimol affinity of most c2-GABA A R conformations, the fraction of non-d-GABA A Rs occupied by 5 nM [ 3 H]muscimol is low and probably highly variable because of differences in high-affinity (desensitized) conformations which could also depend on subunit composition. We therefore decided to measure association and dissociation on selected recombinant receptor subtypes. As observed for native d-GABA A Rs [ 3 H]muscimol association at 22°C was much slower in a6b2d receptors when compared to high-affinity binding to a1b2c2 and a6b2c2 recombinant receptors (Table 2, Fig. 4a). The association rate constant k on for a6b2d subtype was 6.3-11-fold lower when compared to c2-GABA A Rs (p < 0.001, one-way ANOVA).

Dissociation of [ 3 H]muscimol from recombinant GABA A R subtypes
Dissociation of [ 3 H]muscimol from a6b2, a6b2c2 and especially a1b2c2 receptor subtypes was very fast (Table 2, Fig. 4b and c). Dissociation from a1b2 receptors was 'intermediate' while it was very slow from both the a1b2d and a6b2d subtypes, in axb2d significantly slower than dissociation from the corresponding axb2c2 subtypes (p < 0.01, p < 0.001; one-way ANOVA) (Table 2, Fig. 4d). From association and dissociation rates, we calculated K D values of 0.72 nM for a6b3d and 1.3 nM for a6b2d GABA A Rs, which are in excellent agreement with the values observed with native d-GABA A Rs (see Table 2).
Since radioligand binding is frequently performed in an icewater bath (0°C), we decided to compare [ 3 H]muscimol dissociation kinetics at 22°C with unbinding at lower temperature (0°C) on selected c2 and d-GABA A R subtypes. At 0°C dissociation from a6b2d and a6b2c2 were significantly slower than from a1b2c2 GABA A Rs with 70% of [ 3 H]muscimol still remaining bound to a6b2d subtype at 30 min after start of the dissociation (Table 3, Fig. 4d; p < 0.001). [ 3 H]Muscimol dissociated also significantly slower from a6b2d when compared to a6b2c2 GABA A Rs  Association (k on ) and dissociation rate constants (k off ) of [ 3 H]muscimol binding in forebrain samples (association, n = 8, dissociation, n = 4 independent experiments made using individual animal forebrains) and in samples each of pooled from 3 mouse cerebella (n = 3 independent experiments made using pooled samples from 3 individual animal cerebella), and in recombinant receptors (n = 3-6 independent experiments each performed using receptors from independent transfections/ expressions) (mean AE SEM). n.d., not determined. All experiments were performed in triplicate technical replicates. Statistical comparison of forebrain and cerebellar values: *p < 0.05, **p < 0.01, ***p < 0.001, significantly different from the corresponding WT value, unpaired t-test. Statistical comparison of recombinant receptor values: ###p < 0.001, significantly different from the corresponding a1b2c2 value; † † †p < 0.001, significantly different from the corresponding a6b2d value (one-way ANOVA followed by Tukey's post hoc test).
( Table 3, Fig. 4d; p < 0.01, one-way ANOVA followed by Tukey's post hoc test), a difference that was also noted at 22°C (see Fig. 4c). It can be approximated that at 0°C the dissociation of [ 3 H]muscimol binding during the first 15 s after pre-incubation in recombinant a1b2c2 receptors is about 10%, so we can assume that during the 15 s ice-cold washing period the amount dissociated is in that magnitude for a1b2c2 receptors and less for a6b2c2 and a6b2d receptors (Fig. 4d). The binding affinities, number of [ 3 H]muscimol binding sites as well as binding kinetics are in the same range as found in the literature (Wang et al. 1979;Agey and Dunn 1989;Maksay 1990;Negro et al. 1995;Ebert et al. 1999). However, because of the missing high-affinity d-receptors with slow kinetics, our association and dissociation rates in dKO fore/midbrain, dKO cerebellum, and recombinant a1b2c2 receptors are an exception as they were faster than all association rates in the former published studies.

Co-expression of the d subunit leads to sub-nanomolar muscimol currents
To determine the effect of muscimol on expressed recombinant receptors, we compared muscimol dose-response curves evoked with both a4b3d and also binary a4b3 receptors. Fig. 5 shows a representative muscimol concentration-response curve with a4 and b3 subunits either with (Fig. 5a) or without the d subunit (Fig. 5c). Co-expression of the d subunit leads to receptors that respond to much lower muscimol concentrations with a threshold as low as 0.1 nM (Fig. 5a), whereas with a4b3 receptors the threshold moves to about 30 nM muscimol (Fig. 5c), indicating that d coexpression dramatically increases muscimol sensitivity. A closer inspection of the current traces also reveals that muscimol currents look rather different, with a4b3d muscimol evoked currents showing a very slow return to baseline that is absent in a4b3 receptor. Such slow muscimol current deactivation is expected for a high affinity, minimally desensitizing with a slow ligand/muscimol dissociation rate as seen in our binding studies on both native and recombinant GABA A Rs. In our oocyte recording chamber solution exchange takes about 1-2 s, which in many cases is not fast enough to reliably record current kinetics. However, since association rates are concentration dependent and therefore very slow at low nanomolar muscimol concentrations, they actually can be resolved under our perfusion conditions (see current close-up in Fig. 5b), and since these are very high affinity receptors, it takes several minutes for currents to return to baseline.

Discussion
Muscimol has long been known as a general GABA A R agonist, although numerous lines of evidence have emerged over the years that suggested that muscimol and also THIP, both conformationally restricted GABA analogs (see Fig. 1a) have considerable selectivity at low doses for extrasynaptic d subunit-containing receptors. It was shown in brain sections that dKO mice had a complete loss of 6 nM [ 3 H]muscimol binding in the forebrain, with a substantial reduction in binding in the cerebellum (Mihalek et al. 1999, Fig. 1b). Knockout of a6 subunit (a6KO mice) lead to an essentially complete loss of high-affinity [ 3 H]muscimol binding in the cerebellum (M€ akel€ a et al. 1997, Fig. 1b). This suggested that high-affinity muscimol Representative concentration-response data (out of 3 similar recordings made using injections into different batches of oocytes) using muscimol concentrations from 0.1 nM up to 100 lM on (a) a4b3d-or (c) a4b3-injected oocytes. Muscimol concentrations from 0.1 nM to 30 nM activate currents only in a4b3dinjected oocytes, but not in the absence of d subunits in a4b3 injected oocytes. (b) Slow current activation (association rates are slow at these low muscimol concentrations because association is concentrationdependent) and also current deactivation at the lowest doses (expanded in b) and the two-component decay for doses ≥ 10 nM. (d) Superimposed responses to 300 nM muscimol from a4b3d-and a4b3-injected oocytes. The responses were scaled so that the a4b3 300 nM muscimol current fits the fast current component in a4b3d-injected oocytes.
binding to brain sections is dand, in the cerebellum also a6-subunit dependent.
Binding studies are generally performed on ice (0°C), electrophysiological measurements are typically performed at 22°C and in rodent behavioral experiments receptors are studied at body temperature (37°C). Such temperature differences could have a major influence on binding affinities of GABA and GABA analogs. Also, the high-affinity muscimol binding sites have been interpreted to represent desensitized or otherwise non-functional high-affinity conformations (Agey and Dunn 1989;Chandra et al. 2010). In addition, recombinant d-GABA A Rs so far have been shown to be fairly insensitive to muscimol requiring micromolar muscimol concentrations. Given all these uncertainties of temperature influence on binding affinity, conformational binding heterogeneity, and the absence of any evidence for highly muscimolsensitive functional GABA A Rs, it is not surprising that there is still considerable uncertainty of how muscimol affects different GABA A R subtypes.
We studied here GABA A R dKO and WT mice and recombinantly expressed GABA A Rs for high-affinity 5 nM [ 3 H]muscimol binding at 22°C to be able to compare them with electrophysiological data usually collected at 22°C. We show that under these conditions both in the fore/midbrain as well as in the cerebellum dKO animals lose~60% of highaffinity [ 3 H]muscimol 22°C binding, indicating that despite their low abundance, d-GABA A Rs form the majority of highaffinity muscimol-binding sites in the mouse brain.
In mouse forebrain and cerebellar membranes (Fig. 3), the rate of [ 3 H]muscimol dissociation was faster from dKO membranes than from WT membranes ( Table 2) and both forebrain and cerebellar WT membranes have a much slower component for dissociation, that is lacking in dKO membranes. These results are corroborated by our recombinant receptor dissociation experiments, which show much slower muscimol dissociation from expressed d-receptors (see Fig. 4b and c). Analysis of the binding kinetics suggested that the presence of the d subunit decreases association and even more so dissociation rates when compared to non-d GABA A R subtypes, leading to calculated dissociation constants (K D = k off /k on ) of 1.1 nM in the cerebellum and 1.6 nM in fore/mid-brain (see Table 2). However, about 40% (forebrain) of high-affinity binding remains in dKO mice with both association and dissociation faster than those observed for d-GABA A Rs (Figs 2 and 3), but in sum the calculated (from k on and k off ) apparent [ 3 H]muscimol affinities (K D ) for these non-d-GABA A Rs were also around 1 nM (see Table 2). In the cerebellum, relatively high-affinity a6bc GABA A Rs likely make a major contribution to highaffinity binding to non-d-GABA A Rs (see Fig. 1b, M€ akel€ a et al. 1997). The fairly slow dissociation of muscimol from non-d-GABA A Rs may help to explain differences found between [ 3 H]muscimol membrane homogenate binding (Fig. 3) when compared to [ 3 H]muscimol receptor autoradiography studies (Fig. 1b). During short washing procedures, only fairly small amounts of [ 3 H]muscimol dissociate whereas the much longer autoradiography washing periods would allow [ 3 H]muscimol to largely dissociate from non-d-GABA A Rs (mostly a1-5bc2 in the forebrain) and partly also from higher affinity a6bc2 receptors, while the extremely slow dissociation from d-GABA A Rs allows the majority of muscimol to be retained as seen in autoradiographs (M€ akel€ a et al. 1997;Korpi et al. 2002a,b, Fig. 1b).
The residual high-affinity binding to non-d-GABA A Rs in the forebrain still remains somewhat mysterious since there is no evidence for any functional muscimol responses on recombinantly expressed non-d-GABA A Rs at low nanomolar [muscimol]. It should be noted that we estimate that < 10% of total non-d-GABA A Rs are occupied by 5 nM [ 3 H]muscimol (see Fig. 2) under our conditions in the forebrain and therefore contribute to high-affinity binding. Since it has been reported that desensitization reversibly shifts a1b2c2 GABA A Rs into a high-affinity state (Maksay and Ticku 1984;Chang et al. 2002;Newell and Dunn 2002), highaffinity muscimol binding to desensitized GABA A Rs (which do not contribute to muscimol-induced currents), seems to be a plausible explanation. Another (not mutually exclusive) possibility is that such high-affinity binding to non-d-GABA A Rs is due to freezing, since at 22°C room temperature high-affinity binding was lower when never-frozen whole brain membranes were used (Yang and Olsen 1987). The notion that high-affinity c2-GABA A R muscimol sites are non-functional desensitized receptors and/or freezing/ cooling artifacts, is consistent with the observation that behavioral low-dose muscimol sensitivity is dependent on d-GABA A Rs (Chandra et al. 2010).
We show here for the first time that co-expression of the d subunit leads to highly muscimol-sensitive a4b3d currents. Remarkably, the EC 50 for the high-affinity muscimol component shown in Fig. 5a is in the same range as K D for binding at 22°C. In contrast, and despite some high-affinity binding to a fraction of non-d-GABA A Rs, there is no evidence for highly muscimol-sensitive currents in recombinantly expressed ab (Fig. 5) and abc receptors (Adkins et al. 2001;St orustovu and Ebert 2006;Mortensen et al. 2010). With a functional correlate missing for high-affinity [ 3 H]muscimol binding to native non-d-GABA A Rs and recombinant c2-GABA A Rs it is possible that this highaffinity binding to non-d-GABA A Rs is a binding assay artifact and largely irrelevant for functional and behavioral responses. If real, that is, found in native non-d-GABA A Rs, and not non-functional desensitized forms, such high-affinity binding sites could contribute, besides relatively high-affinity a6bc GABA A Rs, to behavioral high dose muscimol (and THIP) effects in dKO mice.
Recombinant expression of functional recombinant d-GABA A Rs is challenging since they generally show biphasic GABA and THIP concentration response curves likely because of incomplete d subunit incorporation into functional receptors (Meera et al. 2010(Meera et al. , 2011Karim et al. 2012;Hoestgaard-Jensen et al. 2014). As seen here in Fig. 5 also the muscimol concentration-response curve on a4b3d receptors shows two components, with the low-sensitivity component similar to what is seen with receptors formed by only a and b subunits, without d subunits (Fig. 5) and a high affinity and slowly deactivating current component. Our highly muscimol-sensitive d-GABA A Rs (Fig. 5) contrast with previous reports of recombinantly expressed a4/6b3"d" receptors: Reported muscimol EC 50 values are 200 nM on a4b3d receptors (Mortensen et al. 2010), 160 nM for a6b3d receptors and 2.28 lM on a4b3d receptors (St orustovu and Ebert 2006). Since these reported EC 50 values are in the same range as we see with a4b3 receptors without d (see Fig. 5c), they are likely the result of low d subunit incorporation into functional receptors in recombinant expression systems. Note that our d-binding data, using a1b2d and a6b2d GABA A Rs shown in Fig. 4 are clear-cut, with only little evidence of biphasic kinetic responses. A plausible and likely explanation is that with high-affinity binding to recombinantly expressed d-GABA A Rs only a very small fraction of contaminating low-muscimol affinity/ sensitivity ab receptors would actually be occupied at 5 nM [ 3 H]muscimol.
Native and recombinantly expressed d-GABA A Rs have been suggested to be activated by relevant low ethanol concentrations (Hanchar et al. 2005). Given that both ethanol and muscimol are d-GABAR selective drugs it may not be surprising that muscimol leads to increased alcohol impairment (Frye and Breese 1982). In addition, chronic ethanol treatment leads to a substantial reduction in high-affinity [ 3 H]muscimol-binding sites (Negro et al. 1995), which meshes well with the notion that chronic alcohol leads to a reduction in d-GABA A R-mediated tonic currents and d-subunit surface expression, a process that likely contributes to alcohol tolerance and the development of alcohol dependence (for review see Olsen and Liang 2017).
Blood-brain barrier (BBB) permeability usually correlates with lipid-solubility and is therefore rather poor for highly water-soluble molecules like GABA, muscimol and THIP. Consistent with a low BBB permeability it has been shown that only around 0.02% (1/5000) of peripherally injected [ 3 H]muscimol actually entered the rat brain (Maggi and Enna 1979). High-affinity muscimol d-GABA A Rs reported here provide a plausible explanation for brain muscimol effects, despite very low effective muscimol concentration in the brain. The program EpiSuite gives the logP (partition coefficient) value À3.60 for GABA, whereas adding hydrophobic ring structures in muscimol (logP = À1.71) and THIP (logP = À0.81) (see Fig. 1) shifts the balance from hydrophilic to more lipophilic (Estimation Programs Interface Suite TM for Microsoftâ Windows, v 4.11, United States Environmental Protection Agency, Washington, DC, USA). It seems therefore likely that GABA has the lowest BBB permeability, followed by muscimol and THIP. Given that THIP affinity for d-GABA A Rs is lower when compared to muscimol (Friemel et al. 2007;Meera et al. 2011) it is tempting to speculate that the higher BBB permeability of THIP compensates to a large extent for its much lower potency on d-GABA A Rs, with both of them having apparently very similar behavioral effects (Chandra et al. 2010). The two GABA A agonist binding sites in GABA A Rs are located at the two extracellular b+ainterfaces (Ernst et al. 2003) and so it is possible that these two GABA/muscimol binding sites do not have same affinities, and also that affinities for GABA site ligands could change once one of the sites is occupied. We show here that substitution of the c2 by d subunit has drastic effects on slowing [ 3 H]muscimol association and even more so dissociation kinetics. While the subunit stoichiometry and organization of d-GABA A Rs has not been resolved unequivocally, there is direct evidence for a simple c2 to d substitution from 2a:2b:c2 to 2a:2b:d (Barrera et al. 2008). Therefore, it is likely that a 2a:2b:d receptor would also have two GABA A agonist/muscimol sites, one at each b+ainterface (-b+-a+-d+-b+-a+), without the d subunit actually directly contributing to the GABAbinding site. This implies that d increases the GABAbinding-site affinity and slows muscimol dissociation in the badba pentamer allosterically. The reciprocal of dissociation rate constant, the drug-target residence time s (= 1/k off ), has been shown to often predict in vivo efficacy better than binding affinity (Pan et al. 2013;Copeland 2016) and may help explain why the d-subunit is required for low dose muscimol behavioral effects.
It appears that in general the GABA analog muscimol is similar to GABA in many aspects, only that it shows about 100-1000 times higher affinity (with THIP having intermediate affinity) across the board for different GABA A R subtypes (with a6-containing GABA A Rs more sensitive). For example, muscimol EC 50 for a1bc2 GABA A Rs is~1 lM, whereas GABA EC 50 is~100 lM (Karim et al. 2013). In contrast, for d-GABARs, the GABA EC 50 is typically~0.3-1 lM, whereas we show here that such d-GABARs not only bind muscimol with low nanomolar K D , but also that co-expression of d (with a4 and b3) induces low nanomolar muscimol currents.
Our results are similar to the other isoxazole GABA A analog THIP, which has been shown to be highly selective for d-GABA A Rs (Meera et al. 2011). This paints a consistent picture in which extrasynaptic d-GABA A Rs are not only exquisitely sensitive to GABA, but also the GABA analogs THIP and muscimol. Since muscimol is a widely used experimental pharmacological tool in neuroscience research, our findings will help to better interpret in vivo and in vitro experiments that involve muscimol. While muscimol itself is unlikely to find therapeutic application, our results could help to characterize GABA analogs and GABA-site ligands for potential therapeutic applications. For example, recent work suggested that a6bd-selective agonists might be useful in the clinic as antitremor medications (Handforth et al. 2018).