Mutations in the Gabrb1 gene promote alcohol consumption through increased tonic inhibition.

Alcohol dependence is a common, complex and debilitating disorder with genetic and environmental influences. Here we show that alcohol consumption increases following mutations to the γ-aminobutyric acidA receptor (GABAAR) β1 subunit gene (Gabrb1). Using N-ethyl-N-nitrosourea mutagenesis on an alcohol-averse background (F1 BALB/cAnN x C3H/HeH), we develop a mouse model exhibiting strong heritable preference for ethanol resulting from a dominant mutation (L285R) in Gabrb1. The mutation causes spontaneous GABA ion channel opening and increases GABA sensitivity of recombinant GABAARs, coupled to increased tonic currents in the nucleus accumbens, a region long-associated with alcohol reward. Mutant mice work harder to obtain ethanol, and are more sensitive to alcohol intoxication. Another spontaneous mutation (P228H) in Gabrb1 also causes high ethanol consumption accompanied by spontaneous GABA ion channel opening and increased accumbal tonic current. Our results provide a new and important link between GABAAR function and increased alcohol consumption that could underlie some forms of alcohol abuse.

INTRODUCTION

Our understanding of the genetic and molecular basis of alcohol dependence is incomplete. Alcohol abuse has long been associated with facilitation of neurotransmission mediated by the brain’s major inhibitory transmitter, GABA, acting via GABAA receptors (GABAARs). Recently, a locus within human chromosome 4, containing GABAAR subunit genes encoding α2, α4, β1 and γ1 subunits has been associated with alcohol dependence in humans [1, 2, 3, 4, 5, 6, 7]. In particular, haplotypic variations in the GABRA2 gene encoding the α2 subunit have been repeatedly linked with alcohol dependence[2, 8, 9, 10]. However, the neurobiological basis by which genetic variation translates into alcohol abuse is largely unknown.

Ionotropic GABAARs are pentameric ligand-gated ion channels, drawn from a family of 19 proteins, which underpins the expression of ~20-30 neuronal GABAAR isoforms[11]. These receptors have distinct physiological and pharmacological properties, are heterogeneously expressed in the mammalian CNS and as a consequence can differentially influence behavioral phenotypes[12, 13]. Synaptic GABAARs mediate phasic inhibition, whereas extrasynaptic GABAARs are activated by ambient concentrations of GABA and mediate a tonic form of inhibition. Recent evidence has suggested roles for both forms of GABAergic transmission in the neurobiology of addiction[14, 15, 16, 17].

With regard to ethanol, both consumption and preference are reduced following disruption of GABA-mediated tonic inhibition in δ subunit knock-out (δ−/−) mice [18] and a similar impact on ethanol drinking was achieved by RNAi-induced suppression of either α4 (a subunit partner of the δ subunit) or δ subunit expression in the rodent nucleus accumbens (NAc)[17, 18]. The reduced ethanol self-administration appeared to be a consequence of the altered reinforcing properties of the drug[17, 18]. These actions on ethanol drinking and self-administration appear to be specific to the activity of α4βδ receptors and not a general effect on manipulating GABAergic activity in the NAc, as no differences in ethanol self-administration have been found in α2−/− or α5−/− mice[19, 20], when compared to wild-type (WT) counterparts.

It has been suggested that ethanol may exert a direct action on δ-GABAARs to enhance their function, which might account for the rodent self-administration data. However, whether ethanol exerts such direct effects is controversial[21, 22, 23, 24]. An alternative interpretation of the behavioral data posits that the activity of NAc extrasynaptic GABAARs influences the activity of neural circuits underlying certain addictive behaviors, such as the desire for alcohol. We were able to explore this possibility by exploiting the availability of two novel mutant mouse lines in which single point mutations in GABAAR β1 subunits have occurred. The first line was identified through a phenotype-driven N-ethyl-N-nitrosourea (ENU) mutagenesis screen[25, 26] for alcohol-preferring mice, whilst the second was a spontaneous mutation identified using a genotype-driven approach[27]. Both mutant mouse lines display increased alcohol consumption and self-administration, and are characterized by GABA channels that can open spontaneously, promoting an increased tonic inhibition in NAc medium spiny neurons (MSNs). These findings strongly suggest that the large tonic conductance of MSNs contributes to the increased preference and intake of ethanol by Gabrb1+/ and Gabrb1+/ mice. These data reveal a novel link between GABAAR function and increased alcohol consumption that could lead to a better understanding of some forms of alcohol abuse.

RESULTS

Alcohol preferring mouse strains

We identified two dominant mutations in Gabrb1 that induced a phenotypic switch from alcohol aversion to a sustained, strongly-heritable alcohol preference. One mutation was generated through phenotype-driven random ENU mutagenesis[25, 26] (Gabrb1), whilst the other was a spontaneous mutation (Gabrb1) identified through gene-driven screening of a DNA/sperm archive[27]. Both lines showed highly-penetrant ethanol preference throughout over eight and five backcross generations respectively.

An ENU-induced Gabrb1 mutation confers alcohol preference

Alcohol-averse male BALB/cAnN mice were exposed to ENU[25] and crossed to WT C3H/HeH females. G1 progeny (n=1047) were screened in a two-bottle choice test for preference for 10% (v/v) ethanol with mice showing ethanol preference backcrossed to C3H/HeH to test heritability. The ENU-induced mutation was localized to a region on mouse chromosome 5 (71.45-73.05 Mb), syntenic with a region on human chromosome 4, containing 11 genes including Gabra4 and Gabrb1 (Supplementary Fig S1). Sequencing identified just one mutation in the β1 subunit (Gabrb1 exon 8), a leucine-to-arginine exchange (L285R) in the highly-conserved third transmembrane domain (M3), near the M2-M3 linker (Supplementary Fig S2A), an important area for GABA receptor activation and ion channel gating. The mutation was absent in both parental strains.

A spontaneous Gabrb1 mutation confers alcohol preference

To determine if other Gabrb1 mutations modified alcohol drinking, we screened a DNA library containing ~10,000 unique samples from ENU mutagenized male mice (F1 C57BL/6JxC3H/HeH) and identified one sample carrying a non-synonymous proline-to-histidine mutation (P228H) within M1 of the β1 subunit. This proline is highly conserved in GABAARs from various species (Supplementary Fig S2B). While absent in both background strains, the mutation occurred in the non-ENU mutagenized C3H DNA strand and therefore had arisen spontaneously. Significantly, as with Gabrb1+/ above, Gabrb1+/ mutants displayed a heritable ethanol preference (Supplementary Fig S3).

Gabrb1+/ mutant mice were significantly smaller than WT (P<0.0001; Supplementary Figure S4). A similar but less pronounced size difference was also observed in Gabrb1+/ mutant mice. Both mutant lines exhibited complete female infertility characterized by underdeveloped corpus luteii. Consequently, we could only study heterozygotes (Gabrb1+/ and Gabrb1+/) and their corresponding wild-type (WT) mice. Reduced body size and impaired fertility most likely reflect hypothalamic-pituitary axis endocrine deficiency and so analysis of anterior pituitary hormone content was performed. This demonstrated that carriage of either mutant Gabrb1 allele (Gabrb1 or Gabrb1) was associated with significant hypothalamic-pituitary axis dysfunction (Supplementary Tables S1 & S2). Mutant mice displayed a significant preference for ethanol (Gabrb1+/: Fig 1A, B; Gabrb1+/: Supplementary Fig S3), but not for similarly presented sucrose, saccharin, or quinine solutions, suggesting taste or calorific requirements were not driving alcohol preference (Supplementary Table S3). Furthermore, all mice consumed similar daily fluid volumes, despite Gabrb1+/ and, to a lesser extent, Gabrb1+/ mutant mice possessing lower body weights (Supplementary Fig S4). The co-segregation of both alcohol preference and body weight traits were observed throughout all eight (for Gabrb1+/) and five (Gabrb1+/) generations studied, being present in 123/130 (93%) of Gabrb1+/ mice effectively excluding the possibility of separate mutations causing these components.

Figure 1

Alcohol Consumption and Behavioral Phenotype of Gabrb1+/L285

(A) Ethanol (10% v/v) preference of male Gabrb1+/ (Mean = 68.37%, 95% CI 65.61-71.13, n = 46) and WT littermates (24.67%, 22.20 −27.13, n = 44; t-test P<0.0001).

(B) Daily ethanol consumption (g/kg body weight): Male Gabrb1+/ (10.16, 95% CI 9.08-11.25, n = 46); WT littermates (1.24, 0.71-1.75; t-test P<0.0001, n = 44). (C) Amounts of ethanol earned (g/kg body weight ± SEM). * P<0.05 (n=8/group, ANOVA plus Bonferroni post-hoc test). (D) Numbers of lever presses ± SEM over 30 min by WT and mutant mice on a fixed ratio 4 (FR4) schedule for sucrose-ethanol reinforcers. Note increased number of lever presses for mutants at reinforcer mixtures of 7% sucrose / 5% ethanol, and 10% sucrose / 10% ethanol (n=8/group). * P<0.05, ANOVA plus Bonferroni post-hoc test.

Gabrb1 and alcohol seeking behavior

The motivation for Gabrb1+/ mice to obtain alcohol was assessed using operant self-administration of fluid coupled with a sucrose-fading technique[19, 20]. Self-administration of ethanol over 1 hour caused ataxia, consistent with alcohol intoxication (Supplementary Fig S5A). To determine whether the features of intoxication observed in mutant mice during the 1 hour operant sessions were due to increased intake or a heightened sensitivity to alcohol, WT and mutant mice were given a standard bodyweight-adjusted dose of ethanol and tested for the presence of ataxia on the rotarod (ethanol 3g/kg) and loss of righting reflex (ethanol 3.5g/kg; Supplementary Fig S5). Although study groups were relatively small (n = 7 - 8), limiting statistical power, there was a strong tendency for the mutant mice to be more impaired than the WT mice (two-way ANOVA, main effect of genotype: F(1,13) = 4.46; p = 0.05), and to recover more slowly (two-way ANOVA, time point*genotype interaction: F(1,13)=1.76; p = 0.08). Genotype did not affect the rate of loss of righting reflex, but there was a trend towards a more rapid recovery in WT mice (806±140 vs. 1466±448 seconds (SEM); t-test, t = 1.41; p < 0.1). Since neither the peak, nor time course of blood or brain alcohol levels differed significantly between mutant and WT mice following ethanol administration (Supplementary Fig S6), these data suggest an increased sensitivity of Gabrb1+/ mice to the ataxic effects of ethanol (Supplementary Figs S5B, S5C and S6).

To ensure ataxia did not affect performance, we analyzed only the first 30 min of each operant self-administration session. Although rates of lever pressing declined with lower sucrose and higher ethanol concentrations (P<0.001, two-way ANOVA n = 8 / group), these were still higher with 7/5% and 10/10% v/v sucrose/ethanol (Fig 1C, D) in Gabrb1+/ mice compared to WT (P<0.01, two-way ANOVA n = 8 / group). Above 5% v/v ethanol, mutant mice worked harder to obtain ethanol, irrespective of the sucrose concentration, tending towards lower lever-pressing rates for unadulterated sucrose (P<0.1, two-way ANOVA n = 8 / group). Thus, the Gabrb1 mutation is specific in affecting the motivation to consume alcohol.

Consistent with higher response rates, inter-response times (IRTs) were significantly shorter in the mutant mice (Supplementary Fig S7). In order to obtain information on patterns of responding within a session, we analysed the pattern of IRTs within sessions. Interestingly, WT mice showed increases in IRTs as the session progressed, consistent with them satiating on alcohol. In contrast, the Gabrb1+/ mice maintained their shorter inter-response times (Supplementary Fig S7). This pattern suggests that, in comparison with the WTs, their desire for alcohol decreased more slowly as they consumed alcohol.

Following the tests of alcohol self-administration, the mice were tested over two sessions in extinction (i.e. lever pressing no longer resulted in fluid presentation). Their response rates declined when alcohol was no longer delivered following lever presses. The higher response rates maintained by Gabrb1+/ mice during extinction sessions (Supplementary Fig S8), were not significant, suggesting that there were no major differences in motivation to obtain ethanol under deprivation conditions. However, mice could not consume ethanol during extinction sessions and so any differences in rate of satiation to ethanol would not influence responding.

Mutant GABAAR β1 subunits increase NAc tonic inhibition

The NAc is an important brain region for understanding the neurobiology of reward and addiction. Implicating GABAARs, specific suppression of either the GABAAR α4 or δ subunits in the NAc decreases both ethanol consumption and preference in rats [16, 17]. We therefore determined how the β1 mutations affected GABAAR function in NAc slices. Whole-cell voltage-clamp (−60mV) of Gabrb1+/ NAc core medium spiny neurons (MSNs) revealed greater membrane current noise (root mean square (RMS)) and holding currents than for WT (Table 1). These effects were GABAAR-mediated as the receptor antagonist bicuculline (30μM) induced an ~6-fold greater outward current and a reduction in RMS for Gabrb1+/ than for WT MSNs (Table 1, Fig 2A, C).

Table 1

NAc Tonic Conductance.

	WT	Gabrb1 +/L285R	WT	Gabrb1 +/P228H
IHOLD (pA)	212 ± 24n = 12	284 ± 14 **n = 19	156 ± 12n = 15	296 ± 20 ***n = 12
RMS (pA) (Control)	4.3 ± 0.2n = 12	7.9 ± 0.2 ***n = 19	3.9 ± 0.1n = 15	7.0 ± 0.2 ***n = 12
RMS (pA) (+ Bic)	2.7 ± 0.2n = 6	4.4 ± 0. 2 ***n = 9	2.6 ± 0.1n = 11	3.5 ± 0. 1 ***n = 8
Δ RMS (pA) (+ Bic)	1.1 ± 0.2n = 6	3.7 ± 0.3 ***n = 9	1.1 ± 0.1n = 11	3.7 ± 0.2 ***n = 8
IBIC. (pA)	22 ± 3n = 6	120 ± 8 ***n = 9	22 ± 3n = 11	115 ± 9 ***n = 8
IGBZ. (pA)	17 ± 4n = 5	19 ± 3 †††n = 4	16 ± 5n = 4	46 ± 10 * †† ‡n = 3
I(pA) Bic + Ptx	27 ± 5n = 4	171 ± 8 ***n = 4	20 ± 10n = 3	137 ± 26 *n = 3
I(pA) Ptx after Bic	4 ± 2n = 4	41 ± 8 **n = 4	−5 ± 3n = 3	19 ± 6 *n = 3
RMS (pA) (Bic + Ptx)	2.9 ± 0.2n = 4	2.9 ± 0.2n = 4	2.6 ± 0.1n = 3	2.8 ± 0.1n = 3
ITHIP (pA)	60 ± 15n = 7	41 ± 9n = 4	N.D.	N.D.

ND = Not determined; Bic = bicuculline; Ptx = picrotoxin; Gbz = gabazine

Statistical significance of mutant relative to its WT:

P<0.05;

P<0.01;

P<0.001.

Statistical significance of IGBZ relative to IBIC:

P<0.01;

P<0.001.

Statistical significance of Gabrb1+/ relative to Gabrb1+/:

P<0.05.

(unpaired Student’s t test used in all cases)

Data presented as ± SEM

Figure 2

β1 Mutations Influence Tonic and Phasic Inhibition of Medium Spiny Neurons

Whole-cell recordings from accumbal MSNs of Gabrb1+/ (A), Gabrb1+/ (B) and WT littermates. Note for both Gabrb1+/ and Gabrb1+/ MSNs the larger outward currents produced by bicuculline (30 μM). Scale bars apply to both A and B. (C) Outward currents induced by bicuculline (30 μM), gabazine (20 μM) and picrotoxin (PTX, 100 μM) from WT, Gabrb1+/ and Gabrb1+/ MSNs (means ± SEM; n = 3-12). (D) Whole-cell recording from a Gabrb1+/ MSN. Gabazine (20 μM) abolishes the mIPSCs, but produces only a relatively small outward current. However, gabazine prevents the further outward current usually produced by bicuculline for this mutant, but not that produced by picrotoxin. (E) Mean outward current induced by GABAAR antagonists for Gabrb1+/ MSNs (means ± SEM; n = 4). Whole-cell recordings from a Gabrb1+/ (F) and a WT MSNs (G). Gabazine (20 μM) abolishes mIPSCs, but additionally produces only a relatively small outward current for both the Gabrb1+/ and WT MSNs. However, for the Gabrb1+/ MSN gabazine prevents a further outward current by subsequent bicuculline, but this antagonist does not prevent the outward current produced by picrotoxin. By contrast, for WT MSNs after gabazine both bicuculline and picrotoxin are inert. Superimposed averaged mIPSCs are shown for WT (black) and Gabrb1+/ , or Gabrb1+/ (I) MSNs (grey). Right-hand traces of each pair are normalized to the mean peak amplitude of the appropriate WT mIPSC. Note the prolonged mIPSCs for Gabrb1+/ and Gabrb1+/ over WT. Scale bars apply to both H and I. *P<0.05, ** P<0.01, *** P<0.001 (unpaired Student’s t-test for mutants relative to WTs).

Accumbal MSNs exhibit a GABA-dependent tonic conductance mediated by extrasynaptic α4βδ receptors[28]. However, the inward current induced by the agonist THIP, at a δ-GABAAR-selective concentration (1 μM) was similar for WT and Gabrb1+/ MSNs, implying the large tonic current for Gabrb1+/ was not caused by increased δ-GABAAR expression (Table 1). In contrast to bicuculline (a partial negative allosteric modulator[29, 30, 31]), the competitive antagonist gabazine (20 μM) only induced small outward currents in both WT and L285R MSNs (Table 1, Fig 2C, D, E), suggesting increased ambient levels of agonist (e.g. GABA, taurine) do not cause the larger tonic current.

The differential influence of GABAAR antagonists on Gabrb1+/L285R neurons may indicate that bicuculline, but not gabazine, shuts spontaneously-open mutant β1 GABAAR channels[29, 32]. Indeed, picrotoxin (100 μM), which is a non-competitive antagonist of GABAAR-gated chloride channels, when co-applied with bicuculline, produced an additional outward current selectively in Gabrb1+/ neurons (Table 1; Fig 2C); whilst gabazine (20 μM), which shares a common binding site with bicuculline, prevented the outward current induced by bicuculline (30 μM), but not by picrotoxin (100 μM; Table 1, Fig 2D, E).

Gabrb1+/ MSNs also exhibited a greater membrane current noise and holding current than WT MSNs (Table 1). In Gabrb1+/ neurons, bicuculline (30 μM) and, to a lesser extent, gabazine (20 μM), induced larger outward currents relative to WT (Fig. 2B, C). In common with the β1L285R mutation, after gabazine, the co-application of picrotoxin (100μM) to Gabrb1+/, but not to WT MSNs, induced an additional outward current (Fig 2F, G). Furthermore, as for β1L285R, gabazine prevented the additional outward current produced by bicuculline in Gabrb1+/ MSNs (Table 1; Fig 2F). Thus, Gabrb1 enhanced the tonic conductance, consistent with the ability of these mutant β1 GABAARs to open spontaneously.

β1 subunit mutations influence NAc phasic inhibition

The β1 subunit mutations also affected GABAAR-mediated phasic inhibition. For Gabrb1+/ and Gabrb1+/ MSNs the frequency of mIPSCs was reduced and their decay times prolonged. Additionally, the mIPSC amplitude was increased for Gabrb1+/ relative to WT MSNs (Fig 2H, 2I; Table 2),which was associated with an increased population of large amplitude mIPSCs, exhibiting slow decays. A scatter plot revealed a cluster of large amplitude (peak amplitude > 105 pA), slowly decaying (T70 > 19ms) events prevalent in Gabrb1+/ MSNs constituting only 0.7% of the total number of events for WT MSNs, but 8.8% (> 10-fold increase) of the Gabrb1+/ MSNs (Supplementary Fig. S9). Conversely, the proportion of events with a peak amplitude < 105 pA and T70 < 19ms decreased from 81.7% for WT to 54.2% for the Gabrb1+/ MSNs. This loss is accounted for not only by the greater percentage of events with a peak amplitude > 105 pA and T70 > 19ms but, additionally, by a higher proportion of events with an amplitude >105 pA (T70 ≤ 19 ms) and of those with a T70 > 19ms (but pA ≤ 105pA) (Supplementary Fig. S9). Such events may originate from a population of mutant postsynaptic β1-GABAARs, with increased open probabilities compared to WT β1-GABAARs.

Table 2

Brain Slice Electrophysiology: Synaptic transmission.

	WTn = 12	Gabrb1+/L285Rn = 13	WTn = 6	Gabrb1+/P228Hn = 7
Peak Amp. (pA)	62 ± 4	88 ± 6 **	53 ± 2	64 ± 5
Rise time (ms)	0.5 ± 0.02	0.5 ± 0.02	0.6± 0.02	0.5 ± 0.02
Tw (ms)	9.5 ± 0.6	14.4 ± 1.3 **	7.8 ± 0.5	10.9 ± 0.9 *
Frequency (Hz)	2.1 ± 0.5	0.6 ± 0.1 **	2.9 ± 0.5	0.8 ± 0.02 ***

Statistical significance of mutant relative to the appropriate WT:

P<0.05;

P<0.01;

P<0.001

(unpaired Student’s t test).

Data presented as ± SEM

Both the L285R and the P228H mutations were associated with a reduced mIPSC frequency compared to WT counterparts (Supplementary Table S2). Activation by THIP of presynaptic δ-GABAARs located either on accumbal interneurons (the main source of MSN somatic input) or on neighboring MSNs, reduces GABA release onto MSNs (mIPSC frequency, Control: 1.9 ± 0.2 Hz; + THIP 1 μM: 1.1 ± 0.2 Hz, data presented as ± SEM, n = 5; p<0.05 paired t-test). Therefore, it is conceivable that pre-synaptically located spontaneously-open β1-GABAARs either on interneurons, or neighboring MSNs would similarly influence GABA release, thereby reducing MSN mIPSC frequency.

Mutant β1 subunit expression and spontaneous channel opening

To explore how the β1L285R and β1P228H mutations affected GABAAR function, we used heterologous expression of WT and mutant recombinant GABAARs in HEK293 cells. Immunocytochemistry was used to assess the expression levels of myc epitope-tagged WT and mutant β1 subunits. The expression levels of β1L285R along with α2 and γ2L subunits, including enhanced green fluorescent protein (eGFP), revealed no differences either for cell surface or overall total fluorescence (Supplementary Fig S10). Similarly, the cell surface membrane and intracellular expression levels for β1P228H were also unaltered compared to WT β1 subunits (Supplementary Fig S11).

To examine whether the β1 subunit mutations (L285R and P228H) altered GABAAR receptor physiology and pharmacology, whole-cell recording was performed on α2β1L285Rγ2 and α2β1P228Hγ2 receptors expressed in HEK293 cells. Relative to WT, both mutations reduced the maximum current density induced by saturating GABA concentrations (1mM; Fig. 3A).

Figure 3

Effects of β1L285R and β1P228H on Recombinant GABAARs

(A) Mean maximal GABA current densities to saturating GABA concentrations for WT and mutant, α2β1γ2 receptors. (B) GABA concentration-response curves for WT and α2β1L285Rγ2 receptors. Responses are normalized to the maximum response from each cell. (C) Mean maximal outward currents produced by picrotoxin (PTX; 100μM) for WT and mutant receptors. Values are percentages of total current (= IPTX/(IPTX + IGABA Max). (D) GABA concentration-response curves for WT and α2β1P228Hγ2L receptors, analysed as in (B). (E) Single channel currents (sampled at 50 kHz) from outside-out patches (−70mV) of HEK293 cells expressing α2β1γ2, α2β1L285Rγ2, and α2β1P228Hγ2 receptors. Spontaneous channel activity is evident in the absence of GABA (traces labelled ‘Krebs’). Both spontaneous and GABA-activated channel openings are inhibited by bicuculline (Bic: 50μM). Scale bars: 2 pA, 100 ms. Single channel amplitude histograms (right panels) are based on 10-30 s epochs for α2β1γ2 (activated by 10μM GABA) and for spontaneously-active α2β1L285Rγ2 and α2β1P228Hγ2. All single channel currents have mean amplitudes of ~ −1.9 pA (~27 pS). Error bars in panels A-D represent the mean ± SEM (n = 7-15 for panel A: n = 4-6 for panels B & D: n = 6-11 for panel C).

As native GABAARs contain two β subunits and because our in vivo studies necessarily used heterozygous mice, a proportion of native GABAARs could comprise a mixture of WT and mutant β1 subunits. To examine the amplitudes of GABA-evoked currents under these conditions, we recreated a binomial mixture of GABAARs in HEK cells by co-expressing α2 and β1 with either β1L285R or β1P228H, and γ2 subunits in an equimolar ratio thereby reproducing native receptor isoforms likely to be present in heterozygotes. Assuming receptor subunit assembly proceeds according to binomial probabilities, we would expect a mixture of pentameric receptors (2α:2β:1γ) to include: WT, full mutant (both β subunits are mutated) and two forms of partial mutant receptors containing only one copy of the mutant β subunit. The maximum current densities induced by saturating concentrations of GABA (up to 1 mM) were larger for receptors containing only one copy of the mutant β subunit than those observed with full mutant receptors, and this approached that of WT receptors (Supplementary Fig S12).

GABA concentration response curves revealed that GABA potency was increased (~2-3-fold) by β1L285R and β1P228H (Fig 3B, D; Supplementary Table S4). Significantly, the holding current (at −40 mV) was greater for many mutant receptor-expressing cells compared to WT (Supplementary Table S4), indicative of spontaneous GABA channel activity[33]. Indeed, picrotoxin (100 μM) induced outward currents in the absence of GABA for cells expressing β1L285R, and to a lesser extent for β1P228H mutant receptors, reflecting spontaneous channel activity (Fig. 3C). The spontaneous current revealed by picrotoxin (IPTX) accounted for ~3-15% of the total current (= IGABA,max + IPTX) for these mutant receptors.

We also examined the level of spontaneous current for α4β1γ2 receptors, with WT or mutant β1 subunits, as an alternative isoform that may populate synaptic and/or extrasynaptic sites. Maximal current densities and holding currents exhibited greater variability for the β1 mutants compared to WT (Supplementary Table S4), and α4β1L285Rγ2 exhibited a spontaneous current revealed by picrotoxin (Fig 3C).

To examine the gating of GABA ion channels underlying the spontaneous current, we used outside-out patches from HEK cells expressing α2β1γ2, α2β1L285Rγ2 or α2β1P228Hγ2 receptors (Fig. 3E). Spontaneous channel activity was evident with mutant receptors, but absent in WT, and abolished by the GABA antagonist, bicuculline (50 μM), acting as a negative allosteric modulator[29] in the absence of GABA. Activating WT α2β1γ2 GABA channels with 10 μM GABA induced single channel currents that were indistinguishable from the spontaneous openings observed with the mutant receptors (Fig. 3E) and with only minor differences in open and shut time durations (Supplementary Table S5). Overall, β1L285R, but to a lesser extent, β1P228H caused spontaneous channel opening and increased receptor sensitivity to GABA when co-assembled with either α2 or α4, and γ2 subunits.

Given the importance of α4 and δ subunits for tonic current in NAc MSNs[16, 17, 28] we explored whether β1 mutants also conferred spontaneous channel activity on δ-GABAARs. Using HEK293 cells, β1 subunits were expressed with either α2 (co-located on same chromosome as β1, and expressed in accumbal synapses[34]), or α4 (co-expressed in the accumbens with δ forming extrasynaptic receptors)[28]. The β1L285R mutation increased the maximum current density to saturating GABA (300 μM) for both α2β1L285Rδ and α4β1L285Rδ receptors, compared to WT equivalents (α2/4β1δ), whereas β1P228H did not (Supplementary Fig. S13). Both β1 mutants increased GABA sensitivity by 3-5 fold at α2β1δ-GABAARs (Fig. 4A, Supplementary Table S4), similar to the increased sensitivity for mutant α2β1γ GABAARs (Fig. 3B), but only β1L285R affected α4β1δ receptor sensitivity to GABA (Fig 4B).

Figure 4

Effects of β1L285R and β1P228H on δ-containing recombinant GABAARs

(A & B) GABA concentration-response curves for WT and mutant α2β1δ and α4β1δ receptors, analysed as in Figure 3B. (C) Representative membrane currents for GABA (black) or picrotoxin (red) recorded for WT and mutant (L285R) α2β1δ and α4β1δ receptors (VH = −40mV). (D) Mean maximal outward currents after PTX (100 μM; expressed as in Figure 3C) applied in the absence of GABA for WT and mutant α2β1δ (solid bar) or α4β1δ (open bar) receptors. (E) Mean GABA responses (EC10) in the presence of 30mM EtOH for WT and mutant receptors (Control EC10 responses =100%). All error bars in panels A, B, D & E represent the mean ± SEM (n = 5-12 for panels A & B: n = 7-12 for panel D: n = 3-13 for panel E).

In the absence of GABA, picrotoxin (100 μM) induced outward currents only for α2/4β1L285Rδ GABAARs (Fig. 4C, D) with α2/4β1P228Hδ showing similar levels of activity to WT. Given that β1L285R conferred substantial spontaneous activity on δ-GABAARs, compared to β1P228H, the increased tonic currents recorded in MSNs of the NAc and caused by the β1 mutants, are most likely due to the presence of α2/4β1L285Rγ2, α2/4β1P228Hγ2 and α2/4β1L285Rδ isoforms.

Alcohol and GABAARs

Although alcohol may directly modulate some GABAARs, this remains controversial[21, 22, 35, 36]. Here, ethanol (30 - 200 mM) co-applied with low concentrations (EC10) of GABA, did not modulate α2/α4β1γ2 GABAARs containing either WT, or mutant β1 subunits in HEK cells (Supplementary Table S4). Alcohol intoxication is associated with increased production of the endogenous neurosteroid allopregnanolone, an allosteric potentiator of GABAARs[35]. However, WT and β1 mutant GABAARs remained equally sensitive to potentiation by allopregnanolone (1 μM) and benzodiazepines (0.5 μM diazepam), and to inhibition by Zn2+ (100 μM) or bicuculline (50 μM); Supplementary Fig S14 and S15. Similarly, neither β1 mutant when co-expressed with α2δ, or α4δ subunits demonstrated sensitivity to 30 or 100mM ethanol (Fig. 4E). Potentiation of αβδ GABAAR function by neurosteroids (THDOC) was also unaffected by the mutation (Supplementary Fig S13). A previous study in the dorsal striatum clearly demonstrated that acute ethanol decreased the amplitude of the electrically-evoked IPSCs recorded from MSNs of young rats, but with no effect on the paired pulse ratio, suggesting a postsynaptic locus[37]. However, in our recordings from mature mouse NAc MSNs, ethanol (50 mM) had no significant effect on the spontaneous IPSCs (paired recordings n = 5 neurons from 5 mice per genotype; Supplementary Figure S16). Note that although there was an apparent trend for ethanol to increase the frequency of sIPSCs in the Gabrb1+/ MSNs, it was not significant in this sample. In further agreement with the recombinant receptor experiments, ethanol (50 mM) had no effect on the holding current of WT or Gabrb1+/ MSNs (paired recordings n=5 neurons from 5 mice per genotype; Supplementary Fig S16). Thus, it is evident that the L285R and P228H mutations do not affect the co-assembly of β1 with α2/4, δ and γ2 subunits, or their pharmacological properties.

DISCUSSION

ENU mutagenesis[25] of alcohol-averse BALB/cAnN mice generated a stable line carrying a non-synonymous dominant mutation in Gabrb1 (L285R) with a strong phenotypic preference for alcohol. Given that the probability of a second ENU-induced mutation within a 5Mb region is estimated to be p < 0.002 (using both a Maximum Likelihood Estimation method, adopting a Poisson model; and a Markov chain Monte Carlo method, with a Poisson Gamma model)[38, 39], the probability of a second functional mutation within the 1.6Mb genomic region of interest is extremely remote, and would continue to halve with each of the subsequent eight backcross generations that have been studied. This possibility has been further excluded by the absence of other exonic mutations following comparative sequencing across the candidate region. Importantly, the identification of a second, independent line with a spontaneous, non-synonymous single base-pair mutation (P228H) in the same gene that also consistently exhibits ethanol preference and similar phenotypic traits over the 5 backcross generations studied, provides further confirmation that the phenotype is caused by mutating Gabrb1.

Mutant and WT mice showed no differences in preference for saccharin or quinine solutions, discounting taste as a driver for high alcohol consumption by the mutants. That Gabrb1+/ mice worked harder than WT to obtain alcohol compared to sucrose solutions indicates that motivation for alcohol was also not driven by calorific requirement or taste. Importantly, mutants undertook this extra work despite possessing lower body weights than their WT counterparts. The mutant mice also showed greater sensitivity to the sedative/ataxic effects of ethanol, which together with increased consumption, contributed to an increased incidence of intoxication.

Previous studies in man demonstrated significant allelic association between the risk of alcohol dependence and GABRA2 and GABRB1 polymorphisms[4, 5, 40, 41, 42]. Although deleting Gabra2 in mice did not alter ethanol self-administration[19], and human polymorphisms and mouse mutations may lack common effects on receptor function, our studies are supportive of the view that the GABAAR β1 subunit is a modifier of alcohol consumption.

The NAc is associated with reward and plays an important role in addiction. Our recordings from MSNs within the accumbal core revealed that the β1 subunit mutations greatly impact upon both synaptic and tonic inhibitory transmission. For Gabrb1+/L285R and Gabrb1+/P228H, mIPSCs were prolonged and less frequent. Given our previous demonstration that MSNs express synaptic α2-GABAARs[34], the changed mIPSC kinetics suggest that β1 subunits may co-assemble with α2 and γ2 subunits at inhibitory synapses. The GABAAR-mediated tonic conductance was greatly increased in the NAc of Gabrb1+/ and Gabrb1+/ mice. The differential effects of GABAAR antagonists suggest that this perturbation results from spontaneous activity of mutant β1-GABAARs, though increased receptor sensitivity to GABA and, for Gabrb1+/, an increase in ambient GABA, may also contribute.

Our recombinant receptor studies indicate that mutant β1 subunits efficiently co-assemble with α and γ or δ subunits and are functionally expressed at the cell surface, with minimal effect on their pharmacological properties. Nevertheless, both β1P228H and β1L285R mutations increased receptor sensitivity to GABA and caused their channels to open spontaneously in the absence of GABA, particularly for L285R. Modeling the likely subunit combinations of synaptic α2- and extrasynaptic δ-containing GABAARs found in accumbal MSNs, we recreated recombinant equivalents of extrasynaptic (α2/4β1δ) and synaptic (α2/4β1γ2)-type receptors. Mutating the β1 subunit initiated spontaneous GABA channel opening, with β1L285R exhibiting a greater degree of spontaneous activity, implying that both αβ1L285Rγ and αβ1L285Rδ receptors are the most significant contributers to the increased spontaneous (tonic) current in the NAc of the Gabrb1+/ mouse. In comparison, β1P228H GABAARs were less spontaneously-active, but such activity was nevertheless clearly evident for α2β1P228Hγ2.

Recombinant GABAARs (α2/α4β1γ2, α2/α4β1δ) incorporating either WT or mutant β1 subunits, were consistently insensitive to ethanol. Similarly, neither tonic, nor phasic inhibition of WT and β1 mutant MSNs was affected by ethanol. Although low ethanol concentrations (~10-30 mM) have been reported to enhance neuronal and recombinant δ-GABA R[21, 23] function, other studies have failed to corroborate such effects[22, 24, 43, 44, 45].

Thus, increased accumbal tonic current is associated with β1 subunit mutations leading to increased ethanol consumption, but not as a consequence of direct effects of ethanol on GABAARs. Notably, both ethanol preference and consumption are reduced in δ−/− mice[18], a genetic manipulation that reduces accumbal tonic currents[28]. The importance of the tonic current for ethanol reward is emphasized by RNAi knock-down of either α4 or δ subunits in the accumbens, which reduced ethanol consumption and preference without affecting water or sucrose intake[16, 17]. Collectively, this strongly suggests that the large tonic conductance of MSNs contributes to the increased preference and intake of alcohol by Gabrb1+/ and Gabrb1+/ mice.

Our study identifies GABAARs containing the β1 subunit as a key element in modulating alcohol consumption and suggests that spontaneous GABA channel opening and increased tonic inhibition in the accumbens are critically important factors in this debilitating behavioral phenotype that is so costly to individuals and society.

MATERIALS AND METHODS

Mice

Mice were housed under standard conditions with a commercial diet (SDS, UK) and drinking water ad libitum. Using alcohol-averse background strains, 12-week old male BALB/cAnN mice were exposed to ENU[25] and crossed to WT C3H/HeH females. G1 progeny (n=1047) were screened in a two-bottle choice test for preference for 10% (v/v) ethanol at age 7-12 weeks. Mice showing ethanol preference (>2.5 standard deviations above a control non-mutagenized BALB/cAnN × C3H cohort) were backcrossed to 8-12 week old C3H/HeH females to test heritability. Our study was approved by the local ethical review panels of the MRC Mammalian Genetics Unit (MRC Harwell), the University of Dundee and the University of Sussex and complies with the UK Animal (Scientific Procedures) Act 1986.

Phenotyping

Male and female 7-12 week old adult mice were singly housed with free choice of water or ethanol (3% or 10% v/v) during two 10-day test periods. Consumption was determined by weighing the drinking bottles. The amount consumed was corrected for leakage and evaporation and expressed as grams ethanol consumed daily per kg bodyweight of the mice measured at the beginning of each period. Preference was calculated as the ratio of ethanol over total amount of liquid imbibed. Taste preference was determined similarly using 15 and 120mM sucrose (caloric value), 0.25mM and 0.4mM saccharin (no caloric value) and 0.05mM quinine. For self-administration of ethanol studies, animals (n=8 group) were trained using a sucrose-fading procedure to self–administer up to 10% v/v ethanol[19, 20]. Intoxication was scored as: normal, mild ataxia, ambulatory impairment, upon removal of the animals from the operant boxes at the end of daily sessions.

Mapping of ENU Mutation

A genome scan was performed on 13 animals (2 G2, 11 G3) displaying high ethanol preference using 86 microsatellite markers distributed throughout the genome to differentiate DNA of BALB/cAnN or C3H/HeH origin. A further 169 animals were used to identify informative recombinants. Fine mapping with additional microsatellite markers in the candidate region was undertaken to narrow the location of the mutation.

Candidate Gene Sequencing

Exons and the exon/intron borders of twelve candidate genes were sequenced from both directions in mutant and wild-type mice. Oligonucleotides were designed using Ensembl v36 predictions (http://ensembl.org). Data were analyzed with BioEdit software. Primer sequences can be found in Supplementary Table S6.

Electrophysiology and analysis of brain slice preparations

Coronal slices (300 μm) containing the NAc were prepared from male Gabrb1+/, Gabrb1+/ and WT littermates (age 2-5 months). As previously described[34], slices were cut in oxygenated ice-cold maintenance solution containing (mM): 140 K gluconate, 15 Na gluconate, 4 NaCl, 10 HEPES, 0.2 EGTA (pH 7.2; 310-320 mOsm), before storage for at least 1hr at room temperature (20-23°C) in oxygenated, extracellular solution (ECS) containing (mM): 126 NaCl, 2.95 KCl, 26 NaHCO3, 1.25 NaH2PO4, 2 CaCl2, 10 D-glucose and 2 MgCl2 (pH 7.4; 300-310 mOsm). MSNs were identified with an Olympus BX51 microscope equipped with DIC/infrared optics. Whole-cell voltage-clamp recordings (−60 mV) were performed at 35°C, using the ECS containing 1 μM strychnine, 2 mM kynurenic acid and 0.5 μM tetrodotoxin (TTX). Patch electrodes (3-4 MΩ) were filled with (mM): 135 CsCl, 10 HEPES, 10 EGTA, 1 CaCl2, 2 MgCl2, 2 Mg-ATP, 5 QX-314 (pH 7.2-7.3 with CsOH, 300-308mOsm).

Recordings were discarded if series resistance changed >20%. Currents were filtered at 2 kHz (8-pole, low pass Bessel filter) and recorded for offline analysis (DTR1205 recorder). Bicuculline methobromide, gabazine, THIP and strychnine hydrochloride were prepared as aqueous stock solutions for dilution in ECS. Picrotoxin was prepared in ECS. All drugs were obtained from Sigma-Aldrich (Poole, UK) or Tocris Bioscience (Bristol, UK). THIP and ethanol were allowed to infiltrate the slice for at least 10 min before data acquisition. Recordings were digitized (NIDQMX, National Instruments) and sampled at 10kHz before analysis (WinEDR/WinWCP). The mIPSCs were threshold detected (-4 pA, duration 3 ms) and visually inspected. A minimum of 50-100 mIPSCs were used for analysis, including: peak amplitude, rise time (10-90% - ≤ 1ms), and decay times. The decay phase of mIPSCs was best fit (98-10% of the peak amplitude) with a bi-exponential [y(t) = Afast.e(-t/τfast) + Aslow.e(-t/τslow)] function, where t is time, A is the amplitude, and τ is the decay time constant. A weighted decay time constant (τw) was calculated from: τw = τ1P1 + τ2P2 where τ1 and τ2 are decay time constants and P1 and P2 the relative proportions of the decay described by each component. The mIPSC frequency was determined in 20s bins for 2 min, detected by the rate of rise (30-50 pA ms−1), excluding spurious noise.

The mean current and associated RMS (root mean square) were calculated over 102.4 ms epochs for 1 min, using a 10 kHz sampling rate. Epochs containing mIPSCs or unstable baseline were excluded. To ensure changes in the holding current reflected a drug effect, two 1 min control holding current periods (C1 and C2) were sampled and a 1 min section after drug equilibration (D). The mean holding current for C1 and C2 were pooled and the standard deviation calculated. The drug effect was accepted if the absolute change in the holding current (D-C2) was greater than twice the standard deviation of the controls (C1, C2). For electrophysiological analysis of brain slice recordings all data reported represent the mean ± SEM of observations derived from a minimum of 3 animals and each determination is derived from an individual slice.

Site-directed mutagenesis

All point mutations in the GABAAR β subunit were introduced using Quikchange (Stratagene). Sequences of the full coding region of mutated murine cDNAs were determined by automated fluorescent sequencing. Plasmid DNAs for transfection were purified using the Hi-Speed Plasmid Midi Kit (Qiagen).

Cell culture and electrophysiology

HEK293 cells (ATCC, Middlesex, UK) were cultured in Dulbecco’s modified Eagle’s medium with 10%v/v fetal calf serum, 2mM glutamine, 100 units/ml penicillin G and 100 mg/ml streptomycin at 37°C in 95% air/5% CO2. Cells were transiently transfected by calcium phosphate co-precipitation using 1 μg of each plasmid DNA encoding for α2, α4, β1, δ or γ2 and 0.5 μg pEnhanced Green Fluorescent Protein (GFP; Clontech) and used for electrophysiology after 20 hours.

Whole-cell currents were recorded at room temperature (20-22 °C) from single HEK293 cells voltage clamped at −40mV, using an Axopatch 200B amplifier (Molecular Devices, USA). Patch electrodes (3-5 MΩ) contained (mM): 144 KCl, 2 MgCl2, 1 CaCl2, 10 HEPES, 11 EGTA and 2 adenosine triphosphate, pH 7.2. Cells were superfused with Krebs solution containing (mM): 140 NaCl, 4.7 KCl, 1.2 MgCl2, 2.5 CaCl2, 10 HEPES and 11 glucose, pH 7.4. Recordings were filtered at 5 kHz (6-pole Bessel, 36dB/octave). Drugs and Krebs solution were applied using a modified U-tube. GABA, bicuculline, picrotoxin and ethanol were dissolved directly into Krebs solution (adjusted to pH7.4). Allopregnanolone and diazepam were diluted in Krebs from a 10 mM stock in DMSO. For whole cell studies of ethanol potentiation, the GABA current evoked by a low GABA concentration (EC10) was initially determined then 30-200 mM ethanol was applied for 30-60 s prior to the co-application of EC10 GABA and ethanol. Allopregnanolone potentiation was determined by co-application with EC10 GABA and the neurosteroid. Dose-response relationship data were fitted with a non-linear least squares fitting routine using Origin 6.1 (Microcal). Data points represent the mean ± SEM of at least three experiments. The relative proportion of spontaneous GABAA receptor activation was ascertained by dividing the amplitude of currents induced by saturating concentrations of PTX (IPTX) by the summed current amplitudes induced by saturating concentrations of PTX and GABA (IGABA + IPTX). Single channel currents were recorded at room temperature from excised outside-out membrane patches maintained at −70 mV (sampling rate 50 kHz; low pass-filtered at 10 kHz during recording, and at 2-5 kHz during analysis). Patches showing multiple simultaneous channel openings (channel stacking) exceeding 2% of all detected openings were discarded.

Single channel data were analysed using WinEDR (v.3.3.8; Strathclyde electrophysiology software, J Dempster) and QuB (2.0.0.13, Buffalo, NY). Amplitude histograms were created by fitting Gaussian components to the amplitude distributions using a non-linear least-squares routine. Open probabilities (Po) were defined as the ratio between the Gaussian areas of open and shut time components. Individual open and shut time durations were idealised using either WinEDR’s 50% threshold detection method or by QuB’s segmental k-means (SKM) fitting routine. Open and shut dwell-time histograms were fitted with a mixture of exponentials using a Levenberg-Marquardt nonlinear least-squares method from which the areas representing the individual exponential components and their relative time constants were obtained. Mean dwell time durations were calculated from the individual open and shut times weighted by their areas[33,46, 47].

Confocal microscopy and immunocytochemistry

GABAAR myc receptor α2, β1myc or β1myc,L285R or β1myc,P228H, and γ2S or γ2L subunits were expressed with eGFP in HEK293 cells and incubated at room temperature with 9E10 antibody (to the extracellular myc epitope; Santa Cruz Biotechnology) followed by TRITC- or Cy5-conjugated secondary antibody. The myc epitope in the β1 subunit is electrophysiologically silent[48]. Transfected HEK293 cells were fixed in phosphate-buffered saline (PBS) containing 4% paraformaldehyde for 15 min before being quenched with 50mM NH4Cl in PBS for 10 min. After washing in PBS, cells were incubated for 45 min at room temperature with 9E10 antibody. Cells were washed in PBS containing 10% v/v fetal calf serum (FCS) and 0.4% w/v bovine serum albumin (BSA) before incubation for 45 min with a TRITC- or Cy5-conjugated secondary antibody. Cells were washed and then mounted in glycerol before imaging using a Zeiss Axiophot confocal microscope (LSM510 Meta). The detector gain, and amplifier offset and gain were set at the same levels for all FITC, TRITC and Cy5 images of wild-type and mutant receptors to compare expression intensities. The scanning slice depth was set to 2.1 μm.

Statistical analysis

Analyses were performed using SPSS v14 and SigmaStat software (SPSS, USA). Data are presented as mean ± SEM or 95% confidence intervals unless otherwise stated. Student’s t-tests were used for two group comparisons (paired or unpaired as appropriate). Multifactorial linear model analysis was used to investigate the effect of genotype on alcohol preference correcting for generation, gender and weight. Two-way ANOVA, with the between-subject factor genotype and within-subject factor session, was used to evaluate potential genotype and reinforcer type effect and any interactions, and lever pressing in extinction. Categorized indices were compared using a Fisher’s Exact Test. The non-parametric Kolmogorov-Smirnoff (KS) test was used to compare cumulative probability distributions of IPSC parameters. Statistical significance was routinely set at p < 0.05, and at p < 0.01 for the KS test.