γ-Hydroxybutyrate and the GABAergic footprint: a metabolomic approach to unpicking the actions of GHB

Authors


Address correspondence and reprint requests to Caroline Rae, Neuroscience Research Australia, Barker St, Randwick, NSW 2031, Australia. E-mail: c.rae@unsw.edu.au

Abstract

J. Neurochem. (2010) 115, 58–67.

Abstract

Gamma-hydroxybutyrate is found both naturally in the brain and self-administered as a drug of abuse. It has been reported to act at endogenous γ-hydroxybutyrate (GHB) receptors and GABA(B) receptors [GABA(B)R], and may also be metabolized to GABA. Here, the metabolic fingerprints of a range of concentrations of GHB were measured in brain cortical tissue slices and compared with those of ligands active at GHB and GABA-R using principal components analysis (PCA) to identify sites of GHB activity. Low concentrations of GHB (1.0 μM) produced fingerprints similar to those of ligands active at GHB receptors and α4-containing GABA(A)R. A total of 10 μM GHB clustered proximate to mainstream GABAergic synapse ligands, such as 1.0 μM baclofen, a GABA(B)R agonist. Higher concentrations of GHB (30 μM) clustered with GABA(C)R agonists and the metabolic responses induced by blockade of the GABA transporter-1 (GAT1). The metabolic responses induced by 60 and 100 μM GHB were mimicked by simultaneous blockade of GAT1 and GAT3, addition of low concentrations of GABA(C)R antagonists, or increasing cytoplasmic GABA concentrations by incubation with the GABA transaminase inhibitor vigabatrin. These data suggest that at concentrations > 30 μM, GHB may be active via metabolism to GABA, which is then acting upon an unidentified GABAergic master switch receptor (possibly a high-affinity extrasynaptic receptor), or GHB may itself be acting directly on an extrasynaptic GABA-R, capable of turning off large numbers of cells. These results offer an explanation for the steep dose–response curve of GHB seen in vivo, and suggest potential target receptors for further investigation.

Abbreviations used:
ACPBPA

(±)-cis-(3-aminocyclopentane) butylphosphinic acid

AMPA

(RS)-α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

GABA(x)R

GABA x receptor

GAT

GABA transporter

GHB

γ-hydroxybutyrate

MCT

monocarboxylate transporter

THCA

trans-gammahydroxycrotonic acid

TPMPA

1,5,6-tetrahydropyridin-4-yl)methylphosphinic acid

The endogenous compound and party drug γ-hydroxybutyrate (GHB) is an intriguing and enigmatic ligand. Although it is present naturally at low concentrations in the brain, and exhibits many of the functions of a neurotransmitter, the role that it plays is unclear. Ingested recreationally it exhibits a steep dose–response curve, producing euphoria and hallucinations, agitation and anxiety, severely distorting judgment, and it may result in a prolonged coma-like state followed by a loss of memory (Tancredi et al. 2003; Zvosec and Smith 2005; Drasbek et al. 2006; Barker et al. 2007).

A deficiency in succinyl semialdehyde dehydrogenase, the enzyme which normally removes succinyl semialdehyde (a precursor of GHB; for a review and further discussion see Maitre 1997) has been associated with a severe but often unrecognized (and under-reported) psychiatric disorder (Pearl et al. 2003; Knerr et al. 2008). It has been, however, claimed that controlled administration of GHB may be beneficial to patients suffering from sleep disorders and/or depression (Mamelak 2009) or as a therapeutic in alcohol use disorders (Addolorato et al. 2009; Caputo et al. 2009). Although GHB has been occasionally applied as an anesthetic, there have been numerous reports of its exploitation as an illegal substance distributed at ‘techno-parties’ or, as publicized by media in the United States and Europe, used in criminal activities ranging from ‘date-rape’ to a ‘perfect murder’.

A receptor for GHB which is insensitive to baclofen has been identified and cloned in both rats (Andriamampandry et al. 2003) and humans (Andriamampandry et al. 2007) with reported Kds in the 100 nM range but there is some controversy as to the exact nature of this putative receptor. It is also clear that many of the actions of GHB are not mediated via this particular mechanism.

At levels above endogenous concentrations, it is generally thought that the actions of GHB occur via GABA(B) receptor [GABA(B)R]-mediated processes, although GHB itself is not a particularly potent agonist at these receptors (Mathivet et al. 1997; Lingenhoehl et al. 1999) and there are differences in the response of nervous tissue to baclofen compared to GHB (Carter et al. 2004, 2009; van Nieuwenhuijzen and McGregor 2009). The sedative actions of GHB are blocked in a large part by antagonists or mimicked by agonists at GABA(B)R (Banerjee and Snead 1995; Hosford et al. 1995; Jensen and Mody 2001; Kaupmann et al. 2003; Schweitzer et al. 2004; Koek et al. 2006; van Nieuwenhuijzen and McGregor 2009) with the effects occurring through both pre- and post-synaptically located GABA(B)R (Jensen and Mody 2001). The orally active GABA(B)R antagonist SCH50911 has been shown to be successful in preventing GHB-induced mortality (Carai et al. 2005). It does remain clear although, that there are other actions of GHB and its metabolites that this mechanism does not explain including effects on dopamine release (Howard and Feigenbaum 1997; Madden and Johnson 1998; Li et al. 2007), modulation by the neutral allosteric modulator at GABA(A)R, flumazenil (Schmidt-Mutter et al. 1998; Koek et al. 2006), or selective effects at the GABA(B)R (Molnar et al. 2009; van Nieuwenhuijzen and McGregor 2009). Administration of GHB has been reported to alter levels of neuroactive steroids (Barbaccia et al. 2005) and, over the long term, to have neurotoxic effects which are mediated via the GHB receptor (Pedraza et al. 2009).

We have developed a Guinea pig cortical tissue slice model which is ideal for examining the mechanism of action of multi-potent (dirty) drugs, such as GHB. In this system, the metabolic fingerprint produced by drug-modulated metabolism of [3-13C]pyruvate is compared against a library of known metabolic fingerprints produced by a range of ligands acting at different GABA-R, GABA transporters (GATs), and by exogenous GABA itself (Nasrallah et al. 2007, 2009, 2010; Rae et al. 2009). In this work, we have investigated the metabolic response of our model system to a range of GHB concentrations and ligands active in the GHB system and compared these with our library of metabolomic responses to ligands active in the GABA system.

Materials and methods

Materials

Guinea pigs (Dunkin-Hartley), weighing 400–800 g, were fed ad libitum on standard Guinea pig/rabbit pellets, with fresh cabbage leaves and lucerne hay roughage. Animals were maintained on a 12 h light/dark cycle. All experiments were conducted in accordance with the guidelines of the National Health and Medical Research Council of Australia and were approved by the institutional (UNSW) Animal Care Ethics Committee.

Sodium [3-13C]pyruvate and sodium [13C]formate were purchased from Cambridge Isotope Laboratories Inc. (Andover, MA, USA). Gammahydroxybutyric acid (4-hydroxybutanoic acid) and (RS)-α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) were obtained from Sigma (St Louis, MO, USA). NCS-382 (6,7,8,9-tetrahydro-5-hydroxy-5H-benzocyclohept-6-ylideneacetic acid), a reported antagonist at the GHB receptor (Godbout et al. 1995; Cook et al. 2002; Castelli et al. 2004), cis-4-aminocrotonic acid an agonist at GABA(C)R, and trans-gammahydroxycrotonic acid (THCA) a GHB analog (Hechler et al. 1990) were purchased from Tocris (Bristol, UK). Other ligands used have been described previously (Nasrallah et al. 2007, 2009, 2010; Rae et al. 2009). (±)-cis-(3-Aminocyclopentane) butylphosphinic acid (ACPBPA), (3-aminopropy1) n-butylphosphinic acid (CGP 36742/SGS 742), and cis-2-(aminomethyl)-1-carboxycyclopropane were synthesized as published (Froestl et al. 1995b; Duke et al. 1998; Hanrahan et al. 2006).

Preparation of brain cortical tissue slices

Following cervical dislocation, guinea pig brains were removed from the cranial vault and 350 μm paraxial cortical slices were obtained using a McIlwain tissue chopper (The Vibratome Company, O’Fallon, MO, USA). The slices were then washed three times in a modified Krebs–Henseleit buffer (Badar-Goffer et al. 1990), re-suspended for 1 h in fresh buffer containing 10 mM unlabeled glucose and gassed with 95% O2/5% CO2 in a shaking water bath, maintained at 37°C, to allow metabolic recovery (McIlwain and Bachelard 1985). Slices were then washed three times in glucose-free buffer and re-suspended in fresh buffer with the substrate of choice.

Modulation of metabolic activity by GHB ligands

To determine the metabolic effects of modulation of metabolism by GHB and related ligands, slices were incubated for 1 h with 2 mM sodium [3-13C]pyruvate (control) and, in the case of ligand treatment groups, a concentration of the ligand.

To determine the effects of GHB, a range of concentrations were employed: 1.0, 10.0, 30.0, 60.0, and 100 μM GHB. We also used 20 and 200 μM NCS-382, a specific antagonist at the GHB receptor (Maitre et al. 1990; Castelli et al. 2004) and 50 nM and 5 μM THCA, a GHB analog (Bourguignon et al. 1988).

An additional experiment was performed where the slices were activated using the glutamate agonist AMPA (5 μM) and the effect of 60 μM GHB on this activity was also tested. The number of samples was N = 4 in all cases.

Preparation of samples and NMR analysis

On completion of the incubation period, slices were removed from the incubation buffer by rapid filtration and extracted in methanol/chloroform according to the method of Le Belle et al. (2002). Extracts were lyophilized and the pellet was retained for protein estimation by the Lowry technique. Lyophilized supernatants were stored at −20°C until required for NMR analysis. Samples were re-suspended in 0.65 mL D2O containing 2 mM sodium [13C]formate as an internal intensity and chemical shift reference (13C δ 171.8). Fully relaxed 1H and 1H[13C-decoupled] spectra (total cycle of 30 s, comprising 24 s relaxation delay, 4 s water suppression, and ∼2 s acquisition time), WURST-40 (Kupce and Freeman 1995) with a 112-step phase cycle (Skinner and Bendall 1997; decoupling during acquisition) were obtained at 600.13 MHz on a Bruker DRX-600 spectrometer (Bruker Biospin, Karlsruhe, Germany) with a 5-mm dual 1H/13C probe, followed by 13C [1H-decoupled] spectra (typically 3000–5000 transients, cycle of 4 s comprising 2 s relaxation delay, and ∼2 s acquisition time, continuous WALTZ-16 decoupling, 131 072 data points). Assignments were made as described previously (Rae et al. 2000).

13C [1H-decoupled] spectra were Fourier transformed using 3 Hz exponential line-broadening and peak areas were determined by integration using standard Bruker software (TOPSPIN, Version 1.3 or 2.5, Bruker Biospin) following baseline correction. Peak areas were adjusted for nuclear Overhauser effect, saturation and natural abundance effects, and quantified by reference to the area of the internal standard resonance of [13C]formate. Glu C3 was not quantified because of possible resonance overlap with pyruvate signal. Metabolite pool sizes (lactate, alanine, GABA, glutamate, glutamine, and aspartate) were determined by integration of resonances in fully relaxed 600 MHz 1H[13C-decoupled] spectra using [13C]formate as the internal intensity reference.

Experimental data (N = 4) are given as means (SD). Statistical analysis was done using anova for comparing ligand-treated metabolism at each receptor with control [N = 4, followed, only where statistical significance was indicated by Scheffe F-test, by a non-parametric (Mann–Whitney U) test (Statview Student)]. The statistical analysis was performed on the raw experimental values, not the relative differences as shown in the graph. Significance was assumed at α = 0.05. Data are shown graphically as change in each variable relative to the mean of that variable in the control experiment in order more clearly to show the metabolic ‘pattern’ generated by the ligand relative to control.

Pattern recognition of the data

Multivariate statistical analysis used the Simca P+ software package (v11.5; Umetrics, Umeå, Sweden). Each dataset for a particular manipulation was imported as the relative change from the average value obtained from the control group for that particular experiment. Data were univariance scaled to standardize variance between the high and low concentration metabolites (Wold et al. 1998), to ensure that the 13C labeling and steady-state pool size concentrations equally contributed to the model.

Here, we imported the data from experiments using GHB, NCS-382, and THCA into a previously generated model of the GABAergic system, a so-called GABAergic ‘footprint’. This footprint was generated by importing datasets acquired under identical conditions using a range of ligands at GABA(A)R, GABA(B)R and GABA(C)R, inhibitors of GABA uptake (GAT inhibitors), and exogenous GABA into SIMCA P+ (Nasrallah et al. 2010). This footprint contained six distinct groupings of metabolic profiles which aligned to particular concentrations of GABA and distinct locales. The GHB-ligand dataset from this work was compared against the background of this footprint using PCA to see which GABAergic subclass was related to which concentration of GHB. The robustness of the resultant model is assessed by a goodness-of-fit algorithm, with R2 > 0.60 representing a model which accounts for the majority of variance in the dataset (Eriksson et al. 1999); here, R2 > 0.75. Only models where the goodness-of-fit (Q2) was greater than 40% were deemed to be statistically robust. All models used in this work had Q2 > 60%.

Rationale for use of [3-13C]pyruvate as substrate

It is well known that glucose is the obligatory brain substrate. Here, we seek to maximize incorporation of label into as many metabolic compartments as possible, to give us the best possible metabolic resolution. [3-13C]Pyruvate at 2 mM is taken up into both neurons and astroctyes [monocarboxylate transporter (MCT) 1 KM = 1.0 mM, MCT2 KM = 0.08 mM; Broer et al. 1999] where it has been shown to support respiration, maintain adequate bioenergetic levels and membrane potential (Woodman and McIlwain 1961), support synaptic metabolism (Tarasenko et al. 2006), as well as brain function in vivo (Gonzalez et al. 2005). Studies have shown that pyruvate is used in multiple brain compartments (e.g., Cruz et al. 2001; Zwingmann et al. 2001) and our own data have shown that the use of [3-13C]pyruvate yields more resolution in the resultant metabolic pattern than does [1-13C]glucose as well as being capable of supporting large metabolic demands (Rae et al. 2006). As noted previously (Cruz et al. 2001; Zwingmann et al. 2001), pyruvate is located in a vital branch point between glycolysis and the Krebs cycle. Key enzymes regulating its concentration have high flux control coefficients ensuring its rapid metabolism and hence incorporation of label into a range of compartments throughout the slice.

Results

The metabolic profiles of the effects of each concentration of GHB on metabolism are shown in Fig. 1. The figure shows the change in concentration for each variable relative to the control average for that variable. The statistically significant changes shown were calculated using the raw (rather than the change relative to control) data. The different concentrations of GHB all produce distinct fingerprints reflecting a range of different metabolic activities, with 60 and 100 μM GHB being most similar (Fig. 1).

Figure 1.

 Relative effect of different concentrations of exogenous γ-hydroxybutyrate (GHB) on net flux of 13C and on total metabolite pool sizes in brain cortical tissue slices incubated 1 h with sodium [3-13C]pyruvate. Data are shown as relative to the control mean, with control metabolism centered at zero. Error bars represent standard deviations. Statistically significant changes (calculated on the raw data not the relative change in flux or pool size) are indicated by * (p < 0.05 (different to control).

The metabolic profiles generated by 20 and 200 μM NCS-382, the GHB receptor antagonist, and 50 nM and 5.0 μM THCA, a GHB analogue, are shown in Fig. 2.

Figure 2.

 Relative effect of ligands active at the γ-hydroxybutyrate (GHB) receptor on net flux of 13C and on total metabolite pool sizes in brain cortical tissue slices incubated 1 h with sodium [3-13C]pyruvate. Data are shown as relative to the control mean, with control metabolism centered at zero. Error bars represent standard deviations. Statistically significant changes (calculated on the raw data not the relative change in flux or pool size) are indicated by * (p < 0.05, different to control) or # (p < 0.05, different to the other concentration of ligand used).

Principal components analysis

Data from the experiments with exogenous GHB (Fig. 1) and GHB receptor ligands (Fig. 2) were combined with data obtained previously using a range of concentrations of exogenous GABA (Nasrallah et al. 2009) and the data were subjected to PCA using Simca P+. The data formed a four component model accounting for 90% of the variance in the data (49%, 22%, 10%, and 9%, respectively; cumulative Q= 69%). The first two components of this model are shown in Fig. 3. The data show good separation for the different concentrations of GHB, with the two higher concentrations (30 and 60 μM) clustering together with a high positive loading on PC1. GHB (1.0 μM) clusters closely with 1.0 μM GABA and the GHB ligands 5 μM THCA and 20 μM NCS-782. GHB (10 μM) also clusters near 50 nM THCA but the higher (30, 60, and 100 μM) concentrations of GHB are not located near any other ligands in the model.

Figure 3.

 Principal components analysis of γ-hydroxybutyrate (GHB) and GHB ligand data with exogenous GABA. These data generated a four component model accounting for 89% of the variance in the data (49%, 22%, 10%, and 9% respectively; Q2 = 0.69), the two major components of which are shown in this diagram. GHB data are shown as red circles, GABA as blue diamonds, and GHB ligands as black squares. The large outer ellipse represents the 95% confidence interval (Hotelling score).

We therefore combined the data shown in Fig. 3 with library data obtained using ligands active at GABA(A), (B), and (C) receptors as well as at GATs (Nasrallah et al. 2007, 2009, 2010; Rae et al. 2009) and reconstructed the PCA model. This generated a three component model accounting for 79% of the variance in the data (46%, 24%, and 9%, respectively; Q= 0.63), which is shown in Fig. 4. Using this model, we can clearly identify ligands which produce similar metabolic responses to each of the concentrations of exogenous GHB.

Figure 4.

 Principal components analysis (PCA) of γ-hydroxybutyrate (GHB) and GHB ligand data shown against metabolic fingerprint data from selected ligands active at GABA(A), GABA(B), and GABA(C) receptors or GABA transporter (GAT) inhibitors and exogenous GABA. These data generated a three component model accounting for 79% of the variance in the data (46%, 24%, and 9%, respectively), the major two components of which are shown in this diagram. GHB data are shown as red circles, NCS-382 as black squares, trans-gammahydroxycrotonic acid (THCA) as black diamonds, SCH 50911 as blue diamonds, and Baclofen as green stars. All other data are represented as gray circles. The large outer ellipse represents the 95% confidence interval (Hotelling score). Insets to the figure show enlargements of the area of the PCA plot immediately surrounding selected concentrations of GHB.

GHB clusters (1.0 μM) with: 1.0 μM GABA and 1.0 nM RO19-4603 (negative allosteric modulator of GABA(A)R receptors with affinity for diazepam-insensitive GABA(A)R; Wong and Skolnick 1992). Other ligands in the vicinity include 5 μM THCA, 20 μM NCS-382, and 0.2 μM SKF 97541 (potent GABA(B)R agonist; Froestl et al. 1995a).

GHB (10 μM) clusters with Class 1 ligands in our GABAergic footprint (Nasrallah et al. 2010). These are characteristic mainstream GABAergic synaptic receptor ligands, and include 1 μM Baclofen (classic GABA(B)R agonist; Davies and Watkins 1974), the GAT1 inhibitors 40 μM NNC-711 (Suzdak et al. 1992), 10 μM guvacine (Krogsgaardlarsen et al. 1978), and 10 μM muscimol (GABA(A)R agonist; Chebib and Johnston 2000). The GHB analogue THCA (50 nM) is located in the same sector but is loaded more negatively on PC2 than 10 μM GHB (Fig. 4).

GHB (30 μM) clusters with Class 5 ligands in our GABAergic footprint. These are mostly ligands active at GABA(C)R, including cis-4-aminocrotonic acid (100 μM) a partial agonist at GABA(C)R (Chebib and Johnston 2000), also active at α6-containing GABA(A)R (Wall 2001) although the expression of the α6 subunit has been reported to be confined to the cerebellum (Kato 1990; Wisden et al. 1992), cis-2-(aminomethyl)-1-carboxycyclopropane, the most selective GABA(C)R agonist discovered to date (Duke et al. 1998), high concentrations of the GABA(C)R selective antagonists 1,5,6-tetrahydropyridin-4-yl)methylphosphinic acid (TPMPA, 100 μM (Ragozzino et al. 1996) at this concentration also active at GABA(A)R) and ACPBPA (40 μM; Chebib et al. 2007). Also found in the vicinity is the orally active GABA(B)R agonist SCH50911 (50 μM; Ong et al. 1998).

GHB (60 and 100 μM) give very similar metabolic profiles (Fig. 1) and cluster closely together just outside the Hotelling circle. The three nearest ligands are all antagonists at GABA(C)R; 4 μM ACPBPA, 10 μM TPMPA, and 100 μM SGS742, which is both an antagonist at GABA(B)R and, at this concentration, at GABA(C)R (Chebib et al. 1997). With an almost identical loading on PC1, although a much different loading on PC2 is the combination of GAT1 and GAT3 inhibitors NNC-711 (0.4 μM) and SNAP-5114 (5 μM). Together these compounds have been shown to inhibit tonic activity (Keros and Hablitz 2005), a result was not found with either inhibitor alone.

Figure 5 shows the effect of 60 μM GHB when added to slices already activated by 5 μM AMPA. The effect of 5.0 μM AMPA on slice metabolism is shown in Fig. 5a with increased net flux into all measured Krebs cycle intermediates. Addition of 60 μM GHB produces a significant decrease in net flux and a decrease in pool sizes. This indicates that the context in which GHB is acting (i.e. what activity is exhibited by the circuitry being inhibited by GHB) is important.

Figure 5.

 Relative effect of activation of slices by 5 μM (RS)-α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and 5 μM AMPA and 60 μM γ-hydroxybutyrate (GHB) on net flux of 13C and on total metabolite pool sizes in brain cortical tissue slices incubated 1 h with sodium [3-13C]pyruvate. Data are shown as relative to the control mean, with control metabolism centered at zero. Error bars represent standard deviations. Statistically significant changes (calculated on the raw data not the relative change in flux or pool size) are indicated by * (p < 0.05, different to control) or # (p < 0.05, different to the other ligand experiment).

Discussion

Actions of lower concentrations of GHB – GHB(R) and α4-GABA(A)R?

We have shown here that different concentrations of GHB produce quite distinct metabolic outcomes. GHB is present naturally in the brain at very low concentrations. Studies using radiolabeled GHB have detected labeled GABA, glutamate, and glycine within two and a half hours of administration of labeled GHB (Gobaille et al. 1999). GHB has recently been reported to enter cells via the monocarboxylate transporter MCT1 as well as via the sodium-dependent lactate transporter (Morris and Felmlee 2008; Cui and Morris 2009). Here, resonances derived from GHB were not detected in the [13C-decoupled]1H NMR spectra and there was no obvious relationship between added GHB and levels of GABA (Fig. 1). This agrees with a recent report suggesting that increased GHB levels result in lowered concentrations of GABA and glutamate (Felmlee et al. 2010) as might be expected given the general decrease in metabolism seen here. Labeling of GABA C2, arguably a better indicator of GABAergic activity than GABA levels (Rae et al. 2009) is initially decreased compared with controls, but increased once GHB levels exceeded 10 μM. These effects are in line with previous reports where low doses of GHB increased cell firing rates in the frontal cortex (Godbout et al. 1995) and decreased cell content levels of GABA (Gobaille et al. 1999). These low-dose effects are reportedly blocked by NCS-382 (Godbout et al. 1995).

The metabolic fingerprint generated by 1.0 μM GHB exhibits increased net flux into Gln C4; indeed all concentrations of GHB result in increased net flux into Gln C4. This is a pattern mostly associated with ligands active at GABA(C)R but also with a subset of ligands which are allosteric modulators of GABA(A)R, including RO19-4603 (Nasrallah et al. 2009), which clusters closely to 1.0 μM GHB (Fig. 4).

The other ligand clustering with 1.0 μM GHB (Fig. 4) is the δ-subunit preferring analogue 4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridine-3-ol hydrochloride (Brown et al. 2002). These results suggest that GHB at 1.0 μM acts at the GHB receptor and/or α4-containing GABA(A)R which may also include δ-subunit containing receptors, such as α4β3δ GABA(A)R. In mice lacking the δ-subunit, GHB fails to induce the synchronous spike and wave discharges which it induces in wild-type mice (Belelli et al. 2009).

The putative actions of GHB at the α4-containing GABA(A)R are of interest. There is some evidence in the literature suggesting that the synergistic effects of GHB with alcohol may be owing to the activity at this receptor (Cook et al. 2006; Caputo et al. 2009).

The analogue THCA has been reported to have both low (Kd = 2 μM) and high (Kd = 7 nM) affinity binding sites in rat brain (Hechler et al. 1990) where it is a naturally occurring compound (Vayer et al. 1985). It is reported to have little effect at GABA(B)R but to increase glutamate release (Castelli et al. 2003). The effect of 50 nM THCA produced a unique fingerprint which did not cluster near to any other ligand we have studied to date (Figs 2 and 4). Higher concentrations of THCA (5 μM) clustered closely to 1.0 μM GHB.

Actions of GHB at GABA(B)R

GHB at slightly higher concentrations (10 μM) produces a metabolic response which is a typical response at mainstream GABA synapses (including α1βγ GABA(A)R, Class 1; Nasrallah et al. 2010). In Fig. 4, 10 μM GHB clusters with patterns from the GAT1 inhibitors guvacine (10 μM) and NNC-711 (40 μM), the classic GABA(A)R agonist muscimol (10 μM) and the classic GABA(B)R agonist baclofen (1 μM). The most likely interpretation of these results, given the known effects of GHB at GABA(B)R (Banerjee and Snead 1995; Mathivet et al. 1997; Lingenhoehl et al. 1999; Schweitzer et al. 2004; Molnar et al. 2009) and the clustering of 10 μM GHB with 1 μM baclofen, is that this concentration of GHB is acting largely at GABA(B)R.

Actions of 30 μM GHB – transition to ionotropic GABA-R?

The bulk metabolic response of cortical brain tissue slices to 30 μM GHB is again significantly different to that of 10 μM GHB. The cluster of data points is found located in Class 5 territory on the GABAergic footprint (Fig. 4) with some overlap with Class 6. Class 5 is dominated by response to ligands active at GABA(C)R, whereas Class 6 is characterized by ligands which block GAT1 but which induce a response with a higher PC1 loading than Class 1 ligands (Nasrallah et al. 2010). The response also overlaps with that to the GABA(B)R antagonist SCH50911 which is known to exhibit resuscitative effects on mice dosed with GHB (Carai et al. 2005).

Higher (60 and 100 μM) GHB

To understand the metabolic response to 60 and 100 μM GHB, first one must appreciate the steady-state activity of the brain tissue slice. In slices, ‘housekeeping’ metabolic activity is lower than that in anesthetized brain (Griffin 1997) which is understandable as the circuitry has no outside efferent or afferent activity. Around 50% of slice activity is because of spontaneous NMDAR, as activity is reduced by ∼50% by the NMDAR blocker MK-801 (Rae et al. 2006). Consequently, inhibitory activity in slices can induce a number of responses depending on the context; i.e. depending on whether the circuits involved are actively driving other circuits or not (Tagamets and Horwitz 2001; Nasrallah et al. 2007). In intact, awake brain, however, these circuits are most generally actively engaged, but in the slice this is not necessarily the case. The metabolic activity induced by the ligands seen at the far right of the PCA plot in Fig. 4 is illustrative of this point. The combination of GAT1 and GAT3 inhibitors (NNC-711 and SNAP-5114) is known to increase tonic inhibition (Keros and Hablitz 2005), presumably through synergistically increasing the amount of GABA in the extrasynaptic space. It is not known which receptors are involved in this, although β3-GABA(A)R (Hentschke et al. 2009), δ-GABA(A)R (Belelli et al. 2009), α4-GABA(A)R (Chandra et al. 2006), and volume transmitting neurogliaform cells (Olah et al. 2009) have all been shown to influence tonic inhibition (Glykys and Mody 2007).

However, as illustrated in Fig. 5, if the slices are activated, in this instance by the addition of the glutamatergic agonist AMPA, the addition of GHB (60 μM) gives a completely different, and probably more realistic metabolic outcome, which is to significantly decrease net flux into Krebs cycle intermediates and significantly decrease metabolic pool sizes. This latter effect suggests that there is a significant volume effect; generally large metabolic pool depletion is associated with widespread deactivation of metabolism (e.g. Cox et al. 1983; Moussa et al. 2007; Rae et al. 2005, 2006, 2009).

The metabolic response of cortical tissue slices in this work to 60 and 100 μM GHB is similar to that caused by blocking GAT1/GAT3 simultaneously with similar loadings on PC2 (Fig. 4). Succinic semialdehyde dehydrogenase deficient mice [which have elevated GHB (40×) and GABA (3×)] have been shown to have elevated extracellular GABA and increased tonic inhibition (Drasbek et al. 2008). These mice have been reported to have no alteration in the receptor density of the GHB(R), as determined by binding of NCS-382 (Ticku and Mehta 2008). It is possible that GHB increases tonic inhibition by increasing extrasynaptic GABA levels. Felmlee et al. (2010) have recently suggested that extrasynaptic metabolism is unlikely to occur because of the lack of correlation between GABA levels and symptoms in rats administered GHB. However, a very small amount of GABA, released in the vicinity of an appropriate receptor can have a striking and disproportionate effect to the ‘total’ GABA level, by acting at an extrasynaptic receptor.

GABA(C)R are known to include high affinity, extrasynaptic receptor populations (Chebib 2004), and the presence of the metabolic profiles for the GABA(C)R antagonists TPMPA, ACPBPA, and the mixed GABA(B)R/GABA(C)R antagonist SGS742 in the vicinity (Fig. 4) suggests that the ρ-subunit may be involved in mediating the effects of this concentration of GHB. It may well be that extracellular GHB is converted to GABA in the vicinity of these receptors, resulting in activation; there is no evidence that GHB is active at GABA(C)R although the compound has not been rigorously tested against all GABA-R.

In summary, this study demonstrates that GHB produces highly distinct effects depending on its concentration. At the lowest applied concentrations, GHB had little or no effect on GABAergic activity and the pattern of metabolic changes suggested activation of a specific GHB receptor and/or a subtype of GABA(A)R particularly sensitive to GHB. Increasing the concentrations first activated GABA(B)R and, when the concentration was further increased, additional GABA(A)/GABA(C)R putatively became involved. The metabolic response to the highest concentrations of GHB, particularly when it was applied to artificially activated slices, indicated a broad inhibitory effect of GHB, involving a range of synaptic and extrasynaptic GABA-R. These findings may thus parallel the observations in vivo and offer a mechanistic explanation (based on metabolic correlates of neuronal activity) of the steep course of the GHB dose–response curve.

Acknowledgements

The authors are grateful to Dr Ken Mewett for synthesis of SGS742 and ACPBPA. The staff of the UNSW Mark Wainwright Analytical Centre, Dr Jim Hook, and Dr Adelle Amoore are thanked for expert technical support. This work was supported by the Australian National Health and Medical Research Council (Grant to C.R. and V.J.B. and Fellowship to C.R.).

Ancillary