Molecular analysis of the site for 2-arachidonylglycerol (2-AG) on the β2 subunit of GABAA receptors


Address correspondence and reprint requests to Erwin Sigel, Bühlstr. 28, Institute of Biochemistry and Molecular Medicine, University of Bern, CH-3012 Bern, Switzerland. E-mail:


2-arachidonyl glycerol (2-AG) allosterically potentiates GABAA receptors via a binding site located in transmembrane segment M4 of the β2 subunit. Two amino acid residues have been described that are essential for this effect. With the aim to further describe this potential drug target, we performed a cysteine scanning of the entire M4 and part of M3. All four residues in M4 affecting the potentiation here and the two already identified residues locate to the same side of the α-helix. This side is exposed to M3, where further residues were identified. From the fact that the important residues span > 18 Å, we conclude that the hydrophobic tail of the bound 2-AG molecule must be near linear and that the site mainly locates to the inner leaflet but stretches far into the membrane. The influence of the structure of the head group of the ligand molecule on the activity of the molecule was also investigated. We present a model of 2-AG docked to the GABAA receptor.

Abbreviations used

1-arachidonyl glycerol


2-arachidonyl glycerol


arachidonic acid














noladin ether


3α, 21-dihydroxy-5α-pregnan-20-one

The major inhibitory neurotransmitter receptors, the GABAA receptors are composed of five subunits surrounding a central chloride ion selective channel (Macdonald and Olsen 1994; Sieghart 1995; Sieghart and Sperk 2002; Sigel and Steinmann 2012). A variety of subunit isoforms of the GABAA receptor has been cloned, leading to a multiplicity of receptor subtypes (Macdonald and Olsen 1994; Barnard et al. 1998; Olsen and Sieghart 2008; Sigel and Steinmann 2012). The major receptor isoform in mammalian brain consists of α1, β2, and γ2 subunits (Olsen and Sieghart 2008). Different approaches have indicated a 2α:2β:1γ subunit stoichiometry for this receptor (Chang et al. 1996; Tretter et al. 1997; Farrar et al. 1999; Baumann et al. 2001, 2002; Baur et al. 2006) with a subunit arrangement γβαβα anti-clockwise as seen from the synaptic cleft (Baumann et al. 2001, 2002; Baur et al. 2006). The pharmacological properties depend both on subunit composition (Sigel et al. 1990) and arrangement (Minier and Sigel 2004).

Recently, direct effects of the endocannabinoid 2-arachidonyl glycerol (2-AG) have been reported on GABAA receptors (Sigel et al. 2011). 2-AG potentiates currents elicited by low concentrations of GABA specifically in β2 subunit-containing receptors. The sequence of β2 was aligned with β1 and β3 to identify residues conferring subunit specificity. Two mutations were found in the inner membrane leaflet of the trans-membrane sequence M4 to strongly affect the potentiation and further two mutations in M3 that affected potentiation to a smaller extent. The binding site for 2-AG on β2 subunit of GABAA receptors offers the potential for pharmacological intervention. To better characterize this binding site, we performed a cysteine scan of M4 and part of M3. We identified five additional residues that affect potentiation by 2-AG. All six residues on M4 locate to the same side of the α-helix M4. The effect of the hydrophilic head group structure on the degree of modulation was also investigated.


Construction of the mutated receptor subunits

The mutant subunits β2M294C, β2L301C, β2V302C, β2Y304C, β2I305C, β2N423C, β2A424C, β2I425C, β2D426C, β2R427C, β2W428C, β2S429C, β2R430C, β2I431C, β2F432C, β2F433C, β2P434C, β2V435C, β2V436C, β2F437C, β2S438C, β2F439C, β2F440C, β2N441C, β2I442C, β2V443C, β2Y444C, β2W445C, β2L446C, and β2Y447C were prepared using the QuikChange™ mutagenesis kit (Stratagene, Agilent Technologies, Basel, Switzerland).

Xenopus surgery

Female Xenopus laevis were kept under a 12 h day/night cycle. The animals were anesthetized by immersion until loss of all reflexes (~ 10–15 min) in pre-chilled water containing 0.2% ethyl 3-aminobenzoate methane sulfate (A5040; Sigma, St. Louis, MO, USA). The female frogs were then laid on wet tissues placed on an ice bed (ventral face up) and kept wet by covering the animal with soaked tissue. The nose of the animal was exposed to air to enable breathing. Through a small abdominal incision (0.5–0.8 cm) lobes of the ovary were pulled out carefully. At least two but not all lobes of the ovary were removed to ensure oocyte regeneration. Follicles were singled out from an ovary lobe using a platinum loop. Follicles were then stored at 18°C in sterile filtered Barth's modified medium containing NaCl (88 mM), KCl (1 mM), NaHCO3 (2.4 mM), HEPES (10 mM, pH 7.5), MgSO4 * 7H2O (0.82 mM), Ca(NO3)2 * 4H2O (0.34 mM), CaCl2 * 2H2O (0.41 mM), and penicillin/streptomycin (100 U/mL).

Expression of GABAA receptors in Xenopus oocytes

Capped cRNAs were synthesized (Ambion, Austin, TX, USA) from the linearized plasmids with a cytomegalovirus promotor (pCMV vectors) containing the different subunits, respectively. A poly-A tail of about 400 residues was added to each transcript using yeast poly-A polymerase (United States Biologicals, Cleveland, OH, USA). The concentration of the cRNA was quantified on a formaldehyde gel using Radiant Red stain (Bio-Rad Laboratories, Reinach, Switzerland) for visualization of the RNA. Known concentrations of RNA ladder (Invitrogen. Life Technologies, Zug, Switzerland) were loaded as standard on the same gel. cRNAs were precipitated in ethanol/isoamylalcohol 19 : 1, the dried pellet dissolved in water and stored at −80°C. cRNA mixtures were prepared from these stock solutions and stored at −80°C. Xenopus laevis oocytes were prepared, injected, and defolliculated as described previously (Sigel 1987; Sigel and Minier 2005). They were injected with 50 nL of the cRNA solution containing wild-type α1 and wild-type or mutated β2 and wild-type γ2 subunits at a concentration of 10 nM : 10 nM : 50 nM (Boileau et al. 2002) and then incubated in modified Barth's solution at +18°C for at least 24 h before the measurements. Where indicated concatenated subunits α12121 or α11122 were used at a concentration of 25 nM : 25 nM, each.

Functional characterization of the GABAA receptors

Currents were measured using a modified two-electrode voltage clamp amplifier Oocyte clamp OC-725 (Warner Instruments, Camden, CT, USA) in combination with a XY-recorder (90% response time 0.1 s) or digitized at 100 Hz using a PowerLab 2/20 (ADInstruments, Spechbach, Germany) using the computer programs Chart (ADInstruments GmbH). Tests with a model oocyte were performed to ensure linearity in the larger current range. The response was linear up to 15 μA.

Electrophysiological experiments were performed using the two-electrode voltage clamp method at a holding potential of −80 mV. The perfusion medium contained 90 mM NaCl, 1 mM KCl, 1 mM MgCl2, 1 mM CaCl2, and 5 mM Na-HEPES (pH 7.4) and was applied by gravity flow 6 mL/min. The perfusion medium was applied through a glass capillary with an inner diameter of 1.35 mm, the mouth of which was placed about 0.4 mm from the surface of the oocyte. Allosteric modulation via the 2-AG site was measured at a GABA concentration eliciting about 1% of the maximal GABA current amplitude (EC1). In each experiment, 1 mM GABA was applied to determine the maximal current amplitude. Subsequently increasing concentrations of GABA were applied until 0.5–1% of the maximal current amplitude was elicited (0.3–3 μM). For modulation experiments, GABA was applied for 20 s alone or in combination with 2-AG. 2-AG was pre-applied for 30 s. Modulation of GABA currents was expressed as (I(modulator + GABA)/IGABA − 1) × 100%. The perfusion system was cleaned between drug applications by washing with dimethylsulfoxide to avoid contamination.

Homology modeling and docking

The homology model of the α1β2γ2 GABAA receptor constructed by M. Ernst (see also Richter et al. 2012) which was based on the crystal structure of the nematode glutamate-gated chloride channel (PDB code 3RIF) was used for molecular docking studies.

The binding pocket was defined as all residues with a maximum distance of 4.5 A to any of the following amino acids: β2M294, β2L301, β2W428, β2S429, β2F432, β2V436, β2F439, and β2V443. Pocket side chains collapsing the binding site were turned into different low energy conformations using the MOE tool Rotamer Explorer. Finally proper protonation states were assigned to the binding site atoms using the MOE tool Protonate 3D.

The structure of 2-AG was built in Molecular Operating Environment (MOE) version 2011.10 (MOE 2011), constraining the double bonds to Z conformation. We then conducted a stochastic conformational search for 2-AG within MOE using default values. The resulting conformational library was then docked with MOE into the previously defined binding site using Triangle Matcher as placement methodology and London dG as scoring function. The 100 best scored poses were refined using the MOE tool LigX energy minimize. For these final 100 poses, the distances between residues β2M294, β2L301, β2V302, β2W428, β2S429, β2F432, β2V436, β2F439, β2V443, and each 2-AG pose were calculated, resulting in a vector of 9 distances for each pose. Finally, a distance cutoff of 6 Å was taken to select poses that were in near proximity to the defined set of residues.


To describe the binding site for 2-AG on the β2 subunit of the GABAA receptor, we mutated 30 residues in the predicted transmembrane segment M4 and in the intracellular portion near M3 and M4 and individually to cysteine. Mutated β2 subunit was co-expressed with wild-type α1 and γ2 subunits in Xenopus oocytes. α1β2F433Cγ2, α1β2P434Cγ2, and α1β2F437Cγ2 failed to express any currents. All other mutations studied here were characterized at a GABA concentration eliciting 0.5–1% of the maximal current amplitude of the corresponding α1β2 mγ2 receptor (0.5–2.0 μM). A cysteine scan was chosen as it allows covalent reaction with numerous reactive molecules to potentially increase the effect seen with the mutation alone. As the mutations alone had pronounced effects, the consequences of this chemical modification were not further studied. The fact that neither expression levels nor the EC50 for GABA was substantially affected by the mutations indicate that assembly is normal in the mutated receptors (not shown).

Modulation by 2-AG: interaction with the hydrophobic tail

Wild-type and mutant α1β2γ2 GABAA receptors were investigated for allosteric potentiation by 5 μM 2-AG. As wild-type receptors varied in the extent of potentiation each day, modulation of wild-type receptors was determined every day (mean of four to six determinations) and modulation of mutant receptors was expressed in percent of this control potentiation. The lowest and the highest potentiation of wild-type receptors amounted to 105 ± 60% (= 6) and 258 ± 21% (= 4), respectively. Current traces shown in Fig. 1a document the potentiation by 2-AG in wild-type receptors. Fig. 1b shows the same experiment in α1β2W428Cγ2 receptors and potentiation is clearly reduced. α1β2W428Cγ2, α1β2S429Cγ2, α1β2F432Cγ2, and α1β2V443Cγ2 receptors, all mutated in M4 of the β2 subunit showed a significantly reduced potentiation by 5 μM 2-AG. Table 1 summarizes our findings at wild-type and mutant receptors. α1β2V436Cγ2 and α1β2F439Cγ2 receptors showed no decrease in the extent of modulation upon the mutation. Mutation of the same amino acid residues to threonine and leucine has previously been shown to drastically affect the modulation (Sigel et al. 2011). The table includes these findings. Figure 2 shows a wheel representation of the α-helix M4. It should be noted that this helix contains a proline residue and is therefore kinked. All of the mutations affecting the modulation are located within a 100° segment of the circle.

Table 1. Effect of point mutations on the standardized current potentiation by 2-AG
ReceptorPotentiation (%) n Statistics
  1. Modulation by 5 μM 2-AG was determined at GABA concentrations eliciting 0.5–1.0% of the maximal current amplitude (EC0.5–1.0). Data are given as mean ± SD. aReported previously. n.e., no expression; n.s., not statistically significant. Data were standardized to the stimulation observed in wild-type receptor analyzed at the same day. Data were analyzed by one-way anova followed by a Dunnett's test. Residues shown in bold face differ with high significance (< 0.01) from wild type receptors.

α 1 β 2 M294Lγ 2 a 36 ± 15 8 p  < 0.001
α 1 β 2 L301Fγ 2 a 45 ± 18 6 p  < 0.001
α1β2L301Cγ2101 ± 444n.s.
α 1 β 2 V302Cγ 2 43 ± 18 5 p  < 0.006
α1β2Y304Cγ2109 ± 174n.s.
α1β2I305Cγ284 ± 94n.s.
α1β2N423Cγ2112 ± 384n.s.
α1β2A424Cγ2116 ± 274n.s.
α1β2I425Cγ2105 ± 354n.s.
α1β2D426Cγ277 ± 84n.s.
α1β2R427Cγ278 ± 108n.s.
α 1 β 2 W428Cγ 2 44 ± 15 9 p  < 0.001
α 1 β 2 S429Cγ 2 49 ± 13 9 p  < 0.001
α1β2R430Cγ2102 ± 435n.s.
α1β2I431Cγ2113 ± 325n.s.
α 1 β 2 F432Cγ 2 63 ± 18 10 p  < 0.001
α1β2V435Cγ2101 ± 178n.s.
α1β2V436Cγ299 ± 277n.s.
α 1 β 2 V436Tγ 2 a 4 ± 16 8 p  < 0.001
α1β2S438Cγ2106 ± 345n.s.
α1β2F439Cγ2107 ± 328n.s.
α 1 β 2 F439Lγ 2 a 31 ± 10 5 p  < 0.001
α1β2F440Cγ298 ± 194n.s.
α1β2N441Cγ2103 ± 264n.s.
α1β2I442Cγ2101 ± 288n.s.
α 1 β 2 V443Cγ 2 49 ± 13 8 p  < 0.001
α1β2Y444Cγ2120 ± 276n.s.
α1β2W445Cγ2134 ± 4611n.s.
α1β2L446Cγ2102 ± 204n.s.
α1β2Y447Cγ2104 ± 195n.s.
Figure 1.

Potentiation of wild-type and mutant receptors by 2-arachidonyl glycerol (2-AG). Receptors were expressed in Xenopus oocytes and currents were measured using electrophysiological techniques at a GABA concentration eliciting 0.5–1.0% of the maximal current amplitude (EC0.5–1.0). Original current traces are shown. (a) Wild-type α1β2γ2 receptors were exposed twice to 1 μM GABA and subsequently to the combination of GABA with 5 μM 2-AG. 2-AG was pre-applied for 30 s. (b) Same experiment at mutant α1β2W428Cγ2 receptors except that 0.5 μM GABA was used. The extent of the current potentiation by 2-AG is strongly reduced in the mutant receptor.

Figure 2.

Wheel representation of the transmembrane α-helix M4. Amino acid residues that upon mutation lead to a decreased potentiation by 2-arachidonyl glycerol (2-AG) are shown in bold if mutated to C or bold/italic if mutated to other residues as shown in Table 1. Cysteine mutant receptors containing the residues shown in weak gray failed to express. Please note that P434 will introduce a kink in M4.

The residues putatively located intracellularly adjacent to M4 failed to affect modulation by 2-AG. As we had previously identified two residues in M3, β2M294L and β2L301F, to affect modulation, we reasoned that the hydrophilic head portion of 2-AG might interact with some amino acid residues of M3 located adjacent to M4. Therefore, we investigated residues β2V302, β2Y304, and β2I305. α1β2V302Cγ2 receptors showed a significantly reduced potentiation by 5 μM 2-AG (Table 1). α1β2L301Cγ2 receptors showed no decrease in the extent of modulation upon the mutation. Mutation of this residue to phenylalanine has previously been shown to decrease the extent of the modulation (Sigel et al. 2011).

A point mutation interfering with the modulation by 2-AG may influence the binding site of the modulator itself or the subsequent conformational changes. We tested all receptors mutated in M4 that show reduced 2-AG potentiation on modulation of 0.5–2 μM GABA responses (about EC1) by 0.2 μM 3α, 21-dihydroxy-5α-pregnan-20-one (THDOC). Modulation by the neurosteroid was in all cases similar to wild-type receptors (Table 2). Thus, modulation by 2-AG and THDOC were determined at similar concentrations of GABA. These results do not entirely rule out effects on conformational changes but makes them less likely.

Table 2. Effect of point mutations on the standardized current potentiation by 0.2 μM THDOC
ReceptorPotentiation (%) n
  1. Data are given as mean ± SEM.

α1β2γ2100 ± 235
α1β2W428Cγ2100 ± 114
α1β2S429Cγ2115 ± 84
α1β2F432Cγ291 ± 124
α1β2V436Tγ292 ± 164
α1β2F439Lγ293 ± 84
α1β2V443Cγ2121 ± 175

DEA is an antagonist of 2-AG

The importance of the hydrophilic portion for the modulation of α1β2γ2 receptors was investigated. In this study, 3 μM each of N-arachidonyl-glycine (NA-glycine), N-arachidonyl-serine (NA-serine), 2-AG, 1-AG, noladin ether (NE), NA-GABA, anandamide (AEA), 2-methyl-2′-F-anandamide (MET-AEA), arachidonic acid (AA), docosatetraenylethanolamide (DEA), and oleamide (Fig. 3a) were tested for current potentiation. NA-glycine and NA-serine were significantly more potent at 3 μM than 2-AG (Fig. 3b). Please note that we do not report maximal potentiation. It was tested if these compounds act as agonists. When applied at 3 μM in the absence of GABA these compounds elicited very small currents amounting to < 0.1% of the maximal current amplitude. Figure 3c compares the concentration response curves of the potentiation by 2-AG and Na-glycine. The curve for 2-AG was fitted with an EC50 of 2.9 ± 0.7 μM (SD, = 3) and a Hill coefficient of 2.0 ± 0.5. The curve for Na-glycine was not fitted, as no saturating responses could be measured. At concentrations of NA-glycine higher than 3 μM current traces were recorded with the typical hallmarks of open channel block, apparent desensitization and off-current. This was not observed for 2-AG.

Figure 3.

Effect of the head group. Current potentiation was determined as described under Fig. 1. (a) Molecular structure of the tested compounds. NA-glycine, N-arachidonyl-glycine; NA-serine, N-arachidonyl-serine; 2-AG, 2-arachidonyl-glycerol; 1-AG, 1-arachidonyl glycerol; NE, noladin ether; NA-GABA, N-arachidonyl-GABA; AEA, anandamide; MET-AEA, 2-methyl-2′-F-anandamide; AA, arachidonic acid; DEA, docosatetraenylethanolamide; oleamide (b) Current potentiation by 3 μM of the compounds shown in (a). Data are shown as mean ± SEM (n = 4). (c) Concentration response curves of the potentiation by 2-AG (circles) and NA-glycine (squares). Data are shown as mean ± SD (n = 3).

For NA-glycine, we investigated if current potentiation occured via the 2-AG site. Observations are complex. In mutant α1β2V436Tγ2 receptors, in which 2-AG potentiation is nearly abolished, current potentiation was also nearly abolished for NA-glycine (not shown). As NA-glycine is very potent and 2-AG relatively weak in stimulating the response by GABA, it would be expected that co-application of NA-glycine with 2-AG results in an intermediate stimulation in case of a similar affinity of 2-AG and NA-glycine. In fact, the two agents in combination (3 μM each) resulted in a stimulation of similar size as NA-glycine alone (not shown). Either the two compounds help to solubilize each other or NA-glycine has a higher affinity than 2-AG or both agents act at different sites. Further experimentation is required to answer this question.

Interestingly, the CB1 receptor agonist DEA antagonized effects by 2-AG. Fig. 4 shows the concentration dependent inhibition of the stimulation by 3 μM 2-AG. From the inhibition curve, a Ki of 1.0 ± 0.3 μM (SD, = 4) was calculated using the Cheng–Prusoff equation (Cheng and Prusoff 1973). The reason for the relatively low slope of the curve is not known. We hypothesized that DEA would bind with different affinity to the two 2-AG binding sites within a receptor pentamer. Therefore, we studied inhibition by DEA of the stimulation by 2-AG in receptor pentamers, only containing one 2-AG site. For this purpose, concatenated receptors, α12121 or α11122, were expressed as β1 does not confer the 2-AG site. Interestingly, DEA inhibited 2-AG effects in these receptors in a similar way to receptors containing two 2-AG sites. This indicates that DEA is able to interact with both 2-AG sites in a similar way. Oleamide was also assayed as potential antagonist. It slightly counteracted stimulation by 3 μM 2-AG, but only at high concentrations (not shown).

Figure 4.

Concentration inhibition curve of docosatetraenylethanolamide (DEA) in α1β2γ2 (circles), and concatenated α12121 (squares) or α11122 receptors (diamonds). Increasing concentrations of DEA were co-applied with 3 μM 2-arachidonyl glycerol (2-AG). Data are shown as mean ± SD (n = 4).

Kinetics of the action of 2-AG

Our results suggested a site of action of 2-AG in the inner leaflet of M4 of the β2 subunit. If 2-AG is added externally to a Xenopus oocyte, it is expected to have to traverse either by diffusion or mediated by a transport system through the membrane. Thus, onset of modulation might be slow. To test this, we exposed an oocyte to GABA followed by GABA+2-AG (Fig. 5). Indeed, onset of modulation was slow and did not reach a steady level within 1 min. Upon switch of the medium to GABA only, a slow decay of modulation was observed. In three of three experiments, a very small current overshoot was observed in this case.

Figure 5.

Time course of the potentiation by 2-arachidonyl glycerol (2-AG). An oocyte expressing α1β2γ2 receptors was sequentially exposed for 20 s to medium alone, 25 s to 1 μM GABA, 60 s to the same concentration of GABA in combination with 5 μM 2-AG, for 65 s to GABA alone and subsequently for 25 s to medium alone. This experiment was repeated two more times with similar results.

As reported previously current potentiation by 2-AG is only observed at low concentrations of GABA (Sigel et al. 2011). A model previously proposed on the basis of other observations (Baumann et al. 2003) predicts that 2-AG promotes isomerization of the singly ligated receptor from the closed to the open state. The model also predicts that sizeable potentiation in this case is limited to small concentrations of GABA (not shown).

Docking of 2-AG

To obtain a better insight into the binding mode of 2-AG at the GABAA receptor, molecular docking studies were performed. As no experimental structure for the GABAA receptor is available we used a homology model of the α1β2γ2 GABAA receptor (Richter et al. 2012; see also Methods).

We docked 2-AG into the presumed binding site at the M3-M4 interface of the β2 subunit using MOE and keeping the 100 best scored poses according to London dG scoring function. We wanted to implement the existing (Sigel et al. 2011) and novel experimental findings from receptor mutagenesis studies in docking pose evaluation, prioritizing poses that were in vicinity to residues that showed reduced 2-AG potentiation after receptor mutagenesis. A set of nine residues was selected showing this reduction in potentiation in experimental studies: β2M294, β2L301, β2V302, β2W428, β2S429, β2F432, β2V436, β2F439, and β2V443. We then calculated the distances between each 2-AG pose and this set of residues. This resulted in a vector of 9 distances per pose. Only one pose was in proximity to all nine residues (distance cutoff = 6 Å). This pose is depicted in Fig. 6. The pose showed pronounced hydrophobic contacts between the fatty acid moiety of 2-AG and the hydrophobic residues β2W428, β2F432, β2V436, β2F439, β2V443, and β2M294, while the polar glycerin head group showed an H-bond interaction with β2S429 (Fig. 6).

Figure 6.

Side-view of the putative 2-arachidonyl glycerol (2-AG) binding mode derived from molecular docking into a GABAA receptor homology model. The ligand 2-AG is illustrated in space-filling mode (blue carbon atoms, red oxygen atoms, white hydrogen atoms). The approximate interface between plasma membrane (PM) and the cytosol (cyto) is indicated by a white dashed line. Residues showing reduced 2-AG potentiation after receptor mutagenesis are (a) depicted in ball and stick representation (yellow carbon atoms) or (b) high-lighted in yellow molecular surface area.

The binding mode found in the β2 subunit could only be partly reproduced in β3. This was seen in in-silico β3 mutations (β2M294L, β2L301F, β2V436T, β2F439L) preserving the binding mode depicted in Fig. 6. The mutations did not allow to retain the pocket shape especially because of altered torsional flexibility. From the physicochemical view, the hydrophilic β2V436T facing the lipophilic tail of 2-AG would lead to additional unfavorable β3 interactions.


The discovery of a direct action of the endocannabinoid 2-AG on a subtype of GABAA receptors offers new possibilities to interfere pharmacologically. A positive allosteric modulator for this site would be expected to potentiate specifically the β2 subunit-containing GABAA receptors. As KO animals for this subunit show hypermotility (Sigel et al. 2011), such a positive allosteric modulator is expeceted to reduce motility and possibly affect sedation and state of anxiety. An antagonist for this site would be expected to counteract 2-AG effects at the GABAA receptor. In this context, it is interesting to characterize the binding site for 2-AG on β2 subunits, that has previously been localized to M4, possibly facing M3. We now performed a cysteine scan of a large part of M4 and observed effects on the degree of allosteric modulation by 2-AG. In principle, a mutation of a residue that induces a change in modulation may reflect direct interaction of 2-AG with this residue, a conformational change ultimately affecting channel opening or an allosteric effect on either of these two processes. In the last two cases, false positives would be identified in a search for the binding site. Residues affecting a conformational change would be expected to be located between the binding site and the channel. The residues identified in the very C-terminal portion of the β2 subunit are unlikely to represent such residues. Also, modulation by neurosteroids was unaffected. Therefore, we think that the corresponding amino acid residues are in direct contact with 2-AG. False negatives would be found if a cysteine replacement of a naturally occurring residue would not interfere with the binding. In fact, we observed three cases, residues β2L301, β2V436, and β2F439, where this is the case.

We found four amino acid residues in M4 of the β2 subunit, β2W428, β2S429, β2F432, and β2V443, in addition to the previously reported residues β2V436 and β2F439. We also identified a further residue in M3, β2V302, in addition to the residues previously described, β2M294 and β2L301. All six residues identified in M4 are located on the same side of the α-helix M4. The residue β2V443 is predicted to be located deep in the membrane, more than 18 Å from the membrane surface. Together, the results imply that 2-AG immerses into the membrane quasi linearly, alike a phospholipid molecule. With this information, we can firmly fix the tail portion of the 2-AG molecule. The fact that three residues on M3 also affect the modulation by 2-AG help to place its binding site between the two transmembrane segments M4 and M3.

In summary, we provide molecular details of the 2-AG site on β2 subunits of GABAA receptors that represents a new potential drug target. It would be attractive to identify specific exogenous ligands for this site. As mentioned above, an allosteric modulator acting at this site would be specific for β2 subunits containing GABAA receptors and an antagonist would block endocannabinoid action on GABAA receptors.


We thank Prof. J. Gertsch for helpful discussions. None of the authors has a conflict of interest. This work was supported by the Swiss National Science Foundation grant 31003A_132806/1 to ES.