In a first set of experiments, we examined the sensitivity of the three most abundant GlyR α subunits (α₁, α₂ and α₃) to different ECs. Most of these molecules have been identified as ECs or putative ECs from brain or spinal cord extracts [21]–[22]. We found that low micromolar concentrations of neutral and acidic ECs modulated currents through homomeric GlyRs expressed in HEK 293 cells elicited at low glycine concentrations (EC₁₀) (Figure 1, Figure S1). Five of the compounds tested contained free carboxyl groups (NA-Gly, N-arachidonoyl glycine; NA-Ser, N-arachidonoyl-L-serine; NALA, N-arachidonoyl-L-alanine; NA-GABA, N-arachidonoyl-GABA; AA, arachidonic acid) and displayed subunit-specificity. For example, NA-Gly induced a significant potentiation of glycine-activated currents through α₁ GlyRs by 101±11% (10 µM, n = 14), whereas currents through α₂ or α₃ GlyRs were inhibited by −56±5% (n = 14) and by −32±3% (n = 14) at 10 µM, respectively. On the other hand, the four hydroxylated, neutral compounds (NOLE, noladin ether; AEA, anandamide; NA-5HT, arachidonyl serotonin; NADA, N-arachidonyl dopamine) showed consistent positive allosteric modulation of all the GlyR isoforms, suggesting that carboxyl groups were required for the inhibition of α₂ and α₃ GlyRs. Interestingly, the basic EC virodhamine (VIR), which possesses an amino group instead of a hydroxyl or carboxyl group, did not significantly alter the α₂ and α₃ GlyR responses and behaved as a very weak α₁ GlyR modulator (Figure 1A–C). These results suggest that hydroxyl groups on ECs are required for the positive modulation of all three GlyR isoforms, whereas carboxyl groups are structural prerequisites for the inhibition of α₂ and α₃ GlyRs. These findings strongly suggest that the existence of specific molecular sites in the different GlyR isoforms which underlie subunit-specific actions of ECs on the GlyR subtypes.

We next aimed at determining the molecular mechanisms and elements underlying the differential allosteric modulation of GlyRs by acidic ECs. To this end, we focused on NA-Gly and examined its actions at different glycine concentrations (Figure S2). Interestingly, the potentiation elicited by NA-Gly on α₁ GlyRs at EC₁₀ of glycine was significantly attenuated at a higher glycine concentration (EC₅₀), whereas the NA-Gly-induced inhibition of α₂ and α₃ GlyRs remained unaltered. These results suggest that the mechanisms and molecular sites involved in potentiation and inhibition are different and non-conserved between subunits. In order to define the molecular sites involved, we next examined the NA-Gly effects using a set of chimeric and mutant GlyRs. To analyze the importance of TM4 and the IL domains, we first studied a pair of chimeric GlyR constructs in which the regions upstream of the IL between α₁ and α₂ were exchanged (Figure 2A). We found that this exchange did not significantly affect the modulation by NA-Gly (Figure 2B–C). Based on these results, we can conclude that IL and TM4 domains do not significantly contribute to the subunit-specific modulation by NA-Gly.

Next, we evaluated the importance of TM regions upstream from the IL. Previous studies have shown that TM regions of GABA_A and GlyRs contain critical residues for several allosteric modulators, such as ethanol, general anesthetics and neurosteroids [11]–[12], [26]–[27]. Furthermore, very recent evidences have been reported that TM residues are critical for the potentiation of recombinant α₁ GlyRs by Δ⁹-tetrahydrocannabinol (THC) and cannabidiol, two phytocannabinoids [28]–[29]. Thus, a rational explanation for the NA-Gly subunit-specific modulation could arise from differences in TM domain composition. An amino acid sequence alignment of the TM regions of α₁ and α₂ GlyR subunits revealed only 3 non-conserved residues: I240, G254 and S296 in α₁ GlyRs or A247, A261 and A303 in α₂ GlyRs (Figure 3A). To investigate the contribution of these residues, we examined the NA-Gly-induced potentiation of triple, double and single mutant α₁ GlyRs in which these amino acids have been substituted with their α₂ GlyR counterparts. Importantly, these mutated receptors as well as all others described in this study responded normally to glycine application (Table 1). Mutation of these non-conserved amino acids (I240V/G254A/S296A) in α₁ GlyR significantly reduced the GlyR current potentiation by NA-Gly to 14±6% (10 µM, n = 7). Subsequent analyses using single and double mutated α₁ GlyRs demonstrated that the G254A mutation reduced the NA-Gly modulation similarly (17±9%, 10 µM, n = 8), indicating that G254 is a critical determinant of α₁ GlyR potentiation (Figure 3B-D). At higher agonist concentrations, no differences were found between wild type and G254A α₁ GlyRs. On the other hand, the reverse A261G mutation in α₂ GlyRs significantly attenuated NA-Gly-induced inhibition at low and high glycine concentrations (Figure 3D), indicating a role for this residue in both positive and negative modulation of α₁ and α₂ GlyRs by NA-Gly. However, the fact that this single mutation did not invert the potentiation into inhibition or vice versa on both GlyRs suggests the existence of other molecular sites that determine the final NA-Gly effects.

Many of the key residues involved in the coupling of agonist binding to channel gating are located in the N-terminal ECD. Several electrophysiological and molecular modeling studies have postulated that residues within the loops 2 and 7 of the ECD (terminology established by Brejc and coworkers) [30] are critical for the events that precede channel gating [27]. In addition, recent studies have proposed that the generation of a pre-open conformation of the ion channel is a key determinant to explain the differences between full and partial agonists on GlyRs [31]. Interestingly, a specific residue within the extracellular loop 2 of the α₁ GlyR (A52) has been implicated in the generation of this pre-open conformation [32]–[33], and furthermore, mutation of this amino acid into its α₂ GlyR counterpart (A52S in human or A52T in rat GlyRs) decreases the sensitivity to the allosteric effects of ethanol [15], [17]. Thus, it is possible that this residue may also contribute to the NA-Gly effects on GlyRs and we directly investigated its importance studying the NA-Gly sensitivity of A52T α₁ GlyRs and the reverse T59A α₂ GlyR (Figure 4A). In α₁ GlyRs, we found that this mutation significantly attenuated positive modulation by NA-Gly, whereas the reverse substitution on α₂ GlyRs did not change NA-Gly-induced inhibition (Figure 4B). At a higher glycine concentration, these mutations did not significantly affect the receptor sensitivity to this putative EC (Figure 4B). Therefore, these results indicate that only the positive allosteric modulation elicited by NA-Gly on α₁ GlyRs required the A52 residue. However, these mutations again did not convert the current potentiation into inhibition or vice versa, suggesting that the extracellular and TM domain elements identified here could jointly determine the NA-Gly sensitivity of both GlyR isoforms.

To explore this idea, we studied receptors in which both TM and loop 2 residues were exchanged (Figure 5A–C). These experiments showed that in α₂ GlyRs, the combined reversal T59A/A261G significantly reduced NA-Gly-induced inhibition to −4±6% (10 µM, n = 10), but still did not change the direction of modulation. Surprisingly, the incorporation of the reversal A303S substitution into this double-mutated GlyR was able to convert NA-Gly from an inhibitor to an allosteric potentiator (44±5%, 10 µM, n = 10) (Figure 5A–C). At a higher glycine concentration, the NA-Gly-induced inhibition on α₂ GlyRs was also significantly attenuated (Figure 5D). On the other hand, 10 µM of NA-Gly did not inhibit α₁ GlyRs containing the reverse mutations (A52T/G254A/S296A) at EC₁₀ of glycine, but showed a partial inhibition at EC₅₀ (Figure 5C). These results suggest that the potentiating effects of NA-Gly on α₁ GlyRs are mainly determined by the residues A52T/G254A/S296A, whereas the negative effect depends on unidentified molecular determinants only present in α₂, but not in α₁ GlyRs.

In order to investigate if these molecular determinants can also reverse the inhibitory actions of NA-Gly on α₃ GlyRs, we first performed amino acid sequence alignments of loop 2 and the critical TM regions. These analyses revealed only one residue, on TM2, which was not conserved between α₁ and α₃ GlyRs (G254 in α₁, A261 in α₂ and A265 in α₃, Figure 6A). Taking into account the findings described in Figure 5, it seems fair to suggest that the introduction of the point-mutation A265G in α₃ GlyRs should be sufficient to convert the NA-Gly-mediated inhibition into a potentiation. Introduction of the A265G mutation in α₃ GlyR significantly reduced the GlyR current inhibition by NA-Gly (-8±4%, 10 µM, n = 6) but did not convert the inhibition into potentiation (Figure 6B-C). At higher agonist concentrations, this mutation also significantly attenuated NA-Gly-induced inhibition (Figure 6C), indicating a common role of this residue in the negative modulation of α₂ and α₃ GlyRs at low and high glycine concentrations (see also Figure 3).

These observations demonstrate that residues in extracellular and TM domains jointly determine the positive modulation of α₁ GlyRs by NA-Gly, and suggest that acidic ECs could share similar mechanisms of action on these receptors. On the other hand, the current inhibition induced by NA-Gly appears to be largely determined by the nature of the 2′ residue in the TM2 of α₂ and α₃ GlyRs (A261 and A265, respectively). However, all these residues appear dispensable for potentiation by hydroxylated ECs (compare Figure 1). Thus, these results indicate that other residues, possibly conserved between the GlyR isoforms, should be critical for the positive allosteric modulation by neutral ECs.

The results presented above suggest that none of the known ECD and TM molecular sites for allosteric modulators are responsible for the positive modulation by hydroxylated ECs on GlyRs. In addition, since AEA effects on GlyRs appear to be different than those TM elements described for ethanol and general anesthetics (S267 in TM2 domain, see ref. [19]), it appears possible that some conserved key residues outside the TM domains are likely to be responsible for the EC effects on GlyRs. Recent studies have explored the role of the large TM3-4 IL for the Cys-loop ion channel function, intracellular regulation and pharmacology. Electrophysiological studies showed that basic residues within the IL of 5-HT₃ receptors are critical determinants of single channel conductance [34], whereas other intracellular basic residues on GlyRs are critical for direct G protein βγ interaction with the ion channel [35]. In GlyRs, the same basic residues are required for the allosteric effects of ethanol [36], demonstrating that intracellular residues, besides TM amino acids, can contribute to allosteric modulation of GlyRs. Amino acid sequence alignment of the IL revealed that the most critical lysine residue for Gβγ modulation of α₁ GlyRs is fully conserved between the three GlyR isoforms (K385 in α₁ GlyRs, Figure 7A). Thus, in order to examine if this amino acid plays a role in the modulation by ECs, we first tested the sensitivity to NA-Gly of K385A mutated GlyRs at EC₁₀ glycine (Figure 7B–C). Our results show that the intracellular K385A mutation significantly attenuated the potentiation by NA-Gly on α₁ GlyRs to 14±4% (5 µM, n = 8), which is similar to those obtained after mutation of extracellular or TM residues (Figure 3–4). In addition, the homologous intracellular mutations (named K385A relative to α₁ GlyR sequence) did not significantly alter the current inhibition induced by 10 µM of NA-Gly on α₂ (−59±4%, n = 8) or α₃ GlyRs (−40±4%, n = 7). The mutation of this residue furthermore did not alter the NA-Gly inhibitory effects at higher glycine concentrations, indicating a selective effect on the positive allosteric effects on α₁ GlyRs (Figure 7C). This selectivity was further confirmed in experiments performed in the triple mutated T59A/A261G/A303S α₂ GlyRs, which are potentiated by NA-Gly (Figure 5). In these receptors, the mutation of the conserved basic lysine residue significantly attenuated the NA-Gly potentiation at EC₁₀ glycine (−1±5%, 10 µM, n = 5, Figure S3) but did not alter the modulation at high glycine concentration (Figure S3).

Because the intracellular lysine residue K385 is conserved between GlyR isoforms and its mutation was able to selectively attenuate the positive allosteric effects of an acidic EC on α₁ GlyRs and on the triple mutated T59A/A261G/A303S α₂ GlyRs, it appears likely that this residue could play a critical role in the potentiation elicited by neutral ECs, which is similar between the three subunits examined (Figure 1). We therefore investigated the effects of two hydroxylated ECs on GlyRs with mutations in the K385 residue (Figure 8A-E). The mutated K385A α₁ GlyR was significantly less potentiated by AEA (14±3%, n = 8, 5 µM). In subsequent experiments we found that the homologous mutations in α₂ and α₃ GlyRs also attenuated AEA-induced potentiation (9±5%, n = 7 and 19±5%, n = 6, 5 µM, respectively), demonstrating a critical role for this amino acid in the positive modulation by AEA in all three GlyR subtypes (Figure 8E). Further analyses showed that the positive modulation by NA-5HT, a synthetic EC analog [24], was also significantly attenuated in K385A-mutated α_1, α₂ and α₃ GlyRs (Figure 8C–E). These data demonstrate that the K385 residue is critical for the positive allosteric modulation of all the GlyR isoforms by both acidic and neutral EC derivatives, but appears to be dispensable for the inhibitory actions of acidic ECs on α₂ and α₃ GlyRs.

HEK 293 cells (CRL-1573; American Type Culture Collection, Manassas, VA, USA) were cultured using standard methods and were transfected using Lipofectamine LTX (Invitrogen, Carlsbad, CA, USA) with 2 µg of DNA for each GlyR and 0.5 µg of EGFP. Expression of EGFP was used as a marker of successfully transfected cells. All recordings were made 18–36 hours after transfection.

Glycine-evoked currents were recorded from transfected HEK 293 cells in the whole-cell voltage-clamp configuration at room temperature (20–24°C) at a holding potential of −60 mV. Patch electrodes were pulled from borosilicate glass and were filled with (in mM): 120 CsCl, 10 EGTA, 10 HEPES (pH 7.4), 4 MgCl2, 0.5 GTP and 2 ATP. The external solution contained (in mM) 150 NaCl, 10 KCl, 2.0 CaCl2, 1.0 MgCl2, 10 HEPES (pH 7.4), and 10 glucose. Recordings were performed with a HEKA EPC-7 amplifier and Patch Master v2.11 software (HEKA Elektronik, Lambrecht-Pfalz, Germany). The amplitude of the glycine current was obtained using a manually applied pulse (3-6 s) of a sub-saturating glycine concentration (EC₁₀ or EC₅₀) for each GlyR subunit, using an outlet tube (200 µm ID) of a custom-designed gravity-fed microperfusion system positioned 50–120 µm of the recorded cell. EC₁₀ or EC₅₀ values for each GlyR studied were obtained experimentally after successive application of 1, 10, 30, 60, 100, 200, 500 and 1000 µM glycine (Table 1). The concentration-response curves parameters (EC₅₀ and Hill coefficients, n_h) were obtained from the curve fits of normalized concentration–response to the equation I_gly  =  I_max (gly)n_h/((gly)n_h + (EC₅₀)n_h). The mean maximal current (I_max) indicated the average maximal current elicited by a concentration of 1 mM glycine. All modulators were first dissolved in ethanol, DMSO, methanol or methyl-acetate and subsequently diluted into the recording solution on the day of the experiment. None of the vehicles produced discernable effects on the glycine-evoked currents. The drugs were co-applied with glycine, without pre-applications. EC effects were tested up to 10 µM in order to avoid significant membrane-associated effects [48]–[49].

Mutations were inserted using the QuickChange™ site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) in cDNA constructs encoding GlyRs in a pCI vector (Promega, Madison, WI, USA) or pcDNA3 (Invitrogen, Carlsbad, CA, USA). All the constructions were confirmed by full sequencing. The procedures involved in the generation of the chimeric GlyRs have been previously published [17]. The GlyR amino acids were numbered according to their position in the mature protein sequence.

All the drugs were obtained from Tocris Bioscience (Bristol, UK) or BioTrend AG (Zurich, CH). N-arachidonoyl-L-alanine (NALA), N-arachidonoyl-L-serine (NA-Ser) and arachidonic acid (AA) were purchased from Cayman Chemical (Ann Arbor, MI, USA).

All values were expressed as mean ± s.e.m of normalized glycine-activated currents. Statistical comparisons were performed using ANOVA. P<0.05 was considered statistically significant. For statistical analysis, at least 6 cells were analyzed. For all the statistical analysis and plots, MicroCal Origin 6.0 (Northampton, MA, USA) software was used.