Loop A

Loop A (Figures 1A and D and 11) was investigated in α₁β₂ receptors expressed in oocytes by Boileau et al. (Reference Boileau, Newell and Czajkowski2002), who performed a substituted-cysteine accessibility method (SCAM) analysis for residues β₂W92–β₂D101. The β₂W92C (forming a hydrophobic core) mutant was not functional, while the greatest effect on EC₅₀ was observed for β₂Y97C, which along with β₂L99C (Figure 11A and B) was protected by GABA and SR-95531 from the methane thiosulfonate (MTS) reagent. Interestingly, mutation β₂L99C resulted in spontaneous activity of β₂ homomers, indicating the role of this residue in receptor gating. The general pattern of modification suggested that β₂W92–β₂D101 formed a β-strand conformation. Further work by Padgett et al. (Reference Padgett, Hanek, Lester, Dougherty and Lummis2007) examined the interactions between the ‘aromatic box’ and ligands, providing evidence of cation–π interactions with β₂Y97 (Figure 11A and B) in α₁β₂ receptors. However, docking of the GABA molecule in models of the α₁β₂ and ρ type GABA_AR indicated that in the latter receptor type, cation–π interactions were present not in loop A but in loop B (Lummis et al. Reference Lummis, Beene, Harrison, Lester and Dougherty2005). Tran et al. (Reference Tran, Laha and Wagner2011) examined a ‘tight coupling’ between aromatic β₂Y97 (and also β₂F200, loop C, Figure 11A and B) and several arginine residues (α₁R67 loop D, α₁R120 β5 strand, α₁R132 loop E, and β₂R207 β10 strand below loop C, Figure 11) which were previously implicated in ligand binding. By using mutant cycle analysis (MCA) on α₁β₂γ₂ GABA_ARs expressed in HEK293 cells, Tran et al. (Reference Tran, Laha and Wagner2011) found that a significant functional coupling took place between β₂Y97 at loop A and the following residues: β₂F200, β₂R207, and α₁R132. They proposed that the tight coupling between β₂Y97 and β₂F200 plays a critical role in mediating GABA binding. In another study, the β4-β5 strand linker (below loop A) was investigated, which is thought to connect the neurotransmitter binding site to the inner β-sheet (Venkatachalan and Czajkowski Reference Venkatachalan and Czajkowski2012). They introduced multiple glycine residues into this region of the β subunit (α₁β₂γ₂ receptor expressed in oocytes) to alter its length and flexibility. Their findings demonstrated that the more glycine residues were incorporated, the less GABA-induced activation was observed, and the competitive antagonist SR-95531 began to act as a partial agonist. This effect was not observed in the α or γ subunits, but mutations in these subunits altered the actions of pentobarbital and flurazepam, suggesting a role in receptor gating. Laha and Tran (Reference Laha and Tran2013) further examined the impact of the tyrosine β₂Y97 mutation to alanine (together with β₂Y157 and β₂Y205, Figure 11A and B) on the expression and kinetics of α₁β₂γ₂ and α₁β₂ receptors in HEK293 cells. Interestingly, the assembly or trafficking of GABA_ARs was impaired when tyrosine mutants were expressed as αβ receptors but not in αβγ assembly. Mutation of tyrosine β₂Y97 resulted in accelerated deactivation and slower binding, providing evidence for the weakening of GABA-binding reaction. However, mutations at β₂Y157 and β₂Y205 were considerably more detrimental than that at β₂Y97, suggesting that interactions involving multiple tyrosine residues are likely to occur during the binding process. Structural studies have confirmed the role of loop A in forming of the GABA-binding site. For instance, Masiulis et al. (Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019) demonstrated that β₃Y97 (in the α₁β₃γ₂ receptor) forms a hydrogen bond, rather than a cation–π interaction, with the GABA amino group. Collectively, these studies highlight the importance of loop A and its key residue β₂Y97 in agonist binding, although the specific contributions of this residue vary between studies (compare Padgett et al. (Reference Padgett, Hanek, Lester, Dougherty and Lummis2007) and Laha and Tran (Reference Laha and Tran2013)). The involvement of β₂Y97 in receptor gating remains to be fully elucidated.

Figure 11. ECD-binding site of GABA_AR. (A) General view of the orthosteric binding site. The binding cavity is lined with antiparallel β-strands of the principal (4, 7, 10, and 9, loops A and B) and complementary (5, 6, 2, 1, and 8, loops E, D, G, and F) subunits and capped by loop C (between β-strands 9 and 10). (B) In the principal subunit side loops A, B, and C are present. Loops A and B are located deep in the binding cavity, whereas loop C between strand β9 and β10 is exposed to the solvent. Residues β₂Y97, β₂Y157, and β₂F200 are forming an ‘aromatic box’ around GABA molecule, whereas β₂E155 form electrostatic interaction with ligand molecule. (C) All loops (E, D, G, and F) of the complementary side of the cavity are fragments of the respective β-strands. Strand β5 and loops E and D are located similarly to loops A and B in the deeper part of the binding site, whereas loops G and F are partially exposed to the extracellular space. The ‘aromatic box’ on the complementary side is formed with α₁F65 and α₁F45 whereas electrostatic interaction is present between agonist and α₁R67.

Loop B

Loop B is located in the outer β-sheet and parallel to loop A (Figures 1A and D and Figure 11). Amin and Weiss (Reference Amin and Weiss1993) examined mutations at key positions within loop B, including β₂Y157F/N and β₂T160S/A, in α₁β₂γ₂ GABA_ARs expressed in oocytes (Figure 11A and B). Their results indicated that these mutations either abolished receptor responses or significantly reduced activation by GABA and muscimol, while activation by pentobarbital remained unaffected. This is consistent with the fact that the binding site for pentobarbital is located closer to the channel gate than the GABA-binding site. Further research by Newell et al. (Reference Newell, McDevitt and Czajkowski2004) used the SCAM method (α₁β₂ assembly expressed in oocytes) to study a large part of loop B (β₂I154–β₂D163). Most mutations within this region affected GABA EC₅₀, but only β₂E155C, β₂S156C, and β₂D163C (Figure 11A and B) also affected SR-95531 EC₅₀, supporting their role in the binding process. Additionally, β₂T160C and β₂D163 mutations slowed the modification rate by 2-aminoethyl MTS hydrobromide–biotin (MTSEA-biotin) upon GABA, SR-95531, and also pentobarbital-induced activation, indicating their role in gating and long-range allosteric interactions. β₂E155 was also postulated to play a role in gating, as its mutation caused spontaneous activity of the receptor.

Padgett et al. (Reference Padgett, Hanek, Lester, Dougherty and Lummis2007) further supported the role of loop B in agonist binding, demonstrating that removal of the hydroxyl group from β₂Y157 significantly increased the GABA EC₅₀. In a study of the loop C region, Venkatachalan and Czajkowski (Reference Venkatachalan and Czajkowski2008) proposed that β₂E153 forms a salt bridge with β₂K196 (loop C) and is thereby involved in shaping the dynamics of this loop. Laha and Tran (Reference Laha and Tran2013) examined α₁β₂γ₂ and α₁β₂ receptors with the β₂Y157A mutation, finding significant effects on GABA EC₅₀, deactivation and binding rate, but only a minor impact on desensitization. This led them to propose that β₂Y157 primarily contributes to agonist binding, with the hydroxyl group essential for receptor functionality. Structural research by Masiulis et al. (Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019) on α₁β₃γ₂ receptors further highlighted the role of loop B in agonist binding. They showed that GABA forms hydrogen bonds with β₃E155 carboxyl and β₃S156 and β₃Y157 (a residue also attributed to binding site ‘aromatic box’) main chain carbonyls. In addition, upon agonist binding, loop B moves toward the complementary subunit, and β₃D163 forms a salt bridge with α₁R85 located in the β3 strand. Considering a particularly large impact of the β₂E155 residue mutation on binding (the largest increase in EC₅₀) and its effect on spontaneous activity, we addressed the impact of the β₂E155C mutation on receptor gating by recording macroscopic current responses to rapid GABA (or muscimol) application, as well as single-channel recordings (Jatczak-Śliwa et al. Reference Jatczak-Śliwa, Kisiel, Czyzewska, Brodzki and Mozrzymas2020). Our data confirmed the involvement of this residue in receptor gating, specifically in the preactivation transition and microscopic desensitization. Additionally, the β₂E155 residue has also been proposed to act as a proton sensor at the GABA-binding site (Michałowski et al. Reference Michałowski, Czyżewska, Iżykowska and Mozrzymas2021). In conclusion, loop B is crucial for ligand binding (with β₂E155 being particularly critical) and also exerts influence on receptor gating.

Loop C

The loop C (Figures 1A and D and 11) of the principal subunit (β) undergoes significant movement upon ligand binding to the orthosteric binding site, a phenomenon termed ‘capping’. This loop has drawn considerable attention due to its high mobility, which appears compatible with the structural rearrangements associated with receptor conformational transitions. Early studies on loop C were based on mutagenesis. Mutations of the β₂Y205 residue (Figure 11A and B) to F, S,N in α₁β₂γ₂ GABA_ARs expressed in oocytes by Amin and Weiss (Reference Amin and Weiss1993) resulted in receptors that were either non-responding or exhibited highly reduced responsiveness to the agonist, thus confirming its important functional role. Czajkowski and Wagner (Reference Czajkowski and Wagner2001) recorded currents mediated by α₁β₂ receptors expressed in oocytes with cysteine mutations at positions β₂V199–β₂S209. GABA or SR-95531 (competitive antagonist) application slowed the MTSEA-biotin modification of β₂S204, β₂Y205, β₂R207 (β10 strand, below loop C, Figure 11A and B), and β₂S209 (close to β₂R207) mutants, indicating a role in forming the binding site. In addition, these studies suggested that the C loop structure was inconsistent with α- or β-helices, suggesting a form of water accessible coil. The general pattern of modifications was interpreted by these authors as β₂G203–β₂S209 being water accessible, whereas β₂V199–β₂T202 were not, although this interpretation was questioned in more recent studies (β9 strand and loop C are solvent-exposed according to structural data, for example, Miller and Aricescu (Reference Miller and Aricescu2014)). Since orthosteric binding sites in α₁β₂γ₂ GABA_ARs are surrounded by different quaternary neighborhood (one is flanked by γ₂ and β₂, while the second is flanked by α₁ and γ₂ subunits), Baumann et al. (Reference Baumann, Baur and Sigel2003) used concatemers expressed in oocytes with β₂Y205S point mutations at specific binding sites and found that GABA molecules bind with a higher affinity to the binding site β(+)/α(−) flanked by α₁ and γ₂ subunits. This non-equivalence of orthosteric binding sites in GABA_ARs is not surprising, as previous kinetic analyses of GABAergic currents have clearly indicated this difference (Maric et al. Reference Maric, Maric, Wen, Fritschy, Sieghart, Barker and Serafini1999; Mozrzymas et al. Reference Mozrzymas, Barberis, Mercik and Zarnowska2003). Residue β₂R207 was thoroughly investigated by Wagner et al. (Reference Wagner, Czajkowski and Jones2004) by analyzing macroscopic responses to rapid agonist applications and single-channel currents of α₁β₂ receptors expressed in HEK293 cells. They found that macroscopic desensitization was unaffected, while deactivation was strongly accelerated. These findings, together with the lack of effect on the maximum open probability, led to the conclusion that this residue is primarily involved in binding but not in gating. Padgett et al. (Reference Padgett, Hanek, Lester, Dougherty and Lummis2007) showed that removal of the hydroxyl group from β₂Y205 significantly increased GABA EC₅₀. Loop C was further examined by Venkatachalan and Czajkowski (Reference Venkatachalan and Czajkowski2008), who performed patch-clamp recordings of α₁β₂γ₂ GABA_ARs expressed in oocytes with various mutations. Specifically, charge reversal or neutralization of residues β₂E153 (β7 strand below loop B, Figure 11A and B) or β₂K196 (β9 strand below loop C, Figure 11A and B) increased GABA EC₅₀, whereas the double charge switch mutation rescued it to the WT level, indicating the presence of a salt bridge between these residues. These observations were confirmed by MCA and modification with the MTS reagent. In addition, these authors proposed that β₂E153 may also interact with β₂E155 and β₂R207. Tran et al. (Reference Tran, Laha and Wagner2011) used MCA on α₁β₂γ₂ GABA_ARs expressed in HEK293 cells and showed coupling between β₂F200 and each of the following residues: β₂Y97 (loop A), α₁R132 (loop E), and β₂R207 (β10 strand, below loop C). The interactions between GABA molecule and positively charged binding site residues (arginine) were investigated by Goldschen-Ohm et al. (Reference Goldschen-ohm, Wagner and Jones2011) using macroscopic and single-channel recordings of α₁β₂ receptors expressed in HEK293 cells. One of them, β₂R207, when mutated to alanine affected the binding and unbinding rates, indicating a major effect on agonist binding but not on gating. The β₂Y205A mutation in both α₁β₂γ₂ and α₁β₂ type receptors expressed in HEK293 cells was investigated by Laha and Tran (Reference Laha and Tran2013), who reported altered GABA EC₅₀, decreased binding rate, accelerated deactivation and only minimal alterations in desensitization, indicating a major effect on agonist binding but not on gating. In addition, these authors have also shown that besides the phenyl ring, as in the case of β₂Y97 (from loop A), the hydroxyl group in this residue is also essential. In our computational study (Michałowski et al. Reference Michałowski, Kraszewski and Mozrzymas2017), we showed that loop C movement is considerably larger in the principal subunit (β₂) than in the complementary (α₁). We may speculate that capping of loop C at the principal subunit might be related to conformational changes upon receptor gating. In another simulation study, it was shown that for successful binding, the ligand must follow the path from the solution to localization ‘underneath’ the loop C (Carpenter and Lightstone Reference Carpenter and Lightstone2016). The loop C closure movement upon GABA binding was shown by Masiulis et al. (Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019) in α₁β₃γ₂ receptors. The authors also attributed loop C residues β₃F200 and β₃Y205 to the binding site ‘aromatic box’, showing that β₃Y205 forms a cation–π interaction with GABA amino group and β₃T202 – a hydrogen bond with ligand’s carboxylate. Taken together, the available evidence clearly demonstrates that C loop is involved in agonist binding, but its role in receptor gating remains unclear. In our recent study (Terejko et al. Reference Terejko, Kaczor, Michałowski, Dąbrowska and Mozrzymas2020), we investigated the role of loop C in gating mechanisms by analyzing the impact of β₂F200 residue mutations in α₁β₂γ₂ GABA_ARs. Extensive analyses and modeling of current responses to saturating agonist applications demonstrated that this mutation strongly affected preactivation, opening, closing, and desensitization, which are all considered gating steps. We also found that this mutation altered receptor sensitivity to the BDZ flurazepam, which we attributed to a change in preactivation kinetics. Thus, like loops A and B, loop C contributes to the formation of the hydrophobic binding cassette on the principal subunits, playing a crucial role not only in agonist binding but also in transmitting the activation signal to the channel gate, thereby influencing receptor gating.

Loop D

It is interesting to note that whereas loops A–C at the principal subunit are predominantly unstructured coils, the loops on the complementary subunit (α) largely form a β-sheet structure with antiparallel β-strands. This arrangement promotes the formation of numerous hydrogen bonds, which enhance the stability and rigidity of the complementary subunit structure (Figures 1A and D and 11). Early mutagenesis studies focused on identifying residues within the complementary subunit that interact with orthosteric ligands and determine the binding process. Sigel et al. (Reference Sigel, Baur, Kellenberger, Malherbel, Malherbe and Malherbel1992) examined α₁β₂γ₂ GABA_ARs expressed in oocytes with point mutations at homologous positions of each subunit building up the receptor, targeting: α₁F64L(r), β₂Y62L, and γ₂F77L (Figure 11A and C). Among these substitutions, only the mutation in the α₁ subunit significantly altered the EC₅₀ for GABA, highlighting its unique role in binding. Further work (Boileau et al. Reference Boileau, Evers, Davis and Czajkowski1999) employed recordings from HEK293 cells in addition to oocytes and α₁β₂ receptors with cysteine substitutions at the α₁Y59(r)–α₁K70(r) positions (Figure 11A and C). Mutations α₁F64C(r) and α₁R66C(r) (Figure 11A and C) increased GABA EC₅₀, and in the case of α₁R66C(r) and α₁S68C, GABA presence protected from MTSEA-biotin modification. The general modification pattern indicated that these residues form a β-strand. Hartvig et al. (Reference Hartvig, Lükensmejer, Liljefors and Dekermendjian2002) examined the role of positively charged arginine residues in α₅β₂γ₂ GABA_ARs expressed in CHO cells: residues α₅R34, α₅R70, α₅R77, α₅R79, α₅R123, α₅R135, α₅R190, and α₅R224 were mutated to lysine. Mutants of α₅R70 (homolog of α₁R67, Figure 11A and C) and α₅R123 (homolog of α₁R120, Figure 11A and C) increased GABA EC₅₀, in agreement with earlier results on α₅β₂γ₂ GABA_ARs. Previous experiments on loop D (Boileau et al. Reference Boileau, Evers, Davis and Czajkowski1999) were extended by the use of various MTS reagents (Holden and Czajkowski Reference Holden and Czajkowski2002), which confirmed earlier results and showed that α₁F64(r) and α₁R66(r) are located in the aqueous core of the binding site. Loop D was also investigated by Baur and Sigel (Reference Baur and Sigel2003), who showed that mutation β₂Y62L (homolog of α₁F65, Figure 11A and C) had a minor effect on receptor activation supporting the role of the β/α and not α/β interface in agonist binding. Using concatemers expressed in oocytes with a point mutation (α₁F65L), Baumann et al. (Reference Baumann, Baur and Sigel2003) have shown that the effects of α₁ subunit substitution were the same in both binding sites. Padgett et al. (Reference Padgett, Hanek, Lester, Dougherty and Lummis2007) demonstrated that no cation–π interaction exists between GABA and α1F65 (Figure 11A and C). Additionally, this residue exhibits tolerance to small chemical modifications, such as fluorination. Using docking and homology modeling, Lummis (Reference Lummis2009) postulated an electrostatic interaction between the agonist’s carboxyl group and α₁R67 residue. Experiments done by Goldschen-Ohm et al. (Reference Goldschen-ohm, Wagner and Jones2011) showed that mutation α₁R67A changed only binding and unbinding kinetics, indicating involvement in binding rather than in gating. Thus, the available data provided strong evidence that loop D and α₁F64(r), in particular, play a key role in ligand binding, but its role in gating remained elusive. To address the issue of the involvement of this loop in receptor gating, our group examined the role of the α₁F64(r) residue in gating properties of the α₁β₂γ₂ GABA_AR by analyzing current responses to rapid agonist applications. Data analysis and modeling have provided evidence that the α₁F64(r) residue is involved not only in binding but also in gating properties (Szczot et al. Reference Szczot, Kisiel, Czyzewska and Mozrzymas2014). In particular, evidence was provided that mutations in this residue strongly affect the so-called flipping transition (preactivation), a state of the receptor previously postulated for other pLGICS (Jadey and Auerbach Reference Jadey and Auerbach2012; Lape et al. Reference Lape, Colquhoun and Sivilotti2008; Mukhtasimova et al. Reference Mukhtasimova, Lee, Wang and Sine2009). Mutations of this residue to L, A and C also increased the spontaneous activity of the receptor (Jatczak-Śliwa et al. Reference Jatczak-Śliwa, Terejko, Brodzki, Michałowski, Czyzewska, Nowicka, Andrzejczak, Srinivasan and Mozrzymas2018), and mutation to C affected the modulatory effect of alkaline pH (Kisiel et al. Reference Kisiel, Jatczak-Śliwa and Mozrzymas2019). This residue was also confirmed to be a part of the binding site ‘aromatic box’, whereas neighboring α₁R67 was shown to form a salt bridge with GABA carboxylate (Masiulis et al. Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019). Studies on the role of residue α₁F64(r) on GABA_AR gating were extended to investigations of the role of preactivation in spontaneous and singly-bound states (observed at low [GABA]), and it was proposed that liganded receptors (singly and doubly bound) undergo flipping transition prior to opening, whereas spontaneous activity does not (Kisiel et al. Reference Kisiel, Jatczak, Brodzki and Mozrzymas2018. Considering that loop D is part of a rigid β-sheet, we hypothesized that this structure is particularly predisposed to convey an activatory signal from the binding site toward the channel gate. To test this possibility, we considered a glycine mutant of the α₁F64(r) residue (an amino acid known to reduce the rigidity and flexibility of polypeptides). The glycine mutation of the α₁F64(r) residue resulted in the strongest impact among all α1F64(r) mutations, affecting both binding and all gating steps considered in our model – preactivation, opening/closing, and desensitization (Kłopotowski et al. Reference Kłopotowski, Michałowski, Gos, Mosiądz, Czyżewska and Mozrzymas2023. This confirms the significant role of the α1F64(r) residue in receptor gating and supports the notion that the rigid β-sheet structure, which hosts the binding site loops at the complementary subunit, may facilitate efficient signal transduction from the binding site to the channel gate.

Loop E and β5 strand

Continuing the description of the orthosteric binding site at the inner sheet of the complementary subunit, adjacent to loop D, there is the loop E, which is a part of β-strand 6 (Figures 1A and D and 11). Studies on α₁β₂γ₂ receptors expressed in Sf-9 insect cells showed that a relatively small change in the primary structure, via the α₁I121V mutation (β5 strand; Westh-Hansen et al. Reference Westh-Hansen, Rasmussen, Hastrup, Nabekura, Noguchi, Akaike, Witt and Nielsen1997), markedly reduced GABA_AR ligand sensitivity. In a subsequent study, the same authors reported that the α₁R120K substitution (Figure 11A and C, β5 strand; Westh-Hansen et al. Reference Westh-Hansen, Witt, Dekermendjian, Liljefors, Rasmussen and Nielsen1999) resulted in a 180-fold increase in GABA EC₅₀. Similarly, Hartvig et al. (Reference Hartvig, Lükensmejer, Liljefors and Dekermendjian2002) demonstrated that the α₅R123K mutation (equivalent to α₁R120) in α₅β₂γ₂ GABA_ARs significantly increased GABA EC₅₀, further confirming the key role of α₅R123K (α₁R120) in agonist binding.

To investigate conformational changes in the ECD upon GABA binding, Muroi et al. (Reference Muroi, Czajkowski and Jackson2006) labeled α₁β₂γ₂ and α₁β₂ receptors with point cysteine mutations at positions α₁E122C(r), β₂P120, and γ₂N135 (β5 and loop E linker; Figure 11A and C), α₁L127C(r), β₂L125, and γ₂ L140 (loop E; Figure 11A and C) using environment-sensitive fluorophores. GABA induced fluorescence changes in both the α₁ and β₂ subunits (α₁β₂ receptor), while the antagonist SR-95531 affected only the α₁. In the α₁β₂γ₂ receptor, reduced fluorescence was observed, likely indicating that less movement was required for receptor activation, and no fluorescence was detected in the γ₂ subunit. In addition, the fluorescence remained unaltered upon macroscopic desensitization onset. In a subsequent study, the same group (Muroi et al. Reference Muroi, Theusch, Czajkowski and Jackson2009) used fluorophore labeling in α₁β₂ GABA_ARs expressed in oocytes to elucidate activation-induced transitions in the ECD and TMD. Labels were placed at residues (mutations to cysteine) α₁L127(r) (loop E top), β₂L125 (same location), and β₂L274 (M2-M3 loop). Together with previous results (Muroi et al. Reference Muroi, Czajkowski and Jackson2006), the authors found that GABA binding increased fluorescence in loop E, while pentobarbital (at high inhibitory concentrations) decreased it through a long-range allosteric mechanism. In addition, the effects on the M2-M3 loop upon GABA or pentobarbital (at lower activating/positively modulating concentrations) were different, indicating distinct activation mechanisms for these ligands.

The region encompassing β4, β5 strands and loop E was further explored by Kloda and Czajkowski (Reference Kloda and Czajkowski2007) using SCAM on α₁M113(r)–α₁L132(r) residues in α₁β₂γ₂ receptors expressed in oocytes. Residues located in the β5 strand: α₁N115(r), α₁L117(r), and in loop E: α₁L127(r), α₁T129(r), and α₁R131(r) (Figure 11A and C) were involved in the binding of both GABA and SR-95531. Moreover, the chemical reactivity of α₁E122(r) (β5 and loop E linker (Figure 11A and C) with MTSEA-biotin was modulated by GABA and pentobarbital, but not by SR-95531, whereas α₁L127C was affected by GABA and SR-95531, but not by pentobarbital. Additionally, α₁E122(r), α₁L127(r), and α₁R131(r) chemical reactivities were all affected by flurazepam. These findings indicate which specific residues play a role only in ligand binding and which are also important for effective gating and modulation by BDZ.

Voltage-clamp fluorometry was used by Akk et al. (Reference Akk, Li, Bracamontes, Wang and Steinbach2011) for the mutant α₁L127C (loop E, Figure 11A and C). Changes in fluorescence were observed upon GABA activation, but remained unaltered during desensitization onset, indicating a clear conformation-sensitivity of this method. Another interaction between ECD residues was found by Laha and Wagner (Reference Laha and Wagner2011): mutations α₁R120A (β5 strand, Figure 11A and C) and β₂D163A (β7–β8 loop; Figure 11A and B) of α₁β₂γ₂ receptors (expressed in HEK293 cells) increased GABA EC₅₀ and accelerated deactivation and unbinding, indicating the existence of a state-dependent salt bridge between these residues at the binding site. The interactions between GABA and binding site residues were investigated using macroscopic and single-channel recordings of α₁β₂ receptors expressed in HEK293 cells by Goldschen-Ohm et al. (Reference Goldschen-ohm, Wagner and Jones2011), who showed that the mutation α₁R132A (bottom of loop E, Figure 11A and C) resulted in a new long-lived open state of the receptor, indicating the role of this residue in the gating process. In addition, the role of the α₁T130 residue in GABA binding (by forming a hydrogen bond with the GABA carboxyl group) was confirmed by structural studies (Masiulis et al. Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019).

Loop F

Loop F may be defined as a structure made of strand β8 or more accurately (according to structural data, e.g., Kim et al. Reference Kim, Gharpure, Teng, Zhuang, Howard, Zhu, Noviello, Walsh, Lindahl and Hibbs2020; Masiulis et al. Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019) of the β8 strand (starting after the loop following the β7 strand), a coil breaking the strand, and the remaining part of the β8 strand (that can be called β8′; Figures 1A and D and 11). In addition, a fragment of loop following the β8′ leading to the β9 strand also can be attributed to loop C. Newell and Czajkowski (Reference Newell and Czajkowski2003) examined loop F (α₁P174(r)–α₁D191(r)) using the SCAM method on α₁β₂ receptors expressed in oocytes. Most of the mutations did not alter GABA EC₅₀; however, among the studied substitutions, α₁D183C had the largest effect on apparent GABA affinity. MTSEA-biotin modification indicated that the loop is unstructured and solvent exposed and in the case of α₁V178C(r), α₁V180C(r), and α₁D183C(r) (Figure 11A and C), both GABA and SR-95531 slowed down the modification rate. Moreover, activation with pentobarbital accelerated the modification of α₁V180C(r) and α₁A181C(r) and slowed that of α₁R186C(r), indicating that this region of the α₁ subunit is involved in channel-gating transitions. Transitions inside the ECD upon ligand binding have also been examined in α₁β₂γ₂ GABA_ARs expressed in oocytes using voltage-clamp fluorometry (Wang et al. Reference Wang, Pless and Lynch2010). Mutations α₁R186C(r), β₂I180C and γ₂S159C (each residue at loop F bottom part) were investigated. In the case of the α₁ subunit, agonist, partial agonist, or antagonist caused a similar increase in fluorescence, indicating binding, but not gating related effect. However, changes in fluorescence in the β₂ and γ₂ subunit mutants were observed only upon agonist binding, indicating an allosteric mechanism or interaction with loop C in the case of the β₂ subunit. In another study, Eaton et al. (Reference Eaton, Lim, Bracamontes, Steinbach and Akk2012) used the SCAM method and found that GABA presence reduced accessibility to γ₂S195 (loop F), once again indicating allosteric effects within the ECD. In a computational study by Carpenter et al. (Reference Carpenter, Lau and Lightstone2012), the authors addressed the role of loop F in two types of GABA_ARs: α₁β₂γ₂ and α₆β₃δ, the latter being characterized by a much higher GABA sensitivity. They found that the major difference between the binding sites in the two considered receptors was the extent of loop F involvement, with a larger contribution in the case of the α₆β₃δ receptor. In addition, free energy calculations confirmed that the α₆β₃δ binding pocket is characterized by an increased affinity for GABA and indicated that the possible role of loop F in the GABA_AR family is to modulate GABA affinity. Mihalik et al. (Reference Mihalik, Pálvölgyi, Bogár, Megyeri, Ling, Barkóczy, Bartha, Martinek, Gacsályi and Antoni2017) studied the action of novel GABA_AR orthosteric antagonists (for α_xβ₂γ₂ receptors) based on a tricyclic oxazolo-2,3-BDZ scaffold. In silico modeling predicted that the test compounds docked in the GABA-binding pocket would interact with various residues at the orthosteric binding site at the α- and β-subunit interface, including an interaction with loop F of the α subunit. This prediction was further investigated by replacing the amino-terminal variable segment of loop F of the α₅ subunit with the corresponding residues of the α₁ and α₂ subunits. When tested with aforementioned inhibitors, the receptors formed by the modified α₅ subunits displayed the pharmacological phenotype of the source of loop-F, thus confirming the key role of this loop in determining the pharmacological profile of studied compounds. A similar approach was employed by Pálvölgyi et al. (Reference Pálvölgyi, Móricz, Pataki, Mihalik, Gigler, Megyeri, Udvari, Gacsályi and Antoni2018), who examined the effect of replacing a short, variable segment of loop F of the GABA_AR (α_xβ₂γ₂ receptors) α₅ subunit with the corresponding segment of the α₂ subunit (GABAA5_LF2) and vice versa (GABAA2-LF5). When compared with their respective wild-type counterparts, GABAA5_LF2 receptors displayed enhanced sensitivity toward GABA (as for α₂ subunit containing receptors), while in GABAA2-LF5 sensitivity was diminished. In summary, loop F was found to play an important role in agonist binding, and the SCAM method provided indirect evidence also for its involvement in gating transitions.

Loop G

Loop G is a neighboring β strand (β1) to loop D (β2) while both strands are oriented in the antiparallel fashion to each other forming a β sheet (Figures 1A and D and 11). Loop G was examined using electrophysiological recordings of currents mediated by α₁β₂γ₂ GABA_ARs (expressed in HEK293 cells) with cysteine substitutions at residues α₁D43(r)–α₁T47(r) (Baptista-Hon et al. Reference Baptista-Hon, Krah, Zachariae and Hales2016). Mutations α₁D43C(r) (Figure 11A and C) and α₁T47C(r) reduced the apparent potency of the agonist and were accessible to MTSEA modification, whereas the α₁F45C(r) mutation (Figure 11A and C) led to biphasic dose–response characteristics and spontaneous activity, but was not accessible for MTSEA modification. However, the presence of GABA reduced MTSEA modification in the α₁D43C(r) and α₁T47C(r) mutants. Thus, these results indicate the movement of loop G during activation of the receptor. In a continuation of their work, the authors showed that α₁D43C(r) and α₁T47R(r) mutations reduced the maximal open probability of the receptor, slowed down desensitization, and accelerated deactivation. In addition, the α₁T47R(r) mutation affected the spontaneous gating of another mutant (β₂L285R), highlighting the role of loop G not only in binding but also in gating (Baptista-Hon et al. Reference Baptista-Hon, Gulbinaite and Hales2017). In a study aimed at turning GABA β₃ homomers into functional receptors responsive to GABA Gottschald Chiodi et al. (Reference Gottschald Chiodi, Baptista-Hon, Hunter and Hales2018), introduced the following mutations in the complementary side of the orthosteric binding site in loop D (Y87F and Q89R), loop E (G152T), and loop G (N66D and A70T). However, electrophysiological investigations indicated that the two substitutions from loops D and E, Q89R and G152T, in β₃ GABA_AR were sufficient to reconstitute GABA-mediated activation, indicating that, at least in this model, the G loop was not essential for converting β₃ GABA_AR into a GABA-activable receptor. To further explore the role of loop G in binding and gating, our group performed a detailed analysis of α₁β₂γ₂ GABA_ARs activity with point mutations of the α₁F45(r) residue (Figure 11A and C) expressed in HEK293 cells (Brodzki et al. Reference Brodzki, Michałowski, Gos and Mozrzymas2020). Macroscopic measurements showed a significant increase of GABA EC₅₀ in the case of mutations to cysteine, glycine, and lysine and a minor increase for leucine substitution, thus confirming the involvement of this residue in the binding process of GABA. However, substitutions with C, G, and K also affected macroscopic desensitization (slower and weaker), thus providing evidence that the gating process was also affected. In single-channel recordings at saturating GABA concentrations, changes were observed in both open- and shut-time distributions. Kinetic modeling clearly indicated that, beyond its role in agonist binding, loop G is important in channel closing/opening and desensitization transitions. Molecular modeling revealed that although α₁F45(r) and loop G do not directly contact with the GABA molecule at the binding site, they play an important role in shaping the interaction between the agonist and loops other than G (Brodzki et al. Reference Brodzki, Michałowski, Gos and Mozrzymas2020). Structural studies, such as those conducted by Kim et al. (Reference Kim, Gharpure, Teng, Zhuang, Howard, Zhu, Noviello, Walsh, Lindahl and Hibbs2020), further confirmed previous conclusions on the role of the loop G in the ECD-binding site. Specifically, residues α₁D44(r) and α₁F45(r) are located parallel to α₁R66(r) and α₁F64(r) (at loop D), respectively, forming the outer layer of the complementary subunit’s β-sheet. It is noteworthy that considering the substantial impact of α₁F64(r) on receptor gating, it was anticipated that the neighboring residue α₁F45(r), located on the same β-sheet, would also participate in GABA_A receptor conformational transitions. This is because the β-sheet likely acts as a structurally cohesive element involved in transducing the activation signal to the channel gate.

Investigations on ρ-type and resistance to dieldrin (RDL) receptor-binding site

Several studies addressed the structure–function relationship of ρ receptors, formerly classified as ‘GABA_C receptors’ but now they are included in the GABA_AR type family. One of the reasons for such interest is that ρ receptors may form agonist-activable functional homomers (in contrast to, e.g., β₃ homomers), considerably facilitating data interpretation. It has been demonstrated that in ρ-type receptors expressed in oocytes, the ρ₁Y198F residue (homolog of β₂Y157 on loop B, Figure 11A and B) forms a cation–π interaction with the GABA molecule (Harrison and Lummis Reference Harrison and Lummis2006a, Reference Harrison and Lummis2006b; Lummis et al. Reference Lummis, Beene, Harrison, Lester and Dougherty2005). In addition, the authors showed that ECD’s arginines are important for receptor function: ρ₁R104 (homolog of α₁R67 on loop D, Figure 11A and C) mutation caused a large increase in GABA EC₅₀ as it forms electrostatic interaction with agonist carboxyl group, ρ₁R158 (homolog of α₁R120 on β5 strand, Figure 11A and C) and ρ₁R170 (homolog of α₁R132 on loop E, Figure 11A and C) mutations resulted in nonfunctional receptors or increased GABA EC₅₀, whereas ρ₁R249 (homolog of β₂R207 on β10 strand, below loop C, Figure 11A and B) had small effects on GABA EC₅₀ and was proposed to be involved in shaping the binding site via hydrogen bonds and/or salt bridges. The prominent role of ρ₁R104 (homolog of α₁R67, loop D, Figure 11A and C) in agonist binding was confirmed by a molecular dynamics study (Melis et al. Reference Melis, Lummis and Molteni2008). Lummis (Reference Lummis2009) using docking and homology modeling confirmed the possibility of a cation–π interaction between GABA and ρ₁Y198 (homolog of β₂Y157 on loop B, Figure 11A and B). The role of loop F in receptor activation was examined by Khatri et al. (Reference Khatri, Sedelnikova and Weiss2009) in ρ type receptors expressed in oocytes using fluorophore labeling (residues ρ₁L216–ρ₁I229, mutations to cysteine), and the loop was proposed to be involved only in binding, but not gating transitions. In addition to ρ-type receptors, RDL receptors, found in insects, was a subject of many investigations. They share similarities with GABA_ARs in that they are activated by GABA and permeable to anions. The cation–π interactions (shown previously to differ between α₁β₂γ₂ and ρ type receptors) were examined in these receptors (Lummis et al. Reference Lummis, McGonigle, Ashby and Dougherty2011) using various unnatural fluorinated phenylalanine derivative mutants. The authors showed that both F206 (homolog of β₂Y157 and ρ₁Y198 on loop B, Figure 11A and B) and Y254 (homolog of β₂Y205 and ρ₁Y247 on loop C, Figure 11A and B) residues form cation–π interaction with the GABA molecule. A molecular dynamics study of the RDL-type receptor confirmed cation–π interactions between GABA and F206 (homologous to β₂Y157 in loop B, Figure 11A and B) and Y254 (homologous to β₂Y205 in loop C, Figure 11A and B). In addition, the authors demonstrated that the agonist forms hydrogen bonds with E204, S205, R111, S176, T251, and exhibits ionic interactions with R111 (homolog of α₁R67, loop D, Figure 11A and C) and E204 (homolog of β₂E155 on loop B, Figure 11A and B) (Ashby et al. Reference Ashby, McGonigle, Price, Cohen, Comitani, Dougherty, Molteni and Lummis2012). Lummis et al. (Reference Lummis, Harrison, Wang, Ashby, Millen, Beene and Dougherty2012) showed that ρ₁Y198 is the only residue that forms a cation–π interaction with GABA. They further examined the roles of other tyrosine residues, showing that ρ₁Y102 (loop D, homolog of α₁F65, Figure 11A and C) and ρ₁Y138 (homolog of β₂Y97 on loop A, Figure 11A and B) stabilize the amino group of the GABA molecule. Additionally, ρ₁Y241 (homologous to β₂F200 on loop C, Figure 11A and B) and ρ₁Y247 (homolog of β₂Y205 on loop C, Figure 11A and B) form a π–π interaction, while ρ₁Y241 and ρ₁R104 (homolog of α₁R67 on loop D, Figure 11A and B) interact via a hydrogen bond. Moreover, ρ₁Y102 was not only identified as crucial for binding but also found to play an integral role in the receptor’s gating mechanism, suggesting a dual function in both ligand recognition and channel activation.

Molecular dynamics simulations of the RDL receptor model, based on the structure of the GluCl receptor, identified key residues involved in interactions with the GABA molecule: R111 (homologous to α₁R67 on loop D, Figure 11A and C) and E204 (homologous to β₂E155 on loop B, Figure 11A and B) form salt bridges, while Y109 (homologous to α₁F65 on loop D, Figure 11A and B), F206 (homologous to β₂Y157 on loop B, Figure 11A and B), and Y254 (homologous to β₂Y205 on loop C, Figure 11A and B) engage in cation–π interactions (Comitani et al. Reference Comitani, Cohen, Ashby, Botten, Lummis and Molteni2014). Naffaa et al. (Reference Naffaa, Absalom, Raja Solomon, Chebib, Hibbs and Hanrahan2016) explored the role of the T244 residue in the ρ-type receptor (homologous to β₂T202 on loop C, Figure 11A and B) by introducing mutations. Substitution with alanine or cysteine resulted in minimal GABA sensitivity, while substitution with serine significantly reduced this sensitivity. Using molecular modeling, the authors proposed that a hydrogen bond between GABA and the hydroxyl group of ρ₁T244 is critical for ligand binding and contributes to the movement of loop C, which closes the binding site. The binding process was further investigated using funnel-metadynamics simulations and the RDL receptor model of the WT and two mutants: R111A (homolog of α₁R67 on loop D, Figure 11A and C) and E204A (homolog of β₂E155 on loop B, Figure 11A and B) were considered (Comitani et al. Reference Comitani, Limongelli and Molteni2016). The authors revealed that in both mutants, the free energy well present in the WT receptor-binding side, was absent, and loop C exhibited significant movement upon ligand binding. The roles of residues ρ₁S168 (homolog of α₁T130 on loop E, Figure 11A and C) and ρ₁S243 (homolog of β₂S201 on loop C, Figure 11A and B) were examined by Naffaa et al. (Reference Naffaa, Hibbs, Chebib and Hanrahan2022) using electrophysiological and computational methods. Mutations in these residues altered the dose–response relationships of GABA and other ligands, leading the authors to propose that these residues engage in a set of state-dependent interactions with other ECD residues.

Top part

Within the top part of the ion pore, 17′ histidine of the β₂ subunit (β₂H267, Figure 13) has attracted considerable interest because of its role in the modulation of receptor function by pH level and Zn²⁺ ions (Krishek et al. Reference Krishek, Moss and Smart1998; Krishek et al. Reference Krishek, Amato, Connolly, Moss and Smart1996). H292A (17′) mutation in murine β₃ subunit in homomeric (β₃) and heteromeric (α₁β₃) GABA_A receptors caused reduction of Zn²⁺ inhibitory effect, indicating thus that H292 in TM2 of murine β₃ subunit is an important component of a Zn²-binding site on the GABA_AR (Wooltorton et al. Reference Wooltorton, Mcdonald, Moss and Smart1997). The 17′ residue of the β subunit has also been implicated in receptor modulation by protons (H⁺). Wilkins et al. (Reference Wilkins, Hosie and Smart2002) showed that H267(A, E) mutations in the β₂ subunit of α₁β₂ receptors abolished the potentiating effect of H⁺ (at GABA concentrations between 1 and 30 μM), whereas the β₂H267K mutant did not affect proton modulation (compared to WT) but increased the rate of macroscopic desensitization of GABA-induced currents. Further investigation revealed that the β₂H267A mutation increased the inhibition of GABA-evoked currents at pH 5.4 for the α₁β₂H267Aγ₂ receptor to a level similar to that observed for α₁β₂H267A (Wilkins et al. Reference Wilkins, Hosie and Smart2005). Horenstein et al. (Reference Horenstein, Wagner, Czajkowski and Akabas2001), using two-electrode voltage-clamp on oocytes and disulfide bond trapping, examined the proximity and mobility of cysteine substitutions at the 20′, 17′, and 6′ positions of the α₁ and β₁ subunits (Figure 13) in resting and activated receptors. The authors found that, with or without an agonist, disulfide bonds were formed between α₁N275C and β₁E270C (20′), as well as between α₁S272C and β₁H267C (17′), indicating significant mobility of this part of the ion pore. The possible role of residue β₂E270 (20′), located above β₂H267, in the interaction with chloride ions was proposed in a molecular dynamics study of α₁β₂γ₂ receptors (Michałowski et al. Reference Michałowski, Kraszewski and Mozrzymas2017). Structural studies (Kim et al. Reference Kim, Gharpure, Teng, Zhuang, Howard, Zhu, Noviello, Walsh, Lindahl and Hibbs2020; Masiulis et al. Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019) showed that the top part of the ion pore is wide even in the non-permeable states of the receptor (desensitized or antagonist-bound), indicating that, in contrast to the other regions of the pore (middle and bottom), no gate of any kind is present there. A recent study reported the crucial role of the β₂H267 residue and its interaction with β₂E270 in the activation of proton-gated chloride channels (Garifulina et al. Reference Garifulina, Friesacher, Stadler, Zangerl-Plessl, Ernst, Stary-Weinzinger, Willam and Hering2022). The authors discovered that homomers of β₂ and β₃ subunits formed chloride channels that were directly activated by protons. However, due to the constitutive activity of β₁ homomers, the direct activation by H⁺ could not be reliably observed. To address this, the β₁S265N (15′) mutation was examined, as it has been previously shown to reduce the spontaneous activity of homomeric β₁ channels (Miko et al. Reference Miko, Werby, Sun, Healey and Zhang2004). Using a two-electrode voltage-clamp technique on oocytes, Garifulina et al. (Reference Garifulina, Friesacher, Stadler, Zangerl-Plessl, Ernst, Stary-Weinzinger, Willam and Hering2022) found that proton-activated chloride currents mediated by homomeric channels (β₁S265N, β2, β3) exhibit slower activation and desensitization kinetics in comparison to αβ receptors. The β₂H267A mutation completely abolished the proton sensitivity of these homomeric channels. Molecular dynamics simulations further revealed that protonation of the 17′ histidine induced hydrogen bonding with the 20′ glutamate of the adjacent subunit. Further evidence of β₂H267–β₂E270 interaction has been provided by homologous substitution (ρ₁G331H–ρ₁A334E) in homomeric ρ₁ receptors. While ρ₁ homomers are typically insensitive to protons, the incorporation of histidine and glutamate at positions 331 and 334 rendered the channel proton-sensitive, whereas a single mutation, ρ₁G331H, was insufficient to induce this sensitivity (Garifulina et al. Reference Garifulina, Friesacher, Stadler, Zangerl-Plessl, Ernst, Stary-Weinzinger, Willam and Hering2022).

Middle part

The residues located at the 9′ and 6′ positions (Figure 13) in the middle part of the ion pore play important roles in ion permeation and GABA_A receptor gating. Gurley et al. (Reference Gurley, Amin, Ross, Weiss and White1995) found that mutations at the 6′ residues in the α₁β₂γ₂ receptor cause insensitivity to the channel blocker PTX. In a subsequent study by this group, it was reported that a single mutation of T6’ to phenylalanine in any subunit of the α₁β₂γ₂ receptor was sufficient to confer this insensitivity (Sedelnikova et al. Reference Sedelnikova, Erkkila, Harris, Zakharkin and Weiss2006). These results underscored thus the key role of the 6′ residue in PTX binding. Studies by Tierney et al. (Reference Tierney, Birnir, Pillai, Clements, Howitt, Cox and Gage1996) demonstrated that mutations at the L9′ position, such as α₁L264T or β₂L259T, in α₁β₂ receptors caused spontaneous activity, indicating an impact on receptor gating. In the α₁β₁ receptor with the α₁L264T mutation (9′), GABA-evoked currents were characterized by slower rise times and decay kinetics compared to native receptors. In contrast, receptors with mutations at the 9′ position of the β₂ subunit or in both α₁ and β₁ subunits were permanently open, suggesting that this residue is key in controlling the channel activation gate. Importantly, other mutations of this conserved M2 leucine in different GABA_AR subunits also led to spontaneously active receptors. Specifically, mutations such as α₁L263S, β₂L259S, or γ₂L274S (all at 9′) shifted the GABA dose-response relationship to the left (Chang and Weiss Reference Chang and Weiss1999). The α₁β₂γ_2L receptors containing one mutated subunit at 9′ (of any type) exhibited spontaneous activity, which was blocked by PTX (an open-channel blocker) and the competitive antagonist bicuculline (Chang and Weiss Reference Chang and Weiss1999).

High-resolution experiments using ultrafast agonist delivery provided detailed insights into the kinetic properties of evoked currents mediated by α₁L264Tβ₂γ₂ receptors (Scheller and Forman Reference Scheller and Forman2002). Receptors with the α₁L9′T substitution exhibited higher GABA sensitivity, spontaneous activity, and PTX sensitivity. Furthermore, currents mediated by these mutated receptors were characterized by extremely slow desensitization. Deactivation of currents mediated by wild-type receptors was described with two time constants, and this process slowed down after the induction of partial macroscopic desensitization (by prolonged agonist exposure), thus showing clear desensitization-deactivation coupling. In contrast, α₁L264Tβ₂γ₂ receptors showed deactivation characterized by a single time constant, which was uncoupled from desensitization. Based on the modeling of macroscopic currents, the α₁L264T mutation was found to increase the gating efficacy of GABA_ARs by slowing the channel closing rate (Scheller and Forman Reference Scheller and Forman2002).

Horenstein et al. (Reference Horenstein, Wagner, Czajkowski and Akabas2001) analyzed the impact of mutating various residues to cysteines of the TM2 segment of the α₁ and β₂ subunits. They found that, with or without GABA, disulfide bonds formed near the extracellular end of TM2 at the 20′ and 17′positions, indicating that this part of TM2 is particularly mobile and/or flexible. In contrast, the disulfide bond between α₁T261C/β₁T256C residues (6′) is formed only in the presence of GABA and locks the channel in the open state (Horenstein et al. Reference Horenstein, Wagner, Czajkowski and Akabas2001). Their results suggest that receptor activation must involve an asymmetric rotation of adjacent subunits toward each other.

The spontaneous activity of α₁β₂259Sγ₂ receptors_, with 9′ residue mutated, was confirmed by Mortensen et al. (Reference Mortensen, Wafford, Wingrove and Ebert2003), who demonstrated a correlation between partial agonist potency and the level of receptor spontaneous activity. Downing et al. (Reference Downing, Lee, Farb and Gibbs2005) further showed that the spontaneous activity of α₁L263Sβ₃γ₂(r) receptors could be potentiated by various allosteric modulators. In addition to 9′ residues, also mutations at the 6′ position (Figure 13) led to spontaneous activity; for instance, the β₁T256C mutation caused spontaneous activity following the formation of an intersubunit disulfide bond between adjacent β₁ subunits in α₁β₁ receptors (Rosen et al. Reference Rosen, Bali, Horenstein and Akabas2007). Importantly, molecular dynamics simulations by Xie et al. (Reference Xie, Wang, Sha and Cheng2013) confirmed the presence of an energy barrier for chloride ion permeation at the 9′ and 6′ residue levels

Further studies by Patel et al. (Reference Patel, Mortensen and Smart2014) focused on α₄β₃δ receptors containing δ subunit and showed that L9′ mutations in the α₄ (L297S), β₃ (L284S), or δ (L288S) subunits increased GABA sensitivity. Current responses to GABA mediated by receptors with any mutated L9′ exhibited prolonged deactivation kinetics compared with the wild-type receptor (α₄β₃δ). Moreover, all receptors with a serine mutation at 9′ in any subunit showed spontaneous activity, which was blocked by PTX (Patel et al. Reference Patel, Mortensen and Smart2014). In contrast to the effect observed in α, β, γ, and δ subunits, Germann et al. (Reference Germann, Burbridge, Pierce and Akk2022) found that a mutation at the 9′ position of the ε subunit (L299S) in α₁β₂ε receptors significantly reduced constitutive activity and slightly shifted the dose–response curve to the right. Moreover, mutation of the 6′ residue to phenylalanine in the ε subunit (S299F), which is known to mediate receptor insensitivity to PTX, increased receptor sensitivity to this channel blocker (Germann et al. Reference Germann, Burbridge, Pierce and Akk2022). Finally, structural studies (Kim et al. Reference Kim, Gharpure, Teng, Zhuang, Howard, Zhu, Noviello, Walsh, Lindahl and Hibbs2020; Masiulis et al. Reference Masiulis, Desai, Uchański, Serna Martin, Laverty, Karia, Malinauskas, Zivanov, Pardon, Kotecha, Steyaert, Miller and Aricescu2019) confirmed that the channel gate forms at the 9′ residue level (α₁L264, β₂L259, and γ₂L274) with pore constriction made of a conserved leucine ring in receptors bound to bicuculline or PTX.

Bottom part

Similar to some residues located in the middle segment of the ion pore, mutations in the bottom regions of the pore have also been found to cause spontaneous activity. Ueno et al. (Reference Ueno, Lin, Nikolaeva, Trudell, John Mihic, Adron Harris and Harrison2000) showed that the α₂V257W mutation at the 2′ position (Figure 13) induces channel opening, enhances sensitivity to GABA, and renders the receptor insensitive to PTX. Furthermore, this mutation was shown to abolish the sensitivity to pregnanolone sulfate (PS) (Akk et al. Reference Akk, Bracamontes and Steinbach2001; Sachidanandan and Bera Reference Sachidanandan and Bera2015). Interestingly, receptors with mutations at the 2′ residues in the β or γ subunits remained susceptible to modulation by this neurosteroid.

A large number of mutations in the M2 segment of GABA_AR are associated with epilepsy (for review see Hernandez and Macdonald Reference Hernandez and Macdonald2019). One severe form of epilepsy is Dravet syndrome, which is predominantly associated with a mutation in the SCN1A gene encoding the α₁ subunit of the sodium channel. Recently, a pore-lining mutation at the -2′ position in the γ₂ subunit (γ₂P302L) has also been associated with Dravet syndrome. The γ₂P302L substitution resulted in a 90% reduction in current amplitude and a rightward shift in the dose-response curve. Currents mediated by α₁β₃γ₂P302L receptors were characterized by slower rise times, increased desensitization and faster deactivation. Moreover, single-channel recordings further revealed that the γ₂P302L mutation increased the frequency of low-conductance openings, reduced the occurrence of the main-conductance openings and decreased the open probability. Measurements of mIPSCs in neurons overexpressing mutated γ₂ subunit revealed reduced current amplitude, slower rise times, and prolonged decay kinetics. Collectively, the γ₂P302L mutation alters gating transitions, leading to increased neuronal excitability, a key factor in the pathology of Dravet syndrome (Hernandez et al. Reference Hernandez, Kong, Hu, Zhang, Shen, Jackson, Liu, Jiang and Macdonald2017).

Binding site at the β/α interface

The analogy to the orthosteric binding site at the β/α interface in the ECD has inspired several studies aimed at characterizing the β/α interface in the upper part of the TMD. Key residues located at the helices forming the binding cavity for general anesthetics have been determined (Bali and Akabas Reference Bali and Akabas2004; Chiara et al. Reference Chiara, Dostalova, Jayakar, Zhou, Miller and Cohen2012; Jayakar et al. Reference Jayakar, Zhou, Chiara, Dostalova, Savechenkov, Bruzik, Dailey, Miller, Eckenhoff and Cohen2014; Li et al. Reference Li, Chiara, Sawyer, Husain, Olsen and Cohen2006, Reference Li, Chiara, Cohen and Olsen2010; Stewart, Hotta, Desai, et al., Reference Stewart, Hotta, Desai and Forman2013a; Stewart, Hotta, Li, et al., Reference Stewart, Hotta, Li, Desai, Chiara, Olsen and Forman2013b): β_2/3M286 and β₃V290 on the M3 helix (principal side) and α₁L232, α₁M236, α₁T237, and α₁I239 on the M1 helix (complementary side). Some of these residues are illustrated in Figure 15A. To further examine the role of specific residues at the M3 and M1 helices as an entrance to the binding cavity, several cysteine mutants at the respective M3 and M1 residues were analyzed, clearly demonstrating the potential for disulfide bond formation between them (Bali et al. Reference Bali, Jansen and Akabas2009; Borghese et al. Reference Borghese, Hicks, Lapid, Trudell and Harris2014), indicating thus their close interaction. In addition, mutagenesis studies revealed that mutations of β₂M286 and α₁M236 (Figure 15A) to tryptophan produced effects on receptor function similar to those of modulation by general anesthetics (Krasowski et al. Reference Krasowski, Nishikawa, Nikolaeva, Lin and Harrison2001; Stewart et al. Reference Stewart, Desai, Cheng, Liu and Forman2008).

Figure 15. TMD-binding sites of the GABA_AR: α₁β₂γ₂ receptor structure viewed from the plane of the membrane. (A): Upper TMD-binding site. The cavity between α-helices M2 and M3 (principal subunit side) and M1 (complementary subunit side) is marked in orange. Crucial for modulatory action, β₂N265 is located in the M2 helix behind β₂M286 (M3). At the same level of the complementary subunit, α₁L232 is present. (B) PAM neurosteroid-binding sites: intersubunit at the bottom, and intrasubunit at the top of the TMD. Important for neurosteroid function, α₁Q242 is located roughly in the middle of the M1 helix (top of the intersubunit side), and on the opposite side of the cavity, β₂Y304 and α₁W246 are located (bottom of the site). (C) NAM neurosteroid-binding site at the bottom of the α subunit.

The M2 helix of the β subunit lines the inner side of the cavity (Figure 15A) and its N265 residue (present in β₂ and β₃, but not in the β₁ subunit, Figure 15A) have been identified as key players in the action of etomidate. Substituting this residue with serine (present in the β₁ subunit) reduced the modulatory effect of etomidate (Belelli et al. Reference Belelli, Lambert, Peters, Wafford and Whiting1997), whereas substitution with methionine eliminated the modulation by etomidate and reduced the effect of propofol (Siegwart et al. Reference Siegwart, Jurd and Rudolph2002, Reference Siegwart, Krähenbühl, Lambert and Rudolph2003). Later research showed that this M2 residue plays an important role in both modulator binding (affinity) and its modulatory action (efficacy). The double mutant β₂N265M + α₁M236C showed weaker protection against thiol modification upon etomidate application than the single α₁M236C mutant (Stewart et al. Reference Stewart, Pierce, Hotta, Stern and Forman2014). In addition, another double mutant β₂N265M + α₁M236W exhibited a reduced level of mimicking the etomidate effect compared to the single mutant (α₁M236W), indicating the role of β₂N265 in modulation (Stewart et al. Reference Stewart, Pierce, Hotta, Stern and Forman2014).

Binding sites at other subunit interfaces

Extensive research has targeted other upper TMD subunit interfaces. Photolabeling studies (Yip et al. Reference Yip, Chen, Edge, Smith, Dickinson, Hohenester, Townsend, Fuchs, Sieghart, Evers and Franks2013) and molecular modeling analysis (Franks Reference Franks2015) indicated that propofol could also bind to the intrasubunit cavity of the β₃ subunit. This binding site is located slightly higher (near the ECD/TMD interface) compared to the cavities at the β/α interfaces, with key roles for the β₃H267 residue (photolabeled site) and β₃F221 (mutation to tryptophan, similar to some residues at the β/α binding sites, resulted in spontaneous activity and eliminated modulation by propofol). This intrasubunit binding site was further investigated using patch-clamp electrophysiology on β₃ homomeric receptors (Eaton et al. Reference Eaton, Cao, Chen, Franks, Evers and Akk2015). The authors identified the following residues as important for propofol action at this site in β₃ homomeric assembly: β₃Y143, β₃F221, β₃Q224, and β₃T266. In a subsequent study on α₁β₃ heteromeric GABA_ARs, Eaton et al. (Reference Eaton, Germann, Arora, Cao, Gao, Shin, Wu, Chiara, Cohen, Steinbach, Evers and Akk2016) summarized possible binding sites for propofol and key residues for modulation: β/α interface with β₃M286, α/β interface with β₃Y143, β₃F221, and β₃Q224, and β/β interface with the same residues as in the α/β interface. Another photolabeling study indicated that photoreactive analogs of general anesthetics can bind to intersubunit binding sites, where β subunit plays complementary role, specifically at the α/β and γ/β interfaces in the αβγ assembly type (Jayakar et al. Reference Jayakar, Zhou, Chiara, Dostalova, Savechenkov, Bruzik, Dailey, Miller, Eckenhoff and Cohen2014), with varying affinities. The photolabeled residue in this case was β₃M227, a homolog of α₁M236 located at the M1 helix.

In a study by the Sigel group, a set of receptor mutants was examined (in α₁β₂γ₂ assembly): β₂N265I (M2 helix, inner binding cavity, Figure 15A) and homologous point mutants: α₂S269I and γ₂S280I (Maldifassi et al. Reference Maldifassi, Baur and Sigel2016). They used two-electrode voltage-clamp recordings on Xenopus oocytes expressing these receptor types to assess the potentiation extent of GABA-evoked current by propofol, etomidate, pentobarbital, and low concentrations of diazepam or tetrahydrodeoxycorticosterone (THDOC). Triple mutants abolished potentiation by anesthetics and pentobarbital, but not by diazepam or THDOC (which do not bind to these sites at low concentrations). In the case of single point mutants, mutations in the β₂ subunits abolished potentiation by propofol and etomidate, confirming previous data. Conversely, mutations in the γ₂ subunit either lowered (for propofol) or increased (for etomidate) the potentiation, indicating the involvement of this subunit in modulation by these drugs. Interestingly, all single mutations decreased current potentiation by pentobarbital, suggesting that pentobarbital may bind to multiple interfaces and/or exert allosteric effects. In addition, the authors showed (using concatenated receptors) that etomidate (but not propofol) binding sites have different properties, and only the first β subunit (neighboring γ subunit) is important for its action. Altogether, these results indicate that the modulation of GABA_AR by these compounds relies, to some extent, on allosteric coupling between the interface cavities. Similar work was performed by the Forman group; the set of mutants (to cysteine and tryptophan) at positions α₁L232 and α₁M236 (M1 helix, binding cavity entrance complementary side, Figure 15A) and their homologs (β₃M227, β₃L231, γ₂L242, and γ₂L246) in the α₁β₃γ₂ receptor were examined using voltage-clamp electrophysiology in Xenopus oocytes (Nourmahnad et al. Reference Nourmahnad, Stern, Hotta, Stewart, Ziemba, Szabo and Forman2016). Tryptophan substitutions at position α₁M236 and its homologs caused spontaneous channel activity and reduced GABA EC₅₀. Cysteine substitutions demonstrated that etomidate protected both α subunit mutations, while mtFd-mPaB (R-5-allyl-1-methyl-5-(m-trifluoromethyl-diazirinylphenyl) barbituric acid) provided protection for both β subunit mutations against thiol modification. Additionally, propofol protected α₁M236C and β₃M227C mutants, whereas alphaxalone (a neurosteroid) did not protect against same modification. Interestingly, none of the tested modulators protected the cysteines substituted in the γ subunit. These data clearly indicated an important phenomenon – each subunit interface cavity plays a role in receptor gating (spontaneous activity induced by tryptophan substitution and effect on GABA dose-response relationship), but the α/γ interface seems to be an orphan interface without any compatible ligands. Electrophysiological recordings of currents in oocytes expressing α₁β₂γ₂ receptors with point mutations β₂M286W (at the β/α binding site, Figure 15A) and β₂γY143W (at the β intrasubunit site) together with Monod–Wyman–Changeux (MWC) kinetic modeling, revealed that propofol binds to at least four distinct sites (four investigated and potentially other different sites) (Shin et al. Reference Shin, Germann, Johnson, Forman, Steinbach and Akk2018). The intersubunit sites (β/α, γ/β, and α/β) were further investigated using competitive photolabeling studies (Jayakar et al. Reference Jayakar, Zhou, Chiara, Jarava-Barrera, Savechenkov, Bruzik, Tortosa, Miller and Cohen2019). The authors found that different photoreactive analogs of general anesthetics bind to all these interfaces, but with different affinities. Most importantly, the etomidate analog demonstrated high-affinity binding to the β/α site and low-affinity binding to the γ/β and α/β sites. In contrast, the propofol analog exhibited high affinity for the β/α site and medium affinity for the γ/β and α/β sites. Conversely, mephobarbital displayed high affinity for the γ/β and α/β interfaces but low affinity for the β/α site. The intrasubunit β site was not examined. In another study, the interplay between different modulators binding in the upper TMD region was examined in concatemeric α₁β₂γ₂ receptors expressed in oocytes (Shin et al. Reference Shin, Germann, Covey, Steinbach and Akk2019). The authors discovered that different neurosteroids (both 5α- and 5β-reduced forms, as well as natural and enantiomeric variants) share a common interaction site, whereas propofol and pentobarbital exhibit partially overlapping binding sites. This topic was further explored in another study by Szabo et al. (Reference Szabo, Nourmahnad, Halpin and Forman2019), who performed electrophysiological recordings of α₁β₃γ₂ receptors expressed in oocytes with point mutations of residues β₃N265 (and homologs in other subunits, Figure 15A) and β₃M286 (and homologs, Figure 15A) and performed MWC modeling analysis. The obtained results confirmed the binding pattern of etomidate (both β/α), barbital (γ/β and α/β) and propofol (all four sites), but contrary to the proposal of Shin et al. (Reference Shin, Germann, Johnson, Forman, Steinbach and Akk2018), some allosteric coupling between the sites was suspected.

Recent cryo-EM structural imaging and molecular dynamics studies (Kim et al. Reference Kim, Gharpure, Teng, Zhuang, Howard, Zhu, Noviello, Walsh, Lindahl and Hibbs2020) have provided important insights into the roles of the respective binding sites in the upper TMD. Several structures of the α₁β₂γ₂ receptor with different bound modulators have been obtained. These structures confirmed the binding of etomidate to both β/α interfaces and the binding of pentobarbital (a barbiturate) to γ/β and α/β interfaces, in agreement with most previous studies. However, propofol was found only in β/α interfaces, whereas most studies by other authors indicated binding of this compound also to other sites (see Table 2). It is important to emphasize that in the case of propofol (but not etomidate), the γ/β and α/β cavities remained expanded and showed the presence of lipids. In addition, no modulators were found in the postulated intrasubunit binding sites in the β₂ subunit or in a previously determined as ‘orphan’ site at the α/γ interface. In agreement with previous studies, the residues lining the modulatory binding site were determined as follows for the β/α interface (homologs of key residues for other subunits in parenthesis, Figure 15A): β₂D282, β₂M286 (α₁S310, γ₂A291), β₂F289 (α₁F304, γ₂Y294) at the M3 helix; β₂N265 (α₁S280, γ₂S270), β₂T262 at the M2 helix; and α₁I228 (β₂L223), α₁L232 (β₂M227), α₁P233 (β₂P228), and α₁M236 (β₂L231) at the M1 helix. A summary of the determined binding sites is presented in Table 2. Another important structural insight was provided by the work of Zhu et al. (Reference Zhu, Sridhar, Teng, Howard, Lindahl and Hibbs2022), who determined the receptor structures with zolpidem (a positive allosteric modulator) and DMCM (a negative allosteric modulator) bound at the α/γ ECD and β/α TMD interface binding sites. The key residues for these ligands binding within TMD are β₂D282, β₂M286, β₂F289, α₁I228, α₁L232, and α₁M236.

Table 2. Localization of the upper TMD modulatory binding sites

Presence of upper TMD modulators in respective intersubunit cavities according to studies listed in the first column.

Neurosteroid binding in the TMD

TMD has also been the subject of numerous investigations because of its role in neurosteroid modulation. Early studies on chimeric receptors (ECD of RDL receptor and TMD of α₁, β₂, and γ₂ GABA receptor subunits, respectively) indicated the key role of the M1 and M2 helices in modulation by neurosteroids. More specifically, the α₁T236I(r) mutation reduced the direct activation by allopregnanolone and THDOC, whereas α₁Q241W(r) (Figure 15B) mutant diminished both the direct activation and modulatory effects of these compounds (Hosie et al. Reference Hosie, Wilkins, Da Silva and Smart2006). Because the α₁Q241(r) residue is expected to form a hydrogen bond with the hydroxyl group of neurosteroids, additional mutants were examined (Hosie et al. Reference Hosie, Wilkins, Da Silva and Smart2006). The α₁Q241(H,N)(r) mutants were similar to the WT, while α₁Q241(S,T,M) exhibited slightly reduced potencies of neurosteroid modulation. In contrast, α₁Q241(I,L)(r) showed no susceptibility to neurosteroids whatsoever, confirming that the ability of α₁Q241(r) to form a hydrogen bond with neurosteroids is crucial. On the basis of homology modeling, studies by Hosie et al. (Reference Hosie, Wilkins, Da Silva and Smart2006) indicated residues that may form contact with the opposite side of the neurosteroid molecule: α₁N407(r) (Figure 15B) and α₁Y410(r) and showed that mutants α₁N407(A,V,D)(r) and α₁Y410F(r) have reduced neurosteroids potencies. Similar analysis was performed for the α₁T236(r) residue and its mutations (to I, V, and S) affected modulation by neurosteroids. In addition, β₂Y284 (its mutation to F affects direct activation by neurosteroids) was found to form a neurosteroid-binding site together with α₁T236(r). Altogether, Hosie et al. (Reference Hosie, Wilkins, Da Silva and Smart2006) proposed two binding sites for neurosteroids: the first made of α₁Q241(r), α₁N407(r), and α₁Y410F(r) and the second composed of α₁T236(r) and β₂Y284, both spanning from the middle of the TMD up to its top part. These studies were continued for different subunit assemblies of the GABA_A receptor (Hosie et al. Reference Hosie, Clarke, da Silva and Smart2009). Residues located at M4: α₁N407(r) and α₁Y410F(r) are highly conserved among receptor subtypes, but α₁Q241(r) is specific for α subunits, indicating the key role of this subunit. All receptors in the α _xβ₃γ₂ assembly (where x represents any of the α subunits) with a point mutation at a residue homologous to α₁Q241(r) had similar phenotypes. Substitution with histidine resulted in a phenotype comparable to WT, while substitutions with threonine, serine, or methionine reduced the potentiation of GABA-evoked currents at EC₁₀. In contrast, substitutions with leucine or tryptophan abolished potentiation (Hosie et al. Reference Hosie, Clarke, da Silva and Smart2009). These results further underlined the key role of α₁Q241(r) and its ability to form hydrogen bonds with neurosteroids. Interesting results were obtained from photolabeling studies, which indicated that the neurosteroid analogs labeled the β₃F301 residue (Figure 15B) located at the bottom of the M3 helix, indicating thus a different binding site for these modulators (Chen et al. Reference Chen, Manion, Townsend, Reichert, Covey, Steinbach, Sieghart, Fuchs and Evers2012)

PAM intersubunit binding site at the β/α interface and NAM intrasubunit sites in α and β subunits

An important insight into neurosteroid binding was provided by structural research on β₃/α₅ chimeric GABA_ARs (Miller et al. Reference Miller, Scott, Masiulis, De Colibus, Pardon, Steyaert and Aricescu2017). The structure of these receptors (homomeric, β₃ ECD and α₅ TMD, α₅T287K background mutation to reduce spontaneous activity) was resolved with bound pregnanolone and in the apo state. A neurosteroid was found between the two subunits (M3 helix of the principal and M1 of the complementary subunit) in the bottom part of the TMD. The α₅Q245 residue (Figure 15B) formed a hydrogen bond with the hydroxyl group of the neurosteroid. The A ring of the modulator was located between α₅I242 and α₅W249 of the M1 helix, while the B and D rings were positioned between α₅I305 of the M3 helix and α₅W249 of the M1 helix. Additionally, the α₅V246 residue in the M1 helix lined the deep part of the binding site, and the ketone of the neurosteroid formed a hydrogen bond with α₅T309 in the M3 helix (Figure 15B). Mutations of these residues (α₅Q245S, α₅V246A, α₅W249L, α₅I305A, and α₅T309A) reduced receptor sensitivity to pregnanolone, further indicating that the binding site containing α₅Q245 (α₁Q242) does not span upwardly, but rather toward the bottom of the TMD. To examine the role of the binding site residues located at the principal subunit in αβγ assembly, the modulation of α₁β₃L297Aγ₂ (corresponding to α₅I305) and α₁β₃LF301Aγ₂ (α₅T309, Figure 15B) receptors by neurosteroids was assessed. Only the α₁β₃L297Aγ₂ mutation showed a smaller effect on modulation by pregnanolone and allopregnanolone compared to the WT, indicating a more important role of the α subunit. Docking showed that pregnanolone, allopregnanolone, and epipregnanolone bind to the same site at the β/α interface, but the epipregnanolone hydroxyl group is oriented upward and cannot coordinate with α₅Q245 (Figure 15B), as in the case of other compounds. Miller et al. (Reference Miller, Scott, Masiulis, De Colibus, Pardon, Steyaert and Aricescu2017) compared also the apo (without ligand) and pregnanalone bound structures and observed that in the neurosteroid bound state the ion pore widens at -2′ residue level and loop M2-M3 moves inward. In addition, other chimeric structures (GLIC ECD and GABA_AR α₁ TMD, α₁G258V(r) to enhance purification and increase desensitization) with THDOC, PS and without any modulators bound were obtained by Laverty et al. (Reference Laverty, Thomas, Field, Andersen, Gold, Biggin, Gielen and Smart2017). Similar to the allopregnanolone in β₃/α₅ chimeric GABA_AR, the THDOC hydroxyl group formed a hydrogen bond with α₁Q241(r) (Figure 15B), with rings C and D aligned parallel to α₁W245 (Figure 15B), and the neurosteroid ketone group formed a hydrogen bond with α₁T305(r) and α₁Y308(r) at the bottom of the binding site. Mutations in these residues (α₁Q241L(r), α₁W245L(r), and α₁T305W(r)) abolished the modulatory effects of THDOC. Interestingly, Laverty et al. (Reference Laverty, Thomas, Field, Andersen, Gold, Biggin, Gielen and Smart2017) determined also the binding site of PS to be placed in the bottom TMD within single subunits M3 and M4 helices. The sulfate group of PS was oriented toward α₁K390(r), and this molecule probably interacts with α₁I391(r), α₁A398(r), and α₁F399(r) (Figure 15C). Mutations of these residues (α₁K390A(r), α₁I391C(r), α₁A398C(r), and α₁F399C(r)) reduced the inhibitory effect of PS. The binding site at the subunit interface was also confirmed by Chen et al. (Reference Chen, Wells, Arjunan, Tillman, Cohen, Xu and Tang2018), who obtained the structure of a chimeric receptor (ELIC’s ECD and α₁’s subunit TMD, homomeric) with bound alphaxalone. The neurosteroid molecule was located between the principal side’s α₁T306 and complementary side’s α₁Q242 (Figure 15B). To further verify the location of the neurosteroid-binding sites, Ziemba et al. (Reference Ziemba, Szabo, Pierce, Haburcak, Stern, Nourmahnad, Halpin and Forman2018) used the SCAM in oocytes expressing α₁β₃γ₂ receptors, with single point mutations aimed at residues located at the M2 and M3 helices (Figure 15A and B: β₃T262, β₃T266, β₃M283, β₃Y284, β₃M286, β₃G287, β₃F289, β₃V290, β₃F293, β₃L297, β₃E298, and β₃F301). Among these, the β₃Y284C mutation caused receptors to be nonfunctional, whereas all other mutations, except β₃T262C, β₃M283C, β₃V290C, and F β₃301C, affected GABA EC₅₀ and/or efficacy, and β₃F289C additionally increased spontaneous activity. Most importantly, the β₃F284C, β₃F2893C, and β₃L297C mutations reduced alphaxalone-induced potentiation, and in the case of β₃V290C, β₃F293C, β₃L297C, and β₃F301C, alphaxalone affected protein modification by 4-Chloromercuribenzenesulfonate (pCMBS). These data showed that the location of the neurosteroid-binding site at the bottom TMD determined by structural studies was correct. Another study confirmed that the PS-binding site consists of M3 and M4 helices of a single subunit, and this intrasubunit-binding site is functional not only within α subunits but also in β subunits (Seljeset et al. Reference Seljeset, Bright, Thomas and Smart2018).

In 2023, a number of GABA_AR structures with bound neurosteroids were published, shedding light on the details of their binding sites and possible mechanisms of modulation. Sun et al. (Reference Sun, Zhu, Clark and Gouaux2023) obtained the structures of the receptor in α₁β₃γ₂ and novel α_2/3β_1/2α₁β_1/2γ₂ and α₁β_1/2α_2/3β_1/2α₁γ₂ (mixture of different types of α and β subunit in single receptor, isolated from mouse brain) assemblies. In addition to the unexpected subunit composition, these receptors contained bound endogenous allopregnanolone molecules, indicating their abundance in the mouse brain. Neurosteroids were present in the intersubunit-binding site at the β/α subunit interface between M3 and M1 α-helices, in agreement with previous studies. Other receptor structures (α₁β₃γ₂ stoichiometry) have been presented by Legesse et al. (Reference Legesse, Fan, Teng, Zhuang, Howard, Noviello, Lindahl and Hibbs2023). The authors confirmed localization of the PAM neurosteroid-binding site at the subunit interface (lined by, e.g., α₁Q242, α₁W246, β₂L301, and β₂Y304, Figure 15B). Surprisingly, NAM neurosteroids (PS and DHEAS) were found at the ion pore (between the 9′ and -2′ residues, Figure 13), rather than at the previously proposed intrasubunit sites between the M3 and M4 helices (e.g., Laverty et al. Reference Laverty, Thomas, Field, Andersen, Gold, Biggin, Gielen and Smart2017; Mortensen et al. Reference Mortensen, Xu, Shehata, Krall, Ernst, Frølund and Smart2023; Seljeset et al. Reference Seljeset, Bright, Thomas and Smart2018). A summary of NS-binding sites can be found in Table 3.

Table 3. Localization of the TMD neurosteroid-binding site

Epipregnanolone and pregnanolone sulfate are NAMs, allopregnanolone is PAM, but its action via β intrasubunit site is NAM-like.

Additional PAM intrasubunit-binding sites in the top part of the TMD

Photolabeling and patch-clamp studies were performed to verify the location and number of neurosteroid-binding sites (Chen et al. Reference Chen, Bracamontes, Budelier, Germann, Shin, Kathiresan, Qian, Manion, Cheng, Reichert, Akk, Covey and Evers2019). They focused on α₁β₃ receptors and investigated the action of allopregnanolone and its photolabeling derivatives. Their findings provided evidence that within the receptor, there are three binding sites for non-sulfated steroids. The intersubunit-binding site is located at the bottom of the TMD and is formed by the M3 helix of the principal subunit and the M1 helix of the complementary M1 subunits (photolabeled residues: β₃L294 and β₃G308, binding site with previously examined α₁Q242). In addition, two intrasubunit sites were located – one within the α subunit (photolabeled residues: α₁N408A and α₁Y415F, with mutants affecting potentiation and direct activation: α₁V227W, α₁N408A, and α₁Y411F – see Figure 15B) and one within the β subunit (photolabeled residues: β₃Y442, β₃V278-A280, with no mutants affecting the neurosteroid action, Figure 15B). Since there were no mutants of the β₃ intrasubunit site that reduced the allopregnanolone effect, this site is probably not functional in the α₁β₃ receptor assembly. As the location of neurosteroid-binding sites has been confirmed using multiple methods, the next step was to determine the roles of the residues of these sites. Photolabeling of the ELIC-α₁ chimeric receptor by allopregnanolone-based reagents confirmed neurosteroid binding to both intra- and intersubunit-binding sites (photolabeled residues: α₁N408, α₁Y415, and α₁Y309, Figure 15B), but surprisingly, mutations α₁Q242L, α₁Q242W, and α₁W246L did not affect the photolabeling efficiency and addition of the allopregnanolone decreased photolabeling, indicating that neurosteroid binding was still possible (Sugasawa et al. Reference Sugasawa, Bracamontes, Krishnan, Covey, Reichert, Akk, Chen, Tang, Evers and Cheng2019). In addition, mutations decreased the thermal stabilization of the protein induced by neurosteroid (measured by UV absorbance intensity in size exclusion chromatography), which is proposed to be correlated with the modulatory effect of the neurosteroid. The authors then checked which residues were photolabeled by the neurosteroid analog (K220) in mutated receptors; in each case (WT and each mutant) α₁N408 (Figure 15B) was labeled, but α₁Y309 only in the WT receptor. Interestingly, for α₁Q242L (Fig. 15B) and α₁Q242W mutations, residue α₁F298 was labeled and in α₁W246L mutant-α₁F298 or α₁S299 were labeled instead of α₁Y309 (Figure 15B). Residue α₁F298 is located significantly above α₁Y309 what together with previous data (mutations not affecting the photolabeling total efficiency) indicates, that mutations of α₁Q242 and α₁W246 are not affecting the neurosteroid affinity, but rather the madulator’s orientation within in the binding site. A summary of the proposed binding site locations is presented in Table 3.

Mechanism of neurosteroid action

So far, structural and electrophysiological studies have identified the locations of neurosteroid-binding sites, but the exact mechanisms of action of these compounds remain elusive. Sugasawa et al. (Reference Sugasawa, Cheng, Bracamontes, Chen, Wang, Germann, Pierce, Senneff, Krishnan, Reichert, Covey, Akk and Evers2020) examined different neurosteroids, including allopregnanolone (3α5α), epipregnanolone (3β5β), and photolabeling analogs (KK148 and KK150; for a detailed description, see Jiang et al. Reference Jiang, Shu, Krishnan, Qian, Taylor, Covey, Zorumski and Mennerick2016). Interestingly, only allopregnanolone potentiated GABA-induced responses, and all of them enhanced (KK148 and KK150 at low level) muscimol binding at the orthosteric binding site, indicating at least two independent mechanisms of modulation. In addition, the application of epipregnanolone during the steady-state phase of the current response induced by high GABA concentrations enhanced the extent of macroscopic desensitization. The latter effect was abolished by the α₁V256S(r) (Figure 13) mutation, which also prevented the inhibitory action of sulfated neurosteroids (PS). Photolabeling indicated that all tested compounds were binding at intrasubunit sites, and all except epipregnanolone were found in the intersubunit site. Altogether, it has been shown that different neurosteroids bind to different sites and exert a mixture of effects on the receptor, including potentiation of the GABA-induced response, upregulation of macroscopic desensitization, and/or muscimol binding enhancement. To attribute these effects to the respective binding sites, the following mutants were examined: α₁Q242L (intersubunit site, Figure 15B), α₁N408A, α₁Y411F, αV227W (α intrasubunit site, Figure 15B), and β₃Y284F (β intrasubunit site, Figure 15B). Allopregnanolone enhanced muscimol binding by interacting with all three sites (mostly the intersubunit), but epipregnanolone and KK148 acted only via both intrasubunit sites. In addition, the 3β5β effect on desensitization was decreased by mutations in both intrasubunit sites. To check whether 3α5α also can enhance desensitization, the background mutation α₁Q242L (to abolish the function of intersubunit site) was introduced. Interestingly, enhancement of desensitization by allopregnanolone was observed and was abolished by mutation in the β intrasubunit site. In summary, the authors proposed that allopregnanolone stabilizes the open state of the receptor by binding to the β/α and α intersubunit sites, and the desensitized state by binding to the β intrasubunit site. In contrast, epipregnanolone stabilizes only the desensitized state by binding to both intrasubunit sites. KK148 stabilized the desensitized state via intrasubunit sites and all states equally via the intersubunit site, and K150 stabilized all states equally via all binding sites. Owing to the novel photolabeling agent ([³H]21-pTFDBzoxy-AP), Jayakar et al. (Reference Jayakar, Chiara, Zhou, Wu, Bruzik, Miller and Cohen2020) were able to detect other residues forming the intersubunit-binding site: β₃P415, β₃L417, β₃T418 (at M4 helix bottom), and β₃R309 (at M3 helix bottom). Using multiple types of neurosteroids, they confirmed that the presence of a free 3α-OH group is crucial for binding at this interface, and that the affinity is affected by the substitution at C17. In addition, they described various effects of neurosteroid NAMs on the modulation of muscimol binding: epipregnanolone and betaxalone (epimer of alphaxalone) upregulated the binding, epiallopregnanolone and PS had no effect, and dehydroepiandrosterone sulfate (DHEAS) showed a downregulating action. Electrophysiological recordings (on α₁β₂γ₂ receptors expressed in oocytes) allowed for a more accurate description of neurosteroids effects on gating (Pierce et al. Reference Pierce, Germann, Steinbach and Akk2022). Both PS and DHEAS reduced the current amplitude, caused faster and deeper macroscopic desensitization when coapplied with GABA, and enhanced the desensitization extent when applied at the steady-state phase of the response to GABA at saturating concentrations. Interestingly, the application of the NAM neurosteroids during the steady state (deepening the desensitization) showed also its own ‘desensitization’ (shift toward higher current values) and a rebound current once application was finished. In addition, the recovery of the current amplitude after subsequent application of GABA was faster if the first one was a co-application of GABA and NAM neurosteroid. They proposed that the closed state to which receptor transitions occur due to PS or DHEAS modulation is not a canonical desensitized state but a novel short-lived shut state. This was further supported by MWC type of kinetic modeling, in which this additional non-conducting state was incorporated. Modeling also indicated that both PS and DHEAS bind to the same site. Tateiwa et al. (Reference Tateiwa, Chintala, Chen, Wang, Amtashar, Bracamontes, Germann, Pierce, Covey, Akk and Evers2023) showed that the enantiomer of allopregnanolone, but not of pregnanolone, has lower potency due to different binding pose (180° rotation) at the intersubunit site.

Based on multiple GABA_AR structures with bound neurosteroids, Sun et al. (Reference Sun, Zhu, Clark and Gouaux2023) proposed that the modulatory mechanism of neurosteroids relies on enforcing the anticlockwise rotation of the TMD, thereby stabilizing the open conformation of the ion pore. Using structural methods, Legesse et al. (Reference Legesse, Fan, Teng, Zhuang, Howard, Noviello, Lindahl and Hibbs2023) noticed that allopregnanolone affected ion pore residues; in the GABA-bound structure (without neurosteroid), β₂E270 points toward α₁N275, whereas in the neurosteroid-bound structure, it moves toward β₂H267 (Figure 13), allowing for increased pore dilation in the top region. Based on molecular dynamics simulations, the authors also proposed that PAM neurosteroids influenced the TMD by lowering the energy barrier of ion movement and ECD via an allosteric mechanism by reducing its spread and making GABA molecules less likely to unbind.

Negative modulation by various PS analogs was examined by Mortensen et al. (Reference Mortensen, Xu, Shehata, Krall, Ernst, Frølund and Smart2023) in HEK293 cells expressing α₁β₃γ₂ receptor. Whereas all investigated compounds retained the general effect of lowering the response amplitude, their effects on macroscopic desensitization differed. Molecular modeling allowed the authors to assess the protonation states of the ligands and to conclude that uncharged analogs had a higher potency to reduce both the amplitude (at steady state) and the macroscopic desensitization time constant. Molecular dynamics simulations indicated that charged analogs interacted with α₁F295(r) residue via cation–π interaction and moved closer toward α₁F399(r) (Figure 15C). α₁F295A and α₁F399A(r) mutations reduced the inhibitory potency of the selected charged and uncharged analogs, but increased only the charged compound effects on the desensitization time constant. To further examine the effects of neurosteroids on desensitization, the authors utilized receptor mutants with decreased (α₁V296L(r)) and increased (γ₂V262F) desensitization, and selected neurosteroids showed varied potencies depending on the type of incorporated mutation, indicating that the uncharged analog is more likely to bind to the receptor in the desensitized state. This finding was further explored using kinetic modeling, which showed that the uncharged analog binds to both GABA-bound closed and desensitized states, whereas the charged analog binds only to the former. In contrast to the results of the study by Mortensen et al. (Reference Mortensen, Xu, Shehata, Krall, Ernst, Frølund and Smart2023), Legesse et al. (Reference Legesse, Fan, Teng, Zhuang, Howard, Noviello, Lindahl and Hibbs2023) located the binding site of PS in the ion pore, indicating, that the mechanism of action of NAM neurosteroids involves channel blocking, rather than negative modulation.