Discrete M3-M4 Intracellular Loop Subdomains Control Specific Aspects of γ-Aminobutyric Acid Type A Receptor Function*

The GABA type A receptor (GABAAR) is a member of the pentameric ligand gated ion channel (pLGIC) family that mediates ionotropic neurotransmission. Residues in the intracellular loop domain (ILD) have recently been shown to define part of the ion permeation pathway in several closely related members of the pentameric ligand gated ion channel family. In this study, we investigated the role the ILD of the GABAAR α1 subunit plays in channel function. Deletion of the α1 ILD resulted in a significant increase in GABA EC50 and maximal current amplitude, suggesting that the ILD must be intact for proper receptor function. To test this hypothesis, we conducted a mutagenic screen of all amino acids harboring ionizable side chains within this domain to investigate the contribution of individual charged residues to ion permeation. Using macroscopic and single channel voltage-clamp recording techniques, we found that mutations within a subdomain of the α1 ILD near M3 altered GABA apparent affinity; interestingly, α1(K312E) exhibited reduced partial agonist efficacy. We introduced point mutations near M4, including α1(K383E) and α1(K384E), that enhanced receptor desensitization. Mutation of 5 charged residues within a 39-residue span contiguous with M4 reduced relative anion permeability of the channel and may represent a weak intracellular selectivity filter. Within this subdomain, the α1(K378E) mutation induced a significant reduction in single channel conductance, consistent with our hypothesis that the GABAAR α1 ILD contributes directly to the permeation pathway.

The GABA type A receptor (GABA A R) is a member of the pentameric ligand gated ion channel (pLGIC) family that mediates ionotropic neurotransmission. Residues in the intracellular loop domain (ILD) have recently been shown to define part of the ion permeation pathway in several closely related members of the pentameric ligand gated ion channel family. In this study, we investigated the role the ILD of the GABA A R ␣1 subunit plays in channel function. Deletion of the ␣1 ILD resulted in a significant increase in GABA EC 50 and maximal current amplitude, suggesting that the ILD must be intact for proper receptor function. To test this hypothesis, we conducted a mutagenic screen of all amino acids harboring ionizable side chains within this domain to investigate the contribution of individual charged residues to ion permeation. Using macroscopic and single channel voltage-clamp recording techniques, we found that mutations within a subdomain of the ␣1 ILD near M3 altered GABA apparent affinity; interestingly, ␣1(K312E) exhibited reduced partial agonist efficacy. We introduced point mutations near M4, including ␣1(K383E) and ␣1(K384E), that enhanced receptor desensitization. Mutation of 5 charged residues within a 39-residue span contiguous with M4 reduced relative anion permeability of the channel and may represent a weak intracellular selectivity filter. Within this subdomain, the ␣1(K378E) mutation induced a significant reduction in single channel conductance, consistent with our hypothesis that the GABA A R ␣1 ILD contributes directly to the permeation pathway.
The GABA type A receptor (GABA A R) 2 is a chloride permeable ion channel that mediates inhibitory neurotransmission and is a member of the Cys-loop family of pentameric ligand gated ion channels (pLGIC). This family of receptors is ubiquitous throughout the nervous system and is essential for neural signaling. Cationic family members include the ionotropic serotonin receptor (5-HT 3 R), the nicotinic acetylcholine receptor (nAChR), the zinc-activated channel, and two bacterial homologs from Gloeobacter violaceus and Erwinia chrysanthemi (1); anionic channels include the glycine receptor (GlyR), heteromeric GABA A receptor, and homomeric GABA C receptor. Each of the five subunits of the pentamer has a modular arrangement with an extracellular ligand binding domain (LBD), four ␣-helical transmembrane domains (TMDs, deemed M1-M4), and a large intracellular loop domain (ILD) connecting M3 and M4. This arrangement of each subunit with respect to the membrane was first postulated in nAChR from Torpedo californica (2) and has recently been confirmed at 2.9 Å resolution in the prokaryotic homolog from G. violaceus (3).
Within the pLGIC family, the major determinants of ligand binding, channel gating, ion conductance, and charge selectivity have been shown to be controlled by residues within the LBD and the TMD. For instance in GABA A Rs, the GABA binding site exists at the two extracellular interfaces between ␣ and ␤ subunits (4), and the structural determinants of gating are predicted to be the pre-M1 segment, M2-3 linker, and loops 2, 7, and 9 at the interface of the LBD and TMD within each subunit (5)(6)(7)(8). Ion permeation and selectivity are proposed to be determined at sites along the central pore in the extracellular vestibule and along M2 (9 -15). These canonical rings of charged residues, contributed by each of the five subunits, have been hypothesized to interact with permeant ions to control the magnitude of channel conductance and to determine which charge species transverse the membrane.
The ILD, on the other hand, is known to play a role in receptor trafficking and activity through control of surface expression (16 -22). The ILD is the substrate for interactions with accessory proteins (19,(23)(24)(25) and kinases that covalently modify the receptor to alter function (17,18,26,27). In contrast to the TMD and LBD, much less is known about the structure of the ILD. The determination of a structure for the nAChR from Torpedo marmorata suggests that M4 is contiguous with a 25-residue membrane-associated helix (denoted MA); however, the remaining ϳ100 residues in the ILD were disordered (28). In addition, this domain contains the highest degree of primary sequence divergence both in terms of domain length and residue identity (Fig. 1). Furthermore, the prokaryotic pLGIC subunits lack an ILD, with M3 and M4 being connected by a short (Ͻ10 residue) linker.
Recent mutagenesis studies in cationic nAChR and homomeric GlyRs have uncovered a role for intracellular charged residues in determining channel conductance (29 -34). This was first determined by comparing the 5-HT 3A and 5-HT 3B subunits of the 5-HT 3 R that have an ϳ40-fold difference in conductance to identify the MA residues, which control this phenomenon (32,35) (Fig. 1). The lack of homology in the ILD makes it difficult to extrapolate these findings across the entire pLGIC family, and the ILD role in permeation has yet to be established for the GABA A R. Furthermore, these studies have focused solely on the MA stretch and have largely ignored the portion between M3 and MA. It, therefore, remains to be seen if the ILD component of permeation is conserved within heteromeric GABA A Rs.
Based on previous work, we hypothesized that the ILD of the GABA A R plays an integral role in defining a novel segment of the permeation pathway. To test this, we chose to focus on the ␣1 subunit. The M3-M4 loops of ␤ and ␥ subunits are known to make important and potentially confounding interactions with trafficking and scaffold proteins such as AP2, GABARAP, gephyrin, BIG-2, and GRIP-1 (17,18,21,23). Therefore, in this study we chose to modify only the ␣1 subunit, leaving the ␤2 and ␥2s subunits intact to establish a role for the ␣1 ILD in controlling GABA A R activity, not its involvement in trafficking and signaling pathways. Specifically, we characterized deletion constructs of the ␣1 ILD and also carried out a systematic sitedirected mutagenesis charge-switch scan, changing basic residues to glutamate and acidic residues to lysine to assay the role that individual charged ␣1 ILD residues play in controlling channel function. Our results indicate that subdomains of the M3-M4 intracellular loop are indeed responsible for controlling separate components of channel function.

EXPERIMENTAL PROCEDURES
Mutagenesis and Cell Culture-Wild type GABA A R complementary DNAs (cDNAs) for the human ␣1, ␤2, and ␥2s sub-units were expressed via the pCIS2 vector with a CMV viral promoter in HEK293 cells (American Type Culture Collection). Deletions and point mutations in the wild type ␣1 subunit cDNA were introduced using the QuikChange site-directed mutagenesis kit (Stratagene) and confirmed by sequencing (MWG/Operon). It is noteworthy to mention that the ␤2 homolog of ␣1(D393) has been shown to be essential for surface expression and, therefore, was preserved in all constructs (36). HEK293 cells were maintained at 37°C and 5% CO 2 , in Eagle minimum essential medium supplemented with 5% FBS (Hyclone), 40 M L-glutamine, 100 units/ml penicillin, and 0.1 mM streptomycin. Cells were passaged weekly with trypsin treatment up to 20 times. Seventy-two h after plating onto poly-D-lysine-coated glass coverslips, HEK293 cells were co-transfected with cDNAs for the GABA A R ␣1, ␤2, and ␥2s subunits as well as adeno-associated virus GFP via a calcium phosphate method (37). Reagents were purchased from Sigma unless otherwise noted.
Whole-cell Recordings-Wild type and mutant GABA A receptors were characterized via whole-cell, voltage-clamp electrophysiology at room temperature 36 -72 h after transfection. Patch pipettes were fabricated from thin-walled borosilicate glass (TW150F-4, World Precision Instruments, Inc.) using a horizontal puller (P-97, Sutter Instrument Co.). The resistance of patch pipettes was 2-5 megaohms when filled with an intracellular solution containing 120 mM KCl, 2 mM MgCl 2 , 10 mM EGTA, and 10 mM HEPES adjusted to pH 7.2 with NaOH. Cells were perfused continuously with an extracellular solution containing 160 mM NaCl, 10 mM HEPES, 6 mM D-glucose, 3 mM KCl, 1 mM MgCl 2 , 1.5 mM CaCl 2 adjusted to pH 7.4 for a near symmetrical chloride gradient with a theoretical reversal potential of Ϫ7.80 mV. Changes in ion selectivity were addressed using extracellular ion replacement solutions. The low sodium saline contained 14 mM NaCl, 146 mM N-methyl-D-glucamine, 10 mM HEPES, 6 mM D-glucose, 3 mM KCl, 1 mM MgCl 2 , 1.5 mM CaCl 2 , and the low chloride saline contained 5 mM NaCl, 155 mM sodium gluconate, 10 mM HEPES, 6 mM D-glucose, 3 mM KCl, 1 mM MgCl 2 , 1.5 mM CaCl 2 ; both solutions were adjusted to pH 7.4. Two 10-channel infusion pumps (KD Scientific) and a rapid solution exchanger (Biologic) were used to apply extracellular salines to the cells. For more rapid solution exchange, to assess the fast desensitization of channel gating, large (ϳ10 picofarads) outside-out patches were pulled and positioned in the outflow of the solution changer. Wholecell currents were recorded at Ϫ60 mV, digitized at 200 Hz, and filtered at 100 Hz with a MultiClamp 700B amplifier and Digi-Data 1322A interface (Molecular Devices). The solution changer was driven by protocols written in pClamp 9.2 (Molecular Devices). To assess current-voltage properties of each receptor construct, a voltage ramp, from Ϫ60 to ϩ60 mV and back to Ϫ60 mV (0.24 mV/ms for total ramp duration of 0.5 s) was applied within the plateau response to each concentration of GABA and in the washout (subtracted to control for basal response of the membrane). Data were gathered from at least eight cells for each mutation from at least two independent transfections. To ensure consistency, several wild type responses were recorded from each transfection. Each wholecell patch was exposed to eight increasing concentrations of , and 5HT3B (NP_006019)). Flanking TMD residues from M3 and M4 are shown as white letters highlighted in black. Charged residues in the GABA A R ␣1 subunit that were mutated in this study are marked with asterisks, and aligned residues are highlighted in gray to show conservation. Putative gephyrin binding sites are underlined with a dotted line (24,25), residues shown to control conductance in 5HT 3 R are boxed (32) and the MA domain for nAChR is underlined (28).
GABA in separate traces with and without the voltage ramp. To determine desensitization from excised patches, 3 mM GABA was applied for 10 s and washed out for 20 s between sweeps.
Whole-cell Analysis-Analysis of recordings was carried out using MatLab (The Math Works, Inc.). Peak currents (I) from GABA exposures spread over 3.5 logarithmic decades were fit to the Hill equation (Equation 1); each mutation was compared with WT using three metrics: GABA apparent affinity (EC 50 ) current magnitude (I max ) and Hill coefficient (nH).
The desensitizing portion of currents elicited with saturating concentrations of GABA (1-3 mM) was defined as the time period from 200 ms after GABA onset to 200 ms before GABA was removed. Any traces still showing a rising phase during this time period were not included in this analysis. This segment was fit with an exponential equation (Equation 2) to determine a weighted rate constant (w) for desensitization (Equation 3).
y͑t͒ ϭ a*e bt ϩ ce dt The S.E. (s) for normalized means (X) was determined from the raw mean () and standard error () values using Equation 4 for WT and Equation 5 for mutants.
Reversal potential (E rev ) values in normal and each ion replacement solution were calculated from ramp data by interpolating the membrane potential when I ϭ 0. Theoretical reversal potentials were determined from the Nernst equation (Equation 6) assuming membranes were permeable only to chloride. At room temperature, 25°C, ϭ RT/F ϭ 25.69 mV (38).
Ramp data were collected in a pair-wise fashion for normal and ion replacement solutions so that relative shifts in reversal potential, E NS and E CS , induced by low sodium and low chloride solutions, respectively, could be determined for each cell. We determined relative permeabilities for gluconate (P G /P Cl ) and sodium (P Na /P Cl ) from the relative reversal potential shifts with the following manipulations of the Goldman-Hodgkin-Katz voltage equation, (Equations 7-10), Subscripts refer to the intracellular (i) and extracellular (o) ion concentrations in standard (S) low sodium (N) and low chloride (C) solutions.
Single Channel Recordings and Analysis-Single channel currents elicited from saturating GABA concentrations (3 mM) were recorded at Ϫ80 mV and filtered at 8 kHz with an Axopatch 200B amplifier (Molecular Devices) in the outsideout patch configuration. Data were digitized at 40 kHz using pClamp 9.0 and 10.0 (Molecular Devices). Patch pipettes were pulled from thick wall borosilicate glass (G150F-4, Warner Instruments) using a vertical puller (PC-10, Narishige). Pipette tips were coated with Sylgard-184 (Dow Corning) then fire polished to a final resistance of 8 -11 megaohms. For analysis, recordings were low-pass filtered at 1 kHz using QuB software and idealized with SKM (39). Unitary current amplitudes were determined from a Gaussian fit of an all-points histogram (0.1 pA/bin) of single channel records that contained no overlapping openings.
To determine burst open probability (P open ), clusters of bursts were first segmented by hand. Using MIL (40), separate bursts were defined and then separated into segments using ChpId (41). Finally, the open probability of each segment was calculated.
Statistical Analysis-All data metrics were compared with wild type receptors by multivariate analysis of variance using SPSS 17.0 software (IBM Co.) and MatLab (The Math Works, Inc.). Significance was held at p Ͻ 0.05 for Student's t test, pairwise test, and Dunnett's post hoc analysis. Tests compared each mutant response individually to the wild type response.

RESULTS
Deletion of the ␣1 ILD Altered the Apparent Affinity for GABA-We created and characterized two different ␣1 ILD deletion constructs ( Fig. 2A). To compare the function of the GABA A R ␣1 subunit with a prokaryotic homolog, we replaced 80 ␣1 ILD residues with the heptapeptide from G. violaceus (IL-BAC). In addition, we replaced the ILD with seven lysine residues to investigate the effect of charge on ion permeation and channel gating (IL-7K). We assessed IL-BAC and IL-7K for changes in the GABA concentration response relative to WT receptors with macroscopic whole-cell recordings from transfected HEK cells (Fig. 2). Replacement of ␣1 ILD with the bacterial residues SQPARAA and the positive charges KKKKKKK, both, caused a significant decrease in GABA apparent affinity. EC 50 values for WT, IL-BAC, and IL-7K were 51.6 Ϯ 14.6, 144 Ϯ 26.0, and 660 Ϯ 172 M, respectively (means Ϯ S.E. of the mean (S.E.), n ϭ 6). There were no changes induced in Hill slope, but both deletion constructs significantly increased current magnitude. IL-7K and IL-BAC elicited an ϳ2 and ϳ3-fold increase in current magnitude with values of Ϫ6250 Ϯ 1210 and Ϫ9810 Ϯ 462 pA, respectively, compared with the WT current value of Ϫ3230 Ϯ 308 pA (n ϭ 6).

Point Mutations Near M3 Decreased GABA Apparent
Affinity-We mutated acidic residues to lysine and basic residues to glutamate in a point-wise fashion. Of the 80 ␣1 ILD residues, 17 are positively charged, and 8 are negatively charged for a total of 25 point mutations. Most of these mutations did not significantly alter the apparent affinity of the receptor for GABA, the maximal response to GABA, or the Hill coefficient for the GABA concentration response relationship (Table 1). However, 6 of the 25 mutants showed a significant right shift in concentration response with an ϳ2-3-fold increase in GABA EC 50 . In contrast, even the largest increase in apparent affinity, observed with ␣1(K329E), was not significantly different from WT (all tests, p Ͼ 0.05, Fig. 3). All but one of the significant shifts in apparent affinity is located in a 20-residue span between positions 312 and 332, the ILD segment contiguous with M3. Charge neutralization also shifted EC 50 values (data not shown). The ␣1(K312E) charge-switch mutation caused a 2.0 Ϯ 0.6-fold increase in EC 50 , whereas the ␣1(K312A) charge neutralization mutation only increased EC 50 1.4 Ϯ 1.7-fold, which was not significantly different from WT (Student's t test, p Ͼ 0.05; n ϭ 3). Interestingly, a significant decrease in EC 50 (14.1 Ϯ 4.69 M) was caused by the ␣1(D332A) charge neutralization mutation compared with WT (Student's t test, p Ͻ 0.05; n ϭ 5).
The ␣1(K312E) Mutation Exhibited Impaired Gating-A shift in EC 50 may be caused by changes in receptor gating or a change in structure propagated globally to alter the GABA binding site. Therefore, we constructed concentration response relationships from the same cell for both the partial agonist piperidine-4-sulfonic acid (P4S) and the full agonist

TABLE 1 Point-wise mutation of charged resides in the ␣1 ILD significantly increased apparent affinity and current magnitude
The indicated lysine or arginine residues were mutated to glutamate. The indicated glutamate or aspartate residues were mutated to lysine. Maximal current (I max ), GABA apparent affinity (EC 50 ), and Hill coefficient (nH) values were determined from fitting peak current values to the Hill equation (see "Experimental Procedures").

Function of GABA A R Intracellular Subdomains
NOVEMBER 4, 2011 • VOLUME 286 • NUMBER 44 GABA (Fig. 4). Cells expressing the ␣1(K312E) mutation yielded an ϳ2-fold increase in EC 50 with no other significant changes in the GABA concentration response (Table 1). P4S exhibited decreased efficacy in both WT and mutant receptor containing cells; however, the relative agonist efficacy (I P4S / I GABA ) for ␣1(K312E) was significantly decreased compared with WT, 0.080 Ϯ 0.035 and 0.190 Ϯ 0.022 respectively (Stu-dent's t test p Ͻ 0.05; n ϭ 3; Fig. 4). This change in I P4S /I GABA is consistent with a gating deficiency being introduced by the ␣1(K312E) mutation.

Mutation of ILD Charged Residues Increased the Speed of Receptor
Desensitization-Desensitization is characterized by channel closure during the prolonged application of agonist. To determine whether charge-switch mutations induced changes in the desensitization component of gating, each macroscopic trace was fit with an exponential equation to determine a weighted time constant from currents elicited by maximally effective concentrations of GABA (see "Experimental Procedures"). Seven mutations had significantly enhanced rates of desensitization compared with WT with up to an ϳ3-fold increase induced by the ␣1(K384E) mutation (Fig. 5A). To confirm these results were not confounded by slow solution exchange rates, we used a rapid application technique to verify one of the mutants (Fig. 5B); the ratio of weighted rate constants between WT and the ␣1(K378E) mutation remained constant at 2.5-fold when determined from both whole-cell and excised patch-generated traces.
Current Voltage Relationship Confirms Anion Selectivity of WT Receptor-To determine the relative permeability of sodium and chloride in WT receptors, we replaced Na ϩ and Cl Ϫ ions with large impermeable ions of the same valence, N-methyl-D-glucamine and gluconate, respectively. WT receptor-containing cells had an experimentally determined reversal potential of Ϫ3.41 Ϯ 1.85 mV in normal saline (n ϭ 34; Fig. 6). There was a ϩ3.82 Ϯ 0.49-mV relative shift from normal to low sodium extracellular saline (E NS ), which was not significantly different from the normal saline and confirmed that the cation permeability of the channel was negligible (n ϭ 17; paired t test p Ͼ 0.05; Fig. 6A). Decreasing the extracellular chloride concentration caused a ϩ35.8 Ϯ 2.66-mV shift in reversal potential relative to the IV response in normal saline (E CS ), which verified  that the WT GABA A R was an anion-permeable channel (n ϭ 17; paired t test p Ͻ 0.05; Fig. 6B). The theoretical shift in reversal potential that should have been caused by decreasing the extracellular chloride was ϩ65.7 mV assuming the channel was impermeable to gluconate. This experimental deviation may be caused by a relative permeability to gluconate (P G /P Cl ) of ϳ20%.
ILD Residues Contribute to Ion Permeation Pathway-We hypothesized that amino acids with ionizable side chains lining the pore lumen make important electrostatic interactions with permeant ions to establish the permeability of the pore. Therefore, switching the charge of such residues should perturb this interaction, causing a shift in reversal potential. For example, an increase in the relative permeability to sodium (P Na /P Cl ) will cause a negative relative shift in current reversal when the extracellular concentration of sodium is decreased. Of the 26 charge-switches made, none resulted in a negative E NS (Fig.  6C). Furthermore, none of the mutations were significantly dif-  . IV data were gathered in each of the ion replacement solutions so that the relative shift in reversal potential could be determined for each patch. Each trace was normalized by its maximal response to GABA at Ϫ60 mV in the normal saline, then the up and down component of each ramp was averaged. Data points depict the mean, and error bars depict the S.E. There was no significant difference in reversal between the normal and low sodium salines; paired t test, p Ͼ 0.05. The low chloride extracellular saline caused a significant shift in reversal potential; paired t test, p Ͻ 0.05. C, relative shifts in current reversal caused by low sodium versus normal (E NS ; white) and low chloride versus normal (E CS ; gray) extracellular salines are shown Ϯ S.E. Dotted lines represent the S.E. margins for WT, n ϭ 17; mutations, n ϭ 5; significance held at p Ͻ 0.05 for Dunnett's post hoc ( ‡) and Student's t test (*). NOVEMBER 4, 2011 • VOLUME 286 • NUMBER 44 JOURNAL OF BIOLOGICAL CHEMISTRY 37995 ferent from WT in regard to relative sodium permeability (Student's t test p Ͼ 0.05).

Function of GABA A R Intracellular Subdomains
Six charge-switch mutations located in both flanks of the ILD significantly increased E CS (Fig. 6C). For example, the ␣1(K320E) and ␣1(K378E) mutations exhibited E CS ϭ ϩ47.9 Ϯ 2.45 and ϩ46.5 Ϯ 2.61 mV relative shift values in low chloride, which may be caused by an ϳ50% decreased relative permeability to gluconate, P G /P Cl ϭ 0.10 Ϯ 0.02 and 0.09 Ϯ 0.02 respectively, compared with WT, P G /P Cl ϭ 0.21 Ϯ 0.03. Interestingly, although the ␣1(E372K) charge-switch mutation positively shifted E CS compared with WT (Fig. 6C), introduction of positive charge at the neighboring neutral ␣1(V373) position did not significantly shift the relative reversal potential in low chloride conditions compared with WT (Student's t test, p Ͼ 0.05; n ϭ 3; data not shown).
Mutations in ILD Altered Single Channel Properties-We observed that charge-switch mutations caused two main changes in channel function; that is, a gating impairment as observed by decreased apparent GABA affinity and a shift in reversal potential in low chloride extracellular saline accompanied by faster desensitization. We chose residues from both groups for further characterization. Specifically, ␣1(K328E) showed a significant increase in EC 50 , and ␣1(K378E) increased the speed of desensitization and altered anion permeability. We recorded single channel responses for WT and each mutation from membrane patches in the outside-out configuration and observed unique differences in single channel activity. First, channels harboring the ␣1(K328E) mutation exhibited an increase in mean burst open probability (Table 2). Second, compared with WT, channels harboring the ␣1(K378E) mutation exhibited an ϳ30% decrease in unitary current amplitude ( Fig. 7 and Table 2).
The ␣1(K378E) Mutant Decreased Conductance-To verify the difference in current amplitude between WT and ␣1(K378E), single channel currents were recorded from outside-out patches at membrane holding potentials from Ϫ80 to ϩ60 mV. The ␣1(K378E) mutation caused a decrease in chord conductance and a modest negative shift in reversal potential compared with WT (Table 2 and Fig. 8).

DISCUSSION
To date the importance of the GABA A R M3-M4 intracellular loop in defining ion channel activity has not been comprehensively addressed. Previous studies of pLGIC ILDs have focused on receptor surface stability, mobility (16,17,22), phosphorylation (26,27), binding interactions with accessory proteins (23), and trafficking machinery (19). Recent studies point to an intracellular component of desensitization (42)(43)(44), and intracellular control of conductance has been linked to the MA stretch (29,34), but this has yet to be confirmed for the GABA A R.
Deletion of the ILD-High resolution structures of prokaryotic pLGICs, which lack the M3-M4 intracellular loop, have brought into question the importance of the ILD for channel activation (1, 3). Deletion of this domain in homomeric 5-HT 3 R and GABA C R resulted in functional, surface-expressing receptors in oocytes, suggesting this domain is not necessary for channel activity (45). However, these constructs exhibited noteworthy changes in channel conductance and agonist apparent affinity, underscoring the importance of this domain for receptor activity. In our study, deletion of the GABA A R ␣1 ILD produced significant changes in channel function (Fig. 2), matching the effects of similar deletions in other pLGICs (45). At present, the theory of ion permeation in pLGICs relies on the

TABLE 2 Charge-switch mutations altered burst open probability and amplitude of single channel current amplitudes
Single channel recordings were gathered from outside-out patches held at Ϫ80 mV in extracellular saline containing 3 mM GABA; means are depicted Ϯ95% confidence interval of fit. Some values were not determined (ND) for both mutants.

Function of GABA A R Intracellular Subdomains
concept that canonical rings of charged residues make integral electrostatic interactions with permeant ions to control permeation and selectivity (10,12). Recent functional data in 5-HT 3 R, nAChR, and GlyR demonstrated that the ILD also plays a role in this process (29,31,32,35). Therefore, we hypothesized that deletion of the ␣1 ILD of the GABA A R would change the effective geometry of the permeation pathway. The results presented here show that the ILD of the GABA A R ␣1 subunit is involved in controlling ion flow across the membrane. The reduced rate of ion permeation observed with the IL-7K construct compared with the mostly neutral bacterial heptapeptide is consistent with the positively charged lysine residues providing a stronger attractive force on the permeant chloride ions. Deletion constructs also exhibited altered GABA apparent affinity as evidenced by large shifts in EC 50 . This suggests that the tertiary structure of the ILD must be intact for proper channel gating to occur. Taken together, the results of our deletion experiments established an important role for the GABA A R ␣1 ILD architecture in controlling the function of the ion channel, as has been observed with other members of the pLGIC family.
Subdomains of the ILD-The ILDs of GABA A R have a higher degree of variability compared with the LBD and the TMD in terms of side chain conservation and loop length. A simple sequence alignment of pLGIC ILDs highlights this variability but also suggests some local conservation (Fig. 1); the first ϳ20 and last ϳ40 residues of the ILD, i.e. those contiguous with M3 and M4, are regions of the protein that contain a large number of strongly conserved charged residues. We observed that charge-switches close to M3 predominantly resulted in receptors with altered gating. On the other hand, charge-switch mutants close to M4 resulted in receptors with enhanced desensitization. Importantly, mutations within both putative subdomains caused a shift in reversal potential in low chloride extracellular saline, which suggests the existence of an intracellular anion selectivity filter. We hypothesized that in the GABA A R these two membrane-associated subdomains, deemed M3A and M4A respectively, define a portion of the permeation pathway (Fig. 1).
It is important to note that a mutation that causes a change in agonist EC 50 may do so by altering either receptor gating or ligand binding. These properties are intrinsically linked as ligand binding and ion channel activation are separate events within a well defined Markovian process with many states (46,47). Binding studies and electrophysiological assays only give a readout of a subset of these states and cannot definitively isolate the individual contributions of different receptor states to mutant receptor function. Therefore, although the results presented here are consistent with changes in receptor gating, we cannot unambiguously rule out that some changes in ligand binding may have also occurred in the LBD as a result of the ILD mutation. Nevertheless, a comparison of the relative efficacy of full and partial agonists acting on WT and ␣1(K312E) containing receptors does suggest that the observed increase in EC 50 was likely due to a gating impairment (Fig. 4). GABA and P4S are structurally similar (Fig. 4A) and bind at the same orthosteric site, but the partial agonist is less efficacious in stabilizing the open state of the ion channel. Furthermore, this subdomain is continuous with M3, which has been shown to move during channel gating (48). Therefore, we propose that charged residues within M3A also undergo movement upon channel activation, and by switching these charges we have impeded this movement.
Mutation of M3A residues also induced shifts in the relative reversal potential caused by decreasing extracellular chloride. For example, the ␣1(K320E) charge-switch mutation positively shifted E CS compared with WT (Fig. 6). Again, this suggests that within the ␣1 subunit of the GABA A R, permeation pathway is not solely defined by residues in the LBD, the M1-M2 linker, and M2.
The M4A Subdomain-M4A was defined by a 39-residue region between positions ␣1(Arg-354) and ␣1(Lys-391) (Fig. 1). Charge-switches throughout the ILD, but mostly near M4, increased the speed of desensitization (Fig. 5). A similar effect has been observed in the 5-HT 3 R, where a single intracellular arginine near M4 has also been shown to control receptor desensitization (43). In our sequence alignment, this arginine is homologous to ␣1(Lys-383) (Fig. 1); the charge-switch mutation, ␣1(K383E), at this position induced a significant increase in desensitization rate (Fig. 5). Interestingly, the neighboring ␣1(K384E) mutation caused the largest fold increase in desensitization. These results suggest that ILD M4A residues modu- late specific components of channel gating, i.e. those that define the desensitization gate.
The mutation of charged residues near M4 has been shown to alter single channel conductance in homomeric 5-HT 3 R and GlyRs; in both cases the mutations were made within a putative homolog of the MA stretch (29,32). In our study ␣1(K378E) induced a significant decrease in single channel chord conductance (Figs. 7 and 8; Table 2). This result confirmed our hypothesis that ILD residues play a key role in coordinating ions as they pass along the permeation pathway and through the ion channel. The linear nature of the single channel chord conductance (Fig. 8) also reveals that there is no voltage dependence to the activation of the ion channel under the conditions used in this study. Although the simplest explanation for the decreased current amplitude is a reduction in the single channel conductance, other explanations are possible. For example, it is conceivable that the mutation alters the ability of the membrane potential to invade the permeation pathway. If the mutation alters the distance over which the membrane potential is distributed or focused, this might have the effect of decreasing local field strength in the permeation pathway, reducing the force on individual ions and, hence, reducing single channel current amplitude. Nevertheless, it is clear from these results that ␣1(K378) is intimately involved with mechanisms directing the permeation of ions through the GABA A R.
There were significant changes in channel permeability to anions caused by mutating GABA A R M4A-charged residues (Fig. 6). Within this subdomain, introduction of a lysine at position ␣1(V373) did not shift E CS , suggesting that changes in permeability were position-specific and not merely due to a global change in the electrostatic environment of the M4A subdomain. Interestingly, charged residues within both ILD subdomains were shown to control anion permeation. Increased relative reversal potentials translated to a decreased permeability to gluconate that may reflect a decrease in the diameter of the intracellular pore entrance. Alternatively, point mutations may have electrostatically occluded the pore to impair gluconate permeation. Our results are consistent with the hypothesis that intracellular residues of the GABA A R ␣1 subunit influence ion permeation and demonstrate that residues within both the M3A and M4A subdomain likely line the channel pore.
In summary, we have shown that the ILD must be intact for proper receptor activity to occur as the charged residues of the GABA A R ␣1 subunit ILD play a critical role in controlling channel gating and ion permeation. Furthermore, our results suggest the existence of novel functional subdomains of the ILD that mediate specific aspects of channel activity. We have shown that specific charged residues within both membraneassociated subdomains control different facets of channel gating; M3A residues contributed to agonist-dependent gating, whereas M4A residues mediated the time course of receptor desensitization. Positions within both subdomains were shown to line a novel portion of the pore lumen by influencing channel conductance and altering anion permeation. The GABA A R is defined by a diverse complement of subunits with unique properties, and future studies will be necessary to determine whether these ILD contributions to channel function are subunit-specific. Our findings have established a novel role for the M3-M4 intracellular loop in the ␣1 subunit of the GABA A R. Importantly, our findings taken together with other recent advances (29 -35) suggest a critical role for ILD-charged residues in pLGIC function.