Anterograde trafficking signals in GABAA subunits are required for functional expression.

Pentameric GABAA receptors are composed from 19 possible subunits. The GABAA β subunit is unique because the β1 and β3 subunits can assemble and traffic to the cell surface as homomers, whereas most of the other subunits, including β2, are heteromers. The intracellular domain (ICD) of the GABAA subunits has been implicated in targeting and clustering GABAA receptors at the plasma membrane. Here, we sought to test whether and how the ICD is involved in functional expression of the β3 subunit. Since θ is the most homologous to β but does not form homomers, we created two reciprocal chimeric subunits, swapping the ICD between the β3 and θ subunits, and expressed them in HEK293 cells. Surface expression was detected with immunofluorescence and functional expression was quantified using whole-cell patch-clamp recording with fast perfusion. Results indicate that, unlike β3, neither the β3/θIC nor the θ/β3IC chimera can traffic to the plasma membrane when expressed alone; however, when expressed in combination with either wild-type α3 or β3, the β3/θIC chimera was functionally expressed. This suggests that the ICD of α3 and β3 each contain essential anterograde trafficking signals that are required to overcome ER retention of assembled GABAA homo- or heteropentamers.

Introduction

Epilepsy, anxiety, neurodevelopmental disorders, and neuropsychiatric disorders collectively affect a significant proportion of the population. One common problem in these disorders is the dysfunction of the GABAA receptor, so it is important to understand how these receptors are functionally regulated. GABAA receptors are the target of anesthetics as well as drugs that are used as anticonvulsives, anxiolytics, and hypnotics [1]. They are pentameric ligand-gated chloride channels in the Cys-Loop superfamily of ligand-gated ion channels. Their role is to mediate fast inhibitory neuronal transmission. The pentameric ion channel is formed from a pool of 19 different GABAA subunits (α1-6, β1-3, γ1-3, δ, ϵ, θ, π, and ρ1-3) [1], with each subunit having the same general structure: a long N-terminal domain (NTD) which creates the extracellular ligand binding domain (LBD), 4 transmembrane domains (TMD) that create the ion channel pore, a variable length intracellular domain (ICD) composed of the TM3-TM4 loop, and a short extracellular C-terminus (CTD) [1,2]. Considering the vast number of possible subunit combinations that could exist, it is important to understand the cellular mechanisms that limit the functional expression (assembly and trafficking) of subunit combinations that might otherwise exist. Most of the known combinations require α and β subunits; however, the rules that govern assembly and regulate trafficking of the receptors are still poorly understood.

Canonical GABAA receptors contain 2 α subunits, 2 β subunits, and a third X subunit arranged counterclockwise around the central pore in the order: β-α-β-α-X; where X is typically a γ subunit; however, it is generally accepted that α, β, δ, or one of the other subunits can replace the γ [1]. Some of the structural elements that are involved in regulated assembly of compatible subunits have been identified, most commonly in the NTD binding loops that form the subunit-subunit interface [3-7]. Assembly of the β subunit is unique in that the β1 and β3 subunits can assemble and traffic to the plasma membrane as homomeric receptors, whereas most of the other subunits, including β2, are obligatory heteromers [8]. β1, β2 and β3 are at minimum 70% homologous but show differences in homomeric assembly. The critical determinant of β3 homomeric functional expression was identified as the so-called GKER sequence, located in the extracellular domain (ECD) binding loop F of β3, which corresponds to a DNTK sequence at the equivalent site in β2[7]. Using chimeric exchanges between β2 and β3, Taylor et. al. (1999) [7] showed that the GKER sequence was necessary for β3 and sufficient to rescue β2. This study implied that the determinants of functional expression are entirely contained in the NTD; however, there may be other determinants of functional expression that were not revealed in that study because they are sufficiently conserved across the β subunits. The TM3-TM4 intracellular loop (intracellular domain, ICD) has been implicated in targeting, anchoring, and clustering of the GABAA receptor at the plasma membrane; however, the mechanism of how this loop is involved in GABAA receptor functional expression is still largely unknown.

Here, we sought to test whether and how the ICD is involved in functional expression of the β3 subunit. We created two reciprocal chimeric subunits, one having the EC and TM domains from β3 and the ICD from θ, and the other having the EC and TM domains from θ and the ICD from β3. The θ subunit was chosen as the chimeric donor because it is most homologous to β but forms heteromeric and not homomeric receptors [9]. Moreover, the θ ICD shares virtually no homology with the β3 ICD or any other subunit. Because there is no evidence for ICD involvement in assembly, we hypothesized that the θ ICD would support the functional expression of β3 as either homomers or heteromers. We used immunofluorescence staining to test plasma membrane expression and whole-cell patch-clamp with fast perfusion to test receptor function.

Results indicate that neither the β3/θIC nor the θ/β3IC chimera can traffic to the plasma membrane when expressed alone, suggesting the β3 ICD is necessary for functional homomeric expression but is insufficient to traffic unassembled θ subunits to the plasma membrane. When expressed in combination with either wild-type α3 or β3, the β3/θIC chimera was rescued to the plasma membrane and the receptor functioned essentially like the wild type β-containing receptor, while the θ/β3IC chimera was not rescued. This suggests that the TM3-TM4 intracellular loops of α3 and β3 each contain essential anterograde trafficking signals that are required to overcome ER retention of assembled homo- or heteropentamers.

Results

The IC loop of θ disrupts β3 homomeric plasma membrane expression

To determine how the intracellular loop is involved in functional expression, we produced a reciprocal pair of chimeric receptors (β3/θIC and θ/β3IC) swapping the β3 and θ ICDs Figure 1(a). Unlike β3, θ subunits do not assemble or express as homomeric receptors on the plasma membrane [9]. To visualize expression, the wild-type β3 and both chimeric constructs were tagged with three consecutive hemagglutinin epitopes (HA) on the N-terminus between amino acids 5 and 6 (HAθ/β3IC) or 6 and 7 (HAβ3, HAβ3/θIC) of the mature protein. Western blotting analysis was performed to confirm that the chimeric constructs are full length and expressed with a similar relative abundance compared to either the HAβ3 or the HAθ construct. As expected, HAβ3/θIC was found to be a similar molecular weight as HAθ (HAβ3/θIC calculated MW = 74 kDa, actual MW = 91 kDa and HAθ calculated MW = 77 kDa, actual MW = 95 kDa). Likewise, HAθ/β3IC was found to be a similar molecular weight as HAβ3 (HAθ/β3IC calculated MW = 61 kDa, actual MW = 63 kDa and HAβ3 calculated MW = 58 kDa, actual MW = 64 kDa). GAPDH was used as a loading control and had a molecular weight of 36 kDa, as expected. Relative to GAPDH, the expression ratio of HAβ3 was 1.87, HAθ was 1.36, HAβ3/θIC was 1.53, and HAθ/β3IC was 1.61 Figure 1(b). Next, we used immunofluorescence to visualize and quantify subunit expression in non-permeabilized cells (surface) and after Triton-X permeabilization (total). When expressed alone in HEK293 cells, HAβ3 was clearly labeled on the surface of many cells, as shown by non-permeabilized staining with an anti-HA antibody Figure 1(c,d). In contrast, HAβ3/θIC was rarely found on the surface (max 1–5 cells per dish compared to ≥ 150 cells for HAβ3); but, when it was seen, the fluorescence intensity was comparable to HAβ3 (HAβ3/θIC = 14.3 ± 2.8 AU, n = 10 images (10 cells) and HAβ3 = 16.9 ± 0.6 AU, n = 6 images (46 cells)). HAθ/β3IC was never found to be labeled on the surface of cells (n = 4 images (0 cells)). Total expression levels of the HAβ3, HAβ3/θIC, and HAθ/β3IC constructs were comparable in terms staining intensity and cell number with the exception of HAθ/β3IC, which showed fewer cells labeled (HAβ3 = 26.9 ± 2.3 AU, n = 8 images (79 cells); HAβ3/θIC = 23.5 ± 3.6 AU, n = 11 images (97 cells); and HAθ/β3IC = 21.8 ± 3.7 AU, n = 9 images (31 cells)) when the cells were re-probed following permeabilization.

Figure 1.

Homomeric expression and function of β3 and chimera. a) Schematic of the HAβ3/θIC and HAθ/β3IC chimeras where the blue portions are from β3 and the red portions are from θ. b) Total protein Western blot of HAβ3 + HAθ, HAβ3/θIC, HAθ/β3IC, and EGFP transfected HEK293 cells. The membrane was probed with primary rabbit anti-HA Epitope Tag (1:5000 dilution) and rabbit anti-GAPDH (1:5000 dilution) and secondary goat anti-rabbit-HRP (1:5000 dilution). HA bands were quantified as fold-changes against GAPDH using ImageJ. Compared to GAPDH, the ratio of HAβ3 = 1.87; HAθ = 1.36; HAβ3/θIC = 1.53; HAθ/β3IC = 1.61. c) Representative IF images at 20x magnification of non-permeabilized (surface) and permeabilized (total) staining of HAβ3, HAβ3/θIC, HAθ/β3IC, or EGFP expressed alone in HEK293 cells. Expression was determined using a rabbit anti-HA Epitope Tag DyLightTM 549 conjugated antibody at 1:1000 dilution. d) Bar graphs portray the mean ± SEM of Fiji ImageJ fluorescence quantification of HAβ3 (nsurf = 6 images (46 cells), ntotal = 8 images (79 cells)); HAβ3/θIC (nsurf = 4 images (0 cells), ntotal = 11 images (97 cells)); and HAθ/β3IC (nsurf = 4 images (0 cells), ntotal = 9 images (31 cells)) IF images from c with individual data points overlaid. e) Representative traces of HAβ3 (n = 12), HAβ3/θIC (n = 7/8), and HAθ/β3IC (n = 8/10) in response to 1 mM GABA and 3 mM histamine applied separately or together. f) Bar graphs portray the mean ± SEM of peak current amplitudes from whole-cell recordings in e with individual data points overlaid. * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001; using 2-way ANOVAs with Bonferroni post hoc comparisons.

Functionally, β3 homomers are gated by histamine and only weakly, if at all, by GABA [10,11]. To better quantify functional expression levels, we used patch clamp recording with fast perfusion to measure whole cell currents in response to 1 mM GABA and 3 mM histamine applied separately or together Figure 1(e,f). As expected, HAβ3 homomeric receptors exhibited robust responses to both histamine and GABA + histamine in all cells recorded. HAβ3 homomeric receptors showed large histamine-evoked currents (404 ± 96 pA, n = 12), which were similar when GABA and histamine were co-applied (420 ± 101 pA, n = 12). GABA-evoked currents were comparatively small (21 ± 10 pA, n = 12) and unreliable. HAβ3/θIC had one non-responsive cell, resulting in 7 out of 8 recordings with measurable responses. From the responding cells, homomeric HAβ3/θIC chimeric receptors generated very small currents in response to histamine (70 ± 17 pA, n = 7/8) or GABA + histamine (74 ± 1 pA, n = 7/8), which were significantly smaller than HAβ3 currents (p = 0.006 and p = 0.003, respectively). HAβ3/θIC also generated very small GABA-evoked currents (43 ± 10 pA, n = 7/8), however these were not significantly different from HAβ3 currents (21 ± 10 pA, n = 12, p > 0.9) or currents from non-transfected cells (21 ± 6 pA, n = 6/16, p > 0.9). We also tested the reciprocal IC chimera, HAθ/β3IC. In this case, 8 out of 10 recordings had measurable responses. Like HAβ3/θIC, homomeric HAθ/β3IC generated very small currents in response to histamine (68 ± 15 pA, n = 8/10) and GABA + Histamine (86 ± 18 pA, n = 8/10), which were significantly smaller than HAβ3 (p = 0.002 for both comparisons). The GABA currents generated by HAθ/β3IC (62 ± 17 pA, n = 8/10) were also not significantly different from HAβ3 or non-transfected cells. Taken together, the immunofluorescence and functional results suggest that the β3 ICD is required for efficient homomeric surface expression but is not sufficient for surface expression in the absence of assembly.

The α3 subunit rescues chimeric functional expression

The reduction of β3 functional expression by the θ ICD raises the question of whether this is caused by an assembly or a trafficking defect. Because heteromeric assembly is more common among GABAA receptors than homomeric assembly, we wanted to know if the θ ICD would also prevent the functional expression of αβ heteromeric receptors. To test this, we co-transfected the untagged α3 subunit with HAβ3, HAβ3/θIC, or HAθ/β3IC in parallel cultures of HEK293 cells. α3(HAβ3) and α3(HAβ3/θIC) showed clear plasma membrane labeling by anti-HA in non-permeabilized conditions Figure 2(a,b). There were similar numbers of surface-labeled cells and the fluorescence intensity trended toward a decrease in α3(HAβ3/θIC) compared to α3(HAβ3) (14.0 ± 1.7 AU, n = 20 images (100 cells) and 23.4 ± 3.0 AU, n = 5 images (82 cells), respectively) but failed to reach significance (p = 0.07). α3(HAθ/β3IC) showed no anti-HA surface labeling (n = 4 images (0 cells)). Total HA expression was equivalent in all three conditions when the cells were re-probed following permeabilization (α3(HAβ3) = 26.5 ± 6.2 AU, n = 4 images (60 cells); α3(HAβ3/θIC) = 26.5 ± 2.8 AU, n = 13 images (121 cells); and α3(HAθ/β3IC) = 23.1 ± 5.4 AU, n = 6 images (17 cells)).

Figure 2.

Heteromeric expression and function of β3 and chimera.a) IF image at 20x magnification of non-permeabilized (surface) and permeabilized (total) staining of HAβ3, HAβ3/θIC, or HAθ/β3IC in combination with α3 co-expressed in HEK293 cells using EGFP as a negative control. Expression was determined using a rabbit anti-HA-549 antibody at 1:1000 dilution. b) Bar graphs portray the mean ± SEM of Fiji ImageJ fluorescence quantification of α3(HAβ3) (nsurf = 5 images (82 cells), ntotal = 4 images (60 cells)); α3(HAβ3/θIC) (nsurf = 20 images (100 cells), ntotal = 13 images (121 cells)); and α3(HAθ/β3IC) (nsurf = 4 images (0 cells), ntotal = 6 images (17 cells)) IF images from a with individual data points overlaid. c) Representative traces of α3(HAβ3) (n = 19), α3(HAβ3/θIC) (n = 19/20), and α3(HAθ/β3IC) (n = 3/9) in response to 1 mM GABA and 3 mM histamine applied separately or together. d) Bar graphs portray the mean ± SEM of peak current amplitudes from whole-cell recordings in c with individual data points overlaid. e) Bar graphs portray the mean ± SEM of the degree of histamine potentiation with individual data points overlaid. * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001; using 2-way ANOVAs (b and d) or 1-way ANOVAs (e) with Bonferroni post hoc comparisons.

Functionally, α3β3 heteromeric receptors are gated by GABA and not by histamine, but the GABA responses are strongly potentiated by histamine [10]. To better quantify functional expression levels, as before, we used patch clamp recording with fast perfusion to measure whole cell currents in response to 1 mM GABA and 3 mM histamine applied separately or together Figure 2(c,d). α3(HAβ3) had measurable responses in all cells recorded, α3(HAβ3/θIC) had measurable responses in 19 out of 20 cells recorded, α3(HAθ/β3IC) had measurable responses in 3 out of 9 cells recorded, and the control, α3-only, had measurable responses in 23 out of 28 cells recorded. Comparing the peak amplitudes of GABA-evoked currents, there was no significant difference between α3(HAβ3) and α3(HAβ3/θIC) (1266 ± 270 pA, n = 19 and 803 ± 155 pA, n = 19/20, respectively) or between α3(HAθ/β3IC) and α3-only (7 ± 6 pA, n = 3/9 and 66 ± 22 pA, n = 23/28, respectively). On average, the combined response to GABA + histamine was significantly smaller in both the α3(HAβ3/θIC) and α3(HAθ/β3IC) heteromeric conditions (1983 ± 363 pA, n = 19/20 and 108 ± 19, n = 3/9, respectively) compared to the α3(HAβ3) heteromers (3483 ± 671 pA, n = 19) (α3(HAβ3/θIC) p = 0.002 and α3(HAθ/β3IC) p = 0.0001). The GABA + histamine response from α3(HAθ/β3IC) was not significantly different from the response from α3-only transfection conditions. There was no response to histamine alone in α3(HAβ3/θIC) (10 ± 6 pA, n = 19/20) or α3(HAθ/β3IC) (32 ± 26 pA, n = 3/9) and a comparatively small response, relative to GABA, in α3(HAβ3) (192 ± 61 pA, n = 19), suggesting a small but measurable population of β3 homomers in the latter condition. For both α3(HAβ3) and α3(HAβ3/θIC), the GABA + histamine response was markedly potentiated compared to GABA alone, which is typical of αβ heteromers. There was no difference in the extent of histamine potentiation of the GABA response between α3(HAβ3) (2.93 ± 0.27-fold, n = 19), α3(HAβ3/θIC) (2.64 ± 0.22-fold, n = 19), and α3-only (4.78 ± 0.96-fold, n = 22). α3(HAθ/β3IC) only responded to GABA in one recording and thus, could not be included in the statistical analysis. Altogether, these data suggest that the θ ICD chimera functions like the wild-type β3 subunit, albeit with marginally lower heteromeric surface expression levels and no homomeric surface expression.

To further explore any functional differences caused by the θ ICD, we compared agonist potencies in αβ heteromers containing either HAβ3 or HAβ3/θIC (Figure 3). HAθ/β3IC was not included in this experiment due to the lack of response seen in Figure 2. Concentration-response curves were constructed from the peak amplitudes of whole cell currents evoked sequentially by increasing concentrations of GABA (from 1 μM to 1 mM), as shown in Figure 3. Similar to the previous experiment, results showed a trend toward lower maximal peak currents from α3(HAβ3/θIC) heteromeric receptors (1509 ± 274.6 pA, n = 19) compared to α3(HAβ3) heteromers (2384 ± 258 pA, n = 20); however, this trend failed to reach significance (p = 0.08; Figure 3(a,b). Concentration-response curves were normalized to the maximum GABA concentration tested, and the comparison of these curves revealed a rightward-shift to a 3-fold higher GABA EC50 for the chimera-containing heteromer. EC50 values were 21 μM (n = 10–20 cells per concentration) for α3(HAβ3) and 64 μM (n = 9–19 cells per concentration) for α3(HAβ3/θIC) Figure 3(c). It is not clear if the 3-fold potency difference represents an effect of the θ ICD on β3 subunit function per se or a difference in αβ subunit stoichiometry between conditions, which could also explain the trend toward smaller current amplitudes.

Figure 3.

GABA concentration-response curves of αβ heteromeric combinations.a) Representative traces of GABA-evoked currents from α3(HAβ3) (n = 10–20 cells per concentration) and α3(HAβ3/θIC) (n = 9–19 cells per concentration) in response to increasing GABA concentrations from 1 μM to 1 mM. b) Raw peak current amplitudes plotted as a function of GABA concentration. Fit parameters: α3(HAβ3) EC50 = 28 μM, Imax = 2384 pA; α3(HAβ3/θIC) EC50 = 86 μM, Imax = 1509 pA. c) Normalized peak amplitudes from B are plotted as a function of GABA concentration. Fit parameters: α3(HAβ3) EC50 = 21 μM; α3(HAβ3/θIC) EC50 = 64 μM.

The β3 subunit also rescues chimeric functional expression

Our data clearly show that the HAβ3/θIC chimera is viable and readily assembles into functional surface-expressed receptors with α3. Since the known structural determinants for assembly are in the EC ligand binding domains and wild-type β3 readily assembles in a homomeric configuration, we reasoned that HAβ3/θIC probably also assembles as homomeric receptors but has a deficit in trafficking introduced by the θ ICD. Such deficit could either result from the loss of an anterograde trafficking signal contained in β ICD or the gain of an ER retention signal in the θ ICD that co-assembly with α3 can overcome. If so, the next logical question was whether co-assembly with wild-type β3, having its natural ICD, could also rescue the θ ICD chimera. To test this, we co-transfected the untagged β3 subunit with HAβ3, HAβ3/θIC, or HAθ/β3IC in parallel cultures of HEK293 cells to produce pseudo-homomeric β3 receptors having either the β ICD on all subunits or a mixture of β and θ ICDs. Remarkably, when co-transfected with wild-type β3, clear surface labeling was seen for both HAβ3 and HAβ3/θIC, as shown by anti-HA staining in non-permeabilized conditions Figure 4(a,b). Wild-type β3 co-expression with both HAβ3 and HAβ3/θIC yielded comparable numbers of surface-labeled cells with similar intensity (β3(HAβ3) = 13.9 ± 0.7 AU, n = 6 images (49 cells) and β3(HAβ3/θIC) = 9.9 ± 1.1 AU, n = 7 images (52 cells)). Co-expression of wild-type β3 with HAθ/β3IC showed no surface labeling by anti-HA (n = 2 images (0 cells)). When the cells were re-probed following permeabilization there was a significantly higher fluorescence intensity but a similar number of labeled cells in β3(HAβ3/θIC) (25.8 ± 2.6 AU, n = 9 (79 cells)) compared to β3(HAβ3) (17.4 ± 1.8 AU, n = 8 (67 cells)) (p = 0.0003). There were not enough fields containing labeled cells for β3(HAθ/β3IC) to be included in the statistical analysis (25.0 AU, n = 2 images (7 cells)).

Figure 4.

Pseudo-homomeric expression and function of wild type and chimeric β3.a) IF images at 20x magnification of non-permeabilized (surface) and permeabilized (total) staining of HAβ3, HAβ3/θIC, or HAθ/β3IC in combination with β3 co-expressed in HEK293 cells using EGFP as a negative control. Expression was determined using a rabbit anti-HA-549 antibody at 1:1000 dilution. b) Bar graphs portray the mean ± SEM of Fiji ImageJ fluorescence quantification of β3(HAβ3) (nsurf = 6 images (49 cells), ntotal = 8 images (67 cells)); β3(HAβ3/θIC) (nsurf = 7 images (52 cells), ntotal = 9 images (79 cells)); and β3(HAθ/β3IC) (nsurf = 2 images (0 cells), ntotal = 2 images (7 cells)) IF images from a with individual data points overlaid. c) Representative traces of β3(HAβ3) (n = 14), β3(HAβ3/θIC) (n = 13/15), and β3(HAθ/β3IC) (n = 5) in response to 1 mM GABA and 3 mM histamine applied separately or together. d) Bar graphs portray the mean ± SEM of peak current amplitudes from whole-cell recordings in c with individual data points overlaid. * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001; using 2-way ANOVAs with Bonferroni post hoc comparisons.

Functionally, we tested the pseudo-homomeric combinations in the same manner as the homomeric receptors and the α3 heteromers Figure 4(c,d). Since the β3(HAβ3) and β3(HAβ3/θIC) combinations both contain the β3 EC ligand binding and TM domains, the pseudo-homomers were expected to behave like β3 homomeric receptors and give histamine-evoked but not GABA-evoked currents. Because the β3(HAθ/β3IC) combination contains the both the β3 and θ EC ligand binding domains, and we would not expect the θ NTD to bind histamine, it was unclear how the heteromers should behave if they were produced. The β3(HAβ3) and β3(HAθ/β3IC) responses were measurable in all cells recorded and the β3(HAβ3/θIC) response was measurable in 13 out of 15 cells. In all conditions, there was a comparable but minimal response to 1 mM GABA (β3(HAβ3) = 8 ± 4 pA, n = 14; β3(HAβ3/θIC) = 3 ± 3 pA, n = 13/15; and β3(HAθ/β3IC) = 0 ± 0 pA, n = 5) and much larger currents evoked by 3 mM histamine (β3(HAβ3) = 533 ± 123 pA, n = 14; β3(HAβ3/θIC) = 213 ± 43 pA, n = 13/15; and β3(HAθ/β3IC) = 342 ± 90 pA, n = 5). There were no significant differences in the GABA or histamine responses for the three conditions, however the histamine-evoked current from β3(HAβ3/θIC) trended toward a decrease compared to β3(HAβ3) (p = 0.08). The currents elicited by GABA + histamine together were significantly different between the β3(HAβ3) and β3(HAβ3/θIC) conditions (580 ± 129 pA, n = 14 and 220 ± 42 pA, n = 13/15, respectively, p = 0.03) while the currents from β3(HAθ/β3IC) were not significantly different from either β3(HAβ3) or β3(HAβ3/θIC) currents (β3(HAθ/β3IC) = 353 ± 88 pA, n = 5). Taken together, the immunofluorescence and whole cell recordings suggest that the β3/θIC chimera is, indeed, surface-expressed in a functional complex with wild-type β3. However, the HAθ/β3IC chimera also gave histamine-evoked currents but no surface immunofluorescence, so we must also consider whether the functional responses are mostly or entirely generated by the wild-type subunits.

A functional tag that could distinguish the contributions of the individual subunits to overall receptor function was required to answer this question. We took advantage of a mutant β3 subunit, zβ3Q64E, which has a 50-fold higher affinity for histamine [11] and compared the histamine potencies at zβ3Q64E homomeric receptors and pseudo-homomeric receptors containing the zβ3Q64E mutant co-expressed with either HAβ3, HAβ3/θIC, or HAθ/β3IC in parallel cultures of HEK293 cells. Concentration-response curves were constructed from whole cell current amplitudes evoked sequentially by increasing concentrations of histamine (from 10 μM to 10 mM), as shown in Figure 5. If the θ ICD-containing chimera is, in fact, a component of the functional surface receptor population in complex with the zβ3Q64E mutant, it would cause a rightward-shift in the concentration-response and the emergence of a biphasic curve because the chimera does not carry the high-affinity mutations. If it is not, then the functional receptors should behave like zβ3Q64E mutant homomers, having apparent high affinity for histamine and only one phase in the curve.

Figure 5.

Histamine dose-response curves of pseudo-homomeric combinations.a) IF images at 20x magnification of non-permeabilized (surface) and permeabilized (total) staining of HAβ3, HAβ3/θIC, or HAθ/β3IC in combination with zβ3Q64E co-expressed in HEK293 cells using EGFP as a negative control. Expression was determined using a rabbit anti-HA-549 antibody at 1:1000 dilution. b) Bar graphs portray the mean ± SEM of Fiji ImageJ fluorescence quantification of zβ3Q64E(HAβ3) (nsurf = 11 images (46 cells), ntotal = 12 images (65 cells)); zβ3Q64E(HAβ3/θIC) (nsurf = 9 images (26 cells), ntotal = 9 images (51 cells)); and zβ3Q64E(HAθ/β3IC) (nsurf = 3 images (0 cells), ntotal = 10 images (74 cells)) IF images from a with individual data points overlaid. c) Representative traces of zβ3Q64E (n = 8), zβ3Q64E(HAβ3) (n = 12), zβ3Q64E(HAβ3/θIC) (n = 12), and zβ3Q64E(HAθ/β3IC) (n = 3) in response to increasing histamine concentrations from 10 μM to 10 mM. d) Normalized peak amplitudes from c plotted as a function of histamine concentration. Fit parameters: zβ3Q64E EC50 = 15 μM; zβ3Q64E(HAβ3) EC501 = 29 μM, EC502 = 2.4 mM; zβ3Q64E(HAβ3/θIC) EC501 = 39 μM, EC502 = 1.3 mM; zβ3Q64E(HAθ/β3IC) EC50 = 22 μM; HAβ3 EC50 = 1.1 mM (n = 4–8 cells per concentration). e) Normalized peak amplitudes from combinations in C using varied cDNA ratios plotted as a function of histamine concentration. Fit parameters: zβ3Q64E(HAβ3/θIC) 2:1 ratio EC501 = 21 μM, EC502 = 330 μM (n = 7); zβ3Q64E(HAβ3/θIC) 1:2 ratio EC501 = 25 μM, EC502 = 1.3 mM (n = 6).

We first looked at non-permeabilized anti-HA staining to confirm that zβ3Q64E could traffic with HAβ3 and HAβ3/θIC to the surface like the wild-type α3 and β3 subunits before. Indeed, the HA-immunofluorescence surface staining showed cells labeled in both the zβ3Q64E(HAβ3) and the zβ3Q64E(HAβ3/θIC) conditions Figure 5(a,b). Like the previous α and β combinations, the surface fluorescence intensity of cells expressing HAβ3/θIC was not significantly different compared to that of cells expressing HAβ3 (zβ3Q64E(HAβ3/θIC) = 7.8 ± 0.8 AU, n = 9 images (26 cells); zβ3Q64E(HAβ3) = 13.3 ± 1.9 AU, n = 11 images (46 cells)). Unsurprisingly, zβ3Q64E(HAθ/β3IC) showed no anti-HA surface labeling (n = 3 images (0 cells)). Permeabilized staining revealed similar numbers and fluorescence intensity of labeled cells (zβ3Q64E(HAβ3) = 20.2 ± 2.2 AU, n = 12 images (65 cells); zβ3Q64E(HAβ3/θIC) = 24.8 ± 4.9 AU, n = 9 images (51 cells); and zβ3Q64E(HAθ/β3IC) = 22.8 ± 3.6 AU, n = 8 images (74 cells)).

Figure 5 also shows that the concentration-response curves were right-shifted for both zβ3Q64E(HAβ3) and zβ3Q64E(HAβ3/θIC) pseudo-homomeric combinations compared to zβ3Q64E alone, while the curve for zβ3Q64E(HAθ/β3IC) was not different from that of zβ3Q64E alone Figure 5(d). As shown previously in Hoerbelt et. al. (2016)[11], zβ3Q64E homomeric receptors have a very low Histamine EC50 (15 μM, n = 8) compared to the wild type β3 (1.1 mM, n = 4–8 cells per concentration). Comparing the maximal peak amplitudes evoked by 10 mM histamine, the whole cell currents from zβ3Q64E(HAβ3/θIC) and zβ3Q64E(HAθ/β3IC) pseudo-homomeric receptors (275 ± 55 pA, n = 12 and 178 ± 36 pA, n = 3, respectively) were smaller than either zβ3Q64E homomers (649 ± 201 pA, n = 8, p < 0.05) or the zβ3Q64E(HAβ3) combination (804 ± 230 pA, n = 12, p < 0.05). Moreover, the concentration-response curves for the zβ3Q64E(HAβ3) and the zβ3Q64E(HAβ3/θIC) combinations were clearly bimodal, having two components with markedly different concentration dependence. The histamine EC50 for homomeric zβ3Q64E was 15 μM (n = 8), which was similar to zβ3Q64E(HAθ/β3IC) (22 μM, n = 3) and the high-affinity component in both zβ3Q64E(HAβ3) heteromers (16 μM, n = 12) and zβ3Q64E(HAβ3/θIC) heteromers (29 μM, n = 12). The low-affinity component for the for zβ3Q64E(HAβ3) curve had an EC50 of 2.4 mM (n = 12) and represented 85% of the total curve fit, while the low-affinity component for the zβ3Q64E(HAβ3/θIC) curve had an EC50 of 1.3 mM (n = 8) and represented 53% of the total curve fit. These results confirm that the HAβ3/θIC chimera indeed assembles with other β3 subunits and contributes to the functional response, but to a slightly lesser extent than wild-type HAβ3 while the HAθ/β3IC chimera remains ER retained.

Figure 5(e) explores the effect of subunit ratios on shifting the concentration-response curve toward the high-affinity or low-affinity phenotypes. To this end, we compared zβ3Q64E(HAβ3/θIC) at 2:1, 1:1, and 1:2 ratios of cDNA. The maximal peak amplitudes evoked by 10 mM histamine were 754 ± 278 (n = 7), 275 ± 58 pA (n = 12), and 114 ± 22 pA (n = 6), respectively, consistent with a reduction in total surface receptors as the number of β ICDs was reduced. Regardless of the subunit ratio, the concentration-response curves retained their bimodality, suggesting that the HA-tagged constructs were assembled with the zβ3Q64E construct. The high-affinity component of the zβ3Q64E(HAβ3/θIC) curves at 2:1, 1:1, and 1:2 ratios had an EC50 of 21 μM (n = 7), 29 μM (n = 12), and 25 μM (n = 6), respectively, which are all comparable to the EC50 for zβ3Q64E (15 μM, n = 8). Similarly, the low-affinity component of the zβ3Q64E(HAβ3/θIC) curves at 1:2 and 1:1 ratios both had an EC50 of 1.3 mM (1:2 ratio n = 6, 1:1 ratio n = 12) which is comparable to the EC50 for wild-type β3 (1.1 mM, n = 4–8 cells per concentration). The low-affinity component accounted for 64% of the zβ3Q64E(HAβ3/θIC) curve at 1:2 ratio compared to 53% at 1:1. Interestingly, the low-affinity component was also 64% of the curve fit for zβ3Q64E(HAβ3/θIC) at 2:1 ratio, but the fit had an EC50 of 330 μM (n = 7), suggesting a difference in the number or functional impact of low-affinity sites present at the surface in this condition. The inverse relationship between maximal current amplitudes and the HAβ3/θIC content implies that multiple copies of the β3 ICD are probably required in the pentamer to overcome ER retention and traffic the receptor to the surface.

Discussion

GABAA receptors are functionally expressed on the cell surface. This requires translation by ER-associated ribosomes, the proper folding of the individual subunit proteins, their assembly into pentameric ion channels, and their trafficking through the secretory pathway to the plasma membrane where they can respond to extracellular signals. The cellular mechanisms that regulate folding, assembly and trafficking are not completely understood. Some of the more general maturation processes involve common ER chaperones like calnexin, BiP, and protein disulfide isomerase, which aid in folding, glycosylation, and formation of intra- or inter-subunit disulfide bonds[12]. Others may be more specific for particular subunits or combinations of subunits. Which subunits are compatible to assemble with one another, for example, appears to be determined by their extracellular ligand binding loops A-E and the GKER motif in loop F of the NTD [3,7,12]. Then, trafficking of the pentameric receptors from ER to the plasma membrane might occur by default but often appears to involve cytosolic chaperone proteins that interact with the cytoplasmic TM3-TM4 loops of different subunits to help guide and cluster the receptors to the appropriate location on the plasma membrane [13,14].

In the present study, we sought to test whether and how the ICD is involved in β3 homomeric and heteromeric functional expression. The β3 subunit can form homo-pentameric ion channels, unlike β2, and previous studies suggested this could be fully explained by differences in an “assembly motif” (GKER/DNTK) in Loop F of their NTD binding domains [7]. Up to now, the β ICD was not known to regulate recombinant functional expression. To explore the role of the ICD, we created reciprocal chimeric exchanges of the TM3-TM4 ICD between the β3 and θ subunits. We expected that the θ ICD would support the functional expression of chimeric β3/θ homomers. It did not. While the θ ICD did not disrupt assembly, our data revealed that the θ ICD could not support ER export and trafficking of receptors unless it was co-assembled with α or β subunits bearing their natural ICD. Of note, it is curious that the HAθ/β3IC chimera was not surface expressed in combination with α as previously reported for the wild-type θ subunit[9], but our data suggest they fail to co-assemble.

Results show that HAβ3 homomeric receptors can traffic to the plasma membrane while neither the HAβ3/θIC chimera nor the HAθ/β3IC chimera could do this. By contrast, we also show that pseudo-homomeric receptors composed of the HAβ3/θIC chimera plus wild-type β3 or the high-affinity zβ3 mutant are functionally expressed on the plasma membrane, as are heteromeric receptors composed of the HAβ3/θIC chimera plus wild-type α3. Taken together, three conclusions can be drawn from these results: (1) the β3 ICD is necessary for functional expression of the homomeric receptor, (2) the α3 ICD likewise promotes functional expression of the heteromeric receptor, and (3) the θ ICD contains neither permissive signals to promote surface expression nor inhibitory signals to prevent it. In the rare cases where homomeric HAβ3/θIC was seen to reach the plasma membrane, these too may be pseudo-homomers assembled with endogenous β3, which is expressed in HEK293 cell cultures at low levels in the culture [15] as a whole, or involve some other factor that varies from cell to cell and leads to a few positively stained cells. Based on the absence of HAθ/β3IC surface staining under any conditions, and the fact that θ does not assemble as a homomeric receptor [9], we can also conclude that the β3 ICD is not sufficient to drive surface expression if the subunits are not assembled. Although this study did not test other subunit ICDs, we propose that β1-3 all contain similar anterograde trafficking signals. This is supported by previous reports by other groups showing that homomeric β1 expression[3], and homomeric β2 mutants having the assembly-permissive GKER motif in their F-loop but still having the β2 ICD were also functionally expressed[7].

Virtually all neurons in the CNS express functional GABAA receptors. Some of the 19 available subunits (e.g., α1, β2-3,γ2) are expressed abundantly throughout the brain and are well characterized in native and recombinant systems[1]. Others are far less common (e.g., γ1, γ3, θ, ϵ, π) and are poorly understood[1]. Heterologous expression systems are often used in combination with mutagenesis to foster understanding of which subunits or combinations of subunits are functionally expressed, how they work and how they are regulated. HEK293 cells are by far the most commonly used mammalian cell line for the study of ligand-gated ion channels (LGIC), including GABAA receptors, because they do not express LGIC subunits in an amount that would confound interpretation [15]. They are considered, in this regard, to be a blank slate. It is important, however, to consider that HEK293 cells are not neurons and, just as they generally do not express neuronal receptors, they also might not express all the proper machinery to process or traffic neuronal receptors in the same manner as neurons. This may be especially true for receptors or other cargo destined to axonal/presynaptic compartments, which HEK293 cells do not have.

Recent studies have made this point clearly with respect to nicotinic acetylcholine receptors (nAChR). The most widely expressed neuronal nAChRs, the α7 and α4β2 subtypes, have been notoriously difficult to study in mammalian cell lines because they are not functionally expressed [16-19]. However, co-expression of the putative chaperone/accessory proteins RIC-3 or TMEM35A/NACHO can permit robust functional expression of these nAChRs in HEK293 cells [17-21].

Our data indicate the HAβ3/θIC chimera alone rarely reaches the plasma membrane, whereas co-expression with either wild-type α3 or β3 rescues its surface expression. The defect in HAβ3/θIC homomeric expression appears to reflect a trafficking error, not an assembly error, because the pseudo-homomeric HAβ3/θIC plus wild-type β3 receptors are functionally expressed. From this we can infer that α3 and β3 subunits both contain anterograde trafficking signals that can overcome ER retention of the assembled receptors. In GABAA subunits, and across all subunits of the pentameric ion channel superfamily, the TM3-TM4 intracellular loop is the most variable region both in terms of length and amino acid composition.

A number of cytosolic chaperones have been proposed to regulate GABAA receptor trafficking by interactions involving the ICD of α, β or γ subunits. The first and best characterized chaperone shown to be involved in GABAA receptor trafficking is the GABAA receptor-associated protein, or GABARAP [22]. GABARAP acts by interacting with the γ2 ICD to promote plasma membrane expression [23]. Other cytosolic chaperones such as BIG2 (brefeldin A-inhibited GDP/GTP exchange factor 2), NSF (N-ethylmaleimide-sensitive factor), and PRIP1 (PLC-related catalytically inactive protein 1) all interact with the ICD of β1-3 [24-27]. Gephryn and PLIC1 (protein linking integrin-associated protein to cytoskeleton-1) interact with both α and β ICDs [14,28]. Except for PRIP1, these chaperone proteins and others are expressed at moderate to high levels in HEK293 cells (GEO accession number GDS5213) [29]. It remains to be determined how these proteins interact with the ICDs and which, if any, are required for functional expression of various subtypes of GABAA receptors in HEK293 cells.

It is perhaps worth noting the relative homology among the synaptically expressed α, β, γ subunits, which are all readily expressed in heterologous systems. The rat β1-3 ICDs share 48–54% amino acid sequence identity, while the α1-3 and α5 ICDs are 36–62% identical, and the γ1-3 ICDs are 52–56% identical by comparing BLAST alignments. In contrast, the θ ICD has no homology in the rat genome and the non-synaptic α4, α6 and δ ICDs are likewise each unique. So, one question raised from this study remains: why do GABAA β homomers and αβ heteromers express in HEK293 cells but the β/θIC chimera does not? If certain nAChRs are a precedent, it may be that the θ ICD requires interactions with a unique chaperone protein found only where θ is naturally expressed. If so, the β/θIC chimera could well express in another system, for example in neurons, or in heterologous cells upon co-expression of appropriate chaperones that are yet to be identified. Trafficking might also be amenable to proteostatic enhancement by small molecules [30]. As a whole, this study shows the involvement of the ICD in β3 homomeric and heteromeric functional expression and raises questions about the ICD of less common GABAA subunits, such as θ. Further work is needed to determine precisely where the critical ICD motifs reside and to understand how the assembly and trafficking of β, θ, and the β/θ chimeric subunits are functionally regulated.

Methods

Drugs and solutions

Recording solution components, buffer components, GABA (cat. #A2129), histamine dihydrochloride (cat. #H7250) were purchased from Sigma-Aldrich (St. Louis, MO). GABA was dissolved in extracellular recording solution at a 1 M stock concentration, then stored at 4°C. Histamine was dissolved at a 30 mM stock concentration and the pH was adjusted to neutral. This stock histamine solution was made freshly or stored at −20°C for up to 72 hr. Stock drug solutions were diluted on the day of the experiment. Concentration-response curve solutions were made by serial dilution.

cDNA and mutagenesis

All cDNA constructs were expressed in the pRK5 vector carrying the CMV promoter and ampicillin-resistance. Subcloning-efficiency E. coli were used as the host for cDNA copy replication. Rat α3 (accession no. L08492.1) and β3 (X15468.1) subunit cDNAs in the pRK5 vector were given generously by Dr. Peter Seeburg (Max Planck Institute for Medical Research, Heidelberg, Germany). Rat θ (AF419333.1) subunit cDNA was generously provided by Dr. Maurice Garrett (University of Bordeaux, Bordeaux, France) in the pcDNA3 vector and transferred into the pRK5 vector between BamHI (5ʹ) and XbaI (3ʹ) upon receipt. The coding region and UTRs were confirmed by Sanger sequencing for all subunit cDNAs and subsequent mutations. We identified two separate point mutations (causing H362N and L400M) within the ICD of our θ cDNA and one point mutation causing S296T in the third TMD of α3 that differ from the published sequences.

HAβ3, HAθ, HAβ3/θIC, and HAθ/β3IC were all created using NEBuilder HiFi DNA assembly kit (New England Biolab, Ipswich, MA, E5520S). The hemagglutinin tag was a triplet of HA epitopes flanked by AgeI (5ʹ) and NotI (3ʹ) endonuclease sites. PCR primers were designed using the NEBuilder Assembly Tool v1.12. To make HAβ3, the triplet HA tag, TGLDYPYDVPDYAGYPYDVPDYAGSYPYDVPDYAAAA, was inserted between amino acids 6 and 7 of the mature β3 protein using the manufacturer’s protocol. In short, PCR was used to isolate and amplify the fragments of interest (linearized β3 and the HA tag). The fragments were then assembled by mixing and incubating at 50°C with the NEBuilder HiFi DNA Assembly Master Mix for ≥ 15min. The assembled product was then transformed into NEB5α competent cells and spread onto ampicillin-resistant plates, clonal colonies were picked and screened by endonuclease digestion. The HAθ/pRK5 construct was made in the same manner as HAβ3/pRK5 except that the HA tag was inserted between amino acids 5 and 6 of the mature protein.

The HAβ3/θIC construct was made by linearizing HAβ3/pRK5 without the M3-M4 loop (IVFPFT … YIFFGR) and isolating the θ M3-M4 loop (RNHRRC … VPKVDR) from HAθ/pRK5 using PCR. Likewise, to make the HAθ/β3IC construct, HAθ/pRK5 was linearized (LFPLSF … YLFFSQ) and the β3 M3-M4 loop (QRQKKL … AIDRWS) was isolated from HAβ3/pRK5 using PCR. Fragments for both chimeras were then assembled and screened in the same manner as above.

The zβ3Q64E construct was made previously and described in Hoerbelt et. al. (2016).

Cell culture and transfections

HEK293 cells (ATCC CRL-1573) were cultured at 37°C with 5% CO2 in Minimal Essential Medium plus glutamine (MEM, Gibco, Gaithersburg, MD) with 10% fetal bovine serum (Gibco) and 5% Penicillin-Streptomycin (Gibco) added. For immunofluorescence experiments and electrophysiological recording, cells were plated at 100,000/dish in Poly-D-Lysine (Sigma-Aldrich) coated 6-well plates (Corning) or 35 mm Nunc dishes (Nalge Nunc, Naperville, IL), left overnight to adhere, and then transfected with a total of the following per dish/well: 1 μg total cDNA (subunit ratios of 1:1 for co-transfections, unless otherwise stated), 0.82 μL Lipofectamine 2000 (Invitrogen), and 100 μL total serum-free MEM. Components were mixed using the manufacturer’s protocol. 100 μL of the transfection mixture was dripped in an outward spiraling motion into each dish or well. For Western blotting experiments, non-coated 60 mm dishes were used and cell density and transfection volumes were scaled up by a ratio of 0.4 due to the increase in surface area of the dish. Cells were used 36–40 hr post-transfection for immunofluorescence and Western blotting experiments and 20–48 hr post-transfection for electrophysiological recording. For electrophysiology transfections, EGFP/pRK5 was co-transfected with the GABAA subunits as 10% of the total cDNA, and only cells expressing the EGFP were targeted to patch.

Western blot

At 36–40h post transfection, HEK293 cells plated in 60 mm dishes were washed twice by PBS then lysed in 2% SDS + 8 mM EDTA. A fraction of the lysates was saved for a BCA analysis while the remaining fraction of the lysates was diluted 1:1 with reducing Laemmeli buffer (with 5% β-Mercaptoethanol). 10–20 μg of total protein was separated using a 10% SDS-PAGE gel. Separated proteins were then transferred to a nitrocellulose membrane. The membranes were washed with tris-buffered saline (pH = 7.4) + 0.1% Tween-20 (TBST) and then blocked with 5% milk in TBST. Blocked membranes were incubated at 4°C overnight in primary rabbit anti-HA Epitope Tag (1 mg/ml, diluted 1:5000 in 5% milk + TBST, Rockland Antibodies & Assays cat. #600-401-384). After washing thoroughly, the membranes were probed with secondary HRP-conjugated goat anti-rabbit IgG (1 mg/ml, diluted 1:5000 in 5% milk + TBST, Invitrogen cat. #ABIN964977) for 1 h at room temperature. Immunoreactivity was visualized using PierceTM ECL Western Blotting Substrate (ThermoFisher Scientific cat. #32,106) in a ChemiDoc Imaging System (BioRad, Hercules, CA, USA). Membranes were then incubated with primary rabbit anti-GAPDH (1 mg/ml, diluted 1:5000 in 5% milk + TBST, Sigma-Aldrich #G9545) for 1 h at room temperature. After washing well, the membranes were probed with the same secondary and visualized again as above. Band intensities and molecular weights were quantified using the ImageJ (NIH) gel analysis tool[31]. Blot images were post processed on a personal computer using Photoshop software for presentation.

Immunofluorescence (IF)

At 36–40h post transfection, HEK293 cells plated in 6-well plates were washed in sucrose solution (290 mM sucrose, 5 mM HEPES, 3 mM KCl, 1.8 mM CaCl2, and 1 mM MgCl2; pH 7.3), then cells were washed in tris-buffered saline (TBS, pH 7.4), and fixed in TBS + 3.7% formaldehyde (pH 7.4). After fixation, cells were washed in TBS, blocked in TBS + 5% goat serum, then incubated for 1 hr in rabbit anti-HA Epitope Tag DyLightTM 549 conjugated antibody (1 mg/ml, diluted 1:1000 in TBS + 5% goat serum, Rockland Antibodies & Assays #600-442-384). After washing with TBS, cells were visualized on an Olympus IX71 fluorescence microscope fitted with a LUCPlanFl 20X/0.4 RC2 (∞/) objective. Non-permeabilized fluorescent photomicrographs were captured at the same exposure with a QImaging QICAM digital camera (1X). After capturing non-permeabilized photomicrographs, the cells were then permeabilized in TBS + 0.1% TritonX-100. After permeabilization, cells were washed with TBS, blocked in TBS + 5% goat serum, re-incubated for 1 hr in the same antibody as above, and washed again with TBS. Cells were then visualized, and fluorescent photomicrographs of permeabilized staining were captured, as above. Both non-permeabilized (surface) and permeabilized (total) photomicrographs were post processed on a personal computer using Photoshop software for presentation.

Immunofluorescence quantification

Average fluorescent intensities were calculated using Fiji ImageJ [31,32]. Using the 192 immunofluorescence photomicrographs captured as above, the fluorescent cells were manually outlined in each image, then the average fluorescent intensity was calculated and the background fluorescence was subtracted. The fluorescent intensities of immune-positive cells were reported in arbitrary units (AU).

Electrophysiological recordings

Whole-cell patch clamp was used to examine the function of GABAA receptors as in Fleck (2002) and Hoerbelt et al. (2015). Briefly, at 20–48 hr post-transfection HEK293 cells were superfused with extracellular recording solution (pH 7.3–7.4; 295–305 mOsm) consisting of 0.1 mg/ml phenol red pH indicator and (in mM): 145 NaCl, 5 HEPES, 3 KCl, 1.8 CaCl2 and 1 MgCl2. Thin-walled borosilicate glass microelectrodes were 3–7MΩ when filled with intracellular recording solution (pH 7.3; 295–305 mOsm) consisting of (in mM): 135 CsCl, 10 CsF, 10 HEPES, 5 EGTA, 1 MgCl2 and 0.5 CaCl2. Patch recordings were conducted in voltage-clamp mode at −80 mV, causing inward chloride currents with Vrev around 0 mV under these ionic conditions. Current signals were recorded and analyzed on a Macintosh computer using Synapse software (Synergistic Research Systems, Silver Spring, MD). Drugs were applied mostly as described in Fleck (2002) and Hoerbelt et al. (2015). Briefly, 4 or 8 syringes were loaded with extracellular recording solution with and without GABA and/or histamine and driven at 1.5 ml/min through a single glass flow-pipe combining these 4 or 8 barrels. The flow pipe had a ~200 mm diameter tip and was placed <1 mm from the target cell to allow complete superfusion of the cell. Rapid solution exchange (5–20ms) was provided by 3-way solenoid valves (Lee Co., Westbrook, CT) controlled by the computer. The protocols used for drug application were consistently 2 sec drug application pulses with at least 5 sec control solution application between each drug pulse.

Data analysis and statistics

Current traces were analyzed with Synapse software (Silver Spring, MD), Kaleidagraph (Synergy Software, Reading, PA), GraphPad Prism 8 (San Diego, CA), and Microsoft Excel. All comparisons were made using cells transfected in parallel and recording dishes were alternated by transfection subtype. Current amplitudes were measured from baseline to peak, where the baseline was taken during control period immediately before the switch to drug application. Cells with noisy or unstable baselines were excluded from analysis. Histamine potentiation was assessed by comparing the current amplitude during combined application of GABA + histamine (or the extent of potentiation) to the current amplitude of the initial pulse of GABA alone. Concentration response curves were fit using the following equation:

Where = maximum current, = concentration at half maximal current, and = Hill coefficient. Current amplitudes for concentration response curves were normalized to the highest concentration of agonist tested then averaged. Biphasic histamine concentration response curves were fit using the following equation:

Where = maximum high affinity current, = Hill coefficient of the high affinity component (constrained to 1.4 based on the control fit to zβ3Q64E), = concentration at half maximal current of the high affinity component, = Hill coefficient of the low affinity component (constrained to 1.0 based on the control fit to wild type β3), and = concentration at half maximal current of the low affinity component. Notably, there is a ~3 mV drop in Cl− driving force at the highest histamine concentration tested (10 mM histamine dihydrochloride) which was not compensated. Half maximal (EC50) values from multiple replicates per group are presented as the mean ± the 95% confidence interval. Other data are shown as mean ± SEM and n values refer to either the number of images analyzed followed by the total number of cells in all images acquired in parentheses or the number of cells recorded. For all statistical comparisons, a p value < 0.05 is considered statistically significant.