Synaptic Localization of α5 GABA (A) Receptors via Gephyrin Interaction Regulates Dendritic Outgrowth and Spine Maturation

Abstract

GABA_A receptor subunit composition is a critical determinant of receptor localization and physiology, with synaptic receptors generating phasic inhibition and extrasynaptic receptors producing tonic inhibition. Extrasynaptically localized α5 GABA_A receptors are largely responsible for tonic inhibition in hippocampal neurons. However, we show here that inhibitory synapses also contain a constant level of α5 GABA_A receptors throughout neuronal development, as measured by its colocalization with gephyrin, the inhibitory postsynaptic scaffolding protein. Immunoprecipitation of the α5 subunit from both cultured neurons and adult rat brain coimmunoprecipitated gephyrin, confirming this interaction in vivo. Furthermore, the α5 subunit can interact with gephyrin independent of other synaptically localized alpha subunits, as shown by immunoprecipitation experiments in HEK cells. By replacing the α5 predicted gephyrin binding domain (Residues 370–385) with either the high affinity gephyrin binding domain of the α2 subunit or homologous residues from the extrasynaptic α4 subunit that does not interact with gephyrin, α5 GABA_A receptor localization shifted into or out of the synapse, respectively. These shifts in the ratio of synaptic/extrasynaptic α5 localization disrupted dendritic outgrowth and spine maturation. In contrast to the predominant view of α5 GABA_A receptors being extrasynaptic and modulating tonic inhibition, we identify an intimate association of the α5 subunit with gephyrin, resulting in constant synaptic levels of α5 GABA_AR throughout circuit formation that regulates neu ronal development.

INTRODUCTION

Gamma-aminobutyric acid type A receptors (GABA_ARs) are pentameric chloride permeable ligand gated ion channels mediating GABAergic inhibition in the brain. Synaptic release of GABA activates postsynaptically localized GABA_ARs to trigger fast inhibitory currents, while ambient GABA results in activation of extrasynaptic receptors and generation of tonic inhibition. The subunit composition of the receptor, typically two α subunits, two β subunits, and either a γ or a δ subunit, determines receptor cell surface localization as well as physiological and pharmacological properties. Receptors containing α1, α2, and/or α3 subunits are primarily located at GABAergic synapses via direct association with gephyrin, the key inhibitory postsynaptic scaffolding protein responsible for GABA_AR and glycine receptor localization (Tyagarajan and Fritschy, 2014). GABA_ARs containing the α4 and α6 subunits do not interact with gephyrin, are extrasynaptically localized, and are responsible for tonic GABA currents (Glykys and Mody, 2007). The α5 subunit assembles with β and γ subunits, is reported to be primarily localized extrasynaptically (Brunig et al., 2002; Christie and de Blas, 2002), and is primarily responsible for generating tonic inhibition in the hippocampus (Glykys et al., 2008). Extrasynaptically, the α5 subunit colocalizes with radixin, a cytoskeletal protein linking actin to the plasma membrane (Loebrich et al., 2006).

The α5 subunit is suggested to play an important role in hippocampal pyramidal neuron development (Marchionni et al., 2007; Giusi et al., 2009). However, the contribution of α5-containing GABA_AR localization to neuronal development is unknown, and the mechanism for synaptic accumulation of the α5 subunit has not been established. We show here that surface α5 GABA_ARs significantly colocalize with gephyrin, with α5 synaptic content being maintained at a constant ratio throughout development. We then demonstrate that the α5 subunit directly interacts with gephyrin, and that disruption of Residues 370–385 of the α5 subunit alters both α5 association with gephyrin and receptor synaptic/extrasynaptic localization, but is not required for cluster formation or maintenance of surface receptor number. Finally, we show that balanced distribution of the α5 subunit at synaptic and extrasynaptic sites is important for proper dendritic outgrowth and spine maturation. Together these results support a synaptic role in neuronal development for the α5 GABA_AR subtype.

METHODS

DNA Constructs and Antibodies

The α5^WT, α5^α4GBD, α5^α2GBD, and RFP (red monomeric fluorescent protein mCherry) tagged gephyrin construct were generated by PCR cloning (Li et al., 2011) from CDNAs (kindly provided by Stephen J. Moss). The pH-sensitive GFP tagged α2 GABA_AR subunit (α2^WT) has been described previously (Tretter et al, 2008). All constructs were fully sequenced. The following antibodies were used: rabbit anti-α5 (Cat # 224503), mouse anti-gephyrin (RRID: AB_887719), rabbit anti-VGAT (RRID: AB_887871), mouse anti-α1 (RRID: AB_10597955), and rabbit anti-α2 (RRID: AB_2108839; Synaptic Systems, Gottingen, Germany); rabbit anti-radixin (Cat # R3653, Sigma, St. Louis, MO); chicken anti-GFP (RRID: AB_10000240, Aves Labs, Tigard, OR); rabbit anti-GFP (RRID: AB_221570, Life Technologies, Carlsbad, CA); mouse anti-RFP (RRID: AB_1141717, Abcam, Cambridge, MA) and secondary antibodies for immunofluorescence (goat anti-rabbit Alexa Fluor 568 RRID: AB_143011; goat anti-rabbit Alexa Fluor 488 RRID: AB_10562715; Goat anti-mouse Alexa Fluor 647 RRID: AB_141725; goat anti-chicken Alexa Fluor 488 Cat # 11309, Life Technologies, Carlsbad, CA). Rabbit IgG (RRID: AB_737197), mouse IgG (RRID: AB_737182), and goat anti-α5 (RRID: AB_2109314) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA).

Cell Culture and Transfection

All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Hippocampal neurons were prepared from embryonic day 18 rats and nucleofected (Lonza) at plating (Jacob et al., 2005). HEK-293 cells were also transfected using nucleofection.

Immunofluorescence, Imaging, and Analysis of Cells

Surface GABA_AR α5 or anti-GFP staining was done under nonpermeabilized conditions, followed by permeabilization and staining for α5, GFP, and/or gephyrin as previously described (Jacob et al., 2005). Images were taken on a Nikon A1 Confocal microscope using a 60× oil immersion objective (NA 1.49) at a 2× zoom. Samples for confocal imaging were sequentially scanned with individual lasers (488, 561, and 640 nm) and an appropriate emission band pass filter (500–550 and 575–625 nm) or long pass filter (650LP) to avoid any spectral bleed-through between channels. Data were analyzed using NIS Elements software (Nikon, NY), with colocalization determined as previously described (Tretter et al., 2008). In the gephyrin and α5 GABA_AR colocalization studies, 20-μm length regions (ROIs) from two to three proximal dendrites per neuron were used for analysis. For comparison between α5^WT, α5^α4GBD, and α5^α2GBD, the researcher was blinded to the experimental conditions. For cluster analysis, thresholding was kept constant between cultures. For total α5 surface fluorescence measurements (both clustered and diffuse receptors), fluorescence values in the same ROIs were measured without thresholding. 3D reconstructions of confocal z series acquired with a 40× oil immersion objective, 1× zoom, were used for Sholl analysis, and analyzed using ImageJ software. Spine morphology studies were performed as previously described (Jacob et al., 2009). Briefly, images were analyzed from 3D reconstructions of confocal z series acquired with a 60× oil immersion objective using a 4× zoom. Spine length was determined using NIS Elements by measuring the length from the spine head to where the spine intersected the dendritic shaft. We classified spines as either mushroom or filopodia, with a mushroom spine having a spine head twice the size of the spine neck, and the remaining spines classified as filopodia.

Immunoprecipitation and Western Blots

For immunoprecipitation from hippocampal neuronal cultures, neurons were nucleofected with α5^WT at plating. At DIV 14, cells were lysed in Buffer A containing 50 mM Tris-HCl (pH 7.6), 50 mM NaCl, 1 mM EDTA, 0.25% Igepal, 10 mM NaF, 2 mM sodium orthovanadate, and protease inhibitor cocktail (Sigma, St. Louis, MO), solubilized for 15 min at 4°C, and spun at 1,000g for 10 min. Extracts were then immunoprecipitated with either rabbit anti-GFP (5 μg), or rabbit IgG as a negative control. For hippocampal and cortical rat tissue, adult female rats were killed using CO₂ followed by cervical dislocation. The hippocampus and cortex were dissected and homogenized as previously described (Brady et al., 2013), diluted approximately fivefold in Buffer A, and immunoprecipitated with rabbit anti-α5 antibody crosslinked to Protein A Sepharose beads with BS³ (Thermo Scientific, Rockford, IL). The beads were eluted by incubating beads twice with 50 mM glycine for 5 min, as previously described (Sousa et al., 2011), and SDS-PAGE loading buffer was added. For immunoprecipitation from HEK cells, HEK cells were transfected with RFP-gephyrin, an untagged β3 subunit and either a α2^WT or α5^WT construct, and lysates were immunoprecipitated with rabbit anti-GFP as described above.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism 5 (San Diego, CA) on 3–4 independent cultures, 12–18 neurons per culture, using two way ANOVA with Tukey post hoc test for dendritic morphology analysis or one-way ANOVA with paired t-test post-hoc tests for all other studies. All values are given as mean ± standard error of the mean.

RESULTS

Inhibitory Synapses Contain a Constant Level of α5-Containing GABA_ARs Throughout Development

GABA_ARs containing the α5 subunit are generally considered to be extrasynaptic; however, previous studies have found some evidence for synaptic localization (Christie and de Blas, 2002; Serwanski et al., 2006). We first examined whether a pool of endogenous α5 GABA_ARs was synaptically located and if receptor localization changes throughout development. Hippocampal neurons were fixed and stained for surface α5 and gephyrin, the postsynaptic inhibitory scaffold protein (Jacob et al., 2005; Flores and Mendez, 2014; Tyagarajan and Fritschy, 2014). Analysis of the number of gephyrin clusters colocalized with α5 found a high level of colocalization, with approximately a 60% association maintained at a uniform level throughout development [DIV 7 62.67 ± 7.38, DIV 14 53.34 ± 4.62, DIV 21 72.34 ± 9.06; Fig. 1(A,C)]. These data support a role for synaptically located α5 GABA_ARs contributing to GABAergic inhibition, in addition to the established role of extrasynaptic α5 GABA_ARs.

Figure 1.

GABA_AR α5 association with gephyrin throughout development in hipppocampal neurons. (A) DIV 7, 14, and 21 neurons were fixed and immunostained for surface α5 GABA_AR (anti-α5, green, non permeabilized conditions) and gephyrin (red, permeabilized conditions). Colocalized clusters are indicated by arrowheads. Scale bar = 10 μm. (B) Neurons expressing α5^WT were fixed and stained for surface tagged GABA_AR (anti-GFP, green, nonpermeabilized conditions) and gephyrin (red). (C) The percentage of gephyrin clusters colocalized with surface clusters of endogenous α5 or tagged α5 GABA_AR remained constant throughout development. (D) The number of surface α5 GABA_AR and gephyrin clusters in 20 μm dendritic segments in nontransfected and transfected cells was compared. No significant difference was found between groups for either α5 or gephyrin.

To further investigate mechanisms of α5 GABA_AR synaptic localization, we then introduced a myc tag and a pH-sensitive GFP variant at the N-terminus of the α5 subunit (α5^WT). It has been shown that insertion of these tags into other GABA_AR subunits are functionally silent (Jacob et al., 2005; Tretter et al., 2008, 2011; Mukherjee et al., 2011). To confirm that these tags do not induce major alterations, we examined the total number of surface α5 containing GABA_AR clusters and the total number of gephyrin clusters between nontransfected and transfected neurons at all developmental stages [Fig. 1(A,B,D)]. We found no significant difference between groups, indicating that GABA_ARs containing tagged α5 subunits do not induce changes in the development of GABAergic synapses. We then examined synaptic localization of the α5^WT construct. Similar to nontransfected neurons, the number of gephyrin clusters associated with α5^WT remains constant throughout development [DIV 7 59.24 ± 7.95, DIV 14 59.25 ± 7.25, DIV 21 62.76 ± 6.45; Fig. 1(B,C)]. As the α5 subunit has been shown to associate with radixin in the extrasynaptic membrane at DIV 13–14 (Loebrich et al., 2006), surface α5^WT colocalization with radixin was also assessed: there was a significant decrease in association of radixin and α5 between DIV 7 and DIV 21 (DIV 7 70.45 ± 10.12%, DIV 21 26.60 ±9.05%, p = 0.0270; Supporting Information Fig. S1). These data suggest that during this developmental period, while α5-radixin association decreases over time, the number of synapses (as represented by gephyrin) with α5-containing GABA_ARs remains constant. In summary, these data show that a major portion of α5 GABA_ARs is observed at synaptic locations throughout the key period of synaptogenesis and in an established hippocampal neuron culture circuit.

α5-Containing GABA_AR and Gephyrin Are Intimately Associated in Cultured Neurons and In Vivo

To further demonstrate a close association of α5 with gephyrin in neurons, we performed immunoprecipitations on hippocampal neuron cultures transfected with the α5^WT construct. Immunprecipitation of the tagged α5 subunit with an anti-GFP antibody also coimmunprecipitated gephyrin [Fig. 2(A)]. To confirm that α5 and gephyrin interact in vivo, endogenous α5 was immunoprecipitated from adult rat hippocampal and cortical tissue, and the association of gephyrin was determined. As shown in Figure 2(B), the α5 subunit coimmunoprecipitated gephyrin, confirming that these two proteins are associated in vivo.

Figure 2.

The GABA_AR α5 subunit is able to directly associate with gephyrin. (A) Cultured hippocampal neurons were transfected with α5^WT and lysates were immunoprecipitated with anti-GFP or anti-IgG. Precipitates were immunoblotted for GFP and gephyrin. Immunoprecipitation of GFP resulted in coimmunoprecipitation of gephyrin. (B) Cortical and hippocampal (HPC) lysates from adult rat brain were immunoprecipitated with either anti-α5 or anti-IgG and immunoblotted for α5 and gephyrin. In both cortex and HPC, gephyrin coimmunoprecipitates with α5 (n = 3 animals). (C) Neurons were transfected with RFP-gephyrin and α2^WT, then fixed, stained under nonpermeabilized conditions with anti-GFP, then permeabilized and stained with anti-RFP. (D) Neurons were transfected with RFP-gephyrin, fixed, permeabilized, and stained for RFP and VGAT. (E) HEK cells were transfected with α5^WT, β3, and RFP-gephyrin. Lysates were then immunoprecipitated for GFP and immunoblotted for RFP and GFP. GFP immunoprecipitation resulted in coimmunoprecipitation of RFP-gephyrin. HEK cells transfected with α2^WT, β3, and RFP-gephyrin were used as a positive control.

The α5 Subunit Intracellular Domain Is Sufficient for Gephyrin Binding

Our data indicate that the α5 subunit is associating with gephyrin, although this could occur solely via a direct interaction of the α5 subunit with gephyrin (via the α5 intracellular domain), or via α5 interactions with other proteins capable of directly interacting with gephyrin, such as another α subunit. A potential weak direct interaction between the α5 intracellular loop and gephyrin was suggested by minimal binding in a GST overlay assay using an α5 intracellular loop GST-fusion protein overlaid with ³⁵S-methionine labeled gephyrin (Mukherjee et al., 2011). To further examine the mechanism for α5 GABA_AR synaptic localization with a complete α5 subunit, we cotransfected the α5^WT and an RFP-gephyrin construct into HEK cells, along with an untagged β3 construct (surface expression of functional GABA_ARs requires a β subunit; Connolly et al., 1996). As a positive control, a pH-sensitive GFP tagged α2^WT subunit was cotransfected with RFP-gephyrin and the β3 subunit in a separate set of HEK cell experiments. In neurons the RFP-gephyrin exhibits equivalent subcellular localization as endogenous gephyrin, being highly colocalized with synaptic α2 GABA_AR and closely apposed to the presynaptic inhibitory terminal maker VGAT [vesicular GABA/glycine transporter; Fig. 2(C,D)], indicating its ability to interact with GABA_ARs. Immunoprecipitation of HEK lysates with anti-GFP resulted in coimmunoprecipitation of RFP-gephyrin by both α2^WT and α5^WT [Fig. 2(E)], indicating that these proteins are associated in a heterologous system, supporting a direct interaction between the α5 and gephyrin rather than via α1 or α2 subunits.

The α5 Gephyrin Binding Domain Contributes to Synaptic Localization of α5 GABA_AR

GABA_AR α1, α2, and α3 subunits localize to the synapse via interaction with gephyrin through specific domains in the subunit intracellular loops between transmembrane Domains 3 and 4 (Tretter et al., 2008, 2011; Mukherjee et al., 2011), whereas extrasynaptic α4 or α6 GABA_ARs do not colocalize or interact with gephyrin (Sun et al., 2004; Kralic et al., 2006). To determine if the α5 sequence (Residues 370–385) homologous to the gephyrin binding domains in other alpha subunits enables α5 interaction with gephyrin, we created two different α5 chimeras; one in which the predicted gephyrin binding domain (GBD; Mukherjee et al., 2011) of α5 was replaced with the homologous region from α4 (α5^α4GBD) and another in which the α5 GBD was replaced with the α2 GBD [α5^α2GBD; Fig. 3(A)]. Chimeric GABA_AR expression levels, assembly and surface trafficking were equivalent to α5^WT in HEK cells (Supporting Information Fig. S2). We then transfected neurons with the α5^WT, α5^α4GBD, or α5^α2GBD constructs and analyzed surface GABA_AR α5 subunit clustering and synaptic localization at DIV 14. In neurons expressing the chimeric receptors, both surface tagged α5 GABA_AR clusters and gephyrin overall cluster attributes were unchanged, indicating normal synaptogenesis: the total number and sum intensity of clusters were not significantly different [Fig. 3(C,D,F,G)]. In addition, the total surface fluorescence of α5 (both clustered and diffuse), was unchanged between groups, indicating no apparent change in receptor trafficking [Fig. 3(E)]. We found a significant decrease in surface receptor colocalization with gephyrin in neurons transfected with α5^α4GBD compared to α5^WT and α5^α2GBD, and a significant increase in colocalization in neurons transfected with α5^α2GBD compared to α5^WT [α5^WT 60.80 ± 3.16%; α5^α4GBD 45.29 ± 2.64%; α5^α2GBD 74.68 ± 5.06%; α5^WT vs. α5^α4GBD p = 0.0055, α5^WT vs. α5^α2GBD p = 0.048, α5^α4GBD vs. α5^α2GBD p = 0.0012, Fig. 3(B,H)]. Gephyrin clusters colocalized with α5 also exhibited corresponding changes: cluster sum area [α5^WT 0.54 ± 0.045; α5^α4GBD 0.32 ± 0.046; α5^α2GBD 1.15 ± 0.063; α5^WT vs. α5^α4GBD p = 0.0096, α5^WT vs. α5^α2GBD p = 0.0002, α5^α4GBD vs. α5^α2GBD p < 0.0001, Fig. 3(B,I)] and sum intensity normalized to control [α5^WT 100; α5^α4GBD 56.89 ± 7.52; α5^α2GBD 189.10 ± 31.81; α5^WT vs. α5^α4GBD p = 0.0004, α5^WT vs. α5^α2GBD p = 0.0086, α5^α4GBD vs. α5^α2GBD p = 0.002, Fig. 3(B,J)]. The colocalization data indicate that replacement of the putative gephyrin binding domain in α5 with α2GBD, representing a high affinity gephyrin binding site, increased α5 synaptic levels, while disruption with the α4GBD decreased, but did not eliminate, α5 synaptic levels. The remaining association of α5 and gephyrin seen here may be due to a gephyrin interaction with an endogenous α1 or α2 subunit, as previous studies have shown mixed subunit populations occurring in vivo (Araujo et al., 1999; del Rio et al., 2001). In summary, these data show that the α5 gephyrin binding domain significantly contributes to α5 GABA_AR synaptic localization.

Figure 3.

Disruption of the gephyrin binding domain in the α5 subunit decreases its association with gephyrin. (A) Alignment of homologous regions of the gephyrin binding domain (GBD) of different alpha subunits, and construction of the α5^α4GBD and α5^α2GBD chimeras. (B) DIV 14 neurons expressing the α5^WT, α5^α4GBD, or α5^α2GBD constructs were fixed and stained for surface GABA_AR (anti-GFP, green) and gephyrin (red). (C–E) In transfected neurons, there was no significant difference in the total number of α5 clusters per 20 μm, the total sum intensity of α5 clusters, or the total α5 surface fluorescence. (F and G) There was no significant difference in the total number of gephyrin clusters per 20 μm or the overall sum intensity of gephyrin between groups. (H) Swapping the α5 GBD for the α4 GBD significantly decreased colocalization of the α5 subunit with gephyrin, while exchanging the α5 GBD for the α2 GBD significantly increased gephyrin colocalization (*p < 0.05, **p < 0.01). (I and J) The sum area and sum intensity of gephyrin clusters colocalized with tagged GABA_AR was significantly decreased in neurons transfected with the α5^α4GBD compared to control, while neurons transfected with α5^α2GBD showed a significant increase in sum area and sum intensity of colocalized gephyrin clusters (**p < 0.01, ***p < 0.001, and ****p < 0.0001).

Proper Localization of α5-Containing GABA_ARs Is Important for Neuronal Development

As previous genetic and pharmacological studies have implicated a role of α5-containing GABA_ARs in proper neuronal development (Curia et al., 2009; Giusi et al., 2009; Fatemi et al., 2010) we examined the effect of α5 redistribution on neuronal morphology at DIV 14 and 21. Neurons were transfected with the α5^WT, α5^α4GBD, or α5^α2GBD constructs at plating, then fixed, permeabilized, and stained with anti-GFP and anti-MAP-2. 3-D reconstructions of confocal z stacks were analyzed by Sholl analysis using ImageJ. At DIV 14, α5^α2GBD neurons exhibited a significant decrease in the number of intersections compared to α5^WT and α5^α4GBD [two-way ANOVA significant effect of transfection p = 0.0116; Tukey post hoc test α5^α2GBD vs. α5^WT p < 0.05; α5^α2GBD vs. α5^α4GBD p < 0.05; Fig. 4(A,C)]. At DIV 21, we observed a significant decrease in the number of intersections between α5^α2GBD neurons compared to α5^WT and α5^α4GBD [two-way ANOVA significant effect of transfection p < 0.0001; Tukey post hoc test α5^α2GBD vs. α5^WT p < 0.05; α5^α2GBD vs. α5^α4GBD p < 0.0001 Fig. 4(B,D)], and a significant increase in the number of intersections between α5^α4GBD and α5^WT [p < 0.05 Fig. 4(B,D)]. Further analysis found no significant difference in number of primary, secondary, or tertiary dendrites, or total dendritic length, either at DIV 14 or 21 (Supporting Information Fig. S3). Together, these data indicate that shifting α5 localization into the synapse via the α2 GBD impairs dendritic outgrowth, while moving the α5 subunit out of the synapse via the α4 GBD increases dendritic outgrowth.

Figure 4.

α5 localization is important in controlling dendritic outgrowth. Sholl analysis was performed on hippocampal cultures transfected with the α5^WT, α5^α4GBD, or α5^α2GBD, then fixed and stained under permeabilized conditions with anti-GFP and anti-MAP-2 antibodies. Confocal z-series were acquired with a 40× objective, and 3D constructions were used to analyze dendritic morphology. (A and C) At DIV 14, neurons transfected with α5^α2GBD exhibited impaired dendritic growth compared to those transfected with α5^WT or α5^α4GBD (*p < 0.05 compared to α5^WT, ⁺p < 0.05 compared to α5^α4GBD; number of dendrites examined: α5^WT 1054, α5^α4GBD 1076, α5^α2GBD 1020). (B and D) At DIV 21, neurons transfected with α5^α4GBD showed more dendritic intersections than either α5^WT or α5^α2GBD, while neurons expressing α5^α2GBD showed fewer dendritic intersections (*p < 0.05 compared to α5^WT, ** p < 0.01 compared to α5^WT, ⁺⁺⁺⁺p < 0.0001 compared to α5^α4GBD; number of dendrites examined: α5^WT 1126, α5^α4GBD 1213, α5^α2GBD 1269).

We also examined the effect of α5 mislocalization on dendritic spine morphology, since alterations in GABA_AR surface levels have been shown to alter spine maturation (Jacob et al., 2009). At DIV 14, while overall spine length and spine density were unchanged (Supporting Information Fig. S3), there was a significant decrease in the number of mushroom spines in α5^α4GBD neurons compared to α5^WT and α5^α2GBD [α5^α4GBD 2.502 ± 0.16 spines/10 μm, α5^WT 4.48 ± 0.46 spines/10 μm, α5^α2GBD 4.50 ± 0.38 spines/10 μm; α5^α4GBD vs. α5^WT p = 0.016; α5^α4GBD vs. α5^α2GBD p = 0.0083, Fig. 5(A,C)], which led to a shift in mushroom/filopodia ratio in α5^α4GBD neurons compared to α5^WT and α5^α2GBD [α5^α4GBD 0.73 ± 0.033, α5^WT 1.66 ± 0.24, α5^α2GBD 2.34 ± 0.38; α5^α4GBD vs. α5^WT p = 0.019; α5^α4GBD vs. α5^α2GBD p = 0.0131, Fig. 5(A,B)]. At DIV 21, the number of mushroom spines in α5^α4GBD neurons were still reduced compared to α5^WT [α5^α4GBD 3.50 ± 0.37, α5^WT 4.90 ± 0.32, α5^α2GBD 4.46 ± 0.23; α5^α4GBD vs. α5^WT p = 0.0461, Fig. 5(E,G)], again leading to a decrease in the mushroom/filopodia ratio [α5^α4GBD 1.02 ± 0.14, α5^WT 2.45 ± 0.34, α5^α2GBD 1.56 ± 0.30; α5^α4GBD vs. α5^WT p = 0.018, Fig. 5(E,F)]. Neither spine density nor lengths were altered at DIV 21 (Supporting Information Fig. S4). The decrease in the mushroom/filopodia ratio suggests that a shift of α5-containing GABA_ARs out of the GABAergic synapse results in a less mature spine, and that this immature phenotype is maintained throughout development.

Figure 5.

Altered ratio of synaptic/extrasynaptic α5 GABA_AR disrupts spine maturation. Neurons were transfected with α5^WT, α5^α4GBD, or α5^α2GBD, then fixed and stained under permeabilized conditions with anti-GFP and anti-MAP-2 antibodies. High magnification confocal z-series through dendritic regions were obtained, and 3D reconstructions were used to analyze spine length, density, and morphology. 3D reconstructions of confocal images from neurons expressing α5^WT, α5^α4GBD, or α5^α2GBD at DIV 14 (A) and DIV 21 (E). (B) At DIV 14, neurons expressing α5^α4GBD exhibited a significantly lower mushroom/filopodia spine ratio compared to either α5^WT or α5^α2GBD (for B–H, *p < 0.05, **p < 0.01; number of spines examined: α5^WT 394, α5^α4GBD 311, α5^α2GBD 358). (C) Neurons expressing α5^α4GBD exhibited significantly fewer mushroom spines per 10 μm compared to either α5^WT or α5^α2GBD. (F) At DIV 21, neurons expressing α5^α4GBD exhibited a significantly lower mushroom/filopodia spine ratio compared to α5^WT (number of spines examined: α5^WT 474, α5^α4GBD 453, α5^α2GBD 455). (G) Neurons expressing α5^α4GBD exhibited significantly fewer mushroom spines per 10 μm compared to α5^WT. There was no significant difference between the number of filopodia spines per 10 μm between constructs at DIV 14 (D) or at DIV 21 (H).

DISCUSSION

α5 GABA_ARs are emerging as key contributors to learning and memory processes and potential targets for pharmacological modulation in treating cognitive and neurodevelopmental disorders, with clinical trials ongoing in Down syndrome treatment (Rudolph and Mohler, 2014). Extrasynaptically localized α5 GABA_ARs generate tonic inhibition in the hippocampus. However, the function and contribution of synaptically localized α5 GABA_AR to inhibition has largely been overlooked. To resolve this issue, we investigated the localization of the α5 GABA_AR sub-type during development, the mechanism of α5 restriction at synapses, and how its regulated distribution at synaptic versus extrasynaptic sites contributes to the function of α5 in neuronal development. We found that a significant portion of GABAergic post-synaptic compartments contain the α5 GABA_AR subunit, both in vitro and in vivo. Surface α5 GABA_ARs are significantly colocalized with gephyrin in hippocampal neurons, with α5 synaptic content being maintained at a constant ratio from 7 to 21 DIV. The association between gephyrin and the α5 subunit occurs independent of other α subunits that bind gephyrin and requires Residues 370–385 of the α5 subunit. By exchanging these residues with the corresponding residues of either the high affinity gephyrin binding domain of the α2 subunit or homologous residues from the extrasynaptic α4 subunit that does not interact with gephyrin, we created chimeras that shifted the localization of the α5 subunit without changing α5 total surface levels, thus altering the synaptic/extrasynaptic ratio of α5. Redistribution of the α5 subunit had significant functional consequences, as these chimeras disrupted proper neuronal dendritic morphology and spine maturation.

A role for α5 GABA_AR in dendritic development is supported from in vitro pharmacological and genetic studies. α5-inverse agonist treatment of hippocampal neurons decreased dendritic arborization and reduced glutamate receptor expression (Giusi et al., 2009). Furthermore, neurodevelopmental disorders, such as autism and Fragile X syndrome, have demonstrated alterations in α5 subunit levels (Curia et al., 2009; Mendez et al., 2013). However, the precise function of α5-containing GABA_AR signaling, particularly of the prevalent synaptic α5 GABA_AR population identified here, in dendritic outgrowth is not known. The α5^α4GBD and α5^α2GBD chimeras provided the opportunity to assess the contributions of α5 synaptic and extrasynaptic signaling to neuronal development. Replacing the α5 GBD with the α2 GBD increased α5-containing GABA_AR synaptic localization similar to levels previously reported for the endogenous α2 subunit (Tretter et al., 2008). We found that shifting the α5 subunit into the synapse reduced the number of dendritic intersections in Sholl analysis compared to wild type at DIV 14 and 21, indicating impaired dendritic growth. Conversely, moving the α5 subunit out of the synapse, by way of the α5^α4GBD chimera, increased the number of intersections at DIV 21. Previous studies have shown that both tonic and phasic GABAergic inhibition can control Ca²⁺ transients in dendrites (Pan and Lipton, 1995; Kanemoto et al., 2011; Hayama et al., 2013), making this a possible mechanism by which altering the synaptic and extrasynaptic α5 levels may influence dendritic outgrowth. While the increased dendritic outgrowth observed at DIV 21 in neurons transfected with the α5^α4GBD chimera implies enhanced neuronal maturation, it is also possible that decreasing the ratio of synaptic/extrasynaptic α5 GABA_AR signaling reduced dendritic pruning. Few studies have examined the role of GABA_AR in dendrite pruning, although one study found that propofol, an anesthetic agent that binds to GABA_AR β subunits and potentiates GABA_AR, results in GABA_AR-dependent neurite retraction in primary cortical cultures (Turina et al., 2008). It is also known that local Ca²⁺ transients, both through ion channels and intracellular Ca²⁺ stores, can help to stabilize dendrites, and alterations in these transients may result in dendrite retraction (Wong and Ghosh, 2002; Kanamori et al., 2013). A recent study found that glycogen synthase kinase 3β (GSK3β) can regulate dendrite retraction in hippocampal cultures between DIV 14–18 by regulating γ2 surface expression (Rui et al., 2013). In neurons expressing α5^α4GBD, the shift of the α5 subunit from the synapse to the extrasynaptic membrane may alter the overall excitatory/inhibitory balance, which in turn alters local Ca²⁺ signaling and affects dendritic pruning.

We also found that altering the ratio of synaptic/extrasynaptic α5 GABA_AR affected spine maturation. Shifting the α5 subunit out of the synapse resulted in a less mature dendritic spine phenotype. This suggests a specific role for α5 synaptic signaling in spine development, as removing extrasynaptic α5 via α5^α2GBD had no significant effect on spine maturity. As with dendritic morphology, the role of synaptic and extrasynaptic GABAergic signaling has not been fully examined in spine development. GABAergic inhibition, primarily through GABAergic synaptic signaling, can focally suppress Ca²⁺ transients on individual spines (Chiu et al., 2013), and induce spine shrinkage or elimination (Hayama et al., 2013). Furthermore, alterations to other GABA_AR subunits, such as the α1 and β3 subunits, disrupted spine maturation in neurons (Heinen et al., 2003; Jacob et al., 2009). Less mature spines show enhanced Ca²⁺ diffusion between the spine head and dendrite resulting in lower spine calcium accumulation (Zito et al., 2009), which could also play a role in the changes in dendritic morphology seen here.

A large body of evidence indicates that alpha subunit type plays a direct role in controlling GABA_AR localization via gephyrin interactions (Tretter et al., 2008, 2011; Saiepour et al., 2010; Mukherjee et al., 2011; Wu et al., 2012). It has been hypothesized that gephyrin trimers provide multiple binding sites (Sola et al., 2004; Fritschy et al., 2008; Tretter et al., 2012; Specht et al., 2013), such that two subunits within a GABA_AR may bind the gephyrin scaffold. Our study supports a direct interaction between the α5 subunit and gephyrin, which would alter receptor avidity for gephyrin and diffusion within and out of the synapse. This has implications for both α5/α5 and mixed alpha subunit GABA_AR, which exist in vivo (Araujo et al., 1999; del Rio et al., 2001). To assess alpha subunit synaptic localization independent of receptor oligomers, a recent study compared α5 or α2 subunit intracellular domains fused to a single pass transmembrane domain, revealing reduced colocalization of the α5 chimera with gephyrin and a higher diffusion rate (Gerrow and Triller, 2014). This has significant implications for the receptor composition of inhibitory synapses, as a mixed α2/α5 receptor would diffuse out of the synapse faster than a homogenous α2/α2 receptor, but slower than an α5/α5 receptor. This subtle tuning of synaptic GABAergic signaling could contribute to neuronal development by modulating local spine and dendritic Ca²⁺ influx.

In conclusion, we found that a significant portion of GABAergic postsynaptic compartments contain the α5 GABA_AR subunit, both in vitro and in vivo. Shifting the synaptic/extrasynaptic localization of α5 GABA_AR through substitution of the α5 gephyrin binding domain disrupts proper neuronal development, indicating that an appropriate balance of phasic and tonic α5 GABAergic inhibition regulates dendritic outgrowth and spine maturation. Further research is needed to elucidate the precise roles of synaptic and extrasynaptic α5 GABA_AR at early stages of neuronal circuit development and later in learning and memory.