Phosphorylation on Ser 359 of the α2 subunit in GABA type A receptors down-regulates their density at inhibitory synapses

GABA type A receptors (GABA A Rs) mediate fast synaptic inhibition and are trafficked to functionally diverse synapses. However, the precise molecular mechanisms that regulate the synaptic targeting of these receptors are unclear. While it has been previously shown that phosphorylation events in α4, β and γ subunits of GABA A Rs govern their function and trafficking, phosphorylation of other subunits has not yet been demonstrated. Here, we show that the α2 subunit of GABA A Rs is phosphorylated at Ser 359 and enables dynamic regulation of GABA A R binding to the scaffolding proteins gephyrin and collybistin. We initially identified Ser 359 phosphorylation by MS analysis, and additional experiments revealed that it is regulated by the activities of cAMP-dependent protein kinase (PKA) and the protein phosphatase 1 (PP1) and/or PP2A. GST-based pulldowns and coimmunoprecipitation experiments demonstrate preferential binding of both gephyrin and collybistin to WT and an S359A phosphonull variant, but not to an S359D phosphomimetic variant. Furthermore, the decreased capacity of the α2 S359D variant to bind collybistin and gephyrin decreased the density of synaptic α2-containing GABA A R clusters and caused an absence of α2 enrichment in the axon initial segment. These results suggest that PKA-mediated phosphorylation and PP1/PP2A-dependent dephosphorylation of the α2 subunit play a role in the dynamic regulation of GABA A R accumulation at inhibitory synapses, thereby regulating the strength of synaptic inhibition. The MS data have been deposited to ProteomeXchange, with the dataset identifier <PXD019597>.

ABSTRACT GABA type A receptors (GABAARs) mediate fast synaptic inhibition and are trafficked to functionally diverse synapses. However, the precise molecular mechanisms that regulate the synaptic targeting of these receptors are unclear. While it has been previously shown that phosphorylation events in α4, β and γ subunits of GABAARs govern their function and trafficking, phosphorylation of other subunits has not yet been demonstrated. Here, we show that the α2 subunit of GABAARs is phosphorylated at Ser 359 and enables dynamic regulation of GABAAR binding to the scaffolding proteins gephyrin and collybistin. We initially identified Ser 359 phosphorylation by MS analysis, and additional experiments revealed that it is regulated by the activities of cAMP-dependent protein kinase (PKA) and the protein phosphatase 1 (PP1) and/or PP2A. GST-based pulldowns and coimmunoprecipitation experiments demonstrate preferential binding of both gephyrin and collybistin to WT and an S359A phosphonull variant, but not to an S359D phosphomimetic variant. Furthermore, the decreased capacity of the α2 S359D variant to bind collybistin and gephyrin decreased the density of synaptic α2containing GABAAR clusters and caused an absence of α2 enrichment in the axon initial segment. These results suggest that PKAmediated phosphorylation and PP1/PP2Adependent dephosphorylation of the α2 subunit play a role in the dynamic regulation of GABAAR accumulation at inhibitory synapses, thereby regulating the strength of synaptic inhibition. The MS data have been deposited to ProteomeXchange, with the dataset identifier <PXD019597>. most synaptic GABAARs comprise two α1, α2 or α3 subunits, two β2 or β3 subunits and one γ2 subunit (3,4). Each subunit is composed of a large N-terminal domain, four transmembrane (TM) domains, and a large intracellular loop between TM3 and 4 that is the site of multiple proteinprotein interactions (5)(6)(7). Despite the high structural homology between GABAAR αsubunits, they are selectively targeted to distinct types of inhibitory synapses. Indeed, α2containing GABAAR are enriched at synapses on the axon initial segment (AIS) whereas α1containing receptors are more evenly distributed between the AIS and dendrites (8,9). It has been proposed that the α-subunits play key roles in the selective subcellular targeting of GABAARs (10). However, the molecular mechanisms that regulate precise targeting of these receptors remains unknown.
The post-translational modification is one mechanism by which GABAARs are regulated. Notably, the phosphorylation of key residues on GABAARs have been previously shown to regulate protein interactions that govern trafficking and surface stability of receptors (11,12); however phosphorylation of GABAAR α2-subunit has not been demonstrated, and its significance is yet to be determined (13).
Protein interactions fundamental to receptor clustering at the postsynaptic domain include GABAAR α2-subunit interactions with gephyrin and collybistin. Gephyrin is the canonical inhibitory synaptic scaffold protein that is highly enriched at GABAergic postsynapses where it colocalises with synaptic GABAARs (14). The critical role for gephyrin in the formation of postsynaptic inhibitory synapses was highlighted through the utilisation of knockout and RNA interference tools which demonstrated the loss of gephyrin and subsequent loss of synaptic GABAAR clusters (15,16). Moreover, GABAAR mutations that reduce gephyrin binding also result in fewer synaptic clusters (17). Collybistin is a Rho guanine nucleotide exchange factor (GEF) that binds both gephyrin and GABAAR α2 (18). The importance of collybistin in targeting gephyrin to the inhibitory synapse was demonstrated in collybistin deficient mice, which exhibit a loss of gephyrin and GABAAR puncta (19). Further, mutations in GABAARs which disrupt or enhance collybistin binding resulted in corresponding decreased or increased GABAAR-containing synapses at the AIS (20,21).
In this study, we utilised mass spectrometry to identify a novel phosphorylation site at serine 359 on the GABAAR α2 subunit. This residue is situated in the large intracellular loop between TM3 and 4 of the α2-subunit, close to the overlapping binding sites of gephyrin and collybistin. S359 is phosphorylated by PKA and dephosphorylation of this site is dependent on PP1/PP2A. Phosphomimetic mutation of this site decreased receptor binding to both gephyrin and collybistin, resulting in a reduced number of synaptic α2 clusters in dendrites and an absence of α2 enrichment in the AIS. These results show that the phospho-regulation of α2 by PKA and PP1/PP2A allows neurons to adjust the efficacy of gephyrin and collybistin binding, which plays a fundamental role in regulating the density of inhibitory synapses at dendrites and the enrichment of inhibitory synapses at the AIS.

Results
Identification of serine 359 on the GABAAR α2 subunit as a novel phosphorylation site. To identify novel phosphorylation sites on the GABAAR α2 subunit we utilised previously characterised myc/pH-sensitive GFP (pHluorin)tagged GABAAR α2 subunit (pHα2) knock-in mice (22). Hippocampi and cortex from age-and sex-matched WT and pHα2 mice were solubilised and incubated with GFP-Trap beads to immunoprecipitate pHα2. After extensive washes, bound material was eluted and subject to SDS-PAGE followed by Coomassie staining. An 80 kDa band corresponding to pHα2 was excised and analysed by LC-MS/MS (Fig. 1A, Supplementary Fig. S1). Phosphorylation sites were identified by the Taplin Facility using the Ascore algorithm (23). Two phosphopeptides were identified corresponding to residues S359 and S379 on the GABAAR α2 subunit ( Fig.  1B/C). Phosphorylated S359 peptides were found in all four independent purifications, whereas S379 was only found in one of four experiments. Both potential phosphorylation sites are located in the large intracellular loop of the receptor. Further, this site is conserved in the human, mouse and rat sequence of the α2 subunit (Fig.  1D). Given the more robust phosphorylation of Ser359 we focused on this site.

Phosphorylation of purified GST-α2.
To further corroborate our MS findings, in vitro kinase assays were performed with GST alone or GST fused to the intracellular loop of the GABAAR α2 subunit. Since the kinase responsible for phosphorylating this site had yet to be identified, brain lysate was used as a source of kinase. GST, GST-α2 WT or GST-α2 S359A phosphomutant proteins were incubated with brain lysate and 32 Pγ-ATP to measure phosphorylation levels at this site. Phosphorylation of GST alone was not detected in these experiments compared to GST-WT and GST-S359A (p<0.0001, Fig. 2A). The mutant form of the GST-α2 subunit, where S359 was converted to an alanine residue by site-specific mutagenesis, showed significantly decreased levels of phosphorylation compared to WT (WT 1.00±0.03, S359A 0.77±0.03, p=0.0023). The decrease in phosphorylation shown for the GST-α2 S359A mutant strongly suggests that S359 represents a site of phosphorylation in the α2 subunit.
Phosphorylation of this site was further examined by phosphopeptide map analysis. Twodimensional phosphopeptide map analysis revealed two major positively charged phosphopeptides (red and blue circle, Fig. 2B), one of which corresponds to a phosphopeptide containing a phosphorylated S359 (red circle). Together, these results support the MS data in the identification a novel phosphorylation site at S359 in the GABAAR α2 subunit.
Serine 359 is dephosphorylation is PP1/PP2Adependent. Since kinases present in brain lysate are capable of phosphorylating GABAAR α2 at S359, a purified phosphorylation-state specific antibody directed toward S359 was produced (PhosphoSolutions). To test the specificity of this antibody, horizontal slices from 8 week old male pHα2 animals were treated for 30 minutes with 1 μM okadaic acid, a broad-spectrum inhibitor of protein phosphatase 1 (PP1) and protein phosphatase 2A (PP2A). Immunoblotting revealed that the pS359 antibody did not recognize the α2 subunit in crude brain lysates ( Fig. 3A/B input lane). Therefore, we first immunoprecipitated pHα2 from these lysates with GFP-Trap. Phosphorylation at S359 was only detected following okadaic acid treatment (no peptide DMSO 0.069±0.07, OA 1±0.08, p<0.0001; non-phosphopeptide DMSO 0.0007±0.0007, OA 0.76±0.08, p<0.0001, Fig.  3A). Moreover, detection of the α2 S359 band was prevented by incubation of the phosphoantibody with synthetic phospho-S359 peptide (0.00±0.00) compared to no peptide+OA and non-phosphopeptide+OA (no peptide+OA 1±0.07, p<0.0001; non-phosphopeptide+OA 0.76±0.08, p<0.0001). However, incubation with non-phosphorylated S359 peptide showed little interference with the immunodetection of the phosphorylated-α2 subunit (p=0.058). Together, these data indicate that S359 dephosphorylation is dependent on PP1 and/or PP2A to maintain phosphorylation at low levels under basal conditions in the mouse brain.
To further validate the phosphoantibody, similar experiments were performed in neuronal cultures. Rat cortical neurons at 19 DIV expressing pHα2 WT or S359A phosphomutant were treated with okadaic acid (Fig. 3B). Similar to the experiments in slices, immunoprecipitation was required before using this antibody. Further, okadaic acid led to an increase in S359 phosphorylation levels (WT−OA 0.01±0.01, WT+OA 1±0.05, p<0.0001) which was abolished in the S359A mutant (SA+OA 0.00±0.00, p<0.0001). Collectively these experiments indicate that we have successfully produced a phosphospecific α2 S359 antibody. Furthermore, these data corroborate that phosphorylation of this site is kept at low levels under basal conditions, and confirm that its dephosphorylation is regulated by PP1 and/or PP2A.

Discussion
Identification of S359 as a novel phosphorylation site on the GABAAR α2 subunit. Efficient phasic inhibition requires the accumulation of specific GABAAR subtypes at precise postsynaptic specializations.
To determine the contribution of phosphorylation to this process, we examined whether the GABAAR α2 subunits were phosphorylated endogenously by utilizing a previously characterised pHα2 mouse that allows rapid receptor purification by guest on November 6, 2020 http://www.jbc.org/ Downloaded from using GFP-Trap (22). MS analysis identified novel phosphorylation sites at S359 and S379 within the predicted major intracellular domain of the GABAAR α2 subunit. Notably, these sites flank the collybistin and gephyrin binding site (18). Our studies focused on S359 as its phosphorylation was more reproducibly detected compared to S379. S359 dephosphorylation is PP1/PP2Adependent. To further study α2 phosphorylation we created a phospho-specific antibody against S359. Application of okadaic acid to brain slices from pHα2 mice showed that blocking PP1/PP2A led to a marked increase in endogenous α2 S359 phosphorylation detected using this antibody. Additionally, in neuronal culture, this increase in phosphorylation by OA was completely abolished in the phosphomutant S359A, highlighting the specificity of this newly developed antibody, and confirming S359 as the first phosphorylation site identified in the α2 subunit (13). These experiments also demonstrated that α2 S359 phosphorylation is continuously maintained at low levels under basal conditions by constitutive cycles of phosphorylation and dephosphorylation. Potentially, this mechanism would allow neurons to rapidly control inhibitory synaptic receptor content and fine-tune synaptic responsiveness by dynamically altering the balance between these opposing processes.

S359 phosphorylation is regulated by PKA.
GST-α2 kinase assays demonstrated that S359 is directly phosphorylated by PKA. These results were corroborated in neurons, where PKA activation by forskolin led to increases in phosphorylation which were abrogated by the addition of the PKA specific inhibitor myr-PKI.
Interestingly, PKA and PP1/PP2A have been previously implicated in regulation of GABAARs through phosphorylation of the β3 subunit, which forms receptors with α2 (3,28), to control receptor surface expression (29)(30)(31). However, how multiple phosphorylation sites on various subunits are independently regulated, and how this influences the fate of the assembled receptors, remains to be established. Potentially, proteins that facilitate phosphorylation and dephosphorylation such as A-kinase anchoring protein 79/150 (AKAP70/150), receptor activated C-kinase 1 (RACK1) and phospholipase C-related inactive protein type 1 (PRIP1) may assist in the precise spatio-temporal targeting of kinases and phosphatases to specific receptor subunits (31)(32)(33), although further work will be required to address this possibility directly.
GABAAR α2 S359D phosphomimetic mutant binds less efficiently to gephyrin and collybistin. Since the α2 S359 phosphorylation site was near the gephyrin and collybistin binding site, we next addressed the role of phosphorylation at this site using in vitro binding assays as well as immunoprecipitation. As evidenced by binding assays, both gephyrin and collybistin bound preferentially to GST-α2 WT and the S359A phosphonull mutant compared to the S359D phosphomimetic. Considering that the phosphorylation of this site is kept at low levels, we might expect WT and S359A to behave in a similar fashion. Correspondingly, when pHα2 was immunoprecipitated from neuronal lysates, levels of coimmunoprecipitation of both gephyrin and collybistin were reduced in S359D phosphomimetic mutants compared to WT and S359A phosphonull mutants.
pHα2 WT and S359A phosphonull expressing neurons have a higher density of α2 clusters in dendrites and AIS. Since previous studies have shown that decreased binding to gephyrin and collybistin resulted in a decreased density of clusters of α1-containing GABAARs in dendrites and α2-containing GABAARs in the AIS, respectively (17,20), we decided to ascertain the consequences of phosphorylation on the synaptic accumulation of GABAAR α2 by measuring the size and density of pHα2 clusters in different neuronal domains. Although there was no difference in cluster area, there were significant differences in cluster density. In both the AIS and dendrites, there was a decrease in pHα2 cluster density with S359D phosphomimetic mutant, compared to WT and S359A mutant, which we attribute to the fact that S359D has a decreased capacity to bind both collybistin and gephyrin. In dendrites, the fewer synaptic α2 puncta observed in S359D-expressing neurons is likely attributed to the decrease in gephyrin's ability to mediate the transient stabilisation of these receptors at the synapse, as previously observed in α1 mutants with compromised binding to gephyrin (17). The decrease in α2 clustering at the AIS is likely due to decreased collybistin binding, since the S359D mutant displays a similar loss of AIS clusters to that reported for an α2 mutant with decreased ability to bind collybistin (20). Moreover, gephyrin-α2 subunit associations are weaker in the AIS, indicating that gephyrin has a comparatively minor role in AIS clustering of α2containing GABAARs (34). Indeed, this seems to be the case in our study, as we also noticed a previously reported (27) lower intensity gephyrin staining in the AIS compared to dendrites.
Collybistin immunostaining is technically challenging, but published studies suggest this protein is not exclusively found at synapses or at the AIS, rather it is associated with intracellular membranes throughout the cells (20). Thus, collybistin may not act as a simple scaffold protein, but may regulate GABAAR intracellular trafficking to facilitate their targeting to synapses.

Enrichment of α2 in the AIS is lost in pHα2 S359D phosphomimetic expressing neurons.
Consistent with a previous report (8), we observed an enrichment of α2 clusters at the AIS compared to dendrites. Further, this enrichment is maintained in pHα2 S359A, but lost in S359D expressing neurons, suggesting that α2 S359D are found at similar levels in both dendrites and the AIS and that phosphorylation of this site limits the enrichment of α2 clusters normally observed at the AIS. Thus, this residue appears to reduce the stability of α2-GABAARs at synaptic sites across the neuron, including at the AIS. This novel regulatory mechanism sheds light on the ways by which neurons control inhibitory neurotransmission.
GABAergic neurotransmission must be tightly and precisely controlled to shape excitatory neurotransmission and generate normal network activity patterns. In the absence of such control, pathological hyperexcitability can result (20,(35)(36)(37). In the case of S359, the phosphoregulation of this site would allow neurons to rapidly alter the affinity of α2containing GABAARs for their stabilisation at synapses; thus, neurons could rapidly respond to local changes in their own excitability and the excitability of the network around them, finetuning the density of α2-GABAAR synaptic clusters and adjusting the inhibitory drive according to their present circumstances. This novel mechanism for rapid response becomes even more integral as it pertains to inhibition at the AIS. The AIS is the site of action potential generation (38), and inhibitory neurotransmission in this area-almost exclusively mediated by synaptic α2-GABAARs-has a tremendous effect on neuronal excitability (8,39,40). Furthermore, the AIS of forebrain pyramidal neurons, such as the ones studied here, are solely innervated by inhibitory chandelier cell interneurons (41). Each chandelier cell is thought to innervate the AIS of hundreds of pyramidal cells, allowing for the coordination of activity across large networks and giving rise to oscillatory activity (42)(43)(44). Further, we have previously shown that disturbing the stability of α2-GABAARs at axo-axonic synapses has large consequences on inhibitory neurotransmission and network excitability (20,21).
In summary, we have identified that phosphorylation of S359 within the α2 subunit acts to reduce the binding to the inhibitory scaffold molecules collybistin and gephyrin, resulting in reduced GABAAR accumulation at synapses. This novel regulatory mechanism may allow neurons to fine tune the activity of a subset of synapses enriched in α2 subunit-containing GABAARs.

Experimental Procedures
Animals-All N numbers refer to the number of animals or the number of individual cell culture preparations. Animal protocols were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Animals (Scientific Procedures) Act 1986, and approved by both the Institutional Animal Care and Use Committee of Tufts University, USA and the Animal Services Unit at Bristol University, UK. Myc-pHluorin GABAAR α2 knock-in (pHα2) mice have previously been described in (22).
Hippocampal slice preparation-Transverse slices were prepared from 8-week old male WT and pHα2 mice. Brains were quickly removed from isoflurane anesthetised mice and put in icecold sucrose-based cutting solution containing (mM): 87 NaCl, 3 KCl, 7 MgCl2⋅6H2O, 1.25 NaH2PO4, 0.5 CaCl2, 25 NaHCO3, 50 sucrose and 25 glucose. 400 μm slices were cut with a vibratome (VT1000S, Leica Microsystems) and were allowed to recover for 1h in warm (35˚C), oxygenated (95% O2/5% CO2) ACSF solution containing (mM): 126 NaCl, 2 MgCl2, 2 CaCl2, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 1.5 pyruvate, 1 L-glutamine and 10 glucose. Mass spectrometry analysis-pHα2 was immunoprecipitated from 8-10 week old mice with GFP-Trap as previously described (22). Trypsin and chymotrypsin digestion, liquid chromatography-tandem mass spectrometry (LC-S/MS) and data analysis was performed by Taplin Mass Spectrometry Facility (Harvard) using peaklist generating software ReAdW.exe (ver. 4.3.1), search engine SEQUEST (ver. 28, rev 13). In all cases, enzymes were not specified for the search. Peptides were filter based on SEQUEST scores (XCorr and ∆Cn) and mass accuracy. Peptides were required to be tryptic, although there were no filters specified for chymotrypsin. Fixed modifications considered were 57.0215 Da on cysteine. Variable modifications considered were methionine oxidation (15.9949 Da) and phosphorylation on serine, threonine and tyrosine (73.9663 Da). The mass tolerance was set to 2 Da for precursor ions and 1 Da for fragment ions. The threshold for accepting individual spectra were that the precursor ppm values were less than 5 ppm from expected, along with manual interpretation. Peptides produced by trypsin would be required to be tryptic peptides. The minimum XCorr values for peptides charged +1, +2, +3 and +4 were >1.5, >1.5 >1.5 and >2 respectively. Each phosphopeptide was analysed using the Ascore algorithm to determine the confidence of a given phosphorylation site (23). Sites were confidently assigned when Ascore values were above 13 (p<0.05).

Antibodies for Western blots-The phospho-α2
Glutathione-S-transferase (GST) fusion protein production and pull-down assay-GST fusion proteins were produced as previously described (22). 19 DIV rat cortical neuronal lysates were incubated with GST fused to the large intracellular loop of GABAAR α2 as performed previously (22). GST pull-down assays for gephyrin utilised 500 μg neuronal lysate per condition and 2% input was loaded, for collybistin 750 μg neuronal lysate was utilized per condition and 4% input was loaded.
Kinase assay and phosphopeptide mapping-GST, GST-α2 WT and GST-α2 S359A immobilised on glutathione sepharose beads (GE Healthcare) were phosphorylated with 10 μCi 32 Pγ-ATP (BLU502A, Perkin Elmer) and 15 μg brain lysate (as a source of kinase) for 7 min at 30˚C. Protein samples were separated by SDS-PAGE and visualised by autoradiography. Bands were analysed with ImageJ (NIH). Phosphopeptide mapping was performed as described in (46). Briefly, 32 P-labelled bands from kinase assays were excised, dried and digested with trypsin (0.3 mg/ml in 50 mM NH4HCO3). Samples were spotted on a cellulose thin layer chormoatography (TLC) plate and subjected to two-dimensional phosphopeptide mapping, first by electrophoresis and then by ascending chromatography. Dried TLC plates were exposed to X-ray film. To detect phosphorylation by PKA, GST-α2 WT and GST-α2 S359A immobilised on glutathione sepharose beads were incubated (30°C, 30 min) with 0.75 μg PKA catalytic subunit (NEB) and NEBuffer for protein kinases supplemented with 200 μM ATP (NEB). Beads were washed four times (2500 × ", 2 min, 4°C) in lysis buffer (below) with 0.1% SDS.
Western blot and analysis-Proteins separated by SDS-PAGE (8% or 10% gel) were transferred to PVDF membranes and blocked with 6% milk in PBST for 1h. Membranes were incubated in primary antibody (1h, 5% milk) followed by 3 x 5 min washes, before incubating with HRPconjugated secondary antibodies for 1h. After a further 5 x 5 min washes, blots were developed on film (CL-Xposure, Thermo Scientific). For membranes incubated with phosphorylated and non-phosphorylated peptides, 2 μg of peptide were added to primary antibodies. Collybistin exists in multiple isoforms (25,26), and appears as a doublet in immunblots. For the purposes of analysis, both bands were taken into account. Films were analyzed using ImageJ. Multiple comparisons were performed with one-way ANOVA followed by Tukey's or Sidak's post hoc test. Post hoc p values are stated in the text, one-way ANOVA F and p values can be found in Supplementary Fig. S3.

Neuronal transfection-Hippocampal neurons were
Lipofectamine 2000 (Invitrogen) transfected at 15 DIV. Neuronal coverslips were rinsed in plain Neurobasal media and transferred to fresh dishes with 1 ml pre-warmed Neurobasal media and placed in the incubator. Two tubes containing 200 μl of Neurobasal media were prepared. In tube 1, 1 μg DNA and in tube 2, 1.5 μl Lipofectamine were added. Both tubes were vortexed and incubated at room temperature (RT) for 5 min, after which tubes were mixed together and incubated at RT for a further 20 min to allow complex formation. The transfection mix was added to cells and placed in the incubator for 45 min, after which they were rinsed with Neurobasal media. Coverslips were replaced into their original dishes and left in the incubator until required.
Image acquisition and analysis-Images were captured on a confocal laser scanning microscope (SP5-AOBS, Leica Microsystems) attached to an inverted epifluorescence microscope (DMI 6000, Lecia Microsystems) with a 63×, NA 1.4, oil immersion objective (Plan Apochromate BL, Leica Microsystems). High resolution images (2048 × 2048, mean 2) were taken as z-series of an average of ~7-8 z-stacks, taken at 0.5 μm intervals. Acquisition parameters were kept the same for all scans within each experiment. Images were collecting using the SP5 system's acquisition software and analysis performed using ImageJ software (47,48). Image stacks were projected using a maximum projection algorithm. For each neuron, one 20 μm section of axon initial segment and one or two 20 μm sections of secondary dendrites were analysed. Clusters of pHα2 on the AIS were measured if they were localised with the ankyrin G marker. Clusters on the dendrite were considered synaptic if colocalised or apposed to gephyrin puncta. To examine the GABAAR α2 puncta, clusters were manually outlined and their area measured by ImageJ. Cluster density was defined as the number of puncta per μm of AIS or dendrite. Multiple comparisons were performed with Kruskal-Wallis or one-way ANOVA followed by Sidak's post hoc test. Post hoc p values are stated in the text, Kruskall-Wallis and one-way ANOVA H, F and p values can be found in Supplementary Figure S3.
Lentiviral production-Lentivirus was produced in HEK293T cells maintained in DMEM (D5796, Sigma) supplemented with 10% fetal bovine serum (F7524, Sigma) (45). Briefly, cells were washed with plain DMEM, polyethylenimine (PEI) transfected with 10 μg lentiviral transfer plasmid pXLG3 WPRE px encoding the expression or knockdown of the protein of interest, 7.5 μg packaging plasmid pΔ8.9, and 2.5 μg envelope plasmid pMD2.G. After 4h, transfection media was replaced with complete HEK media or neuronal feeding media. Viral particles were harvested after ~48 h, spun at 3000 × " for 10 min, dispensed into single use aliquots and stored at -80°C until use.

Data Availability Statement
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository (49) with the dataset identifier <PXD019597>.    were excised, trypsinized, blotted on TLC plates and subjected to electrophoresis followed by ascending chromatography. Coloured circles represent the two major tryptic peptides found in GST-α2 WT. The red circle highlights a tryptic peptide in GST-α2 WT that is missing in S359A. The blue circle highlights a tryptic peptide in GST-α2 WT that is still present in S359A.