GABA_A receptors (GABA_ARs) are ligand-gated hetero-pentameric anion channels assembled from various combinations of 19 subunits: α (1-6), β (1-3), γ (1-3), δ, 𝜖, 𝜃, π, and ρ (1-3), although most GABA_ARs in the brain consist of two α subunits, two β subunits, and one γ or δ subunit (Macdonald and Olsen, 1994; Chang et al., 1996; Sieghart and Sperk, 2002; Mody and Pearce, 2004; Olsen and Sieghart, 2008). These receptors mediate the majority of phasic and tonic inhibition in the adult brain and are critical in the regulation of neuronal excitability and neural network function. In developing neurons, GABA_AR activation can also provide membrane depolarization and increase neuronal activity, resulting from a relatively positive Cl^- reversal potential due to high expression of Na⁺/K⁺/Cl^- co-transporter 1 (NKCC1) and low expression of K⁺/Cl^- co-transporter 2 (KCC2) in immature neurons (Ben-Ari et al., 1997; Owens and Kriegstein, 2002; Ben-Ari et al., 2007).

The role of GABA_AR-mediated signaling in neuronal development and function has been extensively studied (Ben-Ari et al., 1997; Owens and Kriegstein, 2002; Ben-Ari et al., 2007). Early pharmacological experiments have demonstrated that GABA_AR activity can modulate neurogenesis, neuronal migration and differentiation, and synapse development in the immature brain (Owens and Kriegstein, 2002; Ben-Ari et al., 2007). However, pharmacological approaches do not separate the cell-autonomous function of GABA_ARs from indirect neuronal network effects associated with global blockade of the receptors, and also do not address the structural role of GABA_ARs in the regulation of synapse development. Depolarizing GABA_AR activity has also been shown to functionally interact with NMDA receptors (NMDARs) to facilitate NMDAR activation and regulate synapse development in immature neurons (Owens and Kriegstein, 2002; Ben-Ari et al., 2007). In addition, by manipulating NKCC1 or KCC2 expression, the role of depolarizing GABA_AR activity in synapse, and spine development has been inferred (Chudotvorova et al., 2005; Akerman and Cline, 2006; Liu et al., 2006; Wang and Kriegstein, 2008), although recent studies have shown that the regulation of synapse development by KCC2 does not require its ion transport function (Li et al., 2007; Fiumelli et al., 2013). Furthermore, synaptic GABA_ARs in knockdown or conventional knockout (KO) mice of GABA_AR subunits are reduced or lost, leading to the impairment of GABAergic synapse formation and maturation (Fritschy et al., 2006, 2012; Patrizi et al., 2008; Frola et al., 2013), which provides genetic evidence for the role of GABA_AR subunits in synapse development. However, in these genetic models, neurons may adapt to the global absence of the GABA_AR subunits and to altered neural network activities throughout their development. Thus, the cell-autonomous role of GABA_ARs in the regulation of synapse development remains largely unclear.

Here we have utilized the CRISPR-Cas9 approach to perform single-cell genetic deletion (Incontro et al., 2014) of all functional GABA_ARs and examine excitatory and inhibitory synapse development. This is achieved by targeting the β1, β2, and β3 subunits of the GABA_ARs in individual hippocampal neurons. These β subunits are required for the receptor assembly and agonist binding (Connolly et al., 1996; Tretter et al., 1997; Baumann et al., 2001; Olsen and Sieghart, 2008; Nguyen and Nicoll, 2018). We found that in neurons lacking GABA_ARs, GABAergic synapses are strongly impaired without an accompanying change of excitatory transmission and the spine density. Furthermore, combined genetic deletion of GABA_ARs and ionotropic glutamate receptors (iGluRs), including both AMPARs and NMDARs, reveals that signaling mediated by these receptors is critical for inhibitory, but not excitatory, synapse development.

We first examined the expression and subcellular distribution of GABA_AR β subunits (GABRBs) in mouse primary hippocampal neuron cultures. In 18 days in vitro (DIV) neurons in culture, the immunolabeling of GABA_AR β subunits, including β1, β2 and β3, demonstrated that all three β subunits were expressed in hippocampal neurons with substantial synaptic localization at both dendritic (Supplementary Figures S1A,B) and somatic (Supplementary Figures S1C,D) regions, consistent with a recent electron microscopy study in rat hippocampi (Kerti-Szigeti and Nusser, 2016). Our data also indicate that among the three β subunits, β2 exhibits a higher level of synaptic distribution (Supplementary Figure S1).

Expression and synaptic localization of β1-3 subunits in hippocampal neurons indicate that functional KO of GABA_ARs requires genetic deletion of all three β subunits. To this end, we have employed the CRISPR-Cas9 technology to develop single-guide RNAs (gRNAs) to target gene loci of three GABRBs, respectively, that encode GABA_AR β subunits in mouse genome. Among several gRNA candidates for each β subunits, positive gRNAs that also expressed GFP effectively reduced the expression of co-transfected Myc-β1, 2 or 3 in HEK293T cells (Supplementary Figure S2). We thus generated a construct (hereafter β1-3 gRNAs) containing the three positive gRNAs for β1, β2, and β3, respectively (Figure 1A). To test the KO effect of β1-3 gRNAs in neurons, we transfected hippocampal neuronal cultures at DIV2-3 and performed immunocytochemical analysis at DIV13-14. We found that compared with empty gRNA vector (hereafter GFP)-transfected neurons, the immunolabeling of β1, β2, or β3 in β1-3 gRNAs expressing neurons was strongly diminished (Figure 1B–D). Furthermore, electrophysiological recordings demonstrated that miniature inhibitory postsynaptic currents (mIPSCs) was essentially lost in neurons expressing β1-3 gRNAs (Figure 1F), in agreement with a recent report (Nguyen and Nicoll, 2018). In contrast, no change of mIPSCs was observed in neurons expressing the empty gRNA vector (Figure 1E). Furthermore, we treated the cultured neurons with 15 mM K⁺ to enhance synaptic activity and measured GABAergic transmission. We found that although 15 mM K⁺ significantly increased the mIPSC frequency in control neurons, it didn’t change GABAergic transmission in neurons expressing β1-3 gRNAs (Figure 1G). Indeed, mIPSCs were barely detectable in these neurons either before or after 15 mM K⁺ treatment (Figure 1G). In addition, tonic inhibition generated by persistent activation of extrasynaptic GABA_ARs by ambient GABA has been observed in many types of neurons and has a profound effect on neuronal excitability (Farrant and Nusser, 2005; Glykys and Mody, 2007; Belelli et al., 2009). To examine whether extrasynaptic GABA_AR-mediated tonic currents were diminished in neurons expressing β1-3 gRNAs, we measured tonic currents by adding the GABA_AR antagonist, bicuculline, in the perfusion solution. We found that compared with control neurons whereby GABA_AR-mediated tonic currents were readily detected by bicuculline, little tonic currents were observed in neurons expressing β1-3 gRNAs (Figure 1H).

We further performed rescue experiments to characterize the role of individual β subunits in the regulation of GABAergic transmission. To this end, we developed gRNA-resistant β1, β2 or β3 in HEK cells through immunocytochemical experiments (β1^∗, β2^∗, or β3^∗ in Figure 2A,C,E, respectively). We then co-transfected β1^∗ (Figure 2B), β2^∗ (Figure 2D), or β3^∗ (Figure 2F) with β1-3 gRNAs in hippocampal neuronal cultures and measured mIPSCs. We found that the loss of GABAergic transmission in neurons expressing β1-3 gRNAs could be rescued by co-expressing β1^∗, β2^∗, or β3^∗ (Figure 2B,D,F, respectively), indicating that individual β subunits can support basal inhibitory transmission and can substitute for each other for GABAergic transmission in hippocampal neurons. Together, these data demonstrate that we have successfully eliminated GABAergic transmission in single hippocampal neurons and individual β subunits can rescue GABAergic transmission in neurons lacking all endogenous β subunits.

The loss of functional GABA_ARs in individual hippocampal neurons allowed us to examine the cell-autonomous role of these receptors in the regulation of GABAergic synapse development. We found that in neurons expressing β1-3 gRNAs, the immunolabeling of vGAT and gephyrin, the pre- and post-synaptic markers of GABAergic synapses, respectively, at both somatic and dendritic regions, was significantly decreased by ∼50% (Figure 3A–C), indicating a reduction of GABAergic synapse density. Interestingly, there was no change of the puncta density of Neuroligin2 (NL2), a key synaptogenic cell adhesion molecule for GABAergic synapses (Graf et al., 2004; Varoqueaux et al., 2004; Chih et al., 2005; Poulopoulos et al., 2009; Li et al., 2017), in neurons lacking GABA_ARs (Figure 3D), suggesting that NL2 clustering is independent of GABA_ARs and may be an upstream event of GABA_AR-mediated signaling for inhibitory synapse development. Taken together, these data demonstrate that genetic deletion of GABA_ARs at the single-cell level leads to a substantial reduction of GABAergic synapses.

Pharmacological blockade of GABA_ARs can induce homeostatic adaptation of excitatory synaptic transmission in neurons (Turrigiano et al., 1998; Shepherd et al., 2006; Anggono et al., 2011; Diering et al., 2014). Specifically, chronic inhibition of GABA_ARs reduces AMPAR-mediated excitatory transmission. We thus examined how excitatory synapses adapted to the single-cell silencing of GABAergic inhibitory transmission. Surprisingly, recording of miniature excitatory postsynaptic currents (mEPSCs) in the presence of TTX in hippocampal neurons expressing β1-3 gRNAs showed that there was no change of frequency and amplitude of mEPSCs (Figure 4A). In addition, immunocytochemical analysis demonstrated that the surface expression of GluA1, a key AMPAR subunit in hippocampal neurons (Lu et al., 2009), and the puncta of vesicular glutamate transporter 1 (vGluT1) did not change in β1-3 gRNA-expressing neurons (Figure 4B,C), suggesting that AMPAR trafficking to the neuronal surface and excitatory synapse density were not altered in these neurons lacking functional GABA_ARs. Furthermore, we measured the spine density by immunolabeling GFP and found that the spine density or type was not significantly changed in neurons expressing β1-3 gRNAs (Figure 4D,E). Collectively, these data show that single-cell elimination of GABAergic transmission does not induce a homeostatic reduction of excitatory synaptic transmission and also leads to little change of cell biological properties of excitatory synapses, including surface AMPAR expressing levels, and the density of excitatory synapses.

We previously employed a Cre-LoxP system to genetically delete both AMPARs and NMDARs in single hippocampal neurons and found that these iGluRs were dispensable for spinogenesis (Lu et al., 2013). One possibility was that in neurons lacking both AMPARs and NMDARs, remaining GABA_ARs could generate depolarizing drive in developing neurons and thus might provide activity necessary for spine development. We thus combined β1-3 gRNAs with the conditional KO of both AMPARs and NMDARs to genetically remove functional GABA_ARs, AMPARs and NMDARs in individual neurons and examined excitatory and inhibitory synapses. In hippocampal neuronal cultures prepared from GRIA1-3^fl/flGRIN1^fl/fl in which three genes encoding AMPAR subunits (GluA1, GluA2 and GluA3) and the gene encoding the NMDAR obligatory subunit, GluN1, are all conditional alleles (Lu et al., 2013), we co-expressed Cre-mCherry and β1-3 gRNAs through transfection. About 2 weeks after transfection, we performed mIPSCs and mEPSCs recordings in transfected neurons and found the loss of inhibitory synaptic transmission and both AMPAR- and NMDAR-mediated excitatory synaptic transmission in these neurons (Figure 5A,B).

We then measured the spine density in neuronal dendrites and observed no change of spine density or type in Cre-positive, β1-3 gRNA-expressing neurons (Figure 5C,D), as compared to control neurons expressing the empty gRNA vector. We also examined inhibitory and excitatory synapses by the measurement of gephyrin and PSD-95 puncta. We found that compared to neurons expressing the empty gRNA vector, co-expression of both Cre and β1-3 gRNAs in GRIA1-3^fl/flGRIN1^fl/fl neurons led to a large reduction of the density of gephyrin puncta (∼90%) (Figure 5E). However, there was no difference of the density of PSD-95 puncta between control neurons and neurons lacking GABA_ARs, AMPARs and NMDARs (Figure 5F). Therefore, genetic deletion of both GABA_ARs and iGluRs impairs the development of GABAergic synapses but does not change the density of spines or glutamatergic synapses.

To study the role of ionotropic GABA_ARs in the regulation of GABAergic synapse development, we have utilized the CRISPR-Cas9 technology to genetically delete all three β subunits of GABA_ARs in hippocampal neurons. GABA_AR β subunits are required for the receptor assembly and GABA binding (Connolly et al., 1996; Tretter et al., 1997; Baumann et al., 2001; Olsen and Sieghart, 2008; Nguyen and Nicoll, 2018), and, consistently, we found that single-cell genetic deletion of three β subunits leads to a loss of GABAergic transmission, in agreement with a recent report (Nguyen and Nicoll, 2018). The lack of inhibitory synaptic transmission at the level of individual neurons allowed us to investigate the cell-autonomous role of GABA_ARs in the regulation of GABAergic synapse development.

Our data demonstrate that GABA_ARs are critical for GABAergic synapse development at the level of single neurons. Indeed, in hippocampal neurons lacking functional GABA_ARs in a mosaic fashion, gephyrin, and vGAT puncta at both somatic and dendritic areas are significantly reduced. Our data are consistent with previous reports in which germline KO of α1 subunit of GABA_ARs abolished GABAergic transmission in Purkinje cells, and consequently impaired GABAergic synapse formation (Fritschy et al., 2006; Patrizi et al., 2008). Similarly, knockdown or KO of the GABA_AR subunit, γ2, which is important for synaptic clustering of GABA_ARs (Essrich et al., 1998), impaired GABAergic innervation and reduced GABAergic synapse density (Schweizer et al., 2003; Li et al., 2005; Frola et al., 2013). In addition, GABA_AR activity has been shown to be important for GABAergic synapse formation (Chattopadhyaya et al., 2007; Arama et al., 2015; Oh et al., 2016; Lin et al., 2018). It is worth noting that broad genetic deletion or widespread pharmacological inhibition of GABA_AR subunits alter neural network activity, and thus does not separate the cell-autonomous function of GABA_ARs from the indirect effects on network activity in the regulation of synapse development. Our data thus provide the genetic evidence of a critical cell-autonomous role of GABA_ARs for GABAergic synapse development. Currently, the molecular mechanisms underlying the regulation of GABAergic synapse development by GABA_ARs remain largely unclear. It has been reported that the GABA_ARs interact with the synaptic adhesion molecule, neurexins (Zhang et al., 2010). In addition, GABA_ARs may induce Ca²⁺ influx through NMDARs (Owens and Kriegstein, 2002; Ben-Ari et al., 2007) or voltage-gated calcium channels (Oh et al., 2016) to regulate GABAergic synaptogenesis, and may also play a synaptogenic role in GABAergic synapse development (Fuchs et al., 2013; Brown et al., 2016). Interestingly, the puncta density of NL2, a key synaptogenic cell adhesion molecule for GABAergic synapses (Lu et al., 2016), is not altered in neurons lacking GABA_ARs, similar to a previous report (Patrizi et al., 2008), suggesting that NL2 may act upstream of GABA_AR signaling for GABAergic synaptogenesis. Recent studies have identified that NL2 is crucial for synaptic anchorage of GABA_ARs through binding to the GABA_AR-interacting protein, GARLH/LHFPL4 (Davenport et al., 2017; Yamasaki et al., 2017; Wu et al., 2018) and is critical for GABAergic synapse development (Poulopoulos et al., 2009; Li et al., 2017; Panzanelli et al., 2017). Thus, it is plausible that during development NL2 may regulate GABA_AR clustering, which in turn modulates GABAergic synapse formation and maturation. Interestingly, in GARLH/LHFPL4 KO neurons, both NL2 and GABA_AR clustering are impaired (Davenport et al., 2017; Yamasaki et al., 2017; Wu et al., 2018), indicating that GARLH/LHFPL4 may function upstream of both NL2 and GABA_ARs in the regulation of GABAergic synapse development.

A recent elegant study has also employed the CRISPR-Cas9 technique to eliminate GABAergic transmission in hippocampal neurons, although this work did not examine the role of GABA_ARs in inhibitory synapse development (Nguyen and Nicoll, 2018). In this work, both β2 and β3, but not β1, subunits could rescue the loss of GABAergic transmission (Nguyen and Nicoll, 2018). Specifically, β1 subunit rescued ∼45% GABAergic transmission (Nguyen and Nicoll, 2018). In contrast, in our study all three β subunits rescued inhibitory synaptic currents in neurons expressing β1-3 gRNAs with β1 restoring ∼70% mIPSC frequency. Possibilities to explain this discrepancy include the expression levels of recombinant β subunits through different expression techniques (gene-gun mediated transfection vs. calcium phosphate transfection in our study) and different experimental preparations (hippocampal organotypic cultures vs. dissociated neuronal cultures in our study).

One surprising observation from our study is that excitatory synaptic transmission is normal in hippocampal neurons lacking GABA_ARs. It has been well-established that chronic pharmacological inhibition of GABA_ARs induces homeostatic adaptation of excitatory synapses and reduces AMPAR-mediated synaptic transmission (Turrigiano et al., 1998; Shepherd et al., 2006; Anggono et al., 2011; Diering et al., 2014). However, the role of GABA_AR-mediated signaling at the single-cell level in the regulation of glutamatergic transmission remained unclear. Our data now demonstrate that at the level of individual neurons, the complete loss of functional GABA_ARs does not impair AMPAR-mediated excitatory transmission. Indeed, glutamatergic transmission, the expression levels of surface GluA1, a major AMPAR subunit in hippocampus (Lu et al., 2009), and the number of vGluT1 puncta are not altered in neurons lacking GABA_ARs. In addition, our work reveals that GABA_ARs are not absolutely required for spinogenesis in hippocampal neurons in vitro, as the spine density in neurons lacking GABA_ARs is indistinguishable from that in control neurons (Figure 4D). Our data are consistent with early work in α1 KO mice in which IPSCs are lost in cerebellar Purkinje cells, but EPSCs and excitatory synapses are largely intact (Fritschy et al., 2006; Patrizi et al., 2008). In addition, in GARLH/LHFPL4 KO neurons, GABAergic transmission is severely reduced without an accompanying change of glutamatergic transmission (Davenport et al., 2017; Yamasaki et al., 2017). Similarly, single-cell ablation of gephyrin in hippocampal neurons strongly reduces GABAergic transmission, but does not change glutamatergic transmission (Gross et al., 2016). However, it is important to point out that our data do not exclude the possibility that GABA_AR activation can sufficiently induce spine formation (Oh et al., 2016) and can modulate glutamatergic synapse development (Ben-Ari et al., 2007).

Previous studies have also indicated a role of NKCC1 and KCC2 in the regulation of excitatory synapse and spine development (Akerman and Cline, 2006; Liu et al., 2006; Wang and Kriegstein, 2008). During development, NKCC1 and KCC2 play fundamental roles in determining the neuronal intracellular Cl^- concentration and polarity of GABA_AR action (Owens and Kriegstein, 2002; Payne et al., 2003; Ben-Ari et al., 2007), and regulate GABA_AR subunit expression (Succol et al., 2012). Through manipulation of NKCC1 or KCC2 expression in neurons, the role of the depolarizing GABA in excitatory synapse development has been proposed (Akerman and Cline, 2006; Liu et al., 2006; Wang and Kriegstein, 2008). Interestingly, recent studies have indicated that the regulation of synapse or dendritic development by KCC2 is independent of its ion transport function (Li et al., 2007; Fiumelli et al., 2013), suggesting that the effect of KCC2 on neuronal development may be independent of GABA action. Future work toward a more complete understanding of depolarizing GABA in excitatory synapse development will be important to our understanding of the molecular mechanisms underlying synaptogenesis.

We have previously employed a Cre-LoxP system to genetically remove iGluRs, AMPARs and/or NMDARs, in individual hippocampal neurons, and found that excitatory synaptic input is not necessary for development of neuronal spines (Lu et al., 2009, 2013). One possibility is that in developing neurons lacking iGluRs, remaining GABA_ARs that are activated may generate depolarizing drives and provide activities important for neuronal morphological development. Thus, to further investigate the role of these receptors in excitatory or inhibitory synapse development, we have combined the Cre-LoxP system with the CRISPR-Cas9 approach to genetically target both iGluRs and GABA_ARs. We found that in individual hippocampal neurons lacking both iGluRs and GABA_ARs, there was no significant change of the spine density and PSD-95 immunolabeling. These data corroborate a series of recent reports that neurotransmitter release, and thus the activation of postsynaptic neurotransmitter receptors, are not required for spine and excitatory synapse development (Verhage et al., 2000; Varoqueaux et al., 2002; Lu et al., 2013; Sando et al., 2017; Sigler et al., 2017). In contrast, we found that there was a nearly 90% reduction of gephyrin puncta in these neurons lacking both iGluRs and GABA_ARs. Previously, we have shown that genetic deletion of both AMPARs and NMDARs led to a strong reduction of inhibitory transmission in hippocampal CA1 pyramidal neurons (Lu et al., 2013). Recently we have further shown that the NMDAR, but not the AMPAR, acts as an important molecule for controlling GABAergic synaptogenesis during development (Gu et al., 2016b; Gu and Lu, 2018). Together, these data suggest that NMDARs and GABA_ARs may play a synergistic role in the regulation of GABAergic synapse development. Currently, it remains unclear how NMDARs and GABA_ARs work together to control the development of GABAergic connections. It is conceivable that in developing, immature neurons, GABA_AR activity may facilitate NMDAR activation, inducing Ca²⁺ influx and stimulating signaling pathways important for GABAergic synapse development (Owens and Kriegstein, 2002; Ben-Ari et al., 2007; Gu et al., 2016b). It is also possible that GABA_ARs and NMDARs may activate parallel pathways to regulate the formation of GABAergic connections. In the future, it will be imperative to determine the sequential action and functional interplay of signaling pathways mediated by NMDARs and GABA_ARs in the regulation of GABAergic synaptogenesis. It will also be important to determine how these neurotransmitter receptors functionally interact with cell surface molecules important for GABAergic synaptogenesis to regulate formation of inhibitory connections (Kuzirian and Paradis, 2011; Ko et al., 2015; Lu et al., 2016; Krueger-Burg et al., 2017).

In summary, through a single-cell genetic approach in vitro we have provided new insights into the cell-autonomous role of GABA_ARs in developing neurons and discovered a dichotomy in the regulation of synapse development by this prominent Cl^- channel. While inhibitory synapse development is critically regulated by GABA_ARs, establishment of glutamatergic transmission and excitatory synapses are largely independent of GABA_AR-mediated signaling at the level of individual neurons. Furthermore, we have managed to remove all AMPARs, NMDARs and GABA_ARs in single neurons in culture and demonstrated that iGluR- and GABA_AR-mediated signaling are not essential for spinogenesis. Our data thus suggest that other developmental pathways including neurotropic factor-mediated signaling (Reichardt, 2006; Park and Poo, 2013), guidance cues, and their receptors (Klein, 2009; Shen and Cowan, 2010; Koropouli and Kolodkin, 2014), trans-synaptic cell adhesion interactions (Scheiffele, 2003; Dalva et al., 2007; Missler et al., 2012; de Wit and Ghosh, 2016), and other receptors or channels-mediated signaling (Komatsuzaki et al., 2005; Eroglu et al., 2009; Uemura et al., 2010; Lozada et al., 2012a,b; Kozorovitskiy et al., 2015; Sellers et al., 2015) may play key roles in spinogenesis and excitatory synaptogenesis. It is also worth noting that our data were collected in cultured neurons and there are limitations in using in vitro models to study synapse development. Thus, future experiments in vivo will help further establish the role of GABA_ARs in synaptogenesis. Nevertheless, our data demonstrate a remarkable specificity of these ionotropic receptors in mediating signaling important for synapse development in vitro. Given the prominent roles of malfunctions of these receptors in the pathogenesis of many neurodevelopmental disorders (Cellot and Cherubini, 2014; Yuan et al., 2015), our data also highlight the importance of understanding the molecular mechanisms for the regulation of GABAergic synapse development by these receptors.

All experiments using mice were performed in accordance with animal protocols approved by the Institutional Animal Care and Use Committee at National Institute of Neurological Disorders and Stroke (NINDS), National Institutes of Health (NIH).

JD, SP, and WL designed the experiments. JD, DC, JL, and XG cloned and characterized the gRNA constructs. JD, SP, and TL performed the immunocytochemical experiments. JD performed the electrophysiological assays. QT provided critical technical design and problem solving. JD and WL wrote the manuscript. All authors read and commented on the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.