Activity-dependent development of inhibitory synapses and innervation pattern: role of GABA signalling and beyond


  • This report was presented at The Journal of Physiology Symposium on Mechanisms of neocortical development, which took place on 14 November 2008 at the annual meeting of the Society for Neuroscience in Washington, DC. It was commissioned by the Editorial Board and reflects the views of the authors.

Corresponding author Z. J. Huang: Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA.  Email:


GABA-mediated synaptic inhibition is crucial in neural circuit operations. The development of GABAergic inhibitory synapses and innervation pattern in mammalian neocortex is a prolonged process, extending well into the postnatal period, and is regulated by neural activity and experience. Accumulating evidence supports the hypothesis that GABA signalling acts beyond synaptic transmission and regulates inhibitory synapse development; in other words, similar to glutamate signalling at developing excitatory synapses, GABA may coordinate pre- and post-synaptic maturation at inhibitory synapses. These findings raise numerous questions regarding the underlying mechanisms, including the role of GABA receptors and their link to synaptic adhesion molecules. Since synapse formation is a crucial component of axon growth, GABA signalling may also shape the axon arbor and innervation pattern of inhibitory neurons. A mechanism unique to GABAergic neurons is activity-dependent GABA synthesis, largely mediated through activity-regulated transcription of the rate-limiting enzyme GAD67. Such cell-wide as well as synaptic regulation of GABA signalling may constitute a mechanism by which input levels and patterns onto GABAergic neurons shape their innervation pattern during circuit development.

In many areas of the mammalian brain, such as the neocortex, neural circuits rely on inhibition mediated by γ-aminobutyric acid (GABA) from diverse cell types to control the spatiotemporal patterns of electrical signalling (Markram et al. 2004). The inhibitory output of GABAergic neurons is distributed in the network through their axons and synapses, which constitute elaborate and cell-type-specific innervation patterns (Huang et al. 2007). A prominent feature of GABAergic axon arbors in neocortex is their local exuberance: a single interneuron often produces extensive local arbors that innervate hundreds of neurons in its vicinity and form multiple clustered synapses onto each target neuron (Tamas et al. 1997; Wang et al. 2002). Such an innervation pattern probably contributes to their efficient control over the activity patterns in local cell populations. For example, a single parvalbumin-containing (PV) basket interneuron innervates hundreds of pyramidal neurons at the soma and proximal dendrites, and controls the output and synchrony of pyramidal neurons (Fig. 1; Cobb et al. 1995; Miles et al. 1996; Tamas et al. 1997). Furthermore, PV basket cells form extensive mutual innervation (Tamas et al. 2000) and, together with their unique physiological properties, contribute to the generation of coherent network oscillations that might organize functional neural ensembles (Bartos et al. 2007).

Figure 1.

Perisomatic innervation pattern of the neocortical basket interneurons
A, highly exuberant axonal arborization of a neocortical basket interneuron (blue) and one of its many postsynaptic pyramidal cells (red). Although the basket axon overlaps with a large part of the pyramidal basal dendritic tree, all 15 electron microscopically verified synaptic junctions (bottom panel, right) are clustered around the soma or the most proximal dendrites (bottom panel, left) (adapted from Tamas et al. 1997). B, reconstructions of the two PV basket cells connected by both chemical and electrical synapses (presynaptic cell: soma and dendrites, red; axon, green; postsynaptic cell: soma, dendrites, black; axon blue). Cortical layers (I−V) are indicated on the left. The electron microscopically identified synaptic junctions (1 − 4) and gap junctions (5, 6) mediating the interaction between the coupled cells were found nearby on the soma and a proximal dendrite (inset) (adapted from Tamas et al. 2000). C, a schematic showing prominent features of the innervation pattern of cortical basket interneurons. A single basket cell axon (green) innervating the many pyramidal neurons (pink) in its vicinity with clusters of perisomatic synapses (green dots). Basket cells also innervate other basket cells via chemical and electrical (zigzaged lines) synapses. Grey triangles represent pyramidal neurons that are not innervated by these basket cells.

The development of a mature GABAergic innervation pattern is often a prolonged process, extending well into the postnatal period. In the dentate gyrus of hippocampus, basket cell axon arbors undergo marked maturation between the first and fourth week, and increased connectivity among basket cells contributes to the enhanced coherence of gamma oscillation in local networks (Doischer et al. 2008). In primary visual cortex, the maturation of perisomatic inhibition by basket interneurons proceeds into the fifth postnatal week and may contribute to the regulation of the critical period of plasticity (Huang et al. 1999; Morales et al. 2002). Importantly, the maturation of inhibitory innervation in visual and somatosensory cortex is regulated by sensory experience (Morales et al. 2002; Chattopadhyaya et al. 2004; Jiao et al. 2006). Such activity-dependent development of inhibitory synapses and innervation pattern is a major component of neural circuit assembly, yet the underlying cellular and molecular mechanisms are poorly understood.

GABA signalling regulates inhibitory synapse development

As key mediators of neural activity, neurotransmitters are particularly well suited to couple synaptic signalling with synaptic wiring (Zhang & Poo, 2001; Hua & Smith, 2004). Glutamate, the major excitatory transmitter in vertebrate brain, has been implicated in regulating many aspects of synapse formation, maturation and plasticity (Zheng et al. 1994; Shi et al. 1999; Carroll et al. 1999; Wong & Wong, 2001; Bonhoeffer & Yuste, 2002; Malinow & Malenka, 2002; Tashiro et al. 2003). In addition, through regulating synaptogenesis, glutamate receptor signalling contributes to activity-dependent development of axonal and dendritic arbors (Ruthazer et al. 2003; Hua & Smith, 2004; Hua et al. 2005; Cline & Haas, 2008).

Initially discovered as an inhibitory transmitter, GABA has since been implicated in multiple processes of neural development, from cell proliferation to circuit formation (Owens & Kriegstein, 2002). The trophic effects of GABA on neuronal migration and neurite growth during the embryonic and perinatal period are largely explained by its depolarizing action in immature neurons, resulting from chloride ion efflux through the GABAA receptor, which triggers calcium influx and signalling (Ben-Ari et al. 1989; Leinekugel et al. 1995). During the postnatal period, the up-regulation of the chloride transporter KCC2 in neurons results in increased extrusion of intracellular chloride (Rivera et al. 1999), and GABA assumes its classic role as an inhibitory transmitter (Ben-Ari et al. 2007).

Recently, several studies converge and suggest that, in addition to mediating synaptic inhibition in more mature circuits, GABA signalling promotes and coordinates pre- and post- synaptic maturation during activity-dependent development of inhibitory synapses and innervation (Fig. 2). A main line of evidence came from studying the effects of altering GABA synthesis on the development of perisomatic synapses from PV basket interneurons in the visual cortex. The maturation of many features of basket cell axon arbors and perisomatic synapses can be recapitulated in cortical organotypic cultures (Di Cristo et al. 2004) and is strongly regulated by neuronal activity (Klostermann & Wahle, 1999; Chattopadhyaya et al. 2004). Genetic knockdown of GABA synthesis implicates GABA signalling itself in the development of perisomatic synapses (Chattopadhyaya et al. 2007). GABA is synthesized by two glutamate decarboxylases, GAD67 and GAD65 (Soghomonian & Martin, 1998). Of these two enzymes, GAD67 is the rate-limiting enzyme and influences cellular GABA contents in a dosage-dependent manner (Asada et al. 1997; Ji et al. 1999). Knockdown of GAD67 in single GABAergic interneurons, which should have minimum impact on circuit activity levels, results in profound cell autonomous deficits in synapse formation, axon branching and innervation field in cortical organotypic cultures; such deficits were partially rescued by blocking GABA re-uptake or enhancing GABAA or GABAB receptor function (Chattopadhyaya et al. 2007). Similar deficits were found in visual cortex of Gad67 germline heterozygotes, which show ∼40% reduction of GABA levels (Chattopadhyaya et al. 2007). Conversely, overexpression of Gad67 in single basket interneurons promotes the maturation of perisomatic synapses (Chattopadhyaya et al. 2007). These results demonstrate that GABA acts beyond inhibitory transmission in juvenile and adolescent brain, and regulates the maturation of inhibitory synapses and innervation patterns (Fig. 2), thus revealing a new facet of GABA function distinct from its early tropic action in neonatal brain.

Figure 2.

GAD67 and GABA act beyond inhibitory transmission and regulate inhibitory synapse development and innervation patterns
A, GABA signalling may regulate the morphogenesis of inhibitory synapses. B, since synapse formation is an integral part of axon growth and branching, activity-dependent GABA signalling may further influence the development of GABAergic axon arbor and innervation pattern. C, a hypothetical model depicting how GABA–GABA receptor signalling and neuroligin–neurexin adhesion may interact and co-operate to regulate the development of inhibitory synapses. Pentameric GABAARs are assembled in the endoplasmic reticulum. Most GABAARs are first delivered to extrasynaptic locations, they then either diffuse to and become trapped at postsynaptic sites or undergo endocytosis. NL2 and synaptic GABAARs stabilize each other, either through intracellular reciprocal interactions aided by scaffolding proteins such as gephyrin or through extracellular cis interaction. In addition, GABA activation of GABAARs might further stabilize synaptic GABAARs through structural changes or signalling mechanisms. Such activity- and GABA-mediated stabilization of GABAAR might further increase the levels of NL2 at cell–cell contacts and, in turn, stabilize presynaptic terminals through trans-synaptic interactions with neurexins.

Structural role of GABAA receptors: coupling transmission to synapse maturation and stability

Another line of evidence supporting a role of GABA on the development of inhibitory synapses came from studying the effects of manipulating GABAA receptor (GABAAR) subunits. GABAARs are heteropentameric chloride channels composed of several classes of subunits (Michels & Moss, 2007). Although over 19 subunits have been identified, giving rise to a large number of possible subunit combinations, the vast majority of GABAARs consist of α, β and γ2 subunits in a 2 : 2 : 1 stoichiometry. In the mature brain, GABAARs are primarily localized at postsynaptic and extrasynaptic membranes where they mediate phasic and tonic inhibition, respectively.

The γ2 subunit is essential for accumulation of cell surface GABAARs at postsynaptic sites (Essrich et al. 1998; Schweizer et al. 2003). Acute suppression of γ2 expression in cultured hippocampal neurons not only disrupts GABAAR clustering but also results in a profound reduction of GABAergic innervation of γ2-deficient neurons (Li et al. 2005; Fang et al. 2006). Moreover, when palmitoylation of the γ2 subunit was suppressed by knockdown of the Asp-His-His-Cys (DHHC) family palmitoyltransferase GODZ, trafficking of GABARs to postsynaptic sites was perturbed and GABAergic innervation was reduced (Fang et al. 2006). As both presynaptic GABA and postsynaptic GABAA receptors influence GABAergic synapse development, a simple hypothesis is that activity-dependent GABA signalling promotes the differentiation of pre- and post-synaptic sites, and coordinates the maturation and stabilization of inhibitory synapses.

Further evidence regarding the role of GABAARs in synapse formation came from studies of Purkinje neurons in the cerebellum. Purkinje cells are themselves GABAergic neurons but also receive two types of GABAergic inputs: the axo-somatic synapses from basket interneurons and the axo-dendritic synapses from stellate interneurons, both with GABAARs containing the α1 subunit. Deletion of the α1 subunit gene results in a complete loss of functional GABAARs in Purkinje cells by postnatal day 18 (Fritschy et al. 2006). In these α1−/− mice, GABAergic terminals from stellate axons are initially formed normally onto the Purkinje dendritic shaft. However, starting from postnatal day 11, synaptogenesis is significantly reduced and perturbed (Fritschy et al. 2006; Patrizi et al. 2008). Instead, stellate cell terminals form aberrant and mismatched contacts with postsynaptic specialization on the spines of Purkinje dendrites. These results suggest that initial steps of GABAergic synapse formation can proceed in the absence of α1, but GABAA receptors appear crucial for activity-dependent regulation of synapse density, possibly through promoting the stabilization of transient axodendritic contact into mature synapses. The mechanism linking GABA signalling to synapse maturation are still unclear. Activation of GABAARs may result in the local release of trophic factors which promote inhibitory synapse maturation, and/or act as protective signals that prevent synapses elimination. The failure to stabilize presynaptic terminals after postsynaptic loss of GABAARs suggests the presence of a retrograde signal that is regulated by synaptic activity or by association with postsynaptic GABAARs. Amongst the molecular mechanisms that may contribute to such an activity-regulated trans-synaptic signal, the neuroligin and neurexin complex represents one of the plausible candidates.

From GABAA receptors to synaptic adhesion and activity-dependent retrograde signalling

Neuroligins and neurexins are heterophilic synaptic adhesion molecules broadly expressed in the central nervous system (Brose, 1999; Sudhof, 2008). Cell biological studies have revealed potent ‘synaptogenic’ or synapse-organizing activities for these proteins (recently reviewed in (Levinson & El-Husseini, 2005; Craig & Kang, 2007). Postsynaptic neuroligins promote assembly of functional presynaptic specializations in axons, while presynaptic neurexins – through interaction with neuroligins – recruit postsynaptic scaffolding proteins and transmitter receptors in dendrites.

While neuroligin–neurexin complexes are common building blocks of glutamatergic and GABAergic synapses, analysis of mutant mice so far support their particularly critical roles in the organization of GABAergic synapses. Triple knockout mice lacking the three alpha-neurexin transcripts, although they die at birth, show a 50% reduction in the density of GABAergic synapses in the brainstem (Missler et al. 2003). In double knockout mice, some of which reach adulthood, GABAergic synapse density is reduced by 30% whereas glutamatergic synapse density is apparently unchanged (Dudanova et al. 2007). As for neuroligins, mice lacking the three major isoforms (NL1, 2 and 3), also perinatal lethal, show only a relatively small (15–20%) reduction in the number of synapses in the brainstem, but a severe loss of GABAARs and the scaffolding protein gephyrin from postsynaptic sites (Varoqueaux et al. 2006). Among the different isoforms, NL2 is exclusively localized to GABAergic synapses. NL2−/− mice display a selective decrease in the number of inhibitory synapses in the postnatal neocortex (Chubykin et al. 2007). In addition, layer 2/3 neurons in acute cortical slices from NL2−/− mice show a selective impairment of GABAergic transmission whereas glutamatergic transmission is normal. Overexpression of NL2 in cultured neurons increases the density GABAergic terminals (Chih et al. 2005) and the amplitude of inhibitory postsynaptic currents (Chubykin et al. 2007). Notably, this overexpression-induced increase in GABAergic transmission is blocked by pharmacologically reducing network activity in the culture. Therefore, neuronal and synaptic activity might either regulate the presynaptic response to NL2 or postsynaptic stabilization induced by NL2.

As both neuroligins and GABAA receptors play important roles in the maturation of postsynaptic specializations and the differentiation and stabilization of presynaptic terminals at inhibitory synapses, an obvious question is: how do GABA/GABAAR-mediated synaptic signalling and neuroligin/neurexin-mediated synaptic adhesion interact and cooperate to regulate activity-dependent development of inhibitory synapse? It is currently unknown at what stage of their biosynthetic pathway GABAARs first interact with NLs, and how such interactions might be regulated. One possibility is that NL2 and synaptic GABAARs would stabilize each other, either through intracellular reciprocal interactions aided by scaffolding proteins such as gephyrin or through extracellular cis interactions (Fig. 2). In addition, GABA activation of GABAARs might further stabilize GABAARs at synapses through as yet unknown structural or signalling mechanisms. Such activity- and GABA-mediated stabilization of GABAARs might further increase the levels of NL2 at postsynaptic sites; this, in turn, would stabilize the presynaptic terminals through trans-synaptic interactions with neurexins. Evidence consistent with this model include: (1) in vitro studies demonstrated a co-aggregation of NL2 and the GABAARα2 subunit in heterologous cells (Dong et al. 2007); (2) the residence time of GABAARs on the plasma membrane and their targeting to synapses is regulated by synaptic activity (Saliba et al. 2007); (3) pharmacological blockade of neuronal activity in cultured neurons diminishes the synaptogenic activity of NL2 (Chubykin et al. 2007); (4) reduced GABA synthesis and release result in a reduction of inhibitory synapses (Chattopadhyaya et al. 2007). Moreover, there is precedent for such mechanisms in activity-dependent recruitment of glutamate receptor and trans-synaptic signalling at glutamatergic synapses. Local spontaneous activity and glutamate release reduce diffusion exchange of GluR1 between synaptic and extrasynaptic domains, resulting in postsynaptic accumulation of GluR1 (Ehlers et al. 2007). In addition, PSD-95 and NL1 retrogradely modulate presynaptic release probability and may coordinate post- and pre-synaptic morphological changes (Ehrlich et al. 2007; Futai et al. 2007). It remains to be seen whether analogous mechanisms for GABA and NL2 signalling exist at inhibitory synapses.

In an alternative model, the expression and/or localization of NL2 might be regulated by GABA signalling, either through regulating NL2 protein levels or NL2-interacting proteins involved in its synaptic localization. It is also possible that GABA binding to GABAARs might modulate their coupling to NL2, thereby increasing the potency and affinity of NL2 towards neurexin in the presynaptic terminals.

Activity-regulated Gad67 transcription as a cell-wide mechanism for modulating GABA signalling and innervation pattern

A mechanism unique to GABAergic neurons is activity-dependent GABA synthesis. Unlike glutamate, which is both the precursor and product of many essential metabolic and signalling processes in the cell, GABA can only be synthesized by two glutamate decarboxylases, and the main function of GABA is intercellular signalling (Soghomonian & Martin, 1998). In most brain regions, GAD67 activity is rate-limiting for GABA synthesis (Asada et al. 1996; Kash et al. 1997). Since GAD67 is produced at a limiting level in the brain (Asada et al. 1997), alterations in GAD67 levels influence cellular and vesicular GABA content (Murphy et al. 1998; Engel et al. 2001). Unlike GAD65, which is relatively stable, GAD67 protein has a rather quick turn-over rate, with a half-life of several hours (Christgau et al. 1991; Pinal & Tobin, 1998). The major step in the physiological regulation of GAD67 activity is Gad1 transcription, which is dynamically regulated during development (Kiser et al. 1998), by neural activity (Patz et al. 2003; Kinney et al. 2006) and experience (Benson et al. 1989; Benevento et al. 1995; Liang et al. 1996; Gierdalski et al. 2001; Kobori & Dash, 2006). Therefore, activity-dependent transcription may result in adjustment of GAD67 levels and the intracellular GABA pool for release. As alterations in GAD67 and GABA levels profoundly influence interneuron axon growth and synapse formation during the development of inhibitory circuits, neuronal activity might shape the pattern of inhibitory synaptic innervation through GAD67-mediated GABA synthesis (Fig. 3). Such activity-dependent and cell-wide regulation of a ‘transmitter resource’ implies a novel logic for the maturation of inhibitory synapses and innervation pattern. This hypothesis needs to be tested by disrupting the activity regulation of GAD67 transcription in GABAergic neurons and examining the impact on inhibitory synapse development.

Figure 3.

A scheme showing that the level and pattern of neuronal activity may regulate inhibitory synaptic morphogenesis and innervation patterns through GAD67-mediated GABA synthesis and signalling.

More questions than answers

The converging findings that GABA and GABA receptor signalling regulate inhibitory synapse development raise numerous questions regarding the underlying mechanisms and their functional implications. The many steps from GABA signalling to receptor trafficking/stability and neuroligin–neurexin function remain to be defined. In addition, it is unknown whether and how postsynaptic activity in pyramidal neuron might influence the action of GABA signalling on inhibitory synapse development. Furthermore, because cortical GABAergic neurons not only innervate pyramidal neurons but also other GABAergic neurons, an obvious question is whether and how GABA signalling might regulate the development of inhibitory synapses onto inhibitory neurons. Addressing such questions will require methods to visualize inhibitory synapses onto inhibitory neurons. Finally, although activity regulation of GAD67 transcription has been well demonstrated in numerous developmental and plasticity paradigms, its impact on GABA signalling and inhibitory synapse development and plasticity remains to be established in vivo. Compared with our understanding of the role of glutamate in excitatory synapse development, we are only beginning to scratch the surface of the role of GABA in the development of inhibitory synapses. Progress in this area will not only enhance our understanding of activity-dependent development of inhibitory synapses, axon arbors and innervation patterns, but also might have implications in the construction of cortical subnetworks, such as reciprocally connected groups of excitatory and inhibitory neurons (Yoshimura & Callaway, 2005).