The emerging role of GABA_B receptors as regulators of network dynamics: fast actions from a ‘slow’ receptor?

Abstract

Convention holds that ionotropic receptors mediate fast neurotransmission and that ‘slow’ G-protein coupled metabotropic receptors have a secondary, modulatory role in the control of neuronal networks. Here, we discuss recent evidence showing that activation of metabotropic GABA_B receptors in cortical layer 1 can powerfully inhibit principal cell activity and that their activation can rapidly halt ongoing network activity. Inputs from both within and outwith the cortex converge upon layer 1 where they target various populations of interneurons, including neurogliaform cells. We argue that neurogliaform cells are the main effector of a powerful inhibitory circuit that, acting through GABA_B receptors, can be differentially recruited by long-range connections to serve in roles as diverse as conscious perception and memory consolidation.

Introduction

In undergraduate neuroscience courses, students are taught that ionotropic receptors mediate fast neurotransmission and that ‘slow’ G-protein coupled metabotropic receptors play more of a supporting, modulatory role in cortical circuits [1]. While ionotropic glutamate (AMPA, NMDA and kainate) and GABA (GABA_A) receptors are the primary effectors of local synaptic transmission, the role of metabotropic receptors in normal network function is often overlooked. Here, we will review recent evidence suggesting that at least one metabotropic receptor, the GABA_B receptor, can exert a fast, powerful inhibitory influence over cortical networks, capable of rapidly silencing ongoing network activity.

GABA_B receptors are members of the G-protein coupled receptor (GPCR) superfamily, a group of receptors that include metabotropic glutamate, opioid and olfactory receptors. All GPCRs share a common structure comprising seven transmembrane domains: ligand-binding domains are located in the extracellular region and they activate second messenger systems via interactions with G-proteins in the cytosolic region. Functional GABA_B receptors are heterodimers of two GPCR proteins, the GABA_B1 and GABA_B2 subunits, the presence of both being essential for membrane trafficking. Several isoforms of the GABA_B1 subunit exist, but the most abundant are GABA_B1a and GABA_B1b, which vary in the extracellular domain and are conserved throughout the vertebrate family [2]. The isoform of GABA_B1 subunit affects the synaptic location of the GABA_B receptor: receptors containing the GABA_B1a subunit are predominantly found on presynaptic terminals of both excitatory and inhibitory synapses whilst those containing the GABA_B1b subunit predominantly exist in postsynaptic locations [3,4], an observation that holds for all excitatory synapses examined throughout the CNS [5].

Inhibition via GABA_B receptors is mediated through the adenylyl cyclase/protein kinase A (PKA) second messenger pathway (via activation of Gα_i/o-type G proteins) and, through liberation of Gβγ subunits, by activation of G-protein coupled inward-rectifying K⁺ (GIRK) channels and inhibition of voltage-gated Ca²⁺ channels (VGCCs) [2]. Presynaptic GABA_B receptors inhibit transmitter release through inhibition of VGCCs and possibly via interactions with vesicular release machinery [5], while postsynaptic GABA_B receptors exert their effects through multiple pathways. Recent studies have shown that postsynaptic GABA_B receptors inhibit dendritic Ca²⁺ signals mainly by inhibiting VGCCs, with little contribution from GIRK channels [6*]. While GABA_B receptors inhibit dendritic NMDA receptor Ca²⁺ signals through the PKA/cAMP pathway [7], inhibition of dendritic Ca²⁺ spikes appears to occur, at least in part, via a direct interaction between the Gβγ-subunit and L-type Ca²⁺ channels [8].

Traditionally, GABA_B receptors are thought to regulate slow changes in neuronal excitability, with their hypofunction being implicated in disorders such as depression, epilepsy and impaired sleep [2]. Despite their wide distribution throughout the hippocampus, it was believed that GABA_B receptors were primarily activated during periods of strong network activation, via GABA spill-over [9]. Although GABA_B receptors have the potential to affect multiple cellular and network processes through a myriad of mechanisms, including cross-talk with other GPCR systems via the PKA/cAMP pathway, most early studies into their function appeared to preclude a dominant role in controlling fast network processes, perhaps only having relevance in pathological states [10].

‘Slow’ inhibition can rapidly terminate persistent activity

Evidence that GABA_B receptors may have a significant role in normal network function comes from the study of the slow oscillation. During sleep, cortical networks participate in the slow oscillation when neurons display periods of synchronised depolarisation and firing (Up states) punctuated by periods of relative hyperpolarisation (Down states). Up states are involved in memory consolidation and Down states may play a role in regulating neuronal homeostasis [for reviews, see e.g. 11,12,13]. A study using an in vitro model of the oscillation in the medial entorhinal cortex (mEC) demonstrated that GABA_A and GABA_B receptors had different roles in mediating Up states [14]. Up states occurred either spontaneously or in response to electrical stimulation in layer 3, with a mean duration of ~3 seconds. Progressively blocking GABA_A receptors through increasing concentrations of antagonist shortened Up states, with complete blockade inducing epileptiform activity [14]. Given that Up states are sustained by a dynamically-regulated balance between inhibitory and excitatory conductances [15], this result was perhaps not too surprising.

Unexpectedly, however, Mann et al. also found that GABA_B receptors could control Up state termination [14]. Blockade of GABA_B receptors prolonged the duration of Up states by around 50% and, interestingly, electrical stimulation in layer 1 applied 500ms after the onset of the Up state, could immediately evoke a Down state, terminating ongoing activity. Notably, this effect was abolished by application of GABA_B receptor antagonists. This important study demonstrated that GABA_B receptors could be the major determinant of Up state duration, with GABA_B receptors both controlling the spontaneous and afferent-evoked termination of Up states. Given the lack of subtype-specific GABA_B receptor antagonists, pharmacology alone cannot determine whether these two phenomena are controlled by presynaptic or postsynaptic GABA_B receptors, or a combination of both.

Recently, we further studied the role of GABA_B receptors in termination of Up states using mice in which either GABA_B1a or GABA_B1b subunits had been genetically ablated [16**]. We found that, in mice lacking GABA_B1a-containing receptors, electrical stimulation in layer 1 could still terminate Up states, but that blockade of the remaining GABA_B1b-containing receptors did not prolong Up states (Fig. 1A). Conversely, we found that in mice lacking GABA_B1b-containing receptors, electrical stimulation in layer 1 could not actively terminate Up states, but that they were prolonged by blockade of the remaining GABA_B1a-containing receptors (Fig. 1B). These experiments demonstrate that presynaptic GABA_B1a-containing receptors alone are important in controlling the duration of spontaneous Up states, perhaps by regulating neurotransmitter release. They also show that postsynaptic GABA_B1b-containing receptors are essential for afferent-evoked termination of persistent activity, with electrical stimulation in layer 1 activating a powerful mechanism by which interneurons in layer 1, targeting GABA_B receptors, can halt ongoing network activity (see Fig. 1C and D). In addition to the mEC [14,16], prolongation of Up states via GABA_B receptor blockade has been demonstrated in visual [17], prefrontal [18] and somatosensory cortices [Craig et al., unpublished], suggesting that this mechanism is conserved across cortical regions.

Figure 1. Presynaptic GABA_B receptors contribute to spontaneous termination of the Up state whilst postsynaptic GABA_B receptors are necessary for layer 1 stimulus-evoked termination of the Up state.

(A), blockade of GABA_B receptors with the selective antagonist CGP55845 prolongs the Up state (evoked in response to layer 3 stimulation in medial entorhinal cortex) in wildtype mice and those lacking the GABA_B1b subunit, but not those lacking the GABA_B1a subunit, implying that presynaptic GABA_B receptors are involved in the spontaneous termination of the UP state. (B)_, electrical stimulation in layer 1 can terminate an ongoing Up state in wildtype mice and those lacking the GABA_B1a subunit, but not in those lacking the GABA_B1b subunit, demonstrating that afferent-evoked termination of the UP state is mediated by postsynaptic GABA_B receptors. (C) and (D), schematic representation of proposed model: recurrent excitatory and inhibitory (GABA_A receptor-mediated) connections sustain the Up state, with presynaptic GABA_B receptors regulating the duration of the Up state by inhibiting transmitter release. Inputs to layer 1 (perhaps from the thalamus or other cortical regions) activate neurogliaform (NGF) cells, which target dendritic GABA_B receptors to inhibit principal cell firing through reduced calcium entry, causing the network to enter a Down state. Neurogliaform cells may also regulate Up state duration by targeting GABA_B receptors on the presynaptic glutamatergic terminals of the inputs arriving in layer 1. (A) and (B) adapted from [16], with permission.

Layer 1: important regulator of network activity?

Electrical stimulation in layer 1 can halt ongoing network activity, presumably by activating interneurons that target postsynaptic GABA_B1b-containing receptors. While it is remarkable that a ‘slow’ metabotropic receptor can have a profound effect on the state of a network, one has to ask whether this effect is specific to the slow oscillation or whether this experimental manipulation is tapping into a system capable of inhibiting any activity within the cortical circuit?

Early hints of a GABA_B receptor-mediated inhibitory circuit in layer 1 came from a study of GABA_B receptor location and synaptic inhibition. The study found that layer 1 stimulation blocked dendritic Ca²⁺ spike generation in layer 5 pyramidal cells, acting via postsynaptic GABA_B1b-containing receptors, whilst presynaptic GABA release from GABAergic terminals, detected by GABA_A receptor-mediated IPSPs on principal cells, was inhibited by activation of GABA_B1a-containing receptors [3]. This study suggested the presence of a complex circuit in layer 1 that could control the excitability of principal cells, and provides a mechanism by which presynaptic and postsynaptic GABA_B receptors have different roles in terminating persistent activity [16], with presynaptic receptors regulating transmitter release and postsynaptic receptors actively halting Up states by inhibiting the firing of principal cells. This begs the question of whether such an inhibitory circuit might regulate the operation of other processes?

Evidence that the layer 1 inhibitory circuit is also active during sensory processing was provided by an elegant in vivo study demonstrating that interhemispheric inhibition is mediated by postsynaptic GABA_B receptors [19**]. Using a combination of calcium imaging, optogenetics and transgenic mice, it was shown that, in response to hindpaw stimulation, the firing of layer ipsilateral 5 pyramidal cells was disynaptically inhibited by excitatory afferents arising from the contralateral hemisphere that specifically targeted interneurons in layer 1; this effect was mediated by GABA_B1b-containing receptors located on the apical dendrites of the pyramidal cells. Importantly, the GABA_B-receptor mediated inhibition resulted in only a small hyperpolarisation at the soma: due to the contribution of dendritic Ca²⁺ channels to overall pyramidal cell excitability, their blockade provided a form of shunting inhibition apparent only when the cell was spiking.

Palmer et al. also observed that activation of dendritic GABA_B receptors could reduce layer 5 pyramidal cell spiking by up to 50%, even in the absence of excitatory dendritic input [19]. In addition to the axon hillock, a second action potential initiation site is present in the apical tufts of principal cells and Ca²⁺ spikes generated there can actually drive more axonal action potentials than supratheshold input near the soma [reviewed by 20], so inhibition of dendritic Ca²⁺ channels alone can strongly affect the output of principal cells. Taken together this suggests that in addition to halting ongoing network activity, the layer 1 GABA_B receptor-mediated inhibitory circuit can also suppress the activity of principal cells during awake states.

Source(s) of inhibition in layer 1?

All of the studies considered so far point to the existence of a specific circuit in layer 1 that, acting through GABA_B receptors, can inhibit the activity of principal cells by inhibiting dendritic Ca²⁺ channels [3,6–8,19] and terminate persistent activity [14,16]. Although the phenomenon is widely observed throughout many cortical regions, key mechanistic questions remain: is the source of inhibition a single class of interneuron? What are the inputs to this circuit? Whilst these questions remain open, several recent studies have begun to address them.

The class of inhibitory interneurons referred to as neurogliaform cells provide an attractive candidate as the source of inhibition in layer 1, as they inhibit pyramidal cells through a mixed GABA_A/GABA_B receptor-mediated mechanism [21]. Furthermore, a small number of neurogliaform cells have the potential to evoke far-reaching inhibition by activating GABA_B receptors via a seemingly non-discriminatory volume transmission [22,23] although new evidence [24**] suggests that the targets of the neurogliaform cells may be more synapse-specific than previously assumed. Neurogliaform cells can target both postsynaptic GABA_B receptors and presynaptic GABA_B receptors located on glutamatergic terminals [22]. Many studies have described the presence of these late-spiking neurogliaform cells in layer 1 [e.g. 25] and a recent study carried out using connected pairs of layer 1 interneurons and layer 2/3 pyramidal cells demonstrated that neurogliaform cells can strongly inhibit, but do not receive input from, layer 2/3 pyramidal cells [26*]. Importantly, this study found that neurogliaform cells inhibited layer 2/3 pyramidal cells through both GABA_A and GABA_B receptors, whereas other interneurons in layer 1 acted through GABA_A receptors alone.

In a Herculean display of electrophysiological panache, Jiang et al. demonstrated, through quadruple – octruple patch clamp recordings, that a group of layer 1 neurogliaform cells could strongly inhibit layer 5 pyramidal cells via direct inhibition and through electrical connections with three types of layer 2/3 interneurons; the layer 1 neurogliaform cells described appeared different to their counterparts in deeper layers, with elongated processes that spanned multiple cortical columns [27**]. All neurogliaform cells described in the hippocampus and amygdala stain for neuropeptide Y (NPY), and hippocampal neurogliaform cells can be further parsed into two sub-populations differentiated by the expression of nitric oxide synthase (NOS) [28,29]. However, the situation may be more complex in the cortex. While NPY expression has been reported for all cortical neurogliaform cells including layer 1 [30,31*], a recent study reported that NOS-positive elongated neurogliaform cells in layer 1 (morphologically similar to those reported by Jiang et al. [27]) do not stain for NPY, raising the possibility that they represent a third distinct group [32*], perhaps with different embryonic origins. This is not without precedent: while neocortical neurogliaform cells are believed to share common embryonic origins [33–35], hippocampal neurogliaform cells parse into two groups with distinct embryonic origins but similar morphological and electrophysiological properties [36]. It is possible that two (or perhaps three) classes of neurogliaform-like interneurons reside in layer 1: those expressing NPY with ‘typical’ neurogliaform morphology, and an NPY lacking population with a wider dendritic arbour. Interestingly, direct application of NPY to the distal dendrites of layer 5 pyramidal cells is sufficient to inhibit dendritic Ca²⁺ transients, without the need for either GABAergic or glutamatergic transmission [37*].

While current evidence strongly implies neurogliaform cells as the effector of the layer 1 GABA_B receptor-mediated inhibitory circuit, further work is needed to conclusively test this hypothesis. The origin of the inputs to this circuit also remains unclear. Only 10% of the inputs to layer 1 arise from the local circuit, with long-range connections providing the remainder [20]. Layer 1 receives cortico-cortical inputs from the ipsilateral hemisphere [e.g. 38,39,40] and, as already discussed, interhemispheric inhibition relies on inputs to layer 1 from the contralateral hemisphere [19]. Layer 1 receives additional input from subcortical regions: one in vivo study reported that layer 1 interneurons responded to whisker stimulation with sub-10 millisecond latencies, implying direct thalamic input [41], which was anatomically confirmed by a later study [42]. Through expression of channelrhodopsin in the midline thalamus, a recent study found that layer 1 interneurons, especially late-spiking (presumably neurogliaform) cells, could be strongly activated by thalamic input, even driving feed-forward inhibition of other L1 interneurons and deeper pyramidal cells [43**]. Thus, future studies could try to selectively drive neurogliaform cells to actively terminate Up states.

Functional significance?

While the circuitry of the GABA_B receptor-mediated inhibitory system in cortical layer 1 has yet to be fully mapped, it is clear that GABA_B receptors can strongly regulate cortical activity in a state-dependent manner. In the awake animal, the system regulates interhemispheric inhibition [19], and feedback inhibition via long-range connections to layer 1 is important in cognition and conscious perception, with regulation of dendritic calcium signalling being proposed as a key cellular mechanism for association [20].

Slow wave sleep is believed to be important for long-term memory consolidation [e.g. 44,45], where hippocampal sharp waves can drive cortical Up states [46], allowing the activity of neuronal ensembles to be temporally coordinated across different brain regions [47]. As the layer 1 inhibitory circuit can rapidly terminate an ongoing Up state [14,16], this could provide the mechanism for mediating the long-range synchrony of Up-to-Down state transitions observed in vivo [48]. Inhibition via GABA_B receptors could ensure that only the appropriate ensemble is selected for memory consolidation, and may provide a mechanism for a proposed thalamic driver of the slow oscillation [12].

Concluding remarks

This review has focused primarily on GABA_B receptors in layer 1 and in particular, their role in modulation of slow oscillations. However, GABA_B receptors also influence faster hippocampal oscillations, such as those in the theta and gamma frequency ranges [reviewed by 49]. The mechanisms by which this is achieved may be numerous: recent work has shown that postsynaptic GABA_B receptors in the dentate gyrus enhance the function of extrasynaptic GABA_A receptors [50], and that postsynaptic GABA_B receptors inhibit perisomatic-targeting but not dendritic-targeting parvalbumin-positive interneurons [51]. Presynaptic GABA_B receptors, in addition to contributing to spontaneous termination of the Up state [16], mediate disinhibition of dentate granule cell output [52], and may regulate excitability at the hippocampal mossy fibre terminal [53]. Dysfunction of GABA_B receptor-mediated signalling has been implicated in several models of absence epilepsy [54], highlighting the importance of this receptor in normal network function.

Future work should focus on uncovering the nature of the layer 1 inhibitory circuit that is mediated by GABA_B receptors, such as identifying which cell type(s) are the main effectors and mapping the inputs to the circuit. While several existing optogenetic tools will be useful for dissecting this circuitry, the creation of mice with floxed GABA_B1a and GABA_B1b subunits to allow conditional inactivation of pre- or post-synaptic GABA_B receptors in a neuron subtype-specific manner would greatly aid this research, especially given the multitude of mice available that express Cre recombinase in different interneuron subtypes [55]. Additionally, the study of a group of recently-identified GABA_B receptor auxiliary subunits from the KCTD family [56], which have variable expression throughout the brain and modify receptor properties in a subtype-specific manner [56–59], could provide new insights into how GABA_B receptors influence neuronal circuits. Although challenges clearly remain, future advances in understanding how the ‘slow’ GABA_B receptor can effect rapid changes in network state could present exciting new insights into the genesis of synchronous neuronal activity, in both healthy and pathological conditions.

Highlights for Craig & McBain.

• Metabotropic GABA_B receptors can, unexpectedly, mediate rapid termination of persistent network activity in the cortex.

• Layer 1 interneurons modulate network activity by targeting GABA_B receptors on principal cells.

• Dendritic GABA_B receptors can inhibit the firing of principal cells by acting on voltage-gated calcium channels.