Multi-transmitter neurons in the mammalian central nervous system

Highlights

• Neurons are capable of releasing multiple diverse neurotransmitters.

• Multi-transmitter release must be considered when studying neural connectivity.

• Recent results from four examples of multi-transmitter neurons are discussed.

• These include GABA/Ach, Dopamine/GABA, Dopamine/Glutamate, and Glutamate/GABA.

• These examples illustrate different possible mechanisms of multi-transmitter release.

It is firmly established that many mammalian neurons release various combinations of amino acids, their derivatives, and other small molecules from presynaptic terminals in order to signal to their postsynaptic targets. Here we discuss recent findings about four types of multi-transmitter neurons—those that release GABA and acetylcholine (Ach); dopamine (DA) and GABA or glutamate; and glutamate and GABA. The mechanisms of co-release in each class differ and highlight the complex and dynamic nature of neurotransmitter release. Furthermore, identifying the neurotransmitter signature of each neuron and the post-synaptic targets of each neurotransmitter remain challenging. The existence of multi-transmitter neurons complicates the interpretation of connectomic wiring diagrams and poses interesting challenges for our understanding of circuit function in the brain.

Introduction

Our conception of how a particular presynaptic neuron influences downstream cells is determined by which small molecule-based neurotransmitter – for example, 5-HT, Ach, GABA, glutamate, and monoamines – it releases. However, many additional classes of molecules – for example, peptides (substance P, opioids, neuropeptide Y), gases (nitric oxide), and lipophilic molecules (prostaglandins, endocannibinoids) – are released in an activity-dependent manner and used to signal to neighboring neurons. In addition, many neurons release multiple small-molecule neurotransmitters. Thus, the designation of one molecule as a principal neurotransmitter obscures the true diversity of synaptic signaling.

Rather than repeat recent comprehensive reviews [1, 2, 3], here we discuss four examples of multi-transmitter neurons for which new results have recently been published. These are neurons in the entopeduncular nucleus (EP) that release GABA and glutamate, neurons in the substantia nigra pars compacta (SNc) and ventral tegmental area (VTA) that release DA and GABA or glutamate, and vasoactive intestinal peptide (VIP)-expressing cortical interneurons that release GABA and Ach. Although we focus on results from the analysis of mouse brain, some of these neurotransmitter combinations are conserved across brain areas and species. In many species that are not genetically tractable and in studies from the pre-optogenetics era, the functional demonstration of co-transmission is often lacking. Nevertheless, neurons with markers or functional evidence suggesting a dual GABAergic/cholinergic phenotype are found, for example, in C. elegans, birds, and the mammalian retina [4, 5, 6, 7, 8]. Furthermore, in zebrafish, catecholamine-producing neurons likely release GABA or glutamate [9], and DANs in mammalian retina and olfactory bulb also appear to be GABAergic [10, 11]. The examples presented here therefore highlight common motifs shared across species and brain areas.

The entopeduncular nucleus (EP; analogous to the globus pallidus internus (GPi) in primates) is a major output nucleus of the basal ganglia and sends a large projection to lateral habenula (LHb). In monkeys [12] and in mice [13^••], rewarding outcomes reduce and aversive ones increase firing in EP → LHb projecting neurons. This is consistent with the LHb evaluating the outcome of an action and biasing future behavior to maximize positive outcomes and minimize negative ones. Optogenetic studies in mice confirm a causal role for EP → LHb projections in reward motivated behaviors [13^••] and are consistent with a net inhibitory effect of LHb activity on dopaminergic neurons of the VTA [15].

A neuron subtype in the EP projects only to the LHb [16, 17] and releases both GABA and glutamate from the same axon onto individual LHb neurons [18^••]. These neurons contain GABA synthesizing enzymes (Gad1/GAD67 and Gad2/GAD65) as well as vesicular transporters for GABA (Slc32a1/VGAT) and glutamate (Slc17a6/VGLUT2), suggesting that both neurotransmitters are packaged into synaptic vesicles via canonical mechanisms [18••, 19]. Interestingly, GABA and glutamate appear to be packaged into the same synaptic vesicles, resulting in simultaneous release of both neurotransmitters (Figure 1a). This conclusion is based on: (1) the electrophysiological observation of biphasic miniature and spontaneous post-synaptic currents at holding potentials between the reversal potentials of GABA and glutamate; and (2) ultrastructural observation by immunoelectron microscopy of GABA and VGLUT2 in close proximity (<30 nm) to each other in EP axon terminals within the LHb [18^••]. Interestingly, inputs to LHb from the ventral tegmental area also release both glutamate and GABA [20•, 21], but, in this case, apparently from distinct sites [20^•].

In vitro, optogenetic activation of EP axons leads to time-locked spiking of neurons in the LHb [18^••], suggesting a net excitatory effect despite the release of GABA. In this case, simultaneous release of both GABA and glutamate may serve to sharpen the time course of excitation by accelerating the decay of the post-synaptic potential. Alternatively, pre- or post-synaptic mechanisms may dynamically alter the glutamate/GABA current ratios evoked by activity of individual EP → LHb neurons. Indeed, synaptic transmission within the LHb is highly plastic and GABA and glutamate release from EP axons are modulated independently [18••, 22]. Induction of depressive-like states in rodents and their treatment with selective serotonin reuptake inhibitors (SSRIs) or exposure of EP axons to serotonin in vitro, appears to act on the presynaptic terminal to modulate GABA and glutamate release [18••, 23]. On the postsynaptic side, potential independent regulation of GABA and glutamate receptor trafficking to the synapse would allow each LHb neuron to set the relative strength of each input in a continuous fashion from strongly inhibitory to strongly excitatory. Since the LHb appears to lack GABAergic interneurons, such regulation may be necessary for projection neurons to implement gain control.

Alternatively, the graded regulation of synaptic weights of EP inputs could be used by LHb neurons to learn to encode ‘reward prediction errors (RPE)’—that is, to modulate activity in proportion to the degree of unexpectedly bad or good outcomes in specific contexts. Some inputs might predict an aversive outcome and thus need to contribute positively to activity in the LHb whereas others might predict reward and contribute negatively. The sign and strength of each synapse from GABA/glutamate multi-transmitter EP neurons, which carry information about sensory state, motor action, and other environmental variables, may be regulated to assign such negative and positive contributions to the activity of individual LHb neurons. More studies are needed to determine if GABA/glutamate co-transmission serves any of these proposed functions.

Axons of midbrain dopamine neurons (DANs) release GABA in the striatum and nucleus accumbens (NAC) [24]. This is difficult to deduce based on transcriptional and immunohistochemical analyses since DANs in the ventral tegmental area (VTA) and substantia nigra pars compacta (SNC) do not express classical synthesizing enzymes (Gad1/GAD67 and Gad2/GAD65) or vesicular transporters (Slc32a1/VGAT) for GABA and these genes are not essential for GABA release from DANs onto striatal projection neurons (SPNs) [24, 26••]. Instead, the brain vesicular monoamine transporter (Slc18a2/VMAT2) is necessary in DANs for GABA release and can substitute for VGAT to sustain GABA release in classical GABAergic neurons, suggesting that VMAT2 is a vesicular GABA transporter [24]. Nevertheless, biochemical demonstration of GABA transport by VMAT2 across reconstituted membranes is lacking. The GABA packaged and released by DANs is acquired from the extracellular space via action of a plasma membrane GABA transporter (Slc6a1/GAT1) or produced autonomously via Aldh1a1-dependent GABA synthesis [25•, 26••]. We propose that essentially all SNC DANs release GABA, given the near ubiquitous expression of either Aldh1a1 or GAT1 [25•, 26••, 27]. Furthermore, as is true of DA, GABA can be released from DAN axons in an action potential independent manner, for example, following opening of presynaptic nicotinic cholinergic receptors [28, 29].

It is difficult to prove that the molecule released from DA axons that activates GABA_A receptors in SPNs is GABA as opposed to another agonist of the notoriously promiscuous GABA_A receptors. Evidence in favor of GABA is that GABA_A receptor mediated currents in SPNs evoked by optogenetic activation of DAN axons are abolished by antagonists of the plasma-membrane GABA transporter GAT1, reduced by blocking the GABA-producing enzyme Aldh1a1, and sustained by exogenous expression of VGAT in DANs when VMAT2 is inhibited [24, 25•]. Furthermore, given the apparent function of VMAT2 as a vesicular GABA transporter, DA and GABA are likely packaged into and released from the same vesicles (Figure 1b). Many other central monoaminergic neurons, which universally express VMAT2, may be GABAergic depending on the expression of GABA synthetic enzymes or uptake mechanisms.

Recently, the function of GABA release from DA neurons has been probed. As is true of other aldehyde dehydrogenases, Aldh1a1 metabolizes alcohols, which displace and antagonize processing of endogenous substrates; thus ethanol decreases Aldh1a1-dependent GABA production and reduces GABA release from DANs [26^••]. Conversely, via unclear mechanisms, deletion of Aldh1a1 increases ethanol intake in mice. A separate study demonstrated that deletion of Ube3A, the gene underlying Angelman’s syndrome, from tyrosine hydroxylase-expressing neurons decreases GABA release in NAC from VTA DANs without altering DA release [30^••]. Furthermore, this manipulation enhances operant self-stimulation of VTA neurons, an effect that is reversed by restoring GABA release with exogenous over-expression of VGAT in DANs.

These studies provide the first hints that release of GABA from DANs is an important regulator of striatal circuitry. DA increases the spiking rate of active Type 1 DA receptor-expressing direct pathway SPNs (dSPNs) [31]. Therefore, simultaneous release of DA and GABA, even via volume transmission, could enhance spiking in active dSPNs related to ongoing behavior while inhibiting action potential generation of inactive neurons. Depending on one’s theory of DA and basal ganglia function, this would be predicted to favor completion of the current action sequence, increase the vigor with which the action is carried out, or promote repetition of the action in the future [32, 33, 34, 35, 36]. How such potential effects relate to alcohol consumption or self-stimulation of VTA remains unclear.

A subset of DANs in the VTA, but likely not the SNC, express Slc17a6, the gene encoding the vesicular glutamate transporter VGLUT2 and release glutamate in an action-potential-dependent manner (reviewed in Ref. [37]). Early optogenetic studies reported glutamate release from DANs in NAC brain slices [38, 39], and subsequent analyses revealed similar glutamate release in dorsal striatum [24, 25•]. It is possible that DAN-mediated glutamate release in both NAC and striatum originates from VTA axons, since these also enter the dorsal striatum but reflect different aspects of behavior than do SNC DA axons [40]. The presynaptic regulation of glutamate release from DANs is different than that of GABA/DA, suggesting that the release sites are likely distinct, consistent with ultrastructural analyses (Figure 1c) [25•, 41, 42•]. The function of glutamate releasing DA neurons is unknown and whether glutamate/DA neurons also release GABA has not been determined electrophysiologically.

We previously reported the widespread potential for cholinergic neurons of the forebrain, including the major basal forebrain projection to cortex, to co-release GABA [1, 2, 43••, 44]. Another potential source for cortical Ach comes from relatively less studied cholinergic interneurons that are intrinsic to cortex. Immunohistochemical analysis indicates that these cortical cholinergic interneurons are a subset of VIP-expressing interneurons, as all choline acetyltransferase (ChAT)-expressing interneurons label for VIP, and ∼1/3 of all cortical VIP interneurons are positive for ChAT [45, 46, 47, 48, 49, 50]. Several studies reported variable expression of the GABAergic markers GAD65 and GAD67 in cholinergic cortical interneurons [46, 49, 51] indicating they may not be GABAergic. However, to our knowledge, a non-GABAergic population of cortical VIP interneurons has never been described, and this variability may reflect difficulty in reliably immunostaining for GABAergic markers at the cell body rather than the lack of a GABAergic phenotype.

Cortical interneurons that co-express ChAT and VIP have also been identified by single-cell transcriptional analysis of cortical neurons. A study of 3000 single-cell transcriptomes from cortex and hippocampus identified 16 interneuron subclasses, one of which (their #9) co-expresses transcripts for VIP and cholinergic markers [52^•]. Further analysis of their data (http://linnarssonlab.org/cortex/) reveals that this subclass also expresses the vesicular acetylcholine transporter (Slc18a3/VAChT), Ach synthetic enzyme (Chat/ChAT), and membrane choline transporter (Slc5a7/ChT). Independent transcriptional analysis of 1,679 cells from visual cortex of adult male mice with greater read-depth [53^•] identified 23 distinct GABAergic cell populations, with one (cluster 46, consisting of six neurons) distinguished primarily by co-expression of VIP and ChAT. Both of these studies show robust expression of the GABAergic transcripts Gad1/GAD67 and Slc32a1/VGAT, in contrast to the variable GABAergic identity reported by immunostaining. Combined with the previous analyses of protein expression, these results indicate that ChAT interneurons are a distinct subclass of VIP GABAergic interneurons.

Relatively little is known about the functional properties of these VIP/ChAT interneurons. One study explicitly probed their synaptic properties using paired recordings between ChAT interneurons and nearby pyramidal cells [51]. This approach did not reveal a specific post-synaptic effect of Ach release, but instead identified presynaptic modulation of excitatory inputs onto pyramidal neurons via nicotinic Ach receptors (nAchRs). However, the use of a GFP-expressing BAC transgenic line to identify ChAT interneurons may have altered the properties of synaptic release through overexpression of VAChT from the BAC or introduced unclear biases in the selection of presynaptic neurons [54]. The potentially important contributions of GABA release from these cells as well as effects of Ach on non-pyramidal targets were not examined. An independent study, focused on cooperative activity between interneuron classes [55], showed that blocking nAChRs slightly decreases the degree to which firing in VIP interneurons stimulates neighboring VIP interneurons, suggesting that Ach release from VIP/ChAT cells normally excites other VIP neurons. Given previous evidence that GABA release from VIP interneurons inhibits somatostatin-expressing interneurons but not other VIP interneurons [56], this suggests that VIP/ChAT interneurons may release Ach and GABA onto different post-synaptic targets, perhaps from separate synaptic vesicle populations (Figure 1d). However, further experiments are required to explicitly test this possibility.