Astrocytic striatal GABA transporter activity governs dopamine release and shows maladaptive downregulation in early parkinsonism

Striatal dopamine (DA) is critical for action and learning. Recent data show DA release is under tonic inhibition by striatal GABA. Ambient striatal GABA tone on striatal projection neurons can be governed by plasma membrane GABA uptake transporters (GATs) on astrocytes. However, whether striatal GATs and astrocytes determine DA output are unknown. We reveal that DA release in mouse dorsolateral striatum, but not nucleus accumbens core, is governed by GAT-3 and GAT-1. These GATs are partly localized to astrocytes, and are enriched in dorsolateral striatum compared to accumbens core. In a mouse model of parkinsonism, GATs become downregulated and tonic GABAergic inhibition of DA release augmented, despite attenuated GABA co-release from dopaminergic axons. These data define previously unappreciated and important roles for GATs and astrocytes in determining DA release in striatum, and reveal that they underlie maladaptive plasticity in early parkinsonism that impairs DA output in vulnerable striatal regions. Highlights GABA transporters set the level of GABA inhibition of DA output in dorsal striatum Astrocytes facilitate DA release by limiting tonic GABA inhibition Tonic GABA inhibition of DA release is augmented in mouse model of parkinsonism DA and GABA co-release are reduced in mouse model of parkinsonism


Introduction
Dopamine (DA) release in the dorsal and ventral striatum plays key roles in action selection and motivation, and is dysregulated in diverse disorders including Parkinson's disease (PD) and addictions. Striatal DA release is gated locally by axonal mechanisms and striatal neuromodulators that regulate or even drive DA release (Schmitz et al., 2003;Sulzer et al., 2016). It has recently been revealed that DA release is under tonic inhibition by striatal GABA (Lopes et al., 2018), operating through GABA A and GABA B receptors presumably located directly on DA axons (Lopes et al., 2018;Pitman et al., 2014;Schmitz et al., 2002). The striatum contains a high density of GABAergic projection neurons and interneurons and, in addition, receives a source of GABA co-released from mesostriatal DA neurons (Kim et al., 2015;Tritsch et al., 2012Tritsch et al., , 2014. Given the paucity of GABAergic synapses on DA axons (Charara et al., 1999), tonic inhibition of DA release by striatal GABA is presumably mediated through extrasynaptic effects of ambient GABA (Lopes et al., 2018). GABA can spill over for extrasynaptic actions in other nuclei (Farrant and Nusser, 2005), and in the dorsal striatum, provides a sizeable ambient GABA tone on spiny projection neurons (SPNs), evident as a tonic GABA A receptor-mediated inhibitory conductance (Ade et al., 2008;Cepeda et al., 2013;Kirmse et al., 2008Kirmse et al., , 2009Santhakumar et al., 2010).
Tonic inhibition by ambient GABA across the mammalian brain is usually limited by uptake by plasma membrane GABA transporters (GATs) (Brickley and Mody, 2012). There are two isoforms of the GAT in striatum: GAT-1 (Slc6a1), abundant in axons of GABAergic neurons (Augood et al., 1995;Durkin et al., 1995;Ng et al., 2000;Yasumi et al., 1997); and GAT-3 (Slc6a11), expressed moderately (Ficková et al., 1999;Ng et al., 2000;Yasumi et al., 1997) and seen particularly on astrocytes (Chai et al., 2017;Ng et al., 2000;Yu et al., 2018). Emerging transcriptomic data indicate that striatal astrocytes might express both GAT-1 and GAT-3 (Chai et al., 2017;Gokce et al., 2016;Zhang et al., 2014). In addition, mRNA for GAT-1 and for  has been found in midbrain DA neurons and these GATs have been suggested but not confirmed to be located on striatal DA axons to support GABA co-storage and co-release (Tritsch et al., 2014). Ambient GABA tone on SPNs in dorsal striatum is limited by the activity of GAT-3 and GAT-1 (Kirmse et al., 2008(Kirmse et al., , 2009Santhakumar et al., 2010;Wójtowicz et al., 2013), and recent evidence indicates that GAT-3 on striatal astrocytes play a particularly important role: GAT-3 dysregulation results in profound changes to SPN activity and striatum-dependent behavior (Yu et al., 2018). However, whether striatal GAT function is a critical for setting the level of DA output has not previously been examined.
Here we reveal that GAT-3 and GAT-1 strongly regulate striatal DA release in the dorsolateral striatum (DLS) but not in the nucleus accumbens core (NAcC), by limiting tonic inhibition by striatal ambient GABA. We identify a particular role for GATs located on striatal astrocytes in supporting DA release, and furthermore, reveal that maladaptive GAT regulation impairs DA output in the DLS in a mouse model of early parkinsonism.

DA release in DLS and NAcC is tonically inhibited by a GAD-dependent GABA source
We recently reported that axonal DA release in the dorsal striatum is under tonic inhibition by striatal GABA, as GABA A and GABA B receptor antagonists enhanced DA release evoked by single electrical and targeted light pulses (Lopes et al., 2018). Since mechanisms that regulate striatal DA release can diverge between dorsal and ventral striatal territories (Brimblecombe et al., 2015;Britt and McGehee, 2008;Janezic et al., 2013;Shin et al., 2017;Threlfell and Cragg, 2011;Threlfell et al., 2010), we determined whether DA release in NAcC, a part of the ventral striatum, is similarly regulated by tonic GABA inhibition.
We used fast-scan cyclic voltammetry (FSCV) in acute coronal slices of mouse brain to detect extracellular concentration of DA ([DA] o ) at carbon-fiber microelectrodes evoked by single electrical pulses in DLS and NAcC ( Figure 1A). Co-application of GABA A and GABA B receptor antagonists (+)-bicuculline (10 M) and CGP 55845 (4 M) respectively, significantly enhanced electrically evoked [DA] o by ~20% in either the DLS or NAcC, when compared to drug-free time-matched controls (DLS: p = 0.0004, n = 6 experiments/3 mice; NAcC: p = 0.001, n = 5 experiments/3 mice; Figure 1B). These effects were similar in DLS and NAcC ( Figure   1B; p > 0.9; Mann-Whitney tests) and did not require striatal glutamate input (Supplementary Figure S1), nor cholinergic interneuron input to nAChRs (since experiments were conducted in the presence of the nAChR antagonist DHE). Using an optogenetics-based strategy to activate DA axons selectively, we also confirmed that inhibition of DA release by GABA receptors does not require concurrent activation of striatal GABAergic microcircuits ( Figure 1C). GABA receptor antagonism significantly enhanced [DA] o evoked by single light pulses by ~25% in both the DLS and NAcC compared to time-matched drug-free control ( Figure   1D; DLS: p < 0.0001, n = 9 experiments/5 mice; NAcC: p = 0.002, n = 5 experiments/4 mice), and similarly so in DLS vs. NAcC ( Figure 1D; p > 0.9; Mann-Whitney tests). These results confirm that DA release is under tonic inhibition by GABA in striatal regions spanning dorsal to ventral.
We tested whether GABAergic inhibition of DA release arose from GABA co-released by DA axons or from GABA originating from a canonical neuron source (i.e. striatal GABAergic neurons). Mesostriatal DA neurons synthesize, co-store and co-release GABA (Tritsch et al., 2012), with GABA synthesis depending on aldehyde dehydrogenase (ALDH)-1a1 (Kim et al., 2015). In contrast, canonical synthesis of GABA in neurons requires glutamic acid decarboxylase (GAD). We inhibited GABA synthesis in DA axons by pre-treating slices with ALDH inhibitor disulfiram (10 μM) for 2 to 4 hours, which depleted light-evoked GABA currents from DA axons onto SPNs by half (Supplementary Figure S2), as reported previously (Kim et al., 2015). However, ALDH inhibition did not prevent GABA receptor antagonists from enhancing DA release: in the DLS, in the presence of disulfiram, GABA receptor antagonists enhanced light-evoked [DA] o by ~40%, which was a significantly larger effect than seen without disulfiram ( Figure 1E; with disulfiram versus without, p = 0.005, Mann-Whitney test; disulfiram present: n = 6 experiments/5 mice; disulfiram absent: n = 10 experiments/7 mice). These data suggest that GABA co-released from DA axons does not directly inhibit DA release. They also suggest that ALDH-dependent GABA acts indirectly to limit tonic inhibition by a different, ALDH-independent source. To assess whether tonic inhibition of DA release depended instead on a GADdependent GABA source, we pre-treated brain slices with GAD inhibitor 3-mercaptopropionic acid (3-MPA, 500 μM) for 2 to 4 hours, which attenuates electrically evoked GABA transmission onto SPNs by more than half (Kim et al., 2015). After GAD inhibition, the disinhibition of DA release in the DLS by GABA receptor antagonists was attenuated ( Figure 1E; with 3-MPA vs. without, p = 0.008, Mann-Whitney test; 3-MPA: n = 7 experiments/5 mice; 3-MPA absent: n = 10 experiments/7 mice), indicating that a GAD-dependent GABA source provides tonic inhibition of striatal DA release.

GAT inhibition attenuates striatal DA release by increasing GABA receptor tone
We ruled out diminished DA storage as a cause of the attenuation of DA release following GAT inhibition: Striatal DA content measured using high performance liquid chromatography (HPLC) with electrochemical detection was unchanged after incubation in NPA compared to controls ( Figure 2H; p = 0.60, Mann-Whitney test, n = 19 experiments/5 mice per condition). Instead, we confirmed that GAT inhibition modified DA release in a GABA receptor-dependent manner. The acute effects of NPA on evoked [DA] o were prevented in the presence of antagonists for GABA A and GABA B receptors combined (picrotoxin, 100 M, CGP55845, 4 M) ( Figure 2I; without GABA receptor antagonists: ~65% of baseline; with GABA receptor antagonists: ~95% of baseline; p = 0.001, Mann-Whitney test, n = 5 experiments/4 mice), consistent with the effects of GAT on DA release being mediated via extracellular GABA acting on GABA receptors. We have previously shown that activation of GABA receptors can slightly modify the activitydependence of DA release during short stimulus trains (Lopes et al., 2018). Consistent with an increase in  Figure S3) consistent with a decrease in DA release probability (Jennings et al., 2015). Together these data indicate that GAT inhibition attenuates DA release through increasing GABA receptor tone.
We assessed whether the greater effect of GAT inhibition in DLS than NAcC (see Figure 2) was due to differences in GABA receptor function between regions. However, bath application of exogenous GABA (2 mM) attenuated [DA] o evoked by 1p electrical stimulation to a similar degree in DLS and NAcC ( Figure   2K; DLS vs. NAcC: p = 0.33, Mann-Whitney test, DLS n = 6 experiments/4 mice, NAcC, n = 5 experiments/4 mice), ruling out overall differences in GABA receptor function as a factor. These findings suggest that the function of GAT in limiting tonic inhibition differs between DLS and NAcC.
GABA tone in striatum has previously been reported to be action-potential independent i.e. due to spontaneous GABA release (Wójtowicz et al., 2013). We assessed whether GATs were limiting an action potential-independent GABA tone. In the presence of Na v blocker tetrodotoxin (TTX, 1 M), NPA increased the GABA A -mediated holding current in SPNs in the DLS ( Figure 3G, NPA vs. baseline: p = 0.03, Wilcoxon signed-rank test; n = 6 cells/3 mice) to a level not different to that induced by NPA in TTX-free conditions ( Figure 3H; p = 0.43, Mann-Whitney test), confirming that GATs in DLS are limiting a spontaneous GABA tone.
Collectively, these results show that striatal GAT-1 and GAT-3 regulate an ambient GABA tone, which arises from an action-potential independent source, and do so to a greater degree in DLS compared to NAcC. We explored an anatomical basis for the regional heterogeneity in GAT function. Striatal immunoreactivity to GAT-1 and GAT-3 in the DLS and NAcC revealed relative enrichment in the DLS for both GAT1 ( Figure 4A,B, p = 0.009, Wilcoxon signed-rank test, n = 12 hemispheres/6 mice) and GAT-3 ( Figure   4C,D; p = 0.002, Wilcoxon signed-rank test, n = 12 hemispheres/6 mice). We also noted enriched GAT-3 in the medial NAc shell (NAcS) contiguous with the medial septal nucleus (Supplementary Figure S4). This observation prompted us to identify the effects of GAT inhibition on DA release in NAcS. Correspondingly, we noted that GAT inhibition diminished electrically evoked [DA] o in NAcS, unlike NAcC (Supplementary Figure S4), indicating further regional heterogeneity in the role of GATs in limiting tonic inhibition across the striatum.

GAT-1 and GAT-3 on astrocytes are key regulators of ambient GABA inhibition of DA release
Striatal GATs are located on the plasma membranes of a variety of cells that include GABAergic neurons (Augood et al., 1995;Durkin et al., 1995;Ng et al., 2000;Yasumi et al., 1997) and astrocytes (Chai et al., 2017;Ng et al., 2000;Yu et al., 2018). GATs have also been presumed, but not confirmed, to reside on DA axonal membranes to support GABA uptake, co-storage and co-release (Tritsch et al., 2014). To better understand where GAT is located to regulate tonic GABAergic inhibition of DA, we probed two candidate locations, namely DA axons, and astrocytes. We explored whether GAT-1 or GAT-3 could be detected on DA axons using immunofluorescence and confocal microscopy, but we did not find robust evidence for localization of these proteins to DA axons identified by an eYFP reporter (Supplementary Figure S5). As a positive control for our immunofluorescence and imaging, we confirmed that GAT-1 was localized to the neurites of parvalbumin-expressing GABAergic interneurons (Supplementary Figure S6), which are well known to express GAT-1 (Augood et al., 1995). Our observations do not provide support for the assumption that GAT-1 and GAT-3 proteins are localized to DA axons (Tritsch et al., 2014) in order to provide GABA uptake for co-storage and co-release.
In many brain regions, including striatum, astrocytes are thought to regulate ambient GABA levels by active uptake of GABA (Yu et al., 2018). GAT-3 protein expression has been documented on striatal astrocytes (Chai et al., 2017;Ng et al., 2000;Yu et al., 2018), and although GAT-1 is typically associated with neuronal structures (Borden, 1996), recent transcriptomic studies have found RNA for both GAT-3 and GAT-1 in striatal astrocytes (Chai et al., 2017;Gokce et al., 2016;Zhang et al., 2014). We revisited GAT-1 localization, using immunofluorescence and confocal microscopy with antibodies directed against GAT-1 or GAT-3, as well as against the striatal astrocytic marker S100β (Chai et al., 2017) in the DLS and NAcC ( Figure   5A,B). As expected, we found that GAT-3 was robustly co-localized to the plasma membranes of S100βexpressing astrocytes ( Figure 5C, n = 3 mice). We also found some instances of localization of GAT-1 on S100β astrocytes ( Figure 5D, n = 3 mice). These data indicate that, although GAT-1 is commonly expressed by striatal GABAergic interneurons, GAT-1 can also be expressed by some astrocytes in striatum.
We next probed whether GATs on striatal astrocytes could govern tonic GABAergic inhibition of DA release. To address this, we pre-treated striatal slices with fluorocitrate (200 µM for >1hr, or vehicle control), an established approach for metabolically inhibiting astrocytes, thus rendering them inactive and preventing the effects of astrocytic GAT (Boddum et al., 2016;Bonansco et al., 2011). DA release in fluorocitrate-and vehicle-treated slices was then recorded with/without the GAT inhibitor NPA. We first confirmed that we could detect the effects of GAT inhibition on DA release in DLS in control slices.
Accordingly, [DA] o evoked by 1p electrical stimulation across a range of sites in the DLS from slices incubated in NPA (1.5 mM) for 30 min was significantly less than in NPA-free control conditions, as expected ( Figure 5E; p = 0.0003, Mann-Whitney test; n = 24 observations/5 mice for each condition), and 4p/1p ratio (50 Hz) was appropriately enhanced ( Figure 5F; p = 0.014, Mann-Whitney test; n = 8 experiments/5 mice for each condition). By contrast, in slices pre-treated with fluorocitrate to inactivate astrocytes, NPA did not significantly modify [DA] o evoked by 1p ( Figure 5G; p = 0.10, Mann-Whitney test, n = 42 observations/7 mice for each condition), or the 4p/1p ratio (50 Hz), compared to NPA-free conditions ( Figure 5H; p = 0.64, Mann-Whitney test, n = 13 observations/7 mice for each condition). We noted also that evoked [DA] o was lower in fluorocitrate-treated vs vehicle-treated slices ( Figure 5I; p = 0.0001, Mann-Whitney test). Additionally, the inhibition of [DA] o by NPA was attenuated when astrocytes were inhibited compared to not ( Figure 5J; p = 0.004, Mann-Whitney test). Together, these data indicate that GATs on striatal astrocytes regulate the level of inhibition of DA release by ambient GABA.

Tonic inhibition of DA release in the DLS is augmented in a mouse model of parkinsonism
Our data described above provide compelling evidence that GATs regulate DA output in the DLS.
Intriguingly, dysregulation of GATs within the basal ganglia has been implicated in models of neurological disease: in a 6-hydroxydopamine-induced mouse model of Parkinson's, astrocytes in the external globus pallidus have downregulated GAT-3 (Chazalon et al., 2018); and in R6/2 and FVB/N transgenic mouse models of Huntington's disease, GAT expression in striatum is increased and tonic inhibition by ambient GABA decreased (Cepeda et al., 2013;Wójtowicz et al., 2013;Yu et al., 2018). Given that deficits in DA transmission occur in dorsal striatum, but not in NAcC, in several transgenic rodent models of early parkinsonism prior to cell loss (Janezic et al., 2013;Sloan et al., 2016;Taylor et al., 2014), we explored whether tonic GABAergic inhibition of striatal DA release and its regulation by striatal GAT might be affected in a mouse model of early parkinsonism.
We chose to use SNCA-OVX mice, a model of early parkinsonism (Janezic et al., 2013). SNCA-OVX mice are devoid of mouse -synuclein but overexpress human wildtype -synuclein at disease-relevant levels and show early deficits in DA release prior to DA cell loss (Janezic et al., 2013). To address our aims, we made these SNCA-OVX mice 'optogenetics capable', such that they allowed for optical manipulation of DA axons. We crossed Slc6a3 IRES-Cre mice with -synuclein knockout mice to create Slc6a3 IRES-Cre mice devoid of mouse -synuclein, and then crossed these mice with SNCA-OVX mice to generate two cohorts of mice, both devoid of mouse -synuclein: (1) "SNCA+" mice that express Cre recombinase in DA neurons and human -synuclein; and (2) as littermate background controls, "Snca-/-" mice that express Cre recombinase in DA neurons but no human transgene. We confirmed that, as observed in the original SNCA-OVX mice (Janezic et al., 2013), the resulting SNCA+ mice at 4 months exhibited a ~30% deficit in electrically evoked Furthermore, this new optogenetic capable model of parkinsonism allowed us to address for the first time whether DA release deficits in DLS are accompanied by corresponding deficits in GABA co-release from DA axons. To provide a readout of GABA co-release, we used voltage-clamp recordings of currents evoked in SPNs by optical stimulation of DA axons ( Figure 6D). We observed a significantly lower amplitude of light-evoked GABAergic co-release currents in SPNs of SNCA+ mice compared to Snca-/-mice ( Figure   6D,E; t = 2.68, df = 14, p = 0.018; SNCA+: n = 9 cells/4 mice, Snca-/-: n = 7 cells/4 mice). These evoked currents were GABA A receptor-mediated as they were eliminated by picrotoxin (PTX, 100 µM) ( Figure 6F; t = 4.55, df = 10, p = 0.001). This observed difference in GABAergic current amplitudes was not due to differences in series resistance ( Figure 6G; t = 0.23, df = 14, p = 0.82).
We then explored whether tonic GABA inhibition of DA release was modified in the parkinsonian model in DLS and NAcC. We found that GABA R antagonism enhanced [DA] o evoked by single light pulses to a significantly greater degree in SNCA+ mice than in Snca-/-controls in DLS ( Figure 6H; p = 0.0003, Mann-Whitney test; SNCA+: n = 8 experiments/5 mice, Snca-/-: n = 7 experiments/5 mice) but not in NAcC ( Figure   6I; p = 0.09, Mann-Whitney test; SNCA+: n = 8 experiments/5 mice, Snca-/-: n = 9 experiments/5 mice), which was a significant regional difference ( Figure 6J; p = 0.02, Mann-Whitney test). These data indicate that the GABA tone on DA axons is dysregulated in SNCA+ parkinsonian mice and in particular, that tonic GABAergic inhibition of DA release is augmented in DLS.
We tested the hypothesis that elevated tonic inhibition of DA release in the DLS of SNCA+ mice might be due to impaired GAT function. We tested the effect of the non-selective GAT inhibitor NPA on Whitney tests; n = 7 SNCA+ mice, n = 10 Snca -/-mice). Taken together, these data suggest that tonic inhibition of DA release by ambient GABA is augmented in the dorsal striatum in early parkinsonism due to decreased GAT-1 and GAT-3 (Figure 7).

DISCUSSION
We define a major role for striatal GATs and astrocytes in setting the level of DA output in the striatum. We show that GAT-1 and GAT-3, located at least in part on striatal astrocytes, govern tonic GABAergic inhibition of DA release. GATs operate in a heterogeneous manner across the striatum, substantially limiting tonic inhibition of DA release in DLS but not NAcC. Moreover, in a mouse model of early parkinsonism prior to the overt loss of DA neurons, we reveal maladaptive decreases in striatal GAT-1 and GAT-3 expression and consequently, profound augmentation of tonic inhibition of DA release by GABA in the dorsal striatum.

GATs limits the tonic inhibition of DA release
We found that tonic inhibition of DA release by GABA spans dorsal-ventral territories of striatum and arises from a GAD-dependent source of GABA. The source of GABA was not ALDH-dependent e.g. co-release from DA axons, as inhibition of ALDH did not attenuate the tonic inhibition of DA release by GABA, despite attenuating GABA co-release from DA axons, as seen previously (Kim et al., 2015). Conversely, ALDH-inhibition slightly boosted tonic inhibition of DA release, suggesting that ALDH-dependent sources of GABA, such as GABA co-release from DA axons, limit the tonic inhibition by the GAD-dependent GABA network.
Correspondingly, in mice expressing human α-synuclein, in which we found that GABA co-release from DA axons is attenuated, we also found that the level of tonic GABA inhibition on DA release was boosted. We note that Aldh1a1 mutations in humans and deletion in mice lead to alcohol-consuming preferences (Kim et al., 2015;Liu et al., 2011;Sherva et al., 2009), and speculate that a modified DA output might plausibly contribute to this behaviour.
The paucity of GABAergic synapses on DA axons (Charara et al., 1999) suggests that GADdependent GABA tone arises from the extrasynaptic ambient tone that can be detected in striatum (Ade et al., 2008;Cepeda et al., 2013;Kirmse et al., 2008Kirmse et al., , 2009Santhakumar et al., 2010). This tone was action potential-independent, i.e. spontaneous (Kaeser and Regehr, 2013), as reported previously for tonic inhibition of SPNs (Wójtowicz et al., 2013). A spontaneous GABAergic regulation of DA release is not surprising when considering that the axonal arbour of a given nigrostriatal DA neuron (in rat) reaches on average 2.7% of the volume of striatum (Matsuda et al., 2009;Oorschot, 1996), and that such volumes contain ~74,000 GABAergic neurons (calculated from 2.8 million striatal neurons per hemisphere (Oorschot, 1996) of which ~98% are GAD-immunoreactive) and also GAD-positive cholinergic interneurons that can co-release GABA (Lozovaya et al., 2018). Even very low rates of spontaneous vesicle release from a small fraction of GAD-utilizing GABAergic neurons might summate sufficiently to provide a tone at GABA receptors on DA axons that limits DA output. The general functions of this spontaneous GABA tone are not well understood, but could differ from functions of action potential-dependent or synaptic events (Farrant and Nusser, 2005), and could include regulation of DA axonal membrane resistance to modify the impact of other inputs or limit the propagation of action potentials through the axonal arbour for a sparser coding.
We found that GAT-1 and GAT-3 both limit the actions of GABA on DA axons in DLS, and thereby indirectly facilitate DA release. This unprecedented role for the GATs in supporting DA output was heterogeneous: GATs limited tonic GABAergic inhibition of DA release in DLS, and also NAcS, but not NAcC, which corresponded with heterogeneity in GAT-1 and GAT-3 expression. Of note, the positive relationship we find between GAT function and DA output is paralleled by, and provides a candidate explanation for, some clinical effects of GAT inhibitors e.g. tiagibine. When used clinically as antiepileptics drugs to increase extracellular GABA levels, these inhibitors can have parkinsonian-like motor side effects (Zaccara et al., 2004).
We did not find evidence for robust localization of GAT-1 or GAT-3 to DA axons in DLS, despite a previous inference that GATs reside on DA axons to support GABA uptake for co-release (Tritsch et al., 2014). This inference was based on mRNA for GAT-1 (and weakly for GAT-3) being present in the somata of DA neurons in substantia nigra, and on the attenuation of GABA co-release from DA axons after pharmacological inhibition of GATs (Tritsch et al., 2014). However, because subsequent work has shown that there is a tonic GABAergic inhibition of DA release mediated by both GABA A and GABA B receptors (Lopes et al., 2018), which we show here is profoundly limited by the GATs, then the observed dependence of GABA co-release on GAT is very likely a result of the GATs limiting tonic inhibition of GABA co-release, rather than GATs necessarily being required for GABA uptake.
We revealed that astrocytes play a critical role in limiting the tonic inhibition of DA release and therefore supporting DA output. We found that both GAT-3 and, to a lesser extent, GAT-1, could be identified on astrocytes, challenging the previous generalization that GAT-1 is exclusively neuronal (Borden, 1996). The role we find for astrocytes in supporting GABA uptake to minimise tonic inhibition of DA release, indicates a previously unappreciated role for astrocytes in determining the dynamic output of DA. This finding significantly revises current understanding of the striatal mechanisms that can dynamically regulate DA transmission. Astrocytic GATs have recently been shown to regulate tonic GABAergic inhibition of striatal SPNs and striatal-dependent behaviours (Yu et al., 2018), and thus, our collective findings point to GATs and astrocytes as powerful regulators of striatal and DA function that warrant further future investigation.

Striatal GAT dysfunction in a mouse model of Parkinson's disease
To probe the significance of the regulation of striatal DA by striatal GATs, we explored GAT function in a mouse model of early parkinsonism. A recent study in external globus pallidus of dopamine-depleted rats found elevated extracellular GABA resulting from downregulation of GAT-3 on astrocytes, mediated through a loss of DA signalling at D 2 DA receptors (Chazalon et al., 2018). Conversely, striatal GAT-3 levels are upregulated in mouse models of Huntington's disease (Wójtowicz et al., 2013;Yu et al., 2018). We explored potential adaptations to GAT function and tonic GABA inhibition of DA release in the striatum of the human α-synuclein-overexpressing mouse model of PD. This model is a highly physiological, slowly progressing mouse model of parkinsonism, that, in capturing a human disease-relevant genetic burden of α-synuclein overexpression, shows early deficits in DA release restricted to dorsal striatum prior to latestage degeneration of DA neurons, disturbed encoding of behaviour of surviving DA neurons and a motor phenotype (Dodson et al., 2016;Janezic et al., 2013). We firstly ascertained the novel finding that DA transmission deficits in this model in early adulthood are accompanied by a corresponding deficit in GABA co-release from DA axons, which suggests that in early parkinsonism at least, malfunction in nigrostriatal DA is accompanied by malfunction in nigrostriatal GABA. Furthermore, we found an augmentation of tonic GABA inhibition of DA release in the DLS (and not NAcC), which was accompanied by downregulated GAT-1 and GAT-3 expression. Whether these adaptations in GAT are consequential to reduced dopamine signalling, as seen in the globus pallidus after massive depletion of dopamine (Chazalon et al., 2018), or due to a potential interaction between α-synuclein and striatal GAT and/or astrocytes is not yet known.
Regardless, this resulting enhanced tonic inhibition will diminish nigrostriatal DA release, compounding any release deficits underpinned by α-synuclein actions e.g. tighter vesicle clustering at DA release sites (Janezic et al., 2013). These changes in GATs and tonic GABA inhibition in early parkinsonism can be considered 'maladaptive'.
In conclusion, the regulation of striatal GABA-DA interactions via striatal GATs and astrocytes represent loci for governing DA output as well as for maladaptive plasticity in early parkinsonism, which could also provide a novel therapeutic avenue for upregulating DA signalling in PD. involving SNCA+ mice, we used age-and sex-matched Snca-null mice (heterozygous for Slc6a3 IRES-Cre ) as littermate controls. All mice were maintained on a C57BL/6 background, group-housed and maintained on a 12-hr light cycle with ad libitum access to food and water. All transgenic mice used in experiments were homozygous for transgenes or mutant alleles.
Slice preparation. Acute brain slices were obtained from 35-to 80-day-old mice using standard techniques.

Fast-scan cyclic voltammetry (FSCV).
Individual slices were hemisected and transferred to a recording chamber and superfused at ~3.0 mL/min with aCSF at 31-33 °C. A carbon fibre microelectrode (CFM; diameter 7-10 μm, tip length 70-120 μm), fabricated in-house, was inserted 100 μm into the tissue and slices were left to equilibrate and the CFM to charge for 30-60 min prior to recordings. All experiments were carried out either in the dorsolateral quarter (DLS) of the CPu or nucleus accumbens (NAc) core (NAcC; within 100 μm of the anterior commissure) or lateral NAc shell (NAcS), one site per slice (see Supplementary Figure S4). Evoked extracellular DA concentration ([DA] o ) was measured using FSCV at CFMs as described previously (Threlfell et al., 2012). In brief, a triangular voltage waveform was scanned across the microelectrode (-700 to +1300 mV and back vs Ag/AgCl reference, scan rate 800 V/s) using a Millar Voltammeter (Julian Millar, Barts and the London School of Medicine and Dentistry), with a sweep frequency of 8 Hz. Electrical or light stimuli were delivered to the striatal slices at 2.5 min intervals, which allow stable release to be sustained at ~90-95% (see Fig. 1B antagonist at 2 subunit-containing nicotinic acetylcholine receptors (nAChRs), to eliminate cholinergic signalling effects on DA release (Exley and Cragg, 2008;Rice and Cragg, 2004;Threlfell et al., 2012). Release was tetrodotoxin-sensitive as shown previously (Threlfell et al., 2012).
In experiments where [DA] o was evoked by electrical stimulation, a local bipolar concentric Pt/Ir electrode (25 μm diameter; FHC Inc.) was placed approximately 100 μm from the CFMs and stimulus pulses (200 μs duration) were given at 0.6 mA (perimaximal in drug-free control conditions). We applied either single pulses (1p) or 2-10 pulses (2p, 4p, 5p, 10p) at 10 -100 Hz. A frequency of 100 Hz is useful as a tool for exposing changes in short-term plasticity in DA release that arise through changes in initial release probability (Jennings et al., 2015;Rice and Cragg, 2004). In experiments where [DA] o was evoked by light stimulation in slices prepared from Slc6a3 IRES-Cre mice expressing ChR2, DA axons in striatum were activated by TTL-driven (Multi Channel Stimulus II, Multi Channel Systems) brief pulses (2 ms) of blue light (470 nm; 5 mWmm -2 ; OptoLED; Cairn Research), which illuminated the field of view (2.2 mm, x10 water-immersion objective). Epifluorescence used to visualize ChR2-eYFP expression was used sparingly to minimize ChR2 activation before recordings Electrophysiology. Individual slices were hemisected and transferred to a recording chamber and superfused at ~3.0 mL/min with aCSF at 31-33 °C. Cells were visualized through a X40 water-immersion objective with differential interference contrast optics. All whole-cell experiments were recorded using borosilicate glass pipettes with resistances in the 3 -5 MΩ range and were pulled on a Flaming-Brown micropipette puller (P-1000, Sutter Instruments). Whole-cell voltage-clamp electrophysiology recordings were made from spiny projection neurons (SPNs; identified by their membrane properties (Gertler et al., 2008;Planert et al., 2013)) in the DLS or NAcC. SPNs were voltage-clamped at -70 mV using a MultiClamp 700B amplifier (Molecular Devices) and with pipettes filled with a CsCl-based internal solution (in mM 120 CsCl, 15 CsMeSO 3 , 8 NaCl, 0.5 EGTA, 10 HEPES, 2 Mg-ATP, 0.3 Na-GTP, 5 QX-314; pH 7.3 adjusted with CsOH; osmolarity ranging from 305 -310 mOsmkg -1 ). The recording perfusate always contained NBQX (5 μM) and APV (50 μM) to block AMPA and NMDA receptor-mediated inward currents. Errors due to the voltage drop across the series resistance (<20 MΩ) were left uncompensated and membrane potentials were corrected for a ~5 mV liquid junction potential. Cells were discarded from analysis if if series resistance varied by more than 15% or increased over 25 MΩ.
To record tonic GABA A currents, SPNs voltage-clamped at -70 mV were recorded in gap-free mode.
Cells were allowed to stabilize for 5-10 min before drug manipulations: GAT inhibitors were bath applied for 20 -25 min; picrotoxin (100 μM) for an additional 3-5 min. Recordings of light-evoked GABA currents in SPNs from ChR2-expressing DA axons in slices from Slc6a3 IRES-Cre mice were taken 10 min after break-in, and at 30 s intervals for a duration of 10 min from SPNs voltage-clamped at -70 mV. Under these conditions, GABA A receptor-mediated currents appear inward as reported previously (Tritsch et al., 2012). TTL-driven (Multi Channel Stimulus II, Multi Channel Systems) brief pulses (2 ms) of blue light (470 nm; 5 mWmm -2 ; OptoLED; Cairn Research) illuminated the full field of view (2.2 mm, X10 water-immersion objective).
High-performance liquid chromatography. DA content in dorsal striatum was measured by HPLC with electrochemical detection as described previously (Janezic et al., 2013). Tissue punches (2 mm in diameter) were taken from dorsal striatum in two brain slices per animal, snap frozen and stored at −80 °C in 200 μL 0.1 M HClO 4 . On the day of analysis, samples were thawed on ice, homogenized, and centrifuged at 15,000 g for 15 min at 4 °C. The supernatant was analysed for DA content. Analytes were separated using a 4.6 × 250 mm Microsorb C18 reverse-phase column (Varian or Agilent) and detected using a Decade II SDS electrochemical detector with a Glassy carbon working electrode (Antec Leyden) set at + 0.7 V with respect to a Ag/AgCl reference electrode. The mobile phase consisted of 13% methanol (vol/vol), 0.12 M NaH 2 PO 4 , 0.5-4.0 mM octenyl succinic anhydride (OSA), and 0.8 mM EDTA (pH 4.4-4.6), and the flow rate was fixed at 1 mL/min.

Western blot.
Mouse brains were extracted and sliced using the procedures outlined above. One 1.2 mm thick coronal slice containing striatum was prepared from each brain and one tissue punch (2 mm in diameter) of dorsal striatum taken per hemisphere. Striatal tissue samples were snap-frozen and stored at at −80 °C. For analysis, striatal tissue was defrosted on ice, homogenized in RIPA Lysis and Extraction Buffer (Sigma) containing 150 mM NaCl, 1.0% IGEPAL, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris, pH 8.0, with Complete-Mini Protease Inhibitor and PhosStop (Roche), using a Tissue Tearor (Biospec Products, Inc), and soluble fraction isolated by microcentrifugation at 15,000 g for 15 min at 4°C. Total protein content was quantified using a BCA Protein Assay Kit (Thermo Scientific) and equal amounts of total protein were loaded onto 4 -15% Tris-Glycine gels (BioRad). Following electrophoresis (200 V for ~45 min), proteins were transferred onto polyvinylidene fluoride membranes (BioRad). Blots were probed overnight at 4C with 1:1,000 rabbit anti-GABA transporter 1 (Synaptic Systems, 274102) or 1:1,000 rabbit anti-GABA transporter 3 (Abcam, AB181783). Blots were incubated with HRP-conjugated secondary anitbodies at 1:3,000 for 1 h at room temperature and bands developed using ECL Prime Western Blotting Detection Reagent (GE Healthcare). Blots were subsequently incubated with 1:20,000 HRP-conjugated β-actin (Abcam, AB49900) for 1 h at room temperature and bands developed as above. Visualization and imaging of blots was performed with a ChemiDoc Imaging System (BioRad) and bands quantified using Image Lab Software (BioRad). Protein concentration for GAT-1 and GAT-3 were normalized to β-actin. interspersed with 1p stimulations, and then averaged and normalized to 1p stimulations at each recording site, as previously (Lopes et al., 2018;Threlfell et al., 2012).
Membrane currents from voltage-clamp electrophysiology experiments were amplified and lowpass filtered at 5 kHz using a MultiClamp 700B amplifier (Molecular Devices), digitized at 10 kHz and acquired using a Digidata 1550A digitizer (Molecular Devices). Peak amplitude, onset latency, peak latency, 10-90% rise time and decay time were measured from an average of 3 replicate traces recorded before and after drug wash on conditions using Clampfit 10.4.1.4 software (Molecular Devices).
For all experiments, data were collected from a minimum of 3 animals. Data were compared for statistical significance using Prism 7 (Graph Pad) with the following statistical tests (as indicated in the text, and two-tailed): un-paired t-tests, paired t-tests, two-way repeated-measures ANOVA followed by Sidak's multiple comparisons, and where the data were not normally distributed, Mann-Whitney U tests, Kruskal    , before and during bath application of (A-B) GAT inhibitor NPA (blue, 1.5 mM, n = 7 cells/5 mice for DLS in A, n = 6 cells/3 mice for NAcC in B), (D) GAT-1 specific inhibitor SKF89976A (orange, 20 M, n = 6 cells/3 mice), (E) the combined application of SKF89976A and GAT-3 specific inhibitor SNAP5114 (green, 50 M, n = 6 cells/4 mice), or (G) NPA in the presence of TTX (1 M) (red, n = 6 cells/3 mice). GAT inhibitors increase the extracellular GABA A -mediated inward current, revealed by a shift in the holding current, and is reversed upon application of GABA A receptor antagonist picrotoxin (PTX, 100 M). Right, mean (± SEM) holding current in pA recorded in SPNs in control conditions, upon addition of GAT inhibitors and then PTX. (C,F,H) Mean (± SEM) tonic GABA A -receptor-mediated currents induced by GAT inhibition recorded from SPNs, calculated by subtracting pre-drug holding current from GAT block-induced holding current. Friedman's ANOVA on Ranks and Student-Newman-Keuls multiple comparisons (A,B,D,E), Mann-Whitney tests (C,H), Kruskal-Wallis test and Dunn's multiple comparisons (F), Wilcoxon signed-rank test (G) *P < 0.05, **P < 0.01., n.s. not significant.

Figure 4. Enrichment of GAT-1 and GAT-3 expression in the DLS versus NAcC.
(A,C) Representative immunofluorescence signals for GAT-1 (cyan, A) and GAT-3 (green, C) using confocal microscopy in coronal sections across the rostral-caudal limits containing striatum prepared from an individual C57BL/6J mouse with heat maps for striatal GAT intensity. Boxes indicate representative locations for GAT intensity measurements in the dorsolateral striatum (DLS) and nucleus accumbens core (NAcC). Scale bars: 1 mm. Note enriched GAT-3 in the medial NAc shell (NAcS) contiguous with the medial septal nucleus and enriched GAT-3 expression in the claustrum. (B,D) Left, Mean (± SEM) GAT-1 (B) and GAT-3 (D) intensity in DLS and NAcC normalized to total striatum and averaged across rostral-caudal sites for each hemisphere (n = 12 hemispheres/6 mice for each GAT-1 and GAT-3). Right, Representative single plane images of GAT-1 (B) and GAT-3 (D) immunofluorescence from DLS and NAcC; imaging parameters were kept constant across regions. Scale bars: 50 m. Mann-Whitney tests. **P < 0.01. (A-B) Striatal immunofluorescence signals for astrocyte marker S100β (magenta) in dorsolateral striatum (DLS, A) and nucleus accumbens core (NAcC, B). Scale bars: 100 µm, for inset: 10 µm. cc: corpus callosum, ac: anterior commissure. (C-D) GAT-3 (green, C) and GAT-1 (cyan, D) are expressed on plasma membranes of striatal S100β-expressing astrocytes imaged in DLS (n = 3 animals). Note that localization of GATs on astrocytes (white arrows) was more prevalent for GAT-3 than GAT-1. Scale bars: 5 m.  Under normal circumstances (left), GAD-synthesized GABA is released from GABAergic striatal neurons can spillover to act at GABA receptors (GABA A R and GABA B R) located presumably on DA axons, inhibiting (dashed red lines) DA and GABA co-release. The level of GABA spillover and tonic inhibition of DA release is determined by the activity of GABA transporters (GATs) located on astrocytes (gray) and neurons, which remove GABA from the extracellular space. In a mouse model of early Parkinsonism (right), striatal GAT expression is downregulated in dorsal striatum, resulting in augmented tonic inhibition of DA release by GABA. Co-release of GABA from DA axons is also reduced.