GABAergic-like dopamine synapses in the brain.

Dopamine synapses play a crucial role in volitional movement and reward-related behaviors, while dysfunction of dopamine synapses causes various psychiatric and neurological disorders. Despite this significance, the true biological nature of dopamine synapses remains poorly understood. Here, we show that dopamine transmission is strongly correlated with GABA co-transmission across the brain and dopamine synapses are structured and function like GABAergic synapses with marked regional heterogeneity. In addition, GABAergic-like dopamine synapses are clustered on the dendrites, and GABA transmission at dopamine synapses has distinct physiological properties. Interestingly, the knockdown of neuroligin-2, a key postsynaptic protein at GABAergic synapses, unexpectedly does not weaken GABA co-transmission but instead facilitates it at dopamine synapses in the striatal neurons. More importantly, the attenuation of GABA co-transmission precedes deficits in dopaminergic transmission in animal models of Parkinson's disease. Our findings reveal the spatial and functional nature of GABAergic-like dopamine synapses in health and disease.

*Correspondence should be addressed to: jikim220@unist.ac.kr 31 the one of TH-positive axons (Fig. 1b, d, f). Notably, although small, regional heterogeneity and 99 differences were found within the striatum in DAT-positive axons (Fig. 1d, f). 100 We next questioned whether neurotransmitters co-transmission at DA synapses would 101 follow the quantitative pattern of axon fibers across the brain. We selectively expressed 102 channelrhodopsin 2 (ChR2) in DA neurons by crossing DAT-IRES-Cre mice (in which Cre 103 recombinase is selectively expressed in DA transporter-expressing neurons) with transgenic mice 104 containing a conditional floxed allele of ChR2 in the Rosa26 locus (Ai32 mice) 8 . We first 105 performed fast-scanning cyclic voltammetry (FSCV) to measure and compare brain-wide DA 106 transmission among each DA projection region (Fig. 1g). As documented before 11,12 , light-evoked 107 (450 nm) DA transmission was the highest in the dorsal striatum and ventral domains of striatum 108 exhibited relatively smaller DA release than dorsal striatum. Photo-stimulation of DA axons 109 elicited very low DA release in the cortex as well as extrastriatal basal ganglia nuclei ( Fig. 1h-j). 110 We then sought to determine the regional heterogeneity in the co-transmission of GABA  Fig. 1a, b). Unlike GABA co-120 transmission, the highest responses by glutamatergic transmission at DA synapses came from the 121 5 nucleus accumbens (NAc) shell in the ventral striatum Extended Data Fig. 1c,d). 122 Glutamatergic oEPSCs in other striatal domains were much smaller than that of NAc shell and 123 there were almost no EPSC responses from the recorded neurons in the cortex and extrastriatal 124 basal ganglia nuclei. Of particular note, when this regional transmission of DA, GABA, and 125 glutamate at DA synapses was compared with each other, we found a strong correlation between 126 DA and GABA release, whereas no pronounced correlations existed between glutamate and other 127 neurotransmitters (Fig. 1q). In addition, when the area of DA axon fibers and neurotransmitters 128 co-transmission were compared throughout DA target areas, significant correlations were found 129 between DA axonal areas (TH-and DAT-positive) and GABA (or DA) transmission, while there 130 was no meaningful correlation between DA fibers and glutamate transmission (Fig. 1r, Extended 131 Data Fig. 2). Interestingly, the correlation between DAT-positive axonal area and GABA (or DA) 132 transmission was stronger than the one by TH-positive axons both in the whole brain and in the 133 striatal regions. Together, it is very likely that DA transmission reliably accompanies GABA 134 transmission at DA synapses in the brain. Furthermore, regional heterogeneity of DAT-positive 135 axons in the brain, although small, mirrors regional differences in DA and GABA transmissions 136 better than TH-positive axonal areas.

138
Spatial distribution and characteristics of GABAergic-like DA synapses in the brain 139 DA axon terminals make synaptic contacts with neurons in numerous brain areas, although the 140 majority of these synapses reside within the striatum. These multiple brain regions are implicated 141 in diverse neural functions, which are critically mediated by synaptic transmission via DA 142 synapses. It is important to note that not all TH-positive bouton-like structures are able to release 143 DA 6 and postsynaptic DA receptors are typically at extrasynaptic sites 4,5,13 . Thus, unlike 144 glutamatergic and GABAergic synapses, this mismatch between presynaptic and postsynaptic sites, 145 and our lack of knowledge about presynaptic and postsynaptic molecular markers defining 146 functional DA synapses, have hampered so far the comprehensive identification of spatial and 147 functional features of DA synapses throughout the brain. Recently, neuroligin-2 (NL2), a common 148 postsynaptic cell adhesion molecule (CAM) that has been thought to be exclusively localized to 149 inhibitory synapses 14 , was found in the postsynaptic membrane specializations at striatal DA 150 synapses 5 . Despite this finding, however, our understanding of molecular components defining DA 151 6 synapses in the brain is still far from complete. 152 We investigated brain-wide distribution and spatial characteristics of DA synapses by 153 labeling molecular markers for both presynaptic and postsynaptic sites. We utilized NL2 as a 154 postsynaptic marker for DA synapse. We also considered the co-localization of TH and vesicular 155 monoamine transporter 2 (VMAT2) immunofluorescences as a putative marker for dopaminergic 156 presynaptic boutons that are able to co-transmit DA and GABA 7 . In multi-color immunostaining 157 and enhanced confocal microscopy by airyscan, TH-and VMAT2-positive presynaptic boutons, 158 unlike TH-positive DA axons, showed marked regional heterogeneity in their distribution across 159 the nigrostriatal and mesocorticolimbic pathways Extended Data Fig. 3a,. Dorsal 160 striatum largely contained more TH-and VMAT2-positive boutons than ventral striatum. As 161 expected from our results ( Fig. 1b-f), the number of TH-and VMAT2-positive boutons in cortical 162 areas and extrastriatal basal ganglia was much smaller than that of striatal regions (Fig. 2b,163 Extended Data Fig. 3c). Notably, the number of NL2-positive postsynaptic sites also exhibited 164 regional differences across the brain, although the relative difference among brain areas was 165 smaller than that of dopaminergic presynaptic boutons (Fig. 2c, Extended Data Fig. 3d). We then 166 regarded the co-localization of a TH-and VMAT2-positive presynaptic bouton and a NL2-positive 167 postsynaptic site as a potential GABAergic-like DA synapse. When this triple co-localization of 168 TH, VMAT2, and NL2 was compared, the regional heterogeneity in dopaminergic presynaptic

172
It is likely that TH-and VMAT2-positive presynaptic boutons may not represent 173 exclusively functional DA terminals capable of releasing neurotransmitters as VMAT2 can also 174 exist outside synaptic boutons. To confirm this, we examined the spatial properties of GABAergic-175 like DA synapses by labeling presynaptic protein Bassoon as a functional presynaptic marker at 176 DA synapses 6 . When VMAT2 was replaced by Bassoon, relative regional differences in the 177 numbers of both presynaptic boutons and GABAergic-like DA synapses were largely maintained 178 ( Fig. 2g-j), while the estimated numbers of dopaminergic presynaptic boutons and GABAergic-179 like DA synapses were considerably reduced (Extended Data Fig. 3b, f-h). In addition, when 180 Bassoon was used as a presynaptic marker, the difference in the number of GABAergic-like DA 181 7 synapses among brain regions, especially between dorsal and ventral striatum, became more 182 pronounced (Fig. 2j, Extended Data Fig. 3h). We further validated these findings and analyzed 183 GABAergic-like DA synapses by using Gephyrin as a postsynaptic marker for DA synapses.

184
Gephyrin is an assembly protein which anchors inhibitory neurotransmitter receptors to the 185 postsynaptic membrane and has been determined as a major postsynaptic protein of GABAergic 186 synapses 15,16 . Spatial distribution and regional disparity of GABAergic-like DA synapses 187 throughout the brain, revealed by Gephyrin, were also comparable to the pattern identified by NL2, 188 suggesting that regional properties of GABAergic-like DA synapses exist independent of the 189 molecular markers used. Moreover, compared to the co-localization of VMAT2 and Gephyrin, the

202
We then examined the correlation and found that the correlation between GABAergic-like 203 DA synapses and dopaminergic presynaptic boutons was remarkably higher than the one between 204 GABAergic-like DA synapses and GABAergic postsynaptic sites throughout the brain regions ( Fig.   205 2m, n, Extended Data Fig. 6). Of particular note, Bassoon-NL2 combination exhibited the 206 strongest correlation between GABAergic-like DA synapses and dopaminergic presynaptic 207 boutons and the weakest correlation between GABAergic-like DA synapses and GABAergic 208 postsynaptic sites when the correlations were separately calculated and compared from each brain 209 area (Fig. 2o). More importantly, the correlation between GABAergic-like DA synapses and 210 GABA (or DA) transmission was the highest and highly significant in the whole brain as well as  these results indicate that unlike just the area of DA axons, GABAergic-like DA synapses are 223 distributed throughout the brain with significant regional and spatial heterogeneity, irrespective of 224 synaptic markers used for analysis. Furthermore, GABAergic-like DA synapses determined by 225 Bassoon-NL2 combination do correlate best with GABA-DA transmission at DA synapses. throughout the brain, we performed multi-color immunostaining and enhanced confocal imaging 231 via airyscan. We used triple co-localization of TH, Bassoon, and NL2 as a marker for GABAergic-232 like DA synapse. To better capture region-specific spatial clustering of GABAergic-like DA 233 synapses, image area per each brain region was expanded (Fig. 3a). For spatial clustering analysis, 234 we utilized Ripley's H function that is particularly sensitive to point pattern clustering 19,20 . Owing NL2 as a conventional GABAergic synapse (Fig. 4a). We found that the density of GABAergic-279 like DA synapses on the single dendrite of SPNs was markedly higher than that of conventional 280 GABAergic synapses made by local iSPNs. Furthermore, this difference in the density of synapses 281 was maintained irrespective of the specific location on the dendrite (distal vs. proximal) (Fig. 4b). proximal dendrites also exhibited more than 50% rejection rate in DCLF test, this seemed to be 291 caused by the fact that the clustering patterns of GABAergic synapses are more variable than that 292 of GABAergic-like DA synapses on the entire and proximal dendrites of SPNs.

293
Meanwhile, the functional properties of synaptic transmission by DA-containing synaptic 294 vesicles have never been investigated in detail by whole-cell patch clamp recording, which have 295 been mostly studied either by extracellular DA recording through FSCV or by overexpression of 296 G protein-coupled inwardly-rectifying potassium channel (GIRK) to SPNs. Taking advantage of 297 GABA co-transmission at DA synapses, we first examined whether GABA co-transmission at DA 298 synapses and GABA transmission at conventional GABAergic synapses rely on distinct types of 299 voltage-dependent Ca 2+ channels for vesicular release. We recorded GABA co-transmission at DA 300 synapses by stimulating ChR2-expressing DA axons from DAT-Cre;Ai32 mice and evaluated the 301 11 effects of 3 different types of Ca 2+ channels blockers on the amplitude of oIPSC in SPNs. We found 302 that GABA release at DA terminals is dependent on N-type as well as P/Q type Ca 2+ channels in 303 the dorsal striatum, whereas T-type Ca 2+ channels were not implicated in vesicular release (Fig. 4f,304 g, Extended Data Fig. 8b). On the other hand, conventional GABAergic transmission that were 305 examined at GABAergic synapses between iSPN and dSPN by using A2A-Cre;Ai32 mice, was 306 mediated by not only N-and P/Q-type Ca 2+ channels but also T-type channel (Fig. 4h,  both in DAT-Cre;Ai32 and A2A-Cre;Ai32 mice, while GABA co-transmission at DA synapses was 314 less prone to evoke asynchronous release than conventional GABAergic transmission (Fig. 4j, k). 315 Furthermore, the quantal amplitude of GABA co-transmission at DA synapses was markedly 316 smaller than that of conventional GABA transmission between the SPNs (Fig. 4k, l). To search for 317 the presynaptic molecular mechanisms behind the differential susceptibility to asynchronous 318 release by Sr 2+ , we checked the individual isoforms of synaptotagmin enriched at DA and 319 GABAergic terminals, respectively (Fig. 4m, n). We found that DA terminals were preferentially 320 co-localized with synaptotagmin 1, followed by synaptotagmin 7 and 5/9 in the dorsal striatum. In 321 addition, the relative expression levels of each synaptotagmin isoform at conventional GABAergic 322 terminals were similar to DA terminals ( Fig. 4m-o). However, the co-localization of synaptotagmin 323 1 that is critical for synchronous vesicular release at various synaptic terminals 24-26 , was 324 considerably higher at DA terminals than at GABAergic terminals (Fig. 4p), possibly accounting 325 for the observed differential susceptibility to asynchronous release by Sr 2+ at DA and GABAergic 326 terminals. of DAT-Cre mice (Fig. 5a). We first checked the degeneration of GABAergic-like DA synapses in 337 the dorsal striatum by simultaneously labeling TH, Bassoon, and NL2 (Fig. 5b). We found that the 338 number of GABAergic-like DA synapses in the ipsilateral striatum was largely intact one day after 339 6-OHDA injection (Fig. 5c). However, when compared with the contralateral striatum, 340 GABAergic-like DA synapses have begun to decrease substantially 3 days after 6-OHDA injection.

341
The number of GABAergic-like DA synapses was significantly reduced 7 days after 6-OHDA 342 injection (Fig. 5c). In addition, the spatial distribution of GABAergic-like DA synapses became 343 sparser as 6-OHDA-induced Parkinsonism progressed (Fig. 5d). 344 We then examined the temporal alteration of DA transmission 1, 3, and 7 days after 6-345 OHDA injection. To identify DA synapse-specific physiological changes, we unilaterally injected 346 6-OHDA into the MFB of DAT-Cre;Ai32 mice (Fig. 5e). As with the number of GABAergic-like 347 DA synapses over time, DA transmission in the ipsilateral striatum was not affected one day after 348 6-OHDA injection, while the release of DA was dramatically reduced 3 and 7 days after 6-OHDA 349 injection ( Fig. 5f, g). Most interestingly, however, GABA transmission at DA synapses in the 350 ipsilateral striatum begun to significantly diminish even one day after 6-OHDA injection (Fig. 5h,   351 i, Extended Data Fig. 9a). Like DA transmission, GABA transmission measured 3 and 7 days after 352 6-OHDA injection was also considerably weakened when compared with the contralateral striatum.

353
In case of glutamate transmission at DA synapses, no attenuation of oEPSC was observed one day 354 after injection at the ipsilateral side of striatum. Since the magnitude of glutamate co-transmission 355 at DA synapses was relatively small when compared with DA or GABA co-transmission, 356 glutamate transmission measured by oEPSC from SPNs was almost abolished 3 and 7 days after 357 6-OHDA injection (Fig. 5j, k, Extended Data Fig. 9b). These results strongly suggest that although 358 GABA transmission is highly correlated with DA transmission at DA synapses in the normal brain, 359 DA and GABA transmission may change over the course of Parkinsonism with differential 360 temporal susceptibility.  Unlike DA axons immunoreactive for TH and DAT, these GABAergic-like DA synapses 386 exhibit marked regional and spatial heterogeneity in the brain, especially within the striatum.

387
GABAergic-like DA synapses also show region-specific clustering in the striatal subregions.

388
GABAergic-like DA synapses are more dispersed in DLS and NAc core, whereas DMS, NAc shell, 389 and OT seem to exhibit more clustered pattern of theses synapses. This spatial heterogeneity is 390 well suited to provide regionally distinct functions of DA synapses throughout the brain.

391
Importantly, the density of GABAergic-like DA synapses is the highest in the DLS within the 392 striatum, while OT and DMS display a relatively higher degree of clustering of these synapses than 393 DLS. Although the region-specific structure of neural circuits may partly underlie the high degree   whisker plot or mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. washing with PBST and PBS, brain slices were mounted to slides using mounting medium 697 (P36934 or P36935, Invitrogen).

698
For quantitative analysis, images were captured by FV1000 confocal laser scanning 699 microscope (Olympus) using a 60x/1.35 NA oil immersion objective (image size: 512 x 512 pixels),