Astrocytes mediate long-lasting synaptic regulation of ventral tegmental area dopamine neurons

The plasticity of glutamatergic transmission in the ventral tegmental area (VTA) represents a fundamental mechanism in the modulation of dopamine neuron burst firing and phasic dopamine release at target regions. These processes encode basic behavioral responses, including locomotor activity, learning and motivated behaviors. Here we describe a hitherto unidentified mechanism of long-term synaptic plasticity in mouse VTA. We found that the burst firing in individual dopamine neurons induces a long-lasting potentiation of excitatory synapses on adjacent dopamine neurons that crucially depends on Ca 2+ elevations in astrocytes, mediated by endocannabinoid CB1 and dopamine D2 receptors co-localized at the same astrocytic process, and activation of pre-synaptic metabotropic glutamate receptors. Consistent with these findings, selective in vivo activation of astrocytes increases the burst firing of dopamine neurons in the VTA and induces locomotor hyperactivity. Astrocytes play, therefore, a key role in the modulation of VTA dopamine neuron functional activity. Dopamine (DA) neurons of the VTA regulate a wide array of physiological functions, including locomotion, attention, motivation and reward-based learning 1–3 . A fundamental step in these

The plasticity of glutamatergic transmission in the ventral tegmental area (VTA) represents a fundamental mechanism in the modulation of dopamine neuron burst firing and phasic dopamine release at target regions. These processes encode basic behavioral responses, including locomotor activity, learning and motivated behaviors. Here we describe a hitherto unidentified mechanism of long-term synaptic plasticity in mouse VTA. We found that the burst firing in individual dopamine neurons induces a long-lasting potentiation of excitatory synapses on adjacent dopamine neurons that crucially depends on Ca 2+ elevations in astrocytes, mediated by endocannabinoid CB1 and dopamine D2 receptors co-localized at the same astrocytic process, and activation of pre-synaptic metabotropic glutamate receptors. Consistent with these findings, selective in vivo activation of astrocytes increases the burst firing of dopamine neurons in the VTA and induces locomotor hyperactivity. Astrocytes play, therefore, a key role in the modulation of VTA dopamine neuron functional activity. Dopamine (DA) neurons of the VTA regulate a wide array of physiological functions, including locomotion, attention, motivation and reward-based learning [1][2][3] . A fundamental step in these DA-dependent functions is the transition of spiking activity in VTA DA neurons from tonic, low-frequency firing at rest to high-frequency bursts that modulate the action of DA by determining the synaptic phasic release of DA at VTA target areas, such as nucleus accumbens (NAc), medial prefrontal cortex, hippocampus and amygdala [4][5][6][7] . This transition to bursting activity of DA neurons is under crucial control of glutamatergic afferent inputs to the VTA originating from various brain regions 4,7,8 . Importantly, the enduring changes in the strength of these glutamatergic synapses exert profound effects on DA neurons, regulating their Article https://doi.org/10.1038/s41593-022-01193-4 conditions (bsl), 1.045 ± 0.08; 30 minutes after tonic firing, 1.056 ± 0.1; P = 0.827, paired t-test; Extended Data Fig. 1g). The long-lasting potentiation is not observed in age-matched male mice (Fig. 1b,d), in which EPSC amplitude is only transiently increased 3 minutes after bursts (Fig. 1b). We did not further investigate this short-term potentiation, and we focused the present study on the bLTP generation mechanism.

Generation of bLTP requires astrocytic Ca 2+ elevations
To understand whether astrocytes are involved in bLTP generation, we performed experiments in P14-17 type-2 inositol 1,4,5-trisphosphate receptor knockout (IP 3 R2 −/− ) female mice in which G-protein-coupled mediated astrocyte Ca 2+ elevations are largely impaired 19,26,27 . In contrast to the bLTP observed in wild-type (WT) C57BL/6J mice, in VTA slices from IP 3 R2 −/− female mice, DA neuron bursts evoke only a transient potentiation of synaptic transmission (Fig. 1d). Additional experiments in IP 3 R2 −/− and IP 3 R2 +/+ littermates confirm that DA neuron bursts in IP 3 R2 −/− littermates fail to evoke bLTP, whereas IP 3 R2 +/+ littermates show a bLTP similar to that of WT mice (Extended Data Fig. 2). Dialysis of the Ca 2+ chelator BAPTA in the astrocyte syncytium, which blocks Ca 2+ signaling in astrocytes 28 , also prevents bLTP (Fig. 1d). These data suggest that bLTP induction depends on IP 3 R2-mediated astrocytic Ca 2+ elevations induced by signals generated by DA neurons. This hypothesis was directly tested in VTA slices from WT and IP 3 R2 −/− mice loaded with the Ca 2+ fluorescent indicator Fluo-4 and the specific astrocytic marker SR101. To monitor Ca 2+ signals from astrocytes in proximity of soma and dendrites, through a patch pipette we filled DA neurons with the fluorescence tracer neurobiotin (Fig. 1e). We observed that DA neuron bursts evoke in astrocytes of female, but not male, mice Ca 2+ elevations that last for at least 25 minutes (Fig. 1f-h). Furthermore, DA neuron bursts fail to evoke astrocyte Ca 2+ elevations in IP 3 R2 −/− female mice and also in WT female mice after loading the astrocyte syncytium with BAPTA ( Fig. 1h and Extended Data Fig. 2c). Overall, these data suggest that astrocyte IP 3 R2-mediated Ca 2+ elevations are required for bLTP generation.

Generation of bLTP requires eCB, DA and mGluR1 signaling
To gain further insights into the molecular mechanism of bLTP generation, we investigated whether CB1 and/or DA receptors (Rs), activated by eCBs and/or DA locally released by VTA DA neurons, are involved. We found that applications of either the CB1R antagonist AM251 or the D2-type receptor antagonist eticlopride prevent bLTP, whereas the D1R antagonist SCH-23390 hydrochloride is ineffective ( Fig. 2a; compared to controls, AM251 P = 0.046, eticlopride P = 0.045 and SCH-23390 P = 0.916, Mann-Whitney rank-sum test). We also evaluated the role of the N-methyl-D-aspartate receptor (NMDAR), which mediates synaptic plasticity in different brain regions 29 , including the VTA 10 . We found that bLTP is unaffected by the NMDAR antagonist D-AP5 ( Fig. 2a; compared to controls, P = 0.584, unpaired t-test), suggesting that NMDAR is not involved. We then observed that bLTP is abolished by the type-1 metabotropic glutamate receptor (mGluR1) antagonist LY-367385 ( Fig. 2a; compared to controls, P = 0.015, unpaired t-test), indicating that, as previously reported in hippocampal 18,19 and striatal circuitries 20 , the astrocyte action is mediated by mGluR1 receptor activation.
We then asked whether CB1 and D2R activation, which is required for bLTP generation, is also required for DA neuron burst-induced astrocytic Ca 2+ elevations. We found that the astrocyte response is abolished in the presence of either AM251 or eticlopride ( Fig. 2b; compared to controls, AM251 P = 0.026, unpaired t-test; eticlopride P = 0.002, Mann-Whitney rank-sum test). In contrast, the astrocyte Ca 2+ response is unaffected in the presence of the mGluR1 antagonist LY-367385 ( Fig. 2b; compared to controls, P = 0.778, unpaired t-test), suggesting that mGluR1 activation plays its crucial role in bLTP generation downstream astrocytic Ca 2+ signals.
In support of the role of astrocytic CB1 and D2Rs in bLTP, pre-embedding electron microscopy (EM) experiments showed that, besides neurons (Extended Data Fig. 3a and Supplementary Table 1), burst firing mode and DA release at target regions 1,9 . The plasticity of these glutamatergic synapses represents, therefore, a key mechanism in the modulation of DA transmission and DA-dependent behaviors. Although extensive studies highlighted the role of neuronal signals in the synaptic plasticity of VTA circuits 1,10 , the role of astrocytes has been insufficiently investigated.
A recent study reported that optogenetic stimulation of channelrhodopsin-expressing VTA astrocytes alters glutamate transport, favoring DA neuron inhibition and avoidance behavior 11 . However, this type of stimulation depolarizes astrocytes, leading to substantial increase in extracellular K + and increase in neuronal excitation 12 . Whether astrocytes are functionally recruited to the VTA circuitry by neuronal signals and influence the plasticity of glutamatergic synaptic transmission to VTA DA neurons remains totally unexplored.
Astrocytes are active components of brain circuits. Besides their support and metabolic functions, they respond with Ca 2+ elevations to neurotransmitters and, in turn, release gliotransmitters that regulate synaptic transmission and plasticity [13][14][15] . Astrocytes are similarly activated by local signals, such as endocannabinoids (eCBs), released by neurons at somatodendritic levels. In various brain areas, including the VTA, eCBs act as retrograde signals that induce neurotransmitter release depression upon pre-synaptic type-1 cannabinoid receptor (CB1R) activation 16,17 . Studies in hippocampus and dorsal striatum showed that eCBs also target astrocytic CB1Rs, evoking Ca 2+ elevations and glutamate release that potentiates distant excitatory synapses [18][19][20] . Whether this lateral potentiation of synaptic transmission is also operative in the VTA is unknown.
Using ex vivo and in vivo approaches, we investigated whether eCBs released by bursting discharges of VTA DA neurons 21 induce a potentiation of glutamatergic transmission to nearby DA neurons and whether this action is mediated by astrocytes. Because VTA DA neurons, beside eCBs, release DA at somatodendritic levels 22 , we investigated whether DA is also involved in DA neuron-to-astrocyte signaling. Finally, we evaluated the functional consequences of a specific activation of astrocytes in vivo at the level of both VTA DA neuron firing and locomotor activity. Our results unveil a reciprocal functional signaling between DA neurons and astrocytes in VTA circuits.

Lateral LTP of EPSCs in VTA DA neurons of young female mice
We investigated whether the bursting activity in individual DA neurons evokes lateral potentiation of glutamatergic transmission 18 . In VTA slices of postnatal day (P) 14-17 C57BL/6J female mice, we recorded from pairs of neurons showing the typical features of DA neurons ( Fig. 1a and Extended Data Fig. 1a-d). In one neuron of the pair, we monitored excitatory post-synaptic currents (EPSCs) evoked by low-frequency stimulation of rostral glutamatergic afferents. To the second neuron, located 70-120 µm apart, through intracellular current pulses we imposed the burst firing mode that characterizes in vivo DA neuron activity (bursts of five action potentials at 20 Hz, 2-Hz inter-burst frequency and 5-minute duration; Extended Data Fig. 1e) [23][24][25] . After bursting activity, EPSC amplitude from the first DA neuron is significantly increased, and this potentiation is maintained for at least 45 minutes (Fig. 1b,d). We define this novel form of lateral synaptic plasticity as burst-induced long-term potentiation (bLTP). Evaluation of the paired-pulse ratio (PPR) revealed a significant PPR reduction 45 minutes after DA neuron bursts, suggesting a pre-synaptic mechanism in bLTP generation (Fig. 1c). This pre-synaptic locus is confirmed by the relative changes of the coefficient of variation (CV) for EPSCs after DA neuron bursts (Fig. 1c). Notably, the induction of a tonic-like discharge that mimics basal DA neuron activity (2-Hz action potential frequency for 5 minutes; Extended Data Fig. 1e) fails to modify evoked EPSCs in adjacent DA neurons at any timepoint tested (Extended Data Fig. 1f Tables 1 and 2). According to our post-embedding quantitative EM analysis of CB1/D2R immunogold double-labeled astrocytic processes, CB1 and D2Rs co-localize at peri-synaptic processes (Fig. 2d,e; mean distance: 523.57 ± 38.37 nm), indicating that the same astrocyte can sense both eCBs and DA. Notably, we found that the mGluR1β isoform is expressed at axon terminals making asymmetric synaptic contacts (Fig. 2c, Extended Data Fig. 3c and Supplementary Tables 2 and 3), consistent with a pre-synaptic mechanism of bLTP, as suggested by PPR reduction and CV analysis. Notwithstanding our finding that D1-type receptors are not involved in bLTP, interestingly, we found that functional D1Rs are also expressed in VTA astrocytes (Extended Data Fig. 3e-h, Supplementary Tables 4  and 5 and Supplementary Note 1).

D2R, CB1R and mGluR1 expression in female and male young mice
The bLTP is not observed in young male mice, in which VTA astrocytes fail to respond to DA neuron bursts. This lack of astrocyte Ca 2+ responses may be due to absence or low levels of CB1 and/or D2Rs. Our EM experiments show that astrocytes from male mice express CB1, D2, D3 and D4Rs ( Fig. 3a and Extended Data Fig. 3a,d), but the expression of CB1 and D2Rs is higher in astrocytic processes of female mice than of male mice (P < 0.0001; Fig. 3a and Supplementary  µM, n = 7 from four mice, P = 0.853; eticlopride (D2-type R) 1 µM, n = 10 from eight mice, P = 0.495; SCH-23390 (D1-type R) 10 µM, n = 10 from eight mice, P = 0.026; D-AP5 (NMDAR) 50 µM, n = 11 from nine mice, P = 0.038; and LY-367385 (mGluR1) 100 µM, n = 12 from nine mice, P = 0.249). Two-tailed one-sample t-test. b, Time course and bar chart of astrocytic Ca 2+ spike probability per minute in the presence of antagonists that impair bLTP generation (AM251, n = 6 from three mice, P = 0.818; eticlopride, n = 6 from three mice, P = 0.351; and LY-367385, n = 6 from four mice, P = 0.003). Two-tailed paired t-test. c, Pre-embedding EM images from lateral VTA of a young female mouse of CB1, D2 and mGluR1β receptors. Green arrows, immunopositive products in AsP (AsP + ) and AxT (AxT + ) forming asymmetric synaptic contacts (arrowheads) with a dendrite (Den). Scale bar, 300 nm. d, Post-embedding EM images of CB1/D2R immunogold double-labeled astrocytic processes (AsP) in lateral VTA (CB1R, 18-nm gold particles; D2R, 12-nm gold particles). Left panel: a double-labeled AsP expressing CB1 and D2R (arrows) in close apposition to an asymmetric synapse (AxT, axon terminal; Den, dendrite; PSD, post-synaptic density). Right panel: an edge-to-edge separation distance between these receptors ≤ 50 nm (arrowhead). Scale bar, 300 nm. e, Left panel: CB1 and D2R immunogold densities at the membrane of astrocytic processes (AsP, n = 138 from four P16 females; D2R, 32.76 ± 2.14 (D2), CB1R, 26.04 ± 1.53 (CB1) gold particles per µm 2 ) and at neuronal nuclei (b, background values, n = 20; 0.86 ± 0.08 and 0.46 ± 0.05 for 12-nm and 18-nm gold particles, respectively; P < 0.0001, two-tailed Mann-Whitney test). Data are presented as a box and whisker plot. Each box is defined by the 25th and 75th percentiles; the central line indicates the median; and the dot indicates the mean value. The whiskers represent the minimum and maximum values in 1.5× interquartile range. Right panel: distribution of edge-to-edge interdistances (bin, 50 nm) between D2 and CB1R immunogold couples. Except for e, data are represented as mean ± s.e.m. Article https://doi.org/10.1038/s41593-022-01193-4 to DA neurons in male mice is likely due to reduced CB1R expression and different levels of D2/D3Rs in the astrocytic processes. Whether the low pre-synaptic mGluR1β expression in young male mice contributes to the lack of bLTP is also a plausible hypothesis.

Induction of bLTP in young male mice by astrocyte activation
If the absence of bLTP in young male mice is due, at least in part, to a lack of astrocyte Ca 2+ responses to DA neuron bursts, we expect bLTP to be observed after coupling DA neuron bursts with astrocyte Ca 2+ elevations. As a specific astrocyte stimulus, we used chemogenetic activation of Gq-protein-coupled designer receptor exclusively activated by designer drugs (DREADDs, hM3D(Gq)) selectively expressed in astrocytes (Fig. 3b,c and Extended Data Fig. 4). We found that bath perfusion with the hM3D(Gq) agonist clozapine N-oxide (CNO) evokes transient Ca 2+ elevations in astrocytes expressing hM3D(Gq) and the genetically encoded Ca 2+ indicator GCaMP6f (Fig. 3d). In agreement with this transient astrocyte response, parallel experiments performed in the presence of CB1 and D2R antagonists, AM251 and eticlopride, revealed that CNO evokes in male mice expressing hM3Dq in astrocytes, but not in non-injected controls, a short-lasting potentiation of excitatory transmission (Fig. 3e), which becomes full bLTP after coupling CNO with DA neuron bursts (Fig. 3f). Burst firing in DA neurons is, therefore, necessary for bLTP generation, and mGluR1β expression level may be sufficient to mediate bLTP in young male mice. Given that these experiments were performed in the presence of CB1 and D2R antagonists, these results further support that bLTP generation depends on astrocytic, and not neuronal, CB1 and D2Rs.
Previous studies reported that nitric oxide (NO) contributes to long-term synaptic plasticity in different brain circuits 19,30 . The release of NO by burst firing DA neurons may also contribute to bLTP. In female mice, we found that burst firing in DA neurons patched with an NO synthase inhibitor (L-NAME)-containing pipette does not evoke bLTP AxT. AsP − , AsP without immunoreactivity. Scale bar, 300 nm. Bottom panel: quantification and comparison (two-sided contingency Fisher's test) of CB1 (P < 0.0001, n = 479 and 410 total AsP in female and male mice, respectively), D2 (P < 0.0001, n = 554 and 588 total AsP in female and male mice, respectively) and mGlu1βR (P = 0.008, n = 284 and 273 total AxT in female and male mice, respectively) expression in female and male mice. b, Schematic of the AAV-9/2-hGFAP-hM3D(Gq)_mCherry-WPRE-hGHp(A) injection in the VTA of a neonatal male mouse and fluorescence image of a brain slice 2 weeks after injection (yellow, mCherry-hM3D expression). c, Confocal images of the VTA from a mouse injected with AAV-9/2-hGFAP-hM3D(Gq)_mCherry-WPRE-hGHp(A), showing the fluorescence of mCherry-hM3D (red), the nuclear Top-Ro3 (blue) and the specific green fluorescence for either neurons (α-NeuN) or astrocytes (α-S100β). Scale bars, 50 µm. d, GCaMP6f fluorescence images of astrocytes at basal conditions and after CNO (10 µM) perfusion. Scale bar, 50 µm. Lower panels: time course of Ca 2+ elevations evoked by CNO in these astrocytes (left, scale bars, 100%, 30 seconds) and mean change of total Ca 2+ levels in slices (n = 9 from seven mice, mean ± s.e.m.) expressing GCaMP6f and hM3D in astrocytes in response to CNO (right, scale bars, 2%, 30 seconds). e, Upper panel: schematic of the experimental design. Lower panel: CNO-induced transient (over the first 9 minutes) increase in EPSC amplitude of DA neurons in male mice expressing hM3D in astrocytes (n = 9 from six mice, P = 0.007, two-tailed one-sample t-test) but not in non-injected mice (n = 8 from five mice, P = 0.945, two-tailed one-sample Wilcoxon signedrank test). f, Upper panel: schematic of the experimental design. Lower panel: burst firing coupled with CNO application evokes bLTP (30 minutes) in male mice expressing hM3D in astrocytes (n = 9 from eight mice, P = 0.003, two-tailed one-sample t-test) but not in non-injected mice (n = 6 from five mice, P = 0.438, two-tailed one-sample Wilcoxon signed-rank test). Experiments in e and f were performed in the presence of AM251 and eticlopride. Data are represented as mean ± s.e.m. mt, medial terminal.  Fig. 5e; see also Supplementary Note 2 for comments on these conflicting results). Further experiments are, therefore, necessary to clarify the role of NO in bLTP.

Female and male adult mice show astrocyte-induced bLTP
We next investigated whether astrocyte-mediated bLTP observed in young mice is also present in young adulthood. We found that, in VTA slices from adolescent/adult mice (P30-70; to simplify, hereafter termed adult mice), DA neuron bursts evoke in adjacent DA neurons a bLTP that is maintained for at least 30 minutes after bursts, and, in contrast to data obtained from young mice, it is surprisingly expressed not only in female mice but also in male mice (Fig. 4a). The presence of bLTP in adult male mice could be due to developmentally regulated expression of CB1, D2 and/or mGluR1β receptors. Quantitative analysis of pre-embedded materials from adult male mice shows that the levels of CB1Rs at astrocytic processes and mGluR1β at excitatory terminals are, indeed, higher in adult male mice than in young male mice, whereas D2R levels are similar ( Fig. 4b and Supplementary Table 6). In agreement with the presence of bLTP, our results reveal similar mGluR1β, CB1 and D2 receptor levels in female and male adult mice (Fig. 4b, Extended Data Fig. 6a and Supplementary Table 6). Together with data presented in Fig. 3a,f, these results suggest that the absence of bLTP in young male mice (P14-17) is due to the low expression of astrocytic CB1Rs at this developmental stage. The mechanism of bLTP in adult mice is similar to that in young female mice, because bLTP is abolished by specific D2, CB1 or mGluR1 receptor antagonists (Extended Data Fig. 6b). To further confirm that bLTP depends on astrocytic, and not neuronal, D2 and CB1Rs, we injected the AAV9/2-hGFAP-mCherry_iCre-WPRE-hGHp into the VTA of male mice carrying a 'floxed' version of either the Drd2 or the Cnr1 genes, to express the Cre recombinase in VTA astrocytes (Fig. 4c). Immunohistochemical experiments showed that the great majority of mCherry-Cre-immunopositive cells are also GFAP + and only a very few mCherry-Cre-immunopositive cells are NeuN + ( Fig. 4d and Extended Data Fig. 7a,b). We found that bLTP is abolished when the Cre recombinase is expressed in astrocytes containing the Drd2 or Cnr1 floxed gene but not when it is expressed in astrocytes of WT mice ( Fig. 4e and Discussion), validating the central role of astrocytic D2 and CB1Rs in bLTP generation. Finally, as in young female mice, after including the NO synthase inhibitor L-NAME in the patch pipette (Extended Data Fig. 6b), DA neuron bursts induce, rather than bLTP, a small, transient potentiation lasting no more than 6 minutes (EPSC amplitude (%) t 6min = 112.7 ± 4.4, P = 0.028, n = 7). Altogether, these data indicate that the astrocyte-mediated bLTP observed in young female mice is also present in adult female and male mice with similar cellular and molecular mechanism.

Adult IP 3 R2 −/− mice show astrocyte Ca 2+ -dependent bLTP
To further explore the role of astrocyte Ca 2+ signals in bLTP during adulthood, we performed DA neuron-paired recording experiments from VTA slices of adult IP 3 R2 −/− female and male mice. Unexpectedly, bLTP was observed (Fig. 4g), although the statistical significance of the potentiation in these mice (P < 0.05) is lower with respect to WT mice (P < 0.01 and P < 0.001, female and male mice, respectively). These results suggest that, in adult IP 3 R2 −/− mice, the astrocyte Ca 2+ response to DA neuron bursts is, at least in part, maintained. We, thus, evaluated Ca 2+ signals in astrocytes of adult male mice that specifically express GCaMP6f (Extended Data Fig. 7c,d). We found that the frequency of spontaneous events at thin astrocytic processes-that is, the so-called microdomains-is lower in IP 3 R2 −/− mice than in IP 3 R2 +/+ mice (mean event number per minute, IP 3 R2 +/+ , 144.6 ± 21.5; IP 3 R2 −/− , 69 ± 16.3; P < 0.05, t-test) and tends to be reduced also at the soma (Fig. 4k). However, similarly to IP 3 R2 +/+ mice, in IP 3 R2 −/− mice DA neuron bursts induce a significant increase in the number of Ca 2+ microdomains that can account for the presence of bLTP in these mice ( Fig. 4h-j). In both IP 3 R2 +/+ and IP 3 R2 −/− mice, Ca 2+ response at the soma, mean area and duration of Ca 2+ microdomains are unchanged after DA neuron bursts, whereas the amplitude is slightly, although significantly, reduced,

In vivo activation of VTA astrocytes favors DA neuron bursts
Glutamatergic synapses in the VTA circuitry modulate DA neuron firing activity and their potentiation, mainly mediated by NMDARs, Article https://doi.org/10.1038/s41593-022-01193-4 enhance DA neuron bursts, thus playing a key role in DA-dependent function and dysfunction 1,33 . Because astrocytes, as we show here, also induce a potentiation of these glutamatergic synapses, we asked whether in vivo astrocyte activation increases the burst firing mode of VTA DA neurons and eventually affects behavior. We injected AAV9-GFAP-hM3D(Gq)-mCherry or AAV5.GfaABC1D.cyto-tdTomato. SV40 in the VTA of adult male mice (Fig. 5a), specifically targeting astrocytes (Extended Data Fig. 9). Astrocytes were activated through brief pressure pulses applied to a CNO-containing glass pipette (Methods) while recording the firing activity from individual VTA neurons, showing the typical features of DA neurons (Fig. 5a,b). This approach warrants that only VTA astrocytes in proximity of the recorded neuron are stimulated, ruling out the activation of possible mistargeted astrocytes in regions surrounding the VTA. Consistent with the astrocyte-mediated enhancement of glutamatergic transmission to DA neurons observed in VTA slices, astrocyte activation by CNO increases the bursting discharges of all putative DA neurons that persist for at least 10 minutes (Fig. 5c,e,f), and it also increases the overall firing rate in five of seven DA neurons recorded (Fig. 5g,h). In contrast, DA neuron activity in tdTomato-expressing mice is unaffected by CNO, in terms of percentage of spikes in bursts and firing rate (Fig. 5d-h). These in vivo data show that astrocytes exert a direct control on VTA DA neuron firing activity.

In vivo activation of VTA astrocytes induces hyperlocomotion
The dopaminergic system plays a central role in the control of locomotor activity 3 . We, thus, evaluated whether the selective activation of astrocytes, which increases VTA DA neuron burst firing, controls locomotor activity. Locomotion was tested in male mice given bilateral VTA injections of AAV9-GFAP-hM3D(Gq)-mCherry (hM3D) or AAV8-GFAP-GFP (GFP) (Fig. 6a,b). Thirty minutes after intraperitoneal (i.p.) injections, CNO induces a locomotor hyperactivity in hM3D-injected mice as compared to GFP control mice (Fig. 6c-e). The time spent at the center was similar in the two groups (Fig. 6f), suggesting no major effects on anxiety-like phenotypes. Interestingly, 48 hours after CNO, a significant locomotor hyperactivity was still observed in hM3D mice over the first 10 minutes of the task (Fig. 6c-e).
Although we cannot rule out that possible mistargeted astrocytes in the substantia nigra (SN) contribute to the action of astrocytes described above, recent studies reveal that DA neurons of the VTA, and not those of the SN, play a major role in the induction of motor hyperactivity 3,34,35 . Overall, these in vivo data show that activation of astrocytes enhances VTA DA neuron firing activity and induces locomotor hyperactivity.

Discussion
The present study describes an astrocyte-mediated LTP of glutamatergic transmission to DA neurons in the VTA circuitry that we term bLTP. This novel form of synaptic plasticity is evoked by the following sequence of events (Fig. 7). First, DA neuron bursting activity induces the somatodendritic release of eCBs and DA; second, activation of CB1 and D2Rs in astrocytes triggers Ca 2+ elevations; and third, astrocyte activation, coupled with another signal, possibly NO (Supplementary Note 2), released during DA neuron bursts, leads to an LTP of excitatory transmission onto adjacent DA neurons. At the basis of this potentation is the pre-synaptic activation of the mGluR1 receptor that mediates a sustained increase in glutamate release probability. We also show that in vivo astrocyte activation increases burst and overall firing activity of DA neurons and induces hyperlocomotion. These results indicate that astrocytes play a key role in the modulation of VTA DA neuron circuits that control DA-dependent physiological functions. Astrocytes have been shown to respond with Ca 2+ elevations to synaptic neurotransmitters and, in turn, to contribute to sensory information processing and behavioral responses [13][14][15][36][37][38][39] . We show here that, in the VTA, activation of both CB1 and D2Rs is required for astrocyte Ca 2+ responses to DA neuron bursts and that these events are crucial for bLTP generation. This is based on the following observations: (1) CB1 and D2Rs are expressed and closely localized in the same astrocyte; (2) both astrocyte Ca 2+ response and bLTP induction are abolished in the presence of specific antagonists that block either the CB1 or the D2R; (3) DA neuron bursts fail to evoke bLTP after deletion in VTA astrocytes of either the CB1 or the D2R; (4) bLTP could not be evoked after the impairment of astrocyte Ca 2+ elevations downstream CB1 and D2R activation; and (5) in the presence of CB1 and D2R antagonists, bLTP can be induced in young male mice upon selective chemogenetic activation of astrocytes coupled to DA neuron bursts. Notably, the results reported above in (3)-(5) provide evidence that activation of neuronal CB1 and D2Rs is not required for bLTP induction.
It is worthwhile to further comment on results reported in (3). In our experiments on mice carrying the 'floxed' CB1 or D2 gene and injected in the VTA with AAV9/2-hGFAP-mCherry_iCre, we observed that the great majority of Cre + cells were astrocytes and only about 5% were neurons (Extended Data Fig. 7). However, this approach may lead to undetectable expression levels of the Cre recombinase and result in CB1 or D2R deletion in a higher percentage of neurons 40 . It is noteworthy, however, that neuronal CB1 and D2Rs in the VTA are inhibitory. Indeed, activation of pre-synaptic CB1 or D2Rs inhibits excitatory transmission onto VTA DA neurons 9,16 , and activation of post-synaptic D2Rs induces a hyperpolarization that reduces VTA DA neuron excitability 9 . Furthermore, D2R activation in the VTA favors eCB-induced suppression of excitation 16 . These well-established inhibitory actions of neuronal CB1 and D2Rs in VTA circuitry are not consistent with the CB1 and D2R-dependent bLTP that we describe here and further support that activation of astrocytic, rather than neuronal, CB1 and D2Rs is required for bLTP generation.
Recent studies reported that astrocytes in different brain regions, including the VTA, express D1 and D2-type receptors and respond to bath-applied DA stimuli with complex Ca 2+ dynamics, including regulation of basal cytosolic Ca 2+ and repetitive Ca 2+ transients 41-44 .  Most interestingly, astrocytes in the NAc respond to synaptic DA release with D1R-mediated rather than D2R-mediated Ca 2+ elevations 44 . These data confirm that astrocytes of different brain regions and synaptic circuits express different receptors that match the specific signals generated by distinct neuronal activities 13 . Consistent with this view, through the low-affinity D1R, NAc astrocytes can sense the transient, high DA concentrations generated by synaptic DA release 33 . Conversely, through the high-affinity D2R 33 , VTA astrocytes can sense lower DA concentrations mainly generated in the VTA by somatodendritic rather than synaptic release and, thus, be functional targets of DA volume transmission 45 (Supplementary Note 1). An additional specificity of VTA astrocytes is that a cooperativity between CB1 and D2Rs is necessary for the Ca 2+ response to DA neuron bursts, being activation of either CB1 or D2Rs alone insufficient to induce astrocytic Ca 2+ elevations. Our EM immunogold experiments provide an ultrastructural background for this cooperativity, revealing that CB1 and D2Rs are expressed in the same astrocytes, closely localized at astrocytic processes. Quantitative analysis from CB1/D2R double-labeled astrocytic processes also reveals that a group of couples exhibits an edge-to-edge separation ≤50 nm, which suggests physical interactions between CB1 and D2Rs and possible formation of heterodimers. Consistently, previous studies reported that D2 and CB1R co-activation in neurons enhances the formation of CB1/D2R heterodimers 17,46 . Furthermore, we recently showed that co-activation of GABA B and somatostatin receptors in neocortical astrocytes confers signaling specificity between different interneuron subtypes and astrocytes 47 .
Although additional experiments are necessary to fully elucidate the mechanism by which the effect that we observe involves both D2 and CB1 receptors, a cooperativity between different G-protein-coupled receptors may, therefore, be a general functional feature of the astrocyte response to neuronal signals. Overall, CB1/D2R-expressing astrocytes in the VTA are fine-tuned to sense eCB and dopamine-releasing neurons and extent excitation to neighboring DA neurons through lateral potentiation of glutamatergic transmission. These results provide further evidence for circuit-specificity and synapse-specificity of neuron-astrocyte reciprocal signaling in the brain 13 .
The astrocyte-mediated bLTP is absent in young male mice in which DA neuron bursts fail to elevate Ca 2+ in astrocytes. Our data suggest that this failure is most likely due to a lower expression of astrocytic CB1Rs in young with respect to adult male mice showing regular bLTP. Notably, bLTP is observed in young male mice by coupling DA neuron bursts with Ca 2+ elevations evoked by CNO in hM3D-expressing astrocytes. These results further support a crucial role of astrocytic Ca 2+ signals in bLTP induction mechanism.
In hippocampus, dorsal striatum and neocortex, the somatodendritic release of eCBs recruits astrocytes that modulate synaptic transmission through pre-synaptic receptor activation [18][19][20]48 . In our VTA experiments, the presence of the mGluR1β at excitatory axon terminals, the significant changes in PPR and the CV values are also consistent with a pre-synaptic mechanism in bLTP mediated by mGluR activation (Supplementary Note 3). Different astrocytic actions could account for pre-synaptic mGluR modulation, including (1) glutamate release from activated astrocytes, (2) changes in astrocytic glutamate transporter activity that increase extracellular glutamate 49 or (3) rapid, structural rearrangement of astrocytic peri-synaptic processes expressing glutamate transporters 50 . Additional experiments are necessary to specifically address these different hypotheses.
Astrocytic Ca 2+ signaling is required for bLTP generation in young and adult mice. However, although the mechanism downstream astrocyte Ca 2+ signaling remains similar, IP 3 R2 deletion in adult mice is insufficient to abolish bLTP, and, consistently, a Ca 2+ response to DA neuron bursting is observed at astrocyte processes from IP 3 R2 −/− mice. This finding reveals an increased complexity in the regulatory mechanisms of astrocytic Ca 2+ dynamics during development with contribution of signaling pathways other than the IP 3 R2-mediated pathway in IP 3 R2 −/− mice 27,51 . From these observations, it also follows that negative results on the role of astrocytic Ca 2+ signaling in IP 3 R2 −/− adult mice must be interpreted with caution.
Transient and/or persistent potentiation of glutamatergic synapses, fundamentally mediated by NMDAR activation, regulates the burst firing mode of VTA DA neurons that plays a pivotal role in DA-dependent behaviors 1,33 . The novel form of astrocyte-mediated potentiation described here may integrate with these other forms of NMDAR-dependent plasticity that favor the burst firing of DA neurons. Consistent with this view, we report that in vivo astrocyte activation enhances DA neuron bursts and leads to a long-lasting locomotor hyperactivity that recent studies revealed to depend on VTA rather than SN DA neuron activity 3,34,35 . Although the molecular mechanism of astrocyte modulation of glutamatergic transmission, showed in brain slice preparations, is fully consistent with the results presented in Fig.  5, we cannot exclude that other processes, apart from the synaptic mechanism, can contribute to the effects observed in vivo. Additional experiments are necessary to specifically address this issue.
We suggest that activation of astrocytes by burst firing DA neurons and the consequent lateral potentiation of glutamatergic synapses may represent a strategy used by individual DA neurons to expand the burst firing mode to neighboring DA neurons. Hence, it is possible that, through the fundamental recruitment of astrocytes, an isolated, high-bursting DA neuron favors the formation of spatially defined clusters of co-active DA neurons that convey essential information about a specific subset of behavioral variables to target regions 52 .
The present results show that astrocyte signaling induces a long-lasting potentiation of glutamatergic synapses to VTA DA neurons, induces a sustained increase in the burst firing mode of DA neurons and favors locomotor hyperactivity, thereby revealing an astrocyte-mediated mechanism in the control of DA neuron activity and DA-dependent behaviors. Our study also paves the way to future investigations examining whether dysregulations of DA neuron-astrocyte reciprocal communication within the VTA contribute to the development of disease states, including motivation disorders, psychiatric disorders with a strong motor component, such as attention-deficit/ hyperactivity disorder, and drug addiction.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41593-022-01193-4.  Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.  Fig. 2). Separated data from adult non-littermate and littermate mice are reported in Extended Data Fig. 10. Given that results obtained from these groups were similar, data from adult mice were pooled together in Fig. 4. For slice preparations, animals were anesthetized with isofluorane, and the brain was removed and transferred into an ice-cold artificial cerebrospinal fluid (ACSF) containing (in mM): 125 NaCl, 2 KCl, 2

Electrophysiological recordings and extracellular stimulation
Brain slices were continuously perfused in a submerged chamber with recording solution containing (in mM): 120 NaCl, 2 KCl, 1 NaH 2 PO 4 , 26 NaHCO 3 , 1 MgCl 2 , 2 CaCl 2 , 10 glucose, pH 7.4 (with 95% O 2 /5% CO 2 ). Picrotoxin (50 µM) was added to block GABA A receptor currents. When indicated, other antagonists were bath-applied with the recording solution. Cells were visualized with an Olympus FV1000 microscope (Olympus Optical). Conventional VTA DA neurons were recorded in the lateral part of the region medial to the medial terminal nucleus of the accessory optical tract (Fig. 1) 5,57 . DA neurons from the lateral VTA were identified on the basis of their distinct morphology characterized by a large and elongated soma with no particular dendritic orientation and the presence of the following electrophysiological properties: a low-frequency tonic firing, a large I h current elicited by hyperpolarizing steps 16 and a slow depolarizing potential during current step injections 58 (Extended Data Fig. 1b-d). Simultaneous electrophysiological whole-cell patch-clamp recordings from two DA neurons were made (distance of the somata, 70-120 µm; this distance was used in previous studies that revealed astrocyte-mediated lateral potentiation 18 , and it is consistent with the territory occupied by individual astrocytic processes, which has a diameter of 80-100 µm 59 ). Patch electrodes for neuronal recordings (resistance, 3-4 MΩ) were filled with an internal solution containing (in mM): 135 K-gluconate, 70 KCl, 10 HEPES, 1 MgCl 2 and 2 Na 2 ATP (pH 7.4, adjusted with KOH, 280-290 mOsm L −1 ).
Recordings were obtained using a multiclamp-700B amplifier (Molecular Devices). Signals were filtered at 1 kHz and acquired at 10-kHz sampling rate with a DigiData 1440A interface board and pClamp 10 software. Series and input resistances were monitored throughout the experiment using a 5-mV pulse. Recordings were considered stable when the change of series and input resistances were below 20%. Cells that did not meet these criteria were discharged. Theta capillaries filled with recording solution were used for bipolar stimulation. To stimulate glutamatergic afferents, electrodes were connected to an S-900 stimulator through an isolation unit and placed 100-200 µm rostral to the recording electrode (Extended Data Fig. 1a). Paired pulses (50-ms intervals) were delivered at 0.33 Hz. EPSCs were recorded while holding the membrane potential at −70 mV. Stimulus intensity was adjusted to evoke 30-50% maximal EPSC amplitude. The EPSC amplitude was measured as the peak current amplitude (2-9 ms after stimulus) minus the mean baseline current (100 ms before stimulus). To illustrate the mean EPSCs time course, values were grouped in 3-minute bins (that is, mean EPSCs from 60 stimuli). Changes in mean EPSCs in the first DA neuron were monitored after imposing a burst or a tonic firing pattern to the second DA neuron (70-120 µm apart). Burst firing pattern was imposed in current-clamp mode, through injections of intracellular current pulses, five pulses at 20 Hz every 500 ms for 5 minutes (Extended Data Fig. 1e) 23 . Tonic firing was imposed with individual current pulses applied at 2 Hz for 5 minutes 23 (Extended Data Fig. 1e). During the burst/ tonic firing, the extracellular stimulation was switched off. In electrophysiological experiments, time 0 indicates the end of the burst/ tonic firing. For statistical analysis of long-term effects, mean EPSCs from 120 stimuli applied before (basal), 24-30 minutes (indicated as 30-minute timepoint in the bar chart) or 39-45 minutes (indicated as 45-minute timepoint in the bar chart) after the firing protocol were compared. In young adult mice, only the long-term effect at the timepoint of 30 minutes was analyzed owing to the difficulty of obtaining long-lasting recordings in tissues from these mice. PPR was calculated as 2nd EPSC/1st EPSC, and evaluation of the PPR before (mean value from two basal recordings) and after (mean value from recordings at 39 minutes and 45 minutes, indicated as 45 minutes) the burst firing protocol was used to identify the pre-synaptic or post-synaptic locus of the bLTP. For the analysis of the CV (CV = σ/µ) of the EPSCs, we divided the standard deviation (σ) by the mean (µ) of 120 evoked EPSCs before and 39-45 minutes (indicated as 45min timepoint) after the burst firing protocol for each potentiated cell. Then, we calculated (CV −2 45min normalized)/mean 45min normalized for each cell. Values of ((CV −2 45min norm)/mean 45min norm) > 1 support a pre-synaptic locus of plasticity expression 60  depth) were acquired with a 0.5-Hz frame rate for 90 seconds, with time intervals of 5 minutes between recordings. Image sequences were processed with ImageJ software. Regions of interest (ROIs) were drawn around cellular somata using the red SR101 signal. Ca 2+ events were estimated as changes of the Fluo-4 fluorescence signal over baseline (ΔF/F 0 = (F(t)−F 0 )/F 0 ). A fluorescence increase was considered an event when it exceeded two times the standard deviation from the baseline. Astrocyte Ca 2+ responses were quantified by analyzing the probability of occurrence of Ca 2+ spike by detecting the onset of Ca 2+ elevations (Ca 2+ spikes) during the recording period. To investigate the astrocyte response to the burst firing of DA neurons, a DA neuron was patched with an intracellular solution (see details before) containing the fluorescent tracer Neurobiotin 488 (60 µM, Vector Laboratories) to visualize neuronal soma and dendrites. To obtain the time course of the Ca 2+ spike probability index reported in Figs. 1 and 2, the number of astrocytic Ca 2+ spikes for each recording period was divided by the number of SR101 + astrocytes in proximity (around 50 µm) to Neurobiotin 488-filled DA neuron soma and dendrites. After three basal recordings, a burst firing pattern was imposed to the DA neuron (in current-clamp mode, through injections of intracellular current pulses, with five-pulse 20-Hz burst every 500 ms for 5 minutes), and the quantification of the Ca 2+ spike probability was resumed 4.5 minutes after the initiation of the burst firing. In Ca 2+ imaging experiments, time 0 indicates the onset of the burst firing. For statistical analysis, a mean value of the Ca 2+ spike probability per minute per slice was calculated at basal conditions (mean of the three basal recordings) and after DA neuronal burst firing (mean of four consecutive recordings after the burst firing, the first at a timepoint of 4.5 minutes after the burst firing and the last at a timepoint of 24 minutes after the burst firing). To analyze the astrocyte response to D1R activation in young female mice, the D1-type receptor agonist SKF 38393 (1 mM) was locally delivered to SR-101 and Fluo-4 loaded VTA slices by using a pressure ejection unit (PDSE, NPI Electronics) that applies pulses (0.5 bar, 2 seconds) to a SKF 38393-containing pipette. Astrocyte Ca 2+ responses in the presence of TTX (1 µM) were quantified analyzing the Ca 2+ spike probability in 10-second bins. A mean time course of the Ca 2+ spike probability per slice was calculated at basal conditions and after SKF 38393 challenge, from three recordings in each condition (5-minute intervals between recordings). The mean time course of the Ca 2+ spike probability for all the experiments is reported in Extended Data Fig. 3. For statistical analysis, the mean Ca 2+ spike probability per minute was calculated at basal conditions and in the 10-second bin immediately after the SKF 38393 challenge. These experiments were performed in the absence or presence of the D1R antagonist SCH-23390 (10 µM). At the end of the recording session, ATP (4 mM) was locally delivered, and the Ca 2+ spike probability in response to this agonist was calculated. When indicated, Ca 2+ imaging experiments were performed in slices from young mice expressing the genetically encoded Ca 2+ indicator cytoG-CaMP6f and the Gq-protein-coupled DREADD hM3D in astrocytes (for AAV delivery details, see below). In these experiments, Ca 2+ elevations were evoked by bath perfusion of the hM3D agonist CNO (10 µM). When CNO was coupled with DA neuron burst, CNO bath perfusion initiated 2.5 minutes after the start of the burst firing. A mean time course of the Ca 2+ response to CNO was calculated by plotting the ΔF/F 0 of an ROI drawn around the entire recording field. Then, plots were aligned for the Ca 2+ peak to calculate the mean time course of the Ca 2+ response to CNO. In adult male mice, Ca 2+ signals were studied in GCaMP6f and tdTomato co-expressing astrocytes (for AAV delivery details, see below). Calcium imaging experiments in brain slices were performed using a two-photon laser scanning microscope (Multiphoton Imaging System, Scientifica) equipped with a pulsed red laser (Chameleon Ultra 2, Coherent). Power at sample was 10-17 mW. GCaMP6f and tdTomato were excited at 920 nm. Images (12-bit depth) were acquired with a water-immersion lens (Olympus, LUMPlan FI/IR ×20, 1.05 NA), with a field of view of 120 × 120 µm at 1.5-Hz acquisition frame rate. Calcium signal recordings were performed for 2 minutes with 5-minute time intervals. To investigate the astrocyte response to the burst firing of DA neurons, a DA neuron was patched with an intracellular solution containing the fluorescent tracer Alexa Fluor 594 (60 µM, Vector Laboratories), which allows visualization of neuronal soma and dendrites at 800 nm. After three basal recordings, a burst firing pattern was imposed to the DA neuron (in current-clamp mode, through injections of intracellular current pulses, five pulses at 20 Hz every 500 ms for 5 minutes), and the Ca 2+ signal recordings were resumed 4.5 minutes after the initiation of the burst firing. In these experiments, time 0 indicates the onset of the burst firing. To extract Ca 2+ event dynamics at astrocyte processes from the entire field of view in an automated, unbiased, event-based way, we used AQuA 61 . Analysis of Ca 2+ signals at the level of soma was performed by drawing ROIs around cellular somata using the red tdTomato signal. Then, Ca 2+ events were identified with ImageJ and a custom software developed in MATLAB 7.6.0 R2008 A (MathWorks) that essentially combines a threshold measured from the global baseline with a stricter threshold computed from a local baseline (for details, see ref. 47 ). For statistical analysis, the mean value of the number of events per minute at processes and the number of events per minute per soma were calculated at basal conditions (mean of the three basal recordings) and after DA neuronal burst firing (mean of four consecutive recordings after the burst firing, the first at a timepoint of 4.5 minutes after the burst firing onset and the last at a timepoint of 25.5 minutes after the burst firing onset). The area, amplitude (ΔF/F 0 ) and duration (from 10% onset time to 10% offset time) of Ca 2+ events extracted by AQuA at basal conditions and after the burst firing were compared using the Kolmogorov-Smirnov test.
Depth of anesthesia was assured by monitoring respiration rate, eyelid reflex, vibrissae movements and reactions to pinching the tail and toe. After drilling two holes into the skull over the VTA, we bilaterally injected a total volume of 500 nl per hole by using a pulled glass pipette connected to a peristaltic pump, at a rate of 100 nl min −1 . To express GCaMP6f and tdTomato in VTA astrocytes from adult mice, we injected a total volume of 1 µl by using a pulled glass pipette connected to a custom-made pressure injection system. To minimize AAV spreading along the pipette track, in adult mice injections the pipette was kept in the tissue for 10 minutes before slow withdrawal. The spreading to overlying tissue needs to be considered in all studies employing microinjections in subcortical brain regions and subsequent in vivo experiments 40 , whereas this can be hardly a concern for the studies using horizontal brain slice preparations. After injections, the skin was sutured, and mice were revitalized under a heat lamp and returned to their cage. VTA microinjections are astrocyte selective for the expression of the AAV-coded proteins (Results) and fundamentally restricted to the VTA. Some AAV spreading in surrounding VTA tissue is occasionally observed, mainly in young mice. In these latter mice, however, we limited our study to horizontal brain slice preparations where neural circuits and glutamatergic inputs are highly isolated and pharmacologically controlled (that is, experiments were performed in the presence of the GABA A R antagonist picrotoxin). Furthermore, given the maximal length of astrocytic processes of about 50 µm, possible mistargeted astrocytes in regions outside the VTA, such as the SN, might affect the glutamatergic inputs to VTA DA neurons at the SN-VTA border, but they could not reach the lateral VTA region where pair recordings were performed.

Pre-embedding and post-embedding EM
Thirteen P16 and eight P50 C57BL/6 mice (seven females and six males for P16; four females and four males for P50) were used. Mice were anesthetized with an i.p. injection of chloral hydrate (300 mg kg −1 ) and perfused transcardially with a flush of saline solution, followed by 4% freshly depolymerized paraformaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer (PB, pH 7.4). Brains were removed, post-fixed in the same fixative (for 48 hours) and cut on a vibratome in 50-µm serial horizontal sections from the midbrain, which were collected in PB until processing 63 . Horizontal sections were through the dorso-ventral extent of the VTA, resulting in 7-8 sections per series. To verify the dorso-ventral extension of VTA, a pilot series of sections from a male mouse were immuno-processed for tyrosine-hydroxylase (primary antibodies from Millipore, AB1542, RRID:AB_90755 (ref. 64 ); 1:500) and for Nissl staining. In immuno-reacted sections (see 'Data collection and data analysis' section), lateral VTA was identified as the region medial to the medial terminal nucleus of the accessory optical tract 5,57 .
Immunoperoxidase and pre-embedding procedures. Sections were treated with H 2 O 2 (1% in PB, 30 minutes) to remove endogenous peroxidase activity, rinsed in PB and pre-incubated in 10% normal goat serum (NGS, 1 hour, for mGLUR1α, mGlur1β, D2, D3, D4 and D1) or in 10% normal donkey serum (NDS, 1 hour, for CB1). Sections were then incubated in a solution containing primary antibodies (see Supplementary Table 7 for dilutions, 2 hours at room temperature and overnight at 4 °C). The next day, sections were rinsed three times in PB and incubated first in 10% NGS or 10% NDS (15 minutes) and then in a solution containing secondary biotinylated secondary antibodies (see Supplementary Table 7 for dilutions, 1.5 hours at room temperature). Sections were subsequently rinsed in PB, incubated in avidin-biotin peroxidase complex (ABC Elite, PK6100, Vector Laboratories), washed several times in PB and incubated in 3,3′diaminobenzidine tetrahydrochloride (DAB; 0.05% in 0.05 M Tris buffer, pH 7.6 with 0.03% H 2 O 2 ). Method specificity was verified by substituting primary antibodies with PB or NGS. As previously described 63 , after completion of immunoperoxidase procedures, sections were post-fixed in 1% osmium tetroxide in PB for 45 minutes and contrasted with 1% uranyl acetate in maleate buffer (pH 6.0, 1 hour). After dehydration in ethanol and propylene oxide, sections were embedded in Epon/Spurr resin (Electron Microscopy Sciences), flattened between Aclar sheets (Electron Microscopy Sciences) and polymerized at 60 °C for 48 hours. Chips including lateral VTA were selected by light-microscopic inspection, glued to blank epoxy and sectioned with an ultramicrotome (MTX, Arizona Research and Manufacturing Company). The most superficial ultrathin sections (∼60 nm) were collected and mounted on 300 mesh nickel grids, stained with Sato's lead and examined with Philips EM 208 and CM10 electron microscopes coupled to a MegaView 2 high-resolution CCD camera (Olympus Soft Imaging Solutions). To minimize the effects of procedural variables, all material from P16 and P50 females and males was processed in parallel.
Post-embedding procedures. Sections were processed for an osmium-free embedding method 65 . Dehydrated sections were immersed in propylene oxide, infiltrated with a mixture of Epon/Spurr resins, sandwiched between Aclar films and polymerized at 60 °C for 48 hours. After polymerization, chips were cut from the wafers, glued to blank resin blocks and sectioned with an ultramicrotome. Thin sections (60-80 nm) were cut and mounted on 300 mesh nickel grids and processed for immunogold labeling 65,66 . In brief, after treatment with 4% para-phenylenediamine in Tris-buffered saline (0.1 M Tris, pH 7.6, with 0.005% Tergitol NP-10 (TBST)), grids were washed in TBST (pH 7.6), transferred for 15 minutes in 0.25% NDS in TBST (pH 7.6) and then incubated overnight (26 °C) in a solution of TBST (pH 7.6) containing a mixture of anti-D2 and anti-CB1 primary antibodies (see Supplementary  Table 7 for dilutions). Grids were subsequently washed in TBST (pH 8.2), transferred for 10 minutes in 0.5% NDS in TBST (pH 8.2), incubated for 2 hours (26 °C) in TBST (pH 8.2) containing secondary antibodies conjugated to 18-nm and 12-nm gold particles, washed in distilled water and then stained with uranyl acetate and Sato's lead. The optimal concentration of antibodies to D2 and CB1Rs was sought by testing several dilutions; the concentration yielding the lowest level of background labeling and still immunopositive elements was used to perform the final studies. Gold particles were not detected when primary antiserum was omitted. When normal serum was substituted for immune serum, sparse and scattered gold particles were observed, but they did not show any specific relationship to subcellular compartments.
Data collection and analysis. All data were obtained from lateral VTA of immuno-reacted sections 5,57 . For pre-embedding EM, mGluR1α, mGluR1β, CB1, D1, D2, D3 and D4R immuno-reactive profiles were studied in ultrathin sections from the surface of the embedded blocks. Quantitative data derive from the analysis of microscopic fields of lateral VTA (10-12 ultrathin sections per animal) that were selected and captured at original magnifications of ×12,000-×30,000. Microscopical fields from females and males containing positive processes were randomly selected. Acquisition of microscopical fields and analysis of female and male mice were performed under blinded conditions.
For the analysis of the distribution of mGluR1α, mGluR1β, CB1, D1, D2, D3 and D4R positive profiles, subcellular compartments were identified according to well-established criteria 67 (Extended Data  figures and Supplementary Tables 1, 3 and 4). For quantifying mGluR1α or mGluR1β in P16 VTA and mGluR1β in P50 VTA at axon terminals, synapses exclusively characterized by a pre-synaptic terminal with clear and round vesicles nearby the pre-synaptic density, by a synaptic cleft displaying electron dense material, by pre-synaptic and post-synaptic membranes defining the active zone and the post-synaptic specialization and, finally, by a prominent post-synaptic density, the asymmetric synapses 67,68 were sampled (axon terminals making asymmetric synaptic contacts containing one or more dense core vesicles more likely representative of co-release of glutamate and others neurotransmitters [69][70][71] were not included in this group; Supplementary Tables 2 and 6).
For quantifying CB1, D1, D2, D3 and D4R at astrocytic processes in P16 VTA and CB1 and D2R in P50 VTA, astrocytic profiles were identified based on their typical irregular outlines and the paucity of cytoplasmic components (with the exception of ribosomes, glycogen granules and various fibrils 67 ). For post-embedding EM, ultrathin sections (20 ultrathin sections per animal) were examined at ×50,000-×85,000, and fields that included at least one double immuno-labeled astrocytic profile were selected. For determining the relative density of D2 and CB1Rs at the membranes of double-labeled astrocytic profiles, pyramidal cell nuclei were also identified: gold particles within labeled structures were counted, and areas were calculated using ImageJ. Background was calculated by estimating labeling density over pyramidal cell nuclei 66,72 . Particle densities were counted and compared with background labeling. Gold particles were considered associated with plasma membrane if they were within 20 nm of the extracellular side of the membrane. To determine the degree of nearness of D2 and CB1R at the membrane of double-labeled profiles, the edge-to-edge distances between immunogold-labeled D2 and CB1R were measured along the membrane using ImageJ, and the distribution of the separation distances between D2 and CB1R was determined 47, 66,[73][74][75][76] . In the cases in which multiple paths connecting particles gave different inter-distance values, the shortest inter-distance was selected and used for distribution analysis. Given that gold particles with edge-to-edge separation distance ≤50 nm are highly suggestive of physical interactions of two detected proteins (that is, a physical coupling complex 47,66,73-76 ), distribution analysis of the inter-distance between particles was based on bins of 50 nm.

In vivo single-unit recordings
C57BL/6J WT male mice, injected 2-3 weeks before with ssAAV-9/2-hGFAP-hM3D(Gq)_mCherry-WPRE-hGHp(A) or AAV5. GfaABC1D.cyto-tdTomato.SV40, were anesthetized using chloral hydrate (400 mg kg −1 i.p.), supplemented as required to maintain optimal anesthesia throughout the experiment, and placed in the stereotaxic apparatus (Kopf). Their body temperature was maintained at 36 ± 1 °C using a feedback-controlled heating pad. For the placement of a recording electrode, the scalp was retracted, and one burr hole was drilled above the parabrachial pigmented nuclei of the posterior VTA (AP: −3.0 to −3.5 mm from bregma; L: 0.4-0.6 mm from midline; V: 4-5 mm from the cortical surface) according to the Paxinos and Franklin atlas (2004). Extracellular identification of putative DA neurons was based on their location as well as on the set of electrophysiological features that characterize these cells in vivo: (1) a typical triphasic action potential with a marked negative deflection; (2) an action potential width from start to end >2.5 ms; and (3) a slow firing rate (<10 Hz). VTA putative DA neurons were selected only when all the already published criteria were fulfilled [83][84][85][86] . Single-unit activity of putative DA neurons was recorded extracellularly using glass micropipettes filled with 2% Chicago sky blue dissolved in 0.5 M sodium acetate (impedance 3-7 MΩ). An injection pipette (20-40 µm in diameter attached 100-150 µm above the recording tip) was used for simultaneous microinjections of CNO (1 mM). This approach allowed us to specifically activate VTA astrocytes in proximity of the glass pipette tip and evaluate their action on the local VTA circuitry. Signal was pre-amplified, amplified (NeuroLog System, Digitimer), filtered (band-pass 500-5,000 Hz) Article https://doi.org/10.1038/s41593-022-01193-4 and displayed on a digital storage oscilloscope. Experiments were sampled on-line and off-line by a computer connected to CED Power 1401 laboratory interface (Cambridge Electronic Design) running the Spike2 software (Cambridge Electronic Design). Single units were isolated, and the spontaneous activity was recorded for a minimum of 3 minutes before local application of CNO (1 mM). A total volume of 30-100 nl was infused using brief (10-100-ms) pressure pulses (40 psi, Picospritzer). One injection maximum per hemisphere was given. For statistical analysis, we calculated the mean firing rate (number of spikes per second) and the percentage of spikes in burst (SiB) before and after CNO application (in 2-minute bins or in the 10 minutes of recording after CNO application). Bursts were defined as the occurrence of two spikes at an inter-spike interval of <80 ms and terminated when the inter-spike interval exceeded 160 ms 87 . At the end of the experiment, negative DC (15 mA for 5 minutes) was passed through the recording electrode to eject Pontamine sky blue, which allowed the anatomical location of the recorded neuron. Mice were then euthanized, and brains were rapidly removed and fixed in 4% paraformaldehyde solution. The position of the electrodes was identified with a microscope in coronal sections (100 µm). Only recordings in the correct area were considered for analysis.

Behavioral test
Viral Injection. C57BL/6J mice were naive and 2 months old at the time of surgery. All mice were anesthetized with a mix of isoflurane/oxygen 2%/1% by inhalation and mounted into a stereotaxic frame (Kopf). Brain coordinates of viral injections in the VTA were chosen in accordance with the Mouse Brain Atlas: AP: −3 mm; ML: ± 0.50 mm; DV: −4.7 mm. The volume of AAV injection (AAV9-GFAP-hM3D(Gq)-mCherry or AAV8-GFAP-GFP) was 100 nl per hemisphere. We infused virus through a glass micropipette connected to a 10-µl Hamilton syringe. After infusion, the pipette was kept in place for 6 minutes and then slowly withdrawn.
Locomotor activity. Mice were tested during the first 2 hours of the dark phase in an experimental apparatus consisting of four gray, opaque, open-field boxes (40 × 40 × 40 cm) evenly illuminated by overhead lighting (5 ± 1 lux). Each session was video recorded with ANY-maze tracking software (Stoelting) for 1 hour. In the first day of locomotor activity, all animals received an injection of CNO (3 mg kg −1 ) 30 minutes before the beginning of the test; 48 hours later, the animals were tested for a second time in the same apparatus with a saline injection.

Data collection and data analysis
Data collection was performed with Clampex 10.5, ANY-maze tracking software, Spike2 software, Leica Application Suite software 2.5.

Statistical analysis
No statistical methods were used to pre-determine sample size, but our sample sizes are similar to those reported in previous publications 19,[94][95][96][97] . Mice were randomized to groups. Data were not subject to exclusion except in cases of viral vector misplacement. For electrophysiological experiments in slices, recordings were not considered when the change of series and input resistances were above 20%. In EM, immunohistochemical, single-unit recordings in vivo and behavioral experiments, data collection and analysis were blinded to investigators. Experiments in brain slices were not blinded to investigators. However, the paired design of the study, with comparisons to internal control values in all experiments, and the absence of manual scoring during analysis avoid the experimenter bias. Data are expressed as mean ± s.e.m., except for Fig. 2e. In Fig. 2e, data are presented as a box and whisker plot. Each box is defined by the 25th and 75th percentiles; the central line indicates the median; and the dot indicates the mean value. The whiskers represent the minimum and maximum values in 1.5× the interquartile range. Normality test (Shapiro-Wilk test) was applied to the data before running statistical tests. Based on the normality test result, data before and after burst firing were analyzed using either parametric tests (paired t-test and one-sample t-test) or non-parametric tests (Wilcoxon signed-rank test and one-sample Wilcoxon signed-rank test) as appropriate. When indicated, data in the absence and presence of antagonist were compared (unpaired t-test or Mann-Whitney rank-sum test, depending on the data distribution). The area (µm 2 ), amplitude (ΔF/F 0 ) and duration (seconds, from 10% onset time to 10% offset time) of Ca 2+ events extracted by AQuA at basal conditions and after the burst firing were compared using the Kolmogorov-Smirnov test. For EM data analysis, the Mann-Whitney test and contingency Fisher's test were used. For in vivo single-unit recordings and behavioral tests, two-way repeated-measures ANOVA and Bonferroniʼs multiple comparison test with adjustment was used. Two-tailed tests were always performed. Statistical differences were established with P < 0.05 (*), P < 0.01 (**), P < 0.001 (***) and P < 0.0001 (****).

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request. Source data are provided with this paper. Top, bLTP can be evoked in IP 3 R2 +/+ (n = 8 from 8 mice, p = 0.029, two-tailed One Sample t-test), but not in IP 3 R2 −/− (n = 5 from 4 mice, p = 0.524, two-tailed One Sample t-test) young female littermates. Bottom, the bLTP in IP 3 R2 +/+ female mice is accompanied by a reduced PPR (p = 0.031, two-tailed paired t-test), similarly to that observed in C57BL/6J young female mice (see Fig. 1). Analysis of the coefficient of variation of EPSCs, 45 min after burst firing for potentiated cells in IP 3 R2 +/+ young mice (black circle, mean value). c) DA neuron bursts evoke an increase of the Ca 2+ spike probability/min in astrocytes from IP 3 R2 +/+ (n = 9 from 5 mice, p = 0.017, two-tailed paired t-test), but not in astrocytes from IP 3 R2 −/− (n = 6 from 4 mice, p = 0.533, two-tailed paired t-test) young female littermates. Data are represented as mean ± SEM.  Fig. 5 | Effects of the NO synthase inhibitor L-NAME on bLTP and astrocyte Ca 2+ response to DA neuron burst firing. a) Time course and bar chart of EPSC amplitude in the presence of the NO synthase inhibitor L-NAME (100 µM in the patch pipette of the burst firing DA neuron, n = 12 from 9 mice, p = 0.277; two-tailed One sample t-test). b) Mean amplitude of normalized EPSCs in female mice, 6 min after bursts, in the presence of different antagonists (AM251, n = 7 from 4 mice, p = 0.105; eticlopride, n = 10 from 8 mice, p = 0.291; LY-367385, n = 12 from 9 mice, p = 0.215; L-NAME, n = 12, from 9 mice p = 0.044; twotailed One sample t-test). c) Time course and bar chart of astrocytic Ca 2+ spike probability/min in the presence of L-NAME before and after burst firing (100 µM, n = 7 from 4 mice, p = 0.075; two-tailed paired t-test). d) Mean astrocytic Ca 2+ spike probability/min in female mice, at basal conditions and 4.5 min after burst, in the presence of different antagonists (AM251, n = 6 from 3 mice, p = 0.671; eticlopride, n = 6 from 3 mice, p = 0.673; LY-367385, n = 6 from 4 mice, p = 0.048; L-NAME, n = 7 from 4 mice, p = 0.009; two-tailed paired t-test). e) A 5 min bath perfusion of CNO (10 µM), in the absence and presence of DEA NONOate (10 µM), transiently (in the first 9 min) increases EPSC amplitude of DA neurons in male mice expressing hM3D in astrocytes (CNO, n = 7 from 6 mice, p = 0.016, two-tailed One sample Wilcoxon Signed Rank test; CNO + DEA NONOate, n = 13 from 9 mice, p = 0.013, two-tailed One sample t-test). These experiments were performed in the presence of AM251 and eticlopride. Data are represented as mean ± SEM. Fig. 6 | The mechanism of bLTP generation in young female mice is preserved in adult mice. a) Representative EM images of mGluR1β expression at axon terminals (AxT+) forming asymmetric synaptic contacts (arrowheads) with dendrites (Den) and CB1 and D2R localization at astrocytic processes (AsP+) from adult female and male mice. Green and blue arrows indicate the presence of immunopositive products in female and male, respectively. Scale bar, 300 nm. b) Time course and bar chart of the mean amplitude of normalized EPSCs in adult male mice in the presence of different antagonists (L-741,626 (D2R) 10 µM, n = 9 from 7 mice, p = 0.34, two-tailed One Sample t-test; AM251 (CB1R), n = 11 from 8 mice, p = 0.24, two-tailed One Sample Wilcoxon signed Rank test; LY-367385 (mGluR1), n = 8 from 7 mice, p = 0.096, two-tailed One Sample t-test; L-NAME (NO synthase), n = 7 from 5 mice, p = 0.604, two-tailed One Sample t-test). As in young mice, bLTP generation in adult mice requires eCB-DA signaling and mGluR activation. Data are represented as mean ± SEM.

nature research | reporting summary
April 2020 Corresponding author(s): GOMEZ-GONZALO MARTA, CARMIGNOTO GIORGIO Last updated by author(s): Sep 19, 2022 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability The data that support the findings of this study are available from the corresponding author upon reasonable request.