Dysfunction of ventral tegmental area GABA neurons causes mania-like behavior

The ventral tegmental area (VTA), an important source of dopamine, regulates goal- and reward-directed and social behaviors, wakefulness, and sleep. Hyperactivation of dopamine neurons generates behavioral pathologies. But any roles of non-dopamine VTA neurons in psychiatric illness have been little explored. Lesioning or chemogenetically inhibiting VTA GABAergic (VTAVgat) neurons generated persistent wakefulness with mania-like qualities: locomotor activity was increased; sensitivity to D-amphetamine was heightened; immobility times decreased on the tail suspension and forced swim tests; and sucrose preference increased. Furthermore, after sleep deprivation, mice with lesioned VTAVgat neurons did not catch up on lost sleep, even though they were starting from a sleep-deprived baseline, suggesting that sleep homeostasis was bypassed. The mania-like behaviors, including the sleep loss, were reversed by valproate, and re-emerged when treatment was stopped. Lithium salts and lamotrigine, however, had no effect. Low doses of diazepam partially reduced the hyperlocomotion and fully recovered the immobility time during tail suspension. The mania like-behaviors mostly depended on dopamine, because giving D1/D2/D3 receptor antagonists reduced these behaviors, but also partially on VTAVgat projections to the lateral hypothalamus (LH). Optically or chemogenetically inhibiting VTAVgat terminals in the LH elevated locomotion and decreased immobility time during the tail suspension and forced swimming tests. VTAVgat neurons help set an animal’s (and perhaps human’s) mental and physical activity levels. Inputs inhibiting VTAVgat neurons intensify wakefulness (increased activity, enhanced alertness and motivation), qualities useful for acute survival. In the extreme, however, decreased or failed inhibition from VTAVgat neurons produces mania-like qualities (hyperactivity, hedonia, decreased sleep).

Both these themes come together in the ventral tegmental area (VTA). The VTA an important source of dopamine, regulates goal-and reward-directed and social behaviors [25,26], as well as wakefulness and sleep [27][28][29]. Exciting VTA dopamine neurons with well-chosen rhythms can produce mania-like behaviors in the day and euthymia at night [30]. In addition to dopamine neurons, there is a rich heterogeneity of glutamate and GABA neurons in the different anatomical subdomains of the VTA, and neurotransmitter co-release (e.g., dopamine -glutamate, GABAglutamate, GABA-dopamine) from VTA neurons is common [25,[31][32][33].
Here, we characterize the type of wakefulness produced by inhibiting or lesioning the VTA Vgat neurons or exciting VTA Vglut2 neurons. We find that the wakefulness induced by diminishing or removing VTA Vgat neuron inhibition contains behavioral endophenotypes that are mania-like, and are treatable with valproate and diazepam, although not with lithium or lamotrigine. On the other hand, the extended but quiet wakefulness produced by activating midline VTA Vglut2 neurons has no endophenotypes characteristic of mania. We suggest that the mania-like symptoms resulting from diminished VTA Vgat function are generated by changing the E-I balance in both the VTA and the LH.

Drug treatments
For all drug treatments, mice first received vehicle injections, and behavioral tests were performed; 2 weeks later, the same group of mice were given drug injections, and behavioral tests were performed. For the long-term valproate treatments, we further tested mice behaviors 2 weeks after valproate treatment was withdrawn. Baseline behaviors of the mice were tested during this period without any treatment.
Chemogenetics CNO (C0832, Sigma-Aldrich, dissolved in saline, 1 mg/kg) or saline was injected i.p. 30 min before the start of the behavioral tests. For VTA Vgat -hM4Di mice or VTA Vglut2 -hM3Dq mice, CNO or saline was injected during the "lights off" active phase. 1 μl of 1 mM CNO was infused into the LH through the guide cannula at a rate of 0.1 µl min −1 with an injector needle (33-gauge, Hamilton) D-amphetamine D-Amphetamine (2813/100, Tocris Bioscience, dissolved in saline, 2 mg/kg) or saline was injected i.p. into the mice [13,21], and the mice were assessed directly after Damphetamine injection.
Valproate, lithium, diazepam, and lamotrigine Sodium valproate (2815, TOCRIS, dissolved in saline, 200 mg/kg), LiCl (Sigma-Aldrich, dissolved in saline, 100 mg/kg) or saline (vehicle) was injected i.p. into the mice. For acute treatments (repeated injections) with lithium or valproate, mice received vehicle injections, and then behaviors (locomotion and tail-suspension) were tested; after 2 weeks, the same group of mice were given lithium or valproate injections. For the Li-H 2 O treatment, mice were treated with LiCl in drinking water (300 mg/L) for 2 weeks. Mice were first given normal water, and behaviors were tested, and the same group of mice were placed on Li-H 2 O for 2 weeks. Valproate or LiCl treatments were as previously reported [13,17]: valproate or LiCl were injected three times (10:00, 14:00 and 17:00) one day before the behavioral assays; and during the day of the behavioral assay, valproate or LiCl was injected two times (10:00, 14:00). Locomotion or tail-suspension tests were performed 30 min after injection (14:30). After the locomotion or tailsuspension tests (see below), mice received a final valproate injection (17:00), and then the 24-h sleep-wake recordings were then performed (see section "EEG analysis, sleep-wake behavior and sleep deprivation").

Serum lithium
Blood samples were collected before and after lithium injections via transcutaneous cardiac puncture and the serum was separated for determination of lithium levels. Mice were maintained on anesthesia and blood samples were collected directly from the left ventricle before killing them. Serum lithium was analyzed using a Roche Cobas c311 analyzer.

Dopamine receptor antagonists experiment
Dopamine antagonists SCH-23390 (0.03 mg/kg, dissolved in saline) and raclopride (1 mg/kg, dissolved in saline), for D1 and D2/D3 receptors, respectively, were injected serially i.p. into the VTA Vgat -mCherry mice or VTA Vgat -CASP3 mice 20 min before the locomotion, tail-suspension test (TST), or sleep experiment. For the experiments with chemogenetic inhibition combined with dopamine receptor antagonists, SCH-23390 and raclopride were injected serially i.p. into the VTA Vgat -hM4Di mice, and 20 min later, saline or CNO was injected into the antagonists-injected mice, and 30 min later, the locomotion test and TST was performed.

Locomotor activity
The locomotor activity and time spent in stereotypy was detected in an activity test chamber (Med Associates, Inc) with an ANY-maze video tracking system (FUJIFILM co.) and measured by ANY-maze software (Stoelting Co. US.). VTA Vgat -mCherry or VTA Vgat -CASP3 mice were directly put into the activity test chamber; for VTA Vgat -hM4Di mice, the mice were put into the test chamber 30 min after saline or CNO injection; for D-Amphetamine experiments, mice were first put into chamber for 30 min, and the mice received vehicle or amphetamine injections. The locomotor activity was detected straight after injection for 1-h.

Home cage activity
Mice were habituated in the cage for 24 h with mock Neurologgers before recordings. The home cage activities were then recorded using the accelerometer built into the Neurologger 2A devices [44] for a 24-h period and analyzed using Spike2 software.

Tail-suspension test (TST)
The TST was performed as described [13]. After a habituation period in the test room, mice were suspended 60 cm above the floor by their tails by taping the tail tip. Behaviors were video recorded and blindly scored manually and measured by ANY-maze software (Stoelting Co. US).

Forced swimming test (FST)
Mice were placed in a borosilicate glass cylinder (5 L, 18 cm diameter, 27 cm high) filled with water (25°C, water depth 14 cm) for 6 min. The immobility time during the last 4 min was manually measured. Immobility time was defined as the time spent without any movements except for a single limb paddling to maintain flotation.

Sucrose preference test (SPT)
This was a 2-choice test between 1% sucrose and water. Mice were habituated to a dual delivery system (one bottle with water and one bottle with 1% sucrose) for 3 days. Sucrose preference was then assessed over 3 consecutive days. For the chemogenetic experiments, animals were water-restricted overnight before saline or CNO injection. 30 min after saline or CNO injection, the mice were given free access to the 2-water delivery system for 4 h. Sucrose preference (%) was calculated as (weight of sucrose consumed)/(weight of water consumed + weight of sucrose consumed) × 100%.

Elevated plus maze test
The elevated plus maze apparatus (Global Biotech Inc. Shanghai) was opaque and consisted of a central platform (10 cm × 10 cm), two open arms (50 cm × 10 cm), and two closed arms (50 cm × 10 cm) with protective walls 40 cm high, which was 70 cm above the ground. Animals were placed in the central platform facing one open arm of the apparatus and were free to explore the arms for 5 min. The apparatus was cleaned with 75% ethanol before and after each session. The traces were recorded by an overhead camera and shown by average heatmap of each group. The ratio of exploring time in the open arm was calculated by Video Tracking Software (ANY-maze, Stoelting Co., Ltd.).

Timing of behavioral tests
10-12 weeks male mice were given virus injection (see "Surgery" section), and 4-6 weeks after this, behavioral tests were performed (the mice were~14-18 weeks old by this stage), or drug treatments were started. During the drug treatment experiments, after vehicle injections, behavioral tests were performed (the mice were~14-18 weeks old); and 2 weeks later, mice received drug injection, and behavioral tests were performed (by this stage, the mice werẽ 16-20 weeks old). All behavioral tests (locomotion, TST, forced swimming test (FST), or sucrose preference test (SPT)) took place during the "lights-off" phase of the lightdark cycle, when the mice were most active, particularly between 14:00 p.m.-17:00 (2 p.m.-5 p.m.), except for the sleep deprivation experiments (see section "EEG analysis, sleep-wake behavior, and sleep deprivation").

Mouse groupings for behavioral tests
For chemogenetic experiments, VTA Vgat -hM4Di mice were randomly split into two groups that received saline or CNO injection, and the locomotion test, TST, FST, or SPT were then performed. After 1-2 weeks, the same mice were given CNO or saline injection, and the locomotion test, TST, FST, or SPT were again performed. For the saline control experiments, VTA Vgat -hM4Di mice received saline injections, and then the locomotion test, TST, FST, or SPT were performed, and after 1-2 weeks, the same mice were given saline injection, and the locomotion test, TST, FST, or SPT were performed. EEG analysis, sleep-wake behavior, and sleep deprivation EEG and EMG signals were recorded using Neurologger 2A devices [28,44]. NREM sleep and wake states were automatically classified using a sleep analysis software Spike2 and then manually scored. The sleep deprivation protocol was as described previously [45]. At the start of "lights on" period, when sleep pressure is the highest, mice fitted with Neurologgers were put into novel cages, and at 1-h intervals, novel objects were introduced. After 5 h sleep deprivation, mice were then put back into their home cages for 19 h. In total, a 24-h sleep-wake state was recorded. For the sleep experiments with dopamine receptor antagonists, mice were given SCH-23390 (0.03 mg/kg) and raclopride (2 mg/kg) injections in the middle of the "lights off" active period.

Optogenetic stimulation
For the optogenetic behavioral experiments, a fiber patch cord was connected to the laser generator, and dual optic fibers were connected to the fiber patch through a rotary joint (ThinkerTech Nanjing BioScience Inc. China.). Before the experiments, a monofiberoptic cannula was connected to the fiber patch cord. VTA Vgat -eNpHR-mCherry→LH mice or VTA Vgat -ChR2-mCherry→LH mice were bilaterally optostimulated (20 Hz, 0.5 s duration with 0.5 s interval, 593 nm, 20 μW) in the LH during the "lights off" phase. The stimulation was given for the duration of the behavioral tests.

Slice electrophysiology
Three weeks after injection of AAV2/9-hSyn-DIO-hM3Dq-mCherry into the VTA of VGlut2-Cre mice, brain slices containing the VTA were prepared for electrophysiological recordings. Mice were anesthetized with isoflurane and then killed. Brains were rapidly removed and placed in ice-cold oxygenated cutting ACSF (containing in mM: 252 sucrose, 2.5 KCl, 6 MgSO 4 , 0.5 CaCl 2 , 1.2 NaH 2 PO 4 , 2.5 NaHCO 3 , 10 glucose). Horizontal VTA slices (300 μm) were cut using a vibratome (VT1200S, Leica). Brain slices containing the VTA were transferred to recording ACSF with 124 mM NaCl, 2.5 mM KCl, 1.2 mM NaH 2 PO 4 , 2 mM MgSO 4 , 2 mM CaCl 2 , 24 mM NaHCO 3 , 5 mM HEPES and 12.5 mM glucose. Brain slices were incubated at 34°C for 30 min and were then kept at room temperature under the same conditions for 45 min before transfer to the recording chamber at room temperature (22-25°C). The ACSF was perfused at 2 mL/min. The brain slices were visualized with a fixed upright microscope (BX51WI, Olympus) equipped with a water immersion lens (40×/0.8 W) and a digital camera (C13440, Hamamatsu). Patch pipettes were pulled from borosilicate glass capillary tubes using a pipette puller (P97, Sutter). The resistance of pipettes varied between 3 and 5 MΩ.

Quantification and statistics
Statistical tests were run in "Origin 2019b" (Origin Lab). The individual tests, and the number (n) of mice for each experimental group/condition, are given in the figure legends. All data are given as mean ± SEM and 'center values' are the means. We tested for normality and equal variances. When the data were non-normal, we used nonparametric tests (stated in relevant figure legends). Mice were assigned randomly to the experimental and control groups. For chemogenetic experiments, saline or CNO injection was blinded. For analyzing chronic lesioning experiments, VTA Vgat -mCherry and VTA Vgat -CASP3 mice, experimenters were blinded. For drug treatments, vehicle or drug injections were not done blinded. Experimental data analysis, including animal behavior that was scored from videos and the analysis of EEG data, was done blinded.

Results
Chemogenetic inhibition or lesioning of VTA Vgat neurons produces increased locomotor activity and sensitivity to amphetamine We delivered AAV-DIO-hM4Di-mCherry into the VTA of Vgat-ires-cre mice to express hM4Di-mCherry (an inhibitory receptor activated by the CNO ligand [40]) specifically in VTA Vgat neurons to generate VTA Vgat -hM4Di mice (Fig. 1a) [28]. As previously [28], we also ablated VTA Vgat neurons by injecting AAV-DIO-CASP3 and/or AAV-DIO-mCherry (as a control) into the VTA of Vgat-ires-cre mice to generate VTA Vgat -CASP3 mice and VTA Vgat -mCherry mice respectively. After 5-6 weeks post-injection, mCherrypositive VTA Vgat neurons were detected in control mice that had only received AAV-DIO-mCherry injections, whereas mCherry-positive VTA Vgat neurons were mostly absent in caspase-injected mice (Fig. 1b), but tyrosine hydroxylase-positive (dopamine) neurons were still present (Fig. 1b). We next assessed locomotion behaviors of these animals. After CNO injection, VTA Vgat -hM4Di mice had higher locomotor activity in an open-field arena with more distance traveled (Fig. 1c) and had elevated average and maximum speeds over 30 min compared with saline-injected VTA Vgat -hM4Di mice (Supplementary Fig. 1a). As a control, the locomotor baseline activity of VTA Vgat -hM4Di mice did not differ over the extended experimental period: mice that received an initial saline injection, and then another saline injection two weeks later, had the same locomotor activity ( Supplementary Fig. 1b). Consistent with these chemogenetic inhibition results, VTA Vgat -CASP3 mice also produced hyperlocomotion in the open field-more distance traveled (Fig. 1d), and higher average and maximum speeds over 30 min compared with the distance traveled by VTA Vgat -mCherry control mice ( Fig. 1d; Supplementary  Fig. 1c). VTA Vgat -CASP3 mice were also hyperactive in their home cages (Supplementary Fig. 1d). Notably, the hyperlocomotion of the VTA Vgat -CASP3 mice was still seen when the mice were retested in the open field four months post-lesion ( Supplementary Fig. 1e). By the end of this period, the body weight of VTA Vgat -CASP3 mice was reduced compared with that of VTA Vgat -mCherry control mice ( Supplementary Fig. 1f).
Mice suggested to have mania-like characteristics, or patients with bipolar-related mania, are hypersensitive to Damphetamine [13,21], whereas mice and patients posited to have an attention deficit hyperactivity (ADHD)-like disorder become less active with D-amphetamine [46,47]. Treating CNO-injected VTA Vgat -hM4Di mice and VTA Vgat -CASP3 mice with D-amphetamine significantly increased their locomotion speed above their high baseline speed, and increased the distance traveled (Fig. 1e, f) and the time spent in stereotypy ( Supplementary Fig. 1g, h), suggesting that both groups of mice (VTA Vgat -CASP3 and CNO-injected VTA Vgat -hM4Di mice) were not ADHD-like.

Acute inhibition or chronic lesioning of VTA Vgat neurons elevates mood
We assessed mood-related behaviors using TST, FST, SPT, and the elevated plus maze. During the TST, the immobility times of CNO-injected VTA Vgat -hM4Di mice and VTA Vgat -CASP3 mice were greatly decreased compared with saline-injected VTA Vgat -hM4Di control mice and VTA Vgat -mCherry (Fig. 2a, b). Both CNO-injected VTA Vgat -hM4Di mice and VTA Vgat -CASP3 mice also had decreased immobility times during the FST compared with saline-injected or VTA Vgat -mCherry controls (Fig. 2c, d). This could suggest that reducing VTA Vgat neuronal output, by either chemogenetic inhibition or lesioning, produces less depressive-like behaviors. In addition, during the SPT, CNO-injected VTA Vgat -hM4Di mice and VTA Vgat -CASP3 mice consumed more sucrose than their control littermates (Fig. 2e, f), possibly indicating a raised hedonic state. As a control, the baseline behavior of VTA Vgat -hM4Di mice did not differ over the extended experimental period: mice that received an initial saline injection, and then another saline injection 2 weeks later, had the same performance during the TST, FST, or SPT ( Supplementary Fig. 1i). We characterized the anxiety-like behavior of our mouse models using the elevated plus maze. The time spent in the open arms of the maze by both CNO-injected VTA Vgat -hM4Di mice and VTA Vgat -CASP3 mice increased (Fig. 2g, h), suggesting their anxiety was reduced.
Mice with lesioned VTA Vgat neurons have less sleep need after sleep deprivation VTA Vgat -CASP3 mice have a long-term sleep deficit [28]. To further characterize the need of VTA Vgat -CASP3 mice for sleep, we performed a sleep deprivation experiment. After sleep deprivation, there is usually a rebound in lost NREM sleep, a process termed sleep homeostasis [48]. Mice were kept awake for 5 h with novel objects presented each hour and were then tested to see if they caught up on lost sleep in their home cage. Control VTA Vgat -mCherry mice had rebound NREM sleep after 5 h sleep deprivation, and thereby after 19 h they had regained 90% of their sleep loss (Fig. 2i). Surprisingly, the VTA Vgat -CASP3 mice, despite already starting from a chronic sleep-deprived baseline, did not catch up on lost NREM sleep after 5 h continuous sleep deprivation (Fig. 2i).

The mania-like behavior produced by lesioning VTA Vgat neurons can be pharmacologically rescued by valproate and diazepam but not lithium or lamotrigine
We assessed if lithium could treat the manic-like state of VTA Vgat -CASP3 mice. We confirmed that serum lithium was elevated to therapeutic levels during the acute treatment ( Supplementary Fig. 2a), and that the behavioral baselines of VTA Vgat -mCherry or VTA Vgat -CASP3 mice did not change during the treatment period ( Supplementary  Fig. 2b-f). However, neither acute (100 mg/kg) (Supplementary Fig. 3a-e) nor chronic (300 mg/L) treatments ( Supplementary Fig. 3f-j) had any effects (Of note, chronic Li-H 2 O treatment decreased the immobility time of control mice during the TST (Supplementary Fig. 3i)).
We next examined valproate (200 mg/kg) treatments (Fig. 3a). Acute injection of valproate did not affect the locomotor activity of VTA Vgat -mCherry control mice (Fig. 3b). By contrast, the hyperlocomotor activity of VTA Vgat -CASP3 mice in the open field was restored down to control levels by valproate treatment (Fig. 3c). However, 2 weeks after the acute valproate treatment was withdrawal from VTA Vgat -CASP3 mice, their hyperactivity in the open field had returned (Fig. 3c). Similarly, during the tail vsuspension test, valproate treatment did not affect VTA Vgat -mCherry control mice (Fig. 3d), but significantly increased the immobility time of VTA Vgat -CASP3 mice back up to control levels (Fig. 3e). Two weeks after valproate had been removed, however, the abnormally high agitation of VTA Vgat -CASP3 mice had re-emerged (Fig. 3e).
We assessed if valproate could reduce the sustained wakefulness of VTA Vgat -CASP3 mice. The wake time of VTA Vgat -CASP3 mice significantly decreased after treatment with valproate, whereas NREM sleep time substantially increased (Fig. 3f). VTA Vgat -CASP3 mice also have a pathological sleep architecture with fewer episodes of wake and NREM sleep, prolonged duration of each wake episode, and substantially decreased numbers of wake-NREM transitions [28]. Valproate treatment normalized the sleep-wake architecture of the VTA Vgat -CASP3 mice: the episode number ( Supplementary Fig. 4a), episode duration of wake was restored to control levels ( Supplementary  Fig. 4b), as were the number of transitions between wake and NREM sleep (Supplementary Fig. 4c).
We next tested if a low dose of the GABA A receptor positive allosteric modulator diazepam, an anxiolytic, influenced the behavior of VTA Vgat -CASP3 mice (Supplementary Fig. 5a-j). Treatment of VTA Vgat -CASP3 mice with either single or repeated doses of 1 mg/kg diazepam partially reduced the hyperlocomotion ( Supplementary  Fig. 5c, h) and fully recovered the immobility time during the TST (Supplementary Fig. 5e, j). These results suggest that diazepam has a potential treatment effect on VTA Vgat -CASP3 mice.
Note: with the exception of chronic lithium treatment, none of these drug treatments affected the overall body weight of the mice during and after treatments (Supplementary Fig. 7a-d); body weight was reduced during chronic lithium treatment (Supplementary Fig. 7c).

The extended wakefulness produced by activating VTA Vglut2 neurons is not mania-like
Activating VTA Vglut2 neurons also promotes wakefulness [28]. To investigate if this type of artificially-induced wake is also mania-like, we expressed the excitatory chemogenetic hM3Dq receptor in VTA Vglut2 neurons (VTA Vglut2 -hM3Dq mice; Fig. 4a) [8]. We prepared acute brain slices containing the VTA from VTA Vglut2 -hM3Dq mice, and first confirmed that CNO application triggered action potentials in hM3Dq-mCherry-expressing cells in VTA glutamatergic neurons (Fig. 4b). Chemogenetic activation of VTA Vglut2 neurons by CNO injection did not alter the locomotor activity or distance traveled of VTA Vglut2 -hM3Dq mice compared with saline injected mice (Fig. 4c) (see also data in 28). Moreover, the immobility time during the TST and FST did not differ between CNO injected-and saline injected VTA Vglut2 -hM3Dq mice (Fig. 4d, e). This data serves as an internal negative control, showing that not all artificially induced wakefulness in mice will be mania-like: the activation of VTA Vglut2 neurons did not produce hyperactivity and mood-related deficits, suggesting a different kind of wakefulness from that generated by inhibition of VTA Vgat neurons.
VTA Vgat -hM4Di mice (Fig. 5a), but only had a partial effect on VTA Vgat -CASP3 mice (Fig. 5b). In the TST, dopamine receptor antagonists increased the immobility time of both saline-or CNO-injected VTA Vgat -hM4Di mice and VTA Vgat -mCherry control or VTA Vgat -CASP3 mice (Fig. 5c, d). However, the immobility time of CNO-injected VTA Vgat -hM4Di mice or VTA Vgat -CASP3 mice that received dopamine receptor antagonist injection was still significantly lower than saline-injected VTA Vgat -hM4Di mice or VTA Vgat -mCherry mice injected with dopamine receptor antagonists (Fig. 5c, d). We also found that the extended wakefulness of VTA Vgat -CASP3 mice was substantially attenuated by the dopamine receptor antagonists ( Supplementary Fig. 8a, b). The above results suggest that blocking dopamine signaling substantially reduces the mania-like behaviors of VTA Vgat -CASP3 mice. VTA Vgat neurons project prominently to the LH [28]. We conducted optogenetic and chemogenetic experiments to examine if the VTA Vgat to LH projection contributes to hyperactivity and mood-related behaviors. An AAV carrying a Cre-dependent eNpHR3.0-mCherry transgene was injected into the VTA of Vgat-ires-cre mice to express inhibitory halorhodopsin in VTA Vgat neurons to generate VTA Vgat -NpHR mice (Fig. 5e). Optic fibers were placed above the LH of VTA Vgat -NpHR mice, where dense NpHR-mCherry projections arising from the VTA Vgat cell bodies can be seen (Fig. 5e). Optogenetic inhibition of the VTA Vgat to LH projection, by activating NpHR-mCherry in the LH, elevated the locomotion of mice in the open field (Fig. 5f), whereas it decreased the immobility times during the TST (Fig. 5g) and FST (Fig. 5h). We confirmed this using specific chemogenetic inhibition of the VTA Vgat to LH projection. CNO was infused directly into the LH of VTA Vgat -hM4D i mice, which expressed the inhibitory CNO receptor in VTA Vgat neurons ( Supplementary Fig. 9a). CNO infusion increased locomotion speed ( Supplementary Fig. 9b), and decreased immobility time during the TST ( Supplementary  Fig. 9c). These results indicate that inhibiting GABAergic tone from VTA Vgat neurons to the LH produces hyperlocomotion and less depressive-like behaviors. We next injected AAV-DIO-ChR2-mCherry into the VTA of Vgatcre mice to express excitatory channel rhodopsins in VTA GABAergic cells. Although opto-activation of the VTA Vgat to LH projection did not affect the locomotion of mice in the open field (Fig. 5i), it did increase the immobility times during the TST (Fig. 5j) and FST (Fig. 5k). The above results suggest that the VTA Vgat to LH projection, together with VTA Vgat local inhibition of the dopamine system, contributes to generating mania-like behaviors.

Discussion
We found that endophenotypes (hyperactivity, elevated mood, and reduced sleep) resembling aspects of mania could be produced by lesion or acute inhibition of GABAergic VTA Vgat neurons. In mice with lesioned or inhibited VTA Vgat neurons, D-amphetamine further increased the hyperactivity, suggesting the mania-like behavior of these mice could be the type associated with bipolar disorder in humans. The mania effects were partially blocked by D1, D2, and D3 receptor antagonists. Thus, hyperdopaminergia could originate from VTA Vgat cells failing to inhibit dopamine neurons (Fig. 5l). VTA Vgat cells, however, are not simply VTA inhibitory interneurons that reduce dopamine's actions. Part of their actions in promoting increased movement in the FST and TST depends on their inhibitory projections to the LH. Our results on hyperactivity produced by disinhibiting LH circuits are consistent with previous findings. In rats, repeated delivery of subthreshold stimuli (kindling) to the LH induces mania-like behavior [49]. Indeed, the LH  promotes physical activity and motivated behavior [50]; LH lesions in rodents and people produce a state of wakefulness with no motion [51]. The LH contains orexin/glutamate neurons that promote arousal, motivation and energy expenditure [52], and GABAergic neurons whose activation induces wakefulness and locomotion [50,53,54]. It is likely that the VTA Vgat neurons inhibit both the orexin neurons and/or the arousal/locomotor-promoting GABA neurons in the LH [28,36]. Given the hub-like nature of the GABAergic VTA Vgat neurons, the mania phenotype produced by lesioning or inhibiting these neurons is likely a consequence of the disinhibition of multiple targets, e.g., dopamine cells in the VTA, but also to some extent other cell types in the LH. On the other hand, VTA Vgat neurons also locally inhibit VTA glutamate neurons [28], but disinhibition of glutamate neurons in the VTA by lesioning or inhibiting the VTA Vgat neurons probably does not contribute to the mania-phenotype. Direct pharmacogenetic excitation of these glutamate cells produces calm wakefulness (Fig. 4) [28].
Inability to sleep is one of the diagnostic criteria for the mania phase of bipolar disorder [2,3,55], and certainly VTA Vgat mice sleep consistently less than control mice, such that they have 100% wakefulness during the "lights off" phase compared with control mice that take about 4 h NREM sleep during this time [28]. Usually, the longer that wakefulness persists, the stronger the urge to sleep becomes, until sleep is inescapable. A remarkable finding is that VTA Vgat -CASP3 mice bypassed the homeostatic process of having NREM recovery sleep after sleep deprivation -they did not catch up on lost sleep in spite of already starting from a strongly sleep-deprived background. The mechanisms underlying sleep homeostasis are not well understood [48,56,57]. Sleep homeostasis is thought to reflect the function of sleep, (otherwise why catch up on lost sleep?), and it will be interesting to investigate if the chronic lack of sleep of VTA Vgat -CASP mice will be detrimental metabolically.
Sleep deprivation can sometimes trigger mania episodes in humans [3,55,58]. Consequently, an interesting question is whether the chronic sleep deprivation phenotype of VTA Vgat -CASP3 mice actually causes their mania-like symptoms. Mice chronically sleep derived with the flower pot method-the animals stay on top of a raised platform surrounded by water, and when they fall asleep, they fall into water and wake up-also develop mania-like behavior [59]. This type of sleep deprivation is, however, likely to be stressful. In the VTA Vgat -CASP3 mice, the sleepdeprivation phenotype is internally generated within the brain, and could be less stressful per se.
The mania-like behaviors of VTA Vgat -CASP3 mice, including the strong sleep loss and abnormal sleep Fig. 5 Circuit mechanisms underlying the mania-like behaviors mediated by VTA Vgat neurons. a Locomotion speed and distance traveled of VTA Vgat -hM4Di mice that received injections with vehicle or a dopamine receptor antagonist mixture (SCH-23390 and raclopride for D1 and D2/D3 receptors, respectively) and either saline (n = 7 mice) or CNO (n = 7 mice). Two-way ANOVA and Bonferroni-Holm post hoc test. F (saline or CNO) = 13.6; F (vehicle or antagonists) = 91.2; F (interaction) = 7.4; Vehicle + saline vs. antagonists + saline t(12) = 5, ***p = 0.0002; Vehicle + CNO vs. antagonists + CNO t(12) = 8.34, ****p = 2E−6. All error bars represent the SEM. b Locomotion speed and distance traveled of VTA Vgat -mCherry and VTA Vgat -CASP3 mice that received injections with either vehicle (n = 6 mice for mCherry and n = 10 mice for CASP3) or a dopamine receptor antagonist mixture (n = 6 mice for mCherry and n = 13 mice for CASP3). Two-way ANOVA and Bonferroni-Holm post hoc test. F (mCherry or CASP3) = 10.8; F (saline or antagonists) = 4.5; F (interaction) = 0.03; mCherry + vehicle vs. mCherry + antagonists t(10) = 4.14, **p = 0.002; CASP3 + vehicle vs. CASP3 + antagonists t(21) = 1.64, p = 0.11. All error bars represent the SEM. c Time spent immobile for the tail suspension assay of VTA Vgat -hM4Di mice that received injections of either vehicle or the dopamine receptor antagonist mixture and saline or CNO. Two-way ANOVA and Bonferroni-Holm post hoc test. F (saline or CNO) = 36; F (Vehicle or antagonists) = 27; F (interaction) = 0.35; vehicle + saline vs. antagonists + saline t(12) = −7.59, ****p = 6E−6; vehicle + CNO vs. antagonists + CNO t(12) = −2.54, *p = 0.02; antagonists + saline vs. antagonists + CNO t(12) = 3.8, **p = 0.002. All error bars represent the SEM. d Time spent immobile for the tail suspension assay of VTA Vgat -mCherry and VTA Vgat -CASP3 mice that received injections of either vehicle (n = 8 mice for mCherry and n = 9 mice for CASP3) or the dopamine receptor antagonist mixture (n = 8 mice for mCherry and n = 9 mice for CASP3). Two-way ANOVA and Bonferroni-Holm post hoc test. architecture, were reversed by valproate, and re-emerged when valproate treatment was stopped. Valproate has diverse actions: for example, it enhances GABAergic transmission and reduces action potential firing (reviewed in ref. 13), but by inhibiting histone deacetylases [60,61], valproate treatments also change the expression of many genes. We found that low doses of diazepam were also efficacious in reducing the hyperarousal of VTA Vgat -CASP3 mice. Diazepam a GABA A receptor positive allosteric modulator, enhances GABA transmission [62]. At the low doses tested here in mice, diazepam is an anxiolytic, but at higher doses (which we did not test) it induces sleep [62]. Two other important drugs used for treating bipolar disorder, lithium and lamotrigine, had no effect in treating VTA Vgat -CASP3 mice. Although its mechanism of action is unclear, lithium treatment is the first choice to treat mania episodes, although a subset of bipolar patients with rapidly cycling mania and depression phases are resistant to lithium (reviewed in ref 13). In most mouse models of mania, both valproate and lithium are usually effective treatments (reviewed in ref. 1). But similar to our results, mice with SHANK3 overexpression in the neocortex, hippocampus and basal ganglia, have mania-like symptoms treatable with valproate but not lithium [13]. Similar arguments could apply to lamotrigine. It could be that the VTA Vgat -CASP3 mice and SHANK3-overexpressing mice models reflect a specific mania type.
In summary, the model based on reducing VTA Vgat neuronal function could provide further insight into the genesis of some types of mania-like behaviors. One hypothesis is that VTA Vgat neurons help set the level of mental and physical activity. Physiologically, inputs that inhibit VTA Vgat neurons will transiently intensify aspects of wakefulness useful for acute success or survival: increased activity, enhanced alertness and motivation, reduced sleep (Fig. 5l). Taken to the extreme, however, pathology could emerge and decreased or failed inhibition from VTA Vgat neurons will produce mania-like qualities (Fig. 5l).