Crosstalk between AQP4-dependent ATP/Adenosine release and dopamine neurotransmission in depressive behavior during cocaine withdrawal

The brain abundantly expresses adenosine receptors, which are involved in the regulation of neural activity, blood flow, and inflammation. In a previous study using our originally developed adenosine biosensor, we reported that hippocampal astrocytes release ATP upon water influx from the water channel AQP4, which is degraded extracellularly to increase adenosine (Yamashiro et al., 2017). On the other hand, the interaction between adenosine and dopamine is widely known, and when adenosine release from astrocytes is altered by inflammation or other factors, abnormal dopamine neurotransmission and related ataxia and psychiatric disorders may develop. In the present study, we examined pathological changes in adenosine or dopamine release in depressive-like behavior that develops as a symptom of cocaine withdrawal. The results showed that A1 receptor inhibitors and AQP4 gene disruption suppressed depressive-like behavior. In the striatum, AQP4-dependent adenosine release inhibited dopamine release via A1 receptors, and cocaine inhibited dopamine release by increasing this adenosine release. In contrast, in the medial frontal cortex, AQP4-dependently released adenosine enhanced dopamine release via A1 receptors, and cocaine abolished this adenosine effect. Furthermore, adenosine action was restored in AQP4 knockout mice, suggesting that cocaine reduced A1 receptor function via AQP4-dependent adenosine. In conclusion, astrocytes modulate dopaminergic neurotransmission through AQP4-mediated adenosine release, and this disruption leads to depression-like behavior.


Introduction
Adenosine, a nucleic acid metabolite, modulates neural activity through receptor-mediated actions.
Adenosine action is mediated by G protein-coupled receptors, A1, A2A, A2B, and A3, with A1 and A2A receptors being expressed mainly in the brain (Chen et al., 2001). A1 receptors are widely expressed in the brain and are involved in presynaptic inhibition of glutamate and dopamine (Chen et al., 2013;Yabuuchi et al., 2006). In contrast, A2A receptors are Gs-coupled and are abundantly expressed in the basal ganglia, including the striatum, where they influence dopamine neurotransmission (Ferré et al., 1997) and are involved in the maintenance of synaptic plasticity as a cofactor for BDNF (Fontinha et al., 2008). In addition to this, adenosine has sleep-promoting, vasorelaxant, and anti-inflammatory effects in the brain (Haskó et al., 2008;Lazarus et al., 2011;Ralevic and Dunn, 2015).
Much remains unknown regarding the regulation of extracellular adenosine levels in the brain. Using strain cells that exhibit intracellular calcium elevation in response to adenosine as biosensors, our group found that adenosine release in hippocampal slices is mainly derived from two pathways (Yamashiro et al., 2017). Namely, the pathway in which dendrites of pyramidal neurons release adenosine in response to electrical activity and the pathway in which ATP is released in response to hypoosmotic treatment in a manner dependent on the astrocyte-specific expressed water channel AQP4, which is degraded by extracellular enzymes to increase adenosine. Because astrocytes express AQP4 abundantly (Nielsen et al., 1997), osmotic changes are thought to efficiently induce volume changes. In addition, almost all cells, including astrocytes, have high plasma membrane permeability for K + , so if anion channels are activated upon expansion, intracellular ions are quickly released and can adapt to osmotic changes. Such an osmotic adaptation mechanism is called regulatory volume decrease (RVD), and the anion channels that are activated upon expansion are known as volume-regulated anion channels (VRAC) (Okada et al., 2001). The AQP4-dependent ATP release upon hypoosmotic treatment reflects the RVD of astrocytes, and ATP is thought to be released via VRAC. However, since the gene for VRAC has not yet been identified and selective inhibitors have not yet been established, studies of VRAC have been limited to its electrophysiological properties.
The biosensor used in the previous study was a HEK293 cell expressing A1 receptor and Gqi5 (a fusion protein of Gi and Gq, which co-activates Gi-coupled receptors to elevate intracellular calcium) ), which reproducibly shows [Ca 2+ ]i elevation in response to adenosine but has problems with quantitation due to pronounced desensitization. To improve this point, a new biosensor, A1-CHO, was created by expressing the A1 receptor, Gqi5, and the calcium fluorescent protein RCaMP (Inoue et al., 2014) in CHO cells (Palmer et al., 1996), in which phosphorylation of the A1 receptor does not occur and confirmed that there was no desensitization (Masato Kobayashi, Master's thesis 2019). We also used the same method to create D2-CHO, a biosensor for dopamine. By using them, we achieved quantification of adenosine and dopamine with high quantitatively and without the need for calcium fluorochrome staining.
The interaction between adenosine and dopamine is widely known. Dopamine is a neurotransmitter that projects mainly from the ventral tegmental area and substantia nigra of the midbrain to the cerebral cortex, limbic system, and basal ganglia, and is involved in motivation, emotion, learning, motor regulation, and hormone regulation A2A receptors, like D1 and D2 receptors, are abundantly expressed in basal ganglia, including striatum (Chen et al, 2001), and A2A and D2 receptors form an antagonistic heterodimer (Hillion et al., 2002). This has already been exploited in the treatment of Parkinson's disease, an ataxia associated with decreased dopamine neurotransmission in the striatum, using blocking agents of A2A receptors to promote dopamine neurotransmission (Kondo and Mizuno, 2015). In contrast, A1 receptors inhibit dopamine release in the striatum and spinal cord (Acton et al., 2018;Ross and Venton, 2015).
Modulation of dopamine neurotransmission is a known cause of psychiatric disorders such as bipolar disorder and schizophrenia as well as Parkinson's disease. In particular, a number of studies have recently reported that neuroinflammation caused by stress and drug dependence decreases dopamine neurotransmission in the striatum, including the medial frontal cortex and the nucleus accumbens, causing depressive symptoms (Brites and Fernandes, 2015;Cotto et al., 2018;Furuyashiki, 2012).
On the other hand, astrocytes show diverse activation upon inflammation, and it is quite possible that adenosine release is altered in astrocytes activated by neuroinflammation. An example of increased expression of adenosine degrading enzymes in activated astrocytes leading to decreased adenosine and consequently to epilepsy due to increased glutamate release is already known (Fedele et al., 2005). In contrast, astrocyte atrophy has been reported in depressed patients (Zhao et al., 2018), and it is quite possible that an increase in adenosine, the opposite of epilepsy, occurs. Based on the above, the present study hypothesizes that an AQP4-dependent increase in astrocyte adenosine release suppresses dopaminergic neurotransmission and leads to the development of depression. To demonstrate this hypothesis, we used depression-like behavior in mice, which develops as a symptom of cocaine withdrawal (Filip et al., 2006), as a model and investigated adenosine and dopamine release using biosensors, focusing on the medial frontal cortex and striatum, the major dopamine nerve projection sites in the brain. In addition, AQP4 knockout mice were used to examine AQP4-dependent adenosine release as distinct from other release pathways. This study is expected to lead to new research for the diagnosis and treatment of depression if the actual conditions behind the changes in adenosine release and astrocyte activation that underlie depressive-like behavior are elucidated.

Animal experiments
All animal experiments were approved by the institutional animal care and use committee of Kobe University (Permission number: 25-09-04) and performed in compliance with the Kobe University Animal Experimentation Regulations. C57BLC background mice were maintained at 24℃ and under 12h light/dark cycle. Water and food were provided ad libitum. Mice at least 8 weeks old were used for experiment. AQP4 knock out mice (AQP4 -/-) (Kitaura et al., 2009) were provided by Dr. Mika Terumitsu (National Center of Neurology and Psychiatry, Tokyo, Japan). Depressive behavior were examined between 14 days to 17 days after five consecutive days intraperitoneal injection of 15 mg/kg cocaine. For forced swim test, mice were acclimatized in an experiment room with white noise for one hour, then placed in water (24℃), which was 12.5 cm high in a transparent cylinder of 22 cm in diameter (Can et al., 2012). 8 min RGB videos at 15 fps were taken by an web camera, C270n (Logicool, Tokyo, Japan). Mobility of mice were quantified as described by Gao et al (Gao et al. 2014).
Images in red color channel were converted to 8 bit gray-scale and each frame were subtracted by previous ones. Then pixels above a threshold of five digit were counted in each frame and averaged for 15 consecutive frames. Since this value reflect the mobility of mice in one second, immobility time were measured as the duration, in which the value was below a threshold determined by an experimenter for each test. The immobility time was plotted every 60 sec.

Calcium imaging
Slices were perfused with aCSF at 5 mL/min. RCaMP images of biosensor were obtained every one second by using an inverted microscope (IX-70, Olympus, Tokyo, Japan) equipped with an objective (UApo/340 20x / 0.70, Olympus) and a cooled-CCD camera (Orca-R2, Hamamatsu Photonics, Hamamtsu, Japan), and F/F0 was calculated by ImageJ. Electrical stimulation was given by house-made concentric electrode, isolator (ISO-Flex, Funakoshi, Tokyo, Japan) and pulse generator (Electronic Stimulator, Nihon Koden, Tokyo, Japan). For hypo-osmotic treatment, the aCSF of the osmolarity indicated were prepared by reducing NaCl concentration. Slices were pretreated with the low NaCl aCSF, which was supplemented with sucrose to achieve iso-osmolarity.

Statistical analysis
For statistical analysis by Excel 2016, 50 cells were randomly selected from each slice in Fig2BC and 5AC while eight responding cells were randomly selected in Fig3, 4, 5BD, 6, 7 and 8.

Results
The involvement of AQP4-dependent adenosine increases in depressive behavior during cocaine withdrawal.

Adenosine releases in the striatum
Ventral part of striatum including nucleus accumbens were subjected to the analysis of adenosine release by A1-CHO. A 30 Hz and 150 pulses electrical stimulation, which had successfully induced adenosine releases from hippocampal slices (Yamashiro et al., 2017)  suggest that exogenous adenosine, as well as endogenous adenosine tone suppress evoked-dopamine release by A1 receptor. The involvement of AQP4 in the adenosine tone, was examined by using AQP4 -/mice. As a result, dopamine release of AQP4 -/mice (AQP4 -/-) was 181.38±19.54% significantly larger than control (WT) (Fig. 3E). Moreover, DPCPX significantly reduced the dopamine release of Fig. 3F). AQP4 -/mice showed larger dopamine release, and the enhanced dopamine release of in the presence of DPCPX was lacked in AQP4 -/mice. Thus, it was suggested that dopamine release is tonically suppressed by AQP4-dependent adenosine increase and subsequent A1 receptor activation. The further suppression of dopamine release of AQP4 -/mice in the presence of DPCPX propose a facilitation of dopamine release by an A1 receptor subpopulation, which is activated by adenosine release through other pathway rather than the AQP4 dependent pathway as will be discussed later in medial prefrontal cortex results.

Reduced dopamine release during depressive behavior
Dopamine releases in the striatum during depressive behavior was examined. The dopamine release of cocaine-treated mice (WT / Cocaine) was significantly smaller to 68.35±8.53% of control (WT / (-)) ( Fig. 4A). DPCPX (WT / Cocaine / DPCPX) 137.99 ± 18.11% significantly increased the dopamine release of cocaine-treated WT mice (WT / Cocaine / (-)) as in the case of saline-treated WT (Fig. 4B). These results suggests that the suppression of dopamine release by endogenous adenosine via A1 receptor persists during depressive behavior. In contrast, the dopamine release of cocainetreated AQP4 -/mice (AQP4 -/-/ Cocaine) was not significantly different from control (AQP4 -/-/ (-)) suggesting the ablation of reduced dopamine release by cocaine in AQP4 -/mice (Fig. 4C). These results indicate the lack of depressive behavior and associated reduction of dopamine release in AQP4 -/mice. In addition, DPCX did not change the dopamine release of cocaine-treated AQP4 -/mice (AQP4 -/-/ Cocaine / DPCPX and AQP4 -/-/ Cocaine / (-)) ( Fig. 4D), indicating the loss of the A1 receptor-mediated suppression of dopamine release in cocaine treated AQP4 -/mice. From these results, cocaine is supposed to induce depressive behavior by the increase of AQP4-dependent adenosine increase and subsequent suppression of dopamine release by A1 receptors activation. Furthermore, A1 receptor antagonists likely suppressed depressive behavior by restoring adenosine tone in the striatum.

Adenosine releases in mPFC
Since the altered neural activities in the mPFC is known to cause depressive behavior (Furuyashiki, 2012), the AQP4-dependent adenosine release in the mPFC was analyzed by using A1-CHO. Both 150 pulses electrical stimulation at 30 Hz and -82.7 mOsM hypo-osmotic treatment induced adenosine releases in mPFC slices. The AUC after electrical stimulation did not show significant difference between AQP4 -/mice (AQP4 -/-) and control (WT) (Fig. 5A). In contrast, hypoosmotically-induced adenosine release of AQP4 -/mice (AQP4 -/-) was 64.14±9.24% significantly reduced to control (WT) (Fig. 5B). These results indicate that adenosine release in the mPFC is in line with our previous study using hippocampal slices, where both electrical stimulation and hypoosmotic treatment induce adenosine release and the hypo-osmotic release depends on AQP4 (Yamashiro et al., 2017). In order to determine the involvement of mPFC adenosine in depressive behavior, adenosine releases in cocaine-treated mice were examined. As results, both electrically and hypo-osmotically-induced adenosine release of cocaine-treated mice (WT / Cocaine) did not significant difference from those of control (WT / (-)) ( Fig. 5C,D). Thus, it was suggested that these adenosine releases in the PFC were not involved in depressive behavior.

The involvement of adenosine in dopamine release in the mPFC
Since the decreased dopamine neurotransmission in the mPFC during depressive behavior is proposed (Furuyashiki, 2012), the involvement of adenosine in evoked-dopamine release in the mPFC was examined by using D2-CHO. As in the striatum, the dopamine releases were induced by 40 pulses electrical stimulation at 30 Hz (arrow) in the mPFC. Exogenous adenosine 67.12 ± 8.13% significantly reduced dopamine release in the mPFC (WT / Ado) to control (WT / (-)) (Fig. 6A).

Downregulation of A1 receptor in the mPFC during depressive behavior
Dopamine releases in the mPFC during depressive behavior were examined by D2-CHO. The electrically induced dopamine releases in the mPFC slices from cocaine treated mice (WT / Cocaine) was not significantly different from control (WT / (-)) (Fig. 7A). This suggests the dopamine release in mPFC was not affected by the cocaine. However, DPCPX, which increased the dopamine releases in the mPFC of naïve mice, did not change the dopamine release in cocaine treated mice (WT / Cocaine / DPCPX and WT / Cocaine / (-)), suggesting the loss of the DPCPX induced increases of dopamine releases by cocaine (Fig. 7B). Since cocaine did not change adenosine release in the mPFC, the failure of DPCPX to change dopamine release in cocaine-treated mPFC is supposed to reflect downregulation of A1 receptors. This point was examined in AQP4 AQP4 -/mice. Cocaine treatment did not change dopamine release in the mPFC of AQP4 -/mice ( (AQP4 -/-/ (-) and AQP4 -/-/ Cocaine) (Fig. 7C).
AQP4-dependent adenosine tone facilitates dopamine release in the mPFC.
The involvement of AQP4-dependent adenosine tone in evoked-dopamine release in the mPFC was examined by measuring the evoked-dopamine release of AQP4 -/mice using D2-CHO. The dopamine release in the mPFC of AQP4 -/mice (AQP4 -/-) was not significantly different from wild type mice (WT) (Fig. 8A). The dopamine release in the mPFC of AQP4 -/mice in the presence of DPCPX, which decrease the dopamine release in mPFC of WT mice was also not different from control (AQP4 -/-/ (-)) ( Fig. 8B). These results indicate the lack of A1 receptor-mediated reduction of dopamine release in the mPFC of AQP4 -/mice. Since hypo-osmotically-induced adenosine increase was smaller in the mPFC of AQP -/mice, the loss of A1 receptor-mediated reduction of dopamine release is likely due to the loss of A1 receptor rather than the change of AQP4-dependent adenosine tone.
Two adenosine-dependent pathways regulating dopamine release.
The present discrepancy that both exogenous adenosine and DPCPX reduce dopamine release in mPFC was further examined. A1 receptor is commonly involved in pre-synaptic inhibition, but the inhibition of A1 receptor suppressed dopamine release in the mPFC. Pre-synaptic inhibition of dopamine release is mediated not only by A1 receptor, but also other receptors, such as metabotropic glutamate receptor (*) or GABAA receptor (Yonezawa et al., 1998). Thus, one possibility is that a subpopulation of A1 receptor inhibits glutamate or GABA release, which are simultaneously activated with dopamine release by electrical stimulation and pre-synaptically inhibit dopamine release.
To test this possibility, the effects of DPCPX in the presence of mGluR or GABAA receptors antagonist were examined. The effect of DPCPX was examined in the presence of GABAA receptor antagonist, 100 µM picrotoxin. In the presence of picrotoxin and DPCPX, the dopamine release (WT / PTX / DPCPX) was not significantly different from the control (WT / PTX / (-)) (Fig. 9A). Therefore, GABAA receptor antagonist occlude the effect of DPCPX. Since A1 receptors are mostly resided in excitatory synapses rather than inhibitory ones (Chen et al., 2013), DPCPX supposed to affect on glutamatergic synapses. Since Group II metabotropic glutamate receptors (Group II mGluRs) regulate dopamine release in the mPFC and nucleus accumbens (Cartmell et al., 2000;Gupta and Young, 2018), the possibility that DPCPX affects glutamatergic terminal involving Group II mGluR was examined.
The group II mGluR antagonist LY341495 was used to test this hypothesis and the dopamine release in the presence of 500 nM LY341495 and DPCPX and control (WT / LY / DPCPX and WT / LY / (-)) did not show significant difference (Fig. 9B). Hence, group II mGluR antagonist also occluded the effect of DPCPX on the dopamine release. If the DPCPX-sensitive glutamate release acts on ionotropic glutamate receptor and influences on dopaminergic neuron via neural circuit, AMPA receptor antagonist CNQX and NMDA receptor antagonist APV may affect. However, the dopamine release in the presence of DPCPX, together with 10 µM CNQX and 50 µM APV (WT / CNQX+APV / DPCPX) was 59.82 ± 4.12% significantly reduced to control (WT / CNQX+APV / (-)) ( Fig. 9C), excluding the involvement of ionotropic glutamate receptors. These results indicate that the DPCPX affects the glutamate and GABA releases, which act on group II mGLuR or GABAA receptor locating at the presynaptic terminal releasing dopamine, respectively. Finally, the A1 receptor down-regulation by cocaine treatment at dopamine release terminal, which is sensitive to exogenous adenosine in the mPFC. The dopamine release of cocaine treated mice in the presence of exogenous adenosine (WT / Cocaine / Ado) 73.79±6.82% significantly reduced to control (WT / Cocaine / (-)) (Fig. 9D). This result reflects that cocaine occluded the effect of DPCPX while did not influence on effect of exogenous adenosine. Overall, the dopamine release in the mPFC were regulated by two adenosinedependent pathways, one pathways involving glutamate and GABA was down-regulated bycocaine treatment, while another direct adenosine inhibition of dopamine releasing terminal was not affected.

Discussion
In this study, we examined changes in adenosine and dopamine release associated with depression-like behavior using cocaine withdrawal as a model. Evaluation of depression-like behavior by the forced swimming test suggested that AQP4-mediated adenosine release may cause the development of depression-like behavior via A1 receptors, because genetic disruption of AQP4 and A1 receptor inhibition suppressed depression-like behavior. In the striatum, adenosine release upon electrical stimulation was not detected, and adenosine release was detected only by a greater osmotic pressure difference than in other brain regions, and disruption of the AQP4 gene had no effect on adenosine release. This indicates that the dynamics of extracellular adenosine in the striatum, where A2A receptors are highly expressed, differs from that in the cortex and hippocampus. In contrast, slices of medial frontal cortex released adenosine upon electrical stimulation and hypoosmotic treatment, and the adenosine release upon hypoosmotic treatment was AQP4-dependent. This is consistent with two adenosine release pathways derived from neurons and astrocytes found in the hippocampus in a previous study (Yamashiro et al., 2017). In contrast, dopamine release by electrical stimulation was detected in both the striatum and medial frontal cortex. In the striatum, dopamine release was inhibited by adenosine, while it was enhanced by A1 receptor inhibitors, and cocaine decreased dopamine release while increasing adenosine release. This suggests that in the striatum, cocaine enhances A1 receptor activity by promoting adenosine release, resulting in decreased dopamine release. In contrast, in the medial frontal cortex, both adenosine and the A1 receptor inhibitor inhibited dopamine release, and cocaine administration abolished the effect of the A1 receptor inhibitor. AQP4 -/also abolished the effect of cocaine. Thus, in the medial frontal cortex, cocaine may deactivate A1 receptors via an AQP4mediated pathway, resulting in loss of endogenous adenosine modulation of dopamine release. Further investigation of the point at which A1 receptor inhibitors suppressed dopamine release in the medial frontal cortex, despite the fact that in many cases A1 receptors are involved in presynaptic inhibition, suggests that glutamate release modulated by A1 receptors may be mediated by GABAA receptors and gruop II mGluRs to inhibit dopaminergic, suggesting that A1 receptors may inhibit dopaminergic release via GABAA receptors and gruop II mGluRs. Both striatum and medial frontal cortex showed modulation of dopaminergic neurotransmission associated with AQP4 and adenosine, both changes consistent with the onset of depressive-like behavior. However, because inhibition of A1 receptors suppressed depressive-like behavior, the depressive-like behavior observed as a symptom of cocaine withdrawal may be attributed to a failure of the striatum, where adenosine release was increased and A1 receptor activity was enhanced, rather than to the medial frontal cortex, where A1 receptor dysfunction was observed.
The depression-like behavior due to cocaine withdrawal used in this study was quantified by administering cocaine to mice for 5 consecutive days, followed 14 days later by a forced swim test, using immobility time as an index. In preliminary experiments, immobility time did not increase after 7 days of cocaine administration. In AQP4 -/mice, cocaine-induced reduction of depressive behavior was observed. In the same mice, increased dopamine in the striatum measured by microdialysis and associated deficits in learning and motivation have been reported (Szu and Binder, 2016), consistent with the findings of this study that AQP4-mediated adenosine release inhibits dopamine neurotransmission consistent with the findings of this study that AQP4-mediated adenosine release inhibits dopamine neurotransmission. Acute administration of caffeine reduced cocaine-induced depression-like behavior. Caffeine has already been shown to have antidepressant effects, as it has been reported to decrease immobility time in a forced swimming test in normal mice (Szopa et al., 2016).The fact that acute administration of DPCPX decreased cocaine-induced depressive-like behaviors, as well as caffeine, suggests that caffeine is a inhibited depressive-like behavior by blocking A1 receptors. Similar to caffeine, DPCPX has also been reported to decrease immobility time in the forced swim test in normal mice (Szopa et al., 2018). Therefore, it is possible that the effects of caffeine and DPCPX on depression-like behavior in the present study may not have inhibited the changes induced by cocaine.
No adenosine release associated with electrical stimulation was detected in the striatum. This result is consistent with reports of very low detection success rates when similar measurements are made with fast scan cyclic voltammetry (Pajski and Venton, 2013). On the other hand, there have been several reports of detection of adenosine associated with electrical stimulation when adenosine in the striatum of individual animals is measured by microdialysis (Cechova and Venton, 2008;Cechova et al., 2010). A possible reason for the loss of adenosine release associated with electrical activity when sliced could be that blood flow is required for adenosine release via AQP4. In vivo, glucose consumption increases with neural activity, resulting in increased glucose uptake from capillaries via glucose transporters in astrocytes and glycogen degradation within astrocytes. These increases in glucose within astrocytes are thought to increase osmotic pressure, leading to increased water absorption via AQP4 localized at the site of contact between astrocytes and blood vessels, resulting in cell swelling and adenosine release. In contrast, in slices, glucose and water influx from capillaries to astrocytes did not occur because neurons were able to take up glucose directly from extracellular fluid.
Adenosine release from the striatum by hypo-osmotic treatment was not induced at -82.7 mOsM, which is effective in the frontal cortex, but required a concentration of -120 mOsM and was not affected by disruption of the AQP4 gene. Cocaine, on the other hand, increased this adenosine release, but whether this release is dependent on AQP4 has not yet been examined. It is quite possible that cocaine increased AQP4-dependent adenosine release.
In the striatum, exogenous adenosine inhibited dopamine release, while A1 receptor inhibitors enhanced dopamine release. In addition, AQP4 -/showed increased dopamine release. These results strongly suggest that adenosine released in an AQP4-dependent manner in the striatum constanly inhibits dopamine release. Presynaptic inhibition by A1 receptors on dopaminergic nerve endings has been reported in the striatum, supporting the above results (Yabuuchi et al., 2006). It has also been reported that blockers of A2A receptors do not affect dopamine release in the striatum, consistent with our results (Ross and Venton, 2015). On the other hand, we found a decrease in DPCPX-induced dopamine release in AQP4 -/striatum. This suggests that adenosine released independently of AQP4 may facilitate dopamine release. It is highly likely that AQP4-independent adenosine plays a major role in the altered neural activity in AQP4 -/-, especially since no difference in adenosine release between wild-type and AQP4 -/mice was observed.
Dopamine release in the medial frontal cortex was inhibited by exogenous adenosine, and this phenomenon was not affected by cocaine. On the other hand, A1 receptor inhibitors also suppressed dopamine release, and this phenomenon was abolished by cocaine, suggesting the involvement of GABAA receptors and group II mGluRs in the A1 receptor suppression, but the mechanism of action of exogenous adenosine is not clear. Future studies are planned to examine the effects of ion-channel glutamate receptor or GABAA receptor inhibitors on adenosine-induced dopamine release. If these inhibitors have no effect, then adenosine is acting directly on dopamine nerve endings. In that case, only A1 receptors on glutamatergic nerve endings would be activated by endogenous adenosine and deactivated by cocaine. The A1 receptors of dopaminergic nerve endings would not be deactivated by cocaine in either the striatum or medial frontal cortex, whereas they would be activated by endogenous adenosine in the striatum. Analysis of the fine structure of nerve endings and receptor localization would be necessary to test these possibilities.
In AQP4 -/-, dopamine release in the medial frontal cortex was the same as in wild-type mice, but in these mice, the A1 receptor inhibitor abolished the reduction in dopamine release; the differential effects of AQP4 -/and A1 receptor inhibitors also need to be examined in the future. -/-, it is quite possible that these neuronal circuits were affected as a result of decreased endogenous adenosine and increased glutamate and dopamine release. In this sense, the effects of inhibition of GABAA receptors and group II mGluRs in AQP4 -/should be examined.
In the medial frontal cortex, cocaine did not affect adenosine release by electrical stimulation or hypoosmotic treatment, nor dopamine release, but the inhibition of dopamine release by DPCPX was abolished. This suggests that depressive-like behavior may be related to A1 receptor dysfunction rather than adenosine release. AQP4 -/-, which did not exhibit cocaine-induced depressive-like behavior, also showed no change in dopamine release, and DPCPX suppressed dopamine release as in normal mice. This suggests that cocaine causes a reduction in A1 receptor function in the medial frontal cortex that is associated with depression-like behavior, and that this reduction is abolished in AQP4 -/-. In this connection, deep brain stimulation (Bekar et al., 2008) and electroconvulsive therapy (Sadek et al., 2011) have been reported to show antidepressant effects by inducing increased expression of A1 receptors in the cortex. In the hippocampus, it has also been reported that increased A1 receptors increase expression of Homer1a and promote resistance to depression (Serchov et al., 2015). In the nucleus accumbens, A1 receptor expression decreases during cocaine withdrawal (Toda et al., 2003), and a similar mechanism is expected in the cortex. In other words, cocaine may induce depressivelike behavior by decreasing A1 receptors through some pathway via adenosine release via AQP4 in astrocytes.
In the medial frontal cortex, dopaminergic innervation from the ventral tegmental area mainly inputs deep, while glutamatergic innervation mainly inputs near the surface (Van Eden et al., 1987;Jay et al., 1992;Kuroda et al., 1998), and GABAergic inter neurons mediate both (Yonezawa et al., 1998;Zhu et al., 2004). This suggests a glutamate-GABA-dopamine regulatory mechanism by endogenous adenosine. It has also been reported that continuous cocaine administration in the medial frontal cortex downregulates the expression of group II mGluRs (Huang et al., 2007), and the fact that the action of DPCPX was abolished by inhibitors of group II mGluRs suggests that cocaine suppressed the expression of group II mGluRs This study has demonstrated that cocaine abstinence is a key factor in the development of the DPCPX receptor, and that cocaine suppresses the expression of group II mGluRs.
The present study demonstrates that cocaine withdrawal-induced depressive-like behavior is accompanied by a suppression of dopamine release via A1 receptors due to increased adenosine release associated with hypoosmotic processing in the striatum and dysfunction of A1 receptors in the medial frontal cortex. Future work is needed to analyze the molecular mechanisms of AQP4 in the striatum and A1 receptors in the medial frontal cortex, including their expression and functional regulation. It has also been reported that adenosine release pathways differ among brain regions (Pajski and Venton, 2013), and detailed studies on the association between AQP4-independent adenosine and dopamine release will be necessary in the future. While currently widely used antidepressants mainly inhibit serotonin reuptake and increase extracellular serotonin concentration, this study suggests the possibility of a new antidepressant therapy based on dopamine and adenosine regulation. In recent years, a link between neuroinflammation and psychiatric disorders such as depression has been proposed, and it is quite possible that the activation of astrocytes associated with stress or drug-induced inflammation and the anti-inflammatory effects of adenosine may contribute to the pathogenic process.