The role of adenosine receptors in mood and anxiety disorders

Adenosine receptor subtypes, first described 40 years ago, are known to regulate diverse biological functions and have a role in various conditions, such as cerebral and cardiac ischemia, immune and inflammatory disorders and cancer. In the brain, they limit potentially dangerous over excitation, but also regulate mechanisms essential in sleep and psychiatric disorders. In this review, we discuss the role of adenosine receptors in mood and anxiety disorders. Activation of A2A receptors is associated with increased depression‐like symptoms, while increased A1 receptors signaling elicits rapid antidepressant effects. Indeed, several lines of evidence demonstrate that the therapeutic effects of different non‐pharmacological treatments of depression, like sleep deprivation and electroconvulsive therapy are mediated by A1 receptor up‐regulation or activation. In addition, A1 receptors may also play a role in the antidepressant effects of transcranial direct current stimulation and deep brain stimulation. As a potential downstream mechanism, which facilitates the antidepressant effects of A1 receptors, we propose a crosstalk between adenosinergic and glutamatergic systems mediated via synaptic plasticity protein Homer1a and α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic acid receptors. Moreover, adenosine receptors are also involved in the control of circadian rhythms, sleep homeostasis and some neuro‐immunological mechanisms, all of them implicated in mood regulation. Antagonists of adenosine receptors such as caffeine have general anxiogenic effects. In particular, A2A receptors appear to have an important role in the pathophysiology of anxiety disorders. Taken together, the results discussed here indicate that the adenosinergic system is involved in both the etiology and the treatment of mood and anxiety disorders.

Mood disorders including unipolar depressive and bipolar disorders are heterogeneous illnesses, which cause high individual suffering and impose a severe economic burden on society. It is today believed that depression has a complex multifactorial origin in which psychosocial factors interact with neuropsychological factors and a hereditary burden to induce alterations in mechanisms such as neuroplasticity, neurogenesis, and neuroimmunological regulation, the relative impact of which may vary in different subtypes of depressive syndromes (Krishnan and Nestler, 2010). Modern biochemical hypotheses of depression include e.g., alterations in FK506-binding protein (FKBP) 51, a co-chaperone regulating the glucocorticoid receptor (Fries et al., 2017), the central expression of corticotrophin releasing factor (Waters et al., 2015) or alterations in immune parameters (Wohleb et al., 2016). In recent years, the potential role of glutamate signaling in depression has received particular attention since it appears to mediate the rapid antidepressant effects of ketamine (Murrough et al., 2017;van Calker et al., 2018). Glutamate dysfunction in depression is suggested by genetic, post-mortem and in vivo neuroimaging data (Sanacora et al., 2008). On the other hand, facilitation of a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor-dependent glutamate signaling appears to mediate in addition to those of ketamine also the effects of several other antidepressant measures. These include e.g., increased signaling via A 1 receptors, sleep deprivation (SD) and of the muscarinic acetylcholine receptor antagonist scopolamine (Freudenberg et al., 2015;van Calker et al., 2018).
Depression is very often found comorbid with anxiety disorders. The sequenced treatment alternatives to relieve depression study discerned a prevalence of anxious depression of 46% (Fava et al., 2004), and a lower response to treatment in the comorbid group compared with the nondepression group has been identified (Fava et al., 2008;Domschke et al., 2010a). However, even when not comorbid with depression, anxiety disorders are among the most disabling conditions affecting up to 10% of the population (Craske and Stein, 2016) if not treated by pharmacotherapy (Koen and Stein, 2011) or psychotherapy (Otte, 2011). In the pathomechanism of anxiety disorders, both genetic (Gottschalk and Domschke, 2016) and psychological mechanisms such as childhood separation (Milrod et al., 2014) appear to be involved.
We have previously suggested a role of adenosine receptors in the regulation of mood (van Calker and Biber, 2005). However, reliable data indicating a potential role of the purines adenosine and adenosine triphosphate (ATP) in mental disorders have been obtained only recently (Yamada et al., 2014;Ortiz et al., 2015;Krugel, 2016;Cheffer et al., 2018). In this article, we will restrict our discussion to some selected aspects of adenosine receptor function in mood and anxiety disorders since the potential role of purine receptors in psychiatric illness in general has been comprehensively discussed recently (Krugel, 2016;Cheffer et al., 2018).

The adenosinergic system
Physiological effects of adenosine were first described by Drury and Szent-Gyorgyi (Drury and Szent-Gyorgyi, 1929) and later shown to be mediated by extracellular receptors (Degubareff and Sleator, 1965;Sattin and Rall, 1970). The existence of two different types of purine receptors for adenosine and for ATP, respectively, was first described by Burnstock (Burnstock, 1978), who suggested naming the receptors for adenosine as P1 and those for ATP as P2. In the same year, we first described the existence of two different types of receptors for adenosine which mediate the inhibition and stimulation of cyclic adenosine monophosphate accumulation and differ in their pharmacological properties (van Calker et al., 1978). Unaware of Burnstock's nomenclature, we suggested the names A 1 (inhibiting) and A 2 (stimulating) for these receptors (van Calker et al., 1978(van Calker et al., , 1979. The coincidence and independence of these two discoveries led to a somewhat confusing twofold nomenclature (P1 receptors vs. A 1 and A 2 receptors). Almost at the same time Londos and coworkers (Londos et al., 1980) also detected two different types of adenosine receptors that regulated the adenylate cyclase in fat cells which they suggested to be called R i (inhibiting) and R a (activating). However, the nomenclature A 1 and A 2 is now established Fredholm et al., 2011). The original definition of adenosine receptor subtypes by their effects on adenylate cyclase was soon substituted by a re-definition by means of efficacy of agonists and antagonists, since it became clear that adenosine receptors can have effects on various signal transducing systems. A 2 receptors were later found to encompass two different types of receptors, the high affinity A 2A and the low affinity A 2B receptors, and an additional third adenosine receptor subtype (A 3 ) was identified. These four adenosine receptor subtypes A 1 , A 2A , A 2B and A 3 are coupled to G-proteins. A 1 receptors typically act via the G i/o family, whereas A 2A and A 2B receptors act via G s . A 2B receptors can also activate phospholipase C via G q . A 3 receptors act via G i -mediated inhibition of adenylyl cyclase and G q -mediated stimulation of phospholipase C (Fig. 1). The particular structure of these receptors is now ascertained by molecular cloning Fredholm et al., 2011).
A general principle of adenosine's action in the body is its activity as an 'retaliatory metabolite', which signals an disequilibrium between energy supply and demand and triggers counter-balancing measures such as increase in blood flow and/or diminished cellular activity by activation of adenosine receptors. Presently, adenosine receptors are known to fulfill important regulatory functions in many cells and tissues such as the kidney (Vallon et al., 2006), heart (Mubagwa and Flameng, 2001), lungs (Polosa and Blackburn, 2009) and gastrointestinal tract (Colgan et al., 2013) and have also an important role in several malignancies  such as respiratory disease (Caruso et al., 2013), inflammatory disease (Aherne et al., 2011) or cancer (Antonioli et al., 2013). However, perhaps the most important regulatory function of adenosine is in the brain. Here, A 1 receptors, which have high affinity for adenosine, are distributed both pre-and postsynaptically. Presynaptically, they inhibit the release of excitatory and inhibitory neurotransmitters, e.g., glutamate, dopamine, serotonin and acetylcholine. When situated postsynaptically A 1 receptors inhibit neuronal signaling by hyperpolarization and reduce excitability via regulation of potassium channels. A 2A receptors are highly expressed on striatopallidal neurons with lower presence in other parts of the brain such as the cortex and hippocampus. They can form heteromers with A 1 receptors (Ciruela et al., 2006;Ferre et al., 2008;Cristovao-Ferreira et al., 2013) and with dopamine D 2 receptors (Fuxe et al., 2007), which enable adaptive responses in the regulation of synaptic plasticity (Fuxe et al., 2014). Adenosine A 2B and A 3 receptors may play a protective role in brain ischemia (Pedata et al., 2016) and exitotoxicity (Moidunny et al., 2012).

Role of adenosine A 2A receptors in depression
First evidence that A 2A receptors are expressed in the hippocampus and inhibit the activity of A 1 receptors was reported already 1994 (Cunha et al., 1994). Later, evidence for an antidepressant-like effect of adenosine A 2A antagonists and of A 2A deficiency in rodents was provided by El Yacoubi et al (El Yacoubi et al., 2000;El Yacoubi et al., 2001), an effect later confirmed by various groups (El Yacoubi et al., 2003). Thus, over-expression of A 2A receptors in forebrain neurons of transgenic rats is associated with increased depression-like behavior (Coelho et al., 2014) and anhedonia, one of the major pathological features of depression. In rodents, chronic unpredictable mild stress leads to an increase in depression-like behavior and is associated with a decrease in synaptic plasticity, a reduced density of synaptic proteins and an increase of A 2A receptors in the striatum and in glutamatergic terminals in the hippocampus (Crema et al., 2013;Kaster et al., 2015). These behavioral and synaptic alterations induced by chronic unpredictable mild stress appear to be indeed mediated by an increase in adenosine A 2A receptors, since they are prevented by caffeine (a non-selective adenosine antagonist for A 1 /A 2A receptors, which however elicits its effects on mood predominantly via antagonism at adenosine A 2A receptors), by selective A 2A receptor antagonists and by A 2A receptor deletion in forebrain neurons (Kaster et al., 2015). Furthermore, A 2A receptor antagonists evoke antidepressant-like effects in the forced swim test and the tail suspension test in rodents ( Fig. 2) (Hodgson et al., 2009;Yamada et al., 2013). In particular, depression-associated psychomotor slowing, fatigue and anergia are improved by A 2A receptor antagonists (Randall et al., 2011). This particular cluster of symptoms is also improved by modest doses of caffeine (Smith, 2009), apparently acting via antagonism at A 2A receptors (Fig. 2) (Lopez-Cruz et al., 2018). Very recent evidence indicates that blockade of A 2A receptors by a selective antagonist enhances the antidepressant-like activity of antidepressant medications such as tianeptine and agomelatine in mice behavioral despair tests (Szopa et al., 2019). Furthermore, A 2A receptor blockade also reverts stress-induced hippocampal-related deficits induced by maternal separation (Batalha et al., 2013). At first sight, these antidepressant-like effects of A 2A receptor antagonists effects appear to be inconsistent with the reported up-regulation by A 2A receptor agonists of brain-derived neurotrophic factor (BDNF) expression in rat primary cortical neurons (Jeon et al., 2011), since BDNF has well documented antidepressant-like effects (Bjorkholm and Monteggia, 2016;van Calker et al., 2018). However, the effects of adenosine A 2A receptor activation on BDNF appear to be complex (Rombo et al., 2016). Thus, e.g., in the hippocampus adenosine via A 2A receptors influences BDNF actions on gamma-aminobutyric acid (GABA) transmission affecting both glutamatergic inputs to pyramidal neurons and cholinergic inputs to GABA-ergic interneurons. It can also affect A 2A receptor-dependent facilitation of GABA uptake into astrocytes with consequent increase in GABA clearance from the synapses (Rombo et al., 2016). Furthermore, both anti-depressive-like and pro-depressive-like behaviors are associated with BDNF. To what extent one of these two opposite effects on behavior (anti-depressant or pro-depressant) dominates depends on the brain area and the brain cells in which these genes are activated (van Calker et al., 2018). How the predominant antidepressant-like effects of antagonism at A 2A receptors are mediated is unknown. However, since A 2A receptors are often found to inhibit the actions of A 1 receptors (Stockwell et al., 2017), one possible explanation for the antidepressant-like effects of A 2A antagonists is the facilitation of activity of A 1 receptors (Fig. 2). Also genetic variations in the adenosine A 2 receptor gene were shown to modify the risk of depression (Gass et al., 2010). Thus, the TT genotype of an adenosine A 2 receptor gene small nucleotide polymorphism was associated with reduced risk for depression when compared to the CC/CT genotypes (Oliveira et al., 2019).

Role of adenosine A 1 receptors in depression
Antidepressant effects of activation of adenosine A 1 receptors were first suggested by our group (van Calker and Biber, 2005) and later experimentally confirmed by Hines et al. (Hines et al., 2013) and our group (Serchov et al., 2015). Our suggestion (van Calker and Biber, 2005) was based on findings indicating that the therapeutic effects of SD and electroconvulsive therapy (ECT) are closely related to changes in slow wave sleep, cerebral metabolic rate, and cerebral blood flow, parameters that are at least in part regulated by signaling through adenosine A 1 receptors. Hines et al. later indeed demonstrated a significant correlation between the ability of SD to both activate A 1 receptor signaling pathways and to promote antidepressant-like effects (Hines et al., 2013). They showed that A 1 receptors are required for the antidepressant effect of SD and that activation of A 1 receptors leads to sustained antidepressantlike behaviors. These authors also claimed that the antidepressant-like effect of SD is mediated by astrocytes, since the dominant-negative SNAP receptor (dnSNARE) transgene in astrocytes (SNARE proteins mediate fusion of vesicles with their target membrane, a process inhibited by dnSNARE) impaired the ability of SD to reduce immobility time in both the forced swim and tail suspension tests. However, these conclusions have been questioned on the grounds that expression of the dnSNARE transgene was not restricted to astrocytes but also found in cortical neurons (Fujita et al., 2014).
The fact that activation of adenosine A 1 receptors indeed evokes pronounced antidepressant effects was shown by our group in a line of transgenic mice in which an overexpression of A 1 receptors can be switched on and off (Serchov et al., 2015). This antidepressant effect of A 1 receptor activation is, mediated by neuronal A 1 receptors, since the A 1 transgene expression in these mice is restricted to calcium/calmodulin-dependent protein kinase type II forebrain neurons (Serchov et al., 2012;Serchov et al., 2015). Up-regulating A 1 receptors by activation of the transgene in these mice led to pronounced acute and chronic resilience toward depressive-like behavior in various tests. On the other hand, A 1 receptor knockout mice displayed an increased depressive-like behavior and were resistant to the antidepressant effects of SD, indicating that the antidepressant effects of SD are largely mediated by the up-regulation of adenosine A 1 receptors induced by SD ( Fig. 2) (Serchov et al., 2015). Furthermore, we have shown that the antidepressant effects of A 1 receptor activation are mediated by the immediate early gene Homer1a, which is up-regulated by various antidepressant treatments such as SD, imipramine, ketamine as well as A 1 receptor activation (Fig. 2). Indeed, small interfering ribonucleic acid knockdown of Homer1a in the medial prefrontal cortex (mPFC) enhanced depressivelike behavior and prevented the antidepressant effects of A 1 receptor up-regulation, SD, imipramine and ketamine, while viral over-expression of Homer1a in the mPFC exerted antidepressant effects. Thus, Homer1a in the mPFC is a final common pathway mediating the antidepressant effects not only of adenosine A 1 receptor activation but also of different other antidepressant treatments (Serchov et al., 2015;Serchov et al., 2016). Very recently, we have shown that this antidepressant effect of Homer1a activation is due to Homer1a induced constitutive agonist-independent mGluR5 activation, resulting in enhanced AMPA receptor-mediated synaptic transmission (Holz et al., 2019).

Potential role of adenosine receptors in bipolar disorders
The idea that adenosine receptors might be involved in the pathophysiology of bipolar disorder goes back to findings of an increased excretion of uric acid, a metabolite of adenosine, in manic patients (Machado-Vieira et al., 2002). Since then these findings have been confirmed by several groups suggesting a purinergic system dysfunction associated with manic phases of bipolar disorder (Machado-Vieira et al., 2002;De Berardis et al., 2008;Salvadore et al., 2010;Bartoli et al., 2016;Bartoli et al., 2017a;Bartoli et al., 2017b). This may also be related to the efficacy of allopurinol, which increases adenosine levels by inhibiting purine degradation (Marro et al., 2006;Schmidt et al., 2009), in treating acute mania when used adjunctively with lithium (Akhondzadeh et al., 2006;Machado-Vieira et al., 2008) or valproate (Jahangard et al., 2014). This effect was, however, not evident when allopurinol was used in the absence of lithium or valproate (Weiser et al., 2014;Bartoli et al., 2017b). It is, however, still unclear, whether or not these findings, in the periphery, indeed indicate an adenosine dysfunction in bipolar disorder in the brain (Hirota and Kishi, 2013;Ortiz et al., 2015;Gubert et al., 2016). Evidence from association studies does not give any indication that genetically determined variation of the A 1 receptor and its two promoters could play a major role in the development of bipolar affective disorder (Deckert et al., 1998a). Whether or not adenosine A 1 receptors are also involved in manic-like behavior remains to be established. Indeed, SD, which up-regulates A 1 receptors, not only has antidepressant effects but can also trigger symptoms of mania or hypomania in certain bipolar patients (Wehr, 1989;Lewis et al., 2017). Furthermore, there is evidence that carbamazepine, which is approved for the treatment of acute and dysphoric mania (Baldessarini et al., 2019) acts as a specific antagonist of adenosine A 1 receptors ( Van Calker et al., 1991). Via up-regulation of expression of A 1 receptors carbamazepine may also induce a new quality of adenosine A 1 -receptor-mediated signal transduction in cells that initially express low basal A 1 -receptor numbers (Biber et al., 1996;Biber et al., 1999).

Role of adenosine A 1 and A 2A receptors in anxiety disorders
In general, agonistic actions at A 1 receptors appear to promote anxiolytic effects (Jain et al., 1995;Florio et al., 1998;Vincenzi et al., 2016), whereas cyclopentyltheophylline, an A 1 antagonist, had anxiogenic properties (Florio et al., 1998). However, the investigation of other A 1 antagonists gave mixed results (Correa and Font, 2008). Unspecific antagonists of adenosine receptors appear to exert general anxiogenic effects. Thus, non-selective adenosine antagonists like caffeine, theophylline, theobromine (Charney et al., 1985;Lee et al., 1988;Kulkarni et al., 2007;Lopez-Cruz et al., 2014) and isobutylmethylxanthine (Florio et al., 1998) elicit anxiety related behavior. While the effects of caffeine on mood and memory (Kaster et al., 2015) as well as on wakefulness (Huang et al., 2005;Lazarus et al., 2011) appear to be mediated via antagonism at adenosine A 2A receptors (see above), no definitive information is available about the adenosine receptor subtype mediating the anxiogenic effects of caffeine. At least in rodents, the anxiogenic effect of caffeine is not mimicked by selective A 2A receptor antagonists (El Yacoubi et al., 2000), and increased anxietylike behavior is observed not only in A 2A (Ledent et al., 1997;Deckert, 1998) but also in A 1 (Johansson et al., 2001;Gimenez-Llort et al., 2002) receptor knockout mice. Thus, both adenosine receptors subtypes A 1 and A 2A may play a role in anxiety at least in rodents. The effects of A 2A receptors in anxiety in rodents have been investigated in some detail: A 2A receptor knock-out mice exhibit not only increased anxiety-like behavior but also increased c-Fos immunoreactivity in the anterior cingulate cortex and the amygdala as compared to wild-type mice (Lopez-Cruz et al., 2017). However, the effects of A 2A receptors on anxiety-like behavior in rodents are variable and highly dependent on the brain region. Thus, selective downregulation of the A 2A receptor in the basolateral complex of the amygdala by means of a lentivirus with a silencing short hairpin ribonucleic acid impaired fear acquisition as well as Pavlovian fear retrieval (Simoes et al., 2016). On the other hand, adult male rats over-expressing the human A 2A receptor in forebrain neurons not only showed increased depressive-like behavior (see above) but also covered higher distances in the open field test and spent more time in the central zone than wild-type rats (Coelho et al., 2014). While this might indicate reduced anxiety-like behavior, the authors argue that there is a mutual influence between anxiety and locomotor activity even though locomotion and anxiety are differentially regulated by adenosine A 2A receptors. Thus, the reason for the discrepancy between depressive-like behavior on the one hand and increased exploratory behavior on the other remains unexplained (Coelho et al., 2014). Indeed, deletion of A 2A receptors in the forebrain rather inhibited fear conditioning, whereas deletion of A 2A receptors in the striatum facilitated Pavlovian fear conditioning (Wei et al., 2014).
In humans, there is evidence from genetic studies for a potential role of the adenosine A 2A receptor gene in anxiety disorders. The T allele of a silent polymorphism in exon 2 of the adenosine A 2A receptor gene located on chromosome 22q11.23 (small nucleotide polymorphism rs5751876, 1976T>C, formerly 1083T>C, Tyr/Tyr) was consistently found associated with panic disorder (Deckert et al., 1998b;Hamilton et al., 2004;Rogers et al., 2010). However, no such association was discerned in populations of Asian descent (Yamada et al., 2001;Lam et al., 2005). This rs5751876 T risk allelepartly epistatically with another allele (2592 Tins/Tins genotype)has furthermore been observed to significantly influence anxiety response after caffeine as well as amphetamine administration (Alsene et al., 2003;Hohoff et al., 2005;Childs et al., 2008). The mechanism by which this genotype (rs5751876 TT) may increase the risk for anxiety disorders was investigated in healthy probands. The TT genotype was found associated with increased connectivity between the insula and the prefrontal cortex along with heightened interoceptive accuracy (Geiger et al., 2016). Interoception denotes the sense of the internal state of the body as relayed from the body to specific subregions of the brain such as the brainstem, thalamus, insula, and anterior cingulate cortex. Increased interoception can lead to emotional distress, particularly in individuals with higher sensitivity for anxiety, and contribute to the predisposition to anxiety disorders (Domschke et al., 2010b). Furthermore, carriers of the risk genotype mentioned above (rs5751876 TT) showed the highest startle magnitudes after caffeine administration in response to unpleasant pictures in an emotion-potentiated startle paradigm, with this effect arising particularly from the female subgroup (Domschke et al., 2012a). In addition, female homozygous carriers of this genotype showed other distinctive features such as an impaired ability to selectively process very early information and to gate irrelevant sensory information as measured by the prepulse inhibition/facilitation paradigm (Gajewska et al., 2013). These findings in healthy probands could indicate thatunder adverse life conditionscertain genotypes may confer an increased risk to develop one form of anxiety disorders. However, how these particular genotypes may lead to modifications in behavior is unclear, since they are not associated with changes of the amino-acid sequence of the A 2A receptor. Hamilton and colleagues (Hamilton et al., 2004) discuss the possibility that these 'silent' variants may cause functional variation via codon preference during translation. Indeed, recent research has revealed mechanisms how "codon bias" can guide codon usage in translation and thereby alter the efficiency of protein production (Hanson and Coller, 2018).
Several other studies have revealed an interaction of the adenosinergic system with other systems pivotally involved in the pathogenesis of anxiety and panic disorder in particular such as the neuropeptide S system (Domschke et al., 2012b) or the dopaminergic system (Childs et al., 2008). A recent study implied that regular exercise exerts its anxiolytic effect by inhibiting A 2A receptor function via enhancing serotonin 2A receptor signaling in the basolateral amygdala (Leem et al., 2019). In summary, there is converging multi-level evidence for an arousal-, attention-and anxiety-related role of the adenosinergic system (Geiger et al., 2016) suggesting further research into A 2A receptors as promising pharmacological targets in the treatment of anxiety disorders (Yamada et al., 2014).

Alteration of circadian rhythms in mood disorder: effect of adenosine receptors
Clock gene dysfunction has long been considered as one pathogenic factor in mood disorders (McCarthy and Welsh, 2012;Gonzalez, 2014;Landgraf et al., 2014;Landgraf et al., 2016;Beyer and Freund, 2017). Chronic stress exposure, a major cause for several psychiatric disorders, disrupts circadian rhythms (Zaki et al., 2019). Increasing evidence suggests that region-specific circadian oscillations in limbic regions are instrumental regulators of mood (Kim et al., 2015;Logan et al., 2015;Landgraf et al., 2016). Recent evidence indicates that intrinsically photosensitive retinal ganglion cells may be involved in mood regulation (Lazzerini Ospri et al., 2017). Purinergic signaling has been found important in the regulation of circadian rhythms (Reichert et al., 2016;Lindberg et al., 2018), and circadian regulation of clock genes is believed to be involved in the rapid antidepressant actions of ketamine and SD (Bunney et al., 2015). Both SD and ketamine modulate the activity of the clock gene machinery via effects on e.g., N-methyl-Daspartate receptors, AMPA receptors and mammalian target of rapamycin (Bunney et al., 2015). Clock genes including circadian associated repressor of transcription, period circadian regulator 2, neuronal PAS domain protein 4, D-Box binding protein, and RAR related orphan receptor B are down-regulated in both ketamine-and SD-treated mice (Orozco-Solis et al., 2017). Since the antidepressant effect of SD is mediated by increased signaling via adenosine A 1 receptors (Hines et al., 2013;Serchov et al., 2015), the down-regulation of clock genes by SD (Bunney et al., 2015;Orozco-Solis et al., 2017) is probably induced by activation of A 1 receptors (Fig. 3). We have shown that the antidepressant effects of both SD and ketamine are finally mediated by an increase in Homer1a (Serchov et al., 2015). Among the compounds participating in the regulation of Homer1a (van Calker et al., 2018) particularly BDNF appears to be involved in clock gene regulation (Bunney et al., 2015;Bjorkholm and Monteggia, 2016;Serchov and Heumann, 2017), whereas little is known about a potential interaction of Homer1a with clock genes.
However, not only A 1but also A 2Areceptors play an active role in the control of circadian rhythms which may be involved in the pathophysiology of mood disorders (Lindberg et al., 2018). Thus, adenosine signaling via A 2A receptors was shown to regulate striatal cellular and behavioral circadian timing and activity level (Ruby et al., 2014). Both A 1 receptors and particularly A 2A receptors regulate sleep (Huang et al., 2005). However, while A 1 receptors are known to mediate the antidepressant effects of SD (see above), little is known about the potential relationship between the function of A 2A receptors in sleep and their role in depression or anxiety.

Role of adenosine receptors in the effects of SD and chronic sleep restriction on mood and anxiety
As shortly mentioned above SD induces an increase in adenosine (Leenaars et al., 2018) and an up-regulation of adenosine A 1 receptors in the brain (Porkka-Heiskanen et al., 1997;Elmenhorst et al., 2007;Elmenhorst et al., 2009;Elmenhorst et al., 2017), which elicits the sleepiness-inducing effects of prolonged wakefulness and mediates the antidepressant effects of SD (Fig. 3) (Hines et al., 2013;Serchov et al., 2015). The potential effects of SD on A 2A receptors are much less clear. Initially, a downregulation by SD (3 and 6 h) of A 2A receptor messenger ribonucleic acid and receptor binding was found restricted to the olfactory tubercle (Basheer et al., 2001). Chronic sleep restriction was found to lead to A 2A receptor downregulation also only in the olfactory tubercle (Kim et al., 2015). Thus, the time course, brain area and the extent of down-regulation of A 2A receptors (if any) after SD is still unclear. Since A 2A receptor activation induces depressionlike behavior in rodents (see discussion above), downregulation of A 2A receptors may contribute to the antidepressant effects of SD and add to the antidepressant effects of increased A 1 receptor signaling. However, presently no data are available that would support this hypothesis. The increased signaling via A 1 receptors induced by SD leads to an enhanced formation of Homer1a in the mPFC, which mediates the antidepressant effects of SD (Serchov et al., 2015). However, SD in addition to its antidepressant effects also induces impairments in cognitive functions similar to those of ethanol which also induces an upregulation of cerebral A 1 adenosine receptors (Elmenhorst et al., 2018). In addition, SD in humans appears to increase state anxiety (Pires et al., 2016b), but may induce rather a decrease in anxiety-like behavior in preclinical models (Pires et al., 2016a). There are differences in the time courses for impairment of performance and recovery between acute and chronic sleep loss. While the acute upregulation of A 1 receptors induced by SD is accompanied by homeostatic increase in non-rapid eye movement sleep, slow-wave activity and adenosine-dependent inhibition of synaptic activity, prolonged sleep restriction (3 days) caused a reduction in these parameters by reducing the adenosine-tone and attenuated the response to acute sleep deprivation (Clasadonte et al., 2014). Similarly, whereas short time (12 h) SD elicited antidepressant effects, more extended SD (72 h) had no antidepressant-like effects in mice (Hines et al., 2013). Chronic exposure to sleep restriction is rather associated with an increased risk of depression (Baum et al., 2014;Conklin et al., 2018). Moreover, chronic sleep restriction induces long-lasting increase in A 1 R expression in several brain regions and a reduced adenosine A 2A receptor density in one of the three brain areas analyzed (olfactory tubercle) (Kim et al., 2015), which may underlie the negative effects of chronic sleep restriction on mood regulation (Novati et al., 2008). Indeed, as already mentioned above, the consequences of A 1 receptor up-regulation differ dependent on both the duration of sleep restriction and the particular part of the brain investigated. Chronic insufficient sleep duration equivalent to 5.6 h of sleep opportunity per 24 h impairs neurobehavioral performance even without extended wakefulness (McHill et al., 2018). Disturbed sleep also negatively affects the immune system (Irwin and Opp, 2017) and induces elevation in brain inflammatory molecules such as interleukin 1-b (IL-1b) and tumor necrosis factor-a (TNF-a) and inhibition of BDNF (Zielinski et al., 2014). These negative effects of chronic SD on cognitive performance (Elmenhorst et al., 2018) appear to be mediated via effects on both adenosine A 1 and A 2A receptors (Urry and Landolt, 2015) and are at least in part modified by heritable individual differences (Krause et al., 2017). Indeed, there is evidence that prolonged A 1 receptor signaling and its cross-talk with A 2A receptors may form the cellular basis for increased neurotoxicity in neurodegenerative disorders Chen et al., 2016;Stockwell et al., 2017).
Potential role of adenosine receptors in the antidepressant effects of electroconvulsive therapy ECT is predominantly used to treat major depression but less frequently is also applied to treat schizophrenia, catatonia and acute mania (Payne and Prudic, 2009). The neurobiological mechanism of action of ECT is still unknown, but is related to the seizures induced by the treatment. Modern theories comprise e.g. neuroimmunological mechanisms such as low TNF-a (Sorri et al., 2018;Yrondi et al., 2018), alterations in BDNF and vascular endothelial growth factor (Minelli et al., 2011;Polyakova et al., 2015), neuroendocrine mechanisms (Haskett, 2014) and alterations in sortilin-derived propeptide (Roulot et al., 2018). We (van Calker and Biber, 2005) have first suggested a potential role of adenosine and A 1 receptors in the mechanism of action of ECT based on the effects on slow wave sleep, cerebral metabolic rate and cerebral blood flow, since these effects are very similar to those of SD (see above) and a pronounced augmentation of adenosine and adenosine A 1 receptors in the brain after ECT or seizures in general is well known (Lewin and Bleck, 1981;Newman et al., 1984;Gleiter et al., 1989;Boison, 2016). This increase in adenosine signaling evoked by ECT is most probably also responsible for the well-known ECT-induced increase in seizure threshold (Coffey et al., 1995;van Calker and Biber, 2005). In contrast to A 1 -receptors A 2 -receptors are rapidly downregulated after ECT, perhaps contributing to the antidepressant effects (since A 2 receptors rather increase depression, see above) (van Calker and Biber, 2005). Since increased signaling via adenosine A 1 receptors has been shown to have pronounced antidepressant effects (Serchov et al., 2016), the ECT-induced increase in adenosine and A 1 receptors is very likely at least partially responsible for ECT's antidepressant activity. This conclusion is also corroborated by the other effects of ECT downstream to adenosine A 1 receptor activation (Fig. 2). Indeed, similar to SD, which upregulates Homer1a via A 1 receptor activation (Serchov et al., 2015), also ECT upregulates Homer1a expression levels in the cortex (Kato, 2009), most probably mediated by the increased A 1 receptor signaling induced by ECT. Homer1a was therefore proposed to be instrumental for the therapeutic effect of ECT in depression (Kato, 2009;Serchov et al., 2016). In addition to adenosine and A 1 receptors, also purinergic signaling through ATP via P2-receptors was suggested to play a role in ECT (Sadek et al., 2011).

Potential role of adenosine A 1 receptors in the antidepressant effects of transcranial direct current stimulation
Transcranial direct current stimulation (tDCS) is a noninvasive technique of brain stimulation that modulates cortical excitability. It is used in humans in attempts to treat diverse neurological and neuropsychiatric disorders including e.g Parkinson's disease (Fregni et al., 2006), cerebrovascular events (Fregni et al., 2005), neuropathic pain (Mori et al., 2010), epilepsy (San-Juan et al., 2015) and depressive disorders (Meron et al., 2015;Moffa et al., 2018) including bipolar depression (Sampaio-Junior et al., 2018). In experimental animal models, it was shown that the modulation of cortical excitability induced by cathodal tDCS is mediated by adenosine A 1 receptors, since local microinjection of the adenosine A 1 receptor antagonist 8-cyclopentyl-1,3-dipropylxanthine prevented the effects of cathodal tDCS (Marquez-Ruiz et al., 2012). Since activation of adenosine A 1 receptors elicits pronounced antidepressant-like effects (see previous paragraph) (Serchov et al., 2016), it is conceivable that the antidepressant effects of tDCS in some studies (Meron et al., 2015;Moffa et al., 2018) might be mediated by A 1 receptors (Fig. 2).
Potential role of adenosine A 1 receptors in the antidepressant effects of deep brain stimulation in treatment resistant depression Deep brain stimulation (DBS) consists of implanting electrodes in specific brain areas followed by optimized stimulation settings. This technique has long been used for the treatment of a variety of neurological and neuropsychiatric disorders (Ward et al., 2010) including e.g., Parkinson's disease and essential tremor (Benabid et al., 2009a;Benabid et al., 2009b), pain (Hamani et al., 2006;Levy et al., 2010) and obsessive compulsive disorder (Denys and Mantione, 2009). First evidence from small studies indicated that DBS might also improve treatment resistant depression (Mayberg et al., 2005;Giacobbe et al., 2009;Anderson et al., 2012;Berlim et al., 2014) including bipolar depression (Gippert et al., 2017). However, a recent controlled study could not demonstrate a significant effect of DBS in ventral capsule/ ventral striatum, in chronic treatment resistant depression (Dougherty et al., 2015). Other recent controlled studies report limited antidepressant effects of DBS in other brain regions such as the ventral anterior limb of the internal capsule (Bergfeld et al., 2016) and the subcallosal cingulate gyrus (Merkl et al., 2018). Thus, one problem in the analysis of DBS in depression are the different anatomical targets affected by DBS in the various studies including e.g., ventral capsule/ventral striatum, subgenual cingulate cortex, medial forebrain bundle and the lateral habenula (Malone et al., 2009;Bewernick et al., 2010;Kennedy et al., 2011;Bewernick et al., 2012;Holtzheimer et al., 2012;Lozano et al., 2012;Berlim et al., 2014;Schlaepfer et al., 2014;Dougherty et al., 2015;Dandekar et al., 2018;Coenen et al., 2019). To complicate matters further, a potential role of glia in the mechanism of action of DBS appears possible (Anderson et al., 2012;Vedam-Mai et al., 2012;Fenoy et al., 2014;Etievant et al., 2015a;Etievant et al., 2015b;McIntyre and Anderson, 2016). The therapeutic effects of DBS in tremor (Bekar et al., 2008) and epilepsy (Miranda et al., 2014) were shown to be associated with a marked accumulation of adenosine, which mediated an activation of adenosine A 1 receptors. Similarly, also the action of DBS in depression could be due to activation of adenosine A 1 receptors (Fig. 2) (Tawfik et al., 2010;Etievant et al., 2013;Etievant et al., 2015a;Etievant et al., 2015b), in accordance with the pronounced antidepressant-like effects of A 1 receptor activation in mice (see above) (Serchov et al., 2016).

Regulation of adenosine receptor expression in mood disorders: Neuro-immunological mechanisms
In the preceding chapters, we have presented evidence that alteration of adenosine A 2A and A 1 receptor expression and activity differentially influences mood in experimental animals, partly reflecting the A 1 receptor mediated antidepressant effects of SD and ECT in humans (Serchov et al., 2016). Thus it is important to examine how adenosine receptor expression is regulated in the brain under normal conditions and whether or not this regulation might be disturbed in mood disorders. There is very little information concerning the molecular mechanisms in the regulation of adenosine receptor expression, except for the role of nuclear factor (NF)-jB (Ramesh et al., 2007;Sheth et al., 2014). However, there is evidence that adenosine receptors interact with immunological mechanisms in the brain and that chemokines and cytokines such as IL-1b, IL-6, and TNF-a are altered in depressive disorder (Dantzer et al., 2008;Miller et al., 2009;Dowlati et al., 2010;Young et al., 2014;Hodes et al., 2015;Bhattacharya et al., 2016;Slusarczyk et al., 2016;Wohleb et al., 2016;Kakeda et al., 2018;Kohler et al., 2018). Among these, alterations in IL-6 were found by cumulative meta-analyses to be the best documented (Haapakoski et al., 2015). We have shown that the expression of both adenosine A 1 and A 2 receptors in the brain and in neural cells in culture is regulated by interleukin-6 and other cytokines (Biber et al., 2001;Biber et al., 2008;Vazquez et al., 2008;Moidunny et al., 2010). On the other hand, adenosine stimulates via A 2B -and A 2A receptors excretion of IL-6 (Fiebich et al., 1996;Schwaninger et al., 1997;Schwaninger et al., 2000;Fiebich et al., 2005) and IL-1b , both found increased in depression (Ng et al., 2018), and regulates immune functions in the brain (Hasko et al., 2005;Abbracchio and Ceruti, 2007;. Furthermore, there is very robust evidence showing that A 2A receptors control the release of different cytokines in the brain (Rebola et al., 2011). Thus, there appears to exist a reciprocal interconnection between cytokines and adenosine receptors in the brain potentially important in the pathophysiology of depressive disorders. This crosstalk is particularly evident in retinal ganglion cells, where both adenosine A 1 and A 2A receptors interact with IL-6 to mediate cell survival and IL-6 modulates through the regulation of adenosine A 1 and A 2A receptor expression the level of BDNF (Perigolo-Vicente et al., 2013;Perigolo-Vicente et al., 2014), which has a welldocumented role in depression (van Calker et al., 2018). Furthermore, A 2A receptors are also involved in the regulation of the release of BDNF from activated microglia and in the proliferative role of BDNF (Gomes et al., 2013), in accord with the potential role of microglia in psychiatric disorders . Thus, there is reason to believe that adenosine via modulation of the effects of BDNF, IL-6 and perhaps other cytokines might improve the particular subtype(s) of depressive disorders that are regulated by neuroimmunological mechanisms (Wohleb et al., 2016).

Conclusions
As reviewed above, both A 1 and A 2A adenosine receptors are implicated in the etiology and treatment of mood and anxiety disorders. Thus activation of A 1 and inhibition of A 2A receptors elicit antidepressant effects (Fig. 2). The antidepressant effects of enhancement of A 1 receptor signaling occurs through an increase of signaling via Homer1a which leads finally to a modulation of AMPA receptor functioning (Holz et al., 2019). How the antidepressant effects of inhibition of A 2A receptors are mediated is still unknown. In addition to their role in mood disorders, adenosine A 1 and A 2A receptors also regulate anxiety-like behavior. In particular A 2A receptors appear to be important in this regard. Adenosine receptors play an important role in sleep regulation and influence circadian clockwork. Indeed, circadian function and sleep regulation are consistently dysregulated in many mental diseases including depression and anxiety disorders (Fig. 3). Recent evidence has identified neuroimmunological mechanisms that both regulate and are regulated by adenosine receptors. As much as these mechanisms are involved in the pathophysiology of certain types of depression and perhaps also anxiety disorders they may present a promising field of future research. Preclinical studies have begun to assess antidepressant outcomes associated with adenosinergic modulators. Particularly, a therapeutic use of A 2A receptor agonists has been suggested for autismspectrum disorders and schizophrenia, while A 2A receptor antagonists might carry some promise for Alzheimer's disease, Parkinson's disease, attention-deficit hyperactivity disorder, depression and anxiety (Domenici et al., 2019). Future research is, however, needed to explore the therapeutic potential of adenosine receptor modulators in clinical trials. With regard to translational research, the application of new technologiesfor instance, epigenetics and proteomics should be included in future studies. In therapeutic applications, more selective modulators of adenosine receptors should be developed and tested in mood and anxiety disorders.