Adenosine Receptors: Expression, Function and Regulation

Adenosine receptors (ARs) comprise a group of G protein-coupled receptors (GPCR) which mediate the physiological actions of adenosine. To date, four AR subtypes have been cloned and identified in different tissues. These receptors have distinct localization, signal transduction pathways and different means of regulation upon exposure to agonists. This review will describe the biochemical characteristics and signaling cascade associated with each receptor and provide insight into how these receptors are regulated in response to agonists. A key property of some of these receptors is their ability to serve as sensors of cellular oxidative stress, which is transmitted by transcription factors, such as nuclear factor (NF)-κB, to regulate the expression of ARs. Recent observations of oligomerization of these receptors into homo- and heterodimers will be discussed. In addition, the importance of these receptors in the regulation of normal and pathological processes such as sleep, the development of cancers and in protection against hearing loss will be examined.


Desensitization of Adenosine Receptors
Continued or repeated exposures to an agonist usually result in the receptor-mediated response to plateau and then diminish despite the continual presence of agonist. In the field of receptor kinetics, this process is called desensitization [41]. Desensitization of GPCRs is a complex phenomenon involving multiple and temporally distinct components. The mechanisms underlying rapid desensitization often involve receptor phosphorylation by a family of G protein-coupled receptor kinases (GRKs) resulting in their preferential binding to arrestins molecules. This promotes desensitization by uncoupling the receptor from G-protein, leading to impaired receptor function [42]. Following desensitization, many GPCRs internalize by an arrestin-dependent process via clathrin-coated pits which leads to the eventual intracellular dephosphorylation of the receptor, and its re-insertion into the cell membrane to produce a resensitized state [43]. More prolonged agonist activation generally leads to the transfer of internalized receptor to a lysosomal compartment with subsequent down-regulation [44]. The process of desensitization, internalization, resensitization, or down-regulation forms an important part in receptor regulation which eventually controls receptor-mediated signaling pathways.
It has been shown that activation of all four AR subtypes eventually leads to their desensitization by different mechanisms. Most of these observations are made following administration of non-hydrolyzable agonists to cell cultures and animals. However, it is yet unclear how significant a role receptor desensitization plays in regulation of ARs by adenosine in vivo under normal conditions. This process should become more important with the use of selective AR subtype agonists for the treatment of diseases. Studies have shown that upon addition of agonists, the A 1 AR is phosphorylated and internalized slowly, with a half-life of several hours. In contrast, A 2A AR and A 2B AR demonstrate a more rapid rate of down-regulation, usually lasting about an hour. The down-regulation and desensitization of the A 3 AR occurs within minutes [45]. Thus, it is important to understand their regulation in order to design drugs that can exploit or avoid the receptor-mediated signaling to treat diseases.

A 1 AR
Several studies have been carried out to understand the molecular mechanisms underlying A 1 AR desensitization. In an early study, Parson and Stiles [46] showed that the A 1 AR in rat adipocytes desensitized upon chronic administration of A 1 AR agonist, R-phenylisopropyladenosine (R-PIA), over a period of six days by Alzet minipumps (Alza Corporation, Vacaville, CA, USA). A decrease in A 1 AR levels and reduced inhibition of isoproterenol-stimulated adenylyl cyclase activity was detected following R-PIA treatment. These changes were associated with decreased levels of pertussis-sensitive G proteins, but an increase in (cholera toxin-labeled) G s proteins in the plasma membranes. Furthermore, adipocyte membranes obtained from R-PIA-treated rats showed enhance isoproterenol and forskolin-stimulated adenylyl cyclase. These studies suggest the response of adipocytes to chronic activation of the A 1 AR (i.e., loss of inhibition of adenylyl cyclase) is to increase the stimulatory response (i.e., increase in adenylyl cyclase activity). Alternatively, desensitization of the A 1 AR could relieve a tonic inhibitory tone which is reflected by enhanced agonist stimulation of adenylyl cyclase. A subsequent study performed in cultured adipocytes confirmed the in vivo findings and further demonstrated that desensitization of the A 1 AR was associated with desensitization of insulin-dependent glucose transport [47]. This suggests a common link between desensitization of the A 1 AR and insulin receptor. A later study by Longabaugh et al. [48] confirmed the previous findings that desensitization of the A 1 AR to R-PIA was linked to reductions in A 1 AR and G i proteins and an increase in G sα proteins, but showed that the changes in G iα proteins were not linked to alternations in their mRNA levels.
Studies in a clonal cell line, ductus deferens smooth muscle (DDT 1 -MF2 cells), showed differential rates of desensitization of the A 1 AR and A 2A AR, which were not associated with changes in their coupling to G proteins. Increased phosphorylation of the A 1 AR was observed following exposure of these cells to agonists [49]. However, the mechanism(s) underlying phosphorylation was not determined. This study demonstrates that homologous desensitization of the A 1 AR was associated with phosphorylation and uncoupling of the receptor from its G i protein. Similarly, Nie et al. [50] demonstrated rapid translocation of GRK from cytosol to the plasma membrane and subsequent phosphorylation of the A 1 AR within one hour of R-PIA treatment to DDT 1 -MF2 cells. Furthermore, purified preparations of the A 1 AR that were phosphorylated with purified recombinant GRK-2 displayed enhanced affinity for arrestin over G i /G o G-proteins. In contrast to these findings, short-term exposure of CHO cells, stably expressing human A 1 AR, to R-PIA resulted in neither desensitization nor phosphorylation by GRK [51]. The reason for these different findings is not clear at present, but could reflect differences between hamster and human A 1 AR used in these studies. An interesting finding in DDT 1 -MF2 cells was that ectoenzyme adenosine deaminase, was shown to play an important role in agonist-induced A 1 AR regulation by enhancing desensitization and internalization in DDT 1 -MF2 cells after chronic exposure to R-PIA. It was demonstrated that upon exposure to the agonist, adenosine deaminase and A 1 AR colocalize and internalize by the same endocytotic pathway [52].
In primary cultures of cerebellar granule cells, chronic treatment with A 1 AR agonist and antagonist reciprocally regulates A 1 AR [53]. Exposure to A 1 AR agonist, N 6 -cyclopentyladenosine, resulted in time-and concentration-dependent reduction in the density of the A 1 AR and G-protein coupling corresponding to impaired agonist-induced adenylyl cyclase inhibition.
Coelho et al. [54] reported that hypoxia decreases the density of A 1 AR in rat hippocampal slices. This desensitization could be mimicked by 2-chloroadenosine (CADO), and was prevented by adding the A 1 AR antagonist, 1,3-dipropyl-8-cyclopentylxanthine (DPCPX). These results suggest that hypoxia leads to an increase in extracellular adenosine levels, and a subsequent, quite rapid (<90 min) desensitization of A 1 AR.
Jajoo et al. [55] examined the role of β-arrestin1/ extracellular signal-regulated kinase (ERK1/2) MAPK pathway in the regulation of A 1 AR desensitization and recovery in DDT 1 -MF2 cells. They reported that the exposure of A 1 AR agonist, R-PIA, for 24 h resulted in a decrease of A 1 AR membrane protein which was associated with an unexpected 11-fold increase in A 1 AR mRNA. This effect of R-PIA was dependent on β-arrestin1, as knockdown of β-arrestin1 by siRNA blocked R-PIA-mediated down-regulation of the A 1 AR. In addition, β-arrestin1 knockdown by siRNA suppressed R-PIA-dependent ERK1/2 and activator protein-1 (AP-1) activities and reduced the induction of A 1 AR mRNA. Interestingly, withdrawal of the agonist after a 24 h exposure resulted in rapid recovery of plasma membrane A 1 AR, which was dependent on the de novo protein synthesis and on the activity of ERK1/2 but independent of β-arrestin1 and NF-κB. These findings suggest that the β-arrestin1/ERK1/2 pathway, which contributes to the desensitization and down-regulation of A 1 AR membrane protein, is also able to prime the transcriptional machinery for rapid synthesis of the A 1 AR upon withdrawal of the agonist.
Chick embryo retina cultures of eight days treated for 48 h with CGS 21680 and N6-(2 (3,5- showed an increased expression of A 1 AR. This effect was blocked by the selective A 2A AR antagonist, ZM 241385, and PKA inhibitor, H89 [56]. These in vitro findings were confirmed in an in vivo study where chick embryo retinas were treated for 48 h with CGS 21680. In this study, CGS 21680 was able to increase the expression of A 1 AR which was blocked by selective A 2A AR antagonists, SCH 58261 and ZM 241385. Interestingly, under normal conditions, endogenous adenosine produced and released into the environment of the retina appears to control A 2A AR-mediated expression of A 1 AR, since the treatment with ZM 241385 and SCH 58261 alone was capable of reducing the expression of A 1 AR [57]. These findings indicate that up-regulation of A 1 AR by long-term activation of A 2A AR depends on classical AR signaling pathway. A study done in rat hippocampus challenged with hypoxia showed internalization and desensitization of the A 1 AR, but without a reduction in the overall amount of these receptors [54]. Another study in C6 glioma cells subjected to hypoxia (2, 6 and 24 h) showed down-regulation of the A 1 AR and up-regulation of A 2A AR receptor. This effect was shown to be dependent on the release of adenosine, since treatment with adenosine deaminase was able to block this effect. Peculiarly, the effect of the A 2A AR expression induced by hypoxia was inhibited by A 1 AR antagonist, DPCPX, but not by the A 2A AR antagonist, ZM 241385, indicating that the increased expression of the A 2A AR receptor induced by hypoxia depends on the A 1 AR [58].

A 2A AR
A 2A AR desensitization is mediated by multiple, temporally distinct, agonist-dependent processes. Short-term agonist exposure induced a rapid desensitization of A 2A AR-stimulated adenylyl cyclase activity which was associated with diminished receptor-G s coupling, and agonist-stimulated phosphorylation of the A 2A AR receptor protein. Longer agonist treatment, however, resulted in down-regulation in total receptor number and up-regulation of α-subunits of inhibitory G-protein [59]. The structural requirements necessary for the agonist-induced desensitization of A 2A AR resides mainly in its carboxy-terminus. Palmer and Stiles [60] introduced various mutations in the 95 amino acid sequence from the carboxy-terminus of A 2A AR which contains ten different phosphorylation sites and identified threonine-298 to be essential in agonist-induced receptor phosphorylation and short-term desensitization. However, it is not involved in long-term desensitization of A 2A AR function. These findings confirm that short-term and long-term desensitization of A 2A AR could be mediated by structurally distinct regions of the receptor protein and may involve different mechanisms.
GRKs play an important role in mediating agonist-induced phosphorylation and subsequent desensitization of GPCR action. Out of the different isoforms of GRKs, few of them have been shown to regulate A 2A AR desensitization. NG108-15 mouse neuroblastoma/rat glioma cells over-expressing wild-type GRK-2 showed marked reduction in the adenylyl cyclase activity after acute A 2A AR stimulation and enhanced sensitivity of A 2A AR to desensitization. This phenomenon was found to be dependent on the levels of GRK-2. In cells expressing very high levels of GRK-2, low agonist concentration was sufficient to induce GRK-dependent desensitization [61]. This GRK-2-mediated A 2A AR desensitization is reportedly inhibited by tumor necrosis factor (TNF)-α in human monocytoid THP-1 cells representing a novel cross-talk between TNF-α receptor and A 2A AR. TNF-α treatment in THP-1 cells not only reduced the translocation of GRK-2 to the plasma membrane but also decreased GRK-2 association with the plasma membrane preventing A 2A AR activity and enhancing receptor function [62]. Inhibitors of receptor internalization, such as hypertonic sucrose or concanavalin A did not affect agonist-stimulation of A 2A AR or agonist-induced desensitization of receptor response in NG108-15 cells. However, incubation of these cells with sucrose or concanavalin A did affect the resensitization of A 2A AR response following agonist removal [63].

A 2B AR
Point mutation or deletion studies of rat A 2B AR stably expressed in CHO cells revealed that a serine residue (Ser 329 ), close to the carboxy-terminus, is critical for the rapid agonist-induced desensitization and internalization of the receptor [64]. A 2B AR undergoes rapid agonist-induced desensitization and internalization in a GRK-2 and arrestin-dependent manner. Expression of a dominant negative mutant of GRK-2 in NG108-15 cells [65] or antisense-induced inhibition of non-visual arrestins (arrestin-2 and -3) in human embryonic kidney (HEK-293) cells [66] efficiently reduced the rate of agonist-induced desensitization and internalization of endogenous A 2B AR. Recycling of A 2B AR after agonist-induced endocytosis was also affected in cells with reduced arrestin levels. Interestingly, overexpression of arrestin-2 or arrestin-3 rescued A 2B AR internalization and recycling of the receptor protein. Overexpression of arrestin-3, however, showed a significantly faster rate of recycling than arrestin-2, suggesting an isoform specific role of arrestin in regulating A 2B AR trafficking [67]. In another study, A 2B AR was shown to internalize in an arrestin-and dynamin-sensitive manner [68].
On the other hand, second messenger-dependent kinases, such as PKA and PKC, did not seem to be involved in agonist-mediated phosphorylation and subsequent desensitization of A 2B AR [69]. Human astrocytoma cells which endogenously expressed A 2B AR showed that TNFα, a pro-inflammatory cytokine, markedly reduced agonist-dependent receptor phosphorylation on threonine residues and attenuated agonist-mediated A 2B AR desensitization [70]. TNFα-induced inhibition of A 2B AR desensitization could result in prolonged A 2B AR responsiveness and may contribute to the excessive astrocytic activation that occurs in neurodegenerative diseases.
In a recent study, it was proposed that A 2A AR is involved in the surface expression of A 2B AR in human embryonic kidney cell line, AD-293, transfected simultaneously with mouse A 2A AR and A 2B AR. It was found that newly synthesized A 2B AR is retained in the endoplasmic reticulum (ER) and is eventually targeted for degradation by the proteosomes. Inhibition of proteosome activity was not sufficient to enhance maturation and surface expression of the receptor. Furthermore, it was shown that co-transfection of A 2A AR with A 2B AR enhanced surface expression of A 2B AR through F(X) 6 LL motif in A 2A AR carboxy-terminus. These findings support the notion that A 2A AR and A 2B AR form heterodimer complexes for trafficking and function [71].

A 3 AR
Agonist occupation of A 3 AR results in rapid desensitization of receptor function as a result of phosphorylation of the receptor protein by members of the family of GRKs [72,73]. To identify the amino acid residue important for A 3 AR phosphorylation, Palmer and Stiles [74] demonstrated that a triple mutant, Thr 307 /Ala 307 , Thr 318 /Ala 318 and Thr 319 /Ala 319 within the carboxy-terminus, showed marked reduction in agonist-stimulated phosphorylation and desensitization of rat A 3 AR. Individual mutations of each residue showed that Thr 318 and Thr 319 are the two major sites for phosphorylation where Thr 318 appeared to be necessary to observe phosphorylation at Thr 319 , but not vice versa. Moreover, mutation of two palmitoylation sites within the carboxy-terminus, Cys 302 and Cys 305 of rat A 3 AR which controls the GRK phosphorylation sites, displayed a significant level of basal phosphorylation even in the absence of an agonist. This suggests an important regulatory role of these palmitoylation sites in receptor desensitization.
Receptor kinetics of A 3 AR was studied by Trincavelli et al. [75] in human astrocytoma cells. Short-term exposure to the agonist 2-chloro-N6-(3-iodobenzyl)-N-methyl-5'-carbamoyladenosine (Cl-IBMECA) caused rapid receptor desensitization followed by internalization within 15 and 30 min, respectively. Agonist removal resulted in recycling of the A 3 ARs to the cell surface within 120 min. Long-term exposure (1-24 h) resulted in marked receptor down-regulation and the restoration of receptor levels associated with recovery of receptor function was also slow (24 h). MAPK was shown to regulate agonist-induced phosphorylation of A 3 AR where its stimulation mediated activation of ERK1/2 within 5 min of agonist exposure in CHO cells stably expressing A 3 AR. Treatment with PD98059, a well-characterized MAPK kinase inhibitor, showed impaired receptor phosphorylation, desensitization and internalization by inhibiting GRK-2 translocation from cytosol to the plasma membrane. These data suggests that A 3 AR activation is regulated by ERK1/2 in a feedback mechanism which controls GRK-2 activity and prevents receptor phosphorylation [76]. These findings could explain the dual and opposite role of A 3 AR in neuroprotection and neurodegenerative diseases.

NF-κB Regulation of A 1 AR and A 2A AR Expression
Previous studies from our laboratories indicated that the expression of the A 1 AR is positively regulated by oxidative stress. This observation was made from studying the action of cisplatin, a chemotherapeutic agent, on the levels of A 1 AR in the chinchilla cochlea [77]. It was shown that cisplatin, via an oxidative stress pathway, induces A 1 AR in the cochlea which could serve as a last-ditch effort to rescue the cells in the inner ear from apoptosis. More detailed studies in DDT 1 -MF2 cells showed that cisplatin-induced A 1 AR expression resulted from activation of NF-κB by reactive oxygen species (ROS) produced by cisplatin. More importantly, an NF-κB consensus sequence was located 623 base pairs upstream of the start site of the human promoter A construct [78]. Such a mechanism of A 1 AR induction could help to precondition the cell or tissue to subsequent episodes of oxidative stress. For example, noise exposure, which increases ROS in the chinchilla cochlea (via NADPH oxidases), increases A 1 AR in the inner ear [79] while osmotic diuresis increases A 1 AR in the kidneys by a similar mechanism [80]. A similar mechanism could explain the induction of the A 1 AR and A 3 AR in the gut in a rabbit model of ileitis [81] and in the brain after the induction of cerebral ischemia in rats [82].
We have previously shown that activation of NF-κB is also a key regulator of A 2A AR expression. Treatment of PC12 cells with nerve growth factor (NGF) significantly reduced A 2A AR gene expression by an approximately three 3-fold within three days [83]. This effect was associated with a rapid activation of NF-κB via the low affinity p75 NGF receptor and was mimicked by other activators of NF-κB (e.g., ceramide, H 2 O 2 ) [84]. These results are consistent with findings of NF-κB consensus sites present in the A 2A AR gene promoter [85]. However, other mechanisms could be invoked to explain NGF regulation of A 2A AR. These include ERK and stress-activated protein kinase/c-Jun N-terminal kinase (JNK) [86].
To evaluate the significance of NF-κB in the regulation of the A 1 AR expression, we examined the expression of ARs in brain tissues from mice with deletion of the gene for the p50 subunit of NF-κB. This gene knockout (KO) renders these mice immune deficient, but they are viable and able to reproduce [87]. These mice showed reduced levels of A 1 AR in brain cortical plasma membranes, compared to their wild type counterparts. Interestingly, the levels of the G proteins alpha subunits (Gα i3 ) were also significantly reduced in the p50 KO mice, but the levels of Gα i1 were unchanged. The deficit in A 1 AR/G i protein expression was associated with increased neuronal apoptosis [88]. These findings suggest that NF-κB tonically regulates neuronal A 1 AR expression and survival.
Based on the observation that the A 2A AR is negatively regulated by NF-κB, we were interested in the expression of this receptor in p50 KO mice. We observed higher expression of the A 2A AR in striatal tissues from p50 KO mice, compared to wild type mice. As anticipated, less A 1 AR mRNA and protein was detected in the striatal tissues in p50 KO mice as compared with F2 mice [89]. These studies suggest that absence of the NF-κB p50 subunit leads to dysregulation of ARs in the striatum, as observed for the A 1 AR in other brain regions. Overall, these studies suggest an essential role of the NF-κB p50 subunit in the regulation of A 1 and A 2A AR expression in the brain.
Alterations in A 2A AR in striatum could alter locomotor activity, since this receptor exhibits inhibitory actions on dopamine D 2 receptor (D 2 R). The wild type and p50 KO mice did not show any difference in basal locomotor activity, but the p50 KO mice showed a hypersensitivity to caffeine-induced locomotor activity evaluated during the dark phase (when mice are normally active) [90]. The p50 KO mice also demonstrated increased sensitivity to intraperitoneal injections of SCH 58261, an A 2A AR antagonist [90] but not to the selective A 1 AR antagonist, DPCPX. These data suggest that the increase in A 2A AR in p50 KO mice provides larger striatal target (A 2A AR) for inhibition by caffeine, resulting in behavioral hypersensitivity.

Adenosine Receptor Oligomers
GPCRs are usually treated as individual units responsible for a specific signaling pathway. However, it is now known that these receptors can form complex structures with each other through physical links in its protein structure. When binding occurs between the same GPCRs, it forms a homodimer. On the other hand, when the link is between different receptors, it is a heterodimer. The term heteromer is currently defined as a macromolecular complex consisting of two or more functional receptors with different biochemical characteristics of individual receptors [91]. Moreover, the activation of a member of the heteromeric complex alters the binding properties of other receptor(s) present in the complex [91].
In this regard, a variety of papers in the literature indicate that ARs are capable of forming homodimers with each other and heterodimers with other GPCRs. One of the earliest evidence of the formation of homodimers of A 1 ARs has been demonstrated in intact tissue in the cortex of different species by immunoblotting and coimmunoprecipitation techniques [92]. Using immunogold electron microscopy, the authors confirmed the presence of homodimers. The presence of A 1 AR homodimers has also been demonstrated in hippocampal pyramidal neurons and Purkinje cells of the cerebellum [93]. Recently it was shown that HEK-293T cells transfected with the cDNA for the A 1 AR express receptor homodimers which promoted phosphorylation of ERK. The expression of A 1 AR homodimers in the cortex could explain the biphasic effects of the small and high doses of caffeine on motor activity [94]. Homodimers of A 2A AR receptors have been demonstrated in HeLa and HEK-293T cells co-transfected with different constructs of A 2A AR receptor. Such cells produce functional receptors expressed on the cell surface and the formation of the homodimer apparently was not dependent on the activity of the receptor, since treatment with the A 2A AR agonist CGS 21680 did not alter the bioluminescence resonance energy transfer (BRET) signal detected. Therefore, these data suggest that A 2A AR homodimers are constitutively expressed in native tissues and cultured cell lines [95]. However, it is yet unclear how these homodimeric A 1 ARs differ functionally from their monomeric counterparts.
As mentioned above, ARs can form heterodimers with each other. Ciruela and coworkers [96] showed that co-transfection of the A 1 and A 2A AR receptor cDNAs in HEK-293T cells led to the formation of A 1 /A 2A AR heterodimers on the cell surface. Interestingly, activation of the A 2A AR by CGS 21680 in these cells reduced the affinity of the A 1 AR receptor to a selective radioligand. However, activation of A 1 AR by the A 1 AR agonist, R-PIA, did not produce a reciprocal change in the affinity of the A 2A AR [96]. The ability of the A 2A AR to reduce the affinity of the A 1 AR was confirmed in experiments that measured the levels of intracellular calcium. Incubation of the A 1 /A 2A AR-transfected HEK-293T cells with R-PIA increased accumulation of intracellular calcium, which was drastically reduced when these cells were pretreated with CGS 21680 [96]. In vivo studies in rat striatum demonstrated co-localization of A 1 and A 2A ARs in the extrasynaptic terminals of axons as well as in the presynaptic active zone of excitatory glutamatergic neurons, an indication that these receptors could be forming heterodimers. Using synaptosomes enriched in striatal glutamatergic nerve terminals, the investigators showed that the A 1 AR inhibited K + -evoked glutamate release was abolished by A 2A AR activation [96]. In cultures of rat astrocytes, Christovão-Ferreira et al. [97] showed that A 1 and A 2A AR form heterodimers. In these cells, activation of the A 1 AR inhibited GABA uptake while activation of the A 2A AR increased uptake. Surprisingly, the effect of the A 1 AR was blocked not only by the selective antagonist DPCPX, but also by the SCH 58261, a selective A 2A AR antagonist. In turn, the effect of A 2A AR on GABA uptake was blocked by SCH 58261 and DPCPX, indicating cross antagonism and/or a physical interaction between these receptors [97]. Both receptors were shown to be internalized together when exposed to A 1 and A 2A AR agonists separately. Interestingly, the differential responses of A 1 and A 2A AR agonists on GABA uptake involved activation of the G i/0 and G s protein, respectively [97]. Thus, the presence of A 1 -A 2A AR heterodimers in the central nervous system could increase the complexity by which these two receptors regulate neuronal excitability. However, the advantage of A 1 -A 2A AR heterodimers regulation of GABA uptake over the effects of the individual ARs is unclear.
Evidence from the literature suggests that an interaction exists between adenosine receptors and adrenergic receptors. For example, activation of the A 1 AR in myocardium attenuates β 1 -induced cyclic AMP formation and PKA enzymatic activity [98]. In cardiomyocytes, activation of the A 1 AR receptor by selective agonist CCPA translocates protein kinase C ε (PKCε) to the membrane. This effect was attenuated by isoproterenol, a non-selective β-adrenergic receptor agonist or forskolin, an activator of adenylyl cyclase [99]. These results are suggestive of an interaction between A 1 AR and the β-adrenergic receptors. HEK-293T cells co-transfected with A 1 AR, β 1 and β 2 adrenergic receptors show A 1 AR/β 1 AR or A 1 AR/β 2 AR heterodimers which enhance the phosphorylation of ERK induced by isoproterenol or CCPA, as compared to cells transfected with only one type of receptor. In addition, the inhibitory effect of A 1 AR on cyclic AMP production mediated by CCPA was abolished in cells expressing both heterodimers. Analysis of samples of human heart tissue showed the formation of these heterodimers [100]. Thus, the data suggest that A 1 AR and adrenergic receptors are able to form heterodimers with different properties compared to the generally accepted pharmacological and functional properties of individual receptors.
The principal source of production of adenosine into the extracellular environment occurs through metabolism of ATP [11]. Thus, the amount of ATP released into the extracellular medium directly determines the activity of ARs. However, ATP itself influences the activity of its receptors, which are subdivided into P2X, P2Y, P2U and P2Z subtypes. All members of the P2Y family of receptors are G protein-coupled. Studies indicate that the P2Y1 receptor is able to form heterodimers with the A 1 AR in vitro. These heterodimers were less effective in inhibiting forskolin-stimulated cyclic AMP formation than the native A 1 AR [101]. Imaging by BRET in HEK-293T cells confirms the formation of P2Y1/A 1 AR heterodimers [102]. Immunostaining of rat cortical neurons showed that both receptors co-localize in the cell body and dendritic regions. In rat cortex slices these receptors are shown to be present together in cortical, hippocampal, and cerebellar neurons, mainly in the soma and dendrites, indicating that these receptors could be form heterodimers. Moreover, immunoprecipitation analysis and electron microscopy of these specific brain regions have reported an interaction between the A 1 AR and P2Y1 receptors [93,103].
A 2A ARs are highly expressed in the striatum [5], a region that also has an abundance of cannabinoid (CB)-1 receptors [104]. CB1 receptors are coupled to G i/o proteins and belong to a subtype of receptor for exogenous cannabinoids and endocannabinoids. The latter are represented by 2-arachidonoylglycerol (2AG) and N-arachidonoylethanolamine (anandamide), which differ from the majority of neurotransmitters because they are produced in the post-synaptic neuron and released by retrograde transport to the pre-synaptic neuron to perform its functions [104]. Immunofluorescence studies in coronal sections of rat striatum revealed a co-localization between A 2A AR and CB1 receptors in striatal neurons. Subsequent immunoprecipitation experiments confirmed A 2A AR-CB1 in striatal cells [105]. BRET assays in HEK-293 cells, co-transfected with A 2A AR and CB1 receptors, revealed the formation of heterodimers on the cell surface [105,106]. In vitro studies in human neuroblastoma SH-SY5Y cells constitutively expressing A 2A AR and CB1 receptors revealed a functional cross-talk between these receptors. When these cells were treated as selective CB1 receptor agonist, arachidonyl-2'-chloroethylamide (ACEA), the accumulation of cyclic AMP induced by forskolin was reduced. However, in the presence of the selective A 2A AR antagonist, ZM 241385, the decrease in cyclic AMP by ACEA was reduced. These data suggest that effective coupling of the CB1 receptor to G i protein requires prior or simultaneous activation of the A 2A AR [105]. Moreover, CGS 21680 increase cyclic AMP production in SH-SY5Y cells, which is blocked by either ZM 241385 or ACEA. Thus, activation of A 2A AR, leading to the production of cyclic AMP, required concurrent activation of the CB1 receptor [105]. Taken together, the data described above suggest that the A 2A AR can heterodimerize with the CB1 receptor, which controls how each of these receptors responds to their specific agonists [107,108].
The best studied heterodimer of ARs are with dopamine receptors (DRs). Using double immunofluorescence assays, Hillion et al. [109] demonstrated a high degree of co-localization between A 2A AR and D 2 R in the cell membranes of SH-SY5Y cells stably transfected with the human D 2 R and in cultures of striatal rat neurons. The existence of A 2A AR-D 2 R heterodimeric complex was confirmed by co-immunoprecipitation assays. Interestingly, co-administration of A 2A AR and D 2 R agonists to SH-SY5Y cells transfected with D 2 R cDNA resulted in co-aggregation, co-internalization and co-desensitization of the A 2A AR and D 2 R [109]. A 2A AR-D 2 R heterodimers were also revealed in HEK-293T cells co-transfected with the cDNAs encoding these receptors by fluorescence resonance energy transfer (FRET) and co-immunoprecipitation studies [110][111][112]. In HEK-293T cells co-transfected with A 2A AR and D 2 R, Borroto-Escuela et al. [110] demonstrated that quinpirole (D 2 R agonist) induced internalization of the receptor heterodimer complex, which was facilitated by CGS 21680 (A 2A AR agonist). Furthermore, internalization of the complex was shown to be dependent on β-arrestin2 and Akt [110]. Other data also show that the A 2A AR-D 2 R complex can associate with the CB1 receptor [106] or the metabotropic glutamate receptor, mGluR5 [113] to form a trimeric complex. Studies have also shown formation of A 1 AR-D 1 R heterodimers. For example, co-transfection of mouse fibroblasts with A 1 AR and D 1 R cDNAs led to the formation of heterodimers. These heterodimers disappeared when these cells are pretreated with the SKF 38393, a D 1 R agonist. Using immunofluorescence labeling and confocal microscopy detection, co-localization or A 1 AR and D 1 R were observed in fibroblast cultures and cortical neurons [114]. Immunoprecipitation studies of tissues obtained from the rat nucleus accumbens demonstrated the presence of A 1 AR-D 1 R heterodimers in this brain region. Interestingly, the levels of heterodimers were significantly reduced in rats previously treated with cocaine, suggesting that activation of the D 1 R component of the heterodimer complex triggers its dissolution [115]. The presence of heterodimers between ARs and DRs in cells of the central nervous system would suggest some added functional roles of these complexes other than those mediated by activation of the individual receptors. In the striatum, one can envision that the AR-DR heterodimers could help to fine tune the intracellular signaling processes emanating from extracellular adenosine and dopamine or therapeutic agents directed to these receptors. As such, a better appreciation of AR-DR heterodimers could provide better insights to the treatment of diseases involving the striatum, such as Parkinson's and Huntington's disease [116][117][118].

Adenosine Receptors in the Control of Sleep
Adenosine has been shown to serve as a sleep promoting factor. The levels of this nucleoside in the basal forebrain increase during prolonged wakefulness and resolves during sleep [119]. Furthermore, individuals expressing an adenosine deaminase polymorphism, which increases adenosine, show deeper sleep and higher slow-wave activity (SWA) [120]. Infusion of adenosine in the basal forebrains increases sleep in rats [121] and cats [122]. Similarly, inhibition of adenosine intracellular uptake by inhibition of its equilibrative transporter produces electrophysiological and behavioral parameters similar to those produced by sleep deprivation [123]. Several pieces of evidence support a role of the A 1 AR (localized to the basal forebrain) in mediating the sleep promoting actions of adenosine. For example, systemic or cerebroventricular administration of A 1 AR agonists to rats led to increased sleep drive [124,125]. Administration of selective A 1 AR antagonist or A 1 AR antisense oligonucleotides (to reduce A 1 AR expression) increased wakefulness in rats. Interestingly, an increase in A 1 AR density in the basal forebrain was observed following sleep deprivation, which could contribute to the subsequent sleep rebound [124]. It is believed that the source of the excessive levels of adenosine during wakefulness derives from astrocytes and is released via soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE)-dependent gliotransmission [126]. Mice expressing an astrocyte-specific dominant negative SNARE protein show reduced SWA, compared with their wild-type counterparts [126]. This deficit could be produced in wild-type mice by intracerebroventricular infusion of cyclopentyltheophylline. In addition, these investigators observed that mice expressing the dominant negative SNARE protein had a lower level of memory deficit induced by sleep deprivation as compared to their wild-type littermates. This suggests a role of adenosine (produced during sleep deprivation) in mediating the memory deficits [126].
Recent studies using A 1 AR and A 2A AR KO mice, clearly demonstrate a role of the A 2A AR (not the A 1 AR) in mediating both the sleep suppressing and arousal actions of caffeine [127]. However, these investigators showed that the A 1 AR in the tuberomammillary nucleus mediates non-rapid eye movement (non-REM) sleep by inhibiting histamine release [128].
Since the A 1 and A 2A ARs are induced by NF-κB in cell cultures [78,84], we examined the expression of these receptors in the cortex and striatal tissues of mice with a global deletion in the p50 subunit of NF-κB. As anticipated, these mice showed reduced expression of the A 1 AR in different regions of the brain and significant elevations in A 2A AR in the striatum, suggesting that NF-κB activation controls the normal expression of these receptors. We also evaluated the sleep patterns of these mice to determine whether the modest decrease in A 1 AR would produce any changes. These mice exhibited more slow wave and REM sleep under normal conditions than their wild type counterparts (B6129PF2/J strain) [129]. This finding was surprising, based on the purported role of the A 1 AR in regulating sleep (see above) in rats. Accordingly, a lower level of A 1 AR expression should lead to reduced sleep duration. The p50 KO mice also demonstrate a more rapid recovery to normal sleep patterns following sleep deprivation. These data suggest dissociation between A 1 AR expression and normal sleep patterns in mice and in mediating the homeostatic drive following sleep deprivation.
The p50 KO mice also showed increased expression of the A 2A AR in the striatum [89] and possibly in other brain regions. The increased striatal expression of A 2A AR in the KO mice could account for their increased responsiveness to caffeine [89]. Since the A 2A AR has been implicated in mediating the increased sleep propensity to adenosine and arousal actions of caffeine [127], it is tempting to speculate that the increase in cortical A 2A AR could explain the increases in slow wave and REM sleep observed in the P50 KO mice, in addition to their rapid recovery following sleep deprivation (see Figure 1).

Adenosine Receptor and Protection against Hearing Loss
Early studies indicate that adenosine could modulate afferent neurotransmission in hair cell of the frog labyrinth system, a model system for studying hair cell [130]. These investigators showed that adenosine and adenosine uptake inhibitors reduced, while theophylline and adenosine deaminase increased, firing of hair cell afferents [130]. A subsequent study by Dulon et al. [131] demonstrated that application of adenosine to mammalian inner hair cells increased intracellular Ca 2+ levels. Direct demonstration of A 1 AR in the rat cochlea, using several biochemical techniques, was later provided [132]. These receptors did not alter sound-evoked endocochlear potentials but activated antioxidant enzymes in the cochlea, thereby reducing the basal levels of malondialdehyde, a marker of lipid peroxidation and cell damage [77]. Activation of the A 1 AR contributes to the otoprotective actions of the adenosine analog, R-PIA, against cisplatin ototoxicity ( [133,134] see Figure 2) and noise-induced hearing loss [135]. An interesting aspect of the A 1 AR in the cochlea is that it could be induced by oxidative derived from certain chemotherapeutic agents (such as cisplatin and aminoglycoside antibiotics) and noise. Induction of A 1 AR appears to be dependent on the activation of NF-κB by ROS which enhance transcriptional activity of the A 1 AR gene [78]. We propose that such feedback regulation could increase the cytoprotective activity of the A 1 AR in response to oxidative stress. The NADPH oxidase system appears to contribute significantly to cochlear ROS produced by noise trauma and cisplatin. Specifically, the NOX3 isoform of NADPH oxidase is localized primarily to the cochlea and mediates cisplatin ototoxicity [136]. As such, knockdown of NOX3 by trans-tympanic administration of NOX3 siRNA, protected against cisplatin ototoxicity [137]. ROS can increase DNA damage, and modify membrane lipids, cytosolic proteins and receptors on the cell surface [138]. Increased lipid peroxides appear to induce cochlear damage following noise exposure [139]. This explains the protective actions of antioxidants to treat hearing loss induced by noise or therapeutic agents [140]. However, the use of these agents to treat cisplatin-induced ototoxicity is limited by the potential interaction of the antioxidants and cisplatin to reduce its chemotherapeutic efficacy.
Recent studies have revealed another target of ROS, induction of inflammation in the cochlea. ROS can increase inflammatory processes in the cochlea by activating NF-κB [78,79]. NF-κB is a known mediator of inflammatory processes in the body [141]. More recently, we have shown that oxidative stress in the cochlea can contribute to the inflammatory process by activating signal transducer and activator of transcription 1 (STAT1) transcription factor [142]. In addition, STAT1 is able to couple the activation of transient receptor potential vanilloid receptor (TRPV)-1 in the cochlea to the induction of inflammation [143]. Accordingly, down-regulation of STAT1 ameliorates cisplatin-induced ototoxicity in the rats [142]. It is possible that the otoprotective actions of the A 1 AR against cisplatin ototoxicity involve inhibition of NF-κB and STAT1 transcription factors.
The cochlea expresses three subtypes of the ARs [144]. The A 1 ARs are expressed in inner hair cells, Deiters cells and spiral ganglion neurons. The A 2A AR are present in the inner hair cells, Deiters cells, spiral ligament and spiral ganglion. However, the A 3 AR appear to have a more wide-spread expression. While it is clear that the activation of the A 1 AR confers otoprotection, the role of the A 3 AR is less clear. On the other hand, activation of the A 2A AR appears to exacerbate cisplatin-induced ototoxicity [134]. As such, potentially useful AR drugs to treat ototoxicity could include agonists of the A 1 AR and antagonists of the A 2A AR.

Figure 2.
Protection of rat cochlear outer hair cells from cisplatin-induced damage by A 1 AR agonist, R-PIA. Rats were pre-treated with R-PIA (10 µL of a 10 µM solution added to the round window). The remaining liquid was removed after 1 h and rats were administered cisplatin by intraperitoneal injections (16 mg/kg). Panels A-C are electron micrographs of the organ of Corti from rats treated with cisplatin and represent the hook, basal turn and the middle turn, respectively; Panels D-F shows the effect of pre-treatment with R-PIA for the respective regions; Panels G-I show the effect of R-PIA in presence of DPCPX, an A 1 AR antagonist, on cisplatin-induced damage. The inner hair cells are the three rows of cells with the "V" shaped stereociliary bundles. The inner hair cells are barely visible on the top of each micrograph. The data which was confirmed from different cochlear preparations support a protective role of the A 1 AR against cisplatin ototoxicity. (Reprinted with permission from [134], Copyright 2004 Elsevier). Scale bar = 5 µm.

Adenosine Receptors in Cancer
The expression levels and function of A 1 AR in different cancer types is highly debatable. Panjehpour et al. [145] using semi-quantitative RT-PCR, reported that there was no significant difference between the expression levels of A 1 AR in non-neoplastic normal tissues and human breast tumor tissues. However, in another study, Mirza et al. [146] found high levels of A 1 AR in all breast tumor cell lines as well as in 15 of 24 human primary breast cancer tissues compared to normal mammary epithelial cells and matched normal breast tissues, respectively. In this study, knockdown of A 1 AR using siRNA attenuated MDA-MB-468 human breast tumor cell growth and proliferation by impairing the G 1 checkpoint in the cell cycle. High expression of A 1 AR was also found in human colorectal adenocarcinoma [147], human leukemia Jurkat T cells [148] and human melanoma A375 cell lines [149]. Increased A 1 AR was also observed in the peritumoral regions of F98 glioblastoma cells [150]. The source of increased A 1 AR expression was found to be activated astrocytes and microglia which suppressed growth of glioblastoma upon A 1 AR activation. Despite its role in tumorigenesis, few studies have shown A 1 AR to have anti-tumor effects. In CW2 human colon cancer cells, A 1 AR activation increased apoptosis by activating caspases [151]. Anti-proliferative effects of A 1 AR were also reported in LoVo human metastatic colon cancer cells [152] and Sertoli-like TM4 cells [153].
A 2A AR are found to be over-expressed in several cancer cell lines such as MCF-7 human breast cancer cells [154], Jurkat T-cell leukemia [148], A375 melanoma cells [149], U87MG human glioblastoma cells [155], HT29 and DLD-1 colon carcinoma cells [156], SH-SY5Y neuroblastoma cells [157], and non-small cell lung cancer cells [158] where it is shown to affect cell proliferation, apoptosis, angiogenesis and anti-tumor immunity. Activation of the A 2A AR by CGS 21680 was shown to stimulate cell proliferation in MCF-7 breast cancer cells [154]. A 2A AR also induced angiogenesis by promoting endothelial cell proliferation, migration and tube formation which is an important hallmark of cancer [159,160]. In another study, A 2A AR was reported to protect the tumor by inhibiting anti-tumor T cells and A 2A AR antagonist facilitated CD8+ T cell-mediated retardation of tumor growth [161]. These findings warrant the use of A 2A AR antagonists to inhibit tumor growth and angiogenesis. In contrast, A 2A AR activation in A375 melanoma cells reduced cell viability and cell clone formation [162], and activated caspases to induce apoptosis in Caco-2 human colon cancer cells [163].
The role of A 2B AR in cancer can be broadly classified as pro-angiogenic. A 2B AR activation in human retinal endothelial cells stimulated neovascularization through production of vascular endothelial growth factor (VEGF) [164]. A 2B AR knockout mice, administered with Lewis lung carcinoma cells, demonstrated reduced tumor growth and longer survival times, compared to wild type control mice. It was suggested that tumor cells promote their growth by exploiting A 2B AR-dependent regulation of VEGF in host immune cells [165]. In addition, knockdown of A 2B AR by siRNA inhibited the release of pro-angiogenic growth factor, interleukin-8, in A375 human melanoma cells treated with chemotherapeutic agents like etoposide and doxorubicin [166]. The A 2B AR promoter region contains a functional binding site for hypoxia-inducible factor (HIF), which leads to the induction of A 2B AR under hypoxic conditions [167]. A 2B AR is also over-expressed in colorectal carcinoma cells grown under hypoxic state, the inhibition of which significantly attenuated cell proliferation [168]. A 3 ARs are over-expressed in different types of cancer cells, such as PC-3MM human prostate cancer cells [169], colon and breast carcinoma tissues [170], A375 human melanoma cells [146], U87MG human glioblastoma cells [155,171], HL60 human promyelocytic leukemia cells [172], and hepatocellular carcinoma [173] compared to their normal counterparts, suggesting that these receptors could serve as potential molecular markers of these cancers. A role of the A 3 AR in mediating anti-tumor actions has been demonstrated in in vitro and in vivo models. For example, A 3 AR agonists inhibit the growth of melanoma [174], colon [175], leukemia [172] and prostate cancers [169,176] in animal models. A 3 AR agonist also inhibited liver metastasis in mice inoculated with human HCT-116 or murine CT-26 colon carcinoma cells [175]. The anti-tumor action of A 3 AR agonists could be explained by an increase in natural killer (NK) cell activity, which promotes killing of tumor cells [177]. In prostate cancer cells, Jajoo et al. [169] demonstrated that activation of the A 3 AR led to suppression of the high levels of ROS generated by these cells by inhibiting NADPH oxidases ( Figure 3). Interestingly, activation of this receptor reduced ERK1/2 activity which normally controls the phoshorylation and activity of the p47 phox subunit of this enzyme. Figure 3. Activation of A 3 AR suppresses ROS generation in AT6.1 prostate cancer cells. Cells were pretreated with the A 3 AR agonist, IB-MECA, the antagonist, MRS1523, or a combination of IB-MECA + MRS1523. ROS was determined by H 2 DCFDA fluorescence. IB-MECA significantly reduced fluorescence in these cells (A,B) which was reversed by MRS1523, indicating a role of the A 3 AR in this process. Similar effects were obtained with DPI and AEBSF, known inhibitors of NADPH oxidase (C,D). Experiments were replicated four times. Histograms represent the mean ± SEM. Asterisks (*) and (**) indicates statistically significant difference (p < 0.05) from vehicle and IB-MECA treatments, respectively. (Reprinted with permission from [169], Copyright 2009 Neoplasia Press, Inc.).

Conclusions
As described above, adenosine and ARs play a dynamic role in regulating normal cell physiology and also act as modulators in disease processes. A better understanding of the functions of these receptors, especially the newly identified receptor homomers and heteromers, could stimulate development of new therapies for the treatment of diseases.