Purinergic Signalling

, Volume 4, Issue 1, pp 21–37

Internalization and desensitization of adenosine receptors

  • Elisabeth C. Klaasse
  • Adriaan P. IJzerman
  • Willem J. de Grip
  • Margot W. Beukers
Open AccessReview

DOI: 10.1007/s11302-007-9086-7

Cite this article as:
Klaasse, E.C., IJzerman, A.P., de Grip, W.J. et al. Purinergic Signalling (2008) 4: 21. doi:10.1007/s11302-007-9086-7


Until now, more than 800 distinct G protein-coupled receptors (GPCRs) have been identified in the human genome. The four subtypes of the adenosine receptor (A1, A2A, A2B and A3 receptor) belong to this large family of GPCRs that represent the most widely targeted pharmacological protein class. Since adenosine receptors are widespread throughout the body and involved in a variety of physiological processes and diseases, there is great interest in understanding how the different subtypes are regulated, as a basis for designing therapeutic drugs that either avoid or make use of this regulation. The major GPCR regulatory pathway involves phosphorylation of activated receptors by G protein-coupled receptor kinases (GRKs), a process that is followed by binding of arrestin proteins. This prevents receptors from activating downstream heterotrimeric G protein pathways, but at the same time allows activation of arrestin-dependent signalling pathways. Upon agonist treatment, adenosine receptor subtypes are differently regulated. For instance, the A1Rs are not (readily) phosphorylated and internalize slowly, showing a typical half-life of several hours, whereas the A2AR and A2BR undergo much faster downregulation, usually shorter than 1 h. The A3R is subject to even faster downregulation, often a matter of minutes. The fast desensitization of the A3R after agonist exposure may be therapeutically equivalent to antagonist occupancy of the receptor. This review describes the process of desensitization and internalization of the different adenosine subtypes in cell systems, tissues and in vivo studies. In addition, molecular mechanisms involved in adenosine receptor desensitization are discussed.


Adenosine receptorsβ-arrestinsCaveolaeDesensitizationG protein-coupled receptor kinaseLipid raftsInternalizationPalmitoylationPhosphorylation



adenosine A1 receptor


adenosine A2A receptor


adenosine A2B receptor


adenosine A3 receptor


adenosine deaminase


Arf nucleotide site opener


β1 adrenergic receptor






[3H] N6-cyclohexyladenosine






dopamine D1 receptor



GIRK channels

G protein-activated inwardly rectifying K+ channels


G protein-coupled receptor kinase


G protein-coupled receptor




serotonin, 5-hydroxytryptamine



MRS 1523

3-propyl-6-ethyl-5-[(ethylthio)carbonyl]-2 phenyl-4-propyl-3-pyridine carboxylate

MRS 1706





protein kinase A


R-N6-phenylisopropyl adenosine

ZM 241385

(4-(2-[7-amino-2-(2-furyl {1,2,4}-triazolo {2,3-a {1,3,5}triazin-5-yl-aminoethyl)phenol


Adenosine is an important neuromodulator involved in a variety of brain activities and it also serves many different functions in the periphery. This nucleoside, when extracellular, exerts its action via specific G protein-coupled receptors (GPCRs) of the P1 class, divided into four subtypes: A1R, A2AR, A2BR and A3R [1]. GPCRs consist of a single polypeptide, containing seven α-helices which are oriented perpendicular to the membrane. The N terminus is located at the extracellular side of the cell and often contains one or more glycosylation sites. The C terminus is located intracellularly and contains phosphorylation and palmitoylation sites, which are involved in regulation of receptor desensitization and internalization [2]. All adenosine receptors, with the exception of the A2AR, contain a palmitoylation site near the C terminus. The A2AR is the only subtype with an extraordinary long C terminus, 122 amino acids versus 36 amino acids in e.g. the A1R [3]. All the adenosine receptors are glycosylated on the second extracellular loop, although glycosylation does not appear to influence ligand binding. The third intracellular loop and/or the C terminus are involved in coupling the adenosine receptors to G proteins. Phosphorylation of in particular intracellular loop 3 is involved in desensitization and internalization of adenosine receptors [46].

Adenosine and analogues

Adenosine, consisting of a purine ring connected to a ribose group, is the endogenous ligand for the adenosine receptors (Fig. 1). Under normal conditions, adenosine is continuously formed extracellularly by dephosphorylation of ATP, ADP and/or AMP to adenosine by NTPDases (ectonucleoside triphosphate diphosphohydrolases). However, the A3R can also be activated by inosine, a breakdown product from adenosine [5]. Most adenosine receptor agonists are analogues of adenosine, modified by N6, C2 and C8 substitutions at the adenine base, and C5′ modifications of the ribose moiety [5, 6]. Antagonists lack the ribose group and usually possess a mono-, bi- or tricyclic core structure, e.g. caffeïne, which contains a xanthine as basic structure (Fig. 1). For extended reviews on high affinity agonists and antagonists for adenosine receptors and their structure-activity relationships, see Palmer and Stiles [1], Fredholm et al. [5], Jacobson and Gao [7], Beukers et al. [8], Müller [9] and Klotz [10].
Fig. 1

Chemical structures of the endogenous ligand adenosine and the antagonist caffeine

Occurrence and physiological functions of adenosine receptors

The adenosine receptors are widespread throughout the body and exert many different functions both in the CNS and in the periphery.

The A1R is particularly prevalent in the central nervous system, with high levels in the cerebral cortex, hippocampus, cerebellum, thalamus, brain stem and spinal cord. Numerous peripheral tissues also express the A1R, including vas deferens, testis, white adipose tissue, stomach, spleen, pituitary, adrenal gland, heart, aorta, liver, eye and bladder. Low levels are found in the lung, kidney and small intestine [1, 5, 6]. The A1R is involved in cardiovascular effects (e.g. reducing heart rate), inhibition of lipolysis and stimulation of glucose uptake in white adipocytes and the modulation of neurotransmitter release in the CNS [1]. The A1R also plays a role in anxiety, hyperalgesia, bronchoconstriction and the glomerular filtration rate and renin release in the kidney [5, 6, 11].

In the CNS, the A2AR is highly expressed in the striatum and olfactory tubercle [1]. In the periphery, it is highly expressed in the spleen, thymus, leucocytes and blood platelets, and intermediate levels are found in the heart, lung and blood vessels [5, 6]. The A2AR is involved in the onset of vasodilation, inhibition of platelet aggregation, exploratory activity, aggressiveness and hypoalgesia [1]. In addition, A2AR plays a role in Parkinson’s disease, Huntington’s disease, Alzheimer’s disease, ischaemia, attenuation of inflammation and neuroprotection, particularly in peripheral tissues [5, 6]. A2A receptor antagonists slow the neurodegeneration which occurs in Parkinson’s and Huntington’s disease and also prevent toxicity induced by beta-amyloid in the development of Alzheimer’s disease [1214].

The A2BR is widely expressed in the brain, but generally at very low levels. In the periphery, high levels of A2BR were detected in the caecum, large intestine and urinary bladder. Lower levels were observed in the lung, spinal cord, vas deferens, pituitary, adipose tissue, adrenal gland, kidney, liver and ovaries [5, 6]. Since there is a lack of specific agonists for the A2BR, little is known about the functional significance of this receptor. However, the A2BR plays a role in mediating vasodilation in a.o. the aorta, the renal artery and the coronary artery of different species. It is also involved in allergic and inflammatory disorders [5, 6].

The A3R is expressed in the CNS, but at relatively low levels, and only the hypothalamus and the thalamus have been reported to contain A3R [15]. The highest levels of adenosine A3R have been found in the lung and liver, and somewhat lower levels were found in the aorta [1]. In addition, the A3R was found in eosinophils, mast cells, testis, kidney, placenta, heart, spleen, uterus, bladder, jejunum, aorta, proximal colon and eye, although with pronounced differences in expression level between species [5, 6, 16]. The A3R has been implicated in mediating allergic responses, airway inflammation and apoptotic events; however, the latter is dependent on the cell type involved and/or the type of activation [5, 16]. Furthermore, the A3R is involved in the control of the cell cycle and inhibition of tumour growth both in vitro and in vivo [6]. In fact, adenosine A3 receptors have been demonstrated to be more highly expressed in tumours than in healthy cells, suggesting a role for A3R as a tumour marker [17].

Signal transduction of adenosine receptors

G protein-coupling and second messengers

Heterotrimeric G proteins are guanine-nucleotide regulatory protein complexes composed of α and βγ subunits. They are responsible for transmitting signals from G protein-coupled receptors to effectors, e.g. adenylyl cyclase. Until now, 16 α, 5 β and 14 γ isoforms have been reported [18]. G proteins are divided into several subclasses with a specific activity profile: Gs proteins stimulate adenylyl cyclase, Gi proteins inhibit adenylyl cyclase and stimulate GIRK channels, G0 proteins stimulate K+ ion channels, Gq/11 proteins activate phospholipase C, G12 proteins activate Rho guanine-nucleotide exchange factors (GEFs) and the olfactory G protein, Golf, stimulates adenylyl cyclase. Upon receptor activation, both the α subunit and the βγ subunit can signal, but to different effectors [1921]. For a recent review on G proteins, see Milligan and Kostenis [18].

The A1R is usually coupled to a pertussis toxin-sensitive \({\text{G}}_{{{\text{i}}\alpha }} \) protein, which mediates inhibition of adenylyl cyclase and regulates calcium and potassium channels [1, 3, 5, 6, 11]. Both the third intracellular loop and the C-terminal tail of the A1R are involved in \({\text{G}}_{{{\text{i}}\alpha }} \) coupling [5]. In addition, it has been reported that under certain conditions the A1R couples to Gs to stimulate adenylyl cyclase, or to Gq/11 to stimulate inositol phosphate production. Apparently, the specific activity state of the receptor or the nature of the agonist determine which G protein class is activated by the A1R [22, 23]. The A2AR in the periphery is coupled to cholera toxin-sensitive \({\text{G}}_{{{\text{s}}\alpha }} \) proteins, which increase adenylyl cyclase activity upon receptor activation. The A2AR in the striatum is presumably coupled to Golf [5, 6]. The third intracellular loop, but not the C terminus of the A2AR, is involved in \({\text{G}}_{{{\text{s}}\alpha }} \) coupling [5]. The A2BR is coupled to \({\text{G}}_{{{\text{s}}\alpha }} \) proteins leading to stimulation of adenylyl cyclase upon receptor activation [6]. There is quite some evidence that A2BR can activate phospholipase C as well, via Gq/11 proteins [5]. The A3R is coupled to pertussis toxin-sensitive \({\text{G}}_{{{\text{i}}\alpha }} \) proteins, which mediate inhibition of adenylyl cyclase. In addition, A3R can stimulate phospholipase C via Gq/11 proteins [5, 6]. For an extensive overview of adenosine receptor-G protein coupling, see Fredholm et al. [24].

Desensitization and internalization—general principles and players

Mechanisms to dampen GPCR signalling exist at every level in the cell. In this section attention will be paid to the underlying principles of desensitization and internalization and the protein partners involved in these processes.

Receptor localization

Depending on the localization signal, GPCRs in the plasma membrane can be targeted to lipid rafts1 or caveolae2. Different regions of GPCRs can influence not only the targeting to either lipid rafts or caveolae but may also enable an interaction of the receptor with constituents of these rafts and caveolae. For instance the extracellular part of the receptor might interact with GM1 gangliosides (glycosphingolipids) present in lipid rafts/caveolae. In addition, the C-terminal fatty acid acylation or palmitoylation may also affect targeting of GPCRs to either lipid rafts or caveolae. Finally, transmembrane regions may interact with cholesterol in the lipid rafts resulting in a change in conformation of the α-helices. Since the conformation of α-helices depends on the activation state of GPCRs, it may well be that agonist binding to the receptor may affect its localization in lipid rafts by means of molecular transitions leading to receptor activation [25, 26].


Desensitization reduces receptor activity and plays a role in signal duration, intensity and quality. Desensitization is initiated by phosphorylation of serine and/or threonine residues in the third intracellular loop and C terminus of the receptor. Two types of desensitization occur, heterologous and homologous desensitization, and both are the result of receptor phosphorylation. Heterologous desensitization is induced by phosphorylation of the receptor by protein kinase A or C—sometimes even without agonist occupancy. On the other hand, homologous desensitization is specific for agonist-occupied receptors and consists in most cases of two steps. First, the receptor is phosphorylated by one of the G protein-coupled receptor kinases (GRKs 1-7); then it binds to β-arrestin, of which two subtypes exist, which exhibit high affinity for agonist-occupied, phosphorylated receptors. β-Arrestin serves to sterically inhibit G protein coupling, thereby terminating the G protein activation, and may also target the receptor to clathrin-coated pits3 for internalization (Fig. 2) [19, 25, 27, 28].
Fig. 2

Homologous (a) and heterologous (b) desensitization and subsequent internalization of GPCRs. \(\alpha = {\text{G}}_{{{\text{i}}\alpha }} \) subunit, \(\beta \gamma = {\text{G}}_{{{\text{i}}\beta \gamma }} \) subunit, L ligand, GRK G protein-coupled receptor kinase, P phosphorylated amino acids, β-arr β-arrestin, AP2 adaptin, E effector, second messenger, PKA protein kinase A, PKC protein kinase C


Receptor desensitization, initiated by phosphorylation of the receptor by different protein kinases (A or C) or GRKs, can be subsequently followed by receptor internalization. Upon phosphorylation, β-arrestin 1 or 2 is attracted to the receptor [31]. β-Arrestins not only interact with the phosphorylated receptor, but also bind to the heavy chain of clathrin, to the β2-adaptin subunit of the clathrin adaptor protein AP2, and to phosphoinositides. These interactions direct the phosphorylated receptor to punctate clathrin-coated pits in the cell membrane, which are internalized by action of the GTPase dynamin. Upon internalization, receptors can either be rapidly recycled to the plasma membrane, targeted to larger endosomes and slowly recycled, or degraded in lysosomes. The final destination of the internalized receptors largely depends on the β-arrestin subtype (1 or 2) that is recruited by the receptor upon phosphorylation and the duration of β-arrestin binding. In this way, internalization may regulate receptor resensitization and contributes to a positive regulation of receptor signalling [19, 25, 31].

Internalization pathways

From internalization studies with several receptors, it appears that the internalization pathway is specific for receptor type, cell type, metabolic state of the cell, cell-specific factors etc. Receptor trafficking can be regulated in different ways (Fig. 3): (a) the receptor resides mainly in lipid rafts/caveolae and enters the cell via this pathway by default; (b) the receptor is in lipid rafts, but leaves these upon agonist binding to be internalized via clathrin-coated pits; (c) the receptor moves into lipid rafts upon agonist binding and is internalized via this pathway; (d) the receptor moves into lipid rafts after agonist binding to activate certain signalling events, but is eventually moved out of the lipid rafts to be internalized via clathrin-coated pits. Internalization can even be achieved via uncoated vesicles or by a combination of two or more of the aforementioned pathways. For example, β1-AR is internalized via both lipid rafts and clathrin-coated pits. PKA phosphorylation directs β1-AR to a clathrin-coated pit, whereas GRK phosphorylation directs the receptor to lipid raft-mediated internalization [19, 25, 26].
Fig. 3

Different internalization pathways, adapted from Chini and Parenti, 2004 [25]. Internalization via (a) lipid rafts/caveolae, (b) upon agonist binding, the receptor moves to clathrin-coated pits to be internalized, (c) the receptor moves into lipid rafts upon agonist binding and is internalized, (d) the receptor moves into lipid rafts upon agonist binding to activate certain signalling events, but is eventually moved out to be internalized via clathrin-coated pits. Ligand (L, green triangle), clathrin-coated pits (dotted blue lines) and lipid rafts (solid pink lines) are indicated

Function of lipid rafts/caveolae

The existence of lipid rafts/caveolae serves different functions. First of all, lipid rafts act as ‘stations’ in which GPCRs accomplish specific signalling tasks by meeting a selected set of signalling molecules, e.g. G proteins and adenylyl cyclases. Another possible function of lipid rafts is the protection of receptors from rapid constitutive or agonist-induced internalization, thus allowing their coupling to specific signalling pathways. In addition, caveolin may regulate the constitutive activity of receptors [32]. Finally, the endocytic pathways that GPCRs choose may depend on cell-specific factors. Switching the internalization pathway from lipid rafts/caveolae to clathrin-coated pits may alter the final receptor destination [19, 25, 26]. Research on adenosine receptors probably provided the first account of receptor internalization via caveolae and lipid rafts, as an alternative to the well-described β-arrestin pathway [33].

G protein-coupled receptor kinases (GRKs)

Upon agonist binding to GPCRs, not only G proteins are recruited to the activated receptor conformation, but also two other protein families, namely the G protein-coupled receptor kinases (GRKs) and the β-arrestins [31]. The GRK family consists of seven different genes. The GRKs have been divided into three protein subfamilies, based on sequence similarity. GRK1 and GRK7 belong together. The second subfamily consists of GRK2 and GRK3 since their membrane recruitment depends on interaction with Gβγ subunits and phosphatidylinositol 4,5-bisphosphate. GRK2 and GRK3 are primarily responsible for agonist-dependent receptor phosphorylation, β-arrestin recruitment and functional uncoupling of the receptor. GRK4, GRK5 and GRK6 form the third subfamily and are constitutively associated with the membrane. GRK1 and GRK7 are expressed mainly in the retinal rods and cones, whereas GRK4 expression is limited to the cerebellum, testis and kidney. In contrast, GRK2, GRK3, GRK5 and GRK6 are widely expressed in mammalian tissues [31].


β-Arrestins belong to the arrestin family of which four members have been identified. Arrestin 1 and arrestin 4 are expressed in the visual system in retinal photoreceptor cells; arrestin 2 (also called β-arrestin 1) and arrestin 3 (also called β-arrestin 2) are more widely expressed and are involved in the regulation of nonvisual GPCRs. The β-arrestins were originally discovered as molecules that bind to and desensitize the activated and phosphorylated form of G protein-coupled receptors as described before [19, 25, 27, 28, 31]. The β-arrestin protein consists of an N and a C domain, which are almost entirely composed of antiparallel β sheets, connected by a linker of 12 amino acids. A ‘polar core’ is embedded between those two domains, and its disruption by the phosphorylated C terminus of the activated receptor leads to conformational changes in the β-arrestin. The C tail of the β-arrestin is then released, exposing both the clathrin and the AP2 binding domains. Recently, it appeared that β-arrestins not only are involved in the desensitization of G protein-coupled receptors, but also act as signal transducers on their own. As a consequence, they emerge as multifunctional adaptor/scaffold proteins which mediate cellular processes such as chemotaxis, apoptosis and metastasis besides receptor signalling and trafficking. For an extensive and recent review on the interactions of β-arrestins with other cellular proteins, see Lefkowitz et al. [34].

GRKs and β-arrestins orchestrate GPCR activities at three different levels: (1) silencing: the functional uncoupling of the receptor from its G protein by a mechanism known as ‘homologous desensitization’, (2) trafficking, which involves receptor internalization, resensitization and/or degradation, and (3) cross-signalling: activation or inhibition of intracellular signalling pathways, independent of heterotrimeric G proteins [31].

Dampening adenosine receptor signals

In this section we will discuss the evidence for and mechanisms of desensitization and internalization of the four adenosine receptor subtypes. For each receptor we will first summarize results from in vitro, ex vivo and in vivo studies. This will be followed by a more biochemical approach in which we will focus on the molecular mechanisms of adenosine receptor desensitization and internalization by paying attention to the role of accessory proteins, the influence of receptor mutations, etc. We will refer to earlier seminal publications, but mostly focus on more recent evidence for adenosine receptor desensitization and internalization.

A1 receptor

Cellular and physiological studies

The early evidence for adenosine A1 receptor desensitization was largely obtained from primary cells, cell lines, tissues or tissue slices, and intact animals that were exposed to varying concentrations of adenosine receptor agonists (either A1-selective or not), often examined over several time periods.

Stiles and co-workers were among the first to study A1 receptor desensitization in detail (1991). Ramkumar et al. pretreated DDT1 MF-2 cells, a smooth muscle cell line expressing both A1 and A2A receptors, with R-phenylisopropyl adenosine (R-PIA) for up to 24 h, after which the adenylyl cyclase activity was reduced by approximately 50%. This was associated with a significant decrease in cell membrane-bound A1 receptors and a concomitant increase of receptor number in intracellular compartments. The authors also showed an increase in receptor phosphorylation, nicely paralleling the time course of adenylyl cyclase modulation [35]. In a later study, Nie et al. reported similar findings in the same cell line, although desensitization occurred at a somewhat faster pace (4 h) [36]. Interestingly, Palmer et al. were unable to demonstrate desensitization in CHO cells expressing the human A1R [37]. Klaasse et al., however, were able to show internalization of the human A1R tagged with a C-terminal yellow fluorescent protein (hA1YFP-R), stably expressed in CHO cells. Exposure of these cells for 16 h to 400 nM or 4 μM CPA resulted in 25 and 40% receptor internalization, respectively. Addition of 10 μM of the allosteric enhancer PD81,723 did not accelerate the internalization process, but lowered the threshold concentration at which internalization occurred. Under those conditions a small degree of internalization was observed already at a concentration of 40 nM CPA, and at 400 nM CPA, 59% of the receptors internalized [38].

Exposure of cerebellar granule cells, endogenously expressing A1R, to 100 nM R-PIA for 2–48 h led to a blunting of the inhibition of adenylyl cyclase. Along with this observation, a decrease of [3H]cyclohexyladenosine ([3H]CHA) binding to intact cerebellar granule cells and an increase of [3H]CHA binding in microsomes was detected. Simultaneously, a decrease in the steady-state level of \({\text{G}}_{{{\text{i}}\alpha }} \) in plasma membrane and microsomes was observed. These findings point not only to homologous desensitization, but also to subsequent internalization of the A1R in the microsomal fraction of cerebellar granule cells upon long-term agonist exposure. However, no change in mRNA level was observed, suggesting that post-transcriptional regulation underlies receptor desensitization [39].

In hippocampal neurons, A1Rs are present on both presynaptic and postsynaptic terminals, as well as on the cell body and dendrites where they exert different actions. To address the question of whether desensitization was influenced by subcellular localization, neurons were exposed to the agonist 2-chloroadenosine (CADO, 20 μM) from 2 up to 96 h. It was found that upon agonist exposure in cultured hippocampal neurons, the presynaptic A1Rs desensitize less quickly (>12 h) than the postsynaptic A1Rs (2 h). In accordance, the recovery of desensitized presynaptic A1Rs also requires more time (48 h vs 8 h for postsynaptic A1R). Desensitization of the postsynaptic A1Rs apparently occurs at the level of the receptor, because the other elements of the signal transduction machinery appeared to be fully functional upon receptor desensitization. All in all, these results suggest that the extent and the kinetics of agonist-induced desensitization of A1Rs depend on the subcellular localization of the receptors [40].

Using a more intact preparation, Coelho et al. found that the density of A1R in rat hippocampal slices was decreased by 30% upon 60 min of imposed hypoxia. This desensitization could be mimicked by adding the A1R agonist CADO (10 μM) for 60 min and was prevented by adding the A1R antagonist DPCPX. These results suggest that hypoxia leads to an increase in extracellular adenosine levels, and a subsequent, quite rapid (<90 min) desensitization, possibly followed by subsequent internalization of the A1R in nerve terminals [41].

Rat in vivo studies also demonstrated A1R desensitization. In a study by Parsons and Stiles rats were chronically (6 days) infused with R-PIA. Examination of adipocyte membranes of both treated and control rats revealed a 40% lower inhibition of adenylyl cyclase in the treated animals, coinciding with a reduced number of A1Rs (68% agonist-occupied receptors remaining) as determined by radioligand binding [42]. Upon further study it was noticed that the effects on adenylyl cyclase were not A1R specific but were controlled at the level of adipocyte G proteins. Gi levels appeared downregulated, whereas the amount of Gs in the preparation was increased, although not at the level of their mRNA, suggesting a heterologous form of desensitization [43].

Chronic exposure of rats to R-PIA (6 days) also led to A1R desensitization in the brain, and subsequent reduced inhibition of adenylyl cyclase. This loss of response was accompanied by a significant decrease in both total numbers of A1R and \({\text{G}}_{{{\text{i}}\alpha }} \) proteins in synaptic plasma membranes in the brain, paralleling the finding by Stiles and co-workers in adipocytes [44]. As a consequence, a significant increase of A1R was observed in microsomes and coated vesicles, which suggested a role for coated vesicles in the internalization of A1R. Similarly, chronic agonist exposure of rats to NECA (6 days) resulted in a significant decrease of A1R in the high-affinity state in the rat brain, however without changes in adenylyl cyclase activity or inhibition of the \({\text{G}}_{{{\text{i}}\alpha }} \) proteins [45].

Molecular mechanisms

As discussed in the general introduction to this review, receptor posttranslational modifications, receptor phosphorylation, recruitment of arrestins and the formation of clathrin-coated pits are elements of the molecular machinery of desensitization and internalization. In addition, other potential protein partners in the two processes have been studied for the various adenosine receptor subtypes, which will also be discussed.

Effect of receptor posttranslational modifications on desensitization and/or internalization

Gao et al. studied the effects of preventing palmitoylation of the A1R. It appeared that the Cys309Ala mutation, thus removing the palmitoylation site of the human A1R, had no effect on internalization [46]. These findings were later confirmed by Ferguson et al. [47].

GRKs and arrestins

A1R phosphorylation by GRKs has been subject to debate. Palmer et al. and Ferguson et al. failed to demonstrate GRK-2 phosphorylation of A1R; in the latter study it was found that nonphosphorylated A1R redistributes arrestin 3 from the cytoplasm into punctate clusters at the plasma membrane [37, 47]. Nie et al., however, reported that within 1 h of exposure of DDT1MF-2 cells to R-PIA, rapid translocation of GRKs was observed from the cytosol to the cytoplasm [36]. In a further biochemical approach with both purified receptor and GRK-2, a phosphorylated receptor was obtained that showed enhanced affinity for arrestins over G proteins. The same receptor kinase, upon overexpression in FRTL-5 cells, influenced A1R signalling, however, not via \({\text{G}}_{{{\text{i}}\alpha }} \)-mediated adenylyl cyclase but through \({\text{G}}_{{{\text{i}}\beta \gamma }} \)-mediated MAP kinase activation [48]. It may be concluded that under physiological conditions A1R phosphorylation does not (readily) take place, which would be a rationale for the long time periods required for receptor internalization.

Other protein partners

The ectoenzyme adenosine deaminase (ADA) regulates the extracellular adenosine concentration by converting excess adenosine into inosine. However, ADA also plays a role in the desensitization and internalization of A1R in smooth muscle DDT1MF-2 cells [49, 50] and LLC-PK1 epithelial cells [51], acting as a receptor activity-modifying protein (RAMP). Upon agonist exposure (100 nM R-PIA, 2 h), ADA and A1R formed complexes on the cell surface, clustered and internalized together to intracellular compartments. Such clustering of adenosine A1 receptors prior to internalization was also reported by Ciruela et al. and Saura et al. [49, 52]. The intracellular vesicles contained the lipid raft marker protein caveolin. Filipin, an agent that disrupts rafts or caveolae, inhibited A1R internalization. In contrast, acidic treatment disrupting clathrin-coated vesicles did not inhibit agonist-induced internalization of A1R. These results indicate that ADA and A1R form a stable complex in the cell membrane of LLC-PK1 cells, internalizing upon agonist exposure via lipid rafts, in a clathrin-independent pathway. Furthermore, a direct interaction of the C terminus of A1R with caveolin was demonstrated [50]. These and other data suggest that the mode of receptor compartmentalization in response to agonist stimulation may be governed by both receptor subtype and cell type [51].

Another accessory protein, the heat shock cognate protein 73 (hsc73), a member of the hsp70 family, was identified as a cytosolic component able to interact with the third intracellular loop of the A1R. The interaction between hsc73 (30 nM) and A1R led to a marked reduction in affinity of ligands for the A1R and also prevented activation of the G proteins, even more so than the addition of GTP analogues or GTP itself (100 μM). These effects were completely prevented by the addition of 25 nM ADA. A high percentage of A1R was coupled to hsc73 in cell lysate, according to immunoprecipitation experiments. A remarkable feature upon internalization of the receptor in DDT1MF-2 cells was found; A1Rs internalized via two different vesicle types, one in which A1R and hsc73 are colocalized and another in which hsc73 was absent. These observations open the possibility that signalling via GPCRs is regulated at least to some extent by heat shock proteins [53].

Receptor-receptor interactions

Dunwiddie et al. found that activation of A3R with a selective A3R agonist resulted in subsequent heterologous desensitization of the A1R, as observed in electrophysiological experiments in the CA1 region of rat hippocampus. Similar results were obtained after A3R occupancy via a brief superfusion with a high concentration of adenosine [54].

Lopes et al. investigated how activated A2AR influenced A1R function and whether this interaction was modified in aged rats. In hippocampal and cortical nerve terminals, the A2AR agonist CGS 21680 (30 nM) was able to lower the binding affinity of the A1R-selective agonist CPA, and this was taken as proof for A1R desensitization. The effect was counteracted by the addition of the A2AR antagonist ZM 241385 (20 nM). This reduction in A1R function could only be detected in young adult rats (6 weeks), but not in old rats (24 months). The addition of a PKC inhibitor, chelerythrine (6 μM), also prevented the desensitization of A1R [55]. Similarly, Ciruela et al. [56] demonstrated that A1R-A2AR heteromers were able to modulate the glutamatergic neurotransmission in the rat striatum. The main biochemical characteristic of this heteromer is the ability of the agonist-occupied A2AR to reduce the agonist affinity of the A1R [56].

By binding to cannabinoid CB1 receptors in Purkinje fibers in the cerebellum, Δ9-tetrahydrocannabinol inhibits adenylyl cyclase and consequently motor coordination. Long-term Δ9-tetrahydrocannabinol treatment resulted in CB1R downregulation, desensitization of the \({\text{G}}_{{{\text{i}}\alpha }} \) protein and desensitization of adenylyl cyclase. G protein activation by A1R, however, was unaffected. Surprisingly, heterologous attenuation of A1R-mediated inhibition of adenylyl cyclase was observed. These results indicate that long-term Δ9-tetrahydrocannabinol administration produces a disruption of inhibitory receptor control of cerebellar adenylyl cyclase and suggest a potential mechanism of cross-tolerance to the motor effects of cannabinoid and A1 agonists [57].

D1R and A1R have been shown to form functionally interacting heteromeric complexes in engineered cell lines and in cortical neurons in culture. Pretreatment with an A1R agonist caused complex formation of both receptors. Combined pretreatment with selective agonists for both receptors (but not one agonist alone) substantially reduced cAMP accumulation induced via the D1R, indicative for this receptor’s desensitization [58].

A2A receptor

Cellular and physiological studies

DDT1 MF-2 cells, expressing both A1R and A2AR, were exposed to an A2A-selective agonist, which resulted in a rapid loss (t1/2 = 45 min) of agonist-stimulated cAMP production in these cells. This receptor desensitization, however, did not involve a reduction in cell membrane receptor number or a change in ligand affinity [35].

Prolonged exposure of PC12 cells, expressing both A2AR and A2BR, to AR agonists led to a fast (30 min) and significant inhibition of A2AR stimulation by the A2AR-selective agonist CGS21680. This effect appeared to occur at the level of adenylyl cyclase, since no change was observed in receptor number or in CGS21680’s affinity for the receptor. This conclusion was corroborated by the finding of reduced activation of adenylyl cyclase by forskolin. Longer agonist exposure (12–20 h) led to a reduction of \({\text{G}}_{{{\text{s}}\alpha }} \) levels, whereas no changes occurred in the short-term protocol [59].

NG108-15 neuroblastoma x glioma hybrid cells express both A2AR and A2BR [60, 61]. Treatment of the cells with the non-selective agonist NECA followed by its washout led to a rapid (t1/2 = 20 min) and pronounced reduction in cAMP production by the A2AR-selective agonist CGS21680 when compared to vehicle-treated cells.

Palmer et al. [62] generated CHO cells solely expressing the recombinant (canine) A2AR and examined the desensitization process of this receptor. Cells exposed to NECA showed a rapid desensitization of A2AR-stimulated adenylyl cyclase activity with no obvious difference between pretreatment of 30 min or 24 h. This was associated with a slightly reduced affinity of the receptor for the A2AR-selective radiolabelled agonist [3H]CGS21680. Cell surface receptor numbers only diminished significantly (up to 40%) upon longer-term pretreatment (t1/2 = 8 h) [62].

Using a tissue preparation, Conti et al. investigated whether prolonged exposure of A2AR to the non-selective agonist NECA, or to the selective A2AR agonists CGS21680 and 2HE-NECA, influenced A2AR desensitization. The authors used the porcine coronary artery as a sensitive vascular model, expressing among others A2ARs. The arteries were first precontracted by adding 3 μM PGFα. NECA, CGS21680 and 2HE-NECA showed high affinities for the A2AR (EC50’s of 72, 40 and 20 nM, respectively) inducing vasorelaxation. Next, coronary arteries were pretreated with 10 μM NECA for 30 min or 2 h. After a 2-h washout period, the functional response was assessed. It appeared that preincubation with NECA did not hamper the vasorelaxing effects of CGS21680 and 2HE-NECA. However, NECA response curves were shifted to the right after NECA pretreatment [63]. These results seem inconclusive, since NECA pretreatment might ‘hit’ other adenosine receptor subtypes.

A2ARs, next to A1R, are present on presynaptic baroreceptor afferent terminals within the nucleus tractus solitarius (NTS) in the brain [64]. It was observed that these A2AR modulate 5-HT release as a mechanism of baroreflex control mediated by the NTS. Low concentrations of the non-selective agonist NECA (0.3–3 nM), briefly exposed (5 min) to NTS brain slices, induced the release of 5-HT caused by the activation of the A2AR present at presynaptic nerves. This effect could be blocked by the addition of the adenosine A2AR antagonist 8-(3-chlorostyryl)caffeine (CSC; 100 nM). Longer exposure of NECA (20 min) to the NTS slices resulted in inhibition of the 5-HT release, probably caused by quick desensitization of the A2AR (15 min) and subsequent involvement of the A1R [65]. These findings were corroborated in a similar study with CGS21680, an A2AR-selective agonist [66].

The effects on AR fate upon chronic agonist exposure were studied in vivo in rat brain [45]. After 6 days of NECA treatment, the adenylyl cyclase activity in synaptic plasma membranes was decreased, suggesting a desensitization of A2Rs, although the authors did not specify or study which A2R subtype was involved. Gs protein levels were decreased indicating Gs downregulation as the mechanism of desensitization.

Interestingly, Rekik and Mustafa [67] showed that chronic antagonist treatment (3 days) of porcine coronary arteries with ZM241,385 led to a decreased agonist responsiveness. Although A2A receptor expression went up, it appeared that the levels of Gs had decreased, altogether leading to a functional desensitization of the relaxing response by e.g. CGS21680 [67]. It should be kept in mind that ZM241,385 is also a potent antagonist for the adenosine A2B receptor.

Molecular mechanisms

Effect of receptor C terminus on desensitization and/or internalization

To investigate the importance of the (120 amino acid residues) long C terminus of the A2AR receptor in inducing desensitization and internalization, Palmer and Stiles introduced several mutations and deletions into the receptor tail. It appeared that deletion of the last 95 amino acids of the C terminus, containing 10 possible phosphorylation sites, did not have any effect on radioligand binding, adenylyl cyclase activity or desensitization kinetics compared to the wild-type A2AR. However, when two possible phosphorylation sites (Thr 298 and Ser 305) just upstream the 95 deleted amino acids were mutated to Ala, short-term (30 min) agonist-induced desensitization was attenuated, while the long-term (24 h) desensitization was not affected. Single mutations revealed that mutation of Thr 298 alone was sufficient to reduce receptor phosphorylation and agonist-induced short-term desensitization. This study also shows that short-term and long-term desensitization have distinct structural requirements and do not occur via the same mechanism [4].

GRKs and arrestins

Which GRK isoforms are involved in the phosphorylation of the A2AR is not entirely clear yet. However, a putative role for GRK2 and/or GRK5 has been suggested [4]. The role of GRK2 in agonist-induced phosphorylation and subsequent desensitization of A2AR and A2BR was thoroughly investigated by Kelly and co-workers. Wild-type GRK2 was stably expressed in NG108-15 cells, which endogenously express A2AR and A2BR. The acute stimulation of adenylyl cyclase by activation of A2AR was markedly reduced in NG108-15 cells overexpressing wild-type GRK2. This was probably caused by GRK2-dependent pre-desensitization of the A2AR by extracellular adenosine. This effect could be reversed by pretreating the cells 24 h with 0.5 unit/ml ADA [68]. The same research group investigated the effect of a dominant-negative GRK2 mutant on the desensitization of the A2AR [60]. Stable transfection of a GRK2Lys220Arg mutant in NG108-15 cells reduced desensitization of the A2AR by 50% following a 30-min treatment with the adenosine agonist NECA.

Tumour necrosis factor (TNF)-α treatment of human monocytoid THP-1 cells expressing the A2AR prevented desensitization of this receptor, occurring under control conditions upon pretreatment with CGS21680 or NECA [69]. It was discovered that the TNF-α treatment prevented GRK2 translocation to and decreased GRK2 association with the plasma membranes of these cells as a consequence of the activation of a sphingomyelinase-dependent pathway.

In another study, Mundell and Kelly investigated the effect of inhibitors of receptor internalization on desensitization and resensitization of ARs in NG108-15 cells [70]. Before agonist exposure, cells were pretreated with hypertonic sucrose or concanavalin A (ConA), both inhibitors of internalization. This pretreatment did not affect the agonist-induced desensitization of the A2AR. However, the resensitization of the A2AR upon agonist removal was abolished in the presence of ConA or sucrose.

Arrestins 2 and 3 have been implicated in the downstream desensitization process. CGS21680 stimulation of a tagged A2AR transiently transfected in HEK293 cells induced the translocation of GFP-tagged arrestins 2 and 3 towards the plasma membrane. A dominant-negative arrestin 2 mutant inhibited agonist-induced internalization [71].

Other protein partners

The long C terminus of the A2AR has been coined a ‘coincidence detector’ as it recognizes quite a number of other proteins such as α-actinin and ARNO, the Arf nucleotide site opener (for a review see Gsandtner and Freissmuth [72]). α-Actinin may play a role in receptor internalization, not unlikely for a protein involved in cytoskeletal arrangements [71], however ARNO does not [73].

Receptor-receptor interactions

The A2AR colocalizes with D2 dopamine receptors in the basal ganglia, and their interaction has been documented on several occasions [74]. For instance, the protein-protein interaction between A2AR-D2R was confirmed in HEK293T cells. The most likely mode of interaction is that helix 5 and/or helix 6 and the N-terminal portion of I3 in the D2R approach helix 4 and the C terminus of the A2AR [75]. Within the scope of this review fits the observation that activation of the D2 receptor actually sensitizes A2AR-mediated increases in cAMP production, in CHO cells expressing both receptors [76] as well as in CAD and NS20Y neuroblastoma cells in which the D2 receptor was expressed [77]. In SH-SY5Y neuroblastoma cells co-stimulation of A2AR and D2R accelerated D2R desensitization, probably by causing or enhancing D2R internalization [78]. Recently it was found that in these SH-SY5Y neuroblastoma cells the A2AR also formed heteromers with the endogenously expressed CB1R. The CB1R is negatively coupled to adenylyl cyclase and requires previous or simultaneous activation of A2AR to signal in these cells [79].

A2B receptor

Cellular and physiological studies

The adenosine A2B receptor (A2BR) is endogenously expressed on most artificial cell lines, such as COS and HEK293 cells. The exception is formed by CHO cells that lack this adenosine receptor subtype.

The A2BR, like the other three adenosine receptor subtypes, is subject to agonist-induced desensitization. This was measured on the level of cAMP production by Peters et al. [80]. Pretreatment of COS7 cells endogenously expressing A2B receptors with the non-selective agonist NECA (1 μM) for 1 up to 17 h led to a significant reduction in cAMP production upon acute agonist challenge, already after 1 h of pretreatment. Maximal reduction of cAMP production was reached after 4 h of pretreatment. CHO cells stably expressing a 5′ FLAG epitope-tagged A2B receptor showed the same result, albeit that maximal reduction of cAMP response was already achieved after 1 h of pretreatment. Also mouse 3T3-L1 cells and human HEK293 cells, endogenously expressing the A2BR, showed a decreased cAMP response after 2 h and 5.5 h of pretreatment with 1 μM NECA, respectively [80]. In NG108-15 cells, expressing both A2AR and A2BR, the rate of desensitization for both receptor subtypes appeared similar, with a half-life of 15–20 min. Pretreatment with NECA of these cells (0.1 mM, 30 min) reduced stimulation of adenylyl cyclase by A2BR by approximately 50% [60, 61].

Rat pheochromocytoma PC12 cells also endogenously express both A2AR and A2BR. Prolonged exposure (14 h) to 100 nM NECA or 1 μM CGS21680 significantly inhibited the cAMP response of the cells to subsequent stimulation with the A2A-selective agonist CGS21680. Although a 100-fold higher NECA concentration is needed to stimulate the A2BR compared to A2AR, Chern et al. found that the A2B receptor cAMP response was also desensitized upon desensitization of the A2A receptor. This observation can be explained by the fact that the \({\text{G}}_{{{\text{s}}\alpha }} \) protein level and the adenylyl cyclase activity were diminished upon long-term desensitization of the A2AR by CGS21680 [59].

The domain in which the receptor resides is also important for desensitization events, according to Sitaraman et al. NECA (10 μM) added to either the apical or basolateral side of T84 intestinal cells resulted in desensitization of the A2BR on the corresponding side within 2–3 h. However, whereas applying NECA to the apical side of the membrane had no effect on the basolateral A2BR, basolateral NECA induced a complete desensitization of the apical receptor. Since receptor trafficking may play a role in this cross-desensitization process, this effect may contribute to the desensitization and subsequent downregulation of the A2BR [81].

Trincavelli et al. observed a very rapid desensitization (t1/2 = 5 min) of A2BR in a human astrocytoma cell line. Both G protein coupling efficiency and cAMP production were diminished after pretreatment of the cells with NECA in the presence of an A2AR-selective antagonist [82].

Mundell et al. reported that acute exposure of human airway smooth muscle (ASM) cells to adenosine receptor agonists resulted in a rapid accumulation of cAMP, most probably via A2BR. Treatment with adenosine deaminase (ADA) suggested that ASM cells produce adenosine which feeds back on the cells’ A2BRs, thereby regulating basal cAMP levels and inducing a small degree of A2BR desensitization. Chronic treatment with adenosine agonists had a dual effect; both A2BR desensitization and adenylyl cyclase sensitization (an increased responsiveness of adenylyl cyclase upon stimulation) were observed [83].

Haynes et al. [84] studied A2BR desensitization in both pulmonary artery smooth muscle cells and a more integrated preparation, the isolated perfused lung. Pretreatment of the smooth muscle cells with NECA for 45 min abrogated the increase in cAMP response otherwise observed for both NECA and isoproterenol, suggesting heterologous desensitization. Indeed, experiments with cholera toxin showed that the desensitization took place at the level of the Gs protein-adenylyl cyclase complex. In the lung preparation NECA, when dosed acutely, caused a rapid vasodilation, which was not observed when the tissue was treated again with the same compound 45 min after the first dose [84].

Taken together, these results suggest that desensitization of the A2B receptor is a robust phenomenon, rather independent from the cell type or tissue in which it is expressed.

Molecular mechanisms

Effects of receptor mutations on desensitization and/or internalization

Matharu et al. studied the regions of the A2BR which are responsible for the desensitization and internalization of the A2BR [85]. For this purpose they introduced point mutations or deletions in the C terminus of the A2BR. Deleting the final two amino acids (Leu330-stop) did not affect the internalization properties of the A2BR. The Phe328-stop and Gln325-stop mutants, however, were resistant to agonist-induced desensitization and internalization. GFP-tagged arrestin 2 did not translocate from the cytosol to the plasma membrane upon agonist stimulation of these two truncated receptors. From a single point mutation in the C terminus, Ser329Gly, it became clear that this serine residue is responsible for the rapid agonist-induced desensitization and internalization. Surprisingly, a further deletion mutant Ser326-stop was able to undergo rapid agonist-induced desensitization and internalization; however, arrestin 2 was not attracted to the plasma membrane upon agonist stimulation. It appeared that internalization of this truncated receptor occurred via an arrestin- and clathrin-independent pathway, probably via caveolin-mediated internalization since the internalization of Ser326-stop was dynamin dependent. The destination of this truncated receptor was also different: a compartment close to the plasma membrane, instead of early endosomes. Whether the receptor will undergo rapid agonist-induced desensitization and internalization or influence the rate, extent and mechanism of internalization depends on specific sites in the C terminus of the A2BR and can be influenced by their mutation or deletion.

GRKs and arrestins

The A2BR, endogenously expressed in NG108-15 cells, is desensitized through phosphorylation by GRK2. This GRK2 effect was selective for A2AR and A2BR, since the other endogenously expressed receptors (secretin R and IP-prostanoid R) were not desensitized in this way [68]. In addition, stable transfection of a dominant negative GRK2(Lys220Arg) mutant in the NG108-15 cells reduced the NECA-induced (30 min, 100 μM) desensitization of the A2BR by 50% [60]. Peters et al. studied the role of phosphorylation by PKA and PKC rather than GRK2 in agonist-induced desensitization in COS-7 cells transiently expressing the 5′-FLAG A2BR. They concluded that these second messenger-dependent kinases are not involved in A2BR phosphorylation [80]. A similar conclusion for A2BR phosphorylation in NG108-15 cells was reached by Mundell and Kelly [61]. Trincavelli et al. found that treatment of astrocytoma cells with TNF-α increased the A2BR functional response, without changing receptor protein or mRNA levels. TNF-α treatment appeared to markedly reduce agonist-dependent receptor phosphorylation and also attenuated agonist-mediated A2BR desensitization. These effects could be inhibited by the addition of the A2BR antagonist MRS 1706 [82]. These findings resemble data obtained with the A2AR (vide supra).

Mundell et al. found that within the arrestin family, both arrestin 2 and arrestin 3 are involved in the internalization process of the A2B receptor. Within 1 min of agonist exposure, both GFP-tagged arrestin isoforms were translocated to the plasma membrane. Longer agonist exposure (>10 min), however, revealed that only arrestin 2 colocalizes with the receptor in the early endosomes. One explanation for this may be a higher affinity of the A2BR for arrestin 2 compared to arrestin 3. Interestingly, arrestins are not only involved in the internalization of the A2B receptor, but also in its recycling. It appeared that the expression of stable arrestin anti-sense constructs, reducing the levels of endogenous arrestins, not only resulted in less internalization of the receptors, but also impaired the recycling process significantly. In contrast to the internalization process, arrestin 3 induced a significantly faster rate of A2BR recycling compared to arrestin 2 [86].

Penn et al. also studied the behaviour of GFP-tagged arrestins, now in human airway smooth muscle cells. Arrestin 2 and arrestin 3 were able to reduce the cAMP production upon stimulation of endogenous A2BR by NECA and stimulation of β2-AR by isoproterenol. In addition, a punctate clustering was observed in the membrane upon exposure to either NECA (100 μM) or isoproterenol (1 μM), indicating that arrestins play a role in receptor trafficking. On the contrary, signalling and trafficking of the prostaglandin E2-R, another endogenously expressed receptor, was not affected by arrestin 2 or arrestin 3 [27].

A3 receptor

Adenosine A3 receptors (A3R) evoke considerable interest as novel drug targets to address cerebral/cardiac ischaemia, cancer, inflammation, asthma and chronic obstructive pulmonary disease. So far, potent and selective antagonists for the hA3R have been identified; the disadvantage however is that those antagonists show extremely low binding affinity for the rodent A3R, typically 1,000 times lower than in humans. Since rodent models are essential for the pharmacological evaluation of new therapeutic agents, this forms a serious drawback.

In addition, the adenosine A3 receptor (A3R) sequence holds a nuclear localization signal in its C-terminal tail [87]. This typical stretch of four amino acids (KKFK) in helix 8 may direct the receptor to the cell nucleus, not necessarily as a consequence of desensitization or internalization. This should be kept in mind when appreciating the studies described below.

Cellular and physiological studies

Ramkumar et al. studied the characteristics of rat A3R, endogenously expressed on RBL-2H3 mast cells. In one of their experiments the authors pretreated the cells with NECA and found a partial desensitization of the initial Ca2+ response to NECA [88]. Trincavelli et al. studied the relationship between agonist-induced desensitization, internalization and resensitization of hA3R in transfected CHO cells. Agonist-induced endocytosis of hA3R was investigated by immunogold electron microscopy of plasma membranes and intracellular vesicles and shown to occur with a half-life of 17 min. Subsequent removal of the agonist led to recycling of 90% of the receptor population to the cell surface, with a half-life of 35 min. Short-term exposure to agonist caused rapid desensitization, as assessed in cAMP assays. Removal of the agonist led to resensitization of the cAMP signal to 90% of the original signal within 120 min. Internalization did not affect signal transduction, as was demonstrated after blockade of internalization and recycling. Internalization occurred via clathrin-coated pits. These results show that hA3R undergoes agonist-induced endocytosis, which is not responsible for desensitization. Moreover, in the case of hA3R, receptor sequestration rather than desensitization seems to be the first step in a cycle of internalization, dephosphorylation and recycling to the plasma membrane [89]. A similar cell line was used by Palmer et al. In stably transfected CHO cells prolonged treatment with NECA (10 μM, 20 h) induced uncoupling of recombinant hA3R from Gi proteins and a functional desensitization. Upon this A3R desensitization, an approximately twofold increase in adenylyl cyclase activity was found in the presence or absence of forskolin. This sensitization of adenylyl cyclase activity was not caused by an altered level of inhibitory or stimulatory G protein expression. The occurrence of the sensitization was also not due to new protein synthesis, but probably to an increased coupling efficiency between Gs and adenylyl cyclase. Compared to control cells, long-term exposure of stably transfected CHO cells to NECA caused an increase in phosphorylation of the cAMP-responsive element binding protein upon addition of suboptimal concentrations of forskolin. The sensitization of adenylyl cyclase activity upon long-term treatment might provide a molecular basis for the observation that for several adenosine receptor-mediated events, acute agonist exposure produces opposite effects to those after chronic agonist treatment [90]. Ferguson et al. studied the fate of the rat A3R expressed in CHO cells. The receptors internalized rapidly after treatment with NECA or R-PIA over a time frame (t½ = 10 min) that followed receptor phosphorylation (t½ = 1 min) [91].

The desensitization, internalization and downregulation of hA3R was also investigated in human astrocytoma cells, natively expressing hA3R. Short-term (15 min) exposure of the cells to 100 nM Cl-IBMECA, an A3R-selective agonist, caused rapid receptor desensitization, subsequently followed by receptor internalization within 30 min. With the help of immunogold electron microscopy, the localization of the A3R was revealed. After 10 min of exposure, the A3R was found in smooth-surfaced pits and in uncoated vesicles in the cytoplasm. After 30 min of exposure, the A3R was found in vesicular endosomes. Upon removal of the agonist, resensitization of the A3R occurred within 120 min through receptor recycling to the cell surface. Long-term agonist exposure (1–24 h) resulted in a marked downregulation of A3R to 22 ± 3% of control after 24 h. In conclusion, multiple, temporally distinct and sequential processes are involved in the regulation of hA3R upon short- and long-term exposure to agonists [92].

A3Rs were found to be highly expressed on murine B16-F10 melanoma cells [93]. The authors examined the association between A3R trafficking and receptor functionality and tumour growth inhibition upon activation with the A3R-selective agonist IB-MECA. Exposure to 10 nM IB-MECA (5 min) led to rapid internalization of A3R to the cytosol, upon which receptors were directed to the endosomes for recycling or to lysosomes for degradation. The addition of 100 nM MRS 1523 (5 min), an A3R antagonist, was able to counteract the internalization process as well as the modulation of the Wnt pathway leading to proliferation, thereby emphasizing the involvement of A3R in this process. When the melanoma cells were injected into nude mice, tumours rapidly developed. Tumour growth was significantly inhibited after administration of IB-MECA to the animals, paralleled by a decrease in A3R expression in tumour lesions. In hypoxic human A172 and U87MG glioblastoma cell lines it was found that adenosine upregulates the expression of hypoxia-inducible factor-1α (HIF-1α) and vascular endothelial growth factor (VEGF) via the A3 receptor. HIF-1α is a key regulator in the development of tumours. The addition of the A3-antagonist MRE 3008F20 inhibited the adenosine-induced HIF-1α and VEGF accumulation in these hypoxic cells, thereby showing a putative role in inhibiting tumour growth [94].

To circumvent the species differences mentioned in the introduction to the A3R section, Yamano et al. generated A3R humanized (A3ARh/h) mice in which the A3R was replaced by its human counterpart. The expression level of hA3R in the humanized (A3ARh/h) mice was equal to the expression levels of A3R in wild-type mice. A3R agonists were able to elevate the intracellular Ca2+ concentration in bone marrow-derived mast cells from the humanized (A3ARh/h) mice. However, the rate of hA3R internalization was markedly reduced compared with that of mA3R in these mast cells [95].

Molecular mechanisms

Receptor phosphorylation, GRKs and arrestins

Palmer and Stiles as well as Ferguson et al. investigated which amino acid residues in the C terminus are responsible and crucial for the rapid desensitization of the A3R [96, 91]. A triple mutant (Thr307, Thr318 and Thr319 to Ala) exhibited dramatically reduced phosphorylation, desensitization and internalization of the rat A3AR. Individual mutation of each Thr residue showed that Thr318 and Thr319 were the most important sites for phosphorylation. In addition, phosphorylation of Thr318 was necessary to observe phosphorylation of Thr319, but not vice versa. Moreover, changing Thr318 for a negatively charged residue (Glu) was not sufficient to retain phosphorylation at Thr319. Mutating two predicted palmitoylation sites, Cys302,305 to Ala resulted in agonist-independent basal phosphorylation of the rat A3AR. Such findings strongly suggest that palmitoylation of these Cys residues is an important factor in controlling accessibility of the C terminus of the A3R in the process of recruiting GRKs. In fact, the palmitoylation sites are highly conserved between different species, e.g. rat, mouse, human, dog, sheep. Taking these results together, it appears that GRK-mediated phosphorylation of the A3R C terminus follows a sequential mechanism, with the receptor palmitoyl moieties in an important regulatory role, and phosphorylation of Thr318 being particularly crucial as an essential first step. In an earlier study Palmer et al. replaced the carboxyl terminus of the A1R by that of the A3R [37]. This chimaeric construct was able to undergo agonist-stimulated phosphorylation and functional desensitization, similar to A3R. It was also found that purified GRK2, GRK3 and GRK5 were all able to enhance agonist-dependent phosphorylation of A3R as well as the A1-A3 chimaera.

Trincavelli et al. studied the involvement of extracellularly regulated kinases (ERK1 and 2), members of the mitogen-activated protein kinase (MAPK) family, in A3R phosphorylation. It was found that within 5 min of exposure to 10 μM NECA, ERK1 and 2 were already phosphorylated in CHO cells stably expressing hA3R. An inhibitor of MAPK activation (PD98059) also caused inhibition of A3R phosphorylation, desensitization and internalization, probably by preventing the membrane translocation of GRK2. These results indicate that the MAPK cascade is involved in A3R regulation by a feedback mechanism which controls GRK2 activity and probably involves direct receptor phosphorylation [97].

The rat basophilic leukaemia cell 2H3 cell line (RBL-2H3 cells) endogenously expresses equal levels of arrestin 2 and arrestin 3. Both arrestin isoforms also have a high and comparable affinity for clathrin, thereby promoting agonist-induced internalization. RBL-2H3 cells also endogenously express a high level of A3R, which, however, neither recruited arrestin 3 nor arrestin 2 upon stimulation with NECA. Also no changes in A3R distribution were observed. One explanation is that A3R follows an endocytic mechanism upon agonist stimulation that does not involve arrestin-mediated clathrin-coated pit internalization. Another possible explanation is that arrestin recruitment was below the limit of detection in the RBL-2H3 cells [98]. Ferguson et al. observed that upon phosphorylation of A3R by GRK arrestin 3 is redistributed into punctate vesicles both at the plasma membrane and within the cytoplasm. Nevertheless, these authors also noticed that there was no colocalization between A3R and arrestin 3 [47].


The adenosine receptor subtypes are differentially regulated when exposed to agonist treatment. A1Rs are not (readily) phosphorylated and slowly internalize with a typical half-life of several hours. This feature may cause less than average tachyphylaxis upon chronic agonist administration, for instance in type II diabetes. A2AR and A2BR show much faster downregulation with similar kinetics, usually < 1 h for short-term desensitization. Agonists for the A2A receptor, currently profiled in wound healing, may thus suffer from declining efficacy upon chronic administration, in contrast to those for the A1 receptor. The A3R undergoes even faster downregulation, often a matter of minutes. The latter receptor also holds a nuclear localization signal in its carboxy terminal tail, possibly obscuring true agonist-induced downregulation. The fast desensitization of the A3R after agonist exposure may be therapeutically equivalent to antagonist occupancy of the receptor.


Lipid rafts are planar domains of 25–100 nm in cell membranes enriched in specific lipids and proteins. They are in particular characterised by a high cholesterol and glycosphingolipid content in the outer leaflet of the lipid bilayer that gives them a gel-like liquid-ordered organization in comparison with the surrounding phospholipid-rich disordered membrane [26].


Caveolae are flask-shaped invaginations located at or near the plasma membrane with a 50–100 nm diameter. They are considered to be a non-planar subfamily of lipid rafts. The shape and structural organization of caveolae are due to the presence of caveolin−1, −2 and −3, which self-assemble in high-mass oligomers to form a cytoplasmic coat on the membrane invaginations [25, 26].


Clathrin-coated pit is a specialized region of the cell surface that mediates the internalization of extracellular macromolecules and GPCRs. Coated pits derive their name from the presence of a distinctive polygonal lattice that decorates the inner surface of the membrane. This lattice is composed of multiple triskelion-shaped subunits that contain three clathrin heavy chains (180 kD) and three clathrin light chains (30–40 kD) plus a family of associated proteins with molecular sizes of 50 kD and 110 kD [29, 30].


Copyright information

© Springer Science+Business Media B.V. 2007

Authors and Affiliations

  • Elisabeth C. Klaasse
    • 1
  • Adriaan P. IJzerman
    • 1
  • Willem J. de Grip
    • 1
    • 2
  • Margot W. Beukers
    • 1
  1. 1.Department of Medicinal Chemistry, Leiden/Amsterdam Center for Drug ResearchLeiden UniversityLeidenThe Netherlands
  2. 2.Department of Biophysical Organic Chemistry, Leiden Institute of Chemistry, Leiden/Amsterdam Center for Drug ResearchLeiden UniversityLeidenThe Netherlands