Adenosine Signalling and Function in Glial Cells

Abstract

Despite major advances in a variety of neuroscientific research fields, the majority of neurodegenerative and neurological diseases are poorly controlled by currently available drugs, which are largely based on a neurocentric drug design. Research from the past five years has established a central role of glia to determine how neurons function and – consequently – glial dysfunction is implicated in almost every neurodegenerative and neurological disease. Glial cells are key regulators of the brain’s endogenous neuroprotectant and anticonvulsant adenosine. This review will summarize how glial cells contribute to adenosine homeostasis and how glial adenosine receptors affect glial function. We will then move on to discuss how glial cells interact with neurons and the vasculature and outline new methods to study glial function. We will discuss how glial control of adenosine-function affects neuronal cell death and its implications for epilepsy, traumatic brain injury, ischemia, and Parkinson’s disease. Eventually, glial adenosine-modulating drug targets might be an attractive alternative for the treatment of neurodegenerative diseases. There are, however, several major open questions that remain to be tackled.

1. Introduction

It is becoming increasingly clear that inflammatory processes play an important role in neurodegenerative disease, just as inflammation is becoming increasingly implicated in various systemic diseases elsewhere in the body¹. Because the immune responses in brain show uncommon features it was long considered to be “immune privileged”. However, cells that are involved in adaptive immune reactions do enter the brain, and this can result in major CNS pathology. Multiple sclerosis is an important example. Furthermore, immune reactions, often attributable to the innate immune system, feature in many if not all neurodegenerative diseases. Microglial cells, astrocytes, endothelial cells, oligodendrocytes, and neurons all produce signals to orchestrate these reactions. Arachidonic acid metabolites, nitric oxide, cytokines, and chemokines all appear to play some role. Recently, the role of ATP and adenosine as critically important signaling molecules has become appreciated^2,3. This awareness that a multiplicity of signals other than those principally involved in nerve-nerve communication play a role in neurodegenerative disease has also meant that we have to consider other cells than neurons as critically important players. The present brief review will focus on adenosine (and to a lesser extent ATP) signaling, and the role of glial cells in neurodegenerative disease. The review will feature results from the authors’ own work more prominently than would be motivated from an objective standpoint.

2. Glial control of adenosine homeostasis

In mammals, purine de novo synthesis proceeds via formation of IMP, which is then converted into AMP; however, there is no de novo synthesis pathway for adenosine. Physiologically, intracellular adenosine can be formed by either dephosphorylation of AMP by 5′-nucleotidase, or, alternatively, by hydrolysis of S-adenosylhomocysteine, whereas extracellular adenosine can be formed from released adenine nucleotides by a cascade of ectonucleotidases⁴. Two metabolic pathways are responsible for the removal of adenosine: deamination into inosine via adenosine deaminase (ADA; EC 3.5.4.4) and phosphorylation into AMP via adenosine kinase (ADK; EC 2.7.1.20). Based on its low K_M for adenosine, ADK is considered to be the primary route of adenosine metabolism⁵. Recent findings indicate that extracellular levels of adenosine, and consequently the levels close to synapses, are largely regulated by astrocytes^6–10, and an astrocyte-based adenosine-cycle has been proposed^11,12.

2.1. Glial release of ATP as source for extracellular adenosine

ATP can be released from neurons and astrocytes, is identified as a neurotransmitter in both CNS and PNS, and exerts a multitude of largely excitatory effects by activation of specific ATP receptors (P2X and P2Y receptors)¹³. Vesicular release has also been clearly demonstrated from endocrine cells and here the release of ATP may differ in several ways from that of the hormone, because of so called kiss-and-run release¹⁴. Thus, ATP can be released even in situations when the vesicle fusion is too incomplete and transient to allow release of the stored hormone. Perhaps this can occur also in nerves. However, under physiological conditions the vesicular release of ATP, not from neurons, but from astrocytes, has been identified as a major source of synaptic adenosine¹⁵. Transgenic mice that express a dominant-negative SNARE domain selectively in astrocytes were characterized by the loss of the adenosine A₁ receptor mediated tonic inhibition in synaptic slices, indicating that under physiological conditions astrocytic release of ATP (followed by degradation into adenosine via ectonucleotidases) is a major source of adenosine¹⁵ that affects synaptic transmission. A kiss and run like exocytotic release of ATP has been detected in astrocytes¹⁶. Another, possibly related, mechanism involves release of ATP from a lysosome pool in astrocytes¹⁷. However, astrocytes appear to use also other mechanisms to release ATP as a signal. One of those involve connexin hemichannels¹⁸, but recent work provides strong reason to assume that pannexin-1, rather than connexin-43, provides the most responsible channel¹⁹. Another proposal is that ATP is released via maxi-anion channels, and this may be particularly important in astrocytic swelling²⁰. The P2X₇ receptor may in some conformations allow ATP to be released. Finally, ATP may be released whenever there are increases in membrane volume via incorporation of intracellular vesicles into the membrane, or when there is a shedding of small vesicles²¹. In addition to the mechanisms described above, astrocytes can directly release adenosine, especially in response to hypoxic stimulation^6,22, even though release of adenosine per se is more typical of neurons²³. In that case the release depends on export of adenosine via the equilibrative nucleoside transporters.

Since astrocytes can contact thousands of synapses and coordinate synaptic networks^8,24, it is conceivable that astrocytic release of ATP and its subsequent degradation into adenosine has a major regulatory function in setting a global adenosine-mediated inhibitory tone within a neuronal network. In addition, other glial cells can also contribute. For example, in retinal tissue stimulation of glial cells leads to an adenosine-mediated inhibition of neuronal activity, but in this case Muller cells rather than astrocytes appear to be most important²⁵.

The distribution of adenosine formed from breakdown of released ATP will obviously also depend on the distribution of the enzymes that degrade the nucleotide. It was shown long ago that 5′ nucleotidase tended to accumulate in areas of a lesion²⁶ and we now know that CD73, the ecto- 5′ nucleotidase, is highly expressed in microglial cells, as is the ecto-NTPDase, CD39²⁷.

2.2. Elimination of adenosine via astrocytic adenosine kinase

Several lines of evidence indicate that astrocytic ADK is the key regulator for ambient levels of adenosine: (i) In adult brain, ADK is predominantly expressed in astrocytes¹⁰. (ii) Pharmacological inhibition of ADK is sufficient to prevent seizures in various models of epilepsy²⁸. (iii) Genetic knockout or knockdown of ADK in cultured cells induces the secretion of adenosine into the medium^29–32. (iv) Transgenic overexpression of ADK triggers seizures by reduction of ambient adenosine³³. (v) Inhibition of ADK in hippocampal slices increases endogenous adenosine and depresses neuronal firing, whereas inhibition of ADA had no effect³⁴. (vi) A substrate cycle between AMP and adenosine, which involves ADK and 5′-nucleotidase, enables minor changes in ADK activity to rapidly translate into major changes in adenosine³⁵. (vii) ADK activity is regulated in response to brain injury and is subject to developmental regulation^10,36,37. Based on these considerations, and based on the lack of a classical transporter-regulated-reuptake system for adenosine and the ubiquitous presence of bi-directional equilibrative nucleoside transporters³⁸, ADK likely fulfills the role of a metabolic reuptake system for adenosine. Thus, tight regulation of ADK expression levels and of its specific activity becomes a necessity. Therefore, it is not surprising that ADK is highly conserved in evolution, that no naturally occurring mutations of the Adk-gene are known, and that a genetic disruption of the Adk-gene is lethal^5,39.

The fact that astrocytic ADK is of critical importance in regulating extracellular adenosine concentrations implies that the nucleoside transporters that facilitate adenosine uptake into astrocytes are important. Astrocytes express one concentrating (sodium-dependent) and two equilibrative nucleoside transporters⁴⁰. Inhibition of this transport could potentially be used to elevate brain adenosine levels under conditions when extracellular adenosine is derived from extracellular ATP, but they would be less useful under conditions when adenosine is derived from intracellular production⁴¹. It is interesting to note that cannabinoids can block the equilibrative transporter and this effect can partly explain the immunosuppressive effects of these compounds⁴². The central role of astrocytes in regulating extracellular levels of adenosine is demonstrated in Figure 1.

Figure 1.

Extracellular adenosine levels are thought to be regulated by an astrocyte-based adenosine-cycle. Astrocytes can release ATP via vesicular release and/or by direct release through hemichannels (h-ch). Extracellular ATP is rapidly degraded into adenosine (ADO) by a series of ectonucleotidases. Adenosine can also be released directly via equilibrative nucleoside transporters (nt). Intracellularly adenosine levels are largely controlled by adenosine kinase, which is part of a substrate cycle between adenosine and AMP. Small changes in adenosine kinase activity rapidly translate into major changes in adenosine. Intracellular adenosine kinase is considered to be a metabolic reuptake system for adenosine. Only selected mechanisms and pathways are shown; for details please refer to main text.

3. Glial adenosine receptors

There are four types of evolutionarily conserved and pharmacologically well-characterized adenosine receptors called A₁, A_2A, A_2B and A₃⁴³ (Fig. 2). Adenosine is the endogenous agonist at all these receptors, but at A₁ and A₃ receptors inosine can act as a partial agonist^44,45. The A₁ and A₃ receptors couple to the G_i family of G proteins and thus stimulate K⁺ channels, reduce transient voltage dependent Ca²⁺ channels and inhibit cAMP formation; A_2A receptors couple to members of the G_s family (G_olf in striatal neurons), whereas A_2B receptors couple to many G proteins including G_s, G_q and G₁₂. Adenosine is approximately equipotent on A₁, A_2A and A₃ receptors, whereas A_2B receptors require higher agonist concentrations, if cAMP changes is the readout⁴⁵, but if MAP kinase activation is used to measure receptor activation adenosine is virtually equipotent on all four receptors⁴⁶. All four adenosine receptors are detected in astrocytes²², and all have been reported to be expressed in microglial cells or microglial cell lines^47–49.

Figure 2.

Adenosine receptors, their coupling to G-proteins and some of the down-stream consequences of receptor activation.

3.1. A₁ receptors

A₁ receptors on astrocytes reduce their proliferation rate in culture⁵⁰. As in many other types of cells activation of A₁ receptors can not only decrease cAMP accumulation but also stimulate phospholipase C, especially if this pathway is simultaneously activated by other stimuli^51–53. A₁ receptors help protect astrocytes from damage and cell death^22,54,55, partly via activation of PI3K and Erk 1/2 phosphorylation. Nerve activity promotes myelination and it has been shown that this response is dependent on ATP⁵⁶, but ATP acts on oligodendrocytes indirectly, because ATP acts on astrocytes to release leukemia inhibitory factor. This response may be modified by adenosine, as A₁ receptors are present in oligodendroglia⁵⁷ and stimulate their migration. It has been shown that A₁ receptor activation leads to white matter loss, and that A₁ receptors contribute to hypoxia-induced white matter loss⁵⁸. A₁ receptors on microglial cells are reported to reduce excessive activation of microglial cells upon immune activation⁵⁹. Activation of these microglial receptors may secondarily affect oligodendroglial cells⁵⁹ and also astrocyte proliferation⁶⁰, which emphasizes the possibility of an extended glial network of signaling. A₁ receptors on neurons (especially at nerve terminals) are critically important in mediating the dampening effect on neuronal activity mediated by adenosine generated from ATP released from astrocytes^6,61. The highly abundant A₁ receptors at nerve endings may preferentially signal via G_o proteins to inhibit transient calcium channels, whereas the same receptors in nerve cell bodies and dendrites may preferentially regulate potassium channel conductance via G₁ proteins.

3.2. A_2A receptors

In brain, A_2ARs are expressed at high levels in striatal neurons and at low levels in neurons outside of the striatum and in glial cells^62,63. Many functional measurements (such as cAMP levels and cytokine release) coupled with pharmacological tools have clearly demonstrated the presence and function of A_2ARs in glial cells⁶⁴. A_2AR binding densities are at the range ~30–60 fmole/mg protein in primary cultured microglial cells or in sorted microglial cells derived from striatum, as estimated by ³H-ligand binding studies^65,66. Furthermore, the expression of the A_2AR in glial elements in both the striatum and the solitary tract is confirmed by electronmicroscopic studies^63,67. It should be noted that A_2AR expression in microglia and astrocytes is usually low under physiological conditions and frequently below the detection limit of histological methods (i.e. immunohistochemistry, autoradiography, or in situ hybridization)^62,64,68.

Importantly, the expression of A_2ARs in glial cells is induced following brain insults. For example, LPS treatment induced A_2AR mRNA and protein in primary cultures of (mixed) glial cells (mainly in microglial cells) at 16 hours and peaked at 48 hours after the treatment⁶⁵. Recently, using double immunohistochemistry analysis, we demonstrated that A_2AR expression is induced in microglial cells and astrocytes of mouse substantia nigra at 24 hours after 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) intoxication. The induction of A_2ARs in glial cells by brain insults and inflammatory signals, coupled with a local increase in adenosine and pro-inflammatory cytokines (such as IL-1β, which further induces A_2AR expression), may serve as part of an important feed-forward mechanism to locally control neuroinflammatory responses in the brain. It has been shown that adenosine, acting on A_2A receptors, can increase extracellular levels of glutamate, both by reducing glutamate uptake via GLT-1 and by direct release^69,70. Thus, some of the reported effects of A_2A receptors on glutamate release may be based on mechanisms mediated by astrocytes rather than by neurons.

A_2ARs in glial cells may exert complex actions on neuronal cell death (both, potentially deleterious as well as neuroprotective) and possibly other functions such as modulation of synaptic transmission. In astrocytes, activation of A_2ARs by extracellular adenosine increases astrocyte proliferation and activation^71,72, but inhibits the expression of iNOS and the production of NO⁷³, and regulates glutamate efflux by astrocytes⁶⁹. Thus, modulation of astrogliosis by A_2ARs is likely involved in brain repair processes, possibly via the formation of tissue scar. In microglial cells, activation of A_2ARs has mixed effects on microglial proliferation, but has clear facilitating effects on the release of cytokines including up-regulation of cyclo-oxygenase 2 and the release of prostaglandin E2 (PGE2)⁷⁴, and on increases in NOS activity and NO release⁶⁵ and nerve growth factor expression⁷⁵.

3.3. A_2B receptors

It was shown early on that cyclic AMP accumulation in brain slices was due to a different adenosine receptor than that responsible for adenylate cyclase stimulation in striatum, and it is now clear that a major part of the brain slice cAMP response is due to A_2B receptor activation on astrocytes⁷⁶. A_2B receptors can couple to two different classes of G proteins, G_q and G_s. G_q regulates intracellular calcium and vesicular release, whereas G_s affects a plethora of cAMP dependent signaling pathways. The A_2B receptor may also activate phospholipase C⁷⁷ and appears to be responsible for the adenosine-induced stimulation of Il-6 from astrocytes⁷⁸. In airways, A_2B receptors, via cAMP, regulate chloride channels⁷⁹. Since this occurs also in the intestine, it is fair to predict that cells in the CNS likewise alter chloride flux via A_2BR signaling. In particular, the possibility exists that adenosine is an important regulator of astrocytic swelling via modulation of volume-regulated anion channels. A_2B receptors may also play an impiortant role in the development of the nervous system; it was intially proposed that the important regulator netrin required signaling via A_2B receptors⁸⁰, but later studies showed that netrin does not require simulataneous A_2BR signaling⁸¹, but that netrin signaling requires prior A_2B activation because the receptor is thereby regulated⁸².

3.4. A₃ receptors

Astrocytic A₃ receptors appear to regulate chemokine release⁸³. Activation of A₃ receptors by endogenous adenosine protects astrocytes from cell death induced, e.g., by hypoxia²². Microglial cells have functional A₃ receptors coupled to MSAP-kinase and p38 signaling^47,84. Although adenosine itself has little effect on microglial migration, it was shown that the migration induced by ATP⁸⁵, perhaps predominantly via P2Y12 receptors⁸⁶ is lost when ATP hydrolysis via CD39 is eliminated and can be restored by adenosine²⁷. Although it is not absolutely certain that the relevant adenosine receptor is the A₃R, it is tempting to speculate that there is a similarity to the situation in neutrophil leucocytes where ATP acting on a P2Y receptor acts in concert with adenosine acting on A₃ receptors to stimulate migration⁸⁷. Microglial cells (and A₃ receptors) may also be of particular importance in the regulation of chemokine release and chemokine actions⁸⁸.

4. Glial cells and neurovascular coupling

Glial cells play important roles in coupling neuronal function to the cerebral microvasculature that controls cerebral blood flow (CBF)^89,90 in the sense that increased neuronal activity requires corresponding increases in CBF. Apart from large processes that stain for intermediate filaments and give astrocytes their stellar appearance, astrocytes have a multitude of fine processes that have little overlap with processes from other astrocytes and that define individual astrocytic domains, which each contain 300–600 neuronal dendrites and 10⁵ synapses in rodent hippocampus^8,91–93. Thus, a single astrocyte can sense the activity, and integrate the function, of hundreds of neurons within its domain. In addition, each astrocyte extends at least one process with endfeet surrounding blood vessels of the microvasculature. Therefore, astrocytes are uniquely located to adjust regional CBF to regional energy metabolism.

The vasodilator adenosine has been identified as an important mediator that couples cerebral blood flow to neuronal activation⁹⁴. Thus, adenosine was demonstrated to mediate glutamate-induced vasodilation in the cerebral cortex^95,96. Topical application of glutamate dilated pial arterioles, an effect that could be reversed by an A_2AR antagonist, but not by an A_2AR antagonist⁹⁵. Likewise topical superfusion AMPA on the cortical surface through a closed cranial window resulted in increases in pial arteriolar diameter, an effect that could be reversed by A_2A and A_2BR blockade, but not by inhibition of NO synthase, cyclooxygenase-2, or cytochrome P-450 epoxygenase⁹⁶. Apart from the activation of vascular adenosine receptors, adenosine can exert important regulatory functions by activation of astrocytic adenosine receptors⁹⁷. Thus, the adenosine-evoked calcium response in acutely isolated astrocytes was found to be coupled to the A_2B receptor⁷⁷; based on these findings adenosine could be implicated in promoting the propagation of calcium-increases throughout astrocytic processes. Increased Ca²⁺ in turn is also associated with the release of ATP through connexin hemichannels, a process that is potentiated by A_2BR activation^98,99. Through these mechanisms ATP-release and degradation into adenosine via ectonucleotidases appears to mediate arteriolar dilation in response to neuronal activation¹⁰⁰. This process was dependent on astrocytes, since the application of the selective gliotoxin L-AAA, led to complete loss of arteriolar dilation in response to neuronal activation¹⁰¹.

5. Glial control of glutamate and excitotoxicity

Astrocytes play a fundamental role in the pathogenesis of ischemic neuronal death¹⁰². A large body of evidence indicates that astrocytes are involved in the control of glutamate homeostasis and the susceptibility of the brain to excitotoxic injury¹⁰³. Glutamate transporters are expressed in many different types of brain cells, but astrocytes are primarily responsible for glutamate uptake. Studies using genetic deletion or antisense-oligonucleotide mediated knockdown of the astroglial glutamate transporter GLT-1 have demonstrated that this transporter is the predominant subtype responsible for the clearance of extracellular glutamate in the brain^104,105. Affected animals were highly susceptible to glutamate-dependent excitotoxicity and developed epileptic seizures^104,105. After uptake of glutamate into astrocytes the enzyme glutamine synthetase converts glutamate into glutamine, which is then transported into neurons, where it is converted back into glutamate. Interestingly, a loss of glutamine synthetase was found in the sclerotic hippocampus of human patients with temporal lobe epilepsy¹⁰⁶ and the authors of that study concluded that reduced activity of the glutamate-glutamine cycle led to an accumulation of extracellular glutamate.

Apart from the mechanisms described above, astrocytes themselves can be a significant source of extracellular glutamate, which can be released by a variety of mechanisms¹⁰⁷. It has been demonstrated that Ca²⁺ elevations in astrocytes induce the excitotoxic release of glutamate from these cells¹⁰⁸. Most importantly, the Ca²⁺-dependent astrocytic release of glutamate was also dependent on the vesicular glutamate transporters (VGLUT1/2) and the vesicular SNARE protein, cellubrevin, and was consistent with a vesicular release mechanism of glutamate that was similar to synaptic release of glutamate¹⁰⁸. Finally, it was shown that astrocyte-derived glutamate targets synaptic NMDA receptors¹⁰⁹, providing a rational explanation for the astrocyte-based control of neurotoxicity. Given the emerging roles of astrocytes in the control of neuronal excitotoxicity neuroprotective efforts targeting the functional integrity of astrocytes may constitute a superior strategy for future neuroprotection.

6. Novel methods for studying the role of glia

Co-culture of glia and neurons

Many original insights into glio-transmission were discovered using co-culture systems, and then extended to brain slices or the intact brain. For example, the first evidence for gliotransmission came from studies in mixed cultures of astrocytes and neurons demonstrating that experimentally evoked Ca²⁺ elevation in astrocytes evoked the elevation of Ca²⁺ in adjacent neurons^110,111. In this method, cultured cells from postnatal day 1–4 rodent brain are first enriched with one population of either glial or neuronal origin using specific culture conditions and then confirmed by immunohistochemistry. Glial or neuronal cells are then plated separately or together onto a coated substrate. The distinct morphology of glial cells and neurons allows the identification of these distinct cell types and permits the direct application of electric field potentials, micropipette tips, or neurochemical substrates to astrocytes in order to evoke a specific elevation of calcium in glial cells (e.g. in astrocytes)^110,111. The spatiotemporal control over mechanical stimuli has permitted the selective stimulation of single astrocytes in mixed cultures of rat forebrain astrocytes and neurons. Using this strategy it was demonstrated that the elevation of calcium, triggered by focal electric field potentials in single astrocytes, induced a wave of calcium increase that was propagated from astrocyte to astrocyte, and importantly, this wave of calcium increase also evoked large increases in the concentration of cytosolic calcium in neurons depending on those astrocytes^110,111. More recently, mixed co-cultures in multi-compartment dishes were equipped with the capability to provide electronic stimuli selectively to neurons in mixed cultures of oligodendrocytes and dorsal root ganglion neurons. This approach led to the finding that ATP and adenosine, released from neurons, acts as a potent neuron-glial transmitter that inhibited oligodendrocyte progenitor cell proliferation, stimulated their differentiation, and promoted the formation of myelin^56,112.

Transgenic overexpression and targeted knockout of genes in defined glial populations

Glial cells have been shown to release gliotransmitters, including ATP^9,17, glutamate^111,113, and D-serine¹¹⁴ to coordinate synaptic networks. However, neurons and glia share these same chemical signaling molecules, making it difficult to define the role of gliotransmitters. To molecularly dissect the role of glial signaling molecules, genetic approaches have been developed to selectively manipulate the SNARE-dependent release of gliotransmitters using a glia-specific promoter. Pascual and Haydon developed a transgenic mouse line, which uses the tet-off system to allow conditional expression of the cytosolic portion of the SNARE domain of synptobrevin 2 [dominant-negative SNARE (dn-SNARE)] selectively in astrocytes¹⁵. The selective expression of dn-SNARE was achieved by the use of an astrocyte-specific glial fibrillary acid protein promoter. To confirm the cell type selectivity of astrocytic transgene expression, Pascual et al. used EGFP as a reporter system and showed that EGFP was visually detectable in 97% of the dnSNARE-transgene-expressing cultured astrocytes, and that EGFP-positive cells colocalized specifically with the astrocytic marker, but not with neuronal, NG2-glial, or oligodendroglial markers. This transgenic line has successfully been used to demonstrate the functional significance of gliotransmission on synaptic plasticity in hippocampus¹⁵ and more recently in the sleep-wake cycle⁶¹. Similarly, transgenic over-expression of a mutant disease-causing gene in a defined glial cell population has been used to study the role of SOD1 in astrocytes in the development of motor neuron death. Nagai et al demonstrated that the astrocyte selective expression of mutant human SOD1 (but not in spinal motor neurons, microglia or fibroblasts) killed spinal primary and embryonic mouse stem cell-derived motor neurons¹¹⁵.

As a complementary genetic approach, the selective deletion of signaling molecules in defined glial cell populations can be accomplished using the Cre-loxP strategy. For example, Boillee et al generated a transgenic line (LoxSOD1^G37R) that carried a mutant human SOD1 gene flanked at both ends by a 34-base pair LoxP sequence¹¹⁶. These “floxed” mice were then cross-bred to two transgenic lines with expression of the Cre protein under control of (i) the promoter from the Islet-1 transcription factor, which directs the expression exclusively in progenitors of motor and dorsal root ganglion neurons, and (ii) the CD11b promoter, which directs the expression exclusively in the myeloid lineage (including macrophages and microglial cells). The establishment of these two novel transgenic lines allowed to demonstrate that the SOD1 mutation in motor neurons and microglial cells contributes distinctly to the onset and progression of amyotrophic lateral sclerosis: while expression of SOD1 in motor neurons is the primary signal for the initiation of motor neurodegeneration and an early sign of disease progression, the genetic inactivation of SOD in mciroglial cells had little effect on the early disease phase, but markedly attenuated disease progression¹¹⁶. Thus, these genetic approaches to selectively manipulate signaling molecules in defined glial (or neuronal) populations provide critical insights into the distinct role of glial cells in the development of neurodegeneration.

Flow cytometric analysis and fluorescence activated cell sorting (FACS)

The structural complexicity of brain tissues hampers the dissection of unique roles of glial cells under various physiological and pathological conditions. Glial cells are characterized by a unique morphology, which permits the distinction of neuronal versus glial cells in intact brain by immunohistochemistry. However, it is difficult to quantify immunohistochemical changes without performing labor-intensive stereological analysis. Furthermore, there is a critical need to isolate large numbers of pure glial cells from intact brain for detailed molecular analysis such as qPCR and microarray analysis. Flow cytometric and fluorescence activated cell sorting can be adapted to partially circumvent these limitations for the study of glial cell functions in brain.

In the first application, defined glial cell populations (such astrocytes, microglia, or oligodendrocytes) are identified by labeling with a fluorescent antibody directed against specific cell surface markers. Quantitative changes of glial populations in normal and injured brains are determined by flow cytometry. For example, we recently utilized this analysis to evaluate the change of CD11b+ (a cell surface marker for microglial cells) and GFAP+ (a maker for astrocytes) cells in mouse striatum after treatment with MPTP and the A_2AR antagonist KW6002¹¹⁷. This analysis not only provided an improved quantitative assessment of the effect of the A_2AR antagonist on microglial activation at the very early phase of MPTP intoxication, but also identified a specific microglial cell population (i.e. CD11b+ cells with a large cell size representing fully activated microglial cells), which are most sensitive to KW6002 treatment in the brain¹¹⁷.

In the second application, fluorescence activated cell sorting permits isolation and purification of distinct populations of glial cells from neurons from brain tissues using fluorescent antibodies directed against cell surface markers. The sorted glial cell populations can then be used for detailed molecular analyses such as quantitative PCR and microarray analysis. For example, Lovatt et al. successfully performed microarray profiling of sorted astrocytes from mouse cortex using FACS and (surprisingly) demonstrated that most enzymes in the tricarboxylic acid cycle are expressed at higher relative levels in astrocytes than in neurons¹¹⁸.

7. Adenosine signaling in glial cells, excitotoxicity and cell death

As outlined above, synaptic levels of adenosine are largely controlled by an astrocyte-based adenosine cycle and the activity of the astrocyte-based enzyme ADK. Consequently, adenosine signaling in glial cells effects excitotoxicity and cell death in a variety of experimental paradigms. In addition, several forms of brain insult activate microglial cells. Here, ATP release is critically important⁸⁵, but the ATP response in microglial cells is markedly enhanced by adenosine generated from ATP²⁷.

7.1. Epilepsy

The adenosine kinase hypothesis of epileptogenesis implies that dysregulation of ADK is a major contributing factor to the epileptogenic cascade¹² (Fig. 3). Consequently, ADK expression levels (that determine levels of ambient adenosine) determine the brain’s susceptibility to acute seizure-induced cell death. Mice with only moderate transgenic overexpression of ADK in brain (141% of normal) were highly susceptible to acute seizure-induced cell death and did not survive beyond 3 days following status epilepticus³⁷, whereas engineered mice with reduced levels of ADK in forebrain (62% of normal) were completely resistant to seizure-induced cell death³⁷. Resistance to seizure-induced excitotoxic cell death in ADK-deficient mice was dependent on adenosine and increased adenosine A₁R activation, since blockade of A₁Rs with its selective antagonist DPCPX restored wild-type like seizure-induced excitotoxic cell death³⁷.

Figure 3.

Role of the adenosine (ADO) / adenosine kinase (ADK) system in regulating acute and chronic responses to injury. Left: Within hours after brain injury (e.g. stroke, trauma, prolonged seizures) a surge in micromolar levels of ADO results that protects the brain from further injury and from seizures. Hypoxia and trauma can directly lead to a rise in extracellular ATP that is rapidly degraded into adenosine. High levels of adenosine are known to inhibit ADK, further amplifying the adenosine surge. Right: The acute adenosine surge contributes to trigger astrogliosis via a variety of mechanisms that include modulation of astrocytic adenosine receptors (ARs), modulation of inflammatory processes and the release of cytokines. Astrogliosis leads to overexpression of ADK resulting in adenosine-deficiency, which contributes to seizure generation.

Astrogliosis is a pathological hallmark of the epileptic brain and contributes to seizure generation by a variety of mechanisms^12,119,120. A recent study from our lab has identified the enzyme ADK in astrocytes as a molecular link between astrogliosis and neuronal dysfunction in epilepsy³⁷. In a mouse model of CA3-selective epileptogenesis we found spatio-temporal co-localization of astrogliosis, upregulated ADK, and focal spontaneous electrographic seizures that were all restricted to the CA3-region, the site of the epileptogenesis precipitating acute injury; importantly, seizures could be suppressed pharmacologically by ADK inhibition³⁷. In this model, the seizures remained highly localized and restricted to the astrogliotic scar, presumably due to normal adenosinergic control of the surrounding brain tissue. Transgenic overexpression of ADK, as well as genetic disruption of the A₁R were sufficient to trigger spontaneous seizures, indicating that adenosine dysfunction rather than astrogliosis per se was responsible for seizure generation¹²¹. Conversely, mice with a genetically induced reduction of ADK in forebrain were completely resistant to the development of spontaneous seizures³⁷. In vitro studies performed on hippocampal slices have subsequently demonstrated that reduction of the basal tone of adenosine by ADK is permissive to seizure generation, whereas ADK did not limit activity-dependant adenosine-release¹²². Together, these findings provide a neurochemical rationale for adenosine augmentation therapies (AATs). Consequently, several focal AAT-approaches – based on intracerebral adenosine-releasing implants – have demonstrated robust anticonvulsive and possibly antiepileptogenic efficacy in a variety of experimental paradigms that have been reviewed elsewhere^123,124.

7.2. Traumatic brain injury

Traumatic brain injury triggers an acute surge in adenosine, presumably as a consequence of ATP release, and this may represent an endogenous neuroprotective mechanism. In one study, adenosine levels increased 61-fold following controlled cortical impact (CCI) in rats and peaked at 20 min following the impact¹²⁵. The existence of an endogenous protective action of adenosine at A₁ receptors early after experimental TBI was further corroborated by the finding of lethal status epilepticus in A₁R knockout mice subjected to either controlled cortical impact¹²⁶ or to kainic acid induced hippocampal injury¹²⁷. In contrast to A₁R knockout mice, A_2AR knockout mice were largely protected from the adverse effects of CCI¹²⁸. In line with these findings, increases in cerebrospinal fluid caffeine concentration were associated with favorable outcome after severe traumatic brain injury in humans, likely due to caffeine-mediated inhibition of A_2ARs¹²⁹. Likewise, chronic, but not acute caffeine attenuated the consequences of TBI in the mouse CCI model¹³⁰. Although it probably does not contribute to the surge in adenosine following brain injury, acute downregulation of ADK in astrocytes has been described as a consequence of stroke³⁶ or acute seizures¹³¹, and this may prolong the adenosine increase.

7.3. Parkinson’s Disease

The adenosine A_2A receptor is a leading non-dopaminergic therapeutic target in Parkinson’s disease (PD) (Fig. 4). Interest in this receptor within the context of PD derives primarily from two lines of experimental and clinical investigations: First, decade-long preclinical studies demonstrated a unique co-localization of A_2ARs and dopamine D₂Rs in striatopallidal neurons. Antagonistic interactions between A_2ARs and D₂Rs at the molecular, neurochemical and behavioral level explain the motor stimulant effects of A_2AR blockade ^132–134. Thus, A_2AR antagonists such as KW-6002 (istradefylline) and SCH420814 have now completed clinical phase IIB-III trials. Despite some limitations of these clinical trials and admittedly modest effects (“OFF” time reduced by one hour), these studies confirm that selective A_2AR antagonists can stimulate motor activity by potentiating the L-dopa effect in advanced PD patients^135,136. Second, in addition to symptomatic relief, A_2AR antagonists appear to more directly attenuate dopaminergic neurodegeneration, as suggested by convergent epidemiological and experimental evidence. Following an initial report from the Honolulu Heart Program¹³⁷ by Ross and colleagues, several large-cohort prospective studies have confirmed a similar inverse relationship between the consumption of caffeinated coffee and the risk of developing PD. Including the Health Professionals' Follow-Up Study and the Nurses’ Health Study these studies involved a total of 47,351 men and 88,565 women¹³⁸, whereas a more recent study conducted by the Finnish Mobile Clinic Health Examination Survey included 19518 men and women¹³⁹. These studies firmly established a relationship between increased caffeine consumption and decreased risk of developing PD in males. In addition, studies with animal models of PD provide a compelling clue about the potentially protective effects of caffeine, by demonstrating that pharmacological blockade (by caffeine or selective A_2AR antagonists) or genetic depletion of the A_2AR attenuates dopaminergic neurotoxicity and neurodegeneration^140–142. These studies provide a neurobiological basis for the inverse relationship between increased caffeine consumption and reduced risk of developing PD.

Figure 4.

The dual functions of A_2A receptor antagonists in Parkinson’s disease models: A_2AR antagonists act at the A_2AR in striatal neurons to stimulate motor activity. Furthermore, it is postulated that A_2AR antagonists may modulate microglial activation in substantia nigra to exert a possible neuroprotective effect in an animal model of Parkinson’s disease.

Despite consistent demonstration that A_2AR antagonists afford neuroprotection against MPTP or 6-hydroxydopamine -induced dopaminergic neurotoxicity, the mechanism by which A_2AR inactivation protects against the loss of dopaminergic neurons remains unknown. The particular challenge lies in explaining the apparent dichotomy between restricted expression of the A_2AR in striatopallidal neurons and neuroprotection against degeneration of dopaminergic neurons in the substantia nigra where only a scattered expression of A_2ARs is detected. An additional challenge is to identify the cellular mechanism, which allows A_2AR inactivation to protect neurons against a broad spectrum of brain insults, from ischemia to excitotoxicity to mitochondrial toxicity¹⁴³. In this context, the involvement of the glial A_2AR becomes an attractive possibility, since glial function and neuroinflammation is commonly associated with diverse pathological conditions as mentioned above. Indeed, we recently demonstrated that MPTP treatment markedly upregulates A_2AR expression in microglial cells at 24–48 hours after treatment^65,66, which may result in further amplification of the A_2AR-mediated modulation of neuroinflammation in PD models. Consistent with this notion, an immunohistochemical study showed that KW-6002 reduced the loss of striatal dopamine contents and nigral cell bodies, and this coincided with inhibition of microglial activation¹⁴⁴. Furthermore, we demonstrated by flow cytometery that KW-6002 attenuated MPTP-induced microglial activation at 48 hours after MPTP treatment¹¹⁷. In conclusion, A_2AR antagonists may confer neuroprotection by acting at A_2ARs in glial cells, at least in the MPTP model of PD.

7.4. Ischemia

During ischemia, an imbalance between ATP degradation and resynthesis brings about a rapid and marked increase in extracellular level of adenosine in the brain during ischemia ^145–154. In addition hypoxia will increase ATP release, resulting in further adenosine production. While no clinical reports with purinergic compounds in human stroke exist, it is widely believed that adenosine and its receptors function as an endogenous neuroprotectant under these conditions^155–161. Indeed, adenosine¹⁶² or adenosine-potentiating agents (such as inhibitors of ADA or ADK^163–167 or of adenosine transport^{151,166,168–173}) offer protection against ischemic neuronal damage in different in vivo ischemia models. Furthermore, transgenic overexpression of ADK aggravates cell death, while reduction of ADK in the hippocampus increases protection after transient focal ischaemia^36,174.

Such a protective effect is attributed to stimulation of adenosine A₁ receptors that exert a protective role in ischemia by presynaptic reduction of Ca²⁺ influx, by inhibition of the release of excitatory neurotransmitters^175,176, and by postsynaptic hyper-polarization and reduction of neuronal activity through increases in K⁺ and Cl⁻ ion conductances¹⁷⁷. The efficacy of A₁ receptors stimulation on neuroprotection depends on the model used and no protective effect was observed in a global ischemia model¹⁷⁸. Since adenosine does influence glutamate release it is suggested that this is not critically important in some ischemic models. However, adenosine (probably acting at the A_2A receptor) may in fact contribute to neurotoxicity, neuronal damage, and cell death. The potential neuroprotection by A_2AR antagonists was first reported in a global ischemia model with the less selective antagonist CGS 15943^179,180. Further studies substantiated this finding in different models of ischemia with the selective A_2A receptor antagonist 8-(3-chlorostyryl)caffeine (CSC) and SCH 58261^181,147,182 in various animal models of stroke^183,184. Studies in genetically manipulated mice confirmed the neuroprotective role of A_2A receptor antagonists on ischemic brain damage¹⁸⁵. Major protective effects of A_2A receptor antagonists in stroke have been attributed to reduced glutamate outflow^183,186,187. It should be considered however, that in several studies, A_2A receptor agonists have been found protective in the global ischaemia model in the gerbil¹⁸⁰ and that A_2A knockout mice show aggravated hypoxic ischaemic injury in neonatal mice¹⁸⁸. Possible mechanisms are not clear yet, but include A_2AR-mediated protection via inhibition of platelet aggregation, vasodilation^163,189, or anti-inflammatory actions. Lastly, activation of A₃Rs produced mixed results and the exact contribution of A₃Rs to ischemic brain injury is not clear¹⁹⁰.

So far, support for a role for glia in the neuroprotective effect of A_2AR antagonism in ischemia comes from the observation that the A_2A receptor antagonist SCH 58261 reduces p38 MAPK activation in microglial cells¹⁸⁴ and phospho-JNK in neurons and oligodendrocytes¹⁹¹ in the ischaemic hemisphere 24 hours after permanent MCAO. Since p38 MAPK and JNK are activated up to 24 hours after ischemia^192,193 and are involved in neuronal death^194,195, this correlation indicates that A_2AR antagonists may confer neuroprotection against ischemic brain injury through modulation of glial function and neuroinflammation. However, reduced MAPK activation might be secondary to a reduction in the excitotoxic cascade that primes p38 and JNK activation, since reduction glutamate outflow in the ischemic brain by A_2AR blockade is believed to be one of the main underlying mechanisms¹⁹⁶. Further studies with selective manipulation of glial adenosine receptors or glial function in vivo are critical to our understanding to what extend adenosine regulation of glial signalling and function is responsible for ischemic brain injury.

8. Conclusions and major open questions

The discussion above has shown that much is now known about a role of glial cells in mediating the effects of adenosine (and other purines) in different neurodegenerative states. It is clear that many of the actions of endogenous or exogenous adenosine (and ATP) are in fact due to actions on glial cells. This makes it much more complicated to understand precisely how adenosine acts, and it has become apparent that the often diverse actions reported are due to the fact that several different receptors, located on many cell types are involved. In order to get a better understanding, there are several major questions that require an answer. Among the questions that need to be addressed in future research we find:

• Is the acute adenosine surge that follows brain injury a trigger for subsequent astrocyte activation?

• Do all the proposed mechanisms of ATP release from glial cells in fact occur, and are the triggers for astrocytic ATP release via these mechanisms different?

• To what extent does regional variation in the expression of ecto-nucleotidases control the distribution of adenosine in brain?

• Can adenosine-signals propagate within the brain via astrocyte-astrocyte communication?

• Given the major increase in glial cells in humans, how far can we extrapolate from rodents to man?

• When several adenosine receptors, with at least partly opposing signaling, appear to regulate a single biological response, are the receptors located on different cells or on very different parts of the same cell?

• Given that two GPCR molecules can form dimers, but apparently only one of them can actually signal, is there any functional significance of heterodimers between two types of purine receptors?

• Can activation of endothelial cells at the vascular interphase signal to synapses?

• Are microglial cells directly involved in synaptic transmission or do they mainly help prune synaptic contacts?

• Why are effects of A2ARs on inflammation different in CNS and peripheral organs?

Abbreviations

AAT

adenosine augmentation therapy

ADA

adenosine deaminase

ADK

adenosine kinase

AMPA

α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate

AR

adenosine receptor

CBF

cerebral blood flow

CCI

controlled cortical impact

DPCPX

8-cyclopentyl-1,3-dipropylxanthine

EGFP

enhanced green fluorescent protein

DR

dopamine receptor

FACS

fluorescence activated cell sorting

MAP

microtubule associated protein

NMDA

N-methyl-D-aspartic acid

MPTP

1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine

NOS

nitric oxide synthase

NTPDase

nucleoside triphosphate diphosphohydrolase

PD

Parkinson’s disease

SNARE

soluble NSF attachment protein

SOD

superoxide dismutase