Adenosine acts as a neurotransmitter in the brain through the activation of four specific G-protein-coupled receptors (the A1, A2A, A2B, and A3 receptors). The A1 receptor has long been known to mediate neuroprotection, mostly by blockade of Ca2+ influx, which results in inhibition of glutamate release and reduction of its excitatory effects at a postsynaptic level. However, the development of selective A1 receptor agonists as antiischemic agents has been hampered by their major cardiovascular side effects. More recently, apparently deleterious effects have been reported following the activation of other adenosine receptor subtypes, namely, the A2A and the A3 receptors. In particular, selective A2A receptor antagonists have been demonstrated to markedly reduce cell death associated with brain ischemia in the rat, suggesting that the cerebral A2A receptor may indeed contribute to the development of ischemic damage. The beneficial effects evoked by A2A antagonists may be due to blockade of presynaptic A2A receptors (which are stimulatory on glutamate release) and/or to inhibition of A2A receptor-mediated activation of microglial cells. Even more puzzling data have been reported for the A3 receptor subtype, which can indeed mediate both cell protection and cell death, simply depending upon the degree of receptor activation and/or specific pathophysiological conditions. In particular, a mild subthreshold activation of this receptor has been associated with a reinforcement of the cytoskeleton and reduction of spontaneous apoptosis, which may play a role in "ischemic preconditioning" of the brain, according to which a short ischemic period may protect the brain from a subsequent, sustained ischemic insult that would be lethal. In contrast, a robust and prolonged activation of the A3 receptor has been shown to trigger cell death by either necrosis or apoptosis. Such apparently opposing actions may be reconciled by hypothesizing that adenosine-mediated cell killing during ischemia may be aimed at isolating the most damaged areas to favor those parts of the brain that still retain a chance for functional recovery. In fact, both A3 receptor-mediated cell death and A2A receptor-mediated actions may be viewed as an attempt to selectively kill irreversibly damaged cells in the "core" ischemic area, in order to save space and energy for the surrounding live cells in the "pneumbra" area. Hence, the pharmacological modulation of the A2A and A3 receptors via selective ligands may represent a novel strategy in the therapeutic approach to pathologies characterized by acute or chronic neurodegenerative events.