We examined how the endogenous anticonvulsant adenosine might influence gamma-aminobutyric acid type A (GABA(A)) receptor stability and which adenosine receptors (ARs) were involved. Upon repetitive activation (GABA 500 microM), GABA(A) receptors, microtransplanted into Xenopus oocytes from neurosurgically resected epileptic human nervous tissues, exhibited an obvious GABA(A)-current (I(GABA)) run-down, which was consistently and significantly reduced by treatment with the nonselective adenosine receptor antagonist CGS15943 (100 nM) or with adenosine deaminase (ADA) (1 units/ml), that inactivates adenosine. It was also found that selective antagonists of A2B (MRS1706, 10 nM) or A3 (MRS1334, 30 nM) receptors reduced I(GABA) run-down, whereas treatment with the specific A1 receptor antagonist DPCPX (10 nM) was ineffective. The selective A2A receptor antagonist SCH58261 (10 nM) reduced or potentiated I(GABA) run-down in approximately 40% and approximately 20% of tested oocytes, respectively. The ADA-resistant, AR agonist 2-chloroadenosine (2-CA) (10 microM) potentiated I(GABA) run-down but only in approximately 20% of tested oocytes. CGS15943 administration again decreased I(GABA) run-down in patch-clamped neurons from either human or rat neocortex slices. I(GABA) run-down in pyramidal neurons was equivalent in A1 receptor-deficient and wt neurons but much larger in neurons from A2A receptor-deficient mice, indicating that, in mouse cortex, GABA(A)-receptor stability is tonically influenced by A2A but not by A1 receptors. I(GABA) run-down from wt mice was not affected by 2-CA, suggesting maximal ARs activity by endogenous adenosine. Our findings strongly suggest that cortical A2-A3 receptors alter the stability of GABA(A) receptors, which could offer therapeutic opportunities.