Activation of both N-methyl-D-aspartate (NMDA) and non-NMDA receptors releases endogenous adenosine from superfused rat cortical slices. NMDA-evoked adenosine release is Ca(2+)-dependent and results from the extracellular degradation of a released nucleotide, whereas non-NMDA receptor activation releases adenosine per se in a Ca(2+)-independent manner. IBMX selectively inhibits NMDA- but not non-NMDA-evoked adenosine release. Forskolin, but not 1,9-dideoxy-forskolin, produced a slight but significant increase in NMDA-evoked adenosine release, suggesting that the formation of cyclic AMP may somehow be involved. The inhibition of NMDA-evoked adenosine release by IBMX is not accompanied by enhanced cyclic AMP recovery in superfusates, nor is release diminished when cyclic AMP transport is inhibited by probenecid, suggesting that the adenosine is not derived from the extracellular metabolism of released cyclic AMP. It is possible that 5'AMP, derived from the intracellular conversion of cyclic AMP by phosphodiesterase, might be released during NMDA receptor activation. However, more selective inhibitors of the specific phosphodiesterase isozymes known to be located in the cortex failed to diminish NMDA-evoked adenosine release. Therefore, the effects of both forskolin and IBMX on NMDA-evoked adenosine release could be nonspecific, coincidental and unrelated to their actions on cyclic AMP levels in the cortex. However, it is also possible that a novel IBMX-sensitive phosphodiesterase plays a primary role in converting cyclic AMP to 5'AMP intracellularly during NMDA receptor activation; the 5'AMP could then exit the cells and be converted to adenosine extracellularly.