ATP serves not only as an energy source for all cell types but as an 'extracellular messenger' for autocrine and paracrine signalling. It is released from the cell via several different purinergic signal efflux pathways. ATP and its Mg(2+) and/or H(+) salts exist in anionic forms at physiological pH and may exit cells via some anion channel if the pore physically permits this. In this review we survey experimental data providing evidence for and against the release of ATP through anion channels. CFTR has long been considered a probable pathway for ATP release in airway epithelium and other types of cells expressing this protein, although non-CFTR ATP currents have also been observed. Volume-sensitive outwardly rectifying (VSOR) chloride channels are found in virtually all cell types and can physically accommodate or even permeate ATP(4-) in certain experimental conditions. However, pharmacological studies are controversial and argue against the actual involvement of the VSOR channel in significant release of ATP. A large-conductance anion channel whose open probability exhibits a bell-shaped voltage dependence is also ubiquitously expressed and represents a putative pathway for ATP release. This channel, called a maxi-anion channel, has a wide nanoscopic pore suitable for nucleotide transport and possesses an ATP-binding site in the middle of the pore lumen to facilitate the passage of the nucleotide. The maxi-anion channel conducts ATP and displays a pharmacological profile similar to that of ATP release in response to osmotic, ischemic, hypoxic and salt stresses. The relation of some other channels and transporters to the regulated release of ATP is also discussed.