A limitation of current gene therapy efforts aimed at central nervous system disorders concerns distribution of vectors on direct injection into neural tissue. Here we have circumvented this problem by transferring genes to the meninges surrounding the spinal cord, achieving an in vivo gene transfer paradigm for treating chronic pain. The therapeutic vector consisted of a recombinant adenovirus encoding a secreted form of the potent endogenous opioid beta-endorphin. In an inflammation model of persistent pain, administration of the vector into the cerebrospinal fluid (CSF) surrounding the spinal cord transduced meningeal pia mater cells. The resulting increase in beta-endorphin secretion attenuated inflammatory hyperalgesia, yet had no effect on basal nociceptive responses. This demonstration of a gene transfer approach to pain treatment can be generalized to neurodegenerative disorcers in which broad spatial distribution of therapeutic effect is critical.