Multiple sclerosis, an inflammatory, demyelinating disease of the CNS currently lacks an effective therapy. We show here that CNS inflammation and clinical disease in experimental autoimmune encephalomyelitis, an experimental model of multiple sclerosis, could be prevented completely by a replication-defective adenovirus vector expressing the anti-inflammatory cytokine IL-10 (replication-deficient adenovirus expressing human IL-10), but only upon inoculation into the CNS where local infection and high IL-10 levels were achieved. High circulating levels of IL-10 produced by i. v. infection with replication-deficient adenovirus expressing human IL-10 was ineffective, although the immunological pathways for disease are initiated in the periphery in this disease model. In addition to this protective activity, intracranial injection of replication-deficient adenovirus expressing human IL-10 to mice with active disease blocked progression and accelerated disease remission. In a relapsing-remitting disease model, IL-10 gene transfer during remission prevented subsequent relapses. These data help explain the varying outcomes previously reported for systemic delivery of IL-10 in experimental autoimmune encephalomyelitis and show that, for optimum therapeutic activity, IL-10 must either access the CNS from the peripheral circulation or be delivered directly to it by strategies including the gene transfer described here.