Herpes simplex virus (HSV) has several potential advantages as a vector for delivering genes to the nervous system. The virus naturally infects and remains latent in neurons and has evolved the ability of highly efficient retrograde transport from the site of infection at the periphery to the site of latency in the spinal ganglia. HSV is a large virus, potentially allowing the insertion of multiple or very large transgenes. Furthermore, HSV does not integrate into the host chromosome, removing any potential for insertional activation or inactivation of cellular genes. However, the development of HSV vectors for the central nervous system that exploit these properties has been problematical. This has mainly been due to either vector toxicity or an inability to maintain transgene expression. Here we report the development of highly disabled versions of HSV-1 deleted for ICP27, ICP4, and ICP34.5/open reading frame P and with an inactivating mutation in VP16. These viruses express only minimal levels of any of the immediate-early genes in noncomplementing cells. Transgene expression is maintained for extended periods with promoter systems containing elements from the HSV latency-associated transcript promoter (J. A. Palmer et al., J. Virol. 74:5604-5618, 2000). Unlike less-disabled viruses, these vectors allow highly effective gene delivery both to neurons in culture and to the central nervous system in vivo. Gene delivery in vivo is further enhanced by the retrograde transport capabilities of HSV. Here the vector is efficiently transported from the site of inoculation to connected sites within the nervous system. This is demonstrated by gene delivery to both the striatum and substantia nigra following striatal inoculation; to the spinal cord, spinal ganglia, and brainstem following injection into the spinal cord; and to retinal ganglion neurons following injection into the superior colliculus and thalamus.