Due to the lack of any effective therapy, novel approaches are currently being explored for the treatment of primary brain tumours. It has previously been demonstrated that variants of HSV-1 which are deleted in the RL1 gene and fail to produce the virulence factor ICP34.5 are potential candidates for tumour therapy. The RL1 variant 1716 replicates selectively within tumour cells and has the potential to deliver a therapeutic or tumour killing gene directly to the site of tumour growth. As many intracerebral tumours are glial and predominantly astrocytic in origin, we have evaluated the ability of 1716 to deliver a reporter gene specifically to astrocytes in vivo and in vitro using a 2.2 kb fragment which controls expression of the glial fibrillary acidic protein (GFAP), an astrocyte specific protein. Two 1716 variants, 1774 and 1775, were constructed which contain the GFAP-promoter element linked to the E. coli beta-galactosidase gene, inserted into the HSV-1 UL43 and US5 loci, respectively. In primary cultures, human primary tumour cell lines and established tumour cell lines in vitro, 1774 and 1775 gave high levels of expression of beta-galactosidase specifically in astrocytes. In vivo following intracerebral inoculation, both viruses demonstrated high levels of beta-galactosidase expression predominantly in astrocytes. These results indicate that the GFAP promoter element could be used for efficient and selective transgene delivery to human gliomas.