Glial fibrillary acidic protein (GFAP) is an intermediate filament protein abundantly expressed in malignant gliomas. We have constructed a novel oncolytic adenovirus, Ad5-gfa2(B)3-E1, for treatment of these tumors. In this construct, the E1 region is under control of the tissue-specific GFAP promoter (gfa2) with three additional copies of the glial specific 'B' enhancer. Infection of a GFAP-positive cell line with Ad5-gfa2(B)3-E1 resulted in E1A and E1B expression at 75% and 30% of the levels obtained after wtAd5 infection. Q-PCR showed that Ad5-gfa2(B)3-E1 replicated 4.5 times more efficiently in the GFAP-positive than in the GFAP-negative cell lines. Cell viability assays showed efficient elimination of GFAP-positive cells by Ad5-gfa2(B)3-E1, in some cell lines as efficiently as wtAd5, while the elimination was attenuated in GFAP-negative cell lines. When tested in human tumor xenografts in nude mice, Ad5-gfa2(B)3-E1 effectively suppressed the growth of GFAP-positive SNB-19 glial tumors but not of GFAP-negative A549 lung tumors. In Ad5-gfa2(B)3-E1, the E3 region was deleted to create space for future insertion of heterologous therapeutic genes. Experiments with dl7001, an E3-deleted variant of wtAd5, confirmed that the specificity of Ad5-gfa2(B)3-E1 replication was based on the promoter driving E1 and not on the E3 deletion. Strategies to further improve the efficacy of Ad5-gfa2(B)3-E1 for the treatment of malignant gliomas include the insertion of therapeutic genes in E3 or retargeting to receptors that are more abundantly expressed on primary glioma cells than CAR.