We isolated three glial fibrillary acidic protein (GFAP) cDNA clones from a glioma cell line, U-251 MG. One clone isolated from a U-251 MG cDNA library was long, but lacked both ends. Using poly(A)+ RNA and primers synthesized according to the sequence of this clone, we used the polymerase chain reaction-assisted rapid amplification of cDNA ends (PCR-RACE) method, which is a strategy to isolate cDNA ends, and obtained cDNA clones for the 5' and 3' ends. From the sequences of these overlapping clones, the complete nucleotide sequence of human GFAP cDNA was established. The start (ATG) and the stop (TGA) signals were seen at nucleotide positions 15 and 1311, respectively, and divided the entire sequence of 3027 bp into 14 bp of 5' non-coding, 1296 bp of coding and 1717 bp of 3' non-coding regions. Using cDNA probes made from both the coding and the 3' non-coding regions, Northern blot hybridization was performed with two different stringencies on RNAs from human and rodent brains and human GFAP-positive and -negative cells. It was shown that the 3' non-coding region probe was more specific for human GFAP than the coding region probe which was specific only under higher stringency conditions. This was also suggested by homology analysis of the sequence with those of various intermediate filament proteins. Based on these findings, we performed spot blot hybridization of sorted human chromosomes and Southern blot hybridization of PCR-amplified DNAs of a panel of hamster-human somatic cell hybrids and localized the human GFAP gene to chromosome 17.