A distinct nuclear form of human uracil-DNA glycosylase [UNG2, open reading frame (ORF) 313 amino acid residues] from the UNG gene has been identified. UNG2 differs from the previously known form (UNG1, ORF 304 amino acid residues) in the 44 amino acids of the N-terminal sequence, which is not necessary for catalytic activity. The rest of the sequence and the catalytic domain, altogether 269 amino acids, are identical. The alternative N-terminal sequence in UNG2 arises by splicing of a previously unrecognized exon (exon 1A) into a consensus splice site after codon 35 in exon 1B (previously designated exon 1). The UNG1 sequence starts at codon 1 in exon 1B and thus has 35 amino acids not present in UNG2. Coupled transcription/translation in rabbit reticulocyte lysates demonstrated that both proteins are catalytically active. Similar forms of UNG1 and UNG2 are expressed in mouse which has an identical organization of the homologous gene. Constructs that express fusion products of UNG1 or UNG2 and green fluorescent protein (EGFP) were used to study the significance of the N-terminal sequences in UNG1 and UNG2 for subcellular targeting. After transient transfection of HeLa cells, the pUNG1-EGFP-N1 product colocalizes with mitochondria, whereas the pUNG2-EGFP-N1 product is targeted exclusively to nuclei.