Human liver aldehyde dehydrogenase (E3 isozyme), with wide substrate specificity and low K(m) for 4-aminobutyraldehyde, was only recently characterized [Kurys, G., Ambroziak, W. & Pietruszko, R. (1989) J. Biol. Chem. 264, 4715-4721] and in this study we report on its primary structure. Polyclonal antibodies, specific for the E3 isozyme and three oligonucleotide probes derived from amino acid sequence of the E3 protein, were used for isolation of the first cDNA clone encoding the human enzyme (1503 bp; coding for 440 amino acid residues). Additional clones were obtained by using the first isolated clone as a probe. The largest clone of 1635 bp coded for 462 amino acid residues; it was longer at the 3'end of the cDNA non-coding region. The identity of the clone was established by DNA sequencing and by comparison with peptide sequences derived from the E3 protein, which constituted approximately 29% of the total primary structure of the E3 isozyme. The start codon was never encountered despite a variety of different approaches (500 amino acid residues were expected on the basis of SDS-gel molecular-mass determination of the E3 isozyme subunit). Despite the great catalytic similarity between the E3 and El isozymes [Ambroziak, W. & Pietruszko, R. (1991) J. Biol. Chem. 266, 13011-13018], the primary structure of the E3 isozyme has only approximately 40.6% of positional identity with that of the El isozyme. Sequence comparison with GenBank and Protein Identification Resource database sequences indicated no primary structure of aldehyde dehydrogenase more closely resembling the E3 isozyme than that of Escherichia coli betaine aldehyde dehydrogenase (52.7% positional identity), a prokaryotic enzyme specific for betaine aldehyde.