Classification of humans as rapid or slow acetylators is based on hereditary differences in rates of N-acetylation of therapeutic and carcinogenic agents, but N-acetylation of certain arylamine drugs displays no genetic variations. Two highly homologous human genes for N-acetyltransferase (NAT; arylamine acetyltransferase, acetyl CoA:arylamine N-acetyltransferase, EC 2.3.1.5), NAT1 and NAT2, presumably code for the genetically invariant and variant NAT proteins, respectively. In the present investigation, 1.9-kilobase human genomic EcoRI fragments encoding NAT2 were generated by the polymerase chain reaction with liver and leukocyte DNA from seven subjects phenotyped as homozygous and heterozygous acetylators. Direct sequencing revealed multiple point mutations in the coding region of two distinct NAT2 variants. One of these was derived from leukocytes of a slow acetylator and was distinguished by a silent mutation (codon 94) and a separate G----A transition (position 590) leading to replacement of Arg-197 by Gln; the mutated guanine was part of a CpG dinucleotide and a Taq I site. The second NAT2 variant originated from liver with low N-acetylation activity. It was characterized by three nucleotide transitions giving rise to a silent mutation (codon 161), accompanied by obliteration of the sole Kpn I site, and two amino acid substitutions: Thr for Ile (codon 114) and Arg for Lys (codon 268). Heterozygosity was detected in three NAT2 samples: two were heterozygous for the rapid and one of the allelic variants, and the third was a compound heterozygote of both mutant alleles. The results show conclusively that the genetically variant NAT is encoded by NAT2.