We investigated which features of the substrate specificity of human immunodeficiency virus type 1 (HIV-1) integrase could be assigned to the central domain of the 288-residue HIV-1 integrase protein, composed of amino acids 50-212. This domain contains the active site and shares structural homology with a large family of polynucleotidyl transferases. Using model substrates with defined alterations in critical features we found that this domain alone is sufficient for recognition of: 1) the phylogenetically conserved CA/TG base pairs near the viral DNA end; 2) the 5'-terminal dinucleotide that is left unpaired after end processing; and 3) target DNA flanking the site of joining. Future efforts aimed at identifying specific amino acids involved in recognition of these key substrate features can now be targeted at this domain.