The ubiquitous octamer-binding protein oct-1 contains a POU domain required for DNA binding, which can be subdivided into a POU-specific domain and a POU homeo domain. We have overproduced the POU domain and the POU homeo domain in a vaccinia expression system, purified both polypeptides to near homogeneity, and compared their DNA-binding properties. In contrast to the POU domain, the homeo domain protects only part of the octamer sequence in the Ad2 origin against breakdown by DNase I or hydroxyl radicals. Analysis of purine contacts by DMS and DEPC interference assays shows that the Ad2 octamer can be divided into two regions: one that is recognized both by the POU domain and the homeo domain in an identical fashion, and one that is only recognized by the POU domain. This suggests that the POU-specific domain is responsible for the additional contacts located at one side of the octamer. In agreement with this, mutating the first 3 nucleotides (ATG) of the octamer affected binding by the POU domain but not by the homeo domain. The apparent binding affinities to different octamer sites were compared. The homeo domain binds 600-fold less efficiently to the canonical octamer sequence (ATGCAAAT) than the POU domain. The difference is only sevenfold for the Ad2 octamer, whereas both Kd values are almost identical for the HSV ICP4 TAATGARAT motif. Both the POU and homeo domains recognize target sequences for mammalian homeo box proteins. We conclude that the octamer can act as a bipartite recognition sequence for oct-1 and that the POU-specific domain contributes to the binding affinity, as well as to the specificity, by providing additional contacts.