We have used reciprocal competition binding experiments with mutant substrates and chemical modification interference assays to precisely define the sequences within the adeno-associated virus (AAV) terminal repeat (TR) that are involved in site-specific binding to the AAV Rep protein. Mutagenesis experiments were done with a 43-bp oligonucleotide which contained the Rep binding element (RBE) within the A stem of the TR. Experiments in which two adjacent base pairs of the RBE were substituted simultaneously with nucleotides that produced transversions identified a 22-bp sequence (CAGTGAGCGAGCGAGCGCGCAG) in which substitutions measurably affected the binding affinity. Although the 22-bp RBE contains the GAGC motifs that have been found in all known Rep binding sites, our results suggest that the GAGC motifs alone are not the only sequences specifically recognized by Rep. The effects of substitutions within the 22-bp sequence were relatively symmetrical, with nucleotides at the periphery of the RBE having the least effect on binding affinity and those in the middle having the greatest effect. Dinucleotide mutations within 18 (GTGAGCGAGCGAGC) of the 22 bp were found to decrease the binding affinity by at least threefold. Dinucleotide mutations within a 10-bp core sequence (GCGAGCGAGC) were found to decrease binding affinity by more than 10-fold. Single-base substitutions within the 10-bp core sequence lowered the binding affinity by variable amounts (up to fivefold). The results of the mutagenesis analysis suggested that the A-stem RBE contains only a single Rep binding site rather than two or more independent sites. To confirm the results of the mutant analysis and to determine the relative contribution of each base to binding, chemical modification experiments using dimethyl sulfate and hydrazine were performed on both the linear A-stem sequence and the entire AAV TR in both the flip and flop hairpinned configurations. Interference assays on the linear A stem identified the 18-bp sequence described above as essential for binding. G, C, and T residues on both strands contributed to binding, and the interference pattern correlated well with the results of the mutagenesis experiments. Interference assays with complete hairpinned TR substrates also identified the 18-bp sequence as important for binding. However, the interference patterns on the two strands within the RBE and the relative contributions of the individual bases to binding were clearly different between the hairpinned substrates and the linear A-stem binding element. Interference assays also allowed us to search for residues within the small internal palindromes of the TR (B and C) that contribute to binding. The largest effect was seen by modification of two T residues within the sequence CTTTG. This sequence was present in the same position relative to the terminal resolution site (trs) in both the flip and flop orientations of the TR. In addition, the interference pattern suggested that the remaining bases within the CTTTG motif as well as other bases within the B and C palindromes make contacts with the Rep protein, albeit with lower affinities. Regardless of whether the TR was in the flip or flop orientation, most of the contact points were clustered in the small internal palindrome furthest away from the trs. We also determined the relative binding affinity of linear substrates containing a complete RBE with hairpinned substrates and found that linear substrates bound Rep less efficiently. Our results were consistent with our previous model that there are three distinct elements within the hairpinned AAV TR that contribute to binding affinity or to efficient nicking at the trs: the A-stem RBE, the secondary structure element which consists of the B and C palindromes, and the trs.