The crystal structure analyses of the EcoRI-DNA and EcoRV-DNA complexes do not provide clear suggestions as to which amino acid residues are responsible for the activation of water to carry out the DNA cleavage. Based on molecular modeling, we have proposed recently that the attacking water molecule is activated by the negatively charged pro-Rp phosphoryl oxygen of the phosphate group 3' to the scissile phosphodiester bond. We now present experimental evidence to support this proposal. (i) Oligodeoxynucleotide substrates lacking this phosphate group in one strand are cleaved only in the other strand. (ii) Oligodeoxynucleotide substrates carrying an H-phosphonate substitution at this position in both strands and, therefore, lacking a negatively charged oxygen at this position are cleaved at least four orders of magnitude more slowly than the unmodified substrate. These results are supported by other modification studies: oligodeoxynucleotide substrates with a phosphorothioate substitution at this position in both strands are cleaved only if the negatively charged sulfur is in the RP configuration as shown for EcoRI [Koziolkiewicz, M. & Stec, W.J. (1992) Biochemistry 31, 9460-9466] and EcoRV (B. A. Connolly, personal communication). As the phosphate residue 3' to the scissile phosphodiester bond is not needed for strong DNA binding by both enzymes, these findings strongly suggest that this phosphate group plays an active role during catalysis. This proposal, furthermore, gives a straightforward explanation of why in the EcoRI-DNA and EcoRV-DNA complexes the DNA is distorted differently, but in each case the 3' phosphate group closely approaches the phosphate group that is attacked. Finally, an alternative mechanism for DNA cleavage involving two metal ions is unlikely in the light of our finding that both EcoRI and EcoRV need only one Mg2+ per active site for cleavage.