Deoxyinosine (hypoxanthine deoxyribonucleoside) in DNA is produced by the hydrolytic or nitrosative deamination of deoxyadenosine. Hypoxanthine pairs with cytosine during replication resulting in A:T to G:C transition mutation if it is not repaired. Deoxyinosine (dI) is excised in a repair pathway that is initiated by endonuclease V (endo V), the nfi gene product, in Escherichia coli. This enzyme cleaves the second phosphodiester bond 3' to the deoxyinosine. It is suggested that DNA polymerase I and DNA ligase participate in the repair of deoxyinosine but the complete mechanism of deoxyinosine excision after endo V cleavage is not known. We constructed an dI:dG mispair heteroduplex substrate which dI is designed in the XhoI restriction enzyme site. It can be digested by XhoI if dI:dG is repaired to dG:dC. If not, it cannot be recognized by XhoI. We can determine the repair level by evaluating the digested products. Our results showed the repair levels of the dI:dG substrate were not significantly different between mismatch proficient NM522 extracts and mismatch repair-deficient extracts (GM2931, RK1517). We concluded that the involvement of MutHLS pathway in our assay system is insignificant. Furthermore, the repair level of AlkA extracts was similar with that observed in NM522 extracts. Repair of dI heterology was highly dependent on the presence of the nfi gene product in crude extracts. The repair level in endo V deficient extracts decreased to less than 50% of that of endo V-proficient extracts. We also tested the requirement for endo V-mediated repair in E.coli extracts. Mg2+ is indispensable in the repair of dI and addition of exogenous ATP or 4 dNTPs was not required. We also performed the repair reaction with three purified enzymes:endonuclease V, DNA polymerase I, and E. coli DNA ligase. To our surprise, without any additional component from cell extracts, these three enzymes alone were sufficient to reconstitute the repair in vitro. We also replaced DNA polymerase I with klenow fragment 3’→5’ exo- in the reconstitution reaction and found the dI heterology cannot be repaired. This observation clearly demonstrated that 3’→5’ exonuclease activity of DNA polymerase I is essential for the endo V-mediated excision repair assay.