The DNA polymerase III holoenzyme of Escherichia coli contains a potent 3'----5' exonuclease that removes the terminal nucleotide from a synthetic deoxyoligonucleotide primer with a half-life of approximately 2 s. Degradation of primers could not be effectively prevented by permitting the holoenzyme to "idle" at the primer terminus in the presence of limited deoxynucleoside triphosphates. To further characterize this exonuclease and to develop stable primers to facilitate experimental manipulations, we synthesized a series of twelve 25-mer oligonucleotides that differed only in the two 3'-terminal residues. The penultimate position contained either a CMP or a dCMP residue, while at the terminal position either AMP, dAMP, 2',3'-dideoxyAMP, cordycepin (3'-dAMP), dAMP alpha S, or 2',3'-dideoxyAMP alpha S was incorporated. No single change at either the 3'-penultimate or 3'-terminal positions resulted in a decrease in the exonuclease rate greater than 10-fold; however, combined changes at these two sites resulted in a strong synergistic effect. Placing a ribonucleotide at the penultimate position coupled by a phosphorothioate linkage to a terminal 2',3'-dideoxynucleotide reduced the rate of exonucleolytic activity almost 30,000-fold (half-life approximately 16 h). If only the ribonucleotide and phosphorothioate substitutions were made, a primer capable of being efficiently elongated was generated that exhibited a 500-fold increase in stability (half-life = 40 min). The elemental effect observed by substituting a nonbridging oxygen in the terminal phosphodiester bond for sulfur increased from 1.5 to 200 as other substitutions were made that decreased the exonuclease rate. This was consistent with a change in the rate-limiting step of the exonuclease reaction from a conformational change to the chemical step where the covalent bond is cleaved. At least part of this effect appears to be due to perturbations within the enzyme's active site and not solely due to changes in electrophilicity.