In vitro studies have shown that loss of DNA mismatch repair due to lack of either hMSH2 or hMLH1 activity results in low-level resistance to cisplatin but not to oxaliplatin, an analogue that produces a different type of DNA adduct. No information is currently available on whether this low-level resistance is sufficient to result in enrichment of mismatch repair-deficient cells during drug exposure in vitro or to account for clinical failure of treatment in vivo. Mixed populations of cells containing a minority of DNA mismatch repair-deficient cells constitutively expressing green fluorescence protein were exposed repeatedly in vitro to cisplatin and oxaliplatin. Treatment with cisplatin resulted in a gradual enrichment for DNA mismatch repair-deficient cells, whereas treatment with oxaliplatin did not. MSH2-/- and MSH2+/+ embryonic stem cells were established as xenografts in athymic nude mice. Animals were treated 48 h after tumor implantation with a single LD10 dose of either cisplatin or oxaliplatin. MSH2-/- tumors were significantly less responsive to cisplatin than MSH2+/+ tumors, whereas there was no difference in sensitivity to oxaliplatin. These results demonstrate that the degree of cisplatin resistance conferred by loss of DNA mismatch repair is sufficient to produce both enrichment of mismatch repair-deficient cells during treatment in vitro and a large difference in clinical responsiveness in vivo. The results identify loss of DNA mismatch repair as a mechanism of resistance to cisplatin but not oxaliplatin.