The t(8;21)(q22;q22) that results in an AML1-ETO fusion gene is frequently detected in de novo and in therapy-related acute myeloid leukemia (AML) patients. Most t-AML patients with t(8;21) have received treatment with topoisomerase (topo) II inhibitors for primary tumors. In our previous study in 31 de novo leukemia patients with t(8;21) genomic breakpoints clustered in each three breakpoint cluster regions (BCRs) in intron 5 in AML1 and in intron 1b in ETO which tended to colocalize with DNA topo II cleavage and DNase I hypersensitive sites, implicating these chromatin structural elements in the mechanisms leading to the t(8;21). In this study, we cloned genomic breakpoints in AML1 and ETO in both derivative chromosomes in six t-AML leukemia patients with t(8;21), all of whom were treated with topo II inhibitors for previous tumors. Genomic breakpoints in AML1 and ETO in t-AML patients cluster in the same BCRs previously identified in de novo patients with t(8;21). Our results are unexpected because in MLL translocations involving 11q23, the location of breakpoints in MLL in de novo and in t-AML patients is different. There were deletions and duplications at the breakpoints in both AML1 and ETO, and microhomologies or non-templated nucleotides occurred at most of the genomic fusion junctions. No specific recombination motifs were identified at or near the breakpoint junctions. Both deletions and duplications were larger in de novo leukemia patients than in t-AML patients. In each 10 de novo patients, the deletions ranged from 5 to 556 bp (251 bp in average) in AML1 and from 6 to 225 bp (88 bp) in ETO, whereas in t-AML patients the deletions in AML1 ranged from 1-111bp (39 bp) in three and from 2-125 bp (32 bp) in ETO in five patients. Duplications in AML1 in seven de novo patients ranged from 1-141 bp (76 bp) and in ETO in eight patients from 51 to 355 bp (185 bp); in contrast, duplications in AML1 were 1, 2 and 1085 bp in three t-AML patients, and only one t-AML patient had a duplication (1 bp) in ETO. Similarly, only very small (2-7 bp) deletions or duplications were detected in MLL and CBP in t-AML patients with t(11;16). Thus, there are some unexplained differences in non-homologous end joining repair in de novo and therapy-related leukemia. We mapped a scaffold attachment region (SAR) in intron 4 of AML1, and two SARs in intron 1b of ETO that are near the BCRs. Moreover, we analyzed the free energy (delta G) that is needed to unwind double strand DNA of the BCR-bearing introns in AML1 and ETO and found that the topo II cleavage sites we identified in both genes have the lowest delta G value suggesting that topo II cleaves DNA at the point of the lowest free energy. Our study suggests that illegitimate recombination between AML1 and ETO in the t(8;21) in both de novo and therapy-related leukemia patients is similar, and that topo II cleavage sites with the lowest free energy provide vulnerable sites for breakage. Non-homologous end joining repair is a likely mechanism involved in the formation of the t(8;21).