Low NAD(P)H:quinone oxidoreductase activity is associated with increased risk of leukemia with MLL translocations in infants and children

NAD(P)H:quinone oxidoreductase 1 (NQO1) protects cells against oxidative stress and toxic quinones.1,2 A cytosine to thymine (C→T) polymorphism at position 609 in the NQO1 gene (NQO1 C609T)produces a proline to serine substitution that destabilizes and inactivates the enzyme.3 Individuals who are homozygous for NQO1 C609T are completely lacking in NQO1 activity, whereas individuals who are heterozygous have low-to-intermediate NQO1 activity compared with wild-type individuals.4 The frequency of NQO1 C609T is similar in whites and African Americans, but higher in Hispanics and Asians.5-8,NQO1 C609T has been associated with a greater risk of neutropenia in benzene-exposed adult Chinese workers9 and is significantly overrepresented in therapy-related10 and de novo leukemias11 in adults. Recently, Wiemels et al reported that NQO1 C609T conferred susceptibility to infant ALL and acute myeloid leukemia (AML) with MLL translocations in a British population of 36 cases, with the greatest risk in cases of ALL with t(4;11).12 Here we expand upon the findings of Wiemels et al12 by investigating this inactivatingNQO1 polymorphism as a risk factor for pediatric de novo and treatment-related leukemias with MLL translocations in a United States population.

The institutional review boards of our institutions approved this research. Genomic DNAs were prepared from bone marrow or peripheral blood leukemic cells as previously described.13-15 For cytogenetic studies, the cells were cultured for 24 hours without mitogen and karyotypes were prepared by standard methods.16 

There were 39 patients diagnosed with de novo leukemia characterized by translocation of the MLL gene at chromosome band 11q23, and 18 with treatment-related leukemia with MLL translocations. We examined 56 patients with de novo B-lineage ALL withoutMLL gene rearrangement as a comparison population with a common pediatric cancer. The demographic features of these patients are shown in Table 1. The 39MLL(+) de novo cases were diagnosed from birth to 18 years, 5 months of age. The karyotypes were previously described.13,17,18 There were 19 ALL cases, including one case of T-cell ALL; 17 AML cases; and 3 biphenotypic cases. The morphology was French-American-British (FAB) M4 (myelomonocytic) or FAB M5 (monoblastic) in 13 cases of AML and in one of the biphenotypic cases. There were 16 patients with ALL, 12 patients with AML, and all 3 patients with biphenotypic leukemia who were diagnosed before 12 months of age, and 1 patient with AML who was diagnosed at age 15 months; these 32 patients were considered infants. Of the 32 infants, 7 were diagnosed at birth. In all cases there was evidence of MLL gene rearrangement by Southern blot analysis.13 The MLL translocations were characterized cytogenetically and/or, in some cases, by panhandle polymerase chain reaction (PCR) approaches.13,17 19-21 

The 18 MLL(+) treatment-related leukemias and prior DNA topoisomerase II inhibitor exposures have also been described.14,18,22 The ages of the patients ranged from 3 years, 7 months to 17 years, 2 months when the diagnosis of treatment-related leukemia was made. There were 13 treatment-related leukemia patients who presented with AML, 2 with myelodysplasic syndrome (MDS), 2 with ALL, and 1 was biphenotypic. The detection of MLL translocations was the same as in the de novo cases.14,18 22 

The 56 patients with MLL(-) de novo B-lineage ALL were included in prior studies.15 18 The age at diagnosis in 48 of these patients for whom data were available ranged from 1 year, 4 months to 19 years, 11 months.

All laboratory personnel were blinded to case-control status.NQO1 genotypes were analyzed as previously described.10 Wild-type (CC) individuals were assigned to the high activity category. Individuals who were heterozygous (CT) or homozygous (TT) for the C609T polymorphism were assigned to the low/null NQO1 activity category because the homozygous group was too small to analyze alone.

NQO1 genotype frequencies in patients withMLL(+) de novo or treatment-related leukemias were compared with NQO1 genotype frequencies in children withMLL(-) de novo B-lineage ALL. The expected NQO1allele frequencies in a disease-free population ethnically matched to the MLL(+) de novo cases were calculated from previously published data5,8,9 23 and used as a second comparison group. The significance of the difference between groups was determined by constructing 2-by-2 tables and generating crude odds ratios and 95% confidence intervals using Cornfield approximations, and 2-tailedP values were calculated using Fisher exact methods. All results were considered statistically significant if the 2-tailedP value was less than .05. The analysis was carried out using the statistical computer program STATA (Stata Corporation, College Station, TX).

NQO1 genotypes in the 3 groups including pediatric patients with MLL(+) de novo leukemia (n = 39), pediatric patients with MLL(+) treatment-related leukemia (n = 18), and pediatric patients with MLL(-) de novo B-lineage ALL (n = 56), are shown in Table 2. In the patients with MLL(+) de novo leukemia, there was a strong shift toward heterozygosity at the NQO1 C609T allele. These patients were significantly more likely to be heterozygous atNQO1 C609T than patients in the comparison group withMLL(-) de novo B-lineage ALL (OR = 2.77, 95% CI 1.17-6.57; P = .02). Assigning individuals who were heterozygous (CT) or homozygous (TT) for the C609Tpolymorphism to the low/null NQO1 activity category revealed that patients with MLL(+) de novo leukemia were significantly more likely to have low/null NQO1 activity than patients withMLL(-) de novo B-lineage ALL (OR = 2.47, 95% CI 1.08-5.68; P = .033) or than would be expected in an ethnically matched population of disease-free subjects (OR = 2.50,P = .02) (Table 2). This almost identical finding of an increased OR of approximately 2.5 when the MLL(+) group is compared with 2 different groups shows that bias due to ethnic differences or population stratification is unlikely to explain the findings.

There was no difference in the susceptibility of males and females withMLL(+) de novo leukemia (OR = 1.07, 95% CI 0.29-3.86;P = .92), indicating that the NQO1 genotype effect is sex independent. When the MLL(+) de novo cases were analyzed by lineage, a statistically significant association of NQO1 C609T was only observed with ALL (OR = 3.35) (Table 3). The sample size of patients with MLL(+) de novo AML was too small to make strong inferences.

The 39 patients with MLL(+) de novo leukemias were further analyzed by age at diagnosis, and by whether t(4;11) was observed in the karyotype (Table 3). Of the 39 MLL(+) de novo cases, 32 were classified as infant leukemias as described above. The OR forNQO1 low/null genotypes among these 32 infant cases compared with the 56 cases of MLL(-) de novo B-lineage ALL was 2.25, which is similar to the OR of 2.47 for the entire group of patients with MLL(+) de novo leukemia compared with the same comparison group. For the 7 patients who were more than 2 years old at diagnosis of MLL(+) de novo leukemia, the OR for NQO1 low/null genotypes was 3.86 compared with the same comparison group with MLL(-) de novo B-lineage ALL, but this was not significantly different than the OR of 2.25 observed for the infants.

The most common MLL translocation in infant ALL is t(4;11)(q21;q23),24,25 which fuses MLL withAF-4.26 In the de novo leukemias in the present study, the karyotype revealed t(4;11) in 9 cases of ALL, 2 cases of AML, and one biphenotypic leukemia. NQO1 low/null genotypes were present in 10 (83.3%) of these 12 cases, and the OR compared with the control population of children with MLL(-) de novo B-lineage ALL was 7.73 (95% CI 1.7 to infinity) (Table 3). This result is highly statistically significant (P = .006). When only those infants diagnosed with leukemia with t(4;11) before 12 months of age (n = 8) were analyzed, the OR was even higher at 10.82, and also was highly statistically significant (Table 3). Panhandle PCR and/or cDNA panhandle analysis has been performed in 8 of the 12 cases with cytogenetic evidence of t(4;11) and, in each of the 8 cases, there was evidence of a translocation fusing MLL toAF-4.19,21 There were 2 other de novo leukemias with molecular evidence of a translocation fusing MLL toAF-4. In one case the karyotype was normal13,19; in the other case, there were no mitoses in the diagnostic marrow for karyotype analysis but the karyotype at relapse was complex and included evidence of t(4;11).13,17 21 Both cases were infant ALL and both had low/null NQO1 activity, such that when these cases were included in the analysis of leukemias with t(4;11), the association of low/null NQO1 activity with t(4;11) became even stronger in the infants less than 12 months old at diagnosis (OR = 13.91) (Table 3).

The distribution of NQO1 genotypes among patients with treatment-related leukemias was not statistically different from that found in patients with MLL(-) de novo B-lineage ALL with respect to either low/null NQO1 activity (OR = 0.59; 95% CI 0.19-1.85; P = .38) or the frequency of heterozygosity (OR = 0.73; 95% CI 0.23-2.3; P = .6). Moreover, the OR and trend toward heterozygosity appeared to be in the opposite direction compared with that found in pediatric patients withMLL(+) de novo leukemias. This may be related to the predominance of other MLL translocations, especially t(9;11) and t(11;19), in the treatment-related cases included in this study versus the predominance of t(4;11) in the MLL(+) de novo cases or, alternatively, to differing etiologies.

We have shown that the inactivating NQO1 C609T polymorphism is associated with an increased risk of leukemia with MLL translocations in infants and children in a United States population. These findings are almost identical to those of Wiemels et al12 in a population of British infants and further support the hypothesis that low/null NQO1 activity is a risk factor for infant leukemias harboring MLLtranslocations. The odds ratios between 7.7 and 13.9 in the patients with de novo leukemias with t(4;11) compared with a population of children with MLL(-) de novo B-lineage ALL are consistent with the findings of Wiemels et al, who observed an 8-fold increased risk for infant leukemias with t(4;11) when using normal cord blood samples as controls.12 Moreover, in the present study, the odds ratios became even higher (10.82 to 13.91) when considering only infant leukemias with t(4;11). While the total number of subjects studied is still small and the confidence intervals quite wide, this analysis by karyotype may imply that there may be different risk factors for de novo leukemias harboring different MLLtranslocations.

The reason why low/null NQO1 activity appears to be so strongly associated with the t(4;11) translocation in particular is unknown. However, the finding of the same NQO1 genotype-leukemia association in 2, albeit small, independent studies supports a role for NQO1 substrate(s) such as benzoquinone and related compounds and/or oxidative stress as causative factors in leukemias with MLLtranslocations, especially infant leukemias with t(4;11). Because similar MLL translocations are found in leukemias related to chemotherapy with DNA topoisomerase II inhibitors, DNA topoisomerase II has been implicated in the generation of MLL gene rearrangements (reviewed in Felix27). MLLtranslocations in infant leukemias arise in utero,28,29and maternal prenatal consumption of food items containing DNA topoisomerase II inhibitors may increase the risk of infant AML.30,31 Flavonoids such as genistein, quercetin, and daidzein are examples of dietary DNA topoisomerase II inhibitors,32 and it has been shown that quercetin inducesNQO1 gene expression.33 In addition, there are numerous potential dietary and environmental sources of the NQO1 substrate 1,4-benzoquinone,34 which potentially may interfere with the catalytic activity of DNA topoisomerase II35 or interact with DNA directly. Therefore, it is biologically plausible that the NQO1 genotype is a host factor that modulates the risk of leukemia in infants. Exactly how low/null NQO1 activity may contribute to the increased risk remains unknown, but these findings suggest that the pursuit of NQO1 substrate(s) as potential causative factors in infant leukemia is warranted.

The present study also provides new information on the group of pediatric patients with treatment-related leukemias with MLLtranslocations. Although the sample size for this group was small, the results suggest that NQO1 does not have a protective role againstMLL(+) leukemias arising after chemotherapeutic DNA topoisomerase II inhibitors, where the CYP3A4 wild-type genotype has been shown to confer susceptibility.18 

We acknowledge the AML Cell Bank of the Children's Cancer Group for providing several of the de novo leukemias reported on herein.