CYP1A1*2B (Val) allele is overrepresented in a subgroup of acute myeloid leukemia patients with poor-risk karyotype associated with NRAS mutation, but not associated withFLT3 internal tandem duplication

The etiology of acute myeloid leukemia (AML) is unknown for the majority of cases. The identification of biologically well-defined subgroups on the basis of cytogenetic analysis has improved prognostic power and led to risk-directed therapeutic strategies. It is increasingly clear that the etiologic mechanisms underlying the “good-risk” translocational karyotypes t(15:17), inv(16), and t(8:21), which predominate in younger patients, will differ from the largely deletional abnormalities seen in “poor-risk” disease, which are more common in older patients. The karyotypic similarities between therapy-related AML (t-AML) and poor-risk de novo AML1 suggest that an environmental etiology is more likely in the latter group than in the de novo good-risk group. Questionnaire-based case-control studies have suggested associations between exposure to specific environmental toxicants and the development of AML, but cannot define mechanisms of leukemogenesis.2 3 

The cytochrome P450 enzymes are expressed predominantly in the liver and function to detoxify many environmental toxicants (xenobiotics). These phase 1 metabolic enzymes transfer electrons onto substrate toxicants to create highly reactive intermediates, which are then available for detoxification by a variety of phase 2 enzymes including glutathione S-transferases and sulfotransferases.

Several carcinogen-metabolizing genes with functional allelic variants are now known to predispose to t-AML4,5 and also to de novo AML.6-8 Furthermore, this predisposition may be greater or less in subgroups of AML defined by specific genomic damage at the karyotypic level. The paradigm for an AML-predisposition gene encoding a carcinogen-metabolizing enzyme is NADP(H) quinone oxidoreductase (NQO1). TheC609T polymorphic variant, which confers reduced phase 2 metabolism, is associated with a predisposition to therapy-related AML4 and selected cytogenetic subgroups of de novo AML.6 Reduced phase 2 metabolism has the potential to result in an accumulation of reactive intermediates, which in turn oxidize DNA and protein, leading to DNA mutation and/or cell death. Phase 1 enzymes with functionally relevant allelic variants are also predisposition genes for AML and include cytochrome P450 2C19 (CYP2C19) and CYP2D6.8 The proposed mechanism of predisposition for these 2 genes is a consequence of reduced phase 1 metabolism, presumably allowing the accumulation of nonmetabolized toxic carcinogens.

We selected functional polymorphic variants in specific carcinogen-metabolizing enzymes for study. Enzymes were selected on the basis of an established role in the metabolism of known leukemogenic compounds (eg, benzene, cytotoxic chemotherapeutic drugs). We then selected only patients with well-characterized karyotypic abnormalities, to include abnormalities known to occur following genotoxic insult (eg, chromosome 5 and 7 abnormalities) and also good-prognosis translocations for which an environmental etiology may be less likely. Finally, we assayed the 2 most common molecular abnormalities in AML, namely, FLT3 internal tandem duplication (ITD) and NRAS mutation; each of which is likely to arise from a different mechanism.

We demonstrate that the *2B high-inducibility variant ofCYP1A1 is overrepresented in a subgroup of poor-risk AML patients with NRAS mutation and may represent a predisposition allele for these specific forms of genomic damage in AML.

We studied 447 patients with de novo or secondary AML (defined as secondary to chemotherapy, or following an antecedent hematologic disorder), and who were entered into the Medical Research Council (MRC) AML 10 (younger than 60 years, n = 109); AML 11 (older than 55 years, n = 115), and AML 12 (younger than 60 years, n = 223) clinical trials. This patient cohort overlaps with cohorts in previously published studies.6 Median age was 46 years (range, 15-79 years). DNA was extracted from presentation bone marrow by dodecyltrimethylammonium bromide lysis9 and stored at University College London (United Kingdom).

Most cytogenetic studies were done in local laboratories, all of them participants in a central quality control scheme (United Kingdom National External Quality Assessment Scheme). Where this was not available, cytogenetic analyses were done at a central facility in University College London. Standard methods of bone marrow culture were used; 20 metaphases were fully analyzed and reported in accordance with the International System for Human Cytogenetic nomenclature (ISCN) guidelines.10 A hierarchical cytogenetic classification was used to assign all patients to a cytogenetic subgroup as previously described.11 For patients with more than one chromosomal abnormality, the subgroup assigned was determined by the most important abnormality in the following order (with the most important listed first): established translocation, established deletion, established trisomy, nonestablished translocation, nonestablished deletion, nonestablished trisomy.11 Data in this study are now presented for cytogenetic risk groups on the basis of the MRC hierarchical classification12 and defined here as (1) good-risk, t(15;17), t(8;21), inv(16); (2) poor-risk, partial or complete deletion of chromosomes 5 and 7 (5q/7q), abnormalities of 3q; and (3) intermediate-risk, which includes +8, 11q23 abnormalities, +11, 17p−, +21, 20q−, 12p− (but excludes normal karyotype).

Assays for allelic variants at the following loci have been previously described: GSTM1 (null), GSTT1(null),7,CYP2D6 (poor/extensive metabolizer),CYP2C19*2 (681G→A; aberrant splice),CYP1A1*2B allele (2455A→G, Ile→Val),8 and NQO1(609C→T; Pro→Ser).6 In addition, a polymorphism in SULT1A1 (*2, 213G→ A;Arg→His) was assayed by allelic discrimination by means of the polymerase chain reaction (PCR)–based TaqMan technology (PE Applied Biosystems, Foster City, CA). Primer and probe sequences for this assay were as follows: forward primer, 5′-GGTTGAGGAGTTGGCTCTGC-3′ (300 nM); reverse primer, 5′-ACGTGTGCTGAACCATGAAGT-3′ (300 nM) (annealing temperature, 62°C); wild-type (WT) probe, 5′-AGTTTGTGGGGCGCTCCCTG-3′ (100 nM) (Tet labeled); variant probe 5′-AGTTTGTGGGGCACTCCCTGC-3′ (200 nM) (Fam labeled).

Genomic DNA was amplified from exons 12 and 13 of theFLT3 gene.13 This produced a wild-type fragment of 328 bases plus a higher–molecular weight fragment containing internal tandem duplicated sequences (when present) of varying length as previously described.13,14 Bands with higher molecular weight than wild type were cut from the gel and directly sequenced by means of a fluorescent primer adapted chain-termination method15 on an ABI 3100 sequencer (PE Applied Biosystems) to determine the start site and size of each duplication.

Separate assays were developed for mutation detection at “hot spots” in codons 12/13 (exon 1) and codon 61 (exon 2). Oligonucleotide primers amplifying short fragments (241 base pair [bp], exon 1; 201 bp, exon 2) were designed for PCR as follows: N12/13 assay: forward, 5′-GACTGAGTACAAACTGGTGG-3′; reverse, 5′-TGCATAACTGAATGTATACCC-3′. N61 assay: forward, 5′-CAAGTGGTTATAGATGGTGAAACC-3′; reverse, 5′-AAGATCATCCTTTCAGAGAAAATAAT-3′. PCR products were then subjected to denaturing heteroduplex high-pressure liquid chromatography (HPLC) analysis (dHPLC) (Transgenomic WAVE, Crewe, United Kingdom). PCR conditions were as follows: (1) Exon 1, HotStarTaq (Qiagen, Valencia, CA), 0.625 U. Primers (12.5 pmol), N12/13 forward, N12/13 reverse. Denaturing, 95°C for 15 minutes and 94°C for 30 seconds; annealing, 55.5°C for 1 minute; extension, 72°C for 1 minute for 35 cycles; final cycle at 72°C for 10 minutes. (2) Exon 2, HotStarTaq (Qiagen), 0.625 U. Primers (12.5 pmol), N61F forward, N61R reverse. Denaturing, 95°C for 15 minutes; 94°C for 30 seconds; annealing, 55.5°C for 1 minute; extension, 72°C for 1 minute for 35 cycles; final cycle at 72°C for 10 minutes. Heteroduplexes were then generated by means of a thermal cycler as follows: 95°C for 5 minutes; 95°C, reducing at 1°C per 22 seconds, for 70 cycles. Then, 10 μL heteroduplexed PCR product per well was loaded from 96-well plates and analyzed by dHPLC under the following conditions: flow, 0.9 mL/min, 47% to 52% buffer (B) in 0.1 minutes, to 60% B in 4 minutes at 61°C. Representative dHPLC plots are shown in Figure1.

Samples exhibiting an abnormal dHPLC profile were confirmed as mutant by direct sequencing and/or PCR-restriction fragment length polymorphism mutation-sensitive digestion analysis for codons 12 and 13.16,17 If these assays were insufficiently sensitive to confirm mutation, PCR products were cloned (Original TA Cloning Kit; Invitrogen, Groningen, the Netherlands) and sequenced.15 The sensitivity of the dHPLC is such that 15% mutant DNA can be confidently detected in a sample containing wild-type and mutant sequence (M.E.F., manuscript in preparation).

CYP1A1 allellic variants were analyzed as dichotomous variables (Ile/Val plus Val/Val, versusIle/Ile). Odds ratios (ORs) and 95% confidence intervals were calculated from 2 × 2 tables for mutant versus nonmutant(NRAS or FLT3) and Ile/Val plusVal/Val, versus Ile/Ile. Logistic regression was used to test these associations within the entire cohort and also between the 3 cytogenetic risk groups. Power calculations were prospectively computed to inform sample size for the case-control study reported elsewhere6 and were based upon expected enzyme variant polymorphism frequencies. These prospective calculations were not done for NRAS or FLT3 ITD frequency within the patient cohort alone. Retrospective power calculations were performed with the assumption of an expected CYP1A1*2Ballele variant frequency (Ile/Val plus Val/Val)of 11%,8 an FLT3 ITD mutant frequency of 25%,14 and an NRAS mutant frequency of 15%. The sample size of 427 patients was sufficient to provide 80% power at a P value less than .05 to detect an OR of 2.5 for the difference in CYP1A1*2B variant frequency betweenFLT3 ITD mutant versus no ITD, and an OR of 2.9 forNRAS mutation versus no NRASmutation.

FLT3 ITD frequency was 17% (77 of 447) andNRAS mutation was 12% (53 of 443) for the entire cytogenetically selected AML cohort. The 2 mutations were found together in only 4 patients, confirming previous observations.18 

No differences in carcinogen-metabolizing enzyme variant allele frequency were found for GSTT1, GSTM1, CYP2D6, CYP2C19, CYP1A1, SULT1A1, or NQO1, either for the presence versus absence of FLT3 ITD (data for CYP1A1variant alleles shown in Table 1) or in relation to start site or duplication size (data not shown).

Patients with NRAS-mutant AML had a higher frequency of the CYP1A1*2B allele (combined Ile/Val plusVal/Val) compared with nonmutant AML cases (z = 2.0; P < .045) (Table2). When the cohort was subdivided by cytogenetic risk group, this association was confined to the poor-risk karyotype group (Table 2), which was in turn statistically significantly different (logistic regression) from the good-risk group alone (z = 2.78; P < .005); the intermediate-risk group alone (z = 2.0;P < .045); and the combined good- plus intermediate-risk groups (z = 3.026; P < .005).

The poor-risk group comprised patients with abnormalities of chromosomes 5 and/or 7 (5q/7q) and also patients with abnormalities of 3q. The frequency of the *2B allele in patients with 5q/7q abnormalities was 4 of 7 for patients with NRASmutation versus 4 of 79 for patients without the mutation. Corresponding frequencies for the 3q group were 2 of 7 for patients with the NRAS mutation versus 0 of 10 for those without.

No difference in variant allele frequency was found for GSTT1, GSTM1, CYP2D6, CYP2C19, SULT1A1, or NQO1 when patient groups with and without the NRAS AML mutation were compared.

The NRAS mutation spectrum for patients with theCYP1A1*2B allele is presented in Table3. Codon 12 mutations were most frequent; 6 of 7 of those mutations were at hot-spot sites, compared with 24 of 45 mutations at codon 12 among patients lacking the *2Ballele (not significant). The other base change at codon 33 is a novelNRAS mutation in a patient with trisomy 8.

We have previously observed that the allele frequency for theCYP1A1*2B variant is no different in AML patients than in controls8 and cannot therefore be considered a predisposition allele for the development of all types of AML. We now show that the CYP1A1*2B allele may indeed predispose patients to develop a subgroup of AML characterized by specific genomic damage, namely, NRAS mutation and poor-risk karyotype.

CYP1A1 is a phase 1 detoxification enzyme, expressed predominantly in extrahepatic tissue and with a wide variety of substrates, including the carcinogenic polycyclic aromatic hydrocarbons (PAHs). Baseline cellular expression of CYP1A1 is usually low, but high inducibility is mediated through the binding of inducers such as TCDD (2,3,7,8-tetrachlorodibenzo-p-dioxin) and PAH to a cellular protein, the aryl hydrocarbon receptor (AhR). AhR is in turn bound to heat-shock protein 90 (Hsp90) and releases this before translocation to the nucleus to associate with another protein, the Ah receptor nuclear translocator (Arnt). This complex binds an enhancer sequence activating transcriptional induction of CYP1A1 mRNA.19 Thus, phase 1 detoxification capacity of CYP1A1 is influenced at 3 levels: baseline expression level, expression induction, and enzymatic activity.

Several polymorphic allelic variants have now been described within theCYP1A1 gene (reviewed in Wormhoudt et al20). The most common variant allele, designated*2A, introduces an Msp1 restriction site in the 3′ end of the gene. The next most common, and potentially functionally more significant, variant is a single nucleotide polymorphism in exon 7 (A→G) producing an Iso→Val substitution at amino acid 462 (allelic variant designated *2B). These alleles are often, but not always, in linkage disequilibrium.20 While the data linking the *2A allele to high inducibility of CYP1A1 in human systems remain inconclusive,21,22 the *2B allele does appear to increase induced gene expression and enzymatic function in lymphocytes from healthy donors with a significant trend through heterozygotes to homozygotes for the variant (Val)allele.22 In the Crofts et al22study, CYP1A1 expression was particularly high among smokers with at least one *2B allele but was not elevated for smokers with at least one *2A allele. Thus, augmented phase 1 metabolism resulting from high inducibility and increased expression of CYP1A1 upon exposure to environmental inducers and substrates will lead to accumulation of toxic metabolic intermediates should phase 2 metabolic pathways be saturated.

The high concentration of PAH in cigarette smoke has led to extensive studies of CYP1A1 allelic variant frequencies in smoking-associated cancers. Although most studies are small, the*2B allele may be overrepresented in smokers developing lung,23 breast,24 and colorectal25 cancers. Several studies provide plausible epidemiologic support for the association between AML and substrates and inducers of CYP1A1. A cohort of subjects exposed accidentally to the potent CYP1A1 inducer TCDD have shown a 3.8-fold increased risk of myeloid leukemia with a latency of 15 years.26 Exposure to cigarette smoke is a weak but reasonably consistent risk factor for AML in large case-control studies.27 28 Smoking history data were unfortunately not available for most subjects in our study.

NRAS mutations in our patients with the CYP1A1*2Ballele show no signature mutation, and this is perhaps not surprising given the diversity of putative environmental carcinogens in the etiology of AML. It is of note, however, that in those patients reported here and in our extended cohort of AML samples (data not shown), most NRAS mutations are transitions, as has previously been reported in AML and other hematologic malignancies.29 Thus, although PAHs are the most attractive candidate carcinogenic substrate of CYP1A1 in the etiology of AML, on the basis of the mutation spectrum in our study, it is likely that the other substrates are involved. This may indeed be the case in other tumors, including lung cancer. Although a signature p53 mutation spectrum is evident in cohorts of smokers with lung cancer,30 a recent study found a higher prevalence of p53 mutation in lung cancers from heavy smokers (compared with nonsmokers) with variant CYP1A1 alleles (*2A/*2B), but the usual signature mutation spectrum was not found in this subgroup.31 This suggests that the mutation signature may be less evident in patients with variantCYP1A1 alleles. Also in this study, smokers with a variantCYP1A1 allele had an increased frequency of KRASmutation but the authors do not comment upon the mutation spectrum. Cigarette smoke contains a multitude of carcinogenic compounds in addition to both TCDD and PAHs. Exposure to combinations of CYP1A1 inducers and substrates may therefore exacerbate the potential for genomic damage on the background of inheritance of the high-inducibility *2B allele. A final alternative hypothesis is that reactive oxygen species known to be generated as a byproduct of CYP1A1-mediated carcinogen metabolism32 33 may oxidize DNA. Oxidative DNA damage may generate a more diverse mutation spectrum, including transitions (eg, 5OH-cytosine G→A) and transversions (8OH-guanine G→T).

It remains unclear why the association between the CYP1A1*2B variant allele and NRAS mutation should be confined to the poor-risk karyotype groups. In the context of therapy-related AML, however, point mutations in critical regulatory genes are more common within the “deletional” karyotypes, namely, p53 mutation associated with chromosome 5 deletion,34 and RAS mutation associated with chromosome 7 deletion.1 These data suggest that genomic instability leads to AML through different pathways,1 and in the context of de novo AML,CYP1A1 variant allele status may represent one determinant of this instability. It is not surprising that FLT3 ITD is not associated with the CYP1A1*2B allele within these cytogenetic subgroups, as the ITD most likely arises from a mechanism different from that of point mutation, perhaps aberrant recombination or defective double-strand DNA repair.

In conclusion, our data suggest that the variant CYP1A1*2Ballele may predispose to the development of a subgroup of AML patients characterized by poor-risk karyotypes and NRAS mutation. Epidemiologic data also suggest that substrates and inducers of CYP1A1 represent candidates for environmental carcinogens with potential to cause AML, but alternative mechanisms such as CYP1A1-induced oxidative stress are also proposed.

We wish to thank Mrs Helen Walker for her assistance in characterizing the DNA bank samples and Dr Simon Ogston for statistical advice. We also wish to thank Professors Alan Burnett (Chairman, MRC Adult Leukemia Working Party) and David Linch for initiating, administering, and providing access to the DNA bank. Finally, we wish to thank David Baty and Dot Mechan for assistance with dHPLC assays, and Professor Roland Wolf for access to the ABI 7700 TaqMan.