Acute myeloid leukemia (AML) is a malignant neoplasm of hematopoietic cells characterized by diverse clonal genetic aberrations, ranging from gross chromosomal abnormalities to DNA mutations. The etiological significance of DNA lesions in the pathogenesis of AML is further upheld by epidemiological observations that AML, but not acute lymphoid leukemia, is clustered among individuals with congenital defects involving DNA repair genes (e.g., in Fanconi anemia) or those with prior history of genotoxic chemotherapy and/or irradiation.1, 2 Certain common genetic variations of DNA repair genes confer a subtle alteration of the constitutive capacity of the host cells in dealing with DNA lesions and lead to an alteration of susceptibility to specific malignancies or sensitivity to chemo- and radiotherapy. It is of great interest whether certain genetic variations influence the predisposition to and/or treatment response of AML individually or interactively. Several previous studies have revealed associations between AML and genetic variations of certain DNA repair genes.3–6 However, most of these studies focused on genes encoding enzymes involved in drug metabolism such as NQO1/2, MTHFR or GSTT1, and studies were performed separately within patients of different ethnic origins.7 In this study, we performed a systematic screening of 42 nonsynonymous variations in 15 genes functioning in DNA repair for association with AML and therapy response in a cohort of 307 Chinese patients.
Common genetic variations in genes involved in DNA repair or response to genotoxic stress may influence both cancer susceptibility and treatment response individually or interactively. However, in acute myeloid leukemia (AML), the relevance of these genetic variations remains to be fully established. In this study, we analyzed 42 genetic variations among 15 candidate genes in 307 AML patients and 560 age-sex matched controls. Their associations with chemotherapy response were further evaluated in combination with other well-established prognostic factors. An increased risk of AML was found in individuals heterozygous for XPD 2251A>C (rs13181) with an odds ratio (OR) of 1.637 (95% confidence interval [CI]: 1.118–2.395), and the increased risk could be attributed to C allele (OR = 1.505, 95% CI: 1.061–2.134). Postchemotherapy response analysis revealed that AML patients heterozygous for ATM 4138C>T (rs3092856) or GG homozygous for TP53 215C>G (rs1042522) were independently linked to inferior treatment outcomes. These results uncover novel prognostic factors for AML patients treated with chemotherapy and may also indicate an etiological role of XPD in this disease.
Material and Methods
Cases and controls
From March 2001 to March 2007, a total of 307 consecutive patients with newly diagnosed AML were enrolled in this study. The clinical features of these patients are summarized in Table 1. All non-AML-M3 patients received cytosine arabinoside plus anthracyclines-based regimens as described previously.8 Evaluation of treatment response was according to the revised recommendations made by International Working Group for therapeutic trials in AML.9 Classification of cytogenetic risk groups was according to the recent guideline made by the Italian Society of Hematology.10–13 The definition of refractory AML made by Estey et al. was adopted in our study of treatment response.14 Informed-consent documents were obtained from all patients before samples of bone marrow cells were collected for genotyping. Peripheral blood samples obtained from 560 healthy donors were set as the reference population.
Genomic DNA of both case and control population was extracted using Wizard Genomic DNA Purification Kit (Promega, Madison, Wisconsin, USA). A chip-based matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis system (iPLEX™, Sequenom, San Diego, CA, USA) was used to perform high-throughput genotyping assays. The experimental procedures have previously been described.15FLT3-ITD and NPM1 mutations in AML patients were detected by PCR amplification followed by products sequencing.
The frequency distribution at each polymorphic site among control population was tested for their deviation from Hardy-Weinberg Equilibrium using chi-square Goodness-of-Fit test at one degree of freedom. In analyses of all the dichotomous or trichotomous variables of genotype data within AML patients and healthy controls, binary logistic regression was adopted. Odds ratio (OR) was calculated with samples of the most prevalent genotypes set as reference and expressed with 95% confidence interval (CI). Subjects were further stratified according to age, sex and de novo/secondary AML for association with individual genotypes. Binary or multinomial logistic regression was used in the analysis of AML clinical features among patients with different genotypes. For survival analyses of patients post chemotherapies, Kaplan-Meier method was adopted and tested using log-rank algorithm. Cox regression was used to exert multivariate survival analysis. All the statistic analyses were performed using SPSS software package Ver. 13 (SPSS Inc., Chicago, IL, USA).
For this study, we analyzed 42 genetic variations in 15 genes encoding proteins functioning in DNA nucleotide excision repair (ERCC1, XPA, XPD and XPG), base excision repair (APE1, MGMT, PARP1 and XRCC1), nonhomologous end joining (ATM, NBS1, RAD50 and XRCC5), cell-cycle checkpoint regulation (MDM2, TP53) or metabolism of genotoxic materials (GSTP1). Individual sites were selected for this study based on single nucleotide polymorphisms (SNP) entries in the NCBI dbSNP database or reported genetic variations (Supporting Information Table 1).16–20 Twenty one of these sites were subsequently excluded from further analyses due to an absence of polymorphisms within our cohorts. The genotype frequencies of the remaining 21 sites in patients with AML and the control population are listed in Supporting Information Table 2. No significant deviation from Hardy-Weinberg equilibrium was detected among these polymorphic genotypes in healthy individuals. Polymorphisms at MGMT 427A>G (Ile143Val) and 533A>G (Lys178Arg) were identical in both patient and control cohort, indicating a complete linkage between these two polymorphic sites.
Logistic regression analyses revealed an increased risk of AML in XPD 2251AC individuals. The proportion of XPD 2251AC heterozygotes in AML patients (19.1%, 58/303) was significantly higher than that in control population (12.6%, 70/554), with an OR value of 1.637 (95% CI: 1.118–2.395, p = 0.011) (Supporting Information Table 2). The frequency of the XPD 2251C allele among AML patients was higher than that in healthy controls (10.23–7.04%, OR = 1.505, 95% CI: 1.061–2.134, p = 0.021 by Pearson's chi-Squared 2-side test shown in Supporting Information Table 3), further supporting the notion that this variation is linked to a predisposition to develop AML. Possible associations between genotypes of the 21 SNPs and clinical features of AML patients were analyzed. No significant difference was found at these polymorphic sites between patients with de novo and those with secondary AML (Supporting Information Table 4). We did not find any correlation between genetic polymorphisms of these genes and the distribution of patients among different cytogenetic risk groups (Supporting Information Table 5), neither the acquisition of common chromosomal abnormalities including t(8;21), −5/5q−, −7/7q−, +8, t(9;22) or −X/−Y.
To determine whether the genetic variations of the genes functioning in DNA repair or response to genotoxic stress in our study were associated with treatment outcome in AML patients, we compared the response to chemotherapy of patients with different genotypes. Those, who received treatment other than chemotherapy including hematopoietic stem cell transplantation or monoclonal antibody therapy, were removed from response analysis. In total, 159 patients with de novo AML (Table 1) were eligible for treatment response study. Patients were defined to have refractory diseases according to the criteria proposed by Estay et al. Those patients, whose disease could not reach complete remission after 2 cycles of induction therapy (including death during induction therapy), and patients who experienced disease relapse within 12 months post their initial remission, were defined as patients with primary refractory AML. We found that patients with TP53 215GG variant were more likely to have refractory diseases to cytosine arabinoside plus anthracyclines-based chemotherapy (OR: 2.909 for GG homozygotes vs. CC+CG carriers, 95% CI: 1.350–6.268, p = 0.006, data shown in Supporting Information Table 6).
Possible genotype-specific influences on long-term outcome of patients post chemotherapy were further analyzed. Patients with TP53 215CC+CG variants (108/159) showed a significant survival advantage over those TP53 215GG homozygotes (51/159). The estimated 2-year overall survival of patients with CC/CG variants was 45.7%, in comparison with 16.5% for patients with TP53 215GG genotype (p = 0.003, Fig. 1a). Interestingly, survival analysis also revealed that AML patients of ATM 4138CT heterozygotes (8/159) had unanimously shorter overall survival than those of ATM 4138CC homozygotes (151/159), with an estimated median survival of 7 months in heterozygotes comparing with 11 months in homozygotes (p = 0.044, Fig. 1b) and a estimated 2-year survival of 38.1 and 0% for patients with CC and CT genotype, respectively. As there were not significant discrepancies of mean ages, WBC counts or cytogenetic subclassification among patients with different genotypes of ATM 4138C>T and TP53 215C>G (Supporting Information Table 7), it was unlikely that the survival discrepancies were actually resulted from uneven distributions of these clinical factors.
The effects of these genetic variations on patients' survival were examined in combination with other well-established prognostic factors (i.e., age, WBC count, cytogenetic risk groups, NPM1 and FLT3-ITD mutations). As illustrated in Table 2, univariate analysis proved that overall survival of AML patients was significantly associated with hyperleukocytosis and unfavorable cytogenetic abnormalities at diagnosis, and marginally influenced by FLT3-ITD mutation status as well. Nevertheless, multivariate Cox regression analysis revealed that only variation at TP53 215C>G (with a hazard ratio value of 1.807, 95% CI: 1.101–2.966, p = 0.019) and cytogenetic risk groups (with a hazard ratio value of 1.679, 95% CI: 1.121–2.514, p = 0.012) were independently associated with overall survival of patients (Table 2).
Because of their distinctive cellular biological features, hematopoietic cells, especially myelopoietic cells, show an extraordinary vulnerability to malignant transformation induced by genotoxic chemicals or irradiative exposure. This scenario warrants this study on the association between genetic variations of the main DNA repair genes and susceptibility to AML.
XPD (xeroderma pigmentosum group D, also known as ERCC2) encodes a DNA helicase involved in nucleotide excision repair. Because of the biological significance of XPD, the XPD 2251A>C (Lys751Gln) polymorphism has been a common subject of studies in different malignant diseases in recent years. Although the XPD 2251C variant was associated with an increased risk of esophageal cancer and acute lymphoid leukemia,21 its significance in AML was not established. Allan et al. investigated XPD 2251A>C polymorphism in 341 adult British AML patients and observed an excessive risk of disease for individuals with 2251CC genotype in both de novo and secondary AML, but only in the cohort of secondary AML the difference reached statistical significance.22 A similar trend existed in another study composed of pediatric AML patients in USA.23 Of note, the allele frequency of XPD 2251C was 7.0% in our Chinese control cohort, comparing with 35.5 and 36.4% in the healthy British/American controls reported in these two studies.22, 23 Although the low allele frequency of XPD 2251C in Chinese population hindered analysis on CC homozygotes separately in our study, we found XPD 2251AC variant were associated with increased risk of AML, upholding the notion that C allele in XPD 2251 conveyed an increase risk of AML. The considerable variety of the XPD 2251C allele prevalence in different ethnic groups was also noticed by a recent meta-analysis study.21 We believe the underlying discrepancies in the prevalence of the XPD 2251C variant may contribute greatly to the inconsistent results of those association studies based on different ethnic populations. In addition, the functional significance of the Lys751Gln amino acid substitution on the capacity of XPD in nucleotide excision repair has not been established and merits further investigation.
Chemotherapy holds an irreplaceable role in AML, even in settings where molecular targeting therapy (such as retinoids treatment in AML with PML/RARα fusion gene) or allogeneic hematopoietic stem cell transplantation plays an active role for treatment. AML is a heterogeneous disease entity. Chemotherapy confers a high remission rate to 70% patients of certain subtypes, while relapsed or resistant disease remains in more than 80% of patients with specific genetic aberrations even after autologous stem cell supported high-dose chemotherapies.13 Although much of the heterogeneity of AML is still unresolved, approaches aiming to clarify these issues are of unequivocal importance. In patients with AML, the innate genetic and molecular aberrations of leukemic cells have been proved the most relevant factors for prognosis evaluation among common clinical parameters including lineage specificities, maturation stages or even burdens of leukemic cells. So, it is of extraordinary interest whether certain genetic variants of genes involved in DNA repair could affect treatment outcomes in patients with AML receiving chemotherapies. In this study, we discovered for the first time the significance of polymorphisms at TP53 215C>G and ATM 4138C>T in chemotherapy-treated AML patients. The inferior treatment response conveyed by these genetic variations could be proved in short-term response rate and long-term survival analysis, indicating the intrinsic refractoriness to chemotherapies of leukemic cells accounts for the worse treatment outcomes.
The significance of the TP53 215C>G polymorphism, which results in a proline to arginine substitution in the proline rich domain of p53 protein, had been enthusiastically investigated in both biological researches and disease-association studies as its discovery, nonetheless, with inconsistent results.24–28 The heterogeneity of the TP53 deletion and mutation status in different tumors renders additional complexity to the analysis of these results. In contrast, the relatively low frequency of TP53 mutation in AML (about 5% in general) may facilitate functional studies of its genetic variants. In fact, a recent study highlights p53 as a central determinant for myeloid leukemic cells in response to chemotherapy, and the expression and functional status of p53 in AML is differentially modulated by specific cytogenetic abnormalities.29 In this study, we found that TP53 215G variant was correlated with the primary refractoriness to chemotherapy of patients with de novo AML. Long-term survival of AML patients definitely favored patients other than 215GG homozygotes. The hypothesis that p53 Arg72 and Pro72 variants convey differential survival advantage for AML cells merits further experimental validation.
The other SNP site of prognostic significance in our research is ATM 4138C>T (His1380Tyr). The ATM 4138CT genotype has relatively low penetrance in Europeans (0–2% in dbSNP database). This precludes meaningful conclusions regarding tumor susceptibility or treatment response analysis in the early studies, as the majority of disease-associated studies concerning this SNP were performed in European cohorts.30–32 In our study, 4.3% (13/305) of AML patients carried the ATM 4138CT variant. Although we have identified more heterozygote cases than the previous studies outlined above, the disease-associated results could not be interpreted without caveats due to the absolute scarcity of items. However, it is noteworthy that 7/8 chemotherapy treated cases with ATM 4138CT variant presented primary refractory disease, and ATM 4138CT patients generally had inferior overall survival comparing with ATM 4138CC patients. It was proved that a DPAPNPPHFP motif (residues 1373–1382) within the ATM protein is essential for binding and subsequent phosphorylation of c-ABL by ATM, and the ATM Tyr1380 variant was deficient in binding and activation of c-ABL.20, 33 Blocking of the interaction between c-ABL and ATM could interfere with pro-apoptotic signaling in response to DNA double strand break damage, and the contribution of individual ATM variants for the resistance of AML cells to genotoxic agents warrants further investigation.
Even though concurrent therapeutic modalities for patients with AML have evolved considerably comparing with the situations half a century ago, chemotherapy is still the mainstay of treatment options. Because of the heterogeneity of this disease entity, accurate and clinically feasible prognostic systems are imperatively wanted to improve long-term outcomes of patients and avoid unnecessary toxicity of therapies. In this report, we demonstrated the prognostic value of ATM 4138C>T and TP53 215C>G in patients with de novo AML treated with chemotherapy. As the distributions of these genetic variants are independent of other clinical prognostic factors, we believe they could provide additional prognostic information for patients with de novo AML. We also envision future prospective investigations or large-scale researches further validate the significance of these genetic polymorphisms in patients of different ethnic origins, specific pathological characteristics or even patients receiving therapies other than chemotherapies, such as immune mediated therapies with specific antibodies or stem cell transplantations.
The authors would like to thank Dr. Kevin Petrie, Institute of Cancer Research, Surrey, UK and Dr. Guang-Biao Zhou, Institute of Zoology, Chinese Academy of Science, Beijing, China for critical reviews of the manuscript.