The impact of DNMT3A variant allele frequency and two different comutations on patients with de novo cytogenetically normal acute myeloid leukemia

Abstract

To refine the biological and prognostic significance of DNMT3A mutations in acute myeloid leukemia (AML), we assessed the impact of DNMT3A variant allele frequency (VAF) and its comutations in this study. Using targeted next-generation sequencing, we analyzed 171 adult patients with de novo cytogenetically normal AML for DNMT3A mutations and associated comutations. DNMT3A^mut was detected in 35 patients. DNMT3A^mut patients were divided into DNMT3A^High and DNMT3A^Low using a cut-off VAF value of 42%. We observed that DNMT3A^High patients at diagnosis had increasing white blood cell (WBC) counts (p < 0.001) and a higher lactate dehydrogenase (LDH) level (p = 0.027), and were associated with lower complete remission (CR) rate (p = 0.015) and shorter overall survival (OS) (p = 0.032) than DNMT3A^Low patients. We classified two different comutated genetypes, including DNMT3A^mutNPM1^mutFLT3-ITD^mut and DNMT3A^mutIDH1/IDH2^mut. Patients with DNMT3A^mutNPM1^mutFLT3-ITD^mut showed worse OS (p = 0.026) and relapse-free survival (RFS) (p = 0.003) than those with DNMT3A^mutIDH1/IDH2^mut, and showed a shorter OS (p = 0.027) than those with DNMT3A^wtNPM1^mutFLT3-ITD^mut. We also observed that patients with DNMT3A^mutIDH1/IDH2^mut had higher platelet counts (p = 0.009) and a lower BM blast percentage (p = 0.040) than those with DNMT3A^wtIDH1/IDH2^mut. In multivariate analyses, DNMT3A^High was independently associated with a lower CR rate (OR = 5.883; p = 0.004) and shorter OS (HR = 3.768; p < 0.001). DNMT3A^mutNPM1^mutFLT3-ITD^mut independently affected worse OS (HR = 6.030; p < 0.001) and RFS (HR = 8.939; p < 0.001). Our findings might be potentially useful for predicting clinical outcomes.

INTRODUCTION

DNMT3A belongs to a family of DNMTs, including DNMT1, DNMT3A, and DNMT3B, which encodes a DNA methyltransferase that is thought to regulate de novo DNA methylation modification, rather than maintenance methylation.^1,2 By regulating DNA methylation, DNMT3A may regulate the growth of hematopoietic stem cells (HSCs) into a predominantly granulocytic lineage during normal hematopoiesis.^3–5 Studies in mice have demonstrated that deletion of DNMT3A could cause HSC persistent self-renewal and inefficient differentiation.^6–8

Somatic DNMT3A mutations are observed in various types of adult myeloid and lymphoid neoplasms. They are more frequently occurred (23%–37%) in adult patients with cytogenetically normal acute myeloid leukemia (CN-AML),^9–14 but very rarely found in pediatric or adolescent blood cancers.^15–17 DNMT3A mutations are usually heterozygous in AML.^9,18–20 In leukemogenesis, the mutant protein can dimerize with wild-type DNMT3A, but homotetramers with more potent activity cannot be formed. The resulting low level of DNMT3A homotetramer results in significantly reduced methyltransferase activity and genome-wide hypomethylation in patients.^21–24 In contrast to wild-type DNMT3A, more recent research suggests that it could interact with EZH2, which is the catalytic component of Polycomb repressive complex 1, leading to the down-regulation of genes associated with hematopoietic differentiation.²⁵

Although DNMT3A mutations are more common in clonal hematopoiesis and appear to be relatively early events in leukemogenesis, the clinical effect of DNMT3A mutations on CN-AML remains inconclusive. Previous studies vary regarding the impact of DNMT3A mutations. Some studies found a significantly worse overall or event-free survival (EFS).^9,20,26–33 Still, others found no significant association with overall and EFS.^{11,18,34–36} These conflicting results may be due to the genetic heterogeneity of DNMT3A-mutated CN-AML. Hence, it is crucial to further refine the genetic subclassification for a better understanding of the clinical effect of DNMT3A mutations in adult CN-AML.

We analyzed a cohort of 171 adult patients with de novo CN-AML for DNMT3A mutations and associated comutations using targeted next-generation sequencing (NGS). We further investigated the clinical impact of DNMT3A variant allele frequency (VAF) and two different comutated genetypes on these patients, which might help to further refine biological and prognostic implications of DNMT3A mutations in de novo CN-AML.

METHODS Patients group

We conducted a retrospective review of NGS analyses performed on clinical bone marrow samples from adult patients with De novo CN-AML who presented to The Secondary Hospital of Shanxi Medical University Hematology Center in China between February 2017 and January 2021 in this study. We identified 171 newly diagnosed adult patients with de novo CN-AML (90 males and 81 females; median age, 53 years; age range, 19–86 years). The clinical samples were unpaired design. All patients provided informed written consent. This study complied with the Helsinki declaration and was approved by the ethical board of the Second Hospital of Shanxi Medical University. The patient's data including age, sex, hematological parameters, blasts in bone marrow aspirates, and prior history of cytotoxic chemotherapy or radiotherapy were obtained from the medical records at diagnosis. De novo AML defined the patient as having no antecedent myeloid malignancy, cytotoxic therapy or radiotherapy before the diagnosis. A total of 168 patients were subjected to anti-cancer therapies. Among these patients, 125 were treated with high-intensity induction chemotherapy regimens and 43 received low-intensity induction chemotherapy regimens. The dose and course of treatment were performed according to the Chinese Guidelines for the Diagnosis and Treatment of adult acute myeloid leukemia (non-acute promyelocytic leukemia) (2017 edition). Twenty-one patients without concomitant DNMT3A mutations but in the intermediate or adverse-risk group underwent allogous- hematopoietic stem cell transplantation (allo-HSCT) after complete remission (CR). The CR and recurrence were determined according to the ELN2017 recommendations. Overall survival (OS) was calculated as the period from the date of diagnosis to death or to the date of last observation. Relapse-free survival (RFS) was calculated as the period from the first CR to relapse, death or the last observation.

Molecular analysis

A targeted NGS study was used to analyze the DNA of fresh bone marrow samples at initial diagnosis. Blood genomic DNA was isolated by Mini Blood DNA kit (Qiagen or OMEGA) and quantified with the NanoDrop spectrophotometer. We sequenced the mutation hotspots or entire coding regions of 34 genes associated with myeloid leukemia, which contains FLT3, JAK2, KIT, MPL, CALR, CSF3R, PDGFRA, CEBPA, NRAS, KRAS, NPM1, TP53, RUNX1, GATA2, WT1, TET2, DNMT3A, IDH1, IDH2, ASXL1, BOCR, BOCRL1, CBL, ETV6, EZH2, MLL, NOTHCH2, PHF6, SF3B1, SRSF2, SH2B3, SETBP1, U2AF1, ZRSR2. The regions analyzed included mutational hotspots or the coding sequence of 34 genes. In Brief, 50 ng genomic DNA was used for each reaction. DNA samples from all patients were sequenced and analyzed on a high-throughput sequencing platform, the MiSeq next-generation sequencing instrument (Illumina). VAF was observed with a specific DNA sequence variation matching divided by the percentage of the overall coverage of the site. VAF greater than 5% was considered to be the presence of a mutation. VAF cut-off value was to use the optimal cutoff method with the Cutoff Finder web application (http://molpath.charite.de/cutoff). Data of patients with DNMT3A mutations were uploaded from a tab-separated table. The VAF of DNMT3A mutations and OS survival variable were selected from the table. The optimal cut-off point was determined by choosing “ROC curve (Manhattan distance)” method.

Statistical analysis

The clinical features of the patients were described using descriptive statistics. Differences between groups were analyzed by the chi-squared or Fisher exact test for categorical variables and non-parametric Mann–Whitney U test for continuous variables. Survival analysis using the Kaplan–Meier method and the log-rank test, including the OS and RFS. Regarding the multivariate analysis of prognostic factors, a Cox-proportional hazards regression model was used for survival endpoints and a logistic regression model was used for CR. Bilateral p < 0.05 prompt difference was statistically significant. All statistical procedures were performed using SPSS software package version 25.0 and Graphpad Prism™ 8.30.

RESULTS DNMT3A mutations characterization in adult patients with de novo CN-AML

DNMT3A mutations were identified in 35 of 171 (20%) patients with de novo CN-AML. Thirty-two patients carried single mutations in the DNMT3A gene, and three patients harbored two DNMT3A variants (double mutations). R882 missense mutations were more common variants (Figure 1A). Other non-R882 variants were detected in one patient respectively, including W893S,C911Y, E863K,Q842R,T835M, R792C,F755S,A741G,R736C,A571P,R326L,K299Q,P627fs,G511fs,R598X,W581X, A571-575del (Figure 1A).

FIGURE 1. DNMT3A mutations and associated co-mutations in 171 adult patients with de novo CN-AML. (A) Structural diagram showing the location of mutations in DNMT3A. Each patient with DNMT3A mutation is designated with a circle. The color of the circle indicates the different types of mutations. (B) DNMT3A VAFs of all patient in this cohort. Blue bar indicates that the patient has single DNMT3A mutation. Orange bar indicates that the patient has two DNMT3A mutations. (C) Comparison of associated comutations and their frequency between DNMT3Amut and DNMT3Awt patients.

All 35 patients with DNMT3A^mut harbored at least one or more companion comutations, with an average total of three mutations per patient (range:1–6), which was higher than DNMT3A^wt patients with (mean:2; range: 1–6) (p < 0.001). A detailed mutations profiling is provided in Figure 1B,C. Notably, compared to those with DNMT3A^wt, patients with DNMT3A^mut more frequently harbored NPM1 (69% vs. 18%; p < 0.001), IDH1/IDH2 (34% vs. 9%; p < 0.001), FLT3-ITD (26% vs. 12%; p = 0.037), NRAS (29% vs. 8%; p = 0.029) mutations. In a similar fashion, patients with DNMT3A^mut less frequently had CEBPA mutations (6% vs. 32%; p = 0.002). In this cohort, the median VAF value of DNMT3A mutations in 35 patients was 45% (6%–63%). Additionally, we observed DNMT3A mutations were presented as an ancestral mutation with higher or similar VAFs compared with other co-mutations in 35 patients (data not shown). Based on data from comutations and their VAFs, we further identified three comutated patterns with high frequency in 35 patients, including DNMT3A^mutIDH1/IDH2^mut (N = 12, 34%), DNMT3A^mutNPM1^mutFLT3-ITD^mut (N = 9, 26%), and DNMT3A^mutNRAS^mut (N = 8, 23%). Among these three comutations, DNMT3A^mutIDH1/IDH2^mut and DNMT3A^mutNPM1^mutFLT3-ITD^mut existed independently of each other in DNMT3A^mut patients of this cohort, whereas DNMT3A^mutNRAS^mut and DNMT3A^mutIDH1/IDH2^mut had overlaps in two patients, but was mutually exclusive with DNMT3A^mutNPM1^mutFLT3-ITD^mut. Furthermore, we also observed that the ratios of FLT3-ITD were lower than 0.5 in all of nine patients with DNMT3A^mutNPM1^mut FLT3-ITD^mut. In addition, there were no significant differences in the VAFs of DNMT3A mutations between patients with or without NPM1(44.5% vs. 45%; p = 0.320), FLT3-ITD (43% vs. 46.5%; p = 0.503), or IDH1/IDH2 (46.5% vs. 44%; p = 0.469).

Clinical impact of DNMT3A VAF

We used the optimal cut-off method to set the VAF of 42% as the cut-off value, and further divided DNMT3A^mut patients into two subgroups: DNMT3A^High patients (DNMT3A VAF > 42%) and DNMT3A^Low patients (DNMT3A VAF ≤42%). To classify double DNMT3A mutant cases into the DNMT3A^High or DNMT3A^Low group, the higher VAF of the two mutations was used.

Biological features and patient outcomes were compared between these two subgroups (Table 1). To exclude the clinical effects of DNMT3A mutation types and comutations, here we evaluated these mutational features between the both subgroups. Our data showed that the number of patients with R882 or non-R882, comutated genes in the DNMT3A^High subgroup were similarly distributed as those in DNMT3A^Low subgroup. For biological features at diagnosis, patients with DNMT3A^High had significantly increased white blood cell (WBC) counts (median:50.915 vs. 2.33 × 10⁹/L; p < 0.001) and a higher lactate dehydrogenase (LDH) level (median:582.85 vs. 333.5 U/mL; p = 0.027) than those with DNMT3A^Low. Other biological features (including age, sex, Hb levels, platelet counts, BM blast percentages) were not significantly different between DNMT3A^High and DNMT3A^Low subgroups. In this cohort, 32 patients further received therapy, whereas three patients did not receive any therapy. DNMT3A^High subgroup had a greater number of patients receiving high-intensity induction than DNMT3A^Low subgroup, whereas most of patients with DNMT3A^Lowreceiving low-intensity induction due to associated hypoproliferative AML. In addition, none of patients received allo-HSCT in the entire DNMT3A^mut group. Among patients receiving induction therapy, patients with DNMT3A^High showed a significantly lower CR rate (38% vs. 82%; p = 0.015). In terms of survival, patients with DNMT3A^High had a shorter OS (median: 3 vs. 12 months; p = 0.032) than those with DNMT3A^Low (Figure 2 A1), but no statistical difference on RFS (median: 6 vs. 12 months; p = 0.642) was detected between them (Figure 2 B1). In multivariable analyses, DNMT3A^High had independent effects on worse OS (hazard ratio [HR] = 3.768, 95% CI, 1.957–7.255; p < 0.001) (Figure 2 C1), and lower CR rate (odds ratio [OR] = 5.883, 95% CI, 1.733–19.970; p = 0.004) (Figure 2 C3).

TABLE 1 Comparison of clinical impact between DNMT3A^High and DNMT3A^Low.

[Table omitted. See PDF]

Note: The bold values indicate p value < 0.05.

FIGURE 2. Prognosis effects of DNMT3Amut VAF and two comutations on adult patients with de novo CN-AML. Kaplan–Meier survival curves for OS and RFS in DNMT3AHigh versus DNMT3ALow (A-1,B-1); as well as in DNMT3AmutNPM1mutFLT3- ITDmut and DNMT3AmutIDH1/IDH2mut (A-2,B-2), in DNMT3Amut NPM1mutFLT3- ITDmut and DNMT3AwtNPM1mutFLT3-ITDmut (A-3,B-3), in DNMT3Amut IDH1/IDH2mut and DNMT3AwtIDH1/IDH2mut (A-4,B-4). Cox-proportional hazard regression models were analyzed for OS(C-1) and RFS(C-2). Logistic regression models were analyzed for CR(C-3).

Clinical impact of two different DNMT3A^mut comutations

To address the clinical impact of DNMT3A mutations in more detail and because of the high prevalence and mutual exclusion of two comutations of DNMT3A^mut, we next analyzed clinical features and outcomes of these two comutations, including DNMT3A^mutNPM1^mutFLT3-ITD^mut and DNMT3A^mutIDH1/IDH2^mut.

We first compared two subgroups of patients with these two different comutated-genetypes each other (Table 2). Similarly, we also compared DNMT3A^mut VAF and type (R882 and non-R882) between both subgroups, and no significant statistical differences were observed. Our data further showed that these two subgroups were similar for biological features at diagnosis, including age, sex, Hb levels, WBC counts, platelet counts, BM blast percentages, and LDH levels. Regarding treatment types, there were no differences for the number of patients who received different induction therapies between the two subgroups. We observed that patients with DNMT3A^mutNPM1^mutFLT3-ITD^mut were associated with poorer OS (median:7 vs. 15 months; p = 0.026) (Figure 2 A2) and RFS (median: 5.5 vs. 15 months; p = 0.003) (Figure 2 B2) compared to those with DNMT3A^mutIDH1/IDH2^mut, though there was no significant difference on CR rate between these two subgroups (Table 2).

TABLE 2 Comparison of clinical impact between the two DNMT3A^mut comutaions.

[Table omitted. See PDF]

To investigate the interaction impact of DNMT3A mutations and comutations, we also compared biological and clinical outcomes of patients harboring NPM1^mutFLT3- ITD^mut (FLT3-ITD ratio <0.5) genetype when with or without DNMT3A mutations, and patients harboring IDH1/IDH2^mut when with or without DNMT3A mutations (Table 3). When comparing DNMT3A^mutNPM1^mutFLT3-ITD^mut and DNMT3A^wtNPM1^mut FLT3-ITD^mut subgroups, most of the features were similar, including age, sex, Hb levels, WBC counts, platelet counts, BM blast percentages, and LDH levels, treatment, CR rate, RFS, only for OS, patients with DNMT3A^mutNPM1^mutFLT3-ITD^mut revealed a shorter OS (median: 7 vs. 11 months; p = 0.027) than those with DNMT3A^wtNPM1^mut FLT3-ITD^mut (Figure 2 A3). When comparing DNMT3A^mutIDH1/IDH2^mut and DNMT3A^wtIDH1/IDH2^mut subgroups, patients with DNMT3A^mutIDH1/IDH2^mut presented with higher platelet counts (median: 85.0 vs. 37.0 × 10⁹/L; p = 0.009) and a lower BM blast percentages (median: 42.9 vs. 81.5%; p = 0.040) than those with DNMT3A^wtIDH1/IDH2^mut, but no significant differences were detected for other biological features and their impacts on clinical outcomes. In multivariable analyses, DNMT3A^mutNPM1^mutFLT3-ITD^mut independently affected worse OS with a HR for the risk of death of 6.030 (95% CI, 2.656–13.688; p < 0.001) (Figure 2 C1), and worse RFS with an HR of 8.939 (95% CI, 2.952–27.069; p < 0.001) (Figure 2 C2). In contrast, DNMT3A^mutIDH1/IDH2^mut was not an independent factor for impacting CR rates, OS, and RFS (Figure 2 C1–3).

TABLE 3 Comparison of clinical impact between two comutaions with or without DNMT3A^mut.

[Table omitted. See PDF]

Note: The bold values indicate p value < 0.05.

DISCUSSION

In this study, we further assessed the genetic characteristics of DNMT3A mutations in adult patients with de novo CN-AML using targeted NGS with a panel of 34 genes associated with myeloid leukemia. In accordance with previous studies,^{9,10,18,19,26–36} we detected DNMT3A mutations with a frequency of 20% in adult patients with primary CN-AML, and most of the mutations clustered at the R882 site in exon 23. Moreover, all DNMT3A^mut patients of this cohort harbored one or more additional mutations, of which the majority of patients had a relatively higher or similar VAF compared with other comutated genes, strengthening previous data that reported DNMT3A^mut were presented as ancestral clone or preleukemia clone in AML.^12,37 Previous findings reported that DNMT3A mutations had a significant association with NPM1 and IDH1/IDH2 mutations, of which ~60% of DNMT3A^mut cases having NPM1, and more often (~30%) displaying a significant co-mutation pattern with NPM1 and FLT3-ITD mutations, and had a mutually exclusive relationship with CEBPA mutations in AML patients.^18–20 Similarly, our data still showed that DNMT3A mutations had a significant association with NPM1, FLT3-ITD, and IDH1/IDH2 mutations, and but an inverse correlation with CEBPA mutations in CN-AML patients. In addition, we found that DNMT3A mutations had a positive association with NRAS mutations in our patients. Furthermore, by an analysis of comutations, we identified two critical comutated genetypes with a high frequency, including DNMT3A^mutIDH1/IDH2^mut and DNMT3A^mutNPM1^mutFLT3-ITD^mut. Although DNMT3A^mutIDH1/IDH2^mut and DNMT3A^mutNPM1^mutFLT3-ITD^mut comutations had been reported in other studies, their mutation features were rarely described. Here we found that DNMT3A^mutIDH1/IDH2^mut and DNMT3A^mutNPM1^mutFLT3-ITD^mut were mutually exclusive in DNMT3A^mut patients, and that the mutation ratios of FLT3-ITD were lower than 0.5 in all of the patients with DNMT3A^mutNPM1^mutFLT3-ITD^mut. The above data strongly indicated that DNMT3A^mutNPM1^mutFLT3-ITD^mut and DNMT3A^mutIDH1/IDH2^mut might be two different genetic subgroups of DNMT3A^mut patients with CN-AML.

We first investigated the biological and clinical impact of DNMT3A^mut VAF on CN-AML, which be stratified by the cut-off value of 42%. For biological features, we observed that DNMT3A^High patients at diagnosis had a significantly higher WBC counts and a trend for higher BM blast percentage compared with DNMT3A^Low ones, further strengthing a previous report by Narayanan et al.³⁸ that that DNMT3A^High (≥44%) patients presented with leukocytosis and higher blast counts, and they further demonstrated that DNMT3A VAF had a positive correlations with WBC counts in AML patients. Combining these two data sets suggested that higher WBC counts and BM blast percentage might be unique biological features for patients with CN-AML with DNMT3A^High. In addition, we also observed another striking feature with more elevated serum LDH levels in DNMT3A^High cases in our cohort. As far as we know, this is the first description of such an association in DNMT3A^mut AML patients. With regard to clinical outcomes, in univariable and multivariable analyses, we found that DNMT3A^High conferred an unfavorable effect on CR rate and OS, but did not show its negative impact on RFS in these patients. Although the predictive results of DNMT3A VAF are rarely reported, our findings were similar to previous reports. An analysis of 104 patients with DNMT3A-mutated AML led by Narayanan et al.³⁸ showed that high DNMT3A VAF was associated with more inferior OS and EFS, but had no impact on CR rate in univariable analyses, but in multivariable analyses, the adverse effect of DNMT3A^High only on OS but not on EFS in the CN-AML subset. Linch et al.³⁹ reported that high DNMT3A R882^mut VAF (≥47%) also presented worse effects on CR rate and OS in univariable analyses, but were not significant in multivariable analyses. Yuan et al.⁴⁰ reported that higher DNMT3A R882^mut VAF (≥39%) had a shorter OS than those with a lower DNMT3A R882^mut VAF.

Based on the comutations features of DNMT3A^mut, we further investigated the clinical effects of two comutations, including DNMT3A^mutNPM1^mutFLT3-ITD^mut and DNMT3A^mutIDH1/IDH2^mut, which were considered as the two genetic subgroups of DNMT3A^mut patients with de novo CN-AML in this study. We observed that patients with DNMT3A^mutNPM1^mutFLT3-ITD^mut had significantly worse OS and RFS than those with DNMT3A^mutIDH1/IDH2^mut. To better understand the effect of DNMT3A mutations in CN-AML, we compared the clinical impact of NPM1^mutFLT3- ITD^mut and IDH1/IDH2^mut when with or without the genetic context of DNMT3A mutations. Recently, the 2022 ELN guideline has updated FLT3-ITD as an intermediate risk marker irrespective of NPM1 mutational status.⁴¹Our data showed that DNMT3A mutations had a significantly poorer effect on OS on patients with NPM1^mutFLT3-ITD^mut (FLT3-ITD-ratio <0.5) genetype, further suggesting that DNMT3A mutations could reduce favorable prognosis effect of NPM1^mutFLT3-ITD^mut (FLT3-ITD-ratio <0.5) genetype. In contrast, we did not observe significant differences on clinical outcomes between DNMT3A^mutIDH1/IDH2^mut and DNMT3A^wtIDH1/IDH2^mut subgroups, indicating that DNMT3A mutations could not change the clinical prognosis of IDH1/IDH2 mutations. In multivariable analyses, the DNMT3A^mutNPM1^mutFLT3-ITD^mut (FLT3-ITD-ratio <0.5) was still independently associated with worse OS and RFS in this cohort. The above data strongly indicated that DNMT3A mutations could generate different prognostic effects when combined with different comutations. In addition, although DNMT3A^mutIDH1/IDH2^mut had no significant prognosis impact, we found that patients with this mutated genetype presented biological features such as higher platelet counts and a lower BM blast percentage in comparison to those with DNMT3A^wtIDH1/IDH2^mut. To the best of our knowledge, the clinical impact of the DNMT3A^mutIDH1/IDH2^mut was barely reported, whereas the clinical implications of the DNMT3A^mutNPM1^mutFLT3-ITD^mut were described in only a few reports. A report led by Loghavi et al.⁴² showed that DNMT3A^mutNPM1^mutFLT3-ITD^mut had a shorter effect on EFS and a trend for shorter OS among AML old patients in comparison to those within other mutation subgroups. Another recent study cohort conducted by Bezerra et al.⁴³ reported that the DNMT3A^mutNPM1^mutFLT3-ITD^mut had worse effects on OS and DFS, which was similar to our findings.

CONCLUSIONS

In summary, this study more detailly refined the biological and clinical prognostic effects of DNMT3A^mut in adult patients with de novo CN-AML. Our findings highlighted that DNMT3A^mut VAF and its two comutations had their specific clinical consequences. We found that high DNMT3A VAF was associated with higher WBC counts and BM blast percentage than low DNMT3A VAF, and had an independent effect on lower CR rate and shorter OS. We also identified that DNMT3A^mutNPM1^mutFLT3-ITD^mut exerted an independent worse impact on OS and RFS. In contrast, the patients with DNMT3A^mutIDH1/IDH2^mut had relatively favorable prognoses, but manifested as higher platelet counts and a lower BM blast percentage in CN-AML patients than those with DNMT3A^wtIDH1/IDH2^mut. However, there were several limitations to the current study because of a small sample size and its retrospective nature. Our findings need to be validated in a more extensive and prospective cohort, which might be the potentially helpful in predicting clinical outcomes.

AUTHOR CONTRIBUTIONS

Xian Chen: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); writing – original draft (equal). Chuchu Tian: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal). Zhuanghui Hao: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); writing – original draft (equal). Lingang Pan: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); writing – original draft (equal). Minglin Hong: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal). Wei Wei: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal). Daniel Muteb Muyey: Data curation (equal); formal analysis (equal); investigation (equal); methodology (equal). Hongwei Wang: Conceptualization (equal); funding acquisition (equal); project administration (equal); writing – review and editing (equal). Xiuhua Chen: Conceptualization (equal); data curation (equal); formal analysis (equal); funding acquisition (equal); investigation (equal); methodology (equal); project administration (equal); writing – review and editing (equal).

ACKNOWLEDGMENTS

We would like to thank Chunliang Liu (The Second Hospital of Shanxi Medical University) who conducted the statistical analyses in this study.

FUNDING INFORMATION

This work was supported by the National Natural Science Foundation of China (nos 81500104; 81670126), The Shanxi Natural Science Foundation of China (nos 201801D221409; 201801D111003), Graduate Innovation Fund of Shanxi Province.

CONFLICT OF INTEREST STATEMENT

All the authors declare they have no competing interests.