Age-Related Prognostic Impact of Different Types of DNMT3A Mutations in Adults With Primary Cytogenetically Normal Acute Myeloid Leukemia

Abstract

Purpose

To determine the frequency of DNMT3A mutations, their associations with clinical and molecular characteristics and outcome, and the associated gene- and microRNA-expression signatures in primary cytogenetically normal acute myeloid leukemia (CN-AML).

Patients and Methods

Four hundred fifteen previously untreated adults were analyzed for DNMT3A mutations and established prognostic gene mutations and expression markers. Gene- and microRNA-expression profiles were derived using microarrays.

Results

Younger (< 60 years; n = 181) and older (≥ 60 years; n = 234) patients had similar frequencies of DNMT3A mutations (35.3% v 33.3%). Missense mutations affecting arginine codon 882 (R882-DNMT3A) were more common (n = 92; 62%) than those affecting other codons (non–R882-DNMT3A). DNMT3A-mutated patients did not differ regarding complete remission rate, but had shorter disease-free survival (DFS; P = .03) and, by trend, overall survival (OS; P = .07) than DNMT3A–wild-type patients. In multivariable analyses, DNMT3A mutations remained associated with shorter DFS (P = .01), but not with shorter OS. When analyzed separately, the two DNMT3A mutation types had different significance by age group. Younger patients with non–R882-DNMT3A mutations had shorter DFS (P = .002) and OS (P = .02), whereas older patients with R882-DNMT3A mutations had shorter DFS (P = .005) and OS (P = .002) after adjustment for other clinical and molecular prognosticators. Gene- and microRNA-expression signatures did not accurately predict DNMT3A mutational status.

Conclusion

DNMT3A mutations are frequent in CN-AML, and their clinical significance seems to be age dependent. DNMT3A-R882 mutations are associated with adverse prognosis in older patients, and non–R882-DNMT3A mutations are associated with adverse prognosis in younger patients. Low accuracy of gene- and microRNA-expression signatures in predicting DNMT3A mutation status suggested that the role of these mutations in AML remains to be elucidated.

INTRODUCTION

Acute myeloid leukemia (AML) is a genetically heterogeneous disease characterized by nonrandom cytogenetic aberrations^1,2 and, at the submicroscopic level, recurrent gene mutations and changes in gene expression.³ Cytogenetic and molecular alterations not only define distinct biologic entities, but are also relevant for disease classification and treatment guidance.⁴

Cytogenetically normal (CN) AML, comprising 45% to 50% of adults with primary disease,⁵ is one of the best molecularly characterized cytogenetic groups. Some gene mutations recurrent in CN-AML are strong, independent prognosticators (eg, FLT3 internal tandem duplications [FLT3-ITD],^6,7 CEBPA,^8,9 and WT1^10,11 mutations), whereas others affect outcome in distinct molecular or clinical subsets of CN-AML (eg, NPM1,^12,13TET2,¹⁴ and IDH1/IDH2^15,16 mutations) or are of uncertain significance (eg, FLT3-tyrosine kinase domain mutations [FLT3-TKD]^17,18). More intense treatment may modify the prognostic weight of some molecular markers in CN-AML, such as MLL partial tandem duplication (MLL-PTD)^19,20 or FLT3-ITD.²¹ Additionally, altered expression of genes (eg, high BAALC,^22,23ERG,^23,24 and MN1^25–27 levels) and microRNAs (eg, low miR-181a level²⁸) identify high-risk CN-AML patients.

The DNMT3A gene encodes one of the three DNA methyltransferase (DNMT) isoforms. Among these, DNMT1 is the most abundant and preferentially replicates existing DNA methylation patterns, whereas DNMT3A and DNMT3B are responsible for establishing de novo DNA methylation. The process of DNA methylation consists of an enzymatic addition of a methyl group at the carbon 5 position of cytosine in the context of cytosine-guanine dinucleotides. When occurring in the promoter region of a coding gene, it generally results in gene silencing. In AML, all three DNMT enzymes are reportedly overexpressed in malignant blasts compared with normal bone marrow (BM) cells and contribute to leukemogenesis by mediating tumor suppressor gene silencing.²⁹ Somatic DNMT3A mutations in AML were first described by Yamashita et al³⁰ and subsequently by other groups.^31,32 Ley et al³¹ first reported that DNMT3A mutations conferred worse outcome in AML. However, the patients analyzed were heterogeneous for biologic and clinical characteristics and treatment received, and the prognostic value of DNMT3A mutations was not fully evaluated within the context of other known molecular prognosticators.³¹ Recently, Thol et al³³ reported that DNMT3A mutations are associated with shorter overall survival (OS) in cytogenetically diverse patients with AML who are younger than 60 years and with lower complete remission (CR) rates and shorter OS in a CN-AML subset. However, this study included patients with secondary disease and those who received allogeneic stem-cell transplantation (SCT) and did not analyze the prognostic impact of different types of DNMT3A mutations.³³

To our knowledge, our study is the first to investigate the prognostic impact of DNMT3A mutations in a large population of patients diagnosed exclusively with primary CN-AML, comprehensively characterized for other molecular prognosticators, and receiving intensive chemotherapy. Additionally, we analyzed the differential impact of DNMT3A mutations by age group (younger [< 60 years] v older [≥ 60 years]) and mutation type (missense mutations at codon R882 [hereafter called R882-DNMT3A] v mutations at other locations [denoted non–R882-DNMT3A]). Furthermore, to gain insights into the biologic role of DNMT3A mutations in CN-AML, we derived genome-wide DNMT3A mutation-associated gene- and microRNA-expression signatures.

PATIENTS AND METHODS

Patients, Treatment, and Cytogenetic Studies

Pretreatment BM or blood samples were obtained from 415 patients with primary CN-AML, 18 to 83 years of age (181 younger and 234 older), who received intensive first-line therapy on Cancer and Leukemia Group B trials.^34–42 Patients received cytarabine-daunorubicin-based induction chemotherapy; most younger patients received consolidation with high-dose chemotherapy and autologous SCT. Per protocol, no patient received allogeneic SCT during first CR. For details regarding treatment protocols and sample collection, see the Data Supplement. The diagnosis of normal cytogenetics was based on centrally reviewed analysis of ≥ 20 metaphases in BM specimens.⁴³ All patients provided written informed consent; study protocols were in accordance with the Declaration of Helsinki and approved by local institutional review boards.

Mutational Analyses

For DNMT3A mutational analysis, the sequences of exons 18, 19, 21, 22, and 24 to 26 (GenBank reference NM_175629) were analyzed from genomic DNA by polymerase chain reaction and direct sequencing. Patients were also characterized for FLT3-ITD,^7,44FLT3-TKD,¹⁷MLL-PTD,^20,45 mutations in NPM1,¹³CEBPA,⁸WT1,^10,11TET2¹⁴ and IDH1/IDH2,¹⁵ and expression levels of ERG^23,24 and BAALC,^22,23 as previously reported. Molecular analyses were performed at The Ohio State University.

Microarray Experiments

Gene-expression profiling was performed using oligonucleotide microarrays (Affymetrix, Santa Clara, CA), and microRNA-expression profiling was performed using a custom microarray, as previously reported.^13,14,46 Expression signatures were identified by comparing DNMT3A-mutated and DNMT3A–wild-type (DNMT3A-wt) patients, and analyses to predict DNMT3A mutation status were performed (Data Supplement).

Statistical Analyses

Baseline characteristics were compared between DNMT3A-mutated and DNMT3A-wt patients using Fisher's exact test for categorical and the Wilcoxon rank sum test for continuous variables. Clinical end points were defined according to published recommendations (Data Supplement).⁴⁷ For time-to-event analyses, survival estimates were calculated using the Kaplan-Meier method, and groups were compared using the log-rank test. In addition to analyzing all DNMT3A-mutated cases as a combined group, we also evaluated the prognostic significance of R882-DNMT3A and non–R882-DNMT3A mutations separately, in the entire cohort and in the younger and older groups.

In models considering both age groups, we adjusted for an age-group effect (≥ 60 years v < 60 years). We constructed multivariable logistic regression models to analyze factors influencing achievement of CR and multivariable Cox proportional hazards models for factors associated with survival end points (Data Supplement). All analyses were performed by the Alliance for Clinical Trials in Oncology Statistics and Data Center.

RESULTS

Prevalence and Spectrum of DNMT3A Mutations in Primary CN-AML

Excluding known single-nucleotide polymorphisms, 148 nonsynonymous sequence variations (mutations) in DNMT3A were found in 142 (34.2%) of 415 patients (Data Supplement). The frequencies of these mutations were similar in younger (35.3%) and older (33.3%) patients. Six patients had two mutations each, and four mutations appeared homozygous. Ninety-two mutations (62%) were missense changes in codon R882, leading to an amino acid exchange from arginine to histidine (R882H, n = 49), cysteine (R882C, n = 36), proline (R882P, n = 3), serine (R882S, n = 3), or glycine (R882G, n = 1). R882-DNMT3A missense mutations were the most common mutation type among both younger (26%) and older (19%) patients. The remaining non–R882-DNMT3A mutations (n = 56; 38%) included 22 nonsense, frameshift, and splice-site mutations found in 22 different patients. These mutations are predicted to either trigger nonsense-mediated RNA decay or result in a truncated protein and thus are likely to impair protein function.⁴⁸ Two of these 22 patients concomitantly had an R882-DNMT3A mutation (for outcome analyses, these patients were included in the R882-DNMT3A mutation group), and two others concomitantly had another non–R882-DNMT3A missense mutation. Furthermore, there were 32 missense mutations not affecting codon R882 and two short in-frame deletions. All 32 non-R882 missense mutations were predicted to be “disease causing” by the MutationTaster software,⁴⁹ a computational algorithm that evaluates the disease-causing potential of gene mutations on the basis of evolutionary conservation and structural protein features.

Associations of DNMT3A Mutations With Pretreatment Clinical and Molecular Characteristics

No differences in age, sex, or race were observed between patients with and without DNMT3A mutations. However, DNMT3A-mutated patients had higher WBC counts (P < .001) and BM blasts percentages (P = .03) and harbored NPM1 mutations (P < .001) and FLT3-ITD (P = .01) more often and CEBPA mutations (P < .001) less often than those with DNMT3A-wt (Table 1).

Table 1.

Clinical and Molecular Characteristics of 415 Patients With Primary Cytogenetically Normal Acute Myeloid Leukemia According to DNMT3A Mutation Status

Characteristic	DNMT3A Mutated (n = 142)		DNMT3A Wild Type (n = 273)		P*
	No.	%	No.	%
Age, years					.46
Median	61		62
Range	22-82		18-83
Age, years					.68
< 60	64	45	117	43
≥ 60	78	55	156	57
Female sex	72	51	135	49	.84
Race					.49
White	124	89	247	91
Nonwhite	16	11	25	9
Hemoglobin, g/dL					.80
Median	9.4		9.4
Range	4.8-14.5		4.6-15
Platelet count, ×109/L					.55
Median	66		61
Range	4-481		7-850
WBC, ×109/L					< .001
Median	43.4		22.4
Range	0.9-434.1		0.9-450
Percentage of blood blasts					.83
Median	58		57
Range	0-97		0-99
Percentage of bone marrow blasts					.03
Median	70		66
Range	4-97		7-96
FAB category					< .001
M0	1	1	7	4
M1	29	27	54	27
M2	18	17	71	36
M4	33	31	41	21
M5	26	24	21	11
M6	0	0	3	2
NPM1					< .001
Mutated	107	75	146	53
Wild type	35	25	127	47
FLT3-ITD					.01
Present	62	44	85	31
Absent	80	56	188	69
CEBPA					< .001
Mutated	7	5	58	21
Single mutated	4		26
Double mutated	3		32
Wild type	135	95	215	79
ELN genetic group†					.10
Favorable	60	42	140	51
Intermediate-I	82	58	133	49
FLT3-TKD					1.00
Present	10	7	21	8
Absent	128	93	245	92
WT1					.37
Mutated	10	7	28	10
Wild type	132	93	245	90
TET2					.54
Mutated	31	22	67	25
Wild type	110	78	202	75
MLL-PTD					1.00
Present	7	6	15	6
Absent	116	94	228	94
IDH1					.07
R132	22	16	26	10
Wild type	118	84	246	90
IDH2
Mutated	24	17	51	19	.79
Codon R140	18		44
Codon R172	6		7
Wild type	116	83	221	81
ERG expression group‡					.39
High	56	55	92	49
Low	45	45	94	51
BAALC expression group‡					.38
High	45	47	104	53
Low	51	53	93	47

Abbreviations: FAB, French-American-British classification; ELN, European LeukemiaNet; FLT3-ITD, internal tandem duplication of the FLT3 gene; FLT3-TKD, tyrosine kinase domain mutation in the FLT3 gene; MLL-PTD, partial tandem duplication of the MLL gene.

^*P values for categorical variables are from Fisher's exact test; P values for continuous variables are from the Wilcoxon rank sum test.

^†The ELN favorable genetic group includes patients with mutated CEBPA and/or mutated NPM1 without FLT3-ITD.⁴ The ELN intermediate-I risk group comprises the remaining patients with CN-AML who had wild-type CEBPA and wild-type NPM1 with or without FLT3-ITD or mutated NPM1 with FLT3-ITD.

^‡The median expression value was used as a cut point.

Because AML biology and treatment regimens differ between younger and older patients, and it is unclear whether R882-DNMT3A and non–R882-DNMT3A mutations are functionally and clinically equivalent, we performed subgroup analyses taking age and DNMT3A mutation types into account. Younger R882-DNMT3A–mutated patients more often had NPM1 mutations (P = .02) and FLT3-ITD (P = .03) and less often had CEBPA mutations (P < .001), WT1 mutations (P = .02), and low ERG expression (P = .04) than DNMT3A-wt patients (Data Supplement). Younger patients with non–R882-DNMT3A mutations were more frequently NPM1-mutated (P = .02) and showed trends toward a higher frequency of FLT3-ITD (P = .11) and lower frequency of CEBPA mutations (P = .07; Data Supplement). Among older patients, those with R882-DNMT3A mutations showed trends toward a higher frequency of NPM1 mutations (P = .09) and FLT3-ITD (P = .15) and lower frequency of CEBPA mutations (P = .14; Data Supplement). Older patients with non–R882-DNMT3A mutations were more likely NPM1-mutated (P = .003) and, by trend, WT1-mutated (P = .07) and less likely CEBPA-mutated (P = .05; Data Supplement).

Association of DNMT3A Mutation Status With Clinical Outcome

When younger and older patients were considered together in analyses adjusted for age group, DNMT3A mutations were not associated with the probability of CR attainment (P = .42; Table 2). With a median follow-up of 7.5 years (range, 2.3 to 12.4 years) for patients alive, those harboring DNMT3A mutations had shorter disease-free survival (DFS; P = .03) and a trend toward shorter OS (P = .07) than DNMT3A-wt patients (Table 2; Fig 1). In a multivariable analysis for DFS (Table 3), DNMT3A mutations were associated with a 47% increased risk of relapse or death (P = .01), once adjusted for FLT3-ITD, WT1 mutations, MLL-PTD status, and age group. In contrast, once adjusted for other clinical and molecular prognosticators, there was no association of DNMT3A mutation status with OS.

Table 2.

Age Group–Adjusted Analysis of Outcomes of Patients With Primary Cytogenetically Normal Acute Myeloid Leukemia According to DNMT3A Mutation Status

End Point	DNMT3-mut (n = 142)	R882-DNMT3A (n = 92)	non–R882-DNMT3A (n = 50)	DNMT3A-wt (n = 273)	P (DNMT3A-mut vDNMT3A-wt)	P (R882-DNMT3AvDNMT3A-wt)	P (non–R882-DNMT3AvDNMT3A-wt)
Complete remission					.42	.85	.21
Odds ratio	1.22	1.05	1.62	Reference group
95% CI	0.75 to 1.96	0.61 to 1.83	0.76 to 3.43
Disease-free survival					.03	.03	.19
Hazard ratio	1.34	1.42	1.30	Reference group
95% CI	1.02 to 1.75	1.03 to 1.96	0.88 to 1.90
Overall survival					.07	.05	.39
Hazard ratio	1.25	1.32	1.16	Reference group
95% CI	0.98 to 1.57	1.01 to 1.74	0.83 to 1.63

Abbreviations: mut, mutated; wt, wild type.

Fig 1.

Age group–adjusted clinical outcome for patients with and without DNMT3A mutations. (A) Disease-free survival. (B) Overall survival. The curves are adjusted for age group. wt, wild type.

Table 3.

Multivariable Analyses for Outcome in Patients With Primary Cytogenetically Normal Acute Myeloid Leukemia

Group	Disease-Free Survival			Overall Survival
	HR	95% CI	P	HR	95% CI	P
All patients
DNMT3A, mutated v wild type	1.47	1.08 to 2.00	.01	DNMT3A mutation status was not significantly associated with OS upon adjusting for other variables
FLT3-ITD, ITD v no ITD	1.82	1.34 to 2.48	< .001
WT1, mutated v wild type	2.17	1.26 to 3.72	.005
MLL-PTD, present v absent	1.95	1.12 to 3.39	.02
Age group, older v younger	2.53	1.89 to 3.39	< .001
Patients, age < 60 years
DNMT3A, non-R882-mutated v wild type	2.78	1.45 to 5.36	.002	2.24	1.17 to 4.30	.02
NPM1, mutated v wild type	0.52	0.29 to 0.95	.03	0.38	0.23 to 0.64	< .001
FLT3-ITD, ITD v no ITD	1.79	1.07 to 3.02	.03	1.70	1.06 to 2.76	.03
CEBPA, double-mutated v single-mutated or wild type	0.21	0.09 to 0.50	< .001	0.15	0.06 to 0.35	< .001
WT1, mutated v wild type	4.88	2.37 to 10.04	< .001	5.91	3.20 to 10.90	< .001
Patients, age ≥ 60 years
DNMT3A, R882-mutated v wild type	1.85	1.20 to 2.84	.005	1.76	1.24 to 2.49	.002
NPM1, mutated v wild type	0.54	0.36 to 0.80	.002	0.48	0.35 to 0.66	< .001
FLT3-ITD, ITD v no ITD	2.00	1.34 to 2.98	< .001	1.80	1.31 to 2.47	< .001
Age, each 10 year increase	0.96	0.93 to 0.99	.02

Abbreviations: FLT3-ITD, internal tandem duplication of the FLT3 gene; HR, hazard ratio; MLL-PTD, partial tandem duplication of the MLL gene.

Association of Different DNMT3A Mutation Types With Clinical Outcome

We tested the association of the two types of DNMT3A mutations with outcome of younger and older patients separately because these age groups were treated on Cancer and Leukemia Group B protocols that differ in chemotherapy intensity (Data Supplement). Neither type of DNMT3A mutation had an impact on the probability of achieving CR in younger or older patients.

In younger patients, R882-DNMT3A mutations were not significantly associated with DFS or OS (Table 4, Figs 2A and 2B). In contrast, patients harboring non–R882-DNMT3A mutations had a significantly shorter DFS (P = .007; 3-year rates, 20% v 49%; Fig 2A) and a trend toward shorter OS (P = .09; 3-year rates, 29% v 52%; Fig 2B) than DNMT3A-wt patients (Table 4). In a multivariable analysis for DFS (Table 3), patients with non–R882-DNMT3A mutations had an almost three-fold increased risk of relapse or death (P = .002), once adjusted for FLT3-ITD and mutations in NPM1, CEBPA, and WT1. Likewise, in a multivariable model for OS (Table 3), the risk of death of non–R882-DNMT3A–mutated patients was more than twice that of DNMT3A-wt patients (P = .02) after adjustment for FLT3-ITD, NPM1, CEBPA, and WT1 mutation status.

Table 4.

Outcome of Patients With Primary Cytogenetically Normal Acute Myeloid Leukemia, According to Age Group and Type of DNMT3A Mutation

End Point	DNMT3A-mut	R882-DNMT3A	non–R882-DNMT3A	DNMT3A-wt	P (DNMT3A-mut vDNMT3A-wt)	P (R882-DNMT3AvDNMT3A-wt)	P (non–R882-DNMT3AvDNMT3A-wt)
Patients < 60 years of age, no.	64	47	17	117
Complete remission rate, %	81	79	88	96 (82)	1.00	.66	.74
Disease-free survival					.16	.68	.007
Median, years	1.1	1.3	0.7	2.9
% Disease-free at 3 years	37	43	20	49
95% CI	24 to 49	27 to 58	5 to 42	39 to 58
Overall survival					.36	.78	.09
Median, years	1.4	3.5	1.3	3.6
% Alive at 3 years	45	51	29	52
95% CI	33 to 57	36 to 64	11 to 51	43 to 61
Patients ≥ 60 years of age, no.	78	45	33	156
Complete remission rate,%	73	71	76	66	.30	.59	.31
Disease-free survival					.11	.006	.91
Median, years	0.7	0.7	1.0	1.0
% Disease-free at 3 years	11	3	20	21
95% CI	4 to 20	1 to 14	7 to 37	14 to 30
Overall survival					.10	.01	.96
Median, years	1.0	0.9	1.1	1.3
% Alive at 3 years	12	4	24	24
95% CI	6 to 21	1 to 13	11 to 39	17 to 31

Abbreviations: mut, mutated; wt, wild type.

Fig 2.

Kaplan-Meier survival curves according to DNMT3A mutation type (R882-DNMT3A v non–R882-DNMT3A mutations v DNMT3A wild type). (A) Disease-free survival and (B) overall survival of younger (< 60 years) patients. (C) Disease-free survival and (D) overall survival of older (≥ 60 years) patients. mut, mutated; wt, wild type.

In older patients, R882-DNMT3A mutations were associated with significantly shorter DFS (P = .006; 3-year rates, 3% v 21%; Fig 2C) and OS (P = .01; 3-year rates, 4% v 24%; Fig 2D), whereas non–R882-DNMT3A mutations were not (Table 4, Figs 2C and 2D). In a multivariable model for DFS (Table 3), R882-DNMT3A mutations remained associated with an 85% increased risk of relapse or death (P = .005) after adjustment for NPM1 mutation and FLT3-ITD status and age. Similarly, in a multivariable model for OS (Table 3), R882-DNMT3A mutations were associated with a 76% increased risk of death (P = .002) once adjusted for NPM1 mutation and FLT3-ITD status.

Gene- and microRNA-Expression Signatures Associated With DNMT3A Mutations

To gain insights into the biology of DNMT3A-mutated CN-AML, we studied mutation-associated gene-expression signatures in a subset of patients (n = 278) with available material. Clinical and molecular characteristics and outcome of this subset were similar to those of patients not analyzed.

A gene-expression signature associated with DNMT3A mutations comprised 1,886 differentially expressed probe sets: 1,323 were upregulated and 563 downregulated in DNMT3A-mutated patients (Data Supplement). The most upregulated known gene was VCAN, encoding a protein involved in cell adhesion, proliferation, migration, and angiogenesis; the most downregulated gene was ALAS2, involved in the heme biosynthetic pathway. However, the signature had an overall cross-validated accuracy of only 67% for predicting DNMT3A mutation status (62% sensitivity; 70% specificity), thereby suggesting the contributing effect of other associated molecular aberrations.

When we attempted to derive gene-expression signatures associated with specific types of DNMT3A mutations, no significant signature separated patients harboring non–R882-DNMT3A mutations (n = 32) from those with DNMT3A-R882 mutation (n = 60).

For microRNA profiling, younger and older patients were analyzed separately to avoid confounding batch effects. Testing for differentially expressed microRNAs revealed no signature associated with DNMT3A mutations in the younger group. In contrast, we derived a signature consisting of 12 microRNAs associated with DNMT3A mutations in older patients (Data Supplement), with four microRNA probes upregulated, including a member of the miR-10 family reportedly associated with NPM1 mutations,¹³ and eight microRNA probes downregulated, including miR-181c, a member of the miR-181 family associated with CEBPA mutations.⁸ However, these features might reflect confounding as a result of the significant positive association of DNMT3A mutations with NPM1 mutations and the negative association with CEBPA mutations. This microRNA-expression signature had an overall accuracy of only 58% for predicting DNMT3A mutation status (49% sensitivity; 62% specificity).

DISCUSSION

Advanced sequencing technologies have allowed analysis of the whole genome of AML blasts. Application of these technologies has recently identified two novel recurrent gene mutations in CN-AML, first IDH1 mutations⁵⁰ and, more recently, DNMT3A mutations.³¹ As this approach becomes broadly used, it is likely that previously unrecognized mutations in AML will continue to emerge. Because these mutations have the potential to contribute to myeloid leukemogenesis and become prognostic factors and/or therapeutic targets, it is imperative to rapidly test their biologic and clinical impact on patients with AML. However, from previously discovered mutated or aberrantly expressed genes in CN-AML, we have learned that only rarely is testing for a single genetic alteration sufficient for accurate outcome prediction and treatment guidance.¹³ Instead, the clinical impact of most molecular markers is influenced by other, concurrent molecular aberrations.^{12,14–16,21} Therefore, to fully understand the clinical significance of emerging molecular markers, such as DNMT3A mutations, they need to be evaluated in large series of patients homogeneous for age and type of disease (primary v secondary or treatment-related AML), similarly treated and fully characterized for established prognostic markers. To our knowledge, our study analyzed DNMT3A mutations in the largest CN-AML patient cohort to date and is first to report subgroup analyses and multivariable models considering different types of DNMT3A mutations in distinct age groups.

We found that DNMT3A mutations were among the most common mutations in CN-AML, occurring in 34% of patients, with a similar frequency among younger and older patients, and were significantly associated with NPM1 mutations, FLT3-ITD, and wild-type CEBPA. Regarding prognostic significance, we showed that DNMT3A mutations had worse DFS and OS, after adjustment for age. Moreover, we observed that the prognostic significance of DNMT3A mutations depended both on age and the type of mutation (R882-DNMT3A v non–R882-DNMT3A) considered concurrently (see Fig 2 and also Data Supplement). In younger patients, only non–R882-DNMT3A mutations were associated with worse clinical outcome, whereas R882-DNMT3A mutations had no prognostic significance. Conversely, in older patients, only R882-DNMT3A mutations, not non–R882-DNMT3A mutations, were independently associated with worse outcome. The reasons why the prognostic significance of different DNMT3A mutation types varies in younger and older patients are currently unknown. One could postulate that this is related to their association with other prognosticators. Thus, in older patients, the potentially negative prognostic significance of non–R882-DNMT3A mutations might have been somewhat offset by a high incidence (79%) of accompanying NPM1 mutations, known to favorably affect prognosis of older patients.¹³ However, two thirds of older patients with the prognostically adverse R882-DNMT3A mutations also harbored NPM1 mutations, and slight differences in frequencies of other molecular markers between patients harboring R882-DNMT3A and non–R882-DNMT3A mutations, both in the older and younger age groups, do not seem sufficient to account for the differential association of the two DNMT3A mutation types on treatment outcome.

Our results differ somewhat from those reported by Ley et al,³¹ who found a strong, independent association of DNMT3A mutations with OS, and those by Thol et al,³³ who studied only patients younger than 60 years and who found that in the CN-AML subgroup, DNMT3A mutations were associated with a lower CR rate and shorter OS in multivariable analyses. These discrepancies may be related to differences in the patient populations analyzed with respect to their size, cytogenetics, molecular markers, age, disease type, and treatment. Furthermore, previous studies did not include older patients³³ or included only a small proportion of older patients and did not present data on CR rates, DFS, or multivariable analyses for patients with CN-AML.³¹ Therefore, a direct comparison of the findings across studies is not possible.

We also report the first gene- and microRNA-expression signatures associated with DNMT3A mutations. However, the accuracy of the gene-expression signature in predicting DNMT3A mutational status was only 67%. These results are consistent with an unsupervised analysis of gene-expression array data reported by Ley et al,³¹ where no patient cluster was clearly linked to DNMT3A mutation status. Similarly, a microRNA-expression signature derived in older patients with CN-AML comprised microRNAs strongly associated with other markers (ie, NPM1 mutations and wild-type CEBPA) and was not accurate in predicting DNMT3A mutational status. These results suggest that DNMT3A mutations have no strong impact on genome-wide gene- and microRNA-expression profiles in CN-AML. The signatures we identified might at least partially reflect the association of DNMT3A mutation status with other molecular markers that are themselves associated with characteristic gene- and microRNA-expression signatures.

The mechanisms through which DNMT3A mutations contribute to leukemogenesis are not yet characterized. Although two studies^31,33 found no differences in global DNA methylation or changes in gene methylation patterns in DNMT3A-mutated patients, other reports^30,32 suggested that most of the DNMT3A mutations decrease the enzymatic activity of the encoded protein. Uncovering how DNMT3A mutations affect DNA methylation and epigenetic regulation of gene expression may have ramifications for treatment selection because DNA hypomethylating agents, such as decitabine, are increasingly used for up-front or salvage therapies in older patients with AML,⁵¹ and response to these drugs may be affected by alterations in DNMT3A function.⁵²

In summary, testing for the mutations in DNMT3A may provide a new tool for refining age-related risk classification of CN-AML. The strongest prognostic significance was found in older patients harboring R882-DNMT3A mutations, whereas non–R882-DNMT3A mutations were associated with relapse risk in younger patients. The gene- and microRNA-expression signatures were not accurate in predicting DNMT3A mutational status likely because they are affected by other, concurrent molecular markers. Thus the contribution of the DNMT3A mutations to myeloid leukemogenesis requires further investigation, as does the usefulness of DNMT3A mutations for risk stratification both in patients with CN-AML and in other cytogenetic and molecular subsets of AML.

Appendix

The following Cancer and Leukemia Group B (CALGB) institutions, principal investigators, and cytogeneticists participated in this study: Wake Forest University School of Medicine, Winston-Salem, NC: David D. Hurd, P. Nagesh Rao, Wendy L. Flejter, and Mark J. Pettenati (Grant No. CA03927); The The Ohio State University Medical Center, Columbus, OH: Clara D. Bloomfield, Karl S. Theil, Diane Minka, and Nyla A. Heerema (Grant No. CA77658); North Shore–Long Island Jewish Health System, Manhasset, NY: Daniel R. Budman and Prasad R.K. Koduru (Grant No. CA35279); University of Iowa Hospitals, Iowa City, IA: Daniel A. Vaena and Shivanand R. Patil (Grant No. CA47642); Roswell Park Cancer Institute, Buffalo, NY: Ellis G. Levine and AnneMarie W. Block (Grant No. CA02599); Duke University Medical Center, Durham, NC: Jeffrey Crawford, Sandra H. Bigner, Mazin B. Qumsiyeh, John Eyre, and Barbara K. Goodman (Grant No. CA47577); Washington University School of Medicine, St. Louis, MO: Nancy L. Bartlett, Michael S. Watson, Eric C. Crawford, and Jaime Garcia-Heras (Grant No. CA77440); University of Chicago Medical Center, Chicago, IL: Hedy L. Kindler, Diane Roulston, Katrin M. Carlson, Yanming Zhang, and Michelle M. Le Beau (Grant No. CA41287); University of North Carolina, Chapel Hill, NC: Thomas C. Shea and Kathleen W. Rao (Grant No. CA47559); University of Massachusetts Medical Center, Worcester, MA: William V. Walsh, Vikram Jaswaney, Michael J. Mitchell, and Patricia Miron (Grant No. CA37135); Dana-Farber Cancer Institute, Boston, MA: Harold J. Burstein, Ramana Tantravahi, Leonard L. Atkins, Paola Dal Cin, and Cynthia C. Morton (Grant No. CA32291); Vermont Cancer Center, Burlington, VT: Steven M. Grunberg, Elizabeth F. Allen, and Mary Tang (Grant No. CA77406); Weill Medical College of Cornell University, New York, NY: John Leonard, Ram S. Verma, Prasad R.K. Koduru, and Susan Mathew (Grant No. CA07968); Ft. Wayne Medical Oncology/Hematology, Ft. Wayne, IN: Sreenivasa Nattam and Patricia I. Bader; Dartmouth Medical School, Lebanon, NH: Konstantin Dragnev, Doris H. Wurster-Hill, and Thuluvancheri K. Mohandas (Grant No. CA04326); Eastern Maine Medical Center, Bangor, ME: Harvey M. Segal and Laurent J. Beauregard (Grant No. CA35406); Mount Sinai School of Medicine, New York, NY: Lewis R. Silverman and Vesna Najfeld (Grant No. CA04457); Minneapolis VA Medical Center, Minneapolis, MN: Vicki A. Morrison and Sugandhi A. Tharapel (Grant No. CA47555); University of Puerto Rico School of Medicine, San Juan, PR: Eileen I. Pacheco, Leonard L. Atkins, Paola Dal Cin, and Cynthia C. Morton; Rhode Island Hospital, Providence, RI: William Sikov, Teresita Padre-Mendoza, Hon Fong L. Mark, Shelly L. Kerman, and Aurelia Meloni-Ehrig (Grant No. CA08025); SUNY Upstate Medical University, Syracuse, NY: Stephen L. Graziano, Larry Gordon, and Constance K. Stein (Grant No. CA21060); University of California at San Diego: Barbara A. Parker, Renée Bernstein, and Marie L. Dell'Aquila (Grant No. CA11789); Christiana Care Health Services, Newark, DE: Stephen S. Grubbs, Digamber S. Borgaonkar, and Jeanne M. Meck (Grant No. CA45418); Long Island Jewish Medical Center CCOP, Lake Success, NY: Kanti R. Rai and Prasad R. K. Koduru (Grant No. CA11028); Massachusetts General Hospital, Boston, MA: Jeffrey W. Clark, Leonard L. Atkins, Paola Dal Cin, and Cynthia C. Morton (Grant No. CA 12,449); University of Maryland Cancer Center, Baltimore, MD: Martin J. Edelman, Joseph R. Testa, Maimon M. Cohen, Judith Stamberg, and Yi Ning (Grant No. CA31983); University of Minnesota, Minneapolis, MN: Bruce A. Peterson, Diane C. Arthur, and Betsy A. Hirsch (Grant No. CA16450); University of Missouri/Ellis Fischel Cancer Center, Columbia, MO: Michael C. Perry and Tim H. Huang (Grant No. CA12046); University of Nebraska Medical Center, Omaha, NE: Anne Kessinger and Warren G. Sanger (Grant No. CA77298); Western Pennsylvania Hospital, Pittsburgh, PA: John Lister and Gerard R. Diggans; University of Illinois at Chicago: David J. Peace, Maureen M. McCorquodale, and Kathleen E. Richkind (Grant No. CA74811); Walter Reed Army Medical Center, Washington, DC: David C. Van Echo, Rawatmal B. Surana, and Digamber S. Borgaonkar (Grant No. CA26806); Georgetown University Medical Center, Washington, DC: Minnetta C. Liu and Jeanne M. Meck (Grant No. CA77597); McGill Department of Oncology, Montreal, Quebec: J.L. Hutchison and Jacqueline Emond (Grant No. CA31809); Medical University of South Carolina, Charleston, SC: Mark R. Green, G. Shashidhar Pai, and Daynna J. Wolff (Grant No. CA03927); University of Cincinnati Medical Center, Cincinnati, OH: Orlando J. Martelo and Ashok K. Srivastava (Grant No. CA47515); Columbia-Presbyterian Medical Center, New York, NY: Rose R. Ellison and Dorothy Warburton (Grant No. CA12011); University of Minnesota, Minneapolis, MN: Bruce A. Peterson, Diane C. Arthur, and Betsy A. Hirsch (Grant No. CA16450); Virginia Commonwealth University MB CCOP, Richmond, VA: John D. Roberts and Colleen Jackson-Cook (Grant No. CA52784); SUNY Maimonides Medical Center, Brooklyn, NY: Sameer Rafla and Ram S. Verma (Grant No. CA25119); Southern Nevada Cancer Research Foundation CCOP, Las Vegas, NV: John Ellerton and Marie L. Dell'Aquila (Grant No. CA35421); University of California at San Francisco: Charles J. Ryan and Kathleen E. Richkind (Grant No. CA60138).

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Guido Marcucci, Klaus H. Metzeler, Clara D. Bloomfield

Financial support: Guido Marcucci

Provision of study materials or patients: Guido Marcucci, Bayard L. Powell, Thomas H. Carter, Jonathan E. Kolitz, Meir Wetzler, Andrew J. Carroll, Maria R. Baer, Joseph O. Moore, Michael A. Caligiuri, Richard A. Larson

Collection and assembly of data: Guido Marcucci, Klaus H. Metzeler, Sebastian Schwind, Heiko Becker, Krzysztof Mrózek, Susan P. Whitman, Yue-Zhong Wu, Bayard L. Powell, Thomas H. Carter, Jonathan E. Kolitz, Meir Wetzler, Andrew J. Carroll, Maria R. Baer, Joseph O. Moore, Michael A. Caligiuri, Richard A. Larson, Clara D. Bloomfield

Data analysis and interpretation: Guido Marcucci, Klaus H. Metzeler, Sebastian Schwind, Heiko Becker, Kati Maharry, Krzysztof Mrózek, Michael D. Radmacher, Jessica Kohlschmidt, Deedra Nicolet, Clara D. Bloomfield

Manuscript writing: All authors

Final approval of manuscript: All authors