Prognostic impact of DNMT3A mutation in acute myeloid leukemia with mutated NPM1

In recent years, the role of molecular genetics has proven to be essential in deciphering the heterogeneity of acute myeloid leukemia (AML)^1,2  and defining genetic markers of prognostic significance that can guide risk-adapted treatment.³ 

AML with mutations in the nucleophosmin 1 gene (AML-NPM1) forms a specific category in the latest World Health Organization classification because of its singular characteristics.⁴  The cooccurrence of mutated NPM1 (NPM1^mut) and the internal tandem duplication of FLT3 (FLT3-ITD) in de novo AML with intermediate-risk cytogenetics results in a different prognostic impact depending on the FLT3 allelic burden.⁵ -⁷  Previous studies have shown that patients with NPM1^mut and an FLT3-ITD low ratio (FLT3^low; FLT3-ITD/FLT3^WT ratio of <0.5) had overall survival (OS) and risk of relapse (RR) similar to those of patients with NPM1^mut and wild-type (WT) FLT3 (FLT3^WT).^8,9  Since 2012, these findings have been included in our latest protocol (CETLAM [Spanish Cooperative Group for the Diagnosis and Treatment of Acute Myeloid Leukemia and Myelodysplastic Syndromes] AML-12), and patients with FLT3^low-NPM1^mut AML are not considered for allogenic hematopoietic stem cell transplantation (alloHSCT) in first complete remission (CR1). However, a molecular-based measurable residual disease (MRD) monitoring protocol is strictly followed to allow early-intervention strategies.

The DNA (cytosine-5)-methyltransferase 3A (DNMT3A) gene is located on the short arm of chromosome 2 and encodes for a DNA methyltransferase that methylates unmodified DNA cytosine residues modulating the expression of several genes.^10,11  Almost all DNMT3A mutations are heterozygous, and more than two-thirds cluster at the methyltransferase domain in codon R882, causing loss of methylation activity by disturbing DNMT3A tetramerization.¹² -¹⁶  However, although a precise methylation pattern alteration resulting from mutations in DNMT3A has not yet been established,¹⁷ -²¹  a new mechanism of leukemogenesis characterized by the upregulation of the hepatic leukemia factor (a specific leukemic transcription factor) has been shown to be related to the cooccurrence of DNMT3A, NPM1, and FLT3 mutations.²² 

DNMT3A is considered a founder mutation in AML.^23,24  It has been associated with age-related clonal hematopoiesis, with increasing frequency in healthy elderly individuals,²⁵  although a recent study found a correlation of DNMT3A mutations with younger age in NPM1^mut AML.²⁶  Patients with AML and mutated DNMT3A (DNMT3A^mut) are frequently older and present with higher white blood cell (WBC) counts and higher platelet counts compared with WT DNMT3A (DNMT3A^WT).^13,17,27 DNMT3A is the third most frequently mutated gene in AML patients included in intensive chemotherapy trials. It is predominantly observed in AML-NPM1 (73%) and less frequently in patients with mutations in chromatin remodeling genes or genes involved in spliceosome function. Interestingly, a recurrent association of NPM1^mut/DNMT3A^mut/FLT3-ITD has been observed in 6% of AML cases.^28,29  The prognostic significance of mutations in DNMT3A has been controversial; some studies have found no significant influence on survival outcomes,¹⁷  whereas others have suggested that the cooccurrence of NPM1^mut/DNMT3A^mut/FLT3-ITD in AML patients is associated with adverse outcomes.^7,13,30 -³³ 

In 2013, the German AML Study Group described the clinical impact of DNMT3A^mut in younger adults with AML.³⁴  In a univariable exploratory subset analysis, the group showed a significant prognostic impact of DNMT3A^mut in unfavorable European LeukemiaNet (ELN) AML, whereas no impact was observed in favorable ELN AML. In 2016, the proposed AML gene classification by Papaemmanuil et al²⁸  showed a deleterious effect of DNMT3A^mut when specifically associated with FLT3-ITD independently of its allelic ratio.

We analyzed whether this triple-gene alteration led to an unfavorable prognosis in AML-NPM1 patients, with particular attention to those harboring FLT3^low.

Patients with de novo AML and intermediate-risk cytogenetics according to the Medical Research Council,³⁵ NPM1^mut, and available bone marrow sample from diagnosis were selected. All patients were diagnosed between 2003 and 2017 and were included in the CETLAM intensive treatment protocols AML-03 and AML-12 (registered at www.clinicaltrials.gov as #NCT01723657 and #NCT04687098) provided they met the criteria for eligibility. The present study was reviewed and approved by the ethics committee of the Hospital de la Santa Creu i Sant Pau (Comitè ètic d’Investigació Clínica). Informed consent for both bone marrow analysis and treatment was obtained in all cases according to the Declaration of Helsinki.

Diagnostic bone marrow samples from all patients were analyzed for DNMT3A^mut as previously described.¹⁷  The mutational status of the FLT3 gene was also established. In mutated cases, the allelic ratio was calculated by dividing the area under the curve of the FLT3-ITD peak and the area under the curve of the FLT3^WT peak. Patients were stratified into 2 groups: those with a high ratio (FLT3^high) if the ratio of FLT3-ITD/FLT3^WT was ≥0.5 and those with FLT3^low if the ratio of FLT3-ITD/FLT3^WT was <0.5.

Monitoring of NPM1 MRD was performed on bone marrow samples by quantitative reverse transcription polymerase chain reaction (sensitivity 10⁻⁴ to 10⁻⁶) as previously described.³⁶  After each treatment cycle, absolute transcript reduction was estimated, and its logarithm (log10) reduction from diagnosis was also explored. Based on the latest ELN MRD working party recommendations,³⁷  MRD positivity was considered when NPM1 transcripts were amplified in at least 2 of 3 replicates with cycle threshol values of ≤40 at a cycling threshold of 0.1. Molecular relapse was confirmed if the MRD level (in a patient previously MRD⁻) increased ≥1 log10 between 2 consecutive positive samples, and molecular progression was confirmed if copy numbers increased ≥1 log10 between 2 positive samples in a patient with MRD⁺.

Analysis of the relationship between categorical variables was performed using the χ² or Fisher’s exact test. Differences between groups for continuous variables were established through the independent samples t test or Mann-Whitney U test. All tests were 2 sided and considered significant where P < .05. OS was calculated from diagnosis to death, whereas leukemia-free survival (LFS) was calculated from CR to death or relapse; both functions were estimated with the Kaplan-Meier method. Unless specified otherwise, all survival results reported reflect 5-year estimates. A log-rank test was run to determine differences in the survival distribution, with a significance threshold of P ≤ .05. RR was estimated using the cumulative incidence method, defining relapse as the main event and death without relapse as the competitive event. Molecular LFS (molLFS) was estimated from CR to molecular failure, overt hematological relapse, or death. All statistical analyses were performed with SPSS software (version 26; IBM, Armonk, NY) and R statistics (version 3.6.1; R Foundation for Statistical Computing, Vienna, Austria).

A total of 164 patients with AML-NPM1 were included. Clinical and genetic characteristics at diagnosis are described in Table 1. Patients were included in protocols AML-03 (n = 48) and AML-12 (n = 116). DNMT3A^mut was found in 79 patients (48%); in 62 cases (38%), mutations were in codon R882 or were insertions/deletions, whereas 17 (10%) were different missense mutations. Seventy-six patients (46%) harbored FLT3-ITD, 42 of whom had FLT3^high (55%). According to DNMT3A^mut status, patient characteristics were comparable, except for a higher WBC presentation among DNMT3A^mut patients. Of note, the proportion of FLT3-ITD allelic burden subsets (ie, FLT3^WT, FLT3^low, and FLT3^high) was independent of DNMT3A^mut (P = .6).

Eighty-eight percent of patients achieved CR after 1 or 2 cycles of induction therapy (n = 128 and 16, respectively); 6% (n = 10) had refractory disease, and 10 patients died during induction. As consolidation therapy, 58 patients received intensive treatment consisting of 2 to 3 high-dose cytarabine cycles. AlloHSCT was performed in 65 patients (CR1, n = 44; CR2, n = 15; with refractory disease, n = 4); 14 patients underwent autologous transplantation. The median follow-up time was 30 months.

In the entire cohort of AML-NPM1 patients, OS, LFS, and RR were 59% ± 4%, 60% ± 5%, and 27% ± 7%, respectively. FLT3-ITD allelic ratio confirmed its prognostic impact, with a similar outcome for patients with FLT3^WT or FLT3^low and a significantly worst prognosis for cases with FLT3^high. OS was 67% ± 6% vs 62% ± 9% vs 40% ± 8% (P = .002; supplemental Figure 1), respectively; RR was 18% ± 9% vs 27% ± 16% vs 41% ± 17% (P = .008), respectively; and LFS was 71% ± 5% vs 56% ± 9% vs 40% ± 9% (P = .002), respectively. In contrast, DNMT3A^mut did not exert a significant effect on overall outcomes (Figure 1), with OS in DNMT3A^WT vs DNMT3A^mut patients of 62% ± 6% vs 56% ± 6% (P = .2), respectively; LFS of 65% ± 6% vs 54% ± 6% (P = .1), respectively; and RR of 22% ± 11% vs 31% ± 11% (P = .2). The outcomes of DNMT3A subsets among the entire AML-12 cohort are available in supplemental Figure 2.

Additionally, the individual effect of R882 DNMT3A^mut was analyzed separately and did not show any prognostic impact (P = .4; supplemental Figure 3). Multivariate analysis performed for OS included age, protocol, WBC count, DNMT3A^mut status, and FLT3-ITD subsets, the latter being the only statistically significant variable (data not shown).

The effect of the cooccurrence of DNMT3A^mut and FLT3-ITD was analyzed separately. In the DNMT3A^mut cohort, whereas OS between FLT3^WT and FLT3^low was similar, patients with FLT3^high showed particularly worse outcomes (FLT3^WT vs FLT3^low OS, 68% ± 8% vs 56% ± 13%, respectively; P = .3; FLT3^WT/low vs FLT3^high OS, 65% ± 7% vs 28% ± 13%, respectively; P = .015). This was further validated in terms of LFS (FLT3^WT vs FLT3^low, 68% ± 8% vs 50% ± 13%, respectively; P = .1; FLT3^WT/low vs FLT3^high, 62% ± 7 vs 20% ± 16%, respectively; P = .014) and RR in the same groups (23% ± 15% vs 29% ± 23%, respectively; P = .4; 25% ± 13% vs 48% ± 20%, respectively; P = .017; Figure 2).

In patients without DNMT3A^mut, the allelic ratio of FLT3-ITD maintained its prognostic value, with similar outcomes between FLT3^WT and FLT3^low patients. Interestingly, in this group of patients, the deleterious effect of FLT3^high was mildly modulated (FLT3^WT vs FLT3^low OS, 66% ± 8% vs 66% ± 11%, respectively; P = .5; FLT3^WT/low vs FLT3^high OS, 66% ± 7% vs 47% ± 12%, respectively; P = .028). Similar findings were seen in terms of LFS (FLT3^WT vs FLT3^low, 74% ± 8% vs 61% ± 13%, respectively; P = .2; FLT3^WT/low vs FLT3^high, 70% ± 7% vs 52% ± 13%, respectively; P = .083) and RR (14% ± 11% vs 26% ± 21%, respectively; P = .2; 17% ± 12% vs 33% ± 23%, respectively; P = .1; Figure 3).

The direct comparison of DNMT3A^mut status according to each FLT3 status (WT, low, and high) was not statistically different (supplemental Figure 4), although the general analysis of DNMT3A depending on FLT3^WT or FLT3^mut showed statistical differences. This might be due to the low number of patients analyzed when dividing into the 3 groups.

Patients included in the CETLAM AML-12 trial with AML-NPM1 allocated to the favorable ELN category (FLT3^WT or FLT3^low) were not intended to undergo alloHSCT in CR1 unless a molecular failure was identified. Therefore, patients were closely monitored for NPM1 MRD at specific time points (postinduction, after each consolidation cycle, and every 3 months for 3 years after treatment completion) to ensure rapid detection of molecular relapse and subsequent therapy initation.³⁸  Patients with FLT3^high underwent alloHSCT after the first consolidation cycle (C1).

In 94 patients with available MRD data, we further investigated whether DNMT3A^mut status influenced NPM1 MRD. Although the probability of achieving hematological CR was not affected by DNMT3A (75 vs 69 patients in with DNMT3A^WT and DNMT3A^mut, respectively; P = .46), NPM1 MRD kinetics differed according to DNMT3A^mut status. Patients with DNMT3A^mut showed a higher number of absolute NPM1^mut transcripts after induction (P = .012) and C1 (P < .001; Figure 4A). Similar findings were observed after C2 and C3 among patients not intended to undergo alloHSCT in CR1.

Therefore, we explored the relationship between DNMT3A subsets, posttreatment molecular MRD status (positive vs negative), and MRD response depth (log10 reduction; Figure 4B-G). After induction, all but 1 patient remained MRD⁺. Although there were no statistical differences in the number of log10 reductions, a trend toward a more profound molecular response (≥4 log10) was observed in the DNMT3A^WT group (DNMT3A^WT vs DNMT3A^mut, 39% vs 15%, respectively; P not significant).

After C1, none of the DNMT3A^mut patients achieved MRD⁻ status, compared with 32% of DNMT3A^WT patients (P = .001). Of note, patients without DNMT3A^mut presented a deeper MRD reduction (≥4 log10 reduction in 77% of DNMT3A^WT vs 46% of DNMT3A^mut patients; P = .033). The relationship between DNMT3A and NPM1 MRD was also sustained after C2 and C3 (Figure 4B-G; supplemental Figure 5). Additionally, when considering the triple-mutated group (NPM1, FLT3-ITD, and DNMT3A), all patients remained MRD⁺ after induction, C1, and C2 regardless of FLT3-ITD allelic ratio.

Finally, the potential influence of DNMT3A^mut status on molecular failure was explored. Among 85 cases included in the AML-12 protocol not initially considered for alloHSCT in CR1 (AML-NPM1 with FLT3^WT or FLT3^low, n = 63 and 22, respectively), the median molLFS was not reached at a mean follow-up of 30 months (supplemental Figure 6). When stratified by DNMT3A^mut status, patients with the WT form exhibited a trend toward a long-term sustained molecular CR (molLFS, 63% ± 9% vs 50% ± 9% in DNMT3A^WT (n = 42) vs DNMT3A^mut (n = 35), respectively; P = .054; Figure 5).

Eleven patients in the favorable-risk group harbored NPM1^mut/FLT3^low/DNMT3A^mut; of these, only 5 experienced a molecular or hematological relapse and underwent alloHSCT. In total, 23 patients (27%) in this favorable subgroup underwent alloHSCT because of molecular or hematological relapse.

Overall, these findings suggest a deleterious effect of DNMT3A^mut on NPM1 MRD that should be validated in larger studies.

Several studies have been published attempting to elucidate the prognostic impact of DNMT3A^mut, but many have had contradictory results. This may be due to differences in the biological characteristics of the patients included (age, cytogenetics, availability of molecular studies), the treatment protocols, or other factors.³⁹  Of note, even those studies comparing the impact of DNMT3A^mut status based on NPM1^mut and FLT3^mut status have shown contradictory results.^13,17,34  The aim of this study was not to analyze the impact of DNMT3A^mut on AML outcomes, but rather to analyze its effect in the particular subset of patients with NPM1^mut and FLT3-ITD, after the publication of a large study showing that DNMT3A^mut have a deleterious effect on outcomes when cooccurring in this subgroup.²⁸  Our group described the effect of the FLT3-ITD ratio in 2012, and it was incorporated into the new treatment protocol. Consequently, patients with NPM1^mut and FLT3^low did not undergo alloHSCT in CR1. Therefore, we had a long follow-up in this group of patients treated following the ELN 2017 recommendations in which to analyze the possible effect of DNMT3A^mut.

In our study, survival analysis showed that DNMT3A^mut did not have an impact in this particular group and that patients with NPM1^mut and FLT3^low had similar outcomes to patients with NPM1^mut and FLT3^WT regardless of DNMT3A^mut status. However, because an effect of DNMT3A^mut on NPM1 MRD clearance was demonstrated, we investigated the influence of an early intervention planned in the treatment protocol when molecular relapse was detected. In the last few years, several publications analyzing the prognostic value of MRD follow-up based on NPM1 transcript levels have been published. Although there is no consensus regarding the cutoff level or evaluation time points, all of them support the prognostic impact of MRD⁺ persistence, with a higher incidence of relapse and shorter OS.^38,40 -⁴³  The largest study performed⁴⁴  evaluated the impact of MRD⁺ in peripheral blood after the second chemotherapy cycle; it found the same impact on prognosis as previously reported, but the authors also reported that MRD persistence was the only independent prognostic factor for death in multivariate analysis. The ELN recommendations³⁷  also state that in AML-NPM1, rising MRD levels or the failure to achieve MRD⁻ CR is associated with disease relapse and consequently advise that a change in therapy should be considered. Following the same reasoning, it was recently published by our group that an MRD ratio (NPM1^mut/ABL1X100) of ≥0.05 (in bone marrow) after the C1 was associated with significantly lower molLFS and that an early intervention resulted in a favorable outcomes.³⁸  Consequently, using the MRD level to guide postremission therapy can be considered a good strategy.

Interestingly, in the present study, a trend toward worse molLFS was observed in patients with DNMT3A^mut, but without an impact on OS. When only patients in the favorable ELN 2017 risk group were considered, we found that 27% of patients met either cytological or molecular relapse criteria. Of those, 70% underwent alloHSCT in CR1 (in molecular relapse) or CR2. As a result, the effect of this strategy might counteract the negative effect on OS seen in the DNMT3A^mut subgroup. This intervention might be the most important difference between the treatment protocols for our patients and those included in the Papaemmanuil et al²⁸  study, which considered alloHSCT only in patients at high cytogenetic risk, whereas intermediate-risk patients underwent alloHSCT only when a sibling donor was available.⁴⁵ -⁴⁷ 

Considering these findings, close MRD monitoring in DNMT3A^mut AML-NPM1 patients, along with early intervention strategies when a molecular relapse is detected, could be an appropriate approach, with a possible impact on OS.

Patients with DNMT3A^mut and FLT3^high had poorer outcomes than patients in the favorable ELN group (ie, FLT3^WT or FLT3^low). Nonetheless, DNMT3A^mut status did not seem to affect patients with FLT3^high, although a deleterious effect of this triple-mutation status (NPM1^mut/FLT3^high/DNMT3A^mut vs NPM1^mut/FLT3^high/DNMT3A^WT) cannot be definitively excluded because of the small size of the subgroups analyzed. These findings may show a dosage effect on the interaction between FLT3 and DNMT3A^mut in AML-NPM1, highlighting the relevance of considering not only the presence of every single mutation but also the interaction among them.

In conclusion, patients with NPM1-AML with FLT3^low and DNMT3A^mut can be classified as favorable risk, but closer MRD follow-up is recommended to detect a molecular relapse and proceed to a therapeutic intervention.

This work was supported in part by the Biomedical Research Institute (IIB Sant-Pau) and the José Carreras Leukemia Research Institute as well as grants from the Catalan Government (PERIS SLT002/16/0043 and AGAUR 2017 SGR 139) and the Instituto de Salud Carlos III, Ministerio de Economía y Competitividad, Spain (PI17/01246, PI20/01621 and CM20/00061).

G.O. is a PhD candidate at the Autonomous University of Barcelona, and this work is submitted in partial fulfillment of the requirement for a PhD.

Contribution: G.O. and A.A. performed research. M.P. and G.O. designed research, analyzed data, and wrote the paper; J.S., J.F.N., and J.E. supervised research and wrote the paper; M.A., S.V., R.C., M.T., A.S., L.E., O.S., A.G., and J.B. collected and provided the clinical data; A.B., A.G., and M.H. analyzed data; and all authors reviewed the final manuscript.

Conflict-of-interest disclosure: J.E. reports an advisory role and trial investigation for Novartis, Daiichi Sankyo, Astellas, Celgene, Jazz Pharmaceuticals, Roche, Boehringer Ingelheim, and Janssen. J.S. reports personal fees from AbbVie, Vyxeos, Gilead, CSL Behring, Astellas, and Gilead; grants and personal fees from Novartis and Daiichi-Sankyo; and grants from Amgen. The remaining authors declare no competing financial interests.

Correspondence: Marta Pratcorona, Department of Hematology, Hospital de la Santa Creu i Sant Pau, Sant Quintí 89 08046 Barcelona, Spain; e-mail: mpratcorona@santpau.cat.