Do AML patients with DNMT3A exon 23 mutations benefit from idarubicin as compared to daunorubicin? A single center experience.

Mutations in DNMT3A encoding DNA methyltransferase 3A were recently described in patients with acute myeloid leukemia. To assess their prognostic significance, we determined the mutational status of DNMT3A exon 23 in 288 patients with AML excluding acute promyelocytic leukemia, aged from 18 to 65 years and treated in Toulouse University Hospital. A mutation was detected in 39 patients (13.5%). All DNMT3A exon 23+ patients had intermediate-risk cytogenetics. Mutations significantly correlated with a higher WBC count (p less than 0.001), NPM1 and FLT3-ITD mutations (p=0.027). DNMT3A mutations were conserved through xenotransplantation in immunodeficient mice. No difference in outcome between DNMT3A exon 23+ and DNMT3A exon 23- patients was found even if the results were stratified by NPM1 or FLT3-ITD status. However, DNMT3A exon 23+ patients had better median DFS (not reached vs 11.6 months, p=0.009) and OS (not reached vs 14.3 months, p=0.005) as compared to DNMT3A exon 23- patients when treated with idarubicin, whereas patients treated with daunorubicin had similar outcome regardless the DNMT3A status. This study shows that DNMT3A mutations have no impact on outcome but could be a predictive factor for response to idarubicin and thus, could have a direct influence in the way AML patients should be managed.


IntroductIon
Acute myeloid leukemia (AML), most widespread acute leukemia in adults, is characterized by clonal proliferation of oncogenic event-transformed hematopoietic stem cells or progenitors.Despite a high rate of complete remission after treatment with intensive chemotherapy combining cytarabine and anthracyclines in schemas that have little changed during the past 30 years, relapse rates are very high, resulting in a poor outcome in most cases with 5-year overall survival of 40% in younger adults and only 10-15% in elderly patients [1].Nevertheless, therapeutic results drastically vary regarding the main prognostic factors including age, performance status, cytogenetic and molecular abnormalities.Indeed, the knowledge of molecular basis of AML has considerably increased in the past few years mostly through the identification of recurrent mutational events occurring in a substantial number of patients [2].
The leukemic clone emerges from normal hematopoietic stem cells or more mature myeloid progenitors after acquisition of gene mutations affecting cell differentiation and self-renewal (such as the fusion genes PML-RARA, AML1-ETO, CBFb-MYH11 or point mutations targeting the functions of CEBPA, RUNX1, MLL or NPM1), cell signaling (FLT3, RAS or KIT), epigenetic machinery (TET2) and cell metabolism (IDH1 or IDH2) [3][4].These mutations particularly occur in AML without chromosomal abnormalities detected (normal karyotypes) and represent a major issue in the clinical management of patients since they could provide targets for both therapy (i.e., tyrosine kinase inhibitors in FLT3-ITD positive AML) and molecular monitoring of residual disease.However, the major impact of these mutations resides in the stratification of consolidation therapy once complete response is achieved following induction chemotherapy.It is now generally accepted that patients with normal karyotype and favorable genotype (i.e., NPM1 or CEBPA mutations without FLT3-ITD mutations) are no longer referred to allogeneic hematopoietic stemcell transplantation (HSCT) in first complete response since their outcome with chemotherapy alone is similar to those receiving HSCT [5].In addition, patients with an unfavorable genotype are more readily directed to allograft from unrelated donors or cord blood units since their prognosis with chemotherapy alone is dismal.Thus, FLT3-ITD, CEBPA, NPM1 mutations are now part of the initial work up of all AML patients with intermediate cytogenetic risk who undergo intensive treatments [6].However, newly discovered mutations such as isocitrate dehydrogenase 1 and 2 (IDH1/2) mutations have been already proposed to refine this molecular stratification although their prognostic impact remains to be definitely established [7][8].
Even more recently, by sequencing the genome of leukemic cells from a patient with AML, Ley et al have detected somatic mutations in the gene of a DNA methyltransferase (MTase), DNMT3A [9].DNMT3A is a member of the DNA MTases family including DNMT1, DNMT2, DNMT3A and DNMT3B that are involved in the methylation of CpG islands [10].The hypermethylation of CpG islands that contributes to the downregulation of gene expression, and notably of tumor suppressor genes, is also a hallmark of AML [11].In a series of 281 de novo AML, DNMT3A mutations were found in 22% of cases, clustering in intermediate-risk cytogenetic group and associated with a poor outcome.Most of the DNMT3A mutations (60%) found in AML samples were missense mutations localized on the R882 amino acid of the MTase domain.It is noteworthy that R882 mutations were significantly associated with a high white blood cell count as compared with other DNMT3A mutations.The aim of our study was to confirm the prognostic impact of R882 DNMT3A mutations in a series of 288 AML patients treated in our institution and to evaluate its screening usefulness in the medical diagnosis process, in a similar way to KIT exon 17 screening.

nod/scId mice xenograft
Adult NOD/SCID mice (6-8 weeks old) were sublethally irradiated with 250 cGy of total body irradiation 24 h before injection of leukemic cells.Leukemia samples were thawed at room temperature, washed twice and suspended in PBS at a final concentration of 1-2 million cells per 200 µL of PBS per mouse for IV injection.Daily monitoring of mice for symptoms of disease (ruffled coat, hunched back, weakness, or reduced motility) determined the time of killing for injected animals with signs of distress.If no sign of distress was seen, mice were analyzed 12 weeks after injection except as otherwise noted.For assessment of leukemic engraftment, NOD/SCID mice were humanely killed in accordance with IACUC protocols.Bone marrow (mixed from tibias and femurs) and spleen were dissected in a sterile environment, flushed in PBS and made into single cell suspensions for analysis by flow cytometry (FACS Calibur, BD Biosciences, San Jose, CA USA).

Comparisons of patient characteristics (covariates)
were performed using the Mann-Whitney test for continuous variables and the Fisher's exact test for categorical variables.Complete remission (CR) was defined following Cheson criteria [14].In univariate analysis, covariates associated with response to induction therapy or outcome were identified using Fisher's exact test then included in a multivariable logistic model.Overall survival (OS) and disease-free survival (DFS) rates were measured from the date of diagnosis until death and from the date of CR until death or relapse, respectively.Patients alive were censored at the time of last contact.OS and DFS were estimated by the Kaplan-Meier method and compared using the log-rank test.All calculations were

DNMT3A exon 23 mutations are associated with FAb M4/M5 subtypes and a higher Wbc count
The characteristics of intermediate-risk patients according to DNMT3A mutational status are listed in table 1.A bias toward monocytic differentiation (FAB groups M4 and M5) was identified in DNMT3A exon 23+ samples (p<0.0001).A strong association was found between DNMT3A exon 23+ and a higher white blood cell count (p=0.001).Platelet count, hemoglobin concentration and marrow blast percentage at diagnosis were similar in DNMT3A exon 23+ and -groups.
No relationship was identified between DNMT3A exon 23 mutations and CEBPA mutations, IDH1 R132, IDH2 R140 and R172 mutations, FLT3 tyrosine kinase domain or KIT exon 17 mutations.The significant relationship between DNMT3A exon 23 mutations and NPM1 mutations was independent of the FLT3-ITD mutation status (p<0.001 in FLT3-ITD+ or FLT3-ITD-in both groups).In contrast, the relationship between DNMT3A exon 23 mutations and FLT3-ITD mutations was lost in the NPM1 mutation subgroups (p=0.700 in NPM1c -and p=0.043 in NPM1c + patients), suggesting that the relationship between DNMT3A and NPM1 mutations was stronger than the relationship between DNMT3A and FLT3-ITD mutations.

DNMT3A exon 23 mutations are conserved in xenograft mice models
We next analyzed the mutational status of nine primary AML samples which showed engraftment capacities in xenograft NOD/SCID mice model.The presence of FLT3-ITD (seven out of nine, data not shown) correlated with the ability to engraft in these mice as it has been shown earlier [16][17][18].Furthermore, we have also observed that four of these nine specimens carried the DNMT3A exon 23+ mutations that were associated with FLT3-ITD and NPM1 mutations (Table 2).For these four samples, we also analyzed their post-transplantation mutational status and found that DNMT3A exon 23+ mutations were conserved in NOD/SCID mice.Overall, these data show a stability of the DNMT3A mutations in AML engrafted mice and suggest a preferential engraftment of primary AML specimens with triple FLT3-ITD/NPM1/DNMT3A mutations in immunodeficient mice.

AML patients with DNMT3A exon 23 mutations may benefit from idarubicin as compared to daunorubicin
As we did not find any prognostic impact of DNMT3A exon 23+ mutations in contrast to previous studies, we tried to find out whether treatments received by DNMT3A exons 23+ patients could explain these differences.[9;19-20] Because patients who were older than 60 received idarubicin and lomustin but also because they were much less frequently allografted, we focalized our analysis on younger patients (60y or less).The characteristics of patients according to the type of anthracycline used at induction are reported in table 3.There was no difference between patients treated by daunorubicin and those treated by idarubicin in terms of DFS (median DFS, 22.7 months for DNR vs 22.8 months for Ida, p=0.90) and OS (median OS, 28.5 months for DNR vs 24.4 months for Ida, p=0.39).Analysis of covariates associated with both DFS and OS are shown in table S1-S2.However, there was a significant impact of DNMT3A exon 23 mutations in patients treated with idarubicin.DNMT3A exons 23+ patients had better DFS (not reached vs 11.6 months, HR 3.3 , 95% CI, 1.34-8.11,p=0.009) and OS (not reached vs 14.3 months, HR 3.1, 95% CI, 1.39-6.71,p=0.005) as compared to DNMT3A exons 23-patients when treated with idarubicin whereas patients treated with daunorubicin had similar outcome regardless the DNMT3A mutational status (figure 3).Conversely, the outcome of patients with NPM1+/FLT3wt genotype was not impacted by the type of anthracycline used in induction (Figure S1).In patients who received allo-SCT (n=53), there was a trend for improved outcome in DNMT3A exon 23+ patients (DFS and OS medians not reached) as compared to DNMT3A exon 23-patients (median DFS, 19.1 months, p=0.15; median OS, 29.2 months, p=0.1) but the difference did not reach statistical significance (figure 4).To better address the effects of idarubicin vs. daunorubicin along with other variables in DNMT3A mutated patients, we performed univariate and multivariate analysis.As shown in table 4, idarubicin had an independent favorable prognostic effect on OS (HR 0.27, 95% CI, 0.08-0.97,p=0.046) when considering age, karyotype, NPM1/FLT3wt genotype, WBC and allo-SCT in DNMT3A mutated patients.

dIscussIon
In accordance with the first study assessing DNMT3A mutations in AML, we confirm here the high prevalence of DNMT3A exon 23 mutations in a series of 288 de novo adults AML [9].We have also observed that these mutations were exclusively found within the intermediaterisk cytogenetic group, associated with M4/M5 FAB subtypes and hyperleukocytosis.At the molecular level, DNMT3A exon 23 mutations were strongly associated with NPM1 mutations and to a lesser extent with FLT3-ITD mutations whereas no correlation was found with CEBPA, KIT or IDH1/2 mutations.
However, in contrast with previous studies, we failed to show any prognostic impact of DNMT3A exon 23 mutations in AML with intermediate-risk cytogenetic [9;19-20].We acknowledge that our patients had been treated over a long period of time (nine years) and this could have potentially introduced some bias.Indeed, two major changes in the management of AML patients had been undertaken in our center during this period of time: the introduction of primary prophylaxis of invasive fungal infections using azoles (voriconazole from 2003 to 2008, posaconazole thereafter) [21] and the stratification of allo-HSCT indication according to the new molecular classification of intermediate-risk AML using CEBPA, FLT3 and NPM1 mutations.Indeed, patients with wild type FLT3 and CEBPA or NPM1 mutations were no longer allocated to allo-HSCT in first CR since the results published by Schlenk et al. [5].However, since these two measures were applied in 2003 and 2008 respectively, their Allogeneic stem-cell transplantation impact on outcome, if any, could have been negligible in the whole cohort of patients.Furthermore, our induction and consolidation regimen have little changed for fifteen years, particularly the dosing of anthracyclines, with daunorubicin always given at 60 mg/m 2 /day for three days at time of the induction chemotherapy.By contrast, the doses of anthracyclines are not fully described in the study of Ley et al, in which several induction regimen were used, some of them using the infra optimal dose of 45 mg/m 2 /d [22][23][24].Moreover, some patients received hypomethylating agents or lenalidomide which generally induce fewer responses than intensive chemotherapy in AML patients [25][26].Also, complete response and early death rates were not mentioned and this could be of interest as patients with R882 DNMT3A mutations usually have a high white blood cell count, a recognized risk factor for early death.In our study, DNMT3A exon 23 mutations did not impact on both complete response and early death rates.Moreover, we found that patients with DNMT3A exon 23 mutations could specifically benefit from idarubicin although the small number of patients in our study requires confirmation in larger cohorts.The strong impact of DNMT3A mutations on the response to idarubicin as compared to daunorubicin is quite unexpected.However, it has been recently shown that DNMT3A could play a role in anthracyclines-induced apoptosis of colorectal cancer cells.Indeed, the expression of DNMT3A is upregulated at apoptosis-inducing concentrations of doxorubicin and involved in p21 repression thereby blocking senescence [27].Whether this program is induced by idarubicin but not daunorubicin in DNMT3A mutated/haploinsufficient cells remains to be determined.Alternatively, DNMT3A mutations could specifically impact on the expression of genes involved in idarubicin metabolism as compared to daunorubicin.The broader spectrum of activity of idarubicin has been attributed to increased lipophilicity, cellular uptake and improved stabilization of a ternary drug-topoisomerase II-DNA complex [28].Thus, the discrepancy in therapeutic outcome between our series and those previously described could be due to differences in treatment intensity or altered metabolism of anthracyclines.Several clinical trials have already assessed the impact of daunorubicin dose intensification or compared idarubicin vs. daunorubicin in large cohorts of AML patients.Reassessing results of these controlled studies in light of DNMT3A mutational status could easily confirm or not our preliminary observations [12;22-23;29-30].It remains also to be determined if DNMT3A mutations affect in a similar way the metabolism of other compounds that intercalates DNA and inhibits topoisomerase II such as the new quinolone derivative vosaroxin which is currently assessed in AML [31].
In addition, we were also able to demonstrate for the first time that DNMT3A mutations are also found in table 4: univariate and Multivariate Analysis for dFs and os in DNMT3A exon 23 mutated patients.Analysis of covariates associated with DFS and OS.P of the univariate analysis is the p value of the Log rank test.HR is the value of the hazard ratio.95% CI is the 95% confident interval of the hazard ratio.Data of AML patients with DNMT3A exon 23 mutations were complete and were included in the Cox proportional-hazards regression.NPM1: Nucleophosmin; FLT3-ITD: internal tandem duplication of the FLT3 gene; Allo-SCT: Allogeneic stem-cell transplantation; Ida: idarubicin; WBC: white blood cell count.[32].This suggests that engraftment properties of AML samples should be assessed in light of specific molecular lesions.

Univariate analysis
It has been shown that engraftment in xenograft NOD/ SCID-IL2Rγc -/-mice model did not correlate with French-American-British subtype or cytogenetic abnormalities but with the presence of FLT3-ITD [17].More recently, Sarry et al. have observed that a high engraftment level occurred within human primary AML samples carrying at least two mutations (seven out of nine samples had FLT3-ITD and NPM1 mutations) and that these mutations were strongly conserved in the different leukemic stem cell populations sorted as well as still present after the first transplantation into NOD/SCID-IL2Rγc -/-mice [18].In the present study, we show that the DNMT3A mutation is also conserved through the NOD/SCID xenograft model and is associated with higher leukemic engraftment level in cooperation with FLT3-ITD and NPM1 mutations, suggesting that this mutation occurs in the early leukemogenic events and belongs to the main leukemic clone over the course of the disease progression.In summary, our data confirm that DNMT3A exon 23 mutations can be easily evaluated in medical practice and represent a frequent molecular event in intermediate-risk AML associated with NPM1 and FLT3-ITD mutations but have no clear impact on disease-free and overall survival.However, although this finding needs to be confirmed in largest cohort of patients, DNMT3A mutations could be a predictive factor for response to idarubicin and, thus could have a direct influence in the way AML patients should be managed.

Figure 1 :
Figure 1: Flow chart of AML patients treated by intensive chemotherapy between 2000 and 2009.From 2000 to 2009, 457 consecutive patients with de novo AML were treated by intensive chemotherapy.Patients with acute promyelocytic leukemia or with secondary AML were excluded from this study.

table 1 : characteristics of AML patients with intermediate-risk according to DNMT3A exon 23 mutation.
WBC, white blood cell count; FAB, French American British; SCT, stem-cell transplantation.

Multivariate analysis
leukemic samples engrafted in immunodeficient mice.Although we could not demonstrate that the DNMT3A exon 23 mutation is a surrogate marker of engraftment in NOD/SCID mice, it should be noted that all mutated samples tested recapitulated the initial disease in mice.By comparison, it has been shown that only 50% of intermediate-risk cytogenetic samples display engraftment properties in NOD/SCID mice