Landmark Analyses of DNMT3A Mutations in Hematological Malignancies

Abstract 407

CpG methylation is an epigenetic modification that is important for cellular development. The DNMT3A gene, located on chromosome 2p23.3, encodes for a DNA methyltransferase and plays a central role in de novo CpG methylation. Recently, DNMT3A has been reported to be mutated in 22% of AML and 8% of MDS (Ley et al., N Engl J Med, 2010; Walter et al., Leukemia, 2011). Further, DNMT3A mutations were observed to be associated with a short overall survival in both diseases, respectively. In order to determine the role of DNMT3A mutations in leukemia we investigated two different entities by next-generation sequencing: 145 AML patients and 83 cases harboring a T-cell acute lymphoblastic leukemia (T-ALL). We applied an amplicon based deep-sequencing assay (454 Life Sciences, Branford, CT) in combination with the 48.48 Access Array technology (Fluidigm, South San Francisco, CA). The peripheral blood or bone marrow samples were obtained from untreated patients. The AML cohort was restricted to cases with normal karyotype (CN-AML). 87/145 (60%) cases were specifically selected to be wild-type for NPM1, FLT3-ITD, CEBPA, and MLL-PTD, whereas 58/145 (40%) samples were mutated in NPM1 (n=33) or double-mutated in NPM1 and FLT3-ITD (n=25). In our cohort of AML cases without mutations in NPM1, FLT3-ITD, CEBPA, and MLL-PTD, we observed a DNMT3A mutation frequency of 17.2% (15/87 cases). The DNMT3A mutation rate in the NPM1 mutated/FLT3 wild-type cases (16/33, 48.5%, P=0.001) and NPM1/FLT3-ITD mutated cases (19/25, 76%, P<0.001) was significantly higher, confirming the association of DNMT3A mutations with NPM1 and FLT3-ITD mutations that had been reported previously (Ley et al.). Interestingly, also in the cohort of T-ALL we detected patients that carried a DNMT3A mutation (16/83, 19.3%), which is very similar to the mutation frequency in AML, and has not been described yet. To further address the biology of DNMT3A mutations in acute leukemias we combined the AML and T-ALL cohorts and identified in total 31 distinct missense mutations in 65 patients (49 AML, 16 T-ALL). Most frequently, amino acid R882 located in exon 23 was mutated (n=29 cases). In addition, we identified 7 frame-shift alterations, 5 nonsense and 2 splice-site mutations. Moreover, 9 of the 65 mutated cases had two independent mutations. Focusing on AML, only three (6.1%) of the 49 DNMT3A-mutated cases were observed to harbor two different mutations concomitantly. In contrast, in the cohort of T-ALL we detected two different mutations in 6/16 (37.5%, P=0.003) cases. Further, in the cohort of AML, no homozygous mutation was detected, however, in the T-ALL group, two cases harbored a homozygous mutation. Therefore, only 3/49 AML (6.1%) cases, but 8/16 T-ALL (50%) cases showed biallelic mutation status (P<0.001). With respect to overall survival, no association was seen in the complete cohort of CN-AML cases (n=145). After limiting this cohort to the cases without mutations in NPM1, FLT3-ITD, CEBPA and MLL-PTD (n=87), an inferior survival was observed for DNMT3A-mutated patients as compared to DNMT3A wild-type patients (n=15 vs. n=72; alive at 2 years: 27.9% vs. 56.6%; P=0.048). Remarkably, also in the cohort of T-ALL a worse survival for patients with DNMT3A mutations was seen which has not been reported thus far (n=13 vs. n=64; alive at 1 years: 28.6% vs. 80.9%; P=0.001). Subsequently, we were interested whether gain-of-function mutations of the DNMT3A gene were associated with trisomy 2 and acquired uniparental disomy (aUPDs) of the short arm of chromosome 2 where DNMT3A is located. As such, we investigated 9 cases harboring a trisomy 2 (AML n=4, MDS n=4, and CMML n=1) and one MDS patient harboring an aUPD 2p, as confirmed by SNP microarray analyses (SNP Array 6.0, Affymetrix, Santa Clara, CA). Not all, but 3/9 cases with trisomy 2 harbored a DNMT3A mutation (one AML, MDS, and CMML case each), suggesting that duplication of DNMT3A mutations can enhance the effect of the mutation. Moreover, the single case with aUPD 2p also showed a mutation, further suggesting that LOH leading to loss of the wild-type DNMT3A may be another mechanism of disease leading to progression of leukemia. In conclusion, we here report on a high mutation rate of DNMT3A in both AML and T-ALL and independently confirmed an inferior overall survival in these two entities, respectively. This indicates a significant role of DNMT3A alterations in myeloid as well as in lymphoid neoplasms.