In AML Secondary to MDS NPM1 Mutations Are Late Events, Less Frequent, and Associated with a Different Pattern of Molecular Mutations Than in De Novo AML

Background: NPM1 mutations (mut) are considered the most frequent mutations in de novo acute myeloid leukemia (AML) and have been suggested as provisional entity in the WHO classification 2008. It has become clear that nearly all NPM1mut AML have additional mutations that may contribute to onset of AML by affecting different genetic pathways. However, it has not been evaluated yet whether the pattern of additional mutations varies between NPM1mut de novo AML and NPM1mut secondary AML (sAML) arising from a previous myelodysplastic syndrome (MDS).

Aim: To evaluate the 1) genetic pattern associated with NPM1mut de novo AML and sAML, 2) chronologic sequence of mutations from MDS to sAML.

Patients and Methods: 5,545 de novo AML and 504 sAML cases were analyzed during the last 9 years. The de novo cohort was comprised of 2,951 males and 2,594 females, median age was 65.7 years (y; range 17.5-93.1 y). The sAML cohort was comprised of 329 males and 175 females, median age was 71.7 y (range 29.3-91.8 y; p=0.004). All cases were analyzed for NPM1mut by a melting curve analysis. For more detailed analysis from these cohorts 359 NPM1mut de novo AML (162 male, 197 female; median age: 61.4 y, range: 17.8-88.0 y), and 21 sAML (12 male, 9 female; median age: 70.3 y, range: 44.2-87.4 y) were selected for mutation analysis in 13 different genes (ASXL1, CEBPA, DNMT3A, FLT3-ITD, FLT3-TKD, IDH1, IDH2, KRAS, MLL-PTD, NRAS, RUNX1, TET2, TP53, WT1). Paired samples from the diagnostic time points of both MDS and sAML, respectively, were available in all 21 sAML cases. For both time points an NPM1-specific quantitative real time PCR was performed in addition.

Results: First the overall frequency of NPM1mut was calculated from the total cohort. NPM1mut was more frequent in de novo AML (1,737/5,545 cases; 31.2%) than in sAML (67/504; 13.3%) (p<0.001). Frequencies for mutations in all other genes were calculated for the selected NPM1mut subcohorts only. In de novo AML DNMT3A was the most frequently mutated gene (204/359; 56.8%), followed by FLT3-ITD (n=157; 43.7%), TET2 (n=101; 28.1%), IDH2 (n=52; 14.5%), NRAS (n=50; 13.9%), FLT3-TKD (n=49; 13.6%), IDH1 (n=47; 13.1%), CEBPA (single mutated: n=30; 8.4%; no double mutated cases), WT1 (n=23; 6.4%), KRAS (n=16, 4.5%), ASXL1 (n=10; 2.8%) and RUNX1 (n=2; 0.6%). No mutations were detected in TP53 or MLL-PTD. In the sAML cohort of 21 NPM1mut cases the most frequent additional mutations were present in TET2 (n=12, 57.1%), followed by FLT3mut (7 ITD and 1TKD) (n=8, 38.1%), ASXL1 and DNMT3A (n=4, 19.0%, each) and each 2 (9.5%) in IDH1, IDH2, KRAS, NRAS, RUNX1 and WT1, respectively. Thus DNMT3Amut were significantly more frequent cooperating with NPM1mut in de novo AML (56.8% vs. 19.0%, p=0.001). In contrast, mutations in TET2 (57.1% vs 28.1%, p=0.005), ASXL1 (19% vs. 2.8%, p<0.001) and RUNX1 (9.5% vs. 0.6%, p<0.001) were more frequent in sAML. For none of the other mutations any significant difference between de novo and sAML was observed. Next, we evaluated the chronologic sequence of the emergence of the respective mutations by comparing the paired MDS and sAML samples. At MDS phase and at a sensitivity of 10^-7 the NPM1 mutations were undetectable in 6 patients and detectable at a very low level (0.01-1%) in 8 pts. In contrast, in 7 cases the NPM1mut was already detectable at a level of 5-100% (median: 10%). At MDS phase the median number of additional mutations was 1 (range: 1-4), at sAML it increased to 3 (range: 1-5). All 12 TET2mut and all 4 DNMT3Amut cases carried this mutation already at MDS phase, thus these two genes can be regarded as early events. ASXL1 was present in 3 cases at MDS and was gained at AML in 1 case. IDH mutations (n=5) were stable in 3 and gained or lost in 1 patient each. RAS mutations were gained in 3 and lost in 1case. FLT3mut (n=8) were never detected at MDS but gained in all cases at sAML stage and thus can be regarded as late events. Median time from diagnosis of MDS to transformation to sAML was 9.2 months (range: 1.6 - 33.6 months). Median time to transformation was shorter in cases with TET2mut (8.2 vs.16.8 months, p=0.026) than in TET2 wildtype cases. No impact on time to transformation was seen for the other mutations.

Conclusions: NPM1 mutations 1) occur less frequent in sAML than in de novo AML, 2) like FLT3mut are usually late events that drive transformation from MDS to sAML, 3) are frequently associated with TET2, ASXL1 and RUNX1 mutations in sAML whereas in de novo AML most frequently are accompanied by DNMT3A mutations.