Background. The occurrence of clonal hematopoiesis increases with age and is associated with increased risk of hematologic malignancies (Jaiswal et al, NEJM 2014). Recent work identified pre-leukemic mutations in DNMT3A, ASXL1 and TET2 in AML that persist in remission and are not associated with survival (Jongen-Lavrencic et al. NEJM 2018). NPM1 mutation (NPM1mut) identifies a WHO AML entity which accounts for ~30% of all AML and is associated with younger age. Mutant DNMT3A was shown to be the most frequent co-mutation in NPM1mut AML, which is counterintuitive given its known increase with age. We aimed to dissect the age association of known co-mutations in NPM1mut AML vs wild type NPM1 (NPM1wt) AML in a large cohort.

Methods. Between 2005 and 2018, 9010 patients (pts) were diagnosed with AML by cytomorphology at our institution. For 6990/9010 pts (78%) NPM1 was studied at diagnosis by gene scan analysis, melting curve analysis or next-generation sequencing (NGS, NPM1 testing performed: NPM1p). Samples were sequenced for AML-related genes, including: ASXL1, DNMT3A, TET2, IDH1/2, WT1, FLT3, NRAS, PTPN11, SF3B1, SRSF2, TP53, CEBPA, RUNX1. At least two genes other than NPM1 were sequenced in 6872/6990 (98%) cases. FLT3-ITD was analyzed by gene scan in 6851/6990 (98%) pts. For 6488/6990 NPM1p pts (93%), chromosome banding analysis (CBA) was performed. The NPM1p cohort did not differ from the whole AML cohort nor from the general AML population in terms of age distribution (z-score comparison). Differences across age groups were evaluated by chi-square testing and. ANOVA tests were applied on continuous variables.

Results. In the NPM1p cohort, 2097/6990 pts (30%) were NPM1mut. As expected, NPM1mut was more common in younger pts (<60 yrs, 37%) compared to older pts (27%, odds ratio, OR=1.5, p<0.0001). NPM1 mutational frequency peaked in the age group from 30 to 59 years (yrs, 38%), it was significantly lower in younger (20-29 yrs, 18%, p<0.0001) and older pts (60-79 yrs, 28% and >80 yrs 21%, p<0.0001). The most commonly co-mutated genes were: DNMT3A (47%), FLT3-ITD (41%), TET2 (23%) and IDH2 (only R140, 18%), in line with previous reports (Ivey et al. NEJM 2016).

As expected, the common CHIP-related genes ASXL1 and TET2 were more frequently mutated in pts ≥60 yrs vs <60 yrs (ASXL1: 4% vs 0.5%, OR= 8.3, p=0.001; TET2: 29% vs 16%, OR=2.1, p<0.0001). Interestingly, DNMT3A mutations were significantly more frequent in younger pts (52% vs 43%, OR=1.5, p<0.0001). Regarding mutation subtypes, DNMT3A-R882 was more frequent than non-R882 in younger pts compared to the older ones (68% of all DNMT3A+ cases in <60 yrs vs 50% in ≥60 yrs, OR=2, p<0.0001). The mean age at diagnosis was 58 yrs (range: 22-86 yrs) for R882 vs 63.3 yrs (26-88 yrs) for non-R882 (p<0.0001). Other gene mutations associated with younger age included: WT1 (9% vs 2%, OR=4, p<0.0001), NRAS (16% vs 11%, OR=1.5, p=0.04) and PTPN11 (11 vs 1%, OR= 8.5, p=0.024).

To determine the clonal structure, we compared the variant allele frequency (VAF) of NPM1 and DNMT3A±5%, 131/164 (80%) had a DNMT3A VAF higher than NPM1 and only 3 (2%) a DNMT3A VAF lower than NPM1. This suggests that DNMT3A is the first hit and NPM1mut is acquired during clonal evolution in the majority of cases. VAF clonal relationships were constant across age groups.

For 1904/2097 NPM1mut pts with available CBA analysis (91%), 250/1904 (13%) had an aberrant karyotype which was significantly associated with older age in NPM1mut AML (8% in 20-29 yrs old pts vs 20% in ≥80 yrs old pts, p<0.0001; age <60yrs: 10% vs age ≥60 yrs: 15%, p=0.0043).

As a control, we analyzed the age dependency of mutations in NPM1wt AML pts (4893/6990): ASXL1, TET2, DNMT3A (both R882 and non-R882), IDH1 and IDH2R140 mutations increased continuously with age, while FLT3-ITD was more frequent in young pts (16% vs 9%, p<0.0001).

Conclusion. In a large AML cohort, prototypic CHIP mutations such as ASXL1 and TET2 increase with age regardless of NPM1 status, whereas in NPM1mut AML mutated DNMT3A was a phenomenon of the younger patient (essentially due to the R882 subtype). VAF comparison indicates that DNMT3A is a pre-leukemic mutation in NPM1mut AML and thus represents the first hit in NPM1mut AML pathogenesis. We conclude that NPM1mut/DNMT3Amut double mutated AML could represent a distinct biologic entity which needs further investigation.


Cappelli:MLL Munich Leukemia Laboratory: Employment. Meggendorfer:MLL Munich Leukemia Laboratory: Employment. Dicker:MLL Munich Leukemia Laboratory: Employment. Jeromin:MLL Munich Leukemia Laboratory: Employment. Hutter:MLL Munich Leukemia Laboratory: Employment. Kern:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Höllein:MLL Munich Leukemia Laboratory: Employment.

Author notes


Asterisk with author names denotes non-ASH members.