Clonal genetic evolution at relapse of favorable-risk acute myeloid leukemia with NPM1 mutation is associated with phenotypic changes and worse outcomes

Acute myeloid leukemia (AML) is a dynamic disease caused by accumulating, somatically acquired driver mutations generating branching competing clones.1 In favorable-risk AML, high resolution genomic profiling by single nucleotide polymorphism array analysis of paired diagnostic-relapse NPM1 and CBF AML samples revealed increased genomic complexity at relapse but most patients retained founding mutations.32 Furthermore, it has been extensively reported that phenotypic changes are commonly found at relapse in AML patients. It seems plausible that clonal evolution could be reflected in the phenotypic shifts of AML blast cells found at relapse, although the correlation with genetic clonal evolution has not been established.742 The aim of our work was to determine the patterns of genetic clonal evolution occurring from diagnosis to relapse in favorable-risk AML patients by tracking the kinetic behavior of the most frequent co-mutations in paired samples and correlating these with the occurrence of phenotype shifts on blast cells and with the clinical outcome.

We included a total of 26 patients with favorable-risk AML (non-promyelocytic), according to European LeukemiaNet criteria, who relapsed after achieving complete remission. These patients were treated with the intensive chemotherapy schedules standard at the time of diagnosis and experienced a relapse after a median of 17.5 months (range, 4-252) (Table 1). As controls, we studied seven NPM1 AML patients (median age: 46.7 years; range, 22-69) who achieved sustained complete remission after treatment with a median follow-up of 24 months and no evidence of leukemia relapse at last follow-up.

Table 1.Clinical and biological data of 26 patients with favorable-risk acute myeloid leukemia.

Bone marrow-derived genomic DNA was obtained from paired diagnostic-relapse samples. Details of the methods are available in the Online Supplementary Material. At diagnosis, among 16 NPM1 AML patients we found three cases with DNMT3A (18.7%: two R882H and one new mutation, c.2705_2706delTC), two cases with IDH1 (12.5%), two cases with IDH2 (12.5%) and one case with FLT3-TKD (c.2503>T, at low ratio: 0.18). No mutations in RAS and TP53 were found. In contrast, seven non-relapsing NPM1 controls showed less genetic complexity: we detected only one case with IDH1 (14.3%) and no DNMT3A was detected. With regards to CBF-AML (n=8), we detected one case with DNMT3A (12.5%) and one with C-KIT (12.5%) and no mutations were found in two CEBPA patients. At the time of relapse, two patterns of genetic findings were observed: ‘no clonal evolution’, with persistence of mutations of the original founding clone, and ‘clonal evolution’, with changes in the gene mutation profile. No clonal evolution was found in 20 patients (77%): ten from the NPM1 AML group (62.5%), all eight of the CBF-AML group (100%) and both of those with CEBPA-AML (100%). In ten NPM1 AML patients, IDH and DNMT3A remained stable with the same variant allelic fraction (VAF) and no acquisition of TP53 was detected. In the CBF-AML and CEBPA groups, DNMT3A and CKIT remained stable at relapse and acquisition of TP53 was not observed (Figure 1A). Interestingly, the second pattern, clonal evoluation, was found in the remaining six patients (23%), all of who were in the NPM1 AML group (36.5%): loss of NPM1 was confirmed in four cases, evolution of DNMT3A in two cases [one R882H (VAF of 7.1% to 49.1%) and one new mutation p.D876Y (VAF of 0 to 48.4%)], one patient acquired FLT3-ITD and one patient lost a previously present FLT3-TKD. Absence of these newly acquired mutations (2 DNMT3A and 1 FLT3-ITD) in diagnostic samples was confirmed by next-generation sequencing as well as by reverse transcriptase polymerase chain reaction analysis (Figure 1A). By quantitative pyrosequencing analysis we demonstrated that both new DNMT3A mutations (c.2705_2706delTC,p.F902fs from patient 7 and c.2626G>T,p.D876Y from patient 10) were only found in leukemic samples and were not present in bone marrow samples obtained from patients in complete remission or in healthy donors (Online Supplementary Figure S1). From in silico studies, both mutations could alter normal function of native DNMT3A decreasing the activity of DNA methylation (Online Supplementary Figure S2).

Figure 1.Clonal behavior in a series of 26 relapsing patients with favorable-risk acute myeloid leukemia. (A) Mutation analyses in paired diagnostic (D) and relapse (R) samples in 26 patients. Each column represents an individual patient. Colored bars indicate the presence of a mutation, blank bars represent wild-type for the specific gene and beige bars indicate that data are not available. *VAF 0%; **VAF: 48.4%; #VAF 7.1%; ## VAF 44.8%. (B) Immunophenotypic patterns in paired diagnostic (D)/relapse (R) samples from 26 patients. Each column represents an individual patient. Colored bars indicate strong, dim and negative CD antigen detection for each marker. “Strong” means greater than 10⁴, “dim” means between 10³ and 10⁴ and “negative” means lower than 10³. Blank bars indicate missing data.

For immunophenotypic analyses, at least 30,000 leukemic events were acquired, mostly in FACSCalibur or FACSCanto II dual-triple laser flow cytometers, and list modes files were analyzed with CellQuest™, FACSDiva™ or Paint-a-Gate software (Becton Dickinson Biosciences). Multidimensional analyses of immunophenotypes obtained at diagnosis and relapse were performed using the File merge and Automatic Population Separator functions of Infinicyt software (Cytognos SL, Salamanca, Spain). At diagnosis, most NPM1 AML patients displayed strong CD33 and CD13 expression (93.8% and 62.5%, respectively) with strong CD117 and CD34 expression in 43.8% and aberrant CD56 in 12.5%. CBF-AML blast cells expressed CD117 strongly in all cases, CD34 in 87.5%, CD56 and aberrant CD19 in 12.5% of cases. At the time of relapse, complete stability in the expression of all markers was observed in 14 patients (53.8%). By contrast, phenotypic profile evolution (defined as a significant modification of intensity in at least one marker) was confirmed in 12 patients (46%): eight of the 16 NPM1 patients (50%) and four of the group of ten with no NPM1 (40%). More frequently shifted expression was observed in CD15 (58.3% of patients), CD117, CD34 and CD56 (41.6%), CD7 and CD13 (33.3%), CD11b, CD4, CD33 and CD14 (25%) (Figure 1B). When comparing the incidence of phenotypic shifts in both genetic groups, we found that a significant percentage of patients with the ‘no clonal evolution’ pattern still displayed phenotypic shifts (8 out of 20; 40%) and this percentage was even higher among those showing a pattern of clonal evolution (4 out of 6; 66.7%), although the difference was not statistically significant (P=0.3). Altogether, 12 (46.1%) favorable-risk AML patients relapsed maintaining the same mutational and phenotypic profiles. A representative case of phenotypic shift is shown in Online Supplementary Figure S3.

Finally, we analyzed the impact of genetic patterns and phenotypic shifts on outcomes. At the time of analysis 11 patients were alive and in complete remission. The median follow-up after leukemia relapse was 55 months (range, 16.3-96.3) and probability of overall survival was 40.7% ± 10 for the overall series (Figure 2A). Salvage rescue treatment included allogeneic stem cell transplantation after re-induction chemotherapy (n=12; 46.1%) and intensive chemotherapy ± azacitidine (n=11; 42.3%) whereas three patients received only supportive care. Patients who underwent allogeneic stem cell transplantation had a statistical significantly higher probability of overall survival (82.5 ± 11.3% versus 7.1 ± 6.9, P<0.01). The median time from complete remission to relapse was shorter in patients with clonal evolution [12.6 (range, 6-67) months versus 18.5 (range, 6-252) months] than in the ‘no-clonal evolution’ group. Considering only the NPM1 group (n=16), 66.7% of patients showing clonal evolution had undergone allogeneic stem cell transplantation at first complete remission (4 out of 6; 66.7%) compared to 20% (2 out of 10) in the ‘no-clonal evolution’ group (Online Supplementary Table S1). Importantly, favorable-risk AML patients with no clonal evolution at relapse had a significantly higher estimated probability of overall survival compared to that of the group with clonal evolution (48.5 ± 11.5% versus 16.7 ± 15.2%, P=0.003) with a longer, mean estimated overall survival of 53.6 months (95% CI: 34.8;72.4) versus 8 months (95% CI: 0;19.3), respectively. Of note, overall survival probability was identical (48.5% ± 16.3) for AML patients with or without NPM1 within the ‘no-clonal evolution’ group (Figure 2B). In the multivariate analysis, only clonal evolution remained a significant adverse factor and allogeneic stem cell transplantation as salvage treatment of relapse as a favorable clinical factor (Online Supplementary Table S2).

Figure 2.Overall survival. (A) Overall survival of the whole series (n=26). (B) Overall survival comparing patients with NPM1 and clonal evolution (red), NPM1 without clonal evolution (green) and CBF/CEBPA without clonal evolution (blue).

In this study, we addressed genotypic and phenotypic clonal behavior in a series of 26 relapsing favorable-risk AML patients. Our study demonstrated that the main scenario for leukemia relapse is the re-emergence of a founder clone with no clonal evolution (77% of cases), although 40% of such cases displayed phenotypic changes. This finding suggests that conventional chemotherapy protocols may not be able to achieve complete eradication of the founder AML clone, which is capable of regenerating the bulk of leukemic blasts after a variable period of time. This is in agreement with previous reports of genomic profiling studies by single nucleotide polymorphism arrays in AML series including all-risk subtypes or NPM1 cases82 demonstrating increasing genomic complexity at relapse, which showed significantly worse outcomes32 but maintenance of a common ancestral founder clone. Our data suggest the persistence of a rare subset of leukemic stem cells in favorable-risk AML after achievement of complete remission. These leukemic stem cells are capable of remaining quiescent for long periods,9 such as in patient 3 who relapsed with the same genetic and phenotypic profile 20 years after achieving complete remission. In our series, all CBF-AML showed mutational stability, despite displaying phenotypic changes in 40% of cases. By contrast, clonal evolution was present in 36.5% of NPM1 AML and 66% of these cases also displayed phenotypic shifts. In our series, loss of NPM1 at relapse was the most frequent genetic evolution, followed by the acquisition of DNMT3. Loss of NPM1 at relapse was confirmed in four cases (25% of 16 NPM1 cases). Three of them had undergone allogeneic stem cell transplantation at first complete remission and received azacitidine and/or chemotherapy with dismal outcome. These cases can plausibly be considered as “secondary therapy-related” or “clonally unrelated” AML. Importantly, our findings suggest that monitoring for minimal residual disease can be hampered by frequent phenotypic changes and also by the possibility of NPM1 losses. Minimal residual disease monitoring by multiflow cytometry or quantitative reverse transcriptase polymerase chain reaction for a genetic marker11 can, therefore, be complementary and parallel monitoring could be quite useful to avoid false-negative minimal residual disease results, providing useful biological information to trace clonal evolution.1312

Strikingly, in our series, DNMT3A evolved in two patients, one of whom had concurrent loss of NPM1. These findings, also in accordance with those reported by Krönke et al.,2 point out the kinetic complexity of the interactions of DNMT3A and NPM1 in AML patients at relapse, in whom new mutations in this epigenetic modifier occur as a “late event” in some instance.102

In conclusion, a comprehensive assessment of genetic and phenotypic features at relapse in favorable-risk AML provides useful biological information and could have important prognostic implications.