1 INTRODUCTION

The human nucleophosmin 1 (NPM1) gene is multifunctional, including chromatin remodeling, ribosome biogenesis, genomic stability, regulation of tumor suppressors, and transcription factors.^1-3 Given its unique biological and clinical relevance, the 2016 revision of World Health Organization (WHO) classification of myeloid neoplasms and acute leukemia recognized NPM1^mut acute myeloid leukemia (AML) as a distinct entity.⁴ NPM1 mutation (NPM1^mut) is present in approximately one-third de novo AML, and up to ~60% in normal karyotype (NK) AML.^{5, 6} Approximately 40% of NPM1^mut cases coexist with the fms-like tyrosine kinase 3 (FLT3)-internal tandem duplication (ITD). In the latest NCCN guideline⁷ and 2017-European LeukemiaNet recommendation,⁸ NPM1^mut AML is divided into favorable- and intermediate-risk groups according to a presence or absence of concomitant FLT3-ITD and allele ratio (AR) levels. NPM1^mut AML with negative FLT3-ITD and FLT3-ITD low AR (FLT3-ITD^low) genotype is in the favorable-risk category. Yet, a considerable proportion of these patients experience relapse and resistance, ultimately shortening their survival. Even the outcome of NPM1^mut/FLT3-ITD⁽⁻⁾ patients still has considerable heterogeneity.^{9, 10}

Standard-dose (SD) cytarabine arabinoside (Ara-C) combined with anthracyclines (i.e., “7 + 3” scheme) is the most commonly used AML induction protocol. The literature of high-dose (HD) Ara-C used for AML chemotherapy was first published in 1985. Several randomized trials were performed; the majority failed to draw a conclusion of significant efficacy improvement, as reported during 1991–2011.¹¹ However, prolonged follow-up time subsequently identified subsets of patients benefiting from HD-Ara-C, like AML with t(8;21)¹² and Ras-pathway mutations.¹³ Furthermore, most of the earlier studies did not comprehensively integrate the genomics-based prognostication scheme. Thus, the conclusions could not reflect the concept of risk-adapted therapy using HD-Ara-C. Development of massively parallel sequencing methodology has increasingly emphasized the risk-stratification oriented mode of treatment.

It has reported that the cellular uptake of Ara-C was insufficient in FLT3-ITD⁽⁺⁾ K562 cells, with no significant changes of intracellular deoxycytidine kinase (DCK) and cytidine deaminase (CDA) responsible for metabolism of Ara-C, and transcription level of multidrug resistance protein 1 (MDR1) related to drug efflux.¹⁴ Another study described that FLT3-ITD decreased the expression of equilibrative nucleoside transporter 1 (ENT1) by upregulating hypoxia inducible factor-1 alpha subunit (HIF-1α), compromising Ara-C uptake by leukemic cells. In addition, FLT3-ITD also drives a specific Ara-C resistance mechanism via the downstream molecule RUNX3.¹⁵ In view of the linkage between FLT3-ITD⁽⁺⁾ with distinctive Ara-C insensitivity, we wonder whether increasing Ara-C in the induction period can improve outcome of NPM1^mut/FLT3-ITD⁽⁺⁾ populations. According to a review regarding dose-efficacy relationship of Ara-C, a level of ID-Ara-C in each treatment cycle in AML is sufficient, while HD-Ara-C seems to be of little significance.¹¹

In this study, we retrospectively analyzed the clinical characteristics of NPM1^mut patients and data of next-generation sequencing (NGS) mutation spectrum concerning 112 genes related to blood disease, synthetically including SD-Ara-C 100–200 mg/m² or ID-Ara-C 1000–2000 mg/m² induction as covariates. We aimed to refine patient subsets carrying potentially poor remission and prognostic characters from this relatively favorable AML subtype, to guide further risk-adapted treatment.

2 PATIENTS AND METHODS

3 RESULTS

3.1 Selection of patients for outcome analysis

We selected a total of 238 NPM1^mut AML patients (105 men and 133 women; median age, 49 years; range, 15–81 years). Among them, 203 NPM1^mut patients who met the following three criteria were included in the remission and prognosis analysis. First, low-dose and no treatment patients were excluded. Only patients who received SD- or ID-Ara-C containing induction regimen were included. As of the time of writing, patients attaining CR who were induced with one or two cycles had undergone at least one cycle of ID- or HD-Ara-C consolidation. Second, favorable- and adverse-risk karyotypes were excluded. Only NK and intermediate-risk abnormal karyotypes were included. Four patients lack of metaphase were screen by fluorescence in situ hybridization and 43 kinds of fusion transcripts, which ruled out recurrent chromosomal translocations. They were mostly likely NK or intermediate-risk abnormal karyotypes because of the very low probability of complex or monosomy karyotypes. The four patients also participated in the analysis. Third, NPM1^mut missense mutations were excluded. Only those with the insertions/deletions (indels) type were included. Additionally, among 10 NPM1^mut missense cases detected in our study, outcome analysis included two cases with NPM1^mut missense who each occurred concomitant with a type-A mutation. The remaining eight cases with NPM1^mut missense were excluded. The flowchart of the selected and excluded cases for the outcome analysis is presented in Figure 1.

3.5 NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁺⁾ is an unfavorable prognostic subset in NPM1^mut/FLT3-ITD⁽⁻⁾ patients

Given the significance of TET2 mutation, we compared clinical outcomes across the NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁻⁾, NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁺⁾, and NPM1^mut/FLT3-ITD⁽⁺⁾ subgroups. The NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁻⁾ subgroup was superior to NPM1^mut/FLT3-ITD⁽⁺⁾ subgroup in all endpoints (Bonferroni: cCR rate p = 7.5 E-05; EFS p = 4.29 E-7; OS p = 0.010; Figure 2A,B), and to the NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁺⁾ subgroup regarding cCR rate (Bonferroni p = 0.043) and EFS (Bonferroni p = 1.20 E-5; Figure 2A). There were no differences between NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁺⁾ and NPM1^mut/FLT3-ITD⁽⁺⁾ regarding all endpoints (Bonferroni p = 1.000 for all; Figure 2A,B).

3.7 FLT3-ITD combined with induction can further refine EFS stratification of NPM1^mut AML

Finally, as FLT3-ITD and induction regimen were independent predictors on EFS in NPM1^mut patients, further analysis integrating these two factors further refined risk-stratification of EFS in NPM1^mut patients. The FLT3-ITD⁽⁺⁾/SD-Ara-C subgroup had the worst EFS (2-year EFS rate: 29% ± 6%; 3-year EFS rate: 24% ± 7%; median EFS: 14.6 [95% CI 10.3–18.9] months). The EFS rate was significantly different from the other three subgroups (median EFS not reached, Bonferroni p < 0.05 for all; Figure 3). In contrast, the EFS of FLT3-ITD⁽⁺⁾/ID-Ara-C subgroup was similar to that of the FLT3-ITD⁽⁻⁾/SD-Ara-C and FLT3-ITD⁽⁻⁾/ID-Ara-C subgroups (Bonferroni p > 0.05 for both; data not shown).

4 DISCUSSION

NPM1^mut occurs in one-third of de novo AML and is associated with a good prognosis.^{5, 6} Approximately 40% of NPM1^mut cases are concurrent with FLT3-ITD, which impairs the favorable effect of NPM1^mut.^20-24 Despite the increasing use of FLT3 inhibitors, the mainstay of the “7 + 3” regimen in treatment of FLT3-ITD⁽⁺⁾ AML is still unshakable. In developing countries, FLT3 inhibitors cannot typically be prescribed as conventional drugs due to their lack of accessibility, economic burden, and tolerance by some patients. The correlation between FLT3-ITD⁽⁺⁾ with Ara-C specific resistance^{14, 15} prompted the present demonstration that ID-Ara-C induction improved the response rate and mitigated induction failure in NPM1^mut/FLT3-ITD⁽⁺⁾ patients. These findings support the notion that Ara-C uptake deficiency can be partially compensated by increasing the dosage of administered Ara-C.¹⁴ Of the four patients with NPM1^mut/FLT3-ITD⁽⁺⁾ who did not attain CR after a first cycle of ID-Ara-C induction, re-induction CR was noted in only one patient, who had previously obtained PR. This suggests that ID-Ara-C containing chemotherapy could be a first-line option for initial induction, in addition to its role as a main component of second-line rescue protocol. This could allow physicians to identify and ameliorate primary resistance. In this study, ID-Ara-C containing induction benefited the cCR rate and EFS only in the NPM1^mut/FLT3-ITD⁽⁺⁾ group, with no benefits in the NPM1^mut/FLT3-ITD⁽⁻⁾ group. Therefore, NPM1^mut/FLT3-ITD⁽⁺⁾ patients can be promisingly induced by initially increasing Ara-C intensity to improve short-term efficacy. Moreover, the toxicity of the ID- and SD-Ara-C regimens were similar and did not increase the risk of induction-related ED.

Maurillo et al.²⁵ combined Ara-C intensity with MRD in AML patients. The probability of achieving MRD⁽⁻⁾ was similar between the HD- and SD-Ara-C regimens in NPM1⁽⁺⁾ or FLT3-ITD⁽⁺⁾ patients regardless of whether after induction or during consolidation. When restricting analysis to MRD⁽⁻⁾, the authors even revealed the prognostic superiority of the SD-regimen compared to the HD-Ara-C regimen. However, in that study, only patients who achieved morphological CR after induction were selected for survival analysis. This does not exclude the possibility that the outcome analysis included a considerable percentage of patients who reached CR induced with HD-Ara-C, but tended to have primary resistance, which cannot be identified by SD-Ara-C. Furthermore, more patients who did not respond to SD-Ara-C were excluded from the analysis, resulting in an interferential bias. Interestingly, disease-free survival of HD-Ara-C was still better than that of SD-Ara-C if cases were confined to continuous MRD⁽⁺⁾ during consolidation, suggestive of an inhibition effect of HD-Ara-C on resistant leukemic clone to defer the occurrence of relapse and prolong OS.

In addition to passively increasing Ara-C dosage, the manner of administration of Ara-C may also affect the chemotherapeutic efficacy.²⁵ In this study, the first 4 days of the induction schedule of the ID-Ara-C regimen was the same as that of the SD-Ara-C regimen, with ID-Ara-C supplemented in days 5–7. Adding Ara-C dosage at the later stage of induction schedule has several advantages. First, for patients with hyperleukocytosis, later administration of ID-Ara-C can reduce risk of tumor lysis syndrome. Second, there is the opportunity to wait for a turnaround of the FLT3-ITD results before deciding whether to adjust the Ara-C dose. Although the present findings indicate the improvement of ID-Ara-C in NPM1^mut/FLT3-ITD⁽⁺⁾ patients, we assume that this effect may otherwise be extended to all FLT3-ITD⁽⁺⁾ patients regardless of NPM1^mut status. This assumption needs to be confirmed.

FLT3-ITD did not change the degree of DNA damage after daunorubicin chemotherapy, but was linked with an increased level of p53. FLT3-ITD⁽⁺⁾ leukemic cells were desensitized in the presence of p53, while chemosensitivity was restored after knockout of TP53, indicating a dependence on p53 for the resistance of FLT3-ITD⁽⁺⁾ leukemia cells.²⁶ Therefore, the combination of SD-Ara-C with daunorubicin may be suboptimal for induction.¹¹ An alternative of anthracyclines, such as idarubicin (IDA) substituting for DNR, might be considered to potentially improve the efficacy. If applicable, FLT3 inhibitor can be administrated as soon as possible to sensitize cytotoxic chemotherapy and mitigate resistance.^{26, 27} However, in our study, owing to the non-homogenous usage of FLT3 inhibitor, varying agent selection, and initiation timing and duration, it was not a covariate in multivariable outcome analysis in the NPM1^mut/FLT3-ITD⁽⁺⁾ group.

Among the comutations of NPM1^mut AML, TET2⁽⁺⁾ was identified as an independent factor predicting lower cCR rate and inferior EFS in the NPM1^mut/FLT3-ITD⁽⁻⁾ group, consistent with other results.^{28, 29} Moreover, integration of TET2 and FLT3-ITD could better discriminate patients with NPM1^mut AML into three prognostic subsets, with NPM1^mut/FLT3-ITD⁽⁻⁾/TET2⁽⁺⁾ comparable to NPM1^mut/FLT3-ITD⁽⁺⁾ group regarding all outcome endpoints. Our findings differ from the study of Chou et al.,³⁰ who reported no significant impact of TET2 mutation on OS among AML patients with favorable molecular genotypes (NPM1^mut/FLT3-ITD⁽⁻⁾ or CEBPA^double). However, the survival curve indicated a trend of OS difference between the TET2 wild and mutant subgroups, although not significant (p = 0.198). The lack of significance may be due to the limitations of the number of cases and ethnic disparity. Moreover, the authors combined NPM1⁽⁺⁾/FLT3-ITD⁽⁻⁾ and CEBPA^double, and did not analyze the specific NPM1⁽⁺⁾/FLT3-ITD⁽⁻⁾ individually.

According to guidelines,^{7, 8} AML having the NPM1^mut without FLT3-ITD or with FLT3-ITD^low genotype were combined and allocated into a favorable-risk. However, the AMLSG large cohort study²⁰ risk-stratified patients as per the 2017-European LeukemiaNet criteria, demonstrating that OS of NPM1^mut/FLT3-ITD^low group <60 years of age was worse than that of other favorable-risk groups (p = 8.7 × 10⁻⁸) including NPM1^mut/FLT3^wt cases. The OS rate did not significantly differ from that of intermediate-risk group (p = 0.76), which included NPM1^mut/FLT3-ITD^high AR cases. The survival rates in patients ≥60 years of age were also similar across NPM1^mut/FLT3^wt, NPM1^mut/FLT3-ITD^low, and NPM1^mut/FLT3-ITD^high. Additionally, in the small cohort of The Cancer Genome Atlas data, NPM1^mut/FLT3-ITD^low had similar poor OS with NPM1^mut/FLT3^wt patients because of higher median age in both groups. The authors concluded that patients with NPM1^mut/FLT3-ITD^low, as well as those with NPM1^mut/FLT3^wt ≥ 60 years of age, could not be classified into a favorable-risk. Other studies also demonstrated no significant prognostic difference between patients with NPM1^mut/FLT3-ITD^low and NPM1^mut/FLT3-ITD^high.^21-24 Thus, it appeared reasonable to re-allocate NPM1^mut/FLT3-ITD^low to an intermediate-risk. This is embodied in the very latest published recommendations, 2022 European LeukemiaNet risk classification by genetics at initial diagnosis in AML, which categorizes NPM1^mut/FLT3-ITD as intermediate-risk irrespective of FLT3-ITD AR.³¹

CD34⁽⁺⁾ was associated with primary resistance and poorer EFS in NPM1^mut/FLT3-ITD⁽⁺⁾ group, similar to other findings.^{32, 33} The combination of varying features of antigen expression is more informative in predicting survival, where CD34⁽⁺⁾/CD38⁽⁻⁾/CD123⁽⁺⁾ representing the leukemia stem cell phenotype has prognostic relevance.³⁴ Chen et al.³⁵ characterized the antigen expression among 94 NPM1^mut patients by cluster analysis and divided them into two categories according to CD34, CD7, and HLA-DR. The results revealed a significantly unfavorable prognosis for type-II features characterized by CD34⁽⁺⁾/HLA-DR⁽⁺⁾/CD7⁽⁺⁾ compared to type-I features characterized by CD34⁽⁻⁾/CD7⁽⁻⁾. Mason et al.³⁶ analyzed myeloid blast populations excluding monocytic differentiation in NPM1^mut patients. The acute promyelocytic leukemia-like phenotype CD34⁽⁻⁾/HLA-DR⁽⁻⁾/MPO^(str+) was present in nearly half the patients (48%) and beneficially influenced RFS and OS. Therefore, the leukemic immunophenotypes should not be limited to AML diagnosis, but rather additionally play roles in prognostication and MRD monitoring, to guide treatment.

In the present study, AR was detected in only some NPM1^mut/FLT3-ITD⁽⁺⁾ patients, so we did not include FLT3-ITD AR in the analysis. A recent study concerning the relationships between CD34 with NPM1^mut or FLT3-ITD described the distribution of FLT3-ITD^high almost exclusively in the CD34⁽⁺⁾ group and rarely in the CD34⁽⁻⁾ group, the latter entirely presenting FLT3-ITD^low features.³⁷ Thus, CD34 negativity may likely imply an absence of FLT3-ITD or low AR, which partly explains the better prognosis of CD34⁽⁻⁾ patients than those with CD34⁽⁺⁾. Our findings indicate that ID-Ara-C as a first-line induction in NPM1^mut/FLT3-ITD⁽⁺⁾ patients improved the remission rate, especially for CD34⁽⁺⁾ patients at the initial diagnosis. CD34 in NPM1^mut/FLT3-ITD⁽⁻⁾ was not prognostically significant, partly due to a lower CD34⁽⁺⁾ percentage and the accordingly reduced occurrence of endpoints, with no statistical significance. EFS stratification of NPM1^mut AML could be further refined by the combination of FLT3-ITD with an induction scheme. FLT3-ITD⁽⁺⁾/SD-Ara-C displayed the worst EFS. This finding indicates that NPM1^mut/FLT3-ITD⁽⁺⁾ patients can be initially induced by ID-Ara-C, especially CD34⁽⁺⁾ patients, who tended to be enriched for FLT3-ITD^high characteristics.

Consistent with other results,³⁸ the number of mutated genes at initial presentation in our NPM1^mut cohort independently predicted OS, regardless of FLT3-ITD and induction intensity, highlighting an interplay between coexisting genetic lesions. A Japanese report showed that even patients with NPM1^mut/FLT3-ITD⁽⁻⁾ presented an inferior outcome when a high number of mutations coexist.³⁹ This finding highlights the broad coverage and high throughput value of NGS for determining genetic alterations.

In our cohort, some traditional indicators, such as increasing age and high WBC count, had independent impacts on EFS in NPM1^mut/FLT3-ITD⁽⁻⁾ patients, similar to other results.^{20, 23, 40} These “old” parameters may be a comprehensive embodiment of the external effects of varying biological mechanisms intrinsically harbored in leukemia cells. For example, WBC reflects leukemic burden and cell proliferation and viability, and age reflects a distribution tendency toward clonal hematopoiesis-related mutations, as well as performance status and organic comorbidity, which decrease chemotherapeutic tolerance.

There were several limitations in our study. First, this is a retrospectively nonrandomized study design. Its inherent selection bias leads to potential confounders (e.g., divergent salvage protocols and FLT3 inhibitor usage, supportive care heterogeneity), although the analyzed cases received intensive therapy with a curative intent. A second limitation is missing data on immunophenotypic markers, which might compromise the statistical power of other covariates in a multivariate model. Although the transplant-associated biases were adjusted in the survival analysis, discrepant transplant indications, and timing and donor-recipient HLA relationships were present. Also, loss to follow-up for some cases might affect study results. Future prospective studies are necessary to confirm the present findings.

To summarize, patients with NPM1^mut AML have differential predictors according to their FLT3-ITD status. TET2 mutation, age, and WBC count are independently associated with outcome in AML with NPM1^mut/FLT3-ITD⁽⁻⁾. Positive CD34 expression is adversely associated with outcome in AML with NPM1^mut/FLT3-ITD⁽⁺⁾ where initial ID-Ara-C induction and allo-SCT consolidation can benefit efficacy. The number of mutated genes predicts OS, regardless of FLT3-ITD status. We identified high-risk subsets among NPM1^mut AML patients, allowing individually-tailored management strategies in this AML subtype.

AUTHOR CONTRIBUTIONS

Biao Wang: Data curation (equal); formal analysis (equal); software (equal); writing – original draft (equal); writing – review and editing (equal). Xiaoying Hua: Resources (equal); validation (equal); visualization (equal); writing – review and editing (equal). Jihong Zhang: Data curation (equal); methodology (equal); resources (equal); writing – review and editing (equal). Weiying Gu: Conceptualization (equal); funding acquisition (equal); supervision (lead); writing – review and editing (equal). Haiqian Li: Supervision (supporting); validation (equal); writing – review and editing (equal).

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflict of interests.