Introduction

Prognostic risk classification of acute myeloid leukemia (AML) has been subdivided according to chromosomal abnormalities and genetic mutations. The National Comprehensive Center Network (NCCN) and European LeukemiaNet (ELN) continue to revise their prognostic risk classifications for AML [1,2,3,4,5,6,7]. In NCCN guidelines 2017 AML classification, cytogenetically normal (CN)-AML is prognostically stratified by Fms-like tyrosine kinase 3-internal tandem duplication (FLT3-ITD), mutations of nucleophosmin 1 (NPM1) and biallelic mutation of CCAAT/enhancer-binding protein alpha (CEBPA) [2]. Although biallelic mutation of CEBPA (CEBPA-bm) has long been considered to be a better prognostic marker [2, 6, 8], recent studies have shown that a single mutation in the C-terminus basic leucine zipper domain (bZIP) is strongly associated with better prognosis [9, 10]. We determined the clinical features and prognosis of CN-AML by analyzing FLT3-ITD, NPM1 mutation, and CEBPA mutation based on our regional prospective AML cohort.

Patients and methods

Patients

Hokkaido Leukemia Net (HLN) is a multicentric prospective observational cohort study collecting acute leukemia samples from affiliated hospitals of the North Japan Hematology Study Group (NJHSG) covering all of Hokkaido (UMIN000048611). In the HLN prospective protocol, newly diagnosed cases of acute myeloid leukemia excluding acute promyelocytic leukemia (APL) are analyzed for the presence of FLT3-ITD and Wilms tumor gene (WT1) expression, preserving genome DNA and complementary DNA (cDNA), respectively. The results are returned to the clinician in charge and then the clinician reports the patient’s mutation and clinical information including final diagnosis and karyotype. NPM1 and CEBPA or KIT mutations are additionally analyzed for CN-AML or core-binding factor AML, respectively (Fig. 1A). The results of these genetic analyses are returned to the clinician, and this information enables the clinician to stratify AML into three prognostic groups based on the NCCN 2017 prognostic criteria (Table 1). In the HLN protocol, patients aged 16 years or older are eligible for registration without an upper age limit, reflecting regional real-world cohort. Each patient's treatment is at the physician’s discretion. We retrospectively analyzed FLT3-ITD, NPM1 mutation, and the mutation pattern of CEBPA and prognosis in patients with CN-AML. This study was conducted in accordance with the Helsinki Declaration and was approved by the institutional review boards of Hokkaido University Hospital (#015-0344). Written consent was obtained from all patients.

Fig. 1
figure 1

A Flow of analysis of AML patients in HLN. B Venn diagram of FLT3-ITD, NPM1 and CEBPA mutations

Full size image
Table 1 NCCN guidelines 2017
Full size table

Mutational analysis

Patient samples were collected at diagnosis, and genomic DNA was extracted from bone marrow or peripheral blood. FLT3-ITD, NPM1 mutation, and CEBPA mutation were analyzed using a genome DNA template. The relevant region of FLT3 was amplified by polymerase chain reaction (PCR) and gel electrophoresis to verify the presence of the ITD band. NPM1 exon 12 and full-length CEBPA were analyzed by Sanger sequencing. Primers used for PCR and Sanger sequencing are listed in Supplemental Table 1. Sanger sequencing was performed using the BigDye™ Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Waltham, MA, USA) and analyzed by Applied Biosystems SeqStudio Genetic Analyzer (Applied Biosystems, Waltham, MA, USA). The 6-bp duplication polymorphism within the CEBPA TAD2 domain (c.589_590 ins ACCCGC; rs762459325) and single-nucleotide polymorphism (c.690 G > T; rs34529039) were not considered as mutations. Although AML with TP53 mutation is classified as a poor prognosis group in the NCCN 2017 guidelines, mutation analysis of TP53 was omitted because the frequency of TP53 mutations is rare (2–3%) in CN-AML [11, 12].

Statistical analysis

Clinical variables across groups were compared using the χ2 test or 2-sided Fisher’s exact test for categorial variables. The nonparametric Mann–Whitney U test and Kruskal–Wallis test were used for continuous variables. A P value < 0.05 indicated a significant difference. Prognostic analysis was done excluding patients treated by best supportive care only. The log-rank test was used to evaluate overall survival (OS). For multivariable analysis of prognostic factors, Cox proportional hazards regression models were used. Variables were selected based on previous studies and clinical experience. For multiple comparisons of three or more groups, p-values were adjusted using Bonferroni–Holm correction. All statistical analyses were performed with EZR software ver.1.54 (Saitama Medical Center, Jichi Medical University, Saitama, Japan) [13].

Results

A total of 794 patients with AML were enrolled in the HLN from January 2009 to December 2021. CN-AML accounted for 43.7% (342 cases) of the 782 karyotype-identified cases. Of those, 295 patients for whom all data were available were included in the study (Fig. 1A). Patients’ characteristics are shown in Table 2. The median age of the 295 CN-AML patients was 64 years (range, 18–93 years). Age distribution was shown on Supplemental Fig. 1. The patients included 155 males and 140 females. Intensive chemotherapy with anthracycline plus cytarabine was performed in 195 patients (66.1%). After induction chemotherapy, complete remission was achieved in 207 patients (70.2%). Allogeneic hematopoietic stem cell transplantation (allo-SCT) was performed in 97 patients (32.9%). The median follow-up period was 446.5 days (range 0–4219). We excluded patients who did not receive chemotherapy from the prognostic analysis.

Table 2 Patient’s characteristics (N = 295)
Full size table

FLT3-ITD, NPM1 mutation, and CEBPA-bm were positive in 76 patients (25.8%), 113 patients (38.3%), and 30 patients (25.1%), respectively. Cases with NPM1 mutation and CEBPA-bm were completely exclusive (Fig. 1B). The proportion of patients with NPM1 mutations increased with age, while the proportion of patients with CEBPA-bm decreased with age (Fig. 2A). Patients with CEBPA-bm were significantly younger at diagnosis than the other patients (median age: 46 vs 65 years, P < 0.001).

Fig. 2
figure 2

A Ratios of mutations in FLT3-ITD, NPM1, and CEBPA by age. B Survival curves stratified by FLT3-ITD status, NPM1 mutation status, and CEBPA mutation status for patients younger than 65 years. C Survival curves stratified by FLT3-ITD status, NPM1 mutation status, and CEBPA mutation status for patients aged 65 years or more

Full size image

The prognostic impacts of FLT3-ITD, NPM1 mutation, and CEBPA mutations were different between age groups (Fig. 2B, C). Patients who had FLT3-ITD-positive showed significantly poor survival compared to those patients who were FLT3-ITD-negative, whereas the difference is more prominent in patients younger than 65 years. Then, we analyzed the effect of the FLT3-ITD allelic ratio (AR) on the prognosis of FLT3-ITD positive patients. FLT3-ITD AR was retrospectively analyzed for 58 patients. The WBC count at diagnosis was higher in the ARhigh (AR ≥ 0.5) group, but there were no statistically significant differences in other patient characteristics (Supplemental Table 2). Prognostic analysis of 56 patients who underwent induction chemotherapy showed no difference in prognosis between the ARhigh and ARlow (AR < 0.5) groups (Supplemental Fig. 2A). Even combined with NPM1 mutation, the better prognostic impact of ARlow was not evident (Supplemental Fig. 2B).

Although there was no statistically significant difference in prognosis by the NPM1 mutation status alone, NPM1-mutated patients tend to show better prognosis only in patients aged 65 years or more. The 295 CN-AML patients were divided into three groups depending on CEBPA mutation; CEBPA-bm, CEBPA single mutation (CEBPA-sm), and CEBPA wild type (CEBPA-wt). CEBPA-bm showed significantly better survival in patients younger than 65 years; however, the statistical significance is not clear for patients aged 65 years or more due to the limited number of CEBPA-mutated cases. As reported previously [14], patients with CEBPA-bm had a significantly higher rate of complete remission (CR) (CEBPA-bm, 90.0% vs the other patients, 68.0%; P = 0.025) (Supplemental Table 3) as well as a better 5-year OS (CEBPA-bm, 76.4% vs CEBPA-sm, 35.0% vs CEBPA-wt, 32.7%; P = 0.0074).

Because there was no statistically significant difference in 5-year OS for the NPM1 mutation alone, we combined NPM1 mutation with FLT3-ITD status determined in 2 age groups (Age < 65 years, ≥ 65 years) (Fig. 3A, B). FLT3-ITD was a poor prognostic factor in both age groups, but the favorable prognostic impact of NPM1 was more evident in patients aged 65 years or more.

Fig. 3
figure 3

A Survival curves stratified by the presence or absence of FLT3-ITD and NPM1 mutation for patients younger than 65 years. B Survival curves stratified by the presence or absence of FLT3-ITD and NPM1 mutation for patients aged 65 years or more

Full size image

Patients were classified according to the NCCN 2017 guidelines (Fig. 4A, B). Patients’ characteristics are shown on Table 3. Patients younger than 65 years were clearly stratified into 3 prognostic groups. For patients aged 65 years or more, the favorable-risk group had a significantly better prognosis than the intermediate- and poor-risk groups, but there was no difference in prognosis between the intermediate-risk and poor-risk groups in this cohort.

Fig. 4
figure 4

A Survival curves based on the NCCN guidelines 2017 in patients younger than 65 years. B Survival curves based on the NCCN guidelines 2017 in patients aged 65 years or more

Full size image
Table 3 Patient’s characteristics by risk status based on the NCCN guidelines 2017
Full size table

CEBPA bZIP in-frame mutation is good prognostic factor

The domain arrangement of the CEBPA protein is shown in Fig. 5. The CEBPA protein has a TAD1 domain, a TAD2 domain from the N-terminal side and a bZIP domain located at the C-terminus. The location and type of the mutation in the CEBPA gene in this study are also shown. Most of the non-bZIP mutations (71.4%) were frameshift mutations, and most of the mutations in the bZIP domain (82.9%) were in-frame mutations (Fig. 5). CEBPA bZIP mutation was found in 26 (86.7%) of the 30 CEBPA-bm cases and in 4 (9.1%) of the 44 CEBPA-sm cases (Fig. 6A).

Fig. 5
figure 5

Distribution and types of CEBPA mutations detected in this study

Full size image
Fig. 6
figure 6

A Reclassification based on mutations in the CEBPA bZIP domain. B Survival curves based on conventional CEBPA-bm (left) and new classification depend on CEBPA bZIP inframe mutation (right)

Full size image

Then we reclassified the patients with CEBPA bZIP in-frame mutations as “CEBPA-bZIP inf” and those with other mutations as “CEBPA-other” (Fig. 6A). Patients with CEBPA-bZIP inf were significantly younger at diagnosis than the other patients (median age: 46 vs 65 years, P < 0.001). However, there were no differences in other characteristics (gender, WBC count at diagnosis, blast percentage in bone marrow, FLT3-ITD-positive ratio, intensity of induction chemotherapy, and number of patients who received allo-SCT) (Supplemental Table 3). Patients with CEBPA-bZIP inf had a significantly higher rate of CR (CEBPA-bZIP inf vs the other patients, 90.0 vs 68.0%; P = 0.025) (Supplemental Table 3) as well as a higher 5-year OS (CEBPA-bZIP inf, 80.2% vs CEBPA-other, 34.7% vs CEBPA-wt, 33.5%; P = 0.0068) (Fig. 6B).

Factors associated with OS

Table 4 shows the results of univariate and multivariate analyses for OS. Univariate analysis was carried out for age, gender, WBC count, FLT3-ITD, NPM1 mutation, CEBPA-bZIP inf, and intensity of induction chemotherapy, whether CR was achieved by induction chemotherapy, and allo-SCT. Male gender and FLT3-ITD-positive were associated with worse OS (P = 0.020 and 0.0080, respectively; log-rank test). Age < 65 years, intensive chemotherapy, achievement of CR by induction chemotherapy, allo-SCT, and CEBPA-bZIP inf were associated with better OS (P < 0.001, < 0.001, < 0.001, < 0.001, and 0.0036, respectively; log-rank test). Multivariate analysis showed that age < 65 years and CEBPA-bZIP inf were independently associated with better OS (age < 65 years: HR, 0.45; 95% CI 0.31–0.64, P < 0.001; CEBPA-bZIP inf: HR 0.31; 95% CI 0.13–0.78, P = 0.013) and that FLT3-ITD-positive was independently associated with worse OS (FLT3-ITD: HR, 1.77; 95% CI 1.16–2.70, P < 0.001).

Table 4 Univariate and multivariate analysis for overall survival
Full size table

Discussion

CN-AML represents the most frequent category of AML cytogenetics, accounting for 40–50% of all cases of AML [15,16,17,18,19,20,21]. CN-AML is considered as an intermediate-risk group if the risk is classified only by karyotype. As a result of advances in genetic analysis, the prognostic classification of AML has been updated frequently and has become more detailed. Thus, searching for these genetic alterations in addition to identifying the karyotype is very important for prognostic stratification and selection of patient treatment strategies. According to the NCCN AML guidelines 2017, CN-AML cases with biallelic mutations of CEBPA or with NPM1 mutation without FLT3-ITD are classified as a favorable-risk group [2]. However, a comprehensive panel sequence has not been clinically approved in Japan. Therefore, physicians have not been able to use genetic stratification according to international guidelines, and prognostic risk has been evaluated only by cytogenetics and treatment response in daily practice. In this study, without comprehensive genome panel analysis, we showed that mutation analysis of only three genes could clearly stratify a favorable prognostic group in CN-AML. Patients with CEBPA-bm or bZIP inframe mutation had a remarkably favorable prognosis. Patients with NPM1 mutation without FLT3-ITD showed better prognosis than other groups in patients aged 65 years or more. If those patients were evaluated only by cytogenetics, they could be overtreated by upfront allo-SCT. In HLN, the prospective feedback of results of FLT3-ITD, NPM1, and CEBPA analysis to the attending physician would have resulted in a lower rate of upfront allo-SCT in the favorable-risk group and, conversely, a higher rate in the poor-risk group as shown in Table 2. The favorable risk group clearly showed a better prognosis despite a lower allo-SCT rate. On the other hand, we could not find a difference in OS between the intermediate- and poor-risk groups. The lack of statistical difference in prognosis between the intermediate-risk and poor-risk groups may be due to the older age of patients and the lower rate of intense chemotherapy in the intermediate-risk group. The higher allo-SCT rate in the poor-risk group might have improved OS in that group and abolished the difference between the poor-risk and intermediate-risk groups.

In the ELN 2022 classification, FLT3-ITD mutation is no longer considered as a poor prognostic factor, and FLT3-ITD-positive AML cases are classified as an intermediate-risk group regardless of NPM1 mutation status [7]. In this study, FLT3-ITD was an independent poor prognostic factor for the survival of patients with CN-AML. This may be due to the fact that FLT3 inhibitors were not available in Japan until 2018, and the agents are currently approved only for relapsed or refractory cases. We found that FLT3-ITD AR did not stratify the prognosis of FLT3-ITD positive patients in our cohort, which is in contrast to the classification of the ELN 2017 guideline [6]. The prognostic significance of FLT3-ITD AR in the Japanese cohort is not consistent [22, 23]. Future application of upfront FLT3 inhibitors would abrogate the prognostic importance of FLT3-ITD AR.

We confirmed that NPM1 mutation is often coexistent with FLT3-ITD as previously reported [19, 24, 25]. Although NPM1 mutation alone did not show a favorable prognostic impact for all cases of CN-AML, NPM1 mutation showed a favorable impact for FLT3-ITD-negative cases as previously reported in the pre-FLT3 inhibitor era [19, 26,27,28,29,30]. The results of the analysis classified by age showed that the prognostic impacts of FLT3-ITD and NPM1 mutations were different between age groups as previously reported [31]. In CN-AML, genetic mutations such as DNMT3A have been reported to affect prognosis [32], so it is possible that other gene mutations which we did not analyze also contribute to prognosis.

In the ELN 2022 classification, CEBPA-bZIP inf rather than CEBPA-bm has become a criterion of CEBPA as a favorable-risk group. In the 5th edition of the WHO classification, single mutations within the bZIP domain (CEBPA-smbZIP) in addition to CEBPA-bm defined CEBPA criteria for a good prognosis [7, 33]. In our cohort, the CEBPA-bm group and the CEBPA-bZIP inf group accounted for 87% (26/30) of the cases, and the CEBPA-bZIP inf group had an even better prognosis. Further study with a larger number of cases is needed because the number of CEBPA-smbZIP cases was small and followed a relatively short period in our cohort. We welcomed the successful simplification of prognostic criteria for CEBPA mutation because CEBPA is an extremely GC-rich gene (GC: 75%) and a difficult PCR target. Commercially available panel sequences could sometimes fail to detect CEBPA mutation due to lower sequence coverage of the region [34]. Targeting only the bZIP domain reduces sequencing effort because the CEBPA bZIP domain is less GC-rich (GC: 64%) than domains other than bZIP domain (GC: 78%). As previously reported [35, 36], most of the bZIP domain mutations were inframe mutations, whereas most of the mutations other than bZIP CEBPA mutation were frameshift mutations in our series. Frameshift mutation on the N-terminus results in the formation of a short p30-CEBPA protein instead of the normal p42-CEBPA protein. This shortened protein is the dominant-negative form to hamper the function of wild-type CEBPA protein. Most of the C-terminal mutations are in-frame mutations in the bZIP domain, and the DNA-binding and dimer-forming abilities are reduced [37, 38]. In this study, patients with CEBPA-bm or CEBPA-bZIP inf were significantly younger at diagnosis than patients in the other groups as previously described in several reports [9, 10, 39, 40]. One possible explanation is the presence of a subset of the CEBPA-bm population in which one of the CEBPA mutations was a germline mutation. It has been reported that more than 10% of cases of AML with biallelic CEBPA mutations carried an N-terminal frameshift CEBPA germline mutation with acquisition of a C-terminal somatic mutation as a second event in the development of AML [41].

As a limitation, an FLT3 inhibitor was introduced in Japan in the midst of the study period. Application of the FLT3 inhibitor and its future expansion to de-novo FLT3-ITD-positive AML would affect the significance of FLT3-ITD as a prognostic factor. CEBPA-bm out of bZIP and CEBPA-smbZIP consisted of a small number of CEBPA mutations and it is difficult to compare each group directly.

In conclusion, CN-AML prognostic classification was possible by analyzing mutations in three genes, FLT3, NPM1 and CEBPA. Prospective analysis of these gene mutations may enable physicians to choose an appropriate treatment for each patient. CEBPA-bZIP inf had a very good prognosis, while CEBPA mt other than CEBPA-bZIP inf had a prognosis comparable to that of CEBPA-wt. CEBPA-bZIP inf is a better and simpler prognostic marker than CEBPA-bm. The successful simplification of the CEBPA mt prognostic criteria will have a great impact on genetic risk stratification of AML and potentially contribute to elucidation of the functional mechanisms of CEBPA for tumorigenesis in AML.