Heterogeneity analysis of the CEBPAdm AML based on bZIP region mutations

Patients with double-mutated CEBPA (CEBPAdm) AML were stratified into favorable risk group, however, few studies have investigated the heterogeneity of different CEBPAdm types in detail. In this study, we analyzed 2211 newly diagnosed AML and identified CEBPAdm in 10.8% of the patients. Within the CEBPAdm cohort, 225 of 239 patients (94.14%) presented with bZIP region mutations (CEBPAdmbZIP) while 14 of 239 patients (5.86%) without bZIP region mutation (CEBPAdmnonbZIP). Analysis of the accompanied molecular mutations showed statistically different incidences of GATA2 mutations between the CEBPAdmbZIP group and the CEBPAdmnonbZIP group (30.29% vs 0%). In the analysis of outcomes, patients with CEBPAdmnonbZIP were associated with shorter overall survival (OS) censored at hematopoietic stem cell transplantation (HSCT) during CR1 compared to those with CEBPAdmbZIP (hazard ratio (HR) = 3.132, 95% confidence interval (CI) = 1.229–7.979, P = .017). Refractory or relapsed AML (R/RAML) patients with CEBPAdmnonbZIP were associated with shorter OS compared to those with CEBPAdmbZIP (HR = 2.881, 95% CI = 1.021–8.131, P = .046). Taken together, AML with CEBPAdmbZIP and CEBPAdmnonbZIP showed different outcomes and might be regarded as distinctive AML entities.


INTRODUCTION
The mutations of the gene encoding the CCAAT-enhancer binding protein alpha (CEBPA) were found in 6% to 15% of newly-diagnosed Acute Myeloid Leukemia (AML) patients, twothirds of which were double-mutated CEBPA (CEBPAdm). [1][2][3] Initially, CEBPAdm was stratified into favorable risk group. [4][5][6][7] Recently, studies demonstrated that the presence of inframe bZIP mutations of CEBPA was associated with favorable prognosis in newly diagnosed AML 2,8,9 which was updated in the new International Consensus Classification (ICC) of Myeloid Neoplasms and Acute Leukemia. 10 However, CEBPAdm without bZIP region mutation (CEBPAdmnonbZIP) is classified as CEBPA mutated AML in the 5th edition of WHO Classification of Haematolymphoid Tumours 11 but not in ICC classification. Few studies have investigated the impact of different CEBPAdm types in detail, especially for CEBPAdmnonbZIP. To further evaluate the heterogeneity of CEBPAdm, we analyzed the clinical features and outcomes of patients with different mutation types of CEBPAdm.

Patients
Two hundred and thirty-nine newly diagnosed AML patients with CEBPAdm from 2211 consecutive AML (non-APL) patients treated in our center between July 2011 and December 2021 were assessed for the inclusion criteria including the following: AML according to WHO classification (version 2016) 12 ; Patients received at least 2 courses of chemotherapy treatment if patients did not achieve complete response (CR) after the first course while others at least 1 course. The study was approved by the hospital ethics committee (approval number: NSFC2021076-EC-2) and conducted in accordance with the Declaration of Helsinki.

Genetic studies
Genetic analysis was performed on bone marrow samples obtained from patients at initial presentation. Cytogenetics were analyzed by R-banding, and the molecular genetic examination was performed by next-generation sequencing (NGS) and polymerase chain reaction (PCR).

Statistical analysis
Event-free survival (EFS) was defined as the interval from diagnosis to assessment of response after the second course of chemotherapy treatment if patients failed to achieve CR in the first induction chemotherapy, the date of relapse, or the date of death, whichever occurred first. Overall survival (OS) was calculated from the date of AML diagnosis to the date of death or the date of last follow-up for surviving patients. To exclude the bias brought about by early relapse or death ahead of hematopoietic stem cell transplantation (HSCT) during the first complete remission phase (CR1), landmark analysis were used. For considering the impact of HSCT on OS, the landmark was set on the median time from trial-enrollment to CR1-HSCT. Statistical analysis were performed using SPSS (version 21.0) and GraphPad Prism (version 7.0) software. Univariable analysis were using the Cox proportional hazards regression model. Statistical tests were 2-sided with a significance level set at 0.05.

Patients
Totally, 2211 newly-diagnosed AML (non-APL) patients in our center were screened and 242 patients with CEBPAdm were found. Two hundred thirty-nine patients with CEBPAdm were included in the study while 3 patients who did not receive chemotherapy were excluded ( Figure 1). The baseline data of the 239 AML patients with CEBPAdm are shown in Supplementary  Table 1, http://links.lww.com/BS/A55. Of these patients, the median age was 40 years (interquartile range (IQR) = 30-48 years) and the median follow-up period was 42.43 months (IQR = 19.57-60.57 months). In our cohort, 7 + 3 induction regimens which consists of DA (Daunorubicin, Cytarabine) or IA (Idarubicin, Cytarabine) were used as the first induction therapy in 154 patients (64.44%), while HAD (homoharringtonine, cytarabine, daunorubicin) 13,14 was used in 76 patients (31.80%). Thirty patients underwent HSCT during CR1. The median OS and EFS were 33.53 months (IQR = 17.20-59.13 months) and 23.70 months (IQR = 10.30-56.00 months), respectively (the Kaplan-Meier curves are shown in Figure 2).

Comparison of clinical features between CEBPAdmbZIP and CEBPAdmnonbZIP
Based on the location of mutations, CEBPAdm was divided into 2 types: At least 1 of the 2 mutations occurred in bZIP region (CEBPAdmbZIP); Mutations did not occur in bZIP region (CEBPAdmnonbZIP). In our cohort, CEBPAdmbZIP was found in 225 of 239 patients (94.14%) while 14 of 239 patients (5.86%) presented with CEBPAdmnonbZIP. We compared the clinical features of the 2 groups and focused on the accompanied molecular mutations. Table 1 shows that baseline characteristics were balanced among the 2 groups which included age, sex, white blood cell count (WBC), hemoglobin (HB) content, platelet count (PLT), and cytogenetics (P > .05). Intriguingly, CEBPAdmnonbZIP patients had no GATA2 mutation at all, while 53 of 175 (30.29%) CEBPAdm bZIP group patients happened (P = .034). There was no statistical difference in the rest of the molecular markers and treatments between the 2 groups. The CR rate after induction chemotherapy and the measurable residual disease (MRD)-positive rate of CEBPAdmbZIP group and CEBPAdmnonbZIP group presented no statistical difference (CR rate: 90.67% vs 85.71%, P = .887; MRD-positive rate: 11.58% vs 8.33%, P = 1.000).

Comparison of outcomes between CEBPAdmbZIP and CEBPAdmnonbZIP
Further, we investigated the impact of CEBPAdm type on outcomes. In univariate analysis by the Cox proportional hazards regression model, OS and EFS present no statistical difference (OS: hazard ratio (HR) = 2.027, 95% confidence interval (CI) = 0.802-5.121, P = .135; EFS: HR = 1.147, 95% CI = 0.501-2.626, P = .746, the Kaplan-Meier curves are shown in Figure 3A, B). When HSCT during CR1 was considered as a censored event, patients with CEBPAdmnonbZIP were associated with shorter OS censored at HSCT compared to those with CEBPAdmbZIP (HR = 3.132, 95% CI = 1.229-7.979, P = .017, the Kaplan-Meier curves are shown in Figure 3C). However, no statistical difference was found in EFS censored at HSCT between CEBPAdmnonbZIP and CEBPAdmbZIP (HR = 1.437, 95% CI = 0.627-3.293, P = .391, the Kaplan-Meier curves are shown in Figure 3D). The discordant results for OS and EFS remind us to focus on outcomes of patients with refractory or relapsed AML (R/RAML). Eighty-five patients were R/RAML in the cohort (80 patients with CEBPAdmbZIP, and 5 patients with CEBPAdmnonbZIP). Univariate analysis showed that R/RAML patients with CEBPAdmnonbZIP were associated with shorter OS compared to those with CEBPAdmbZIP (HR = 2.881, 95% CI = 1.021-8.131, P = .046, the Kaplan-Meier curves are shown in Figure 4).
To identify the effect of transplantation, we compared OS by landmark analysis in each group according to whether underwent HSCT during CR1. The Kaplan-Meier curves in Figure 5 suggested that HSCT during CR1 did not improve OS in CEBPAdmbZIP group, but was associated with a trend to prolong OS in CEBPAdmnonbZIP group (CEBPAdmbZIP: P = .227; CEBPAdmnonbZIP: P = .095).

DISCUSSION
In this study, we analyzed 2211 newly-diagnosed AML and identified CEBPAdm in 10.8% of the patients. We mainly focus on the heterogeneity of CEBPAdm, therefore we performed a detailed analysis of the different types of CEBPAdm according to bZIP region mutations. We found that the presence of CEBPAdmnonbZIP is associated with a lower prevalence of GATA2 mutations and poor outcome.
Previous studies showed that GATA2 mutations occurred in 18% to 41% of patients with CEBPAdm AML, which were obviously more common than those of CEBPAwt group or CEBPAsm group. [15][16][17] Similarly, in our CEBPAdm cohort, GATA2 mutations occurred in 28.04% of patients. Studies also found that CEBPAdm shared similar biologic features with CEBPA bZIP including the high prevalence of GATA2 mutations. 8,9 Intriguingly, we found that different incidences of  The bold P value indicates statistical significance with P less than 0.05. CR1-HSCT = HSCT during the first complete remission phase, DA = daunorubicin and cytarabine, HAD = homoharringtonine, cytarabine, and daunorubicin, HB = hemoglobin, IA = idarubicin and cytarabine, IQR = interquartile range; PLT = platelet, R/RAML = refractory or relapsed AML, WBC = white blood cell.
GATA2 mutations existed between the CEBPAdmbZIP group and the CEBPAdmnonbZIP group (30.29% vs 0%). This finding suggests that AML bearing CEBPAdm with bZIP and without bZIP might have different pathogenesis and be classified as distinctive entities. Recent studies demonstrated that CEBPA mutations involving bZIP region were associated with favorable prognosis in both biallelic and single mutations. 2,8,9 The favorable category in the new ICC of myeloid neoplasms and acute leukemia 10 has been updated to CEBPA with inframe bZIP mutations instead of biallelic mutations. Simultaneously, the definition has changed to AML with CEBPA mutation, which includes biallelic as well as single mutation located in the bZIP region in the 5th edition of WHO Classification of Haematolymphoid Tumours. 11 The difference between the 2 classifications mentioned above was whether CEBPAdmnonbZIP was included. Until now, few studies have been reported about the outcomes of CEBPAdmnonbZIP patients. Although we did not find difference in OS and EFS between CEBPAdmnonbZIP group and CEBPAdmbZIP group, CEBPAdmnonbZIP was associated with shorter OS when censored at HSCT during CR1. However, we did not find statistical difference in EFS censored at HSCT during CR1, which prompted us to investigate the outcomes of R/RAML patients. Further analysis showed that R/RAML with CEBPAdmnonbZIP were associated with poorer OS compared to those with CEBPAdmbZIP. Schlenk et al 18 revealed that the outcome of relapsed CEBPAdm patients could be improved  by a high rate of second CR followed by allo-HSCT extending the survival to more than 2 years after relapse. Similarly, the median interval from relapse or refractoriness to death of 27 R/RAML patients with CEBPAdmbZIP who underwent allo-HSCT after relapse was 29.67 months in our study. In R/ RAML with CEBPAdmnonbZIP group, 1 patient underwent allo-HSCT in CR2 and 1 patient received allo-HSCT after 2 induction treatments for NR, the interval from relapse or refractoriness to death was 14.30 and 5.50 months, respectively. Hence, a potential difference between the CEBPAdmnonbZIP group and CEBPAdmbZIP group was the outcome after relapse and refractoriness, but not in CR1. Thus, these data further suggested that AML with CEBPAdmbZIP and CEBPAdmnonbZIP should be regarded as distinctive AML entities.
Several limitations existed in our study. This was a retrospective study and the treatment regimens could not be uniform, although the induction chemotherapy regimens were balanced between the 2 groups in our cohort. Another limitation was the limited sample size of patients with CEBPAdmnonbZIP. Future larger and prospective studies are needed to clarify the results.
In conclusion, CEBPAdmnonbZIP AML was associated with inferior outcome mainly due to shorter survival after refractoriness or relapse compared to CEBPAdmbZIP AML, which suggested that they were distinctive entities.