Validation of mutated CEBPA bZIP as a distinct prognosis entity in acute myeloid leukemia: a study by the Spanish PETHEMA registry

The prognosis for patients diagnosed with acute myeloid leukemia (AML) suitable for intensive chemotherapy, is defined by the presence of specific genetic abnormalities.^1,2 Among these, mutations in CCAAT/enhancer binding protein a (CEBPA) gene have classically classified as favorable risk.³ The frequency of CEBPA gene mutations ranges from 7% to 20%,^4,5 being present mostly in cytogenetically normal patients. While wild-type CEBPA (CEBPAwt) or CEBPA single-mutation (CEBPAsm) patients have ~60% risk of relapse, this risk is ~40% in those with CEBPA double mutation (CEBPAdm).⁵ These findings led to the inclusion of CEBPAdm AML as a distinct diagnostic entity in the 2016 World health Organization (WHO) classification⁶ and as favorable risk group by 2017 European Leukemia Net (ELN) risk classification.⁷ However, a study in children and young adults enrolled in Children’s Oncology Group trials showed that CEBPA mutations in bZIP region conferred favorable prognosis, regardless of whether they are CEBPAsm or CEPBAdm.⁸ This finding was confirmed in adult patients enrolled in protocols of the Study Alliance Leukemia,⁹ where bZIP mutations were associated with higher overall survival (OS) and complete remission (CR) rate. This data led to refinement of 2022 ELN risk classification,² defining as favorable risk only the presence of inframe bZIP CEBPA mutation (CEBPA-bZIP-inf).

In this study, we aim to describe the incidence, clinical-biological features, and prognosis of CEBPA mutations, including CEBPA-bZIP-inf, in a large series of real-life consecutive patients, homogeneously studied with harmonized next-generation sequencing (NGS) methodologies. For this purpose, we conducted a retrospective, non-interventional, multicenter study in the PETHEMA epidemiologic registry (N=2,434 consecutive patients with available centralized NGS) involving seven Spanish central-core laboratories (PLATAFO-LMA protocol; clinicaltrials gov. Identifier: NCT04446741). The consortium members are included in the Online Supplementary Appendix.

Of them, a total of 696 intensively treated AML (IT AML) patients (≥18 years) diagnosed with AML according to WHO 2016⁶ criteria since October 2017, with treatment and survival data were included. The ELN2017 was used for risk stratification.⁷ The intensive schedules consisted mainly in anthracycline plus cytarabine (Ara-C)-based regimens, such as 3+7 (idarubicin or daunorubicin and Ara-C) (N=493, 66.5%), mitoxantrone plus Ara-C, FLAG-IDA, FLAT (fludarabine, Ara-C, and topotecan), or ICE (idarubicin, Ara-C, and etoposide).

Genomic DNA, extracted from bone marrow (or peripheral blood) of each patient at the time of diagnosis, was shipped and analyzed at reference hospitals. The AML PETHEMA diagnostic network employs harmonized NGS protocols for analysis and reporting with external quality control rounds.¹⁰ All reference laboratories performed the analysis of at least 32 genes established by consensus due to their importance in AML In all cases, CEBPA gene was entirely sequenced.¹⁰ Mutation in bZIP-inf was considered if they are multiples of 3 bp and affect DNA binding, fork or bZIP from amino acid position 278 to C-terminus as previously stated.⁹

A total of 82 of 696 IT AML patients (11.8%) harbored CEBPA gene mutations by NGS. Among them, 45 had mutations within bZIP domain and 40 fulfilled criteria of CEBPA-bZIPinf (5.7%). Among CEBPA-bZIP-inf, 22 were CEBPAdm and 18 CEBPAsm. Online Supplementary Figure 1S shows patients flow chart classified from the detection of any CEBPA gene mutation to the final categorization as CEBPA-bZIP-inf. Main characteristics of the entire cohort according to CEBPA status (i.e, bZIP-inf vs. other CEBPA mutations [CEBPA other mut] vs. CEBPAwt are detailed in Table 1. Patients with CEBPA-bZIP-inf were significantly younger than other CEBPAmut and CEBPAwt (49.6 vs. 60.6 vs. 57.8 years respectively; P=0.009). Patients harboring CEBPA-bZIP-inf mutation had an estimated 3-year survival of 83.3% (95% confidence interval [CI]: 58.3-100) better than those with CEBPA other mut (54.3%, 95% CI: 34.9-84.4) and those with CEBPAwt (47.2%, 95% CI: 41.5-53.7) albeit no statistical differences were reached (P=0.17 for both comparisons) (Figure 1A). In order to seek the prognosis importance of being strictly “inframe” bZIP mutations, we performed similar OS analyses grouping all bZIP mutations (including not-inframe). Thus, patients harboring grouped CEBPA-bZIP mutations had also a 3-year survival (87.5%, 95% CI: 67.3-100) better than those with CEBPA other mut (47.9%, 95% CI: 27.9-82.2) and those with CEBPAwt (47.2%, 95% CI: 41.5-53.7; P=0.068 for both comparisons) (Figure 1B).

The mutational landscape found in patients with bZIP-inf (median number of mutations 1.5, range 0-5) compared CEBPA other mut (median number of mutations 2.5, range 0-7) is displayed in Figure 2A. We identified at least one mutation in any of the genes included in the study panel in 82 patients with CEBPA mutations (median 2; range, 0-7). Only 17 patients (13.69%) had no additional mutations. From these 17 patients (47% CEBPA-bZIP-inf), karyotype data was only available for five patients (2 patients had intermediate risk, 1 had favorable risk and 2 had normal karyotype). We identified significantly higher percentage of mutations in WT1, GATA2 y C-KIT in patients with CEBPA-bZIP-inf compared to CEBPA other mut (20% vs. 4.8%, 20% vs. 7.1% and 5% vs. 0%, respectively). By contrast, CEBPA other mut patients harbored significantly higher percentage of mutations in ASXL1 and NPM1 genes than CEBPA-bZIP-inf (19% vs. 2.5%, 28.5% vs. 2.5%, and 14.3% vs. 5% respectively). These differences are maintained when grouping all CEBPA-bZIP mutations compared to CEBPA other mut.

Table 1.Demographic and clinical characteristics of intensively treated acute myeloid leukemia patients including wild-type CEBPA, bZIP in-frame CEBPA mutation and other CEBPA mutation.

Additionally, we performed analysis to infer the timing of co-mutation occurrence among all CEBPA-mutated AML patients by using the Bradley-Terry model.¹¹ As displayed in Figure 2B, TP53 mutations, chromatin modificators mutations (ASXL1, EZH2), epigenetic regulators (TET2, DNMT3A) and splicing machinery mutations (SF3B1, UA2F1) seem to occur earlier. By contrast, mutations in NPM1 and signaling pathways (NRAS, KRAS and FLT3) seem to occur later.

Finally, we studied the impact on OS of the presence of co-mutations in CEBPA-bZIP-inf and CEBPA other mut. The presence of mutations in WT1 and GATA2 genes did not modify the prognosis of CEBPA-bZIP-inf patients. In the same way, no statistical differences were found analyzing the impact of mutations in the myelodysplasia-associated genes SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, RUNX1 or STAG2 in patients with CEBPA-bZIP-inf. However, the presence of mutations in TET2 genes conferred worse outcomes to CEBPA-bZIP-inf (P=0.064) and FLT3 mutations conferred significant worse outcomes only to CEBPA other mut patients (P=0.042). These differences are maintained when grouping all CEBPA-bZIP mutations compared to CEBPA other mut.

Since the first description of mutations in CEBPA gene in AML, the definition of which type had diagnosis entity and clinical prognostic impact, has evolved.³ Currently, WHO 2022 includes biallelic CEBPA mutations (independent of the gene region) and single mutations located in the bZIP region,¹² but ICC only accepts a narrower definition CEBPAbZIP-inf mutations (independent of the allelic state)¹³ as defined also as favorable prognosis category in ELN 2022.² Therefore, this recent step forward of ICC and ELN2002 statements implies a meaningful paradigm shift in the way AML with CEBPA mutations must be diagnosed and prognostically defined. This notable change has been made based mainly on the results of two large series of pediatric and adult patients intensively treated in clinical protocols and analyzed using different methologies.^8,9

We confirm that CEBPA-bZIP-inf is associated with favorable prognosis among fit AML patients intensively treated, but we also suggest that all CEBPA-bZIP (inframe and others) could be categorized as favorable risk. These findings were also in agreement with those reported by Taube et al. when analyzed only bZIP-inf and all bZIP mutations differentially.⁹ Importantly, other studies have reported favorable outcomes when grouping all bZIP mutations.^8,14 Altogether, these results question whether the restriction to bZIP-inf mutations as defined by ICC and ELN2022 has a meaningful clinical or diagnostic impact, while it is sure that it could increase complexity when reporting and interpreting these mutations. Moreover, although some data suggest that CEBPA bZIP mutant does not downregulate miR-182 and this incapability could be restricted to typical bZIP-inf mutations,¹⁵ there is no a clear evidence of CEBPA-bZIP-inf as a biological distinct entity.

Figure 1.Overall survival probability curves (%) of intensively treated acute myeloid leukemia patients. (A) Patients with bZIP in-frame (inf) CEBPA mutation CEBPA-bZIP-inf (red line), with other CEBPA mutation (mut) (green line) and wild-type CEBPA (CEBPAwt) mutation (blue line). (B) Patients with all CEBPA-bZIP mutations including non-inframe (CEBPA bZIP, red line), with CEBPA other mut (green line) and CEBPAwt (blue line).

Additionally, we observed that CEBPA-bZIP-inf carried more frequently well-known co-mutations as WT1 and GATA2, whereas in CEBPA other mut, ASXL1 mutations and NPM1 mutations were more frequent. It is important to remark the relatively frequent co-existence of mutations in NPM1 and CEBPA, both defining diagnostic entities in current WHO and ICC classifications, which is homogeneously found in up to 5% of CEBPA-bZIP-inf cases in our series and others.^8,9,14 This finding opens the question of the real-independent diagnosis entity of CEBPA-bZIP-inf which should be mutually exclusive with other genetically defined AML entities. The role of co-mutations in CEBPA-mutated patients has been extensively analyzed with discordant results. Prior studies reported inferior outcomes among CEBPAdm with G ATA or WT1 co-mutations.^3,5 However, when restricting the analyses to the bZIP-inf, the negative impact of these co-mutations is less clear since conflicting results have been published.^8,9,14 In our series, neither G ATA nor WT1 mutations adversely impact clinical outcome. Interestingly, the presence of mutations in TET2 gene could negatively impact prognosis in CEBPA-bZIP-inf patients in agreement with Taube et al.⁹

Figure 2.Genomic characterization of CEBPA-mutated patients. (A) Co-mutational spectrum of the 82 patients with CEBPA mutations. Each column represents 1 subject. (B) Comparison of percentage of patients with additional specific gene mutations between bZIP in-frame CEBPA mutation (CEBPA-bZIP-inf) and other CEBPA mutation (CEBPA other mut). (B) Bradley-Terry model to infer the timing of co-mutation occurrence by using the Bradley-Terry model.¹¹ As displayed TP53 mutations, chromatin modificators mutations (ASXL1, EZH2), epigenetic regulators (TET2, DNMT3A) and splicing machinery mutations (SF3B1, UA2F1) seem to occur earlier. By contrast, mutations in NPM1 and signaling pathways (NRAS, KRAS and FLT3) seem to occur later. MDS: myelodysplatic syndrome; TF: tissue factor.

The strengths of our study are to be a very large series of real-life consecutive patients, homogeneously analyzed within a harmonized NGS nationwide platform. In summary, CEBPA-bZIP-inf confer a favorable OS, compared to CEBPA other mut, but the narrow definition of in-frame could not be clinically relevant while increasing complexity for routine practice. although larger series are undoubtedly needed to firmly conclude this statement.

Footnotes

• Received November 11, 2023
• Accepted April 10, 2024

Correspondence

*EPdlT and JS contributed equally as first authors

J. SÁNCHEZ GARCÍA - joaquin.sanchez@cheerful.com

Disclosures

No conflicts of interest to disclose.

Funding

Acknowledgments

The authors would like to thank María D. García, Carlos Pastorini, Asier Laria, Yolanda Mendizabal, and Teresa Martinez Sena for data collection and management.