Prognostic impact of CEBPA mutational subgroups in adult AML

Despite recent refinements in the diagnostic and prognostic assessment of CEBPA mutations in AML, several questions remain open, i.e. implications of different types of basic region leucin zipper (bZIP) mutations, the role of co-mutations and the allelic state. Using pooled primary data analysis on 1010 CEBPA-mutant adult AML patients, a comparison was performed taking into account the type of mutation (bZIP: either typical in-frame insertion/deletion (InDel) mutations (bZIPInDel), frameshift InDel or nonsense mutations inducing translational stop (bZIPSTOP) or single base-pair missense alterations (bZIPms), and transcription activation domain (TAD) mutations) and the allelic state (single (smCEBPA) vs. double mutant (dmCEBPA)). Only bZIPInDel patients had significantly higher rates of complete remission and longer relapse free and overall survival (OS) compared with all other CEBPA-mutant subgroups. Moreover, co-mutations in bZIPInDel patients (e.g. GATA2, FLT3, WT1 as well as ELN2022 adverse risk aberrations) had no independent impact on OS, whereas in non-bZIPInDel patients, grouping according to ELN2022 recommendations added significant prognostic information. In conclusion, these results demonstrate bZIPInDel mutations to be the major independent determinant of outcome in CEBPA-mutant AML, thereby refining current classifications according to WHO (including all dmCEBPA and smCEBPA bZIP) as well as ELN2022 and ICC recommendations (including CEBPA bZIPms).


INTRODUCTION
Mutations of the gene encoding the CCAAT-enhancer binding protein alpha (CEBPA) transcription factor are common genetic alterations in acute myeloid leukemia (AML).Since their first description [1], several groups have investigated CEBPA mutations in AML (reviewed in [2]).After initial studies suggested that all patients with CEBPA mutations carry a more favorable outcome [3][4][5][6][7], subsequent analyses have consistently demonstrated that this improved prognosis is confined to biallelic or double mutations (dmCEBPA).DmCEBPA was shown to be associated with a distinct biology and to confer a more favorable clinical outcome, including higher rates of complete remission (CR), reduced relapse risk, and increased overall survival (OS), whereas single allele mutations were considered prognostically irrelevant [8][9][10][11][12][13][14].However, several recent reports looking in more detail for the impact of individual mutations in CEBPA suggested that the specific clinical and molecular characteristics as well as the favorable prognosis were restricted to mutations within the basic leucine zipper region (bZIP) region of CEBPA, irrespective of their occurrence as double or single mutation [15][16][17].Gene expression analysis further supports a unique biology of CEBPA bZIP mutations in AML [15,16].However, there is evidence that even within this mutational subgroup, biological differences may exist depending on the particular type of CEBPA bZIP mutation.Our group observed significant differences in outcome and molecular profiles when comparing patients with in-frame CEBPA bZIP mutations (either in-frame insertions/deletions or single base pair missense mutations) and patients with frameshift or nonsense mutations [16].
These findings have provided the basis for a refined biological and clinical classification of CEBPA mutations.CEBPA bZIP in-frame mutations are now being classified as favorable risk entity in the 2022 update of the European LeukemiaNet (ELN) recommendations on genetic risk classification [18] and in the International Consensus Classification [19].However, the 2022 WHO classification still defines the CEBPA mutational class via the presence of a double mutant allelic status or a single mutant bZIP mutation regardless of the type of mutation(s)/exact DNA alteration [20].This persistent ambiguity might be due to the fact that few reports have investigated the impact of different CEBPA mutational constellations in more detail, so the evidence supporting either of these modifications is still limited.
To gain further insights into the impact of different mutational subtypes, in particular the spectrum of CEBPA bZIP mutations, a pooled primary data analysis was performed involving detailed sequencing data as well as clinical variables and treatment outcome from more than 1000 CEBPA mutant AML patients.The main objective of this study was to investigate potential differences between different types of bZIP mutations and to examine the relevance of the allelic status (double vs. single mutant).
Our results, generated in the largest cohorts of patients with CEBPA mutant AML reported so far, strongly support the notion that the CEBPA bZIP InDel genotype introduced in this work (bZIP inframe insertions/deletions, double and single mutant) shows a specific biology and favorable prognostic implications, whereas the other CEBPA mutational subgroups, i.e.TAD mutations as well as bZIP missense and frameshift/nonsense mutations, differ substantially with respect to most clinical as well as molecular factors studied.
This analysis establishes the basis for a more accurate refinement of current classifications and highlights the need for additional research efforts to elucidate the specific biological effects of CEBPA bZIP InDel mutations and their role in leukemogenesis.S1).Patients were treated in prospective trials (details on study protocols are given in the supplement) involving risk stratified post induction therapy according to cytogenetic risk groups, including the option for an allogeneic hematopoietic cell transplantation (alloHCT) in CR1, or recruited to AML registries and biorepositories.Only 40 patients (3.9%) were treated after the release of the ELN 2016 guidelines, which proposed stratification by allelic status (dmCEBPA vs. smCEBPA) within the group of CEBPA-mutant patients.Otherwise, CEBPA mutations were not used for risk stratification in any of the trials.For each individual patient, a predefined minimal data set was collected, including clinical variables, i.e. age, sex, date of AML diagnosis, type of AML (de novo or secondary/therapy-related), bone marrow (BM) blast count, white blood cell (WBC) count, type of and response to induction chemotherapy, date of alloHCT in CR1, date of alloHCT beyond CR1 and events (i.e.induction failure, relapse, death) as well as genetic variables (karyotype, mutational status of NPM1, FLT3, GATA2, DNMT3A, IDH1, IDH2, WT1 and other genes, if available).Co-mutational data sufficient for genetic risk stratification according to the ELN2022 guidelines were available for 645 patients (63.8%).

MATERIALS/SUBJECTS AND METHODS
The information collected included complete sequencing results (performed either by Sanger sequencing or next generation sequencing; NGS) of the entire CEBPA gene (Genbank Accession No. NM_004364.2).All retrieved CEBPA sequences were evaluated for the precise localization of the mutation, i.e. bZIP vs. transcription activation domains [17] 1 and 2, allelic status (smCEBPA vs. dmCEBPA) as well as the type of mutation, i.e. insertions/deletions either in-frame or frameshift, missense mutations as well as nonsense mutations.
This study was performed in accordance with the Declaration of Helsinki, all clinical studies and data registries were approved by the local institutional review boards, and written informed consent was obtained from all patients through the participating centers.
For statistical analysis, comparisons of categorical variables between groups were done with the Chi-squared test.Continuous variables were compared with the Kruskal-Wallis-Test between groups.OS was calculated from date of study entry until death, or last follow-up visit.Relapse free survival (RFS) was calculated from date of first remission until date of relapse, date of death, or date of last follow-up visit.Survival endpoints were analyzed with the Kaplan-Meier method.Cox regression models were fitted to estimate hazard ratios.AlloHCT as adjusting variable in multiple models was modeled as time-dependent covariate.Univariate and multivariate logistic regression models were used to estimate odds ratios for achievement of CR1.Individual patient data of the different study groups were pooled.All Cox regression models were stratified for study group.In logistic regression models study group was modeled as a factor.To estimate variability between study groups, all analyses were conducted per study group and interaction of CEBPA categories with study group was assessed.Additionally, the estimates from the study groups were pooled via inverse variance method and heterogeneity statistics were estimated.AlloHCT was modeled as time-dependent covariate.A landmark of 3 months from study entry was applied to reduce bias due to early deaths disqualifying patients for alloHCT.

CEBPA mutational status
Of the 1010 patients, 661 patients (65.4%) showed mutations affecting the bZIP-domain of CEBPA encompassing amino acids (AA) 272-358 [1].As illustrated in Fig. 1, mutations in the CEBPA bZIP region were typically in-frame insertions or deletions (n = 491 pts; 74.3%; referred to as bZIP InDel ), i.e. (multiples of) 3 bp affecting the DNA-binding-, fork-or bZIP-region.Frameshift insertions/deletions or nonsense mutations in bZIP, causing premature termination of transcription (referred to as bZIP STOP ), as well as missense mutations (bZIP ms ), causing single AA changes, were less common and were observed in 81 (12.3%) and 89 (13.5%) of patients, respectively.Interestingly, the different types of mutations showed a non-random distribution, with bZIP ms mainly clustering in certain critical AA positions in the DNAbinding basic region, i.e.AA297 and AA300.In contrast, most bZIP InDel mutations affected the hinge/fork region of the bZIPdomain, i.e., the 14 AAs preceding the first leucine residue of the leucine zipper at position 317 [21], with the most frequent InDel mutations at AA312 and AA313 (Fig. 1).In contrast to the bZIP region, the TAD domains almost uniformly harbored frameshift InDel or nonsense mutations, creating a premature termination codon.

CEBPA mutational subgroups and clinical characteristics
As outlined in Table 1, the association of several clinical parameters differed substantially between the defined mutational subgroups.Patients with bZIP InDel mutations, i.e., Gr1 and Gr5, were significantly younger (median age Gr1 42.2 years [IQR 31-54.9];Gr5 47 years [IQR 39-58]) than those without bZIP InDel   mutations (groups 2-4 and groups 6-8) (median age 52-64 years).They also had a higher prevalence of de novo AML (98% and 96%) compared to patients without bZIP InDel mutations (groups 2-4 and groups 6-8), with the latter more frequently evolving as secondary disease after prior MDS or as tAML (rate of de novo AML 81-94%) (Table 1).Categorizing age in 10-year intervals (Fig. 3, Table S2), a continuous decrease in the occurrence of bZIP InDel mutations (especially dmCEBPA bZIP InDel ) was seen with increasing age, whereas bZIP STOP and bZIP ms mutations and alterations affecting only the TAD regions were particularly common in older individuals and less prevalent in patients up to the age of 40 years.Other clinical parameters did not differ significantly between subgroups.

CEBPA mutational subgroups and response to treatment
Treatment response data were available for 992 patients (98.2%).Outcome analysis showed comparability of CR rates as well as RFS and OS between study groups (Table S1).Regarding initial response to therapy within the eight different subgroups, patients with bZIP InDel mutations achieved the highest rates of CR1, with 94.3% in Gr1 (OR 6.38 [3.83-10.63],p < 0.001) and 92.1% in Gr5 (OR 4.51 [1.71-11.86],p = 0.002) compared to CR-rates of 73.1-79.6%observed in the other subgroups (Table 1, Table S4).
Impact of ELN2022 mutational subgroups in CEBPA mutant patients without bZIP InDel (subgroups 2-4 and 6-8) Recent analysis suggested that certain co-mutations, in particular mutant NPM1, might have an effect in patients with smCEBPA [29].
To gain further insights on the impact of co-mutations in patients without bZIP InDel mutations, i.e. groups 2-4 and 6-8 (CEBPA other ), a combined analysis based on the ELN2022 guidelines was performed for these patients.A total of 345 patients (63.8%) had sufficient cytogenetic and molecular data to allow reclassification according to the ELN2022 risk groups.Most of these patients (n = 183/345) (53%) were assigned to the adverse risk group, predominantly due to ASXL1 (n = 85), RUNX1 (n = 63), SRSF2 (n = 56) and STAG2 (n = 55) mutations, while only 27 patients (14.8%) had poor risk cytogenetics.
In the patients re-assigned to the favorable risk group (n = 88), all but one patient (showing a t(8;21)) had NPM1-mutations.As depicted in Fig. 6B, the outcome of these groups showed statistically significant differences, with the median RFS not reached and a median OS of 154 months in the CEBPA other / ELN2022 favorable risk group compared to 16 months and 31 months in the CEBPA other /ELN2022 intermediate risk group and 12 months and 16 months in the CEBPA other /ELN2022 adverse risk group (p < 0.001; for multivariate analysis see Tables S16 and S17).

DISCUSSION
This study examined the prognostic impact of different CEBPA mutational subgroups in detail in a large cohort of patients.The prevalence of mutations in CEBPA ranges between 5 and 10 [8], therefore the 1010 CEBPA-mutant patients investigated in this study Fig. 4 Frequency distribution of additional gene mutations identified in the different CEBPA-mutant subgroups.Frequencies are shown for genes with a prevalence of at least 10% in the total CEBPA-mutant cohort.
correspond to a total of 10.000-20.000adult patients with AML.Based on these large numbers, this analysis for the first time allowed us to address several questions which had remained unclear or controversial in previous investigations, namely the impact of missense mutations in the bZIP region, the impact of co-mutations and cytogenetics in patients with bZIP InDel mutations as well as the prognosis of CEBPA mutant patients without CEBPA bZIP InDel mutations.
An important conclusion from our study is that it clearly supports the previous findings on the unique behavior of bZIP mutations compared to other types of mutations in CEBPA.The results confirm that irrespective of the allelic state, CEBPA bZIP InDel mutations define a distinct subgroup characterized by younger age and a specific comutational profile, including a high rate of GATA2 and WT1 mutations and mutual exclusiveness of other subtype-defining lesions like mutations in NPM1.In addition, CEBPA bZIP InDel mutations were associated with a very favorable outcome.Patients with bZIP InDel mutations, especially those with double mutant CEBPA, demonstrated a 5-year overall survival rate above 75%, indicating that these patients should be considered as one of the AML subgroups with the best response to conventional treatment.In line with this, alloHCT performed in CR1 did not improve outcome in this group.However, relapsed bZIP InDel patients appear to benefit from alloHCT as part of salvage treatment which is most strikingly demonstrated in the small subgroup of patients with CEBPA bZIP InDel mutations and CSF3R comutations, which showed a highly significantly decreased RFS, but no difference in OS due to successful salvage treatment.These results confirm similar observations in pediatric patients [15].Co-mutations in several other genes have been associated with prognosis in CEBPAmutant AML.In particular, GATA2 co-mutations were reported to confer a better prognosis in dmCEBPA patients by some groups [23,30], although this was not confirmed by others [22,24].The data presented here provide a possible explanation for these discrepant reports.GATA2 mutations were predominantly found in patients with bZIP InDel mutations, with a significantly lower prevalence in patients with other dmCEBPA mutations (Gr.2-4).Given that GATA2 mutations had no impact on outcome when analysis was restricted to bZIP InDel patients (Fig. S5), this suggests that GATA2-mutation status might be a surrogate parameter for bZIP InDel -mutations, and therefore associated with better outcome in some studies.The same might be true for several other mutations, i.e., FLT3 and TET2, which have a significantly higher prevalence in non-bZIP InDel patients and had previously been shown to be associated with inferior outcome in CEBPA-mutant patients in some studies [23,28,[31][32][33].Very recently Tet2-mutations have been demonstrated to enhance aggressiveness of Cebpa-mutant disease in animal models [34].Although significant differences were observed for mutant TET2 and FLT3 in univariate analysis, multivariable analysis did not confirm an independent effect of these alterations in bZIP InDel patients.In addition, a combined analysis of adverse molecular and cytogenetic prognostic factors according to ELN 2022 recommendations failed to indicate a significant prognostic impact in the bZIP InDel group.Taken together, these data suggest that CEBPA bZIP InDel -mutant patients represent a unique subgroup of patients with AML.
In contrast, patients with bZIP STOP , bZIP ms or TAD mutations, irrespective of allelic status, showed a different biology and worse outcome.In our previous analysis, bZIP ms mutations were grouped and analyzed together with bZIP InDel as "bZIP in-frame", which corresponds to the definitions of the current ELN and ICC guidelines [18,19].However, in the current analysis of a larger cohort of patients, bZIP ms mutations were clearly associated with an inferior outcome when evaluated separately, and were clinically and molecularly distinct from bZIP InDel mutations, while sharing more commonality with the other CEBPA subgroups.
WHO continues to include biallelic CEBPA mutations as a defined subgroup [20].However, our results indicate that patients with dmCEBPA without bZIP InDel mutations, i.e. patients showing either two TAD mutations, TAD and bZIP STOP or TAD and bZIP ms , differ substantially in biology and outcome, suggesting that only bZIP InDel mutations and not bZIP ms mutations or any other dmCEBPA mutations should be included in this specific AML subgroup (Fig. 7).This extends previously published data by El-Sharkawi et al. which already provided evidence for a differential effect of different double mutant constellations [35].
Interestingly, our data indicate that the different CEBPA bZIP mutational subtypes, i.e. in-frame InDel mutations, InDel mutations inducing frameshift and missense mutations are distributed in a non-random way in the bZIP region, raising the possibility that mutation location impacts on the functional consequences.As illustrated in Fig. 1, CEBPA InDel mutations significantly clustered in the first part of the leucine zipper (between AA310-317) as well as the fork region (AA303-309), whereas missense mutations were significantly more common in the basic region (especially in several highly conserved amino acids, e.g.R300 and R297), which are directly involved in DNA-binding (reviewed in [36]).In contrast, the fork or hinge-region of bZIP-proteins is considered to be especially important for the spacing of the two alpha-helices of the leucine zipper, which in turn could influence either binding specificity and/or affinity of the DNA-binding [21] as well as the interaction with other proteins.CEBP proteins bind DNA as homo and heterodimers, and the CEBPA-interactome appears to be complex and still not completely understood [37].Even though the presence of adverse risk aberrations according to the ELN 2022 recommendations was rare in CEBPA bZIP InDel mutant patients, knowledge of their prognostic implications is crucial for the choice of post-remission treatment, as it might abrogate the presumed prognostic advantage and low risk of relapse in these AML patients.In this analysis, patients with CEBPA bZIP InDel demonstrated superior survival irrespective of concurrent high-risk features.However, it is important to note that the impact of adverse genetic factors according to ELN 2022 in CEBPA mutant AML may vary depending on the specific chromosomal or molecular abnormalities.
Aside from CEBPA bZIP InDel , the other mutational subgroups do not appear to have an independent prognostic value.Analysis based on concomitant cytogenetic and molecular alterations according to current ELN 2022 recommendations within the CEBPA other patient group showed that they conformed to the expected risk stratification group, with no evidence that the CEBPA mutation had substantially changed the outcome.For example, a more favorable outcome in these groups was usually attributable to a concomitant NPM1 mutation.
Although our study represents the largest cohort of CEBPAmutant AML, the analysis also has some limitations, in particular the retrospective nature of the analysis covering a period of almost three decades in which patients were treated.Consequently, none of the patients included were treated with novel targeted agents, e.g.tyrosine kinase inhibitors or Venetoclax/ HMA-based therapies, which might affect outcome, at least in subgroups, and the impact of such agents on the different subgroups will be an important issue for future analyses.
In conclusion, CEBPA bZIP InDel -mutant AML represent a subset of AML with profoundly distinct disease biology and clinical outcomes.Further research efforts aimed at elucidating the underlying molecular mechanisms and identifying additional genetic and epigenetic alterations that interact with CEBPA mutations are necessary to harness the full potential of CEBPA bZIP InDel mutations in improving the management and prognosis of these AML patients.

Fig. 1
Fig. 1 Illustration of type and localization of mutations affecting the bZIP region of the CEBPA gene.Mutations in the bZIP region (AA272-358) are typically in-frame insertion or deletion mutations (bZIP InDel ), whereas frameshift insertions/deletions or nonsense mutations causing a premature translational termination (bZIP STOP ), and single base-pair missense mutations (bZIP ms ) are less common.

Fig. 2
Fig.2Schematic illustration of the predominant mutational constellations in the eight different CEBPA-mutant subgroups.Subgrouping was performed according to different types of mutations, i.e.TAD or bZIP mutations (in-frame insertions/deletions, frameshift insertions/ deletions and nonsense mutations, missense mutations) and allelic status (double vs. single mutant).

Fig. 5
Fig. 5 Impact of different CEBPA-mutant subgroups on outcome.Kaplan Meyer estimates for RFS and OS in A Gr1-8 and B Gr1 and Gr5 (dmCEBPA bZIP InDel and smCEBPA bZIP InDel) vs. Gr2-4 and Gr6-8 (CEBPA other ).HR and p values were calculated by Cox regression analysis.

Fig. 6
Fig. 6 Impact of ELN2022 risk factors in CEBPA-mutant subgroups.Kaplan Meyer estimates for RFS and OS of A CEBPA bZIP InDel patients with vs. without ELN2022 adverse risk factors and B different ELN2022 risk groups within CEBPA other patients.HR and p values were calculated by Cox regression analysis.

Fig. 7
Fig. 7 Re-grouping of CEBPA-mutant patients according to localization and type of mutation(s).Illustration of the frequency of different CEBPA mutational subtypes and their allocation to the proposed subgroups CEBPA bZIP InDel and CEBPA other .

Table 1 .
Clinical and genetic characteristics of the eight different CEBPA mutant subgroups.
a according to karyotype and comutations, irrespective of CEBPA mutational status.