CEBPA mutations in 4708 patients with acute myeloid leukemia: differential impact of bZIP and TAD mutations on outcome

Patients with normal karyotype represent the largest subgroup of patients with acute myeloid leukemia (AML).¹ Research performed during the past 20 years has revealed several novel abnormalities that are particularly common in this subgroup of patients (eg, activating mutations of the FLT3 receptor tyrosine kinase^2-5 and mutations of the NPM1 gene encoding the nucleophosmin protein.^6-8) In addition, mutations of the gene encoding CCAAT/enhancer-binding protein-α (CEBPA) have been described in patients with AML and normal karyotype.^9-12 The CEBPA gene encodes a transcription factor that serves as a master regulator of granulopoiesis.¹³ Targeted disruption of the gene in mice is associated with a block of granulocyte development and downregulation of target genes such as granulocyte colony-stimulating factor receptor.¹³ The intronless CEBPA gene on chromosome 19q13.1 encodes 2 major protein isoforms, the 42-kDa full-length protein (p42), which has the full transcriptional activity, and a shorter, 30-kDa isoform (p30), produced from a second, alternative start codon and shown to have complex functions, including an inhibitory effect on⁹ and increased degradation¹⁴ of the p42 full-length protein.

Since the initial report,⁹ several groups have investigated CEBPA mutations in AML.^11,12,15-18 Besides the reproducible association with certain morphological and clinical features (eg, FAB M1 and M2 morphology, high CD34 expression on blasts, and predominance in normal karyotype), nearly all studies have shown that patients with CEBPA mutations have a more favorable outcome. However, more recent results have indicated that a good prognosis is confined to patients with biallelic or double mutations (CEBPA^bi),^19-21 whereas patients with monoallelic CEBPA mutations (CEBPA^sm) did not differ in their response to treatment from patients with wild-type CEBPA (CEBPA^wt) and had a less favorable outcome.^19-21 This finding led to the inclusion of biallelic CEBPA mutations as an independent entity in the most recent World Health Organization classification,²² as well as a favorable prognostic group in the ELN2017 recommendations.²³ However, the impact of monoallelic CEBPA mutations has been investigated in more detail in only a few studies, especially in light of the different biological effects of N- and C-terminal mutations.^21,24,25

To investigate the prevalence and prognostic role of CEBPA mutations, in particular CEBPA^sm mutations, in adult patients with AML, we studied 4708 patients with newly diagnosed AML. In contrast to many previous studies, we included patients of all age groups, as well as secondary AML (sAML) after prior myelodysplasia (MDS) or therapy-related AML (tAML). Because most CEBPA mutations reported so far consist of insertions or deletions,^26-29 we used high-resolution fragment analysis for mutation screening and found CEBPA mutations in 5.1% of patients. The results of targeted next-generation sequencing (NGS) indicated profound differences in the co-mutational spectrum of individual CEBPA mutations. Our results point to a differential effect of CEBPA^smbZIP mutations that appear to be associated with similar clinical parameters, co-mutations, and outcome compared with CEBPA^bi mutations. If confirmed, these results, generated in one of the largest cohorts of patients with AML analyzed so far, could build the basis of a refined clinical classification of CEBPA mutations.

We screened 4708 adult patients with newly diagnosed AML (n = 3729 with de novo AML; n = 644 with AML and a history of MDS; and n = 335 with tAML) for the presence of mutations in the CEBPA gene. Most individuals (n = 3104; 67.45%) were treated in prospective studies, including the AML96 (n = 1457), AML2003 (n = 1081), AML60⁺ (n = 359), and SORAML (n = 207) protocols of the Study Alliance Leukemia (SAL). The remaining patients (n = 1604) were recruited to the SAL registry and biorepository. Detailed treatment protocols have been published^30-33 and are summarized in the supplemental Data (available on the Blood Web site), including the number of patients treated in each protocol. All studies involved risk-stratified consolidation therapy according to cytogenetic risk groups (CEBPA mutations were not used for risk stratification in any of those studies). Patients <60 years of age received standard cytosine arabinoside (Ara-C)/anthracycline (DA 3 + 7)–based, double-induction chemotherapy followed by consolidation with high-dose Ara-C in patients who had favorable risk. Patients in the intermediate- and high-risk groups had the option of upfront allogeneic transplantation and underwent autologous transplantation or chemotherapy-based consolidation in the absence of suitable donors. In older patients treated with curative intention, the chemotherapy regimen was adjusted according to predefined algorithms that integrate performance status and organ function.

This study was approved by the ethics board of the Technical University Dresden. Each patient gave written informed consent to participate in the respective study protocols.

All materials investigated were obtained at the time of diagnosis. Bone marrow was used whenever available; in all other cases, peripheral blood samples were examined. Genomic DNA was extracted from mononuclear cells by using standard procedures (DNA blood minikit; Qiagen, Hilden, Germany).

All polymerase chain reaction (PCR) analyses were performed on genomic DNA. Details of the PCR primers and cycling conditions are given in supplemental Tables 1 and 2 and supplemental Figure 1.

PCR-amplified mutant samples were purified and sequenced on an ABI3130xl instrument. Sequences were compared with the WT-CEBPA messenger RNA (mRNA) sequence (U34070).

Profiling of mutations was achieved by targeted NGS-based resequencing with the TruSight Myeloid assay (Illumina, Chesterford, United Kingdom) covering 54 genes frequently mutated in AML, as described recently³⁴ (details in the supplemental Data). Data alignment of demultiplexed FastQ files, variant calling, and filtering were performed with the Sequence Pilot software package (JSI Medical Systems GmbH, Ettenheim, Germany), with default settings and a 5% variant allele frequency cutoff. FLT3-internal tandem duplication (ITD) and NPM1 mutations were evaluated as reported previously.^5,8

RNA sequencing (RNA-Seq) was performed on total RNA isolated at diagnosis from 20 patients with a CEBPA mutation (5 CEBPA^bi-inf, 5 CEBPA^sm-inf, 5 CEBPA^sm-other, and 5 CEBPA^bi-other), by using strand-specific RNA-Seq library preparation (Ultra II Directional RNA Library Prep; New England Biolabs) and sequenced on an Illumina NovaSeq 6000 instrument. The complete workflow and the bioinformatic analyses are detailed in the supplemental Data.

Clinical variables across groups were compared by using the χ² test or 2-sided Fisher’s exact test for categorial variables. The nonparametric Mann-Whitney U test was applied for continuous variables. P < .05 indicated a significant difference. Numerical variables are expressed as the median with interquartile range (IQR). Univariate analyses for the influence of the CEBPA mutational status on complete response (CR) rates were performed by using the χ² test. The log-rank test was used to evaluate OS and EFS. For multivariable analysis of prognostic factors, Cox proportional hazards regression models were used. Allogeneic hematopoietic stem cell transplantation (allo-HSCT) was modeled as a time-dependent covariate. P-values of association analyses of CEBPA mutations with clinical variables and other molecular abnormalities were adjusted for multiplicity by using the Bonferroni-Holm correction. All statistical analyses were performed with the SPSS software package, version 26 (SPSS, Chicago, IL) and R version 3.5.3 (https://www.R-project.org/).

A total of 371 individual CEBPA mutations were identified in 240 of the 4708 patients (5.1%) by the screening procedure. In all patients with a mutation identified in any part of CEBPA, Sanger sequencing of the entire gene was performed to assess the presence of additional mutations, including single nucleotide variants, and to confirm the alterations identified using fragment analysis. Patients showing the previously described 6-bp polymorphism in TAD2²⁶ were included in the WT-CEBPA group.

As illustrated in supplemental Figure 2 and supplemental Table 3 and reported in previous studies, mutations largely clustered in the N-terminal first transactivation domain (TAD1) and the DNA-binding and basic leucine zipper region (bZIP) in the C-terminal part of CEBPA. In our cohort, 131 of 240 patients (54.6%) presented with 2 CEBPA mutations, mostly consisting of combined mutations in bZIP and TAD, denoted as CEBPA^bi. CEBPA^sm mutations were found in 109 of 240 patients (45.4%), of which 60 had N-terminal (CEBPA^smTAD) and 49 C-terminal (CEBPA^smbZIP) mutations. As reported before, most of the mutations in the TAD domains caused a frameshift, whereas mutations in the bZIP-region were predominantly in-frame insertions and duplications. Eight patients had monoallelic mutations with high variant allele frequency (4 CEBPA^smTAD and 4 CEBPA^smbZIP), indicating a homozygous state.

The association of clinical parameters according to the localization of the CEBPA mutation (ie, patients with single N- or C-terminal mutations and patients with biallelic mutations) is summarized in Table 1. Compared with patients bearing CEBPA^wt, those with CEBPA^bi were significantly younger (median age, 46 years; IQR, 38-59) at diagnosis, similar to those with CEBPA^smbZIP (median age, 50 years; IQR, 39-57), whereas patients with CEBPA^smTAD were significantly older (median age, 63 years; IQR, 55-69.3) and were more comparable to the CEBPA^wt group (median, 57 years; IQR, 46-67; P < .001). When patients were categorized in 10-year age intervals (Figure 1), a continuous decrease in CEBPA^bi and CEBPA^smbZIP mutations was seen with increasing age, whereas CEBPA^smTAD alterations increased. In line with this finding, only a single patient with a CEBPA^smTAD mutation was observed in the group of patients <30 years of age.

As outlined in Table 1, patients with CEBPA^smTAD also had significantly lower WBC counts and significantly lower rates of CD34 positivity, compared with the CEBPA^bi and CEBPA^smbZIP groups, and more commonly had prior myelodysplasia syndrome (MDS) or tAML that evolved as a secondary disease.

Based on the targeted NGS approach, additional mutations were identified in 208 of 240 patients with a CEBPA mutation (86.7%). The most frequently mutated genes were TET2 (70 of 240; 29.2%), GATA2 (68 of 240; 28.3%), DNMT3A (45 of 240; 18.8%), FLT3-ITD (39 of 240; 16.3%), NPM1 and NRAS (31 of 240; 12.9% each), and WT1 (30 of 240; 12.5%).

Figure 2 illustrates the distribution of co-mutations in the 3 CEBPA subgroups. Significant differences were observed for several genes, the most striking being GATA2, which was mutated in 35.1% of patents with CEBPA^bi, 36.7% of those with CEBPA^smbZIP, but in only 6.7% of those with CEBPA^smTAD (P = .001), and NPM1, which was mutated in 3.1% of patients with CEBPA^bi and 8.2% of those with CEBPA^smbZIP, but in 38.3% of patients with CEBPA^smTAD (P < .001). Significant differences were also found for mutations in DNTM3A, FLT3-TKD, IDH1, and IDH2, as well as in SRSF2 and WT1 (Figure 2B). In general, the spectrum of mutations of CEBPA^smbZIP was more comparable to that of patients with CEBPA^bi and differed markedly from the CEBPA^smTAD group, the latter being more similar to patients with CEBPA^wt, who frequently carried mutations in genes associated with AML after prior MDS, such as spliceosome mutations (ie, SRSF2, SF3B1, or U2AF1) and alterations associated with DNA methylation (ie, DNMT3A, ASXL1, TET2, IDH1, and IDH2). Interestingly, only 7 patients in this analysis had mutations in CSF3R, with no significant differences found between the groups (3.1% CEBPA^bi, 2% CEBPA^smbZIP, and 3.3% CEBPA^smTAD).

The prognostic relevance of CEBPA mutations was analyzed in 4461 intensively treated patients, with a median follow-up time of patients remaining alive of 61 months (IQR, 36-96 months). When analyzed according to the current recommendations (ie, separating only CEBPA^bi and CEBPA^sm), only patients with CEBPA^bi had a significantly higher rate of complete remission (CR; CEBPA^bi, 95% vs CEBPA^sm, 75% and CEBPA^wt, 73%; P < .001), as well as a higher median OS (CEBPA^bi, 103.2 months; 95% confidence interval [CI], 70.7-inf vs CEBPA^sm, 21.9 months; 95% CI, 12.7-54, and CEBPA^wt, 19.3 months; 95% CI, 17.9-21; P < .001) and EFS (CEBPA^bi, 20.7 months; 95% CI, 13.1-101.1 vs CEBPA^sm, 9.4 months; 95% CI, 5.7-15.3, and CEBPA^wt, 7.0 months; 95% CI, 6.5-7.6; P < .001; Figure 3), which was confirmed in multivariable analyses (Table 2).

Given the marked differences in the clinical and molecular alterations associated with the different localization of single-mutant CEBPA, we also looked for the impact of the localization of CEBPA^sm on outcome. Patients carrying CEBPA^smbZIP showed a significantly higher CR rate (CEBPA^bi 95% vs CEBPA^smbZIP 86% vs CEBPA^smTAD 66% vs CEBPA^wt 73%) and a significantly longer OS (CEBPA^bi 103.2 months; 95% CI, 70.7-inf, vs CEBPA^smbZIP, 63.3 months; 95% CI, 20.5-inf, vs CEBPA^smTAD, 12.7 months; 95% CI, 7.5-36.5, and CEBPA^wt, 17.9 months; 95% CI, 17.9-21) and EFS (CEBPA^bi, 20.7 months; 95% CI, 13.1-101.1 vs CEBPA^smbZIP, 17.1 months; 95% CI, 8.3-inf vs CEBPA^smTAD, 5.7 months; 95% CI, 2.6-10, and CEBPA^wt, 7.0 months; 95% CI, 6.5-7.6; Figure 4A-B). Multivariable analysis confirmed that CEBPA^smbZIP mutations represented an independent favorable risk for outcome (Table 3).

We compared the effect of co-mutations on the survival of the patients according to the mutational status of the 3 most common mutations previously associated with outcome (ie, GATA2, TET2, and WT1 in the 3 CEBPA mutation subgroups (CEBPA^bi, CEBPA^smbZIP, and CEBPA^smTAD). Although the presence of GATA2 mutations was associated with improved OS and EFS in the CEBPA^bi and CEBPA^smbZIP groups (Figure 5A-B); however, these differences were not significant in pairwise comparisons (all P-values were adjusted for multiple testing with the Bonferroni-Holm procedure). However, for TET2^mut (Figure 5B-C) a significant difference was found between CEBPA^bi/TET2^wt and TET2^mut (median OS: CEBPA^bi/TET2^wt, not reached; 95% CI, 101.1-inf, vs CEBPA^bi/TET2^mut, 22.5 months; 95% CI, 13.9-inf; P = .012; median EFS: CEBPA^bi/TET2^wt, 34.7 months; 95% CI, 13.5-inf vs CEBPA^bi/TET2^mut, 9.5 months; 95% CI, 5.23-50.2; P = .02) and a trend for a different EFS in patients with CEBPA^smbZIP/TET2^wt or CEBPA^smbZIP/TET2^mut (median EFS CEBPA^bZIP/TET2^wt, not reached; 95% CI, 9.59-inf vs CEBPA^bZIP/TET2^mut, 9.2 months; 95% CI, 2.40-66.2; P = .06). WT1 mutations were associated with lower probability of survival (supplemental Figure 4); however, the comparison between the groups did not reach statistical significance.

NPM1 mutations, the most common co-mutation in patients carrying CEBPA^smTAD, were associated with slightly better OS and EFS; however, these differences were also not significant (supplemental Figure 5).

Mutations in the CEBPA-bZIP region are typically in-frame (ie, multiples of 3 bp) and affect the DNA-binding-, fork-, or bZIP-region (for simplicity, summarized as bZIP) between amino acid positions 278 and 345 of the CEBPA protein (supplemental Figure 3), whereas frameshift mutations, the hallmark of mutations affecting the TAD1 and TAD2 domains, are less common in C-terminal mutations. To investigate whether the presence of typical in-frame bZIP mutations (irrespective of the biallelic or monoallelic status) actually represents the decisive molecular factor for the favorable outcome observed, we regrouped the 240 patients with CEBPA^mut according to the presence or absence of these mutations. Typical bZIP mutations according to this classification were found in 118 of 131 who had CEBPA^bi and 39 of 49 of those with CEBPA^smbZIP (denoted CEBPA^bi-inf and CEBPA^sm-inf). Thirteen of 131 CEBPA^bi had nontypical biallelic mutations (eg, consisting of 2 different TAD domain mutations or a combination of TAD mutations and frameshift or nonsense bZIP mutations, denoted CEBPA^bi-other). Patients with mutation in TAD as well as patients with a nontypical single-allele bZIP mutation were grouped as “other single-allele mutations” (CEBPA^sm-other). We again observed profound differences in the age distribution, with typical bZIP mutations (double and single allele) predominating in younger adults, whereas other mutations, especially the few nontypical biallelic mutations, were predominantly detected in patients >50 years of age (supplemental Table 4; supplemental Figure 6).

To further characterize potential biological similarities between CEBPA^bi-inf and CEBPA^sm-inf, we performed RNA-seq of 20 samples from patients with CEBPA^bi-inf, CEBPA^sm-inf, CEBPA^bi-other, and CEBPA^sm-other (5 samples per group). As illustrated in the volcano-plots in supplemental Figure 7, we found between 34 and 129 differentially expressed genes in pairwise comparisons between all groups, the only exception being CEBPA^bi-inf and CEBPA^sm-inf, where no significant differences were found at a false discovery rate of 5% (supplemental Figure 7C).

The survival analysis performed also revealed profound differences. Superimposable, favorable outcomes (OS and EFS) were confined to patients in the CEBPA^bi-inf and the CEBPA^sm-inf groups, whereas patients with other mutational constellations, in particular patients with nontypical biallelic CEBPA mutations, showed an inferior outcome (supplemental Table 5; supplemental Figure 8). Co-mutations retained their prognostic impact in patients carrying CEBPA^bi-inf and CEBPA^sm-inf (as illustrated for the most common GATA2 mutations in supplemental Figure 9).

Because these results strongly support the notion that mutation constellations containing typical in-frame bZIP mutations (ie, CEBPA^bi-inf and CEBPA^sm-inf) are comparable with respect to most clinical and molecular factors studied, we combined these 2 mutation types (denoted CEBPA^bZIP-inf) and compared this new group to those patients without these mutations (CEBPA^other; allocation summarized in Figure 6).

As illustrated in Table 4, the clinical associations already observed for the individual groups were even more pronounced in this analysis, especially the highly divergent median age (46 vs 62 years; P < .001), the difference with respect to the association with sAML (4% vs 17%; P < .001), the median CD34 positivity (51% vs 34%; P < .001), and the median WBC counts (25 × 10⁹/L vs 14.7 × 10⁹/L; P < .001). As expected, the mutational spectrum of patients belonging to these 2 subgroups showed highly significant differences. Only 1 of 157 patients with CEBPA^bZIP-inf (0.6%) showed an NPM1 mutation, compared with 30 of 83 patients with CEBPA^other mutations (36.1%; P < .001; Figure 7C; supplemental Figure 9). Other mutations that were significantly more common in the CEBPA^other subgroup were alterations affecting DNA methylation (43 of 157, 27.4% vs 60 of 83, 72.3%; P < .001) and the spliceosome (4 of 157, 2.6% vs 21 of 83, 25.3%; P < .001).

Outcome analyses performed for CEBPA^bZIP-inf patients confirmed a strong prognostic impact of this subgroup, whereas CEBPA mutations in the other patients did not show significant differences compared with CEBPA^wt (Figure 7A-B). Interestingly, patients with CEBPA^bZIP-inf appear not to benefit from HSCT performed in the first complete remission (CR1; supplemental Figure 10). Multivariable analysis performed for this classification confirmed that CEBPA^bZIP-inf positivity was the strongest predictor for achievement of CR (OR: 6.06; 95% CI, 2.78-13.23, P < .001) and a strong prognostic factor for OS (hazards ratio [HR], 0.57; 95% CI, 0.46-0.71; P < .001) and EFS (HR, 0.53; 95% CI, 0.43-0.64, P < .001; Table 5). We also looked for the effect of mutant GATA2, TET2, and WT1 on outcome in the novel subgroups. As observed in the previous analyses, mutant GATA2 and the absence of TET2 and WT1 mutations were associated with improved OS and EFS in patients carrying CEBPA^bZIP-inf but not in those with CEBPA^other, although these differences were not significant (supplemental Figure 11).

We analyzed 4708 adult patients with newly diagnosed AML for CEBPA alterations and identified mutations in CEBPA in 5.1%. Our cohort differed from patient cohorts in previous studies investigating the role of CEBPA mutations,^{9-12,15,21,24,35-42} because we did not select for age, cytogenetic subgroups, or disease status. Within the entire cohort, 49% showed an aberrant karyotype, and the median patient age at diagnosis was 57 years (IQR, 46-67 years), more accurately reflecting the entirety of patients with AML in general. This fact may explain why the prevalence of CEBPA mutations found in our study is at the lower end of the previously reported range of 4% to 20%.^{9-12,15,21,24,35-42}

Our main focus was the role of monoallelic CEBPA mutations; therefore, we performed an extensive evaluation of this subgroup and a detailed analysis of the individual type and localization of the mutation.

A first aspect observed with respect to the different mutational subgroups (CEBPA^bi, CEBPA^smbZIP, and CEBPA^smTAD) was the highly significant difference in the age distribution. Whereas CEBPA^bi and CEBPA^smbZIP were predominantly found in younger adults and decreased with age, CEBPA^smTAD mutations were rare in patients up to the age of 40 years and were particularly common in older individuals. The lower age of patients with CEBPA^bi mutations has already been reported in previous studies.^21,24,41 In contrast, the major age difference between CEBPA^smTAD and CEBPA^smbZIP mutations has not been reported before. Interestingly, a very recently published work in pediatric AML found only monoallelic CEBPA^bZIP mutations in their analysis, indicating that these mutations are indeed significantly more prevalent in younger individuals.^43,CEBPA^bi and CEBPA^smbZIP also showed overlapping laboratory profiles, with higher rates of CD34 positivity and higher WBC counts.⁴³ We also observed a favorable prognostic impact of CEBPA^smbZIP comparable to biallelic CEBPA mutations. This finding again is in concordance with the data reported by Tarlock et al in pediatric AML, showing that the outcome of patients with CEBPA^bi or CEBPA^smbZIP mutations did not differ.⁴³ 

Our results indicate a similar spectrum of co-mutations for CEBPA^bi and CEBPA^smbZIP (ie, significantly higher rates of GATA2 and WT1 mutations and a mere lack of NPM1 alterations). GATA2 mutations in CEBPA^bi have been described previously,⁴⁴ whereas a similar association of CEBPA^smbZIP with GATA2 mutations has not been reported so far. The presence of GATA2, TET2, and WT1 mutations significantly affected the outcome of patients carrying CEBPA^bi, whereas CEBPA^smbZIP mutations showed a significant effect on outcome only for mutations in GATA2. The prognostic impact of mutations in GATA2, TET2, and WT1 has been described recently,^45,46 but was confined to biallelic CEBPA mutations. In contrast, our data indicate that these prognostic associations also apply to patients with CEBPA^smbZIP, but not to those with CEBPA^smTAD. Clustering of co-mutations according to functional groups further highlighted a differential spectrum of co-occurring mutations, with CEBPA^smTAD mutations showing a significantly higher prevalence of mutations affecting proteins involved in DNA methylation (ie, DNMT3A, TET2, IDH1, and IDH2) as well as RNA splicing (ie, SRSF2, SF3B1, and ZRSR2), which was not seen in patients with CEBPA^bi and CEBPA^smbZIP (Figure 4B). Additional analyses looking in more detail for the individual mutations in CEBPA in our patients suggested that the identified clinical and molecular associations as well as the association with outcome were restricted to typical in-frame mutations within the bZIP region.

The reason for this differential behavior is unclear at present, but based on previous in vitro as well as animal experiments, TAD1 and bZIP mutations have clear functional differences (reviewed in Pulikkan et al⁴⁷). The frameshift mutations typically observed in the TAD1 domain induce the consecutive translation of the shorter CEBPA^p30 protein, instead of the CEBPA^p42 full-length protein. In contrast, with both CEBPA^p42 and CEBPA^p30, mutations in the DBD and bZIP domains result in loss of DNA binding as well as dimerization. Recent animal data modeling the disease suggest functional disparity of different bZIP mutations. Lethally irradiated mice undergoing transplant of hematopoietic stem cells homozygous for a point mutation in bZIP (designated BRM2) start to develop a myeloproliferative disease that transforms into overt AML.⁴⁸ By transplanting transgenic cells carrying the most common mutation in bZIP (K313dup; K allele), alone or in combination with a TAD1 mutation (designated L-allele), Bereshchenko et al⁴⁹ documented that cells from mice carrying either the K/K or the K/L genotype, showed similarities in their mRNA expression profiles and higher expansion of immature blasts in the BM, which was distinct from the L/L genotype, as well as CEBPA-WT cells. CEBPA-mutated AML is characterized by specific RNA,⁵⁰ as well as miRNA^35,51 expression profiles. The expression of several key miRNAs, such as miR-34, miR-182, and miR-223, which are involved in stem cell self-renewal, cell migration, and granulocytic differentiation, are physiologically regulated by CEBPA (reviewed in Stavast et al⁵²). Interestingly, recent data suggest that bZIP-mutant CEBPA does not downregulate miR-182, leading to a block of granulocytic differentiation.⁵³ This incapability appears to be mainly restricted to typical in-frame bZIP mutations clustering around the core mutated amino acids 312 and 313 of the CEBPA protein.⁵³ In support of this, one of the genes we found most highly deregulated between CEBPA^bi-inf/CEBPA^sm-inf and CEBPA^sm-other, OSTL/RNF217, coding for a highly conserved RING-finger ubiquitin ligase overexpressed in various leukemia entities,⁵⁴ has been shown to be regulated by the miRNA cluster miR-183-96-182.⁵⁵ 

Taken together, the CEBPA^bZIP-inf genotype identified in our analysis describes a subgroup of CEBPA mutations predominantly found in younger adults, indicating characteristics of a more immature and proliferative disease that has an overall prevalence of 3.3% in adult patients with AML, but is found in up to 7% of patients ≤40 years, thus representing a relevant subgroup, especially in younger adults. The mere absence of NPM1 mutations (1 of 157 patients; 0.6%) and the high prevalence of GATA2 mutations in these patients, which was found to be associated with an even better prognosis and a long-term survival in up to 80%, further highlights the special biology of these leukemias. The 2016 World Health Organization classification defines the CEBPA mutational class by the presence of a biallelic mutation, regardless of the localization within the gene. Given that 90% of the patients with the biallelic CEBPA mutation in this analysis actually carried typical bZIP CEBPA mutations, there is obviously a high degree of overlap, which may explain why this difference was undetected in most previous analyses, although there was some evidence in another study.²⁵ However, that only the 90% of patients with biallelic CEBPA mutations containing typical bZIP mutations in fact show a better outcome indicates that only those should be assigned to the favorable risk group.

The authors thank all centers and participating physicians of the Study Alliance Leukemia who entered their patients into the study (see  Appendix). They also thank M. Böhm and M. Hartwig for skillful technical assistance.

This study was supported in part by grants from the Federal Ministry of Education and Research 01GS0872 (C.T.).

Contribution: F.T., J.A.G., and C.T. designed the study; performed the research; collected, assembled, analyzed, interpreted the data; and wrote the manuscript; F.T., J.A.G., S. Stasik., A.P., R.M.-L., and P.J.M.V., performed the molecular analyses and analyzed the data; S.H. performed the molecular analyses; M. Kramer and J.S. performed the statistical analyses; J.M.M., C.R., U.K., A.K., S. Scholl, A.H., T.H.B., R. Naumann, B.S., H.E., M.S., A.B., A.N., K.S.-E., C.S., S.W.K., M.H., R. Nopenney., U.K., C.D.B., M. Kaufmann, F.S., K.S., M.v.B., C.M.-T., U.P., W.E.B., H.S., G.E., M.B., and J.S., treated the patients and collected the clinical data; and all authors approved the final version of the manuscript.

Conflict-of-interest disclosure: C.T. is CEO and co-owner of AgenDix GmbH. The remaining authors declare no competing financial interests.

Correspondence: Christian Thiede, Medizinische Klinik und Poliklinik I, Universitätsklinikum Carl Gustav Carus der Technischen Universität, Fetscherstr 74, 01307 Dresden, Germany; e-mail: christian.thiede@uniklinikum-dresden.de.

Tim H. Brümmendorf, Universitätsklinikum Aachen der RWTH, Germany; Armin Schulz-Abelius, Klinikum Altenburger Land GmbH, Altenburg, Germany; Martin Trepel, Klinikum Augsburg, Augsburg, Germany; Martina Teichmann, Sozialstiftung Bamberg, Bamberg, Germany; Alexander Kiani, Klinikum Bayreuth GmbH, Bayreuth, Germany; Bertram Glass, Helios Klinikum Berlin-Buch, Berlin, Germany; Martin Görner, Städt Kliniken Bielefeld gGmbH, Bielefeld, Germany; Dirk Behringer, Augusta-Kranken-Anstalt gGmbH, Bochum, Germany; Johannes Kullmer, Ev Diakonie-Krankenhaus gGmbH, Bremen, Germany; Mathias Hänel, Klinikum Chemnitz gGmbH, Chemnitz, Germany; Christof Lamberti, MVZ Coburg, Coburg, Germany; Martin Schmidt-Hieber, Carl-Thiem-Klinikum Cottbus gGmbH, Cottbus, Germany; Christoph Röllig, Universitätsklinikum Dresden, Dresden, Germany; Michael Flasshove, Krankenhaus Düren gGmbH, Düren, Germany; Andreas Mackensen, Universitätsklinikum Erlangen-Nürnberg, Erlangen, Germany; Maher Hanoun, Universitätsklinikum Essen, Essen, Germany; Michael Kiehl, Klinikum Frankfurt (Oder) GmbH, Frankfurt (Oder), Germany; Hubert Serve, Klinikum der J. W. Goethe Universität, Frankfurt (Main), Germany; H.-G. Höffkes, Klinikum Fulda, Fulda, Germany; Lutz Peter Müller, Universitätsklinikum Halle (Saale), Halle (Saale) Germany; Heinz Albert Dürk, St Barbara-Klinik Hamm, Hamm, Germany; Carsten Müller-Tidow, Universitätsklinikum Heidelberg, Heidelberg, Germany; Ulrich Kaiser, St Bernward Krankenhaus, Hildesheim, Germany; Andreas Hochhaus, Universitätsklinikum Jena, Jena, Germany; Gerhard Held, Westpfalz-Klinikum GmbH, Kaiserslautern, Germany; Roland Peter Repp, Städt Krankenhaus Kiel, Kiel, Germany; Claudia Baldus, Universitätsklinikum Schleswig-Holstein, Kiel, Germany; Jens-Marcus Chemnitz, Gemeinschaftsklinikum Mittelrhein gGmbH, Koblenz, Germany; Uwe Platzbecker, Universitätsklinikum Leipzig, Leipzig, Germany; Stefan Klein, Klinikum Mannheim gGmbH, Mannheim, Germany; Andreas Neubauer, Universitätsklinikum Marburg GmbH, Marburg, Germany; Wolfgang E. Berdel, Universitätsklinikum Münster, Münster, Germany; Martin Wilhelm, Klinikum Nürnberg Nord, Nürnberg, Germany; Jörg Schubert, Elblandklinikum Riesa, Riesa, Germany; Achim Meinhardt, Agaplesion Diakonieklinikum Rotenburg gGmbH, Rotenburg (Wümme), Germany; Thomas Geer, Diakonie-Krankenhaus, Schwäbisch-Hall, Germany; Markus Ritter, Klinikum Sindelfingen-Böblingen, Sindelfingen, Germany; Walter Erich Aulitzky, Robert-Bosch-Krankenhaus, Stuttgart, Germany; Natalia Heinz, HSK Wiesbaden, Wiesbaden, Germany; Markus Schaich, Rems-Murr-Klinikum Winnenden, Winnenden, Germany; and Hermann Einsele, Universitätsklinikum Würzburg, Würzburg, Germany.