CEBPA (CCAAT/enhancer-binding protein alpha) is a member of the C/EBP family of bZIP transcription factors encoding two different translational protein isoforms. The CEBPA transcription factor is involved in cell cycle arrest, repression of self-renewal and myeloid differentiation during normal hematopoiesis. In acute myeloid leukemia (AML), mutations in CEBPA result in a cellular differentiation block.1 They occur in around 8% of normal karyotype AML (CN-AML). Approximately half of the patients harbor two CEPBA mutations associated to a favorable prognosis. CEBPA function can also be affected by promoter methylation2 or alterations in other oncogenes, for example through the t(8;21)(q22;q22)/RUNX1–RUNX1T1, which suppresses CEBPA mRNA expression. The t(8;21)(q22;q22) translocation replaces the C terminus, including the transactivation domain (TAD) of RUNX1, with RUNX1T1.3 RUNX1–RUNX1T1 in t(8;21) blocks CEBPA-dependent activation of its own (CEBPA) promoter and thereby inhibits autoregulation.4 Other leukemic fusion proteins involving core-binding factor (CBF) family members, for example t(3;21)(q26;q22)/RUNX1–EVI1 or inv(16)(p13q22)/CBFB–MYH11, did not suppress CEBPA mRNA, indicating a RUNX1–RUNX1T1-specific effect on CEBPA transcriptional control.1

Intragenic RUNX1 mutations confer an adverse prognosis in AML. Previously, we identified RUNX1 mutations in 32.7% of CN-AML or with non-complex chromosomal imbalances.5 RUNX1 mutations are absent in CBF–AML and acute promyelocytic leukemia.6 They are inversely correlated with CEBPA mutations.5, 6

To clarify whether intragenic RUNX1 mutations such as RUNX1–RUNXT1 fusions also result in CEBPA mRNA downregulation, we investigated 359 AML patients consisting of two independent cohorts: cohort 1 with 209 AML cases (109 males/100 females; median age 65.4 years; 19.7–88.1 years) from different cytogenetic subgroups (normal karyotype (n=93), t(8;21)(q22;q22)/RUNX1–RUNX1T1 (n=16), t(15;17)(q22;q12)/PML-RARA (n=15), sole +8 (n=12), sole +13 (n=10), complex karyotypes (n=10) and other rare noncomplex genetic abnormalities (n=53)). Cohort 2 comprised 150 normal karyotype patients selected according to RUNX1 mutation status (92 males/58 females; median age, 69.7 years; 18.3–88.1 years; 81/150 (54.0%) RUNX1 mutated) with survival data in 124 cases. Bone marrow and/or peripheral blood samples were sent to the MLL Munich Leukemia Laboratory in 2005–2011. All patients gave their written informed consent to genetic analysis and scientific studies.

Chromosome-banding analysis was performed in all cases, when needed, combined with fluorescent in situ hybridization. RUNX1 mutations were analyzed by Sanger sequencing or an amplicon-based high-throughput deep-sequencing assay (454 Life Sciences, Branford, CT, USA). CEBPA (mRNA) expression was quantified in cohort 1 by gene expression microarray profiling (Affymetrix HG-U133 Plus 2.0 microarrays; Santa Clara, CA, USA). The gene expression raw data were processed according to the manufacturer’s recommendations. Detection calls, that is present, marginal, or absent expression, were determined by default parameters. For measurement of CEBPA expression in cohort 2, a quantitative real-time reverse transcriptase PCR (RT-PCR) assay was established (Taqman, Life Technologies, Carlsbad, CA, USA; CEBPA TaqMan Gene Expression Assay: HS00269972_S1). mRNA expression of CEBPA was normalized against expression of ABL1; ratios were given as %CEBPA/ABL1.

First, we investigated 209 AML cases from different cytogenetic subgroups using gene expression microarray profiling (Table 1a). The RUNX1 mutation status was analyzed in 178 cases (RUNX1–RUNX1T1 or PML-RARA-mutated cases had been excluded), in 41/178 (23%) of patients RUNX1 was mutated. The median CEBPA expression intensity value in all patients was 670 (range 48–5244). RUNX1-mutated cases showed a lower CEBPA expression than RUNX1 wild-type cases (n=41 vs 137, mean±s.d. 429±395 vs 998±717; P<0.001). Cases harboring a t(8;21)/RUNX1–RUNX1T1 presented a lower CEBPA expression than patients without (n=16 vs 193, mean±s.d. 292±216 vs 950±808; P<0.001), whereas t(15;17)/PML-RARA-mutated cases showed enhanced CEBPA expression (n=15 vs 194, mean±s.d. 1940±1290 vs 819±690; P=0.005) (Figure 1a). As reported previously, cases with a sole +13 showed lower expression than cases without (n=10 vs 199, mean±s.d. 326±406 vs 929±803; P=0.020); however, all +13 cases were RUNX1-mutated.

Table 1a CEBPA expression levels were determined by microarray gene expression analysis in cohort 1 including 209 AML patients from different cytogenetic subgroups
Figure 1
figure 1

(a) CEBPA expression as determined by microarray gene expression analysis in different genetic subgroups of cohort 1. (b) Distribution of the RUNX1 mutations in cohort 2 with regards to the different domains. (c) CEBPA expression as determined by real-time PCR in cohort 2.

For validation, an independent cohort of 150 normal karyotype AML was investigated for RUNX1 mutations (Table 1b), CEBPA expression was quantified using real-time RT-PCR. RUNX1 mutations were detected in 81/150 (54.0%) (Figure 1b). Although this cohort was selected according to karyotype and RUNX1 mutation status, we compared the overall survival (OS) and event-free survival (EFS) from RUNX1-mutated and RUNX1 wild-type cases. Median OS was 19.9 and 12.2 months, OS at 3 years was 35.3 and 17.4%, P=0.049; median EFS was 18.8 and 6.9 months, OS at 3 years was 34.2 and 4.7%, P=0.007. Median CEBPA expression intensity was 148 (range: 21–960). Correspondingly to the data obtained from the first cohort, CEBPA expression was lower in RUNX1-mutated cases as compared with RUNX1 wild-type patients (mean±s.d. 155±98 vs 222±183; P=0.007) (Figure 1c). When separating the cohort in cases with ⩽148 and >148 CEBPA expression (using the median CEBPA expression as threshold in the cohort), we observed no association with OS and EFS (median OS was 16.8 and 18.8 months, OS at 3 years was 25.6 and 24.3%, P=0.507; median EFS was 11.0 and 9.2 months, OS at 3 years was 19.1 and 15.0%; P=0.541).

Table 1b Investigation of CEBPA expression as determined by quantitative real-time PCR in cohort 2 of 150 patients with normal karyotype AML

In addition, we investigated whether the localization of the RUNX1 mutations had any impact on CEBPA expression. However, no significant difference was detected between CEBPA expression levels, when RUNX1 mutations were located either outside or within the DNA-binding domain (RUNT) (n=46 cases), or behind the RUNT and within the TAD (n=35) (mean±s.d. 161±107 vs 148±87; P=NS), respectively. Also when comparing missense mutations (n=19) against in-frame, frameshift and nonsense alterations (n=62), no significant difference was detectable (mean±s.d. 177±114 vs 148±93; P=NS). In contrast, separating cases in single-mutated (n=60) or double-mutated/homozygous (n=21) RUNX1 mutations, we observed a non-significant trend towards a lower CEBPA expression in cases with double-mutated/homozygous mutations (mean±s.d. 126±89 vs 165±100; P=0.116) (Figure 1c).

In murine experiments, RUNX1 gene deletion was reducing CEBPA mRNA in lineage-negative marrow cells in granulocyte–monocyte progenitors or common myeloid progenitors.7 Here, we demonstrated a negative effect of RUNX1 mutations on CEBPA expression levels in AML patients, similar to the RUNX1–RUNX1T1 fusion, which have previously been reported.4 By gene expression profiling, downregulation of different hematological transcription regulators such as CEBPA or ETV6 had been described by Silva et al.8 in AML FAB M0, which is closely associated with RUNX1 mutations. RUNX1 mutation localization seems to have no impact on CEBPA expression. In summary, downregulation of CEBPA expression may contribute to leukemogenesis in RUNX1-mutated AML.