Approximately 50% of acute myeloid leukemia (AML) have no karyotype changes or those with yet unknown prognostic significance. They are usually pooled together into the prognostically intermediate group. Here we approached the role of CEBPA mutations within this AML subgroup. In total, 255 AML, 237 with normal and 18 with “other” intermediate risk group karyotypes were screened for CEBPA mutations by sequencing. The total incidence of CEBPA mutations was 51/255 (20%) (48/237 (20.3%) in the normal and 3/18 (16.7%) in the “other” karyotypes). Most of the patients showed an M1 (n=16), or M2 (n=25) morphology, but there were also some with FAB M0 (n=1), M4 (n=4), M5 (n=3), and M6 (n=2). CEBPA+ cases were younger as compared to the CEBPA- cases (54.7 vs. 60.0, p=0.023). Leukocyte und platelet counts were similar. Clinical follow up data were available for 191 (37 mutated, 154 unmutated) patients. OS and EFS were significantly better in the patients with compared to those without CEBPA mutations (median 1092 vs. 259 days, p=0.0072; 375 vs. 218 days, p=0.0102, respectively). In addition, 18/42 (42.9%) of CEBPA+ cases had an FLT3-LM, 4/40 (10%) an FLT3-TKD, 4/41 (9.8%) an MLL-PTD, 3/34 (8.8%) an NRAS, 2/40 (5%) a KITD816 mutation. In four cases 2 additional mutations were detected: 1 x FLT3-LM+KITD816, 1 x FLT3-LM+FLT3-TKD, and 2 x MLL-PTD+FLT3-LM. The favorable prognostic impact of CEBPA mutations was not affected by additional mutations. Furthermore, 22 of the CEBPA+ case were analyzed by microarray analysis using the U133A+B array set (Affymetrix) and compared to the expression profile of 131 CEBPA- normal karyotype AML, as well as to 204 AML characterized by the reciprocal translocations t(15;17) (n=43), t(8;21) (n=36), inv(16) (n=48), t(11q23) (n=50), inv(3) (n=27). The discrimination of CEBPA+ cases and reciprocal translocations revealed a classification accuracy of 94.7% with 75% sensitivity and 98.5% specificity. However, the CEPBA+ cases did not show a specific expression pattern within the total group with normal karyotype and could not be discriminated from CEBPA- cases. By use of PCA and hierachical cluster analysis it was obvious that the CEBPA+ cases separated into two domains. One subcluster (cluster 1) was distributed among the cases with CEBPA- normal karyotype AML. A second cluster (cluster 2) was very close to the t(8;21) cases. Accordingly, cases of cluster 2 similar to t(8;21) and in contrast to cluster 1 highly expressed MPO and had low expression of HOXA3, HOXA7, HOXA9, HOXB4, HOXB6, and PBX3. Using the top 100 differentially expressed genes and applying 100 runs of SVM with 2/3 of samples being randomly selected as training set and 1/3 as test set samples, groups A and B could be classified with an overall accuracy of 100% (sensitivity 100% and specificity 100%). A detailed analysis of the two subclusters showed that all 8 cases of cluster 1 revealed mutations in the TAD2 domain of CEBPA and 6 of these had an FLT3-LM in addition. In contrast, 12/14 cases of cluster 2 had mutations that lead to an N-terminal stop and only 2 had an FLT3-LM. Thus these two subclusters have biological differences that may explain the different gene expression patterns. Despite the different functional consequences of the mutations in the two CEBPA-clusters no differences with respect to FAB type and prognosis were found between cluster 1 and 2.

Author notes

Corresponding author