Distinct gene expression profiles of acute myeloid/T-lymphoid leukemia with silenced CEBPA and mutations in NOTCH1

Abstract

Gene expression profiling of acute myeloid leukemia (AML) allows the discovery of previously unrecognized molecular entities. Here, we identified a specific subgroup of AML, defined by an expression profile resembling that of AMLs with mutations in the myeloid transcription factor CCAAT/enhancer-binding protein alpha (C/EBPalpha), while lacking such mutations. We found that in these leukemias, the CEBPA gene was silenced, which was associated with frequent promoter hypermethylation. The leukemias phenotypically showed aberrant expression of T-cell genes, of which CD7 was most consistent. We identified 2 mechanisms that may contribute to this phenotype. First, absence of Cebpa led to up-regulation of specific T-cell transcripts (ie, Cd7 and Lck) in hematopoietic stem cells isolated from conditional Cebpa knockout mice. Second, the enhanced expression of TRIB2, which we identify here as a direct target of the T-cell commitment factor NOTCH1, suggested aberrantly activated Notch signaling. Putatively activating NOTCH1 mutations were found in several specimens of the newly identified subgroup, while a large set of control AMLs was mutation negative. A gene expression prediction signature allowed the detection of similar cases of leukemia in independent series of AML.

Figure 1

CEBPA silencing and promoter hypermethylation are associated with AMLs sharing a CEBPA mutant gene expression signature in GEP cluster no. 4. (A) Pairwise correlations between gene expression profiles of 285 AML samples calculated on the basis of 2856 probe sets are displayed as described. Colors of boxes visualize Pearson correlation coefficient: deeper red indicates higher positive correlation; deeper blue indicates higher negative correlation. Sixteen distinct clusters were previously distinguished, which can be recognized by the red blocks showing high correlation along the diagonal. Cluster no. 4 and cluster no. 15, associated with CEBPA mutations, are enlarged. The bar and histogram next to each patient represent CEBPA mutation status and CEBPA expression level, respectively. CEBPA mutation status: presence (“mut,” red) or absence (“wt,” green) of mutations in bZIP region and/or N-terminus. These data indicate that in leukemias in cluster no. 4 lacking mutations, CEBPA expression is low or absent. In cluster no. 4, the order of samples, from top to bottom, is as follows: no. 3327, no. 2242 (both CEBPA mutant); no. 2668, no. 2238, no. 3314, no. 2686, no. 3483, no. 3491 (all 6 without CEBPA mutation); no. 2218, no. 1316, no. 2273, no. 2545, no. 2169, no. 2753, and no. 2192 (all 7 CEBPA mutant). (B, upper part) Schematic representation of the chromosomal region surrounding the transcriptional start of the CEBPA gene. Numbers indicate position relative to CEBPA transcriptional start. Vertical lines represent CpG dinucleotides; “transcription” indicates transcriptional start; “ATG,” translational start site. (Lower part) Level of cytosine methylation in the region surrounding the CEBPA transcriptional start site of the 6 AML cases in cluster no. 4 with low CEBPA expression, with patient numbers on the left. Every cytosine in a CpG dinucleotide is depicted as a circle. For each of these cytosines, the fraction of methylated residues was determined, which is visualized by the color of the circle: methylated (> 75% of all cytosines methylated, black), partly methylated (25%-75% methylated, gray), unmethylated (< 25% methylated, white).

Figure 2

AML cases with silenced CEBPA in cluster no. 4 simultaneously express myeloid- and T-lymphoid–specific genes and lineage markers. (A) Pairwise correlations between samples are displayed as explained in the legend to Figure 1, and GEP cluster no. 4 is enlarged in the box on the right. Histograms next to each patient display expression of selected genes with significantly elevated expression in cluster no. 4 cases with silenced CEBPA. Expression levels for probe sets of the following genes are visualized: CD7, CD3D, CD3Z, CD3G, LCK, TRD@, and NOTCH1. Corresponding expression levels and Affymetrix probe set identifiers are depicted in Table S4. (B) Representative dot plot images from flow cytometric analysis of samples obtained from 2 individual patients (ie, patients no. 2238 and no. 3491) demonstrating that the majority of cells from these patients simultaneously express CD34, CD13, CD7, and terminal deoxynucleotidyltransferase (TdT). The tumor population was identified by its weak expression of CD45, depicted in black, whereas CD45^high cells, which are considered to be mature lymphocytes, are colored in gray.

Figure 3

Absence of Cebpa in murine HSCs up-regulates Cd7 and Lck. (A) Lin⁻c-Kit⁺Sca1⁺ hematopoietic stem cells (HSCs) were isolated and purified from poly I:C–treated Cebpa^wt/wt Mx1-Cre mice (wt cre⁺) and Cebpa^F/F Mx1-Cre knockout mice (F/F cre⁺). mRNA expression levels of Cd7 and Lck were determined by RQ-PCR. Data are presented relative to expression of wt cre⁺ animals as mean plus standard deviation of at least 3 individual mice in both groups. (B) Purified HSCs from poly I:C–treated Cebpa^F/F Mx1-Cre mice were transduced with either empty vector (vector) or Cebpa expression construct (Cebpa). Cd7 and Lck mRNA expression levels were determined in GFP⁺ infected cells 1 day after transduction. Data are presented relative to expression of empty vector–transduced cells as mean plus standard deviation of 4 reactions.

Figure 4

Trib2 is a direct target gene of Notch1. (A) Reverse transcriptase (RT)–PCR analysis of Trib2 expression in Scid.adh cells treated with GSI (+) or DMSO as vehicle control (−). Cells were transduced with an empty vector expressing only the NGFR selection marker (NGFR) or a vector expressing ICN1 (ICN1). Trib2 mRNA expression was assessed. Hprt is an internal loading control. Results are representative of triplicate experiments. (B) Schematic representation of the Trib2 promoter region. Four putative CSL-binding sites were identified in the region shown, including a tandem CSL site at − 6 kb and 3 single binding sites at − 22 kb, − 6.1 kb, and − 3 kb relative to the translational start site. (C) Notch1 binds to CSL-binding sites in the Trib2 promoter. Chromatin immunoprecipitates were performed on cross-linked fragmented DNAs prepared from Scid.adh cells. Immunoprecipitations were carried out with antibodies against Myc as a negative control (control), acetylated histone 4 (AcHIS4), and Notch1. PCR was performed using primers directed against indicated CSL-binding site regions at − 22, − 6, and − 3 kb from the ATG translational start site of Trib2. PCR for the Hes1 promoter region is shown as a positive control. The figure shown is representative of duplicate experiments. For the − 6-kb region, which shows the strongest enrichment, results are representative of duplicate samples and triplicate experiments. (D) After ChIP as described in panel C, using either Notch1 or normal rabbit control IgG antibody, RQ-PCR was performed using primer sets flanking putative CSL-binding sites in the Trib2 promoter. RQ-PCR was performed using the primers at − 6 kb to quantify enrichment in this region, and also using primers flanking a conserved CSL-binding site at − 6.1 kb. Graphs represent mean of the ratios of the amount of IP DNA/input from values of duplicate wells plus or minus standard deviation. Data are representative of 3 independent experiments.

Figure 5

A gene expression prediction signature identifies new leukemias with the silenced CEBPA phenotype in an independent cohort of AML. (A) For 262 samples analyzed on Affymetrix HGU133Plus2.0 GeneChips, log-transformed (base 2) and mean-centered expression levels for 13 probe sets are depicted (left panel) for an arbitrary range from − 4 to + 4 (corresponding to 16-fold lower to 16-fold higher expression relative to the mean, respectively). The ordering of patients in the figure is arbitrary. In the right panel, these data are enlarged for 31 of these 262 leukemias, representing 3 groups: (I) the 6 cases with silenced CEBPA previously identified, with from top to bottom cases no. 2668, no. 2238, no. 3314, no. 2686, no. 3483, and no. 3491; (II) a variable selection of 10 AMLs from distinct GEP clusters for which also NOTCH1 mutational analysis and CEBPA promoter bisulfite sequencing were performed; and (III) 15 AMLs with CEBPA mutations, originating from either GEP cluster no. 4 (upper 9 samples) or cluster no. 15 (lower 6 samples). The 9 probe sets on the left side constitute the most predictive gene expression signature for group I, as determined by PAM: 216191_s_at (TRA@/TRD@ 1), 217143_s_at (TRA@/TRD@ 2), 213830_at (TRD@), 225681_at (CTHRC1), 1565809_x_at (no annotation), 1560018_at (ARPP-21), 210844_x_at (CTNNA1 1), 200764_s_at (CTNNA1 2), and 200765_x_at (CTNNA1 3). To the right, 4 additional probe sets are indicated, that is, 204039_at (CEBPA), 218902_at (NOTCH1), 202478_at (TRIB2), and 214551_s_at (CD7). Mutational data for NOTCH1 and CEBPA, and methylation status of the CEBPA promoter are depicted next to the normalized hybridization intensities of the probe sets. (B) Two hundred sixty-eight samples obtained from a second cohort of AML were hybridized to HGU133Plus2.0 GeneChips. The 9 probe set signature was used to identify leukemias with a profile similar to group I (A), resulting in the detection of group IV (from top to bottom cases no. 6376, no. 6735, no. 6947, no. 7053, no. 7076, and no. 7120).