|Home | About | Journals | Submit | Contact Us | Français|
C/EBPs are a family of transcription factors that regulate growth control and differentiation of various tissues. We found that C/EBPγ is highly upregulated in a subset of acute myeloid leukemia (AML) samples characterized by C/EBPα hypermethylation/silencing. Similarly, C/EBPγ was upregulated in murine hematopoietic stem/progenitor cells lacking C/EBPα, as C/EBPα mediates C/EBPγ suppression. Studies in myeloid cells demonstrated that CEBPG overexpression blocked neutrophilic differentiation. Further, downregulation of Cebpg in murine Cebpa-deficient stem/progenitor cells or in human CEBPA-silenced AML samples restored granulocytic differentiation. In addition, treatment of these leukemias with demethylating agents restored the C/EBPα-C/EBPγ balance and upregulated the expression of myeloid differentiation markers. Our results indicate that C/EBPγ mediates the myeloid differentiation arrest induced by C/EBPα deficiency and that targeting the C/EBPα-C/EBPγ axis rescues neutrophilic differentiation in this unique subset of AMLs.
The temporal and spatial control of cell-specific transcription factors and their abundance determines to a large extent the cell fate. The hematopoietic system, due to its well characterized cell hierarchy and well defined cell types, has been extensively used to study the function of transcription factors in cell renewal and differentiation (1, 2). Balancing these 2 processes is crucial for homeostasis, and defects in lineage-specific transcription factors, often caused by translocations or mutations, have been described in human leukemia (3–5).
CCAAT/enhancer-binding proteins (C/EBPs) are a family of basic region leucine zipper DNA–binding proteins of which 6 core members have been identified: C/EBPα, C/EBPβ, C/EBPγ, C/EBPδ, C/EBPε, and C/EBPζ (5–8). In the hematopoietic system, C/EBPα plays a crucial role in the commitment of multipotent progenitor cells into the myeloid lineage, and mice deficient in C/EBPα are characterized by a specific block in the transition from common myeloid progenitors (CMP) to granulocyte-macrophage progenitors (GMP), resulting in the lack of mature and functional granulocytes (9, 10). Defects in CEBPA, such as differentiation-deficient mutations, posttranscriptional modifications, posttranslational inhibition, and epigenetic regulation, have been shown to inhibit C/EBPα function in acute myeloid leukemia (AML) (11–17). C/EBPβ plays a role in macrophage activation (18) and is required during emergency granulopoiesis following administration of cytokines, a process that involves rapid production of granulocytes (19). C/EBPε, C/EBPδ, and C/EBPζ have also been implicated at different stages of granulocytic development (20, 21), and aberrant expression of these C/EBP members has been observed in AML and secondary granule deficiency syndromes (22–25). Thus, C/EBPγ is the only C/EBP member that has not been studied in the context of myeloid differentiation or human AML.
C/EBPγ is ubiquitously expressed and was first identified by its affinity for cis-regulatory sites in the immunoglobulin heavy chain promoter and enhancer (26). Structurally, C/EBPγ is characterized by the absence of transactivation domains, although it retains the basic region and the leucine zipper domains, required for DNA binding and homo/heterodimerization, respectively (27, 28). C/EBPγ alone neither activates nor represses transactivation; however, C/EBPγ can inhibit transcriptional activation by other C/EBP members in a cell-specific manner (27, 29). C/EBPγ-KO mice show a high mortality rate within 48 hours after birth, and analysis of the lymphoid lineage demonstrated that C/EBPγ plays a critical role in the functional maturation of natural killer cells (30). In addition, overexpression of C/EBPγ in transgenic mice causes a block in definitive erythropoiesis, leading to severe anemia and fetal lethality (31).
In the present study, we identify a unique subset of AML patients with increased CEBPG expression levels and methylation/silencing of the CEBPA gene. We demonstrate that C/EBPγ is a suppressor of myeloid differentiation in these AML samples and that targeting the C/EBPα-C/EBPγ axis represents a novel therapeutic approach in this particular subtype of AML.
In order to determine whether CEBPG is aberrantly expressed in certain cases of AML, we analyzed the gene expression profile (GEP) of 526 AML cases. This analysis revealed that CEBPG mRNA was significantly upregulated in a small subset of patients (n = 8) (Figure (Figure1A).1A). The 526 AML cases include a previously reported cohort of n = 285 samples (32), and as found in the original study, unsupervised clustering of the 526 cases defined a group of patients characterized by defects in CEBPA (Figure (Figure1A).1A). The defects in CEBPA were caused by either CEBPA biallelic mutations (n = 21) or CEBPA silencing, primarily due to hypermethylation (n = 9) (Figure (Figure1A1A and refs. 17, 33). These later cases showed very low or no CEBPA mRNA levels compared with all other AML cases (Supplemental Figure 1A; supplemental material available online with this article; doi: 10.1172/JCI65102DS1), and C/EBPα protein could not be detected (17). In addition, it was reported that these leukemias highly express Trib2, a protein that degrades C/EBPα p42 protein, contributing to the absence of C/EBPα in the silenced cases (34). Interestingly, CEBPG expression levels were particularly high in 8 out of 9 cases characterized by CEBPA silencing in comparison with the other AML samples, normal bone marrow, and CD34+ specimens (Figure (Figure1A).1A). We verified these changes in CEBPG RNA levels by quantitative RT-PCR (Supplemental Figure 1B). In summary, we identified a subset of AML patients characterized by silencing of CEBPA and upregulation of CEBPG.
Since we identified a subgroup of AML patients characterized by silencing of C/EBPα and upregulation of CEBPG, we asked whether ablation of Cebpa in vivo would lead to an increase of Cebpg expression, in particular in the early hematopoietic stem/progenitor cells (lineage–c-kit+ Sca-1+ [LKS]). For these experiments we used a well defined Cebpa conditional KO model previously developed in our laboratory (10). CebpaFlox/Flox Mx1Cre+ conditional KO mice and CebpaFlox/Flox Mx1Cre– control mice (WT) were treated with polyinosinic-polycytidylic acid (pI:pC) to induce excision of the single exon Cebpa gene (10). Quantitative RT-PCR analysis revealed that KO LKS cells lost detectable expression of Cebpa (P = 1.1 × 107) and that the Cebpg transcripts had increased significantly compared with control LKS (P = 1.6 × 10–6) (Figure (Figure1B).1B). Further, we determined that LKS cells from Cebpa heterozygous mice (CebpaFlox/+ Mx1Cre+) did not present increased Cebpg levels compared with the CebpaFlox/+ Mx1Cre– control mice upon pI:pC treatment (Supplemental Figure 1C). Reintroduction of C/EBPα into C/EBPα-KO LKS cells using retroviral infection significantly reduced Cebpg expression (P = 0.0065) (Figure (Figure1C).1C). Similarly, reintroduction of the C/EBPα p30 isoform also diminished the Cebpg mRNA levels (P = 0.034) (Supplemental Figure 1D). In summary, our data suggest that C/EBPα negatively regulates Cebpg expression not only during later stages of granulopoiesis, but also in murine hematopoietic stem/progenitor cells. Further, the CebpaFlox/Flox Mx1Cre+ conditional KO mice mimic the human AMLs characterized by silencing of C/EBPα and upregulation of Cebpg, providing an excellent model to study the contribution of C/EBPγ to the development of those leukemias.
Since we observed that CEBPG and C/EBPA expression levels are inversely correlated in a subset of AML patient samples and in the Cebpa conditional KO mice, we questioned whether CEBPG repression is directed by C/EBPα. First, we studied C/EBPα interaction with Cebpg regulatory regions in murine 32D/G-CSF–R cells transduced with a β-estradiol–inducible (E2) C/EBPα-ER expression construct. When stimulated with E2 to induce nuclear translocation of C/EBPα, these cells differentiate toward neutrophils within 3–4 days of culture, even in the presence of IL-3 (data not shown). After 4 hours of treatment, we performed ChIP followed by promoter array analysis and identified 4826 chromosomal regions, sized between 310 and 7399 nucleotides (median 1084), enriched in C/EBPα-ER cells compared with ER control cells at P < 1 × 10–5 (false discovery rate 2.3%). The transcriptional start site of 1064 unique RefSeq genes was located within 1 kb downstream of 1 or more of the 4826 fragments. Among those, we identified previously reported genes to be directly or indirectly regulated by C/EBPs, such as Mpo (35), Hp (36), C3 (36), PU.1 (37), and IL-6rα (38) (data not shown). A strong interaction with the Cebpg promoter was observed (Figure (Figure2A),2A), whereas almost none of the other loci in that region appeared to be enriched by ChIP. Next, ChIP was performed in 32D/G-CSF–R cells to determine whether endogenous C/EBPα could immunoprecipitate Cebpg promoter. Similarly to the C/EBPα-ER overexpressing system, we observed immunoprecipitation of Cebpg promoter by endogenous C/EBPα (Supplemental Figure 2A), but not of a control upstream region (data not shown). To further investigate the binding of C/EBPα to the CEBPG promoter, we used K562 cells stably transfected with C/EBPα-ER as a model for human granulocytic differentiation (39). We identified a putative C/EBPα-binding site in the human CEBPG proximal promoter, which was highly conserved between mouse and human (–871 bp to –762 bp, oligos A, corresponding to the region amplified by oligos no. 2 in the murine promoter, Supplemental Figure 2B). ChIP assays showed binding of C/EBPα to this region (oligos A), but not to a control upstream region (oligos B) (Supplemental Figure 2, C and D). Together, these results demonstrate that C/EBPα can bind to the proximal promoter of CEBPG and potentially downregulate CEBPG expression.
In order to determine whether C/EBPα could negatively regulate CEBPG promoter, we performed luciferase reporter assays. We generated a human CEBPG promoter construct (–354 bp to +1 bp from TSS, Supplemental Figure 2B) in the pXP2 reporter vector and performed luciferase assays to analyze CEBPG promoter activity. Cotransfection of the CEBPG reporter construct with increasing amounts of a C/EBPα expression plasmid showed that C/EBPα was able to repress CEBPG promoter reporter gene activity (Figure (Figure2B).2B). Of note, both a p42-C/EBPα plasmid and a p30-C/EBPα plasmid inhibited the CEBPG reporter gene activity in a dose-dependent manner and with similar efficiency (Figure (Figure2B).2B). As a positive control for C/EBPα transactivation, we cotransfected the C/EBPα expression vector with a reporter for the gene encoding the G-CSF receptor (CSF3R), denoted here as the G-CSF–R reporter. Luciferase assays showed that C/EBPα transactivated the G-CSF–R reporter gene, indicating that negative regulation by C/EBPα is specific to the CEBPG promoter (data not shown). Based on the AML microarray (Figure (Figure1A)1A) C/EBPγ is only upregulated in the C/EBPα silenced leukemias, suggesting that mutated C/EBPα proteins are still capable of repressing C/EBPγ expression. Luciferase assays with the C/EBPγ reporter construct and increasing amounts of distinct mutated C/EBPα expression plasmids showed that the N-terminal mutants repress C/EBPγ luciferase activity (Supplemental Figure 2E). These results demonstrate that WT p42 and p30 C/EBPα forms as well as N-terminal mutant C/EBPα proteins are able to negatively regulate the C/EBPγ promoter.
To identify the cis-acting elements on the CEBPG promoter that respond to C/EBPα, we first compared 2 human CEBPG reporter constructs: one corresponding to –957 bp to +1 bp of the CEBPG proximal promoter and the other corresponding to a shorter version (from –354 bp to +1 bp), which is missing the first 2 potential C/EBP-binding sites (Supplemental Figure 2B). Luciferase assays with increasing amounts of C/EBPα expression vector showed similar C/EBPα-mediated repression in both C/EBPγ reporter constructs, indicating that the first 2 potential C/EBP-binding sites are not necessary for the repression (data not shown). Next, we investigated whether the other 2 potential C/EBP-binding sites present in the shorter reporter construct were necessary to repress the luciferase activity. We therefore generated a CEBPG reporter construct with mutated C/EBP-binding sites in the context of the shorter reporter. These 2 C/EBP sites appeared not to be required, since C/EBPα was still able to repress the luciferase activity (Figure (Figure2C).2C). Thus C/EBPα represses either through other binding sites and/or in an indirect manner, through interaction with other transcriptional regulators. In fact, we and others have previously reported that C/EBPα can act as a transcriptional repressor through E2F-binding sites (40, 41). We determined the presence of 3 E2F-binding sites in the C/EBPγ luciferase constructs (Supplemental Figure 2B). To determine whether the E2F sites were required for C/EBPα repression, we generated a luciferase construct containing solely these 3 sites (–139 bp to +1 bp from TSS) and observed that C/EBPα could still significantly repress luciferase activity (Figure (Figure2D).2D). Deletion or mutation of these sites (Supplemental Figure 2F) resulted in a lack of reporter activity, pointing to a critical role of E2F in transcriptional control of CEBPG (data not shown). Several mechanisms may explain how C/EBPα regulates CEBPG through these E2F-binding sites. First, we assessed whether E2F binding to the CEBPG promoter could be affected by C/EBPα. ChIP performed on K562-C/EBPαER cells at distinct time points upon E2 treatment showed that C/EBPα binds to the proximal promoter of CEBPG without displacing E2F1 (Supplemental Figure 2G). This result is in agreement with a previous publication by Johansen et al. demonstrating that C/EBPα had no effect on E2F1 binding to a consensus E2F DNA–binding site (40). Alternatively, C/EBPα may interfere with E2F transcriptional activity and repress the CEBPG promoter. We demonstrated that E2F1 protein (but not other E2F family members) could transactivate the CEBPG promoter and that E2F1-induced reporter activity could be reduced by increasing amounts of C/EBPα expression plasmid (Figure (Figure2E).2E). Together, these results indicate that C/EBPα can bind to CEBPG proximal promoter and downregulate C/EBPγ expression by affecting E2F1 transcriptional activity.
Next, we investigated whether changes in chromatin structure, which regulate gene expression, would be present in the C/EBPα-KO cells versus WT. DNA methylation HpaII tiny fragment enrichment by ligation mediated PCR (HELP) arrays in lineage–c-kit+ murine bone marrow cells demonstrated no differences in the Cebpg DNA methylation status from KO versus WT mice (Supplemental Figure 2H). Similarly, HELP arrays in human AML patient samples showed no differences in the CEBPG DNA methylation profile of samples with CEBPA silencing and other AML subtypes (data not shown). Further, ChIP using H3K4me3 and H3K27me3 antibodies in lineage–c-kit+ cells demonstrated increased H3K4me3 enrichment in the Cebpg promoter of C/EBPα-KO compared with WT cells (Supplemental Figure 2I), whereas no difference was observed in the enrichment of the H3K27me3 mark (data not shown). Together, these experiments indicate that the active histone modification mark H3K4me3, but not changes in DNA methylation, contributes to the elevated Cebpg levels observed in the C/EBPα-KO cells.
To understand the role of C/EBPγ in the pathogenesis of AML, we first investigated the expression of this transcription factor in normal hematopoiesis and determined Cebpg expression levels in distinct hematopoietic progenitor cell populations. RT-PCR showed that Cebpg transcripts are expressed in the LKS (lineage–c-kit+ Sca-1+) population, which is enriched in both long- and short-term HSCs. Cebpg expression is still high in the CD150–CMP subset, but decreases significantly as cells commit to the myeloid lineage and reach the CD150+CMP and GMP stages (Figure (Figure3A).3A). Cebpg mRNA levels are also high in megakaryocyte-erythroid progenitors (MEP), independently of endoglin and CD150 markers, and in common lymphoid progenitors (CLP).
As C/EBPα is a crucial factor for myelopoiesis, we analyzed Cebpa expression in parallel in the same sorted populations. We observed that Cebpa is also expressed in LKS cells, however, with further lineage commitment, the expression of Cebpa and Cebpg followed a reciprocal pattern, showing upregulation of Cebpa in CD150+CMP and GMP cells (Figure (Figure3A3A and Supplemental Figure 3A). This inverse correlation was also seen in CD150–CMP and MEP, as well as CLP, where Cebpg levels were increased and Cebpa reduced (Figure (Figure3A3A and Supplemental Figure 3A). In summary, Cebpg is expressed at the earliest hematopoietic stem/progenitor stage and at highest levels in MEP and CLP, with relatively lower levels in myeloid cells.
To further investigate C/EBPγ expression during granulocytic differentiation, we made use of the 32D/G-CSF–R murine cell line model. 32D/G-CSF–R cells proliferate in the presence of IL-3 and can fully differentiate toward mature neutrophils upon G-CSF stimulation (42, 43). Quantitative RT-PCR and Western blot analysis demonstrated that C/EBPγ is highly expressed in 32D/G-CSF–R cells while proliferating under the control of IL-3 (day 0) (Figure (Figure3B3B and Supplemental Figure 3B). This expression level rapidly decreases as cells differentiate in the presence of G-CSF (day 2 to 8). In contrast, C/EBPα expression levels are low in the presence of IL-3, but gradually increase during G-CSF–induced differentiation (Figure (Figure3B3B and Supplemental Figure 3C). These data demonstrate that C/EBPγ and C/EBPα levels are inversely correlated during neutrophilic differentiation and that C/EBPγ is more abundant in immature cells, similar to the bone marrow–sorted subsets.
To investigate whether downregulation of C/EBPγ is required for granulocytic differentiation, we overexpressed C/EBPγ in the 32D/G-CSF–R system described above. We introduced a tamoxifen-inducible form of C/EBPγ (C/EBPγ-ER) into 32D/G-CSF–R cells. ER control clones (containing the estrogen receptor peptide without C/EBPγ) and C/EBPγ-ER–expressing clones were cultured in the presence of G-CSF with or without 4-hydroxytamoxifen. All clones cultured in the presence of G-CSF fully differentiated toward granulocytes; however, a complete block of differentiation was observed in the C/EBPγ-ER–expressing cells when cultured in the presence of G-CSF and 4-hydroxytamoxifen (Figure (Figure3,3, C and D). Accordingly, upregulation of granulocytic markers such as myeloperoxidase (Mpo), Cebpe, and Csf3r (G-CSF–R) was observed when 32D/G-CSF–R/CEBPγ-ER cells were cultured with G-CSF, but not when 4-hydroxytamoxifen was added to the cultures (Supplemental Figure 3D). ER control clones differentiated toward neutrophils when stimulated with G-CSF+4-hydroxytamoxifen and stopped proliferating after 4–5 days of culture, whereas the C/EBPγ-ER–expressing clones remained proliferative in the same conditions, with no signs of differentiation observed at any given time point (cultures were kept till day 19) (Supplemental Figure 3E).
To investigate the mechanism by which C/EBPγ promotes a block of granulocytic differentiation, we made use of the 32D/G-CSF–R cell line expressing either C/EBPγ-ER or the ER control alone. Cells were treated for 4 hours with 4-hydroxytamoxifen and ChIP was performed using an ER-specific antibody or an IgG control antibody. Oligos were designed on the proximal promoter of putative target genes, such as Csf3r (44) and Cebpe (45), which presented potential C/EBP-binding sites and were downregulated upon 4-hydroxytamoxifen treatment. Quantitative RT-PCR showed enrichment of the Csf3r and Cebpe promoters upon 4-hydroxytamoxifen treatment in the C/EBPγ-ER–expressing cells, but not in the absence of 4-hydroxytamoxifen or the ER control cells (Figure (Figure3E).3E). Further, enrichment was not observed in the IgG antibody control (Figure (Figure3E)3E) or in deserted regions (data not shown).
Next, we investigate whether enforced C/EBPγ expression would affect neutrophilic differentiation of murine bone marrow cells. Cebpg and pMSCV empty vector control were introduced into WT lineage depleted bone marrow cells, and overexpression was verified by quantitative RT-PCR (Supplemental Figure 3F). GFP– (noninfected) and GFP+ (infected) cells were cultured in semi-solid medium, and morphological analysis determined that the C/EBPγ overexpressing cells had reduced number of granulocytic-like colonies (G-CFU), whereas the ability to form other colony types was not affected (Figure (Figure3F).3F). Morphological analysis of cytocentrifuged cells from the semi-solid cultures demonstrated that overexpression of C/EBPγ resulted in a lack of mature neutrophils (Figure (Figure3,3, G and H). Together, these experiments demonstrate that C/EBPγ overexpression induces a block in granulocytic differentiation, suggesting that downregulation of C/EBPγ is required for myeloid differentiation.
C/EBPα-KO mice are characterized by the lack of mature neutrophils in the bone marrow and blood (9, 10). Our observations that LKS from this mouse model have high levels of Cebpg mRNA and that sustained C/EBPγ expression blocks granulocytic differentiation (Figure (Figure1B1B and Figure Figure3,3, C–H) led us to hypothesize that downregulation of C/EBPγ is required for neutrophilic differentiation in vivo. To study the contribution of C/EBPγ to the block in neutrophilic differentiation, we designed several shRNA constructs specifically targeting Cebpg (and not other C/EBPs) and demonstrated their efficiency at protein and RNA levels (Figure (Figure4A4A and Supplemental Figure 4, A and B). C/EBPα-KO LKS cells were infected with a lentivirus expressing either a C/EBPγ-specific shRNA (sh#1) or a nonsilencing control shRNA (NSC#1). Cebpg was significantly downregulated in sh#1-infected LKS cells in comparison with cells infected with the NSC#1 (Figure (Figure4B).4B). Furthermore, Cebpa expression could not be detected in these cells (data not shown). Each individual colony in semi-solid culture was picked, and morphological analysis of cytocentrifuged cells revealed that downregulation of C/EBPγ gave rise to neutrophil-containing colonies (Figure (Figure4,4, C and D). Quantitative RT-PCR of these cultures showed upregulation of granulocyte-specific genes such as neutrophil elastase (Elane), Mpo, Cebpe, and G-CSF–R (Csf3r) in C/EBPγ shRNA-infected cells (Figure (Figure4E).4E). Serial replating experiments demonstrated that while C/EBPα-KO NSC#1-infected cells could be replated for up to 4 times, downregulation of C/EBPγ in these cells abolished their replating abilities (Supplemental Figure 4C). Further, C/EBPα-KO infected cells were also placed in liquid culture, and flow cytometric analysis revealed a decrease in immature cell-surface markers such as c-kit and Sca-1 and an increase in mature markers such as Gr-1 and Mac-1 in sh#1-infected cells in comparison with NSC#1-infected cells (Supplemental Figure 4D). Morphological analysis on Wright-Giemsa–stained cytospins showed a more differentiated phenotype in cells lacking C/EBPα and downregulation of C/EBPγ (Supplemental Figure 4E). Quantitative RT-PCR was used to verify absence of Cebpa in those liquid cultures and downregulation of Cebpg in sh#1-infected cells in comparison with NSC#1-infected cells (Supplemental Figure 4, F–G, and data not shown). Similar experiments were performed using a retroviral vector (46) coexpressing another shRNA specifically targeting Cebpg, shRNA#2, and a GFP reporter (Figure (Figure4A4A and Supplemental Figure 4, A and B). Semi-solid cultures of sorted GFP+ infected cells showed the presence of neutrophil-containing colonies when C/EBPγ was downregulated in C/EBPα-deficient LKS cells in comparison with the NSC#2 colonies (Supplemental Figure 4, H and I). Together, these results show that downregulation of C/EBPγ levels in murine C/EBPα-KO LKS cells is sufficient to restore neutrophilic differentiation in vitro. In addition, serial replating experiments suggest that downregulation of C/EBPγ restricts the increased self-renewal properties of C/EBPα-KO cells (10).
We next studied the effects of knocking down C/EBPγ in the absence of C/EBPα in vivo. For these experiments we made use of the shRNA retroviral vector containing a GFP reporter. C/EBPα-KO LKS cells were isolated and infected with retrovirus containing either shRNA#2, targeting C/EBPγ, or a control scrambled shRNA (NSC#2) and transplanted into lethally irradiated recipient mice (Figure (Figure5A).5A). Figure Figure5B5B shows flow cytometry data from 2 representative mice, transplanted with either sh#2- or NSC#2-infected cells. Plots were gated on Ly5.2+ cells (C/EBPα-KO cells), and further divided into GFP– (noninfected) and GFP+ (infected) cells. We observed that mice transplanted with sh#2 had a high level of Gr-1/Mac-1+ cells (neutrophils) in the GFP+ fraction, whereas an absence of Gr-1/Mac-1+ cells was observed in the sh#2 GFP– fraction or in the NSC#2 transplanted mice (Figure (Figure5,5, B and C, and Supplemental Figure 5A) (n = 8 for NSC#2 and n = 11 for C/EBPγ shRNA#2). Of note, 9 out of 11 mice transplanted with C/EBPγ shRNA#2 contain granulocytes in the GFP+ fraction, whereas 0 out of 8 control mice did. In Supplemental Figure 5A, we measured the percentage of Gr-1/Mac-1+ cells present in blood from reconstituted mice 6 weeks after transplantation. At this time point, Ly5.2+ blood cells were sorted into GFP– and GFP+ cells, and genomic DNA was extracted. Complete excision of CebpaFlox/Flox alleles was observed by PCR of genomic DNA in all transplanted animals and, accordingly, a PCR-specific excision band was detected (Supplemental Figure 5B). Further, GFP– and GFP+ cells from NSC#2 and sh#2 transplanted mice gave rise to B220/IgM+ cells, indicating that downregulation of C/EBPγ in the absence of C/EBPα does not affect the production of B-lymphocytes (Supplemental Figure 6A). Animals were sacrificed 6–10 weeks after transplantation, and the absence of Cebpa was determined by RT-PCR in peripheral blood, total bone marrow, and LKS cells (Supplemental Figure 6, B and C, and data not shown). In summary, these transplantation experiments demonstrate that downregulation of C/EBPγ can rescue granulopoiesis in C/EBPα-KO mice, demonstrating that upregulation of C/EBPγ contributes to the myeloid differentiation arrest.
Since we have identified a subset of patient samples characterized by silencing of CEBPA and upregulation of CEBPG (Figure (Figure11 and Supplemental Figure 1), we next investigated whether downregulation of C/EBPγ in human CEBPA-silenced AML samples would restore granulocytic differentiation. First, we designed and examined the effect of an shRNA sequence specifically targeting human CEBPG (Supplemental Figure 7, A and B). A CEBPA-silenced AML patient sample was then infected with either this C/EBPγ shRNA or an NSC shRNA lentivirus, and cells were transplanted into sublethally irradiated NSG mice. Analysis of the recipient bone marrow cells 16 weeks after transplantation revealed the presence of human CD15-positive cells exclusively in the C/EBPγ shRNA–infected GFP+ cells (Figure (Figure6).6). Further, analysis of the GFP– fraction in these recipients and the NSC transplanted mice showed absence of CD15 expression, although myeloid identity was observed by CD33 expression (Figure (Figure6A).6A). Morphological analysis of human CD45+GFP– and CD45+GFP+ sorted cells showed the presence of mature neutrophils only in the C/EBPγ shRNA GFP+ infected cells (Figure (Figure6B).6B). Similarly to the flow cytometry data, differential counting of the hCD45+ C/EBPγ infected cytospun cells showed 39% versus 1% mature neutrophils in the GFP+ and GFP– fractions, respectively. These experiments indicate that downregulation of C/EBPγ in human CEBPA-silenced AML samples restores granulocytic differentiation in vivo.
Since a subset of the AML patient samples we identified are characterized by silencing of CEBPA due to promoter hypermethylation (refs. 17, 33, Figure Figure1,1, and Supplemental Figure 1), we next investigated whether these AML cases could benefit from treatment with demethylating agents. We observed that CEBPA-silenced AML samples treated with 1 μM 5-aza-2′-deoxycitidine (DAC) showed upregulation of CEBPA mRNA expression in comparison with mock control treatment (Figure (Figure7A).7A). In line with our observations above, we determined that this CEBPA upregulation results in a significant reduction of the CEBPG levels in DAC-treated cells (Figure (Figure7A).7A). Further, we observed that these changes in CEBPA and CEBPG expression in C/EBPα-silenced AML upon DAC treatment result in upregulation of myeloid differentiation markers such as CSF3R, Gelatinase A, CEBPE, and neutrophil elastase (Figure (Figure7B).7B). Similarly, flow cytometric analysis showed upregulation of mature granulocytic markers such as CD15 and CD11b upon DAC treatment in the C/EBPα-silenced AML (Figure (Figure7C).7C). In addition, AML patient samples lacking CEBPA promoter hypermethylation did not respond to DAC treatment, and neither CEBPA nor CEBPG levels significantly changed (Supplemental Figure 7C). Bisulfite sequencing demonstrated DAC induced demethylation of the CEBPA promoter (Supplemental Figure 7, D and E). Together, these data indicate that demethylating agents can restore the C/EBPα-C/EBPγ balance and promote myeloid differentiation in certain AML samples.
Here, in a unique subset of AML defined by the absence of C/EBPα, we found increased CEBPG expression levels. Similar to these human AML cases, in Cebpa conditional KO mice, Cebpa ablation caused an upregulation of Cebpg, demonstrating a direct relation between Cebpg expression and absence of C/EBPα. We showed that overexpression of C/EBPγ in a cell-culture model of granulopoiesis and in murine bone marrow cells induces a neutrophilic differentiation arrest. Accordingly, in C/EBPα-KO mice and in CEBPA silenced human AML, both characterized by a block of neutrophilic development, downregulation of C/EBPγ was sufficient to completely restore granulocytic differentiation (Figure (Figure8).8). These data point to a role for C/EBPγ in impairment of granulocytic development in a subset of human AML. We therefore hypothesize that these leukemia patients could benefit from therapeutic approaches meant to reduce C/EBPγ levels. We also demonstrate that treatment of these AML patient samples with demethylating agents such as DAC results in hypomethylation and reexpression of C/EBPα, which, as we predicted based in our findings in the murine and cell line models, contributes to the downregulation of C/EBPγ and upregulation of myeloid differentiation markers. This specific subset of AML cases is characterized by an aberrant DNA hypermethylation signature (17, 33), and therefore we assume that the beneficial effect of DAC is not restricted to restoring the C/EBPα-C/EBPγ balance. However, our data support an important role of this balance in myeloid differentiation. Currently, demethylating agents are the standard of care for patients with high-risk myelodysplastic syndromes (MDS) (47, 48), reducing the risk of transformation to AML and improving overall survival. But although these agents are being used as therapeutic agents, very few specific gene targets have been identified, and certainly it is difficult to explain the effect on AML cells based on previously identified methylation targets (49, 50). Therefore, identification of the C/EBPα-C/EBPγ axis as a specific target of DAC contributes to our understanding of the mechanisms of action of these drugs.
It is intriguing to ask why CEBPG upregulation is restricted to AML patients with CEBPA methylation/silencing. If CEBPG upregulation is a direct result of defects in C/EBPα, we might expect that other AML subtypes characterized by C/EBPα inactivation would also present elevated CEBPG levels. On one hand, AML cases harboring differentiation-deficient mutations in CEBPA have increased expression of CEBPA and low CEBPG RNA levels. Here, we demonstrate that the N-terminal C/EBPα mutants and the C/EBPα p30 isoform, which is expressed and active in leukemias with mutations in CEBPA (14, 51), are able to suppress Cebpg upregulation and transactivation. On the other hand, FLT3-ITD– and AML1-ETO–positive leukemias present reduced C/EBPα levels and activity (14, 52), but do not show upregulation of CEBPG. Our data indicate that elevated CEBPG levels are directly related to the total absence of CEBPA expression exclusive of CEBPA-silenced leukemias, whereas in the FLT3-ITD and AML1-ETO cases, the CEBPA mRNA levels are reduced, but still present. Accordingly, we demonstrate that in CebpaWT/Flox Mx1Cre+ LKS cells, which harbor reduced Cebpa levels, Cebpg levels are not increased. Therefore, the critical difference between this particular subtype of AML and the rest of leukemias with C/EBPα inactivation is the absence of C/EBPα versus low or mutated C/EBPα. Nevertheless, we cannot exclude that a yet-unidentified factor in CEBPA-silenced leukemias also contributes to CEBPG upregulation. Our data support that modifications at the chromatin level also regulate murine C/EBPγ expression. It would be of interest to study whether regulation at this level also occurs in the distinct AML types.
We previously reported that activating mutations in NOTCH1 are a key hallmark in CEBPA-silenced leukemias (17). These leukemias are phenotypically characterized by the simultaneous expression of myeloid as well as T-lymphoid markers. Our expression data indicate that Cebpg is highly expressed in lymphoid progenitor cells and that C/EBPα is involved in the active repression of Cebpg during myeloid differentiation. Accordingly, C/EPPγ-KO murine models have shown that C/EBPγ is relevant for proper lymphoid differentiation (30). Together, these observations highlight the possibility that C/EBPγ might be directly involved in the development of these biphenotypic leukemias, either by upregulating lymphoid markers and/or by repressing myeloid differentiation.
Our experiments indicate that granulocytic differentiation requires gradual upregulation of C/EBPα and downregulation of C/EBPγ, as seen in defined progenitor bone marrow populations and in a murine myeloid cell model. These results are in agreement with previous observations showing that C/EBPγ expression is reduced during myeloid differentiation (53). The reciprocal expression of C/EBPα and C/EBPγ suggests that these 2 transcription factors may regulate each other. In fact, our data demonstrate that C/EBPα represses Cebpg expression by affecting E2F1 transcriptional activity, which is in agreement with previous publications (40, 41). Although several reports state that only the C/EBPα p42 isoform represses E2F transcriptional activity (41), it has been reported that C/EBPα p30 can also suppress E2F activity (54). In fact, the luciferase reporter assays using the C/EBPγ proximal promoter indicate that both C/EBPα isoforms can repress C/EBPγ transactivation, consistent with our previously published studies demonstrating that p30 and p42 C/EBPα retain the ability to interact with E2F proteins in myeloid cell extracts (39). These different observations could be explained by a cell-type–dependent effect (54), and indeed the C/EBPα-C/EBPγ ratio seems to correlate with lineage commitment, rather than with cell-cycle activity, suggesting that other mechanisms yet to be identified, and that could work in a cell-specific manner, may also be involved in C/EBPγ repression. In contrast, our preliminary data (unpublished observations) suggest that this regulation is not reciprocal, since we do not observe changes in Cebpa in LKS cells that overexpress C/EBPγ. However, we cannot rule out the possibility that this regulation is cell-type– or differentiation-stage specific. Additionally, other factors could be involved in this C/EBPα-C/EBPγ regulation. In fact, since C/EBPα levels are reduced at the later stages of differentiation and C/EBPγ levels need to be low in order to accomplish neutrophilic differentiation, we hypothesize that other transcription factors may be involved in C/EBPγ downregulation. Further, we previously reported that C/EBPα-deficient progenitor/stem cells have increased Bmi-1 expression (10). Bmi-1 is a component of the polycomb group complex and regulates hematopoietic stem cell self renewal by repressing differentiation-related gene expression (55). Given the similarity between Bmi-1 and C/EBPγ expression during myeloid differentiation and in the C/EBPα conditional KO murine model, it will be intriguing to determine how these components interact during differentiation and stem cell self renewal.
It has previously been shown that ectopic expression of other C/EBP members, such as C/EBPα or C/EBPβ, can induce granulocytic differentiation of cell lines or primary cells in vitro (56–59). In contrast, our experiments demonstrate that C/EBPγ overexpression, either induced by ectopic expression in cell culture or by excision of C/EBPα in KO mice, results in a block of neutrophilic differentiation. Therefore, the effect of C/EBPγ overexpression is opposite to the one observed with other C/EBP members, suggesting different mechanisms of action and activation of distinct target genes. Structurally, C/EBPγ resembles the C/EBPβ isoform liver inhibitory protein (LIP) (7). C/EBPγ and LIP retain the basic region-leucine zipper DNA–binding domain, characteristic of the C/EBP family of transcription factors, but lack the transactivation domains. Similarly to LIP, C/EBPγ acts as a transdominant negative inhibitor of other C/EBP members (27, 29). In addition, C/EBPγ could also functionally mimic the C/EBPα mutant proteins identified in AML patient samples. We and others have shown that loss of C/EBPα is not the key event in leukemogenesis, but rather the synergistic activity of C/EBPα p30 and C-terminal mutants (14, 51). Therefore, C/EBPγ could contribute to the granulocytic differentiation arrest characteristic of AML by performing as a C/EBPα mutant protein. The results reported here demonstrate that downregulation of C/EBPγ by an shRNA approach can induce C/EBPα-independent granulopoiesis in vitro and in vivo. This is in agreement with previous publications in which we and others have shown that C/EBPα-KO cells can generate granulocytes in response to cytokines (19, 38, 60) and that this effect was mediated by C/EBPβ (19). Further, expression of C/EBPβ from the Cebpa locus results in normal granulopoiesis, indicating that C/EBPβ can drive granulopoiesis in the absence of C/EBPα (61). In the present study, we indeed observed that C/EBPγ can regulate C/EBPβ levels in murine LKS (Supplemental Figure 8A), and in line with these results,we observed low C/EBPβ mRNA levels in the AML cases with increased C/EBPγ expression (Supplemental Figure 8B). In addition, C/EBPγ modulates C/EBPβ transactivation activity (Supplemental Figure 8C). Next, we demonstrated that expression of C/EBPβ was able to drive neutrophilic differentiation of C/EBPα KO LKS cells (Supplemental Figure 8, D and E). Thus, we hypothesize that the rescue we observed upon C/EBPγ knockdown in C/EBPα-KO mice as well as in CEBPA-silenced leukemias might be partially mediated by C/EBPβ. Supporting our hypothesis, it was previously reported that C/EBPγ forms heterodimers with C/EBPβ and acts as a dominant negative inhibitor of C/EBPβ (27, 29). Further, since C/EBPγ can bind to G-CSF promoter elements (62), we hypothesize that C/EBPγ could also directly regulate G-CSF production and consequently neutrophilic differentiation. In line with this hypothesis, our experiments demonstrate that C/EBPγ binds to the proximal promoters of Csf3r and Cebpe and that enforced C/EBPγ expression prevents upregulation of these myeloid differentiation factors, which in normal hematopoiesis contribute to the granulocytic differentiation.
In summary, we have here identified C/EBPγ as a protooncogene in AML. Sustained C/EBPγ expression in myeloid precursors prevents myeloid differentiation toward mature granulocytes, and downregulation of C/EBPγ can restore neutrophilic differentiation in a C/EBPα-independent manner. Finally, we demonstrated that DAC treatment restores the C/EBPα-C/EBPγ balance affected in a subset of AML cases, contributing to the upregulation of myeloid differentiation markers in vitro (Figure (Figure8).8). To date, treatment results with demethylating agents in AML have not been overwhelmingly positive (63, 64). However, these disappointing results could be explained if only certain subsets of AML patients are responsive. A precedent for this is found in treatment of AML patients with retinoids, in which only a specific subset of patients, those harboring the PML/RARα translocation, are responsive (65, 66). Our data supports DAC treatment and C/EBPα-C/EBPγ axis target approaches as a therapy in AML patients with CEBPA hypermethylation and upregulation of CEBPG.
We made use of the leukemic cell specimens of 2 independent and representative cohorts of AML patients (17, 32, 67). The first cohort included 264 samples and the second 262. Patients not treated according to the Hovon study were excluded from this analysis. GEP of 526 AML cases was carried out on Affymetrix HGU133A Plus2.0 GeneChips. Further details on data analysis have previously been described (17, 32, 67, 68) (GEO GSE6891).
32D/G-CSF–R cells were cultured as previously described (42, 43). K562 and 32D/G-CSF–R cell lines stably expressing C/EBPα-ER, C/EBPγ-ER, or ER alone were maintained in phenol red–free RPMI medium supplemented with 10% charcoal-stripped FBS. Independent clones were analyzed, and induction of nuclear localization of C/EBPα-ER or C/EBPγ-ER fusion protein was achieved by addition of 1 μM β-estradiol or 50 nM 4-hydroxytamoxifen (Sigma-Aldrich), as indicated, into the culture medium.
For ChIP on DNA promoter microarrays (ChIP on chip), 50 to 100 million cells were cultured in the presence of 1 μM 17β-estradiol (E2) (Sigma-Aldrich) for 4 hours. Protein and DNA were crosslinked by incubation with formaldehyde (final concentration 1%; Sigma-Aldrich) for 10 minutes, followed by quenching with glycine (final concentration 0.125 M; Sigma-Aldrich) for 5 minutes. Cells were pelleted, washed twice with ice-cold PBS and 3 times with lysis buffer (10 mM Tris-HCl, 10 mM NaCl, 3 mM MgCl2, 0.5% IGEPAL (Sigma-Aldrich), 1 mM PMSF), and then resuspended in pre–IP dilution buffer (10 mM Tris-HCl, 10 mM NaCl, 3 mM MgCl2, 1 mM CaCl2, 4% IGEPAL, 1 mM PMSF; Sigma-Aldrich]). Material was sonicated in a total volume of 1 ml using a Soniprep 150 device (MSE) for 6 cycles of 20 seconds pulse/90 seconds rest at an amplitude of 6 microns. Cellular debris were removed by pelleting; 10% of the supernatant was kept aside as input control, and the remainder was used for IP after addition of 5 volumes of IP dilution buffer (20 mM Tris-Hcl, 2 mM EDTA, 1% Triton X-100 [Roche], 150 mM NaCl and fresh protease inhibitor [Roche]). Chromatin was precleared by incubation for 30 minutes at 4°C with preequilibrated protein G–coupled magnetic beads (Dynabeads; Invitrogen). IP was carried out overnight at 4°C using 20 μg rabbit polyclonal IgG anti-ERα (sc543X; Santa Cruz Biotechnology Inc.). Samples were incubated for 2.5 hours at room temperature in the presence of 200 μl of preequilibrated magnetic beads. The beads were washed twice with ChIP wash 1 (20 mM Tris-HCl, 2 mM EDTA, 1% Triton X-100, 150 mM NaCl, and 1 mM PMSF), once with ChIP wash 2 (20 mM Tris-HCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS, 500 mM NaCl, and 1 mM PMSF), once with ChIP wash 3 (10 mM Tris-HCl, 1 mM EDTA, 0.25 M LiCl, 0.5% IGEPAL, and 0.5% deoxycholate), and twice with TE. Protein/DNA complexes were eluted into 400 μl elution buffer (25 mM Tris-HCl, 10 mM EDTA, 0.5% SDS) by heating at 65°C for 30 minutes. Crosslinks were reversed overnight at 65°C in the presence of 20 μl proteinase K (20 mg/ml; New England Biolabs). The non-IP 10% input control sample was processed similarly. De-crosslinked material was purified using a QIAGEN PCR Purification Kit and eluted into 20 μl buffer EB.
Other ChIP experiments were performed using ChIP assay kits (Upstate Biotechnology). Briefly, 1 million lineage–c-kit+ bone marrow cells from WT or C/EBPα KO mice were sorted, or 10 million 32D/G-CSF–R and K562 cells stably transfected with C/EBPα-ER, C/EBPγ-ER, or control ER were treated with 1 μM E2 or 50 nM 4-hydroxytamoxifen (Sigma-Aldrich) as indicated. IP was carried out using 20 μg rabbit polyclonal IgG anti-ERα (sc543X; Santa Cruz Biotechnology Inc.), 10 μg anti-C/EBPα (14AA) (Santa Cruz Biotechnology Inc.), 10 μg anti–E2F-1 (clones KH20 and KH95; Millipore), 10 μg anti-H3K4me3 (clone MC315; Millipore), or 10 μg anti-H3K27me3 (6002; Abcam). Oligonucleotide sequences are listed below.
10 μl of DNA isolated by ChIP in 32D/G-CSF–R cells was used for amplification using a Whole Genome Amplification (WGA2) kit (QIAGEN). From 5 to 10 μg of amplified DNA was fragmented following the protocol of Affymetrix SNP array kits, which was confirmed on an Agilent Bioanalyzer. Labeling, hybridization to Affymetrix Mouse Promoter 1.0R arrays, staining, washing, and scanning were performed according to the manufacturer’s recommendations. Raw promoter array data were processed using Model based Analysis of Tiling arrays (MAT) software (69). In these analyses, 2 ER samples and 3 C/EBPα-ER samples were included. Enriched fragments at the 0.0001 significance level were considered to be target regions at a bandwidth of 300, maxgap of 300, and minprobe of 8. The enriched chromosomal regions were mapped to NCBI murine Genome Build 36 (assembled February 2006) and annotated using Cis-regulatory Element Annotation System (CEAS) ( http://ceas.cbi.pku.edu.cn/) (70). CEAS was also used to search for transcription factor–binding motifs from the TRANSFAC and JASPAR databases (70–72). Visualization of enriched regions was carried out using Affymetrix Integrated Genome Browser (IGB) software.
High-molecular-weight DNA was isolated from sorted lineage–ckit+ murine bone marrow cells (n = 3 animals per group) using the PureGene kit from QIAGEN and the HELP assay was carried out as previously described (73, 74). All samples for microarray hybridization were processed at the Roche NimbleGen Service Laboratory. Samples were labeled using Cy-labeled random primers (9 mers) and then hybridized onto a mouse custom-designed oligonucleotide array (50–mers) covering 25,720 HpaII amplifiable fragments (HAF) (>50,000 CpGs), annotated to 15,465 unique gene symbols (Roche NimbleGen, Design name: 2006-10-26_MM5_HELP_Promoter Design ID=4803). HpaII amplifiable fragments are defined as genomic sequences contained between 2 flanking HpaII sites found within 200–2,000 bp of each other, and each one is represented on the array by 15 individual probes randomly distributed across the microarray slide. Scanning was performed using a GenePix 4000B scanner (Molecular Devices) as previously described (75). Quality control and preprocessing of HELP microarrays was performed as described in Thompson et al. (76). Analysis of normalized data revealed the presence of a bimodal distribution. For each sample, a cutoff was selected at the point that more clearly separated these 2 populations, and the data were centered around this point. Each fragment was then categorized as either methylated, if the centered log HpaII/MspI ratio was less than zero, or hypomethylated, if on the other hand the log ratio was greater than zero.
Distinct fragments of the human CEBPG promoter were cloned into the pXP2 firefly luciferase reporter vector: long construct (from –957 bp to +1 bp), short construct (from –354 bp to +1 bp), and E2F-binding sites construct (from –139 to +1 bp). All distances are relative to transcriptional start site. The C/EBP binding sites were mutated into TTATAATA (5′ site) and TAGTTACC (3′ site), using a QuikChange Site-Directed Mutagenesis Kit and following manufacturer’s instructions (Stratagene). C/EBPβ pCDNA3 plasmid was donated by Peter F. Johnson (Center for Cancer Research, National Cancer Institute, Frederick, Maryland, USA). Luciferase experiments were done as previously described (40). Briefly, 1.5 × 105 293T cells, which endogenously express C/EBPγ, were cotransfected with the reporter construct (100 ng) and increasing amounts of C/EBP plasmids (10, 25, 50, 100, and 200 ng). When a luciferase reporter vector containing 4 C/EBP binding sites (77) was used, reporter activity was upregulated by 1 ng C/EBPβ and increasing amounts of C/EBPγ expression plasmid were added to the luciferase reporter assay as indicated. DNA input was equalized by addition of pDNA3 plasmid. Transfections were normalized with the 20 ng of renilla luciferase vector as an internal control.
RNA was isolated by TriReagent (MRC Inc.), treated with DNAseI, and reverse-transcribed into cDNA (Invitrogen). Quantitative RT-PCR was performed using iQ Sybr Green supermix (Bio-Rad). Amplification was done with a Corbett Rotor Gene 6000 (QIAGEN) using the following parameters: 95°C (10 minutes), 45 cycles of 95°C (15 seconds), and 60°C (1 minute). For Taqman Analysis, Hotstart Probe One-step qRT-PCR Master Mix was used. RNA was reverse transcribed for 10 minutes at 50°C, followed by 2 minutes at 95°C,and 45 cycles of 95°C (15 seconds) and 60°C (1 minute). Primer and probe sequences are depicted below.
Briefly, single-cell suspensions were lysed with modified RIPA buffer and whole-cell lysates separated on 10% SDS-PAGE gels. Immunoblots were stained with the following primary antibodies: rabbit C/EBPα (14AA, 1:1000; Santa Cruz Biotechnology Inc.), rabbit C/EBPγ (1:1000; donated by Peter F. Johnson), ER (HC-20, 1:1000; Santa Cruz Biotechnology Inc.), HA (Y-11, 1:1000; Santa Cruz Biotechnology Inc.), HSP90 (1:2000, BD Bioscience), and β-actin (1:10,000; Sigma-Aldrich). All secondary antibodies were HRP conjugated (Santa Cruz Biotechnology Inc.) and diluted 1:5000 for rabbit-HRP and 1:3000 for mouse-HRP.
LKS and progenitor cells were isolated from murine bone marrow as described previously (78). For analysis of peripheral blood and bone marrow, lysis of red blood cells was followed by staining with PE-conjugated anti-CD45 (Ly5.2), Pacific blue–conjugated anti-Gr1, and allophycocyanin-conjugated anti–Mac-1 or allophycocyanin-conjugated anti-CD45 (Ly5.2), PE-conjugated anti-IgM and Pacific blue–conjugated anti-B220. Detection and analysis of human cells in murine bone marrow NSG recipient mice was performed 16 weeks after transplantation by the following staining: PE-conjugated anti-human CD45, PE-Cy7–conjugated anti-human CD33 and Pacific blue–conjugated anti-human CD15. Staining of liquid cultures was done using fluorescein isothiocyanate–conjugated Gr-1 and allophycocyanin-conjugated Mac-1 or allophycocyanin-conjugated Sca-1 and PE-conjugated CD117/c-kit. All antibodies were from eBioscience. Exclusion of dead cells was done by addition of DAPI. Cell sorting was performed using FACSAria and immunophenotyping was done on an LSRII flow cytometer (BD Biosciences). Data were analyzed with FlowJo software (Treestar Inc.).
Murine p42 C/EBPα was cloned into the MSCV-IReS-GFP retroviral vector (17), and p30 C/EBPα MSCV was donated by Ting Xi Liu’s lab (Shanghai Stem Cell Institute, Shanghai Jiao Tong University School of Medicine, Shanghai, China). NSC#2 control shRNA and C/EBPγ-specific shRNA#2 were cloned into the retroviral vector RSG-SIN (46). C/EBPβ was cloned into the MIGR1 retroviral vector. Retrovirus was produced as previously described (17) and concentrated using a Centricon Plus-70 100000 MWCO column (Millipore). Retroviral transduction was performed as described (17). For in vitro cultures, infected GFP+ cells were sorted prior to culture. NSC#1 control shRNA and murine C/EBPγ-specific shRNA#1 into pLKO lentiviral vector were purchased from Open Biosystems. Human C/EBPγ-specific shRNA and an NSC shRNA were cloned into the lentiviral vector pGhU6. 293T cells were cotransfected using lipofectamine 2000 with lentiviral constructs (Gag-Pol and Env). Virus was harvested and concentrated as mentioned above. A single lentiviral transduction was performed in culture dishes (Falcon 1008; BD) in the presence of polybrene (8 μg/ml) (Sigma-Aldrich). Puromycin (2 μg/ml) was added to the cultures 2 days after infection with pLKO constructs. Human AML samples were infected with an MOI = 20 for 6 hours, washed, and transplanted into NSG mice (1.25 × 106 cells/recipient).
Colony assay and replating (79) of sorted lineage-negative BM cells or infected LKS cells was performed with Methocult M3434 medium (Stem Cell Technologies). One day after infection, retrovirally infected GFP+ cells were sorted and placed in Methocult, and lentivirally infected cells were placed in medium supplemented with puromycin (2 μg/ml) (Invivogen). Single colonies were picked up 12 days after culture and cytospun; slides were stained with Wright-Giemsa (Diff-Quick; Baxter Healthcare). 10,000 cells/dish were used for replating assays.
CebpaFlox/Flox Mx1–Cre–, CebpaFlox/Flox Mx1–Cre+, CebpaWT/Flox Mx1–Cre–, CebpaWT/Flox Mx1–Cre+, C57BL/6J (CD45.2+), and congenic B6.SJL-PtprcaPep3b/BoyJ (CD45.1+ congenic C57BL/6J) mice were bred and maintained in the Animal Research Facility at the Center for Life Sciences in accordance with institutional guidelines. CebpaFlox/Flox Mx1–Cre mice were genotyped as previously described (10). Cebpa excision was induced by 3 poly I:C injections (600 μg/mouse, every other day) (Sigma-Aldrich). For adult bone marrow transplantation, 8- to 12-week-old CD45.1+ congenic C57BL/6J recipient mice were lethally irradiated with 9.5 Gy from a 137Cs irradiator (Gammacell 40 Exactor; Nordion International). Cells were transplanted by retroorbital injection. Peripheral blood samples were obtained from facial vein at different time points. Human AML patient samples were transplanted into sublethally irradiated (1.5 Gy) NSG (NOD.Cg-Prkdcscid IL2rgtm1Wjl/szJ) mice (Jackson Laboratory).
AML patient samples were cultured for 96 hours (or as indicated) in BSF60 medium supplemented with 50 ng/ml hSCF, 50 ng/ml hFLT3L, 10 ng/ml hIL-3, 10 ng/ml hTPO and 10% FBS, and in the presence of 1 μM DAC or 0.005% acetic acid. Medium was refreshed 48 hours after culture. RNA (for RT-PCR) and gDNA (for bisulfite sequencing) were isolated after 96 hours of treatment. Bisulfite sequencing was performed following the manufacture’s instructions. Briefly, genomic DNA was isolated, treated with sodium bisulfite using the EZ DNA Methylation kit (Zymo Research) according to the manufacturer’s recommendations, and used as a template for methylation analysis of the Cebpa promoter region (–428 to +64 relative to ATG) in a nested PCR approach, as described previously (17). Sequencing results were analyzed using BiQ analyzer software (80). Primer sets used are indicated below.
Sequences were as follows: sequence NSC#1 (in pLKO vector): ATCTCGCTTGGGCGAGAGTAAG; sequence shRNA#1 (in pLKO vector): GGAAGAGAATGAACGGTTGGAA; sequence NSC#2 (in RSG-SIN vector): TCGCTTGGGCGAGAGTAAG; sequence shRNA#2 (in RSG-SIN vector): GGATCGGAATAGTGACGAA; sequence NSC (in pGhU6 vector): ATCTCGCTTGGGCGAGAGTAA; sequence C/EBPγ shRNA (in pGhU6 vector): GAAGAGAATGAACGGTTGGAA.
Sequences were as follows: murine endogenous C/EBPα forward (F): 5′-GACCATTAGCCTTGTGTGTACTGTATG-3′; murine endogenous C/EBPα reverse (R): 5′-TGGATCGATTGTGCTTCAAGTT-3′; murine C/EBPα overexpression F: 5′-ACGAGTTCCTGGCCGACCT-3′; murine C/EBPα overexpression R: 5′-GGGCTCCCGGGTAGTCAAAG-3′; murine C/EBPγ F: 5′-GCGCAGAGAGCGGAACAA-3′; murine C/EBPγ R: 5′-GTATCTTGAGCTTTCTGCTTGCT-3′; murine ELA 2 F: 5′-CACCATCAGTCAGGTCTTCC-3′; murine ELA 2 R: 5′-AGTCTCCGAAGCATATGCC-3′; murine MPO F: 5′-ATGCAGTGGGGACAGTTTCTG-3′; murine MPO R: 5′-GTCGTTGTAGGATCGGTACTG-3′; murine C/EBPε F: 5′-AAGGCCAAGAGGCGCATT-3′; murine C/EBPε R: 5′-CGCTCGTTTTCAGCCATGTA-3′; murine G-CSF–R F: 5′-TCCGTCACCCTAAACATCTC-3′; murine G-CSF–R R: 5′-TGGAAGGTTTCCTCTGTCAT-3′; murine Trib2 F: 5′-AGCCCGACTGTTCTACCAGA-3′; murine Trib2 R: 5′-AGCGTCTTCCAAACTCTCCA-3′; murine GAPDH F: 5′-CCAGCCTCGTCCCGTAGAC-3′; murine GAPDH R: 5′-CCCTTGACTGTGCCGTTG-3′; 18S F: 5′-ACTGGAATTACCGCCTGGCAC-3′; 18S R: 5′-CGGCTACCACATCCAAGGAAG-3′; human GAPDH F: 5′-CCACATCGCTCAGACACCAT-3′; human GAPDH R: 5′-CCAGGCGCCCAATACG-3′; human G-CSF–R F: 5′-TTTCAGGAACTTCTCTTGACGAGAA-3′; human G-CSF–R R: 5′-CGAGCCGAGCCTCAGTTTC-3′; human C/EBPε F: 5′-CTCCGATCTCTTTGCCGTGAA-3′; human C/EBPε R: 5′-TGGGCCGAAGGTATGTGGA-3; human C/EBPα F: 5′-TCGGTGGACAAGAACAG-3′; human C/EBPα R: 5′-GCAGGCGGTCATTG-3′; human C/EBPα taqman probe: 5′-TGGAGACGCAGCAGAAGGTG-3′; human C/EBPγ F: 5′-GGCTTGAATGTTAAAGGTGTGACC-3′; human C/EBPγ R: 5′-TTGAGTCATGGAAATGGACAACTT-3′; human C/EBPγ F (to verify microarray data): 5′-GGCTAGAGGAGCAGGTACAT-3′; human C/EBPγ R (to verify microarray data): 5′-GCCTGGGTATGGATAACACTA-3′; human C/EBPγ TaqMan probe: 5′-CCGACACCACTCATGTCAATGGCTG-3′; human gelatinase A F: 5′-GTGGGACAAGAACCAGATCACAT-3′; human gelatinase A R: 5′-GTCTGCCTCTCCATCATGGATT-3′; human neutrophil elastase F: 5′-CCACCCGGCAGGTGTTC-3′; and human neutrophil elastase R: 5′-GTGGCCGACCCGTTGAG-3′.
Sequences were as follows: murine oligos set 1 F: 5′-CTTCCCCGACTGTGGTGAG-3′; murine oligos set 1 R: 5′-GGCGTCTGATGCAACCTG-3′; murine oligos set 2 F: 5′-CGAAAGGTTTGATTGCTTC-3′; murine oligos set 2 R: 5′-TATGTGGAGAGCTGCTCT-3′; murine oligos Csf3r F: 5′-TTGCATTACAGGGTCATAGCAC-3′; murine oligos Csf3r R: 5′-TCCTAGGGGTTCCTGGTTTT-3′; murine oligos Cebpe F: 5′-TGGCTTGACACCTCACTCTG-3′; murine oligos Cebpe R: 5′-GGAGGGGTGCTTAGCAGTTA-3′; human Oligos A F: 5′-CCATTGTTCACCGTTGTGACC-3′; human oligos A R: 5′-AGCGTTTTAACACCAGGAAGC-3′; human oligos B F: 5′-CTGTGCCAATCCAAGAGTCCT-3′; human oligos B R: 5′-TATGCTTTCTGCTGGCACAT-3′.
Sequences were as follows: C/EBPα F1: 5′-AAACAAACCTAATTCTAACTTAAA-3′; C/EBPα R1: 5′-GTTAGTTGTTTGGTTTTATTTTTTT-3′; C/EBPα F2: 5′-CCTAAAACAAACAAAAAAAAAAAAC-3′; C/EBPα R2: 5′-GGTTTTGTAGGTGGTTGTTTAT-3′.
We used 2-sided unpaired Student’s t test to determine the statistical significance of experimental results. When indicated, a Wilcoxon or a Mann-Whitney 1-sided test analysis was applied. P value < 0.05 was considered significant.
Patients’ informed consent was obtained in accordance with the Declaration of Helsinki. The study was approved by the Institutional Review Board: Committee on Clinical Investigations of Beth Israel Deaconess Medical Center (Boston, Massachusetts, USA).
We thank Britta Will, David Gonzalez, Hideo Hirai, and all members of the Tenen laboratory for helpful discussions; Pu Zhang, and Susumu S. Kobayashi for careful reading of the manuscript and suggestions; Carol Stocking and Peter F. Johnson for kindly providing us with the RSG-SIN retroviral vector (Stocking), the murine C/EBPγ antibody (Johnson), and the C/EBPβ pCDNA3 plasmid (Johnson); Heike Pahl for providing us with the pGhU6 lentiviral vector; and Li Chai for her generous donation of the NSG mice. We thank Min Deng for providing us with the p30 MSCV plasmid; and John Tigges, Vasilis Toxavidis, and Heidi Mariani from the Harvard Stem Cell Institute/Beth Israel Deaconess Center flow cytometry facility for their help and assistance during cell sorting. This work was supported by NIH R01 grants CA118316 to D.G. Tenen and R. Delwel, HL56745 to D.G. Tenen, a Collegio Ghislieri’s fellowship to G. Amabile, and an EHA research fellowship from the European Hematology Association to M. Alberich-Jordà.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2012;122(12):4490–4504. doi:10.1172/JCI65102.