Genome-Scale Analysis of the Pu.1 Driven Transcriptional Programme That Overcomes the Differentiation Block in a Novel Pu.1-Inducible AML Cell Line

Transcription factors (TFs) play a key role in determining normal haematopoiesis; haematological cancers often being characterised by TF dysregulation. Pu.1 is a TF that plays a major role in myeloid differentiation, with mutations and down regulation of Pu.1 activity described in human acute myeloid leukemia (AML). To investigate the role of Pu.1 in normal and leukemic myeloid fate, we developed a model of inducible Pu.1 rescue in a murine AML cell line, and utilised it to assess genome-wide Pu.1-DNA-binding profiles and the subsequent effects on histone remodelling, gene expression and binding of the cooperative TF CEBPA.

We used a previously described radiation-induced murine AML cell line known to be deficient in functional Pu.1 (Cook et al, Blood 2004; 104:3437-3444) and transduced cells with 4-hydroxy-tamoxifen (OHT)-inducible Pu.1 (PuER) or empty vector (EV). Clonal populations were produced by limiting dilution, and genotype and protein phenotype confirmed by PCR and Western blotting respectively. Incubation in 100nM OHT caused cessation of proliferation, and increased expression of the maturation markers CD11b and Gr-1.

To assess the effects of Pu.1 binding, cells were incubated for 2 hours in 100nM OHT and cross-linked chromatin harvested for ChIP with antibodies against Pu.1, CEBPA and H3K27Ac. ChIP samples were amplified and sequenced, reads were mapped to the mouse reference genome, binding peaks identified, and binding levels normalised for peak length and total read count. RNA from was extracted from the same samples and analysed by microarray expression profiling.

Binding peak analysis revealed an increase in Pu.1 binding sites from 5283 to 15675 following Pu.1 induction, showing that the mutant Pu.1 is able to perform limited DNA-binding, but not in a manner sufficient to induce transcription of genes necessary for normal differentiation. To investigate the effect of Pu.1 binding on histone remodelling, we selected a set of 613 peaks with ≥4-fold increase in Pu.1 read count, ≥4fold increase in H3K27Ac read count, and post-induction minimal H3K27Ac read count of ≥50 reads/region. Interestingly, these highly “dynamic” Pu.1 binding peaks were restricted to non-promoter regions, suggesting increased binding at distal enhancer sites.

These 613 peaks mapped to 553 genes, which showed a small but significant overexpression post-Pu.1 induction, compared to genes adjacent to peaks with no change in Pu.1 (p = 5.357e-16). These genes were intersected with the 284 genes overexpressed in the gene expression array post Pu.1-induction, producing a list of 46 genes of relevant Pu.1 targets. Using Gene Ontology and Functional Term Enrichment analysis, we discovered a significant enrichment for genes overexpressed in bone marrow macrophages post-lipopolysaccharide stimulation (z score = -2.19). Investigating the effect of Pu.1 induction on CEBPA binding, we found that increased Pu.1 binding was enriched in CEBPA peaks which were ≥4fold increased post-Pu.1 induction (p< 0.001), suggesting recruitment of CEBPA in a significant proportion of regions. Further evidence of the relationship between Pu.1 and CEBPA comes from motif analysis of the ≥4fold Pu.1 peaks which showed CEBP motifs present in 30.0% of targets. Similarly, analysis of ≥4fold CEBPA peaks showed Ets motifs in 34.5% of these peaks with ≥ 4fold Pu.1 binding, but only 7.5% in those without.

In conclusion, we present a novel model system where we have examined at genome-scale the effects of Pu.1-DNA binding to overcome the differentiation block in a Pu.1-deficient AML model. Functional Pu.1 binding rapidly induces changes in H3K27Ac profile at key enhancer sites, which subsequently drives the cell into a maturation pathway. The interaction between Pu.1 and CEBPA is also confirmed, with evidence of Pu.1 recruiting CEBPA to DNA-binding sites.