Key regulators control distinct transcriptional programmes in blood progenitor and mast cells

Abstract

Despite major advances in the generation of genome-wide binding maps, the mechanisms by which transcription factors (TFs) regulate cell type identity have remained largely obscure. Through comparative analysis of 10 key haematopoietic TFs in both mast cells and blood progenitors, we demonstrate that the largely cell type-specific binding profiles are not opportunistic, but instead contribute to cell type-specific transcriptional control, because (i) mathematical modelling of differential binding of shared TFs can explain differential gene expression, (ii) consensus binding sites are important for cell type-specific binding and (iii) knock-down of blood stem cell regulators in mast cells reveals mast cell-specific genes as direct targets. Finally, we show that the known mast cell regulators Mitf and c-fos likely contribute to the global reorganisation of TF binding profiles. Taken together therefore, our study elucidates how key regulatory TFs contribute to transcriptional programmes in several distinct mammalian cell types.

Keywords: gene regulation; haematopoiesis; mast cells; progenitors.

Figure 1. RNA-seq gene expression profiling of HPC7 and mast cells

1. Scatterplot of fpkm values for all genes in both cell types. Each dot on the scatterplot is coloured based on 4 categories: (I) non-cell-type-specific, (II) HPC7-specific, (III) mast-specific and (IV) not expressed.

2. Gene set enrichment analysis of genes in category (II) and (III) against the BioGPS expression dataset for CMP and mast cells. NES denotes normalised enrichment score.

3. Genome browser screenshot of a gene expressed in haematopoietic progenitors (Eng) and a non-cell-type-specific gene (Cdk9). The y-axis represents read density along regions of known transcripts. Higher values correspond to more RNA transcripts.

4. The same scatterplot as in (A) but only all known transcription factors are shown (data from RIKEN transcription factor database: http://genome.gsc.riken.jp/TFdb/). A large proportion of transcription factors are expressed at similar levels in HPC7 and mast cells, while a smaller subset are cell type-specific transcription factors.

5. fpkm values of 10 key haematopoietic transcription factors.

Figure 2. ChIP-Seq binding profile of 10 key haematopoietic transcription factors

1. Comparison of binding sites on the Kit gene locus in HPC7 and mast cells shows many binding site differences between cell types.

2. Global, binding site comparison between HPC7 and mast cells.

3. Hierarchical clustering of the 20 global binding profiles. Each box in the heatmap corresponds to pairwise correlation coefficient of peak coverage data between pairs of samples in the row and column. Boxes on the diagonal indicate perfect correlation of a sample with itself. Ordering of samples in columns is identical to ordering in rows.

Figure 3. Mathematical modelling of gene expression and transcription factor variability

1. The genome browser screenshot shows the Pdgfrb gene locus, a gene expressed in haematopoietic progenitors. This locus illustrates regions where differential scores were calculated. Differential transcription factor binding (ΔTF) was used as predictor variables in linear regression models to predict differential gene expression (ΔGE) as described in the equation on the right.

2. Bar chart indicates the average R² values with standard error from multiple linear regression models with tenfold cross-validation. R² refers to the amount of variability explained by the model. Each model examines all genes bound by at least 1 up to 5 of the 10 shared TFs. 58.6% of genes expressed in either cell type are bound by at least 1 shared TF (Supplementary Fig S2A). The model with ≥ 5 transcription factors has the highest R² value, which corresponds to a high correlation coefficient between observed and predicted values as shown in the scatterplot. Scatterplot corresponding to genes bound by at least 1 shared TF is shown in Supplementary Fig S2A.

3. Average R² values for generalised additive models on all genes bound by ≥ 2 transcription factors based on tenfold cross-validation. A higher R² value was achieved using a GAM with interaction terms. The scatterplot shows the correlation between observed and predicted values in this model.

Figure 4. Motif content analysis of cell type-specific and common binding regions

1. Method for conducting motif content analysis.

2. Heatmap of a subset of all motifs tested. Blue rectangles denote significant (q ≤ 0.05) similarity of enriched consensus sequence in a given dataset (column) to a known motif (row). See Supplementary Fig S3 for all motifs tested. Rectangles are coloured based on the percentage of regions that contained a given motif.

Figure 5. Mast cell-specific transcription factors, Mitf and c-fos

1. Genome browser screenshots of the mast cell-specific gene loci Mcpt4 and Mcpt8 and a shared TF gene locus (Erg). On the left, mast-specific TFs and Gata2 bind to a new region in mast cells that is absent in HPC7. On the right, Gata2 binding is present in HPC7 but a new region in mast cells (arrowheads) is co-occupied with the mast cell-specific TFs.

2. De novo motif analysis of Mitf and c-fos peaks.

3. Co-occupancy between mast cell-specific and shared transcription factors in cell type-specific and common binding regions. For any given TF, TF_i, the total number of peaks in each type of binding region (e.g. mast cell-specific) was taken as 100%.

4. Observed versus predicted values scatterplot for the generalised additive model with pairwise TF interactions. From left to right, the scatterplot shows all genes, Mitf-bound genes and c-fos-bound genes, respectively.

Figure 6. Perturbation of key haematopoietic TFs, Erg and Gata2

1. HPC7 versus mast scatterplot of RNA-seq fpkm values. From left to right, the scatterplot shows all genes, Erg-regulated targets and Gata2-regulated targets, respectively. Points on the scatterplot are coloured based on the log2 fold change of shErg or shGata2 compared to control. Only genes that are differentially expressed in the knock-down (absolute log FC > 0.38, P-value ≤ 0.05) are shown. The complete list of regulated targets and expression values can be found in Supplementary Table S10.

2. Nucleotide sequence alignment of the Cx3cr1 promoter region. Conserved GATA (red) and Ets (blue) motifs are highlighted.

3. MST and FMP6- cells were electroporated with luciferase reporter constructs containing either the promoter or 2 different mutant versions of the promoter (GATA motif or simultaneous mutation of 4 conserved ETS motifs). Mean and SEM for two independent transfections (each one performed in triplicate) are shown. Values are expressed relative to the pGL2-Basic vector (Control).

4. Real-time PCR analysis of ChIP for FLI1, GATA2 and PU.1 following knock-down of the same genes in the Cx3cr1 promoter region. Experiments were performed in MST mast cell line and results are expressed relative to total DNA input.