Cancer cell line specific co-factors modulate the FOXM1 cistrome

Abstract

ChIP-seq has been commonly applied to identify genomic occupation of transcription factors (TFs) in a context-specific manner. It is generally assumed that a TF should have similar binding patterns in cells from the same or closely related tissues. Surprisingly, this assumption has not been carefully examined. To this end, we systematically compared the genomic binding of the cell cycle regulator FOXM1 in eight cell lines from seven different human tissues at binding signal, peaks and target genes levels. We found that FOXM1 binding in ER-positive breast cancer cell line MCF-7 are distinct comparing to those in not only other non-breast cell lines, but also MDA-MB-231, ER-negative breast cancer cell line. However, binding sites in MDA-MB-231 and non-breast cell lines were highly consistent. The recruitment of estrogen receptor alpha (ERα) caused the unique FOXM1 binding patterns in MCF-7. Moreover, the activity of FOXM1 in MCF-7 reflects the regulatory functions of ERα, while in MDA-MB-231 and non-breast cell lines, FOXM1 activities regulate cell proliferation. Our results suggest that tissue similarity, in some specific contexts, does not hold precedence over TF-cofactors interactions in determining transcriptional states and that the genomic binding of a TF can be dramatically affected by a particular co-factor under certain conditions.

Keywords: ChIP-seq; FOXM1 reprogramming; breast cancer prognosis; genomic binding; transcription factor.

Figure 1. Schematic depicting the comparison of FOXM1 binding in different cell lines

We compared the difference based on three levels, (A) the raw signal profiles, (B) binding peaks and (C) target genes, to show the different binding of FOXM1 in different cells. (D) We applied the target gene profiles to infer FOXM1 activity and further compared the difference.

Figure 2. Comparison of FOXM1 binding events in different cell lines

(A) PCA analysis of the normalized binding signal of FOXM1 in different ChIP-seq experiments. Colored dots represent different ChIP-seq experiments. The first PC explains 41.13% variation and the second PC explains 14.79% variation. (B) Peak overlap analysis based on the called binding peak in different ChIP-seq experiments. The color bars in left and top represent different ChIP-seq experiments. (C) Genomic regions distribution of FOXM1 binding peaks in different ChIP-seq experiments. (D) Two specific examples of FOXM1 binding.

Figure 3. Comparison of enriched motifs of FOXM1 in different cells

(A) Motif enrichment analyses across all FOXM1 ChIP-seq experiments. The color bars in the left represent different cell lines. The value in heatmap is log 2 transferred enrichment score (ES). (B) Five specific examples show the different binding of FOXM1 in different ChIP-seq experiments. Barplot was performed in log2 transferred enrichment scores. * represents the significance of enrichment (FDR < 0.01). Mann Whitney Wilcoxon Test p-value was showed in the FOXA1 and ESR1 examples.

Figure 4. Comparison of target genes of FOXM1 in different cells

(A) Heatmap of the enrichment of the target genes of pair-wised ChIP-seq experiments. The color bars around the heat map represent different cells. (B) Heatmap of pathway enrichment results based on negative log 10 transferred p-value. The color bars in the left represent different cells. (C) Enrichment scores comparison of two pathways, the cell cycle and the ER nongenomic, in MCF-7 and other cell lines. Colored bars represent corresponding cells. Mann Whitney Wilcoxon Test p-value was showed.

Figure 5. Associations between different FOXM1 activities and primary breast cancer sample prognosis

(A) Boxplot for iRAS_MCF-7 of ER+ and ER- patients. (B) Survival curve for ER+ and ER- patients with high or low iRAS_MCF. (C) Boxplot for iRAS_MDA of ER+ and ER- patients. (D) Survival curve for ER+ and ER- patients with high or low iRAS_MDA. (E) Distribution of patients based on both iRAS_MCF-7 and iRAS_MDA. Red dots: patients with both positive iRAS_MCF-7 and iRAS_MDA. Blue dots: patients with both negative iRAS_MCF-7 and iRAS_MDA. Green dots: patients with positive iRAS_MCF-7 and negative iRAS_MDA. Pink dots: patients with negative iRAS_MCF-7 and positive iRAS_MDA. The percentages at four corners are the fractions of patients with ER+ in the corresponding group. The number 1~4 mapped to the survival curves. (F) Survival curves for the four patient groups.

Figure 6. Co-binding patterns for FOXM1 in different cells

(A) Two possible co-binding pattern of ERa and FOXM1 in MCF-7 cell lines. (B) Four possible co-binding patterns in MDA-MB-231 and other non-breast cell lines.