Transcription factor-associated combinatorial epigenetic pattern reveals higher transcriptional activity of TCF7L2-regulated intragenic enhancers

Abstract

Background: Recent studies have suggested that combinations of multiple epigenetic modifications are essential for controlling gene expression. Despite numerous computational approaches have been developed to decipher the combinatorial epigenetic patterns or "epigenetic code", none of them has explicitly addressed the relationship between a specific transcription factor (TF) and the patterns.

Methods: Here, we developed a novel computational method, T-cep, for annotating chromatin states associated with a specific TF. T-cep is composed of three key consecutive modules: (i) Data preprocessing, (ii) HMM training, and (iii) Potential TF-states calling.

Results: We evaluated T-cep on a TCF7L2-omics data. Unexpectedly, our method has uncovered a novel set of TCF7L2-regulated intragenic enhancers missed by other software tools, where the associated genes exert the highest gene expression. We further used siRNA knockdown, Co-transfection, RT-qPCR and Luciferase Reporter Assay not only to validate the accuracy and efficiency of prediction by T-cep, but also to confirm the functionality of TCF7L2-regulated enhancers in both MCF7 and PANC1 cells respectively.

Conclusions: Our study for the first time at a genome-wide scale reveals the enhanced transcriptional activity of cell-type-specific TCF7L2 intragenic enhancers in regulating gene expression.

Keywords: Intragenic enhancer; MCF7; PANC1; T-cep; TCF7L2.

Fig. 1

T-cep algorithm and training results for TCF7L2-omics data. a The workflow of T-cep algorithm. Shown is a schematic summary of the four steps needed for de novo identifying transcription factor associated chromatin state (TF-state). The approach begins with preprocessing the ChIP-seq data into an alphabet of 2ⁿ observations, builds a HMM to generate a best selected model without infrequent state, re-estimate transition and emission probabilities and then annotate biological meaningful TF-states. b Result of T-cep approach. The emission probabilities of each mark are independent of others for the selected 18-state HMM. The mark probability of greater than 0.1 is considered to be associated with a chromatin state. c A screenshot of a genomic region in Chr8 of MCF-7 annotated by T-cep in UCSC genome browser. ChIP-seq of TCF7L2 and histone marks are in the first nine lines, while the tenth track represents the main output of the T-cep, annotated different states represented by different colors. RefSeq gene (HG19) genomic positions are shown in the last line

Fig. 2

The annotation of chromatin states for TCF7L2-omics data. a The correlation with other genomic features, the distribution of each state bins as well as the average of gene expression for each of 18 states. The average gene expression is normalized by z-score and % total bin shows the percentage of each state in the human genome. b The biological interpretations of the HMM states and identification of four TCF7L2-assoicated States according to the emission probability and other genomic information. c The distribution of four TF-associated States on the gene structure. All genes’ length is normalized, representing the 5’–3’ region as gene body region and extending 90 kb of up/down steam for gene surrounding distribution. d Summary of bin distribution of three TCF7L2-associated states in five cancer cell types

Fig. 3

Genes with TCF7L2-associated states in MCF7 and PANC1 cells. a Identification of gene associated with TCF7L2-states in NCI-60 cell showing cancer type specificity. b A Venn diagram showing an overlapping set of genes among TCFL2-associated States in two cancer cell lines. c Boxplot of gene expression level showing TCF7L2-associated intragenic enhancers with the highest expression among three TCF7L2-associated states. Expression levels were log2-transformed, with genes having expression level <1 considered to have an expression level of 1. d The KEGG pathways for TCF7L2 promoters, TCF7L2 distal enhancers and TCF7L2 intragenic enhancers

Fig. 4

A comparison of T-cep with ChromHMM. a The emission probabilities of 18 states trained by T-cep and ChromHMM respectively. b The common and unique number of TF-state bins in each of three TCF7L2-associated functional states between T-cep and ChromHMM. c Boxplot of gene expression associated with those unique bins in each of three TF-states for T-cep and ChromHMM respectively. d Three TF-states’ segmentation (regions) identified by T-cep corresponding to states identified by ChromHMM in PANC1 cells

Fig. 5

Functional validations of the predicted TCF7L2 enhancers by T-cep. a Quantitative PCR analysis of selected gene associated with TCF7L2-enhancers in MCF7 and PANC1 cells after TCF7L2 knockdown. Selected genes contain both of TCF7L2 intragenic and distal enhancers for further analysis. All data are normalized by GAPDH mRNA and represent the average of three independent experiments with p-values less than 0.05 (t-test). b Luciferase reporter activity assay in TCF7L2 knockdown cells. MCF7 and PANC1 cells are transient transfected with intragenic or distal TFE-pGL3 vector at the TCF7L2-siRNA or non-targeting siRNA conditions. c Luciferase reporter activity assay in TCF7L2 overexpression cells. Intragenic or distal TF-enhancer-pGL3 vector and TCF7L2-pcDNA or empty vector are transient transfected into MCF7 and PANC1 cells