Promoter/enhancer-based controllability of regulatory networks

Abstract

Understanding the mechanisms of tissue-specific transcriptional regulation is crucial as mis-regulation can cause a broad range of diseases. Here, we investigated transcription factors (TF) that are indispensable for the topological control of tissue specific and cell-type specific regulatory networks as a function of their binding to regulatory elements on promoters and enhancers of corresponding target genes. In particular, we found that promoter-binding TFs that were indispensable for regulatory network control regulate genes that are tissue-specifically expressed and overexpressed in corresponding cancer types. In turn, indispensable, enhancer-binding TFs were enriched with disease and signaling genes as they control an increasing number of cell-type specific regulatory networks. Their target genes were cell-type specific for blood and immune-related cell-types and over-expressed in blood-related cancers. Notably, target genes of indispensable enhancer-binding TFs in cell-type specific regulatory networks were enriched with cancer drug targets, while target genes of indispensable promoter-binding TFs were bona-fide targets of cancer drugs in corresponding tissues. Our results emphasize the significant role control analysis of regulatory networks plays in our understanding of transcriptional regulation, demonstrating potential therapeutic implications in tissue-specific drug discovery research.

Figure 1

Characteristics of regulatory networks. (A) In different tissues, we counted both promoter-and enhancer-binding regulatory interactions between transcription factors (TF) and target genes. Specifically, we observed that promoter-binding regulatory links tend to occur in more tissues while enhancer-binding interactions were more likely to be tissue-specific. (B) In the schematic toy model of our controllability framework we considered a regulatory network as a bipartite graph, where TFs and target genes in regulatory interactions represent partitions. Specifically, the application of a maximum matching algorithm allowed us to find three different topological configurations with N_D = 3 driver nodes. To find TFs that are indispensable for the control of the underlying regulatory network, we separately eliminated corresponding promoter/enhancer-binding interactions of a given TF. Determining the number of driver nodes in the network thus obtained, we observed that the elimination of promoter or enhancer-binding interactions of TF_a did not change the number of driver nodes. However, the removal of promoter-binding interactions of TF_b in our toy model increased the number of driver nodes, suggesting that the promoter-binding TF_b is indispensable for the control of the underlying regulatory network. In (C), we counted the number of TFs that were indispensable for the control of tissue and cell-type specific regulatory networks as a function of promoter and enhancer binding regulatory interactions. In particular, we significantly found that cell-type specific networks were dominated by TFs that exerted their control through enhancer binding, while we observed the opposite in tissue-specific networks. (D) In turn, we counted the number of cell-type and tissue (inset) specific regulatory networks where a given TF was found indispensable for their control, pointing to cumulative frequency distributions that generally followed exponential decays.

Figure 2

Tissue and cancer specificity of target genes of promoter-binding indispensable transcription factors (TF). (A) Investigating target genes that were regulated by indispensable promoter-binding TFs, we determined their enrichment in sets of genes that were significantly expressed in 30 different tissues (FDR < 0.05, Fisher’s exact test). In (B), we corroborated our findings by determining the enrichment of cancer-specific genes in 16 different types of cancers in sets of target genes of indispensable promoter-binding TFs. In particular, we observed that cancer-specific genes were indeed enriched in all corresponding tissue-specific regulatory networks. In (C), we determined the enrichment of the families of druggable genes, FDA-approved drug-targets, ion channels and G protein-coupled receptors (GPCR) in sets of target genes that were regulated by indispensable promoter/enhancer-binding TFs in each tissue-specific network. (FDR < 0.05, Fisher’s exact test). Generally, we found that target genes of indispensable promoter-binding TFs were enriched with such gene sets in more tissue-specific regulatory networks than target genes of enhancer-binding or generally indispensable TFs. (D) Focusing on a list of drugs that were approved for different cancers from the National Cancer Institute (NCI) we determined the enrichment of their drug targets in the sets of target genes of indispensable promoter-binding TFs in tissues closest to corresponding cancer types. Except for Brain Tumor and Urothelial Cancer, we found that drug targets significantly appeared in sets of target genes of indispensable promoter-binding TFs in the corresponding tissue-specific regulatory networks.

Figure 3

Enrichment of drug targets of individual cancer drugs in sets of target genes of indispensable promoter-binding TFs in disease affected tissues. (A) The gene targets of Fluorouracil, a drug approved for colorectal, pancreatic, stomach and breast cancer, were broadly enriched in target sets of indispensable promoter-binding TFs in tissue-related regulatory networks (except fetal rectum tissue in colorectal cancer). (B) Drug targets of Docetaxel were enriched in sets of target genes of indispensable promoter-binding TFs in tissues that developed head and neck, prostate, stomach, and breast cancers. Even though this drug was approved for lung cancer, we found a dilution of its drug targets in lung-specific tissues. (C) Drug targets of Cetuximab were enriched in sets of target genes of indispensable promoter-binding TFs in colon-specific tissues affected by colorectal cancer. Although Cetuximab is approved for head and neck cancer, we did not find any enrichment signals in gland tissues. (D) We found mixed results when we considered drug targets of Bleomycin, a drug that was approved for head and neck, cervical and testicular cancer. (E) Etoposide drug targets were enriched in tissues that were associated with testicular cancer and lung cancer. (F) Aldesleukin drug targets were weakly enriched in sets of target genes of indispensable promoter-binding TFs in tissues related to renal cancer and melanoma.

Figure 4

Characteristics of indispensable promoter/enhancer-binding transcription factors (TF) in cell-type specific regulatory networks. Randomly sampling sets of indispensable promoter and enhancer-binding TFs we found that enhancer-binding TFs that were indispensable for an increasing number of cell-type networks were (A) enriched with disease-causing genes, (B) cancer genes and (C) signaling functions. In turn, we found the opposite, when we considered indispensable promoter-binding TFs.

Figure 5

Tissue and cancer specificity of target genes regulated by indispensable enhancer-binding TFs in cell-type specific regulatory networks. In (A) we determined the enrichment of cell-type specifically expressed genes in sets of target genes of indispensable enhancer-binding TFs in the corresponding regulatory cell-type network (FDR < 0.05, Fisher’s exact test). We observed that such target genes were enriched with genes expressed in blood and immune-related cell types and endothelial cell types. (B) We corroborated our findings by observing the enrichment of genes associated with blood-related cancers such as Acute Myeloid Leukemia (LAML) and Lymphoid Neoplasm Diffuse Large B-cell Lymphoma (DLBC) in sets of target genes of indispensable enhancer-binding TFs in cell-type networks of blood and immune cell. (C) Determining the enrichment of gene targets of drugs approved for Acute Myeloid Leukemia and Non-Hodgkin Lymphoma, we found an enrichment in sets of target genes of indispensable enhancer-binding TFs in the cell-type networks of blood and immune cells.