Beyond the AHRE: the Role of Epigenomics in Gene Regulation by the AHR (or, Varied Applications of Computational Modeling in Toxicology and Ingredient Safety) Sudin Bhattacharya Institute for Integrative Toxicology
Canonical AhR signaling pathway Denison MS, Nagy SR, Annu. Rev. Pharmacol. Toxicol (2003)
AhR ChIP-enriched genes and DREs DRE Core Sequence: 5 -GCGTG-3 ChIP-enriched regions associated with a target gene if between 10kb upstream of TSS and 3 UTR DREs considered positive if Matrix Similarity score high compared to consensus derived from bona fide DRE sequences Dere et al Chem. Res. Tox. 2011 3
Inferring AhR Transcriptional Regulatory Network Gene-expression data for multiple time points Identify significant differentially-expressed (DE) genes Identify liganded receptor (AhR)-DNA interactions genome-wide (ChIP-chip / ChIP-seq) Does DE gene have bound AhR? Yes No Is AhR bound to high-scoring DRE on DE gene promoter? X Yes DE gene is direct binding target of AhR TF No DE gene is indirect binding target of AhR TF Co-TF Identify regulatory TFs upstream of DE gene TRANSFAC /CHEA database
Inferring AhR Transcriptional Regulatory Network Does DE gene have bound TF? Yes No Is TF bound to consensus response element on DE gene promoter? X Yes DE gene is direct binding target of TF TF No DE gene is indirect binding target of TF TF Co-TF TRANSFAC database Identify regulatory TFs upstream of DE gene TF TF X Co-TF Assemble into network
AhR Transcriptional Regulatory Network Of 1407 differentially expressed genes, only 632 were AHR-bound, out of which only 144 were directly-bound (AHR binding to positive DRE) - The classical AHRE model seems to account for only ~10% of differential gene expression Gene expression and AhR ChIP dataset: Dere et al. BMC Genomics 2011
Coregulation and Coexpression in the AhR network - Kohonen Self-organizing Maps - Genes grouped according to similarity of regulation by TFs
Coregulation and Coexpression in the AhR network - Kohonen Self-organizing Maps 2 h 4 h 8 h 12 h 18 h 24 h 72 h 168 h - Color of each SOM unit scaled to median log fold change of genes in that unit
https://www.encodeproject.org/
> 150 cell lines > 120 TFs > 13 histone modifications
CYP1A1 regulation in HepG2 cells (ENCODE) H3K4me3 ChIP-seq active promoter signature H3K4me1 ChIP-seq active promoter signature DNase-seq accessible chromatin Putative DREs AHR ChIP-seq (in primary hepatocytes) RNA-pol II ChIP-seq RXRa ChIP-seq JUND ChIP-seq HNF4a ChIP-seq HNF4g ChIP-seq
CYP1A1 regulation in HepG2 vs. Primary B cells Dnase-seq accessible chromatin Hepatocytes ChIP-seq AhR binding High-scoring DREs Primary B cells Dnase-seq accessible chromatin High-scoring DREs
Long-range Chromatin Interactions CYP1A1-1A2 locus Explanations for basal CYP1A2 vs. 1A1 induction?
Goal: Enquire AHR-mediated gene regulation genome-wide using epigenomic features Working hypothesis: Tissue-specific gene regulation by AHR is determined by a combination of: (i) AHR binding to AHREs in regulatory regions, (ii) Chromatin accessibility, and (iii) AHR-mediated long-range chromatin interactions.
Prediction of valid AhR binding sites / AhR-responsive genes AhR binding Core Sequence: 5 -GCGTG-3 Flanking sequences around the 5 -GCGTG-3 core matter for AhR- Arnt binding to DNA Can we detect a predictive signal from data in the absence of explicit rules? Current additive models like PWMs don t account for neighbor site / higher order interactions
Prediction of valid AhR binding sites / AhR-responsive genes
CYP1A1 regulation in HepG2 cells (ENCODE) H3K4me3 ChIP-seq active promoter signature H3K4me1 ChIP-seq active promoter signature DNase-seq accessible chromatin Putative DREs AHR ChIP-seq (in primary hepatocytes) RNA-pol II ChIP-seq RXRa ChIP-seq JUND ChIP-seq HNF4a ChIP-seq HNF4g ChIP-seq
Prediction of valid AhR binding sites / AhR-responsive genes Only ~ 1% of DREs in open chromatin regions are under AHR peaks 19
Modeling the Liver Lobule Periportal (PP) end Centrilobular (CL) end Benhamouche et al, Dev. Cell, 2006
Virtual Liver Lobule created with a Cluster Aggregation Algorithm
Intracellular signaling model Intra-hepatocyte signaling Apc Β-catenin dapc dt dβcat dt dahr dt = a 2 b 2 APC 1 + k 21 Wnt n 2 = a 1 b 1 βcat APC n 1 APC n n 1 + Kd 1 11 n 2 Wnt n 2 + Kd 21 = a 3 + k 31 βcat βcat + Kd 31 k 32 TCDD Ahr + k 33 AhrTCDD b 3 Ahr AhR Cyp1A1 dahrtcdd dt dcyp1a1 dt = k 32 TCDD Ahr k 33 AhrTCDD b 3 AhrTCDD = a 4 + k 41 βcat + k 42 AhrTCDD n3 βcat + Kd 41 AhrTCDD n n 3 + Kd 3 b 4 Cyp1A1 42
Liver Lobule in CompuCell3D Intra-hepatocyte signaling Apc Β-catenin AhR Cyp1A1 Swat et al, Computational Methods in Cell Biology, Methods in Cell Biology (2012)
Basal Cyp1A1 Gradient across Lobule Intra-hepatocyte signaling Apc Β-catenin AhR Cyp1A1
Induced Cyp1A1 Gradient across Lobule Intra-hepatocyte signaling Apc Β-catenin AhR Cyp1A1
Cyp1A1 Induction across Lobule with increasing TCDD dose Heterogeneous acinar induction at low doses Uniform pan-acinar induction at high doses
Agent-Based Computational Modeling of Cell Culture: Understanding Dosimetry In Vitro BEAS-2B cells transfected with the fluorescent reporter rogfp.
Agent-Based Computational Modeling of Cell Culture: Understanding Dosimetry In Vitro
Confluence Vs. Time
Surface Area Distribution among Cells t(mcs) = 0 t(mcs) = 500 t(mcs) = 1000 t(mcs) = 1500 t(mcs) = 2000 t(mcs) = 2500 t(mcs) = 2750 t(mcs) = 3000
Modeling Diet Microbiota Intestinal Health Link Develop a predictive computational model of the effects of dietary fiber on relative compositions of intestinal microbiota Predict intestinal barrier function from the microbiota composition using regression models. Sarah Comstock Adam Moeser Examine the natural clustering of gut microbiota species in the weaning piglet and its alteration by dietary fiber. Use partial least squares regression to model intestinal permeability/barrier function as a function of microbiota composition and host intestinal gene expression. Identify the most important diet-driven factors (microbiota species, specific host genes) contributing to alterations in intestinal permeability/barrier function.
Acknowledgements Norbert Kaminski (MSU) Sarah Comstock Adam Moeser EPA STAR Program NIEHS Superfund Program at MSU Melvin Andersen (The Hamner Institutes / Scitovation) Navya Shilpa Ramya Kumari Max Harlacher Rory Conolly (US EPA) Wan-Yun Cheng