Labelled regulatory elements are pervasive features of the macrophage genome and are dynamically utilized by classical and alternative polarization signals.

Attila Horvath,Zoltan Simandi,Bence Daniel,Laszlo Nagy,Zsolt Czimmerer,Lajos Szeles,Ixchelt Cuaranta-Monroy,Benedek Nagy,Szilard Poliska,Mate Kiss,Lilla Ozgyin,Laszlo Steiner,Emanuele Raineri,Ivo G Gut,Nikolas Giannakis

doi:10.1093/nar/gkz118

Abstract

The concept of tissue-specific gene expression posits that lineage-determining transcription factors (LDTFs) determine the open chromatin profile of a cell via collaborative binding, providing molecular beacons to signal-dependent transcription factors (SDTFs). However, the guiding principles of LDTF binding, chromatin accessibility and enhancer activity have not yet been systematically evaluated. We sought to study these features of the macrophage genome by the combination of experimental (ChIP-seq, ATAC-seq and GRO-seq) and computational approaches. We show that Random Forest and Support Vector Regression machine learning methods can accurately predict chromatin accessibility using the binding patterns of the LDTF PU.1 and four other key TFs of macrophages (IRF8, JUNB, CEBPA and RUNX1). Any of these TFs alone were not sufficient to predict open chromatin, indicating that TF binding is widespread at closed or weakly opened chromatin regions. Analysis of the PU.1 cistrome revealed that two-thirds of PU.1 binding occurs at low accessible chromatin. We termed these sites labelled regulatory elements (LREs), which may represent a dormant state of a future enhancer and contribute to macrophage cellular plasticity. Collectively, our work demonstrates the existence of LREs occupied by various key TFs, regulating specific gene expression programs triggered by divergent macrophage polarizing stimuli.

Highlights

Specification of cellular identity and function are controlled by lineage-determining transcription factors (LDTFs) [1]
We show that Random Forest and Support Vector Regression machine learning methods can accurately predict chromatin accessibility using the binding patterns of the LDTF PU.1 and four other key TFs of macrophages (IRF8, JUNB, CEBPA and RUNX1)
The genomic binding sites of PU.1 and other TFs in mouse macrophages have been characterized in detail [4,5,33,34,35], their relationship to chromatin openness have not yet been comprehensively studied and it is not known whether their binding shows a strong correlation with chromatin openness or they can bind to low accessible regions as well

Summary

Introduction

Specification of cellular identity and function are controlled by lineage-determining transcription factors (LDTFs) [1]. The master regulator of myeloid differentiation, PU. ( known as SPI1), is an essential LDTF and a potential pioneer factor, which is responsible for enhancer selection upon differentiation [3]. Recent studies in macrophages and B cells suggest a distinct model in which PU., and a relatively small set of additional LDTFs, act collaboratively to bind chromatin in a cell type-specific manner. These collaborative interactions set the stage for signal-dependent transcription factors (SDTFs) to initiate their genomic programs [5,6]

Methods

Results

Conclusion