A multi-tool recipe to identify regions of protein-DNA binding and their influence on associated gene expression

Introduction

The genetic make-up of an organism plays a key role in gene regulation, especially during early cell differentiation and development. We can observe this phenomenon in siblings who possess different eye and hair color as a result of differing genetic code. However, epigenetic mechanisms, such as histone modifications, transcription factor binding and DNA methylation, also contribute to the complexity of individuals’ phenotypes as is observed in identical twins who possess the same genetic code while having slightly different features. Phenotypic differences associated with disease and varying stages of development have been mapped to changing patterns in gene regulation; and phenotype can often be attributed to a changing epigenetic landscape rather than hard-coded genetic features.

In order to decode these epigenetic differences, biologists often turn to an analysis based on two experimental assays; RNA sequencing (RNA-Seq) (Nagalakshmi et al., 2008; Wilhelm et al., 2008), which quantifies the amount of (usually messenger) RNA in a cell, and Chromatin Immuno-precipitation sequencing (ChIP-seq) (Johnson et al., 2007; Robertson et al., 2007), which shows where a particular protein binds the genome. Commonly, this protein is expected to have some influence on the mRNA expression of nearby genes (i.e., it is a transcription factor). Thus, by knocking out the gene that codes for the DNA binding protein and observing changes in mRNA expression, the biologist can infer the direct effect of the protein on expression.

When analyzing genomic data, today’s computational biologist may utilize a variety of different tools specific to each step of their analysis process. Not only must they be able to create the perfect marriage between the type of data and the tool, but they must be able to correctly manipulate the output, both for interpretation and for format conversion between tools. For the non-programming biologist, smooth integration of many of these tools is provided through GenomeSpace (Qu et al., 2016, www.genomespace.org) and its user-friendly “recipes” (recipes.genomespace.org). GenomeSpace is a web-based visual workbench that supports a diverse range of bioinformatics tools and data resources popularly used in genomic analyses. Because GenomeSpace provides the ability to reformat data as it moves between software tools, one can create easy to use step-by-step workflows specific to a given analysis task. We refer to these published workflows as “recipes”.

We present one such recipe, currently available in GenomeSpace, which identifies differentially expressed genes between two samples, and compares that gene list with differential transcription factor occupancy from a ChIP-Seq experiment. This recipe is designed to elucidate which DNA-protein binding events are responsible for an observed change in mRNA expression. By identifying protein occupancies throughout the genome and comparing them to observed differences in mRNA expression, we can support hypotheses of functional regulation.

Methods

This recipe takes as input the aligned reads from a differential RNA-seq transcription factor knockout experiment, and aligned reads from a ChIP-Seq experiment for the transcription factor that was knocked out. The output is a visualization of the genomic regions containing both differentially expressed genes and a binding site for the transcription factor. Since all tools used in this recipe are hosted remotely, running the recipe has no system requirements beyond an internet connection. We describe the individual steps of the recipe here.

Use case

We applied the recipe described above to an example dataset from Laurent et al. (2015), accession GSE6328, from NCBI’s Gene Expression Omnibus (GEO) database (Barrett et al., 2013; Edgar et al., 2002). We can identify the interplay between the epigenetics and transcriptomics of mouse embryonic stems cells by observing how the binding of the transcription factor, Prep1, influences gene expression. Prep1 is known for its contribution in embryonic development (Laurent et al., 2015). In comparing genome-wide maps of mouse embryonic cells expressing Prep1 to those that do not, we can identify potential target genes that are being differentially regulated by these binding events. One such example of this is illustrated in Figure 1. Here, the transcription factor binding site has been identified and shown to up-regulate the expression of the gene Igf2.

Figure 1. Epigenetic landscape of Prep1 binding and associated regulation of Igf2.

The left panel illustrates the binding of the Prep1 transcription factor. In the right panel, we see the up-regulation of the gene, Igf2, as a result of this binding event.

Variations of this recipe

This recipe can be used, not only to identify the regulation of genes by transcription factor binding, but also to identify any epigenetic mechanism that can be analyzed by ChIP-sequencing. For example, we can identify regions in the genome where histone modifications have occurred, and match those regions to observed changes in expression presumably resulting from the histone modifications. However, we must consider the nature of the data when selecting parameters in the MACS2 tool in Galaxy. For example, when performing peak enrichment on histone modification occupancies, a user must select an advanced option to include broader regions, since histone modifications are represented by a much broader peak area along the genome.

Data availability

The original ChIP-seq and RNA-seq data of this experiment have been deposited in GEO, with accession number GSE63282. The recipe providing all the detailed steps and corresponding videos associated with this process is accessible at: http://recipes.genomespace.org/view/69.