Epigenetic features improve TALE target prediction

Annett Erkes,Jan Grau,Maik Reschke,Stefanie Mücke,Jens Boch

doi:10.1186/s12864-021-08210-z

Abstract

BackgroundThe yield of many crop plants can be substantially reduced by plant-pathogenic Xanthomonas bacteria. The infection strategy of many Xanthomonas strains is based on transcription activator-like effectors (TALEs), which are secreted into the host cells and act as transcriptional activators of plant genes that are beneficial for the bacteria.The modular DNA binding domain of TALEs contains tandem repeats, each comprising two hyper-variable amino acids. These repeat-variable diresidues (RVDs) bind to their target box and determine the specificity of a TALE.All available tools for the prediction of TALE targets within the host plant suffer from many false positives. In this paper we propose a strategy to improve prediction accuracy by considering the epigenetic state of the host plant genome in the region of the target box.ResultsTo this end, we extend our previously published tool PrediTALE by considering two epigenetic features: (i) chromatin accessibility of potentially bound regions and (ii) DNA methylation of cytosines within target boxes. Here, we determine the epigenetic features from publicly available DNase-seq, ATAC-seq, and WGBS data in rice.We benchmark the utility of both epigenetic features separately and in combination, deriving ground-truth from RNA-seq data of infections studies in rice. We find an improvement for each individual epigenetic feature, but especially the combination of both.Having established an advantage in TALE target predicting considering epigenetic features, we use these data for promoterome and genome-wide scans by our new tool EpiTALE, leading to several novel putative virulence targets.ConclusionsOur results suggest that it would be worthwhile to collect condition-specific chromatin accessibility data and methylation information when studying putative virulence targets of Xanthomonas TALEs.

Highlights

The yield of many crop plants can be substantially reduced by plant-pathogenic Xanthomonas bacteria
Evaluation of prediction results To compare the impact of the two epigenetic features, we evaluate the following prediction variants: original PrediTALE model without epigenetic features (P), EpiTALE with consideration of methylation (P + Methyl), EpiTALE with filtering based on the accessibility filter criterion (P + Access) and EpiTALE with methylation and accessibility filtering (P + Methyl + Access)
An accessibility-based filter that predominantly excludes previous false positive predictions of target boxes may achieve a better enrichment of true positives among the top predictions of EpiTALE

Summary

Introduction

The yield of many crop plants can be substantially reduced by plant-pathogenic Xanthomonas bacteria. The infection strategy of many Xanthomonas strains is based on transcription activator-like effectors (TALEs), which are secreted into the host cells and act as transcriptional activators of plant genes that are beneficial for the bacteria. The modular DNA binding domain of TALEs contains tandem repeats, each comprising two hyper-variable amino acids These repeat-variable diresidues (RVDs) bind to their target box and determine the specificity of a TALE. Transcription activator-like effectors (TALEs) are sequence-specific DNA-binding proteins that bind to host promoters to activate the expression of downstream genes. If such genes promote disease, they are termed susceptibility genes

Methods

Results

Conclusion