HeRA: an atlas of enhancer RNAs across human tissues

Zhao Zhang, Wei Hong, Hang Ruan, Ying Jing, Shengli Li, Yaoming Liu, Jun Wang, Wenbo Li, Lixia Diao, Leng HanNucleic Acids Res 2021 · DOI

TL;DR

This paper introduces HeRA, a comprehensive database and web portal that maps the expression landscape of enhancer RNAs (eRNAs) across 54 human tissues. By integrating multiple large-scale datasets, HeRA provides a valuable resource for understanding eRNA function, their associations with human traits, and their regulatory roles in gene expression.

What Problem Does This Paper Address?

Enhancer RNAs (eRNAs) are a class of non-coding RNAs transcribed from enhancer regions, playing crucial roles in gene regulation. Despite their known importance, the comprehensive expression patterns of eRNAs in normal human tissues and their associations with various traits remained largely unexplored. This lack of a centralized, accessible resource hindered functional and mechanistic investigations into eRNAs. The authors aimed to fill this gap by creating a detailed atlas of eRNA expression and regulatory interactions across a wide range of human tissues.

Methods

The researchers employed a multi-step computational pipeline to identify and characterize eRNAs:

  • eRNA Annotation and Quantification: They integrated enhancer annotations from ENCODE, FANTOM, and the Roadmap Epigenomics Project, filtering for regions likely to produce eRNAs. RNA-sequencing data from the Genotype-Tissue Expression (GTEx) project was used to quantify eRNA expression. A stringent filtering process was applied to exclude regions overlapping with known transcripts and to identify detectable eRNAs with a reads per million (RPM) value greater than 1.
  • Trait Association Analysis: They analyzed associations between eRNA expression and human traits (gender, race, age, height, weight, BMI) using statistical tests (t-tests, ANOVA, Spearman correlation) on GTEx data.
  • eRNA Regulator Identification: Putative transcription factor (TF) regulators were identified by analyzing co-expression between eRNAs and known TFs across tissues, using Spearman correlation. TF binding site (TFBS) data from ENCODE was used to validate these associations.
  • eRNA Target Gene Identification: Putative target genes were identified based on proximity (within 1MB) and significant co-expression with eRNAs in each tissue.
  • Database Development: A user-friendly web portal, HeRA, was developed using Bootstrap 4, HTML, CSS, and JavaScript, with a Python 2.7.2 backend and Django web framework. Data analysis and plotting were performed using R.

Datasets

The study utilized several large-scale public datasets:

  • GTEx (Genotype-Tissue Expression) Project: Provided RNA-seq data for 9,577 samples across 54 human tissues, along with trait information (gender, race, age, height, weight, BMI).
  • ENCODE (Encyclopedia of DNA Elements) Project: Contributed enhancer annotations and ChIP-seq data for TF binding sites.
  • FANTOM (Functional Annotation of the Mammalian Genome) Project: Provided enhancer annotations.
  • Roadmap Epigenomics Project: Contributed enhancer annotations.
  • TF Databases: AnimalTFDB, DBD, JASPAR, and TF2DNA were used for transcription factor annotations.
  • Gene Annotation Databases: Ensembl, GENCODE, and UCSC were used for gene annotations.

Key Insights and Ideas

  • eRNAs are Ubiquitous and Tissue-Specific: The study identified over 45,000 detectable eRNAs, revealing significant variation in their expression levels and presence across different human tissues, suggesting tissue-specific regulatory roles.
  • eRNAs Associate with Human Traits: The research uncovered tens of thousands of associations between eRNA expression and traits like gender, race, and age, opening avenues for understanding the molecular basis of these variations.
  • Co-expression Networks Reveal Regulatory Relationships: The construction of eRNA co-expression networks allowed for the identification of millions of potential eRNA regulators (TFs) and target genes, providing a framework for dissecting gene regulatory mechanisms.
  • HeRA as a Centralized Resource: The development of the HeRA database and portal is a significant contribution, making complex eRNA data accessible and queryable for the broader research community, thereby facilitating further discoveries.

Key Results

  • 45,411 detectable eRNAs were identified across 54 human tissues.
  • Tens of thousands of associations were found between eRNAs and human traits (gender, race, age).
  • Millions of putative eRNA regulators and target genes were identified through co-expression network analysis.
  • The HeRA database provides access to eRNA expression profiles, trait-related eRNAs, and eRNA co-expression networks for browsing and downloading.

Limitations and Future Work

  • Reference Mapping Bias: The authors acknowledge that using SNP-intolerant alignment, as employed by GTEx, might introduce reference mapping bias, potentially leading to false positive co-expression patterns. They suggest that future studies might need to assess the impact of this bias and consider allele-specific RNA-seq alignment.
  • Normalization Methods: They note that standard normalization methods (RPM, RPKM, FPKM) can be vulnerable to confounding artifacts in large-scale association studies. They mention principal component normalization as a potential solution for future work.
  • Cross-Species Conservation: Investigating the cross-species conservation of eRNAs is identified as an interesting future direction, although it is acknowledged as challenging.
  • Continuous Updates: The authors plan to continuously update the HeRA database with new data and analyses to enhance its utility for the research community.

Summary updated

BibTeX
@article{zhang2021hera,
  author = {Zhao Zhang and Wei Hong and Hang Ruan and Ying Jing and Shengli Li and Yaoming Liu and Jun Wang and Wenbo Li and Lixia Diao and Leng Han},
  title = {{HeRA: an atlas of enhancer RNAs across human tissues}},
  journal = {Nucleic Acids Res.},
  year = {2021},
  doi = {10.1093/nar/gkaa940},
}
Figure