Integrative analysis of single cell expression data reveals distinct regulatory states in bidirectional promoters

Fatemeh Behjati Ardakani, Kathrin Kattler, Karl Nordström, Nina Gasparoni, Gilles Gasparoni, Sarah Fuchs, Anupam Sinha, Matthias Barann, Peter Ebert, Jonas Fischer, Barbara Hutter, Gideon Zipprich, Bärbel Felder, Roland Eils, Benedikt Brors, Thomas Lengauer, Thomas Manke, Philip Rosenstiel, Jörn Walter, Marcel H. SchulzbioRxiv 2018 · DOI

TL;DR

Bidirectional promoters (BPs), where two genes are transcribed in opposite directions from a single regulatory region, are common but poorly understood. This study leverages single-cell RNA sequencing (scRNA-seq) to reveal four distinct regulatory states of BPs, showing how structural features and epigenetic marks like histone modifications and TF binding are intricately linked to these states.

What Problem Does This Paper Address?

The regulation of bidirectional promoters (BPs) is a complex puzzle. While we know BPs are widespread in eukaryotic genomes, how cells integrate various epigenetic signals (like transcription factor binding and chromatin modifications) to precisely control gene expression from these dual-output regions remains largely unclear. Traditional bulk RNA sequencing masks the heterogeneity of gene expression within individual cells, making it difficult to dissect the nuanced regulatory mechanisms at play. This study aims to bridge this gap by using single-cell resolution to uncover distinct regulatory states of BPs and their associated epigenetic landscapes.

Methods

The core of this study lies in the integrative analysis of single-cell RNA sequencing (scRNA-seq) data with bulk epigenomic datasets.

  • Single-cell RNA-seq (scRNA-seq): The researchers generated novel scRNA-seq data for human HepG2 and K562 cell lines. They employed a Smart-seq2 protocol for library preparation and sequencing.
  • Data Processing and Matrix Construction: After quality control and imputation of drop-out events (using scImpute), they constructed a "swapped BP matrix." This matrix arranges the expression of the higher-expressed gene (H) on the right and the lower-expressed gene (L) on the left for each BP across all cells. This standardization facilitates direct comparison and clustering.
  • Hierarchical Clustering: BPs were clustered based on their single-cell expression patterns in the swapped matrix, revealing four distinct transcription states: Bidirectional Lowly Expressed (BLE), Bidirectional Weak Difference (BWD), Bidirectional Strong Difference (BSD), and Bidirectional No Difference (BND).
  • Epigenomic Data Integration: They integrated this scRNA-seq data with publicly available and newly generated bulk ChIP-seq (for histone modifications and transcription factors) and DNaseI-seq data.
  • BP-Specific Enrichment Analysis: A novel method was developed to calculate TF enrichment scores that account for the spatial distribution of ChIP-seq signals around both TSSs of a BP.
  • Structural Feature Analysis: They analyzed structural features like TSS distance and transcript span (the genomic region covered by transcripts originating near the TSS).
  • Statistical Tests: Various statistical tests, including hypergeometric tests, Mann-Whitney tests, and t-tests, were used to assess enrichment, significance of differences, and correlations.

Datasets

  • Single-cell RNA-seq: Novel scRNA-seq data generated for HepG2 and K562 cell lines.
  • Bulk Epigenomic Data:
    • Histone modification data (H3K4me1, H3K4me3, H3K36me3, H3K27me3, H3K9me3, H3K27ac, H3K122ac) for HepG2 cells from the DEEP consortium.
  • DNaseI-seq data for HepG2 cells (DEEP consortium) and K562 cells (ENCODE).
  • TF ChIP-seq data for HepG2 (44 TFs) and K562 (50 TFs) from ENCODE.
  • CAGE data for HepG2 and K562 from ENCODE.
  • Small RNA data for HepG2 and K562 from ENCODE.
  • DNA methylation data for K562 from ENCODE.

Key Insights and Ideas

  • Hidden Heterogeneity in BPs: Bulk RNA-seq obscures the true regulatory diversity of BPs. Single-cell resolution is crucial for uncovering distinct transcriptional states.
  • Four Distinct BP Regulatory States: The study identifies four reproducible states (BLE, BWD, BSD, BND) that are largely conserved between HepG2 and K562 cell lines, suggesting these are fundamental regulatory configurations.
  • Structural Features Dictate State: The distance between Transcription Start Sites (TSSs) and the length of transcribed regions (transcript span) are strongly associated with these states. For example, BND states (high co-expression) have smaller TSS distances, potentially facilitating coordinated regulation.
  • Epigenetic Signatures Align with States: Specific patterns of histone modifications (e.g., repression marks in BLE, active marks in BND) and DNA accessibility (DNaseI-seq) clearly correlate with each of the four identified states.
  • TF Binding Landscape Varies: While many TFs are broadly associated with BPs, the BND state, characterized by high co-expression, shows the highest percentage of positively enriched TFs, suggesting a more complex regulatory network is required for coordinated gene output.
  • Potential for Deconvoluting Bulk Data: The findings suggest that scRNA-seq can serve as a powerful tool to deconvolute masked subpopulations in bulk epigenomic data, enabling a more nuanced understanding of complex regulatory regions.

Key Results

  • Four Transcription States Identified: Hierarchical clustering of scRNA-seq data revealed four distinct states: BLE (both genes low), BWD (H slightly higher than L), BSD (H much higher than L), and BND (similar expression for both genes).
  • State Conservation: 88% of BPs showed the same transcription state in both HepG2 and K562 cell lines, indicating stability across cell types.
  • Structural Associations:
    • BLE states exhibited significantly larger TSS distances.
  • BND states had the smallest median TSS distance.
  • BSD and BWD states showed shorter transcript spans for the higher-expressed (H) gene compared to the lower-expressed (L) gene.
  • Epigenetic Associations:
    • BLE states were enriched for repression-associated histone marks (H3K27me3, H3K9me3) and depleted for active marks.
  • BND states showed the opposite pattern, with high levels of active marks and low DNA methylation.
  • DNaseI-seq signals were widest around TSSs in the BND state.
  • TF Binding: BND states showed the highest percentage of positively enriched TFs per BP.

Limitations and Future Work

  • Post-transcriptional Regulation: The study acknowledges that scRNA-seq measures steady-state mRNA, and post-transcriptional regulation might influence the observed gene clustering.
  • Limited Single-Cell Epigenomics: The study relies on bulk epigenomic data, as single-cell epigenomic technologies are not yet as mature. The authors suggest that future advances in single-cell sequencing technologies that can measure RNA expression, TF binding, and histone modifications simultaneously will be crucial for a complete understanding of BP regulation.
  • Novel Statistical Methods: The paper implies that novel statistical methods could be developed to deconvolute masked subpopulations in bulk epigenomic assays using scRNA-seq data.
  • Mechanistic Insights: While the study identifies strong correlations between states and epigenetic features, further experimental validation would be needed to establish direct causal relationships.
  • Nascent Transcription: The authors note that while they observed differences in steady-state RNA expression, nascent transcription data (like GRO-cap) suggests initiation might be more similar, highlighting the importance of elongation and transcript stability in shaping the final expression output. Future studies could investigate elongation and transcriptional initiation directly using single-cell nascent RNA sequencing.

Summary updated

Abstract

Abstract Background Bidirectional promoters (BPs) are prevalent in eukaryotic genomes. However, it is poorly understood how the cell integrates different epigenomic information, such as transcription factor (TF) binding and chromatin marks, to drive gene expression at BPs. Single cell sequencing technologies are revolutionizing the field of genome biology. Therefore, this study focuses on the integration of single cell RNA-seq data with bulk ChIP-seq and other epigenetics data, for which single cell technologies are not yet established, in the context of BPs. Results We performed integrative analyses of novel human single cell RNA-seq (scRNA-seq) data with bulk ChIP-seq and other epigenetics data. scRNA-seq data revealed distinct transcription states of BPs that were previously not recognized. We find associations between these transcription states to distinct patterns in structural gene features, DNA accessibility, histone modification, DNA methylation and TF binding profiles. Conclusions Our results suggest that a complex interplay of all of these elements is required to achieve BP-specific transcriptional output in this specialized promoter configuration. Further, our study implies that novel statistical methods can be developed to deconvolute masked subpopulations of cells measured with different bulk epigenomic assays using scRNA-seq data.
BibTeX
@misc{ardakani2018integrative,
  author = {Fatemeh Behjati Ardakani and Kathrin Kattler and Karl Nordström and Nina Gasparoni and Gilles Gasparoni and Sarah Fuchs and Anupam Sinha and Matthias Barann and Peter Ebert and Jonas Fischer and Barbara Hutter and Gideon Zipprich and Bärbel Felder and Roland Eils and Benedikt Brors and Thomas Lengauer and Thomas Manke and Philip Rosenstiel and Jörn Walter and Marcel H. Schulz},
  title = {{Integrative analysis of single cell expression data reveals distinct regulatory states in bidirectional promoters}},
  year = {2018},
  doi = {10.1101/315937},
  howpublished = {bioRxiv},
}
Figure