Comparative analysis of the transcriptome across distant species

Mark Gerstein, Joel Rozowsky, Koon‐Kiu Yan, Daifeng Wang, Chao Cheng, James B. Brown, Carrie Davis, LaDeana Hillier, Cristina Sisu, Jingyi Jessica Li, Baikang Pei, Arif Harmanci, Michael O. Duff, Sarah Djebali, Roger P. Alexander, B. Alver, Raymond K. Auerbach, Kimberly Bell, Peter J. Bickel, Max E. Boeck, Nathan P. Boley, Benjamin W. Booth, Lucy Cherbas, Peter Cherbas, Chao Di, Alexander Dobin, Jörg Drenkow, Brent Ewing, Gang Fang, Megan Fastuca, Elise A. Feingold, Adam Frankish, Guanjun Gao, Peter J. Good, Roderic Guigó, Ann S. Hammonds, Jennifer Harrow, Roger A. Hoskins, Cédric Howald, Long Hu, Haiyan Huang, Tim Hubbard, Chau Huynh, Sonali Jha, Dionna M. Kasper, Masaomi Kato, Thomas C. Kaufman, Robert R. Kitchen, Erik Ladewig, Julien Lagarde, Eric C. Lai, Jing Leng, Zhi John Lu, Michael J. MacCoss, Gemma E. May, Rebecca McWhirter, Gennifer E. Merrihew, David M. Miller, A Mortazavi, Rabi Murad, Brian Oliver, Sara Olson, Peter J. Park, Michael J. Pazin, Norbert Perrimon, Dmitri D. Pervouchine, V Reinke, Alexandre Reymond, Garrett Robinson, Anastasia Samsonova, Gary Saunders, Felix Schlesinger, Anurag Sethi, Frank J. Slack, William C. Spencer, Marcus H. Stoiber, Pnina Strasbourger, Andrea Tanzer, Owen Thompson, Kenneth H. Wan, Guilin Wang, Huaien Wang, Kathie L. Watkins, Jiayu Wen, Kejia Wen, Chenghai Xue, Li Yang, Kevin Y. Yip, Chris Zaleski, Yan Zhang, Henry Zheng, Steven E. Brenner, Brenton R. Graveley, S Celniker, T Gingeras, R WaterstonNature 2014 · DOI

Abstract

By harmonizing massive datasets from the ENCODE and modENCODE projects, researchers have performed the first large-scale comparative analysis of the transcriptomes of humans, flies, and worms. They discovered that despite hundreds of millions of years of evolution, these distant species share "universal" rules for gene expression and follow a remarkably similar "hourglass" pattern of development.

Introduction

The transcriptome—the complete set of RNA transcripts produced by a genome—is the functional readout of an organism's genetic blueprint. While significant progress has been made in mapping the transcriptomes of individual species, comparing them across vast evolutionary distances, such as between humans and insects, has been historically difficult.

The primary challenge lies in the "noise" of evolution: variations in gene lengths, differing numbers of non-coding RNAs, and unique developmental timelines. This study sought to determine if fundamental, ancient principles of transcription remain constant across the animal kingdom, or if every phylum has evolved its own unique regulatory logic.

Methodology and Datasets

The research team utilized a massive resource of 575 RNA-sequencing experiments containing over 67 billion reads, drawing from the ENCODE (human) and modENCODE (Drosophila melanogaster and Caenorhabditis elegans) consortia. The human data included various tissues and cell lines, while the fly and worm data featured dense time-course coverage of embryonic, larval, pupal (for flies), and adult stages.

To ensure a "fair" comparison, the authors employed several novel strategies:

  1. Uniform Processing: A standardized pipeline was developed to annotate and process data across all three species, ensuring that observed differences were biological rather than technical.
  2. Multilayer Co-expression Networks: Using simulated annealing algorithms, the researchers identified "modules"—groups of genes that work together—and linked them across species via orthology.
  3. The Universal Model: A multivariate statistical model was built to predict gene expression levels based on chromatin features (chemical marks on DNA-wrapping proteins).

Conserved Gene Modules

A major finding of the study was the identification of 16 sets of co-expressed genes shared across all three species. These modules are heavily enriched for fundamental biological processes such as the spliceosome, morphogenesis, and signal transduction.

Conserved co-expression modules across human, worm, and fly.

The models achieved high predictive accuracy, with Pearson correlations ranging from 0.73 to 0.84 when predicting mRNA expression from histone marks. Notably, an organism-independent "universal model" performed nearly as well as species-specific models, proving that the logic of the promoter is a conserved feature of animal life.

The Developmental Hourglass and Unexpected Alignments

The researchers confirmed the "hourglass hypothesis" at the molecular level. Embryos of different species appear most distinct at the beginning and end of development but converge on a highly similar "phylotypic stage" in the middle, where gene expression is most synchronized across species.

Developmental alignment showing the phylotypic stage and secondary alignments between species.

Beyond expected embryo-to-embryo pairings, a novel alignment was discovered: worm embryos match fly pupae. This suggests that the massive tissue remodeling occurring during fly pupation shares an ancient genetic program with late-stage worm development.

Consistency in Non-coding Transcription

Approximately one-third of the genome in all three species is transcriptionally active in non-protein-coding ways. Surprisingly, the extent of this non-canonical transcription is nearly identical across species when measured per base pair. While many of these Transcriptionally Active Regions (TARs) overlap with known enhancers, the specific functions of the majority of non-canonical transcription remain to be elucidated.

Conclusion

This comparative analysis reveals that the "math" of gene regulation is remarkably consistent across hundreds of millions of years of evolution. Future work will focus on expanding this universal model to include more species and complex regulatory elements, such as distant enhancers, to further map the fundamental rules of animal life.

Summary updated

BibTeX
@inproceedings{gerstein2014comparative,
  author = {Mark Gerstein and Joel Rozowsky and Koon‐Kiu Yan and Daifeng Wang and Chao Cheng and James B. Brown and Carrie Davis and LaDeana Hillier and Cristina Sisu and Jingyi Jessica Li and Baikang Pei and Arif Harmanci and Michael O. Duff and Sarah Djebali and Roger P. Alexander and B. Alver and Raymond K. Auerbach and Kimberly Bell and Peter J. Bickel and Max E. Boeck and Nathan P. Boley and Benjamin W. Booth and Lucy Cherbas and Peter Cherbas and Chao Di and Alexander Dobin and Jörg Drenkow and Brent Ewing and Gang Fang and Megan Fastuca and Elise A. Feingold and Adam Frankish and Guanjun Gao and Peter J. Good and Roderic Guigó and Ann S. Hammonds and Jennifer Harrow and Roger A. Hoskins and Cédric Howald and Long Hu and Haiyan Huang and Tim Hubbard and Chau Huynh and Sonali Jha and Dionna M. Kasper and Masaomi Kato and Thomas C. Kaufman and Robert R. Kitchen and Erik Ladewig and Julien Lagarde and Eric C. Lai and Jing Leng and Zhi John Lu and Michael J. MacCoss and Gemma E. May and Rebecca McWhirter and Gennifer E. Merrihew and David M. Miller and A Mortazavi and Rabi Murad and Brian Oliver and Sara Olson and Peter J. Park and Michael J. Pazin and Norbert Perrimon and Dmitri D. Pervouchine and V Reinke and Alexandre Reymond and Garrett Robinson and Anastasia Samsonova and Gary Saunders and Felix Schlesinger and Anurag Sethi and Frank J. Slack and William C. Spencer and Marcus H. Stoiber and Pnina Strasbourger and Andrea Tanzer and Owen Thompson and Kenneth H. Wan and Guilin Wang and Huaien Wang and Kathie L. Watkins and Jiayu Wen and Kejia Wen and Chenghai Xue and Li Yang and Kevin Y. Yip and Chris Zaleski and Yan Zhang and Henry Zheng and Steven E. Brenner and Brenton R. Graveley and S Celniker and T Gingeras and R Waterston},
  title = {{Comparative analysis of the transcriptome across distant species}},
  booktitle = {Nature},
  year = {2014},
  doi = {10.1038/nature13424},
}
Figure