In silico analysis of the transcriptional regulatory logic of neuronal identity specification throughout the C. elegans nervous system

Lori Glenwinkel, Seth R. Taylor, Kasper Langebeck‐Jensen, Laura Pereira, Molly B. Reilly, Manasa Basavaraju, Ibnul Rafi, Eviatar Yemini, Roger Pocock, Nenad Šestan, Marc Hammarlund, David M. Miller, Oliver HoberteLife 2021 · DOI

TL;DR

This study uses computational analysis of gene expression and DNA binding motifs to map the regulatory logic that specifies neuronal identity across the entire C. elegans nervous system. The findings strongly support the "terminal selector" model, where a few key transcription factors orchestrate the expression of gene batteries that define each neuron type, and reveal that this combinatorial control is widespread.

What Problem Does This Paper Address?

The nervous system is incredibly diverse, with hundreds of distinct neuron types. Understanding how this diversity arises from a relatively small number of genes is a fundamental question in developmental biology. Specifically, how is the unique set of genes (the "gene battery") that defines each neuron type turned on and off in a precise manner? Previous work in C. elegans suggested a model where a few "terminal selector" transcription factors (TFs) directly control large sets of these identity-defining genes. However, the generality of this model across the entire nervous system and the precise mechanisms of this control were not fully understood. This paper aims to systematically investigate this regulatory logic across all neuron types in C. elegans.

Methods

The core of this study relies on a computational pipeline called TargetOrtho2. This pipeline integrates two key types of data:

  1. Single-cell RNA sequencing (scRNA-seq) data: This provides a detailed map of gene expression for each of the 118 neuron classes in C. elegans.
  2. DNA binding motif data: This consists of experimentally determined DNA sequences that specific TFs bind to.

TargetOrtho2 uses a phylogenetic footprinting approach. It looks for DNA binding motifs that are conserved across multiple nematode species within the regulatory regions (upstream and introns) of genes. The rationale is that conserved binding sites are more likely to be functionally important for gene regulation.

A key innovation is the use of supervised learning within TargetOrtho2. Previously validated TF target genes (from existing literature) were used to train a classifier (specifically, a Gaussian Process Classifier) to predict novel TF target genes. This classifier ranks candidate target genes based on features like motif conservation, frequency, and binding site strength across multiple species.

The study also employed in vivo experimental validation, including creating transgenic reporter lines and performing genetic loss-of-function experiments to confirm predicted TF-target gene relationships and the functional importance of specific DNA binding sites.

Datasets

  • Single-cell RNA sequencing (scRNA-seq) data: From Taylor et al., 2021, providing differential gene expression profiles for all 118 C. elegans neuron classes.
  • Brain Atlas data: A curated collection of published reporter gene expression patterns from Wormbase.org, covering all 118 neuron classes.
  • CISBP database: Containing DNA recognition sites (motifs) for 395 C. elegans TFs.
  • Experimentally derived DNA binding motifs: From various sources, including protein binding microarrays (PBMs) and validated in vivo studies.

Key Insights and Ideas

  • Widespread "Coordinated Regulation": The study found that the vast majority of neuron classes (117 out of 118) exhibit a "coordinated regulatory" motif signature, strongly supporting the terminal selector model. This means a few TFs likely control large sets of identity genes.
  • Combinatorial Control is the Norm: Instead of single TFs acting alone, the analysis revealed that multiple TFs often co-regulate the same target genes. This suggests that neuronal identity is established through complex "TF collectives" or "terminal selector complexes."
  • Phylogenetic Footprinting as a Powerful Predictor: The conservation of TF binding sites across related species is a strong indicator of functional relevance, enabling accurate prediction of TF targets.
  • In Vitro Motifs Translate to In Vivo Function: The study demonstrated that DNA binding motifs determined from in vitro experiments (like PBMs) accurately predict in vivo regulatory relationships, validating the use of these motifs for large-scale analysis.
  • A "Blueprint" for Neuronal Specification: The comprehensive mapping of TF-gene relationships across the entire nervous system provides a foundational resource for understanding how neuronal diversity is genetically specified.

Key Results

  • 117 of 118 neuron classes predicted to have terminal selectors: This indicates a highly conserved regulatory logic across the nervous system.
  • 124 TFs identified as contributing to "coordinated regulatory" motif signatures: This highlights the extensive network of regulators involved.
  • 98% of co-expressed TF pairs showed common target genes: This strongly supports combinatorial control and shared regulatory mechanisms.
  • Experimental validation confirmed predicted TF-target gene relationships and binding site function for several TFs (e.g., ODR-7, CFI-1, UNC-86), demonstrating the predictive power of the TargetOrtho2 pipeline.
  • TargetOrtho2 pipeline significantly improved computation time (approx. 50x faster than previous version), enabling this large-scale analysis.

Limitations and Future Work

  • Reliance on Known TF Binding Sites: The pipeline requires pre-defined TF binding motifs. It cannot discover novel motifs de novo, meaning the analysis may underestimate the number of regulators if their binding sites are unknown.
  • Ambiguous Motifs: 94 DNA binding motifs were excluded due to high similarity to other TFs, potentially leading to underestimation of regulators where multiple TFs share binding sites.
  • In Vitro vs. In Vivo Binding: While the study shows good agreement, in vitro binding sites may not perfectly reflect in vivo conditions.
  • Indirect Regulation: The study acknowledges that some TF effects might be mediated through other TFs ("feedforward loops"), which are not directly captured by binding site analysis alone.
  • Future Work: The authors suggest that this phylogenetic analysis approach can be applied to other well-conserved genomes to study neuronal regulation in different organisms. They also highlight the potential for further experimental validation of predicted TF-target gene relationships and combinatorial interactions.

Summary updated

Abstract

The generation of the enormous diversity of neuronal cell types in a differentiating nervous system entails the activation of neuron type-specific gene batteries. To examine the regulatory logic that controls the expression of neuron type-specific gene batteries, we interrogate single cell expression profiles of all 118 neuron classes of the <i>Caenorhabditis elegans</i> nervous system for the presence of DNA binding motifs of 136 neuronally expressed <i>C. elegans</i> transcription factors. Using a phylogenetic footprinting pipeline, we identify <i>cis-</i>regulatory motif enrichments among neuron class-specific gene batteries and we identify cognate transcription factors for 117 of the 118 neuron classes. In addition to predicting novel regulators of neuronal identities, our nervous system-wide analysis at single cell resolution supports the hypothesis that many transcription factors directly co-regulate the cohort of effector genes that define a neuron type, thereby corroborating the concept of so-called terminal selectors of neuronal identity. Our analysis provides a blueprint for how individual components of an entire nervous system are genetically specified.
BibTeX
@inproceedings{glenwinkel2021silico,
  author = {Lori Glenwinkel and Seth R. Taylor and Kasper Langebeck‐Jensen and Laura Pereira and Molly B. Reilly and Manasa Basavaraju and Ibnul Rafi and Eviatar Yemini and Roger Pocock and Nenad Šestan and Marc Hammarlund and David M. Miller and Oliver Hobert},
  title = {{In silico analysis of the transcriptional regulatory logic of neuronal identity specification throughout the C. elegans nervous system}},
  booktitle = {eLife},
  year = {2021},
  doi = {10.7554/elife.64906},
}
Figure