The GRN concept as a guide for evolutionary developmental biology

Abstract

Organismal phenotypes result largely from inherited developmental programs, usually executed during embryonic and juvenile life stages. These programs are not blank slates onto which natural selection can draw arbitrary forms. Rather, the mechanisms of development play an integral role in shaping phenotypic diversity and help determine the evolutionary trajectories of species. Modern evolutionary biology must, therefore, account for these mechanisms in both theory and in practice. The gene regulatory network (GRN) concept represents a potent tool for achieving this goal whose utility has grown in tandem with advances in "omic" technologies and experimental techniques. However, while the GRN concept is widely utilized, it is often less clear what practical implications it has for conducting research in evolutionary developmental biology. In this Perspective, we attempt to provide clarity by discussing how experiments and projects can be designed in light of the GRN concept. We first map familiar biological notions onto the more abstract components of GRN models. We then review how diverse functional genomic approaches can be directed toward the goal of constructing such models and discuss current methods for functionally testing evolutionary hypotheses that arise from them. Finally, we show how the major steps of GRN model construction and experimental validation suggest generalizable workflows that can serve as a scaffold for project design. Taken together, the practical implications that we draw from the GRN concept provide a set of guideposts for studies aiming at unraveling the molecular basis of phenotypic diversity.

Keywords: CRISPR; cis-regulatory element; evolutionary developmental biology; gene expression; gene regulatory networks.

FIGURE 1

Gene regulatory network (GRN) models. In a GRN model, network nodes represent genes or gene products, and edges represent regulatory interactions between them. These interactions can drive upregulation (pointed arrow heads) or downregulation (flat arrow heads). In a biological network, a single edge typically encodes two types of regulatory information: inputs to and outputs from an underlying cis-regulatory element. Inputs come in the form of transcription factor binding from an upstream “source” node, while outputs are represented by cis-regulatory activity on a downstream “target” node (i.e., enhancer looping and its impact on transcription)

FIGURE 2

Transcriptomics and GRN inference. DGE analyses can be used to compare expression (a) between tissues within an organism as the light and dark dorsal stripes of the African striped mouse; (b) within a tissue under different treatment regimes; and (c) across timepoints during a developmental process. Alternatives to DGE that better-accommodate longitudinal data without discrete replicate groups include (d) segmented regression or (e) correlation network analyses, such as WGCNA, which directly predicts networks from transcriptomes. Here, a heatmap shows module eigengene activity per sample, with samples sorted by developmental stage. Thus, the highlighted network module represents a group of highly-correlated genes whose expression is highest early in development. DGE, differential gene expression; GRN, gene regulatory network; vs, versus; WGCNA, weighted gene coexpression network analysis

FIGURE 3

From GRN construction to project design. A hypothetical example of an EvoDevo research project applying diverse experimental approaches, organized in terms of the major steps of GRN model construction. In this example, a researcher aims to unravel the molecular basis of variable tail fin morphology between related species of killifish. They first perform RNA-Seq tissue samples spanning tail fin development in their principal model species and use it to infer correlation networks with WGCNA. A module correlated with developmental progression is identified and flagged for further refinement. The researcher then uses ATAC-Seq to identify candidate cis-regulatory elements near genes in the module and applies transcription factor footprinting together with Micro-C to confirm direct regulatory interactions and to prune away edges unsupported by empirical evidence, disconnecting some genes from the network in the process. Because genomes are available for relevant killifish species, they use comparative genomics to identify enhancers under positive selection in a species with a derived tail fin morphology. Several such enhancers contain transcription factor motifs bearing nucleotide changes predicted to impact edges in the network model. The researcher then uses CRISPR-Cas9 to alter this binding site in their principal model species and characterizes resulting phenotypes. GRN, gene regulatory network; TF, transcription factor; WGCNA, weighted gene coexpression network analysis