What a Gene Regulatory Network Is — and What It Is Not !

Pedro Martinez. Departament de Genètica, Microbiologia i Estadística, Universitat de Barcelona

Gene Regulatory Networks: An Introduction and Their Historical and Conceptual Context.

The concept of a gene regulatory network has, by now, a long history. It was essentially developed in the papers that Britten and Davidson published in 1969 and later. In these works, they proposed the idea that genes—particularly transcription factors—constitute the underlying mechanisms that causally explain development and evolution. This paradigm shifted earlier views of developmental processes as “linear” epistatic relationships, which had incorporated concepts such as “master regulatory genes” positioned at the top of epistatic hierarchies. Although ideas like pleiotropy had long been recognized, implying that linear sequences of gene action were unrealistic representations of how genes build structures, the notion that development could still be depicted as a set of linear processes in which transcription factors act on one another remained attractive and was, in some sense, reinforced by classical developmental genetics.

The advent of genomic technologies and the ability to study interactions between numerous transcription factors and cis-regulatory sequences enabled a more complex view of development. In this framework, many factors act through mutual interactions, in hierarchical architectures, driving processes forward in a largely directional manner. Gene regulatory networks (GRNs, from now on) were instrumental in giving mechanistic form to older concepts such as developmental trajectories in Waddingtonian landscapes, where developmental decisions were visualized as the movement of a system downstream along a valley shaped by both external physical constraints and internal constraints built into the architecture of the network. Moreover, GRNs offered hints about what “emergent properties” may depend on—phenomena in which novel, nonlinear properties arise at different biological scales, even though the proposed underlying explanations were never entirely clear.

However, as with many concepts introduced in science, the use of the term gene regulatory network has gradually lost precision. In a substantial fraction of articles, it is employed in a very loose way. This trend has run parallel to the pressure to use “fashionable” terminology in papers, grant proposals, reports, and other scientific outputs—a problem that is not foreign to many disciplines.

Gene Regulatory Networks: Clarifying What Counts—and What Does Not

A Gene Regulatory Network is a graphical representation of the physical interactions between transcription factors (TFs) and their target cis-regulatory regions that drive a specific developmental process. The representation aims at completeness: it should indicate the full set of transcription factors involved, as well as all the cis-regulatory elements to which they bind (often located in other transcription factors). Needless to say, TFs also regulate other classes of genes, including those coding for signalling molecules and structural proteins. These gene products are, of course, essential for establishing intercellular communication and for generating the specific phenotypes of cells and tissues in which they are expressed. However, TF networks are the most relevant component because they constitute the driving engine of the process. TFs are the true, physical effectors that modulate new gene activities and are ultimately responsible for pushing development forward.

While all of the above may seem almost self-evident, what is striking in many papers is that the links between TFs and their regulatory binding sites are often hardly demonstrated. There is no such thing as a network of factors if the interactions are not experimentally established; otherwise, the implied causality collapses. Once again, we end up with mere correlations between TF activities, without any explanation of how the interactions drive specific parts of the developmental process. A network cannot be based on the assumption that correlations between expression patterns—whether in natural conditions or after gene perturbations—and the presence of putative open chromatin sites are sufficient, even if all these data appear together in correlated datasets (e.g., ATAC-seq profiles or transcriptomic datasets). TF binding relationships must be demonstrated. Moreover, the binding events must not only occur but also be functionally significant. For instance, binding sites may be occupied without being functional.

In this context, only those representations whose nodes and edges have been experimentally verified can truly be called networks. Moreover, we should be cautious about calling a small set of genes a network. There is an obvious impossibility in representing a bona fide network with a very limited number of components. A good network should aspire to completeness, and this requires a substantial number of experimentally demonstrated interconnections.

Alternatively, what we often have is merely a collection of correlated data, which may generate preliminary hypotheses. These hypotheses must then be rigorously tested if they are to become genuine GRNs with any degree of predictive power.

It is the completeness of a GRN that provides a genuinely mechanistic understanding of a developmental process—not merely the description of the actors involved, even when some interactions can be demonstrated. By definition, causal explanatory frameworks (Johansson et al., 2024) should always include:

1. Linking antecedent causes to subsequent effects;
2. Describing the pathways from cause to effect (not just establishing a correlation);
3. A clear temporal sequence, in which the cause must precede the effect; and
4. An explanation of the “why” of a process, offering reasons for the observed events or states.

In this context, only GRNs reveal the causal and mechanistic relationships needed for a veritable (and verifiable) understanding of development. This does not mean that GRNs are the only explanatory tool—physical forces and complex cell behaviors also account for complementary aspects of developmental processes. However, these should be integrated into a holistic framework in which the core explanatory level is provided by the nuclear events that contribute to build a GRN.

REFERENCES

Britten RJ, Davidson EH. Gene regulation for higher cells: a theory. Science. 1969 Jul 25;165(3891):349-57. doi: 10.1126/science.165.3891.349.

Johansson, LG. et al. (2024). Causal Explanations. In: A Primer to Causal Reasoning About a Complex World. Springer Briefs in Philosophy. Springer, Cham. https://doi.org/10.1007/978-3-031-59135-8_8

Peter, I. S., & Davidson, E. H. (2015). Genomic control process : development and evolution / Isabelle S. Peter, Eric H. Davidson. (1st ed.). Elsevier. (for a general review)

(No Ratings Yet)

Loading...