Evolutionary Constraints from Regulatory Gene Networks

Evolutionary biology is full of great stories about why life on Earth is the way it is, but one of the field's ultimate goals is to actually understand the process itself — that is, how the evolutionary process works, and its potential and limitations — rather than simply explaining the history of life on this planet. Evolutionary stories are an important and useful way to approach that question, but they're not the only tool at our disposal.

Computational modelling provides researchers with the tools to consider familiar questions from an unfamiliar perspective and to handle questions which might otherwise be out of reach. Essentially, the system you want to study has to be expressed in mathematical form, and then computers are used to explore how the system behaves under different configurations, usually by running simulations.

It's an approach that offers several advantages. Building a mathematical representation forces you to make your reasoning precise and your assumptions explicit; there isn't the wiggle-room for the kind of hand-waving that easily slips into verbal models. It's also an excellent opportunity to try to simplify your conception of the system and include just the bare essentials. Since the mathematical model only has the elements you put into it, this is a great way to figure out which elements of your mental model are actually necessary for things to work. Finally, running simluations enables you to set up experiments and make observations that would be hard to actually carry out, some of which might offer important new insights.

That's precisely what Alba Jiménez, Andreea Munteanu, and James Sharpe did in a recent paper in PNAS. Their aim was to understand how the mechanisms of a gene network affect the evolutionary options available to it. In response to a chemical gradient, a network of three genes can generate an output pattern where one of the genes is on in the middle of the gradient (but off when the chemical is high or low) — a "middle stripe" pattern. An earlier study by James Sharpe and James Cotterell showed that there are six different ways to connect the three genes in order to produce such a pattern; in other words, there are six different regulatory mechanisms that generate the same output. Now the team wanted to understand whether these networks differed in terms of their ability to evolve.

To find out, they simulated the different networks and systematically mutated all the interactions. This showed them the output patterns that were accessible via mutation from each network configuration and how frequent they were. For example, the mutated network might produce a pattern where the gene is on whenever the chemical is below a certain threshold (a "righthand stripe" pattern), or above a threshold ("lefthand stripe"), or even a pattern with multiple stripes. By examining which new patterns were available by mutating each network and how frequently they occurred, the team measured the 'evolvability' of each network.

In many cases, the mutations didn't change the network's output. The interesting result is that the networks differed in terms of how many mutations resulted in a different pattern. For example, in one network 10% of the mutations changed the output, while a network with different connections only produced a different output pattern for 1% of the mutations. The networks also differed in terms of which new patterns were easily accessible -- in some cases left- or right-hand stripes might be the most common change a mutation could cause, while in other cases multiple stripes were more common.

This is a pretty exciting insight about the interaction between evolution and development, and it's just the kind of question that's readily accessible to computational modelling but quite challenging to address experimentally. All six gene networks configurations produce the same output pattern, so there isn't an inherent reason why an organism might evolve to use one or another during its development. In fact, the paper cites examples of different organisms using the different configurations. While their output is similar, the network used by a particular developmental process ends up constraining its evolutionary options, helping shape its evolutionary course. This is hardly the first time I've discussed evolutionary constraints on Accumulating Glitches; in this case, though, the constraints don't arise from developmental trade-offs or complexity, but from the dynamics of the regulatory mechanism. Evolution and development are both intricate, complex processes, so it's hardly surprising that their interaction leads to constraints. However, it's important to recognize these constraints, and even more important to understand their basis if we want to know how the evolutionary process works, rather than simply describe how evolution happened on Earth.

Ref
Jiménez, A, Munteanu, A, and Sharpe J. Dynamics of gene circuits shapes evolvability. PNAS early edition. (2015) doi:10.1073/pnas.1411065112