Nothing in evolution makes sense except in the light of parasites

Simon J. Hickinbotham, Susan Stepney, Paulien Hogeweg

Posted on: 9 July 2021

Preprint posted on 25 February 2021

Parasites in a RNA world model system of evolution. Come check out my new #preLight on @franticspider and collaborators' work, explaining how evolution can make sense if you have parasites in the picture. @preLights #preprint #AcademicTwitter

Selected by Tamara Sternlieb

Categories: bioinformatics, evolutionary biology

Background

Some of the most basic questions asked in the study of evolution are “what is the most simplified system we can imagine that would simulate living systems?” and “how could this simplest of systems drive the huge diversity of life we have today?” From these questions, a theory was born named “The RNA world hypothesis” ¹, in which the life units are RNA replicators. They store the necessary information that identifies the replicator in its environment, and also are able to catalyze the reaction and the replication, which mimics the capacity of RNA to catalyze cellular processes by itself. The ancestors of current living systems might have actually taken such a simpler form; and DNA and protein may be the evolutionary “latecomers”, which took over most of the functions of RNA and beyond. Adding some initial variables or constraints to the systems in which these RNA replicators are modelled, one can run a number of replication cycles and let the results “talk back” and show how complexity arises.

The authors sought to study the co-evolution of replicators and parasites, and the way complexity in replication evolves as replicators develop ways to compete with parasites’ faster replication rate.

They used the Stringmol automata chemistry (https://stringmol.york.ac.uk/) as the environment and model to study these replicators. A replicator is a short computer program, consisting of a string of symbols, sometimes with an associated functionality (opcodes), other times without (no-ops). In this model, one can define how the replication is executed, and replicators can replicate other strings, but do not auto replicate. To replicate, the strings have to bind, which leads to a reaction, meaning the execution of the opcodes program. Point mutations in the replication process are also added stochastically with a pre-set probability.

In this work Stringmol is for the first time embedded in a spatial arrangement. This limits the interaction of replicators with their immediate neighbors in a 2D space (Fig 1). Also, replication is explicit, executing one opcode at a time and taking a defined period of time to execute. This last condition allows evolution to happen directly on the replication process.

As replicators begin binding and copying each other, and point mutations begin to occur, parasite strings rapidly evolve into existence as shorter and more quickly replicated strings. A parasitic pair of strings R,P is one where R can replicate P but P cannot replicate R. As parasites begin outcompeting replicators, replicators must evolve new defense mechanisms.

Since replication in this model takes time and is executed step by step, it favors the accumulation of short strings and would inevitably lead to the extinction of longer more complex ones. This can be avoided by embedding the replicators in a spatial pattern formation, like the authors do (see Figure 1).

Depiction of the spatial arrangement for replicators in the model system and the result of the replication process. — Figure 1 (from Figure 8 of the preprint) – Spatial arrangement for replicators’ reactions in Stringmol. Occupied cells are shown in grey. Empty cells shown in white. Blue and Pink filled cells are two replicators who share a Moore neighborhood (an eight cell 2D grid around a central cell), indicated by the dashed blue and pink lines. They react by random selection, one becomes the replicator and the other the template, and after the template gets copied, the new string (green cell with hatched border) is placed in an empy cell in the Moore neighborhood of the replicator. If there are no empty cells, the product string is discarded.

Although in most cases extinction still happens in the early stages of evolution, in the systems that survive long enough, a wave alternating pattern of replicator-parasite advantage emerges, showing fascinating evolution.

Findings

They carried out 20 runs, of which 12 went extinct, and 8 remained executing for two million execution steps. In a first glance of these 8 runs, replication efficiency increased at the beginning due to shortening of the strings, which is expected. But as the runs progressed, population and variety of species increased and started to show fluctuating ratios. Execution times also became longer, contrary to what would be expected if reproduction efficiency determined productivity. This poses the question as to how these systems increase productivity without decreasing reproduction time.

To explain this, the authors zoomed in on a single run and showed the dominant reactions and spatial distribution of strings by their lengths at 6 different time points of the run (Figure 2). Through these time points they analyzed the most abundant replicator strategies and parasites that evolved in this system. One of the first modified qualities was a loss of complementary binding of strings. This allowed parasites to bind to other parasites, which lead to no replication and inhibition from binding to a replicator for a period of time. They found that replicators developed systems of self-scanning, which slow down replication, thus reducing the increased rate of replication of parasites. Another strategy involved switching the replication execution between both strings (the replicator and the template). This includes “toggle” and “move” sequences that allow strings to check if the template is identical to the replicator. This adds a requirement for replication, as parasites needed these sequences to be replicated. After all this events, the model showed that parasites were hugely reduced in number, and the defense mechanisms started to get lost in evolution due to the reduced selection pressure. Until, finally, new parasites emerged from mutations on more complex replicators.

Figure 2 – Spatial arrangement of opcodes and most abundant reactions between replicators and parasites at each time point. This figure combines Figure 3 and Figure 4 of the preprint, showing on the left side how strings, colored by their length, are spatially arranged at times t1 to t6 during the run, and on the right showing the key reactions between replicators at those time points. See key guide at the bottom left.

As the authors elegantly explain: “We see that parasites, while traditionally seen as a threat to evolving replicator systems, are instead the means by which complexity can evolve, provided that spatial pattern formation prevents global extinction.”

What I like about this preprint

For such a complex simulation, the preprint explains it very clearly, even for readers like myself who don’t have any past knowledge on automata simulations of evolution.

This work introduced me to artificial evolutionary models and the possibilities they bring to study evolution diversity and complexity. Then it showed how parasites are not only an inevitable consequence of a replicating system, but also that they are the causative agents of some of that beautiful complexity, and the reason why some systems don’t always seem to work in the most intuitively efficient way. My personal conclusion is that this shows how parasites are the cause and consequence of the plasticity of biological systems to adapt and evolve. And while they are among us, it means we are always evolving.

Questions for authors

Prephrasing the following answers the authors stated that “Whilst we have an intuitive notion of what complexity is, it is difficult to give it a precise definition that would lead to a measure of “how much” complexity there is in the system“.

Do you believe replication complexity would increase indefinitely in this system?

We have to remember that this is a model system with a fixed chance of death (and limited energy and space). The fixed chance of death imposes a selection pressure for fast replication – in order to fix in the system a replicator must have a good chance of making a copy of itself before it is destroyed. Parasitism acts against this pressure since simple fast replicators are easier to ‘fool’ into copying parasites. If complexity could be increased indefinitely without corresponding decrease in the replication rate and increasing resource requirements, then the answer is yes. If this is not possible (which is more likely in this configuration of Stringmol), then there must be level of complexity at which ‘parasitic pressure’ balances the limits in the model system. We do not know if the system has reached this limit or if the limit is ‘reachable’ by evolution within the system – further work in this area is needed.

Is there any way the complexity itself could become a disadvantage and lead to extinction by parasites which find a way to circumvent the defense mechanisms?

We may already see this in the system, but it wouldn’t tend to lead to extinction because interactions between entities are strictly local – emergent highly complex behaviours such as those in the question can arise, but if it is disadvantageous, the complex community wouldn’t increase in number much before dying out via competition for space with neighbouring communities.

In t6 time point we see that new strings are created during parasitism reactions, which are not replicators, nor parasites. Could parasites eventually create new replicators as well?

This could happen in a number of ways. Firstly, some entities in the system can be a parasite when bound to one species of molecule, but a mutual replicator when bound to another. Indeed, this led us to change our analysis to count parasitic reactions instead of parasitic molecules. Secondly, most parasites are closely related to the replicators that they exploit in this system, so successful parasites existing in large numbers could have mutations back to the replicator behaviour they are descended from. Finally (and this is probably the main point of your question) the attraction of using Stringmol to study replication is that there is nothing hard-coded into the system that prohibits the emergence of new forms of replication. We do see that the ‘copy loop’ that was hand-written into the original replicator tends to be strongly conserved. In another run (not detailed in the main paper), we saw an extra n-op inserted into the copy loop which slowed down replication rate in the same way that self-scan does in the paper. We haven’t checked whether parasites are members of that lineage but this is a strong possibility as macromutations tend to happen via cascades of reactions which may involve some parasitic stages.