Constructive neutral evolution

Constructive neutral evolution '' is a theory that seeks to explain how complex systems can evolve through neutral transitions and spread through a population by chance fixation. Constructive neutral evolution is a competitor for both adaptationist explanations for the emergence of complex traits and hypotheses positing that a complex trait emerged as a response to a deleterious development in an organism. Constructive neutral evolution often leads to irreversible or "irremediable" complexity and produces systems which, instead of being finely adapted for performing a task, represent an excess complexity that has been described with terms such as "runaway bureaucracy" or even a "Rube Goldberg machine".
The groundworks for the concept of CNE were laid by two papers in the 1990s, although first explicitly proposed by Arlin Stoltzfus in 1999. The first proposals for the role CNE was in the evolutionary origins of complex macromolecular machines such as the spliceosome, RNA editing machinery, supernumerary ribosomal proteins, chaperones, and more. Since then and as an emerging trend of studies in molecular evolution, CNE has been applied to broader features of biology and evolutionary history including some models of eukaryogenesis, the emergence of complex interdependence in microbial communities, and de novo formation of functional elements from non-functional transcripts of junk DNA. Several approaches propose a combination of neutral and adaptive contributions in the evolutionary origins of various traits.
Many evolutionary biologists posit that CNE must be the null hypothesis when explaining the emergence of complex systems to avoid assuming that a trait arose for an adaptive benefit. A trait may have arisen neutrally, even if later co-opted for another function. This approach stresses the need for rigorous demonstrations of adaptive explanations when describing the emergence of traits. This avoids the "adaptationist fallacy" which assumes that all traits emerge because they are adaptively favoured by natural selection.

Principles

Excess capacity, presuppression, and ratcheting

Conceptually, there are two components A and B that interact with each other. A, which performs a function for the system, does not depend on its interaction with B for its functionality, and the interaction itself may have randomly arisen in an individual with the ability to disappear without an effect on the fitness of A. This present yet currently unnecessary interaction is therefore called an "excess capacity" of the system. A mutation may then occur which compromises the ability of A to perform its function independently. However, the A:B interaction that has already emerged sustains the capacity of A to perform its initial function. Therefore, the emergence of the A:B interaction "presuppresses" the deleterious nature of the mutation, making it a neutral change in the genome that is capable of spreading through the population via random genetic drift. Hence, A has gained a dependency on its interaction with B. In this case, the loss of B or the A:B interaction would have a negative effect on fitness and so purifying selection would eliminate individuals where this occurs. While each of these steps are individually reversible, a random sequence of mutations tends to further reduce the capacity of A to function independently and a random walk through the dependency space may very well result in a configuration in which a return to functional independence of A is far too unlikely to occur, making CNE a one-directional or "ratchet-like" process.

Biases on the production of variation

CNE models of systematic complexification may rely crucially on some systematic bias in the generation of variation. This is explained relative to the original set of CNE models as follows:

In the gene-scrambling and RNA pan-editing cases, and in the fragmentation of introns, the initial state of the system is unique or rare with regard to some extensive set of combinatorial possibilities that may be reached by mutation and fixation. The resulting systemic bias drives a departure from the improbable initial state to one of many alternative states. In the editing model, a deletion:insertion mutational bias plays a subsidiary role. In the gene duplication model, as well as in the explanation for loss of self-splicing and for the origin of protein dependencies in splicing, it is assumed that mutations that reduce activity or affinity or stability are much more common than those with the opposite effect. The resulting directionality consists in duplicate genes undergoing reductions in activity, and introns losing self-splicing ability, becoming dependent on available proteins as well as trans-acting intron fragments.

That is, some of the models have a component of long-term directionality that reflects biases in variation. A population-genetic effect of bias in the introduction process, which appeared as a verbal theory in the original CNE proposal, was later articulated and demonstrated formally . This kind of effect does not require neutral evolution, lending credence to the suggestion that the components of CNE models may be considered in a general theory of complexification not specifically linked to neutrality.

Subfunctionalization

A case of CNE is subfunctionalization. The concept of subfunctionalization is that one original gene gives rise to two paralogous copies of that gene, where each copy can only carry out part of the function of the original gene. First, a gene undergoes a gene duplication event. This event produces a new copy of the same gene known as a paralog. After the duplication, deleterious mutations are accrued in both copies of the gene. These mutations may compromise the capacity of the gene to produce a product that can complete the desired function, or it may result in the product fully losing one of its functions. In the first scenario, the desired function may still be carried out because the two copies of the gene together can still produce sufficient product for the job. The organism is now dependent on having two copies of this gene which are both slightly degenerated versions of their ancestor. In the second scenario, the genes may undergo mutations where they lose complementary functions. That is to say, one protein may lose only one of its two functions whereas the other protein only loses the other of its two functions. In this case, the two genes now only carry out the individual subfunctions of the original gene, and the organism is dependent on having each gene to carry out each individual subfunction.
Paralogues that functionally interact to maintain the ancestral function can be termed "paralogous heteromers". One high-throughput study confirmed that the rise of such interactions between paralogous proteins as one possible long-term fate of paralogues was frequent in yeast, and the same study further found that paralogous heteromers accounted for eukaryotic protein-protein interaction networks. One specific mechanism for the evolution of paralogous heteromers is by the duplication of an ancestral protein interacting with other copies of itself. To inspect the role of this process in the origins of paralogous heteromers, it was found that ohnologs that form paralogous heteromers in Saccharomyces cerevisiae are more likely to have homomeric orthologues than ohnologs in Schizosaccharomyces pombe. Similar patterns were found in the PPI networks of humans and the model plant Arabidopsis thaliana.

Examples of CNE

Identification and testability

To positively identify features as having evolved through CNE, several approaches are possible. The basic notion of CNE is that features which have evolved through CNE are complex ones but do not provide an advantage in fitness over their simpler ancestors. That is to say, an unnecessary complexification has occurred. In some cases, phylogeny can be used to inspect ancestral versions of systems and to see if those ancestral versions were simpler and, if they were, if the rise in complexity came with an advantage in fitness. While it is not straight forward to identify how adaptive the emergence of a complex feature was, some methods are available. If the more complex system has the same downstream effects in its biochemical pathway as the ancestral and simpler system, this suggests that the complexification did not carry with it any increase in fitness. This approach is simpler when analyzing complex traits of which evolved more recently and are taxonomically restricted in a few lineages because "derived features can be more easily compared to their sisters and inferred ancestors". The 'gold standard' approach for identifying cases of CNE involves direct experimentation, where ancestral versions of genes and systems are reconstructed and their properties directly identified. The first example of this involved analysis of components of a V-ATPase proton pump in fungal lineages.

RNA editing

systems have patchy phylogenetic distributions, indicating that they are derived traits. RNA editing is required when a genome needs to have its mRNA edited through various substitutions, deletions, and insertions prior to translation. Guide RNA molecules derived from separate semicircular strands of DNA provide the correct sequence for the RNA editing complex to make the corresponding edits. The RNA editing complex in Kinetoplastida can comprise over 70 proteins in some taxonomically restricted lineages, and mediate thousands of edits. Another taxonomically restricted case of a different form of RNA editing system is found in land plants. In kinetoplastids, RNA editing involves the addition of thousands of nucleotides and deletion of several hundreds. However, the necessity of this highly complex system is questionable. The large majority of organisms do not rely on RNA editing systems, and in the ones that do have it, the need for it is unclear as the optimal solution would be for the DNA sequence to not contain the wrong nucleotides at several thousand sites to begin with. Furthermore, it is difficult to argue that the RNA editing system emerged only in response and to correct a genome faulty to this degree, as such a genome would have been highly deleterious to the host and eliminated through purifying selection to begin with. However, a scenario where a primitive RNA editing system gratuitously arose prior to the introduction of errors into the genome is more parsimonious. Once the RNA editing system arose, the original mitochondrial genome would be able to tolerate previously deleterious substitutions, deletions, and additions without an effect on fitness. Once a sufficient number of these deleterious mutations took place, the organism would by this point have developed a dependency on the RNA editing system to faithfully correct any inaccurate sequences.