RNA base-pairing rules determine adaptive evolution success

Researchers at the University of Zurich and the Barcelona Collaboratorium for Modelling and Predictive Biology have demonstrated that the fundamental biochemical rules governing RNA base-pairing significantly influence the success of adaptive evolution. In a study published in PLOS Computational Biology, Leander Goldbach, Nora S. Martin, and Andreas Wagner computationally analyzed how ten distinct RNA base-pairing rules affect the topography of fitness landscapes. The team found that the specific rules determining which RNA bases can form hydrogen bonds strongly dictate the likelihood of an evolving population reaching the global peak of a fitness landscape. This work provides a mechanistic understanding of how molecular interaction constraints shape evolutionary outcomes.

The study focuses on the relationship between genotype and phenotype in RNA molecules, a relationship that is central to understanding the evolution of biomolecules. The topography of fitness landscapes, which represents the distribution of fitness values across all possible genotypes, is determined by how genotypic changes affect phenotypes such as RNA secondary structure. By simulating adaptive evolution on fitness landscapes where an RNA’s fitness is defined by its secondary structure, the researchers were able to isolate the impact of base-pairing rules on evolutionary dynamics. The canonical base-pairing rule found in most biological RNA molecules, which allows A-U, G-C, and G-U pairs, was identified as one of the most favorable rules for adaptive evolution.

To conduct this analysis, the team developed a sequence-free method to predict the minimum free energy secondary structure for RNA molecules of length 12. This approach was necessary because standard prediction tools like ViennaRNA rely on sequence-specific energy parameters that are unavailable for non-canonical RNA alphabets. The researchers validated their method by comparing its predictions for canonical RNA with those from the ViennaRNA package, finding that the sequence-free method accurately reproduced global genotype-phenotype map properties such as phenotype bias and neutral network sizes. This validation confirmed that the method could be reliably applied to the nine non-canonical base-pairing rules studied in the paper.

The ten base-pairing rules were modeled as simple graphs where nodes represent the four RNA bases and edges indicate possible hydrogen bonds. The researchers identified two elementary properties of these rules that explain their varying effects on evolution: promiscuity and degeneracy. Promiscuity is defined as the number of allowed base-pairings as a fraction of all six possible pairings between four nucleotides. Degeneracy is the fraction of nucleotide pairs that share an identical set of base-pairing partners. These properties influence the structure of the genotype-phenotype map, which in turn affects the navigability of the fitness landscape for evolving populations.

The study revealed that the canonical base-pairing rule is among the top three most favorable rules for adaptive evolution, with over 80% of evolving populations finding the landscape’s highest fitness peak. Base-pairing rules 3 and 5 also exhibited high navigability, with similar success rates. In contrast, the four least favorable rules, specifically maps 6, 7, 9, and 10, resulted in fitness landscapes where evolving populations were significantly less likely to reach the global peak. This variation in navigability is driven by differences in the frequency of global and local peaks within the genotype space.

Phenotype bias, which refers to the uneven distribution of phenotypes in genotype space, was found to increase with the promiscuity of the base-pairing rule. More promiscuous rules allow for a larger number of permissible secondary structures, which leads to a stronger phenotype bias. This bias means that some phenotypes are encoded by many more genotypes than others, making rare phenotypes difficult for Darwinian evolution to discover. The researchers noted that while phenotype bias is pervasive across all ten maps, it has a particularly strong impact on the variation in navigability for the most promiscuous rules.

Neutral networks, which are sets of genotypes with the same phenotype connected by single nucleotide changes, play a crucial role in facilitating evolution by allowing populations to explore genotype space without losing fitness. The study found that the size distribution of neutral networks varies widely between the ten genotype-phenotype maps, falling into three broad categories. However, the size distribution alone does not fully explain the differences in fitness landscape navigability. Instead, the diversity of phenotypes in the immediate neighborhood of a neutral network is a more significant factor in determining whether a network becomes a local peak.

Local peaks are fitness plateaus surrounded by genotypes of lower fitness, which can trap evolving populations and prevent them from reaching the global peak. The researchers found that a higher fraction of genotypes contained in local peaks strongly and negatively correlates with the average navigability of fitness landscapes. This effect is driven by low phenotypic diversity in the neighborhood of neutral networks. When the 1-mutant neighborhood of a neutral network contains few different phenotypes, the probability that the network’s phenotype is the fittest among them increases, thereby increasing the chance that the network constitutes a local peak.

Base-pairing rule degeneracy was identified as a key factor in reducing neighborhood diversity. In degenerate rules, different nucleotides may have identical sets of base-pairing partners, which means that genotypes differing at a single site may still map to the same phenotype. This reduces the phenotypic diversity in the neighborhood of neutral networks, thereby increasing the prevalence of local peaks. The study showed that degenerate base-pairing rules consistently reduce fitness landscape navigability by increasing the fraction of genotypes trapped in local peaks.

The findings suggest that the canonical RNA base-pairing rule is well-suited for adaptive evolution not because it is uniquely optimal, but because it balances promiscuity and degeneracy effectively. Rules with low-to-moderate promiscuity and degeneracy entail higher and less variable fitness landscape navigability.

The study has several limitations, including its restriction to short RNA sequences of length 12 due to computational constraints. The analysis also assumes that the bases of all ten RNA alphabets differ only in their hydrogen-bonding preferences, ignoring potential biochemical differences that could affect RNA folding. Furthermore, the study focuses on whether populations reach a global peak rather than the time required to do so, which could be influenced by the size of neutral networks. Despite these limitations, the research provides valuable insights into how simple molecular interaction rules can have profound effects on the potential for adaptive evolution.

This work contributes to a broader understanding of the constraints that shape molecular evolution. By demonstrating that the interaction of nucleotides within an alphabet can affect adaptive evolution, the study highlights the importance of fundamental biochemical rules in determining evolutionary outcomes. The findings may have implications for the directed evolution of biotechnologically relevant molecules, where the choice of molecular alphabet and interaction rules can influence the efficiency of evolutionary processes.

The computational simulations were conducted under the strong selection weak mutation regime, a framework appropriate for populations with modest sizes and low per-nucleotide mutation rates. In this model, evolution proceeds as an adaptive walk where a monomorphic population acquires single point mutations that become fixed before new mutations arise. The fixation probability of each mutation was calculated using Kimura’s formula, assuming a population size of 10^6. Each adaptive walk began from a randomly chosen genotype and continued until either 1,000 fixed mutations occurred or the population reached the global fitness peak. This approach allowed the researchers to quantify the fraction of successful walks, termed navigability, across 200 random fitness landscapes for each of the ten genotype-phenotype maps.

Validation of the sequence-free prediction method revealed that it closely mirrors the established ViennaRNA package for canonical RNA. The two methods produced highly similar phenotype frequency distributions, with the 27 most frequent phenotypes showing nearly identical frequencies. Discrepancies appeared only for low-frequency phenotypes, which collectively represented fewer than 0.07% of all genotypes. Similarly, the size distributions of the 210 largest neutral networks were comparable between the two maps. These findings confirm that the sequence-free method accurately reproduces the global properties of the genotype-phenotype map, such as phenotype bias and neutral network sizes, despite lacking sequence-specific energy parameters.

The study also examined the role of neutral network neighborhood diversity in shaping evolutionary outcomes. In most genotype-phenotype maps, the average neutral network had more than 22 different phenotypes in its 1-mutant neighborhood. However, maps 6, 7, and 9 exhibited significantly lower diversity, with only 10 to 14 distinct phenotypes in these neighborhoods. This reduced diversity increases the probability that a neutral network’s phenotype is the fittest among its neighbors, thereby raising the likelihood that the network becomes a local peak. For maps 6 and 7, this effect was compounded by the presence of larger neutral networks, which contributed more genotypes to the local peaks compared to map 9, where many networks were small.

The authors note that their findings align with previous research on alternative genetic codes, which found that the standard genetic code is not uniquely robust but is one of several codes that support evolvability. Similarly, the canonical RNA base-pairing rule is not the only rule that facilitates adaptive evolution; rules 3 and 5 also supported high navigability. The study highlights that low-to-moderate promiscuity and degeneracy are key factors in ensuring that evolving populations can effectively explore genotype space and reach high-fitness phenotypes.

Subject of Research: The promiscuity and degeneracy of ten unique RNA base-pairing rules can determine the success of adaptive RNA evolution

Article Title: The promiscuity and degeneracy of ten unique RNA base-pairing rules can determine the success of adaptive RNA evolution

Article References: Goldbach, L., Martin, N. S., & Wagner, A. (2026). The promiscuity and degeneracy of ten unique RNA base-pairing rules can determine the success of adaptive RNA evolution. PLOS Computational Biology, 22(11), e1014914. https://doi.org/10.1371/journal.pcbi.1014914

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014914

Keywords: promiscuity, degeneracy, unique, base-pairing, rules, determine, success, adaptive, evolution, scientific research