Mutational Robustness of Ribosomal Protein Genes

See allHide authors and affiliations

Science  05 Nov 2010:
Vol. 330, Issue 6005, pp. 825-827
DOI: 10.1126/science.1194617

Keeping Fit

Mutations may be deleterious, neutral, or advantageous. Understanding the relative effect of a new mutation on an organism's fitness is important for many systems from complex diseases to conservation biology. Lind et al. (p. 825) used the bacterium Salmonella typhimurium to study the effects of random mutations in two ribosomal proteins on fitness. Most mutations, whether synonymous or nonsynonymous, had significant fitness costs, thus overturning the prevailing dogma that most point mutations are either neutral or lethal and indicating that the mutations influenced messenger RNA structure and/or stability.


The distribution of fitness effects (DFE) of mutations is of fundamental importance for understanding evolutionary dynamics and complex diseases and for conserving threatened species. DFEs estimated from DNA sequences have rarely been subject to direct experimental tests. We used a bacterial system in which the fitness effects of a large number of defined single mutations in two ribosomal proteins were measured with high sensitivity. The obtained DFE appears to be unimodal, where most mutations (120 out of 126) are weakly deleterious and the remaining ones are potentially neutral. The DFEs for synonymous and nonsynonymous substitutions are similar, suggesting that in some genes, strong fitness constraints are present at the level of the messenger RNA.

The distribution of fitness effects (DFE) of mutations is important in evolutionary biology; for example, in the the degradation of genetic information due to Muller’s ratchet (1), molecular clocks (2), genetic variation at the molecular level (3), and the impact of selection and genetic drift in natural populations (4). The DFE is also of importance for understanding quantitative traits, complex diseases, the evolution of antibiotic resistance, and predicting the minimal populations sizes needed for maintaining healthy populations of endangered species and in breeding programs (58). Mutations may be deleterious, neutral, or beneficial, and the proportion of mutations in each category will depend on several factors (9). Advantageous mutations are rare and appear to be exponentially distributed (6, 1012). The DFE of deleterious mutations often appears to be bimodal, with one low-fitness peak (including lethal mutations) and a second peak close to wild-type fitness, with weakly deleterious mutations (13). Information on the DFE for deleterious mutations comes mostly from analyses of DNA sequence data (2, 5, 14) or mutation accumulation and mutagenesis experiments (1518). The indirect estimates of fitness values made from sequence data suffer the drawback that strongly deleterious mutations (with |s|Ne >>1) (|s|, absolute value of selection coefficient; Ne, effective population size) are poorly represented, and many details of the DFE are unresolved (19). Direct measurements, on the other hand, are difficult at the high-fitness end, where deleterious effects are smaller than the percent level (in this study, selection coefficients |s| < 3 × 10−3 could not be detected). Nevertheless, experimental studies of the fitness effects of defined single-point mutations have proven useful, because they allow an assumption-free determination of the underlying DFE, including the frequencies of strongly deleterious mutations. A few such studies in viruses have demonstrated a bimodal DFE, with most mutations being either neutral or lethal (13, 20, 21).

We used Salmonella typhimurium to study the DFE of random base-pair substitutions in the protein synthesis machinery. A total of 126 random base-pair substitutions were engineered into the rpsT and rplA genes, encoding the ribosomal proteins S20 and L1, respectively (22). These two proteins are nonessential, but deletion mutants lacking either of these ribosomal proteins have severely reduced fitness. Thus, putative mutational effects on fitness can be measured over a large range, and the fitness of complete loss-of-function mutations is known and is larger than zero. We used bacterial growth rate to measure the fitness effects of the mutations. The involvement of ribosomal proteins in translation and the direct relation of translation rates to exponential growth rates (23) ensure that fitness effects will be directly correlated to the quality and quantity of available ribosomes.

In total, 70 single base-pair mutants of rpsT (S20), 56 mutants of rplA (L1), and four independent wild-type controls for each gene were constructed. The 126 random mutations comprised 38 synonymous base-pair substitutions and 88 nonsynonymous substitutions. The identity and fitness data of the individual mutants are available as supporting online material (SOM) (tables S1 to S3). Exponential growth rates were reduced for most of the substitution mutants, with an average relative growth rate of 0.92 (range, 0.61 to 1.01) for synonymous and 0.94 (0.72 to 1.00) for nonsynonymous mutants in rich growth medium (LB), as compared to wild-type controls (relative growth rate set to 1) and a median of 0.96 and 0.95, respectively (Fig. 1, A and B, and figs. S1 and S2). In the poorer M9 glucose minimal medium, the average growth rates were 0.95 (median 0.96) for both synonymous (0.89 to 1.01) and nonsynonymous (0.72 to 1.00) substitutions (Fig. 1, C and D, and figs. S3 and S4). As expected, no significantly advantageous mutations were found. No mutations caused a complete loss of function, and the mutants with the most pronounced fitness reduction (0.61 and 0.74) were still much more fit than the deletion mutants, which showed a relative fitness of 0.29 (rpsT, S20) and 0.25 (rplA, L1) in LB (24).

Fig. 1

Relative exponential growth rates of single-substitution mutants relative to wild-type controls set to 1. (A) Synonymous substitutions (in LB). (B) Nonsynonymous substitutions (in LB). (C) Synonymous substitutions (in M9 medium). (D) Nonsynonymous substitutions (in M9 medium).

The growth rate measurements have limited sensitivity, detecting changes in s > 0.03 (3%), and measure fitness only in the exponential growth phase. To increase sensitivity, we performed competition experiments (24), in which we measured a composite fitness during the entire growth curve. In these competitions, sensitivity is much improved, allowing the detection of fitness differences as small as s = 0.003 (0.3%) (24). The average selection coefficients in this experimental setup were –0.0096 (–0.0279 to 0, Fig. 2A) for synonymous and –0.0131 (–0.0763 to 0, Fig. 2B) for nonsynonymous substitutions. The large majority of the mutations, including synonymous substitutions, caused significant fitness costs that confer strong counterselection (25). Mutations were evenly distributed throughout the genes, and no correlation between gene position and fitness was found (Fig. 3, A and B). The DFEs (for –s) of the mutations were estimated by fitting of commonly used univariate distributions (13) to the data using a maximum-likelihood method (Fig. 2, C and D, figs. S7 to S14, table S5, and SOM text).

Fig. 2

Selection coefficients of single-substitution mutants relative to isogenic wild-type controls (defined as s = 0, SD = 0.001 within experiment variation for competition controls) during growth in M9 medium. (A) Synonymous substitutions. (B) Nonsynonymous substitutions. (C) Cumulative distribution function of fitness costs (–s) for synonymous substitutions, with x’s showing the experimental data and the line showing a fitted gamma function. (D) Cumulative distribution function of fitness costs (–s) for nonsynonymous substitutions, with x’s showing the experimental data and the line showing a fitted gamma function.

Fig. 3

Analyses of fitness costs. (A) Position of substitutions in rpsT (S20) and fitness costs. (B) Position of substitutions in rplA (L1) and fitness costs. (C) Influence of changes in predicted protein free energy (DDG is destabilizing) on fitness costs of nonsynonymous substitutions. (D) Evolutionary conservation versus fitness costs for nonsynonymous mutations. (E) Logarithmic (ln) change in the frequency of codons used in ribosomal protein (RP) genes for synonymous substitutions. (F) Absolute changes in predicted mRNA free energy versus fitness costs for synonymous substitutions.

To explain the similarities in DFEs for synonymous and nonsynonymous changes, we assume that the mechanistic cause(s) of the fitness effects are the same, especially if the effects of most amino acid changes are generally very small but occasionally large, accounting for the outliers in the distribution of nonsynonymous substitutions. This idea is supported by the absence of any significant correlation between fitness and predicted changes in protein free energy (Fig. 3C), conservation (Fig. 3D), and ribosomal RNA contacts (table S4) for the whole set of nonsynonymous mutations. Instead, our data suggest that the fitness reduction produced by base-pair substitutions is primarily conferred by changes in mRNA, which could be manifested at several levels, including, for example, suboptimal codon usage or altered mRNA structure and/or stability. The former is unlikely for several reasons. First, even though several of the synonymous mutations with large fitness costs involve more rarely used codons, this is true for only a limited subset of mutations (Fig. 3E and table S4). Second, we could not find any overall correlation between relative codon usage and fitness effects, using either all S. typhimurium codons or only those used in ribosomal protein genes (Fig. 3E and table S4). This is also consistent with our previous experimental work in which substitution of the S. typhimurium rpsT and rplA genes with genes from other species with different codon usage caused very small fitness effects (24).

These small fitness costs suggest that the fitness constraints on the mRNA for the two ribosomal protein genes are highly conserved between related bacterial species and that this functional conservation is largely independent of codon usage. Selection coefficients determined from the competition experiments were plotted as a function of the absolute values of the predicted free energy change for the mRNA of the synonymous mutants. A weak but significant correlation [correlation coefficient (r) = 0.47, P = 0.0027, n = 38 synonymous mutants] was found, indicating a general connection between changed mRNA structure and fitness (Fig. 3F and table S5). However, no significant changes in mRNA levels could be detected by quantitative real-time fluorescence polymerase chain reaction for synonymous mutants with large fitness costs (SOM text). Studies of synonymous substitutions usually involve large changes in codon usage or particular examples of substitutions with large effects. Mutagenesis studies of single proteins rarely include the use of high-sensitivity assays of fitness and analysis of synonymous substitutions (SOM references).

Studies of fitness effects of defined base substitutions in viruses have focused on the DFE at the whole-genome level, whereas we studied two specific bacterial genes. However, the viruses examined are small and encode only 5 to 11 genes, meaning that there are many independently engineered mutations for each virus gene and that the DFE can also be studied at the level of the individual genes (13, 20, 21). Comparing the shape of the distributions obtained here with those from similar experiments in viruses reveals two differences that are valid both when the viral DFEs are analyzed at the level of the whole genome and of individual genes. First, for viruses, the most frequently found mutational type was lethal (up to 40%) (13, 20, 21), whereas most of the mutations examined here had only small effects on fitness (91% had s values between –0.003 and –0.03). Thus, the compact virus genomes appear to be highly constrained with regard to which sequence changes are acceptable for phage viability (13).

The second difference is the rarity of apparently neutral mutations found here as compared to the viruses examined (13, 20, 21). For the ribosomal protein genes, 6 of 126 mutants (4.8%) had |s| < 0.003, whereas for the viruses, 25% appeared neutral. One reason for this difference could be that the higher sensitivity of our fitness assays allows mutations with small fitness effects to be distinguished from neutral mutations and that a similar peak of weakly deleterious mutations might also exist in the viral systems. Thus, it is conceivable that the relatively high frequencies of apparently neutral mutations observed in certain experimental systems (13, 20, 21) are mainly a consequence of the limited sensitivity of the assays and that the proportion of deleterious mutations is very high even when synonymous substitutions are included.

Supporting Online Material

Materials and Methods

Figs. S1 to S14

Tables S1 to S7


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. This work was supported by grants from the Swedish Research Council to D.I.A. and O.G.B. We thank D. Hughes and P. B. Rainey for comments on the manuscript.
View Abstract

Navigate This Article