On the conservative nature of intragenic recombination

Abstract

Intragenic recombination rapidly creates protein sequence diversity compared with random mutation, but little is known about the relative effects of recombination and mutation on protein function. Here, we compare recombination of the distantly related beta-lactamases PSE-4 and TEM-1 to mutation of PSE-4. We show that, among beta-lactamase variants containing the same number of amino acid substitutions, variants created by recombination retain function with a significantly higher probability than those generated by random mutagenesis. We present a simple model that accurately captures the differing effects of mutation and recombination in real and simulated proteins with only four parameters: (i) the amino acid sequence distance between parents, (ii) the number of substitutions, (iii) the average probability that random substitutions will preserve function, and (iv) the average probability that substitutions generated by recombination will preserve function. Our results expose a fundamental functional enrichment in regions of protein sequence space accessible by recombination and provide a framework for evaluating whether the relative rates of mutation and recombination observed in nature reflect the underlying imbalance in their effects on protein function.

Fig. 1.

Effects of recombination and mutation on lactamase function. (A) Recombination results in a higher fraction of functional lactamase variants than mutation. The (minimum) fractions of functional chimeras (▪) in each bin of substitution levels m are shown relative to PSE-4 (m = 0) and TEM-1 (m = 150) (see Methods). Eq. 3 using the best-fit value ρ = 0.79 ± 0.02 (dashed line) agrees well with these data. Mutation produces a lower fraction of functional variants (Eq. 2 with a best-fit value of ν, solid line) than recombination at all values of m. (B) Error-prone PCR mutagenesis of PSE-4 results in exponentially declining retention of lactamase function with increasing substitutions. The fractions of functional PSE-4 random mutants in each of four libraries and a no-mutation control (▪) are plotted against each library's average nucleotide mutation level 〈m_nt〉± SE. The exponential best-fit of the random mutation data to Eq. 1 yields ν = 0.54 ± 0.03 (solid line).

Fig. 2.

Lattice protein results mirror experimental findings. Shown are average fractions of functional chimeras over 50 replicates using parents sharing 20–80% sequence identity (D = 20, 15, 10, or 5) for a high-ν structure, #1080 (A), and a low-ν structure, #873 (B) (see Supporting Text). Independent fits for ρ and ν are plotted. (Inset) Mutation data for each structure collected from homologs used to construct A and B. Curves show four independent best fits to Eqs. 2 and 3 (see Methods); error bars are ±1 SE.

Fig. 3.

Neutrality ν is correlated with recombinational tolerance ρ for lattice proteins. Results are from 10 different structures. Error bars show SD of averages of ν and ρ taken at four values of sequence identity (20%, 40%, 60%, and 80%, as in Fig. 2). For the two lowest-neutrality structures, error bars reflect two and three sequence identities, respectively, because no highly diverged homologs were found.

Fig. 4.

Chimeras occupy a functionally enriched ridge in sequence space. Surface height, the product of Eqs. 2 and 3 (see text), represents the probability of retaining parental function given independent random and homologous substitutions. Mutants lie along the near and far edges (slope determined by ν), chimeras lie on the ridge (slope determined by ρ), and mutated chimeras lie on the hillsides.