Nucleic acid evolution and minimization by nonhomologous random recombination

Abstract

We have developed a simple method for exploring nucleic acid sequence space by nonhomologous random recombination (NRR) that enables DNA fragments to randomly recombine in a length-controlled manner without the need for sequence homology. We compared the results of using NRR and error-prone PCR to evolve DNA aptamers that bind streptavidin. Starting with two parental sequences of modest avidin affinity, evolution using NRR resulted in aptamers with 15- to 20-fold higher affinity than the highest-affinity aptamers evolved using error-prone PCR, and 27- or 46-fold higher affinities than parental sequences derived using systematic evolution of ligands by exponential enrichment (SELEX). NRR also facilitates the identification of functional regions within evolved sequences. Inspection of a small number of NRR-evolved clones identified a 40-base DNA sequence, present in multiple copies in each clone, that binds streptavidin. Our findings suggest that NRR may enhance the effectiveness of nucleic acid evolution and the ease of identifying structure-activity relationships among evolved sequences.

Figure 1

Diversification methods for nucleic acid evolution. Starting with parental sequences, pure SELEX enriches active sequences. Error-prone PCR yields parental sequences with point mutations. Homologous recombination methods, such as DNA shuffling, allow crossovers between sequences only in regions of homology. Nonhomologous random recombination allows many other changes, including random crossovers between any sequences, repetition, reordering, reorienting, elimination of subsequences, or any combination of these changes.

Figure 2

Overview of the nonhomologous random recombination (NRR) method. (A) Starting DNA sequences are randomly digested with DNase I, blunt-ended with T4 DNA polymerase, and recombined with T4 DNA ligase under conditions that strongly favor intermolecular ligation over intramolecular circularization. (B) A defined stoichiometry of hairpin DNA added to the ligation reaction controls the average length of the recombined products. The completed ligation reaction is digested with a restriction endonuclease to provide a library of double-stranded recombined DNA flanked by defined primer-binding sequences.

Figure 3

Nucleic acid evolution by NRR or error-prone PCR. A random pool of DNA was subjected to three rounds of SELEX for streptavidin binding. Two resulting sequences (S3–13 and S3–16) with moderate streptavidin affinity were diversified using either error-prone PCR or NRR, and the resulting three libraries were separately subjected to identical selection conditions.

Figure 4

Streptavidin binding affinities of DNA aptamers evolved by error-prone PCR (EPPCR) or by NRR. (A) Representative streptavidin binding data. (B) Assay results showing average K_d value and standard deviation of three to six independent trials. 13E, clones from the error-prone PCR of S3–13; 16E, clones from the error-prone PCR of S3–16; 13 × 16, clones from recombining S3–13 and S3–16 using NRR; 13×16-9 min, clones from minimizing 13×16#9 using NRR.

Figure 5

Diversification by error-prone PCR and by NRR. (A) Representative sequences arising from diversification by error-prone PCR. Mutations are shown in red. (B) Representative sequences generated by NRR compared with parental sequences. The parentage of recombined subsequences generated by NRR is shown in various colors.

Figure 6

Isolation of a minimal streptavidin binding motif by inspection of NRR-generated clones. (A) All of the sequenced clones evolved by NRR share a motif of ~40 bases despite their otherwise very different sequences. (B) This 40-mer, both with and without flanking primer binding sequences, has streptavidin binding activity and is predicted to fold into a stem–dumbbell structure. (C, D) The predicted secondary structures of full-length NRR-evolved clones invariably include one or more of these stem-dumbbell motifs.