Introduction

Recombination can occur during PCR as a consequence of the tendency of Taq and other DNA polymerases to jump from one template to another. Thus, whenever a heterogeneous pool of similar sequences is being amplified, the potential exists to observe recombination, even under PCR conditions that would be considered standard for the error-free amplification of homogenous templates. Reports of the fraction of chimeric PCR products, which can be quantified as the recombination frequency (rf), range from <1% to as much as 7% (1,2,3,4). In fact, amplification of genes from polyploid sources has recently been shown to give rf values as high as 31% (5).

Recombination during PCR results from the existence of incompletely extended DNA oligomers after the extension (typically 72°C) step. Upon reannealing, these partial products can hybridize with a fully extended molecule, and if the two elements of these duplexes are not identical in sequence as a consequence of having been derived from different template molecules used to seed the PCR, a recombination event takes place. Subsequent rounds of PCR can amplify these recombinants and restore full sequence complementarity. Recombinatory PCR conditions exacerbate this phenomenon by lessening the probability of full-length extension; for example, the staggered extension process (StEP) utilizes 50–80 rounds containing no extension step, coupled with brief 0- to 10-s annealing steps, to create an abundance of abortive extension products (6). However, in principle, any phenomenon that causes the polymerase to pause or dissociate from a single-stranded DNA template will create partial extension products and promote recombination. Judo et al. (4) noted this in their demonstration that standard PCR amplification of an equimolar mixture of two homologous plasmid-derived sequences results in 1% or 7% recombinants between markers separated by 282 bp when Taq or Vent DNA polymerases, respectively, are used. The cycling conditions used in that study were 25 rounds of 45-s steps at 94°, 50°, and 72°C. These and other authors noted that an increase of extension time to 3 min reduced recombination rates 2- to 4-fold by both polymerases (1,4).

Researchers who are using PCR to create DNA templates for either the in vivo or in vitro transcription of RNA should be acutely aware of these findings. On the one hand, reverse transcription PCR (RT-PCR) amplification from potentially heterogeneous genomic RNA sources, such as retro-viral samples, can lead to erroneous conclusions about populational diversity if artifactual recombination is not taken into account. On the other hand, the use of recombinatory PCR can be deliberately performed during in vitro evolution to exploit the tremendous advantages that recombination offers in exploring sequence space, for both catalytic RNAs and proteins (6,7,8,9,10,11,12,13,14,15).

We have been investigating the efficacy of recombination among catalytic RNA molecules in enhancing the evolutionary process in vitro. Unlike the templates assayed previously for their propensities to recombine during standard PCR protocols, templates for catalytic RNAs have the capacity to form strong secondary structures. These structures, though not identical to those of the RNAs that they encode, include thermodynamically stable elements such as paired regions and tetraloops that could stall DNA polymerases even at 72°C. Here we employ an improved version of the restriction fragment-length polymorphism (RFLP)-based assay of Judo et al. (4) to investigate the recombination of RNA templates during both typical and recombinatory PCR. We find that recombination of the 421-bp template for the Tetrahymena group I intron is remarkably common during standard PCRs, and we test various parameters designed to minimize the recombination. Moreover, we show that other RNA templates, such as those for Escherichia coli 16S ribosomal RNA (rRNA) and eukaryotic (seal) messenger RNA (mRNA), that should contain different types of secondary structure are also quite prone to PCR-generated recombination.

Materials and Methods

Template Preparation

For the Tetrahymena ribozyme, the sources of the two parental molecules were plasmids, each of which encoded a Tetrahymena group I intron L-21 mutant. These sequences were both obtained from in vitro selection experiments for calcium-dependent activity (16). The sequence in plasmid NL137 (P1) has a total of four mutations, one of which (C47A) destroys the wild-type ScrFI restriction site at position 55. The ribozyme sequence in NL151 (P2) contained a C260A mutation that creates a new restriction site HpaI at position 266 while retaining the wild-type ScrFI restriction site. Each plasmid was amplified using PCR (see below for protocol), and the products were digested with Bme 1390I and KspAI restriction endonucleases (isoschizomers of ScrFI and HpaI, respectively) to ensure the plasmids were homogeneous. For the 16S E. coli rRNA, cloned bacterial genomic DNA was amplified using R519 and F8 primers that span 528 bp of the 5′ portion of the rDNA gene. The naturally occurring sequence of this gene contains neither EcoRI nor EcoRV restriction sites and could thus be used as the P1 sequence. To engineer P2, primers were designed with mismatches that create double mutations engendering EcoRV and EcoRI restriction sites at positions 63 and 503, respectively, in the resulting PCR products ((Figure 1)B). To insure homogeneity of the P2 sample, the original PCR product was diluted 1000-fold, reamplified with R519 and F8 primers, and digested with EcoRV and EcoRI, both separately and together. The complete absence of any undigested material was indicative of P2 DNA uncontaminated with P1 DNA. For the seal major histocompatibility (MHC) mRNA, genomic DNA from a Weddell seal with a homozygous genotype W6/W6 at the DQA locus was amplified with primers that generate a 925-bp product spanning approximately 475 bp of exon sequence and 450 bp of intron sequence (17). This gene contains no PvuII restriction sites and could thus be used as the P1 sequence. To engineer P2, a leopard seal with homozygous genotype L1/L1 at this locus was used as a starting point because it contained a single intrinsic PvuII restriction site 80 bp from the 5′end of the gene. The QuickChange® Mutagenesis Kit (Stratagene, La Jolla, CA, USA) was used to create a second PvuII restriction site 707 bp from the 5′ end of the gene ((Figure 1)C). PCR amplification of a single bacterial colony was performed with this mutation to produce a pure population of the P2 genotype.