Downloaded from genome.cshlp.org on October 6, 2021 - Published by Cold Spring Harbor Laboratory Press DNA Polymerase Fidelity and the Polymerase Chain Reaction Kristin A. Eckert and Thomas A. Kunkel Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709 The distinctive ability of the because errors generated during can be controlled experimentally. polymerase chain reaction (PCR) to amplification may interfere with the produce a substantial quantity of DNA interpretation of data. ERROR ACCUMULATION DURING from an initially small amount of ge- During enzymatic DNA amplifica- PCR netic material is revolutionizing tion, the majority of changes in nucle- Error rates in PCR vary according to molecular biology. The variety of high- otide sequence can be attributed to er- the precise DNA sequence and the in ly successful applications of PCR tech- rors made by the DNA polymerase. vitro conditions of DNA synthesis. nology and the ease with which PCR While PCR is a recent development in Several laboratories have estimated the can be performed should not be the field of molecular biology, the fi- error frequency (mutations per misinterpreted as evidence that the delity of DNA polymerases has been nucleotide per cycle) during PCR PCR is a simple process. In fact, quite studied biochemically for almost 30 catalyzed by the thermostable Thermus the opposite is true. A single cycle of years. In this review, we will attempt to aquaticus (Taq) DNA polymerase by DNA polymerization is a complex pro- relate our current understanding of the cloning individual DNA molecules cess requiring the precise interaction of mechanisms of polymerase fidelity in from the amplified population and several components. The repetitive vitro to the accuracy of enzymatic determining the number of DNA se- cycles characteristic of PCR provide an DNA amplification. We begin by de- quence changes. (6-12) As seen from the additional layer of complexity to the scribing steps in polymerase error dis- data in Table 1, observed error frequen- technique. The final PCR product can- crimination and variables important cies during PCR vary more than 10- not be considered a unique entity, as for determining fidelity. The error rates fold, from -2 x 10 -4 to <1 x 10 -5 . For even a discrete DNA band on an of several DNA polymerases that have applications of PCR products that util- agarose gel may contain a variety of been used for PCR will be described, ize the population of amplified DNA DNA molecules differing from the first as determined in model "one- molecules, random errors at these fre- original genetic information by one or cycle" fidelity assays and then during quencies will not interfere with the more nucleotides. actual PCR. Our intent is to focus on biochemical analyses. However, poly- How serious a concern is the fidel- the variables known to influence the merase-mediated errors at a frequency ity of PCR? The answer depends on the fidelity of DNA synthesis in vitro that of 1 mutation per 10,000 nucleotides precise nature of the application. In many instances, for example, direct characterization of the amplified popu- lation by DNA sequence analysis or TABLE 1 Estimates of Fidelity during PCR Catalyzed by Taq Polymerase nucleic acid hybridization, random er- rors in nucleotide sequence that may Error Error rate Target Number frequency per nucleotide be produced during PCR are of little sequence of cycles per nucleotide per cycle a Reference concern. However, some PCR applica- tions involve the characterization of Apolipoprotein B 30 22/8,000 1/5,600 6 individual DNA molecules or rare HLA-DPf~ 30 17/6,692 1/5,900 7 molecules present in a heterogeneous R-Antit rypsin 33 7/4,700 1/ 11,000 8 population. Examples include the HPRT 32 16/15,000 1/15,O00 9 study of allelic polymorphism in indi- HLA-A,B 30 20/21,870 1/16,000 10 vidual mRNA transcripts O,2) and the 20 17/30,710 1/18,000 10 characterization of the allelic states of TCR V8 25 1/1,500 1/19,000 11 single sperm cells (3) or single DNA HIV gag/env 30 < 1/5,411 < 1/83,0OO 12 molecules. (4,s) In these circumstances, the fidelity (error rate per nucleotide) aError rate = I/[(observed error frequency/number of cycles) x 2], assuming a constant ef- of PCR is an important consideration, ficiency per cycle. U) 1:1 7-249 by Cold Spring Harbor Laboratory Press ISSN 1054-9803/91 $3.00 PCR Methods and Applications 1 7 Downloaded from genome.cshlp.org on October 6, 2021 - Published by Cold Spring Harbor Laboratory Press per cycle are an important considera- larger than average number of purified polymerases in vitro increase tion for any PCR application that be- mutants. ~ls) The remainder of this as the length of the repeated sequence gins with a small amount of starting review explores the ways in which the increases. (16) Alternative models have material (i.e., less than a total of polymerase error rate, p, can be also been proposed, including dele- 10,000 nucleotides of target DNA) or manipulated experimentally as a tions initiated by base misinsertion. (19) that focuses on individual DNA means of controlling the PCR error fre- Substantial evidence from model as- molecules present in the final PCR quency. says demonstrates that polymerases population. also can generate deletions of a few to The proportion of DNA molecules MOLECULAR NATURE OF hundreds of nucleotides during that contain sequence changes is a MUTATIONS polymerization in vitro. These dele- function of the error rate per nucleo- A variety of changes in DNA sequence tions often involve directly repeated tide per cycle, the number of amplifi- can occur during enzymatic DNA sequences and palindromic or quasi- cation cycles, and the starting popula- amplification. These mutations vary palindromic sequences that have the tion size.t13,14) The population of from single base-substitution muta- potential to form stern-and-loop DNA altered DNA molecules arises during tions to deletions and insertions. 16-12) structures.~177 PCR from two sources: (1) new errors The most familiar type of mutation is In addition to intramolecular errors at each cycle and (2) the amplification the base substitution that results from during DNA synthesis, deletions and of DNA molecules containing errors the misincorporation of an incorrect other types of rearrangements can oc- from previous cycles. On average, the dNTP during DNA synthesis. Twelve cur between DNA molecules. Such in- same number of DNA sequence distinct base mispairs are possible, termolecular events are thought to changes in the final population results varying in composition and in the arise when the polymerase terminates from polymerase errors during the last template base versus the incoming synthesis on one DNA strand and con- amplification cycle as result from the dNTP (e.g., T'dGTP versus G'dTFP). tinues synthesis after priming occurs amplification of errors that occurred Together with the effects of neighbor- on a complementary DNA strand during previous cycles. The formula f = ing template sequence, this creates a (strand switching or jumping PCR). (2~ np/2 describes the average mutation variety of possible molecular structures Both the presence of a high degree of frequency (f) for PCR amplification as a among which any polymerase must sequence homology between regions function of the polymerase error rate discriminate. All DNA polymerases ex- of DNA molecules and a high DNA per nucleotide per cycle (p) and the amined to date also generate muta- concentration increase the incidence number of cycles (n), assuming that p tions during DNA polymerization in of this type of error in PCR. (1~176 is constant at each cycle (for more vitro wherein one or more nucleotides detailed mathematical analyses, see is lost or added. One factor that will in- ERROR DISCRIMINATION BY DNA refs. 13 and 14). Theoretically, the fre- fluence the types of DNA sequence POLYMERASFS AND FIDELITY IN quency of DNA sequence changes can changes (base substitution versus dele- MODEL SYSTEMS be controlled by altering the number tion mutation) likely to occur during Since the majority of DNA sequence of cycles (n) and/or the polymerase er- enzymatic amplification is the se- changes introduced during PCR may ror rate per nucleotide (p) (see Table 6 quence of the target DNA. be polymerase-mediated, we will in ref. 13). For example, as the number The frequency of deletion muta- review what is known about DNA of cycles increases, the error frequency tions is sequence dependent and is in- polymerase error rates from model in is expected to increase. Keeping p con- creased in repetitive DNA sequences vitro systems. Much of our under- stant at one error per 10,000 (for reviews, see refs. 16 and 17) Many standing of fidelity is derived from as- nucleotides per cycle, the error fre- examples exist for the repetitive nature says that quantitate errors made by quency when n = 20 cycles is one of DNA, from simple reiteration of a DNA polymerases during synthesis change per 1000 nucleotides and in- single base, to alternating dinucleotide equivalent to a single PCR cycle. (21-24) creases to one change per 400 tracts in potential Z-DNA structure, to Among these, the broadest description nucleotides after 50 amplification short direct repeats. The generation of of fidelity comes from the M13mp2 cycles. Alternatively, the final error fre- deletion mutations is likely to be forward mutation assay, (24) which quency can be varied by changing p, determined by the stability of the detects all possible base substitution er- the polymerase error rate. Keeping a mutational intermediates.(16,17) The rors along with a variety of base dele- constant cycle number of n = 50, the most common polymerase-mediated tion and addition errors at a large final mutation frequency is only deletion error is the loss of one number of sites.