JOURNALOFBIOSCIENCEANDBIOENGINEERING Vol. 96, No. 4, 3 17-323. 2003 REVIEW Bias and Artifacts in Multitemplate Polymerase Chain Reactions (PCR) TAKAHIRO KANAGAWA��� Institute for Biological Resources and Functions, National Institute of Advanced Industrial Science and Technologv Central 6, I-I-I Higashi, TSukuba, Ibaraki 305-8566, Japan��� Received 1 July 2003/Accepted 2 July 2003 Polymerase chain sea&on (PCR) is often used for the amplification of a mixture of homologous genes. PCR bias and artifact formation can occur in multitemplate PCR, and provide incorrect information on the abundance and diversity of genes. PCR bias and artifact formation occur at a higher rate during the last few cycles of the reaction, and therefore can be avoided by stopping the PCR earlier. [Key words: PCR bias, chimera, heteroduplex, microbial community analysis, 16s rRNA gene] PCR has become widely used in molecular biology, ap- plied microbiology, and microbial ecology. In many cases, the template is a mixture of homologous genes. In multi- template PCR, some factors give large bias and the relative abundance of homologs in the final amplification products does not necessarily reflect the gene ratio in the starting mixture (l-4). Moreover, artifacts are sometimes produced during the amplification and give the sequence data of non- existent genes (5-1.5). However, many researchers seem to ignore these problems, and there have been only a few re- ports on PCR bias and artifacts (16). In microbial ecology, to reveal microbial community struc- tures in environmental samples, the 16s rRNA gene (16s rDNA) is often amplified by PCR with universal primers which have a sequence common to all microorganisms. The PCR product is a mixture of 16s rDNA from various micro- organisms, and is analyzed by cloning and sequencing (17), denaturing gradient gel electrophoresis (DGGE) (18, 19), terminal restriction fragment length polymorphism (T-RFLP) (20, 21), and other methods (22). In these analyses, PCR bias results in incorrect population data on microorganisms, and artifacts show nonexistent varieties of microorganisms. However, these problems are not fully understood, and in many cases, PCR has been performed with little attention to these issues. To obtain reliable data from such PCR-based analyses, the elimination of PCR bias and artifacts is essen- tial. The present review focuses on the mechanism of PCR bias and artifact formation in multitemplate PCR, and dis- cusses suitable methods for the elimination of these prob- lems to increase the reliability of the data from PCR-based analyses. e-mail: kanagawa-taka@aist.go.jp phone: +81-(0)29-861-6026 fax: +81-(0)29-861-6400 I. PCR BIAS PCR bias due to differences in primer binding energy The primer set used in multitemplate PCR should have a se- quence common to the targets. However, often there are no common sequences in the targets. If a primer has one mis- match with some targets, the amplification efficiency is usu- ally very low, and therefore, a large bias in the amplification will occur. To avoid this bias, a degenerate primer which is a mixture of primers with a nucleotide sequence correspond- ing to the variation among homologs is often used. Polz and Cavanaugh (3) examined the extent of bias in PCR amplifi- cation with degenerate primers. They amplified bacterial 16s rDNA with the very popular degenerate primers 27F and 1492R, each of which contains a single degenerate position as shown in Table 1. First, they prepared the mutagenized Escherichiu coli 16s rDNA fragments Eco(GC), Eco(AT), and Eco(AT)m as templates. Eco(GC) and Eco(AT) differed at a single degenerate position in each of the priming sites for primers 27F and 1492R. Eco(AT)m was made from Eco(AT) by changing six nucleotides in the middle of the molecules for the quantification of Eco(AT)m in the mixture of Eco(AT) plus Eco(AT)m or Eco(GC) plus Eco(AT)m. When the mixture of Eco(AT) and Eco(AT)m at ratios of 1: 1, 1: 5, 1: IO, and 1: 20 was amplified by PCR with prim- ers 27F and 1492R, the ratio was maintained in the final PCR products after 25 cycles. This result shows that the six nucleotide difference in the middle of the template did not influence the product ratio. When an equal amount of Eco(GC) and Eco(AT)m was amplified with primers 27F and 149213, Eco(GC) was amplified better than Eco(AT)m at the product ratios of 1.4, 1.7, and 2.2 after 15, 25, and 35 cycles, respectively. This result shows the gene fragments containing G or C in the degenerate position were prefer- ably amplified probably because of a higher binding energy between G and C than A and T. Next, a mixture of equal 317

31X KANAGAWA J. Brosc~. BIOEN~.. TABLE I. Degenerate primers used by Polz and Cavanaugh (3) ���Templates Primers of perfect match Forward primer (27F)��� Reverse primer (I 492R) _____- __~- .-~ Eco(GC), B. subtilis 16s rDNA AGAGTTTGATCCTGGCTCAG TACGGCTACCTTGTTACGACTT EcolAT). EcofAT)m. CI fischeri 16s rDNA AGAGTTTGATCATGGCTCAG TACGGTTACCTTGTTACGACTT \ , , . ,,J a The underlined nucleotides are at the degenerate positions. amounts of total genomic DNA of Bacillus subtilis and Kbriofischeri was amplified with primers 27F and 1492R, and as a result, B. subtilis rDNA containing G at the de- generate positions was amplified with higher efficiency than Vjscheri rDNA containing A and T at the degenerate posi- tions. The product from the B. subtilis genome was 2.3 times more abundant than that from the Vfischeri genome after 25 PCR cycles. Considering the rrn operon number and the genome size of both bacteria, this means that 1.7 times more product was obtained from one B. subtilis rrn operon than from one V$scheri rrn operon. Thus, the tem- plate with the GC-rich permutation in the priming site was amplified better than that with the AT-rich permutation. The extent of the bias in PCR amplification would depend on the PCR conditions. In general, stricter annealing conditions cause a larger bias (23). Therefore, the extent of the bias should be examined in each case when degenerate primers are used. Whenever possible, degeneracy should be avoided. PCR bias to a 1:l ratio Suzuki and Giovannoni (1) have reported a strong bias towards 1: 1 mixtures of genes in final PCR products, regardless of the initial ratio of the templates. They prepared mixtures of two 16s rDNA frag- ments (27F- 1492R fragment) at 1: 4,2 : 3,3 : 2, and 4 : 1, and amplified them by PCR with primers 27F and 338R. The PCR product ratio after 35 cycles was nearly 1 : 1 in all cases. Mathieu-DaudC et al. (2) has also found this bias in the amplification of cDNA. The bias was strongly depen- dent on the cycle number. The original differences in con- centration decreased as the number of PCR cycles in- creased. This bias is explained by rehybridization of the PCR products. When the amplification reaction proceeds, the concentration of PCR products becomes high enough to allow the rehybridization of the products to some extent while the temperature is lower than the DNA melting point. If multiple PCR products are amplified in the same tube, rehybridization occurs faster for the more abundant PCR product and interferes with primer binding or extension. Consequently, the amplification rate for abundant PCR prod- ucts declines faster than that for less abundant products in the later PCR cycles, and the difference in starting template concentrations decreases. This bias can be eliminated by limiting the number of PCR cycles. II. HETERODUPLEXES Formation of a heteroduplex At the denaturing tem- perature in the PCR cycle, all DNA will be single stranded. When the temperature decreases for annealing, three kinds of duplexes can be formed, a homoduplex between comple- mentary strands, a heteroduplex caused by the cross-hy- bridization of heterologous sequences, and a duplex be- tween primers and templates. When the primer concentra- tion is much higher than the template concentration, the duplex between primers and templates is most abundant at the annealing step, and then, the homoduplex is produced by primer extension. Therefore, after the extension step, the homoduplex is dominant. On the other hand, in later PCR cycles, most primers are used up, and the amplified product concentrations are very high. In these conditions, a signifi- cant amount of heteroduplex would be formed. Problems caused by the heteroduplex Heteroduplexes cause several problems in PCR-based analyses such as clon- ing and sequencing, DGGE, and T-RFLP (24). When hetero- duplexes are cloned, the mismatch repair system in the host (E. co/i) can convert a heteroduplex into a homoduplex (25- 27). Since there is no way for the repair enzymes to identify the parent strand of the heteroduplex, the enzymes indepen- dently choose either strand as a template for resynthesis of the complementary base. This random choice can occur at each repairing position in one heteroduplex. As a result, the repaired sequence is a mixture of the two parent hetero- logous sequences (28). This result gives an artifactual se- quence and increases the sequence diversity. Speksnijder et al. (13) mixed seven closely related 16s rDNA fragments in equal quantities, amplified the mixture by PCR, and ana- lyzed the sequence by cloning and sequencing. Two of 66 clones showed an artifactual sequence resulting from the mismatch repair of the heteroduplex. In the electrophoretic analysis of PCR products, a hetero- duplex can give extra bands or peaks. Qiu et al. (14) have found an extra band on polyacrylamide gel. They equally mixed cloned 16s rDNAs from two strains, amplified near- ly the entire 16s rDNA by PCR, and analyzed the PCR products by PAGE. The migration of heteroduplexes was re- tarded and showed an extra band on the gel. The decrease in the mobility of the heteroduplex was inversely proportional to the sequence similarity of the two parental molecules. This decrease would be due to the bulky form of the mis- match sites. Thompson et al. (28) have found extra peaks in capillary electrophoresis. They mixed genomic DNA from three bacterial species, Vibrio cholera, J? parahaemolyticus, and c! vulnzjkus, amplified a 114-bp region of the 16s rDNA by PCR, and separated the PCR products by constant denaturant capillary electrophoresis (CDCE). As a result, six peaks appeared. Three peaks corresponded with homo- duplexes of the genes from the three species, while the three extra peaks corresponded with heteroduplexes among the genes. Thus, heteroduplexes showed a nonexistent diversity of genes. When heteroduplexes are applied to DGGE, they will surely give extra bands on the gel, and may lead to wrong conclusions. Judo et al. (12) experienced difficulty in restriction frag- ment length polymorphism (RFLP) analysis because of a significant amount of heteroduplex. They used a 590-bp