Reducing PCR Bias in Methylation Analysis

Many analyses of DNA methylation commence with bisulfite treatment of genomic DNA, which converts unmethylated cytosines to uracils, leaving methylated cytosines unchanged. Bisulfite-induced sequence changes can be characterized, following PCR, by a number of approaches, including sequencing and restriction enzyme analysis. It has been reported that PCR amplification of bisulfite-treated DNA results in bias in the form of preferred amplification of unmethylated sequences, presumably as a result of secondary structure effects on amplification of the more GC-rich methylated sequences. Shen et al. address this issue by optimizing PCR annealing temperature and analyzing sequence changes using pyrosequencing. They find that, independent of the primer system used, higher annealing temperatures significantly improve the efficiency of amplification of unmethylated DNA sequences. Implementation of this approach may serve to advance the field of epigenetics by significantly improving the accuracy of DNA methylation analysis. -Page 48

In mammals, the target of DNA methylation is the C5 position of cytosine at the CpG dinucleotides. About 70% of CpG dinucleotides within the genome are methylated, and most unmethylated CpG dinucleotides are found in GC-rich sequences termed CpG islands (1). A large proportion of CpG islands is found in the promoter region of genes, and DNA methylation at these CpG islands can lead to transcriptional silencing of the associated genes. DNA methylation and associated gene silencing has essential normal functions for cell differentiation, imprinting, X-chromosome inactivation, and the suppression of parasitic DNA sequences (2). Aberrant DNA methylation plays an important role in both cancer initiation and progression, and this process is also implicated in other diseases, including imprinting disorders, diseases with trinucleotide expansions, and aging-related diseases (3,4,5). The central role of DNA methylation in maintaining cellular function, and the broad implications of DNA methylation in diseases have created a strong need for techniques to detect and measure DNA methylation reliably and quantitatively.

Currently, most approaches for measuring DNA methylation are based on sodium bisulfite treatment, which creates sequence differences by converting unmethylated cytosines to uracils, but leaving methylated cytosine unchanged. The differences can then be detected quantitatively by several techniques, such as sequencing of subclones or PCR products (6), restriction-digestion (COBRA) (7), or pyrosequencing (8). The main concern for PCR-based quantitative DNA methylation analysis is PCR bias, which is due to the fact that methylated and unmethylated DNA molecules sometimes amplify with greatly differing efficiencies. Using bisulfite PCR and restriction enzyme digestion, Warnecke et al. (9) reported a strong bias of amplification for unmethylated DNA in two human genes, p16 and Rb, and this bias occurred specifically for the primers directed to top strand DNA. They hypothesized that this bias could be due to the lower PCR efficiency through secondary structure formation for methylated DNA. However, they were unable to overcome this PCR bias by changing PCR conditions, such as extension time, annealing and denaturation temperature, MgCl₂ concentration, and the addition of varying concentrations of secondary structure inhibitors, such as dimethyl sulfoxide (DMSO) and formamide. Recently, we have used bisulfite PCR and pyrosequencing to study the methylation status at CpG island promoters for several genes. Surprisingly, we find that in most genes we analyzed, the amplification efficiency for the methylated DNA, and thus PCR bias, could be affected by changing annealing temperature for PCR, and this effect was independent of primers designed for top or bottom strand DNA. Using a mixing experiment, we find that much of the bias can be resolved by optimizing annealing temperature.

Bisulfite-pyrosequencing is a recently developed quantitative technique to detect methylation changes; it relies on bisulfite-induced C to T polymorphisms, which can be detected by a pyrosequencer using a sequencing-by-synthesis method. This technique has the advantages of analyzing several methylation sites, introduces an internal control (DNA sequence including a control for unconverted cytosines), and allows accurate quantitation of multiple CpG methylation sites in the same reaction. There are essentially two steps involved in this technique (i) PCR following bisulfite treatment and (ii) pyrosequencing to measure the degree of methylation at each CpG site within the sequencing region. For primer design, we first identified the region of interest and CpG island for each gene and virtually converted the genomic sequence to the bisulfite-treated sequence by replacing all CG with YG (Y stands for C/T), then all remaining C with T. Based on the converted sequences, we used Pyrosequencing™ Assay Design software (Biotage, Uppsala, Sweden) to design both PCR primers and sequencing primers. We tried to avoid CpGs within the primer sequences. If we could not find suitable primers this way, we included CpGs in each primer, but put them in the 5' end of the primer and synthesized them as Y (C/T) in the forward strand and R (G/A) in the reverse stand. The primer melting temperature (T_m) for PCR was calculated by the nearest neighbor formula, and the optimal annealing temperature was suggested at 55°C.

We examined methylation patterns at the CpG island promoters of seven genes, FGF11, LOC388407, FANK1, SOX11, P2RX5, TNFSF7, and FEZ1 in a colon cancer cell line RKO. Bisulfite treatment of 2 µg genomic DNA was performed as previously described (6). After bisulfite treatment, double-stranded DNA (top and bottom strands) are no longer complementary and can be analyzed separately by designing primers to amplify either top strand sequences or bottom strand sequences. For FGF11, LOC388407, FANK1, and SOX11 genes, the PCR primers were designed for top strand sequences; for P2RX5 and TNFSF7 genes, PCR primers were designed for the bottom strand sequences; and for FEZ1 gene, we designed two sets of PCR primers for both top and bottom strands. PCR product length, CpG density, and GC content within the PCR product are summarized in Table 1. Primer sequences and sequencing regions are shown in Table 2. For PCR amplification, we used a universal biotinylated primer with hot start PCR as previously reported (8). The PCR was carried out in a 50-µL solution containing 2 µL bisulfite-treated DNA, 1× PCR buffer, 1.25 mM dNTPs, 0.1 µM forward primer, 0.01 µM reverse primer with universal overhang, 0.09 µM universal biotin primer, and 1 U Taq DNA polymerase (New England BioLabs, Ipswich, MA, USA). For each assay, three different PCR conditions were performed separately for the same sample by varying annealing temperature at either 50°, 55°, or 60°C. PCR cycling conditions were 95°C for 5 min, followed by 50 cycles of 95°C for 30 s, varying annealing temperature for 45 s, and 72°C for 45 s, and a final incubation at 72°C for 4 min. In order to analyze as many CpG sites as possible, we usually divide PCR products for multiple sequencing and use 10 µL PCR product for pyrosequencing. Pyrosequencing was carried out with Pyro Gold reagents and a PSQ(tm) HS 96 pyrosequencer (both from Biotage), following the manufacturer's recommendations. The methylation levels at different C sites were averaged to represent the degree of methylation in each gene. All the experiments, including bisulfite treatment, PCR, and pyrosequencing, were done independently at least two times. As shown in Figure 1, at the lowest annealing temperature for PCR (at 50°C), five genes (FGF11, LOC388407, FANK1, FEZ1, and P2RX5) showed moderate methylation (40%–60%) in RKO, and two genes, SOX11 and TNFSF7, showed heavy methylation (78%–80%) in this cell line. Surprisingly, for the five genes with moderate methylation, we observed substantially higher methylation levels when the annealing temperature for PCR increased from 50° to 55° or 60°C, while for those two genes with heavy methylation, we did not see any changes for the measured methylation levels at different temperatures. Figure 2 shows representative pyrosequencing results for single CpG sites of FEZ1 gene. When we analyzed all the C sites individually (Figure 2A shows the pyrograms for 12 CpG sites on the top strand, and Figure 2B shows the pyrograms for 7 CpG sites on the bottom strand), we found that the increased methylation measurement by higher annealing temperature affected all C sites equally, suggesting an effect on allele-specific amplification (PCR bias), and we did not find any strand specificity to this effect that occurred to assays designed for top strands or bottom strands. These results indicate that during PCR amplification, methylated and unmethylated DNA has variable efficiency depending on the annealing temperature. These results suggested that we could potentially overcome PCR bias in DNA methylation analyses by optimizing annealing temperature.