Full Text (PDF)
In the range of currently used methodologies in methylation studies (reviewed in Reference (1), sequencing of bisulfite-treated DNA (2) can be considered the gold standard, as it reveals the methylation status of each CpG dinuclotide. Nevertheless, this technique is relatively expensive, therefore it has rarely been used in large-scale experiments. By contrast, the simplicity of design and easy performance of methylation-specific PCR (MSP) (3) has made it the most widely used technique in the investigation of methylation status of the genes. However, MSP is prone to false positive results and may lead to an overestimation of the number of methylated samples (4).
Methylation-sensitive melting curve analyses (MS-MCAs) (5) are an alternative that allow rapid and simple screening of methylation markers. The major advantage of the technique is that when an unambiguous melting profile is obtained, there is no need to confirm the result by a second method, which should be performed for all MSP-positive results.
Following the standard protocol for DNA methylation profiling by melting curve analyses published previously (6), we have encountered two major problems that significantly complicated our experiments. First, the stipulation that primers for MS-MCA should not discriminate between the methylated and unmethylated allele limits the primer binding sites to regions of the template without CpG dinucleotides. Finding suitable primer binding sites, especially within sequences with a high CpG content, is very difficult and can be impossible to achieve. A second problem is the issue of PCR bias. For the majority of our assays, which were designed to determine the qualitative (all or nothing) methylation status of a gene of interest, the PCR amplification showed a significant bias toward the unmethylated allele due to the difference in CG content after bisulfite treatment (7). We tested each assay for PCR bias using different ratios of methylated and unmethylated DNA [CpGenome™ Universal Methylated DNA (Chemicon, Temecula, CA, USA) and peripheral blood DNA as methylated and unmethylated control, respectively]. The majority of assays showed a positive methylation specific signal only in samples containing >50% methylated template. Representative results are shown in Figure 1A, in which the target is a portion of the putative tumor suppressor gene PPP3CC.
We introduced two changes in an attempt to address these limitations. First, we redesigned the primers to contain a limited number of CpG sites, thereby to anneal with a higher efficiency to the methylated than to the unmethylated template. Figure 2 shows the original and redesigned primers for methylation screening experiments of PPP3CC. The new primer sets efficiently amplified both methylated and unmethylated alleles (Figure 1B). Moreover, depending on the primer annealing temperature (Ta) during PCR, we were able to control the bias of the PCR amplification. At a relatively low Ta, both the methylated and unmethylated templates were amplified with comparable efficiency (Figure 1B), whereas elevating the Ta resulted in a significantly more efficient amplification of the methylated template (Figure 1C). In summary, by adjusting the Ta of the PCR amplification we were able to improve the sensitivity of our assay.
Validation of our approach was performed using four assays previously designed according to Reference (6). The assays were developed to screen for methylation of putative tumor suppressor genes in prostate cancer samples, and two of them showed methylation in <15% of samples in our panel. The redesigned primers were tested using templates containing defined mixtures of methylated and unmethylated DNA in experiments analogous to those shown in Figure 1. In all cases, the assays followed the pattern shown here for PPP3CC gene (i.e., depending on Ta, the same primer set was able to amplify the methylated allele of the gene of interest with higher efficiency; see Supplementary Figure S1, available online at www.BioTechniques.com). For the two assays that previously had been used to detect methylation in clinical samples, the modified procedure allowed detection of 40%–50% more methylated samples from the same sample panel (unpublished data). Our interpretation is that observed differences in methylation detection levels were due to increased sensitivity of the redesigned assays.
Our analyses of primer sequences used during these experiments showed that more than two CpG sites in the sequence of a single primer resulted in no amplification of unmethylated template (data not shown) and that the primer CpG sites should be kept as close as possible to the 5′ end of the primer sequence.
The new primer design approach—allowing CpG nucleotides into the primer sequences—significantly increases the flexibility in selection of amplification targets suitable for analyses, which is crucial for the experiment as the sequences subjected to MS-MCAs ideally should contain only one melting domain to ensure unambiguous results in the post-PCR melting analysis (8). Furthermore, the use of one primer set that is able to amplify both the methylated and unmethylated templates overcomes the problem of heterogeneous methylation of the primer binding sites, as primers always amplify the sequence of interest, no matter the methylation status of CpG sites included in primer binding sites.
This modified MS-MCA methodology serves as an alternative to other qualitative techniques currently used in methylation studies. By allowing users to control for the bias of PCR amplification, this method has potential in semiquantitative applications. The use of intercalating dyes reduces the costs of the experiments, as there is no need for the use of expensive probes. The method is characterized by high sensitivity; in our experiments we used a DNA bisulfite treatment protocol requiring only 200 ng genomic DNA (see Supplementary Material). This feature is especially relevant for diagnostics and cancer research experiments, which are frequently limited by a low quantity of tumor DNA.
This work was supported by The Danish Research Council. We would like to acknowledge Kenneth Frederiksen for technical help in preparation of the manuscript and Alexander Dobrovic for critical review of the manuscript.
The authors declare no competing interests.
