A real-time PCR assay for DNA-methylation using methylation-specific b
http://www.100md.com
《核酸研究医学期刊》
Epigenomics Inc, 1000 Seneca Street, Suite 300 Seattle, WA 98101, USA and 1 Epigenomics AG, Science Department, Kastanienallee 24, 10435 Berlin, Germany
*To whom correspondence should be addressed. Tel: +49 30 2434 5303; Fax: +49 30 2434 5299; Email: berlin@epigenomics.com
ABSTRACT
DNA methylation-based biomarkers have been discovered that could potentially be used for the diagnosis of cancer by detection of circulating, tumor-derived DNA in bodily fluids. Any methylation detection assay that would be applied to these samples must be capable of detecting small amounts of tumor DNA in the presence of background normal DNA. We have developed a real-time PCR assay, called HeavyMethyl, that is well suited for this application. HeavyMethyl uses methylation-specific oligonucleotide blockers and a methylation-specific probe to achieve methylation-specific amplification and detection. We tested the assays on unmethylated and artificially methylated DNA in order to determine the limit of detection. After careful optimization, our glutathione-S-transferase pi1 and Calcitonin assays can amplify as little as 30 and 60 pg of methylated DNA, respectively, and neither assay amplifies unmethylated DNA. The Calcitonin assay showed a highly significant methylation difference between normal colon and colon adenocarcinomas, and methylation was also detected in serum DNA from colon cancer patients. These assays show that HeavyMethyl technology can be successfully employed for the analysis of very low concentrations of methylated DNA, e.g. in serum of patients with tumors.
INTRODUCTION
CpG dinucleotides are underrepresented in the genome and appear at only 20% of the expected frequency. In regions referred to as CpG islands, the frequency of this dinucleotide is much higher, approaching or exceeding the statistically expected frequency of 1 in 16 dinucleotides. Typically these CpG islands are found near genes and stretch over 500–2000 bp (1). In normal tissues, most cytosines in the context of CpG dinucleotides within CpG islands are unmethylated, but some methylation occurs in imprinted genes and inactive X-chromosomal genes. In addition, somatic de novo methylation of CpG islands can occur in genes involved in controlling cell growth and division. This aberrant methylation is often associated with reduced transcription of the associated gene, and CpG island hypermethylation is a common mechanism of gene inactivation during tumorigenesis (2).
Specific patterns of DNA methylation have been found to characterize different cancer tissues (3–5). Methylation of the glutathione-S-transferase pi 1 (GSTP1) gene for example, has been found in prostate, breast and liver tumors but not bladder, stomach or head and neck tumors (5). Due to the high rates of methylation and the stability of the signal, many of these methylated genes have been investigated as potential biomarkers for early detection of cancer, and preliminary studies suggest that some of these markers may even be useful for detection of tumor-derived DNA in diagnostic samples such as urine, serum and plasma (6–9). Since there is normally a limited amount of total DNA in these types of samples and only a fraction of this DNA is derived from tumor cells, this application relies on methylation detection assays with high analytical sensitivities. Also, the detection assay must have a high relative sensitivity that favors the tumor DNA over the background DNA, in order to avoid false-positive results when a high amount of non-tumor DNA is present in the sample.
In addition to sensitivity, an early detection assay based on methylation markers must meet other requirements. All components of the assay will, after bisulfite treatment, ideally be combined in one tube in order to minimize the amount of labor as well as the risk of operator error and sample contamination. Furthermore, since methylation is not uniform within a heterogeneous DNA sample, the assay must be able to quantitatively detect the aberrantly methylated DNA in order to provide the maximum amount of information.
The first and most common assay applied to sensitive detection of methylation in tissues and body fluids is called methylation-specific PCR (MSP; 8). For a given DNA sample, treatment with a bisulfite reagent leads to the conversion of all unmethylated cytosines to uracils, leaving only methylated cytosines unchanged (3,10). MSP is now carried out analogously to an allele-specific PCR reaction. The primers bind either to methylated or unmethylated DNA only and therefore selectively amplify only DNA with a defined methylation. In contrast to the allele-specific PCR used for single nucleotide polymorphism (SNP) analysis, MSP primer binding sites include several CpG sites within bisulfite-treated DNA, so that binding and extension are methylation specific. MSP primers are usually designed to detect targets that are either methylated at all of the CpG sites in the binding site or unmethylated at all CpGs. Users of this assay have reported high sensitivity (11,12), and the widespread application of MSP has enabled an increase in research on methylation. Addition of a methylation-specific real-time detection probe (MethyLight) makes the assay both homogenous and quantitative (12–15). The quantitative nature of MethyLight was particularly useful in recently published research distinguishing benign prostate hypertrophy from prostate cancer based on GSTP1 methylation (16).
This paper introduces a novel real-time methylation assay called HeavyMethyl. Like MSP, the priming is methylation specific, but non-extendable oligonucleotide blockers provide this specificity instead of the primers themselves. The blockers bind to bisulfite-treated DNA in a methylation-specific manner, and their binding sites overlap the primer binding sites (Fig. 1). When the blocker is bound, the primer cannot bind and therefore the amplicon is not generated. When the blocker is not bound, the primer-binding site is accessible and the amplicon is generated. Similar blocking approaches have been used to achieve high specificity in primer extension and chip assays for mutation detection (17,18), but so far blocker oligonucleotides have not been successfully employed in methylation analysis. As a proof of concept, we have designed assays for two genes: GSTP1, which is frequently methylated in prostate cancer (16), and Calcitonin (CALCA), which is hypermethylated in multiple types of cancer (19,20). Here we demonstrate that HeavyMethyl in combination with real-time detection with methylation-specific fluorogenic probes is capable of sensitive and specific detection of DNA methylation. Furthermore, our initial application of these assays to tumor and serum samples suggests that this format may be suitable for clinical studies.
Figure 1. Principle of HeavyMethyl. (A) When the DNA is methylated, the blocker oligonucleotides (solid black) do not bind, leaving the primer- binding site accessible for the primers (gray arrows) to bind and amplify the target. The amplification is detected with a methylation specific oligonucleotide probe . (B) When the DNA is unmethylated, the blocker oligonucleotides bind, blocking the access of the primers to their binding sites. No PCR product is generated.
MATERIALS AND METHODS
GSTP1 assay
PCR amplification of a GSTP1Exon fragment (nucleotides 1241–1394 in GenBank M24485 .1) was performed in a 25 μl reaction volume with 500 nM GSTP1Exon forward primer (GTTTTYGTTATTAGTGAGT), 500 nM GSTP1Exon reverse primer (TCCTAAATCCCCTAAACC), 0.2 mM dNTPs, 10 μM GSTP1Exon blocker 1 (GTGAGTATGTGTGGTTTGTGT-phos; where phos = phosphorylated), 10 μM GSTP1Exon blocker 2 (TAAACCCCCATCCCAAATCTCA-phos), 1x Qiagen reaction buffer, 1.5 mM MgCl2, and 1 U Hotstar Polymerase (Qiagen). Thermocycling conditions were an initial step at 95°C (15 min), then 45 cycles of the following steps: 95°C (45 s), 52°C (45 s) and 72°C (20 s). After an additional step of 10 min at 72°C, 10 μl of the PCR reaction was analyzed on a 1.5% agarose gel.
Real-time PCR experiments were performed using a LightCycler (Roche Diagnostics; 21). The 20 μl PCR reactions contained 0.31 μM GSTP1Exon forward primer, 0.31 μM GSTP1Exon reverse primer, 0.25 g/l bovine serum albumin (BSA, Sigma), 0.25 mM dNTPs, 0.25 μM GSTP1Exon anchor probe (GTTTAGAGTTTTTAGTATGGGGTTAATT-Fluo; where Fluo = fluorescein), 0.25 μM GSTP1Exon probe I, specific for methylated bisulfite-treated DNA (LCRed640-TAGTATTAGGTTCGGGTTTTCGG-phos) or probe II, specific for unmethylated bisulfite-treated DNA (LCRed705-TAGTATTAGGTTTGGGTTTTTGG-phos; all probes obtained from TIB-Molbiol), 10 μM GSTP1Exon blocker 3 oligonucleotide (GTGAGTATGTGTGGTTTGTGTT-phos), 1x Qiagen reaction buffer, 1.5 mM MgCl2 and 1 U Hotstar polymerase. A total of 0.1 or 100 ng of bisulfite-treated template DNA was used. For melting-peak analysis of GSTP1Exon in the PCR, SYBR Green I was used instead of specific probes. Thermocycling conditions were an initial step at 95°C (15 min), then 50 cycles of the following steps: 95°C (4 s), 52°C (30 s) with a single fluorescence detection after 10 s in each cycle, and 72°C (20 s). Melt peaks were obtained by plotting the negative derivate of SYBR Green I fluorescence with respect to temperature (–dF/dT) against temperature.
The HeavyMethyl assay of the GSTP1promoter (Prom) fragment (GenBank M24485 .1, position 1014–1171) was performed in 20 μl using the LightCycler. The real-time PCR reaction mix contained 0.30 μM GSTP1Prom forward primer (GGGAAAGAGGGAAAGGTTTTTT), 0.30 μM GSTP1Prom reverse primer (CCCCAATACTAAATCAC), 0.25 g/l BSA (Sigma), 0.2 mM dNTPs, 0.15 μM GSTP1Prom anchor probe (CGGTCGACGTTCGGG-Fluo), 0.15 μM GSTP1Prom detection probe (LCRed640-TAGCGGtCGtCGGGG-phos), 10 μM GSTP1Prom blocker (CTAAATCACAACACCAACCACTCTTC-phos), 1x Qiagen reaction buffer, 1.75 mM MgCl2 and 2 U Hotstar polymerase (Qiagen). Both GSTP1Prom anchor and detection probes were specific for methylated bisulfite-treated template DNA. Thermocycling conditions began with an initial step at 95°C (15 min), then 55 cycles of the following steps: 95°C (10 s), 48°C (30 s) and 72°C (20 s). Fluorescence was detected after the annealing phase at 48°C in each cycle. Underlined nucleotides indicate the overlap between primer and blocker oligonucleotides.
Calcitonin assay
The Calcitonin assay is a 5' exonuclease assay (22) that targets a region of the genome on 11p15 1200 bp upstream from the transcription start site (GenBank X15943 , position 476–728) with primers (GGATGTGAGAGTTGTTGAGGTTA and (ACACACCCAAACCCATTACTATCT; Integrated DNA Technologies). The blocker (TGTTGAGGTTATGTGTAATTGGGTGTGA-phos) overlaps the forward primer sequence by 11 bp, spans two CpGs, and then covers another 13 bp. In addition to the 5' fluorescein label, the probe Fluo-ACCTCCGAATCTCTCGAACGATCGC-BHQ is labeled with a Black Hole Quencher (BHQ) at the 3'-end (Biosearch Technologies).
The Calcitonin assay was run with 300 nM primers, 5 μM blocker, 450 nM probe, 2 U Amplitaq Gold, 3.5 mM MgCl2 and 0.4 mM dNTPs (all Applied Biosystems) in a final reaction volume of 20 μl. First, the reactions were held at 95°C for 10 min to activate the enzyme. Then, we used two-step cycling to amplify the DNA (95°C for 15 s, 67°C for 60 s and repeat three times; 95°C for 15 s, 64°C for 60 s and repeat three times; 95°C, 15 s, 62°C for 60 s and repeat three times; 95°C, 15 s, 60°C for 60 s and repeat 40 times). Data were collected during the extension phase at 60°C on an ABI Prism 7900 instrument.
A positive control reaction for ?-actin was also used to control for input DNA when analyzing the colon samples. This assay has been described previously (20). The same cycling conditions were used as for the Calcitonin assay.
DNA preparation and bisulfite treatment
Colon samples from patients (from 21 to 91 years old) with stage I, II or II colon carcinoma were obtained during surgery (approval from Western Institutional Review Board). DNA was isolated from 100–500 mg of tissue using Genomic Tip 500 kit (Qiagen). DNA was isolated from 1 ml of serum collected before surgery using UltraSens Viral Nucleic Acids kit (Qiagen).
Two sources of unmethylated control DNA were used. For the Calcitonin analysis, sperm DNA was isolated from semen using a protocol supplied by Qiagen. For the GSTP1 analysis, unmethylated DNA from human peripheral lymphocytes (Promega) was used. The methylation status of the lymphocyte DNA was confirmed by bisulfite sequencing. Methylated DNA was created by treating genomic DNA (Promega) with SssI methyltransferase (New England Biolabs) in the presence of a methyl donor S-adenosyl-methionine according to the manufacturer’s instructions.
For bisulfite treatment the denatured DNA was embedded in a 2% agarose bead and was further processed as described earlier (10). Between 5 and 10 μl of this DNA was used per reaction.
Data analysis
Percent Methylated Reference (PMR) was calculated using a procedure previously described (20). The Cts (cycle thresholds) for all samples were converted to a relative quantity scale using a standard curve of diluted methylated DNA. The ratio of the ‘quantity’ of Calcitonin-methylated DNA to total ?-actin DNA is calculated for all samples of interest as well as the fully methylated control DNA. The ratio of these ratios is the PMR. A Mann–Whitney test was used for all statistical tests of significance. ROC (Receiver Operating Characteristic) curves were generated by ranking all of the tissue samples by their PMR values. The tumor sample population overlaps the normal sample population, so a clear threshold could not be determined. The sensitivity and false-positive rate were calculated for each possible cut-off point between the normal and tumor classes and these values were plotted. As the sensitivity increases, the specificity decreases. The best markers will have very little reduction in specificity as sensitivity increases, so the area under the ROC curves (AUC) will be close to 1. A marker with no predictive value will have an AUC of 0.5.
RESULTS
Methylation-specific blocking oligonucleotides prevent amplification of unmethylated DNA
The GSTP1 gene exon 1 is frequently methylated in prostate cancer (4). The GSTP1Exon primers used in our HeavyMethyl assay amplify a 153 bp fragment. In order to prevent amplification of unmethylated DNA, different blocker oligonucleotides were designed to overlap the 3'-end of the binding site of the primer as well as downstream CpGs positions. Therefore, the blocker binds to the target only when the DNA is unmethylated at these CpGs, allowing the primer to bind only when the DNA is partially or fully methylated.
We investigated the effect of the following parameters on the performance of our GSTP1Exon assay: number of blockers (one per assay or one per primer), the melting temperature of the blockers, the melting temperature difference between blocker annealing to the unmethylated and methylated target DNA, the number of CpGs within the blockers, the number of overlapping nucleotides between primers and blockers, and the ratio of the concentrations of primers and blockers. (data not shown).
For the final GSTP1Exon HeavyMethyl assay, the blockers 1 and 2 cover five and three CpGs, respectively. A primer/blocker ratio of 1:20 repressed the GSTP1Exon amplification on 50 ng of unmethylated DNA, but amplification of 0.075 ng methylated template DNA was not significantly inhibited (Fig. 2A, lanes 7 and 12). The GSTP1Exon HeavyMethyl assay detected 100 pg methylated template DNA in a background of 100 ng unmethylated DNA, representing a relative analytical sensitivity of 1:1000. The products obtained with and without blocker were sequenced to confirm their methylated and unmethylated status, respectively (Fig. 2B). The effect of the blocker is seen more clearly with melt peak analysis on a LightCycler (Fig. 2C). GSTP1Exon products of methylated and unmethylated bisulfite-treated template DNA generate traces with melting peaks of 84 and 79°C, respectively. On template DNA representing a mixture of 100 pg SssI methylated and 100 ng unmethylated bisulfite-treated DNA, only methylated GSTP1Exon products were generated by the HeavyMethyl assay in the presence of blocker. Without blocker only unmethylated GSTP1Exon PCR products were found.
Figure 2. Relative and absolute sensitivity of GSTP1 HeavyMethyl assay. GSTP1Exon amplification was performed with and without blockers. (A) 1.5% agarose gel analysis of GSTP1Exon PCRs performed without (–B; lanes 1–6, 13) and with blocker (+B; lanes 7–12) on 50 ng bisulfite-treated unmethylated DNA (–CH3; lanes 1, 7), different amounts of bisulfite-treated SssI-methylated DNA (CH3; 1.2 ng, lanes 2 and 8; 0.6 ng, lanes 3 and 9; 0.3 ng, lanes 4 and 10; 0.15 ng, lanes 5 and 11; 75 pg, lanes 6 and 12) and no DNA ( lane 13). (B) Sequence traces of GSTP1Exon products generated on a mixture of 100 pg methylated in a background of 100 ng unmethylated template DNA (relative sensitivity 1:1000) in the presence and absence of blockers. (C) The discrimination of GSTP1Exon products generated by HeavyMethyl assay with (+B) and without (–B) blocker on bisulfite-treated SssI-methylated (+CH3), unmethylated (–CH3) and a mixture of 100 pg methylated in 100 ng unmethylated (1:1000) template DNA by melting peak analysis.
Real-time HeavyMethyl assays are sensitive and specific
A gel-based detection of HeavyMethyl amplification allows us to visualize the effect of blockers on amplification of unmethylated DNA. However, real-time detection of amplification is preferable because the assay is both quantitative and homogenous. For the probe-based real-time HeavyMethyl assay, blocker 3 sufficiently blocked the amplification of the GSTP1Exon fragment using up to 50 ng unmethylated DNA (Fig. 3B): in the presence of blocker no product was detectable with the detection probe II designed for unmethylated DNA. In addition, the PCR with 15 ng bisulfite-treated methylated template DNA was not affected by the blocker (Fig. 3A). Figure 3A also shows that the detection probe I specifically detected methylated DNA, whereas unmethylated DNA amplified either with or without blocker was not identified. The analytical sensitivity of this HeavyMethyl assay was 30 pg (six genome equivalents, data not shown). Furthermore, 30 pg of methylated DNA was reproducibly detectable in a background of 50 ng unmethylated DNA with the HeavyMethyl assay (Fig. 3C). In contrast, the assay without blocker did not detect the spiked methylated DNA (Fig. 3C). This HeavyMethyl assay showed a relative sensitivity of 1:1600.
Figure 3. HeavyMethyl real-time assays for GSTP1 exon and promoter. (A–C) HeavyMethyl GSTP1 exon assay. Amplification plots of 15 ng SssI-methylated DNA indicated by solid or broken lines and 50 ng of unmethylated DNA indicated by lines with circles (A). (B) 50 ng of unmethylated DNA and 30 pg methylated DNA spiked into 50 ng unmethylated DNA representing a relative sensitivity of 1:1600 (C), which were bisulfite treated and subsequently used as template for the HeavyMethyl assay. The LightCycler detection probes were specific for bisulfite-treated unmethylated (B) or methylated DNA (A, C). HeavyMethyl assays performed with and without blocker are indicated by solid and broken lines, respectively. In (A), the HeavyMethyl assay with and without blocker is also indicated by filled and open circles, respectively. (D) HeavyMethyl GSTP1 promoter assay. The GSTP1 promoter fragment was amplified on 400 ng unmethylated (five replicates indicated by broken lines), 100 pg SssI-methylated DNA (black diamonds) and mixtures containing 100 pg (six replicates indicated by solid lines) or 50 pg (six replicates indicated by lines labeled with open diamonds) methylated DNA in 400 ng unmethylated DNA. All template DNA samples were bisulfite treated prior to real-time PCR experiments. The products were specifically detected via the FRET signal generated by the Fluorescein- and LCRed640-labeled detection probes specific for methylated bisulfite-treated DNA.
The second GSTP1 HeavyMethyl assay, established on the promoter region, showed an even better relative sensitivity. In Figure 3D, this HeavyMethyl assay reproducibly detected methylated copies of a GSTP1Prom fragment in a background of 400 ng unmethylated bisulfite-treated human DNA. This real-time HeavyMethyl assay showed a relative sensitivity of up to 1:8000. In addition, products were generated on 25 and 12.5 pg methylated bisulfite-treated template DNA in 16 out of 16 and 9 out of 16 real-time PCRs, respectively (data not shown).
To determine whether HeavyMethyl assays can be designed for an additional real-time PCR format, we created a 5' exonuclease real-time assay for a CpG island upstream of the Calcitonin gene. In previous reports, CpGs near the transcription start site of the gene have been shown to be methylated in colon cancer and leukemia (17,18). We designed an assay for a CpG island further upstream. The blocker oligonucleotide targets two CpGs just beyond the 3' end of the forward primer, and the blocker binding site overlaps the primer binding site by 11 bp. Since the probe-binding site contains four CpGs, the binding of the probe is also methylation specific.
The Calcitonin HeavyMethyl real-time assay amplified methylated target in the presence of 500-fold excess unmethylated DNA, and in these wells the total amount of methylated DNA is estimated to be <60 pg. In contrast, there was never any significant amplification of unmethylated DNA. Series dilutions of bisulfite-treated methylated into bisulfite-treated unmethylated DNA showed that the assay is linear over at least three orders of magnitude and there is little deviation of the points from the line (R2 = 0.981). The slope, –3.30 (standard error = 0.263), suggests that this assay is highly efficient and the amount of target doubles with each cycle (data not shown).
HeavyMethyl analysis of clinical samples
We used the Calcitonin assay to detect methylation in colon samples to demonstrate that HeavyMethyl assays yield clinically relevant information. We extracted and bisulfite-treated DNA from colon adenocarcinoma samples and normal matched colon tissue. The methylation was measured with our Calcitonin assay, and the total amount of input DNA was measured with a control assay for ?-actin. The control assay does not contain CpGs and therefore amplifies all bisulfite-converted DNA regardless of methylation status. For each sample, the ratio of the quantity of methylated Calcitonin divided by the total amount of DNA was normalized to the same ratio calculated for artificially methylated DNA, the PMR (20).
The mean PMR of the adenocarcinoma samples (n = 33) was 4.76, while the mean PMR for the normal samples (n = 43) was only 0.05. This methylation difference is highly significant (P < 0.0001). There was no effect of age or disease stage on methylation levels in either the normal or tumor samples (Fig. 4A).
Figure 4. Methylation of Calcitonin in colon cancer tissue and serum samples. (A) The PMR is plotted for colon cancer tumor and normal colon tissue samples. The PMR is also plotted for five serum samples from individuals with methylated tumors and 11 serum samples from healthy control donors. (B) Methylation of Calcitonin in normal (n = 43) and adenocarcinoma (n = 33) colon samples was measured by both HeavyMethyl and MSP. ROC curves were generated by plotting the sensitivity against the false-negative rate (1 – specificity). The area under the curve (0.84 for HeavyMethyl and 0.81 for MSP) is used to evaluate the performance of the assay.
Performance of the assay was compared to an MSP-based MethyLight assay covering the same genomic region, using the same CpGs for a methylation-specific probe. The two assays showed similar sensitivity and specificity with the same set of samples. The mean PMR values of the MSP-based MethyLight assay (0.19 for healthy and 5.04 for tumor samples) were very similar to the values for the HeavyMethyl assay (Fig. 4A). Since the two assays do not cover exactly the same CpG positions, methylation biology will strongly influence the results. Figure 4B shows the ROC curves for the HeavyMethyl and MSP MethyLight assays for the Calcitonin gene. The area under the curve is 0.84 for the HeavyMethyl assay and 0.81 for the MSP MethyLight assay. The high analytical sensitivity of our assay suggests that it is suitable for application to serum analysis. We collected serum samples from 11 healthy individuals and found minimal detectable methylation in these samples. The mean PMR was 0.028 ± 0.025. Five of our methylated colon adenocarcinoma samples had matched serum samples taken before surgery, and all five of these had methylation levels at least 6 standard deviations above the mean of the healthy population (Fig. 4A). While the sample size is small in this experiment, the data suggest that HeavyMethyl assays have the sensitivity and specificity necessary for a serum-based application.
DISCUSSION
We have presented data confirming that the introduced HeavyMethyl methodology has high potential for the analysis of DNA methylation in clinical samples. A successful assay format for the detection of methylation in body fluids must fulfill several criteria. Most important is analytical sensitivity: in a given volume of a body fluid, only very limited amounts of DNA may be present. In healthy patients, for example, a typical amount of non cell-bound, free-floating DNA within a milliliter of serum is 10–50 ng (23,24). Although there is evidence that this amount may be increased in tumor patients because of apoptosis or still unknown mechanisms, the amount of DNA that is available for analysis is typically even smaller because of low extraction yields, the small average size of free-floating DNA, DNA degradation during bisulfite conversion and, most importantly, the varying ratio of tumor to healthy free-floating DNA (23–25).
Ideally, an assay would therefore have the sensitivity and specificity to detect a single copy of methylated DNA within a given volume of body fluid. In our hands, we were able to design and conduct PCR reactions that reproducibly amplify as little as 30 pg (DNA from six cells) of bisulfite-converted, previously methylated DNA, with an excess of blocker molecules binding to unmethylated DNA. The presence of blocker molecules did not influence the analytical sensitivity of the PCR reaction significantly. Thirty picograms is far below the expected amounts of tumor DNA in symptomatic cancer patients, as determined, for example, for lung and colon cancer patients (50–500 ng/ml serum; 23–25); however, one would like to be able to detect as few aberrantly methylated DNA molecules as possible for an early cancer screening test. Since HeavyMethyl can reproducibly analyze the DNA from six cells, the assay should be sufficient for application to clinical samples.
Since only a fraction of the sample DNA will be derived from tumor cells, it is at least equally important that background DNA derived from healthy cells with normal methylation status (typically unmethylated) be reliably blocked from amplification, as false-positive results could be obtained otherwise. In the experiments shown, we were able to block a 1600-fold excess of unmethylated DNA from amplification. This was demonstrated by measuring a dilution series of fully methylated DNA into unmethylated DNA. Even if only 1 in 1600 molecules was methylated, HeavyMethyl exclusively amplified the methylated DNA as confirmed by sequencing of the products. No amplification was observed when applying the assay to completely unmethylated reference DNA. In summary, we have demonstrated that the HeavyMethyl technology meets the requirements of both analytical sensitivity and specificity.
The experiments conducted also demonstrate the clinical applicability of HeavyMethyl for two example genes, GSTP1 and CALCA. Highly significant (P < 0.0001) differences were measured between the PMR of colon adenocarcinoma and normal samples using the HeavyMethyl technology on the Calcitonin gene. Initial data obtained on matched serum samples indicate that the technology can also be used in an early detection assay. Calcitonin is known to encode a peptide hormone that plays a role in maintenance of calcium levels in blood serum, but it is unclear why genomic regions near this gene become methylated in cancer. However, a complete understanding of the biological basis of Calcitonin methylation is not necessary for evaluation of the gene as a biomarker. In future research we intend to test this assay on an expanded set of samples, and possibly combine the assay with other methylation markers to create an optimized epigenetic marker panel for early detection of colon cancer.
In contrast to MSP, the HeavyMethyl technology uses blocker molecules to achieve methylation sensitivity, while the primers alone would equally amplify methylated and unmethylated DNA. The difference in the mechanism of methylation specificity of the HeavyMethyl and MSP assays has important implications for their performance. A false-positive signal (the erroneous amplification of unmethylated DNA) can, in principle, occur with MSP if a primer binds to some extent to the unmethylated template and an extension occurs. If this happens only once in the early PCR cycles, the primer is incorporated into the copy produced and this copy and any complementary strand will, in further cycles, serve as a legitimate template. Therefore, methylation specificity in MSP reactions relies on complete specificity of primer binding and extension during each cycle. The situation is different with a HeavyMethyl reaction since the blocker provides methylation specificity of the amplification in every cycle equally. Should any unmethylated DNA serve as the template due to insufficient binding of the blocker, the blocker will still suppress the further amplification of the extension product in all future cycles. Therefore, one can assume that the methylation specificity of HeavyMethyl will increase with rising cycle numbers, whereas in MSP any signal would be amplified with increasing cycle numbers.
Compared to the more established MSP method, HeavyMethyl offers advantages as to the flexibility of the primer and probe design. It is, for example, possible to design probes for bisulfite-specific primers that are not methylation specific and use the same optimized PCR for both methylation-sensitive and -insensitive amplifications. This is especially advantageous if a high analytical sensitivity is needed and an appropriate PCR fragment has already been optimized. It is also possible to optimize the PCR independent of its methylation specificity, whereas for MSP both properties are linked within one primer design. This will allow the adaptation of sets of multiplexed bisulfite-specific PCRs to methylation sensitivity.
One additional advantage is that HeavyMethyl assays have the necessary specificity and sensitivity with only one blocker per PCR. Only the annealing of one primer to unmethylated templates is blocked. MSP usually requires two methylation sensitive primers and therefore will only be applicable if the methylation status of a genomic region is sufficiently uniform over the designed fragment. The advantage of MSP remains that fewer components have to be used in the assay. Also, very high relative sensitivities have been reported for MSP, and MSP primer design has been achieved successfully for a broad range of genes.
We have developed a new assay for the sensitive and specific analysis of DNA methylation patterns. HeavyMethyl fulfills the requirement for a quantitative and homogenous format, as demonstrated with Taqman and LightCycler real-time amplifications. Data confirm that this assay can be successfully employed for the analysis of very low concentrations of methylated DNA, e.g. in serum of tumor patients, with a high background of unmethylated normal DNA present. The assay was applied to two genes relevant for tumor diagnosis, GSTP1 and CALCA. HeavyMethyl fulfills the performance criteria for future methylation-based diagnostic tests and will be applied to new candidate genes in future work.
REFERENCES
Takai,D. and Jones,P.A. (2002) Comprehensive analysis of CpG islands in human chromosomes 21 and 22. Proc. Natl Acad. Sci. USA, 99, 3740–3745.
Jones,P.A. and Laird,P.W. (1999) Cancer epigenetics comes of age. Nature Genet., 21, 163–167.
Adorjan,P., Distler,J., Lipsche,E., Model,F., Mueller,J., Pelet,C., Braun,A., Florl,A.R., Gutig,D., Grabs,G. et al. (2002) Tumour class prediction and discovery by microarray-based DNA methylation analysis. Nucleic Acids Res., 30, e21.
Costello,J.F., Fruhwald,M.C., Smiraglia,D.J., Rush,L.J., Robertson,G.P., Gao,X., Wright,F.A., Feramisco,J.D., Peltomaki,P., Lang,J.C. et al. (2000) Aberrant CpG-island methylation has non-random and tumour-type-specific patterns. Nature Genet., 24, 132–138.
Esteller,M., Corn,P.G., Baylin,S.B. and Herman,J.G. (2001) A gene hypermethylation profile of human cancer. Cancer Res., 61, 3225–3229.
Usade,H., Brabender,J., Danenberg,K,D,, Jeronimo,C., Harden,S., Engles,J., Danenburg,P.V., Yang,S. and Sidransky,D. (2002) Quantitative adenomatous polyposis coli promoter methylation analysis in tumor tissue, serum, and plasma DNA of patients with lung cancer. Cancer Res., 62, 371–375.
Wong,I.H., Lo,Y.M., Zhang,J., Liew,C.T., Ng,M.H., Wong,N., Lai,P.B., Lau,W.Y., Hjelm,N.M. and Johnson,P.J. (1999) Detection of aberrant p16 methylation in the plasma and serum of liver cancer patients. Cancer Res., 59, 71–73.
Goessl,C., Muller,M., Heicappell,R., Krause,H. and Miller,K. (2001) DNA-based detection of prostate cancer in blood, urine, and ejaculates. Ann. N. Y. Acad. Sci., 945, 51–58.
Zou,H.Z., Yu,B.M., Wang,Z.W., Sun,J.Y., Cang,H., Gao,F., Li,D.H., Zhao,R., Feng,G.G. and Yi,J. (2002) Detection of aberrant p16 methylation in the serum of colorectal cancer patients. Clin. Cancer Res., 8, 188–191.
Olek,A., Oswald,J. and Walter,J. (1996) A modified and improved method for bisulphite based cytosine methylation analysis. Nucleic Acids Res., 24, 5064–5066.
Herman,J.G., Graff,J.R., Myohanen,S., Nelkin,B.D. and Baylin,S.B. (1996) Methylation-specific PCR: a novel PCR assay for methylation status of CpG islands. Proc. Natl Acad. Sci. USA, 93, 9821–9826.
Palmisano,W.A., Divine,K.K., Saccomanno,G., Gilliland,F.D., Baylin,S.B., Herman,J.G. and Belinsky,S.A. (2000) Predicting lung cancer by detecting aberrant promoter methylation in sputum. Cancer Res., 60, 5954–5958.
Eads,C.A., Danenberg,K.D., Kawakami,K., Saltz,L.B., Danenberg,P.V. and Laird,P.W. (1999) CpG island hypermethylation in human colorectal tumors is not associated with DNA methyltransferase overexpression. Cancer Res., 59, 2302–2306.
Eads,C.A., Danenberg,K.D., Kawakami,K., Saltz,L.B., Blake,C., Shibata,D., Danenberg,P.V. and Laird,P.W. (2002) MethyLight: a high-throughput assay to measure DNA methylation. Nucleic Acids Res., 28, e32.
Lo,Y.M., Wong,I.H., Zhang,J., Tein,M.S., Ng,M.H. and Hjelm,N.M. (1999) Quantitative analysis of aberrant p16 methylation using real-time quantitative methylation-specific polymerase chain reaction. Cancer Res., 59, 3899–3903.
Jeronimo,C., Usade,H., Henrique,R., Oliveira,J., Lopes,C., Nelson,W.G. and Sidransky,D. (2001) Quantitation of GSTP1 methylation in non-neoplastic prostatic tissue and organ-confined prostate adenocarcinoma. J. Natl Cancer Inst., 93, 1747–1752.
Sun,X., Hung,K., Wu,L., Sidransky,D. and Guo,B. (2002) Detection of tumor mutations in the presence of excess amounts of normal DNA. Nat. Biotechnol., 20, 186–189.
Prix,L., Uciechowski,P., Bockmann,B., Giesing,M. and Schuetz,A.J. (2002) Diagnostic biochip array for fast and sensitive detection of K-ras mutations in stool. Clin. Chem., 48, 428–435.
Melki,J.R., Vincent,P.C. and Clark,S.J. (1999) Concurrent DNA hypermethylation of multiple genes in acute myeloid leukemia. Cancer Res., 59, 3730–3740.
Silverman,A.L., Park,J.G., Hamilton,S.R., Gazdar,A.F., Luk,G.D. and Baylin,S.B. (1989) Abnormal methylation of the Calcitonin gene in human colonic neoplasms. Cancer Res., 49, 3468–8473.
Wittwer,C.T., Ririe,K.M., Andrew,R.V., David,D.A., Gundry,R.A. and Balis,U.J. (1997) The LightCycler: a microvolume multisample fluorimeter with rapid temperature control. Biotechniques, 22, 176–181.
Holland,P.M., Abramson,R.D., Watson,R. and Gelfand,D.H. (1991) Detection of specific polymerase chain reaction product by utilizing the 5'-3' exonuclease activity of Thermus aquaticus DNA polymerase. Proc. Natl Acad. Sci. USA, 88, 7276–7280.
Sozzi,G., Conte,D., Mariani,L., Lo Vull,S., Roz,L., Lombardo,C., Pierotti,M.A. and Tavecchio,L. (2001) Analysis of circulating tumor DNA in plasma at diagnosis and during follow-up of lung cancer patients. Cancer Res., 61, 4675–4678.
Jahr,S., Hentze,H., Englisch,S., Hardt,D., Fackelmayer,F.O., Hesch,R.D. and Knippers,R. (2001) DNA fragments in the blood plasma of cancer patients: quantitations and evidence for their origin from apoptotic and necrotic cells. Cancer Res., 61, 1659–1665.
Cottrell,S.E. and Laird,P.W. (2003) Sensitive detection of DNA methylation. Ann. N. Y. Acad. Sci., 983, 120–130.(Susan E. Cottrell, Jürgen Distler1, Nanc)
*To whom correspondence should be addressed. Tel: +49 30 2434 5303; Fax: +49 30 2434 5299; Email: berlin@epigenomics.com
ABSTRACT
DNA methylation-based biomarkers have been discovered that could potentially be used for the diagnosis of cancer by detection of circulating, tumor-derived DNA in bodily fluids. Any methylation detection assay that would be applied to these samples must be capable of detecting small amounts of tumor DNA in the presence of background normal DNA. We have developed a real-time PCR assay, called HeavyMethyl, that is well suited for this application. HeavyMethyl uses methylation-specific oligonucleotide blockers and a methylation-specific probe to achieve methylation-specific amplification and detection. We tested the assays on unmethylated and artificially methylated DNA in order to determine the limit of detection. After careful optimization, our glutathione-S-transferase pi1 and Calcitonin assays can amplify as little as 30 and 60 pg of methylated DNA, respectively, and neither assay amplifies unmethylated DNA. The Calcitonin assay showed a highly significant methylation difference between normal colon and colon adenocarcinomas, and methylation was also detected in serum DNA from colon cancer patients. These assays show that HeavyMethyl technology can be successfully employed for the analysis of very low concentrations of methylated DNA, e.g. in serum of patients with tumors.
INTRODUCTION
CpG dinucleotides are underrepresented in the genome and appear at only 20% of the expected frequency. In regions referred to as CpG islands, the frequency of this dinucleotide is much higher, approaching or exceeding the statistically expected frequency of 1 in 16 dinucleotides. Typically these CpG islands are found near genes and stretch over 500–2000 bp (1). In normal tissues, most cytosines in the context of CpG dinucleotides within CpG islands are unmethylated, but some methylation occurs in imprinted genes and inactive X-chromosomal genes. In addition, somatic de novo methylation of CpG islands can occur in genes involved in controlling cell growth and division. This aberrant methylation is often associated with reduced transcription of the associated gene, and CpG island hypermethylation is a common mechanism of gene inactivation during tumorigenesis (2).
Specific patterns of DNA methylation have been found to characterize different cancer tissues (3–5). Methylation of the glutathione-S-transferase pi 1 (GSTP1) gene for example, has been found in prostate, breast and liver tumors but not bladder, stomach or head and neck tumors (5). Due to the high rates of methylation and the stability of the signal, many of these methylated genes have been investigated as potential biomarkers for early detection of cancer, and preliminary studies suggest that some of these markers may even be useful for detection of tumor-derived DNA in diagnostic samples such as urine, serum and plasma (6–9). Since there is normally a limited amount of total DNA in these types of samples and only a fraction of this DNA is derived from tumor cells, this application relies on methylation detection assays with high analytical sensitivities. Also, the detection assay must have a high relative sensitivity that favors the tumor DNA over the background DNA, in order to avoid false-positive results when a high amount of non-tumor DNA is present in the sample.
In addition to sensitivity, an early detection assay based on methylation markers must meet other requirements. All components of the assay will, after bisulfite treatment, ideally be combined in one tube in order to minimize the amount of labor as well as the risk of operator error and sample contamination. Furthermore, since methylation is not uniform within a heterogeneous DNA sample, the assay must be able to quantitatively detect the aberrantly methylated DNA in order to provide the maximum amount of information.
The first and most common assay applied to sensitive detection of methylation in tissues and body fluids is called methylation-specific PCR (MSP; 8). For a given DNA sample, treatment with a bisulfite reagent leads to the conversion of all unmethylated cytosines to uracils, leaving only methylated cytosines unchanged (3,10). MSP is now carried out analogously to an allele-specific PCR reaction. The primers bind either to methylated or unmethylated DNA only and therefore selectively amplify only DNA with a defined methylation. In contrast to the allele-specific PCR used for single nucleotide polymorphism (SNP) analysis, MSP primer binding sites include several CpG sites within bisulfite-treated DNA, so that binding and extension are methylation specific. MSP primers are usually designed to detect targets that are either methylated at all of the CpG sites in the binding site or unmethylated at all CpGs. Users of this assay have reported high sensitivity (11,12), and the widespread application of MSP has enabled an increase in research on methylation. Addition of a methylation-specific real-time detection probe (MethyLight) makes the assay both homogenous and quantitative (12–15). The quantitative nature of MethyLight was particularly useful in recently published research distinguishing benign prostate hypertrophy from prostate cancer based on GSTP1 methylation (16).
This paper introduces a novel real-time methylation assay called HeavyMethyl. Like MSP, the priming is methylation specific, but non-extendable oligonucleotide blockers provide this specificity instead of the primers themselves. The blockers bind to bisulfite-treated DNA in a methylation-specific manner, and their binding sites overlap the primer binding sites (Fig. 1). When the blocker is bound, the primer cannot bind and therefore the amplicon is not generated. When the blocker is not bound, the primer-binding site is accessible and the amplicon is generated. Similar blocking approaches have been used to achieve high specificity in primer extension and chip assays for mutation detection (17,18), but so far blocker oligonucleotides have not been successfully employed in methylation analysis. As a proof of concept, we have designed assays for two genes: GSTP1, which is frequently methylated in prostate cancer (16), and Calcitonin (CALCA), which is hypermethylated in multiple types of cancer (19,20). Here we demonstrate that HeavyMethyl in combination with real-time detection with methylation-specific fluorogenic probes is capable of sensitive and specific detection of DNA methylation. Furthermore, our initial application of these assays to tumor and serum samples suggests that this format may be suitable for clinical studies.
Figure 1. Principle of HeavyMethyl. (A) When the DNA is methylated, the blocker oligonucleotides (solid black) do not bind, leaving the primer- binding site accessible for the primers (gray arrows) to bind and amplify the target. The amplification is detected with a methylation specific oligonucleotide probe . (B) When the DNA is unmethylated, the blocker oligonucleotides bind, blocking the access of the primers to their binding sites. No PCR product is generated.
MATERIALS AND METHODS
GSTP1 assay
PCR amplification of a GSTP1Exon fragment (nucleotides 1241–1394 in GenBank M24485 .1) was performed in a 25 μl reaction volume with 500 nM GSTP1Exon forward primer (GTTTTYGTTATTAGTGAGT), 500 nM GSTP1Exon reverse primer (TCCTAAATCCCCTAAACC), 0.2 mM dNTPs, 10 μM GSTP1Exon blocker 1 (GTGAGTATGTGTGGTTTGTGT-phos; where phos = phosphorylated), 10 μM GSTP1Exon blocker 2 (TAAACCCCCATCCCAAATCTCA-phos), 1x Qiagen reaction buffer, 1.5 mM MgCl2, and 1 U Hotstar Polymerase (Qiagen). Thermocycling conditions were an initial step at 95°C (15 min), then 45 cycles of the following steps: 95°C (45 s), 52°C (45 s) and 72°C (20 s). After an additional step of 10 min at 72°C, 10 μl of the PCR reaction was analyzed on a 1.5% agarose gel.
Real-time PCR experiments were performed using a LightCycler (Roche Diagnostics; 21). The 20 μl PCR reactions contained 0.31 μM GSTP1Exon forward primer, 0.31 μM GSTP1Exon reverse primer, 0.25 g/l bovine serum albumin (BSA, Sigma), 0.25 mM dNTPs, 0.25 μM GSTP1Exon anchor probe (GTTTAGAGTTTTTAGTATGGGGTTAATT-Fluo; where Fluo = fluorescein), 0.25 μM GSTP1Exon probe I, specific for methylated bisulfite-treated DNA (LCRed640-TAGTATTAGGTTCGGGTTTTCGG-phos) or probe II, specific for unmethylated bisulfite-treated DNA (LCRed705-TAGTATTAGGTTTGGGTTTTTGG-phos; all probes obtained from TIB-Molbiol), 10 μM GSTP1Exon blocker 3 oligonucleotide (GTGAGTATGTGTGGTTTGTGTT-phos), 1x Qiagen reaction buffer, 1.5 mM MgCl2 and 1 U Hotstar polymerase. A total of 0.1 or 100 ng of bisulfite-treated template DNA was used. For melting-peak analysis of GSTP1Exon in the PCR, SYBR Green I was used instead of specific probes. Thermocycling conditions were an initial step at 95°C (15 min), then 50 cycles of the following steps: 95°C (4 s), 52°C (30 s) with a single fluorescence detection after 10 s in each cycle, and 72°C (20 s). Melt peaks were obtained by plotting the negative derivate of SYBR Green I fluorescence with respect to temperature (–dF/dT) against temperature.
The HeavyMethyl assay of the GSTP1promoter (Prom) fragment (GenBank M24485 .1, position 1014–1171) was performed in 20 μl using the LightCycler. The real-time PCR reaction mix contained 0.30 μM GSTP1Prom forward primer (GGGAAAGAGGGAAAGGTTTTTT), 0.30 μM GSTP1Prom reverse primer (CCCCAATACTAAATCAC), 0.25 g/l BSA (Sigma), 0.2 mM dNTPs, 0.15 μM GSTP1Prom anchor probe (CGGTCGACGTTCGGG-Fluo), 0.15 μM GSTP1Prom detection probe (LCRed640-TAGCGGtCGtCGGGG-phos), 10 μM GSTP1Prom blocker (CTAAATCACAACACCAACCACTCTTC-phos), 1x Qiagen reaction buffer, 1.75 mM MgCl2 and 2 U Hotstar polymerase (Qiagen). Both GSTP1Prom anchor and detection probes were specific for methylated bisulfite-treated template DNA. Thermocycling conditions began with an initial step at 95°C (15 min), then 55 cycles of the following steps: 95°C (10 s), 48°C (30 s) and 72°C (20 s). Fluorescence was detected after the annealing phase at 48°C in each cycle. Underlined nucleotides indicate the overlap between primer and blocker oligonucleotides.
Calcitonin assay
The Calcitonin assay is a 5' exonuclease assay (22) that targets a region of the genome on 11p15 1200 bp upstream from the transcription start site (GenBank X15943 , position 476–728) with primers (GGATGTGAGAGTTGTTGAGGTTA and (ACACACCCAAACCCATTACTATCT; Integrated DNA Technologies). The blocker (TGTTGAGGTTATGTGTAATTGGGTGTGA-phos) overlaps the forward primer sequence by 11 bp, spans two CpGs, and then covers another 13 bp. In addition to the 5' fluorescein label, the probe Fluo-ACCTCCGAATCTCTCGAACGATCGC-BHQ is labeled with a Black Hole Quencher (BHQ) at the 3'-end (Biosearch Technologies).
The Calcitonin assay was run with 300 nM primers, 5 μM blocker, 450 nM probe, 2 U Amplitaq Gold, 3.5 mM MgCl2 and 0.4 mM dNTPs (all Applied Biosystems) in a final reaction volume of 20 μl. First, the reactions were held at 95°C for 10 min to activate the enzyme. Then, we used two-step cycling to amplify the DNA (95°C for 15 s, 67°C for 60 s and repeat three times; 95°C for 15 s, 64°C for 60 s and repeat three times; 95°C, 15 s, 62°C for 60 s and repeat three times; 95°C, 15 s, 60°C for 60 s and repeat 40 times). Data were collected during the extension phase at 60°C on an ABI Prism 7900 instrument.
A positive control reaction for ?-actin was also used to control for input DNA when analyzing the colon samples. This assay has been described previously (20). The same cycling conditions were used as for the Calcitonin assay.
DNA preparation and bisulfite treatment
Colon samples from patients (from 21 to 91 years old) with stage I, II or II colon carcinoma were obtained during surgery (approval from Western Institutional Review Board). DNA was isolated from 100–500 mg of tissue using Genomic Tip 500 kit (Qiagen). DNA was isolated from 1 ml of serum collected before surgery using UltraSens Viral Nucleic Acids kit (Qiagen).
Two sources of unmethylated control DNA were used. For the Calcitonin analysis, sperm DNA was isolated from semen using a protocol supplied by Qiagen. For the GSTP1 analysis, unmethylated DNA from human peripheral lymphocytes (Promega) was used. The methylation status of the lymphocyte DNA was confirmed by bisulfite sequencing. Methylated DNA was created by treating genomic DNA (Promega) with SssI methyltransferase (New England Biolabs) in the presence of a methyl donor S-adenosyl-methionine according to the manufacturer’s instructions.
For bisulfite treatment the denatured DNA was embedded in a 2% agarose bead and was further processed as described earlier (10). Between 5 and 10 μl of this DNA was used per reaction.
Data analysis
Percent Methylated Reference (PMR) was calculated using a procedure previously described (20). The Cts (cycle thresholds) for all samples were converted to a relative quantity scale using a standard curve of diluted methylated DNA. The ratio of the ‘quantity’ of Calcitonin-methylated DNA to total ?-actin DNA is calculated for all samples of interest as well as the fully methylated control DNA. The ratio of these ratios is the PMR. A Mann–Whitney test was used for all statistical tests of significance. ROC (Receiver Operating Characteristic) curves were generated by ranking all of the tissue samples by their PMR values. The tumor sample population overlaps the normal sample population, so a clear threshold could not be determined. The sensitivity and false-positive rate were calculated for each possible cut-off point between the normal and tumor classes and these values were plotted. As the sensitivity increases, the specificity decreases. The best markers will have very little reduction in specificity as sensitivity increases, so the area under the ROC curves (AUC) will be close to 1. A marker with no predictive value will have an AUC of 0.5.
RESULTS
Methylation-specific blocking oligonucleotides prevent amplification of unmethylated DNA
The GSTP1 gene exon 1 is frequently methylated in prostate cancer (4). The GSTP1Exon primers used in our HeavyMethyl assay amplify a 153 bp fragment. In order to prevent amplification of unmethylated DNA, different blocker oligonucleotides were designed to overlap the 3'-end of the binding site of the primer as well as downstream CpGs positions. Therefore, the blocker binds to the target only when the DNA is unmethylated at these CpGs, allowing the primer to bind only when the DNA is partially or fully methylated.
We investigated the effect of the following parameters on the performance of our GSTP1Exon assay: number of blockers (one per assay or one per primer), the melting temperature of the blockers, the melting temperature difference between blocker annealing to the unmethylated and methylated target DNA, the number of CpGs within the blockers, the number of overlapping nucleotides between primers and blockers, and the ratio of the concentrations of primers and blockers. (data not shown).
For the final GSTP1Exon HeavyMethyl assay, the blockers 1 and 2 cover five and three CpGs, respectively. A primer/blocker ratio of 1:20 repressed the GSTP1Exon amplification on 50 ng of unmethylated DNA, but amplification of 0.075 ng methylated template DNA was not significantly inhibited (Fig. 2A, lanes 7 and 12). The GSTP1Exon HeavyMethyl assay detected 100 pg methylated template DNA in a background of 100 ng unmethylated DNA, representing a relative analytical sensitivity of 1:1000. The products obtained with and without blocker were sequenced to confirm their methylated and unmethylated status, respectively (Fig. 2B). The effect of the blocker is seen more clearly with melt peak analysis on a LightCycler (Fig. 2C). GSTP1Exon products of methylated and unmethylated bisulfite-treated template DNA generate traces with melting peaks of 84 and 79°C, respectively. On template DNA representing a mixture of 100 pg SssI methylated and 100 ng unmethylated bisulfite-treated DNA, only methylated GSTP1Exon products were generated by the HeavyMethyl assay in the presence of blocker. Without blocker only unmethylated GSTP1Exon PCR products were found.
Figure 2. Relative and absolute sensitivity of GSTP1 HeavyMethyl assay. GSTP1Exon amplification was performed with and without blockers. (A) 1.5% agarose gel analysis of GSTP1Exon PCRs performed without (–B; lanes 1–6, 13) and with blocker (+B; lanes 7–12) on 50 ng bisulfite-treated unmethylated DNA (–CH3; lanes 1, 7), different amounts of bisulfite-treated SssI-methylated DNA (CH3; 1.2 ng, lanes 2 and 8; 0.6 ng, lanes 3 and 9; 0.3 ng, lanes 4 and 10; 0.15 ng, lanes 5 and 11; 75 pg, lanes 6 and 12) and no DNA ( lane 13). (B) Sequence traces of GSTP1Exon products generated on a mixture of 100 pg methylated in a background of 100 ng unmethylated template DNA (relative sensitivity 1:1000) in the presence and absence of blockers. (C) The discrimination of GSTP1Exon products generated by HeavyMethyl assay with (+B) and without (–B) blocker on bisulfite-treated SssI-methylated (+CH3), unmethylated (–CH3) and a mixture of 100 pg methylated in 100 ng unmethylated (1:1000) template DNA by melting peak analysis.
Real-time HeavyMethyl assays are sensitive and specific
A gel-based detection of HeavyMethyl amplification allows us to visualize the effect of blockers on amplification of unmethylated DNA. However, real-time detection of amplification is preferable because the assay is both quantitative and homogenous. For the probe-based real-time HeavyMethyl assay, blocker 3 sufficiently blocked the amplification of the GSTP1Exon fragment using up to 50 ng unmethylated DNA (Fig. 3B): in the presence of blocker no product was detectable with the detection probe II designed for unmethylated DNA. In addition, the PCR with 15 ng bisulfite-treated methylated template DNA was not affected by the blocker (Fig. 3A). Figure 3A also shows that the detection probe I specifically detected methylated DNA, whereas unmethylated DNA amplified either with or without blocker was not identified. The analytical sensitivity of this HeavyMethyl assay was 30 pg (six genome equivalents, data not shown). Furthermore, 30 pg of methylated DNA was reproducibly detectable in a background of 50 ng unmethylated DNA with the HeavyMethyl assay (Fig. 3C). In contrast, the assay without blocker did not detect the spiked methylated DNA (Fig. 3C). This HeavyMethyl assay showed a relative sensitivity of 1:1600.
Figure 3. HeavyMethyl real-time assays for GSTP1 exon and promoter. (A–C) HeavyMethyl GSTP1 exon assay. Amplification plots of 15 ng SssI-methylated DNA indicated by solid or broken lines and 50 ng of unmethylated DNA indicated by lines with circles (A). (B) 50 ng of unmethylated DNA and 30 pg methylated DNA spiked into 50 ng unmethylated DNA representing a relative sensitivity of 1:1600 (C), which were bisulfite treated and subsequently used as template for the HeavyMethyl assay. The LightCycler detection probes were specific for bisulfite-treated unmethylated (B) or methylated DNA (A, C). HeavyMethyl assays performed with and without blocker are indicated by solid and broken lines, respectively. In (A), the HeavyMethyl assay with and without blocker is also indicated by filled and open circles, respectively. (D) HeavyMethyl GSTP1 promoter assay. The GSTP1 promoter fragment was amplified on 400 ng unmethylated (five replicates indicated by broken lines), 100 pg SssI-methylated DNA (black diamonds) and mixtures containing 100 pg (six replicates indicated by solid lines) or 50 pg (six replicates indicated by lines labeled with open diamonds) methylated DNA in 400 ng unmethylated DNA. All template DNA samples were bisulfite treated prior to real-time PCR experiments. The products were specifically detected via the FRET signal generated by the Fluorescein- and LCRed640-labeled detection probes specific for methylated bisulfite-treated DNA.
The second GSTP1 HeavyMethyl assay, established on the promoter region, showed an even better relative sensitivity. In Figure 3D, this HeavyMethyl assay reproducibly detected methylated copies of a GSTP1Prom fragment in a background of 400 ng unmethylated bisulfite-treated human DNA. This real-time HeavyMethyl assay showed a relative sensitivity of up to 1:8000. In addition, products were generated on 25 and 12.5 pg methylated bisulfite-treated template DNA in 16 out of 16 and 9 out of 16 real-time PCRs, respectively (data not shown).
To determine whether HeavyMethyl assays can be designed for an additional real-time PCR format, we created a 5' exonuclease real-time assay for a CpG island upstream of the Calcitonin gene. In previous reports, CpGs near the transcription start site of the gene have been shown to be methylated in colon cancer and leukemia (17,18). We designed an assay for a CpG island further upstream. The blocker oligonucleotide targets two CpGs just beyond the 3' end of the forward primer, and the blocker binding site overlaps the primer binding site by 11 bp. Since the probe-binding site contains four CpGs, the binding of the probe is also methylation specific.
The Calcitonin HeavyMethyl real-time assay amplified methylated target in the presence of 500-fold excess unmethylated DNA, and in these wells the total amount of methylated DNA is estimated to be <60 pg. In contrast, there was never any significant amplification of unmethylated DNA. Series dilutions of bisulfite-treated methylated into bisulfite-treated unmethylated DNA showed that the assay is linear over at least three orders of magnitude and there is little deviation of the points from the line (R2 = 0.981). The slope, –3.30 (standard error = 0.263), suggests that this assay is highly efficient and the amount of target doubles with each cycle (data not shown).
HeavyMethyl analysis of clinical samples
We used the Calcitonin assay to detect methylation in colon samples to demonstrate that HeavyMethyl assays yield clinically relevant information. We extracted and bisulfite-treated DNA from colon adenocarcinoma samples and normal matched colon tissue. The methylation was measured with our Calcitonin assay, and the total amount of input DNA was measured with a control assay for ?-actin. The control assay does not contain CpGs and therefore amplifies all bisulfite-converted DNA regardless of methylation status. For each sample, the ratio of the quantity of methylated Calcitonin divided by the total amount of DNA was normalized to the same ratio calculated for artificially methylated DNA, the PMR (20).
The mean PMR of the adenocarcinoma samples (n = 33) was 4.76, while the mean PMR for the normal samples (n = 43) was only 0.05. This methylation difference is highly significant (P < 0.0001). There was no effect of age or disease stage on methylation levels in either the normal or tumor samples (Fig. 4A).
Figure 4. Methylation of Calcitonin in colon cancer tissue and serum samples. (A) The PMR is plotted for colon cancer tumor and normal colon tissue samples. The PMR is also plotted for five serum samples from individuals with methylated tumors and 11 serum samples from healthy control donors. (B) Methylation of Calcitonin in normal (n = 43) and adenocarcinoma (n = 33) colon samples was measured by both HeavyMethyl and MSP. ROC curves were generated by plotting the sensitivity against the false-negative rate (1 – specificity). The area under the curve (0.84 for HeavyMethyl and 0.81 for MSP) is used to evaluate the performance of the assay.
Performance of the assay was compared to an MSP-based MethyLight assay covering the same genomic region, using the same CpGs for a methylation-specific probe. The two assays showed similar sensitivity and specificity with the same set of samples. The mean PMR values of the MSP-based MethyLight assay (0.19 for healthy and 5.04 for tumor samples) were very similar to the values for the HeavyMethyl assay (Fig. 4A). Since the two assays do not cover exactly the same CpG positions, methylation biology will strongly influence the results. Figure 4B shows the ROC curves for the HeavyMethyl and MSP MethyLight assays for the Calcitonin gene. The area under the curve is 0.84 for the HeavyMethyl assay and 0.81 for the MSP MethyLight assay. The high analytical sensitivity of our assay suggests that it is suitable for application to serum analysis. We collected serum samples from 11 healthy individuals and found minimal detectable methylation in these samples. The mean PMR was 0.028 ± 0.025. Five of our methylated colon adenocarcinoma samples had matched serum samples taken before surgery, and all five of these had methylation levels at least 6 standard deviations above the mean of the healthy population (Fig. 4A). While the sample size is small in this experiment, the data suggest that HeavyMethyl assays have the sensitivity and specificity necessary for a serum-based application.
DISCUSSION
We have presented data confirming that the introduced HeavyMethyl methodology has high potential for the analysis of DNA methylation in clinical samples. A successful assay format for the detection of methylation in body fluids must fulfill several criteria. Most important is analytical sensitivity: in a given volume of a body fluid, only very limited amounts of DNA may be present. In healthy patients, for example, a typical amount of non cell-bound, free-floating DNA within a milliliter of serum is 10–50 ng (23,24). Although there is evidence that this amount may be increased in tumor patients because of apoptosis or still unknown mechanisms, the amount of DNA that is available for analysis is typically even smaller because of low extraction yields, the small average size of free-floating DNA, DNA degradation during bisulfite conversion and, most importantly, the varying ratio of tumor to healthy free-floating DNA (23–25).
Ideally, an assay would therefore have the sensitivity and specificity to detect a single copy of methylated DNA within a given volume of body fluid. In our hands, we were able to design and conduct PCR reactions that reproducibly amplify as little as 30 pg (DNA from six cells) of bisulfite-converted, previously methylated DNA, with an excess of blocker molecules binding to unmethylated DNA. The presence of blocker molecules did not influence the analytical sensitivity of the PCR reaction significantly. Thirty picograms is far below the expected amounts of tumor DNA in symptomatic cancer patients, as determined, for example, for lung and colon cancer patients (50–500 ng/ml serum; 23–25); however, one would like to be able to detect as few aberrantly methylated DNA molecules as possible for an early cancer screening test. Since HeavyMethyl can reproducibly analyze the DNA from six cells, the assay should be sufficient for application to clinical samples.
Since only a fraction of the sample DNA will be derived from tumor cells, it is at least equally important that background DNA derived from healthy cells with normal methylation status (typically unmethylated) be reliably blocked from amplification, as false-positive results could be obtained otherwise. In the experiments shown, we were able to block a 1600-fold excess of unmethylated DNA from amplification. This was demonstrated by measuring a dilution series of fully methylated DNA into unmethylated DNA. Even if only 1 in 1600 molecules was methylated, HeavyMethyl exclusively amplified the methylated DNA as confirmed by sequencing of the products. No amplification was observed when applying the assay to completely unmethylated reference DNA. In summary, we have demonstrated that the HeavyMethyl technology meets the requirements of both analytical sensitivity and specificity.
The experiments conducted also demonstrate the clinical applicability of HeavyMethyl for two example genes, GSTP1 and CALCA. Highly significant (P < 0.0001) differences were measured between the PMR of colon adenocarcinoma and normal samples using the HeavyMethyl technology on the Calcitonin gene. Initial data obtained on matched serum samples indicate that the technology can also be used in an early detection assay. Calcitonin is known to encode a peptide hormone that plays a role in maintenance of calcium levels in blood serum, but it is unclear why genomic regions near this gene become methylated in cancer. However, a complete understanding of the biological basis of Calcitonin methylation is not necessary for evaluation of the gene as a biomarker. In future research we intend to test this assay on an expanded set of samples, and possibly combine the assay with other methylation markers to create an optimized epigenetic marker panel for early detection of colon cancer.
In contrast to MSP, the HeavyMethyl technology uses blocker molecules to achieve methylation sensitivity, while the primers alone would equally amplify methylated and unmethylated DNA. The difference in the mechanism of methylation specificity of the HeavyMethyl and MSP assays has important implications for their performance. A false-positive signal (the erroneous amplification of unmethylated DNA) can, in principle, occur with MSP if a primer binds to some extent to the unmethylated template and an extension occurs. If this happens only once in the early PCR cycles, the primer is incorporated into the copy produced and this copy and any complementary strand will, in further cycles, serve as a legitimate template. Therefore, methylation specificity in MSP reactions relies on complete specificity of primer binding and extension during each cycle. The situation is different with a HeavyMethyl reaction since the blocker provides methylation specificity of the amplification in every cycle equally. Should any unmethylated DNA serve as the template due to insufficient binding of the blocker, the blocker will still suppress the further amplification of the extension product in all future cycles. Therefore, one can assume that the methylation specificity of HeavyMethyl will increase with rising cycle numbers, whereas in MSP any signal would be amplified with increasing cycle numbers.
Compared to the more established MSP method, HeavyMethyl offers advantages as to the flexibility of the primer and probe design. It is, for example, possible to design probes for bisulfite-specific primers that are not methylation specific and use the same optimized PCR for both methylation-sensitive and -insensitive amplifications. This is especially advantageous if a high analytical sensitivity is needed and an appropriate PCR fragment has already been optimized. It is also possible to optimize the PCR independent of its methylation specificity, whereas for MSP both properties are linked within one primer design. This will allow the adaptation of sets of multiplexed bisulfite-specific PCRs to methylation sensitivity.
One additional advantage is that HeavyMethyl assays have the necessary specificity and sensitivity with only one blocker per PCR. Only the annealing of one primer to unmethylated templates is blocked. MSP usually requires two methylation sensitive primers and therefore will only be applicable if the methylation status of a genomic region is sufficiently uniform over the designed fragment. The advantage of MSP remains that fewer components have to be used in the assay. Also, very high relative sensitivities have been reported for MSP, and MSP primer design has been achieved successfully for a broad range of genes.
We have developed a new assay for the sensitive and specific analysis of DNA methylation patterns. HeavyMethyl fulfills the requirement for a quantitative and homogenous format, as demonstrated with Taqman and LightCycler real-time amplifications. Data confirm that this assay can be successfully employed for the analysis of very low concentrations of methylated DNA, e.g. in serum of tumor patients, with a high background of unmethylated normal DNA present. The assay was applied to two genes relevant for tumor diagnosis, GSTP1 and CALCA. HeavyMethyl fulfills the performance criteria for future methylation-based diagnostic tests and will be applied to new candidate genes in future work.
REFERENCES
Takai,D. and Jones,P.A. (2002) Comprehensive analysis of CpG islands in human chromosomes 21 and 22. Proc. Natl Acad. Sci. USA, 99, 3740–3745.
Jones,P.A. and Laird,P.W. (1999) Cancer epigenetics comes of age. Nature Genet., 21, 163–167.
Adorjan,P., Distler,J., Lipsche,E., Model,F., Mueller,J., Pelet,C., Braun,A., Florl,A.R., Gutig,D., Grabs,G. et al. (2002) Tumour class prediction and discovery by microarray-based DNA methylation analysis. Nucleic Acids Res., 30, e21.
Costello,J.F., Fruhwald,M.C., Smiraglia,D.J., Rush,L.J., Robertson,G.P., Gao,X., Wright,F.A., Feramisco,J.D., Peltomaki,P., Lang,J.C. et al. (2000) Aberrant CpG-island methylation has non-random and tumour-type-specific patterns. Nature Genet., 24, 132–138.
Esteller,M., Corn,P.G., Baylin,S.B. and Herman,J.G. (2001) A gene hypermethylation profile of human cancer. Cancer Res., 61, 3225–3229.
Usade,H., Brabender,J., Danenberg,K,D,, Jeronimo,C., Harden,S., Engles,J., Danenburg,P.V., Yang,S. and Sidransky,D. (2002) Quantitative adenomatous polyposis coli promoter methylation analysis in tumor tissue, serum, and plasma DNA of patients with lung cancer. Cancer Res., 62, 371–375.
Wong,I.H., Lo,Y.M., Zhang,J., Liew,C.T., Ng,M.H., Wong,N., Lai,P.B., Lau,W.Y., Hjelm,N.M. and Johnson,P.J. (1999) Detection of aberrant p16 methylation in the plasma and serum of liver cancer patients. Cancer Res., 59, 71–73.
Goessl,C., Muller,M., Heicappell,R., Krause,H. and Miller,K. (2001) DNA-based detection of prostate cancer in blood, urine, and ejaculates. Ann. N. Y. Acad. Sci., 945, 51–58.
Zou,H.Z., Yu,B.M., Wang,Z.W., Sun,J.Y., Cang,H., Gao,F., Li,D.H., Zhao,R., Feng,G.G. and Yi,J. (2002) Detection of aberrant p16 methylation in the serum of colorectal cancer patients. Clin. Cancer Res., 8, 188–191.
Olek,A., Oswald,J. and Walter,J. (1996) A modified and improved method for bisulphite based cytosine methylation analysis. Nucleic Acids Res., 24, 5064–5066.
Herman,J.G., Graff,J.R., Myohanen,S., Nelkin,B.D. and Baylin,S.B. (1996) Methylation-specific PCR: a novel PCR assay for methylation status of CpG islands. Proc. Natl Acad. Sci. USA, 93, 9821–9826.
Palmisano,W.A., Divine,K.K., Saccomanno,G., Gilliland,F.D., Baylin,S.B., Herman,J.G. and Belinsky,S.A. (2000) Predicting lung cancer by detecting aberrant promoter methylation in sputum. Cancer Res., 60, 5954–5958.
Eads,C.A., Danenberg,K.D., Kawakami,K., Saltz,L.B., Danenberg,P.V. and Laird,P.W. (1999) CpG island hypermethylation in human colorectal tumors is not associated with DNA methyltransferase overexpression. Cancer Res., 59, 2302–2306.
Eads,C.A., Danenberg,K.D., Kawakami,K., Saltz,L.B., Blake,C., Shibata,D., Danenberg,P.V. and Laird,P.W. (2002) MethyLight: a high-throughput assay to measure DNA methylation. Nucleic Acids Res., 28, e32.
Lo,Y.M., Wong,I.H., Zhang,J., Tein,M.S., Ng,M.H. and Hjelm,N.M. (1999) Quantitative analysis of aberrant p16 methylation using real-time quantitative methylation-specific polymerase chain reaction. Cancer Res., 59, 3899–3903.
Jeronimo,C., Usade,H., Henrique,R., Oliveira,J., Lopes,C., Nelson,W.G. and Sidransky,D. (2001) Quantitation of GSTP1 methylation in non-neoplastic prostatic tissue and organ-confined prostate adenocarcinoma. J. Natl Cancer Inst., 93, 1747–1752.
Sun,X., Hung,K., Wu,L., Sidransky,D. and Guo,B. (2002) Detection of tumor mutations in the presence of excess amounts of normal DNA. Nat. Biotechnol., 20, 186–189.
Prix,L., Uciechowski,P., Bockmann,B., Giesing,M. and Schuetz,A.J. (2002) Diagnostic biochip array for fast and sensitive detection of K-ras mutations in stool. Clin. Chem., 48, 428–435.
Melki,J.R., Vincent,P.C. and Clark,S.J. (1999) Concurrent DNA hypermethylation of multiple genes in acute myeloid leukemia. Cancer Res., 59, 3730–3740.
Silverman,A.L., Park,J.G., Hamilton,S.R., Gazdar,A.F., Luk,G.D. and Baylin,S.B. (1989) Abnormal methylation of the Calcitonin gene in human colonic neoplasms. Cancer Res., 49, 3468–8473.
Wittwer,C.T., Ririe,K.M., Andrew,R.V., David,D.A., Gundry,R.A. and Balis,U.J. (1997) The LightCycler: a microvolume multisample fluorimeter with rapid temperature control. Biotechniques, 22, 176–181.
Holland,P.M., Abramson,R.D., Watson,R. and Gelfand,D.H. (1991) Detection of specific polymerase chain reaction product by utilizing the 5'-3' exonuclease activity of Thermus aquaticus DNA polymerase. Proc. Natl Acad. Sci. USA, 88, 7276–7280.
Sozzi,G., Conte,D., Mariani,L., Lo Vull,S., Roz,L., Lombardo,C., Pierotti,M.A. and Tavecchio,L. (2001) Analysis of circulating tumor DNA in plasma at diagnosis and during follow-up of lung cancer patients. Cancer Res., 61, 4675–4678.
Jahr,S., Hentze,H., Englisch,S., Hardt,D., Fackelmayer,F.O., Hesch,R.D. and Knippers,R. (2001) DNA fragments in the blood plasma of cancer patients: quantitations and evidence for their origin from apoptotic and necrotic cells. Cancer Res., 61, 1659–1665.
Cottrell,S.E. and Laird,P.W. (2003) Sensitive detection of DNA methylation. Ann. N. Y. Acad. Sci., 983, 120–130.(Susan E. Cottrell, Jürgen Distler1, Nanc)