COMPARE-MS overview and rationale. Genomic DNA is digested with AluI with or without the methylation-sensitive restriction enzyme HpaII. After digestion, either the MBD2-MBD captured methylated DNA or all digested DNA are subjected to real-time PCR at a gene-specific locus. We hypothesized that enrichment of methylated DNA by methylation-sensitive restriction enzyme digestion alone or by MBD2-MBD capture of methylated DNA alone may result in false positives associated with incomplete digestion or nonspecific capture, respectively, while the combination of the two approaches (COMPARE-MS) would maintain sensitivity while minimizing false-positive results.

Determination of MBD2-MBD affinity for symmetrically methylated hairpin olignucleotide ligands. (A–C) Fluorescence anisotropy measurements were plotted as a function of MBD2-MBD concentration in order to estimate the relative affinity of MBD2-MBD for fluorescently labeled hairpin oligonucleotides containing three pairs (A), two pairs (B) and one pair (C) of symmetrically methylated CpG dinucleotides. The EC₅₀, defined as the MBD2-MBD concentration required to achieve half maximal binding of 10 nM hairpin oligonucleotides, is shown for each case, along with the corresponding SEM. Data shown represent mean ± SEM for triplicate measurements.

Validation of COMPARE-MS by analysis of hypermethylation at six gene-specific CGIs in multiple prostate cell lines. (A) Methylation index (MI), defined as the ratio of the amount of methylated alleles in a given sample to the amount of methylated alleles in the same input quantity of M.SssI-treated WBC DNA, as determined by COMPARE-MS, for six cancer-related genes (GSTP1, PTGS2, MDR1, ESR1, DAPK1 and CDH1) in 20 ng of genomic DNA from six prostate cancer cell lines (LNCaP, C42B, PC-3, DU-145, LAPC-4 and CWR22Rv1), one primary culture model of non-malignant prostate epithelial cells (PrEC) and untreated WBC negative control. (B) MI for the same set of CGIs and samples as determined by MethyLight in a previous study (33). With few exceptions, the CGI hypermethylation pattern obtained from COMPARE-MS (A) is highly similar to those obtained from MethyLight (B). Asterisk denotes MI <0.2 but at least 3 SD greater than the background level seen in 10 identical replicates of WBC samples. These data demonstrate the applicability of COMPARE-MS to multiple genes and heterogeneous samples. (C) Bisulfite genomic sequencing of the GSTP1 CGI in PrEC, LNCaP and CWR22Rv1 cell lines. Bisulfite sequencing shows that COMPARE-MS was accurate in identifying that the GSTP1 CGI in CWR22Rv1 cells is highly methylated, and demonstrates that MethyLight failed to detect this because many of the CpG dinucleotides interrogated by the methylation specific primers and probe were mostly unmethylated. Both COMPARE-MS and MethyLight were able to correctly identify that LNCaP and PrEC cells were homogeneously methylated and unmethylated, respectively. The indicated bisulfite sequencing start and end positions are relative to the transcriptional start site.

Comparison of ROC curves obtained by COMPARE-MS with those obtained by MethyLight. ROC curves for hypermethylation at the GSTP1, MDR1 and PTGS2 CGIs in distinguishing between benign and malignant prostate as determined by COMPARE-MS were comparable with those generated by MethyLight (33). An ideal assay would perfectly distinguish between true positives and true negatives and would have an area under the ROC curve of 1.0. The dashed lines represent the ROC curve for a hypothetical test that cannot distinguish between these two groups, giving an AUC of 0.5. CGI hypermethylation at the GSTP1, MDR1 and PTGS2 genes as determined by COMPARE-MS distinguish benign prostate from prostate cancer with high sensitivity and specificity, with AUCs extremely close to the ideal case.

Baculovirus-mediated expression of C-terminal His₆-tagged human MBD2-MBD in Sf9 insect cells and purification by Ni-NTA beads. Unbound lysate, washes and eluted purified protein were run on an SDS–PAGE gel and stained with Coomassie Blue R-250. The eluate contained the purified ∼9.8 kDa MBD2-MBD-His₆.

COMPARE-MS assay performance. (A and B) Plots showing the measured amount of methylated GSTP1 CGIs in M.SssI treated and untreated WBC genomic DNA versus the amount of input DNA after enriching for methylated DNA by methylation-sensitive restriction enzymes alone (A) or MBD2-MBD capture alone (B). (C) Plot of relative enrichment of M.SssI treated or untreated WBC DNA with or without MBD2-MBD, anti-His antibody and protein G magnetic beads. The degree of capture of unmethylated DNA (untreated WBC DNA) in the presence of MBD2-MBD is less than or equal to the capture of DNA in the absence of MBD2-MBD or anti-His antibody, showing that capture of unmethylated DNA during the DNA capture step of COMPARE-MS is almost completely due to low amounts of non-specific binding to the protein G magnetic beads, as opposed to low-level binding of the MBD2-MBD to unmethylated DNA. (D) Plots showing the measured amount of methylated GSTP1 CGIs in M.SssI treated and untreated WBC genomic DNA versus the amount of input DNA after enriching for methylated DNA by the combination of methylation-sensitive restriction enzyme digestion and MBD2-MBD capture (COMPARE-MS) followed by real-time PCR. When 20–100 ng of input DNA are used, COMPARE-MS has a >5000-fold dynamic range, which is ∼500-fold higher than that of methylation-sensitive restriction enzyme used alone and ∼5- to 10-fold higher than that of MBD2-MBD capture used alone. (E) Plot showing measured output methylated GSTP1 CGIs as determined by COMPARE-MS when decreasing amounts of M.SssI-treated WBC DNA is diluted in 20 ng of untreated WBC genomic DNA. The dashed line is a reference representing the mean COMPARE-MS output (0.0065 ± 0.0023 ng) when four identical replicates of 100 ng of untreated, unmixed WBC genomic DNA were analyzed. COMPARE-MS performance in this series of simulated heterogeneous samples (E) is highly linear for almost four orders of magnitude and nearly identical to that seen with homogeneously methylated samples (D), showing robust reproducibility and sensitivity. Data in (A–E) represent mean ± SEM for triplicate measurements.

COMPARE-MS applied to heterogeneous prostate tissues. (A–C) MI at the GSTP1, MDR1, PTGS2 and ESR1 CGIs in 20 ng of genomic DNA from benign prostate tissues obtained from 13 organ donors, who had no evidence of prostatic malignancies (A), 20 ng of genomic DNA from tumor-adjacent benign prostate cancer tissues isolated from 12 of the 130 men from whom prostates were obtained during radical prostatectomy for treatment of primary prostate cancer (B) and 20 ng of genomic DNA from primary prostate cancer tissues from 130 primary prostate cancer patients undergoing radical prostatectomy for treatment of their disease (C). Asterisk denotes MI <0.2 but greater than the threshold determined by ROC curve analysis. These data demonstrate the applicability of COMPARE-MS to heterogeneous human tissue samples.