Material and methods
Purification of ccfDNA from human serum or plasma samples
Unless otherwise stated, 0.5 mL (Eppendorf, Tokyo, Japan) or 2.0 mL (Eppendorf) Protein LoBind tubes were used. The QIAamp MinElute ccfDNA Mini Kit (Qiagen, Tokyo, Japan) was used to extract ccfDNA from 4 mL of serum, plasma, or culture supernatant, according to the manufacturer’s instructions. The concentration of ccfDNA was measured using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Yokohama, Japan).
Bisulfite conversion of ccfDNA as a template for droplet digital methylation-specific PCR (ddMSP)
Bisulfite conversion was performed using the EZ DNA Methylation-Lightning Kit (Zymo Research, Orange, CA, USA), Premium Bisulfite Kit (Diagenode Diagnostics, Liège, Belgium), and EpiJET Bisulfite Conversion Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The concentration of bisulfite-treated ccfDNA was measured using the Qubit ssDNA Assay Kit (Thermo Fisher Scientific).
ddMSP
The sequence of the DKK3 CpG island (position: chr11:12008191–12009294, band: 11p15.3, and genomic size: 1,104 bp), was downloaded from the UCSC genome browser (http://genome.ucsc.edu/) and analyzed using MethPrimer 2.0 (http://www.urogene.org/methprimer2/tester-invitation.html) [9]. The primers and Taqman-MGB probes were designed in our laboratory and synthesized by Thermo Fisher Scientific. Unmethylated DKK3-derived sequences were amplified using the DKK3_island_U assay, and methylated DKK3-derived sequences were amplified using the DKK3_island_M assay (Fig. 1C). ddPCR was performed using the QX200 system (Bio-Rad Laboratories, Hercules, CA, USA). EpiScope-Methylated HCT116 gDNA (Takara Bio Inc., Otsu, Japan) and unmethylated HCT116 DKO gDNA (Takara-Bio) were both subject to bisulfite conversion and used as positive and negative controls for methylation-specific PCR. The total volume of the PCR mixture in the assay was 20 µL, comprising 10 µL of ddPCR Supermix for Probes (No dUTP) (Bio-Rad Laboratories), 0.9 µM of each primer, 0.25 µM of each probe, and 200 µM of dNTP. The following PCR conditions were used for ddPCR: 10 min at 95 °C for DNA polymerase activation, followed by 40 cycles of 30 s at 94 °C for denaturation and 1 min at 50 and 56 °C (DKK3_island_U assay and DKK3_island_M assay, respectively) for annealing and extension, and termination at 98 °C for 10 min for DNA polymerase deactivation. Methylation- and non-methylation-specific PCR was performed using a C1000 Touch Thermal Cycler with 96 Deep well reaction modules (Bio-Rad Laboratories) in each methylation-specific PCR. The PCR products were read and analyzed using the QX-200 droplet reader (Bio-Rad Laboratories) and QuantaSoft analysis software (Version 1.7.4) (Bio-Rad Laboratories).
Clinical samples
Normal human serum (#S1-100ML, Lot_3291313; Millipore, Japan, Tokyo, Japan) and normal human plasma samples (#12271430, Lot_BJ12440A; Tennessee Blood Services Corporation, Memphis, TN, USA) were used as normal controls. Peripheral blood samples were collected from 21 patients with malignant mesothelioma. The details are shown in Additional files 2 and 3. Patients’ serum was pooled, and ccfDNA was purified from 4 mL.
Results and discussion
Summary of the assay developed in this study
A scheme of the novel method developed for detecting the methylation state of DKK3 via ddMSP using liquid biopsy is presented in Fig. 1A. First, we collected approximately 4 mL of pooled serum samples from patients with malignant mesothelioma. We then extracted ccfDNA using the QIAamp MinElute ccfDNA Mini Kit. The ccfDNA was subjected to bisulfite conversion using the EpiJET Bisulfite Conversion Kit. Finally, we quantified the copy number of methylated and unmethylated regions in the CpG island DKK3 promoter region via ddMSP using the QX200 system.
Primer and probe design for methylated and unmethylated DKK3 promoter regions
The primer and probe designs used for methylation-specific-PCR are crucial, so they were designed in the following flow. We analyzed the CpG islands (position: chr11:12008191–12009294) of the DKK3 promoter using MethPrimer software and designed methylation-specific primers (Fig. 1B) based on four considerations: (i) multiple CpG sites were included in the target amplicon to increase the selectivity; (ii) the size was as small as possible to increase the sensitivity of ccfDNA detection; (iii) the frequency of single nucleotide polymorphisms (SNPs) was relatively low in the target sequences; (iv) the nearly identical region of the genome was used for PCR in both cases of methylation and methylation alleles. Consequently, we selected one potential PCR amplification region for ddMSP including seven CpG sites (Fig. 1B). Furthermore, the methylation of cytosine immediately adjacent to these target sequences was observed among ten types of cancers with significant differences [10] (Additional file 1). This amplification region had no common SNPs based on Short Genetic Variants from dbSNP release 153 (https://www.ncbi.nlm.nih.gov/snp/). Therefore, we designed primers and TaqMan probe pairs based on these observations (Fig. 1C).
Specificity of the ddMSP assay
Since the sequences of amplicon for methylation and unmethylation are very similar, it is important to confirm the specificity. Confirming the specificity of ddMSP is very important for detecting a small percentage of existence of ctDNA. As a validation study, we first determined the specificity of each ddMSP reaction using matched and mismatched templates with optimization of the annealing temperature. We confirmed the selectivity of each PCR in the DKK3_island_U and DKK3_island_M assays (Fig. 2A). Non-specific reactions were not observed in reaction conditions using the mismatched template. We also optimized the annealing temperature for each PCR. The optimized temperatures were 56 and 50 °C for methylated and unmethylated PCRs, respectively. In addition, we confirmed the specificity of the amplicon size via electrophoresis (Fig. 2B), and the amplicon sequence was verified (data not shown).
Comparison of three commercial bisulfite conversion kits for converting ccfDNA derived from ccfDNA
The bisulfite treatment causes further fragmentation of the DNA, which is expected to have a particular impact on the detection system in the case of ccfDNA. Therefore, it is important to perform ddPCR after appropriate bisulfite treatment. We compared three major commercial bisulfite conversion kits to evaluate the recovery rate in determination of copies of unmethylated DKK3 derived from ccfDNA (Fig. 2C). An undiluted solution of ccfDNA extracted from serum and a sample diluted two or four times with nuclease-free water was subjected to bisulfite conversion. The EZ DNA Methylation-Lightning Kit and Premium Bisulfite Kit showed a negative correlation between the copy numbers of ccfDNA before bisulfite conversion and after bisulfite conversion, implying that impurities after bisulfite treatment may affect the detection system. In contrast, the EpiJET Bisulfite Conversion kit showed a positive correlation with the recovery rate (approximately 50%) and the copies of input. Although the reason for this finding is not known, it was established that the assay using EpiJET bisulfite transformation does not significantly impact our detection system. Consequently, we selected the EpiJET Bisulfite Conversion kit for use in the ddMSP system.
Measurement of ctDNA in malignant mesothelioma cell lines and mesothelial cells
To validate the ddMSP analysis system, we analyzed the methylation status of genomic DNA extracted from cultured mesothelioma cells (MSTO-211H, NCI-H28, NCI-H226, NCI-H2052, and NCI-H2452) and mesothelial cells (Met5A). Sources of cells and culture methods are shown in Additional file 2. The results of reduced bisulfite sequencing against DKK3 CpG island (hg19: chr11: 12030008–12030507) were downloaded and compared as a reference (Fig. 2D). Methylated DKK3 in genomic DNA copies was detected in NCI-H28 and NCI-H226 cells, and in NCI-H2052 cells to a lesser extent (Fig. 2E). Methylated DKK3 sequences were not detected in MSTO-211H and NCI-H2452 cells. In comparison with the cancer cell line encyclopedia database, the trend of methylation rate was very similar to the results of this experiment. No copies of the methylated genome were detected in the assay using the normal mesothelial cell line Met5A. Furthermore, when ccfDNA derived from normal serum and normal plasma (collected from different individuals) was examined, no copies of the methylated genome were detected. These results suggested that these MSP regions were specific to the tumor tissue.
Detection and quantification limits of the ddMSP assay
A low detection limit is critical for the detection of ctDNA in liquid biopsies. Therefore, we quantified the lower detection limit of the ddMSP assay. We spiked normal human serum with NCI-H28 cell-free DNA or fully methylated genome, whose copy number was previously quantified using TaqMan Copy Number Reference Assay RNase P (Thermo Fisher Scientific) (Fig. 3A), and determined the recovery rate after the purification of ccfDNA following bisulfite conversion by quantifying the number of copies of fully methylated DNA. We estimated that most cancers show more than 40 copies of ctDNA in 4.0 mL of patient serum [11]. Therefore, if the assay can detect 30 copies of ctDNA in 4.0 mL, the sensitivity is considered sufficient. Our system quantified approximately 30 copies of cell-free DNA per 4 mL, which is sufficient for detecting ctDNA (Fig. 3B). The data were also analyzed to determine whether a certain amount of unmethylated DKK3 could be recovered. The EpiJET Bisulfite Conversion kit showed a lower coefficient of variation value (0.18) (Fig. 3C) than that of the EZ DNA Methylation-Lightning Kit (0.58). We were able to detect a very small amount (< 0.2%) of methylated DNA, which is considered to be sufficiently sensitive.
Determination of DKK3 copies in serum samples from malignant mesothelioma via ddMSP
To verify that clinical samples can be measured using our measurement system, we quantified methylated and unmethylated DKK3 copies in ccfDNA from 21 patients with malignant mesothelioma. Patient characteristics are shown in Additional file 4. We detected 5–500 copies of methylated DKK3 in 4 mL of serum samples of eight patients. In contrast, no copies were detected in serum samples of 13 patients. Furthermore, the number of methylated copies tended to increase as the stage progressed, possibly reflecting changes in tumor size or changes in tumor properties (Fig. 3D). Differences based on disease type cannot be discussed because of the small number of cases (Fig. 3E). Based on these results, this assay system can be used to analyze clinical samples.
Conclusion
Using our novel method, we selectively and quantitatively measured methylated and unmethylated DKK3 in ccfDNA.
Limitations
One limitation of this study is that we did not evaluate the concordance rate with local tumor samples; therefore, the extent to which the peripheral blood results reflect the methylation status in the local tumor is not known. In the future, data such as the positive concordance rate with local test results such as FFPE samples should be evaluated.