Circulating cell-free DNA methylation patterns as non-invasive biomarkers to monitor colorectal cancer treatment efficacy without referencing primary site mutation profiles

This study investigates methylation patterns in circulating cell-free DNA (ccfDNA) for their potential role in colorectal cancer (CRC) detection and the monitoring of treatment response. Through methylation microarrays and quantitative PCR assays, we analyzed 440 samples from The Cancer Genome Atlas (TCGA) and an additional 949 CRC samples. We detected partial or extensive methylation in over 85% of cases within three biomarkers: EFEMP1, SFRP2, and UNC5C. A methylation score for at least one of the six candidate regions within these genes' promoters was present in over 95% of CRC cases, suggesting a viable detection method. In evaluating ccfDNA from 97 CRC patients and 62 control subjects, a difference in methylation and recovery signatures was observed. The combined score, integrating both methylation and recovery metrics, showed high diagnostic accuracy, evidenced by an area under the ROC curve of 0.90 (95% CI = 0.86 to 0.94). While correlating with tumor burden, this score gave early insight into disease progression in a small patient cohort. Our results suggest that DNA methylation in ccfDNA could serve as a sensitive biomarker for CRC, offering a less invasive and potentially more cost-effective approach to augment existing cancer detection and monitoring modalities, possibly supporting comprehensive genetic mutation profiling. Supplementary Information The online version contains supplementary material available at 10.1186/s12943-023-01910-y.

The primer sequences and restriction enzymes used for these analyses are listed in Table S10.
The PCR products were digested with HhaI (New England BioLabs, Ipswitch, MA, USA) and loaded simultaneously onto a SeqStudio Genetic Analyzer (Thermo Fisher Scientific).The unique fluorescent PCR signal distinguished individual PCR products for each target, and their fragment length and the data were analyzed using GeneMapper software 5 (Applied Biosystems, Foster City, CA, USA).The ratio of methylated CpG sites (digested by restriction enzymes) was calculated by the balance between the restriction enzyme-cleaved PCR products and the total amount of PCR product in each locus.
Methylation positivity was defined as the ratio of methylated alleles at 0.05 (5.0%) or more.The number of markers methylated subsequently determined the methylation score.By this definition, the methylation score in the CRC sample ranged between 0 to 6.

Blood specimen collection from a prospective cohort of patients with CRC
We prospectively recruited CRC patients between 2020 and 2022 at the Kochi Medical Centre in Kochi, Japan (Fig. S1C).A cohort of 97 blood samples from patients with CRC patients was collected before surgical resection.For the control group, blood samples were obtained from 70 patients who underwent curative resection for Union for International Cancer Control (UICC) stage I to III CRC with no evidence of recurrence at least one year after surgical resection, as confirmed by computed tomography (CT) and with no corresponding elevation of serum carcinoembryonic antigen (CEA) levels (ng/mL) throughout the post-surgery duration between 2016 and 2018 at Okayama University Hospital.All control subjects underwent colonoscopy after blood collection.Of the 70 cases, eight were excluded from further analysis for the following reasons: two cases with anal squamous cell cancer treated after CRT, one case with residual tumor treated after Endoscopic Submucosal Dissection, one case without colonoscopy examination, one case with metastasis detected by CT, and three cases with insufficient blood collection due to hemolysis.Finally, the remaining 62 cases were categorized as the control group.Of these 62 cases, 16 were confirmed to have adenomatous polyps in the colon and rectum by colonoscopy.
Therefore, the 62 control subjects were divided into 46 patients with no neoplasia (NN) and 16 with adenomatous polyps (AP) (Fig. S1D).
To examine whether our methylation biomarkers could effectively monitor tumor response during chemotherapy treatment, 126 blood samples were collected from six metastatic CRC (mCRC) patients before each treatment cycle of systemic chemotherapy at Okayama University Hospital.
A cohort of 97 blood samples from patients with CRC and 62 control subjects were collected with PAXgene Blood ccfDNA tubes (Qiagen NV, Hilden, Netherlands).The blood specimens from six mCRC patients were collected with an EDTA collection tube before each chemotherapy treatment and processed for plasma isolation immediately after collection.
According to the manufacturer's protocol, the plasma was separated from whole blood and centrifuged at 3,000 g for 15 minutes at 4°C.Prepared plasma samples were stored at −80 °C until use.

Evaluation of tumor burden by radiographical examination and blood CEA levels
The control subjects were evaluated for tumor recurrence events by CT at several time points, typically every 3 to 6 months after curative resection, as decided by the physician or prescribed by a clinical trial conducted at the hospital.
Tumor response to systemic chemotherapies for the six mCRC patients was evaluated based on CT or MRI scans every eight weeks, according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1.This study defined the imaging tumor burden as the sum of the maximum equator of up to six metastatic lesions.Serum CEA levels were evaluated using blood collected at the same time as the blood used in this study by a diagnostic laboratory at Okayama University Hospital or Kochi Medical Centre.

Extraction and bisulfite conversion of cfDNA
The cfDNA in serum was extracted using a QIAamp MinElute ccfDNA Kit (Qiagen), and the concentration of cfDNA was evaluated using a Qubit 4 Fluorometer (Thermo Fisher Scientific, MA) with the dsDNA H.S. Assay Kit (high sensitivity, 0.2 to 100 ng) following the manufacturer's instructions.Bisulfite modification of cfDNA (total amount of cfDNA was 5 ng per sample) was performed using an EZ DNA Methylation-Lightning Kit (Zymo Research).

Detection of methylated and unmethylated alleles in circulating cell-free DNA
To assess the methylation status in ccfDNA, we adopted a high-sensitivity assay for bisulfite DNA (Hi-SA) to recover and confirm methylated alleles from cancer cells [4].We optimized a multiplex-PCR strategy to recover all six loci in the EFEMP1, SFRP2, and UNC5C promoter simultaneously.Hi-SA can detect methylation ratios as low as 0.01 of methylated alleles per unmethylated allele using internal methylation-specific primers; hence, the methylation positivity was defined as the ratio of methylated CpG sites at 0.01 (1.0%) or more in ccfDNA.Similar to the fluorescent COBRA analysis of CRC tissues, a methylation score was given by the number of markers methylated in each case.Additionally, a recovery score was provided by the number of loci amplified from ccfDNA.When the sum of the fluorescence intensity of the methylated and unmethylated alleles exceeded 100, we judged that the target locus was successfully 'recovered.'Thus, methylation and recovery scores ranged from 0 to 6 at a given time.Hi-SA was performed twice per blood sample to confirm the reproducibility and improve the detection capacity.Thus, methylation and recovery scores ranged from 0 to 6 at one inspection and 0 to 12 at the sum of two-time analysis (two-time sum of methylation or recovery score).

Statistical Analysis
All statistical analyses were performed using JMP Genomics software (version 10.2; SAS Institute, Inc., Cary, NC, USA).The chi-squared test was used to examine the associations between categorical variables.The comparison of the mean beta (β) value between tumor and normal mucosa calculated from the TCGA database in each probe was analyzed by analysis of variance.The methylation ratios of the EFEMP1, SFRP2, and UNC5C promoters were analyzed as continuous and categorical variables.In analyzing CRC tissues, we assessed the relationship between methylation ratios in Region 1 and Region 2 of three genes by employing linear regression models.We calculated Spearman's rank correlation coefficients (ρ) to quantify this association.Furthermore, we assigned a numerical score to each CRC sample to represent the count of methylated loci.We used the Wilcoxon rank-sum test to compare the average methylation scores across different subgroups.
In the ccfDNA analysis, we employed linear regression models to estimate the relationship of methylation ratios at each locus between the first and second Hi-SA inspections.In addition, we determined the association of methylation rates at each locus between the two inspections by calculating Spearman's rank correlation coefficients (ρ).We also assigned a numerical score to each ccfDNA sample, representing the number of recovered and methylated loci.To perform nonparametric comparisons, we used the Dunn method for joint ranking, comparing ccfDNA samples with 62 control subjects or 46 NN subjects as the control group.This method calculates ranks for the entire dataset, not just the comparison pairs, resulting in P values reflecting a Bonferroni adjustment that Dunn's test reported.
We estimated the combination score (Fc), a function based on parameter estimates obtained from multiple logistic regression.The receiver operating characteristic (ROC) curve was plotted for potential cut-off values based on the two-time methylation score, the two-time recovery score, and the combination score (Fc) for the sensitivity of ccfDNA in patients with CRC.Similarly, the specificity was determined for control subjects.The area under the ROC curve (AUC) was measured to compare the screening efficiency of each score.The ROC curve for the combination score (Fc) was compared to that for the 5-fold cross-validation evaluated by both recovery and methylation scores, which gave identical results.A nonparametric approach was used to compare the AUCs, estimated by the combination score (Fc) and the 5-fold cross-validation [5].All P values reported were calculated in two-sided tests, and values less than 0.05 were considered statistically significant.Linear regression models were employed to estimate the relationship between methylation ratios at each locus in the first and second Hi-SA inspections.The association between methylation rates at each locus in the two inspections was determined by calculating Spearman's rank correlation coefficients (ρ).In the two independent Hi-SA analyses, the first and second methylation rates demonstrated a statistically significant positive correlation across all six loci.The conbination score (Fc) Table S8.The cut-off values of scores for sensitivity in patients with colorectal cancer (CRC) who are correctly identified as having the disease (true positives), as well as their 1-specificity for the proportion of individuals without the disease who are incorrectly identified as having the disease (false positives).

Fig. S4 .
Fig. S4.Summary of Blood Sample Analyses.(A) The concentration of ccfDNA, (B) the two-time methylation score, (C) the two-time recovery score, and (D) the combination score (Fc) was analyzed between CRC patients divided into the stages, subjects with colorectal adenomatous polyps (AP) and subjects with no evidence of neoplastic disease (NN).The box plot diagrams show the median as a horizontal line within each box, the interquartile ranges as the box limits, and the maximum and minimum values as the whiskers.P values were calculated using Dunn's test.

Table S7 . Clinical characteristics of CRC patients and control subjects for ccfDNA methylation analyses.
The pathological stage (pStage) is a classification based on pathological findings.The clinical stage (cStage) is a classification based on pre-treatment clinical findings.Control subjects categorized as adenomatous polyps (AP) were confirmed to have adenomatous polyps in the colon and rectum by colonoscopy.Control subjects categorized as patients with no neoplasia (NN) were confirmed to have no adenomatous polyps in the colon and rectum by colonoscopy.