Methylome from liquid biopsy cell-free DNA using enzymatic conversion of unmethylated cysteines

Methylome of tumor cell-free DNA (cfDNA) has emerged as a powerful non-invasive technique for cancer subtyping and prognosis. However, its application is frequently hampered by the quality and total cfDNA yield. Here we demonstrate the feasibility of very low-input cfDNA for whole-methylome and copy-number profiling studies using enzymatic conversion of unmethylated cysteines (EMSEQ) to better preserve DNA integrity. We demonstrate that cfDNA methylation of our cohort was comparable with in situ cohorts and built a model that accurately predict MYCN amplification or 11q deletion in validation cohorts. RASSF1A methylation precedes this divergent methylome signature, supporting common and different epigenetic origins. After stratification by 11q, eight CpG methylations show additional prognostic value and two reveal known preclinical targets (CSF1R, CCR7). Methylation in CSF1R and CCR7 associated with differential expression levels and poorer prognosis. Noteworthy, haploinsufficiency of genes located on 11q that repair DNA double-strand breaks display reduced expression by a mechanism independent of promoter methylation. Conclusions: RASSF1A is an early hypermethylation defect that may predispose to later genetic and epigenetic alterations specific to MYCN-amplified or 11q-deleted high-risk neuroblastomas. CSF1R and CCR7 hypermethylation are a hallmark of high-risk 11q-deleted neuroblastomas that could be exploited therapeutically with immunomodulators to override immunosupression and relapse. Our novel findings provide support for use of liquid biopsy as an optimal strategy to assess methylomes of neuroblastoma patients for a precision medicine approach. To prepare libraries for comprehensive methylome representation within a cfcDNA we used the NEBNext Enzymatic Methyl-seq Kit (New England Biolabs, NEB), following manufacturer's recommendations with minor modification that the cfcDNA was not sheared. The method utilizes 10 ng starting material of cfcDNA and relies on two step conversion of the cytosines. The first step uses TET2 oxidation of 5-methylcytosine that provides protection to the modified cytosines from conversion by APOBEC deamination, so only cytosines but not 5-methylcytosines are converted to uracil. Ultimately, cytosines are sequenced as thymines and 5-methylcytosines are sequenced as cytosines. The final 8 cycles of PCR library amplification were carried out using the NEB Unique dual index primer pairs (NEBNext multiplex oligos for Illumina). The libraries were sequenced on NovaSeq 6000 (Illumina) in paired-end with a read length of 2x151bp according to the manufacturer’s protocol for dual indexing. Image analysis, base calling and quality scoring of the run are processed using the manufacturer’s software Real Time Analysis (RTA 3.3.3) and followed by generation of FASTQ sequence files. The EMSEQ reads were processed using the gemBS pipeline v4.03 using as reference GRCh38[23]. Reads with MAPQ scores < 20 and read pairs mapping to the same start and end points on the genome were filtered out after the alignment step. The first 5 bases from each read were trimmed before the variant and methylation calling step to avoid artifacts due to end repair. For each sample, CpG sites were selected where both bases were called with a Phred score of at least 20, corresponding to an estimated genotype error level of <=1%, and where the total number of reads informative for methylation from both strands combined was >=6. Sites with >500x coverage depth were excluded to avoid centromeric/telomeric repetitive regions. Sequencing depth for the analysis of copy number was calculated using samtools v1.15.