Differential methylation of circulating free DNA assessed through cfMeDiP as a new tool for breast cancer diagnosis and detection of BRCA1/2 mutation

Abstract

Background: Recent studies have highlighted the importance of the cell-free DNA (cfDNA) methylation profile in detecting breast cancer (BC) and its different subtypes. We investigated whether plasma cfDNA methylation, using cell-free Methylated DNA Immunoprecipitation and High-Throughput Sequencing (cfMeDIP-seq), may be informative in characterizing breast cancer in patients with BRCA1/2 germline mutations for early cancer detection and response to therapy.

Methods: We enrolled 23 BC patients with germline mutation of BRCA1 and BRCA2 genes, 19 healthy controls without BRCA1/2 mutation, and two healthy individuals who carried BRCA1/2 mutations. Blood samples were collected for all study subjects at the diagnosis, and plasma was isolated by centrifugation. Cell-free DNA was extracted from 1 mL of plasma, and cfMeDIP-seq was performed for each sample. Shallow whole genome sequencing was performed on the immuno-precipitated samples. Then, the differentially methylated 300-bp regions (DMRs) between 25 BRCA germline mutation carriers and 19 non-carriers were identified. DMRs were compared with tumor-specific regions from public datasets to perform an unbiased analysis. Finally, two statistical classifiers were trained based on the GLMnet and random forest model to evaluate if the identified DMRs could discriminate BRCA-positive from healthy samples.

Results: We identified 7,095 hypermethylated and 212 hypomethylated regions in 25 BRCA germline mutation carriers compared to 19 controls. These regions discriminate tumors from healthy samples with high accuracy and sensitivity. We show that the circulating tumor DNA of BRCA1/2 mutant breast cancers is characterized by the hypomethylation of genes involved in DNA repair and cell cycle. We uncovered the TFs associated with these DRMs and identified that proteins of the Erythroblast Transformation Specific (ETS) family are particularly active in the hypermethylated regions. Finally, we assessed that these regions could discriminate between BRCA positives from healthy samples with an AUC of 0.95, a sensitivity of 88%, and a specificity of 94.74%.

Conclusions: Our study emphasizes the importance of tumor cell-derived DNA methylation in BC, reporting a different methylation profile between patients carrying mutations in BRCA1, BRCA2, and wild-type controls. Our minimally invasive approach could allow early cancer diagnosis, assessment of minimal residual disease, and monitoring of response to therapy.

Keywords: BRCA1; BRCA2; Breast cancer; Cell-free DNA; DMRs; cfMeDIP-seq.

Fig. 1

Identification and annotation of Differentially Methylated Regions (DMRs). (A) The volcano plot shows the DMRs. The x-axis represents the log2 fold change (FC) of methylation, and the y-axis represents the -log10(p-value). The regions with a p-value < 0.01 and |log2(FC)| >= 2 were selected as differentially methylated. The yellow points represent the hypermethylated regions (7,095), the blue points represent the hypomethylated regions (212), the dark gray points represent the not significant regions. (B) Principal Component Analysis of TCGA-BRCA methylation data overlapping the DMRs. The red points represent the primary tissue samples, while the blue ones represent the solid tissue normal samples. (C) Proportion of DMRs according to Genic Annotations. The distribution of DMRs across the 3UTRs (3’ untranslated regions), CDS (coding DNA sequences), enhancers_fantom (enhancers from FANTOM5), 5UTRs (5’ untranslated regions), 1to5kb (1 to 5 kb regions upstream of transcription start site, TSS), promoters (regions with a length of less than 1 kb upstream of TSS) were explored. (D) Proportion of DMRs according to CpG Annotations

Fig. 2

Functional annotation of hyper- DMRs in BRCA-carriers Top 20 biological processes enriched by the genes overlapping hypermethylated regions in BRCA-carriers and selected according to p-value. The x axis corresponds to − Log10 (p-value), and the Y axis corresponds to the gene ontology terms

Fig. 3

Transcription factor enrichment analysis. (A) Barplot of enriched binding motifs (y-axis), ordered based on -log10 p-value (x-axis), encoded by the hyperDMRs (B) Volcano plot shows the differentially active TFs estimated comparing 41 BRCA + tumor samples versus 41 solid tissue normal tissue samples of the TCGA-BRCA cohort. The positively dysregulated TFs are shown in red while the negatively dysregulated TFs are shown in green. The TFs with an absolute value of delta greater or equal to 1 and False Discovery Rate (FDR) < 0.05 were selected as dysregulated. The dysregulated TFs identified with VIPER and HOMER were labeled in the volcano plot

Fig. 4

Performance of classifiers built using the DMRs as selected features. (A) Performance of the classifier based on the GLMnet model (B) Performance of the classifier based on the random forest model