Comparative analysis of pyrosequencing and methylation profiling for assessing LINE-1 and Alu methylation in response to low-level benzene exposure

LINE-1 (L1) and Alu methylation has been used as a surrogate marker for global DNA methylation, traditionally assessed via bisulfite-PCR and pyrosequencing (pyroseq). However, this method interrogates few CpGs, likely overlooking wider methylation patterns. Instead, array-based technologies (e.g. Illumina EPIC assay), profile thousands of loci residing in repetitive elements (RE), offering a detailed yet selected view of methylation landscapes. Benzene exposure has been linked to L1 and Alu hypomethylation, yet the extent to which various assays capture these changes is still unclear. This study aims to compare mean methylation derived from pyroseq and EPIC array in adults with low-level occupational benzene exposure, assessing the concordance between methods in detecting exposure-related changes. We studied 211 adults from Milan (77 gas station attendants, 77 traffic police officers, 57 office workers). Personal benzene exposure was measured by passive samplers for a full work shift. Peripheral blood DNA was bisulfite-treated and used for L1 and Alu methylation quantification by PCR/pyroseq and profiling by the Infinium MethylationEPIC v1.0 BeadChip. Raw IDATs were processed to β-values after standard QC. Probes overlapping L1 and Alu (RepeatMasker) were used to derive EPIC-based means employed for comparison. Methylation distributions were summarized (median, IQR), and concordance between methods was assessed using correlation, Lin’s CCC, and variance component analysis. Analyses were performed on R. L1 and Alu mean methylation estimates differed markedly between methods (L1 EPIC μ=0.812 vs pyroseq μ=0.717; Alu EPIC μ=0.828 vs pyroseq μ=0.271). The assays yielded shifted methylation distributions and weak correlations (L1 r=−0.156, p=0.033; ρ=−0.136, p=0.065; Alu r=0.056, p=0.428; ρ=−0.015, p=0.835). Lin’s CCC (L1 −0.009; Alu 8.5×10⁵) indicated poor concordance and a platform effect. Probe-level heterogeneity explained most of variance in EPIC-derived estimates. Our findings show a lack of concordance between pyroseq and EPIC-derived mean methylation estimates for L1 and Alu. This discrepancy highlights the different aspects of RE methylation captured by these assays, questioning the validity of using them equally for global DNA methylation analysis, notably in exposure studies.