Prenatal exposures and cell type proportions are main drivers of FKBP5 DNA methylation in maltreated and non-maltreated children

Abstract

DNA methylation in peripheral tissues may be a relevant biomarker of risk for developing mental disorders after exposure to early life adversity. Genes involved in HPA axis regulation, such as FKBP5, might play a key role. In this study, we aimed to identify the main drivers of salivary FKBP5 methylation in a cohort of 162 maltreated and non-maltreated children aged 3-5 years at two measurement timepoints. We combined data from a targeted bisulfite sequencing approach for fine-mapping 49 CpGs in regulatory regions of FKBP5 and epigenetic scores for exposure to alcohol, cigarette smoke, and glucocorticoids derived from the EPICv1 microarray. Most variability of methylation in the FKBP5 locus was explained by estimated cell type proportions as well as epigenetic exposure scores, most prominently by the glucocorticoid exposure score. While not surviving correction for multiple testing, we replicated previously reported associations of FKBP5 methylation with CM. We also detected synergistic effects of both rs1360780 genotype and the glucocorticoid exposure score on FKBP5 hypomethylation. These effects were identified in the 3'TAD, a distal regulatory region of FKBP5 which is not extensively covered in Illumina arrays, emphasizing the need for fine mapping approaches. Additionally, the epigenetic glucocorticoid exposure score was associated with childhood maltreatment, maternal mental disorders, and pregnancy complications, thereby highlighting the role of glucocorticoid signaling in the epigenetic consequences of early adversity. These results underscore the need to assess cell type heterogeneity in targeted assessments of DNA methylation and show the impact of exposures beyond just childhood maltreatment such as glucocorticoid exposure.

Keywords: DNA methylation; Early-life adversity; FKBP5; Glucocorticoids; HAM-TBS.

Graphical abstract

Fig. 1

Genomic locations of HAM-TBS amplicons. Note. Genome browser view of FKBP5. Track a): transcripts located within the locus. Track b): locations of CTCF factor-mediated chromatin interactions (ChIA-PET data) extracted from lymphoblastoid cell line (GM12878, Tang et al., 2015). Track c) and d): transcription factor binding (CTCF & GR) obtained from chromatin immunoprecipitation (ChIP) experiments in multiple cell lines from the ENCODE project. Track e): positions of targeted bisulfite sequencing amplicons chosen for this project. Panels f) – n): exact positions of sequenced CpGs via HAM-TBS (top track) and via EPICv1 microarray (bottom track) per amplicon.

Fig. 2

Drivers of variance in FKBP5 DNAm. Note. Percentage of explained variance (partial R²) for all predictors per sequenced position. The R² estimates result from linear regression models including all predictors, time, and a random subject effect to account for intra-individual variability from T0 to T2.

Fig. 3

Genotype-dependent CTCF binding site methylation. Note. Two CpGs within the 3′TAD (chr6:35490608 and chr6:35490619) showed significantly lower methylation levels in carriers of the risk allele (CT/TT, shown in dark red) as compared to carriers of the protective genotype (CC, shown in pink). Both CpGs were highly correlated with r = 0.83 (Spearman correlation, p < 2.2∗10⁻¹⁶).

Fig. 4

Genotype-glucocorticoid interaction. Note. FDR-corrected significant additive effects of GC exposure and rs1360780 genotype in the 3′TAD at T0. Carrying the risk allele (CT/TT, shown in dark red) and an increased GC exposure score were associated with hypomethylation.

Fig. 5

Correlation of exposure scores across timepoints. Note. Panel a): Correlation matrix of epigenetic exposure scores, estimated cell type proportions, and biodata at T0 and T2. Panel b): Correlation matrix of epigenetic exposure scores residualized for estimated cell type proportions at T0 and T2. All depicted coefficients are Spearman correlations. Nominally significant correlations (uncorrected p < .05) are denoted with ∗.