DOES IT MATTER WHICH CELLS WE USE IN DNA METHYLATION ANALYSES OF GLUCOCORTICOID SIGNALING GENES

Background Numerous studies have been using peripheral tissues (e.g., blood or saliva) for DNA methylation analyses in psychiatric epigenetics. The glucocorticoid receptor gene (NR3C1) was among the first ones to show DNA methylation changes due to adverse social environment. Importantly, the NR3C1 gene has several alternative promoters, out of which the 1-F showed increased DNA methylation level at a transcription factor binding site. Afterwards, its co-chaperone molecule, FK506-binding protein 5 (FKBP5) gained interest, showing allele-specific demetylation in its glucocorticoid response element of the 7th intron. This decreased methylation at FKBP5 can lead to NR3C1 resistance, because of an impaired intracellular negative feedback loop. Methods While it is now standard for epigenome-wide DNA methylation analyses to control for cell heterogeneity using cell count estimates derived from the DNA methylation profiles, this important technical variable of cell composition is rarely included in candidate gene analyses. Therefore, we aimed to (1) develop a panel of pyrosequencing assays for cell-specific CG-sites, and (2) evaluate the effect of cell ratio on DNA methylation levels of candidate gene regions in frequently used peripheral tissues, such as Whole Blood (WB), Peripheral Blood Mononuclear Cell (PBMC), as well as non-invasively collected buccal, mouthwash, and saliva samples in two cohorts. Samples were collected from 51 adults participating in a Montreal-based longitudinal cohort, out of which 16 men had been assessed twice. In addition, 38 female patients with borderline personality disorder and 21 control women were recruited at a psychiatric clinic in Budapest. Candidate gene regions of the NR3C1-1F and FKBP5-intron7, as well as tissue-specific CG-sites were assessed by pyrosequencing at both sites. Results The DNA methylation level of the CG-island in the NR3C1-1F promoter region showed little variance in all of the tissue types we investigated. However, the low CG-density region of FKBP5-intron7 was highly variable across samples, especially in saliva. Notably, methylation level of the epithelial cell marker PTPN7 cg18384097 varied widely among the mouth-related samples in both cohorts (ranging from 9%-95%), even in buccal swab samples, where there was ~ 80% buccal cell ratio on average with 10–12% SD. The effect of epithelial cell composition was substantial on FKBP5-intron7 methylation levels in saliva and mouthwash samples (r=0.5–0.7), but it was less pronounced in buccal swab samples (r=0.2–0.4). Higher ratio of epithelial cells resulted in lower methylation level at the studied FKBP5 CG-sites. Importantly, the stability of the DNA methylation level of this gene region was significant both in saliva and buccal swab samples (r=0.6–0.8, p Discussion Our preliminary results show that cell composition should be taken into account at candidate gene DNA methylation analyses, especially when measuring low CG-density regions in heterogeneous peripheral tissues, such as whole blood or saliva. Using more homogeneous biological samples, e.g., PBMC or purified leukocytes would result in less technical variability. The non-invasively obtained buccal swabs can also yield a relatively homogeneous surrogate sample in psychiatric epigenetics.