Widespread promoter methylation of synaptic plasticity genes in long-term potentiation in the adult brain in vivo

Abstract

Background: DNA methylation is a key modulator of gene expression in mammalian development and cellular differentiation, including neurons. To date, the role of DNA modifications in long-term potentiation (LTP) has not been explored.

Results: To investigate the occurrence of DNA methylation changes in LTP, we undertook the first detailed study to describe the methylation status of all known LTP-associated genes during LTP induction in the dentate gyrus of live rats. Using a methylated DNA immunoprecipitation (MeDIP)-array, together with previously published matched RNA-seq and public histone modification data, we discover widespread changes in methylation status of LTP-genes. We further show that the expression of many LTP-genes is correlated with their methylation status. We show that these correlated genes are enriched for RNA-processing, active histone marks, and specific transcription factors. These data reveal that the synaptic activity-evoked methylation changes correlates with pre-existing activation of the chromatin landscape. Finally, we show that methylation of Brain-derived neurotrophic factor (Bdnf) CpG-islands correlates with isoform switching from transcripts containing exon IV to exon I.

Conclusions: Together, these data provide the first evidence of widespread regulation of methylation status in LTP-associated genes.

Keywords: DNA-methylation; Epigenetics; LTP; Long-term potentiation; MeDIP; Neuroepigenetics; Synaptic plasticity.

Fig. 1

Design of MeDIP-array for previously LTP associated genes. a Genes identified as differentially expressed in Maag et al. 2015, with an absolute logFC ≥ 0.9 and FDR ≤ 0.05, were targeted in this array. Briefly, probes were designed around an area of 3 kb of the TSS and CpG islands associated with these genes. The intensity of each probe was calculated using limma. In total, 38,119 probes were created for this experiment. b Each probe was collapsed into their specific group based on their location near their associated gene. Probes could be located in promoters, CpG-islands, shore, or a combination of all the previous elements. The majority of probes fell into the shores region

Fig. 2

Differential methylation analysis reveals the majority of regions being downregulated in response to high-frequency stimulation. Number of differentially methylated (DM) (a) probes and (b) regions when comparing the stimulated samples with the controls. c Volcano plot showing the log2 fold change value plotted against the p-value for all regions per time point compared to control. Non-significantly methylated regions are coloured grey. The colours represent the different locations for each probe and region. The name of the top 20 differentially methylated regions for each time point are shown. d Venn diagram of DMRs from (c). e Comparison of methylation changes between LTP, single electroconvulsive stimulation [18], and contextual fear with our without shock [17]. f Total numbers of differentially methylated array-specific genes in each study. Only genes present in our methylation array were included, and all different methylated regions were collapsed into corresponding unique gene names. Bars in purple are overlaps with our LTP results

Fig. 3

Correlation between methylation and expression status shows enrichment for RNA-splicing, kinase, and 3-UTR binding. a Effect of correlation between regions and their corresponding genes (b) Density plot of highly correlated genes (r ≥ 0.7 and p-value ≤ 0.05) (c) association between correlation value and p-value. Examples of LTP-associated genes (d) Arl4d and (e) Arc and the correlation between expression and methylation. Enrichment of (f) biological processes and (g) molecular functions of Methylation/expression correlated genes against non-correlated genes in the array

Fig. 4

Motif search in promoter region of correlated genes show enrichment for CG-rich motifs. a Motifs enriched in significantly methylation/expression correlated 1 kb upstream of the TSS compared to non-correlated genes. b Average number of CpG-islands present in 1 kb upstream from the promoter in both groups compared to number of promoters in each groups. c Boxplots to the right show the GC frequency of the significantly and non-significantly correlated genes. μ represents the mean for each group. p-value was calculated using the Students’ t-test

Fig. 5

Chromatin marks show enrichment over correlated genes in normal rat brain. Public ChIP-seq control data sets were downloaded and analysed over the correlated and non-correlated genes with the respect of their (a) TSS ± 3 kb or (b) the whole gene body ± 3 kb from TSS and TES. H3K27ac, H2K4me2, H3K4me1, and Matrin 3 was obtained from a growth hormone (GH)- expressing rat pituitary cell line [49]), H3K9me3 from rat hippocampal tissue [50], PolII from rat neuronal cultures [51], 5hmc-capture from rat brainstem [52], and H4K5ac from rat dorsal striatum [53]. Error margin shows ± sem. FDR was calculated using Benjamini-Hochberg correction after using Wilcox-test

Fig. 6

Time dependent differences in methylation status correlates with isoform switching of the Bdnf gene. a The Bdnf loci with the 4 major isoforms expressed in this study with associated CpG-islands. b The expression (left) and fraction (right) of the major Bdnf isoforms. c The correlation between methylation and expression of the NM_012513 (left) and NM_001270633 (right) isoforms of Bdnf and their associated CpG-island