Astrocytic abnormalities and global DNA methylation patterns in depression and suicide

Abstract

Astrocytes are glial cells specific to the central nervous system and involved in numerous brain functions, including regulation of synaptic transmission and of immune reactions. There is mounting evidence suggesting astrocytic dysfunction in psychopathologies such as major depression, however, little is known about the underlying etiological mechanisms. Here we report a two-stage study investigating genome-wide DNA methylation associated with astrocytic markers in depressive psychopathology. We first characterized prefrontal cortex samples from 121 individuals (76 who died during a depressive episode and 45 healthy controls) for the astrocytic markers GFAP, ALDH1L1, SOX9, GLUL, SCL1A3, GJA1 and GJB6. A subset of 22 cases with consistently downregulated astrocytic markers was then compared with 17 matched controls using methylation binding domain-2 (MBD2) sequencing followed by validation with high-resolution melting and bisulfite Sanger sequencing. With these data, we generated a genome-wide methylation map unique to altered astrocyte-associated depressive psychopathology. The map revealed differentially methylated regions (DMRs) between cases and controls, the majority of which displayed reduced methylation levels in cases. Among intragenic DMRs, those found in GRIK2 (glutamate receptor, ionotropic kainate 2) and BEGAIN (brain-enriched guanylate kinase-associated protein) were most significant and also showed significant correlations with gene expression. Cell-sorted fractions were investigated and demonstrated an important non-neuronal contribution of methylation status in BEGAIN. Functional cell assays revealed promoter and enhancer-like properties in this region that were markedly decreased by methylation. Furthermore, a large number of our DMRs overlapped known Encyclopedia of DNA elements (ENCODE)-identified regulatory elements. Taken together, our data indicate significant differences in the methylation patterns specific to astrocytic dysfunction associated with depressive psychopathology, providing a potential framework for better understanding this disease phenotype.

Figure 1. Screening cases and controls for astrocytic genes

Strong decrease in expression of astrocyte markers in BA 10 from cases and controls (N=121) using quantitative PCR a) GFAP (mean ± SEM), case (1.17 ± 0.13), control (2.27 ± 0.40) p=0.0068 and b) ALDH1L1 case (1.12 ±0.13), control (2.08 ± 0.31), p=0.0023 c) SOX9 case (1.48± 0.22), control (2.41 ±0.64) p=0.03, d) GLUL case (1.02 ± 0.12), control (1.46 ± 0.21), p=0.01, e) SCL1A3 case (1.63 ± 0.28), control (2.12 ± 0.52), p=0.04, f) GJA1 case (1.17 ± 0.14), p=0.03, g) GJB6 case (1.48 ± 0.23), control (2.41 ± 0.64), p=0.05., h) Expression of RBFOX3 (aka NeuN, a standard marker for mature neuronal identity) shows no difference between groups (p=0.86). (Tests performed were either Student T-test or Mann Whitney, depending on the distribution of the data)

Figure 2. Estimated methylation levels across the genome and level of methylation in case relative to controls

a) Comparison of DNA methylation levels between cases and controls reveals region specific effects in the human genome. The overall methylation levels were estimated from region read counts. The Wilcoxin rank-sum test was used to determine significance of methylation differences between pairs of region types. Only introns were not different than 3′ regions, all other pairs of regions types were significantly differentially methylated (p< 2E10–22). TSS, transcription start site; TES, transcription end site, imprinted genes are based on geneimprint.com, CpG islands are defined as regions with an expected CpG frequency of greater than 60%, a GC content of greater than 50% and a length greater than 200bp. b) the percentage summary of DMRs in known genome regions. c) cases with decreased expression of astrocyte associated genes show a genome-wide pattern of hypomethylation. 63.5% of DMRs were hypomethylated in cases compared to controls.

Figure 3. DMRs overlap with ENCODE data

Strong overlap of DMRs with ENCODE data. a) 44% of DMRs overlap with at least one regulatory feature. a) the breakdown of DMRs overlapping multiple ENCODE features b) the proportion of all queried regulatory features represented within our sample. c) the breakdown of DMRs overlapping with active or repressive regulatory marks.

Figure 4. Gene specific validation of DMRs shows inverse correlation to gene expression levels and FAC sorting shows specific non-neuronal contribution

Schematic diagrams of each gene highlighting region of methylation difference, the size of amplicon and the number and distribution of CpGs for each a) GRIK2 and b) BEGAIN, respectively. c) HRM results for GRIK2 show a decreased methylation in cases (Unpaired t test, p=0.025). d) Bs-cloning supports MDB-Seq and HRM results showing a cluster of the first 9 CpGs as significantly less methylated in cases (n=21, avg. 17 clones/sample) than controls (n=20, avg. 17 clones/sample) mixed model regression analysis was perform to assess significance of methylation at each CpG, *p<0.05, **p<0.01, ***p<0.001 exact values for pairwise comparison are in SI table 9a and cluster analysis in 9b). e) GRIK2 expression as measured by relative quantitation using TAQMAN probes shows an increasing in expression in cases (student t-test p= 0.012). f) GRIK2 expression correlates with the cluster of significantly differentially methylated CpGs (average methylation at each CpG, spearman r= −0.37, p=0.012). g) HRM results for BEGAIN show more methylation in cases compared to controls with (Unpaired t test with Welch’s correction, p=0.007). h) Bs-cloning for BEGAIN shows striking increase in methylation of cases (n=17, avg. 11 clones/sample) compared to controls (n=18, avg. 11 clones/sample) mixed model regression analysis was perform to assess significance of methylation at each CpG, *p<0.05, **p<0.01, #p<0.00, ϕ trend 0.053, exact values for pairwise comparison are in SI table 9c and cluster analysis in 9d). Cases showed on average a three-fold increase in methylation within cluster on BEGAIN. i) BEGAIN FAC sorted samples show a strong contribution of methylation from the non-neuronal cell fraction in case samples, whereas there is no difference in controls sample methylation between the two fractions (n=9 for all groups, avg. 17 clones/sample, one-way Anova F (3.32)=10.74, p<0.0001, Tukey’s pos-hoc, *p<0.05, ***p<0.001, ****p<0.0001). j) Analysis of BEGAIN variant expression in brain and blood. Experiment was independently repeated 3 times with 2 endogeneous controls per experiment, geometric mean of endogenous control was calculated and expression was normalized to this value. We qualitatively show that variant 1 is more expressed in brain and almost not present in blood, while we still detect variant 2 in blood. k) The astrocytic dyfunction group show a 2.3 fold decrease in the expression of BEGAIN Variant-1 compared to controls (Mann Whitney test p<0.0001), l) whereas variant-2 shows no change (Student t-test, p=0.26) m) The decrease in variant-1 expression correlates with the significantly differentially methylated cluster of CpGs in BEGAIN (average methylation at each CpG, spearman r=−0.37, p=0.019).

Figure 5. Promoter and enhancer properties repressed by DNA methylation in BEGAIN 474bp amplicon

Schematic diagram of CpG free lucia reporter vector with a) no endogeneous promoter and b) behind the human EF-1α promoter and placement of 474bp insert into multiple cloning site c) The unmethylated insert shows promoter activity which is completely abolished by methylation (10 replicates/group, One-way Anova F (2,27)=106.8, p<0.0001, Tukey’s post-hoc ****p<0.0001). d) Placed in front of the promoter, the 474 bp insert showed enhancer activity again showing complete repression of activity by methylation (10 replicates/group, One-way Anova F (2,23)=17.80, p<0.0001, Holm-Sidak’s post-hoc, ***p<0.001 ****p<0.0001).