Age-related profiling of DNA methylation in CD8+ T cells reveals changes in immune response and transcriptional regulator genes

Human ageing affects the immune system resulting in an overall decline in immunocompetence. Although all immune cells are affected during aging, the functional capacity of T cells is most influenced and is linked to decreased responsiveness to infections and impaired differentiation. We studied age-related changes in DNA methylation and gene expression in CD4+ and CD8+ T cells from younger and older individuals. We observed marked difference between T cell subsets, with increased number of methylation changes and higher methylome variation in CD8+ T cells with age. The majority of age-related hypermethylated sites were located at CpG islands of silent genes and enriched for repressive histone marks. Specifically, in CD8+ T cell subset we identified strong inverse correlation between methylation and expression levels in genes associated with T cell mediated immune response (LGALS1, IFNG, CCL5, GZMH, CCR7, CD27 and CD248) and differentiation (SATB1, TCF7, BCL11B and RUNX3). Our results thus suggest the link between age-related epigenetic changes and impaired T cell function.

Scientific RepoRts | 5:13107 | DOi: 10.1038/srep13107 Age-related transcriptome marker analyses in peripheral blood leukocytes (PBL) or mononuclear cells have revealed changes in gene expression levels and aberrant mRNA splicing [12][13][14] , with gender-specific effects 15 and dependence upon CMV phenotype 16 . Several studies have used peripheral blood cells for the analysis of DNA methylation changes during ageing, and have revealed differentially methylated CpG sites 17,18 , hypermethylation of CpG islands in promoters 19,20 , existence of mega-base scale hypomethylated blocks 21 , and proposed methylation biomarkers as predictors of age [22][23][24] . However, it should be noted that the transcriptional and epigenetic analyses in human peripheral blood samples are affected by the numerical alteration of specific cell subsets in blood including T cells, B cells, monocytes and granulocytes 25,26 . Even if normalization tools are applied 27 , the use of purified cell subsets is superior to the identification of methylome changes and their correlation to cell-type specific functions.
In light of these considerations we focused on sorted CD4+ and CD8+ T cells and aimed to identify whether the DNA methylation changes are involved in development of age-related decrease of T cell responsiveness. We purified DNA from PBL, CD4+ and CD8+ T cells from younger and older individuals, and compared their genome-wide DNA methylation differences. We correlated the age-related methylation changes with gene expression levels and histone modifications in the same regions. Our genome-wide analyses indicate for the first time in CD8+ T cell subset a strong age-related correlation of DNA methylation and the expression of genes having functional role in controlling T cell immune responses and their differentiation.

Results
Age-related methylation changes in PBL, CD4+ and CD8+ T cells. We studied the methylomes in PBL, and CD4+ and CD8+ T cells from 50 younger (age 22-34) and 50 older (age 73-84) individuals using the Illumina HumanMethylation450 arrays. For the differential methylation analysis, we used a linear model to remove confounding effects (see Materials and Methods for details). As the numbers of leukocyte subsets in blood vary greatly with age and skew the analysis of PBL samples, we applied additional normalization based on cell numbers in PBL samples to remove the effect of variable numbers of granulocyte, lymphocyte and monocyte subsets in each PBL sample based on measured and estimated hematological cell counts. After the normalization for variation in cell counts, we confirmed only a small proportion of the differentially methylated sites in PBL (7725 before and 806 after the normalization). In contrast, we identified 12,275 and 48,876 differentially methylated CpG sites in the purified CD4+ and CD8+ T cells (Fig. 1A). The full list of methylation changes is accessible via Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/; accession number GSE59065). The overall direction (hyper-or hypomethylation) of age-related methylation changes in CD4+ and CD8+ T cells was in strong correlation (r > 0.9) (Fig. 1B). Less than 2% of the age-related methylation sites in CD4+ and CD8+ T cells could be detected in PBLs, and majority of them were hypermethylated in older population. Nevertheless, all top sites with the highest methylation differences between younger and older individuals belonged to this group shared by CD4+ , CD8+ T cells and PBLs (Fig. 1C).

Increased methylome variation in aged CD8+ T cells. The average methylation levels between
CD4+ and CD8+ T cells from younger and older individuals were similar across different genomic locations ( Supplementary Fig. S1A). However, given the high number of methylation changes and expansion of terminally differentiated CD8+ T cells with age, we expected the CD8+ T cell methylome to be more variable among the old age group. Indeed, we found increased age-related variability among majority of CpG sites in CD8+ T cells ( Fig. 2A; note shift in methylome variability of CD8+ T cells from older individuals). This increased variation of CpG methylation in CD8+ T cells from older individuals was present in all gene and CpG island subregions ( Supplementary Fig. S1B). On average, the CpG methylation variation levels in CD4+ T cells changed less with age, with the exception of a subset of CpG sites enriched for genes involved in wound healing according to Gene Ontology (GO) analysis ( Fig. 2A, Supplementary Table S1). We then tested whether the CpG sites that were differentially methylated in CD8+ T cells of older individuals are in general more variable in somatic cells and compared our results with the methylomes from 17 human somatic tissues 28 . We found that the differentially methylated CpG sites had also higher variability in other tissues (Fig. 2B). Together the results showed that CD8+ T cells had higher age-related methylation variation than CD4+ T cells and that the majority of age-related changes in CD8+ T cells occur at CpG sites that are variably methylated in multiple tissues.
Age-related hypermethylation occurs at CpG islands and regions with repressive histone marks. We found a large proportion (> 80%) of the hypermethylated CpG sites located in CpG islands, while the hypomethylated CpG sites were more located at the borders of the CpG islands (CpG island shelves) and in gene bodies in both CD4+ and CD8+ T cells (Fig. 3A). Hypermethylation was also increased at gene regulatory (TSS200, 5´ UTR and 1 st exon) regions whereas hypomethylation was preferentially decreased in gene regulatory regions. This suggested that the hypermethylated regions are enriched in CpG islands located at 5´gene regulatory regions. We then analysed the histone modifications around the differentially methylated sites. To this end, we overlapped the differentially methylated sites in T cells with the reported locations of seven histone modifications in naïve and memory CD4+ and CD8+ T cells, available from the Roadmap Epigenomics project (Fig. 3B). The comparison showed similar distribution of histone marks in naïve and memory subsets of CD4+ and CD8+ T Scientific RepoRts | 5:13107 | DOi: 10.1038/srep13107 cells. Both age-related hypo-and hypermethylated regions were enriched in repressed chromatin marks (H3K27me3 and H3K9me3). In agreement, the differentially methylated sites had also lower levels of histone marks associated with active promoters (H3K9ac and H3K4me3), enhancers (H3K27ac) or exon regions (H3K36me3). However, another enhancer-associated modification, H3K4me1, which has been also associated with transposable elements 29 , was abundant at hypermethylated regions in older individuals. In summary, these results showed the enrichment of age-related hypermethylation at CpG islands and in regions with repressive histone marks. Age-related differential methylation occurs at repressed genes but in a subset of genes correlates inversely with expression. To show that the age-related CpG methylation occurs near transcriptionally repressed genes, we analysed the expression profiles of the CD4+ and CD8+ T cells using the HumanHT-12 v4 Expression array. We compared the differentially methylated sites in respect to the average expression level of their nearest gene. Figure 4A shows that irrespective of the direction of methylation change, the differentially methylated sites were preferentially associated with genes having low or no expression.
We then studied the correlation of DNA methylation and expression in CD4+ and CD8+ T cells (Supplementary Table S2). Importantly, in a subset of genes that were expressed in CD8+ T cells (> log 6), we observed an inverse correlation between gene expression and DNA methylation (Fig. 4B). The inverse correlation was stronger among methylation sites that were located in gene regulatory and promoter regions and among genes with higher expression levels (> log 8). Functional annotation of the genes with the inverse correlation revealed their association with cell activation involved in immune response, tyrosine phosphorylation and lymphocyte differentiation among the hypomethylated genes ( Fig. 4C, Supplementary Table S3). We then searched the promoter regions of the genes with inverse   correlation of methylation and expression in CD8+ T cells for binding sites of transcription factors, which could be involved in the recruitment of DNA methylation modifier enzymes. We found a significant overrepresentation of sites for C2H2 zinc finger family transcription factors KLF4, MZF1 and SP1 (Supplementary Table S3), regardless of the expression levels of the genes. Together, the expression analysis confirmed that the majority of the differentially methylated sites are located near genes with low transcriptional activity. However, among the genes that were differentially expressed during ageing, we found significant inverse correlation between methylation and gene expression.

DNA methylation correlates with the expression of genes involved in immune response and lineage differentiation.
To assess the potential functional effect of DNA methylation changes in aged T cells, we focused on the genes with inverse correlation between methylation and expression: 10 genes in CD4+ and 272 genes in CD8+ T cells (Supplementary Table S2). We found strong inverse correlation with the LGALS1 gene encoding galectin 1, which is known to have a strong suppressive effect on T cell mediated immune responses due to its activity to induce apoptosis of activated T cells 30 . The increased expression of LGALS1 with decreased methylation at its promoter region was present in both aged CD8+ and CD4+ T cells (Fig. 5). The other known genes with decreased methylation and increased expression in aged CD8+ T cells were the proinflammatory mediators IFNG and CCL5, as well as the cytolysis enzyme granzyme GZMH involved in CD8+ T cells effector functions ( Supplementary Fig. S2). By contrast, older individuals showed increased methylation and decreased expression of the chemokine receptor CCR7 responsible for T cell homing to lymph nodes and activation 31 , the membrane surface marker CD27 involved in T cell expansion and induction of long-term memory 32,33 and CD248 which regulates the proliferation of T cells 34 . Furthermore, we observed negative correlation for several master transcriptional regulators of the T cell lineage such as SATB1, a chromatin organizer in T cells, TCF7, a T cell specific factor that induces the T cell differentiation program and controls the expression of the CD3E molecule, as well as BCL11B and RUNX3, the crucial regulators of the T cell lineage. Among other  Supplementary  Table S2 for details). top genes with inverse correlation between methylation and gene expression we found RGS10, a regulator of G-protein signalling acting as an inhibitor of T cell adhesion and KLRD1/CD94, a receptor for the HLA-E molecules. To confirm our array results, we verified the correlations for several genes (IFNG, GZMH, CCR7, CD27, CD248, and SATB1) using the Sequenom MassARRAY and real-time quantitative PCR in a smaller subset of samples from younger and older persons (Supplementary Fig. S3).

Discussion
We here report age-related methylation changes that were identified in CD4+ and CD8+ T cells by analyzing more than 400,000 CpG sites from a total of 100 individuals.
The tissue-specificity of age-related DNA methylation has been reported previously 35 , yet several genome-wide studies have demonstrated similar, but not identical, age-related methylation changes in different tissues [36][37][38][39] . When we applied normalization based on measured granulocyte, monocyte and lymphocyte counts in our donor individuals to the blood DNA methylation profiles, the number of differential methylation sites in PBL samples decreased remarkably, indicating that the blood cell proportions that differ between individuals have impact on overall analysis of DNA methylation. However, the most significant methylation changes remained -they were shared by PBL and T cell samples and cannot be explained by the variation of blood cell proportions. This finding is in agreement with the recent study showing that most prominent epigenetic changes identified in blood cells retain their significance even after the analysis is adjusted for shifts in blood cell subtypes 21 . The top age-related DNA methylation changes found in any leukocyte subclass could accumulate in precursor cells at earlier stages of differentiation, for example during haematopoiesis, or reflect more general phenomenon of epigenetic drift with time. Nevertheless, our data show that although both CD4+ and CD8+ T cells share age-related methylation changes with PBL, many additional DNA methylation changes specific to T cells occur with age.
It should be noted that thymic involution influences T cell population during ageing 40 . T cells from older persons tend to have decreased percentages of naive cells and increased proportion of memory cells, which in CD8+ T cell compartment accumulate as terminally differentiated effector memory cells 41 . One of the limitations of our study is that the proportions of the naïve and memory cells differ between young and old people, and this could partly explain some of the methylation changes seen in our analysis. In agreement with the oligoclonal expansion and accumulation of terminally differentiated CD8+ cells, we found higher number of DNA methylation changes and increased methylation variation in aged CD8+ T cells in comparison to CD4+ cells. Whether the increased differential methylation is associated with the proliferation of CD8+ T cells in response to chronic CMV infection needs further studies.
Majority of the hypermethylated CpG sites were located in CpG islands, at silent gene promoter regions and were enriched for repressive marks such as H3K27me3, confirming the earlier reported links between age-related hypermethylation, gene inactivation and chromatin condensation 42 . Indeed, majority of age-related methylation changes seem not to affect the expression of nearby-positioned genes. However, in a subset of genes expressed in CD8+ T cells, we found a negative correlation between DNA methylation and transcription levels. Among those we identified genes with critical roles in T cell mediated immune responses. We found decreased methylation and increased expression of galectin 1 gene (LGALS1) that has multiple functions in suppressing immune responses, including controlling of T cell survival, TCR mediated signalling, regulatory T cell function, and by inhibiting anti-tumor T cell responses 30,43,44 . In addition, we found in older people the hypomethylation at the IFNG promoter, which correlated with the higher expression of the gene in their CD8+ T cells. Previous studies have shown high production of IFNγ the major proinflammatory cytokine, by activated CD8+ T cells after the stimulation by CMV antigens 10,45 . Decreased levels of DNA methylation and H3K4me3 repressive marks have been found at Ifng gene after the activation and differentiation of mouse CD8+ T cells in response to infections 46 . We also found hypomethylation and increased expression of the proinflammatory chemokine CCL5, the plasma levels of which are known to increase with age 10 , and of the GZMH gene, which is upregulated in effector T cells during infections and in chronic inflammatory diseases 47 . Demethylation of the Ifng, Ccl5 and granzyme genes occurs during viral infection-induced differentiation of mouse effector and memory CD8+ T cells 48,49 . Furthermore, our finding of promoter hypermethylation of costimulatory CD27 and chemokine CCR7 receptor genes is in agreement with their downregulation in terminally differentiated anergic CD8+ T cells observed in aged individuals 41 . In this light, it is tempting to speculate that age-related chronic viral infections, such as CMV, may induce extensive oligoclonal proliferation of CD8+ T cells and result in changed DNA methylation profiles at genes involved in T cell responses to viral infections and in chronic inflammation.
T cell differentiation programme to specialized effector cells is guided by the action of several distinct transcription regulators. In CD8+ T cells, we identified age-related hypermethylation at several transcriptional regulator genes required for T cell lineage differentiation. SATB1, the T lineage-enriched global chromatin organizer, has important roles in T cell development and proliferation and ensures proper development of the lineage 50,51 . Furthermore, three other genes regulating T cell differentiation, TCF7, BCL11B and RUNX3 52-55 were hypermethylated and had lower expression level in CD8+ T cells from older individuals. The expression of TCF7, BCL11B and RUNX3 genes is required for the mature CD8+ T cell differentiation and is decreased with the acquisition of effector cell phenotype 56,57 . Our data thus show that ageing is associated with decreased expression and DNA hypermethylation of central T cell specific transcriptional regulator genes with fundamental roles in CD8+ T cell differentiation. Together, our results support the concept that the silencing of transcriptional regulator genes by DNA hypermethylation during ageing directs the gene expression profile towards the terminally differentiated effector CD8+ T cells.
In conclusion, although epigenome-wide studies with PBL have identified genes with methylation changes associated with age, the purification of cell subtypes allows more precise investigation of changes relevant to the altered function of specific cells. Our study shows that most of the gains in methylation that occur in ageing T cells are in transcriptionally repressed genes. However, the DNA methylation changes that are accompanied with gene expression changes affect many genes that are essential for the differentiation and function of T cells and shed light on the possible causes of the age-related decline in immune response. In addition, our study forms the basis to further evaluate the potential use of the identified DNA methylation changes as clinical markers of immunosenescence in older individuals.

Materials and Methods
Ethics statement. The study was approved by the Ethics Review Committee of Human Research of the University of Tartu, Estonia (permission no 206/T-4, date of issue 25.08.2011) and it was carried out in compliance with the Helsinki Declaration. All of the participants were older than 18 and a written informed consent to participate in the study was obtained from each individual prior to recruitment. All participants were healthy donors of the Estonian Genome Center of the University of Tartu. All methods were carried out in accordance with approved guidelines.
Purification of cell populations. Peripheral Supplementary Table S4. RNA extraction, labelling, hybridization, and qPCR. RNA was extracted from the purified T-cells using the miRNeasy Mini Kit combined with recommended RNase-free DNase I treatment (both from Qiagen). RNA from whole blood was purified using the MagMAX ™ for Stabilized Blood Tubes RNA Isolation Kit. RNA was concentrated using the Heraeus vacuum centrifugation system without heating. RNA was labeled and amplified using the TargetAmp-Nano Labeling Kit for Illumina Expression BeadChip (Epicentre Biotechnologies) with SuperScript III Reverse Transcriptase (Life Technologies), followed by purification with the RNeasy MinElute Cleanup Kit (Qiagen). RNA quality was assessed after extraction and after labelling using an Agilent 2100 Bioanalyzer and Agilent RNA 6000 Nano Kit (all from Agilent Technologies). Labeled RNA was hybridized to the HumanHT-12 v4 Expression BeadChip (Illumina) according to the manufacturer's instructions. Verification of the expression of changed genes in CD4+ and CD8+ T cells was performed by qPCR using primers listed in Supplementary Table S4. Normalization of methylation data. Data pre-processing and quality control analyses were performed in R with the Bioconductor package minfi, using the original IDAT files extracted from the HiScanSQ scanner. 'Raw' pre-processing was used to convert the intensities from the red and the green channels into methylated and unmethylated signals. Beta values were computed using Illumina's formula [beta = M/(M + U + 100)]. The difference in the distribution of beta values for type I and type II probes was corrected using "SWAN" 58 , a normalization method to address systematic changes between type I and type II probes. Detection p-values were obtained for every CpG probe in each sample. Failed positions were defined as signal levels lower than background from both the methylated and unmethylated channels.
Samples with detection p-value > 0.01 in more than 10% of the CpG sites and possibly contaminated or mixed up samples (determined using the 65 SNPs present on the methylation beadchip) were discarded. We mapped all the probe sequences to the human genome (build hg19) and discarded all probes that did not map uniquely, mapped to CpG sites with SNPs up to 3 bp upstream of the site, or mapped to the X/Y chromosomes. After data preprocessing, the total number of samples left for analysis was 296 (97+99+100). The total number of retained CpG sites was 399,653. The full data is stored at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) with accession number GSE59065.
Differential methylation analysis. Before performing differential methylation analysis the effects of known confounding variables were removed. For samples from CD4+ and CD8+ cells the effect of sex and array was corrected for. In the case of PBLs the levels of erythrocytes, platelets, monocytes, lymphocytes and granulocytes that were measured by blood biochemistry were additionally included as covariates. The correction was performed by fitting a linear model with these covariates to every CpG site and using the residuals for further analysis. The differential methylation analysis of younger and older persons was performed using moderated t-test from the Bioconductor limma package 59 . CpG sites with Scientific RepoRts | 5:13107 | DOi: 10.1038/srep13107 an average methylation difference > 0.05 between the younger and the older subjects, and false discovery rate (FDR) < 0.05 were considered differentially methylated. Differential expression analysis. Differentially expressed genes were identified using a moderated t-test from the Bioconductor limma package 59 . The same confounding factors used in the methylation analysis were applied for the different cell types. FDR < 0.05 was used as threshold in all comparisons.
Gene Ontology analysis. All Gene Ontology enrichment analyses were performed using the g:Profiler web toolkit 60 and R package GOsummaries (http://www.bioconductor.org/packages/release/bioc/ html/GOsummaries.html). To display less redundant and more specific results, GOsummaries uses several non-default filtering options in g:Profiler. First, we limited the number of genes in a category; the upper limit of the number of genes was set to 1000 and the lower limit was 50 genes. Second, the program uses a "Best per parent" hierarchical filtering option to remove categories that are closely related on a GO tree. Finally, the results from "Cellular Component" and "Molecular Function" sections of GO were removed, as too general and not informative. The statistical significance threshold was 0.05 for the false discovery rate provided by g:Profiler.
Gene part overlaps. The CpG site associations with gene subregions were obtained from the annotation file provided by Illumina. In the file one site can be annotated to multiple regions if it is associated with more than one gene/island subregion.
Histone modification overlaps. Roadmap Epigenomics 61 histone modification data of naïve and memory CD4+ and CD8+ T cells was downloaded in Wig format from http://www.roadmapepigenomics.org/data . The read counts were extracted from the files for all differentially methylated and 20,000 non-differentially methylated CpG sites using the rtracklayer R package 62 . To make the scores comparable between modifications and cell types, these were log2 transformed (using pseudocount 1) and converted into z-scores based on the background distribution of the 20,000 sites. The coefficients and confidence intervals shown in Fig. 3 were obtained from a linear model that predicted the scores using the direction of differential methylation. In order to account for the preferential distribution of histone modifications along the genes and CpG islands, this information was also included in the linear models.
Correlation between methylation and gene expression. The CpG sites were associated with genes based on annotated gene names (provided by Illumina). The gene associations to gene expression probes were obtained using gConvert tool of the g:Profiler web toolkit 60 . In the correlation plots differentially methylated/expressed genes were used (described above). The plots with gene transcripts were drawn using the Gviz package (http://www.bioconductor.org/packages/devel/bioc/html/Gviz.html), and were based on the known genes table from the UCSC genome browser, genome build hg19.