Aging is associated with highly defined epigenetic changes in the human epidermis
© Raddatz et al.; licensee BioMed Central Ltd. 2013
Received: 25 September 2013
Accepted: 8 October 2013
Published: 31 October 2013
Skip to main content
© Raddatz et al.; licensee BioMed Central Ltd. 2013
Received: 25 September 2013
Accepted: 8 October 2013
Published: 31 October 2013
Altered DNA methylation patterns represent an attractive mechanism for understanding the phenotypic changes associated with human aging. Several studies have described global and complex age-related methylation changes, but their structural and functional significance has remained largely unclear.
We have used transcriptome sequencing to characterize age-related gene expression changes in the human epidermis. The results revealed a significant set of 75 differentially expressed genes with a strong functional relationship to skin homeostasis. We then used whole-genome bisulfite sequencing to identify age-related methylation changes at single-base resolution. Data analysis revealed no global aberrations, but rather highly localized methylation changes, particularly in promoter and enhancer regions that were associated with altered transcriptional activity.
Our results suggest that the core developmental program of human skin is stably maintained through the aging process and that aging is associated with a limited destabilization of the epigenome at gene regulatory elements.
Epigenetic mechanisms regulate the interpretation of genetic information and are thus intricately linked to cellular differentiation and tissue specification. Epigenetic mechanisms include DNA methylation and covalent histone modifications [1, 2]. These mechanisms work in a coordinated manner during development and differentiation to execute specific gene expression programs [3, 4].
DNA methylation is a conserved epigenetic mechanism with a well-known role in cell fate specification [5, 6]. In mice, this function appears to be essential for organismal development, as genetic deficiencies for DNA methyltransferases cause severe developmental phenotypes and embryonic lethality [7, 8]. More recently, significant attention has also been focused on adaptive functions of DNA methylation, which could provide the foundation for the integration of environmental signals . Initially observed as epigenetic variations between monozygotic twins , age-related methylation differences have now been described in several independent studies and tissues . The results suggested that aging induces global and complex changes in the human methylation landscape [12–15]. However, it is presently unclear how these epigenetic changes affect the gene expression patterns of aging human tissues.
We have previously used array-based methylation profiling of skin samples obtained from young and old volunteers and identified a defined set of genes that were hypermethylated in aged skin . In the course of this study, two important features were identified that established the epidermis as an important tissue for further analysis: i) epidermis samples can be obtained by non-invasive procedures and are characterized by a very high degree of cell type homogeneity (>90% keratinocytes), which greatly facilitates the identification of defined methylation patterns; and ii) the epidermis represents a tissue with high functional relevance for human health and disease and shows a well-known aging phenotype . Furthermore, it has also been shown that the DNMT1 DNA methyltransferase is functionally required for proper tissue homeostasis in the human epidermis .
Aging is a multifactorial process that results in a progressive loss of regenerative capacity and tissue functionality. It has previously been suggested that these pathological changes are underpinned by a genome-wide loss of methylation marks . We have now used transcriptome sequencing to identify genes that are differentially expressed in the young (18–24 years) and old (70–75 years) human epidermis. This revealed a highly specific set of 75 genes with strong functional relevance for human skin homeostasis. We also used whole-genome bisulfite sequencing to generate methylation maps from the same tissue samples. We find that the aging epidermis methylome is characterized by considerable stability and does not show any global methylation loss. However, local age-related methylation changes could be observed, and were enriched at weak or poised promoters and at enhancer regions, thus suggesting a functional relevance of epigenetic mechanisms for skin aging.
Number of read pairs
Young (n = 5)
Old (n = 5)
Expression levels of epidermis genes and DNA methylation factors
Epidermis structural factors
Fibroblast markers (dermis)
DNA methylation factors
Further analysis of differentially expressed genes identified several factors that have previously been implied in skin homeostasis. For example, connective tissue growth factor (CTGF), which has been shown to be an important factor in skin aging , was expressed at distinctly lower levels in the old epidermis (Figure 1C). We also used pathway analysis of differentially expressed genes to further characterize the functional effects of age-related differential gene expression. This revealed that genes involved in cell migration (P = 1.17 × 10-16), cancer (P = 9.55 × 10-13), dermatological diseases (P = 2.07 × 10-11) and cell proliferation (P = 1.4 × 10-10), constituted the categories with the most significant enrichment (Figure 1D). In addition, functional annotation of differentially expressed genes also showed a highly significant enrichment of genes involved in the development of the epidermis (P = 9.63 × 10-4), which also included downregulation of SPRR1A and B, KRT16 and KRT17 in old samples. In agreement with this finding, expression of genes involved in differentiation of keratinocytes was also altered in old samples, indicating an overall downregulation of differentiation (P = 1.43 × 10-7). Altogether, our results thus suggest that deregulation of a relatively small set of genes could contribute to the phenotypic changes associated with human skin aging.
Having shown that aging is associated with the deregulation of a highly defined set of genes, we used whole-genome bisulfite sequencing to establish DNA methylation maps at single-base resolution. DNA was purified from the same epidermis samples that were used for transcriptome sequencing (Additional file 1: Table S1). Pooling of samples was necessary to achieve sufficient amounts of DNA for library preparation and has been previously used to reduce the effects of stochastic epigenetic variation . Paired-end sequencing on an Illumina HiSeq 2000 platform with read-lengths of 105 bases generated 137 Gb of DNA sequence. After trimming to a maximal read length of 80 bases and a minimum base quality of a 30 Phred score, sequence reads were mapped to the GRCh37/hg19 reference sequence using a mapping tool based on BSMAP 2.0. The resulting average strand-specific genome coverage was 11.3x (young) and 11.9x (old). We also determined the bisulfite conversion rate by analyzing mitochondrial sequences that were co-purified during the sample preparation and that we considered as unmethylated. These sequences showed a bisulfite conversion rate of >99.8% (Table 1), suggesting highly effective bisulfite treatment.
Overall, the methylation patterns of the young and old samples appeared very similar. This was evident not only by the average methylation ratios of individual genome compartments (Figure 3C), but also in comparisons of the global methylation landscapes (Figure 3D). A sliding window approach identified only 50 differentially methylated windows of 100 kb (methylation difference >0.15), with an equal number of hypomethylated and hypermethylated windows (Figure 3E). Similarly, a more local analysis with sliding windows of 5 CpGs did not reveal any directional changes in global methylation patterns (Figure 3F). Together, these findings strongly suggest that the global age-related methylation loss observed in T-cells  is not conserved in the epidermis.
A visual inspection of the young and old methylation landscapes also indicated the presence of small clusters of differentially methylated CpG dinucleotides. To systematically identify differentially methylated regions (DMRs), Fisher’s exact test was used to determine the CpG dinucleotides with a statistically significant (P <0.05) methylation difference. These differentially methylated CpGs (DMCs) were subsequently collapsed to identify regions of local, coordinated methylation changes. DMRs were defined as clusters of ≥8 DMCs with a distance of ≤50 bp between neighboring DMCs and a net region-wide methylation change of ≥8 DMCs. Only DMRs with an average sequencing coverage of ≥8 and methylation difference of ≥10% were used for further analysis. This identified 2,409 DMRs, of which 1,437 were more strongly methylated in the old sample, and 972 were more strongly methylated in the young sample. DMRs were comparably small (<150 bp) and associated with various gene regions. Notably, 1,156 of these DMRs overlapped with the variably methylated regions that were recently identified through a comprehensive analysis of 42 human methylomes . This represents a robust (2.5-fold) enrichment over the genome average and suggests that age-related methylation changes affect epigenetic regulatory elements.
A visual inspection of DMRs provided further insight into their characteristic features. For example, the ERBB receptor feedback inhibitor 1 (ERRFI1) promoter region harbors a DMR in the shore region of the promoter-associated CpG island (Figure 4C). ERRFI1 is required for proper epidermal homeostasis  and was found to be hypermethylated and expressed at lower levels in old epidermis samples (Figure 4C). Another example for a DMR was identified in the low-density lipoprotein receptor (LDLR) gene region (Figure 4D), and was located in an annotated active enhancer element (Additional file 4: Figure S1). The DMR showed complex, but coordinated age-related methylation changes that were associated with lower expression levels of LDLR in the old sample (Figure 4D). Defects in the LDLR gene are the cause of familial hypercholesterolemia, which underlies the formation of Xanthelasma , a dermatological lesion often found in the elderly population. These examples illustrate how age-related methylation changes can have relatively subtle, but significant expression changes on genes that are relevant for proper skin homeostasis. Further experiments will be required to determine the functional relevance of individual differentially methylated and expressed genes for the aging phenotype.
Whole-genome bisulfite sequencing represents a powerful method to generate genome-wide methylation maps at single-base resolution [24, 32]. This approach was recently used to characterize the methylomes of purified CD4+ T-cells from a newborn and a centenarian . The results showed a pronounced destabilization of the aging epigenome, which was characterized by a widespread loss of methylation marks. Our analysis failed to detect any quantitative differences in the global methylation levels of young and old individuals. This can most likely be attributed to the different tissues used in both studies: while our analysis is based on primary epidermis samples, Heyn et al.  used purified CD4+ T-cells. Based on the specific features of T-cell differentiation and priming, it is conceivable that the epigenetic program of these cells is characterized by a particularly high degree of plasticity.
On a global level, the transcriptomes and methylomes of the young and old epidermis appeared to be substantially similar. This is an important finding, because it illustrates the fundamental stability of the tissue-specific methylation landscape, which is required for the stable maintenance of cell type identity. Age-related methylation changes were limited to specific local alterations, which confirms and expands our previous observations . With a size of 100–150 bp, age-related DMRs were considerably smaller than other known structures of the human methylome, such as DNA methylation valleys  and partially methylated domains [23–25, 33, 34], which extend over tens and hundreds of kilobases of DNA sequence, respectively. Further analysis will be required to understand the epigenetic regulatory function(s) of these elements.
Interestingly, our data also suggest that a significant fraction of the DMRs might represent enhancers that become aberrantly methylated during aging. The methylation status of enhancer-associated CpG dinucleotides has recently been described to be closely associated with epigenetic gene deregulation in human cancers [35, 36]. Our data identify similar elements at single-base resolution and suggest that the differential methylation of enhancers might be involved in age-related gene deregulation.
Finally, our analysis identified several examples for hypermethylated DMRs in promoter regions. Promoter hypermethylation has been closely associated with gene silencing and plays an important role in the etiology of human tumors [37, 38]. Furthermore, our data indicate an association of hypermethylated DMRs with bivalent chromatin structures. Similar results have been described in independent studies investigating age-related methylation changes in other tissues [13, 14]. Bivalent chromatin modifications are a specific feature of stem cells and are not usually found in differentiated tissues like epidermis. The particular enrichment of age-related hypermethylation in promoters that are annotated as “bivalent” might therefore reflect epigenetic changes in aging epidermal stem cells, and may underpin the decreased regenerative capacity of aging stem cells .
Age-related DNA methylation changes have been described in various studies but their defining features and functional significance remain to be established. Our results show that the age-related genome-wide loss of DNA methylation observed in T-cells  is not conserved in the human epidermis, suggesting that the core developmental program of at least some human tissues is maintained through the aging process. Furthermore, our study represents the first to use transcriptome sequencing to explore the consequences of age-related epigenetic changes. Our results identified limited transcriptional changes that may underpin the aging phenotype. Finally, we identified several hundred highly localized elements with robust methylation differences between young and old. Interestingly, these elements were enriched for gene regulatory chromatin marks and many of them were associated with promoters and enhancers. Our study thus provides an important mechanistic framework for understanding age-related epigenetic deregulation.
Epidermal suction blister samples were collected according to the current version of the Declaration of Helsinki and the guideline of the International Conference on Harmonization Good Clinical Practice (ICH GCP) was observed as applicable to a non-drug study. All volunteers provided written, informed consent. Suction blister samples were obtained at the study center of Beiersdorf AG and approved by the Beiersdorf AG Legal Review Board. Briefly, epidermis samples were detached from the forearms of 10 healthy female volunteers by applying a negative pressure of 150–250 mmHg for 2 h. Subsequently, suction blister roofs were taken and immediately stored in liquid nitrogen. For nucleic acid isolation, suction blister samples were washed in DPBS (Cambrex, Verviers, Belgium) and homogenized using a TissueLyser (Retsch, Haan, Germany). DNA and RNA from suction blister samples were isolated using the QIAamp DNA Investigator Kit (Qiagen, Hilden, Germany) and RNeasy Fibrous Tissue Kit (Qiagen), respectively, according to the manufacturer’s instructions. The Poly(A)Purist MAG Kit (Ambion, Darmstadt, Germany) was used for mRNA selection.
Library preparation for bisulfite sequencing was performed as described previously . Transcriptome sequencing libraries were prepared using the TruSeq RNA Sample Preparation Kit (Illumina, San Diego, USA), according to the manufacturer’s instructions. Paired-end sequencing was performed on an Illumina HiSeq system with read lengths of 105 base pairs and an average insert size of 200 bp.
RNA-seq reads were trimmed to a maximal length of 80 bp and stretches of bases having a quality score <30 at the ends of the reads were removed. Reads were mapped using Tophat 2.0.6 . As reference sequence for the transcriptome mapping we used the current assembly of the human genome (hg19). Differential expression was quantified using DESeq 1.10.1  applying the built-in procedures for library normalization and estimation of variance and with Cuffdiff 2.0 . The resulting P values were subjected to multiple testing correction using built-in functions available in DESeq and Cuffdiff, respectively. Genes with a q value smaller than 0.05 were considered as differentially expressed.
Pathway analysis of differentially expressed genes was performed using IPA (Ingenuity Systems), using a P value cutoff of <0.05 on differential expression and a log fold-change of at least 0.263, corresponding to a minimum expression change of 30% between young and old samples.
Total RNA was isolated from suction blister samples (n = 18) using the RNeasy Fibrous Tissue Kit (Qiagen) according to the manufacturer’s instructions. After reverse transcription with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Darmstadt, Germany), samples were analyzed by TaqMan-PCR using the 7900HT Fast-Real-Time PCR System (Applied Biosystems). The following assays were used, according to the manufacturer’s recommendations: CTGF (Hs01026927_g1), SPRR1A (Hs00954595_s1), SPRR1B (Hs00234164_m1), KRT16 (Hs00373910_g1), KRT17 (Hs01588578_m1), VCL (Hs00247826_m1), HAS3 (Hs00193436_m1), FDZ10 (Hs00273077_s1), NPR3 (Hs00168558_m1), VPS37B (Hs00226582_m1), ERRFI1 (Hs00219060_m1), PLIN2 (Hs00605340_m1), RHPN2 (Hs00369111_m1), RGMA (Hs00297192_m1), PI4KB (Hs01090927_m1). Data were analyzed utilizing the Sequence detector version 2.3 software supplied with the 7900 Sequence Detector and RQ Manager 1.2. Quantification was achieved using the ΔCt method which indicates expression of the target gene relative to an endogenous reference (GAPDH; Hs99999905_m1). ΔCt values were averaged for all replicates of a gene/subject combination and for every gene the mean value over all subjects was subtracted, thus adjusting the average gene-specific ΔCt value to zero. The processed ΔCt values were visualized as heat maps. ΔCt values exceeding a threshold of 2.5 were set to 2.5.
Reads were trimmed to a maximal length of 80 bp and stretches of bases having a quality score <30 at the ends of the reads were removed. Reads were mapped using BSMAP 2.02 . As a reference sequence for the bisulfite mapping we used the current assembly of the human genome (hg19). Only reads mapping with both partners of the read pairs at the correct distance were used. The CpG-specificity was calculated by determining the number of cytosines called in all mapped reads at all non-CpG positions and dividing it by the number of all bases in all mapped reads at all non-CpG positions. Methylation ratios were determined using a Python script (methratio.py) distributed together with the BSMAP package. For both the forward and reverse strands, all cytosine bases in CG context were called independently.
Fisher’s exact test was used to identify 3,004,806 CpG dinucleotides with a statistically significant (P <0.05) difference in their methylation ratios between the young and the old datasets. Differentially methylated CpGs were subsequently collapsed to identify regions of local, coordinated methylation change. DMRs were defined as clusters of ≥8 DMCs with a distance of ≤50 bp between neighboring DMCs and a net region-wide methylation change of ≥8 DMCs. Only DMRs with an average sequencing coverage of ≥8 and methylation difference of ≥10% were used for further analysis. ChIP-seq peaks of histone modification marks in NHEK cells were downloaded from the ENCODE website  and averaged in 50 bp windows surrounding the DMR sites. Loess smoothing with a span of 10% was applied. For the enrichment analysis of DMRs in defined chromatin states, the observed frequency (%) of DMRs centered in given states (ChromHMM annotation data in NHEK cells, ENCODE website) was divided by the genomic frequency of the state.
Sequencing data have been deposited in the GEO database under the accession number GSE46487.
Connective tissue growth factor
Differentially methylated CpG
Differentially methylated regions
ERBB receptor feedback inhibitor 1
Low-density lipoprotein receptor.
We thank André Leischwitz and the DKFZ Genomics and Proteomics Core Facility for sequencing services. We also thank Franz Stäb and Horst Wenck for their support. This work was funded by the BMBF-GerontoSys program through the project AGENET (FKZ0315898).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.