Epigenetic Mechanisms of the Aging Human Retina.

Degenerative retinal diseases, such as glaucoma, age-related macular degeneration, and diabetic retinopathy, have complex etiologies with environmental, genetic, and epigenetic contributions to disease pathology. Much effort has gone into elucidating both the genetic and the environmental risk factors for these retinal diseases. However, little is known about how these genetic and environmental risk factors bring about molecular changes that lead to pathology. Epigenetic mechanisms have received extensive attention of late for their promise of bridging the gap between environmental exposures and disease development via their influence on gene expression. Recent studies have identified epigenetic changes that associate with the incidence and/or progression of each of these retinal diseases. Therefore, these epigenetic modifications may be involved in the underlying pathological mechanisms leading to blindness. Further genome-wide epigenetic studies that incorporate well-characterized tissue samples, consider challenges similar to those relevant to gene expression studies, and combine the genome-wide epigenetic data with genome-wide genetic and expression data to identify additional potentially causative agents of disease are needed. Such studies will allow researchers to create much-needed therapeutics to prevent and/or intervene in disease progression. Improved therapeutics will greatly enhance the quality of life and reduce the burden of disease management for millions of patients living with these potentially blinding conditions.

Background

Proper control of gene expression is fundamental to an organism’s viability, affecting every function of each cell, from maintaining homeostasis to responding to environmental changes. One level of control used by cells to regulate gene expression is to alter the structure of the DNA–protein–RNA complex known as chromatin (Table 1). Gene expression changes in response to environmental stimuli do not necessarily require remodeling of chromatin. However, mechanisms that underlie such gene expression changes by modifying the chromatin structure without changing the nucleotide sequence are called epigenetic mechanisms. To better understand the role of epigenetic mechanisms in mammalian gene regulation, one must first understand the role of chromatin structure in the eukaryotic cell.

Table 1

Definitions of key terminology.

TERM	DEFINITION
Acetylation	The addition of an acetyl group to lysine amino acid residues
Chromatin	The DNA molecule in combination with the proteins and RNA bound to it
Epigenetic modifications	Stable changes to chromatin structure that do not change the nucleic acid sequence, such as DNA methylation and histone modifications
Epigenetics	The study of epigenetic modifications
Fundus photograph	Picture of the back of the eye, including the retina, macula, fovea, and optic disk
H3K9me3	Histone modification associated with transcriptionally inactive chromatin: three methyl groups attached to the lysine at amino acid position 9 of histone H3
Histone	One of the five proteins that are the primary protein components of chromatin
Heterochromatin	Structurally condensed, transcriptionally inactive regions of chromatin
HP1	Heterochromatin protein 1, protein associated with heterochromatin which maintains the heterochromatic state
Methylation	The addition of a methyl group to cytosine DNA residues; or the addition of one or more methyl groups to histone lysine or arginine amino acid residues
Optical coherence tomography	A method for capturing a cross-sectional image of tissue
Phosphorylation	The addition of a phosphoryl group to lysine, threonine, or tyrosine amino acid residues
Pioneering transcription factors	Transcription factors which bind to repressed, heterochromatic DNA to facilitate activation of a gene
Transcription factor	1) any protein involved in transcription; 2) sequence-specific DNA binding proteins which facilitate the transcriptional activation or repression of target genes
Ubiquitination	The addition of one or more ubiquitin groups to lysine amino acid residues

In eukaryotic cells, various proteins play important roles in establishing and maintaining the structure of DNA within a cell. Histone proteins are the primary protein components of chromatin and are responsible for condensing the DNA into nucleosomes that form a beads-on-a-string conformation (Fig. 1). Nucleosomes are formed by the wrapping of ~150 bp of the DNA double helix around histone octamers comprising two each of the core histone proteins H2A, H2B, H3, and H4 (Fig. 1). An additional histone, H1, is known as the linker histone because it binds to the DNA between the histone octamer beads and contributes to further condensation of the DNA. The positions of nucleosomes along the DNA helix affect the ability of other DNA-binding proteins, such as transcription factors (TFs), to access specific sequences of DNA, thereby influencing the expression of genes. Nucleosomes are highly repressive to transcription by their restriction of DNA accessibility.1

Figure 1

Chromatin modifications affect gene expression. Changes in epigenetic marks, such as DNA methylation and histone modifications, each contribute to the regulation of gene expression. DNA methylation in the promoter regions of genes is generally associated with decreased gene expression. Histone modifications can be either activating or repressive. Histone acetylation and phosphorylation are generally associated with active genes; histone methylation and ubiquitination arrangements are associated with either active or repressed genes.

Additionally, the N-terminal ends, or tails, of the core histones are less structured and have regulatory roles (Fig. 1). Posttranslational modifications (PTMs) primarily to the N-terminal tails of histone proteins—including methylation, phosphorylation, acetylation, and ubiquitination—affect the conformation of the histone-DNA complex and, therefore, the accessibility of the DNA. For example, acetylation of histone tails is associated with actively expressed genes and generally results in the histones being more negatively charged and, therefore, binding to the DNA and each other less tightly.2 This allows more TF–DNA binding and local expression of relevant genes.

Histone modifications also serve as markers recognized by specific TFs involved in activation or repression of eukaryotic gene expression. Much effort of late has gone into deciphering the so-called histone code, and thereby understanding the roles of each of these marks at each of their many respective positions on histone tails within the context of the genome.3 Acetylation marks are associated with active promoters and are among the most well-studied histone modifications. Histone phosphorylation is also associated with active transcription.4 Methylation or ubiquitination of histone tails is often associated with either activation or repression of gene expression, depending on the specific amino acid residue being modified, the extent to which that given residue is modified, and the position of that residue within a gene or extragenic region.4,5

A number of writers, readers, and erasers are involved in adding, interpreting, and/or removing epigenetic modifications on chromatin. For example, histone acetyl-transferases add (write) and histone deacetylases (HDACs) remove (erase) acetyl groups to/from histone lysine residues. Bromodomain-containing DNA-binding proteins, such as the TFIID subunit TAF1, specifically recognize acetylated lysine residues on histone tails (read) to facilitate promoter recognition and transcriptional activation.6,7 Aberrant histone acetylation has been implicated in various pathologies, and HDAC inhibitors are in clinical trial to treat cancer and have been suggested for the treatment of retinal degenerative diseases.8 One class of HDACs known as sirtuins have been implicated in the aging process, including evidence that suggests that they facilitate the lifespan-extending effects of calorie restriction in model organisms and that their activation may be beneficial for age-related diseases, including neurode-generative diseases.9

In addition to histone modifications, modifications to the DNA itself have profound effects on gene expression. DNA methylation—the addition of a methyl group to the C5 position of cytosine bases that are followed by guanosine bases within the DNA sequence (5′-CpG-3′)—is strongly associated with the repression of gene expression. CpG-rich regions of the genome, referred to as CpG islands, are often found in the 5′ regulatory regions of genes.10 Actively expressed genes generally have unmethylated CpG islands near their transcription start sites, whereas unexpressed genes generally have methylated CpG islands near their transcription start sites. Methylation patterns can be heritable across both meiotic and mitotic cell divisions. In genomic imprinting, for example, methylation is used to ensure that single copies of particular genes are repressed during egg or sperm cell generation in a maternal- or paternal-specific pattern, which persists into the adult.11,12 DNA methylation is also used during development to program cell differentiation by specifying the particular subset of genes to be expressed by each cell type.13 Of particular relevance to the development of age-related disease is that environmental factors induce DNA methylation changes throughout an organism’s lifespan.14 Figure 2 illustrates a number of environmental factors associated with retinal degenerations such as age-related macular degeneration (AMD) that have also been shown to associate with changes in DNA methylation, including tobacco smoke, alcohol, diet (particularly omega-3 fatty acids and antioxidants), and obesity.15–27 The cumulative effects of each of these factors on the epigenetic landscape of an individual’s genome may gradually perturb the expression profile of a tissue sufficiently to transform it from an otherwise healthy state to a diseased state.

Figure 2

Epigenetics in aging and age-related disease. Stable epigenetic marks may be the link between genetic and environmental processes involved in the development of age-related diseases, such as AMD. Environmental influences contribute to epigenetic changes that accumulate with age. Risk factors, such as diet, obesity, smoking, sun exposure, and age, may elicit epigenetic changes that accumulate over a lifetime, eventually resulting in altered expression of genes involved in the disease process. These environmental influences contribute to epigenetic modifications, such as DNA methylation (green ovals), histone methylation (purple pentagons), histone acetylation (orange pentagons), histone ubiquitination (blue pentagons), and histone phosphorylation (red pentagons). The epigenetic changes that accumulate throughout the genome may associate with transcriptional changes at the affected genomic loci. Such expression changes at disease-relevant loci may promote either protection against or progression of age-related disease. The sum of these effects over time may then perturb the normal, healthy homeostasis enough to result in the development and/or progression of diseases such as AMD.

Such gradual epigenetic modifications may play significant roles in the development of age-related retinal diseases. The retina is a region at the back of the eye onto which the lens focuses an image and which contains the photoreceptor cells that transmit the image as a signal to the brain. The macula is the region of the retina responsible for central vision, with the fovea at its center. Figure 3A shows a fundus image (photograph of the back of the eye) and an optical coherence tomography (OCT) image (which shows a cross-section of the macula) of a healthy retina. In the OCT image, the retina is the upper cellular layer, with the retinal pigment epithelium (RPE) and choroid below (posterior to) the retina. The choroid comprises blood vessels that support the retina, with the RPE serving as the inner blood–retinal barrier through which nutrients and waste must pass to transfer between the retina and the choroidal blood supply.28

Figure 3

Progression of retinal disease. (A) Fundus and OCT images of normal eye with anatomy labels. (B) Fundus image of an eye with glaucoma. Note the optic nerve damage. (C) Fundus images of normal, intermediate, and advanced (GA and neovascular) AMD. Note the drusen deposits, atrophy, and neovascularization. (D) Fundus image of an eye with DR. Note the abnormal blood vessels and hard exudates. Photographs taken from DeAngelis laboratory patient cohorts. The study protocol was reviewed and approved by the Institutional Review Board at the University of Utah and conforms to the tenets of the Declaration of Helsinki.

Definition of Epigenetics

As stated earlier, changes to DNA structure that do not involve changes to the DNA sequence are considered epigenetic. The definition of the term epigenetic has been met with some controversy as the study of chromatin modifications has advanced. The more conservative definitions limit the term to refer specifically to inheritable changes in chromatin structure, and the more liberal definitions use the term to encompass any nonsequence changes to chromatin structure. Two distinct definitions are worth noting. A consensus definition of epigenetics from a 2008 meeting at Cold Spring Harbor Laboratories resulted in the following: “An epigenetic trait is a stably heritable phenotype resulting from changes in a chromosome without alterations in the DNA sequence.”29 Note the requirement in this definition for the changes to be stably heritable. In contrast, the NIH Roadmap Epigenomics Mapping Consortium defines epigenetics for their purposes as, “both heritable changes in gene activity and expression (in the progeny of cells or of individuals) and also stable, long-term alterations in the transcriptional potential of a cell that are not necessarily heritable.”30 The use of the term epigenetics to refer to non-heritable chromatin changes in postmitotic cells, as discussed herein, violates the conservative requirement for heritability and therefore requires a more liberal application of the term in line with the NIH Roadmap Consortium definition.

The epigenetic landscape of a particular gene or genome can be affected by various factors, including genetic and environmental influences, as noted earlier. The activation and repression of gene expression is a complex interplay of numerous factors that must be tightly regulated to ensure proper response to environmental signals. Upon detection of an environmental signal, sequence-specific DNA-binding transcription activators or repressors (TFs) are activated alternatively by various methods, such as: (1) Directly binding to the signal, e.g., the nuclear receptor family of TFs in metazoans, including the estrogen receptor alpha and the retinoic acid receptor.31,32 (2) Sensing a disruption to the cellular homeostasis caused by the environmental signal, e.g., protein kinase R-like endoplasmic reticulum kinase (PERK) sensing misfolded proteins in the endoplasmic reticulum of eukaryotes, leading to activation of TFs involved in apoptosis.33 (3) Through a signaling cascade propagated from the extracellular signal-binding receptor, e.g., vascular endothelial growth factor (VEGF) binding to VEGF receptors to induce downstream activation of several TFs in metazoans.34 Once activated, the TFs bind to target sequences within regulatory regions of the genome to modulate the transcription of target genes. Transcriptional activators recruit RNA polymerase II (RNA pol II) and other transcriptional machinery, including chromatin remodelers, to the gene to initiate transcription. The so-called pioneering TFs bind to their target sequence even when the chromatin is in a repressed heterochromatic state, in which the chromatin is structurally condensed and transcriptionally inactive, and can initiate events that lead to the opening of the chromatin and activation of the affiliated gene(s).35

Alternatively, transcriptional repressors can bind to their target sequences and recruit chromatin remodeling proteins to convert the local chromatin to, or maintain it in, a heterochromatic state. For example, the Krüppel-associated box (KRAB) zinc finger proteins are sequence-specific DNA-binding transcriptional repressors that induce the formation of heterochromatin through the recruitment of KRAB-associated protein 1 (KAP1), which then recruits histone methyltransferases (ie, SETDB1) and deacetylases and promotes the binding of heterochromatin protein 1 (HP1), which maintains the heterochromatic state.36–38

In addition to the function of gene-specific TFs, the process of transcription itself is generally associated with the exchange of histones in chromatin. This histone exchange is mediated by adenosine triphosphate (ATP)-dependent chromatin remodelers, such as SWI/SNF, FACT, and Spt6, and facilitates the progression of RNA pol II through the chromatin.39 In addition to allowing the progression of RNA pol II along the DNA template, the dynamic deposition and removal of histones from the DNA allows for dynamic control over nucleosome composition. The replication-independent, transcription-dependent deposition of noncanonical histones, such as H3.3 and H2A.Z, provides one example of changing histone composition.40,41 Additionally, the continued removal and deposition of nucleosomes requires the continual proximity of protein complexes involved in that process, therefore providing a mechanism for rapid changes in response to environmental changes, either continued induction through continual recruitment of nucleosome remodelers or rapid shutoff upon loss of chromatin remodeler recruitment. The turnover of the nucleosomes also provides a potential mechanism for conservation of a histone modification pattern by copying a pattern from a redeposited H3/H4 heterotetramer or H2A/H2B dimer to a newly deposited counterpart. The turnover also allows for the dilution of histone marks in the absence of the continued activity of histone-modifying enzymes by incorporation of unmodified histones.

The expression of canonical histones (H3, H4, H2A, and H2B) is limited to the S-phase of the cell cycle, when the histone proteins are needed for incorporation into the daughter chromosomes.42,43 The noncanonical histones, such as H3.3 and H2A.Z, are expressed throughout the cell cycle and are deposited in the wake of RNA pol II during transcription-dependent histone exchange.44 This likely contributes to the observation that, in postmitotic cells, the noncanonical histones accumulate, while the canonical histones are gradually lost from the cell with age.39

The continued development of robust methods for the genome-wide assessment of DNA methylation and other epigenetic marks is providing a means for extensive analysis of differential epigenetic patterns in various tissues at various ages and disease states.45,46 Table 2 presents a list of genome-wide epigenetic studies in human complex disease to date that may have overlapping pathophysiology with diseases of the posterior eye, including glaucoma, AMD, and/or diabetic retinopathy (DR). Also, several resources are available online to access and analyze genome-wide epigenetic data (Table 3). However, this field remains dynamic with more work to be done, particularly with respect to the role postmortem time plays in tissue integrity for whole-genome epigenetic studies. The quality of tissue, for example, would influence downstream analysis of epigenomic data. The fresher the donor tissue is, the more meaningful the downstream interpretation of the statistical analysis will be.47–50 This is similar to studies of human gene expression from donor eye tissue.51

Table 2

Genome-wide epigenetic studies of retinal disease, risk factors, and related pathologies.

STUDY	SYSTEM	DISEASE	METHODOLOGY	RNA EXPRESSION	SAMPLE SIZE	TISSUE SAMPLE(S)	DEATH/COLLECTION TO PRESERVATION TIME	NUMBER OF DIFFERENTIALLY METHYLATED PROBES OR REGIONS (DMPs OR DMRs)/GENES	PATHWAYS IDENTIFIED IN DMR LOCI	VALIDATED LOCI WITH DIFFERENTIAL METHYLATION
Aavik et al, 2015173	Cardiovascular	Atherosclerosis	MeCP2-seq	Affymetrix HGU133 Plus2 microarray; qPCR; immunohisto-chemistry; western blots	Femoral artery ather-ectomy samples (22); mammary arteries (9)	Femoral artery atherectomy samples; mammary arteries	Not provided	4779 promoter DMPs/3740 genes, 8604 exonic DMPs/4939 genes, 17782 intronic DMPs/7972 genes	Intermediate filament-related genes, ATP-binding proteins, cytoskeleton, chromatin regulators, cell adhesion, morphogenesis	RTL1
Allione et al, 2015174	Multiple	Smoking	Illumina Infinium Human-Methylation450K BeadChip	None	20 MZ twin pairs	Whole blood	Not provided	22 DMPs	GTPase regulatory activity, phospholipid binding, vitamin D response element binding, transcriptional repressor complex	None
Bell et al, 2010175	Endocrine	Type 2 diabetes	MeDIP-chip Nimblegen; H3K4me1/2/3, CTCF, and H3K9me1 ChIP-chip	RNAseq	Discovery: 30 cases, 30 controls; validation: original 60 + 20 plus 14 donor brain tissue	Whole blood; donor brain tissue	Not provided	0 DMR between T2D and controls	None	None
Binder et al, 2015176	Endocrine	Gestational diabetes mellitus (GBM)	Discovery: Illumina Infinium HumanMethylation450K BeadChip; targeted replication with bisulfite-pyrosequencing	Affymetrix Human Transcriptome Array 2.0	Discovery: 41 cases, 41 controls; validation: not provided	Placental tissues (maternal side)	Not provided	Not provided	Immune response, cellular metabolism, response to external stimuli	CCDC181, HLA-H/HLA-J, HLA-DOA, SNRPN/SNURF
Boks et al, 200963	Multiple	Age, gender, genotype	Illumina GoldenGate DNA Methylation Cancer Panel 1 assay	None	23 MZ twin pairs, 23 DZ twin pairs, 96 controls	Whole blood	Not provided	Age: 58 DMPs; gender: 56 auto-somal DMPs; heritability: 96 DMPs; genotype: 11 DMPs	None	None
Dayeh et al, 2014155	Endocrine	Type 2 diabetes	Illumina Infinium Human-Methylation450K BeadChip	Affymetrix GeneChip Human Gene 1.0 ST arrays	15 cases, 34 controls	Donor pancreatic islets	Cultured 2.7 days prior to DNA/RNA extraction	1649 DMPs/853 genes	Cancer, axon guidance, MAPK signaling, focal adhesion, ECM-receptor interactions, regulation of actin cytoskeleton	None
Guida et al, 2015178	Multiple	Smoking	Illumina Infinium Human-Methylation450K BeadChip	Illumina HumanWG-6 chip v3; Illumina HT-12 chip	745 women	Whole blood	Not provided	461 DMPs between current smokers and never smokers; 3 DMPs between former smokers and never smokers; 751 DMPs associated with time since cessation of smoking	None	None
Haas et al, 2013179	Cardiovascular	Dilated cardiomyopathy	Discovery: Illumina Infinium HumanMehtylation27 Bead-Chip; replication: MassArray and bisulfite sequencing	RT-qPCR	Discovery: 9 cases, 8 controls; replication: 30 cases, 128 controls	Left ventricular tissue	Immediately washed and flash frozen	90 genes	Cardiovascular disease, nutritional & metabolic diseases, pathological conditions, signs and symptoms	LY75, ERBB3, HOXB13, ADORA2A
Hannum et al, 201365	Multiple	Aging	Illumina Infinium Human-Methylation450K BeadChip	Used data from Emilsson et al, 2008177	Discovery: 482; validation: 174 additional individuals; verification: Heyn et al, 201264 data	Whole blood	Not provided	70,387	AD, cancer, tissue degradation, DNA damage, oxidative stress	None-used validation sets to validate their model of aging from initial set
Heyn et al, 201264	Multiple	Aging	Discovery: whole genome bisulfite sequencing (WGBS); replication 1: WGBS and Illumina Infinium HumanMethylation450K BeadChip; replication 2: Illumina Infinium Human-Methylation450K BeadChip: validation of discovery + replication 1 combined and replication 2: targeted bisulfite sequencing validation	RT-qPCR	Discovery: 1 new-born, 1 centenarian; replication 1: 1 middle aged person; replication 2: 19 newborns, 19 nonagenarians	Discovery & replication 1: CD4+ T cells; replication 2: cord blood for newborns and peripheral blood mononuclear cells for nonagenarians	Not provided	Discovery & replication 1: 1149 DMPs/615 genes; replication 2: confirmed 214 DMPs plus additional 5774 DMPs	Not provided	AIM2, TNFRSF9, IGSF9B, PTPRE, ZRP-1, FHL2, miR-21
Hidalgo et al, 2014180	Renal	Diabetes	Illumina Infinium Human-Methylation450K BeadChip	None	Discovery: 544; replication: 293	CD4+ T cells	From frozen	Presented top 5 DMPs	Not provided	ABCG1
Irvin et al, 2014182	Multiple	Fasting VLDL and TGs	Illumina Infinium Human-Methylation450K BeadChip	Discovery: RT-qPCR of single hit; replication: Affymetrix HumanExon 1.0 ST GeneChip	Discovery: 991; replication 1261	CD4+ T cells	Not provided	4 DMPs/1 gene	Lipid metabolism	CPT1A
Ko et al, 2013183	Renal	Kidney fibrosis	Discovery: HELP-Nimblegen whole genome covering microarray; validation: bisulfite-Sequenom MassArray EpiTYPER; replication: Illumina Infinium HumanMethylation450K BeadChip	None	Discovery/validation: 12 hypertensive or diabetic chronic kidney disease, 14 controls; replication: 21 diabetic chronic kidney disease, 66 controls	Human kidney tubule epithelial cell tissue samples from healthy living transplants or surgical nephrectomies	Not provided	Discovery: 4751 DMRs/1535 genes; replication: 1061 genes	Cell adhesion, development	COLIVA1, DSCAM, DPT
Markunas et al, 2014184	Multiple	Maternal smoking	Discovery: Illumina Infinium HumanMethylation450K BeadChip; replication using Joubert et al, 2012181 published data	None	287 infants whose mothers smoked during 1st trimester, 602 infants whose mothers did not smoke during pregnancy	Whole blood	Not provided	185 DMPs/110 genes	Nicotine dependence, smoking cessation, placental & embryonic development	FRMD4A, ATP9A, GALNT2, MEG3, GNG12, Loc284998, VGLL4, CUX2, FTO, KIF26B
McClay et al, 201467	Multiple	Aging	Discovery: MBD-seq; replication: pyrosequencing	None	Discovery: 718 aged 25–92 years; replication: 558	Whole blood	Not provided	11 DMRs	Protocadherins, homeo-box genes, MAPKs, ryanodine receptors	LSAMP, ZEB2, MEIS7, GRIA2
Menni et al, 201366	Multiple	Aging	Discovery: Illumina Infinium HumanMehtylation27 BeadChip; replication: Illumina Infinium Human-Methylation450K BeadChip	None	Discovery: 172 female twins; replication: 350 twins	Fasting whole blood	Not provided	3 DMPs associated with C-glyTrp levels (an age-associated metabolite)	Protein synthesis, cell cycle, embryonic development, renal inflammation & hypertension, age-related retinal degeneration	WDR85
Miao et al, 2008185	Endocrine	Type 1 diabetes	H3K9me2 ChIP-chip, including human 12 k cDNA array, human 12 k CpG island array, and Nimblegen human promoter tiling array	None	9 T1D, 7 controls	Blood lymphocytes and monocytes	Not provided	Lymphoblasts: 193 DMPs 12 k cDNA array, 213 DMPx 12 k CpG island array, not specified for the tiling array; monocytes: none	Autoimmunity, inflammation, TGFbeta, NFkappa-B, p38 MAPK, toll-like receptor, IL-6 pathways, T1D genes	HOX13B, NKXB, CPLX4, CXCR4, FAM19A5, AR, PLAZG4B, CD55, IGSF4B, CTLA4
Milenkovic et al, 2014186	Cardiovascular	Flavanols & cardiovascular disease	Discovery: Illumina Infinium HumanMethylation450K BeadChip; validation with pyrosequencing	Agilent G4845A Human GE 4×44K v2 microarray	13 male smokers before and after flavanol treatment	Leukocytes	Freshly stored in RNALater	None	None	CXCL12, SCRIB, PDGFRL, FERMT3, ICAM1, VCAM1 (not significantly DM at 10% change)
Monick et al, 2012187	Multiple	Smoking	Illumina Infinium Human-Methylation450K BeadChip	RT-qPCR (transformed lymphoblasts)	10 smokers, 9 controls	Lung alveolar macrophage	Not provided	1381 DMPs	Wound healing, inflammation, G-protein/Ras signaling	AHRR
Movassagh et al, 2010188	Cardiovascular	Cardiomyopathy	Discovery: MeDIP-chip with the Nimblegen CpG island and promoter microarray; validation: bisulfite PCR and sequencing	RT-qPCR	Discovery: 3 (ischemic & idiopathic) cardiomyopathic tissue from transplant surgeries; 1 normal tissue from donors; validation: 4 (ischemic & idiopathic) cardiomyopathic tissue from transplant surgeries; 4 normal tissue from donors	Left ventricular tissue	Not provided	3 genes	Angiogenesis	AMOTL2, ARH-GAP24, PECAM1
Movassagh et al, 2011189	Cardiovascular	Cardiomyopathy	MeDIP-seq; H3K36me3 ChIP-seq	RT-qPCR	4 (ischemic & idiopathic) cardiomyopathic tissue from transplant surgeries; 4 normal tissue from donors	Left ventricular tissue	Not provided	Not provided	Not provided	DUX4
Nazarenko et al, 2015190	Cardiovascular	Coronary Heart Disease	Illumina Infinium HumanMehtylation27 BeadChip; pyrosequencing validation	No	3 tissues from 6 patients for microarray; 3 tissues from 21 patients for validation	Human right coronary artery in the area of advanced atherosclerotic plaques (CAP); internal mammary arteries (IMA); great saphenous veins (GSV)	Processed and snap frozen immediately after surgery	22 (CAP-IMA); 27 (GSV-IMA)	Inflammation and immune response; development	HOXD4
Nilsson et al, 2014191	Endocrine	Type 2 diabetes	Illumina Infinium Human-Methylation450K BeadChip	RT-qPCR; Affymetrix GeneChip Human Gene 1.0 ST arrays	14 discordant MZ twin pairs; cohort 1: 120 case-controls; cohort 2: 28 cases, 28 controls	Subcutaneous adipose tissue biopsies	Frozen immediately in liquid nitrogen	15,627 DMPs/7,046 genes in cohort 2, 1410 of which DMPs also in MZ twins	Oxidative phosphorylation; mitogen-activated protein kinase signaling; carbohydrate, amino acid, and lipid metabolism; inflammation; Wnt signaling; glycan degradation; cancer	None
Nitert et al, 2012192	Endocrine & musculoskel et al	Diabetes & exercise	Discovery: MeDIP-chip; replication: Illumina Infinium HumanMethylation450K BeadChip; validation: bisulfite-Sequenom Epi-TYPER, luciferase assays with or without methylation	Affymetrix Custom-Array NuGO-Hs1a520180 GeneChip	Discovery: 15 men with 1st degree relative with diabetes (FH+)/13 men w/o relative with diabetes (FH-); replication: 9 MZ twin pairs discordant for T2D	Skeletal muscle biopsies from vastus lateralis after fasting and without exercise for 48 hrs prior to biopsy	Not provided	65 genes DM FH+/FH-before exercise; 38 genes DM FH+/FH-after exercise; 134 genes DM before/after exercise combined FH+/−	FH+/−: MAP signaling, insulin signaling, calcium signaling, Wnt signaling, starch and sucrose metabolism, sphingolipid metabolism; before/after exercise: purine metabolism, retinol metabolism, calcium signaling, muscle, T2D, gly/ser/thr metabolism, glycolysis, gluconeogenesis, starch and sucrose metabolism, insulin signaling	MSI2, THADA, MEF2A, RUNX1, NDUFC2
Olsson et al, 2014193	Endocrine	Insulin secretion	Discovery: Illumina Infinium HumanMethylation450K BeadChip; validation: pyrosequencing	Affymetrix GeneChip Human Gene 1.0 ST	89 pancreatic islet donors	Human pancreatic islets	Cultured in vitro 4 days	14 DMPs potentially mediating relationship between SNPs and mRNA level	Pancreatic islet function, type 1 diabetes, cell adhesion, ECM interaction, folate biosynthesis	GPX7, GST-1, SNX19
Rakyan et al, 2011194	Endocrine	Type 1 diabetes	Illumina Infinium HumanMehtylation27 BeadChip	None	19 discordant MZ twin pairs; 7 singletons before and after diagnosis	CD14+ monocytes	Not provided	132 DMPs	Type 2 diabetes, immune responses	HLA-DQB1, RFXAP, NFKB1A, TNF, GAD2
Ronn et al, 2015195	Endocrine, etc.	Aging, BMI, HbA1c	Discovery: Illumina Infinium HumanMethylation450K BeadChip; validation: pyrosequencing	Affymetrix GeneChip Human Gene 1.0 ST	Discovery: 96 males, 94 females; validation: 37 males, 67 females	Discovery: human adipose tissue biopsies (fasted state); validation: human adipose tissue biopsies and blood (fasted state)	Not provided	Age: 1050 genes methylation and expression; BMI: 2825 genes methylation and expression; HbA1c: 711 genes methylation	Cancer, type 2 diabetes, cardiovascular disease; signal transduction	None
Sapienza et al, 2011196	Endocrine	Diabetes	Illumina Infinium HumanMehtylation27 BeadChip	None	23 diabetics with end stage renal disease; 23 diabetics without nephropathy	Saliva	Not provided	389 DMPs	Inflammation, oxidative stress, ubiquitination, fibrosis, drug metabolism, cancer, cardiotoxicity, hepatotoxicity, nephrotoxicity	None
Sharma et al, 2014197	Cardiovascular	Coronary artery disease	Discovery: 12 k Human CpG Island Microarray; replication: bisulfite deep sequencing	None	Discovery: 18 cases, 18 controls; replication: 48 cases, 48 controls	Peripheral blood	Within 1 hour	72 DMRs	Signal transduction, transcription regulation, organelle development, transport, cell regulation, ion channel activity, E-box binding, mitochondria, plasma membrane, nuclear chromatin	STRADA, C1QL4, HSP90B3P
Steenaard et al, 2015198	Cardiovascular	Tobacco smoking (coronary artery disease)	Illumina Infinium HumanMethylation450K BeadChip	Illumina HumanHT12v4 Expression BeadChip	724 subjects: 195 current smokers, 201 never smokers, 328 prior smokers	Whole blood	Not provided	15 DMPs of 3669 CpGs/1169 CAD genes analyzed	Only looked at CAD related genes	None
Stringhini et al, 2015199	Immune System	Inflammation	Illumina Infinium Human-Methylation450K BeadChip	None	857	Whole blood	Not provided	41 DMPs/10 genes associated with household occupational position; 12 DMPs/6 genes associated with socioeconomic trajectories	Only looked at candidate genes involved in socio-economic status-related inflammation	None
Sun et al, 201468	Multiple	Aging	Illumina Infinium HumanMethylation450K BeadChip	None	100 male, 100 female nonagenarians/centenarians	Peripheral blood mononuclear cells	Not provided	850 DMPs/564 genes	Hormone regulation, neuron projection, disease-related pathways, cellular component organization, cell morphogenesis, cell-cell junctions, CNS development, ECM-receptor interactions, axon guidance, cell adhesion molecules	None
Suter et al, 2011200	Multiple	Maternal tobacco exposure	Discovery: Illumina Infinium HumanMehtylation27 BeadChip; validation: bisulfite PCR sequencing	Illumina HG-12, RT-qPCR	Discovery: 18 smokers, 18 non-smokers; validation: 9 smokers, 9 non-smokers	Placental tissues	Collected immediately after delivery and flash frozen	1024 DMPs	Oxidative stress pathways, oxidative phosphorylation, mitochondrial dysfunction, HIF1alpha signaling, cell death, cell morphology, cell-to-cell signaling	PURA, GTF2H2, GCA, GPR135, HKR1
Tsaprouni et al, 2014201	Multiple	Smoking	Illumina Infinium HumanMethylation450K BeadChip	RNAseq	Discovery: 22 current smokers, 263 former smokers, 179 never smokers; replication: 41 current, 104 former, 211 never	Peripheral blood	Not provided	30 DMPs/15 loci	Cardiovascular disease; cancer; connective tissue and developmental disorders; cell death, cell survival, and cell-cell interactions; heme biosynthesis; aryl hydrocarbon receptor signaling	None
Wan et al, 2012202	Multiple	Smoking	Discovery: Illumina Infinium HumanMethylation27 BeadChip: validation: bisulfite pyrosequencing	None	Discovery: 1085; replication: 369	Peripheral blood leukocytes	Not provided	15 DMPs current vs. former smokers; 2 DMPs cumulative exposure; 3 DMPs time since quitting	Not provided	F2RL3, GPR15
Wang et al, 2013203	Cardiovascular	Hypertension	Discovery: Illumina Infinium HumanMehtylation27 BeadChip; replication: bisulfite sequencing	None	Discovery: 8 cases, 8 controls 14–23 years old; replication 1: 36 cases, 60 controls 14–30 years old; replication 2: 36 cases, 34 controls 15.8–40 years old	Blood leukocytes	Not provided	1 gene	Inflammation	SULF1
Xiao et al, 201569	Multiple	Aging	Discovery: MeDIP-seq; replication: whole genome bisulfite sequencing	None	Discovery: 4 centenarians, 4 middle aged; replication: additional 1 centenarian, 1 middle aged	Peripheral blood	Not provided	Discovery: 626 DMRs/251 genes; replication: 156 genes	Age-related diseases: T2D, cardiovascular disease, stroke, coronary artery disease, AD; developmental processes; cell adhesion; signal transduction; cell communication; cell adhesion; cadherin signaling; Wnt signaling; AD-presenilin	None
Xu et al, 2013204	Multiple	Obesity	Illumina Infinium Human-Methylation450K BeadChip	None	48 obese cases, 28 lean controls, all 14–20 years old	Fasting peripheral blood leukocytes	Not provided	23305 DMPs	DNA binding, development, regulation of neurogenesis, cell differentiation, transcription regulation, obesity-related diseases	None
Yang et al, 201570	Neurological	Aging	Illumina Infinium Human-Methylation450K BeadChip	None	740 aged 66 to 108 years old	Postmortem dorsolateral prefrontal cortex	Not provided	4263 DMPs associate with age	Not provided	None
Yuan et al, 2014205	Endocrine	Type 2 diabetes	Discovery: MeDIP-seq; replication: Illumina Infinium HumanMethylation450K BeadChip	Illumina HumanHT12v3 Expression BeadChip	Discovery: 27 MZ twin pairs (17 T2D discordant, 2 T2D concordant, 7 control concordant pairs); replication: 42 unrelated T2D cases, 221 controls	Whole blood	Blood drawn into EDTA tubes	1355 DMRs	T2D GWAS genes, imprinted genes	GPR61, PRKCB, MALT1
Zaina et al, 2014206	Cardiovascular	Atherosclerosis	Discovery: whole-genome shotgun bisulfite sequencing; replication: Illumina Infinium HumanMethylation450K BeadChip; validation: pyrosequencing and conventional bisulfite sequencing	RT-qPCR	Discovery: 1 matched pair of atherosclerotic and normal tissue from the same donor; replication: 15 matched pairs of atherosclerotic and normal tissues each from the same donor; validation: (multiple) 5 aortas/24 donor pairs/15 donor pairs from original cohort/1 donor pair	Postmortem aortas and carotid tissues	3–26 hours	Discovery: 54625 DMRs; replication: 1895 DMPs	Genes related to processes related to endothelial cells and vascular smooth muscle cells; vascular smooth muscle cell contraction; extracellular matrix formation	HOX family members, PDGFA, PLAT, PRRX1, PXDN, MIR23b
Zawada et al, 2012207	Cardiovascular	Chronic kidney disease-associated cardiovascular disease	Discovery: SuperTAG methylation-specific digital karyotyping; validation: bisulfite pyrosequencing	RT-qPCR	10 cases, 10 controls	Peripheral blood	Not provided	4288 DMRs	Lipid metabolism & transport, cell proliferation, cell-cycle regulation, angiogenesis, inflammation	METTL2B

Table 3

Epigenetic resource databases.

DATABASE	WEBSITE	TISSUE DATA SET	SOURCE
Epigenome Browser	http://www.epigenomebrowser.org/	Varied	ENCODE
Roadmap Epigenomics Visualization Hub (VizHub)	http://vizhub.wustl.edu/	Stem cells & healthy primary tissues	NIH Roadmap Epigenomics Mapping Consortium
WashU EpiGenome Browser	http://epigenomegateway.wustl.edu/browser/	Varied	Roadmap Epigenomics; ENCODE
NCBI	http://www.ncbi.nlm.nih.gov/epigenomics	Varied	Varied
Emsembl	http://useast.ensembl.org/info/website/tutorials/encode.html	Varied cell lines	Varied
MethylomeDB	http://www.neuroepigenomics.org/methylomedb/	Human and mouse brain	Haghighi Lab
MethBase	http://smithlabresearch.org/software/methbase/	Varied	Smith lab
NGSMethDB	http://bioinfo2.ugr.es:8080/NGSmethDB/	Varied	Repository
DiseaseMeth	http://202.97.205.78/diseasemeth/	Varied human disease	Varied
Gene Expression Omnibus (GEO)	http://www.ncbi.nlm.nih.gov/geo/roadmap/epigenomics/	Varied tissues and cell lines	NIH Roadmap Epigenomics data
The Epigenome Atlas	http://www.genboree.org/epigenomeatlas/index.rhtml	Human reference epigenomes	NIH Roadmap Epigenomics Mapping Consortium
Canadian Epigenomics, Environment and Health Research Consortium Platform (CEEHRC)	http://epigenomes.ca/	Varied	CEEHRC
Epigenie	http://epigenie.com/epigenetic-tools-and-databases/	List of epigenetic resources	Varied

Aging

Throughout development, cells acquire chromatin changes that regulate the transcriptional output of each cell. In gametogenesis and embryogenesis, DNA methylation is erased and reset at specific reprograming stages to ensure appropriate gene expression and totipotency of embryonic cells.52 This reprograming is used by the cell for processes such as X-chromosome inactivation for dosage compensation, repression of invasive DNA such as retrotransposons, imprinting to ensure parental allele-specific expression, and to otherwise control the expression of genes relevant to the developmental process. Additionally, the process of cellular differentiation, in which individual cells acquire specialized gene expression profiles specific to their individual tissue types and functions, requires further changes to the chromatin landscape (reviewed for various tissues such as skeletal muscle,53 myeloid cells,54 and the lens and retina55). Much effort of late has focused on defining the tissue-specific differential methylation patterns in differentiated tissues in relation to differential gene expression patterns in an attempt to understand the underlying mechanisms of tissue-specific processes.56–60 Even after a cell has reached its terminally differentiated state, further changes that associate with the aging process can continue to accrue throughout the lifespan of the cell. Methylation levels generally decrease with aging, although individual genes may have increased methylation levels with age.61–70 Replication-independent histone exchange results in the accumulation of histone variants, such as H3.3 and H2A.Z, in older cells.39 Cells from aged human beings have been shown to have reduced levels of both the heterochromatin maintenance protein HP1 and the H3K9me3 histone modification known to recruit HP1 to chromatin.71 Experiments in model organisms showing that the reversal of age-dependent chromatin changes results in extension of lifespan suggest that at least some of the age-related chromatin changes are indeed causative of aging, rather than just a result of aging.72–74 Much work is still needed to fully understand the role of acquired chromatin changes in the aging process.

Sirtuins are a class of HDACs implicated in lifespan regulation and the promotion of healthy aging through various epigenetic and nonepigenetic cellular roles, including telomere maintenance, DNA repair, metabolism, stress tolerance, cellular differentiation, apoptosis, and inflammation.75,76 Sirtuins have also been implicated in age-related diseases, such as diabetes, cardiovascular disease (CVD), neurodegenerative diseases, and some cancers.9 Roles for sirtuins have been observed in ocular aging. The involvement of sirtuin 1 (SIRT1) in ocular aging, including retinal degeneration, has been the most well characterized of the seven mammalian sirtuins. SIRT1 has been associated with retinal stem cell self-renewal, and decreased expression of SIRT1 with age has been suggested to contribute to the aging process.77 SIRT1 also exhibits neuroprotective effects in the retina at least partially through its antioxidant, energy balancing, and antiapoptotic functions.78 Abnormal SIRT1 localization is also thought to promote the apoptosis of photoreceptor cells and precocious aging in the rd10 mouse model of retinal degeneration.79 SIRT6, which is regulated by SIRT1, has also been implicated in retinal aging, and SIRT6 deficiency in mice results in increased levels of retinal cell apoptosis.80,81 Altered TF functions have also been observed in the aging mammalian retina. For example, the function of the oxidative stress response master regulator nuclear factor (erythroid-derived 2)-like 2 (Nrf2) was shown to be altered in aged C57Bl6/J mice compared with young C57Bl6/J mice.82 The aged mice in this study exhibited higher uninduced (basal) expression levels of Nrf2 antioxidant targets compared with the young mice. Upon induction with an oxidative stressor, the Nrf2-dependent antioxidant genes were robustly induced in young mice, but not in aged mice. Nrf2 was also induced by the oxidative stressor in young mice, but not in aged mice. Expression of Nrf2 is regulated by DNA methylation and modulated by dietary factors such as sulforaphane; differential methylation of the Nrf2 promoter is also associated with various diseases.83

The question of whether complex age-related diseases are the inevitable eventual outcome of the aging process or whether they are, in fact, the result of aberrant aging is yet to be elucidated and is of great interest in the field today. Along with the extension of lifespan achieved to date by modern medical advances has come an increase in the prevalence of age-related conditions that severely impact the quality of life of older adults. Degenerative diseases of the retina, such as glaucoma, AMD, and DR, affect >15 million people >40 years old in the US and are the leading causes of blindness in older adults.84–86 As of 2013, 11 million individuals were affected with AMD, per year, in the United States alone. Because aging is the greatest risk factor, this number is anticipated to increase to 22 million by 2050 as a result of the population aging (http://www.brightfocus.org/).

We present here what is known of epigenetic effects in major retinal diseases associated with aging: glaucoma, AMD, and DR. Further work in this field has promise to revolutionize the way we approach therapeutic interventions by revealing points of intervention with potential long-term influences on the progression of disease.

Glaucoma

Glaucoma is the leading cause of irreversible blindness in the US and worldwide, according to the World Health Organization (2015); it ranks second overall after cataract, which is reversible.87–89 Primary open-angle glaucoma is considered the silent thief of vision, as it is initially asymptomatic, progresses slowly, and eventually, causes irreversible vision loss due to optic nerve damage and loss of retinal ganglion cells (Fig. 3B). Early diagnosis and treatment can control glaucoma before vision loss occurs. At present, lowering intraocular pressure (IOP) is the only proven method of slowing glaucoma-associated visual field deficit progression. Glaucoma is a progressive neurodegenerative disease and is thought to progress as follows: fibrosis in the trabecular meshwork (TM) and in the lamina cribrosa at the optical nerve head reduces outflow of the aqueous humor (AH), which increases the IOP and induces hypoxia. Subsequent damage to the optic nerve head is associated with retinal ganglion cell loss and vision loss.90,91 The molecular details of disease etiology remain unclear to date. Risk factors for glaucoma include age, IOP, and central corneal thickness.92 Oxidative stress, systemic hypertension, and family history have also been associated with the development of glaucoma.93,94 At least 29 loci have been associated with genetic risk of glaucoma, of which the best characterized are myocilin (MYOC), optineurin (OPTN), and WD repeat domain 36 (WDR36).95 MYOC acts in cyto-skeletal organization, extracellular matrix remodeling, and mitochondrial function.96,97 Accumulation of MYOC aggregates within the TM is thought to cause TM cell apoptosis and ECM changes, which blocks the AH outflow and leads to increased IOP.98 OPTN is protective against oxidative stress and stabilizes MYOC mRNA.99,100 WDR36 variants in mouse alter retinal ganglion cell axon growth that leads to retinal degeneration.101 However, the role of this gene in the development of human disease remains controversial.95

Studies of gene expression in donor eye samples affected with glaucoma have shown upregulation of profibrotic factors in the TM and AH, including transforming growth factor beta (TGF-β), thrombospondin-1 (TSP1), and connective tissue growth factor, compared with unaffected controls.102–104 Extensive investigations of epigenetic mechanisms have yet to be performed in human tissues affected by glaucoma. However, epigenetic processes involved in fibrosis in other cell types/disease processes give insight into likely processes involved in the glaucoma-associated fibrosis. Hypermethylation of TSP1 in colorectal cancer suppresses TSP1-mediated activation of TGF-β.105 TGF-β-mediated regulation, in turn, is sensitive to the chromatin modifications, including DNA methylation and histone acetylation, of its downstream target genes.106–108

Hypoxia is thought to contribute to the degeneration of retinal ganglion cells in glaucoma.109 The expression of hypoxia-induced factor 1 alpha (HIF1α), which induces the adaptive transcriptional response to hypoxia,110,111 is greater in the retina and optic nerve head of glaucomatous eyes compared with controls.110 HIF1α is stabilized under hypoxic conditions by the loss of a specific oxygen-dependent degradation mechanism involving proline hydroxylation of HIF1α, 112,113 which allows HIF1α to dimerize with HIF1β to form an active HIF1 heterodimer.114 Hypoxia induces the translocation of HIF1α from the cytoplasm to the nucleus, where it recruits the histone acetyltransferase CBP/p300 coactivator and regulates transcription of target genes.115 HIF1α activity is regulated epigenetically in a few ways. The HIF1α promoter contains a hypoxia response element with a CpG that is methylated under hypoxic conditions; a DNA methyltransferase (DNMT) inhibitor increased hypoxia-induced gene expression; methylation of the hypoxia response element influences HIFα binding; and various histone modifications at hypoxia responsive genes are specific to hypoxic conditions.116

Further evidence for epigenetic effects in hypoxia has been provided by Shahrzad et al who showed an inverse relationship between DNA methylation and hypoxia in primary cancer cells and in normal human dermal fibroblasts.117 Also, under chronic hypoxia in prostate epithelial cells, the induced HIF1α stability is lost, and both H3K9 hyperacetylation and DNA hypermethylation accumulate, which are proposed to maintain the hypoxia-conditioned state in the absence of HIF1α.111

Genome-wide epigenetic studies utilizing well-characterized/phenotyped glaucoma-affected tissues and matched normal controls will be important for elucidating the epigenetic changes that occur within the eye throughout the progression of the disease, from normal to the onset of increased IOP, to the damaging effects of that pressure on the optic nerve. Additionally, the use of monozygotic twin pairs discordant for glaucoma for genome-wide epigenetic analysis will further highlight epigenetic processes associated with the development of glaucoma. The information gained by combining rigorous genome-wide epigenetic, genetic, and expression data will lead to improved understanding of the disease and, therefore, better means to treat it.

Age-related Macular Degeneration

Similar to glaucoma, AMD is also a multifactorial, progressive, neurodegenerative disease. AMD is characterized by the sub-RPE accumulation of yellowish lipid-rich, protein-containing drusen deposits and the progressive degeneration of the RPE and the neural retina. Although the whole retina can be affected, the disease preferentially affects the macular region with or without foveal involvement (Fig. 3C). The presence, size, and number of drusen, along with the presence or absence of hypo- or hyperpigmentary changes in the RPE, determine the severity of early or intermediate stages of the disease and the risk for the development of either form of advanced AMD: geographic atrophy (GA; dry AMD) or neovascular (wet) AMD.118–121 GA involves the regional loss of RPE and choroid accompanied by a gradual loss of photoreceptors and central vision.122,123 Neovascular AMD involves the growth of new abnormal blood vessels from the choroid into the normally avascular sub-RPE and subretinal regions. These new vessels are prone to hemorrhage and exudation, which can lead to rapid loss of vision and accounts for the majority of blindness associated with AMD.124 One model for the development of neovascularization in AMD involves the accumulation of drusen, which disrupts the connection between the RPE and the choroidal blood supply, causing hypoxia. The hypoxia in turn induces the expression of VEGF and the formation of new vessels.125 Current therapies for AMD target the advanced neovascular stage, have modest results, and require frequent in-office intraocular injections. Much effort is ongoing in the development of new therapeutics that may ultimately improve outcomes as well as reduce the burden of treatment on affected individuals. Advances in genetic and epigenetic studies have the potential to reveal additional targets for intervention. Additionally, the discovery of molecular mechanisms for the progression from early to intermediate to advanced forms of the disease will provide points for earlier interventions to prevent advancement to late-stage disease and the accompanying vision loss.

Genome-wide association studies (GWAS) for AMD risk alleles have identified 19 genetic variants that associate with AMD at genome-wide significance levels, including genes involved in the complement pathway, lipid metabolism, angiogenesis, and atherosclerosis, among other cellular processes.126 A recent exome chip study identified 34 loci representing both common and rare variants independently associated with AMD.127 This study found evidence for causal roles for three very rare coding variants in three genes (complement factor H [CFH], CHI, and TIMP3) and one splice variant (SC16A8). The two loci with the strongest association with AMD risk are the CFH and ARMS2/HTRA1 loci. The role of CFH in the complement pathway is well characterized.128 However, the role of the ARMS2/HTRA1 locus remains elusive, with ongoing debates regarding which of the two genes at that locus is responsible for the associated genetic risk for AMD.129

Twin studies estimate the genetic component of AMD risk at ~40%–70% of the total risk.130 Known environmental risks include smoking, obesity, and diet.25 Environmental contributions to disease potentially involve epigenetic changes that affect the expression of genes involved in the generation of disease and may also convert environmental exposures into heritable phenotypes.131 This relationship between epigenetics, environment, and heritability may also lead to increased measures of heritability in monozygotic twin studies—due to the shared zygotic epigenetic history of monozygotic twins— than would arise from genetic heritability alone.131 Therefore, epigenetic factors may contribute to both the heritable and nonheritable components of risk. This is consistent with observations that some chromatin changes are conserved through gametogenesis, whereas others are erased in the process.52 The idea that environmental forces introduce small epigenetic effects that accumulate with time (illustrated in Fig. 2) is also consistent with both the observations that age is the primary risk factor for AMD and that siblings with identical AMD risk profiles from major AMD genetic loci, smoking, body mass index (BMI), and CVD can still be extremely discordant for disease development (Fig. 4).

Figure 4

Siblings extremely discordant for AMD. These fundus images illustrate the retinas of discordant siblings with identical risk profiles for major AMD genetic loci, smoking, BMI, and CVD. Note the severe neovascular AMD in one sibling and normal macula in the other at a comparable age. Environmentally induced epigenetic changes may explain the occurrence of such discordance for AMD or other age-related disease. OD, oculus dexter (right eye); OS, oculus sinister (left eye). Photographs taken from DeAngelis laboratory patient cohorts. The study protocol was reviewed and approved by the Institutional Review Board at the University of Utah and conforms to the tenets of the Declaration of Helsinki.

Evidence for the involvement of epigenetic processes associated with macular degeneration has been accumulating in the field. Early work looking at proteins involved in the modification of chromatin showed requirements for these functions in relation to AMD. Suuronen et al demonstrated that HDAC inhibitors increased the expression and secretion of clusterin, the major protein component of drusen, in a human RPE cell line,132 suggesting that histone deacetylation activity is important to limit the amount of clusterin produced in RPE cells to prevent its accumulation in drusen. Chen and Cepko provided further evidence for the role of histone deacetylation in the maintenance of a healthy retina when they showed that HDAC4 promotes the survival of retinal neurons in mouse.133(p4) Given that neovascularization is thought to be a response to hypoxia caused by drusen deposits interrupting the flow of oxygen from the choroid through the RPE and into the retinal cells, it is also relevant that knockout of the noncanonical H2A.X histone in mice reduced the neovascular response to hypoxia.134

In addition to functional studies looking at the effects of histone-modifying activities, studies have looked at the association of chromatin modifications at genes relevant to the AMD disease process in human tissue samples. Using a DNA methylation microarray and bisulfite pyrosequencing of frozen human donor RPE/choroid samples with a postmortem interval of up to nine hours, Hunter et al identified the glutathione S-transferase isoforms mu1 and mu5 (GSTM1 and GSTM5) as hypermethylated in AMD samples compared with age-matched controls.135 The function of these two proteins in reducing oxidative stress is likely relevant to the development of AMD, as oxidative stress is hypothesized to contribute to the underlying pathophysiology of AMD. In addition to differential methylation of GSTM1 and GSTM5, the study also looked at exon expression microarray data from the same RPE/choroid samples plus the corresponding neural retinas and observed reduced mRNA levels of both genes in AMD samples compared with controls. DNA hypermethylation at the GSTM1 promoter in RPE/choroid samples correlated significantly with the reduced expression of both the GSTM1 gene and the adjoining, downstream GSTM5 gene.

Another early example of differential methylation in AMD correlating inversely with gene expression came from Wei et al’s report of hypomethylation in the promoter of IL17RC from peripheral blood mononuclear cells of patients with AMD compared with their unaffected twins, which corresponded to increased expression of IL17RC in the retina and RPA/choroid from formalin-fixed, paraffin-embedded ocular sections of donor eyes.136 However, Oliver et al were unable to replicate this finding in peripheral blood of patients with AMD compared with controls.137 This question will need to be settled by utilizing tissue samples from the RPE/choroid, neural retina, and blood from the same donors to accurately assess the relationship between differential methylation and AMD disease status, as well as through further molecular characterization of the role of IL17RC in AMD.

As a step in that direction, Oliver et al investigated DNA methylation levels in peripheral blood samples and frozen sucrose gradient-treated peripheral retinas (postmortem interval of donor eye tissue was not reported) of AMD patients with either GA or neovascularization compared with unaffected control patients.138 In the only genome-wide epigenetic study of AMD to date, they observed hypomethylation at the ARMS2/HTRA1 locus and hypermethylation at the PRSS50 locus in AMD patients compared with controls. As described above, the ARMS2/HTRA1 locus is one of the top two loci genetically associated with AMD. The finding that hypomethylation at the ARMS2/HTRA1 locus associates with AMD supports a role for either or both of those genes in the development of disease. The PRSS50 locus had not previously been associated with AMD risk.

The genome-wide methylation study by Oliver et al in AMD-affected RPE/choroid and retina might be a step toward understanding the epigenetic mechanisms underlying gene expression of AMD. Building on their study, researchers will need to go on to evaluate the epigenetic changes that occur throughout the progression from normal to severe disease. This will require well-characterized tissue samples with short postmortem intervals to tease out epigenetic changes at each stage of the disease: preclinical through advanced stages. This will allow for the identification of early changes that may influence the progression of AMD and, therefore, present early therapeutic interventions to prevent the progression to severe forms of the disease.

Further methodological improvements will also need to be incorporated to accomplish this goal. Increased quality of well-characterized donor eyes for each stage of the disease, along with well-characterized phenotypically normal controls, will increase the power of the study to identify changes associated with each step of the disease process. Also, the postmortem interval between death and preservation of the tissue must be carefully selected to preserve the integrity of the sample and minimize the noise in the genome-wide data.48 It is also important that these studies be carried out in a fresh tissue that is directly involved and relevant for the specific disease process being studied. Similar to gene expression profiles, DNA methylation profiles differ among tissues.13,14 Therefore, it cannot be taken for granted that an easily accessible tissue, such as peripheral blood, will adequately substitute or replicate for the relevant, disease-involved tissue.

Diabetic Retinopathy

In 2012, ~10% of Americans had diabetes, according to the American Diabetes Association.139 Every diabetic patient is at risk of developing DR, which is the most common cause of blindness among working-age adults in developed countries.140 The risk for DR increases with increasing time since diagnosis of diabetes.141,142 Similar to the age-related risk for AMD and glaucoma, the time component of risk for DR is also consistent with the idea that environmental influences introduce small epigenetic effects that accumulate with time (Fig. 5).

Figure 5

Why do some patients progress beyond and others stop at an intermediate stage of disease? The complex functional networks at play within a cell or tissue are likely susceptible to small perturbations exerted by environmental influences over long periods of time. The gradual accumulation of altered expression or function of key factors within a system may then lead to dysfunction and disease. Disease-promoting factors (such as smoking, obesity, and poor diet) work to alter the networks in ways that lead to pathological states, whereas disease-preventing factors (such as antioxidant consumption and healthy BMI) oppose the progression of disease by keeping the system within the bounds of healthy function. The ratio of disease-promoting and -preventing influences may balance at different equilibrium points, leading to individual cases that progress only so far along the pathological pathway before reaching that balance point. By discovering the various promoting and preventing factors, along with their respective significances at each point along the disease progression timeline, researchers (and eventually clinicians and patients) will be able to balance the networks to prevent or reverse the progression of disease. Photographs taken from the DeAngelis laboratory cohorts. The study protocol was reviewed and approved by the Institutional Review Board at the University of Utah and conforms to the tenets of the Declaration of Helsinki.

Type 2 diabetes is characterized by hyperglycemia due to impaired insulin secretion by pancreatic β-cells and/or to insulin resistance in peripheral tissues, including skeletal muscle, adipose, and liver. Aging is a major risk factor for type 2 diabetes, along with obesity, low physical activity levels, poor diet, and genetics. DNA methylation at the type 2 diabetes risk genes, NDUFB6 and COX7A1, which encode components of the mitochondrial electron transport chain complexes one and four, respectively, were found to increase with age in skeletal muscle of unaffected individuals.143,144 This increased DNA methylation was also associated with decreased gene expression (RNA and protein for NDUFB6, RNA for COX7A1) in nondiabetic elderly patients, consistent with the DNA methylation having a functional effect on gene expression. Both NDUFB6 and COX7A1 were also shown to have decreased expression in skeletal muscle of type 2 diabetes patients compared with controls.145 Ling et al found that a single-nucleotide polymorphism (SNP, rs629566) that introduces an additional CpG methylation site in the NDUFB6 promoter accounted for the increased methylation in non-diabetic elderly subjects.143 Although decreased expression of NDUFB6 is associated with insulin insensitivity in skeletal muscle, the NDUFB6 expression is decreased in type 2 diabetic patients,145 the authors were unable to find an association between the rs629566 SNP and the risk of developing type 2 diabetes. However, their data did suggest that rs629566 may be associated with insulin-stimulated glucose uptake in elderly subjects. Additionally, a small association with disease was found for another NDUFB6 locus SNP (rs540467) in their data and for two other NDUFB6 locus SNPs in GWAS data mined from other groups. Further work is still needed to elucidate the mechanistic relationships between DNA methylation and genetic variants of NDUFB6 and the development of type 2 diabetes. This gene exemplifies the complex relationships among genetic, epigenetic, and nongenetic contributions to gene expression in aging and the development of chronic pathology.

Many studies have shown links between epigenetics and the development of type 2 diabetes. An early study linked nutrition to heritable, potentially epigenetic changes that increased or decreased the risk for type 2 diabetes.146 Exposure to famine during the paternal grandfather’s childhood is protective against, whereas access to abundant food supplies associates with increased risk for diabetes as a cause of death for his grandchild. Also, studies have shown that diabetes during pregnancy increases the risk for that child to be diagnosed with diabetes as an adult.147

In addition to the multigenerational heritability of epigenetic changes involved in the development of type 2 diabetes, many studies of late have shown evidence for long-term epigenetic changes within an individual with hyperglycemia. The observation of a metabolic memory in which diabetic patients continue to develop complications even after the onset of glycemic control following prior hyperglycemia is consistent with stable epigenetic changes in response to hyperglycemia.148,149 A prospective study comparing patients with an intensive therapy for glycemic control compared to a traditional therapy showed cardiovascular benefits of intensive therapy. These benefits continued in the intensive therapy group 10 years after the trial even though mean glycated hemoglobin levels equilibrated between the two groups within one year following the trial.150 Thus, the early hypoglycemia initiated prephenotypic events that would lead to later diabetic complications independent of later glycemic control.

A genome-wide methylation study looking at differential DNA methylation in skeletal muscle biopsies from type 2 diabetic patients and age-matched controls identified 838 promoter regions that were differentially methylated between the two groups, including the promoter for PGC-1α,151 which functions as a coactivator for the transcription of genes involved in mitochondrial function and biogenesis.152 Further characterization of the PGC-1α promoter by Barrès et al showed hypermethylation in type 2 diabetic patients compared with controls, which negatively correlated with expression of PGC-1α and mitochondrial density.151 Interestingly, the hypermethylation at this locus was found mostly at non-CpG cytosines, which is less common and less well characterized than CpG methylation.151,153 Furthermore, the hypermethylation and associated expression changes were responsive to environmental influences, such as free fatty acid exposure and exercise.151,154

A more recent genome-wide methylation study identified 1,649 genes with differential DNA methylation in pancreatic islets between subjects with type 2 diabetes and normal controls, most of which involved modest changes in percent methylation in the range of 5%–10%.155 Of these 1,649 differentially methylated genes, 102 also showed differential gene expression in type 2 diabetics compared with controls. Pathway clustering of these genes identified pathways enriched within these gene sets, providing insight into the possible pathological mechanisms. Genome-wide studies such as this provide targets for further functional analysis that will lead both to a better understanding of the disease process and to novel therapeutic targets for improved management of the disease.

Numerous PTMs to histone proteins have also been associated with the development of type 2 diabetes and accompanying complications.148,149 One example is that the TNF-α and COX2 promoters have increased H3K9 acetylation in blood monocytes from patients with either type 1 or type 2 diabetes compared with healthy controls.156 H3K9 acetylation is associated with open chromatin structures and increased gene expression. These in vivo data, along with cell culture data, show that diabetic conditions are capable of inducing chromatin modifications to induce the expression of inflammatory genes involved in the diabetic disease process. The complicated interplay of causes and effects in metabolic and epigenetic changes involved in diabetes is likely to be an ongoing research focus for some time that will ultimately provide additional therapeutic targets not only for the stabilization of blood glucose levels but also possibly for the prevention of disease progression and the corresponding complications.

DR is a common complication of diabetes, affecting greater than half of patients having type 2 diabetes for ≥10 years and is the leading cause of blindness among working-age adults.142 Although glaucoma and AMD each have associations with processes, including immune function, which suggest that they may have systemic contributions to the disease processes, DR is clearly the result of systemic disease. DR develops via hyperglycemic induction of vasoactive and inflammatory factors that leads to vascular damage and neuronal injury, further leading to ischemia and the subsequent formation of new vessels, which are prone to bleeding and lead to retinal detachment and ultimately vision loss (Fig. 3D).157,158 Specific epigenetic modifications associated with the development of DR include decreased H3K4me2 in the promoter and enhancer regions of the manganese superoxide dismutase gene, Sod2, in retinas from human donors with DR compared with age-matched nondiabetic controls.159 H3K4me2 is associated with transcriptional activation, and Sod2 expression is also decreased in retinas from human donors with DR compared with age-matched nondiabetic controls. Additionally, the expression of lysine-specific histone demethylase-1, which was shown to demethylate H3K4 at the Sod2 promoter in bovine retinal endothelial cells, was increased in the retina of diabetic donor eyes. Zhong and Kowluru also observed increased H3K9 acetylation and the accompanying NFκB subunit p65 binding at the matrix metalloproteinase-9 (MMP9) promoter in retinas of human donors with DR compared with age-matched controls, consistent with MMP9 expression being increased in DR retinas compared with controls.160 Another study from the Kowluru laboratory used donor retinas with a postmortem time of six to eight hours from patients with established DR and age-matched nondiabetic controls and observed increased expression of Kelch-like erythroid cell-derived protein with Cap’n’colar (CNC) homology (ECH)-associated protein 1 (Keap1), which is an inhibitor of the antioxidative stress agent Nrf2 in the DR retinas compared with controls.161 The increased expression of Keap1 was accompanied by increased Sp1 binding at the Keap1 promoter, which the authors showed to be dependent on the methyltransferase SetD7 in bovine retinal endo-thelial cells, suggesting that monomethylation of H3K4 at the Keap1 promoter is necessary for the increased Keap1 expression in DR.

These studies have provided insights into the epigenetic mechanisms that may underlie gene expression in type 2 diabetes and DR. Genome-wide epigenetic studies are needed to further elucidate the epigenetic processes involved in the progression of DR. In particular, investigations comparing the epigenetic profiles of monozygotic twins discordant for DR will potentially highlight differentially modified loci with significant influences on disease development. Such studies will improve the understanding of disease progression and identify potential opportunities for therapeutic intervention.

Future Directions: Epigenetics as Therapeutics

Studies seeking to understand the various epigenetic changes—and their relationship to gene expression—involved in retinal diseases will inform researchers regarding the intricate molecular mechanisms and processes that lead to the development and progression of complex blinding diseases. With an understanding of the disease mechanisms, researchers will develop treatments to effectively manage the full spectrum from preclinical to advanced stages of disease, including both interventions that prevent clinical manifestations in asymptomatic patients and those that prevent the progression to blinding advanced stages and/or promote the regression of existing disease phenotypes.

Data indicating which genes undergo epigenetic modifications associated with the development of retinal disease will provide potential therapeutic targets for interventions that alternatively fix the aberrant chromatin modification(s), target the particular gene/gene product whose expression is pathologically altered as a result of the aberrant chromatin modification(s), or target proteins or RNAs that interact with or facilitate the effects of the affected gene/gene product. Therapeutics that target chromatin-modifying enzymes, such as HDAC inhibitors and DNMT inhibitors, were initially developed as treatments for cancers, and some are currently in clinical use or clinical trials (see the NIH website for current clinical trial information: ClinicalTrials.gov).162–164 More recently, epigenetic therapeutic agents are being considered as potential treatments for noncancerous complex diseases, such as CVD and diabetes, which share common risk factors with age-related retinal diseases.165–168 Additionally, preclinical studies are investigating the potential use of epigenetic therapeutics for the treatment of AMD and DR.169 However, at the time of writing this report, clinical studies of therapeutic agents that directly modulate epigenetic processes have yet to be executed.

A potential epigenetic intervention was observed in a recent clinical trial looking at specific B-vitamin supplementation. The study looked at the association of folic acid, vitamin B6, and vitamin B12 supplementation on the development of AMD in women with either preexisting CVD or at least three risk factors for CVD, which are also risk factors for AMD.170 The authors found an association between folic acid/vitamin B6/vitamin B12 supplementation and protection against AMD development. An ongoing clinical study is also investigating the effects of folic acid, vitamin B6, and vitamin B12 supplementation on the progression of DR.171 Folic acid is an upstream methyl donor for DNA methylation; vitamins B6 and B12 are involved as cofactors in the process of transferring the methyl group indirectly from folic acid to the DNA. Recent data suggest an association between dietary folic acid intake and global methylation levels.172 Therefore, the association of folic acid and vitamins B6 and B12 supplementation with decreased incidence of AMD may indicate an epigenetic influence on disease progression and, therefore, an epigenetic modulation of disease risk via supplementation. However, the effect of folic acid and vitamins B6 and B12 supplementation on disease risk may result from alternative molecular mechanisms, such as direct antioxidant effects.170 Therefore, additional studies are needed to elucidate the molecular mechanism of folic acid and vitamins B6 and B12 supplementation in the progress of AMD and DR.

Further studies into the epigenetics of blinding retinal diseases will provide therapeutic targets for improved treatment strategies. Current treatments for neovascular AMD have limited results and require frequent intraocular injections, which are not ideal for the patient. The development of therapies that reduce or eliminate the need for injections will greatly reduce the burden of care on patients and the health-care system at large. For example, administration of pharmacological agents orally or via eye drops will not only reduce the pain and trouble associated with current therapies but could also improve vision over the long term.

Perspectives

The progression of complex disease is a multifactorial process influenced by various inputs, including genetic, epigenetic, and environmental elements. Extensive genome-wide analyses are insufficient to explain the development and progression of the complex disease from phenotypically normal through advanced disease. Studies utilizing well-characterized collections of tissue, which take into account tissue integrity and represent each stage of the disease, may help us to understand the molecular changes taking place as the disease progresses. Genome-wide epigenetic and expression data for each stage of the disease will provide candidates for involvement in the progression of disease. However, whether these factors are causes or effects of disease progression will have to be teased out with additional studies designed to determine the mechanistic role of each of these factors.

An ideal experimental setup for genome-wide epigenetic experiments will need to incorporate various features to ensure sufficient power and noise reduction in the resulting data (Fig. 6). Each of the nine characteristics illustrated in Figure 6 contributes to the production of meaningful epigenetic data that can be used to identify genes that are differentially methylated/modified in specific disease states. The generation of high-quality genome-wide epigenetic data pertinent to the development or progression of disease requires that relevant tissue(s) be sampled from both thoroughly phenotyped diseased subjects and well-characterized control subjects who are appropriately age and gender matched. Sampling from both affected and unaffected tissues for both the diseased and the control subjects will also allow for determination of changes that are isolated to the relevant tissue(s). Furthermore, the postmortem interval (time from death to preservation of the tissue sample) is critical for the extraction of accurate information. Both the number and quality of the samples influence the power to identify statistically significant differences associated with disease.

Figure 6

Ideal experimental setup for genome-wide epigenetic experiments. Each of these nine characteristics contributes to the production of meaningful epigenetic data that can be used to identify genes that are differentially methylated/modified in specific disease states. The combination of robust epigenetic data with RNAseq expression data provides additional information pertaining to the mechanism of disease by correlating expression changes with underlying chromatin modifying processes and providing initial insights into the mechanisms of disease. These studies identify genes and/or regulatory regions that potentially have causative roles in the development or progression of disease. Further mechanistic studies focusing on the roles of each correlated gene/regulatory region will then provide additional understanding of the disease and targets for therapeutic interventions. Incorporation of genetic and lifestyle data will also improve analysis by integrating the full spectrum of risk into studies of disease mechanisms.

The combination of robust high-resolution epigenetic data with RNAseq expression data provides additional information pertaining to the mechanism of disease by correlating expression changes with underlying chromatin modifying processes and providing initial insights into the mechanisms of disease. These studies identify genes and/or regulatory regions that potentially have causative roles in the development or progression of disease. Further mechanistic studies focusing on the roles of each correlated gene/regulatory region will then provide additional understanding of the disease pathways, networks, and points for therapeutic interventions. Incorporation of genetic, lifestyle, and other environmental data will also improve analysis by integrating the full spectrum of risk into studies of disease mechanisms.

Why do some patients’ diseases progress and others’ stop at an intermediate stage? The complex functional networks at play within a cell or tissue are likely susceptible to small perturbations exerted by environmental influences over long periods of time (Fig. 5). The gradual accumulation of altered expression or function of key factors within a system may then lead to dysfunction and disease. Disease-promoting factors (such as smoking, obesity, and poor nutrition) work to alter the networks/pathways in ways that lead to pathological states, whereas disease-preventing factors (such as antioxidant consumption and healthy BMI) oppose the progression of disease by keeping the system within the bounds of healthy function. The ratio of disease-promoting and -preventing influences may balance at different equilibrium points, leading to individual cases that progress only so far along the pathological pathway before reaching that equilibrium point. By discovering the various promoting and preventing factors, along with their respective significances at each point along the disease progression timeline, researchers (and eventually clinicians and patients) will be able to balance the networks and pathways to prevent or reverse the progression of disease. Such interventions are desperately needed to address the growing costs of complex diseases on the quality and duration of life of millions of people. Additionally, therapeutics capable of preventing progression to advanced, debilitating stages of disease will also decrease the overall financial cost of managing these patients’ diseases.

Summary

Scientists in the field need to tease out the various relationships of cause and effect among genetic, environmental, and epigenetic contributions on both gene expression changes and protein interactions which lead to or result from pathological processes, including potentially circular effects of self-propagating cycles of decline.

With an understanding of the complex relationships between genetic, epigenetic, and environmental factors involved in the development of complex diseases, such as degenerative retinal diseases, preventative and therapeutic interventions will follow that will both decrease incidence of debilitating diseases and also reduce the severity of and complications arising from such diseases, thereby improving the quality of life for older patients.