Printer Friendly

Forensic individual age estimation with DNA: From initial approaches to methylation tests.

TABLE OF CONTENTS

 INTRODUCTION                                                   122
 Forensic Context                                               122
 Initial Approaches for Age Estimation                          123
 Human Aging and Epigenetics                                    124
 Epigenetic Marker of Aging: DNA Methylation                    125
I. DNA METHYLATION PROFILING TECHNOLOGIES                       126
 A. Bisulfite Conversion                                        127
 B. Genomewide Coverage: Discovery of Suitable Age-Informative
 Methylation Sites                                              129
 C. Intermediate Coverage: Validation                           130
 D. Low Coverage: Forensic Implementation                       131
II. FORENSIC AGE ESTIMATION USING DNA METHYLATION TESTS         131
 A. Current Forensic Age-Prediction Models                      131
 B. Candidate Genes                                             133
 C. Intertissue Variability                                     136
 D. Intergroup Variability                                      138
 CONCLUDING REMARKS                                             138
 ACKNOWLEDGMENTS                                                139
 REFERENCES                                                     139
 ABOUT THE AUTHORS                                              144


REFERENCE: Freire-Aradas A, Phillips C, Lareu MV: Forensic individual age estimation with DNA: from initial approaches to methylation tests; Forensic Sci Rev 29:121; 2017.

ABSTRACT: Individual age estimation is a key factor in forensic science analysis that can provide very useful information applicable to criminal, legal, and anthropological investigations. Forensic age inference was initially based on morphological inspection or radiography and only later began to adopt molecular approaches. However, a lack of accuracy or technical problems hampered the introduction of these DNA-based methodologies in casework analysis. A turning point occurred when the epigenetic signature of DNA methylation was observed to gradually change during an individual's lifespan. In the last four years, the number of publications reporting DNA methylation age-correlated changes has gradually risen and the forensic community now has a range of age methylation tests applicable to forensic casework. Most forensic age predictor models have been developed based on blood DNA samples, but additional tissues are now also being explored. This review assesses the most widely adopted genes harboring methylation sites, detection technologies, statistical age-predictive analyses, and potential causes of variation in age estimates. Despite the need for further work to improve predictive accuracy and establishing a broader range of tissues for which tests can analyze the most appropriate methylation sites, several forensic age predictors have now been reported that provide consistency in their prediction accuracies (predictive error of [+ or -]4 years); this makes them compelling tools with the potential to contribute key information to help guide criminal investigations.

KEYWORDS: Age estimation, bisulfite conversion, DNA methylation, epigenetic age, epigenetic clock, forensic, statistical regression models.

INTRODUCTION

Forensic Context

Forensic age estimation can be used to gain information relevant to criminal, legal, or anthropological investigations [129]. First, the age of a person from analysis of their remains or from the biological traces they leave behind can help guide individual identification of a missing person or at the crime scene [2]. Multiple samples can be collected from the scene including hairs, bloodstains, semen, cigarette butts, and personal items such as toothbrushes. However, if DNA profiling does not lead to a match with any DNA database entries, no suspects can be identified. To overcome these barriers, forensic genetics is aiming to develop and validate new DNA intelligence tools. Rather than providing an individual-specific genetic profile, these new tests give information about characteristics that are shared by a much smaller group of potential suspects. Such characteristics include biogeographic ancestry [120] and externally visible characteristics (EVCs), alternatively termed phenotypic traits [78], that can create a more manageable pool of individuals sought by an investigation. Forensic DNA analysis could gain the same kind of advantage from the inference of individual age, as it is a clear feature of human appearance that is difficult to disguise. Second, the prediction of age-related physical traits, such as hair color or early-onset male pattern baldness, is improved by knowing the age of the DNA donor. Third, unidentified remains (of the missing, of victims of mass disasters, or of casualties in regions of conflict) are commonly encountered and are often challenging to identify if surviving relatives are not available [156]. In particular, mass disasters involve numerous samples that require detailed analysis in a short timeframe. Depending on the nature of the catastrophe, the remains will range in condition from relatively well preserved to highly degraded. Age estimation from DNA could therefore be used as a screening tool in order to accelerate the procedures or as supporting data in complex identifications. Fourth, the inference of age from DNA has considerable potential to add detail to the analysis of archeological remains [99]. Forensic anthropologists currently aim to identify a range of physical traits and the ancestry of ancient skeletal remains using established forensic DNA tests that could be extended to estimations of the age-at-death. Lastly; legal hearings could be supported by the inference of age from DNA samples taken from individuals whose age is in dispute, such as the likely age of an asylum seeker or the penalty applicable to young offenders [130]. This is a contentious issue, as the ascertainment of age in such matters requires a minimum level of accuracy and the way an estimate is obtained is invariably challenged in court on the basis of the technique's range of estimation values derived from its predictive error.

Initial Approaches for Age Estimation

Forensic age determination was originally based on the examination of bones and teeth. Different methods were proposed that examined adult skeletal remains and their details have been outlined by Baccino and Schmitt [8]. Selection of the best methodology applicable to each age-estimation scenario depends on the material available to the investigators. Morphological inspections of the pubic symphysis, sacropelvic surface, fourth rib and clavicle from bones, or root translucency from teeth are informative for the age-at-death. A recent study used the symphyseal face of pubic bone as a critical means to determine the age in the identification of a historical figure [50]. It is important to note that whatever the method of assessment used, the main drawback of use of skeletal features is their lack of informativeness for the elderly since, after 65 years of age, the accuracy of age inference from bones and teeth drops markedly. The estimation of age from the skeletal remains of children and adolescents can be additionally determined from the stage of dental or skeletal maturation [87]. In these cases, the degree of mineralization of the dentition is the preferred method. Furthermore, evaluation of the stage of emergence of the third molar is especially relevant in determining whether a subject is under or over the 18-year threshold [96]. Although it is compelling to apply this technique to the morphological analysis of undocumented living subjects of young age, the associated use of X-radiation is a major drawback that entails ethical issues. For these reasons, alternative methods using nonionizing imaging techniques such as MRI are currently being explored [32].

In addition to skeletal methods, molecular approaches for age estimation have increasingly been reported in the last five years. These are based on the gradual alterations to biomolecules that occur during the life cycle of an individual as a result of the aging process. Five main groups of molecular alteration have been assessed: (1) mitochondrial DNA deletions, (2) shortening of telomeres, (3) advanced glycation end-products (AGEs), (4) aspartic acid racemization (AAR), and (5) signal-joint T-cell receptor excision circles (herein sjTRECs). The analyses of (1), (2), and (5) are DNA-based, whereas those of (3) and (4) are protein-based methodologies. As groups 1 through 4 have been previously reviewed in detail [106], we only briefly describe them here.

Technique 1 measures the accumulation of mitochondrial DNA (mtDNA) deletions that increase with age. The basis for this correlation is the process of energy production in mitochondria from the respiratory chain mechanism. Oxidation of glucose and lipids in the mitochondria to produce ATP releases free radicals (also called reactive oxygen species) and generates damage in the mtDNA leading to accumulation of deletions. In particular, the 4977 nucleotide (nt) deletion has been explored as a candidate for age estimation. However, technical problems and a lack of accuracy have hampered its application in forensic cases, since wide confidence intervals were necessary and the comparison was close to the extremes of age: young (under 20 years) versus old (over 70 years) [107].

Technique 2 examines the characteristics of telomeres forming the structures located at the end of the chromosomes that consist of up to several thousand tandem repeats of the sequence TTAGGG. Shortening of the telomeres occurs in every cell after each mitotic division. In order to control this, cells can induce the expression of the enzyme telomerase, which adds repeat sequences to the end of the telomeres. However, this process loses efficacy with the passage of time. Although age-estimation tests have sought to measure telomere shortening, as with mtDNA deletions, technical reasons lead to imprecise results with a general lack of reproducibility [7].

Technique 3 examines AGEs that form a heterogeneous group of compounds produced from the glycation of proteins. They accumulate during the lifetime of an individual in different tissues and have been found to create the commonly observed color changes (yellow-to-brown) measurable using a colorimeter or spectrophotometer. The extent of color changes can be used to infer the age of the individual. Nevertheless, a lack of standardized procedures and the high heterogeneity of this group of biomolecules have hindered its use for age estimation. Moreover, for the method to have sufficient accuracy it is only applicable to the analysis of persons aged 45 years and over [122].

Technically, the most accurate age-estimation system is obtained by technique 4 examining AAR with an estimation error of [+ or -]3 years [112]. Racemization is a chemical process that converts pure enantiomers (L or D) to a mixture of both. In mammals, exclusively L-amino acids are incorporated during protein synthesis. As the individual ages, racemization takes place and the level of D-amino acid enantiomers starts to increase. AAR is based on the chromatographic measure of D-aspartic acid and this technique is particularly applicable to dentine as an optimal tissue for analysis. Although the accuracy of the method is high, the technique is destructive and so lacks wide-scale application. Furthermore, exposure of the teeth to high temperatures introduces bias to the results, notably in the identification of burned bodies.

Technique 5, which analyzes levels of sjTRECs, provides possibly the most viable molecular technique based on the presence of a particular species of DNA in the blood. The sjTRECs detected are circularized DNA molecules that result from T-lymphocyte DNA rearrangement. An age-correlated decline in the levels of [delta]Rec-[psi]J[alpha] sjTREC occurs during the individual's lifespan. Although the corresponding accuracy of the test is moderate (the estimation error is approximately [+ or -]9 years), the technique is restricted to blood samples and disorders affecting the immune system can potentially interfere with the estimates made [167].

The above methodologies comprise the molecular framework for forensic age estimation from its origins up to 2010. Since then, significant advances have been achieved in developing new DNA tests to infer individual age from a biological trace. The change of emphasis that enabled these advances was a shift in focus from characteristics of DNA and proteins toward epigenetic changes to DNA, exploiting increased knowledge of human epigenetic patterns from clinical studies. This shift has opened the door to the identification and development of age-correlated markers that have provided robust age-prediction models for forensic use.

Human Aging and Epigenetics

Aging has been described as "a slow, time-dependent decline of a set of multiple biological functions" [75]. It is a natural process characterized by a progressive loss of physiological integrity caused by the accumulation of cellular damage throughout an individual's lifespan; over time, this triggers the development of age-related diseases, including cancer, neurodegenerative disorders, diabetes, cardiovascular diseases, and others [95]. With accompanying internal phenotypic alterations with time, human aging also involves evident changes to physical appearance. Hair graying, progressive balding in males, decrease of muscle elasticity, as well as collagen loss leading to wrinkles; are the general visible characteristics that can help to describe the stage in the lifespan of a subject. A detailed description of the cellular and molecular features behind the internal or external aging alterations has been well described elsewhere [95]. From this exhaustive review, changes in patterns of the epigenetic signatures (also termed: epigenetic marks/patterns/modifications/modulators) are indicated to be the primary events impacting the aging process.

Different epigenetic marks shape the genome: chromatin remodeling [127], posttranslational modification of histones [135], noncoding RNAs [85], and DNA methylation [74]. All such marks undergo dynamic alteration in order to modulate gene expression. They do not act independently, but sequentially or in combination as a network, and the downstream outcome of the timing of their modulation is what constitutes the human epigenome or epigenetic code [148].

Epigenetic patterns cover the reversible, mitotically heritable changes in gene regulation that occur without any alteration to the underlying DNA sequence [126]. This definition leaves out any meiotic process, when recombination alters the arrangement of DNA in the chromosomes. The concept of transgenerational epigenetics, also referred to as the epigenetic inheritance, although currently under debate [76], is gaining experimental evidence and thus is becoming accepted as a process [4-6,25,110,118,133]. This non-Mendelian inheritance suggests that epigenetic changes that occurred in progenitors (e.g., environmentally induced epigenetic marks [70]), might not be entirely reset by epigenetic reprogramming after fertilization and therefore could be transmitted to offspring between generations through the germ line [31]. Transgenerational inheritance of longevity was reported for the first time in 2011 in Caenorhabditis elegans, supporting the possibility of enlarged lifespan due to epigenetic changes inherited from a limited number of previous generations [49]. Due to the high impact that this concept could have on society, much more research is needed and although inviting, this issue goes beyond the scope of this review.

Changes in epigenetic patterns are found to occur either during development or childhood/adulthood. The main difference is that during embryonic and germ cell development, changes to the epigenetic landscape are biologically reprogrammed and necessary for lineage determination and cellular identity [24,125], whereas in adult somatic cells, this remodeling of epigenetic marks reflects age-associated deleterious events [139]. Therefore, epigenetic changes have a major influence on the human aging process [115]. These modifications--grouped under the term "epigenetic clock" or "epigenetic drift" [73]--are the result of the continuous interactions between genetics and environment that occur during the individual's lifespan.

The epigenetic clock is a directional phenomenon caused by nonstochastic events and involves specific regions of the genome associated with age [115]. In contrast, epigenetic drift entails genomewide variations due to stochastic events [141] and subsequently leads to unpredictable and inconsistent differences in the epigenome between aging individuals. Interindividual differences are best understood in studies monitoring changes across the lifetime of monozygotic (MZ) twins [100]. MZ twins are the ideal model for exploring the influence of environmental factors compared to genetics, as each twin pair shares identical genomes. Fraga et al. [42] reported that during early years of life, MZ twins are epigenetically similar. Epigenetic differences emerge as both individuals become older, giving rise to variations in susceptibilities to disease as well as creating distinguishing anthropomorphic features. Although the underlying molecular pathways of epigenetic drift are still unclear [89], it is known that these variations can be caused by different lifestyle characteristics, including smoking habits [146], alcohol intake [51], physical activity [9], diet [1], and others.

The cumulative changes created by the epigenetic clock and epigenetic drift during a lifetime can be measured and, therefore, will be informative of the age of an individual. This measure will be referred to from now on as the epigenetic age. For forensic investigations, the epigenetic age should aim to match the chronological age as closely as possible. So the chronological age, as time elapsed since birth [12], is different from the biological age, which is a measure of the aging process in the individual correlated to life expectancy and influenced by factors that are independent of the passage of time alone. Differences between chronological and biological age are primarily influenced by the occurrence of health disorders. In such a way, epigenetic age acceleration is commonly observed to be associated with age-related diseases and mortality [26,119]. In addition, there are strong effects from characteristics such as trisomy 21 (Down syndrome) or the occurrence of HIV infection [59,61]. Tissues undergoing neoplastic processes also show accelerated epigenetic aging [52]. Furthermore, some rare human genetic disorders such as Werner and Hutchinson-Gilford progeria syndromes [81] are characterized by premature aging phenotypes with a shortened lifespan and therefore, are suitable models to gain insight into the biology of aging in humans [22]. It is noteworthy that centenarians presenting moderate levels of human aging compared to the average, display lower epigenetic age [62]. Similarly, a decelerated aging process is observed in young females with the X syndrome (children that "evade aging") [154].

In the last five years, the estimation of epigenetic age as a predictor of chronological age has grown substantially and among the features of epigenetic changes, DNA methylation sites have become established as the most informative marker for aging.

Epigenetic Marker of Aging: DNA Methylation

DNA methylation is the best understood epigenetic signature in the genome [67,134,137]. It is a chemical modification that, in mammalian genomes, predominantly occurs at the 5' carbon atom [C5] at cytosine residues that are followed by guanine (termed CpG dinucleotides), leading to 5-methylcytosine. Less frequently, DNA methylation can occur in a non-CpG site (CpH; H = A, C, T) in a limited number of specific cells [54]. These include embryonic stem cells, where non-CpG methylation (NCGM) is lost upon cell differentiation [92]; neurons (NCPM accumulates through early childhood and adolescence) [91]; and oocytes (NCPM disappears by the blastocyst stage) [144]. CpG dinucleotides (alternatively, CpG sites or CpG positions) in mammalian genomes are characterized by a global methylation potential, which is defined as the ability to undergo methylation at positions in all categories of DNA sequence, including genes, transposons, and intergenic DNA [123]. However, the DNA methylation landscape varies according to the genomic region. In the bulk of the genome, methylation occurs at a high rate and is especially abundant in repetitive elements, while it is less common in so-called CpG islands [15,33]. CpG islands are CpG-rich regions (%CG > 55%) of more than 500 nts in length that account for 1-2% of total CpG sites in the genome. About 60% of the CpG islands span gene promoters, while the remaining 40% are localized in intergenic or intragenic regions and are defined as "orphan CpG islands". CpG islands found at the promoters of many housekeeping or developmentally regulated genes are constitutively hypomethylated, defined as below average levels of methylation per CpG site, and this characteristic promotes active transcription. Conversely, some repressed genes have methylated promoter CpG islands whose shutdown is maintained in the long term [74]. For this reason, DNA methylation is generally regarded as an indication of gene silencing [131]. The repressive role of this epigenetic signature contributes to the formation of heterochromatin and has several biological functions such as genomic imprinting [159], X-chromosome inactivation [29], or silencing of repetitive DNA [147]. Either hypermethylated (above-average methylation levels) or hypomethylated DNA patterns are established during early embryonic development and are maintained throughout life in the somatic cells. The mechanism underlying the DNA methylation process is performed by specific enzymes called DNA methyltransferases (DNMTs) [13] that covalently modify target CpG sites by addition of a methyl group using S-adenosyl-L-methionine (SAM) as a cofactor. Three DNMTs have been described for mammals: DNMT1 is dedicated to maintaining the existing DNA methylation level, while de novo methylation processes are mediated by DNMT3A and DNMT3B [113]. In such a way, DNMT1 precisely maintains the symmetrical DNA methylation present in both DNA strands. For this purpose, during mitosis, it recognizes hemimethylated CpG palindromes (mirrored sequence across both strands) and incorporates a methyl group to the complementary cytosine. DNMT3A and DNMT3B are focused on the waves of remethylation that occur during epigenetic reprogramming. Hypomethylated DNA patterns also require maintenance, and for this process histone posttranslational modifications and transcription factors help to moderate the action of DNMTs [38,128].

With increasing age, DNA methylation patterns gradually display a genomewide DNA hypomethylation (affecting promoters, exonic, intronic, and intergenic regions), while in specific regions there is localized DNA hypermethylation (certain promoter-associated CpG islands) [19,75,161]. The locations of age-associated hyper- and hypomethylation differ, which indicates their underlying functions also differ [101]. Results from Marttila et al. [101] show that the most hypermethylated sites were clustered around processes of development and morphogenesis, gene expression, and nucleotide metabolism, while hypomethylated sites did not belong to a specific group. Although DNA hypermethylation has been largely linked to the silencing of CpG island promoters of tumor-suppression genes [83,93,94,97,143], repression of additional promoters related to the aging process--including the immune response, coagulation, and connective tissue homeostasis--has also been reported [161]. On the other hand, gradual DNA hypomethylation has been found at repetitive elements of the human genome, e.g., promoting activation of retrotransposons (discrete pieces of DNA that can be independently duplicated and move within the genome) [115]. In young individuals, retrotransposons are epigenetically silenced; however, when the individual ages a decrease in the bulk of histones leads to a reconfiguration of the chromatin. Retrotransposons are then located in an open state, leading to transcription and transposition elsewhere into the genome; a frequent event found in cells undergoing neoplastic [30] and neurodegenerative processes [140]. Decreases in levels of DNA methylation with aging also contribute to the activation of repetitive elements that are maintained in a constitutive heterochromatin state by accumulation of methyl groups in younger epigenomes in all cell types, which leads to increasing genomic instability with age.

Although DNA methylation patterns are disrupted in some common age-associated diseases, it is important to note that human development and healthy aging also reveal age-related variation in methylation patterns that accumulate with age [3,11,17,40,52,58,72,79,103,124, 151]. As previously described, a decline in CpG methylation is predominant during the aging process and has been observed in all human chromosomes [55]. It has been suggested that age-associated hypomethylation is a passive process caused by stochastic or environmental effects and associated with biological age [101]. This contrasts with the proposal that hypermethylation is an active process, caused by programmed or pseudoprogrammed aging mechanisms, and strongly associated with chronological age [101]. Similar conclusions were reported by Bell et al. [11], since their findings indicate that age-related alterations do not appear to be random events. It is important to take into account that hypermethylation in age-related diseases affects all CpG sites within a CpG island, whereas the methylation pattern in normal elderly individuals is partial and heterogeneous [161].

Since age-related DNA methylation changes are a well-established process, taking place in the human genome during an individual's lifetime, the next aspect is to know when differences in the epigenetic clock arise. These variations emerge before adulthood and thereafter they increase [58]. Epigenetic changes increase exponentially up to adulthood and then slow down to a linear pattern of change later in life [77]. Although pediatric cohorts have not been widely studied to date, some studies provide more detail about age-related DNA methylation changes occurring during the early years of life [3,47,132,157]. These changes occur more rapidly during childhood at a logarithmic scale of change [3] and are associated with developmental and immune ontological functions [3,149], in accordance with biological processes likely to occur in children. Furthermore, a higher rate of change is generally detected in early versus later childhood, e.g., greater DNA methylation differences were observed between 2 and 10 years compared to 10 to 16 years in a follow-up study [47].

In summary, DNA methylation analysis is currently the best way to predict chronological age. The analysis of specific regions of the genome displaying a direct relationship between their methylation changes during aging, and in a common manner in all individuals, can be used as an informative system for estimating chronological age in both adults and children. For this purpose, cross-sectional and longitudinal studies have been performed using different technologies. Table 1 shows the most relevant studies made to date that cover age-correlated epigenetic patterns and constituting a valuable source of data for the selection of human age markers.

I. DNA METHYLATION PROFILING TECHNOLOGIES

Multiple technologies have been described for the detection and analysis of DNA methylation [82]. They can be divided into three main subgroups according to the pretreatment of the DNA prior to analysis: (a) enzyme digestion (using restriction endonucleases); (b) affinity enrichment (using antibodies or methyl-binding proteins specific for methylated CpG sites); and (c) sodium bisulfite conversion. In this review we restrict our discussion to the third pretreatment because nearly all methylation-analysis technologies relevant to forensic tests, from age-marker discovery to test implementation, use bisulfite conversion.

A. Bisulfite Conversion

Bisulfite conversion was discovered in the 1990s and represented a breakthrough for the detection and analysis of DNA methylation [27,45]. Denatured genomic DNA (ssDNA) is treated with sodium bisulfite that deaminates unmethylated cytosine residues to uracil. After PCR, uracil is amplified as thymine, while 5-methycytosine residues are amplified as cytosine. This chemical reaction causes substantial DNA degradation and therefore, subsequent purification is necessary to remove the sodium bisulfite. A detailed protocol has been described elsewhere [28]. Additionally, several commercial kits are currently available in order to proceed with this pretreatment. The quantity of input DNA suggested by the manufacturers, although variable and dependent on the subsequent technology applied, is higher than for STR typing (see Table 2). An average of 200-500 ng of input DNA is normally used to compensate for the degradation that the sodium bisulfite produces in the DNA. Currently, some laboratories are internally testing minor input material and therefore, subsequent optimization of the protocols could reduce the quantity of DNA used.

Bisulfite-based methods provide single-base-pair resolution and accurate quantitative data. Therefore, the majority of new DNA methylation data is based on this methodology. The conversion step is then followed by DNA amplification with target-specific primers. A source of bias can be incomplete bisulfite conversion; however, this can be overcome by the parallel analysis of DNA methylation controls or internal converted non-CpG cytosines to measure the success of the conversion reaction. Differential PCR efficiency for methylated versus unmethylated portions of identical sequence can also introduce some bias. It should also be noted that bisulfite conversion cannot differentiate between 5-methycytosine (5-mC) and 5-hydroxymethycytosine (5-hmC) [63]. The nucleotide 5-hmC has recently received much attention after its detection in high levels in Purkinje neurons and embryonic stem cells [80], as well as the simultaneous discovery of a group of enzymes that catalyze the hydroxylation of 5-mC to 5-hmC (e.g., Ten-Eleven-Translocation oxygenases or TETs [138]). As a result, this has introduced a new residue in the genome, whose role in gene regulation still needs to be elucidated [102]. Although 5-hmC is beyond the scope of the present review, it is important to bear in mind that future studies will focus on deciphering its biological function. New protocols that can detect 5-hmC have been further optimized [20,160] and eventually it will be known whether age-progressive changes in levels of this residue could also contribute to age estimation [23].

The proportion of DNA methylation at a particular CpG site is called the methylation [beta]-value. This value is assessed by taking the ratio of methylated (C) to unmethylated (T) signal. The fraction is generall y calculated with the formula: M/(M+U), where M represents the signal for methylated molecules and U the signal for unmethylated molecules. DNA methylation is measured on a scale of 0 to 1 (or 0 to 100%). A [beta]-value of 0 indicates complete absence of methylation, while 1 indicates full methylation at the studied CpG site. Since DNA methylation is measured as an average from a pool of cells and because progressive alterations are experienced in these cells during the individual's lifetime, a bimodal distribution for [beta]-values is not seen in age-related markers, but gradual values between 0 and 1.

B. Genomewide Coverage: Discovery of Suitable Age-Informative Methylation Sites

Whole-genome bisulfite sequencing (WGBS) and use of array-based DNA analysis such as the Illumina Infinium HumanMethylation array are technologies that allow genomewide scanning of DNA methylation patterns.

WGBS combines large-scale DNA bisulfite treatment with high-throughput sequencing to produce what is known as the "methylome" [92,164]. A methylome provides a complete map of the methylation patterns of all cytosines in the genome of an organism or a cell. The key advantage of this method is full CpG coverage (~28 million CpG sites in the human genome) including those non-CpG or CpH (H = A, C, T) sites, all detected at single-nucleotide resolution. In the context of human aging studies, WGBS was applied in the first place to explore the methylomes of newborns and centenarians, and detected a predominantly global age-related hypomethylation [55]. Data obtained from WGBS is unbiased and accurately mirrors the methylome landscape of a genome, since the whole sequence is mapped. However, the main drawback for full genomewide coverage is the prohibitive cost of studies covering large sample sizes. Furthermore, processing and managing the quantities of data generated is time-consuming and requires specialized computational tools. Targeted bisulfite sequencing was derived from WGBS as a method that reduces the costs by selective capture of a small fraction of the cytosines [165]. Because most of the genome is depleted of CpGs, many full-coverage studies lack relevant information; therefore, covering specific parts of the genome may lead to the retrieval of more informative epigenetic variation.

As an alternative to sequencing, the Infinium BeadChip array was created (Illumina, CA, USA). Array-based analysis uses sodium bisulfite conversion of DNA and subsequent single-base resolution of targeted CpG sites using capture probes complimentary to the sequences of interest and arranged on a microarray [90]. Chip-based arrays are cost-effective and the time of analysis is substantially reduced. The HumanMethylation27K BeadChip (HM27) was introduced in 2008 and interrogates over 27,000 CpG sites. This array was superseded in 2011 by the HumanMethylation450K BeadChip (HM450) comprising over 450,000 CpG markers [14]. The improved coverage of this chip provided a widely applicable tool and became the platform of choice for many epigenome-wide association studies (accession code: GPL13534 in online GEO data sets in NCBI). However, one drawback of the HM450 chip is that only about 1.5% of overall genomic CpGs are represented on the chip and the selection of sites is biased toward promoters and CpG islands. CpG sites in distal regulatory elements--i.e., enhancers--are strongly underrepresented. In order to compensate for this, an updated Illumina microarray has recently been released: the MethylationEPIC BeadChip [65]. This EPIC array covers 853,307 CpG sites and contains >90% of the 450K sites, but adds 333,265 CpGs located in regulatory regions. A large proportion of these additional sites lie in enhancer regions (sequences that impact transcription while lying distal to the transcription start site located in regions of poor CpG content) and the array is also useful for detecting 5-hydroxymethylcytosine [109]. However, the proportion of regulatory elements is still limited and comethylation of multiple sites within a small region cannot be assumed from a single CpG probe per element.

When data from WGBS and HM450/EPIC studies are directly compared, methylation levels obtained from both platforms are generally concordant and well correlated [55,121]. Whereas WGBS finds mostly hypomethylated sites with age [55], Infinium arrays predominantly find hypermethylation [11,17,40,58]; this can be accounted for by the bias each method shows toward specific regions in the genome. Additionally, alternative technologies that cover large-scale methylomes (e.g., affinity enrichment) are also consistent with the pattern of hypomethylation of CpG sites with aging [103]. When directly comparing the methylation patterns obtained from affinity-enrichment methods at CpG islands, hypermethylation is prevalent, indicating concordance between technologies when targeting similar sites.

In summary, genomewide technologies cover a high number of CpG sites throughout the genome. Although they are not forensically manageable at this scale, bioinformatic analysis of the data from key studies, which is accessible from public databases, may be analyzed to find correlations between given patterns of methylation and the recorded age of the donor at the time of the study. This provides a valuable discovery step to identify candidate age-informative markers without the need for large-scale studies.

C. Intermediate Coverage: Validation

Apart from Pyrosequencing, two main methods of methylation detection are applicable to the validation of DNA methylation markers of potential use for age estimation. These are the mass spectrometry--based EpiTYPER[R] system (Agena Bioscience) and massively parallel sequencing (MPS, alternatively next-generation sequencing or NGS). The coverage they both offer is intermediate, i.e., hundreds to thousands of CpG sites can be simultaneously detected with each system.

The Agena Bioscience EpiTYPER[R] system (San Diego, CA, USA; formerly Sequenom) is a bisulfite treatment--based method coupled with MALDI-TOF mass spectrometry [36]. Methylation levels are detected as CpG sets, comprising one, or multiple CpG positions in the same DNA fragment (obtained by base-specific cleavage of the DNA with specific enzymes). Therefore, multiple CpG positions in a set are detectable when they are closely positioned. Detailed protocol guidelines can be found in Suchiman et al. [136]. This technology targets genomic regions of about 200-600 nt in length, has a high degree of automation and is particularly useful for measuring large numbers of samples or regions in one analysis. The study of candidate regions or the validation of markers obtained from previous genomewide coverage data is both feasible and relatively affordable with the EpiTYPER[R] system. Technical replication between EpiTYPER[R] and Infinium BeadChip arrays indicates reproducibility between both systems [17]. Because of these advantages, the forensic community has already shown interest in developing age-prediction methods based on data generated with EpiTYPER[R] technology [44,48,158]. Although single-nucleotide resolution is achievable, some CpGs very close to each other are detected as a block and if individual analysis is necessary, additional techniques must be applied.

MPS platforms are covered by high-throughput sequencing technologies. There are multiple methods for MPS and subsequently, their capability of sequencing varies according to the selected methodology. Whereas WGBS aims to capture the whole genome from a sample, alternative MPS technologies were developed for targeting specific regions of the genome of reduced size (e.g., amplicon bisulfite sequencing). MPS platforms adapted to forensic DNA analysis and designed to be used as high-throughput technologies are gaining increasing traction [21]. Two main suppliers, Illumina (using the MiSeq detector) and Thermo Fisher (using Ion Torrent-based detectors of Ion PGM and Ion S5), offer compact sequence detection systems where the markers are analyzed by large-scale parallel sequencing of multiple short fragments in automated workflows. Both systems sequence the bisulfite-converted DNA in order to detect the targeted DNA methylation regions. MiSeq technology uses sequencing-by-synthesis: where a fluorescently labeled reversible terminator is imaged as each dNTP is added, and then cleaved to allow incorporation of the next base [66]. Ion Torrent MPS chemistry is based on semiconductor sequencing: each time a nucleotide is incorporated onto the growing DNA strand, a proton is released and variation in pH is measured [142]. Both methods provide similar maximum sequence read lengths (about 300-400 nt) and it should be noted that the underlying technology for MiSeq offers slightly better detection of homopolymeric stretches. Signal detection by Ion Torrent systems is performed in real time and therefore the corresponding run time is shorter than MiSeq, where the signal detection is made by imaging. However, more time is required for sample preparation using Ion Torrent technology. Both methods, although initially launched for SNP genotyping [34,35,104], are able to analyze DNA methylation, although this application is relatively new and optimization of the protocols is still required. One of the main advantages of targeted MPS approaches are minimal amounts of input DNA [150]. Although it is still early to establish technical comparisons with previous systems, some initial work exploring MPS data versus Illumina BeadChip data has been reported to provide high levels of reproducibility [114]. Additionally, direct comparison between DNA methylation data from age-related markers between MiSeq and Ion PGM has showed that although coverage for MiSeq is largely increased, methylation values from both platforms may be replicated (David Ballard, Kings College: London, UK; personal communication). Furthermore, an initial approach to predict individual age based on the MiSeq MPS system has recently been reported [152].

MPS platforms are part of the present review in the validation category, but they are promising technologies for full forensic implementation in the not-too-distant future. Although further optimization and consensus in the forensic community on appropriate markers is still required, much development work is currently in progress.

D. Low Coverage: Forensic Implementation

Once selected and validated, forensic markers need to be implemented in techniques that can deal with DNA typical of forensic biological traces: often comprising poor-quality and/or low-level DNA. In order to implement the analysis of locus-specific DNA methylation sites, pyrosequencing has been suggested as the "gold standard". However, it has some limitations that are largely overcome by single-base-extension (SBE).

Pyrosequencing was discovered by Pal Nyren in 1987 [111]. Currently, this method has been described as an extremely quantitative sequencing-by-synthesis technology [86]. Pyrosequencing utilizes the production of light after pyrophosphate release (luminescence), when a nucleotide is incorporated onto the growing DNA strand, in an enzymatic cascade system. The input DNA is bisulfite-treated. For a full technical description, see Tost and Gut [145]. Technical replication between pyrosequencing and Infinium BeadChip arrays has been reported to have high reproducibility [17], as well as with EpiTYPER[R] [16]. In the forensic context, pyrosequencing was the method of choice for the implementation of initial age prediction models [10,116,155,163]. Although read length for pyrosequencing is relatively short (less than 100 nt), it is not considered a disadvantage for forensic implementation, as the markers of choice are usually single CpG sites. However, pyrosequencing does not properly work when multiplexing, although some attempts to achieve this have been carried out. PCR multiplex reactions obviously desirable when multiple markers located in different genomic regions are targeted in tests aiming to analyze samples with low-level DNA. To solve this problem, single-base-extension can be applied.

SBE (also called mini-sequencing) consists of the annealing of an unlabeled oligonucleotide that matches the sequence immediately adjacent to the targeted site. Subsequent incorporation of a single complementary fluorescently labeled terminator (ddNTP) produces a sequence strand extended by one nucleotide. SBE has been extensively used in forensic genetics for SNP genotyping using the SNaPshot[R] assay [41,105]. By addition of bisulfite treatment to the protocol, DNA methylation can be readily detected using the same chemistry. The main advantage SBE offers is its capacity for multiplexing, which allows simultaneous detection of multiple CpG sites in a single reaction. A preliminary study using this technique for age estimation in sperm has recently been reported [84]. Although DNA methylation levels using SNaPshot[R] are quantitative results, particular care is required when assessing the fluorescence signals, since not all fluorophores display identical intensities. Therefore, it is recommended that the analysis be restricted as much as possible to C/T nucleotide combinations (yellow/red dyes) rather than G/A combinations (blue/green), since C and T signals have the most similar fluorescence intensities and so C/T peak height ratios are invariably more balanced. Although DNA methylation values obtained by SBE that were initially compared with Infinium BeadChip technology gave a high correlation [84], replication studies are strongly recommended to establish consistency between both techniques and the approaches developed by different laboratories.

II. FORENSIC AGE ESTIMATION USING DNA METHYLATION TESTS

Studies of DNA methylation specifically for forensic age estimation have continued to grow in numbers in the last five years. Several age-prediction models have been developed using partially overlapping genes, as described below. It is now the time for the forensic community to reach consensus on the best age-informative CpG sites covering the need to analyze different tissues, populations, age ranges, and detection technologies, with the ultimate aim of a universal age-prediction system.

A. Current Forensic Age-Prediction Models

To date, several age-prediction models applicable to analysis of forensic DNA have been developed based on a range of genes, tissues, and technologies. These are summarized in Table 2. The first application of DNA methylation for age estimation using a discrete number of CpG sites was that of Weidner et al. in 2014 [155]. Their study used a 3-CpG model comprising one CpG site located in three independent genomic regions. In this way, three sites identified as: cg02228185 (ASPA), cg25809905 (ITGA2B), and a CpG site close to cg17861230 (PDE4C), were used to construct an age-prediction model from the DNA of blood samples. Although cg17861230 was initially the target of analysis, the simultaneous detection of closely positioned CpG sites by pyrosequencing led to the detection of another CpG site 14 nt apart, with better association to age. Thus, a prediction accuracy of [+ or -]5.43 years was reported for the developed 3-CpG model that was then adapted to buccal swab samples (described in detail in section II.C), by using identical markers and technology, and obtaining a similar accuracy [37]. Even the single use of the PDE4C marker in buccal samples was proposed to be sufficiently informative to achieve reliable age predictions. However, at the time of this review, more detailed exploration of this single CpG model is required. A single gene for age estimation was also investigated by Zbiec-Piekarska et al. [162], focused on DNA methylation levels of ELOVL2 in blood samples using pyrosequencing. In this study, strong correlations with age were found for two CpG sites at the promoter region of ELOVL2. However, a further enhancement from additional markers was suggested for improving the suggested age-prediction model. Taking this into account, the first study was followed up with an updated age-prediction model using five CpG sites (ELOVL2, C1orf132, TRIM59, KLF14, and FHL2) [163]. The accuracy provided by this model improved previous results by achieving a median absolute deviation (MAD) predictive value of [+ or -]3.40 years. Open access for custom use of the age-prediction model from this study can be found at: http://www.agecalculator.ies.krakow.pl/. Once again, the simultaneous detection by pyrosequencing of closely positioned CpG sites allowed the discovery of additional markers that were even more closely associated with age than the original target CpG sites. Due to the high correlation to age shown by ELOVL2 CpG sites, this gene has been included in all subsequent forensic age-prediction systems. Park et al. [116] reported a blood-based model using DNA methylation levels of three CpG sites in ZNF423, ELOVL2, and CCDC102B (MAD: [+ or -]3.16 years). CCDC102B was already identified by Zbiec-Piekarska et al. as a highly informative gene, although not included in that study's final model. In parallel to Park et al.'s study, Zubakov et al. [166] included ELOVL2, FHL2, DUSP27, and ORAOV1 in an 8-CpG blood-based model reporting a MAD: [+ or -]5.09 years. In a third study completed at the same time by Giuliani et al. [48], both ELOVL2 and FHL2 plus PENK were brought together in an age-prediction system using dental samples. The number of markers used in Giuliani et al.'s model varied between 5-13 CpG sites, depending on the area of the teeth assessed. Interestingly, the most informative gene of ELOVL2 was not included in the preliminary age-prediction model constructed by Lee et al. using semen samples [84]. The tissue used as the DNA source for test development and for building the predictive model is a key factor. As described below, DNA methylation events occur in germinal cells that subsequently follow different pathways depending on the tissue differentiation steps.

Independent of the CpG sites or tissues evaluated, a common characteristic of all statistical prediction systems developed in forensic tests is the use of multivariate linear regression models. Linear regression is an approach for modeling the relationship between an interrogated variable (in this case, chronological age) and one or more observed variables (i.e., DNA methylation levels from one or multiple CpG sites). In addition, it is important to note that data for linear regression models needs to fulfil several underlying assumptions about the data: (a) absence of collinearity; (b) absence of heteroscedasticity; and (c) presence of normality. Of these three, the presence of collinearity is acceptable if predictions are made from the same combinations of predictor variables derived from the "training data" used to build the model [53]. However, if normality is absent and/or heteroscedasticity is present, the use of simple linear models becomes restricted. Although the presence of heteroscedasticity has not been directly reported in any of the above studies' prediction models, it is reflected by an absence of uniformity in the predictive errors. That means, although the associated MAD per model is a single value that measures the error in all predictions; in reality the predictive error proportionally increases with an individual's age. Most previous studies handle this effect by dividing the specific ages of study subjects into discrete groups corresponding to age ranges. When this is done, the predictive error shown by young subjects is always below that of elderly individuals. Bekaert et al. [10] directly show how the prediction error is age-dependent (see Figure 4 in [10]); their study divided the sample group into four 20-year age-span categories. The age predictor developed by Bekaert et al. was constructed using DNA methylation levels of blood samples at four CpG sites in ASPA, PDE4C, ELOVL2, and EDARADD genes. The EDARADD gene had also been selected by Weidner et al. [155], and was identified in the earlier study of Freire-Aradas et al. [44], as an informative age-prediction gene. However, technical problems prevented the use of this marker. It is also noteworthy that in the model developed by Bekaert et al., the data derived from ELOVL2 was treated statistically differently from the other genes used. A quadratic regression model was adopted for ELOVL2, in which the methylation levels of the gene were squared to maximize accuracy in the results (MAD: [+ or -]3.75 years). Although most selected CpG sites generally show linear correlations with chronological age, ELOVL2 does not, and thus a quadratic model was suggested as a way to improve predictions. Similar observations that the correlation between DNA methylation levels and age is not always explained by a simple linear regression were also highlighted by the study of Xu et al. [158]. This study developed an age-prediction model using six CpG sites from genes: ADAR, AQP11, ITGA2B, and PDE4C. Although quadratic regression was initially tested, the most accurate results were obtained assuming a support-vector regression model. Use of the regression-analysis alternative to a linear model was also applied in the age-prediction system developed by Freire-Aradas et al. [44]. This age predictor was based on a total of seven CpG sites located in seven genomic regions: ELOVL2, ASPA, PDE4C, FHL2, CCDC102B, C1orf132, and chr16:85395429 (no gene associated with the latter position). The study of Freire-Aradas obtained an average prediction error of [+ or -]3.07 years. The model was developed using DNA blood samples from a large training set of more than 700 individuals, which were uniformly distributed across both age and sex. The presence of heteroscedasticity was observed in the data (see Figure S3 in Freire-Aradas et al. [44]), so a multivariate quantile regression analysis was applied as the most appropriate prediction model for the data. The main advantage of using quantile regression instead of simple linear regression is that the former establishes age-specific prediction intervals each time new data contributes to the model. While a single interval is displayed in linear regression, prediction intervals in a quantile regression model are relative and age-specific. Free access to the corresponding online age-predictor tool developed by Freire-Aradas et al. can be found in the Snipper forensic classification website at: http://mathgene.usc.es/cgi-bin/snps/processmethylation.cgi. Lastly, a recent report from Vidaki et al. [152] introduced the use of artificial neural networks and showed consistent improvement over linear regression. The age-prediction model developed for this study counts 16 CpG sites and provides an MAD value of [+ or -]7.45 years.

In summary, in the space of just two to three years a considerable number of forensic age predictors have been developed, based mainly on blood samples, but with a growing number now covering additional tissues. Independently of the methylation-detection technology used or the statistical analysis applied, similarities in prediction accuracies are consistently reported. This trend confirms the clear finding that detection of DNA methylation levels leads to highly accurate age-prediction models. Although even higher levels of precision are desirable, it is more important at the present time, where laboratories have completed their candidate gene discoveries, to reach a consensus regarding which CpG sites should be combined into an agreed methylation panel before this is adapted for MPS-based analysis. B. Candidate Genes

In the search for methylation candidate genes, the predominant data reported in the public domain has come from the Infinium BeadChip array. Especially relevant is the Illumina HM450, because of its high coverage of methylation sites and use in a plethora of human gene expression studies. In order to capture the most age-informative genes described below, the criteria applied were: (a) reported DNA methylation levels show correlation with age; (b) multiple identification as a marker in current forensic age-prediction models; and (c) presence in methylation data derived from HM450 studies. The compilation of candidate genes in the HM450 array allows their methylation sites to be readily scrutinized both in existing and newly completed studies. The most comprehensive human expression dataset using HM450 chips is that of GSE87571, released at the end of 2016, and this data has been used in the review to generate new figures to demonstrate patterns obtained with the most age-associated candidate genes. A summary of the characteristics of the genes and their CpG sites used to describe age correlations are compiled in Table 3. It should be noted that other regions in the genome that are not covered by current Infinium BeadChip arrays could provide many additional age-correlated CpG sites equally or more informative than those currently described.

The top of the list of principal candidate genes for age prediction is headed by ELOVL2 (ELOVL fatty acid elongase 2). The correlation of chronological age with DNA methylation levels in the ELOVL2 promoter reaches the highest values and is consistently identified as the best predictor in numerous studies [40,46,52,72]. This gene belongs to the ELOVL gene family (elongation-of-very-long-chain-fatty-acids), i.e., a pool of elongase protein enzymes that have a role in human lipid metabolism from control of the elongation of fatty acids [69]. To date, seven ELOVL enzymes have been described (ELOVL#1-7) and all are covered by HM450 analyzing more than 100 CpG sites. Of the seven fatty acid elongation genes, ELOVL2 stands out as the best age predictor. ELOVL2 is located in chromosome 6 (approximately 6:10980500 to 6:11046600, GRCh37 coordinates) and encodes a transmembrane enzyme of the endoplasmic reticulum specifically involved in the elongation of polyunsaturated fatty acids (PUFA). During aging, the promoter of ELOVL2 becomes hypermethylated. Fifteen CpG sites in ELOVL2 are captured by HM450, of which five are located in the promoter region (cg16323298, cg16867657, cg21572722, cg24724428, and cg25151806). From these promoter CpG positions, cg16867657, cg21572722, and cg24724428 give the highest methylation-level correlation with age (Spearman correlation coefficients, [r.sub.s]: 0.9522, 0.9121, and 0.8857, respectively). The corresponding dispersion diagrams for these sites are shown in Figure 1A. However, DNA hypermethylation is slightly increased by cg16323298 ([r.sub.s]: 0.4704) and a constant pattern is observed for cg25151806 ([r.sub.s]: 0.1250). As shown in Table 3, the first three ELOVL2 CpGs of the above five are extremely close to each other (maximum distance: 17 nt between the two outlier sites). The distance increases for cg16323298 (about 80 nt) and cg25151806 (> 400 nt). Due to the high age correlation and reduced intervariability displayed by CpG sites in this genomic region, the most informative forensic age prediction systems developed so far have all included ELOVL2 markers in their models [10,44,48,116,162,163].

Descriptions of the five widely adopted genes--ASPA, EDARADD, FHL2, ITGA2B, and PDE4C--that are additional to ELOVL2 follow in alphabetic order.

The ASPA gene (aspartoacylase, alternatively named: SPATA22) is located on chromosome 17 (17: 3375500-3406600), and plays a role in the catabolism of proteins and peptides. It encodes an enzyme specifically catalyzing the conversion of N-acetyl-L-aspartic acid (NAA) to aspartate and acetate. ASPA is covered by HM450 with 10 CpG sites, from which cg02228185 is the most notable, since it has been included in several forensic age-prediction models [10,37,44,155]. Figure 1B shows the dispersion diagrams for the most age-correlated CpG sites from ASPA (cg02228185, [r.sub.s]: -0.7658 and cg12317815, [r.sub.s]: 0.7562). These plots show both CpGs are negatively correlated with age, and although more dispersion is observed at older ages, ASPA [beta]-values are more uniformly distributed in young adults.

The EDARADD gene (EDAR associated death domain) is located on chromosome 1 (1:236511600-236647900). The corresponding protein is found to interact with EDAR, a death domain receptor involved in morphological evolution, since it is required for the development of hair, teeth, and other ectodermal-derived tissues. Age correlation was initially reported by Bocklandt et al. [17], especially informative for the EDARADD CpG site cg09809672, and later explored as a forensic marker in [10,44,155]. A total of 21 CpG sites related to EDARADD are included in HM450, from these; cg09809672 ([r.sub.s]: -0.8182) together with cg18964582 ([r.sub.s]: -0.7271) are the most age-correlated CpGs from this gene. The dispersion diagrams for EDARADD in Figure 1C indicate that a loss of DNA methylation occurs in both CpG positions, although this is a stronger trend in the more widely studied cg09809672 site.

The FHL2 gene (four and a half LIM domains 2) is located on chromosome 2 (2:105974256-106054900) and encodes for a LIM domain protein. LIM domains are protein-interaction domains that regulate cell proliferation, apoptosis, and gene expression [71]. FHL2 protein is thought to have a role in the assembly of extracellular membranes. HM450 targets 34 CpG sites in this gene and from these, cg06639320 ([r.sub.s]: 0.9362), cg22454769 ([r.sub.s]: 0.9339) and cg24079702 ([r.sub.s]: 0.8999) have been repeatedly reported as highly age-correlated positions [40,44,46,48,163]. Figure 1D shows increasing hypermethylation occurs in FHL2, with progressive accumulation of methyl groups.

The ITGA2B gene (integrin subunit alpha 2b) is located on chromosome 17 (17:42449400-42468400). This gene plays a key role in the blood coagulation system, by mediating platelet aggregation. Twenty-two CpGs from ITGA2B are covered by HM450. Although this gene has been included in some forensic age-prediction models [37,155,158], the correlation with age is low-level, as can be observed in Figure 1E. The sites cg00062245 ([r.sub.s]: 0.6427) and cg25809905 ([r.sub.s]: -0.5247) are the only CpG positions where slight patterns of hyper- or hypomethylation with age have been found. However, further exploration of this gene could add much more detail about the usefulness of ITGA2B for age estimation in additional situations, including different tissues to blood or the age-range extremes. Additionally, it is important to note that the coexistence of simultaneously contrasting patterns of hyper- and hypomethylation that are each correlated with age for the same gene, is very rare. Nevertheless, it should also be noted that both CpG sites are a considerable distance apart (more than 5000 nt, see Table 3). They are therefore located in different regions of the same gene and subsequently, both sites could be part of quite distinct aging mechanisms.

The PDE4C gene (phosphodiesterase 4C) is located on chromosome 19 (19:18318900-18360800) and encodes a protein that regulates the cellular concentration of cAMP, a mediator of a cell's responses to extracellular signals. Of all the genes described in this review, PDE4C presents the highest coverage by HM450 (45 CpG sites in total). Figure 1F shows that the most age-correlated CpG sites in PDE4C (cg17861230 and cg20119148; [r.sub.s]: 0.7084 and 0.7114, respectively) display increased methylation levels with age. Several forensic age predictors have included PDE4C in their models [10,37,44,155,158], but from these studies it is noteworthy that CpG found 14 nt apart from the a commonly used cg17861230 has been detected as the most correlated with age. This relates to the point made in the introduction of this section, that HM450 is a powerful platform generating huge amounts of data, however additional information recovered from alternative technologies can often provide more informative CpG sites.

In summary, the genes briefly summarized here are good candidate loci that have been widely studied for the development of forensic age-estimation tests, because of their high levels of correlation between DNA methylation patterns and chronological age. This has been reflected in their consistent adoption as key components of many age-predictor models already reported by the forensic community.

C. Intertissue Variability

DNA methylation is also known to be an informative way to identify the body fluid or tissue source of forensic traces [153]. DNA methylation patterns are usually tissue/cell-type--specific and this property can be used in forensic scenarios to ascertain the biological source of many DNA traces. Because of this characteristic, a challenge in age-estimation tests that still needs further exploration is to fully evaluate the intertissue variability of DNA methylation levels; since it can be the case that principal predictors in one tissue type, such as blood, lack the necessary variation in other tissues, or lack a sufficiently close association to age. This issue was initially addressed by a multitissue age predictor developed by Horvath [58]. DNA methylation data from more than 50 tissues and cell types derived from studies based on HM27/MH450 was explored in [58]. From these analyses, a pool of 353 CpG sites was selected to construct a prediction system that could be uniformly applied to different tissues (MAD: [+ or -]3.6 years). Forensically relevant tissues examined by Horvath [58] included blood, saliva, and semen. From the data, lower epigenetic age estimates were obtained for semen compared to other parts of the human body. Similar results were found by Lee et al. [84] and consequently a specific age predictor for semen was proposed. Since semen samples are germinal cells, maintenance of epigenetic patterns could be governed by different biological pathways to those acting on somatic cells. Independently of this, to have a universal age predictor that could be applied simultaneously to blood or saliva samples would be advantageous per se for a forensic test. Nevertheless, the number of CpG sites necessary for the test makes a general test unwieldy. For this reason, Horvath's epigenetic clock, adapted in a variety of studies [59,62,119,132,154], is an ineffective forensic tests of age when poor-quality/quantity DNA specimens are analyzed. In order to be applied to forensic samples, a reduced number of markers should be compiled, and this constraint on the number of CpG positions analyzed leads to a lack of tissue universality in the test.

As previously described, Weidner et al. [155] developed an age-prediction model based on blood DNA methylation data from a restricted number of just three CpG sites (in genes: ASPA, ITGA2B, and PDE4C). The plots of predicted versus observed chronological ages showed high accuracy in blood-sample analyses (MAD: [+ or -]5.43 years). Two years later, the performance of this 3-CpG model was tested in buccal swab samples [37]. Results showed a high correlation between the predicted and chronological age ([R.sup.2]: 0:91). However, the data from buccal/saliva samples led to age being on average 14.6 years over their true values (see Figure 1B in Eipel et al. [37]). Since this is a consistent overestimation rather than a random effect, a retraining step for the previous model was adopted and a 3-CpG model was then reported (MAD: [+ or -]4.3 years). Furthermore, Eipel et al. [37] tested additional data sets in more detail which were derived from blood, saliva, and buccal swabs and a low correlation of DNA methylation with age was found for two of the CpG sites included in the model: cg02228185 (ASPA) and cg25809905 (ITGA2B). Conversely, cg17861230 (PDE4C) produced much stronger correlations with age in saliva and buccal swabs than in blood. This highlights the need for full evaluations of methylation patterns in a range of tissue types and the adaptation of the original underlying prediction model used in initial test development from a particular tissue, since differences invariably emerge. It could be suggested that these observations contradict Horvath's model; nonetheless, tissue-specific effects can be partly compensated by using a higher number of age-associated markers.

Parallel assessment of tissues was also covered by the study of Bekaert et al. [10]. In this study, a forensic age-predictor model was developed based on blood DNA methylation levels obtained from four genes (ASPA, PDE4C, ELOVL2, and EDARADD), and subsequently was tested on dentine samples. Prediction errors were slightly increased (MAD: [+ or -]4.84 in teeth versus [+ or -]3.75 in blood) and therefore, to increase the overall prediction accuracy of the test, a detailed search of teeth-specific age-associated markers was made. It should be noted that sample-size limitations might have influenced the results. Teeth samples were later explored in more detail by Giuliani et al. [48]. This study tested modern teeth for patterns in the three age-associated genes of ELOVL2, FHL2, and PENK, previously reported for blood [46]. Although sample-size constraints also applied to this study's data, preliminary results showed similar DNA methylation levels in teeth to those observed in blood and subsequently generated accurate predictions (MAD: [+ or -]1.20-7.07 years depending on the teeth area analyzed).

The analysis of teeth and bones deserves particular attention, since skeletal remains represent the most abundant sources of DNA for forensic anthropologists. As little data has so far been reported for epigenetic age tests for these tissues, additional investigations in either modern or ancient DNA samples is required to know if similar prediction models could be applied to historical samples. The handling of ancient DNA is more challenging, as deamination processes naturally occur from time of death, i.e., methylated cytosines tend to decay to thymines while nonmethylated cytosines tend to decay to uracils, over time in postmortem samples [57]. As stated by Pedersen et al. [117]: "epigenetic information is available from contemporary organisms, but it is difficult to track back in evolutionary time". The slowing or acceleration of the epigenetic clock at different CpG sites due to evolutionary variation in human diet or environment could differentially impact ancient and contemporary populations and their methylation patterns. In spite of these limitations, preliminary inquiries were explored from a sample of hair shafts from a 4,000-year-old Paleo-Eskimo belonging to the Saqqaq culture; where the analysis of two CpG sites (TRIM58 and KCNQ1DN) from the isolated DNA led to an age estimate between 44.1 and 69.3 years old [117]. Although these results need to be assessed cautiously, they might be considered as a prompt to investigate additional tissues that would be more useful in anthropology and also as evidence frequently found in forensic scenarios.

Bloodstains are also a common form of evidence in criminal cases and some studies have confirmed the stability of DNA methylation in such specimens. No statistical differences between the predicted age using blood and bloodstains were found by Huang et al. [64]. Additionally, age predictions for old bloodstains stored for up to four months at room conditions were correctly ascertained [64]. Longer periods of time were explored by Zbiec-Piekarska et al. [162], e.g., for bloodstains kept at room temperature for 5, 10, and 15 years. Although with time the DNA concentrations progressively decreased in such bloodstains, opposite to the degradation status of the sample, which gradually increased; those samples which could be amplified, generated similar correct age-prediction rates independently to the time of storage, confirming the stability of DNA methylation patterns.

Apart from testing for potential DNA methylation variation between tissues, potential intratissue variation under the influence of cell-type composition has also been suggested to be an important factor. Whole blood is a heterogeneous collection of multiple cell types, displaying different DNA methylation profiles, whose distribution and proportion are age-dependent [77]. For this reason, cellular composition has been regarded as a confounding factor that has been proposed should be taken into account in age-prediction models [68]. Nevertheless, the effects of cellular composition of blood were explored by Weidner et al. [155] under the 3-CpG model and no clear associations with cell type were found. These findings were in agreement with Horvath's epigenetic clock that only found minor differences across sorted blood cells [58]. Cell-type composition could also have a role in additional specimens such as buccal swabs--i.e., heterogeneous mixtures of buccal epithelial cells and leukocytes. In order to improve results in age estimation from these biological sources, two cell-type--specific CpG sites (CD6 and SERPINB5) were included in the model from Eipel et al. [37] adjusting predictions for cellular distributions, although prediction accuracy was only slightly increased.

Since at the present time several accurate age-prediction models have been developed, intertissue validation as a stepwise follow-up study is required for each model. It might be that the ideal result of a unique forensic multitissue methylation test is not a realistic outcome. In this case, independent models should be refined, especially for those tissues not fully studied so far, but frequently found in forensic or anthropological analyses, including semen, hairs, teeth, or bones.

D. Intergroup Variability

Potential variation in epigenetic age that could be linked to different individual characteristics that are unrelated to age must also be assessed in full. To ensure a forensic test's consistency and robustness, it is important to evaluate the variation discernible by dividing individuals into groups based on: (a) sex, (b) ancestry, and (c) the occurrence of disease in an individual's lifespan.

Sex as a variable that can influence DNA methylation patterns has been widely studied. However, contradictory results have been reported. Some studies detected sex-correlated DNA methylation patterns [18,52,60], while others did not find significant associations [43,79]. From such conflicting observations it could be the case that although the methylation patterns of some CpG sites are potentially affected by hormonal alterations directly linked to sex, those CpG positions targeted for age estimation could be neutral to this effect, and thus lacking such associations. Forensic age predictors developed so far have also provided different results. Whereas Bekaert et al. [10] and Freire-Aradas et al. [44] failed to detect any effect of sex, Weidner et al. [155] and Zbiec-Piekarska et al. [163] found slight contributions of this parameter to rates of aging.

Different populations originating from widely distributed regions of the world are exposed to various environmental factors, including varied diets, levels of UV radiation or air pollution exposure, and others. Since environment is directly linked to epigenetics, it could be the case that different DNA methylation patterns are found at the same CpG sites, when different ancestries are examined. Most of the studies carried out on age estimation have so far been based on European populations. Thus, it is necessary to explore the full range of worldwide population groups in order to ascertain if common age predictors can be applied independently of the population-of-origin; or if population-specific adjustments are required for the prediction model used.

Variation of DNA methylation across human populations was initially explored in European (CEU) versus African (YRI) HapMap individuals; with differences evident in hundreds of CpG sites [43,108]. In a subsequent study, even more marked variations in DNA methylation patterns were observed across three global populations (Caucasian-American, African-American, and Han Chinese-American) [56]. Nevertheless; it is important to note that from more than 400 differentially methylated CpG sites reported, none match the age-correlated candidate genes already described. However, additional research should be made to confirm these results. Horvath's epigenetic clock was recently assessed in five different population groups (Africans, Caucasians, Hispanics, East Asians, and Tsimane Amerindians) [60]. Although variation in the epigenetic age predictions was detected in different ancestries, larger datasets should be studied to balance the sample size of each population group analyzed. Independent of these results, it is important to note that the main age predictor gene ELOVL2 was also identified in an Asian study [116], as well as CCDC102B also included in the age-prediction model developed by Freire-Aradas et al. [44]. An additional Asian investigation also incorporated ITGA2B and PDE4C into their age-prediction model [158], previously reported by Weidner et al. [155]. This opens the door for the application of common forensic age-predictive marker sets around the world.

The presence of disease has been associated with altered DNA methylation patterns. Cancer in humans stands out as the single most important life event influencing wholesale epigenetic alterations. In fact, DNA methylation patterns have been considered a signature of cancer; when patterns of genomewide DNA hypomethylation, as well as DNA hypermethylation at promoter-associated CpG islands are present in tumors compared to healthy tissue counterparts [39]. Such cancer-induced effects can mimic epigenetic signatures of natural aging. Relevant for current age-prediction models, particularly those developed from blood samples, is the ascertainment of DNA methylation levels detected in tissues affected by the presence of tumors in different human organs other than blood. Evidence suggests that DNA methylation changes in blood might represent early events in the development of cancer or health-related outcomes, such as cardiovascular disease [88,98,119]. Therefore, caution should be exercised, and the adoption of age-prediction intervals related to particular ages can help reduce false predictions, while maintaining meaningful information.

CONCLUDING REMARKS

Forensic age estimation has received considerable attention in the last three to four years. The correlation of DNA methylation levels with chronological age has been the key step in improved prediction accuracy from such tests and their accelerated development for forensic use. Detailed research on CpG methylation levels underlies the major achievements in developing these tests, based in the first instance, on analysis of DNA from blood samples. Bloodstains form an important part of forensic evidence in many crime cases, but in addition, evidence often comprises hairs or semen and saliva traces. For this reason, further detailed examination of these tissues is necessary. Similarly, teeth and bones also require studies to develop tests useful for anthropological studies of skeletal material. Additional issues also need further examination, including the extension of methylation-pattern studies beyond European subjects and detailed analyses of life events and their influence on age-correlated methylation levels among individuals that are otherwise comparable in all other factors. Many worldwide forensic laboratories are now maximizing their efforts to validate a range of published models and sets of most informative CpG sites. With the increasing pace and scope of this work, a position of consensus on the optimum selection of CpG sites and the DNA analysis systems that best characterize them can be expected in the near future.

ACKNOWLEDGMENTS

The work that contributed to this review of methylation analysis was supported by CHRONOGEN (BIO2013-42188-R), a research project funded by the Ministry of Economy and Competitiveness, Spain, and cofinanced with ERDF funds, with MVL as PI. AFA was supported by postdoctorate funding awarded by the Xunta de Galicia, Spain, as part of the Plan Galego de Investigacion, Innovacion e Crecemento 2011-2015, Axudas de apoio a etapa de formacion postdoutoral, Plan I2C.

REFERENCES

(1.) Adaikalakoteswari A, Finer S, Voyias PD, McCarthy CM, Vatish M, Moore J, Smart-Halajko M, Bawazeer N, Al-Daghri NM, McTernan PG, et al.: Vitamin B12 insufficiency induces cholesterol biosynthesis by limiting s-adenosylmethionine and modulating the methylation of SREBF1 and LDLR genes; Clin Epigenetics 7:14; 2015.

(2.) Agudelo J, Halamkova L, Brunelle E, Rodrigues R, Huynh C, Halamek J: Ages at a crime scene: Simultaneous estimation of the time since deposition and age of its originator; Anal Chem 88:6479; 2016.

(3.) Alisch RS, Barwick BG, Chopra P, Myrick LK, Satten GA, Conneely KN, Warren ST: Age-associated DNA methylation in pediatric populations; Genome Res 22:623; 2012.

(4.) Anway MD; Cupp A, Uzumcu S, Skinner MK: Epigenetic transgenerational action of endocrine disruptors and male Fertility; Science 308(5727):1466; 2005.

(5.) Anway MD, Leathers C, Skinner MK: Endocrine disruptor vinclozolin induced epigenetic transgenerational adult-onset disease; Endocrinology 147:5515; 2006.

(6.) Anway MD, Rekow SS, Skinner MK: Transgenerational epigenetic programming of the embryonic testis transcrip-tome; Genomics 91:30; 2008.

(7.) Aubert G, Lansdorp PM: Telomeres and aging; Physiol Rev 88:557; 2008.

(8.) Baccino E, Schmitt A: Determination of adult age at death in the forensic context; In Schmitt A, Cunga E, Pinheiro J (Eds): Forensic Anthropology and Medicine: Complementary Sciences from Recovery to Cause of Death; Humana Press: Totiwa, NJ; 2006.

(9.) Barres R, Yan J, Egan B, Treebak JT, Rasmussen M, Fritz T, Caidahl K, Krook A, O'Gorman DJ, Zierath JR: Acute exercise remodels promoter methylation in human skeletal muscle; Cell Metab 15:405; 2012.

(10.) Bekaert B, Kamalandua A, Zapico SC, Van De Voorde W, Decorte R: Improved age determination of blood and teeth samples using a selected set of DNA methylation markers; Epigenetics 10:922; 2015.

(11.) Bell JT, Tsai PC, Yang TP, Pidsley R, Nisbet J, Glass D, Mangino M, Zhai G, Zhang F, Valdes A, et al.: Epigenomewide scans identify differentially methylated regions for age and age-related phenotypes in a healthy ageing population; PLoS Genet 8(4):e1002629; 2012.

(12.) Benayoun BA, Pollina EA, Brunet A: Epigenetic regulation of ageing: linking environmental inputs to genomic stability; Nat Rev Mol Cell Biol 16:593; 2015.

(13.) Bestor TH: The DNA methyltransferases of mammals; Hum Mol Genet 9:2395; 2000.

(14.) Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, Delano D, Zhang L, Schroth GP, Gunderson KL, et al.: High density DNA methylation array with single CpG site resolution; Genomics 98:288; 2011.

(15.) Bird A: CpG-rich islands and the function of DNA methylation; Nature 321:209; 1986.

(16.) Bock C, Halbritter F, Carmona FJ, Tierling S, Datlinger P, Assenov Y, Berdasco M, Bergmann AK, Booher K, Busato F, et al.: Quantitative comparison of DNA methylation assays for biomarker development and clinical applications; Nat Biotechnol 34; 2016.

(17.) Bocklandt S, Lin W, Sehl M, Sanchez F, Sinsheimer J, Horvath S, Vilain E: Epigenetic predictor of age; PLoS One 6(6):e14821; 2011.

(18.) Boks MP, Derks EM, Weisenberger DJ, Strengman E, Janson E, Sommer IE, Kahn RS, Ophoff RA: The relationship of DNA methylation with age, gender and genotype in twins and healthy controls; PLoS One 4(8):e6767; 2009.

(19.) Booth LN, Brunet A: The aging epigenome; Mol Cell 62:728; 2016.

(20.) Booth MJ, Branco MR, Ficz G, Oxley D, Krueger F, Reik W, Balasubramanian S: Quantitative sequencing of 5-methylcytosine and 5-hydroxymethylcytosine at single-base resolution; Science 336(6083):934; 2012.

(21.) Borsting C, Morling N: Next generation sequencing and its applications in forensic genetics; Forensic Sci Int Genet 18:78; 2015.

(22.) Burtner CR, Kennedy BK: Progeria syndromes and ageing: what is the connection? Nat Rev Mol Cell Biol 11:567; 2010.

(23.) Buscarlet M, Tessier A, Provost S, Mollica L, Busque L: Human blood cell levels of 5-hydroxymethylcytosine (5hmC) decline with age, partly related to acquired mutations in TET2; Exp Hematol 44:1072; 2016.

(24.) Cantone I, Fisher AG: Epigenetic programming and reprogramming during development; Nat Sructural Mol Biol 20:282; 2013.

(25.) Cavalli G, Paro R: The Drosophila Fab-7 chromosomal element conveys epigenetic inheritance during mitosis and meiosis; Cell 93:505; 1998.

(26.) Chen BH, Marioni RE, Colicino E, Peters MJ, Ward-Caviness CK, Tsai P, Roetker NS, Just AC, Demerath EW, Guan W, et al.: DNA methylation-based measures of biological age: Meta-analysis predicting time to death; Aging (Albany, NY) 8:1; 2016.

(27.) Clark SJ, Harrison J, Paul CL, Frommer M: High sensitivity mapping of methylated cytosines; Nucleic Acids Res 22:2990; 1994.

(28.) Clark SJ, Statham A, Stirzaker C, Molloy PL, Frommer M: DNA methylation: Bisulphite modification and analysis; Nat Protoc 1:2353; 2006.

(29.) Cotton AM, Price EM, Jones MJ, Balaton BP, Kobor MS, Brown CJ: Landscape of DNA methylation on the X chromosome reflects CpG density, functional chromatin state and X-chromosome inactivation; Hum Mol Genet 24:1528; 2015.

(30.) Criscione SW, Zhang Y, Thompson W, Sedivy JM, Neretti N: Transcriptional landscape of repetitive elements in normal and cancer human cells; BMC Genomics 15:583; 2014. 31. Daxinger L, Whitelaw E: Understanding transgenerational epigenetic inheritance via the gametes in mammals; Nat Rev Genet 13:153; 2012.

(32.) De Tobel J, Hillewig E, Verstraete K: Forensic age estimation based on magnetic resonance imaging of third molars: converting 2D staging into 3D staging; Ann Hum Biol 44:121; 2017.

(33.) Eckhardt F, Lewin J, Cortese R, Rakyan V, Attwood J, Burger M, Burton J, Cox T, Davies R, Down T, et al.: DNA methylation profiling of human chromosomes 6, 20 and 22; Nat Genet 38:1378; 2006.

(34.) Eduardoff M, Gross TE, Santos C, De La Puente M, Ballard D, Strobl C, Borsting C, Morling N, Fusco L, Hussing C, et al.: Inter-laboratory evaluation of the EUROFORGEN Global ancestry-informative SNP panel by massively parallel sequencing using the Ion PGM[TM]; Forensic Sci Int Genet 23:178; 2016.

(35.) Eduardoff M, Santos C, De La Puente M, Gross TE, Fondevila M, Strobl C, Sobrino B, Ballard D, Schneider PM, Carracedo A, et al.: Inter-laboratory evaluation of SNP-based forensic identification by massively parallel sequencing using the Ion PGM[TM]; Forensic Sci Int Genet 17:110; 2015.

(36.) Ehrich M, Correll D, van den Boom D: Introduction to EpiTYPER for quantitative DNA methylation analysis using the MassARRAY[R] System; Sequenom, Inc. (San Diego, CA) Product Preview Note (Doc. no. 8876-007); 2005; https://www.garvan.org.au/research/capabilities/molecular-genetics/documents/introduction-to-dna-methylation-using-the-massarray-system.pdf (accessed June 7, 2017).

(37.) Eipel M, Mayer F, Arent T, Ferreira MRP, Birkhofer C, Gerstenmaier U, Costa IG, Ritz-Timme S, Wagner W: Epigenetic age predictions based on buccal swabs are more precise in combination with cell type-specific DNA methylation signatures; Aging (Albany, NY) 8:1034; 2016.

(38.) Fedoriw AM, Stein P, Svoboda P, Schultz RM, Bartolomei MS: Transgenic RNAi reveals essential function for CTCF in H19 gene imprinting; Science 303(5655):238; 2004.

(39.) Fernandez A, Assenov Y, Martin-Subero J, Balint B, Siebert R, Taniguchi H, Yamamoto H, Hidalgo M, Tan A, Galm O, et al.: A DNA methylation fingerprint of 1628 human samples; Genome Res 22:407; 2012.

(40.) Florath I, Butterbach K, Muller H, Bewerunge-hudler M, Brenner H: Cross-sectional and longitudinal changes in DNA methylation with age: An epigenome-wide analysis revealing over 60 novel age-associated CpG sites; Hum Mol Genet 23:1186; 2014.

(41.) Fondevila M, Borsting C, Phillips C, de la Puente MCE, Carracedo A, Morling N, Lareu M: Forensic SNP genotyping with SNaPshot: Technical considerations for the development and optimization of multiplexed SNP assays; Forensic Sci Rev 29:57; 2017.

(42.) Fraga MF, Ballestar E, Paz MF, Ropero S, Setien F, Ballestar ML, Heine-Suner D, Cigudosa JC, Urioste M, Benitez J, et al.: Epigenetic differences arise during the lifetime of monozygotic twins; Proc Natl Acad Sci U S A 102:10604; 2005.

(43.) Fraser HB, Lam LL, Neumann SM, Kobor MS: Population-specificity of human DNA methylation; Genome Biol 13:R8; 2012.

(44.) Freire-Aradas A, Phillips C, Mosquera-Miguel A, Giron-Santamaria L, Gomez-Tato A, Casares De Cal M, Alvarez-Dios J, Ansede-Bermejo J, Torres-Espanol M, Schneider PM, et al.: Development of a methylation marker set for forensic age estimation using analysis of public methylation data and the Agena Bioscience EpiTYPER system; Forensic Sci Int Genet 24:65; 2016.

(45.) Frommer M, McDonald LE, Millar DS, Collis CM, Watt F, Grigg GW, Molloy PL, Paul CL: A genomic sequencing protocol that yields a positive display of 5-methylcytosine residues in individual DNA strands; Proc Natl Acad Sci U S A 89:1827; 1992.

(46.) Garagnani P, Bacalini MG, Pirazzini C, Gori D, Giuliani C, Mari D, Di Blasio AM, Gentilini D, Vitale G, Collino S, et al.: Methylation of ELOVL2 gene as a new epigenetic marker of age; Aging Cell 11:1132; 2012.

(47.) Gervin K, Andreassen BK, Hjorthaug HS, Carlsen KCL, Carlsen K-H, Undlien DE, Lyle R, Munthe-Kaas MC: Intra-individual changes in DNA methylation not mediated by cell-type composition are correlated with aging during childhood; Clin Epigenetics 8; 2016.

(48.) Giuliani C, Cilli E, Bacalini MG, Pirazzini C, Sazzini M, Gruppioni G, Franceschi C, Garagnani P, Luiselli D: Inferring chronological age from DNA methylation patterns of human teeth; Am J Phys Anthropol 159:585; 2016.

(49.) Greer EL, Maures TJ, Ucar D, Hauswirth AG, Mancini E, Lim JP, Benayoun BA, Shi Y, Brunet A: Transgenerational epigenetic inheritance of longevity in Caenorhabditis elegans; Nature 479:365; 2011.

(50.) Haeusler M, Haas C, Losch S, Moghaddam N, Villa IM, Walsh S, Kayser M, Seiler R, Ruehli F, Janosa M, et al.: Multidisciplinary identification of the controversial Freedom Fighter Jorg Jenatsch, assassinated 1639 in Chur, Switzerland; PLoS One 11(12):e0168014; 2016.

(51.) Hagerty SL, Bidwell LC, Harlaar N, Hutchison KE: An exploratory association study of alcohol use disorder and DNA methylation; Alcohol Clin Exp Res 40:1633; 2016.

(52.) Hannum G, Guinney J, Zhao L, Zhang L, Hughes G, Sadda S, Klotzle B, Bibikova M, Fan JB, Gao Y, et al.: Genomewide methylation profiles reveal quantitative views of human aging rates; Mol Cell 49:359; 2013.

(53.) Harrell F: Regression Modelling Strategies; Springer-Verlag: New York, NY; 2001.

(54.) He Y, Ecker JR: Non-CG methylation in the human genome; Annu Rev Genomics Hum Genet 16:55; 2015.

(55.) Heyn H, Li N, Ferreira HHJ, Moran S, Pisano DG, Gomez A, Diez J: Distinct DNA methylomes of newborns and centenarians; Proc Natl Acad Sci U S A 109:10522; 2012.

(56.) Heyn H, Moran S, Hernando-herraez I, Res G, Sayols S, Gomez A, Sandoval J, Monk D, Hata K, Marques-bonet T, et al.: DNA methylation contributes to natural human variation DNA methylation contributes to natural human variation; Genome Res 23:1363; 2013.

(57.) Hofreiter M, Jaenicke V, Serre D, Haeseler A von, Paabo S: DNA sequences from multiple amplifications reveal artifacts induced by cytosine deamination in ancient DNA; Nucleic Acids Res 29:4793; 2001.

(58.) Horvath S: DNA methylation age of human tissues and cell types; Genome Biol 14:R115; 2013.

(59.) Horvath S, Garagnani P, Bacalini MG, Pirazzini C, Salvioli S, Gentilini D, Di Blasio AM, Giuliani C, Tung S, Vinters H V, et al.: Accelerated epigenetic aging in Down syndrome; Aging Cell 14:491; 2015.

(60.) Horvath S, Gurven M, Levine ME, Trumble BC, Kaplan H, Allayee H, Ritz BR, Chen B, Lu AT, Rickabaugh TM, et al.: An epigenetic clock analysis of race/ethnicity, sex, and coronary heart disease; Genome Biol 17:171; 2016.

(61.) Horvath S, Levine AJ: HIV-1 infection accelerates age according to the epigenetic clock; J Infect Dis 212:1563; 2015.

(62.) Horvath S, Pirazzini C, Bacalini MG, Gentilini D, Di Blasio AM, Delledonne M, Mari D, Arosio B, Monti D, Passarino G, et al.: Decreased epigenetic age of PBMCs from Italian semi-supercentenarians and their offspring; Aging (Albany, NY) 7:1159; 2015.

(63.) Huang Y, Pastor WA, Shen Y, Tahiliani M, Liu DR, Rao A: The behaviour of 5-hydroxymethylcytosine in bisulfite sequencing; PLoS One 5(1):e8888; 2010.

(64.) Huang Y, Yan J, Hou J, Fu X, Li L, Hou Y: Developing a DNA methylation assay for human age prediction in blood and bloodstain; Forensic Sci Int Genet 17:129; 2015.

(65.) Illumina: MethylationEPIC Beadchip; https://www.illumina.com/products/by-type/microarray-kits/infinium-methylation-epic.html (Accessed May 25, 2017).

(66.) Illumina: MiSeq.; https://www.illumina.com/technology/next-generation-sequencing/sequencing-technology.html (Accessed May 25, 2017).

(67.) Jaenisch R, Bird A: Epigenetic regulation of gene expression: how the genome integrates intrinsic and environmental signals; Nat Genet 33 Suppl:245; 2003.

(68.) Jaffe AE, Irizarry RA: Accounting for cellular heterogeneity is critical in epigenome-wide association studies; Genome Biol 15:R31; 2014.

(69.) Jakobsson A, Westerberg R, Jacobsson A: Fatty acid elongases in mammals: Their regulation and roles in metabolism; Prog Lipid Res 45:237; 2006.

(70.) Jirtle RL, Skinner MK: Environmental epigenomics and disease susceptibility; Nat Rev Genet 8:253; 2007.

(71.) Johannessen M, Moler S, Hansen T, Moens U, Van Ghelue M: The multifunctional roles of the four-and-a-half-LIM only protein FHL2; Cell Mol Life Sci 63:268; 2006.

(72.) Johansson [Angstrom], Enroth S, Gyllensten U: Continuous aging of the human DNA methylome throughout the human Lifespan; PLoS One 8(6):e67378; 2013.

(73.) Jones MJ, Goodman SJ, Kobor MS: DNA methylation and healthy human aging; Aging Cell 14:924; 2015.

(74.) Jones PA: Functions of DNA methylation: Islands, start sites, gene bodies and beyond; Nat Rev Genet 13:484; 2012.

(75.) Jung M, Pfeifer GP: Aging and DNA methylation; BMC Biol 13:7; 2015.

(76.) Kaiser J: The epigenetics heretic; Science 343(6169):361; 2014.

(77.) Kananen L, Marttila S, Nevalainen T, Kummola L, Junttila I, Mononen N, Kahonen M, Raitakari OT, Hervonen A, Jylha M, et al.: The trajectory of the blood DNA methylome ageing rate is largely set before adulthood: Evidence from two longitudinal studies; Age (Omaha, NE) 38:65; 2016.

(78.) Kayser M: Forensic DNA phenotyping: Predicting human appearance from crime scene material for investigative purposes; Forensic Sci Int Genet 18:33; 2015.

(79.) Koch CM, Wagner W: Epigenetic aging signature to determine age in different tissues; Aging (Albany, NY) 3:1018; 2011.

(80.) Kriaucionis S, Heintz N: The nuclear DNA base 5-hydroxymethylcytosine is present in Purkinje neurons and the brain; Science 324(5929):929; 2009.

(81.) Kudlow BA, Kennedy BK, Monnat RJ: Werner and Hutchinson-Gilford progeria syndromes: mechanistic basis of human progeroid diseases; Nat Rev Mol Cell Biol 8:394; 2007.

(82.) Laird PW: Principles and challenges of genomewide DNA methylation analysis; Nat Rev Genet 11:191; 2010.

(83.) Lau QC, Raja E, Salto-Tellez M, Liu Q, Ito K, Inoue M, Putti TC, Loh M, Ko TK, Huang C, et al.: RUNX3 is frequently inactivated by dual mechanisms of protein mislocalization and promoter hypermethylation in breast cancer; Cancer Res 66:6512; 2006.

(84.) Lee HY, Jung SE, Oh YN, Choi A, Yang WI, Shin KJ: Epigenetic age signatures in the forensically relevant body fluid of semen: A preliminary study; Forensic Sci Int Genet 19:28; 2015.

(85.) Lee JT: Epigenetic regulation by long noncoding RNAs; Science 338(6113):1435; 2012.

(86.) Lehmann U, Tost J: Pyrosequencing, Methods and Protocols; Humana Press: Clifton, NJ; 2015.

(87.) Lewis ME, Flavel A: Age assessment of child skeletal remains in forensic contexts; In Schmitt A, Cunha E, Pinheiro J (Eds): Forensic Anthropology and Medicine: Complementary Sciences from Recovery to Cause of Death; Humana Press: Totiwa, NJ; 2006.

(88.) Li Q, Weidner CI, Costa IG, Marioni RE, Ferreira MR, Deary IJ, Wagner W: DNA methylation levels at individual age-associated CpG sites can be indicative for life expectancy; Aging (Albany, NY) 8:394; 2016.

(89.) Li Y, Tollefsbol TO: Age-related epigenetic drift and phenotypic plasticity loss: Implications in prevention of age-related human diseases; Epigenomics 8:1637; 2016.

(90.) Lin Q, Wagner W, Zenke M: Analysis of genome-wide DNA methylation profiles by BeadChip technology; Methods Mol Biol 1049:21; 2013.

(91.) Lister R, Mukamel EA, Nery JR, Urich M, Puddifoot CA, Nicholas D, Lucero J, Huang Y, Dwork AJ, Schultz MD, et al.: Global epigenomic reconfiguration during mammalian brain development; Science 341(6146):629; 2013.

(92.) Lister R, Pelizzola M, Dowen RH, Hawkins RD, Hon G, Tonti-Filippini J, Nery JR, Lee L, Ye Z, Ngo Q-M, et al.: Human DNA methylomes at base resolution show widespread epigenomic differences; Nature 462:315; 2009.

(93.) Liu Q, Jin J, Ying J, Sun M, Cui Y, Zhang L, Xu B, Fan Y, Zhang Q: Frequent epigenetic suppression of tumor suppressor gene glutathione peroxidase 3 by promoter hypermethylation and its clinical implication in clear cell renal cell carcinoma; Int J Mol Sci 16:10636; 2015.

(94.) Loginov VI, Dmitriev AA, Senchenko VN, Pronina IV, Khodyrev DS, Kudryavtseva AV, Krasnov GS, Gerashchenko GV, Chashchina LI, Kazubskaya TP, et al.: Tumor suppressor function of the SEMA3B gene in human lung and renal cancers; PLoS One 10(5):e0123369; 2015.

(95.) Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G: The hallmarks of aging; Cell 153:1194; 2013.

(96.) Lucas VS, McDonald F, Andiappan M, Roberts G: Dental age estimation--Root Pulp Visibility (RPV) patterns: A reliable Mandibular Maturity Marker at the 18 year threshold; Forensic Sci Int 270:98; 2017.

(97.) Man CH, Fung TK, Wan H, Cher CY, Fan A, Ng N, Ho C, Wan TSK, Tanaka T, Wai C, et al.: Suppression of SOX7 by DNA methylation and its tumor suppressor function in acute myeloid leukemia; Blood 125:3928; 2015.

(98.) Marioni RE, Shah S, McRae AF, Chen BH, Colicino E, Harris SE, Gibson J, Henders AK, Redmond P, Cox SR, et al.: DNA methylation age of blood predicts all-cause mortality in later life; Genome Biol 16:25; 2015.

(99.) Marquez-Grant N: An overview of age estimation in forensic anthropology: Perspectives and practical considerations; Ann Hum Biol 42:308; 2015.

(100.) Martin GM: Epigenetic drift in aging identical twins; Proc Natl Acad Sci U S A 102:10413; 2005.

(101.) Marttila S, Kananen L, Hayrynen S, Jylhava J, Nevalainen T, Hervonen A, Jylha M, Nykter M, Hurme M: Ageing-associated changes in the human DNA methylome: genomic locations and effects on gene expression; BMC Genomics 16:1; 2015.

(102.) Matarese F, Carrillo-de Santa Pau E, Stunnenberg HG: 5-Hydroxymethylcytosine: A new kid on the epigenetic block? Mol Syst Biol 7:562; 2011.

(103.) Mcclay JL, Aberg KA, Clark SL, Nerella S, Kumar G, Xie LY, Hudson AD, Harada A, Hultman CM, Magnusson PKE, et al.: A methylome-wide study of aging using massively parallel sequencing of the methyl-CpG-enriched genomic fraction from blood in over 700 subjects; Hum Mol Genet 23:1175; 2014.

(104.) Mehta B, Daniel R, Phillips C, Doyle S, Elvidge G, McNevin D: Massively parallel sequencing of customised forensically informative SNP panels on the MiSeq; Electrophoresis 37:2832; 2016.

(105.) Mehta B, Daniel R, Phillips C, McNevin D: Forensically relevant SNaPshot assays for human DNA SNP analysis: a review; Int J Legal Med 131:21; 2017.

(106.) Meissner C, Ritz-Timme S: Molecular pathology and age estimation; Forensic Sci Int 203:34; 2010.

(107.) Meissner C, Von Wurmb N, Schimansky B, Oehmichen M: Estimation of age at death based on quantitation of the 4977-bp deletion of human mitochondrial DNA in skeletal muscle; Forensic Sci Int 105:115; 1999.

(108.) Moen EL, Zhang X, Mu W, Delaney SM, Wing C, Mcquade J, Myers J, Godley LA, Dolan ME, Zhang W: Genomewide variation of cytosine modifications between European and African populations and the implications for complex traits; Genetics 194:987; 2013.

(109.) Moran S, Arribas C, Esteller M: Validation of a DNA methylation microarray for 850,000 CpG sites of the human genome enriched in enhancer sequences; Epigenomics 8:389; 2016.

(110.) Morgan HD, Sutherland HG, Martin DI, Whitelaw E: Epigenetic inheritance at the agouti locus in the mouse; Nat Genet 23:314; 1999.

(111.) Nyren P: Enzymatic method for continuous monitoring of DNA polymerase activity; Anal Biochem 167:235; 1987.

(112.) Ohtani S, Yamamoto T: Age estimation by amino acid racemization in human teeth; J Forensic Sci 55:1630; 2010.

(113.) Okano M, Bell DW, Haber DA, Li E: DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development; Cell 99:247; 1999.

(114.) Pabinger S, Ernst K, Pulverer W, Kallmeyer R, Valdes AM, Metrustry S, Katic D, Nuzzo A, Kriegner A, Vierlinger K, et al.: Analysis and visualization tool for targeted amplicon bisulfite sequencing on ion torrent sequencers; PLoS One 11(7):e0160227; 2016.

(115.) Pal S, Tyler JK: Epigenetics and aging; Sci Adv 2(7): e1600584; 2016.

(116.) Park JL, Kim JH, Seo E, Bae DH, Kim SY, Lee HC, Woo KM, Kim YS: Identification and evaluation of age-correlated DNA methylation markers for forensic use; Forensic Sci Int Genet 23:64; 2016.

(117.) Pedersen JS, Valen E, Velazquez AMV, Pedersen JS, Valen E, Velazquez AMV, Parker BJ, Rasmussen M, Lindgreen S, Lilje B, et al.: Genome-wide nucleosome map and cytosine methylation levels of an ancient human genome; Genome Res 24:454; 2014.

(118.) Pembrey ME, Bygren LO, Kaati G, Edvinsson S, Northstone K, Sjostrom M, Golding J, Alspac T, Team S: Sex-specific, male-line transgenerational responses in humans; Eur J Hum Genet 14:159; 2006.

(119.) Perna L, Zhang Y, Mons U, Holleczek B, Saum K-U, Brenner H: Epigenetic age acceleration predicts cancer, cardiovascular, and all-cause mortality in a German case cohort; Clin Epigenetics 8:64; 2016.

(120.) Phillips C: Forensic genetic analysis of bio-geographical ancestry; Forensic Sci Int Genet 18:49; 2015.

(121.) Pidsley R, Zotenko E, Peters TJ, Lawrence MG, Risbridger GP, Molloy P, Van Djik S, Muhlhausler B, Stirzaker C, Clark SJ, et al.: Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling; Genome Biol 17:208; 2016.

(122.) Pilin A, Pudil F, Bencko V: Changes in colour of different human tissues as a marker of age; Int J Legal Med 121:158; 2007.

(123.) Rabinowicz PD, Palmer LE, May BP, Hemann MT, Lowe SW, McCombie WR, Martienssen RA: Genes and transposons are differentially methylated in plants, but not in mammals; Genome Res 13:2658; 2003.

(124.) Rakyan VK, Down TA, Maslau S, Andrew T, Yang TP, Beyan H, Whittaker P, McCann OT, Finer S, Valdes AM, et al.: Human aging-associated DNA hypermethylation occurs preferentially at bivalent chromatin domains; Genome Res 20:434; 2010.

(125.) Reik W, Dean W, Walter J: Epigenetic reprogramming in mammalian development; Science 293(5532):1089; 2001.

(126.) Riggs A, Russo V, Martienssen R: Epigenetic mechanisms of gene regulation; Cold Spring Harbor Laboratory Press: Plainview, NY; 1996.

(127.) Saha A, Wittmeyer J, Cairns BR: Chromatin remodelling: The industrial revolution of DNA around histones; Nat Rev Mol Cell Biol 7:437; 2006.

(128.) Sakaguchi R, Okamura E, Matsuzaki H, Fukamizu A, Tanimoto K: Sox-Oct motifs contribute to maintenance of the unmethylated H19 ICR in YAC transgenic mice; Hum Mol Genet 22:4627; 2013.

(129.) Schmeling A, Geserick G, Reisinger W, Olze A: Age estimation; Forensic Sci Int 165:178; 2007.

(130.) Schmeling A, Dettmeyer R, Rudolf E, Vieth V, Geserick G: Forensic age estimation; Dtsch Arzteblatt Int 113:44; 2016.

(131.) Schubeler D: Function and information content of DNA methylation; Nature 517:321; 2015.

(132.) Simpkin AJ, Howe LD, Tilling K, Gaunt TR, Lyttleton O, McArdle WL, Ring SM, Horvath S, Smith GD, Relton CL: The epigenetic clock and physical development during childhood and adolescence: Longitudinal analysis from a UK birth cohort; Int J Epidemiol (In press); 2017.

(133.) Skinner MK, Haque CGBM, Nilsson E, Bhandari R, McCarrey JR: Environmentally induced transgenerational epigenetic reprogramming of primordial germ cells and the subsequent germ line; PLoS One 8(7):e66318; 2013.

(134.) Smith ZD, Meissner A: DNA methylation: Roles in mammalian development; Nat Rev Genet 14:204; 2013.

(135.) Strahl BD, Allis CD: The language of covalent histone modifications; Nature 403:41; 2000.

(136.) Suchiman HED, Slieker RC, Kremer D, Slagboom PE, Heijmans BT, Tobi EW: Design, measurement and processing of region-specific DNA methylation assays: The mass spectrometry-based method EpiTYPER; Front Genet 6:1; 2015.

(137.) Suzuki MM, Bird A: DNA methylation landscapes: Provocative insights from epigenomics; Nat Rev Genet 9:465; 2008.

(138.) Tahiliani M, Koh KP, Shen Y, Pastor WA, Bandukwala H, Brudno Y, Agarwal S, Iyer LM, Liu DR, Aravind L, et al.: Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1; Science 324(5929):930; 2009.

(139.) Talens RP, Christensen K, Putter H, Willemsen G, Christiansen L, Kremer D, Suchiman HED, Slagboom PE, Boomsma DI, Heijmans BT: Epigenetic variation during the adult lifespan: Cross-sectional and longitudinal data on monozygotic twin pairs; Aging Cell 11:694; 2012.

(140.) Tan H, Qurashi A, Poidevin M, Nelson DL, Li H, Jin P: Retrotransposon activation contributes to fragile X premutation rCGG-mediated neurodegeneration; Hum Mol Genet 21:57; 2012.

(141.) Tan Q, Heijmans BT, Hjelmborg JvB, Soerensen M, Christensen K, Christiansen L: Epigenetic drift in the aging genome: A ten-year follow-up in an elderly twin cohort; Int J Epidemiol 45:1146; 2016.

(142.) ThermoFischer: Ion Torrent PGM; https://www.thermofisher.com/de/de/home/life-science/sequencing/next-generation-sequencing/ion-torrent-next-generation-sequencing-technology.html (Accessed May 25, 2017).

(143.) Tiffen J, Wilson S, Gallagher SJ, Hersey P, Filipp FV: Somatic copy number amplification and hyperactivating somatic mutations of EZH2 correlate with DNA methylation and drive epigenetic silencing of genes involved in tumor suppression and immune responses in melanoma; Neoplasia 18:121; 2016.

(144.) Tomizawa S, Kobayashi H, Watanabe T, Andrews S, Hata K, Kelsey G, Sasaki H: Dynamic stage-specific changes in imprinted differentially methylated regions during early mammalian development and prevalence of non-CpG methylation in oocytes; Development 138:811; 2011.

(145.) Tost J, Gut IG: DNA methylation analysis by pyrosequencing; Nat Protoc 2:2265; 2007.

(146.) Tsaprouni LG, Yang TP, Bell J, Dick KJ, Kanoni S, Nisbet J, Vinuela A, Grundberg E, Nelson CP, Meduri E, et al.: Cigarette smoking reduces DNA methylation levels at multiple genomic loci but the effect is partially reversible upon cessation; Epigenetics 9:1382; 2014.

(147.) Turelli P, Castro-diaz N, Marzetta F, Kapopoulou A, Duc J, Tieng V, Quenneville S, Trono D: Interplay of TRIM28 and DNA methylation in controlling human endogenous retroelements Interplay of TRIM28 and DNA methylation in controlling human endogenous retroelements; Genome Res 24:1260; 2014.

(148.) Turner BM: Defining an epigenetic code; Nat Cell Biol 9:2; 2007.

(149.) Urdinguio RG, Torro MI, Bayon GF, Alvarez-Pitti J, Fernandez AF, Redon P, Fraga MF, Lurbe E: Longitudinal study of DNA methylation during the first 5 years of life; J Transl Med 14:160; 2016.

(150.) Vaca-Paniagua F, Oliver J, Nogueira da Costa A, Merle P, McKay J, Herceg Z, Holmila R: Targeted deep DNA methylation analysis of circulating cell-free DNA in plasma using massively parallel semiconductor sequencing; Epigenomics 7:353; 2015.

(151.) van Dongen J, Nivard MG, Willemsen G, Hottenga J-J, Helmer Q, Dolan CV, Ehli EA, Davies GE, van Iterson M, Breeze CE, et al.: Genetic and environmental influences interact with age and sex in shaping the human methylome; Nat Commun 7:11115; 2016.

(152.) Vidaki A, Ballard D, Aliferi A, Miller TH, Barron LP, Syndercombe Court D: DNA methylation-based forensic age prediction using artificial neural networks and next generation sequencing; Forensic Sci Int Genet 28:225; 2017.

(153.) Vidaki A, Daniel B, Syndercombe Court D: Forensic DNA methylation profiling--Potential opportunities and challenges; Forensic Sci Int Genet 7:499; 2013.

(154.) Walker RF, Liu JS, Peters BA, Ritz BR, Wu T, Ophoff RA, Horvath S: Epigenetic age analysis of children who seem to evade aging; Aging (Albany, NY) 7:334; 2015.

(155.) Weidner CI, Lin Q, Koch CM, Eisele L, Beier F, Ziegler P, Bauerschlag DO, Jockel K-H, Erbel R, Muhleisen TW, et al.: Aging of blood can be tracked by DNA methylation changes at just three CpG sites; Genome Biol 15:R24; 2014.

(156.) Wright K, Mundorff A, Chaseling J, Forrest A, Maguire C, Crane DI: A new disaster victim identification management strategy targeting "near identification-threshold" cases: Experiences from the Boxing Day tsunami; Forensic Sci Int 250:91; 2015.

(157.) Xu C-J, Bonder MJ, Soderhall C, Bustamante M, Baiz N, Gehring U, Jankipersadsing SA, van der Vlies P, van Diemen CC, van Rijkom B, et al.: The emerging landscape of dynamic DNA methylation in early childhood; BMC Genomics 18:25; 2017.

(158.) Xu C, Qu H, Wang G, Xie B, Shi Y, Yang Y, Zhao Z, Hu L, Fang X, Yan J, et al.: A novel strategy for forensic age prediction by DNA methylation and support vector regression model; Sci Rep 5:17788; 2015.

(159.) Yang Y, Hu JF, Ulaner GA, Li T, Yao X, Vu TH, Hoffman AR: Epigenetic regulation of Igf2/H19 imprinting at CTCF insulator binding sites; J Cell Biochem 90:1038; 2003.

(160.) Yu M, Hon GC, Szulwach KE, Song C-X, Jin P, Ren B, He C: Tet-assisted bisulfite sequencing of 5-hydroxymethylcytosine; Nat Protoc 7:2159; 2012.

(161.) Zampieri M, Ciccarone F, Calabrese R, Franceschi C, Burkle A, Caiafa P: Reconfiguration of DNA methylation in aging; Mech Ageing Dev 151:60; 2015.

(162.) Zbiec-Piekarska R, Spolnicka M, Kupiec T, Makowska Z, Spas A, Parys-Proszek A, Kucharczyk K, Ploski R, Branicki W: Examination of DNA methylation status of the ELOVL2 marker may be useful for human age prediction in forensic science; Forensic Sci Int Genet 14:161; 2015.

(163.) Zbiec-Piekarska R, Spolnicka M, Kupiec T, Parys-Proszek A, Makowska Z, Paleczka A, Kucharczyk K, Ploski R, Branicki W: Development of a forensically useful age prediction method based on DNA methylation analysis; Forensic Sci Int Genet 17:173; 2015.

(164.) Ziller MJ, Gu H, Muller F, Donaghey J, Tsai LT-Y, Kohlbacher O, De Jager PL, Rosen ED, Bennett DA, Bernstein BE, et al.: Charting a dynamic DNA methylation landscape of the human genome; Nature 500:477; 2013.

(165.) Ziller MJ, Stamenova EK, Gu H, Gnirke A, Meissner A: Targeted bisulfite sequencing of the dynamic DNA methylome; Epigenetics Chromatin 9:55; 2016.

(166.) Zubakov D, Liu F, Kokmeijer I, Choi Y, van Meurs JBJ, van IJcken WFJ, Uitterlinden AG, Hofman A, Broer L, van Duijn CM, et al.: Human age estimation from blood using mRNA, DNA methylation, DNA rearrangement, and telomere length; Forensic Sci Int Genet 24:33; 2016.

(167.) Zubakov D, Liu F, Van Zelm MC, Vermeulen J, Oostra BA, Van Duijn CM, Driessen GJ, Van Dongen JJM, Kayser M, Langerak AW: Estimating human age from T-cell DNA rearrangements; Curr Biol 20:R970; 2010.

A. Freire-Aradas (*), C. Phillips, M. V. Lareu

Forensic Genetics Unit, Institute of Forensic Sciences University of Santiago de Compostela Santiago de Compostela, Galicia Spain

(*) Corresponding author: Dr. Ana Freire-Aradas, Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Santiago de Compostela, Spain; + 34 881 812 206 (voice); ana.freire3@hotmail.com.

ABOUT THE AUTHORS

A. Freire-Aradas; C. Phillips; M. V. Lareu

Ana Freire-Aradas obtained her B.Sc. degree in pharmacy from the University of Santiago de Compostela (Spain) in 2006. In the same year she started her scientific research in the Forensic Genetics Unit, Institute of Forensic Sciences at the same university; obtaining her M.Sc. in molecular medicine in 2008 and her Ph.D. degree in 2013. After completing her Ph.D., Dr. Freire-Aradas continued her research in the same institution. Since 2015 she received a funding grant awarded by the Xunta de Galicia, Spain, as part of the Plan Galego de Investigacion, Innovacion e Crecemento 2011-2015; for supporting her postdoctoral research in both the Institute of Legal Medicine, University of Cologne (Germany), as host institution and the Institute of Forensic Sciences, University of Santiago de Compostela (Spain), as return institution. Research interests include SNP analysis for inference of biogeographic ancestry and externally visible characteristics; study of epigenetic markers such as DNA methylation with forensic applications, e.g., age estimation; evaluation of degraded DNA; and bioinformatic tools for assessment of DNA-based prediction models.

Christopher Phillips studied genetics at Birmingham University (Birmingham, UK) between 1974 and 1977 and in 1978 obtained his M.Sc. degree in applied genetics at the same institute. Mr. Phillips is currently a researcher in the Forensic Genetics Unit of the University of Santiago de Compostela. Mr. Phillips started his forensic genetics career in 1979 at the Biochemistry Division of the Metropolitan Police Forensic Science Laboratory (London, UK). He then moved to the Forensic Haematology Department, Barts Health NHS Trust (London, UK), and the London School of Medicine and Dentistry (London, UK) and worked there until 2001. Since 2001 he has been a full-time researcher in the Forensic Genetics Unit of the University of Santiago de Compostela. Mr. Phillips's research interests include SNP analysis applied to medical, population, and forensic genetics, the development of novel forensic polymorphisms, and building open-access online genomics search tools for the genetics and forensic communities.

Maria Victoria Lareu received her M.D. degree from the University of Santiago de Compostela in 1984. Dr. Lareu is professor of legal medicine at the University of Santiago de Compostela and the director of the Institute of Forensic Sciences at Santiago, directing the forensic science services provided by the Forensic Genetics Unit. Dr. Lareu is the principal investigator of many research projects undertaken in the department and directed the Ph.D. studies of the coauthors of this article: Ana Freire-Aradas and Christopher Phillips.
Table 1. Summary of relevant studies that explored the variation of DNA
methylation with chronological age at genomewide coverage

Year  Data access code           Tissue       n (a)  Age range (b)

2010  GSE20236, GSE20242         Blood          93   49-75
2011  GSE28746                   Saliva         68   21-55
2012  GSE27097                   Blood         398    3-17
2012  GSE58045                   Blood         172   32-80
2012  GSE31263, GSE30870,        Blood           3    0, 26 & 103
      GSE33233
2013  GSE40279                   Blood         656   19-101
2013  Additional file 1 in [58]  Multitissue  7844    0-100
2013  GSE87571                   Blood         421   14-94
2014  dbGaP                      Blood         718   25-92
2014  -- (d)                     Blood         965   50-75
2015  EGAD00010000887            Blood        2603   17-79
2016  GSE73115                   Blood          86   73-82

Year  Technique    Ref.

2010  HM27         [124]
2011  HM27          [17]
2012  HM27           [3]
2012  HM27          [11]
2012  WGBS          [55]

2013  HM450         [52]
2013  HM27/HM450    [58]
2013  HM450         [72]
2014  MBD-seq (c)  [103]
2014  HM450         [40]
2015  HM450        [151]
2016  HM450        [141]

(a) n = Sample size.
(b) Age in year.
(c) MBD-seq = Methyl-CpG binding domain-based capture and sequencing.
(d) No information.

Table 2. Summary of the current forensic age-prediction models based on
DNA methylation

              Age                                  CpG
Year  Tissue  range (a)/n (b))  Techniquec         coverage

2014  Blood   20-75/82          Pyrosequencing     3 CpGs
                                (500 ng gDNA)

2015  Blood   2-75/303          Pyrosequencing     2 CpGs
                                (2 [micro]g gDNA)
2015  Blood   2-75/300          Pyrosequencing     5 CpGs
                                (2 [micro]g gDNA)



2015  Blood   0-91/206          Pyrosequencing     4 CpGs
                                (200 ng gDNA)


2015  Teeth   19-70/29          Pyrosequencing     7 CpGs
                                (200 ng gDNA)





2015  Semen   20-73/31          SNaPshot           3 CpGs
                                (200 ng gDNA)

2015  Blood   20-80/49          EpiTYPER           6 CpGs
                                (1 [micro]g gDNA)




2016  Blood   11-90/535         Pyrosequencing     3 CpGs
                                (500 ng gDNA)

2016  Blood   18-104/725        EpiTYPER           7 CpGs
                                (300 ng gDNA)





2016  Buccal  1-85/55           Pyrosequencing     3 CpGs
      swab                      (500 ng gDNA)

2016  Buccal  1-85/55           Pyrosequencing     1 CpGs
      swab                      (500 ng gDNA)
2016  Blood   4-82/216          EpiTYPER           8 CpGs
                                (500 ng gDNA)






2016  Teeth   17-77/22          EpiTYPER           5-13
                                (200 ng gDNA)      CpGs (f)

2017  Blood   11-76/46          MiSeq              16 CpGs
                                (500 ng gDNA)


                                            somal position
Year  Gene_CpG_code        CpG_ID           chr17:3476273

2014  ASPA                 cg02228185       chr17:44390360
      ITGA2B               cg25809905       chr19:18233105
      PDE4C                None             chr6:11044642
2015  ELOVL2_C5            None             chr6:11044634
      ELOVL2_C7            None             chr6:11044634
2015  ELOVL2_C7            None             chr1: 207823681
      C1orf132_C1          None             chr3: 160450199
      TRIM59_C7            None             chr7: 130734355
      KLF14_C1             cg14361627       chr2: 105399288
      FHL2_C2              None             chr17:3476273
2015  ASPA_C1              cg02228185       chr19:18233078
      PDE4C_C1             None             chr6:11044407
      ELOVL2_C6            none             chr1:236394382
      EDARADD_C1           cg09809672       chr19:18233105
2015  PDE4C_C4             None             chr6:11044655
      ELOVL2_C2            cg24724428       chr6:11044640
      ELOVL2_C6            None             chr6:11044634
      ELOVL2_C7            None             chr6:11044628
      ELOVL2_C8            None             chr6:11044625
      ELOVL2_C9            None             chr1:236394395
      EDARADD_C2           None             chr14:90817262
2015  TTC7B                cg06304190       chr7:35260617
      No gene associated   cg12837463       chr11:89589683
      NOX4                 cg06979108       chr1:154609711
2015  ADAR_X25             None             chr1:154609812
      ADAR_X28             None             chr17:44390412
      ITGA2B_X77           None             chr19:18233105
      PDE4C_X92            None             chr19:18233127-31-33
      PDE4C_X93            None             chr19:18233193
      PDE4C_X95            None             chr16:49491896
2016  ZNF423_C1            cg04208403       chr6:11044661
      ELOVL2_C1            cg21572772       chr18:68722183
      CCDC102B_C1          cg19283806       chr6:11044661
2016  ELOVL2_C9            cg21572722       chr17:3476273
      ASPA_C3              cg02228185       chr19:18233127/31/33
      PDE4C_C27.28.29      None             chr2:105399282
      FHL2_C3              cg06639320       chr18:68722183
      CCDC102B_C2          cg19283806       chr1:207823715
      C1orf132_C11         None             chr16:85395429
      chr16:85395429_C3    cg07082267       chr17:3476273
2016  ASPA                 cg02228185       chr17:44390360
      ITGA2B               cg25809905       chr19:18233105
      PDE4C                None             chr19:18233105
2016  PDE4C                None
                                            chr6:11044634-31-28
2016  ELOVL2_C15.16.17     None             chr6:11044590-87-85
      ELOVL2_C22.23.24     None             chr6:11044655
      ELOVL2_C10           cg24724428       chr1:167128477-80
      DUSP27_C7.8          None             chr6:11044647-44-42-40
      ELOVL2_C11.12.13.14  None             chr11:69662722
      ORAOV1_C6            None             chr6:11044661
      ELOVL2_C9            cg21572722       chr11:69662752-54
      ORAOV1_C7.8          None             CpG ID
2016  ELOVL2               CpG ID           Chromosomal
      FHL2                 Combination (f)  position (f)
      PENK                                  chr17:82274220
2017  CSNK1D               cg19761273       chr21:32413126
      C21orf63             cg27544190       chr15:44288775
      CASC4                cg03286783       chr11:57336157
      SSRP1                cg01511567       chr9:69035321
      FXN                  cg07158339       chr22:21014721
      P2RXL1               cg05442902       chr1:206507825
      RASSF5               cg24450312       chr21:38661968
      ERG                  cg17274064       chr19:6739181
      TRIP10               cg02085507       chr7:73434151
      FZD9                 cg20692569       chr7:130733488
      KLF14                cg04528819       chr15:96330802
      NR2F2                cg08370996       chr7:101165768
      VGF                  cg04084157       chr6:18122488
      NHLRC1               cg22736354       chr6:25652374
      SCGN                 cg06493994       chr19:4769641
      C19orf30             cg02479575

      model (d)  MAD (e)        Ref.

Year  MLRM       [+ or -]5.43   [155]

2014

      MLRM       [+ or -]5.03   [162]
2015
      MLRM       [+ or -]3.40   [163]
2015



      MQDRM      [+ or -]3.75    [10]
2015


      MQDRM      [+ or -]4.84    [10]
2015





      MLRM       [+ or -]4.20    [84]
2015

      SVRM       [+ or -]2.80   [158]
2015




      MLRM       [+ or -]3.16   [116]
2016

      MQTRM      [+ or -]3.07    [44]
2016





      MLRM       [+ or -]4.3     [37]
2016

      MLRM       [+ or -]5.2     [37]
2016
      MLRM       [+ or -]5.09   [167]
2016






      MLRM       [+ or -]1.20-   [48]
2016             7.07 (f)

      ANN        [+ or -]7.45   [152]
2017


(a) Age in year.
(b) n = Sample size.
(c) Input DNA for bisulfite conversion.
(d) MLRM = Multivariate linear regression model; MQDRM = multivariate
quadratic regression model; SVRG = support vector regression model;
MQTRM = multivariate quantile regression model; ANN = Artificial
neural network.
(e) MAD = Median absolute deviation. (f) Depending on which teeth
area is analyzed.

Table 3. Summary of the characteristics of candidate genes and CpG
sites correlated with chronological age selected using the criteria
described in text (Section II. B.)

                     GRCh38 chromo-  UCSC RefGene
Gene     CpGID       somal position  group         Aging pattern

ELOVL2   cg16323298  chr6:11044741   TSS1500       No correlation
         cg16867657  chr6:11044644   TSS1500       Hypermethylation
         cg21572722  chr6:11044661   TSS1500       Hypermethylation
         cg24724428  chr6:11044655   TSS1500       Hypermethylation
         cg25151806  chr6:11045137   TSS1500       No correlation
ASPA     cg02228185  chr17:3476273   NA            Hypomethylation
         cg12317815  chr17:3475989   NA            Hypomethylation
EDARADD  cg09809672  chr1:236394382  NA            Hypomethylation
         cg18964582  chr1:236393882  TSS1500       Hypomethylation
FHL2     cg06639320  chr2:105399282  NA            Hypermethylation
         cg22454769  chr2:105399310  NA            Hypermethylation
         cg24079702  chr2:105399314  NA            Hypermethylation
ITGA2B   cg00062245  chr17:44384892  Body          Hypermethylation
         cg25809905  chr17:44390360  TSS1500       Hypomethylation
PDE4C    cg17861230  chr19:18233091  Body          Hypermethylation
         cg20119148  chr19:18233385  5'UTR         Hypermethylation
COPYRIGHT 2017 Central Police University
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Freire-Aradas, A.; Phillips, C.; Lareu, M.V.
Publication:Forensic Science Review
Article Type:Report
Date:Jul 1, 2017
Words:18328
Previous Article:Disposition of Toxic Drugs and Chemicals in Man.
Next Article:Social and ethical aspects of forensic genetics: A critical review.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters