An epigenetic clock review by committee

This 2019 worldwide review of epigenetic clocks was a semi-anonymous mishmash of opinions, facts, hypotheses, unwarranted extrapolations, and beliefs. The diversity of viewpoints among the 21 coauthors wasn’t evident.

1. Citations of the coauthors’ works seemed excessive, and they apologized for omissions. However, Challenge 5 was titled “Single-cell analysis of aging changes and disease” and Table 1 “Major biological and analytic issues with epigenetic DNA methylation clocks” had single-cell analysis as the Proposed solution to five of the Significant issues. Yet studies such as High-Resolution Single-Cell DNA Methylation Measurements Reveal Epigenetically Distinct Hematopoietic Stem Cell Subpopulations were unmentioned.

2. Some coauthors semi-anonymously expressed faith that using current flawed methodologies in the future – only more thoroughly, with newer equipment, etc. – would yield better results. If the 21 coauthors were asked their viewpoints of Proposed solutions to the top three Significant issues of epigenetic clocks, what would they emphasize when quoted?

3. Techniques were praised:

“Given the precision with which DNA methylation clock age can be estimated and evolving measures of biological, phenotype-, and disease-related age (e.g., PhenoAge, GrimAge)..”

Exactly why these techniques have at times produced inexplicable results wasn’t examined, though. Two examples:

  • In Reversal of aging and immunosenescent trends, the Levine PhenoAge methodology estimated that the 51-65 year old subjects’ biological ages at the beginning of the study averaged 17.5 years less than their chronological age. Comparing that to the Horvath average biological age of 3.95 years less raised the question: exactly why did PhenoAge show such a large difference?
  • The paper mentioned the GrimAge methodology findings about “smoking-related changes.” But it didn’t explain why the GrimAge methylation findings most closely associated with smoking history also accurately predicted future disease risk with non-smokers.

Eluding explanations for these types of findings didn’t help build confidence in the methodologies.

4. A more readable approach to review by committee would have coauthors answer discussion questions, as Reversing epigenetic T cell exhaustion did with 18 experts.

https://genomebiology.biomedcentral.com/articles/10.1186/s13059-019-1824-y “DNA methylation aging clocks: challenges and recommendations”

A review of fetal adverse events

This 2019 Australian review subject was fetal adversities:

“Adversity during the perinatal period is a significant risk factor for the development of neurodevelopmental disorders long after the causative event. Despite stemming from a variety of causes, perinatal compromise appears to have similar effects on the developing brain, thereby resulting in behavioural disorders of a similar nature.

These behavioural disorders occur in a sex‐dependent manner, with males affected more by externalizing behaviours such as attention deficit hyperactivity disorder (ADHD) and females by internalizing behaviours such as anxiety. The term ‘perinatal compromise’ serves as an umbrella term for intrauterine growth restriction, maternal immune activation, prenatal stress, early life stress, premature birth, placental dysfunction, and perinatal hypoxia.

The above conditions are associated with imbalanced excitatory-inhibitory pathways resulting from reduced GABAergic signalling. Methylation of the GAD1/GAD67 gene, which encodes the key glutamate‐to‐GABA synthesizing enzyme Glutamate Decarboxylase 1, resulting in increased levels of glutamate is one epigenetic mechanism that may account for a tendency towards excitation in disorders such as ADHD.

The posterior cerebellum’s role in higher executive functioning is becoming well established due to its connections with the prefrontal cortex, association cortices, and limbic system. It is now suggested that disruptions to cerebellar development, which can occur due to late gestation compromises such as preterm birth, can have a major impact on the region of the brain to which it projects.

Activation of the maternal hypothalamic-pituitary adrenal (HPA) axis and placental protection. Psychological stress is perceived by the maternal HPA axis, which stimulates cortisol release from the maternal adrenal gland.

High levels of maternal cortisol are normally prevented from reaching the fetus by the 11β-hydroxysteroid dehydrogenase 2 (HSD11B2) enzyme, which converts cortisol to the much less active cortisone. Under conditions of high maternal stress, this protective mechanism can be overwhelmed, with the gene encoding the enzyme becoming methylated, which reduces its expression allowing cortisol to cross the placenta and reach the fetus.”


The reviewers extrapolated many animal study findings to humans, although most of their own work was with guinea pigs. The “suggest” and “may” qualifiers were used often – 22 and 37 times, respectively. More frequent use of the “appears,” “hypothesize,” “propose,” and “possible” terms was justified.

As a result, many reviewed items such as the above graphic and caption should be viewed as hypothetical for humans rather than reflecting solid evidence from quality human studies.

The reviewers focused on the prenatal (before birth) period more than the perinatal (last trimester of pregnancy to one month after birth) period. There were fewer mentions of birth and early infancy adversities.

https://onlinelibrary.wiley.com/doi/abs/10.1111/jne.12814 “Perinatal compromise contributes to programming of GABAergic and Glutamatergic systems leading to long-term effects on offspring behaviour” (not freely available)

A GWAS meta-analysis of two epigenetic clocks

This 2019 UK human study conducted a meta-analysis of genome-wide association studies of two epigenetic clocks using 13,493 European-ancestry individuals aged between ten and 98 years:

“Horvath-EAA, described in previous publications as ‘intrinsic’ epigenetic age acceleration (IEAA), can be interpreted as a measure of cell-intrinsic ageing that exhibits preservation across multiple tissues, appears unrelated to lifestyle factors, and probably indicates a fundamental cell ageing process that is largely conserved across cell types.

In contrast, Hannum-EAA, referred to in previous studies as ‘extrinsic’ epigenetic age acceleration (EEAA), can be considered a biomarker of immune system ageing, explicitly incorporating aspects of immune system decline such as age-related changes in blood cell counts, correlating with lifestyle and health-span related characteristics, and thus yielding a stronger predictor of all-cause mortality.

The meta-analysis of Horvath-EAA identified ten independent associated SNPs [single nucleotide polymorphisms], doubling the number reported to date, and highlighted 21 genes involved in Horvath-based epigenetic ageing. Four of the ten Horvath-EAA-associated SNPs are mQTL [methylation quantitative trait loci] for CpGs used in the Horvath/Hannum epigenetic clocks. A possible interpretation of this is that the functional mechanism by which these SNPs influence the rate of biological ageing is via altering methylation levels.

Father’s age at death, a rough proxy for lifespan, was nominally significantly correlated with both EAA measures, and parents’ age at death was additionally correlated with Hannum-EAA. Aside from these, genetic correlations with age-related traits were surprisingly few: it is possible that this could reflect an overly conservative correction for the multiple tests carried out, or low statistical power, rather than a genuine lack of correlations.

Genetic correlation analysis should be restricted to GWAS with a heritability Z-score of 4 or more, on the grounds of interpretability and power, so the Horvath-based results particularly should be interpreted with caution.”


A non-apologetic way to explain the above graphic is that NONE of these 218 “health and behavioral traits” were any more associated with the studied genetic measurements than would be expected by chance!

Fervent believers in the GWAS methodology’s capability to exactly predict individual phenotypes eventually become victims of the scientific method. These GWAS researchers griped about “overly conservative correction, or low statistical power” and other predictable shortfalls, and ended a long limitations statement with:

“While we have identified a number of SNPs and genes significantly associated with EAA, including genes already known to be related to ageing, the analyses presented here fall short of providing a mechanistic explanation for how these variants and genes act to influence biological age.”

Outside of beliefs, it’s hard to understand why research money keeps pouring into the GWAS dead end. If these researchers and their employing institution and sponsors want to make a difference in human lives, they need to get busy in other areas.

These researchers were employed by the same institution that couldn’t be bothered to scrape together six more weeks of funds to study the transgenerational damaging effects of acetaminophen – an analgesic available to billions of people – in Epigenetics research that was designed to fall one step short of wonderful.

https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1008104 “A meta-analysis of genome-wide association studies of epigenetic age acceleration”

Organismal aging and cellular senescence

I’ll curate this 2019 German review through its figures:

“With the discovery of beneficial aspects of cellular senescence and evidence of senescence being not limited to replicative cellular states, a redefinition of our comprehension of aging and senescence appears scientifically overdue.

Figure 1. Current determinants and relevant open questions, marking the processes of aging and senescence as discussed in the text. Aspects represented in green are considered as broadly accepted or scientifically consolidated. Novel aspects that are yet unproven, or are under debate, are highlighted in red.

SASP = senescence-associated secretory phenotype. AASP = putative aging-associated secretory phenotype as suggested in the text.

Figure 2. Theories on the causality and purpose of aging. Graphically summarized are four contrasting concepts crystallized from current evidence addressing the inductive driving force of aging. Apart from a stochastic deleteriome, there are arguments for a pseudo-programmed, programmed or at least partially programmed nature of aging.

Figure 3. Comparative representation of the aging and senescence processes highlighting different levels of interaction and putative sites of interventions.

(1) As discussed in the text, causative mechanisms of aging are still not well understood, however, multiple factors including genetic, epigenetic and stress-related effects seem to have an orchestrated role in the progression of aging. Senescence on the other hand, is seen as a programmed response to different kinds of stressors, which proceed in defined stages. Whether, in analogy, aging also follows a defined program or sequential stages is not known.

(2) Senescence involves autocrine and paracrine factors, which are responsible for a ‘seno-infection’ or bystander effect in neighboring cells. There is currently no direct evidence for a similar factor composition propagating the aging process via a kind of ‘gero-infection’.

(3) Accumulation of senescent cells has been described as a hallmark of aging; however, whether they are a causative factor or a consequence of tissue and organismal aging is still unknown. As discussed in the text, it appears possible that aging and senescence mutually influence each other through positive feedback at this level, leading to accelerated tissue damage and aging.

(4,5) Clearance of senescent or aging cells might constitute putative targets for interventional approaches aimed to reduce or reverse the impact of aging and improve cell and tissue homeostasis by inducing a ‘rejuvenation’ process.

Figure 4. Pathological and beneficial functions of aging and senescence, according to current knowledge. In red are represented pathological consequences and in green beneficial functions of aging and senescence.

The impact of aging has mainly been described at the organismal level, since a complete cellular functional profile has not yet been established. Accordingly, whether beneficial consequences of the aging process exist at the cellular level is unclear.”


The reviewers’ position on Figure 2 was:

“In our view, recent evidence that senescence is based on an unterminated developmental growth program and the finding that the concept of post-mitotic senescence requires the activation of expansion, or ‘growth’ factors as a second hit, favor the assumption that aging underlies a grating of genetic determination similarly to what is summarized above under the pseudo-programmed causative approach.”

Their position on Figure 4’s beneficial effects of aging began with the sentence:

“If we assume that aging already starts before birth, it can be considered simply a developmental stage, required to complete the evolutionary program associated with species-intrinsic biological functions such as reproduction, survival, and selection.”

Cited studies included:

https://www.mdpi.com/2073-4409/8/11/1446 “Dissecting Aging and Senescence-Current Concepts and Open Lessons”

A strawman argument against epigenetic clocks

This 2019 review of epigenetic clocks by Washington cancer researchers repeatedly returned to an argument for randomness as a cause for aging and disease:

“A time-dependent stochastic event process, like epigenetic drift, could lead to cancer formation through the accumulation of random epigenetic alterations that, through chance, eventually alter epigenetic driver gene expression leading to a clone of cells destined to become cancer..

It is plausible that the stochastic process inherent in epigenetic drift can induce aberrant methylation events that accumulate in normal cells and eventually induce cancer formation.

Epigenetic drift relates to a biological process that changes the DNA methylome with age via stochastic gains or losses of DNA methylation. Epigenetic drift can be understood in terms of errors in DNA methylation maintenance during DNA-replication.

The phenomenon of (epi)genetic drift is generally associated with phenotypic neutrality.

For patients who develop cancer around age 80, the most likely initiation time for the founder adenoma cell is predicted to be very early in life, roughly between the ages 15 to 20 years. This unexpected and provocative finding suggests that the optimal age-range for prevention of colorectal cancer may be in adolescence and early adulthood (and ideally through lifelong) dietary and lifestyle interventions.”


The reviewers’ strawman arguments intentionally mischaracterized aspects of the epigenetic clock:

1. The epigenetic clock founder’s actual view on aging was in The epigenetic clock theory of aging:

“The proposed epigenetic clock theory of ageing views biological ageing as an unintended consequence of both developmental programmes and maintenance programmes, the molecular footprints of which give rise to DNAm age estimators.”

The reviewers omitted this intrinsic view of aging, which didn’t fit into their block labeled Extrinsic per the above graphic.

2. Another misrepresentation was:

“In contrast to epigenetic clocks, epigenetic drift refers to a stochastic process that involves both gains and losses of the methylation state of CpG dinucleotides over time.”

A reader of the original 2013 epigenetic clock study would understand that epigenetic clocks measure “both gains and losses of methylation” as in:

“The 193 positively and 160 negatively correlated CpGs get hypermethylated and hypomethylated with age, respectively.”

3. The reviewers omitted recent epigenetic clock significant developments. For example, there was no mention of the GrimAge study, although it was published before the review was submitted.

4. Epigenetic drift as the cause of aging and disease has abundant contrary evidence. The reviewers tossed in a little toward the end of their directed narrative:

“We found only a small number of drift-related CpG island-gene pairs for which drift correlated positively and significantly with gene expression.

The functional consequences of epigenetic drift need to be further elucidated.”

However, they didn’t acknowledge the elephant in the room: The epigenetic drift paradigm is generally inapplicable to humans because the vast majority of our cells don’t divide/proliferate!

https://cancerres.aacrjournals.org/content/early/2019/11/06/0008-5472.CAN-19-0924 “Epigenetic aging: more than just a clock when it comes to cancer” (not freely available)

Do genes or maternal environments shape fetal brains?

This 2019 Singapore human study used Diffusion Tensor Imaging on 5-to-17-day old infants to find:

“Our findings showed evidence for region-specific effects of genotype and GxE on individual differences in human fetal development of the hippocampus and amygdala. Gene x Environment models outcompeted models containing genotype or environment only, to best explain the majority of measures but some, especially of the amygdaloid microstructure, were best explained by genotype only.

Models including DNA methylation measured in the neonate umbilical cords outcompeted the Gene and Gene x Environment models for the majority of amygdaloid measures and minority of hippocampal measures. The fact that methylation models outcompeted gene x environment models in many instances is compatible with the idea that DNA methylation is a product of GxE.

A genome-wide association study of SNP [single nucleotide polymorphism] interactions with the prenatal environments (GxE) yielded genome wide significance for 13 gene x environment models. The majority (10) explained hippocampal measures in interaction with prenatal maternal mental health and SES [socioeconomic status]. The three genome-wide significant models predicting amygdaloid measures, explained right amygdala volume in interaction with maternal depression.

The transcription factor CUX1 was implicated in the genotypic variation interaction with prenatal maternal health to shape the amygdala. It was also a central node in the subnetworks formed by genes mapping to the CpGs in neonatal umbilical cord DNA methylation data associating with both amygdala and hippocampus structure and substructure.

Our results implicated the glucocorticoid receptor (NR3C1) in population variance of neonatal amygdala structure and microstructure.

Estrogen in the hippocampus affects learning, memory, neurogenesis, synapse density and plasticity. In the brain testosterone is commonly aromatized to estradiol and thus the estrogen receptor mediates not only the effects of estrogen, but also that of testosterone.”

https://onlinelibrary.wiley.com/doi/full/10.1111/gbb.12576 “Neonatal amygdalae and hippocampi are influenced by genotype and prenatal environment, and reflected in the neonatal DNA methylome” (not freely available)

Emotional responses and BDNF methylation

This 2019 German human study found:

“A critical role of BDNF [brain-derived neurotrophic factor] methylation in human amygdala response to negative emotional stimuli, whereby:

  • High BDNF methylation rates were for the first time shown to be associated with a high reactivity in the amygdala; and
  • High BDNF methylation and high amygdala reactivity were associated with low novelty seeking.

There was no interaction or main effect of the Val66Met polymorphism on amygdala reactivity.

Our data adds evidence to the hypothesis that epigenetic modifications of BDNF can result in an endophenotype associated with anxiety and mood disorders. However, since correlations do not prove causality:

  • A direct link between human BDNF mRNA/protein levels, methylation, amygdala reactivity and psychiatric disorders is still missing, demanding further research.
  • Determining the underlying directions of the relations between BDNF methylation, amygdala reactivity, and NS [novelty seeking] cannot be accomplished based on our data and must await further research.

The fact that our results mainly involve the right amygdala is in line with previous studies. Recent reviews suggest a general right hemisphere dominance for all kinds of emotions, and, more specifically, a critical role of the right amygdala in the early assessment of emotional stimuli.

The experimental fMRI paradigm utilized a face‐processing task (faces with anger or fear expressions), alternating with a sensorimotor control task. Harm avoidance, novelty seeking, and reward dependence were measured using the Tridimensional Personality Questionnaire.”

https://onlinelibrary.wiley.com/doi/full/10.1002/hbm.24825 “The role of BDNF methylation and Val 66 Met in amygdala reactivity during emotion processing”