NOTE-1264: Transgenerational epigenetic inheritance: a critical perspective

Contribution

A short critical essay, ten years after Heard and Martienssen, by a DNA-methylation biologist. It reviews the mammalian evidence since 2014, adds new negative results (the Agouti diet effect, transposon screens, Dppa2), and is the only one of the record’s three sceptical readings to name the human trauma studies and say what is wrong with them.

Key insight

Germline reprogramming is not a barrier the environment occasionally fails to cross. It is a function: “rather than eagerly harnessing environmental inputs from the life experiences of their ancestors via epigenetic mechanisms, organisms strive to prevent contamination of a new generation with the accumulated epigenetic baggage of the previous one” (p. 3). Where marks persist in mammals, it is when the eraser is broken (Dppa2 mutants) or a mutation keeps re-creating them (Takahashi et al. 2023).

Assumptions

• A broad definition of epigenetics: “the structural adaptation of chromosomal regions so as to register, signal or perpetuate altered activity states” (Bird 2007; p. 1). Heritability is not part of it.
• Inter- and transgenerational are not separated (p. 1). The essay also cites intragenerational effects. The distinction is left to Fitz-James and Cavalli (2022) and Horsthemke (2018).
• Adaptation as a test. A real inheritance system should be expected to pass on useful responses; one that passes on only disease needs an explanation (p. 4).

Key results

• Environment and epigenome (p. 2). Toxins change DNA methylation, but the change may be a consequence of a programmed stress response, in which “transcription factors are the prime movers, with epigenetic marks behaving as their clients”. DNA-binding proteins block methylation, and TET enzymes demethylate together with transcription factors.
• Twins (p. 2). Monozygotic twins acquire de novo mutations as they develop (Jonsson et al. 2021), so their differences are not evidence of epigenetic causes. Development can also vary stochastically: Trim28 haploinsufficient mice are either obese or not, with “identical” genotypes.
• Darwinian cases (pp. 2–3). An RNA-based system gives C. elegans offspring resistance to local pathogens; Bird compares colostrum in humans. These are evolved responses. Fission yeast’s toxin-induced silencing is lost quickly once the toxin is removed. Histone-mark inheritance in worms and yeast appears only in mutants that cannot remove the marks. In mice, inheritance of epigenetic marks is seen only when Dppa2 is mutated (Carlini et al. 2022).
• Mammals (p. 3).
  • Agouti: modest but convincing heritability of transposon methylation. The early report that diet changes it heritably (Wolff et al. 1998) failed to validate with larger samples; a screen for similar loci was “largely negative”.
  • “There is as yet no validated mechanism by which the external environment can directly communicate with the epigenome.” Sperm RNA “lacks robust experimental validation”. DNA methylation is largely stripped from gametes and early embryos.
  • The one stable multigenerational methylation in mice (two CpG islands, Takahashi et al. 2023) was triggered by a genetic deletion and could not be reproduced by adding methylation without it.
  • Effects in the progeny of protein-deprived male mice were undetectable in the second generation. Endocrine-disruptor effects in rats were not always replicated, in outbred animals where selection of resistant germ cells cannot be ruled out.
  • Robust evidence of an inter-generational paternal effect: sperm from males whose microbiome was ablated led to smaller placentas and some progeny deaths (Argaw-Denboba et al. 2024). But the authors found it “highly unlikely” to be transmitted via the sperm epigenome; Bird suggests the mother may withhold placental resources.
• Human claims (pp. 3–4).
  • Överkalix. Pembrey et al. 2006, “the Swedish famine”, “has achieved almost iconic status”. It and studies like it “ignored the potential effects of cultural transmission of behaviours within families”. The descendants of Leningrad siege survivors ate differently from controls, including “excessive red meat consumption”: “It is surely conceivable that cultural influence in the home played a role here.” A larger cohort (Vågerö et al. 2018) failed to reproduce the paternal grandfather effects on diabetes and heart disease. Such analyses risk multiple testing.
  • Direction. Most putative human cases transmit defects (obesity, diabetes, impaired memory). An anticipatory “thrifty” adaptation would backfire in long-lived mammals if the famine ends, which the Barker hypothesis itself predicts for those starved in utero.
  • Holocaust. Yehuda et al. 2016, reporting adverse consequences for survivors’ children, “made headline news, despite numerous technical weakness (e.g., a sample of only 32 people, limited controls and unclear methodology/interpretation)”. FKBP5 is named only in the reference’s title.
  • The Dutch Hunger Winter is not named; only the Barker hypothesis.
• Why the idea persists (p. 4). “Cultural resistance to the deterministic view” that human characteristics are hard-wired. Against it, genome sequencing keeps finding genetic causes: hundreds of mutations for intellectual disability; adults whose conditions were attributed to birth trauma or life experience who carry known causal mutations; polygenic risk scores.
• What would move him (p. 5). No criteria are listed. He grants “some evidence that severe environmental insults can impact the succeeding generation”, doubts it is adaptive, and calls the magnitude and persistence of disease effects “inconclusive”. He suggests acquired immunity, and the large datasets from the SARS-CoV-2 pandemic, as places to look. “Of course, the answer may be no.”

Claims

id	claim	strength	support
C1	Environmentally induced epigenetic changes are usually downstream of transcription-factor responses, not their cause	moderate: an argument from known mechanisms; no direct test cited	p. 2
C2	Organisms actively prevent the transmission of epigenetic marks; persistence appears when the erasing machinery is mutated or a mutation re-creates the mark	moderate: worm, yeast and mouse mutant studies, and one mouse deletion case	pp. 2–3
C3	No validated mechanism links the external environment to a heritable mammalian epigenome	strong as a statement of the literature’s state; Bird is a party to it	p. 3
C4	The Swedish (Överkalix) findings are unreliable: cultural transmission was ignored, a larger cohort did not replicate the key effect, and multiple testing is likely	moderate: three distinct objections, but the cultural one is argued from an analogy (Leningrad), and the replication’s positive findings are not mentioned	pp. 3–4
C5	The Holocaust FKBP5 study is too weak to support its public reception	moderate: sample size and controls named; no analysis of the paper	p. 4
C6	The persistence of the idea owes much to cultural resistance to genetic determinism	weak: a sociological conjecture, offered as “a potential driver”	p. 4

Concepts

• TEI, in Bird’s usage: any transmission of an environmentally induced epigenetic state to later generations, inter- or transgenerational.
• Darwinian TEI: an evolved, selected mechanism that happens to use epigenetic machinery (C. elegans antiviral RNA), as against a general channel for experience.
• epigenetic baggage: marks accumulated in a lifetime, which reprogramming is there to clear.
• marks as clients: epigenetic marks consolidate decisions made by transcription factors rather than make them.

Connections

• Heard and Martienssen (LIT-1745) is cited for directionless plant epialleles. Heard reviewed this article, with Anne Ferguson-Smith; Bird commented on Heard’s 2014 manuscript. Horsthemke (LIT-1762) is cited as “a recent critical review” and co-authored with Bird in 2023. The three readings are one school, not three independent judgements.
• Pembrey et al. (LIT-1764) and Yehuda et al. (LIT-1692) are the two human studies Bird disputes. Both are in the record, unread.
• Cultural transmission inside families (the Leningrad example) is the same route as the question’s parenting route, and the record’s meta-analyses of Holocaust offspring (LIT-1718, LIT-1770) bear on whether it carries harm.
• Inherited state in bacteria (THEORY-179). Bird’s fission-yeast case is the closest analogue: an induced state that persists while the stimulus lasts and dissipates soon after. THEORY-179’s E. coli state fades by about seven generations. Both are inherited physiology in organisms without a sequestered germline, and both are short.

Bearing on the record

• The question (QUESTION-297).
  • Germline route. Bird gives the record’s first reasons against its two named human germline cases. Överkalix: cultural confounding, non-replication in a larger cohort, multiple testing. The Holocaust study: 32 people, limited controls. Both are claims the question currently states without rebuttal.
  • A correction to the essay’s use (an inference). Yehuda et al. 2016 concerns survivors’ children. If, as the war’s dates suggest, they were conceived after the trauma, then by Heard and Martienssen’s rule they are F1 of an exposed parent, an intergenerational case. Bird files it under TEI because he does not separate the two. The question should cite it as an intergenerational claim.
  • Cultural transmission as the confound. Bird’s main objection to Överkalix is that families pass on eating habits. What the sceptic calls a confound is, for the question, one of the routes.
• The owner’s thesis (CLAIM-491). Indirect, and an inference. The weaker the germline route, the more of what is called intergenerational trauma must be carried socially: by care, habit and story. That raises the share left to the identity route without showing that the identity route carries it; Bird’s own example of cultural transmission (diet) is habit, not identification. No note on the claim is proposed.
• Proposed dated note on QUESTION-297 (below, section 3).
• No THEORY filed.
• No ML instruction.

Limitations

• The inter/transgenerational distinction is set aside, so the human examples mix F1 effects (Yehuda et al. 2016) with F2 ones (Överkalix).
• The replication is reported one-sidedly. Bird cites Vågerö et al. 2018 only for failing to reproduce paternal-grandfather effects on diabetes and heart disease. The title of that paper, in his reference list, is “Paternal grandfather’s access to food predicts all-cause and cancer mortality in grandsons”: it reports a paternal-grandfather effect on other outcomes.
• “The Swedish famine” describes Pembrey et al. 2006 loosely; the record’s LIT describes its exposure as grandparents’ food supply in mid-childhood, not famine as such. To check when LIT-1764 is read.
• The cultural objection is by analogy. The Leningrad diet study is about a different population; no cultural variable in the Swedish data is examined.
• Opinion is mixed with evidence. The sociological diagnosis (p. 4) and the case-report evidence of misdiagnosed genetic disease (a single case report, Langenfeld et al. 2021, for “a high proportion”) are rhetorical support.
• A blog post is cited for multiple testing, by URL in the text, not in the reference list.
• Typos. “are reported be transmissible” (p. 3); “numerous technical weakness” (p. 4); “indetectable” (p. 3).

Open questions

• Does the Holocaust FKBP5 study survive the objections Bird names? The record should read it (LIT-1692) before weighting either.
• Did Vågerö et al. 2018 replicate Överkalix or not? It seems to have confirmed one male-line effect and failed on another.
• If cultural transmission of behaviour explains family resemblance in diet, does the same route explain resemblance in fear and grief, and is it habit or identification?