Contribution
A review and manifesto for “transgenerational epidemiology”. It summarises the human studies of grandparental and paternal exposures, chief among them the authors’ own Överkalix series, and the animal experiments that bear on how such effects could be carried. Its aim is practical: that birth cohorts collect parents’ and grandparents’ exposure data, by life stage, and test each mode of transmission separately. Two of its three authors, Pembrey and Bygren, are authors of the 2006 paper it is read in place of.
Key insight
The timing of the ancestor’s exposure is the authors’ argument against culture. A grandfather’s food supply mattered for his grandsons when it fell in his “slow growth period”, the years before the prepubertal growth spurt, and not later in childhood. They take that as “compelling evidence for some other form of transmission between generations” than cultural or genetic inheritance alone (Conclusions from human observations to date). Their methodological rule follows: “assuming cultural inheritance automatically when results make genetic inheritance implausible (and vice versa) is no longer acceptable” (Extending epidemiology).
Assumptions
- A broad sense of “transgenerational” (Framing future research). It covers any measurable outcome in offspring of an exposure in the mother or father before conception, or in their parents or ancestors. The stricter usage is noted (ref. 18) and set aside. A fetus exposed in utero counts as exposed, so for a pregnant mother the outcome of interest is in her child’s children.
- The exposure, not a phenotype, is what passes. The exposed or intermediate generation need show no phenotype, and the outcome need not be harmful (Framing future research).
- Mechanism-agnostic. “Studies in the field should not be predicated upon a particular mode of transmission” (Framing future research).
- Social patterning can be “accounted for” in the right setting, which would separate cultural from biological transmission (Framing future research). How is not said.
Key results
- Origin of the word (Framing future research). “Transgenerational” was first used in this context in the 1980s for the transmission of Holocaust survivors’ trauma to their unexposed offspring (Nadler et al. 1985).
- Överkalix (Food supply, paternal line; Table 1). Historical records
of harvests and food supply in northern Sweden, linked to probands born
in set years. Five studies:
- Bygren et al. 2001: cohort 1905 (n = 94). The paternal grandfather’s food supply just before the prepubertal spurt was inversely associated with proband longevity. This defined the mid-childhood “slow growth period”.
- Kaati et al. 2002: three cohorts. Father’s poor and mother’s good mid-childhood food supply went with less cardiovascular death; paternal grandfather’s good supply with more diabetes death. The sensitive period was confirmed in two of the three. Each diabetic proband had a different paternal grandfather, a point made in the 2006 exchange with Senn (“Epigenetics or ephemeral genetics?”) cited in Table 1.
- Pembrey et al. 2006 (the original): cohorts 1895 (n = 107), 1905 (n = 99) and 1920 (n = 111). Grandsons’ mortality was linked only to the paternal grandfather’s mid-childhood food supply, granddaughters’ only to the paternal grandmother’s. Rate ratios against moderate supply: grandsons 1.67 (good) and 0.65 (poor); granddaughters 2.13 and 0.72. Sex stratification was suggested by early ALSPAC results. The mid-childhood period, not (pre)puberty, was confirmed.
- Kaati et al. 2007: the sex-specific pattern persisted with the grandchild’s early-life circumstances taken into account, and a father-to-son effect on longevity appeared.
- Bygren et al. 2014: a sharp change in the paternal grandmother’s food supply between 0 and 13 years went with cardiovascular death in granddaughters. Biological samples exist for none of the cohorts except a 1935 one (What is needed for future human studies?).
- Other famines (Food supply). A German discussion paper on the 1916–1918 famine (van den Berg and Pinger 2014): third-generation males had better mental health scores if the paternal grandfather was exposed (boys at 9–12, girls at 8–10). Dutch Hunger Winter studies of the grandchildren of women exposed in utero gave mixed birth-weight results and an association with childhood adiposity.
- ALSPAC smoking (Smoking, paternal line). Designed to test the Överkalix hypothesis. The earlier the father began smoking, the greater his sons’ BMI at 9, not his daughters’; strongest for a start before 11. To 17 years, those sons’ excess BMI, waist and fat mass grew with age, significant from 13 (about 5–10 kg more fat mass). Daughters showed less lean mass. Maternal-line results: sons of non-smoking mothers exposed in utero were 61 g heavier at birth; two other cohorts found 244 g and 45 g. Asthma findings conflict and need replication.
- Betel nut (Other paternal line exposures). Paternal betel-quid chewing, usually begun about 18, went with early metabolic syndrome in offspring who never chewed (Taiwan), following a mouse result.
- Their answer to the critics (Conclusions from human observations).
- Multiple testing: “clear prior hypotheses (see table 1) and coherence between different studies/populations”.
- “Research conclusions that extend the public health paradigm need a higher level of evidence than normal, and further replication is needed.”
- The associations are “most unlikely to be due to ‘cultural’ inheritance/social patterning or genetic inheritance in isolation”.
- Sex differences (Lessons from animal experiments). Rodent studies report sex-specific effects in either or both sexes, so the human sex differences are “not unexpected”. The Överkalix pattern fits X and Y transmission, but some mouse sex-specific effects do not.
- Routes other than the germline epigenome (Figure 1; Transmission between generations). Maternal routes: egg cytoplasm, the womb, breast milk, microbiota, and a “metabolic cascade” through the next uterus. Paternal routes need not pass through meiosis: Sertoli–spermatid junctions and seminal fluid. Effects beyond the second generation need germline involvement.
- Mechanisms (A role for epigenetics; ncRNAs). Marks are erased twice in mammals; imprinted genes escape one wave, and a few repeat-linked loci escape the other. Dias and Ressler’s odour study and a mouse prediabetes model are cited as plausible gametic mechanisms. Injected sperm RNA from stressed males reproduced the offspring phenotype, yet F3 mice inherited it from F2 fathers with normal sperm miRNA (Gapp et al. 2014). Genotype moderates transmission (MLH1 epimutations are genetically driven in some families).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | In Överkalix, a paternal grandparent’s mid-childhood food supply is associated, sex-specifically, with the grandchild’s mortality | weak to moderate: about 100 probands per cohort, historical records, one population; some confirmation across cohorts | Table 1; Food supply |
| C2 | The effect has an exposure-sensitive period before puberty, not in adolescence | weak to moderate: the same cohorts, plus ALSPAC smoking and one German discussion paper | Food supply; Smoking |
| C3 | The associations are unlikely to be due to cultural or genetic inheritance alone | weak: argued from the pattern of timing and sex, not from measured cultural variables; the one adjustment is for the grandchild’s early-life circumstances | Conclusions from human observations |
| C4 | Multiple testing is answered by prior hypotheses and coherence across studies | weak: the priors are stated in Table 1, but several were set after earlier Överkalix results | Conclusions from human observations; Table 1 |
| C5 | Transgenerational responses are plausibly a determinant of common disease and of its sex differences | a conjecture, offered as “highly plausible” | Sex differences |
| C6 | Cohorts should collect ancestral exposure data and test each mode of transmission independently | a recommendation | Framing; What is needed |
Concepts
- transgenerational (broad sense): an outcome in offspring of a parent’s pre-conception exposure or an ancestor’s exposure, by any primarily non-genetic route.
- slow growth period: the mid-childhood years before the prepubertal growth spurt; the proposed exposure-sensitive window.
- cultural transmission: parental nurturing and “social patterning of all kinds”; the authors grant it “does play an important role”.
- metabolic cascade / propagational programming: an exposure alters one generation’s physiology, which programs the next, with no mark passed in a gamete.
- transgenerational epidemiology: life-course epidemiology extended to exposures in earlier generations.
Connections
- Pembrey et al. 2006 (LIT-1764), the original. This paper gives its findings, its cohorts and its rate ratios, by two of its authors.
- Bird (NOTE-1264) disputes Överkalix for ignoring cultural transmission, for non-replication (Vågerö et al. 2018) and for multiple testing. This paper answers the first and third in advance; the second postdates it. Denny Vågerö is a member of the Network that wrote it.
- Heard and Martienssen (NOTE-1254) and Horsthemke (NOTE-1265) use the strict rule: F2 after an exposed male. The Överkalix grandchildren are F2 of grandparents exposed in childhood, so they meet that rule. Horsthemke cites Kaati et al. 2007, the fourth row of Table 1.
- Dias and Ressler (NOTE-1281) is cited as a plausible mechanism; Gapp et al. 2014, which Horsthemke also cites, as the sperm-RNA case.
- The Dutch Hunger Winter (LIT-1751) appears only for its grandchildren’s birth weight and adiposity.
Bearing on the record
- The question (QUESTION-297), germline route. Three things.
- Överkalix is not a trauma case. The exposure was year-to-year food supply in a grandparent’s childhood. Higher grandchild mortality went with good supply (1.67), lower with poor (0.65). Bird’s “the Swedish famine” misdescribes it. The question cites it as germline evidence for trauma; it is evidence, if at all, for transmission of a nutritional state.
- Male line is not germline. The authors keep an open mind on mechanism and name paternal routes that bypass the gamete’s epigenome: seminal fluid, Sertoli-cell contacts, a metabolic cascade. The question should cite Överkalix as a male-line case, not a germline one.
- What it says on culture and replication. The authors grant that cultural transmission plays “an important role” and that replication is needed. Their case against culture is the pattern: a window in the grandparent’s mid-childhood and sex-specific lines. They measured no family culture (diet, child-rearing, stories); they adjusted only for the grandchild’s early-life circumstances. So Bird’s charge stands as a charge of omission, and the authors’ reply stands as an argument from the pattern. Neither side has data on family culture.
- An inference on the cultural route. A family habit or story would have to explain why a grandfather’s food supply at the age before his growth spurt, and not later, predicts his grandson’s death, and only down the male line. That is hard for habit and harder for identity: nothing in the pattern looks like remembering. If the pattern holds, it is the strongest human sign that something other than culture passes. It holds on about 100 people per cohort.
- The owner’s thesis (CLAIM-491). Not touched. What would pass here is physiology, not identification with an ancestor. No note on the claim is proposed.
- Proposed dated note on QUESTION-297 (below, section 4).
- No THEORY filed.
- No ML instruction.
Limitations
- Not the 2006 paper. Its methods, food-supply classification, confidence intervals, model, ALSPAC sample and discussion are not here; only its headline result, its cohort sizes and its rate ratios.
- Partisan. The authors are summarising their own studies. The network that wrote it met four times at the Karolinska Institute.
- Small cohorts. About 100 probands each; for diabetes deaths, each proband had a different paternal grandfather (Table 1 comment).
- Priors partly post hoc. Table 1’s “prior research questions” grow from earlier Överkalix results, so the coherence argument is partly internal to one dataset.
- Inconsistencies.
- Table 1 gives the 1905 cohort as n = 94 in the first study and n = 99 later.
- Table 1 says the 2006 paper’s sex stratification was “suggested by early ALSPAC results” citing Northstone et al. 2014, a paper eight years later.
- Table 2 gives ABIS as 17 045 participants and 17 055 children.
- Typos. “Cochester” for Colchester (correspondence); “Aarhas” for Aarhus (Table 2); “only one allele is expression”; “needs to considered”; “aimed at recruit” (Table 2); “measureable”; “Non-genetic biological transmission … spell out”; “0—13ys” (Table 1); “fo” for “of” in reference 32’s title; reference 17’s pages as “257–1”; the ALSPAC Study Team as “Team AS” in reference 29.
Open questions
- Did Vågerö et al. 2018 replicate the Överkalix male-line effect? It came after this paper; Bird reports its failures, its title reports a success.
- Can a cultural route explain an effect keyed to the grandparent’s age at exposure? A test would measure family diet and child-rearing across the three generations in a cohort with the same exposure data.
- Does any human trauma study meet the strict F2 rule, as Överkalix does for food supply?