Contribution
A proof of principle. Animal work had shown that early nutrition can change epigenetic marks for life. This paper asks whether the same can be seen in humans, using a famine with sharp start and end dates, documented rations and intact birth registries. It measures one well-characterised imprinted locus in exposed people and their siblings six decades on. It finds a difference for exposure around conception and none for exposure late in pregnancy.
Key insight
When the exposure falls matters more than how hard it hits the fetus. Late-gestation exposure cut birth weight and left IGF2 methylation alone. Periconceptional exposure left birth weight normal and lowered IGF2 methylation. So birth weight, the usual proxy in developmental-origins research, is “not necessarily sufficient to test the involvement of epigenetics” (p. 63). The authors attribute the window either to the epigenetic reprogramming of very early development or to the hematopoietic system being laid down early, and leave the choice to future work (p. 63).
Assumptions
- Siblings as controls. Same-sex siblings, born before or conceived after the famine, give “partial genetic matching”. This matters because IGF2 DMR methylation is highly heritable in the authors’ earlier twin study (p. 57).
- The mark is stable. The same twin study found IGF2 DMR methylation largely genetic and stable to middle age, so a difference at 58 is taken to be a difference since early life (pp. 57–58). The paper has one blood sample per person, so it cannot show when the difference arose.
- Blood stands for the body. Whole-blood DNA is the only tissue measured.
- Famine is the cause. The match with rat protein-restriction results “substantiates that famine is the main culprit”, though “cold and emotional stress” cannot be ruled out (p. 62).
- Last menstrual period dates conception. Periconceptional exposure is defined by the mother’s last menstrual period falling between 28 November 1944 and 15 May 1945 (p. 65).
Key results
- Cohort (pp. 64–65). The Hunger Winter Families Study: singleton births in 1945 and early 1946 at three institutions in famine cities (Amsterdam, Rotterdam, Leiden). Clinical examination and blood for 311 exposed people, 311 same-sex siblings and 349 unrelated controls. This study uses only exposed people and their siblings.
- Measurement (p. 65). Five CpG sites in the IGF2 DMR by mass spectrometry (Epityper), three resolved singly and two as a pair; each measurement in triplicate; sibships bisulfite-treated on the same plate; 93% complete data. Linear mixed models with sibship as a random effect, adjusted for age.
- Periconceptional exposure (Table 1, p. 59; 60 pairs).
- Average methylation 0.488 (SD 0.047) exposed vs 0.515 (SD 0.055) siblings: −5.2%, −0.48 SD, P = 5.9 × 10⁻⁵.
- By site: CpG 1, −6.9%, P = 1.5 × 10⁻⁴; CpG 2&3, −4.7%, P = 8.1 × 10⁻³; CpG 5, −7.2%, P = 1.4 × 10⁻³; CpG 4, −2.3%, P = 0.41.
- No sex interaction (P = 0.20).
- Lower in the exposed sibling in 72% (43/60) of sibships, across the whole famine period (Fig. 1A).
- Re-measured from bisulfite treatment onward in 46 of the 60 pairs: −5.6%, P = 2.1 × 10⁻³ (p. 59). This repeats the assay on the same people; it is a technical check, not a replication.
- Late-gestation exposure (Table 2, p. 61; 62 pairs, born 28 January to 30 May 1945, so exposed for at least 10 weeks). Average 0.514 vs 0.519, −0.9%, P = 0.64. No site differed.
- Timing (Table 3, p. 62). With both exposed groups and all 122 siblings in one model: periconceptional P = 1.5 × 10⁻⁵, late P = 0.69, timing × exposure P = 4.7 × 10⁻³.
- Birth weight (p. 61). Late-exposed: 3126 g, 296 g below 324 reference births of 1943 at the same institutions (95% CI −420 to −170 g). Periconceptionally exposed: 3612 g, not lower (95% CI +15 to +365 g). Methylation was not associated with birth weight (P = 0.39). Siblings’ birth weights were not available.
- Scale (p. 62). Among the 122 siblings, aged 43 to 70, ten years of age went with 3.2% lower methylation (P = 0.037). By my arithmetic, not the authors’, the famine difference equals about 16 years of that age-related decline.
- Rations (p. 59). Official daily rations averaged 667 kcal (SD 151), set weekly and the same for everyone.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | People exposed to famine around conception have lower IGF2 DMR methylation in blood six decades later than their same-sex siblings | moderate: 60 sibling pairs, a consistent direction at four of five sites and in 72% of pairs, small P values; one locus, one tissue, one cohort, a technical re-run but no independent replication | Table 1; Fig. 1A; p. 59 |
| C2 | The association is specific to periconceptional exposure; late-gestation exposure shows none | moderate: a null in 62 pairs and a significant interaction (P = 4.7 × 10⁻³), but the two windows also differ in how long exposure lasted | Tables 2–3 |
| C3 | Transient conditions early in human gestation “can be recorded as persistent changes in epigenetic information” | weak to moderate as a causal claim: an observational sibling comparison; one measurement per person, so persistence is inferred from the twin study’s stability result | Discussion, pp. 62–63 |
| C4 | Birth weight is a poor proxy for testing epigenetic involvement in developmental origins of disease | moderate: methylation differed where birth weight did not and not where it did; birth weight unrelated to methylation (P = 0.39) | p. 61; p. 63 |
| C5 | Famine, rather than cold or emotional stress, is the main cause | weak: an argument from similarity of size to rat diet studies; the other stressors are not measured | p. 62 |
Concepts
- IGF2 DMR: a differentially methylated region of the maternally imprinted insulin-like growth factor II gene. Its hypomethylation leads to expression from both alleles (p. 57).
- periconceptional exposure: famine at the time of the mother’s last menstrual period, so covering the final weeks of the egg, conception and the first weeks of the embryo.
- same-sex sibling control: a brother or sister of the same sex, born before or conceived after the famine, to match partly for genes and family.
- developmental origins hypothesis: adverse conditions during development contribute to adult disease risk; epigenetic dysregulation is one proposed mechanism.
Connections
- The question it is filed for (QUESTION-297). It is the paper behind the question’s “womb” route.
- Horsthemke (NOTE-1265) classifies the Dutch Hunger Winter results as fetal programming, “a direct consequence of the exposure in utero”. This paper fits that reading and does not claim otherwise: it studies only the exposed generation.
- Heard and Martienssen (NOTE-1254) draw the line this paper sits on the near side of: an effect counts as transgenerational only in a generation never exposed.
- Costa et al. (NOTE-1255) cite this paper; their finding that a mother’s late-gestation nutrition decides whether a father’s hardship shows in sons puts the womb inside the paternal story. Here late-gestation famine left no IGF2 mark, which suggests, as an inference only, that if the Civil War effect runs through the womb it is not through this kind of mark.
- Dias and Ressler (NOTE-1281). The mouse case concerns a mark in a parent’s germ cells; this one concerns a mark in the exposed person’s own blood. The two are often cited together and show different things.
- The thesis it is reprinted in extends the measurement to fifteen more loci (its chapter 4, “common and timing- and sex-specific” differences). That chapter was not read for this NOTE.
Bearing on the record
- QUESTION-297. This is the womb route at its strongest, and it
is narrower than the question’s one sentence on it. These are
inferences from the paper.
- It is real but small. A 5.2% difference at one gene, in blood, in 60 pairs, with no expression, disease or behaviour outcome. The paper shows a lasting mark, not a lasting harm.
- It is a famine, not a trauma. The exposure is undernutrition; the authors name emotional stress only as a possible extra contributor they cannot rule out (p. 62). Filed under intergenerational trauma, it shows that prenatal adversity can leave a mark, not that a mother’s psychological trauma does.
- It is not inheritance. The exposure is the person’s own, around their own conception, and no later generation is measured. The question’s classification is borne out.
- The window blurs “womb” and “germline” at one point. The sensitive period starts at the mother’s last menstrual period, so it takes in the final maturation of the egg and the embryo’s first reprogramming, before any womb in the ordinary sense. The paper does not say which stage carries the effect. This does not reopen the germline route for descendants: the egg was itself exposed.
- On the identity route it is silent. What passes is a methylation level in a growth gene. Nothing in it needs the exposed person to know of the famine, or to represent a mother or a nation.
- The owner’s thesis (CLAIM-491). Idle, as on the other biological routes. The paper shows a channel by which a past event reaches a later person with no self-identification involved. That is the question’s “state, not an identification” again.
- The triviality objection (CLAIM-492). Not addressed.
- A dated note is proposed for QUESTION-297 (section 3). No dated note on CLAIM-491.
- No THEORY filed. One locus, one cohort; the follow-up loci in the thesis are unread.
- No ML instruction.
Limitations
- One locus, one tissue, one time point. A candidate gene in whole blood, measured once per person in late middle age.
- Siblings differ in more than famine. Controls were born before or conceived after the famine, so they differ in age, birth order and the mother’s age. Age is adjusted for; the others are not discussed.
- Survivors. Only those who survived to about 58 and took part are studied.
- The two windows differ in more than timing. Periconceptional exposure was defined by the last menstrual period across the whole famine, so some had only weeks of it; late exposure by at least ten weeks before birth. Duration is not matched between them.
- No siblings’ birth weights, so the birth-weight comparison is with 1943 reference births, not within families.
- No replication in independent people; the re-measurement repeats the assay on 46 of the same pairs. The authors themselves say such studies “may need to be large and include replication” (p. 64).
- An understatement. The periconceptionally exposed are said to be “not lower” in birth weight than reference births, but the reported interval (+15 to +365 g; P = 0.03) means they were heavier.
- Table 1 does not quite match its own means. For CpG 1 the raw means (0.436 vs 0.470) give −7.2%, not the printed −6.9%. The relative changes may come from the model; the text does not say.
- Typos in the reprint.
- Fig. 1 title “data of birth” and caption “data of the last menstrual period”, for “date”; “A lowes curve” for “lowess”.
- “between January January 28 and May 30 1945” (p. 65).
- “pr enatally” (p. 64); “from the three institution” (p. 64).
- Ref. 7: “Epigentically”; ref. 23 gives an author as “de R, Sr.”
Open questions
- Does the difference arise at conception and stay, or accumulate with age? A second blood sample from the same pairs would tell. The thesis’s discussion makes the same point.
- Does the mark do anything? The paper names schizophrenia and coronary heart disease, both linked to early-gestation famine exposure, as the outcomes to test.
- Do the children of the periconceptionally exposed carry it? That is the only test that could move this work from the womb route towards the germline one, and it would need the F2/F3 standard of Heard and Martienssen.