Food abundance in men before puberty predicts a range of cancers in grandsons

The study investigates whether pre-pubertal food abundance in ancestors triggers transgenerational effects on cancer risk, focusing on a male-line pathway from paternal grandfathers to grandsons. Prior animal and emerging human evidence suggests early-life nutritional environments can induce heritable epigenetic alterations (e.g., DNA methylation, histone modifications, small RNAs) that influence offspring phenotypes. Human cohort observations and famine studies imply sex-specific, male-line transgenerational responses linking early-life nutrition to later-life diseases. Building on earlier work indicating paternal grandfather food abundance predicts cancer mortality in grandsons, this study tests the hypothesis that abundant food access in the paternal grandfather’s pre-puberty increases cancer occurrence in grandsons (G2 men), examines specificity by grandparental line and sex, and evaluates whether similar effects appear in the parental (G1) generation.

Animal models (e.g., Agouti mice) show nutrition-induced epigenetic changes can persist across multiple generations, with paternal diet and obesity altering offspring metabolism, potentially via sperm tsRNAs and other epigenetic carriers. Human sperm epigenome content and small RNAs rapidly respond to dietary changes, suggesting a mechanism for soma-to-germline communication. Human studies of severe deprivation (Dutch Hunger Winter, Ukrainian and Chinese famines) link prenatal malnutrition to epigenetic marks (e.g., IGF2 methylation) and adult disease. Observational cohort studies associate childhood nutrition with adult cancer, and prior historical cohorts suggest male-line effects of early-life nutrition on mortality and cardiometabolic outcomes in descendants. A previous study reported paternal grandfather food abundance predicting adult-onset cancer mortality in grandsons. Collectively, literature supports examining sex- and lineage-specific transgenerational associations between ancestral nutrition and descendant cancer risk.

Design and data sources: Historical cohort study using the Uppsala Multigenerational Study. G1 comprises all live births at the Academic Hospital in Uppsala from 1915–1929. Their children who survived to the 1960 Census form G2 (born 1932–1990). Grandparents (G0) were traced for those with ancestry in the Uppland region; exposure ascertainment was possible for G0 born 1865–1900. Exposure: Regional harvest yields (Statistics Sweden, 1874–1910) on a 0–10 scale were used as contextual proxies for food access. Categories were defined as good/abundant (≥8.5), poor/very poor (<5.0), with 5.0–<8.5 as intermediate (reference). A G0 was considered exposed if experiencing at least one year in the specified category during pre-puberty (ages 9–12 for boys, 8–10 for girls). Urban vs rural upbringing was defined by growing up in one of the ten largest cities (urban) versus countryside (rural), with rural G0s considered more dependent on local harvests. G1 food access was derived from a national harvest index (1–5), dichotomized at ≥3.3; this was used only as a control. Outcomes: Cancer events were defined as primary cancers recorded in the Swedish Cancer Registry or cancer listed as the underlying cause of death in the Cause-of-Death Registry (ICD-7 groupings used across the entire period). For grouped analyses, each individual’s first event within a given ICD-7 group counted toward that group; first primary only sensitivity analyses yielded similar results. Cohort sizes and events: In G2 there were 3422 cancer events (men and women combined). Subsets with full covariate data included 8431 G2 with rural ancestry (4310 men, 4121 women) and 3754 with urban/mixed ancestry (919 men, 1839 women). Among G1 with rural ancestry there were 3781 individuals (1960 men, 1821 women) and 1599 cancer events in those with rural G0 ancestors. Additional person-years and event counts by ancestry and sex are provided in Table 1 of the article. Statistical analysis: Cox proportional hazards models with age as the underlying time scale were used to estimate hazard ratios (HRs) and 95% CIs, with sibling-cluster robust standard errors. Analyses of G2 controlled for G2 birth year (five-year bands), sibling size and order, parental death before age 18, parental (G1) education (elementary vs more), 1970 income quintiles, 1960 social class (non-manual, manual, farmers/entrepreneurs, unknown), and G0 birth year (linear trend). Food access of the other grandparent (maternal/paternal counterpart) and G1 parent’s food access were included. Additional models controlled for G0 paternal age at G1 birth. Analyses of G1 cancer controlled for G1 birth year (five-year groups), parental social class and marital status at birth, and sibling position. Interaction by G2 sex was tested. Multiple testing: No correction was applied for the a priori primary hypothesis (Table 2), while Bonferroni correction (0.05/24 = 0.0021) was applied for grouped cancer outcomes (Table 3). Analyses were conducted in R 4.1.x. Ethics approval was obtained from the Regional Ethical Review Board of Stockholm; all data were anonymized and linkage performed by Statistics Sweden.

• Total G2 cancer: Variation in G0 food access predicted cancer occurrence in G2 in a lineage- and sex-specific way. Abundant pre-pubertal food access in paternal grandfathers (but not other grandparents) was associated with increased cancer in grandsons (G2 men). Associations in granddaughters were limited and sporadic. A strong G2–G1 interaction (p ≈ 0.00002) indicated that effects were present in grandchildren but not in the parental generation.
• Cancer group-specific results (Bonferroni-corrected analyses): Evidence indicated specific male cancer groups responding to ancestral food abundance, including digestive system/peritoneum and male reproductive system cancers. Additional elevated risks were reported for some groups (e.g., buccal cavity/pharynx; other/unspecified sites) in grandsons of paternal grandfathers with abundant food access, with some findings surviving Bonferroni correction.
• No transgenerational signal in G1: Analyses of G1 cancer (men and women) by their fathers’ (G0) food access showed no consistent or significant associations across cancer groups (HRs generally near null, wide CIs).
• Robustness: Results were adjusted for extensive social and demographic covariates, other grandparental and G1 food access, and G0 paternal age; sensitivity analyses counting only first primary cancers yielded similar HRs.

Findings support a male-line, preconceptional transgenerational pathway linking paternal grandfather pre-pubertal food abundance to increased cancer susceptibility in grandsons. The use of year- and region-specific harvest fluctuations as a quasi-natural experiment reduces bias from confounding, as harvest variation includes a random component unrelated to individual family characteristics. Extensive adjustment for social and demographic factors, alongside controls for the other grandparent’s food access and parental food access, did not explain the observed associations. The generational specificity (effects in G2 but not G1) and sex specificity (male-line to grandsons) are consistent with epigenetic mechanisms acting through the male germline. Prior literature shows sperm epigenome responsiveness to diet, including alterations in DNA methylation, histone marks, and small RNAs (e.g., tsRNAs) acquired during spermatogenesis and epididymal transit. The authors posit that ancestral nutritional signals carried via sperm or seminal fluid might influence epigenetic reprogramming in early embryogenesis in G2, potentially altering imprinting or methylation of tumor suppressor or DNA repair genes, thereby increasing multi-organ cancer susceptibility in grandsons.

Abundant ancestral nutrition in the paternal grandfather’s pre-pubertal period predicts increased susceptibility to a range of cancers in grandsons, consistent with a male-line transgenerational effect. Associations were not attributable to measured social or demographic confounding, bias, or chance. The proposed mechanism involves early molecular signals in the male germline influencing epigenetic reprogramming in the next generation. Effects in women were limited. Future research should elucidate the specific molecular mediators (e.g., sperm small RNAs, DNA methylation, histone modifications), refine exposure timing and dose-response relationships, and replicate findings in other populations with high-quality multigenerational data.

• Exposure proxy: Harvest yields are contextual proxies for individual dietary intake; potential for exposure misclassification remains, especially for urban G0s less dependent on local harvests.
• Outcome classification: Long follow-up across ICD revisions can introduce misclassification; while ICD-7 groupings were harmonized, site-specific analyses suffered from small numbers, limiting precision and interpretability.
• Small counts for many site-specific cancers reduce power and increase uncertainty; multiple comparisons were addressed variably (no correction for the a priori primary test, Bonferroni for grouped cancers), risking type I/II errors.
• Susceptibility window definition (pre-puberty) is approximate and subject to individual and secular variation.
• Observational design with residual confounding possible despite extensive covariate adjustment.
• Generalizability may be limited to similar historical and geographic contexts; analyses were restricted to G0 born 1865–1900 with traceable ancestry and full covariate data.