Genetic contributions to variation in human stature in prehistoric Europe

Abstract

The relative contributions of genetics and environment to temporal and geographic variation in human height remain largely unknown. Ancient DNA has identified changes in genetic ancestry over time, but it is not clear whether those changes in ancestry are associated with changes in height. Here, we directly test whether changes over the past 38,000 y in European height predicted using DNA from 1,071 ancient individuals are consistent with changes observed in 1,159 skeletal remains from comparable populations. We show that the observed decrease in height between the Early Upper Paleolithic and the Mesolithic is qualitatively predicted by genetics. Similarly, both skeletal and genetic height remained constant between the Mesolithic and Neolithic and increased between the Neolithic and Bronze Age. Sitting height changes much less than standing height-consistent with genetic predictions-although genetics predicts a small post-Neolithic increase that is not observed in skeletal remains. Geographic variation in stature is also qualitatively consistent with genetic predictions, particularly with respect to latitude. Finally, we hypothesize that an observed decrease in genetic heel bone mineral density in the Neolithic reflects adaptation to the decreased mobility indicated by decreased femoral bending strength. This study provides a model for interpreting phenotypic changes predicted from ancient DNA and demonstrates how they can be combined with phenotypic measurements to understand the relative contribution of genetic and developmentally plastic responses to environmental change.

Keywords: ancient DNA; evolution; height; stature.

Fig. 1.

Changes in height PRS and stature through time. Each point is an ancient individual, white lines show fitted values, gray area is the 95% confidence interval, and boxes show parameter estimates and P values for difference in means (δ) and slopes (β). (A–C) PRS(GWAS) (A), PRS(GWAS/Sibs) (B), and skeletal stature (C) with constant values in the EUP, LUP-Neolithic, and post-Neolithic. (D–F) PRS(GWAS) (D), PRS(GWAS/Sibs) (E), and skeletal stature (F) showing a linear trend between EUP and Neolithic and a different trend in the post-Neolithic.

Fig. 2.

Changes in sitting-height PRS and sitting height through time. Each point is an ancient individual, lines show fitted values, gray area is the 95% confidence interval, and boxes show parameter estimates and P values for difference in means (δ) and slopes (β). (A–C) PRS(GWAS) (A), PRS(GWAS/Sibs) (B), and skeletal sitting height (C), with constant values in the EUP, LUP-Neolithic, and post-Neolithic. (D–F) PRS(GWAS) (D), PRS(GWAS/Sibs) (E), and skeletal sitting height (F) showing a linear trend between EUP and Neolithic and a different trend in the post-Neolithic.

Fig. 3.

Geographic variation in PRS and skeletal standing height. Residuals for the linear height model (Fig. 1 D–F) against (A–C) latitude and (D–F) longitude. Each point is an ancient individual, lines show fitted values, gray area is the 95% confidence interval, and boxes show parameter estimates (β) and P values for slopes.

Fig. 4.

Changes in heel bone mineral density (hBMD) PRS and femur bending strength (FZx) through time. Each point is an ancient individual, lines show fitted values, gray area is the 95% confidence interval, and boxes show parameter estimates and P values for difference in means (δ) and slopes (β). (A and B) PRS(GWAS) (A) and PRS(GWAS/Sibs) (B) for hBMD, with constant values in the EUP-Mesolithic and Neolithic–post-Neolithic. (C) FZx constant in the EUP-Mesolithic, Neolithic, and post-Neolithic. (D and E) PRS(GWAS) (D) and PRS(GWAS/Sibs) (E) for hBMD showing a linear trend between EUP and Mesolithic and a different trend in the Neolithic–post-Neolithic. (F) FZx with a linear trend between EUP and Mesolithic and a different trend in the Neolithic–post-Neolithic.

Fig. 5.

Signals of selection on standing height, sitting height, and bone mineral density. (A–C) −Log₁₀ bootstrap P values for the Q_x statistics (y axis, capped at 4) for GWAS signals. We tested each pair of adjacent populations, and the combination of all of them (“All”). We ordered PRS SNPs by increasing P value and tested the significance of Q_x for increasing numbers of SNPs (x axis). (D) Distribution of Q_x statistics in simulated data (Methods). Observed height values for 6,800 SNPs shown by vertical lines.