Using adopted individuals to partition indirect maternal genetic effects into prenatal and postnatal effects on offspring phenotypes

  1. Liang-Dar Hwang
  2. Gunn-Helen Moen
  3. David M Evans  Is a corresponding author
  1. Institute for Molecular Bioscience, The University of Queensland, Australia
  2. The University of Queensland Diamantina Institute, The University of Queensland, Australia
  3. Institute for Clinical Medicine, Faculty of Medicine, University of Oslo, Norway
  4. K.G. Jebsen Center for Genetic Epidemiology, Department of Public Health and Nursing, Norwegian University of Science and Technology, Norway
  5. Population Health Science, Bristol Medical School, University of Bristol, United Kingdom
  6. MRC Integrative Epidemiology Unit, University of Bristol, United Kingdom
15 figures, 5 tables and 8 additional files

Figures

Path diagrams illustrating the structural equation models (SEM) underlying the seven family structures modelled in this manuscript (G1 – G7).

Causal relationships are represented by one headed arrows. Two headed arrows represents correlational relationships. Observed variables and latent variables are shown in squares and circles, …

Power to detect prenatal maternal genetic effects (γm) (top) or postnatal maternal genetic effects (βm) (bottom) whilst also varying the size of prenatal and postnatal maternal genetic effects, paternal genetic effects (βp) or offspring genetic effects (βo).

Effect sizes are parameterized using the path coefficients β and γ. Power was calculated assuming sample sizes approximating the number of white European individuals in the UK Biobank with …

Power to detect prenatal maternal (γm) or postnatal maternal genetic (βm) effects whilst varying the numbers of each family structure, with the sample sizes of other family structures approximating numbers of white European individuals in the UK Biobank reporting educational attainment (1000 biological trios, 4000 biological mother-offspring pairs, 1800 biological father-offspring pairs, 300,000 singletons, 6000 singletons, and 50 biological mother-adopted offspring pairs).

Path coefficients representing postnatal or prenatal maternal genetic effects, paternal genetic effects (βp) and offspring genetic effects (βo) were fixed to 0.1. The covariance between maternal and …

Diagram illustrating a Mendelian randomization study designed to estimate the causal effect of a maternal prenatal environmental exposure on an offspring later life outcome.

Maternal genotype proxies a perinatal exposure of interest and is used as an instrumental variable to estimate the causal effect of the perinatal exposure on the offspring outcome. The pathway of …

Appendix 1—figure 1
Path diagrams illustrating the structural equation models (SEM) underlying four additional family structures (G8 – G11) where biological and adoptive parents are related.

Observed variables and latent variables are shown in squares and circles respectively. Causal relationships are represented by one headed arrows. Two headed arrows represents correlational …

Appendix 1—figure 2
Power to detect prenatal maternal (γm) or postnatal maternal (βm) genetic effects whilst varying the correlation (rg) between maternal and paternal genotypes.

Power was calculated using sample sizes approximating the number of white European individuals in the UK Biobank reporting their own educational attainment (1000 biological trios, 4000 biological …

Appendix 1—figure 3
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive mothers being siblings (top) or cousins (bottom) of biological mothers.

An SEM which did not correctly model this relationship was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 …

Appendix 1—figure 4
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive mothers being siblings (top) or cousins (bottom) of biological fathers.

An SEM which did not correctly model this relationship was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 …

Appendix 1—figure 5
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive fathers being siblings (top) or cousins (bottom) of biological fathers.

An SEM which did not correctly model this relationship was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 …

Appendix 1—figure 6
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive fathers being siblings (top) or cousins (bottom) of biological mothers.

An SEM which did not correctly model this relationship was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 …

Appendix 1—figure 7
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive mothers being siblings (top) or cousins (bottom) of biological mothers.

An SEM which modelled this relationship correctly was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 biological …

Appendix 1—figure 8
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive mothers being siblings (top) or cousins (bottom) of biological fathers.

An SEM which modelled this relationship correctly was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 biological …

Appendix 1—figure 9
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive fathers being siblings (top) or cousins (bottom) of biological fathers.

An SEM which modelled this relationship correctly was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 biological …

Appendix 1—figure 10
Estimated prenatal (γm) and postnatal (βm) maternal effects and type 1 error rates whilst varying the percentage of adoptive fathers being siblings (top) or cousins (bottom) of biological mothers.

An SEM which modelled this relationship correctly was fit to the data. Power was calculated using simulated data with 1000 biological trios, 4000 biological mother-offspring pairs, 1800 biological …

Appendix 1—figure 11
Power to detect prenatal (γm) and postnatal maternal (βm) genetic effects when the relationship between adoptive and biological parents are correctly specified in the model.

Power was calculated using simulated data assuming 1000 biological parent-offspring trios, 4000 biological mother-offspring pairs, 1800 biological father-offspring pairs, 300,000 singletons from …

Tables

Table 1
Modelling results of birth weight in the UK Biobank.
Unweighed PRSWeighted PRS
EstimateStd ErrorP-valueEstimateStd ErrorP-value
Offspring effect–0.0080.0040.058–0.1730.1150.132
Prenatal maternal effect0.0350.0130.0060.8930.3830.020
Postnatal maternal effect–0.0170.0110.130–0.4190.3350.211
Postnatal paternal effect0.0030.0050.5880.0480.1510.751
  1. PRS: polygenic risk score constructed using 20 SNPs showing maternal effects on birth weight from Warrington et al., 2019; Std Error: standard error.

Table 2
Modelling results of educational attainment in the UK Biobank.
Full sample
Unweighted PRSWeighted PRS
EstimateStd Errorp-valueEstimateStd Errorp-value
Offspring effect0.0260.0053.18 × 10 -72.4740.4853.40 × 10 -7
Prenatal maternal effect0.0270.0110.0131.7220.7030.014
Postnatal maternal effect–0.0060.0080.484–0.2880.5140.575
Postnatal paternal effect0.0180.0070.0061.2330.6780.069
Excluding adopted individuals with breastfeeding information
Unweighted PRSWeighted PRS
EstimateStd Errorp-valueEstimateStd Errorp-value
Offspring effect0.0250.0055.52 × 10–72.4410.3725.15 × 10–11
Prenatal maternal effect0.0170.0120.1400.9090.7050.197
Postnatal maternal effect0.0040.0090.6820.5490.5950.357
Postnatal paternal effect0.0190.0070.0051.2730.5520.021
  1. PRS: polygenic risk score constructed using 1,267 SNPs from Lee et al., 2018; Std Error: standard error.

Author response table 1
EffectmatrixEstimateStd.ErrorP-value
GRSV25.3170.1900.000
Fetal effectB_FY0.0000.0020.997
Postnatal maternal effectB_MY0.0070.0040.113
Paternal effectB_PY-0.0010.0030.697
Prenatal maternal effectG_MY0.0030.0050.586
e1E10.2060.0010.000
e2E20.1790.0050.000
rhoR-0.0450.3690.907
Author response table 2
Adopted individuals who knew their breastfeeding statusAdopted individuals who did not know their breastfeeding status
Years of education (mean (SD))13.97 (4.98)13.15 (5.00)
Birthweight (kg; mean (SD))3.30 (0.43)3.34 (0.42)
Author response table 3
EffectmatrixAdopted individuals with age<60 removed (N_G5=2756)Adopted individuals with age>=60 removed (N_G5=2422)
EstimateStd.ErrorP-valueEstimateStd.ErrorP-value
GRSV571.8854.7470.000571.9435.0060.000
Fetal effectB_FY0.0260.0050.0000.0250.0050.000
Postnatal maternal effectB_MY-0.0190.0090.0460.0070.01000.482
Paternal effectB_PY0.0180.0070.0080.0190.0070.004
Prenatal maternal effectG_MY0.0390.0110.0010.0140.0120.235
e1E123.8970.1190.00023.8700.1260.000
e2E221.5440.8570.00024.3350.7270.000
rhoR58.2089.3970.00058.2629.9560.000

Additional files

Supplementary file 1

SNPs used to construct unweighted polygenic scores of educational attainment.

https://cdn.elifesciences.org/articles/73671/elife-73671-supp1-v2.xlsx
Supplementary file 2

SNPs used to construct unweighted polygenic scores of educational attainment.

https://cdn.elifesciences.org/articles/73671/elife-73671-supp2-v2.xlsx
Supplementary file 3

Comparison of power estimated from asymptotic calculations and simulations using varying sizes of prenatal maternal genetic effects (γm), paternal genetic effects (βp) or offspring genetic effects (βo) with sample sizes approximating the number of individuals in UK Biobank with educational attainment information.

1000 biological parent-offspring trios, 4000 biological mother-offspring pairs, 1800 biological father-offspring pairs, 300,000 singletons, 6000 adopted individuals, and 50 biological mother-adopted offspring pairs. Covariance between parental genotypes was fixed at 0.

https://cdn.elifesciences.org/articles/73671/elife-73671-supp3-v2.xlsx
Supplementary file 4

Comparison of power to detect pre-natal (γm) and post-natal (βm) maternal genetic effects estimated from asymptotic calculations and simulations using varying sample size for each of the 7 family structures and the remaining family structures approximating the number of individuals in the UK Biobank with educational attainment data.

1000 biological parent-offspring trios, 4000 biological mother-offspring pairs, 1800 biological father-offspring pairs, 300,000 singletons, 6000 adopted individuals, and 50 biological mother-adopted offspring pairs.

Path coefficients representing postnatal or prenatal maternal genetic effects, paternal genetic effects (βp) and offspring genetic effects (βo) were fixed to 0.1. Covariance between parental genotypes was fixed to 0.

https://cdn.elifesciences.org/articles/73671/elife-73671-supp4-v2.xlsx
Supplementary file 5

Sensitivity analysis for the modelling of educational attainment in the UK Biobank using SNPs from the Okbay et al., 2016.

https://cdn.elifesciences.org/articles/73671/elife-73671-supp5-v2.xlsx
Supplementary file 6

Some limitations/assumptions of our Structural Equation Model (SEM) and its application to the UK Biobank.

Source Code File. R scripts that fit Structural Equation Modelling to partition maternal genetic effects into pre- and postnatal effects on offspring phenotypes.

https://cdn.elifesciences.org/articles/73671/elife-73671-supp6-v2.docx
Transparent reporting form
https://cdn.elifesciences.org/articles/73671/elife-73671-transrepform1-v2.docx
Source code 1

Adoption_SEM.R.

https://cdn.elifesciences.org/articles/73671/elife-73671-code1-v2.zip

Download links