Methylation profiles at birth linked to early childhood obesity

Delphine Larivière, Sarah J. C. Craig, Ian M. Paul, Emily E. Hohman, Jennifer S. Savage, Robert O. Wright, Francesca Chiaromonte, Kateryna D. Makova, Matthew Reimherr

This paper, recently published as a preprint on medRxiv, explores the epigenetic roots of childhood obesity by analyzing DNA methylation at the very beginning of life.

TL;DR

Researchers identified specific DNA methylation patterns in cord blood and placenta at birth that are strongly associated with infant weight gain and BMI at six months. By focusing on gene-body methylation, the team developed a "Methylation Risk Score" (MRS) that could eventually serve as an early-warning biomarker for childhood obesity risk.

What Problem Does This Paper Address?

Obesity is a massive public health challenge, yet our ability to predict it early in life remains limited. While genetics (DNA sequence) accounts for some risk, traditional Genome-Wide Association Studies (GWAS) only explain a small fraction of the variation in Body Mass Index (BMI)—a phenomenon known as "missing heritability."

Epigenetics, specifically DNA methylation (the addition of methyl groups to DNA that can turn genes "on" or "off"), offers a missing link. Unlike your static genetic code, the "epigenome" is modifiable by the environment (diet, stress, etc.) and can be passed down. The authors sought to determine if epigenetic markers present at birth could predict growth trajectories in the first six months of life, a critical window for metabolic programming.

Methods

The study followed 48 infants from the "SIBSIGHT" cohort. At birth, the team collected two types of tissue: cord blood and placenta.

  • Epigenetic Profiling: They used the Illumina Infinium MethylationEPIC chip to measure methylation at over 575,000 CpG sites.
  • Dimensionality Reduction: To make the massive dataset manageable, they grouped individual CpG sites by gene, averaging the signals to focus on "gene-body methylation" across ~20,000 genes.
  • Statistical Modeling: The researchers employed LASSO (Least Absolute Shrinkage and Selection Operator) regression. This is a machine learning-adjacent technique that performs both variable selection and regularization to identify which genes are the most "predictive" of three weight outcomes: Conditional Weight Gain (CWG), BMI, and weight-for-length ratio.
  • Risk Scoring: They used the results to create a Methylation Risk Score (MRS), a weighted linear combination of the most significant methylation states.

Datasets

  • Discovery Cohort (SIBSIGHT): 48 infants, primarily white and non-Hispanic, with detailed longitudinal data on growth, maternal health, and environmental factors.
  • Validation Cohort (PROGRESS): A much larger, independent dataset of 1,001 individuals from Mexico City. This allowed the researchers to see if their findings held up across different populations and environments.

Key Insights and Ideas

  • Tissue Specificity: Cord blood provided a much clearer signal than the placenta. The placenta showed highly heterogeneous methylation profiles, likely due to the mix of different cell types within the tissue, making it harder to find consistent biomarkers.
  • The "Big Three" Genes: Three specific genes in cord blood—PLIN4, UBE2F, and PPP1R16B—emerged as "master predictors," showing significant associations with all three weight outcomes.
  • PLIN4 is involved in lipid storage.
  • UBE2F is linked to protein degradation and has been previously associated with BMI in children.
  • PPP1R16B plays a role in endothelial (blood vessel) stability and has been linked to diabetes.
  • Predictive Persistence: Interestingly, the MRS calculated from birth tissues was not only associated with weight at six months but remained significantly correlated with weight outcomes at one and two years of age.

Key Results

  • Predictive Power: In the discovery cohort, the cord blood MRS for Conditional Weight Gain (CWG) achieved an adjusted R-squared of 0.874 ($p \le 2.2 \times 10^{-16}$), indicating a very strong fit for the model.
  • Gene Identification: The study identified 23 relevant genes in cord blood and 10 in the placenta.
  • Directionality: Higher methylation in PLIN4 and UBE2F was associated with increased weight, while higher methylation in PPP1R16B was associated with decreased weight.

Limitations and Future Work

The authors are transparent about several hurdles:

  1. Sample Size: With only 48 infants in the discovery cohort, the study is small for an 'omics' analysis. While the LASSO method helps mitigate this, larger studies are needed to confirm the results.
  2. Generalizability: While the "Big Three" genes showed signals in the Mexican (PROGRESS) cohort, the overall MRS did not transfer perfectly. This suggests that ethnic, demographic, or environmental differences (like the "external factors" influencing methylation) play a major role in how these scores perform across different populations.
  3. Resolution: By grouping data at the gene level to reduce noise, the researchers may have missed specific, high-impact individual CpG sites.
  4. Causality: This study shows correlation, not causation. Future functional studies are required to determine if these methylation changes actually drive obesity or are simply markers of an underlying process.

Summary updated

Abstract

Childhood obesity represents a significant global health concern and identifying its risk factors is crucial for developing intervention programs. Many "omics" factors associated with the risk of developing obesity have been identified, including genomic, microbiomic, and epigenomic factors. Here, using a sample of 48 infants, we investigated how the methylation profiles in cord blood and placenta at birth were associated with weight outcomes (specifically, conditional weight gain, body mass index, and weight-for-length ratio) at age six months. We characterized genome-wide DNA methylation profiles using the Illumina Infinium MethylationEpic chip, and incorporated information on child and maternal health, and various environmental factors into the analysis. We used regression analysis to identify genes with methylation profiles most predictive of infant weight outcomes, finding a total of 23 relevant genes in cord blood and 10 in placenta. Notably, in cord blood, the methylation profiles of three genes (PLIN4, UBE2F, and PPP1R16B) were associated with all three weight outcomes, which are also associated with weight outcomes in an independent cohort suggesting a strong relationship with weight trajectories in the first six months after birth. Additionally, we developed a Methylation Risk Score (MRS) that could be used to identify children most at risk for developing childhood obesity. While many of the genes identified by our analysis have been associated with weight-related traits (e.g., glucose metabolism, BMI, or hip-to-waist ratio) in previous genome-wide association and variant studies, our analysis implicated several others, whose involvement in the obesity phenotype should be evaluated in future functional investigations.
BibTeX
@article{lariviere2024methylation,
  author = {Delphine Larivière and Sarah J. C. Craig and Ian M. Paul and Emily E. Hohman and Jennifer S. Savage and Robert O. Wright and Francesca Chiaromonte and Kateryna D. Makova and Matthew Reimherr},
  title = {{Methylation profiles at birth linked to early childhood obesity}},
  journal = {Journal of Developmental Origins of Health and Disease},
  year = {2024},
  doi = {10.1017/s2040174424000060},
}

Loading related content...

Figure