Cardenas et al. Reply to “DNA Methylation and Prenatal Exposures”

We thank Ladd-Acosta and Fallin for their thoughtful commentary (1) regarding our work on prenatal maternal smoking and placenta DNA methylation mediators (2). They raise several critical points regarding biomarkers and mediators in epigenetic epidemiology. As noted in their commentary, a promising feature of epigenetic research is the malleability of some epigenetic markers as potential biological targets that might offset adverse health outcomes. This was in part our motivation in testing epigenetic mediators, in addition to expanding our molecular understanding of the genomic fingerprint of prenatal smoking on fetal growth. While we agree that much work is needed to capitalize on this promise as a public health intervention, there is encouraging preliminary evidence that supports nutritional interventions that offset epigenetic toxicity associated with environmental exposures. For example, a randomized crossover trial of subjects exposed to fine particulate matter demonstrated that B-vitamin supplementation nearly prevented DNA methylation alterations compared with placebo (3). For prenatal smoking, a randomized controlled trial of vitamin-C supplementation showed that the intervention group was able to restore DNA methylation levels of candidate genes associated with prenatal maternal smoking (4).

As suggested, studies in ethnically diverse populations with larger sample sizes and increased statistical power are needed to detect moderate to small associations, likely missed in our study. Our study was limited in size, within a relatively homogeneous population. Indeed, for these efforts, we encourage investigators and cohorts with available data to participate in the Pregnancy and Childhood Epigenetics (PACE) consortium, a global consortium of researchers pooling data in meta-analyses of DNA methylation studies across several cohorts (5). We concur that refining the timing of measurements for both the molecular marker and prenatal maternal smoking are needed to obtain better estimates and avoid cross-sectional comparisons. In addition, the integration of multiple molecular markers (e.g., genetics, epigenetics, transcriptomics, and proteomics) is needed to fully understand the interplay and downstream consequences of the DNA methylation associations found for prenatal smoking across tissues.

A major challenge highlighted by the authors is the tissue specificity of epigenetic marks and the need to use disease- and exposure-relevant tissues or cells (1). While blood, cord blood, and now placental DNA are widely collected and being used in epigenetic studies, the tissue-specificity challenge presents an opportunity to study target epigenetic marks relevant to the question. In addition to asking causal questions, tissue-specific DNA methylation signatures might persist years after an exposure, enabling the discovery of biosensors of environmental exposures (6).

Last, the use of parametric and other statistical mediation analyses has been a topic of debate recently (7). As pointed out by the authors, it is important to identify the question before performing mediation analyses (1). Our intent was to answer an explanatory question as well as identifying methylation sites that were significant mediators of the association between maternal smoking and birth weight. As others have pointed out, a DNA methylation mediator could be a proxy for the causal change and not necessarily a driver of the association. For example, DNA methylation changes might reflect altered transcription binding (8) or serve as surrogate markers of cell type or cell subtype expansion (9). We concur that multiple lines of evidence are still needed to provide conclusive evidence and design public health interventions to target epigenetic biomarkers. Replication across studies, methods, and robustness of results are required before interventions can be recommended.