Multigenerational effects of maternal undernutrition

Abstract

Intrauterine exposure to reduced nutrient availability can have major effects in determining susceptibility to chronic disease later in life. Martínez et al. (2014) demonstrate multigenerational effects of poor maternal nutrition and evidence of germ-line transmission through alterations in DNA methylation.

Mounting evidence suggests that an individual’s early life environment can set a trajectory of health or disease across the life span (Barker and Thornburg, 2013). Adverse exposures during development may alter the establishment of epigenetic markings that are subsequently maintained through replication. Modifications induced early in life may have greater phenotypic impact than changes induced later in life as they are amplified to a greater extent by the high rate of cellular replications and cell fate decisions occurring during development. In this edition of Cell Metabolism, Martínez et. al. (2014) take the developmental origins of adult disease concept a step further by showing the multigenerational effects of poor maternal nutrition. They demonstrate that intrauterine undernutrition in male mice induces alterations in lipid metabolism in the livers of F2 generation offspring exposed to normal intrauterine conditions, challenging the long held view that acquired traits (i.e. those induced by environmental exposures) are not transmitted to subsequent generations. The authors show that altered lipid metabolism is in part due to reduced expression of Liver X receptor-alpha (Lxra), a key lipogenic transcription factor, in the liver of F2 males. In addition, hypomethylation in the leader sequence of Lxra, is found in F1 sperm, F2 fetal liver and F2 adult liver and skeletal muscle, suggesting that these phenotypic traits may be transmitted to subsequent generations through modifications in epigenetic marks in gametes.

The work by Martínez et. al. is paradigm shifting in terms of the multigenerational implications that environmental exposures in pregnancy may have. At the same time, it raises many new questions about how the transmission of acquired phenotypic traits to subsequent generations might occur. As the authors discuss, recent evidence suggests that methylation may not be completely erased during gametogenesis as previously thought (Wang et al., 2014). Additionally, the establishment of de novo methylation patterns during development likely does not occur in isolation, but rather DNA methyltransferases function within the context of nucleosomal templates (Chedin, 2011). Although the DNA methylation changes associated with Lxra are quite small, in the order of 2–5%, the significance of these differences is more credible because Martínez et. al. also demonstrate enrichment of repressive histone marks, reduction of active histone marks and involvement of transcription factors binding sites that would contribute to reduce expression of Lxra. It is increasingly clear that epigenetic regulators cannot be considered singly as static markings on a linear strand of helical DNA, but rather are best considered as components of a moving 3-dimensional structure with multiple levels of interacting features.

As is commonly seen in epigenetic studies, the interpretation of differential methylation is also complicated by the fact that each cell type has a distinct methylome. At any given CpG site, the cystosine residue will be either methylated or unmethylated. Although methylation itself is binary, results of methylation assays are generally presented as percentages that represent the percent of sampled cells that are methylated at a given locus of interest. Comparison of quantitative measures of methylation between two samples is only straightforward when the distribution of cell types among the samples remains constant. Differentially methylated regions (DMRs) can be used to identify specific cell lineages with high reliability (Baron et al., 2006). Conversely, shifts in the distribution of cellular subpopulations can be inferred from genome-wide methylation data so that statistical adjustments can be made to account for cell mixture effects (Houseman et al., 2012). New analytical methods that are not dependent on availability of reference methylation datasets are also being developed, which is particularly important for the study of placenta and other tissues comprised of cell types for which reference datasets do not exist (Houseman et al., 2014). Tools such as these hold tremendous promise moving forward for studying whole tissue and epigenome-wide association studies that will enable us to capitalize on the large number of banked blood and tissue samples from longitudinal human cohort studies.

If the methylation changes described are not a result of cellular subpopulation artifact, understanding how such a small shift in methylation status in minor proportion of cells disrupts metabolic function is still perplexing, particularly given the robustness of the phenotype of the exposed offspring and the reproducibility of maternal undernutrition models. Quantification of DNA methylation gives us a snapshot of methylation status at one given time point and may not capture the natural or stochastic variability of epigenetic regulation that occurs over time. High variability in regulatory controls may provide the potential for adaptability to a changing environment, but variability may be detrimental when occurring in pathways critical to cell survival. Comparisons of DNA methylation between groups are generally aimed at identifying differences in average methylation levels in a population of cells from one individual, which is then averaged amongst the other individuals within an experimental group. Measurement variability is often considered “noise” stemming from technical sources but variability may also occur when perturbations cause disruptions without consistent directional change (hyper- or hypomethylation) in different individuals. Standard statistical methods may not be aimed at identifying unusually variable regions or these may not be interpreted as potential contributors to regulatory dysfunction. Examining the presence and distribution of variability in epigenetic regulatory marks may enable us to identify the parts of the genome that are most susceptible perturbation by environmental exposures.

In summary, Martínez et. al. provide convincing evidence that maternal health has significant repercussions not only for the metabolic health of offspring but also for the subsequent generation as epigenetic regulatory modifications may be transmitted through gametes. The work prompts us to critically examine the evidence supporting widely accepted fundamental concepts in biology and to make use of an ever-expanding armamentarium of investigative tools available to us. Finally, the demonstration of the multigenerational effects of poor maternal nutrition gives us to reason to reflect on how mother-child health is prioritized from public health policy perspective.