Epigenetic studies of child neurodevelopment: what can we learn from a developmental science perspective?

Epigenetic studies are increasingly being conducted to test developmental questions, including the role of DNA methylation as a contributor to child neurodevelopment. Epigenetic modifications are intriguing biological mechanisms to study in relation to child neurodevelopment because they are dynamic and known to be influenced by many of the same internal (e.g., genetics) and external (e.g., environment) factors that are associated with neurodevelopmental processes. To move the field forward, research could benefit from a deeper consideration of the core tenets of developmental science. The goal of this commentary is to describe three of these core tenets and explain how they can be applied to testing novel epigenetic questions. Integrating these tenets into ongoing and future studies of epigenetics and neurodevelopment could open the door to answering novel research questions and foster new paradigms concerning the role of epigenetics in child development.

Research on epigenetic mechanisms in child development grew out of a candidate gene approach. Early studies investigated how environmental conditions such as exposure to stressors impacted DNA methylation of single genes, and in turn, how methylation of these genes was associated with child developmental outcomes. This approach was entirely reliant on prior knowledge of genes known to contribute to neurodevelopment or the stress-response system. Perhaps the most common candidate gene in this area is NR3C1, a gene that encodes for the glucocorticoid receptor that binds with cortisol, a well-studied stress hormone. Studies have consistently shown that exposure to environmental stressors including prenatal substance exposure, maternal mood disorders and childhood maltreatment is related to increased methylation of NR3C1 [1]. In turn, increased methylation of NR3C1 has been shown to predict abnormal stress reactivity, dysregulated neurobehavior and the development of psychopathology [2–4]. With the advent of accessible array-based technologies for genome-wide profiling of DNA methylation, and advancements in whole-genome bisulfite sequencing technologies, epigenome-wide association studies have surpassed candidate gene studies in popularity in recent years. Results from these studies have identified DNA methylation at novel loci in relation to both environmental experience and child outcomes, and in some cases point to DNA methylation as a mediator underlying associations between environmental exposures and neurodevelopmental outcomes [5].

Despite great advances in utilizing epigenetic data to study developmental questions, there are limitations to the current state of epigenetic research [6] and we are far from realizing a fully developmental behavioral epigenetics paradigm [7]. Theories of modern developmental science share several core ideas [8] that could be integrated into current epigenetic research in order to further this area of study. These ideas include first, development as multiply determined; second, development as guided by experience; and third, transactions between individuals and environments as drivers of development. In the sections below, we describe these ideas in more detail, along with the new epigenetic questions they raise. Of note, it remains unknown whether DNA methylation is causally related to neurodevelopment, whether it is merely associated, or whether it plays a mediating or moderating role. Each of these models gives us unique information about the role of epigenetics in neurodevelopment, and we touch upon each of them as pertinent in the following sections. We end with a discussion of current challenges to advancing the field of developmental behavioral epigenetics.

Three tenets of developmental science & their application to epigenetics

Development as multiply determined

First, development is determined not just by environment or epigenetics but by multiple influences, both internal and external to the individual. These influences span multiple levels and include genes, neurons, behavior, cognition, families, neighborhoods and society [9]. Research is needed that aims to contextualize how epigenetics relates to neurodevelopment, to better understand for whom and under what conditions we see links between epigenetic signatures and child outcomes. As a specific example, our work has shown significant associations between neonatal DNA methylation at sites across the epigenome and neonatal and child medical and neurobehavioral outcomes, in the context of children born very preterm [10,11], a unique group who experience most or all of the third trimester developing outside the intrauterine environment. An open question is the extent to which the same epigenetic differences that we have highlighted in our work would be observed among children born at term or with other prenatal and perinatal risk factors. On a broader scale, epigenetic studies are needed that not only control for child- and family-level factors (e.g., child sex, family socioeconomic status) but that explicitly test independent and/or joint influences across multiple levels of analysis. New frameworks such as the exposome have been developed as an approach for studying the totality of environmental exposures that impact health across the lifespan [12], and in the field of environmental health there has been substantial development of mixture methods to identify joint effects of multiple exposures on health outcomes [13]. Consideration of multi-omics data that includes genomic, transcriptomic, proteomic and metabolomic levels could also provide critical contextual information for understanding associations of epigenetics and neurodevelopment. These new approaches could productively be applied to epigenetic studies to better understand how the accumulation of environmental experience alongside epigenetic and other internal factors (e.g., multi-omics, brain structure/function) impact health and development.

Development as guided by experience

The second tenet, which very much follows from the first, is that development is guided by experience, particularly early experience and experiences occurring during critical and/or sensitive periods. Research has established that many of the same environmental influences that impact neurodevelopment (e.g., adversity, smoking, lead) also have an impact on the epigenome, but our understanding of critical or sensitive periods underlying these relationships remains incomplete. The importance of the first 1000 days (spanning conception to age 2 years) has recently been emphasized, both to draw attention to this early window of development that requires future study, and as a rallying cry for public policy aimed at optimizing health and opportunity during this period of rapid brain development [14].

Interestingly, epigenetic research has also shown that DNA methylation is most dynamic during the earliest years of life [15,16] and that early-life (i.e., neonatal) DNA methylation is perhaps more predictive of child outcomes compared with DNA methylation measured concurrently [17]. While much more investigation of longitudinal epigenetic patterns is warranted, current evidence implicates the earliest years of life as a potential sensitive period during which individual differences in epigenetic marks are established and exert disproportionate influence on later life outcomes. We do not fully understand the critical early experiences that establish individual differences in epigenetic patterns. Borrowing from what we know about children’s brain and behavioral development, these critical experiences are likely to include prenatal medical and psychological factors (e.g., nutritional status, immune activation, psychosocial stress, mood disorders) alongside postnatal socioenvironmental influences (e.g., caregiving quality, socioeconomic conditions, adversity/threat). The duration, intensity and timing of these experiences may further influence their epigenetic and developmental impact. While the early years of life seem to be particularly influential, it is also important to understand whether there are additional sensitive periods later in life. For example, the beginning of formal schooling or the pubertal years are salient transition periods when reorganization of environmental and biological systems might make epigenetic processes more dynamic, susceptible or influential.

Development as transactions between individuals & their environment

The third tenet is the idea that transactions between individuals and their environment are the driving force of developmental change. Not only does environmental experience shape the course of development, as described above, but individuals also play a role in their own development, in part because they shape their environments in critical ways. These interactions can occur through active processes, such an individual choices and actions (e.g., selecting a peer group), or through passive means (e.g., different characteristics of children elicit different types of parenting behavior). We can apply this idea to advancing epigenetic research in several ways.

First, studies are needed to elucidate the unique, joint and/or mutually influencing contributions of individual and environmental variables on epigenetic patterns and associated neurodevelopmental outcomes. Children’s characteristics at birth, such as their medical or neurodevelopmental characteristics, may interact with features of their environment to predict their early epigenetic patterns or trajectories. Or, akin to gene-by-environment interactions, it is possible that a child’s early epigenetic patterns interact with subsequent environmental experiences and exposures to differentially predict their developmental trajectory. For example, individual differences in DNA methylation at birth (potentially influenced by prenatal conditions) may make some children more or less susceptible to the influence of later environmental conditions (e.g., caregiving), with implications for downstream neurodevelopmental outcomes.

Unlike genetic variations, epigenetic marks are dynamic and reversible [18]. This means that in addition to epigenome–environment interactions, it is possible (even likely) that individuals’ behavior and their epigenomes bidirectionally influence one another across the life course. That is, while epigenetics could impact neurobehavior and development, an individual’s behavior could directly or indirectly cause changes in their own epigenetic landscape (e.g., by choosing different health behaviors or eliciting different types of caregiving) in ways that could also impact neurodevelopment. Epigenetics might also mediate the effects of environmental experience on development, potentially explaining how environments ‘get under the skin’ to impact child health and developmental outcomes. How these transactions between epigenetics, environments and individuals unfold will likely look different for different domains of development, at different life stages and in different populations (e.g., children born preterm vs term, children with neurodevelopmental or psychiatric disorders). Better elucidating the nature of these relationships will require longitudinal data and statistical methods that can accommodate mutual and/or bidirectional associations, such as cross-lagged panel models and structural equation models.

Current challenges to advancing a developmental behavioral epigenetic paradigm

Incorporating the above three tenets into the current landscape of epigenetic research clearly opens the door to testing novel research questions and hypotheses. Along with these vast opportunities for advancing developmentally informed epigenetic research, there are unique challenges to be addressed. Developmental science is first and foremost concerned with describing change over time. Longitudinal studies with repeated assessment of epigenetic (or multi-omic), environmental and phenotypic data are therefore critical, but they remain rare in practice. In part this is due to the difficulty of funding data collection, assays and analysis across all these modalities in the context of existing grant mechanisms and funding caps. This challenge is compounded in developmental science, where effect sizes tend to be smaller [19], thus necessitating larger samples to have power to detect significant effects. Now more than ever, interdisciplinary teams are needed that can tackle long-term, multi-arm studies with complementary aims and areas of expertise. National and international consortia that combine individual cohort studies are invaluable for increasing sample sizes for discovery and/or replication datasets. Practically, these efforts are often limited in scope due to different timing and types of assessments and due to inconsistencies in data collection, processing and analysis practices. It is also worth noting that individuals with certain risk factors (e.g., children born very preterm, children with prenatal substance exposure) and with racially and sexually minoritized identities have historically been inadequately represented in epigenetic research. This under-representation limits the generalizability of research utilizing existing databases and can make it challenging to identify a suitable replication sample for populations of children at risk.

Another difficulty related to studying epigenetic change over time is the lack of available statistical methodology or recommendations for analyzing longitudinal epigenome-wide data, where hundreds of thousands of features are measured at each time point. Common statistical methods in developmental science, such as mixed models and latent growth curves, are not well suited for these types of high-dimensional data. Prescreening of available epigenetic data could possibly result in a smaller number of variables that could be analyzed using existing methods. For example, one might limit their analysis to a subset of CpG sites that demonstrate significant change over time or that have strong prospective associations with a developmental outcome or trajectory of interest. Unfortunately, there is little guidance for prescreening in this way, nor are there any available simulation data that compare different methods for prescreening prior to longitudinal analysis. Alternatively, new methods for tackling ‘big data’ in the context of epigenetic studies, such as machine learning methods for developing polyepigenetic scores, could be adapted to tackling developmental questions. For example, epigenetic clocks could be applied to longitudinal DNA methylation data to examine changes in epigenetic age and/or age acceleration as a contributor to neurodevelopmental trajectories. Methodologists could also explore adaptations to existing statistical techniques to accommodate high-dimensional data. As an example, there has been increased interest in methods for conducting high-dimensional mediation analyses, and at least ten different techniques have now been developed [20]. These techniques could be applied to studying developmental questions such as the role of epigenetics as a mechanism linking environmental experience to developmental trajectories. Similar innovation in high-dimensional longitudinal modeling is now needed in order to study epigenetic change over time. New techniques that can accommodate different functional forms of change over time (e.g., linear, quadratic or spline models), are especially critical because it is likely that some epigenetic changes will follow complex or unanticipated nonlinear patterns.

The value of using peripheral tissue to measure epigenetic marks in developmental samples has also been questioned, especially given the lack of available resources for contextualizing results found in these tissues. There are multiple tissues of special interest to developmental epigenetic studies, including placenta and cord blood samples that can typically only be collected once (i.e., at birth). There are also a handful of minimally invasive tissue samples such as saliva or buccal swabs, or somewhat more invasive samples such as peripheral blood, that can be collected repeatedly from birth onward. Samples collected at birth are most informative for understanding the impact of prenatal stress and exposures, but most studies across different tissue types suggest there are systematic differences in the effects of these exposures on epigenetic patterns in different tissues. Practically, it is unclear how one might model change over time using DNA methylation assayed from different tissue samples. While there are resources for comparing methylation in peripheral versus brain tissue, using samples collected from the same individuals [21], these have primarily been developed using adult samples and thus are not as informative for developmental studies. If similar databases could be developed using epigenetic data from the first 1000 days, this would serve as an invaluable resource for better understanding tissue-specific epigenetic patterns and changes during a critical developmental window. These efforts will require special attention to the unique ethical and practical challenges inherent in the collection of neonatal brain tissue [22]. Moreover, it will be important for future research to investigate changes in cell composition for bulk tissue across this early developmental window, both in typical development and in relation to specific environmental exposures and/or neurodevelopmental disorders.

Developmental behavioral epigenetics: envisioning the future

In spite of these challenges, we anticipate that future years will see the realization of a truly developmental behavioral epigenetic paradigm, inspired and informed by core tenets from developmental science and taking advantage of new methodological innovations. We envision a future research landscape where much more is known about the magnitude and rate of change in epigenetic patterns across the early years of life, including the extent to which different environmental experiences and exposures at different times impact epigenetic change, and whether individual differences in epigenetic change relate to neurodevelopmental trajectories. We hope for a more nuanced consideration of epigenetics together with other levels of analysis (e.g., genetics, neural systems, behavior, proximal and distal environment) to predict typical and atypical neurodevelopment. While we have espoused the importance of longitudinal cohort studies, observational studies cannot typically disentangle correlational versus causal associations, as unmeasured confounding and reverse causality remain possible. Experimental (e.g., in vitro, animal studies), quasi-experimental and genetically informed research designs, combined with modern methods and resources for causal inference analysis (e.g., Mendelian randomization, causal mediation analysis), are needed to establish causal pathways [23]. These studies are especially critical if a future goal is to work toward epigenetically informed interventions. Embracing a developmental behavioral epigenetic paradigm will undoubtedly allow us to uncover more complex integration of epigenetics and development than has previously been possible.