According to the developmental origins of health and disease hypothesis, in utero experiences reprogram an individual for immediate adaptation to gestational perturbations, with the sequelae of later-in-life risk of metabolic disease. An altered gestational milieu with resultant adult metabolic disease has been observed in instances of both in utero constraint (e.g., from famine or uteroplacental insufficiency) and overt caloric abundance (e.g., from a maternal high-fat, caloric-dense diet). The commonality of the adult metabolic phenotype begs the question of how diverse in utero experiences (i.e., reprogramming events) converge on common metabolic pathways and how the memory of these events is maintained across the lifespan. We and others have investigated the molecular mechanisms underlying fetal programming and observed that epigenetic modifications to the fetal and placental epigenome accompany these reprogramming events. Based on several lines of emerging data in human and nonhuman primates, it is now felt that modified epigenetic signature—and the histone code in particular—underlies alterations in postnatal gene expression and metabolic pathways central to accurate functioning and maintenance of health. Because of the tissue lineage specificity of many of these modifications, nonhuman primates serve as an apt model system for the capacity to recapitulate human gene expression and regulation during development. This review summarizes recent epigenetic advances using rodent and primate (both human and nonhuman) models during in utero development and contributing to adult diseases later in life.
Genomic DNA is the template of all organisms’ heredity, and the coordination and regulation of its expression result in the wide complexity and diversity of those organisms. The realization that factors other than the DNA sequence per se regulate gene expression resulted in a burgeoning of studies on epigenetics. The epigenome can be considered as a code written (or rewritten and revised) as a reflection of a modified genomic code, which in turn regulates gene expression as a result of its essential role in regulating accessibility of the transcriptional machinery. Epigenetics has had a tremendous impact on the molecular understanding of gene regulation during development. The widely studied correlation between diet, environment, and genetics has enabled scientists to move to the next step of understanding how the gestational milieu affects our genome and programs for health and disease (Figure 1). This review summarizes the seminal studies on in utero factors affecting the epigenetic regulation and programming for adult diseases such as obesity, hypertension, type 2 diabetes, and cardiovascular disease. We emphasize the unique importance of human and nonhuman primate (NHP1) models in understanding the role of epigenomic modifications in unraveling the molecular mysteries underlying the developmental origins of adult disease.
Epidemiological Studies Linking Early Origins with Adult Diseases
Early Retrospective Cohort Studies
Among the population-based studies on the developmental origins of health and disease, the most extensively documented and well-studied epidemiological studies employ subjects from the Dutch hunger winter. Unlike subsequent famines (e.g., the African famines of the 20th century), the Dutch famine occurred during a very short and well-defined period, enabling fairly precise exposure modeling.
In the winter of 1944–1945, the Axis alliance blocked all food convoys and confiscated stored supplies throughout the western and northern Netherlands, leading to a famine (daily intake of <1000 kcal and, for many, 500 calories/day) that resulted in the death from starvation of more than 18,000 women and children (Lumey et al. 2007). Despite significant poverty and tremendous hardship, the healthcare and data records were preserved, with the maintenance of meticulous records on periconception, early- and late-gestation maternal weight, relative nutrition, and fetal birthweight and gestation. Follow-up studies on children born during this period provided clear evidence that maternal undernutrition during the critical development phase was associated with the children's eventual development of schizophrenia (Susser and Lin 1992), elevated cholesterol and triglyceride levels (Lumey et al. 2009), impaired glucose tolerance (de Rooij et al. 2006; Ravelli et al. 1998), higher body mass index (Roseboom et al. 2000a, 2000c), and higher risk of coronary heart disease (Ravelli et al. 1999; Roseboom et al. 2000b). Barker's hypothesis emerged from these early cohort descriptions and has served as a cornerstone of developmental origins of health and disease; it posits that disturbances during critical gestational windows are associated with, and likely account for, the development of certain metabolic diseases later in life (Miller et al. 2010; Tice 2010).
Numerous subsequent epidemiological studies have been conducted to understand and support these initial observations, and several paramount studies are underway. Similar findings of a predisposition to adult metabolic disease have been found in studies of individuals exposed in utero and through early childhood to famine in Nigeria. Famine struck millions of Biafrans during the Nigerian Civil War of 1967–1970. During the first year of the civil war, Biafra lost most of its territory and access to airports and seaports, its oil refineries, and most of its oilfields, resulting in mass starvation of its population. From late 1968 until the final surrender in January 1970, Biafran refugees in packed, fetid camps were kept alive through one of the largest privately organized relief operations in African history. Akin to the Dutch winter famine, this short interval of mass starvation was documented clearly and was unequivocal in its devastating scope and long-term implications. Forty years since the famine, individuals exposed to it prenatally show hypertension, impaired glucose tolerance, increased waist circumference, and overweight phenotypes compared with individuals born after the famine (Hult et al. 2010).
Recent Longitudinal Cohort Studies
Other longitudinal cohort and population-based analyses have been undertaken in the past two decades and are summarized in Table 1. Initiated in 1998, the Southampton Women's Survey is a prospective cohort study that has assessed the diet, body composition (attained anthropometric measurements), physical activity, and social circumstances of a large cohort of pregravid women aged 20 to 34 years living in Southampton, United Kingdom. Women who subsequently became pregnant were studied at 11, 19, and 34 weeks of gestation, and ongoing follow-up of their offspring during infancy and childhood is underway (Inskip et al. 2006). Although their work is ongoing, initial data gathered from principle component analysis revealed the presence of two distinct occurring cohorts: “prudent diet” and “high-energy diet” (Crozier et al. 2006). Of note, when analyzed by maternal weight gain in pregnancy, “appropriate” (per 2009 Institute of Medicine guidelines) gain was linked to lower levels of adiposity in the offspring, whereas excessive weight gain was associated with greater neonatal fat mass and marginally associated with fat mass at 6 years of age (Crozier et al. 2010). OBELIX, a second large European cohort, aimed to study whether prenatal and early postnatal exposure to endocrine-disrupting chemicals can be a risk factor for obesity and metabolic syndrome–related diseases later in life (Legler et al. 2011).
In a similar vein, Project Viva is a US-based initiative to determine the relationship between maternal diet and behavior with offspring health through infancy and early childhood. The Project Viva cohort arises from slightly more than 2100 gravidae recruited at the time of their first prenatal visit (<22 weeks gestation) from community-based practice groups in eastern Massachusetts. Studies conducted on Project Viva cohorts have established the association between enhanced maternal–fetal n-3 polyunsaturated fatty acid status and lower childhood adiposity (Donahue et al. 2011). Finally, the heretofore unparalleled efforts of the National Children's Study2 will offer comprehensive, true population-based analyses among household-based enrollment of pregravidae with planned follow-up through 21 years of age. Ultimately, it is our opinion that the National Children's Study has the potential to become one of the richest and vibrant research efforts geared toward studying children's health and development through critical windows in development.
Epigenomic Modifications That Contribute to the Developmental Landscape and Fetal Programming
The aforementioned epidemiological studies collectively establish an association between perturbations during human gestation and altered risk of disease later in life. However, as with all observational studies, it is crucial to decipher the biological underpinnings and molecular mechanisms to establish a causal relationship between so-called fetal programming and later-in-life disease.
A fundamental tenet of modern biology is that the environment shapes the organism. At the species level, the process of evolution involves selection of the fittest DNA specimens over time according to the principles outlined by Charles Darwin. The mechanism involved is the inheritance of genetic traits first described by Gregor Mendel and later understood to be attributable to differences in DNA sequence. Since these early days of understanding that genes and chromosomes function as the backbone of our genetic heredity, the view of heredity has been written in the language of DNA, with the assumption that genetic mutations and changes in the nucleotide backbone have driven most descriptions of how phenotypic traits and diseases are handed down from one generation to another.
It is now known that this mechanism is only partially correct. Although selective pressures certainly can result in population genetic drift (generally through either single nucleotide polymorphisms or copy number variants), evidence from the past decade suggests that the environment largely shapes the organism through epigenetic mechanisms without effecting changes in DNA sequence (Bocock and Aagaard-Tillery 2009; Suter and Aagaard-Tillery 2009). Epigenetic mechanisms such as DNA methylation, histone acetylation, and RNA interference and their effects in gene activation and inactivation are increasingly understood to have a profound effect on alteration of an individual's appearance, transmission of a specific congenital anomaly, and even one's lifetime risk of common diseases such as obesity. Thus we have arrived at the point in our current understanding of genetics whereby we acknowledge that, although genomic DNA is the template of our heredity, it is the orchestration and regulation of its expression by epigenetic mechanisms in chromatin remodeling, DNA methylation, and microRNA (miRNA1) variation that ultimately result in the complexity and diversity among individuals observed in nature.
So what are epigenetic modifications? As schematically outlined in Figure 2, the nucleosome serves as the fundamental unit structuring chromatin. Two copies of histone proteins H3 and H4 form the tetramer, and two copies of H2A and H2B dimers join to form the histone octamer. Around this octamer, 147 bases of DNA are wrapped approximately 1.7 times in a superhelix to form the nucleosome (Luger et al. 1997). Adenosine triphosphate–dependent enzymes remodel the chromatin to provide access for transcription machinery (Ho and Crabtree 2010). Several histone modifications, such as acetylation, methylation, ubiquitylation, phosphorylation, and sumolyation, have been studied (Bernstein et al. 2005; Jenuwein and Allis 2001). DNA methylation is addition of methyl groups to the cytosine bases mediated through DNA methyltransferases (DNMTs1). DNMTs are classified as maintenance (DNMT1) and de novo methyltransferases (DNMT3a and DNMT3b), but all of them can be active in both states (Bestor 1992; Bestor et al. 1988; Chen et al. 2003). DNA methylation is inversely correlated with gene expression. Histone modifications are thought to act as guides for DNA methylation, thereby affecting the gene expression (Reik 2007; Vire et al. 2006). During preimplantation, the genome is demethylated and, after implantation, gets remethylated progressively in patterns structured ultimately by species and tissue type (cellular differentiation). Thus, in utero, maternal, and environmental factors have the unequivocal potential to program the fetal epigenomic code. However, to what extent this is largely proscribed and prescribed by species, degree of pluripotency, and lineage of differentiation is still being deciphered.
One of the most exciting attributes of epigenetic inheritance is the potential to meaningfully alter the pace of change. The in utero environment can theoretically lead to early potentially adaptive changes in the phenotype of the organism, which then bear the potential to be passed on to subsequent generations. Ergo, epigenetic mechanisms of inheritance offer potential plasticity to the organism's response to environmental challenges and are particularly responsive to environmental challenges during reproduction and development. However, to avoid “epigenomic entropy” as the methylome is actively demethylated and remethylated during early embryogenesis and development, there must be structure to this process. One such form of structure is differentially methylated regions and imprinted loci that maintain their methylation marks from the maternal and paternal gametes.
Genomic imprinting is an epigenetic phenomenon that silences expression of paternally or maternally derived alleles. It is postulated that the imprint is established in the parental germ lines and then maintained during early zygote formation when the remainder of the methylome is demethylated to regulate expression of imprinted genes that are essential to support proper embryonic development (reviewed in Feng 2010). Of the 120 or so identified imprinted loci in primates, the majority are essential for embryonic development, especially placental formation, and others regulate metabolism, behavior, and physiological functions. In humans, disruption of genomic imprinting causes several diseases, including cancer and (of interest herein) metabolic disorders.
Striking evidence for epigenetic mechanisms in the etiology of obesity is observed in individuals with the imprinting disorder Prader-Willi Syndrome (Jiang et al. 1998). Prader-Willi Syndrome is a development disorder characterized by low birth weight followed by hyperphagia and obesity in childhood. Prader-Willi Syndrome arises from a disruption of a parent-of-origin expression of alleles located within a 2 megabase region of chromosome 15 (15q11q13), and mapping and characterization of this region has helped to elucidate the epigenetic mechanisms involved in the etiology of Prader-Willi Syndrome (Lee et al. 2008).
The region on chromosome 15 associated with Prader-Willi Syndrome is not the only imprinted region implicated in the etiology of obesity. Aberrant DNA methylation within the differentially methylated region of the GNAS gene on chromosome 20 (20q13.11) renders pseudohypoparathyroidism type 1b (Kelsey 2010). Familial pseudohypoparathyroidism type 1b is a maternally transmitted syndrome characterized by obesity and resistance to parathyroid hormone and thyroid stimulating hormone. However, sporadic cases of pseudohypoparathyroidism type 1b caused by a loss of methylation within the differentially methylated region of the maternally derived chromosome with no apparent genetic changes (Linglart et al. 2007; Liu et al. 2000) have been reported. Again, alterations of DNA methylation patterns (imprinting) without a known genetic abnormality are associated with an obesogenic phenotype. However, it is not known if aberrant imprinting is implicated in the current obesity epidemic (Stoger 2008). Moreover, evolutionary genomic imprinting patterns and expression in mammals are highly conserved and observed only in marsupials and placental mammals, although imprinting in the human placenta has been shown to have limited evolutionary conservation (Monk et al. 2006). To date, an NHP model to study genomic imprinting in association with obesity syndromes has not been identified.
Primate Models That Recapitulate Population-Based Studies to Provide Mechanistic Insight
What advantages do primate animal models offer over rodent models? Rodent models have historically served as primary tools in studies of human diseases because of their fast generation time, large litter size, easier handling, and small size and because of the ability to perform transgenic studies (i.e., over at least three generations) using them. However, these advantages are accompanied by significant limitations. For example, because of their faster generation time, the evolutionary rate appears more rapid than has occurred among hominids, and more recent data suggest that there are wide structural and physiological differences between humans and rodents (Raaum et al. 2005). As a second pertinent example, maternal and fetal metabolic physiology comparatively differs among rodents and primates, as demonstrated by the regulation of satiety and appetite being developed postnatally in rodents and prenatally in humans and NHPs (Bouret et al. 2004; Grayson et al. 2006; Grove et al. 2003; Grove and Smith 2003; Koutcherov et al. 2002). As a third example, placental structure and function are distinct among primates, which is of importance in light of the essential role of placentation in regulation of fetal nutrient transport and fatty acid and lipid metabolism (Maltepe et al. 2010). The overall aim of NHP models is to generate an animal model that is robust and that both mimics the gravid physiology unique to primates (generally one uterus, one placenta, and one fetus) and recapitulates human gene function and regulation.
The closest genomic relatives to humans belong to Hominidae, with clades of humans, chimpanzee, and gorilla being of closest genomic relation (human to chimp bearing greatest sequence similarity), and the more distant clades of orangutan and rhesus still sharing 98% or more identity. The remarkable phenotypic variance is highlighted by recalling the significant eras of divergence, with human to chimpanzee having diverged in the neighborhood of 3.5 to 5 million years ago, whereas orangutan and macaque diverged 14 million years ago and 23 million years ago (Raaum et al. 2005). Thus, it is intuitive that these NHP models would serve as excellent animal models to study developmental origins of adult diseases because, despite our genomic similarity and recent divergence, our phenotypic variance is profound.
The landscape of fetal programming and the role of epigenomic modifications differ spatially and temporally with tissue and development stage-specific gene expression. For example, investigators delving into the evolution of cognitive capacity have observed well-conserved and predictive patterns of gene expression in the human brain compared with the chimpanzee brain (Caceres et al. 2003; Enard and Paabo 2004; Gu and Gu 2003; Khaitovich et al. 2006), with early studies suggesting hypermethylation of the cerebrum and thymus occurred with species specificity (Gama-Sosa et al. 1983). Subsequent studies capable of interrogating discriminate changes have demonstrated that the methylation pattern is conserved between humans and chimpanzees, with prior observed differences in gene expression being at least in part explained by differential methylation (Pai et al. 2011).
Which cellular pathways seem to be primarily affected across primate species? Cell cycle–related kinase (CCRK ) gene is involved in cell cycle growth, and transcriptional regulation has been identified to be differentially methylated and expressed in human and chimpanzee brain cortices. The intraspecific differential promoter methylation of CCRK in humans and chimpanzees has been postulated to be connected to interindividual differences in brain development. Thus, the divergence of DNA methylation in CpG islands can be considered as an epigenetic “footprint” of genes crucial for human development compared with NHPs (Farcas et al. 2009). In a baboon model of reduced maternal nutrient availability, it has been observed that the fetal hepatic PCK1 gene encoding phosphoenolpyruvate caroxykinase 1 has a hypomethylated promoter (Nijland et al. 2010). In the same model, it has been demonstrated that the juvenile offspring of mothers with reduced maternal nutrient availability exhibit increased fasting glucose, increased fasting insulin, and b cell responsiveness (Choi et al. 2011). Interestingly, Rodriguez and colleagues (2011) have demonstrated that exposure to prenatal betamethasone (synthetic glucocorticoid administered in case of premature labor to pregnant women) in maternal baboon model results in sex-specific effects in reversal learning and attention in the juvenile offspring. Maternal nutrient restriction during pregnancy in baboons altered fetal hepatic insulin-like growth factor (Li et al. 2009). Comparative epigenetic regulation with the closest relative of humans (NHPs) could be the crucial element in improving our understanding of phenotypic differences, cognition development, and susceptibility to adult diseases such as obesity, metabolic syndrome, and cardiovascular disease (Table 2).
NHPs have been evaluated to study different aspects of reproductive biology. For instance, baboons, rhesus macaques, pig-tailed macaques, and cynomolgus monkeys have been characterized to study endometriosis, which is a common cause of infertility in women (Hastings et al. 2006). Leveraging the baboon model to study endometriosis, it has been observed that together the inhibition of histone deacetylase and progesterone lead to expression of glycodelin gene (decreased expression observed in endometriosis) (Jaffe et al. 2007). Placental methylation for several genes, including APC, SFRP2, CYP24A1, and DNMT1, was conserved in primate species (with variation in distribution and absolute methylation levels) as opposed to nonprimate species (Ng et al. 2010).
NHPs have been also been explored as models to study metabolism with respect to enzymatic methylation of arsenic (carconigenic) compounds, which acts as a detoxification mechanism. Out of 17 NHPs, only four species demonstrated hepatic arsenite methyltransferase activity, thereby suggesting further evaluation of alternative detoxification mechanisms in these primates (Wildfang et al. 2001). The summary of studies performed in other arenas, such as metabolism, reproductive biology, and neurology, are summarized in Table 3.
Maternal Factors That Shape the Fetal Epigenome
Malnutrition results from either under- or over-intake of nutrients. We have understood for decades that there are fetal consequences to relative paucity of maternal intake of folate, vitamin B6, and vitamin B12 (neural tube malformations) and protein restriction (muscle wasting and intrauterine growth restriction). In the past decade, the implications of a caloric-dense, high-fat diet (namely, metabolic syndrome and obesity) have gained attention because of the alarming increase in the rate of adult and childhood obesity. Obesity results from higher caloric intake than energy expenditure. We will now look at the epigenetic effects on the fetus caused be maternal one-carbon supplementation, protein restriction, and high-fat diet (HFD1) exposure.
Maternal One-Carbon Metabolism
In recent years, more detailed information that suggests a crucial role for the one-carbon pathway intermediates in mediating epigenomic modifications, primarily methylation patterning, has emerged. Briefly, methionine is an essential amino acid that is converted to S-adenosyl methionine, which acts as a single-carbon donor and gets converted to S-adenosyl homocysteine. S-adenosyl homocysteine then gets converted to homocysteine. Folate, vitamin B6, and vitamin B12 are involved in the one-carbon metabolism pathway, and thus under- and/or over-intake of nutrients can affect the one-carbon metabolism pathway. Increased levels of plasma homocysteine levels have been associated with cardiovascular disease, placental abruptions, and neural tube defects (Ananth et al. 2007; Eskes 1998; Hague 2003; Mills et al. 1995; Refsum et al. 1998; Steegers-Theunissen 1991). Carbon donor supplementation during pregnancy is felt to bear beneficial effects toward the offspring. Waterland and colleagues (2008) have demonstrated that methyl donor supplementation in maternal diet prevents inheritance of obesity. Using the well-characterized yellow agouti (Avy) model, which contains a transposable element in agouti gene, they observed that maternal dietary methyl intake alters the phenotype of offspring by the transposable element, resulting in increased DNA methylation at the Avylocus (Waterland and Jirtle 2003). Histone modifications have also been reported at the proximal metastable epiallele, resulting in epigenetic alteration and gene expression (Dolinoy et al. 2010). Taken together, these lines of evidence support the notion that supplementation with one-carbon intermediates has the potential to alter methylation patterns in utero.
Maternal Protein Restriction: Rodent Models
In the rodent, several models of maternal low protein (MLP1) diet exposure have been useful in characterizing epigenetic and metabolic changes in the offspring. In utero exposure to MLP leads to decreased body weight of the offspring, predisposition to hypertension, and altered glucose metabolism (Langley and Jackson 1994; Langley-Evans et al. 1996). Bertram and colleagues (2001) have reported that glucocorticoid receptor (GR1) gene expression is increased at least twofold in the liver, kidney, lung, and brain of fetuses exposed to MLP in utero. This increase in expression is maintained into adult life compared with unexposed controls. Peroxisome proliferator-activated receptor (PPARa) expression is also altered in rat liver with MLP exposure (Lillycrop et al. 2005). It is now established that the GR and PPARa are epigenetically regulated by MLP diet exposure. PPARa is important for lipid and carbohydrate homeostasis, and GR is important for blood pressure regulation. Lillycrop and colleagues (2005, 2007) reported that in the rat liver, the promoter region of the GR, GR110, shows a 33% decrease in methylation and expression of GR is increased by 84% in offspring exposed to MLP. Expression of the DNA methyltransferase Dnmt1 is also lowered in MLP offspring. Further characterization of GR110 revealed altered histone modification patterns in MLP offspring, with an increase in H3K9 acetylation and a decrease in H3K9 dimethylation and trimethylation. PPARa has a 21% decrease in promoter methylation (Lillycrop et al. 2005).
Interventions to alleviate the adverse phenotypes of MLP exposure have similarly been studied. In rodent models, offspring exposed to MLP diet supplemented with folic acid do not show a propensity for hypertension in adulthood (Torrens et al. 2006). On a molecular level, the offspring also did not show a difference in promoter methylation in the GR or PPARa promoters compared with control diet animals (Lillycrop et al. 2005). It appears that the supplementation reverted both the phenotypic and epigenetic changes of MLP. Timing of folic acid supplementation is important. If folic acid is given to protein-restricted offspring in the juvenile period (postnatal days 35–40), changes in metabolism and epigenotype are not normalized (Burdge et al. 2009). MLP exposure increases the expression of the transcription factor CCAAT/enhancer-binding protein in the skeletal muscle of female offspring but not male offspring (Zheng et al. 2011). Analysis of the CCAAT/enhancer-binding protein promoter revealed increased acetylated histones H3 and H4 in the female offspring but not the male offspring. A study of genes that are involved in predisposition to hypertension showed gene-specific changes in expression accompanied by altered promoter methylation (Bogdarina et al. 2007). The angiotensin receptor, AT1b, shows increased messenger RNA and protein levels in the adrenal gland in 1-week-old rats exposed to MLP. This increased expression is accompanied by a decrease in promoter methylation. However, changes in gene expression caused by MLP exposure are not always concomitant with changes in DNA methylation. The glucokinase gene in rat is reduced in 6-month-old MLP rat liver, but there is no differential promoter methylation between control animals and MLP-exposed animals (Bogdarina et al. 2004).
One intriguing aspect of the MLP rat model system was the discovery that the adverse phenotypes are transmitted transgenerationally even when the exposure is eliminated in subsequent gestations. Exposure to the MLP diet during gestation is associated with both hypertension and insulin resistance in the subsequent F1 generation, even when the F1 rat is on a control diet postweaning. Their offspring (the F2 generation) have also been reported to have high blood pressure and insulin resistance in adulthood, and disruptions in glucose homeostasis have been reported in the F3 generation (Benyshek et al. 2006; Martin et al. 2000; Zambrano et al. 2005). Study of the adult male F2 liver showed decreased PPARa and GR promoter methylation (Burdge et al. 2007). The methylation status of the hepatic promoters was passed through a transgenerational model even though the F2 generation was never exposed in utero to the low protein diet and consumed only a control diet in adulthood. Transgenerational inheritance patterns are provocative evidence for a highly heritable, epigenetic-mediated mode of phenotypic transmission.
Maternal HFD Exposure: A Model For Understanding Epigenomic Modifications and In Utero Programming
Obesity is a well-characterized, multifactorial disease associated with higher risk for hypertension, insulin sensitivity, and cardiovascular disease. These risk factors together are known as metabolic syndrome. Epigenetics is one of the factors that explain the differential gene expression contributing to higher risk for disease state. For example, increased DNA methylation in peripheral blood leukocytes has been correlated with cardiovascular risk (Kim et al. 2010). Similarly, DNA methylation and histone acetylation have been shown to influence the regulation of somatic angiotensin converting enzyme regulation, which is crucial for maintaining cardiovascular homeostasis (Riviere et al. 2011). Histone modifications (e.g., histone lysine methylation) leads to dysregulation of genes involved in diabetes and related complications (Villeneuve et al. 2011).
It has long been postulated that obesity begets obesity. However, distinguishing factors associated with obesity (i.e., caloric intake) from obesity per se is not an easy undertaking. Along with reports of the effects of maternal undernutrition, protein restriction, and intrauterine growth restriction, studies of maternal overnutrition and high caloric density are becoming of immense importance for their potential elucidations on the current clinical epidemic of obesity. Animal models of maternal HFD exposure on offspring outcome show that, indeed, obesity begets obesity. In an NHP model of maternal obesity, fetal offspring exhibit the pathology of nonalcoholic fatty liver disease in utero (McCurdy et al. 2009). In a rat model of maternal obesity, although pups from obese dams did not weigh more than those from lean dams, they gained greater body weight and body fat when fed an HFD postweaning than those from lean dams (Shankar et al. 2008). Exposure to HFD in utero in animal models has been shown to make offspring more susceptible to type 2 diabetes, altered leptin sensitivity, hypertension, and obesity (Armitage et al. 2005; Chang et al. 2008; Ferezou-Viala et al. 2007; Gniuli et al. 2008; Kirk et al. 2009; Mitra et al. 2009; Samuelsson et al. 2008).
Histone Code Variations
Various epigenetic changes have been associated with exposure to a maternal HFD, with histone modifications serving as a cornerstone to these observations. As one such example, we have demonstrated in an NHP model of maternal obesity that fetuses exposed to a maternal HFD in utero have increased levels of hepatic H3K14 acetylation, concomitant with decreased levels of HDAC1 protein, messenger RNA, and activity; these changes are accompanied by persistent changes postnatally (Aagaard-Tillery et al. 2008). Employing techniques of chromatin immunoprecipitation followed by differential display polymerase chain reaction, we have identified the peripheral circadian regulator Npas2 (a transcription factor essential in maintaining oscillatory circadian rhythms) as a fetal gene reprogrammed by H3K14ac in HFD-exposed animals but not by maternal obesity per se. Of interest, although no differences were seen in Npas2 promoter DNA methylation, promoter H3K14ac occupation was increased in HFD-exposed fetal animals; this was not true in obese animals reverted to a control diet. This alteration in promoter occupancy resulted in alterations in gene expression of Npas2 in HFD-exposed animals (Suter et al. 2011). It was also reported that a postweaning HFD significantly disrupted Npas2 expression even without HFD exposure in utero.
Our observations are of interest to the discussion herein because alterations and disruptions of circadian rhythms are correlated with obesity and cardiovascular disease; the contribution of maternal diet to reprogramming the fetal circadian rhythm is of interest in furthering our understanding of current childhood obesity trends. Glucose and lipid homeostasis are known to display circadian variation, and an HFD in a murine model amplifies the diurnal variation in glucose tolerance in a Clock-dependent fashion. Similar to glucose homeostasis, characterization of peripheral Clock family members in liver and adipose tissue demonstrates robust and coordinated expression of the circadian oscillation system as well as Clock-controlled downstream effectors (Zvonic et al. 2006). Temporally restricted feeding or alterations in diet content (caloric dense/high fat) results in a coordinated phase shift in circadian expression of the major oscillator genes and their downstream targets in adipose tissues (Turek et al. 2005; Zvonic et al. 2006). Our evidence served as a fundamental description that the peripheral circadian metabolic clock in the developing primate fetus can be altered in utero by virtue of maternal dietary exposure, as evidenced by our observed changes in circadian gene expression patterns in fetal liver (Shankar et al. 2008).
Although we did not observe a role for differential methylation in the regulation of Npas2, other investigators have shown that DNA methylation changes have been reported in a mouse model of offspring exposed to an HFD in utero. Male mice aged 18 to 24 weeks exposed in utero to a maternal HFD showed an increased preference for glucose and fat. Gene expression changes were reported in the brain, including alterations in dopamine reuptake transporter, µ-opiod receptor, and preproenkephalin. Global and gene-specific DNA methylation patterns in the brain were altered in HFD-exposed animals (Vucetic et al. 2010). HFD exposure altered gene expression, epigenetic marks, and food preference in animals maintained on a control diet.
Noncoding miRNA levels have also been reported to be altered by HFD exposure. Fifteen-week-old mice maintained on a postweaning chow diet exposed in utero to HFD were studied for alterations in hepatic gene expression and miRNA levels. HFD-exposed animals had reduced hepatic let-7c. Expression of approximately 5.7% of all miRNAs was altered with HFD exposure. Furthermore, expression levels of metabolic regulators PPARa and CPT-1a were altered (Zhang et al. 2009). Aberrant DNA methylation and miRNA regulation have also been observed in diabetic nephropathy (Villeneuve et al. 2011).
In sum, the recent epidemic of childhood obesity has directed efforts toward understanding contributing factors. Extensive molecular studies from the Dutch hunger winter demonstrated that siblings exposed to the famine had decreased DNA methylation at the imprinted insulin-like growth factor 2 locus decades after the famine (Heijmans et al. 2008). The fetuses of obese mothers had greater body fat percentage, cord leptin, and interleukin 6 involved in inflammation than fetuses of lean mothers (Catalano et al. 2009). Maternal HFD in NHPs is also associated with decreased plasma n-3 fatty acids and fetal hepatic apoptosis (Grant et al. 2011). Thus, as established by above and many other studies, obesity begets obesity.
In addition to alternations in maternal nutrition, other in utero exposures modify and reprogram the developing fetus. The best characterized of these include alcohol, tobacco, and nicotine, alongside bisphenol A and organophosphates. Adverse effects of these substances are well documented, but the involvement of epigenetic mechanisms in these effects is not yet understood completely. Waddington hypothesized that the development path tends to follow a pathway termed as “canalization.” In the more modern version of Waddington's sentinel hypotheses, it is generally assumed that because of developmental plasticity, the fetus adapts to multiple environmental signals in utero and programs for the later life (reviewed in Burdge and Lillycrop 2010).
Maternal alcohol exposure results in fetal alcohol syndrome, which is associated with aberrant epigenetic programming. Altered DNA methylation leading to phenotypic changes has been demonstrated by several investigators (Haycock 2009; Kaminen-Ahola et al. 2010; Liu et al. 2009; Miranda et al. 2010; Ouko et al. 2009; Pandey et al. 2008; Zhou et al. 2011). Alcohol interferes with one-carbon metabolism in a manner similar to vitamin B12 deficiency causing disruption of DNA methylation by inhibiting methionine synthase, an enzyme that converts homocysteine to methionine (Cravo and Camilo 2000). Mice that were administered alcohol during midgestation showed increased levels of acetylated histone H3K9/18, followed by apoptosis in fetal lungs (Wang et al. 2010). Another study demonstrated genome-wide hypomethylation in fetuses of pregnant mice administered alcohol (Garro et al. 1991).
As a second example of a common exposure and the effects on fetal programming, significant advances have been made in understanding the underlying genomic and epigenomic mechanisms with maternal tobacco exposure. In more than four decades of research, epidemiological studies have established maternal tobacco consumption during pregnancy as the paramount risk factor for intrauterine growth restriction and small-for-gestational-age births in developed and developing nations. Nicotine, a compound present in tobacco smoke, is known to cause adverse fetal development issues by mediating intrauterine vessel constriction, leading to decreased placental blood flow (reviewed in Suter and Aagaard-Tillery 2009). In utero tobacco exposure is known to upregulate CYP1A1 placental expression in association with differential methylation at the xenobiotic response element transcription factor site (Suter et al. 2010). Methylation of PKCε gene promoter occurs because of prolonged nicotine exposure, resulting in aberrant fetal heart development (Lawrence et al. 2011). Children exposed in utero to tobacco were observed to have a 2.3% increase in DNA methylation in AXL, a tyrosine kinase that is highly heritable and is important in cancer and immune function (Breton et al. 2011).
Maternal smoking is also associated with poor fetal outcomes and aberrant miRNA expression. A study on maternal smoking in humans found that the development-associated miRNAs (miR16, miR21, and miR146a) were downregulated in placenta and miR146 downregulation was dose dependent (Maccani et al. 2010). Prenatal exposure to maternal cigarette smoking is also known to affect fetal brain and behavior development. For example, a study by Toledo-Rodriguez and colleagues (2010) found that adolescents who were exposed to maternal smoking in utero had increased DNA methylation in the brain-derived neurotrophic factor gene. Maternal smoking exposure in utero affects global and gene-specific DNA methylation (Breton et al. 2009). Although the detrimental effects of maternal tobacco smoke exposure to the developing fetus have been well documented, multiple population-based studies have identified maternal tobacco use as the strongest modifiable risk factor for intrauterine growth restriction. It is also implicated with other adverse pregnancy outcomes such as preterm birth, placental abruption, placenta previa, amniotic fluid emboli, and maternal deep vein thrombi. In spite of increased public awareness for more than three decades, the prevalence of smoking during pregnancy is reported to be persistently as high as 12–20%.
Numerous reports have found an association between maternal smoking during pregnancy and being overweight or obese in childhood (Gorog et al. 2011; Griffiths et al. 2010; Kleiser et al. 2009; Mangrio et al. 2010; Mizutani et al. 2007). Studies have shown that children born to smoking mothers trend toward higher serum glucose and have a predisposition to develop diabetes (Huang et al. 2007; Montgomery and Ekbom 2002). Exposure to environmental tobacco smoke is also a risk factor for childhood obesity. Numerous reports have shown that childhood obesity is more prevalent in children who were exposed in utero to environmental tobacco smoke through having a father who smokes (Kwok et al. 2010; Leary et al. 2006; von Kries et al. 2008).
Animal models have similarly been used to determine the molecular mechanism involved in these adverse effects to smoke exposure. Of the 4000 chemicals found in tobacco smoke, most studies have focused on the effects of prenatal nicotine exposure on the offspring development. The results have not been clear cut. Using a rat model of maternal nicotine exposure, it has been reported that there is no effect on offspring growth (Franke et al. 2008; Navarro et al. 1989); increased propensity to adiposity, b-cell dysfunction, and insulin resistance caused by maternal nicotine exposure has also been reported (Bruin et al. 2007, 2008; Holloway et al. 2005). The discrepancy may be caused by length of exposure (prenatal and/or lactation), dose, or a combination of both. More studies are necessary to determine the underlying etiology of tobacco smoke exposure and predisposition to obesity.
There is a fundamental dearth of studies on the effects of tobacco smoke associated with epigenetic changes that may be a key in understanding the origins of childhood and adult disease. Although there are numerous studies on changes in DNA methylation in a cancer paradigm with smoke exposure, little has been done on the fetal origins of disease. However, studies have shown that some morbidities associated with prenatal smoke exposure are transgenerational, suggesting an epigenetic mechanism of inheritance (Holloway et al. 2007; Li et al. 2005). A study of buccal cell DNA methylation in kindergarten and first-grade children who had been exposed prenatally to tobacco smoke showed a decrease in methylation of AluYb8 repetitive elements (Breton et al. 2009). We have demonstrated that placental promoter methylation of an important enzyme necessary for processing the polycyclic aromatic hydrocarbons found in tobacco smoke, CYP1A1, is decreased by 10% in mothers who smoke, concomitant with a fourfold increase in CYP1A1 gene expression (Suter et al. 2010).
As we move forward in the ability to derive meaningful molecular and phenotypic associations associated with the epidemic of obesity, in utero tobacco exposure may serve as an ideal disease model with which to explore the role of epigenomics in relation to pertinent environmental exposures. We now have more than 40 years of clinical evidence suggesting a causal relationship between tobacco use and delivery of infants who are small for gestational age. Studies dating back to 1957 by Simpson and Linda showed a significant 200 g difference in birth weight among infants who were exposed to as few as 10 cigarettes per day. Yet the mechanisms leading to the attenuation of fetal birth weight and adverse pregnancy outcomes are complex, involving multiple epidemiologic, genetic or epigenetic, and sociodemographic factors. Current proposed efforts aimed at understanding the potential genetic, epigenetic, and metabolic basis for susceptibility to the more common exposures, such as tobacco, are of paramount importance in clinical and translational medicine.
Maternal Behavioral and Developmental Outcome
Decades of research on maternal behavior and psychological state have provided evidence for an association between transmission of maternal behavior and characteristic of the offspring. For example, licking and grooming behavior in rodents has been characterized as maternal behavior toward pups within the first week of life (Francis et al. 1999; Liu et al. 1997; Weaver et al. 2004a, 2004b). The adult offspring born to low licking and grooming mothers demonstrated enhanced stress reactivity (Francis et al. 1999; Liu et al. 1997; Weaver et al. 2004a). It has been established that maternal care affects expression levels of neuron-specific GR by altering DNA methylation (McGowan et al. 2008).
Differential DNA methylation has been characterized in candidate genes involved in brain and cognition development. Schizophrenic patients are known to have differential DNA methylation of genes involved in the GABAergic system, serotonin system, dopaminergic system, and brain-derived neurotrophic factor (BDNF) gene (reviewed in Roth et al. 2009). Glutamic acid decarboxylase, an enzyme involved in GABA synthesis, is decreased in schizophrenic patients (Akbarian et al. 1995; Heckers et al. 2002). Understanding the histone modifications in schizophrenic patients is a less-studied arena; however, seminal studies have demonstrated histone modifications in schizophrenic patients compared with healthy individuals (Akbarian et al. 2005). Increased HDAC1 activity has also been observed in schizophrenic patients (Sharma et al. 2008). The synapsin 3 gene promoter has differential DNA methylation in humans, cynomolgous monkeys, and baboons but was not found to be associated with schizophrenic condition in humans (Murphy et al. 2008). In infant rhesus macaques, it has been demonstrated that higher CpG methylation of 5-HTT (serotonin transporter) but not the rh5-HTTLPR promoter polymorphism is associated with exacerbating effects caused by exposure of early life stress on behavioral reactivity in rhesus infants (Kinnally et al. 2010).
Depression morbidity rates have increased in the past decades. Dysregulation of DNA methylation of genes such as BDNF has been observed in depressive patients (Angelucci et al. 2005). Epigenetic regulation of the BDNF gene has been shown in fear learning (Lubin et al. 2008). Maternal mood and mental state is related to fetal brain development. Prenatal exposure to maternal depression leads to methylation of the GR gene in neonates and infant cortisol stress responses (Oberlander et al. 2008). Substance abuse during pregnancy is known to affect fetal development and epigenetic programming. Repeated exposure to cocaine is known to affect histone modifications (reviewed in Maze and Nestler 2011; Suter and Aagaard-Tillery 2009.
In terms of emerging arenas of likely paramount interest in understanding fetal programming, studies on the establishment of the perinatal microbiome are emerging as potential high impact. By way of overview, the microbiome (our so-called second genome) is associated with states of health and well-being throughout our lifetime. To date, the dominant paradigm in Western medicine has been to consider the majority of microbes as “foreign” and has led to the prevailing view that the treatment of overtly dominant microorganisms will result in disease cures. Such a view seemingly ignores long-standing observations that humans are remarkable hosts to microbes that have, in fact, coevolved as highly plethoric communities (Srinivasan et al. 2009; Turnbaugh and Gordon 2009; Xu and Gordon 2003). Human-associated microorganisms are present in numbers exceeding the quantities of human cells by at least 10-fold beginning in the neonatal period, and the collective genome (the microbiome) exceeds the human genome in terms of gene content by more than 100-fold. Scientists have recently come to appreciate that the microbiota are a metabolically and antigenically vibrant diverse community that may function as mutualists (symbiotically beneficial), commensals (of neither harm nor benefit), or pathogens (of host detriment) (Ley et al. 2005; Ley, Hamady, et al. 2008; Ley, Lozupone, et al. 2008; Turnbaugh et al. 2007, 2009; Vael and Desager 2009).
Although current major national research efforts (i.e., the National Institutes of Health Road Map initiative known as the Human Microbiome Project) will enable sequence-based comprehensive characterization of the adult human microbiota (Turnbaugh et al. 2007), how and when these diverse microbiota communities take up residence in the host and how they vary in reproductive life are unexplored at a population-wide level (Ley et al. 2005, 2006; Ley, Hamady, et al. 2008; Ley, Lozupone et al. 2008; Vael and Desager 2009). In other words, although researchers may soon know what constitutes the adult human microbial guild, how it is established will not be known. Moreover, the impact of the microbiome on the development of common perinatal disorders, such as preterm birth, will also not be known. Characterization of the pregnancy and neonatal microbiomes will allow a number of crucial gaps in understanding to begin to be addressed. For example, does this core microbiome change in association with the mode of delivery or timing of birth? What is the impact of the environment on the microbiome, and what is the impact of the microbiome at birth on the lifelong health status of the individual? Can microbial specimens acquired longitudinally from a human perinatal population base be interrogated to analyze microbiomes with efficient and reliable sequencing and computational methodologies in the future?
In an attempt to characterize the core gut microbiome, studies have demonstrated varying effects of diet, ethnicity, and physiological state on the core microbiome of an individual. An established link is present between the maternal immune status and the development of the fetal immune system (Figure 3). Other investigators working in a murine model of colitis have shown that maternal methyl donor supplementation increased susceptibility to colitis in offspring, and this occurred in association with both an altered microbiome in the offspring and epigenetic modifications in key immune response genes (Schaible et al. 2011). Intrauterine growth restriction also affects the offspring intestinal microbiome and leads to persistent alterations in the developing offspring gut microbiome (Fança-Berthon et al. 2010). Obesity has also been linked to human microbiomic composition and short-chain fatty acid metabolism (Dzielinska et al. 2010). Maternal diabetes is also linked to changes in microbiome composition in the offspring, although future studies are critical to corroborate the association of maternal diet versus diabetes and the microbiome and immunity of the offspring (Kaplan et al. 2011; Vassallo and Walker 2008). All of these data should be taken with a note of caution because a definite causal relationship has not been established between variations in the microbiome profile and lifelong risk of obesity and diabetes due to confounding factors such as ethnicity, age, diet, sex, and other lifestyle factors (reviewed in Musso et al. 2010). Undoubtedly, significant data will emerge in coming years and shed light on the essential emerging role of the metagenomics community in the establishment of lifelong metabolic health.
In summary, we have argued that in accordance with the developmental origins of adult disease hypothesis, perturbations in the in utero environment influence the development of diseases later in life. This was first observed to occur in response to maternal nutritional constraints, which resulted in growth-restricted infants who later were observed to develop profound changes, including obesity, insulin resistance, hypertension, heart disease, and lipid disorders. Working in animal models, multiple and converging lines of evidence have shown that these lifelong disease risks occur through the static reprogramming of gene expression by epigenetic modifications to the regulation of gene expression.
The field of epigenetic effects on reproduction and development is young but robust. Scientists have discovered a few of the earliest clues in this complex epigenomic code, with many more left to be unraveled. Primates as a whole and NHPs in particular will continue to serve as apt models to accurately identify the crucial temporal, spatial, and developmental landscape alterations in the epigenome. The employment of recent advances in metagenomics-based research to now adapt “gene, epigene, germ, and environment” interactions into our emerging understanding on what shapes the fetal program will be integral to defining the role of gene and environment in shaping this landscape.
This work was supported by a National Institutes of Health Director New Innovator Pioneer award (DP21DP2OD001500-01 to K.M. Aagaard) and the Burroughs Wellcome Fund Preterm Birth Initiative 1008819.01. We thank the members of the Aagaard laboratory—Dr. Melissa Suter, Dr. Min Hu, Ms. Lori Showalter, Ms. Cynthia Shope, and Dr. Aishe Chen—for helpful discussions regarding the manuscript.