Extensive experimental animal studies and epidemiological observations have shown that environmental influences during early development affect the risk of later pathophysiological processes associated with chronic, especially noncommunicable, disease (NCD). This field is recognized as the developmental origins of health and disease (DOHaD). We discuss the extent to which DOHaD represents the result of the physiological processes of developmental plasticity, which may have potential adverse consequences in terms of NCD risk later, or whether it is the manifestation of pathophysiological processes acting in early life but only becoming apparent as disease later. We argue that the evidence suggests the former, through the operation of conditioning processes induced across the normal range of developmental environments, and we summarize current knowledge of the physiological processes involved. The adaptive pathway to later risk accords with current concepts in evolutionary developmental biology, especially those concerning parental effects. Outside the normal range, effects on development can result in nonadaptive processes, and we review their underlying mechanisms and consequences. New concepts concerning the underlying epigenetic and other mechanisms involved in both disruptive and nondisruptive pathways to disease are reviewed, including the evidence for transgenerational passage of risk from both maternal and paternal lines. These concepts have wider implications for understanding the causes and possible prevention of NCDs such as type 2 diabetes and cardiovascular disease, for broader social policy and for the increasing attention paid in public health to the lifecourse approach to NCD prevention.
I. INTRODUCTION: THE DOHaD CONCEPT
This review is concerned with understanding the physiological and pathophysiological basis for how environmental influences acting during early human development influence the risk of later chronic, especially noncommunicable, disease (NCD). This field of biomedical science and public health has become recognized as the developmental origins of health and disease (DOHaD) (142, 211). The focus of this review is to consider the extent to which early conditioning (Table 1) mechanisms in humans may represent the physiological processes of developmental plasticity (Table 1) operating in early life, but having potential adverse consequences later, or whether they are the result of pathophysiological processes acting in early life but manifesting as disease later in life. We will argue that the evidence supports the former concept. The majority of this evidence comes from animals, but there is increasing evidence from human physiology which suggests that the concepts have wider relevance. In addition, they accord with an emerging understanding of the principles of evolutionary developmental biology (evo-devo) (Table 1) and evolutionary medicine (196, 423). We will also discuss how disruptive processes during development can also lead to later disease, especially if they are novel from an evolutionary point of view. These concepts have significant implications for understanding the epidemiology of NCDs such as type 2 diabetes and cardiovascular disease and hence for their prevention. While the end result of these processes is disease in the modern world, the understanding of the underlying biology which is necessary if we are to devise preventative measures includes not only proximal pathophysiological mechanisms but also more ultimate (Table 1) mechanistic physiological considerations. Insights into these can be found in the broader biological fields of evo-devo (196, 646), and more specifically that of maternal or parental effects (Table 1) (382a, 417, 599).
While the concept of “parental effects” is widely accepted for lower animals and some plant species, a wide range of animal models have confirmed that such processes also occur in mammals, and this has allowed exploration of the underlying physiological mechanisms (for reviews, see Refs. 48, 264, 338, 399). Parental effects are generally considered to operate within the normative range of environments for a species and when the effect on the next generation may well have adaptive value (176). Some influences in one generation may however have nonadaptive (Table 1) effects on the next, and this is more likely in mammals where exposure of the pregnant mother to an environmental agent may have teratogenic effects on her developing offspring. This distinction between potentially adaptive parental effects, which are common across taxa including mammals, and the clearly nonadaptive consequences of developmentally disruptive conditions, which are primarily a feature of mammals, is important but is often not considered in interpreting the experimental and comparative literature. In focusing on this dimension, the present review both builds on previous reviews that have described the mechanistic insights arising from animal studies, generally with specific reference to the development of obesity, insulin resistance, and cardiovascular disease (223, 369, 398) but also develops concepts related to ultimate causation which emerge from evolutionary considerations and which may be more widely applicable. We do not rehearse in detail the physiological processes discussed in earlier reviews, but focus on recent advances in the elucidation of underlying developmental mechanisms, particularly epigenetic (Table 1) processes. This integration of different fields shifts the focus of our understanding of the processes underlying DOHaD away from pathophysiology and towards physiology, and provides the rationale for a greater focus on life-course perspectives in relation to the prevention of NCDs.
NCDs, in particular diabetes, cardiovascular disease, chronic lung disease, and some forms of cancer, account for ∼63% of all deaths globally (9, 659). Eighty percent of these NCD-related deaths occur in low and middle income countries (LMICs), particularly as such countries undergo socioeconomic transition following reductions in the prevalence of communicable diseases. The greatest burden of these diseases is currently in Asia, but in the near future, the prevalence will rise in sub-Saharan Africa and parts of Central and South America. The risks of NCDs are exacerbated by Western lifestyle, urbanization, migration, and factors associated with greater economic prosperity such as nutritional changes and sedentary lifestyle, all of which produce a “mismatch” (Table 1) between human evolved biology and our contemporary habitat (205, 212). In addition, demographic changes leading to increased longevity and smaller family size are shifting the age distribution of many populations upwards (e.g., Refs. 53, 524), increasing the relative burden of such chronic diseases further. These issues are central to the consideration of public health policies in the post-2015 era (600), which will need to build on the continued efforts aimed at meeting the Millennium Development Goals (662).
The potential economic costs of NCDs are predicted to be greater than those of communicable diseases, perhaps equivalent to those of climate change, and approaching those of a global fiscal meltdown (657). The rising global challenge of NCDs prompted the special high level meeting of the United Nations General Assembly in September 2011, at which the importance of developmental processes was recognized, since “. . .maternal and child health is inextricably linked with NCDs and their risk factors. . .” (clause 26) (601). This statement represented a substantial change in public policy focus from the earlier WHO Global status report on NCDs (659) reflecting the impact of rapid progress in addressing a number of the conceptual and scientific issues discussed below.
One of the most important risk factors for NCDs is obesity or overweight, and the problem of the rising burden of NCDs is often conflated with that of obesity via such neologisms as “globesity” (658) or “diabesity” (146). It is however important to recognize that there are marked population differences in the relationship between body mass index (BMI) and the risk of developing NCDs such as type 2 diabetes mellitus (T2D): people of South Asian ancestry have about three times the risk, and those of Chinese ancestry about twice the risk, of developing T2D at a high BMI compared with people of Caucasian ancestry (94). These differences in risk can be seen even at lower BMI.
The currently dominant scientific and public health model for NCDs assumes that they are a consequence of genetic predisposition, coupled with adult lifestyle choices that are largely under voluntary control. However, attempts to reduce the burden of NCDs by promoting weight loss and changing adult lifestyles have been relatively unsuccessful at a public health level. Beyond cultural, behavioral, and public policy issues, there are physiological reasons why maintenance of weight loss is difficult. For example, neuroendocrine drivers to sustain hunger and promote food intake persist following enforced weight loss (573). While attempts to promote a less obesogenic environment must not be discouraged, it is illogical to ignore the importance of underlying physiological processes.
If the processes producing propensity to obesity and insulin resistance are established early in life, then interventions in adults are likely to come too late to be very effective (209). Indeed, as Taubes (584) has pointed out, the physiological basis for obesity, in terms of the physiological control of metabolism, appetite, etc., has been overlooked in recent years in favor of a simple energy balance concept: viz. that obesity arises from excessive caloric intake in relation to daily energy expenditure. Again, this is too simple a view.
The DOHaD concept emerged over the last two decades, although there were many historical precedents for it which were overlooked, partly through declining interest in integrative, and especially developmental, physiology during the latter part of the 20th century as molecular biology became dominant (204, 206). There are a range of historical reasons for the slow acceptance of the concept (see Table 2) and, as they form themes which run through this review, we note them at the outset.
The first issue was a lack of a conceptual framework. The NCDs which first drew the attention of researchers in the field were coronary heart disease, hypertension, and T2D. The dominant paradigm was that these diseases, which develop gradually over many years in adult life, were the result of lifestyle choices which led to cumulative damage to, for example, the blood vessels, as a result of repeated exposure to high levels of glucose, cholesterol, or low-density lipoprotein (LDL)/high-density lipoprotein (HDL) ratio, nicotine, and carbon monoxide (CO) from tobacco smoke, etc. This model assumed that all individuals start life with the same risk, but diverge only according to their later lifestyle. The physiological basis for individual or population-based differences in susceptibility to lifestyle challenges was little addressed. It was however appreciated that the risk of such disease had a heritable component in the broadest sense, at least in terms of ethnicity or family history of disease (577) but that a substantial part of this was explicable simply as shared environmental risk. There is limited animal data to support this concept, although it has been shown that genetically distinct strains of rodents show different susceptibility to features of the metabolic syndrome induced by maternal dietary restriction in pregnancy (321).
The rise of human genomics was associated with a widely held conviction that such variation in risk was explicable in terms of genetic predisposition, specifically small mutations producing single nucleotide polymorphisms (SNPs), gene copy number variation, or other fixed genomic variation. Whilst some SNPs associated with risk of NCDs have been found, those associated with large effects in individuals are uncommon in the population as a whole; the search for more common variants with smaller effects continues in larger cohorts and via wider screening across the exome. The “missing heritability” of chronic disease is still widely viewed as mysterious (376), but it is now recognized that the early optimism about finding the genetic basis for chronic disease was misplaced (303). Moreover, SNPs such as those in the FTO gene known to predispose to adiposity show either no (409) or a complex relation with BMI in childhood (559), so they do not easily account for the life-course pattern of risk. As simply summarized by Watson (635), the contributions of inherited alleles, diet, and levels of exercise are hard to determine even in common diseases.
As workers in the DOHaD field moved in the last decade from the study of simple associations between early life events and subsequent pathology to investigation of underlying mechanisms, new concepts that are central to this review emerged. The processes we appreciate to underlie DOHaD are now seen to operate at the interface between genetic and developmental environmental influences, largely through the resurgence of interest in epigenetic processes which provide a proximate (Table 1) explanation. Even so it required an evo-devo perspective to put this in context (30, 32, 214, 220, 221) and to place the empirical observations within a coherent framework. We explore this further in sections IIIC and VIA.
The second reason for the slow acceptance of the DOHaD concept was that the epidemiological observations of an association between early life attributes, in particular birthweight, and later chronic disease (23, 27) (see below for full discussion), led to the erroneous assumption that such disease processes were initiated during development, when in fact epidemiological observations of this nature can only show correlation, not causation. The intense attention given to consideration of how low birth weight might lead to disease (23) was thus misplaced; indeed, it led to consideration of low birth weight as necessarily lying on the causal pathway. The relatively low incidence of low birth weight in many populations showing high prevalence of NCDs is evidence that, if it exists, the role of such a putative pathway is relatively minor.
The epidemiological observations indicated not only that NCD risk in later life was affected by development, but also that this risk was graded across the entire range of early developmental experiences, even when captured by the relatively crude proxy of birthweight or weight at 1 yr of age. Of key importance, but generally overlooked, was the finding that this included infants whose birth size was unequivocally within the normal range. Recent studies confirm the graded nature of the phenomenon, and for a wider set of processes such as neurocognitive development (623). Other studies have clearly shown that the long-term effects associated with more subtle changes in the fetal environment such as variation in maternal nutrition can be independent of birthweight (185, 229).
As research into what was termed until 2003 the “fetal origins of adult disease” proceeded, competing theories arose about the pathways involved (197). While this led to new insights into developmental effects on tissues, organs, and systems, the concept that such effects form part of an integrated physiological response of the organism to developmental signals was not much discussed. For example, a “mismatch” experiment in which pregnant rats were fed a restricted diet and their offspring fed a high-fat diet was shown to affect central nervous components such as appetite control and voluntary exercise capacity as well as peripheral components including body composition, liver metabolism, and cardiovascular function (611, 612). In addition, there is striking consistency in the adult phenotype, involving multiple physiological systems, in a variety of animal models where the prenatal or early postnatal environment was altered by a range of nutritional or hormonal manipulations (397).
The widespread adoption of the term programming did not help to gain acceptance of DOHaD, because its deterministic implications of “programming of disease” or “programming of function along a pathway” are reminiscent of the genetic program for development (285). This slowed acceptance of the importance of the normative, holistic nature of developmental plasticity and its role in affecting sensitivity to later environments (204). As the essence of the DOHaD concept lies in the induction (Table 1) of phenotypic changes, usually within the normal physiological range, which permit altered responses to later challenges, usually also within the normal physiological range, we prefer the use of terms such as “priming” (Table 1) (68), “induction” (29), or “conditioning”: the echo in the latter of the concept of conditioned reflexes (96) or of conditional growth based on predicted later nutritional or other conditions is not unhelpful in this respect.
From the time of early exposition of the “fetal origins of adult disease” concept, the lack of plausible biological mechanisms limited acceptance of the idea (125). The long latency of the effect, from prenatal life to presentation of adult disease over possibly more than 60 years, was challenging from a mechanistic point of view. In many ways, it was not until the advent of developmental epigenetics (see sect. IIIE) that this particular challenge was addressed (219). Even so, this has until very recently limited the application of DOHaD concepts within clinical medicine and public health and led to misinterpretation of the underlying conceptual model: we will return to this point in section VI, D and E.
In this review, we first discuss the current state of knowledge on the epidemiological and clinical physiological observations that form the basis of DOHaD. Second, to review new understanding of how developmental physiological processes can predispose to the pathophysiology of NCDs, as opposed to inducing pathology in early life as implied by earlier theories, we will consider DOHaD processes from a lifecourse perspective (Figure 1). Linked to this is the issue of the ultimate versus the proximate causation of such effects, and this is considered in the context of evolutionary and developmental biology. Based on this, we go on to distinguish between developmental effects that are potentially adaptive (i.e., physiological; sect. IV) versus those that are nonadaptive (pathophysiological; sect. V). We use a strict evolutionary definition of “adaptive”; adaptive attributes are the result of processes that have been selected during evolution since they confer an advantage in terms of survival to reproduce for the individual, more strictly therefore of Darwinian fitness (Table 1) The adaptationist argument can limit understanding if it becomes merely teleological (319): we aim to avoid this, but do not discuss its limitations in detail here (for more extensive discussion, see Refs. 217, 220).
Adaptive effects of relevance to DOHaD involve the physiological processes of developmental plasticity and broadly include the categories of immediately adaptive and predictive or anticipatory adaptive responses (Table 1). They can lead to greater risks of lifestyle-associated disease through the occurrence of “developmental mismatch” later in the lifecourse (Figure 2). Nonadaptive processes often involve an element of evolutionary novelty, either because of the nature of the challenge or because it operates at a level which in humans exceeds the experience of the hominin lineage during its evolutionary history; in both instances, defenses have not evolved to meet the challenge completely. Nonadaptive processes can therefore lead to an increased NCD risk even without later mismatch. Examples that we discuss include those associated with maternal obesity, gestational diabetes, Caesarean delivery, infant overfeeding, and exposure to toxins, pollutants, or tobacco smoke. Finally, we discuss the implications of this broader understanding of the role of development and the opportunities which it offers for preventing NCDs and promoting public health (see sect. VI, D and E).
II. EPIDEMIOLOGICAL AND CLINICAL PHYSIOLOGICAL EVIDENCE FOR DOHaD
A. Conditioning by a “Poor Start to Life”
We cannot date when the concept that early life influences produce long-term effects on health and disease was first appreciated, but the idea is referred to in the writings of the ancients (e.g., Hippocrates). More recently, Freud (183) noted the importance of early life in his theories of psychosexual development and was aware that even aspects of fetal behavior might set the scene for later mental conditions. Critically, a link between poor living conditions in childhood and later premature mortality was made in 1934 by Kermack et al. (311), and this idea was extended by Forsdahl (171) who reported that a poor childhood environment was associated with adult cardiovascular disease, even when the adult environment was not poor, thus focusing attention on development itself as the initiating period. Dörner and his group (see Refs. 143, 473 for review) studied a range of developmental effects, in particular hormonal influences on sexual behavior, which became controversially applied to issues such as homosexuality, but his work also extended to developmental effects on later obesity, atherosclerosis, and diabetes. Dörner was the first to use the term programming (“programmierung”) to describe these effects (323), a term adopted later by Lucas (370), Barker (22), and many others (Table 1) (398). Metaphors, while frequently used in biology, can be problematic if they restrict the challenging of preconceptions (32, 177, 416). The metaphor of “programming” implies that the complete instructions to carry out a task, e.g., a developmental strategy, are defined from the outset: this accords more with a computer program than the processes of developmental plasticity. In addition, once running, a program either continues to completion or crashes. This ballistic model, in which development proceeds to a preset destination, is too deterministic to describe the observations accurately. As will be discussed, the effects of different developmental trajectories induced by environmental cues can change the magnitude (222, 611) and, indeed, sometimes the direction (222) of a response to a later environment.
Freinkel (182) introduced the term fuel-mediated teratogenesis to describe how prenatal nutrition might produce effects on subsequent development. While this provided an important insight by indicating how nutrition might cue alterations in the development of metabolic homeostasis, as will be discussed in section IIIA, the long-term effects of nutritional exposures within the ecologically normal range during development cannot be viewed as teratogenic, i.e., disruptive to the developmental trajectory. Such terms are not viewed as helpful today.
Experimental studies by van Assche and colleagues in the 1970s (4, 130) first highlighted the long-term consequences for metabolic control in rats which had experienced developmental challenges (264), but there was a delay of more than a decade before the broader importance of this work was appreciated.
The most influential early epidemiological observations in the DOHaD field concerned the cardiovascular system. In 1980, Higgins et al. (260) reported an association between pregnancy complications such as preeclampsia and the blood pressure of the offspring as they grew through adolescence: this effect was only seen in the index pregnancy, it became more exaggerated as the offspring aged, and it persisted after correction for mother's blood pressure. Higgins et al. (260) concluded that prenatal environmental, rather than genetic, effects were therefore involved and suggested that future research should involve the prospective study of women and their children. Interestingly, the effect was more apparent in male children, and sex differences are now commonly seen in both experimental and clinical studies of DOHaD (see sect. IVB). This prescient study, which received little recognition, thus predates much later work in the DOHaD field. Wadsworth et al. (621) subsequently reported an inverse association between birth weight, parental social status, and systolic blood pressure in young men and women, and an inverse association between size at birth and later diastolic blood pressure was reported for males by Gennser et al. (192). The effects may appear small, but elevations of blood pressure are known to track from childhood into adulthood and to predict hypertension (20).
The greatest impetus to the emergence of the DOHaD concept came from the series of epidemiological studies by Barker1 and colleagues from 1986 onwards (25–27, 452) in which links were established between low birthweight, weight at 1 yr of age, and higher prevalence of hypertension, coronary heart disease, stroke, and metabolic syndrome in adulthood. Studies which followed confirmed the association between low birthweight and a range of cardiovascular disease markers or mortality (173, 350, 385, 500, 546). Barker's studies arose from the observations that people born in areas of high perinatal mortality, where poverty, poor diet, overcrowding, and infection led to poor development and early growth, had increased risk of death from coronary heart disease as adults and that, as in Forsdahl's observations, the effect persisted even in those who subsequently moved to more affluent areas. These observations challenged the widely-held belief that conditions such as coronary heart disease are solely the result of affluence although, as discussed below, affluence in adult life may contribute to greater risk of disease through “mismatch.”
The relative importance of the developmental component reported by Barker and colleagues was challenged, for example, in a meta-analysis of available data on the relation between birth weight and adult blood pressure (275). This was countered by studies showing that the link is stronger when the outcome is clinical disease rather than a risk factor such as blood pressure (114, 115). Much of this confusion related to the emphasis on birth weight, largely because this was the only proxy available to assess prenatal development. In addition, the older cohorts, such as those which Barker studied initially, came from an era when maternal obesity, gestational diabetes, and fetal macrosomia were uncommon so the low birth weight relationship seemed clear. Similarly, survival following extreme preterm delivery or growth restriction was poor, so conflicting effects at this end of the spectrum from a range of causes were also removed. As younger cohorts were added, the picture became more complex because maternal obesity and gestational diabetes became important influences on NCD risk in the offspring, but these are associated with higher birthweight. As will be discussed in section V, these effects probably involve different mechanisms and have different ultimate explanations.
This early epidemiological work was given support by studies of the consequences of historical events such as war, albeit that they maintained the focus on extreme changes in the developmental environment rather than on more ecologically normal situations. These studies are not always easy to interpret as wartime conditions produce changes not only in nutrition but also in many other factors. Additionally, the timing of the developmental exposure is likely to be important. During the Dutch Hunger Winter of 1944–1945, food supply to the Western part of the Netherlands was cut dramatically (with reported daily intake falling to 400–800 calories). Individuals who were in utero, at least in the later parts of pregnancy during this famine, had a low birth weight and a decreased glucose tolerance at the age of 50 compared with individuals born the year before or after the famine (489). Subsequent studies of this population are inconsistent in relation to whether fetal exposure to famine was associated with elevated blood pressure in adulthood (131, 507, 565). In a study of adults born before, during, or after the civil war in Nigeria (1967–1970), an increased risk of hypertension was reported in those exposed to war-time famine during fetal and early postnatal life (273). In contrast, there was no significant difference in blood pressure between adults who were exposed in utero to malnutrition during the siege of Leningrad in World War II versus those who were born at the same time but outside the area under siege (563). A further example is provided by the Great Famine from 1959–1961 in China, where one study showed an association between hypertension following exposure to famine during fetal development (361), whereas another study did not (270). More recently, Wang et al. (628) reported that the risk of hypertension was 1.36-fold higher in those exposed to this famine during only the first trimester of pregnancy. The risks were 1.83- and 1.31-fold higher in those exposed only during infancy and in those exposed during both fetal development and infancy, respectively. One of the problems with such studies is the assumption that the inducing cues are those of malnutrition, when clearly the population was exposed to considerable stress and often had substantial changes in the constitution of their diet as well as its quantity; for example, the Dutch Hunger Winter diet included unusual items such as tulip bulbs which may have contained toxins. A further distinction between these famines is that, whereas the Dutch famine was acutely imposed on a previously well-nourished population and then rapidly relieved, this was not the case in the other recent historical famines, which emerged gradually and took many months to be resolved.
Notwithstanding the debates about the relative roles of early life (as measured by birthweight) and adult lifestyle in the etiology of NCDs, and the rapid increase in the prevalence of NCDs in many populations, the heritable risk of NCDs has continued to be viewed as largely genetic. Substantial differences between populations in the relation between measures such as waist-hip ratio or BMI and prevalence of diabetes are well-known (e.g., Refs. 94, 395). Such variations have been assumed to be due to fixed genetic differences. The original conceptualization of this was Neel's (421) proposition of “thrifty genes,” derived from his studies of the Pima native American people living in the states of New Mexico and Arizona, who have a very high incidence of glucose intolerance and T2D even in early adulthood. Until the mid 20th century their lifestyle involved subsistence farming in a harsh environment, so Neel (421) proposed that genes associated with metabolic “thrift,” in particular for insulin resistance, had undergone positive selection during presumed historical periods of famine. Confronted with the more abundant and high glycemic index diet of the contemporary United States, such alleles would confer greater risk of glucose intolerance and diabetes. Neel's concept was widely accepted, although more recently it has been subject to considerable critique (e.g., Ref. 333). As will be discussed below, fixed genetic variation which depends on ancestry is not the sole explanation for transgenerational (Table 1) heritable influences on risk of disease, and there is increasing interest in the potential for both direct and indirect epigenetic inheritance as mediators of such influences.
More widespread acceptance of the DOHaD concept occurred following the plethora of studies which replicated the association between birth size and adult disease, for ischemic heart disease (325), stroke (500), insulin resistance and T2D (425, 504). The adult phenotype associated with lower birthweight includes central adiposity and low skeletal muscle mass (310) reminiscent of the thin/fat neonate found in India, a country with a high prevalence of both low birthweight and T2D (666), and there is now substantial evidence for DOHaD from many populations (223). In addition, very large cohorts have been used to replicate the concept (46, 188, 304, 325, 344, 500, 608). Lastly, whilst most epidemiological studies in the DOHaD field have concerned the metabolic syndrome, or its components of cardiovascular disease or diabetes, and have relied on birth weight or infant size as a proxy index of the developmental exposure, there have been associations found for low birthweight and reduced adult bone density (251), schizophrenia (427), and asthma and atopic conditions (564). Some of these observations, for example, for schizophrenia, have also been linked to the studies of famine (562, 575). There are also however associations between high birthweight and some forms of breast cancer (403).
As we have intimated above, the focus on low birth weight distorted understanding of the underlying phenomena, and it has only been more recently that attention has shifted from extremes of birth weight to considerations of how and why conditioning can operate across all pregnancies and with birth weights across the normal range. While the epidemiological studies of Barker and colleagues (23, 25, 26) showed clear long-term effects on disease risk associated with birthweights within the normal range, the implications of this have often been ignored. In considering the effects of the fetal environment for babies with birth size within the normal range, the processes of maternal constraint, which operates in all pregnancies, are relevant (217, 218, 245). Maternal constraint involves a set of uteroplacental mechanisms by which fetal growth is restricted from reaching its genetic potential, necessary to permit successful passage through the pelvic canal at delivery. Maternal constraint is greater in mothers of short stature, in adolescents, and in primigravida, as the pelvic canal is smaller in such pregnancies (215, 607). Two recent studies have now shown that, in terms of perinatal survival, optimal birthweight is considerably higher than the mean birthweight for the population (180, 607). Mechanistically, even some appropriately grown fetuses show signs from Doppler ultrasonography of being challenged (410), although whether this is the consequence of maternal constraint is not known.
Thus in recent years a shift has been made from the study of the long-term effects of extremes in birthweight to that of cohorts of apparently normal pregnancies. For such studies, accurate measures of gestational age are necessary, as both growth and length of gestation can be influenced by environmental factors that can make the most widely used methods of ultrasound dating inaccurate (295). Effects of maternal diet and body composition within the normal range of fetal growth have been shown on liver blood flow, ductus venosus shunting, and fat deposition in the fetus (313). There are additional effects on hepatic (151), celiac, and splenic (149) arterial hemodynamics and on the portal venous (150) and the intrahepatic circulation (312). There are a few studies (e.g., Ref. 227) on the postnatal sequelae of these fetal patterns of development, but more are needed. Mechanistic studies, for example, those of alterations in vasculogenesis in low birthweight infants (364), need to be extended to cover the normal range.
B. Conditioning by Overnutrition
It has only relatively recently been realized that the DOHaD concept also operates in the context of a prenatally or neonatally excessive or rich nutritional environment (13, 481, 638). Early evidence of this comes from the 1946 United Kingdom birth cohort (247), from the Pima native American population (465) and from the meta-analysis of the Nordic cohorts (188) which show a U-shaped relation between birth weight and blood pressure (Figure 3). More recently, both maternal obesity and gestational diabetes mellitus (GDM) have been implicated in leading to a later risk of obesity and cardiometabolic disease risk in the offspring (343, 498). There are also reported associations between excessive weight gain in pregnancy and offspring blood pressure and adiposity (181, 191, 379).
A range of epidemiological studies have shown effects of maternal prepregnancy BMI on offspring's blood pressure or autonomic function (156, 169, 446, 645). Maternal obesity is associated with increased risk of pregnancy complications including hypertension, preeclampsia, and GDM, and a child of an obese mother is also more likely to become obese (85, 649) and to develop asthma and atopic disorders such as dermatitis (248, 349, 470). Offspring's adiposity at age 4 is related to mother's BMI across the entire range from very thin (BMI <18.5) to obese (BMI >40) (649). While the data are less robust, there is the suggestion that infant feeding with formula (which tends to be more calorie-dense and for which intake can be more directly driven by parental behavior than by infant demand) rather than human breast milk also conveys a greater risk of later obesity (12, 568, 616).
An increasing concern is the transmission of risk of diabetes across generations. An increasing number of women develop T2D during their reproductive years (168, 493) and children of type 1 or 2 diabetic mothers are more likely to become obese and to develop diabetes (479, 541). The effect on the offspring is related to intrauterine exposure to hyperglycemia because, among siblings, the risk of diabetes is greater in those born after the mother developed diabetes (116). In turn, the incidence of GDM is rising rapidly in many countries, particularly in Asia (168, 374), and the offspring of gestational diabetics have been shown to have both a greater risk of developing obesity (540) and insulin resistance. While the focus here has been on the clinical condition, again the data show evidence of effects occurring across a range, as even offspring exposed to mild hyperglycemia during pregnancy have increased lean and fat mass and greater risk of diabetes (246, 578). The problem of “diabetes begetting diabetes” is thus of major concern (374, 667).
III. HOW ORGANISMS RESPOND TO THE DEVELOPMENTAL ENVIRONMENT
A. What Is “Normal” Development: Disruption Versus Plasticity
The term plasticity (Table 1) has different connotations in different contexts. We favor Malabou's (377) distinction between plasticity, which involves a change in phenotype from one condition to another, versus resilience in which the phenotype does not change in the face of a certain range of challenges, and flexibility in which a challenge produces some change but a return to the original state when the challenge is removed. The concept of developmental plasticity is based on the understanding that the phenotype and genotype do not have a fixed relationship and that the phenotypic attributes of individuals are affected by developmental processes (32, 646). Such differences exist even between monozygotic twins (560) and in asexually reproducing species isogenic clones can show environmentally induced variation, for example, in crayfish genetically identical individuals show variation in body markings, growth, and reproduction (614). Such developmentally induced phenotypic variations are normative, and the adaptive advantage of the conservation across evolution of the processes of developmental plasticity has been discussed extensively (499, 671). While there may be some stochastic elements, particularly within species with very high reproductive rates such as asexually reproducing bacteria (40), in the context of more slowly reproducing mammalian species such as the human the emphasis needs to be on the relationship between environmental influences and the consequential phenotypic change which, as will be discussed below, has directional components of potential adaptive value. Indeed, the importance of including the life history strategy of the species in considerations of developmental plasticity cannot be overemphasized (382a). The nature of developmental responses to environmental cues is greatly influenced by the reproductive strategy of the species (382a) and, in particular, on the effects on parent or offspring fitness. With this in mind, in the following discussion we focus only on responses relevant to the human.
While environmentally induced developmental plasticity is an evolved and normative process, many external environmental influences can be said also to “disrupt” development (30, 193, 221). Such use of the word disrupt is appropriate when the agent produces a somatic mutation with phenotypic consequences, or leads to a toxic effect that disrupts the normal pattern of development. Many teratogens are known: most are evolutionarily novel substances such as thalidomide or synthetic steroids, others are infective agents such as rubella, and still other disruptions involve gross deficiencies of micronutrients such as folate or gross excesses such as of vitamin A. For each of these examples we can assume that they were sufficiently uncommon in the evolutionary past for the hominin lineage not to have evolved effective defenses against the developmental disruption they can cause. Likewise, the loss of the ability to synthesize vitamin C in primates is not associated with detrimental antioxidant effects when the micronutrient is plentiful in the diet (44). It can however be argued that phenotypic accommodation (Table 1), i.e., phenotypic responses to the disruption, can allow survival and thus have some adaptive (Table 1) potential (32).
The distinction between developmental plasticity and developmental disruption is not just semantic (207). It has important conceptual and mechanistic implications, viz. that plasticity is an evolved response aimed at enhancing fitness, while disruption is a passive response that in general impairs fitness. However, some cues can exert either a plastic or a disruptive influence depending on their magnitude (30) in a dose-dependent manner. For example, mild hyperglycemia in utero changes the rate of growth and has adaptive potential (see below), but more severe hyperglycemia can disrupt cardiac development leading to congenital heart anomalies (510), and derangement of maternal lipid profile can also lead to increased risk of congenital heart disease (552).
B. Development in Individuals and Populations: Inheritance and Heritability
In moving from consideration of individuals to populations, the concept of variation in phenotype becomes important, because this is one of the necessary prerequisites for Darwinian (both natural and sexual) selection. It was soon realized that the variation in most characteristics does not follow Mendelian rules of inheritance (see Provine 1971 referred to in Ref. 468), so the underlying processes must be more complicated than simply random allocation of genetically determined characteristics. Selection operates on the phenotype, not the genotype, but it is the latter that changes through evolution. The range of phenotypes which a particular genotype can produce in a particular environment is termed the “reaction norm” (Table 1) (654).
It is important to remember that inheritance is a much broader concept than genetic inheritance. The traditional breeder's equation of population genetics assumes independent genetic and environmental factors (39), and measures of heritability attempt to describe how much variation in a particular trait in a population is accounted for by genetic variation. But such measures depend on the environment in which the population resides (Figure 4) (357). For example, height has a lower heritability in populations where childhood nutrition is variable, leading to environmentally induced stunting in some children who are not well nourished. If the genetic determinants of heritability are low, then classical models of selection have less explanatory power with respect to the evolution of the phenotype. These arguments and others have led to the concept that the multiple processes of plasticity itself may be subject to the processes of evolution (337) and that they evolved because they confer a fitness advantage (32). While beyond the focus of this review, there are also some data showing that, under some conditions, plastic changes in the phenotype can become genotypically fixed (assimilated) and thus incorporated into evolutionary processes (32, 646).
For much of the past 100 years it has been assumed that there are only two kinds of inheritance: genomic and cultural. However, in recent years there has been an explosion of interest in three other forms of inheritance. The first concerns maternal (or paternal) effects, the subject of much of this review, by which the environment in one generation can condition the phenotype of the next, and possibility subsequent, generations. The second, related concept concerns indirect epigenetic inheritance whereby maternal effects induce epigenetic changes in the offspring of the next generation and in turn these epigenetic effects recapitulate the maternal phenotype which induced them; thus the offspring themselves induce the same maternal effects in the next generation, and so on (218). This scenario is most well demonstrated in studies of female rats with high versus low grooming behavior towards their offspring (637) where the indirect nature of the inheritance has been demonstrated from cross-rearing experiments. Third, recently a body of evidence has suggested the potential for trans-meiotic epigenetic inheritance (see Ref. 283 for review) including that from experimental studies in rodents (239, 240, 428) and compatible human observations (462, 463).
In the context of this review, the transgenerational transfer of the gut microbiome and its potential longer term consequences illustrate the importance of taking a broader view of inheritance. Mice reared in a germ-free environment are protected from dietary-induced obesity (18). Breast-fed and formula-fed infants have different patterns of gut microbiota, and while the data has been challenged (326), there is a body of evidence suggesting that breast feeding is associated with a lower risk of obesity. There is also a relation between how the gut is colonized depending on the mode of delivery (vaginal/caesarean) (598) and different disease risks in children. Those born by caesarean section have a higher rate of types 1 and 2 diabetes (81) and allergy.
C. Developmental Plasticity and Canalization: Developmental Constraints Versus Sources of Phenotypic Variation
There are multiple processes of developmental plasticity acting at various levels of organization to produce phenotypic variation. Equally there are multiple mechanisms restricting developmental variation, and so inducing robustness (Table 1) in the phenotype. These should not necessarily be seen as opposing processes as they reflect integrated mechanisms to optimize the outcomes of development from an adaptive perspective (for review, see Ref. 32).
There are two important points regarding how these processes work together to enhance fitness. The first is the relative robustness of the developmental “program,” such that individuals with phenotypic traits of proven evolutionary fitness reproduce this reliably in each generation despite variations in the individual developmental conditions. The second is the ability to respond to changes in such conditions within the normal range, to produce a degree of phenotypic variation through developmental plasticity. Just as for the first process, the second confers a fitness advantage and has thus been preserved through evolution. As the fitness advantage is reflected in the ability to survive to reproduce, it can be manifest at any stage of the lifecourse up to peak reproductive potential, starting from early development. Whyte (650) coined the term internal selection to explain how effects on the early embryo would be more likely to have dramatic effects than later challenges and so some might be deselected, although at that time the idea was couched in terms of the lethal effects of mutations.
Of the mechanisms seen as conferring robustness, the best recognized is that of canalization, which permits maintenance of a phenotype in the face of variations in the environment or genotype. Canalization is often associated with the theoretical and empirical work of Conrad Waddington, who coined the term epigenetics in about 1942. His well-known epigenetic landscape (Figure 5) shows not only how developmental processes are canalized to buffer against variations in genotype and certain developmental exposures (the steepness of the walls of the valleys representing the degree of canalization) but also that epigenetic processes can induce switches in a developmental pathway, for example, to induce different cell types from a pluripotent stem cell line, or at the organism level to induce a different morph in a polymorphic species.
From a physiological point of view, it has been proposed that canalization (Table 1) utilizes processes controlled by factors including heat shock protein (HSP) 90, at least in some species (407), and HSP90 has been suggested to act as a “capacitor” to alter the degree of buffering of phenotypic effects (483). In rodents, the transgenerational passage of changes in the expression of the metabolic genes phosphoenolpyruvate carboxykinase (PEPCK) and increased PPARα and carnitine palmitoyltransferase-1 expression, induced by a sustained switch to a higher energy diet, were accompanied by changes in HSP90 expression and epigenetic changes (70). HSP90 acts a chaperone for a wide range of genes and can modulate effects of steroid hormones (512) that are known to be involved in developmental conditioning.
The term genetic accommodation is sometimes used to reflect the processes of phenotypic canalization that act to reduce the impact of randomly occurring genetic mutations (646). Such buffering may prevent their affecting the phenotype, but they can remain in the genotype as cryptic genetic variations (195) that may become manifest in later generations under different circumstances. For example, in the classic experiments of Belyaev using the wild silver fox, where animals were selected artificially on the basis of reduced fear of humans, cryptic variation in a number of phenotypic features appearing early in development, such as coat color as well as head and ear shape, was exposed (596). Presumably the selection on aspects of familiarity with humans had epistatic (Table 1) effects reducing the degree of canalization of coat color and other morphology in the offspring.
D. The Temporal Dimension of Developmental Responses to Environmental Change
Developmental plasticity can also be placed in the context of different time scales of environmental change (measured with respect to the intergenerational period of a species) (332) (see Figure 6). The most rapid response of an adult organism to a change in its environment constitutes physiological homeostasis, operating to achieve constancy of the milieu interieur (235). This operates over a time scale of seconds to hours and involves changes in metabolism, cell signaling, and neuroendocrine responses. At the other temporal extreme, natural selection acts to change the genotype of a lineage in response to a sustained shift in the environment acting over many generations. Developmental plasticity occupies an intermediate position between these two extremes. It allows an organism to change its phenotypic development from that of its evolved lineage in response to an experienced or anticipated environmental change, but where the change is neither consistent over generations nor sufficiently prolonged to induce evolutionary change. The presence of maternal effects allows such plastic responses to persist over more than one generation but not to be sustained indefinitely in subsequent generations if the environmental shift is not permanent. The fitness advantages of such an intermediate response to environments which fluctuate over a longer time base than the generation time is apparent (334).
E. Nature Versus Nurture: The Rebirth of Epigenetics
The dichotomy of distinct influences inherent in the “nature versus nurture” phrase was first put in place by Charles Darwin's cousin Francis Galton (187). It was such concepts that paved the way for the eugenics movement in the first part of the 20th century and which continued into the post-war years (243). Following the rise in the science of genetics, “nature” became associated with heritable genetic traits and “nurture” with environmental or cultural influences, especially those operating during childhood. Unfortunately, the debate persists even today, with widespread belief that it is possible to separate the genetic from the developmental environment component of many traits, including disease susceptibility (308). Increasingly, our understanding of development shows the artificial nature of the distinction and the limitations of attempting to separate genetic and environmental influences, because they interact continuously across the lifecourse.
Furthermore, the concept of the gene as a physical entity in the process of the Central Dogma (111) is now seen to be flawed; indeed, “gene” now has varied and quite different definitions (415). Genes can be defined as segments of DNA, that is as structural elements, or alternatively as functional entities, although this depends on what definition of function is used (32, 416). The problem has become evident most recently in the current debate over the interpretation of the ENCODE data (152), in the sense that “intergenic” regions formerly thought to be “junk” DNA with little functional importance were shown to be potentially expressed, even if their functions remain unknown. In turn, the functional operation of a gene, and indeed its definition according to some interpretations, involves considerable input and influence from the environment. As Fox Keller has pointed out (308), the idea that the two processes of nature and nurture are distinct is neither helpful nor correct. The ways in which integrative physiological thinking is challenging the fundamental tenets of neo-Darwinism as enshrined in the Modern Synthesis (274) has recently been elegantly described (433). In some respects, this indicates a return to Bateson's definition of genetics as “the physiology of descent” (35).
The essential role of development in phenotypic variation was widely recognized by the early 20th century Russian school of biologists, as exemplified by the writings of Schmalhausen (523). Unfortunately, these were appropriated, misapplied, and then suppressed by Lysenko. In the Western world it was the work of Waddington (619), perhaps influenced by Schmalhausen, that placed similar emphasis on this concept. Waddington introduced the term epigenetics at a time when the underlying molecular mechanisms were unknown. It was only with the more recent understanding of the molecular basis of epigenetics that the importance of these early scholars' work was recognized.
There is now a plethora of reviews on the part played by epigenetics in phenotypic development (228). The majority of studies have focused on DNA methylation acting at CpG base pairs although there is growing interest, especially in cancer biology, in effects on the covalent modification of histone protein tails, which affects DNA compaction (74), and on the role of noncoding RNAs (400). One source of confusion is the very different biological roles which epigenetic mechanisms play, even leaving aside their role in the pathological (Table 1) processes of cancer; for example, during evolution the epigenetic mechanisms which control gene expression under normal circumstances appear to have been coopted to regulate gene dosage (497). Thus the term epigenetics can encompass very distinct biological processes including transposon silencing, cellular differentiation (essential for the evolution of metazoa), chromosomal silencing in sex determination, and the imprinting of alleles depending on parental origin which evolved in marsupials (496) but is also found in plants (389). Epigenetic inheritance has been shown over eight generations in Arabidopsis (293). Between the appearance of metazoa and marsupials, the same biochemical pathway has been coopted to serve several quite disparate functions. Adaptive developmental plasticity, which is present in species as widely different as Daphnia and humans, and is the physiological mechanism of central interest for this review, is another distinct class of epigenetic function (216).
DNA methylation and hydroxymethylation affect gene expression by altering the access of transcription factors, which may then either augment or repress transcription, or through promoting the binding of the methyl CpG-binding proteins to methylated cytosines which recruit histone-modifying complexes to the DNA. These include histone acetyltransferases (HATs), which add acetyl groups to the histone tails and make the chromatin transcriptionally active, and histone methyltransferases (HMTs) such as Suv39H, which methylates lysine 9 on H3, and produces a closed chromatin structure and silencing of transcription. Most studies have examined such methylation across large regions of the genome, sometimes called differentially methylated regions (DMRs) (167), or at major clusters of CpGs called CpG islands and shores. However, these different patterns of CpG distribution have distinct biological roles that are only now being elucidated (570), an example being the role of DMRs in determining cell differentiation (318).
Recent evidence however stresses the importance of examining methylation at individual CpGs and patterns of methylation in small genomic regions (229). Methylation of CpGs de novo is catalyzed by DNA methyltransferases (Dnmts) 3a and 3b, which remethylate parts of the genome after the widespread demethylation that occurs following fertilisation (412). The pattern of DNA methylation is then maintained through mitosis by methylation of hemimethylated DNA by Dnmts. DNA methylation patterns induced during development were originally thought not to change greatly later, except in epimutations associated with cancer (138), but are now known to be fundamental to the processes of developmental plasticity. In monozygotic (MZ) twins, epigenetic variability increases with age (655), and this change is affected by lifestyle (178). DNA methylation differences are apparent even at age 5 yr in MZ twins (655). Thus, while the epigenome is most flexible from the perspective of developmental plasticity in early life, some degree of flexibility extends well after birth.
There is now evidence that environmental cues can alter patterns of DNA methylation in the embryo (316), and this is of relevance to the observations of an apparent increased incidence of imprinting disorders in children conceived using assisted reproductive technology (ART) (110, 133, 316, 451; for review, see Ref. 604). There is also evidence that ART can lead to long-term phenotypic effects in the offspring including greater risks of obesity, insulin resistance, and hypertension (250).
In addition to the processes of methylation which produce 5-methylcytosine (5mC), there are also well-regulated demethylation processes, for example, TET (ten-eleven-translocation) which catalyzes conversion of 5mC to 5-hydroxymethylcytosine (5hmC) (279, 663, 664) which may be an intermediate in the removal of 5mC. As 5hmC levels across the genome are low, it may be short-lived, or it may act as another epigenetic modification, inducing chromatin and transcriptional modifications. The distribution of 5hmC relative to 5mC (664), and its presence at exons strengthens the idea that it is associated with gene expression.
There is a high degree of specificity in the processes of DNA methylation despite the small number of methylation-related enzymes involved. It is assumed that specificity is conferred by microRNAs acting in concert with these enzymatic processes, but this is poorly elucidated. A further complication is the phenomenon of allele specific methylation whereby a SNP may impact on the pattern of methylation in a CpG cis to that SNP (538). The underlying mechanism may relate to the capacity for that SNP to affect small RNA production. It is thus important to use advanced bioinformatic techniques to identify methylation changes that are responsive to the environment and additionally those that are secondary to a fixed genetic mutation.
Some epigenetic processes, and particularly those related to developmental plasticity, are susceptible to nutritional influences, in part because processes such as DNA or histone methylation require a source of methyl groups from 1-carbon metabolism. These are not usually in short supply in the diet and can be derived from the nonessential amino acid glycine. However, the demands for this in the late gestation human fetus are such that glycine becomes a conditionally essential amino acid. Animal studies show that the phenotypic effects on the offspring of a maternal low-protein diet can be prevented by supplementation with glycine (64, 284). Folate is also required for the provision of methyl groups, and this is largely derived from the maternal diet. Folate plays a major role in many aspects of fetal developmental (606) pathways via epigenetic mechanisms (238), and imbalance can lead to the accumulation of homocysteine, which has damaging effects on many organs including the kidney and cardiovascular system and which can act via increased oxidative stress (641).
In rodents, folic acid supplementation of the low-protein diet fed to the dam prevents epigenetic and phenotypic effects on her offspring (365), and paternal dietary folate level produces epigenetic effects in sperm and affects pregnancy outcome (336). Provision of methyl group donors in the diet also prevents the induction of epigenetic effects via retrotransposons in the agouti mouse (630). In humans, dietary supplementation with folic acid is known to reduce the risk of neural tube and other “congenital” defects (127), and the balance between folate and vitamin B12 status is linked to diabetes risk (668). The critical period when epigenetic changes can be reversed by folic acid supplementation is not known, although in rodents it may extend into the postweaning prepubertal period (72).
Nutritional influences on the developmental epigenetic state are however broader than specific influences on one-carbon metabolism. Taurine has long been considered to be an important factor (5), and recent studies implicate its deficiency in endoplasmic reticular (ER) stress (36a), which can have a range of downstream effects, including on sirtuins and ageing (92). Given that epigenetic mechanisms underpin developmental plasticity and that nutritional cues are important with respect to adaptive developmental plasticity, it is to be expected that a variety of nutritional cues in development will have specific epigenetic consequences. Most experimental studies of nutritional effects on the offspring have used an unbalanced maternal diet with a low protein-to-carbohydrate ratio or a global dietary restriction during pregnancy. However, unbalanced diets with high levels of dietary fat or cafeteria diets simulating the Western “junk food” diet have also been explored (38, 39).
Epigenetic effects on the developing offspring can also be induced by altered maternal behavior. Offspring of rat dams that exhibit high levels of licking and grooming were shown to have lower levels of methylation at specific CpG dinucleotides within the hippocampal GR promoter, concomitant with higher expression of the gene product which regulates hypothalamic-pituitary-adrenal activity (636). These offspring were better able to cope with stressors in adulthood, demonstrating a functional link. Maternal care-dependent gene expression changes in the hippocampus could be reversed by intracranial administration of the histone deacetylase inhibitor Trichostatin A and l-methionine, which acts as a methyl donor (637).
As Ozanne has pointed out (e.g., Ref. 147), the regulation of transcription factors by epigenetic processes is an effective way of altering the developing phenotype via manipulation of physiological control processes: examples include effects on RXRA (229), Pdx1 (458), and Hnf4α (518).
Prenatal epigenetic effects are not confined to the developing fetus. Some effects are mediated by epigenetic processes in the placenta, for example, at the level of amino acid transporters (352) and imprinted genes that in some species affect fetal growth (177, 294). There may also be effects on glucose transporters (298). The interaction between fetal growth and placental vascular function via the NO system appears also to be mediated in part by epigenetic processes (328).
From a physiological point of view, it is now important that a range of processes have been discovered which affect gene expression and thus phenotype beyond structural mutations such as SNPs. In addition to the processes discussed above, we should note gene copy number variations, the occurrence of which in the offspring can also be affected by maternal diet (75). Evidence that epigenetic processes can affect the probability of mutations through the hypermutability of CpG sites should promote greater acceptance of the previously heretical notion that physiological processes can affect the genotype; this idea has implications not only for developmental but for evolutionary biology (434). Finally, it is now becoming clear that studies of developmental conditioning will require measurement of the interaction between genetic and epigenetic processes.
F. Developmental Strategies
As Hertzman has summarized (259), there are several ways of thinking about DOHaD phenomena. In practice, they are hard to separate. One, which accords with the concept of critical periods in development (see also Ref. 330), proposes latent effects whereby early life environment produces effects on later health and disease risk independently of intervening environmental influences. This has similarities with aspects of neural development, such as the visual system, but less so with cardiometabolic control. The second involves pathway effects, whereby the response to a cue at a particular point in time depends on the developmental pathway leading to that time, and this is currently the most favored explanation. The third possibility uses a cumulative effects model for the detrimental consequences on health of repeated and graded exposure to adverse environmental conditions. While this concept is incorporated into current DOHaD thinking (see Figure 1), on its own it does not differ from the damage accumulation/adult lifestyle model that does not account for the demographic patterns of disease. For example, although the prevalence of obesity has continued to rise in most countries over the last years, there has been a steady decline in global mean blood pressure, and interventions to reduce blood pressure, glucose, and cholesterol account for only 50% of the excess risk of coronary heart disease (201).
The phenotype emerging from the interaction between genetic and environmental factors, and mediated via epigenetic processes, will be adaptive if aspects of the phenotype induced affect either the individual's survival until reproductive age, or reproductive capacity and fecundity. Strictly, fitness is measured in terms of the number of viable offspring which themselves reach reproductive age, or better the number of resulting grand-offspring. Simple examples can be found in egg-laying species including fish, birds, and insects, where environmental factors such as nutrient availability or number of predators affect both the number of eggs laid and their size, which in turn influences the survival chances of the offspring (155). We can see that there may be a potential conflict here between maternal use of food resources for metabolism and investment in the fitness of the next generation (this concept of maternal-offspring conflict is discussed in sect. IVD). In addition, mechanisms which evolved to confer fitness may not have the same effects in contemporary conditions (34, 591).
The role of maternal effects or, where there is also an influence of the father, parental effects and their evolutionary significance has been reviewed (382a, 599). They can be quite dramatic in terms of the offspring phenotype induced. East African grasshoppers that are normally green develop into a black morph after a grassland fire (509). Another excellent example is the desert locust (Schistocerca gregaria). This species has two distinct phenotypes (or morphs): solitary and migratory. They differ dramatically in terms of behavior, with the shy, nocturnal, solitary form contrasting with the gregarious, diurnal, migratory form; in nutritional physiology, with the solitary form having a restricted range of food plant species whilst the migratory form is far more omnivorous; and in metabolism, with only the migratory form being capable of long flight. The morph that forms following larval hatching is influenced by the environmental conditions. A high population density affects the composition of the secretion produced by the mother around the eggs, and it is this chemical stimulus which influences the phenotype, inducing the migratory morph which is more likely to move to a less well-populated area (392). An equivalent mammalian phenomenon has been reported in the red squirrel, in which high population density and low food abundance produces a glucocorticoid response in the mother which is associated with accelerated growth in the offspring (120), permitting them to reach reproductive age more quickly.
The direct role of nutrition as an inducer of phenotype in the developing offspring is exemplified by the honey bee (Apis mellifera) where the maturation of the female larva into a queen versus a worker can be produced by feeding it exclusively on royal jelly for a longer period: the active component in the royal jelly responsible for this induction is not known, but the polyphenism is mediated by epigenetic mechanisms as it can be altered by use of an siRNA to suppress Dnmt3 activity (329).
It is not known whether such polyphenisms are relevant to mammals. We have argued that there is some evidence in the rodent for developmental bifurcations in response to maternal nutrition which might suggest an echo of distinct morphs (222). Furthermore, the distinct forms of response of children to severe undernutrition, which have different adaptive consequences and which appear to originate prenatally, suggest that some aspects of human physiology do not follow a continuous reaction norm but may be discontinuous, reminiscent of polyphenism (170). However, in mammals, most developmental plastic responses do produce a continuous reaction norm (522), for example, in linear growth, birth weight, gestational length, or metabolic function. The upper and lower limits of that reaction norm are limited by various tradeoffs; for example, gestational length is limited by viability of the offspring at one extreme and by placental senescence and the ability of the fetus to pass through the pelvic canal at the other.
G. Predicting the Future
The concept that phenotypic traits formed during development, and which confer fitness advantage in some way will, if they have a heritable component, be selected during evolution is uncontentious. Similarly the concept that plasticity itself is an evolved and conserved process is generally accepted. But a relatively new concept concerns the role of developmental anticipation, i.e., that the developing organism can detect critical aspects of the current environment and use these as cues for the induction of a developmental path which can potentially confer adaptive benefit later. For example, frog larvae exposed to chemical signals suggesting the presence of a predator will develop into tadpoles with tail morphology allowing greater swimming speed and thus greater chance of survival (393). That this might confer a later life disadvantage is traded-off against the adaptive advantage of the changes to tadpole morphology because they provide a greater chance of survival to reproduce.
Such anticipatory responses are common across all biological taxa and can span generations. Other well-known examples are the helmet in Daphnia, the shell morphology of the whelk, and the thicker body of the crucian carp that form in response to signals of predation (66, 141). Studies in Daphnia show the critical importance of the developmental stage in determining the phenotypic outcome (408). The example given earlier of the desert locust indicates the use of an early cue to predict later overcrowding and thus the development of a morph able to migrate to a more favorable environment. The examples differ in that in Daphnia the cue is passed directly from the environment to the developing individual; in the locust, it is mediated by the mother when she lays her eggs. For mammals, the role of the mother both before birth and in infancy is even more important as a transducer of the current environment. Licking and grooming behavior of the rat dam, which may convey information about her stress level, has prolonged effects on the stress responses of her pups (95). Belding's ground squirrels developing in open grassland habitats, where the risk of predation is higher, develop into more anxious adults than do those developing in woodland habitats (386). Similar effects of parental stress levels on the offspring are seen in arctic species such as the lemming and snowshoe hare (535, 536). A mammalian example that resembles the responses of the desert locust is reported in the North American red squirrel, where greater population density leads to an offspring phenotype characterized by faster growth and earlier puberty. That the effect is induced by population density detected by the mother is shown by its occurrence when recorded calls of squirrels in dense populations are played (120). The effect may be mediated by maternal glucocorticoids (see sect. IVC) as their metabolite levels in feces are increased. Maternal effects may integrate a range of environmental aspects, as well as affecting a range of phenotypic characteristics in the offspring (134).
We have termed the class of developmental response that underpins these adaptive anticipatory traits predictive adaptive responses (PARs) (Table 1) (210, 220). There is growing support for the concept from a range of species, (e.g., Ref. 513; see also Ref. 33). The prediction does not need to be perfectly accurate for the capacity to respond in anticipation to have evolved: provided the prediction is more often correct than incorrect, the capacity to predict will have adaptive potential (204, 282). Several studies have modeled the adaptive advantage of a predictive response in early development and shown that, despite the considerable potential for error in the prediction, it confers advantage as long as there is some inertia in the cue, i.e., that the environmental change is interpreted as persisting over a significant portion of the lifecourse (411, 572). Nettle et al. (424) have recently subdivided PARs into external, based on environmental cues, and internal, based on an organism's functional phenotype. We argue that this is an artificial distinction and that “internal PARs” are equivalent to immediately adaptive responses (IARs). The suggestion that PARs are only effective when the environment is relatively stable is countered by the model of del Giudice et al. (134), which takes account of shorter and longer term environmental effects with varying levels of autocorrelation. However, detailed analysis of the stability of the epigenetic processes underlying PARs requires consideration of their costs (256), following Nishimura's (431) discussion of the optimal delay between an environmental cue and the induction of a phenotype.
There is likely to be a directional bias in the maladaptive effects of an inaccurate prediction. Prediction of condition A but being exposed later to condition B need not have the same consequences as the reverse, and thus there may be an advantage in evolving mechanisms which favor one set of predictions over another. As we will discuss below, a good example is that there is a greater fitness disadvantage in humans predicting a plentiful nutritional environment but ending up in a famine than the reverse: the former leads to an individual who is poorly adapted to restricted nutrition and reproductive function, at least in the female, is likely to be compromised. On the other hand, even though a woman who is better adapted to a low nutrition environment may become obese if she lives in a high nutrition environment, her fertility will be compromised only at extreme levels of such obesity (297).
We recognize that the use of the term predictive may seem teleological; however, as with the terms forecasting or anticipatory, it is not intended to imply a conscious decision but it does capture the importance of the response, hence why it may have evolved.
PARs differ from the thrifty phenotype model of Hales and Barker (241) in several ways. First, the thrifty phenotype concept rested on the immediate advantage to the developing individual of economizing on nutrient consumption by reducing growth: this is more akin to an IAR. If there were unintended detrimental health consequences later, this was unfortunate. In contrast, PARs do not necessarily need to have any IAR nature and, indeed, can be induced by stimuli that do not of themselves necessitate a developmental response and that are not necessarily accompanied by a reduction in growth: PARs are induced to gain a fitness advantage in the postnatal period until the time of reproductive competence. Furthermore, some attributes of a PAR can be the opposite of “thrifty,” for example, a greater initial insulin sensitivity in early postnatal life (401) which may promote adipose deposition, greater postnatal muscle deposition especially in males, greater appetite, and accelerated body growth. In addition, a range of behavioral attributes may be affected, which confer advantage in the predicted environment, including appetite (611) and physical activity. Just as larvae of the butterfly Bicyclus anynana develop stronger thoracic muscles if poorly nourished, enabling them to fly to potentially better habitats (513), so offspring of undernourished rat dams will be more sedentary when kept in standard cages (612) but will run greater distances than offspring of well-nourished dams if given a wheel in the cage (406). These induced responses are not easily viewed as thrifty.
It is important to note that not all responses of a mammalian fetus which have phenotypic consequences are predictive in nature: where the developmental cue is evolutionarily novel or extreme, it may induce a disruptive response (see above); where it is severe but not disruptive, it may induce an immediate “coping” or adaptive response (IAR), usually a reduction in growth which allows the fetus to survive (221); and lastly, the effect on the fetus may be neutral in terms of both fetal survival and longer term effects, depending on the circumstances. Intrauterine growth retardation in response to maternal undernutrition or impaired placental function is a good example of an IAR because, unless the fetus slows its somatic growth, the high energetic demands of the fetal heart and brain cannot be met. The effects of reduced nutrition or an environmental cue mediated by glucocorticoids to reduce nephron number is however an example of a PAR. In humans and sheep, complete complement of nephrons is established before birth, and the relatively low metabolic requirements of the kidney may not confer a great survival advantage from reducing this number before birth. Postnatally, however, the kidneys receive 25% of the cardiac output, so reduced nephron number may be adaptive in a poor nutritional environment. The longer term deleterious effects on cardiovascular function of reduced number are not manifest until later (19). PARs can occur independently of IARs or both may be induced, and this is why fetal size often correlates with longer term phenotypic outcomes that are associated with disease risk, even though the processes which set the risk may differ from those which restricted growth and growth may not be only the causal pathway to later risk.
The distinction between IARs and PARs may seem somewhat semantic but is important because some critics of the PARs concept (643) have failed to appreciate the conceptual significance of the clinical observations which show that long-term consequences for the fetus can occur in the absence of IARs (i.e., in the absence of reduced birth weight). If the long-term phenotypic changes associated with later disease risk in the offspring were simply a by-product of a maternal adaptive response, then an IAR would be expected always to be demonstrable: this is not the case for such effects occurring in the absence of an effect on birth weight (185).
Two questions become important in considering the role of adaptive plasticity in the pathways to altered disease risk in humans: the first is when in the lifecourse the adaptive advantage is demonstrated, and the second concerns the causes and consequences of faulty prediction. Both of these have been subject to some confusion from some commentators.
Williams (651) recognized that the pressure generated by natural selection declines with age; survival to reproductive age and fecundity are particularly important in defining human fitness (299). Indeed, the well-accepted evolutionary concept of antagonistic pleiotropy (652) is based on the selection for traits that have value early in life even if they have postreproductive deleterious consequences. Thus, in seeking the adaptive advantage of putative PARs, one would expect to see any advantage manifest in childhood and/or adolescence, not necessarily in adulthood. Indeed, PARs only have to give survival advantage into the next stage of the lifecourse after they are induced to have adaptive value; for example, if they are induced early in fetal life, their potential impact may be on infant and child survival. This key point was not emphasized in the earliest expositions of the PARs hypothesis (204).
A range of studies in humans has shown that maternal body composition affects a range of characteristics in the offspring, including body composition, neurocognitive behavior, kidney size, and reproductive function (223, 413, 551). However, a striking example of a PAR in human comes from a comparison of Jamaican children who develop marasmus or kwashiorkor in childhood famine. Kwashiorkor is more metabolically wasteful, and such children have a higher mortality (169a); they have a higher birthweight than children who develop marasmus in the same community (170), suggesting that their prenatal conditions were better, leaving them less well prepared to meet famine. The more economical metabolic phenotype in the marasmic children was not only associated with greater chance of survival, but was manifest as insulin resistance and glucose intolerance in adulthood (179).
Misleading predictions can occur either because the environment changes over the lifecourse or because cues are inaccurate or misinterpreted. Considering the nutritional environment as an example, maternal ill health, placental dysfunction, or unbalanced maternal diets affecting nutrient transfer can all signal a poor nutritional environment, when in fact the nutritional environment for the population as a whole is adequate or even abundant. Conversely, infant overfeeding with formula can signal a rich environment when in reality it is more nutritionally constrained.
We have already introduced the concept of maternal constraint, and this has important implications for how nutritionally induced PARs operate in humans. In humans, maternal constraint mechanisms generally limit fetal growth to match maternal (i.e., pelvic) size and their evolutionary significance has been discussed elsewhere (215). In polytocous species, the operation of maternal constraint may be linked to the limits imposed by multiple fetuses. One important component may be the saturable nature of placental amino acid transporters (100), another the clearance of insulin-like growth factors by the placenta (28, 104). But, in contrast to nutrients such as amino acids and fatty acids, there is no limit to the capacity of glucose to cross the human placenta to the fetus, where it drives insulin release and in turn promotes both somatic, and particularly adipose, tissue growth. Hence, in untreated gestational diabetes mellitus, fetal overgrowth leading to potential obstructed labor may occur. That humans did not evolve a limit on placental glucose transfer suggests that maternal obesity and GDM were uncommon in our evolutionary past (375). Type 1 diabetes and GDM can be associated with changes in placental glucose transporters, although this appears not to be so in late gestation (289).
Thus we can see two distinct kinds of “mismatch” operating with respect to developmental conditioning, one based on postnatal evolutionary novelty (evolutionary mismatch) and one a consequence of the qualitatively different nutritional signals operating between fetal and postnatal life (developmental mismatch). The modern diet of high glycemic index and energy-dense processed foods is qualitatively very different from the range of diets that humans consumed until recently (478). The modern diet exceeds the range that can be predicted in terms of developmental conditioning because of its evolutionary novelty, and thus a mismatch is likely between the diet predicted in utero or in infancy and that which the individual is likely to consume later. Such evolutionary mismatch can also have consequences for the development of the next generation as for maternal obesity and gestational diabetes. There is some evidence for the operation of a PAR induced by abundant nutrition in utero, which may make the individual less able to cope with a poor environment later. There is anecdotal evidence for lower survival rates in larger individuals in concentration camps (29), and in Ethiopia higher birthweight was associated with greater risk of rickets (88).
H. Transgenerational Effects
The evolutionary significance of the adaptive pathways by which environmental information is passed across generations via the phenotype has been discussed above. We and others have emphasized nutrition as the most obvious example, and examined how epigenetic changes are amplified or modified across generations in the face of a sustained nutritional challenge (70, 358). But there are also a range of other cues that can induce transgenerational effects.
One may be the parental influences on the developing offspring produced by behavior. Animal data for this come from research showing that the licking and grooming behavior of rats produces effects on the behavior of their offspring, an effect mediated via epigenetic changes in the hippocampus of the pups (368) involving the GABA system (677). As cross-fostering in infancy abolished the effect, this can be considered an example of indirect epigenetic inheritance (218). In the wider context, the effect has similarities with niche construction (445). In humans, a range of behavioral attributes are passed between generations and, while it has been generally considered that these are culturally transmitted, the animal observations also raise the possibility of epigenetic effects. Indeed, a recent and somewhat controversial study has suggested that suicide victims who had been abused as children have the same epigenetic changes in their hippocampus as the low groomed rats (394). Such epigenetic dysregulation has also been observed in cord blood from neonates born to mothers who experienced depression late in pregnancy (443). At age 3 mo, assessment of salivary cortisol levels following exposure to a stressor showed that these infants had heightened HPA stress responses. This is an area requiring careful research across generations to resolve the extent to which human behavior may be conditioned by prior generations or by biological processes in utero or infancy.
Female-line transmission from F0 to F2 generations could be produced by an effect on the primordial ovarian follicles developing in the fetus, as they contain the maternal genetic information which will pass to the grand-offspring at the time of fertilization. There is growing evidence that such effects can be mediated by epigenetic processes (672). When such a process operates, a stimulus to the F0 generation may produce effects on the F2 offspring phenotype but should not affect F3 or subsequent generations unless the environmental cue is repeated during the development of these generations. There are several examples of such transmission (70, 144). For the process to be termed truly transgenerational as opposed to intergenerational therefore, it should be manifest in the F3 generation without the environmental cue persisting into the F1 generation (547).
There is growing interest in the intergenerational transmission of effects via the male line (236, 272, 670), although it is unclear how prevalent such effects might be. However, it is now clear that some DNA methylation and chromatin marks can persist in the sperm (84) and that the small amount of cytoplasm remaining in the sperm contains microRNAs that have been shown experimentally to have phenotypic effects (488). For example, injection of RNA from sperm of mice heterozygotic for the Kit paramutation which results in a white spotted phenotype into wild-type embryos led to their developing the spotted phenotype (487). Injection of a miRNA associated with cardiac hypertrophy into embryos led to transmission of the effect to the F3 generation (622). Additionally such transmission appears to involve DNA methylation (317). Dramatic male line transmission effects have been shown for chemical exposures, in particular to endocrine disruptor chemicals such as vinclozolin which have anti-estrogenic and pro-testosteronic properties. Administration of these chemicals to male rats can produce effects on sperm counts and markers of testicular cancer up to the fifth generation, mediated by the male line (11). Manikkam et al. (380) recently showed that endocrine disruptors such as BPA, DEHP, and DBP derived from plastics can induce epigenetic transgenerational inheritance of conditions such as obesity and reproductive disorders and that these are accompanied by sperm epimutations. In mice, fear responses to a chemical stimulus, conditioned by an electric shock, are passed from male rats to their offspring even when the dam was unexposed (136). In rats, dexamethasone prenatally administered to males leads to metabolic effects in offspring (144), and paternal obesity induces pancreatic dysfunction in female offspring (426). In the latter study, myriad gene expression changes were detected in offspring islets; the largest change was seen for a gene encoding a component of the Jak-Stat signal transduction pathway and was coincident with hypomethylation at its putative regulatory region. Paternal prediabetes in mice leads to disrupted glucose metabolism in offspring that persists to at least the F2 generation, and induces an array of methylome alterations in offspring pancreatic islets (639). Most interestingly, many of these epigenomic changes were reflected in paternal sperm. In humans, fathers of low-birth-weight infants are more likely to have metabolic disease (262), while a poorer lipid profile in childhood has been linked to advanced paternal age (519). In an historial study of a Swedish rural population, poor nutrition before puberty in men was associated with lower incidence of diabetes or cardiovascular disease in their grandsons (302). Smoking in fathers in adolescence is associated with increased risk of obesity in their sons (463). Recent evidence (382) shows that smoking can affect miRNAs in sperm, raising the prospect that aspects of paternal behavior can have epigenetic effects on the next generation. Other mechanisms of transgenerational transmission via the male line may include changes in the packaging of DNA around protamine, a cue in the semen, because offspring of male mice lacking seminal vesicles are fatter (65), or even prions, which can induce heritable traits in yeast (242).
IV. PHYSIOLOGICAL (ADAPTIVE) PATHWAYS IN DOHaD
The discussion above distinguishes between ultimate and proximate mechanisms that are encompassed within the study of DOHaD-related phenomena. Lack of clarity over the multiple underlying mechanisms and pathways has slowed progress in this field.
The acceptance of the DOHaD concept was delayed to an extent by the lack of physiological insights into underlying mechanisms: this necessitated detailed animal investigations (360, 398, 479, 480, 495, 576) in which it is possible to define and control the stimulus used to induce the response, and to measure that response in a range of tissues and at various stages during development and throughout the lifecourse. The animal models used included the rat, mouse, guinea pig, sheep, pig, and non-human primate, and in each the DOHaD concept has been confirmed. However, while much of the early work focused on the physiological mechanisms, it was both the development of conceptual models based on evolutionary principles (221, 241) and the recognition of the role of developmental plasticity (30, 32) and the associated epigenetic processes (214) that gave the field a solid mechanistic basis and led to the acceptance of the concept of “mismatch” (469) (Figure 2). We emphasize that it is important to distinguish between mechanisms operating within the normal range of developmental exposure, which are likely to operate via adaptive mechanisms, and those associated with evolutionary novelty that are likely to be nonadaptive in origin.
Adaptive responses operate at multiple levels in the organism to produce an integrated phenotype (217). The levels are summarized in Figure 7. If the stimulus occurs during early development, the responses include effects on stem cell lineages that will in turn affect the growth of individual organs and tissues. They produce effects on mitochondria (3, 543, 586, 674) which affect metabolic function. Subsequent interactions between metabolic demands of tissues, the levels of growth hormones, and the interaction between blood supply and organ function then affect the relative growth of organs. Finally, they affect control systems such as stress responses involving the HPA axis and the sympathetic nervous system (474), hypothalamic function, and appetite (61, 480, 495, 501, 571). These effects can be viewed as changes in the speed of maturation of organs and systems in addition to changes in growth. The effectors operate on multiple pathways, for example, atrial natriuretic peptide has been shown to play a role in vascular and renal effects, possibly via alterations in placental perfusion (290), hormone production (418), cytokines (117), endothelin-1 (139), and the RAS following changes in AT1R expression (673). For many of these processes, changes in placental function are responsible for mediating the effects (418, 526). Because the physiological processes operating in such responses interact, effects on one system can be compensated for by another. This makes measurement of the phenotype complex, especially where there are multiple processes involved in regulating a function. It can also explain why some apparently straightforward interventions do not achieve the desired result, as noted earlier for energy balance considerations with respect to obesity (584).
Where the effects concern behavior, they may also have consequences for relatives or other members of the population. Where there are commensal species, such as the gut microbiota, these may be affected too, and this may then have further reciprocal consequences for the individual (7). The mechanisms thought to operate at these different levels are discussed in detail below.
B. Physiological Mechanisms: Effects
These are summarized in Figure 7.
Initially most animal studies involved investigation of effects of unbalanced or poor maternal diet and body composition on cardiovascular and metabolic function in the offspring (69, 481, 592). The most commonly used challenge was a low-protein diet fed to the mother during pregnancy (554). This produced low birth weight and later growth (unless the animals were cross-fostered) and produced effects on offspring blood pressure (6, 48, 429, 453), the heart (91, 162), and blood vessel structure and function (50, 64, 593, 669). There are also studies using global dietary restriction (253, 430, 611, 656) and low-salt diet (36). The effects on blood vessels differ between beds (595). They can be viewed as changes in the maturation of control of vascular tone which involve a shift in the balance between NO and endothelial hyperpolarizing factor (56) and which can be affected by a high-salt (55) or high-fat diet and which interacts with oxidative stress (438).
Dietary manipulation of postnatal growth in the offspring provided data that led to the mismatch hypothesis, since accelerated growth is associated with shortened lifespan (292, 454–456), accompanied by shorter telomere length in the kidney, pancreas, and aorta (580, 581) and reduced antioxidant protection (582). Unbalanced nutrition in utero affects a range of organs and systems, including brain metabolism and function (186, 612), appetite (38), the sympathetic nervous system (517), HPA axis (476), circadian rhythms (78), kidney (59, 398), pancreas (267, 583), liver (68, 441), skeletal muscle (108, 109), brown and white adipose tissue (372), and bone (340). Vascular effects on the offspring include endothelial dysfunction and effects on intraventricular wall thickness in the heart. There is considerable information about the processes involved in cardiomyocyte development because of its relevance to heart failure in adults. The maturation of the fetal heart in late gestation, triggered by glucocorticoids and thyroid hormone, changes the balance of the α and β isoforms of the myosin heavy chain (MyHC), the former being faster acting and therefore capable of greater force development (640); maturing cardiomyocytes become binucleated and no longer divide. Subsequent changes in afterload, as systemic arterial pressure increases at birth and with vascular disease in later life, produce hypertrophic thickening of cardiomyocytes. In heart failure, reexpression of fetal genes, mediated by epigenetic processes involving a cardiac-specific microRNA (605), can produce temporary compensatory processes.
In humans it is becoming increasingly apparent that long-term effects on health are induced in early gestation, measurable for example in terms of fetal growth in the first trimester and cardiovascular risk in childhood (286). Similar considerations apply to the preconception period (603).
Adaptive processes can be induced experimentally in the preconception and peri-implantation period via manipulation of maternal diet or body composition (76, 101, 102, 112, 595). In both rodents and sheep, such short-duration dietary manipulation preimplantation produces sustained cardiometabolic effects on the adult offspring (631). These effects can be induced in the embryo in vitro, subsequently implanted into the uterus of a recipient dam which is fed a balanced diet (632). The effects are manifest even if the dietary challenge is induced during oocyte maturation (634). The results in both large and small animals again support the PARs model (102, 633). The metabolic effects on the offspring are sex dependent (73) and influence postnatal growth (71). Similarly conditioning effects can be induced postnatally in rodents, for example, by changing litter size (620), and conditioning can be reversed by neonatal intervention (613). Distinguishing between the effects of unbalanced nutrition in early versus later pregnancy is not easy in the rat or mouse; however, in the guinea pig such effects have been shown (47) on both cardiovascular and stress responses in a manner reminiscent of the trimester-dependent effects of famine on those exposed in utero to the Dutch Hunger Winter (507).
As these developmental effects are part of an adaptive strategy (Figure 8) rather than the early inception of a pathological process, they reflect lifecourse biology theory and generate an integrated and somewhat stereotypic phenotype (218). From a physiological point of view, this phenotype has reductions in the size of metabolically active organs such as the kidney, giving the individual a reduced reserve capacity to cope with the naturally occurring detrimental processes of ageing and unhealthy lifestyle. The trade-off includes faster postnatal growth in the face of abundant nutrition and earlier puberty, (102, 213, 551). In line with many evolved traits, such effects show sexual dimorphism, for example, in relation to body composition and metabolic control (475). This is consistent with the concept of PARs as Darwinian fitness traits are often sexually dimorphic. Such effects, especially in relation to the timing of menarche, growth in height, and adiposity have been known in humans for many years (see seminal studies of Tanner and colleagues, e.g., Ref. 579).
How long after birth conditioning effects can be induced in humans is an open question, but it is clear that they can persist through infancy. It has been argued that it is the catch-up growth in infancy rather than the environment of the fetus that induces metabolic conditioning (545), but this ignores the fact that by definition catch-up growth follows impaired fetal growth. The issue is rather whether reversal of the changed later disease risk is possible by postnatal manipulation of growth, a matter yet to be resolved although there is some evidence to support the concept (544). Most data on postnatal growth relate to postnatal overnutrition (see sect. VB), which is likely to involve a nonadaptive pathway, or to behavior where the distinction between biologically conditioned and behaviorally learned mechanisms is complex.
There are a number of maternal and demographic factors that can influence the risk of the mismatch pathway being induced. Godfrey and co-workers confirmed the relation between maternal BMI and fat mass reported in the child (649) in adult offspring aged 28–32 yr (499). Interestingly, however, the effect was greatly magnified in primi- versus multiparous pregnancies. Other studies have also shown both greater obesity, impaired insulin sensitivity, and higher blood pressure (16) in first-born children. This accords with the mismatch theory as the processes of maternal constraint would be expected to be greater in first-born children and therefore the potential for an inappropriate PAR is larger. Similar considerations may apply to twins and multiple conceptions (291) and to preterm infants (265). They may also apply in a complex way to maternal age (567). Evidence is now accumulating for a greater NCD risk in offspring born post-term (43). It is possible that this also represents the consequence of the fetus outgrowing its placenta or that function declines in the post-term placenta (617) as apoptosis is reported in such placentas (553).
In many high-income countries, maternal age at first pregnancy is increasing as couples have access to contraception and wish to establish careers or have a range of experience before starting a family. The relation between increasing maternal age and adverse pregnancy outcomes has been investigated (342). Older women are more likely to have declining fertility and to seek ART (see sect. VG). There is also some evidence that risk of diabetes in the offspring increases with maternal age (567), but this has not been examined extensively.
C. Role of the Placenta
Human placental function differs in some respects from that of the more extensively studied animal species such as the sheep or mouse, e.g., with respect to amino acid transport (63, 354) and the operation of novel exchange mechanisms (98, 99) or specific gene transcripts (104). The functional capacity of the placenta, affected by its size (24), surface area for exchange (e.g., Ref. 539), and perfusion (87) are major determinants of fetal nutrient provision, and this varies across the normal range. From a gross anatomical point of view, the placenta is fully formed before the peak in fetal growth and nutrient demand (508). A range of factors affect placental development from early in gestation, the concept of the placental exposome (353). As for fetal development, the role of oxidative and nitrosative stress in early placental development, particularly after the rise in intervillous space perfusion at about 12 wk gestation in the human, may be critical to subsequent structure and function (77). In late gestation, there is currently interest in whether increasing placental perfusion may promote fetal growth in growth-restricted fetuses; in animal models, the administration of an NO donor promotes such growth (137), and increased VEGF expression using an adenovirus vector promotes fetal growth, although whether this is via effects on uterine perfusion is not clear (83).
The placenta is sensitive to fetal demands, but its responses are also conditioned by maternal factors. In the mouse, the paternally imprinted insulin-like growth factor 2 (IGF2) P0 transcript is responsible for upregulating placental nutrient transport and thus fetal growth, a process limited by maternally driven IGF clearance (104). Placentas deficient in the placental specific IGF2 P0 transcript produce effects on the responses to nutritional challenges in pregnancy, for example, in being less able than wild-type placentas to upregulate amino acid transport (532). Thus a mismatch between fetal nutrient demand and placental supply, for example, in the capacity of the small Igf2P0 placenta to supply nutrients at a sufficient rate to meet the demands for nutrients for growth by the normal fetus, can have long-term effects on the offspring; this has recently been shown for increased anxiety behavior responses (404).
Recent studies have shown that placental function is affected by caloric restriction (93), obesogenic diet (533), and dexamethasone administration (609), all factors known to produce a range of effects on the offspring.
Although the majority of work has been conducted on amino acid transport, studies are now ongoing for the placental handling of fatty acids (356) and calcium transport in relation to bone mineral accrual in the fetus (383) and its later consequences for risks of osteoporosis and fracture (105, 106), again with links to maternal nutrition.
Small increases in maternal and thus fetal glucose would be expected to increase fetal growth if this situation represented the relaxation of maternal constraint, and mild hyperglycemia produces a small increase in lean body as well as fat mass (132, 153, 574, 644). This may be viewed as potentially adaptive, as in infancy relative adiposity provides metabolic reserves for thermogenesis, critical organ function, and growth in the event of inadequate maternal care (331). In addition, it appears that epigenetic control of glucose transport (436) and adiponectin expression in the placenta is related to maternal blood glucose across the range from normal to pathological (60), indicating a role in regulating placental transport and growth of the fetus in relation to maternal nutrient status. Changes in the placental epigenome are now being studied systematically in relation to gestational age as well as maternal and environmental factors (437) and in twins (449).
The placenta plays a role in regulating the response of the fetus to glucocorticoids and in integrating the nutritional response with a hormonal response (418), initially as an adaptive response (Figure 9; Ref. 418).
The balance of enzymes that convert glucocorticoids to the active form cortisol (corticosterone in the rat), i.e., 11βHSD1, or which inactivate it, 11βHSD2, controls the fetal exposure to glucocorticoids of maternal origin (527). There are parallels here with the role of these enzymes in protecting the mineralocorticoid receptor of the kidney and other tissues from activation by glucocorticoids (529). In rodents, experimental inactivation of the 11βHSD2 enzyme by carbonoxalone produces effects on the offspring including elevated blood pressure and hyperglycemia (367). Interestingly, glycorrhizin is a natural product in licorice with similar properties, and studies in Finland, where licorice is widely consumed, have shown that eating it in large amounts in pregnancy produces effects on the timing of labor and offspring HPA axis function (485). Poor maternal nutrition in rats affects the concentrations of placental 11βHSD2, thus potentially exposing the fetus to higher glucocorticoid levels (339).
Within the normal range, the adaptive responses of the placenta will contribute to the PARs of the fetus. Outside the normal range, pregnancy conditions such as preeclampsia or gestational diabetes have pronounced effects on placental function (135, 492). These in turn produce substantial effects on the factors that are likely to operate in the nonadaptive range. The mechanisms involved include oxidative stress and inflammatory pathways (484). There are now a range of studies reporting the epigenetic effects of gestational diabetes on specific gene pathways including those controlling growth and metabolism (511). Ongoing studies are examining the extent to which maternal body composition, including obesity, muscle mass, and diet, affects placental development and function (89, 355, 440). There are some data showing altered methylation of a cardiovascular disease-associated gene, ABCA1, in placenta and cord blood samples from women diagnosed with impaired glucose tolerance during pregnancy (268).
D. Conflicting Demands
It has been argued that effects of environment on the mother-offspring dyad have more adaptive significance for the mother than for her offspring (643) because, in the face of conflict for resources such as nutrition, it makes more sense teleologically for the mother to prevail; she has passed a Darwinian fitness test in reproducing, whilst the fitness of her offspring is untested, and her survival will allow her to capitalize on her fitness by reproducing again. While this argument may apply in rapidly reproducing species which produce large numbers of offspring (382a), there is no evidence for it in the human. Indeed, considerations of life history theory make it an unlikely strategy for a slow reproducing species where there is large maternal investment in each pregnancy (207), all of which will be wasted if the offspring does not survive.
Moreover, available data for human populations do not support the suggestion from conflict theory that maternal interests should be prioritized over those of her offspring at times when resources are limited. Fecundity in humans is only modestly reduced during famine (566), and under conditions of severe undernutrition, there is only a small reduction in fetal growth (506). Even lactation, which places substantial nutritional demands on the mother, continues during severe famine. Conversely, as discussed earlier, fetal conditioning has long-term effects and occurs in the absence of any obvious fetal compromise (185), and this similarly counters the maternal advantage argument (226). Fitness is not easy to measure in humans, as we have great control over our reproduction and nurture even weaker members of our species; despite this, a model which involves “détente” between mother and offspring seems more appropriate, as it puts into context both maternal needs and the opportunities which an exchange of information between mother and offspring offer for the subsequent fitness of the latter.
E. Transduction of Cues
There are two, not mutually exclusive, ways of looking at the possible cues on which the developing embryo, fetus, and neonate can base their predictions of the future. One is to argue that the most effective cues, or reliable markers, are those aspects of the environment on which survival and fitness depend, i.e., oxygen tension, nutrition, water availability, and so on. There are many examples of how such cues operate, from the fetal responses to hypoxia (234) or hyperoxia (51), micronutrients (390), macronutrients (225), and the responses of species such as amphibians to thermal stress and dehydration (49).
The second way concerns the advantage of signaling indirectly aspects of the environment on which fitness depends, e.g., population density (392), predator numbers (58), or other stress levels. Interest here is focused on the use of endocrine signals that might provide adaptive cues. Species that are seasonal breeders also utilize cues related to day length (348) via transplacental melatonin signaling. Anxiety levels or the need for vigilance in escaping predation are thought to be signaled via glucocorticoids produced by the maternal adrenal gland and crossing the placenta or being present in the milk. In rodents, stress induced in pregnancy or infancy leads to altered behavior, stress responses, and cardiovascular function in adult offspring (129, 277). In humans, there are several studies showing an association between maternal anxiety or stress level in pregnancy and cognitive, behavioral, or emotional problems in the child (249, 602). There are even possible effects on ageing mediated by shortening of telomere length (163). There has been recent discussion of the adaptive significance of such processes in humans (537), and del Giudice et al. (134) have proposed the adaptive calibration model by which the stress system coordinates and encodes, as well as regulates, the offspring's response to a variety of physical and psychosocial challenges and conditions, with implications for later adolescent behavior (161), for example.
At the level of tissue and organ growth, physiological control is exerted through the action of growth factors, particularly the insulin-like growth factors (IGFs) (208). Their actions can be modulated in turn by levels of IGF binding proteins in the plasma (37), which also convey them to the cell surface to interact with IGF receptors. Fetal development is much less dependent on growth hormone than on IGF. The mechanisms of action of all these agents are complex, however, including effects not only on mitosis but on blood flow, glucose uptake, carbohydrate and lipid metabolism, and oxidative stress.
Moving down the physiological chain of mechanistic effects, there is a large body of work concerning oxidative and nitrosative stress. The balance between pro- and antioxidant species at any site depends on the level of oxygenation (in turn dependent on blood flow and capillary Po2), tissue metabolism, and the provision of antioxidant protective mechanisms (122, 199). Because of its pivotal role in integrating stimulus and response at the cellular level, oxidative stress is a fundamental regulator of development and has been proposed as a process which “sculpts” development. O2·− and nitric oxide (NO) interact to modulate peripheral vascular responses to hypoxia in the fetus (414, 588), effects which can be altered by antioxidant treatment (257, 305, 587) or statins (160, 594). In addition to hypoxia, under- and overnutrition, infection, and exposure to exogenous glucocorticoids all produce an increased level of oxidative stress (122, 199, 461, 542, 589). There are a range of studies which show effects on tissues, especially the vasculature (280, 502, 593, 624, 678), but there are also effects on the heart (2, 258, 432, 627) and on brain development (590). The mechanism by which oxidative stress operates at the cellular level in these organs is now beginning to be elucidated. In the heart, the reduced expression of the cardioprotective gene protein kinase C (PKC)-epsilon in hypoxia is prevented by inhibition of DNA methylation (460, 461) or by N-acetyl-cysteine. The process may be initiated by norepinephrine via activation of Nox1-dependent reactive oxygen species generation (665).
Circadian rhythms and metabolic control processes interact in both physiological and pathophysiological states, as shown by the perturbations in feeding and activity patterns in rodents whose dams were fed unbalanced diets (66) and the role of clock genes in diabetes (381). There may be links with oxidative stress here, as the role of Rev-erb in adipogenesis is affected by heme, which in turn is modulated by oxidative stress (322). The role of oxidative stress in conditions such as diabetes is not clear however; for example, it has been hypothesized that the beneficial role of exercise is to promote oxidative stress that stabilizes protein structure (635).
Obesity is associated with inflammation; higher levels of interleukin (IL)-6, tumor necrosis factor (TNF)-α, and C-reactive protein (CRP); and oxidative stress (128, 486), and this can have adverse effects on the placenta (448) and cardiovascular system of the offspring (627). The inflammatory process in a range of tissues, including adipose tissue, may involve infiltration and polarization of macrophages from M1 to M2 forms (373) which are associated with a proinflammatory state and the secretion of inflammatory cytokines (642).
As for the effects of hypoxia, the role of oxidative stress associated with obesity is shown by the protective effects on the offspring of maternal treatment with quercetin (362), vitamin E (420), or a combination of antioxidants (531).
Oxidative stress is also linked to sirtuin-mediated ageing processes such as telomere shortening, which can be conditioned by the developmental environment (292). SIRT-1 is a histone deacetylase that senses cellular energy levels via the NAD/NADH ratio (52), so it can mediate epigenetic effects of nutrient provision and also metabolism. This pathway, leading to changes in peroxisomal proliferator-γ coactivator (PGC) 1α, is linked to fat deposition in the liver of offspring of rat dams globally undernourished during pregnancy (653). Changes in PGC-1α, ER-α, ERR-α, and HNF-4α can also be induced in rat offspring by feeding their dams a methyl donor-deficient diet in pregnancy (477). This condition resembles nonalcoholic fatty liver disease (NAFLD) in humans and can also be produced in rodents by feeding a high-fat diet in pregnancy or postnatally (68). A high-fat/high-sugar diet fed to dams produces epigenetic changes in the regulation of the glycerol-3-phosphate acyltransferase 1 gene, which is activated by sterol regulatory element-binding protein-1c (SREBP-1c), a transcription factor that acts as a master regulator of lipogenesis (154). Further evidence for the role of epigenetic processes in this pathway in humans is that methylation of specific CpGs in the promoter region of PGC-1α in the peripheral blood leukocytes of children aged 5–7 yr predicts their level of adiposity at age 14 yr (97). These processes may remain dynamic throughout life, as PGC-1α levels change rapidly following exercise in skeletal muscle, possibly linked to mitochondrial biogenesis (17).
Other processes by which maternal obesity may induce effects on the offspring can result from the elevated levels of insulin and glucose in the offspring's blood. Insulin is associated with increased growth (261), and hyperglycemia can have effects on the fetal heart (107).
There may also be elevations in RAS activity and in sympathetic activity (517). These interact with leptin to induce changes in appetite. Prolonged sympathetic nervous system activation can desensitize β receptors in the heart, for example, initiating a series of processes leading to cardiomyocyte apoptosis (320). Again epigenetic repression of PKC-ε, induced by NO, may involve Nox1-dependent ROS production (665).
Because the two types of cues referred to above are not mutually exclusive, it is possible that they overlap. The evolution of the glucocorticoid system may have arisen through the advantage of a stereotypical response to an environmental challenge, for example, from a noxious stimulus, predator attack, or overcrowding (422). In mammals, the glucocorticoid system has a role in response to other challenges such as unbalanced maternal nutrition (528), hypoxia (200), or infection (231). Glucocorticoids are involved in the maturation of many fetal organ systems (175), preparation for birth (363), and the control of fetal growth (174). They can induce a very useful set of responses that prepare the offspring for the predicted future and may be induced from early gestation. Thus, in the sheep while mild undernutrition in the periconceptional period and early gestation prolongs the duration of gestation (102), more severe undernutrition shortens gestation (54).
For both of these types of cue, attention has largely been focused on the parents in signaling aspects of their contemporary environment. However, another feature of the signaling may be of environmental history over a longer period, integrating this into a cue that drives the optimal adaptive strategy for the species (331). Indeed, some level of inertia in the system seems inevitable so that the fetus responds to either sustained or repeated stimuli. This has not been extensively investigated. However, in the studies of the effects of intrauterine growth retardation on epigenetic changes in pancreatic β cell Pdx1 (459), it was noted that histone changes occurred first and were later consolidated by more stable DNA methylation changes. It may be that the more labile histone system responds first, and it is only with persistent or repeated cues that a more stable change in epigenetic state is induced.
V. PATHOPHYSIOLOGICAL (NONADAPTIVE) PATHWAYS IN DOHaD
These are summarized in Figure 10.
A. Maternal Overnutrition, Obesity, and Excessive Gestational Weight Gain
A perspective derived from the principles of evolutionary medicine (202) is useful in that it highlights the importance of evolutionary novelty as a pathway to disease risk. The underlying concept is that humans are unlikely to have evolved protective mechanisms against novel challenges: the modern western diet can be seen as one such challenge.
As for the adaptive processes discussed above, insights into nonadaptive processes have largely come from animal studies and then been applied to humans. Most studies involved overnutrition, especially the use of a high-fat, a cafeteria, or a high-fructose diet (611). There have been some studies using non-human primates, in which feeding a highly palatable diet to produce obesity in the mother induces placental effects and changes in fetal growth (166) and differential expression of cardiac microRNAs (378), and feeding a high-fat diet across generations produces aspects of the metabolic syndrome in the offspring (165). Other studies have used sheep given the similarities to humans in degree of maturity at birth. While sheep do not readily consume a cafeteria diet, dietary-induced obesity in the ewe induces detrimental effects on the cardiac function of their lambs (271, 627).
The majority of studies have been performed in rodents. A high-fat diet during pregnancy produces adiposity, metabolic and vascular effects in adult offspring, changes in aortic and cardiac lipid composition (159, 194, 315, 549, 586, 676), and also early signs of fatty liver disease (159, 442), reminiscent of the NAFLD now thought to be an early marker of metabolic syndrome. Changes in mitochondrial function are present (68). There is also evidence for left ventricular hypertrophy in this model and for the disruption of the cyclin signaling pathway (14), which may also include mitochondrial effects and epigenetic changes (145). Although many of these studies involved exposure of the dam to a high-fat diet only during pregnancy/lactation, longer-term exposure which mimics a human nutritional transition produces similar effects (159, 516). Interestingly there is evidence that prenatal undernutrition followed by dietary-induced obesity in rodents advances the timing of puberty in their offspring (550), a feature of transitional human societies, especially migrants (457).
Most recently, it has been demonstrated that, as for folate supplementation of the maternal low-protein diet, methyl donor supplementation blocks the epigenetic effects of a maternal high-fat diet on offspring prefrontal cortex, accompanied by effects on weight gain and food preference (82). In the vasculature the effects of the quantity and the nature of maternal dietary fat on offspring vascular function involve epigenetically mediated changes in FADS1 leading to changes in prostanoid production (309). Ectopic fat accumulation is reported in many sites in the metabolic syndrome, and such deposition even occurs in bone (340). Some effects on offspring are mimicked by exposure of the pregnant dam to endocrine disruptors (79, 230), and there are effects on circadian rhythms as well as stress responses (see Ref. 78).
There is much evidence that maternal obesity and/or excessive weight gain in pregnancy is associated with offspring adiposity. This seems likely to have a nonadaptive basis as extreme maternal body size and nutrition is likely to have been relatively rare in our prehistory (375). Once again it is clear that the transmission of phenotype from mother to offspring occurs across a wide range, both for maternal weight gain in pregnancy and for maternal blood glucose. At the extreme, fetal macrosomia is associated with a greater risk of shoulder dystocia, obstructed labor and, in the absence of modern medicine, stillbirth, genitourinary fistula, and possible maternal death (62). Such problems contribute substantially to maternal and fetal deaths globally. But such processes have not apparently been subject completely to negative selection through evolution, suggesting they have been traded-off against the advantages of bipedalism and large head size in our evolutionary past. Although pregnancy produces relative maternal insulin resistance, which promotes glucose transfer to the fetus (335), incidence of maternal obesity, high pregnancy weight gain, and gestational diabetes would presumably have been low in hominid evolution where high glycemic index foods were not available. In support of this, and as discussed earlier, it appears that there is no transport maximum for glucose in the human placenta, as there is for many other nutrients (126). Maternal obesity is likely to induce fetal hyperinsulinemia (86) leading to greater fetal adipogenesis (see below), although effects via other systems studied in animals as discussed above have yet to be undertaken in humans.
There is now accumulating evidence from several cohort studies that maternal obesity and excessive gestational weight gain are associated with increased risk of asthma and atopic disease in the offspring (248, 349, 469). Because the effect is amplified by infant weight gain (371), it gives another example of the consequences of mismatch (469), so arguably they are not inherently nonadaptive. The role of epigenetic processes in producing effects on immune function (384) provides clues to underlying mechanisms that could be viewed as physiological, in shifting the balance of innate versus acquired immune function. However, the exacerbating effects of unbalanced maternal diet, smoking, or exposure to toxins or pollutants indicate their pathophysiological nature.
B. Infant Overfeeding
Infant overfeeding is essentially an evolutionary novelty made possible by infant formula products. Children showing rapid growth in early infancy or in early childhood are at greater risk of adulthood cardiovascular and metabolic disease, especially if they were small at birth (164, 172). In premature infants who are fed with infant formula, rapid growth is associated with risk factors for cardiovascular disease in adolescence (545). Rapid weight gain in infancy is associated with childhood (447, 450, 569) and adolescent (198) obesity and with childhood cardiovascular dysfunction (184). Formula feeding is widely believed to produce rapid growth in early infancy (13), although the validity of these data has been challenged (327). Meta-analysis shows a reduced incidence of childhood or adulthood obesity in breast- versus formula-fed infants (279). The protective effects of breast feeding may extend to allergy/atopy (444, 525) and cardiorespiratory function (212). However, the protection against obesity afforded by breast feeding remains controversial (520) and may be explained by different effects of breast feeding in normal compared with overweight children (29). A further issue which merits consideration is the composition of the formula, especially its protein content (556).
C. Preeclampsia and Hypertension in Pregnancy
There is a limited amount of evidence that the offspring of preeclamptic pregnancies have elevated blood pressure and BMI (123) and also greater risk of later cardiovascular disease and stroke (124). Such offspring usually have lower birthweight and are often preterm, although there appears to be an effect of preeclampsia on offspring cardiovascular and metabolic risk independent of gestational age. The relative contributions of reduced growth, undernutrition, hypoxia, as well as anti-angiogenic and immune factors to the effects on the offspring are not known. In addition, there may be effects related to maternal hypertension and preterm birth (345) (see below).
D. Preterm Birth
A range of studies have shown an association between preterm birth and higher risk of cardiovascular disease, insulin resistance, and T2D in the offspring as children or adults (265, 278, 351, 387, 471). The assumption has been that this is largely due to the small size of the infant at birth as shown from a large historical cohort (57). We have discussed above why this explanation does not seem adequate. A recent study (626) has, however, separated out these components in a prospective cohort in which births were classified as early preterm (<34 wk), late preterm (34–36 wk), early term (37–38 wk), and full term (=39 wk gestation). Wang et al. (626) confirmed the association between plasma insulin level in children and their size at birth (SGA vs. AGA) but also found an independent association with gestational age in their four categories. This persisted after correcting for ethnicity, maternal smoking, BMI prepregnancy, parity, pre- and gestational diabetes, and also infant's sex and birthweight for gestational age. Moreover, the plasma insulin levels at birth tracked into those in the children in early childhood (up to 6.5 yr old). The study thus provides evidence for prenatal factors conditioning postnatal insulin sensitivity, although as yet it is not known how this is mediated, or whether it depends on the conditions leading to preterm delivery. It does however raise the question of whether late (and possibly some early) preterm birth should be viewed as part of an adaptive response; the consequences of extremely preterm birth and very low birthweight (269) are likely to be an IAR.
E. Gestational Diabetes Mellitus
Pregnancy is associated with some degree of insulin resistance due to the actions of placental lactogen (306) and growth hormone (21). This is thought to induce a homeorhetic change in maternal physiology to favor glucose transfer to the fetus. But as discussed above, there is no effective barrier to placental glucose transfer and thus, if maternal glucose rises excessively so will fetal glucose, leading to fetal hyperinsulinemia and excess somatic and particularly adipose growth. There are also distinct epigenetic effects (157). The consequences if untreated can be dystocia and fetal and maternal death; thus we have argued that gestational diabetes mellitus (GDM) is an evolutionary novelty and that the consequences for the next generation have nonadaptive origin.
GDM is more common in obese women and in those with a preexisting predisposition to insulin resistance. Women of lower birth weight themselves have a greater risk of developing GDM, reflecting their already conditioned predisposition to insulin resistance (153), whilst genetic variants associated with T2D are also associated with increased risk of GDM (324, 341). The evolutionarily novel situation of abundant nutrition and low levels of physical activity plays an important role in perpetuating a vicious cycle of disease. New criteria for defining GDM (402) have now produced a dramatic upward revision in the number of women suffering from this condition, the prevalence of which is approaching 30% in some parts of the world (244), especially in populations which have gone through rapid socioeconomic and nutritional transition such as parts of Asia (263), the Middle East, and some Pacific islands (629). Indeed, in a Canadian First Nation Peoples population, the proportion of T2D incidence explicable in terms of GDM in the previous generation was 30% (148, 671).
Apart from GDM, the link between maternal diabetes and risk of metabolic syndrome in the offspring is well established (see Ref. 648). Epidemiological studies show greater risk of diabetes in children of type 1 or 2 diabetic mothers (307, 347, 465, 555). This risk of diabetes is greater in children born after the mother is diagnosed with diabetes, presumably due to prenatal exposure to hyperglycemia (116). These offspring are also more likely to develop diabetes at a younger age (466). Recent studies show that even individuals exposed to mild hyperglycemia prenatally have greater adiposity and risk of later diabetes and cardiometabolic disease (237, 578), and there is evidence from a meta-analysis that GDM is associated with higher blood pressure in offspring (1).
F. Caesarean Delivery
There is a small but clear increased risk of diabetes in offspring born by caesarean section (81), which is of course an evolutionary novelty. One possible mechanism concerns the gut microbiota, which colonize the infant gut from the maternal vaginal and gastrointestinal tracts and which differs according to the mode of delivery (7). Interestingly, use of antibiotics in the first six months of life in a large Danish cohort of mother-child dyads is associated with childhood overweight in offspring of normal weight, but not of obese, mothers perhaps as a result of effects on the gut microbiota (7).
G. Assisted Reproductive Technologies
An issue that is now receiving much attention concerns the apparently greater risk of NCDs in offspring conceived by ART (250), also an evolutionary novelty. It is of note that the effects seem to be independent of the medical or other reasons for the couple seeking fertility treatment. Animal studies show that short-term culture of the early embryo can induce NCD risk-related phenotypes (632), and this appears to be the case even for effects of maternal dietary manipulation during the period of oocyte maturation (634). In humans, the hormonal treatments used to induce ovulation prior to ART may have longer-term effects. Some of these effects may be mediated by epigenetic processes (604).
H. Effects of Drugs
Relatively few drugs are prescribed in pregnancy, and the teratogenic (Table 1) effects of thalidomide, prescribed to control hyperemesis, are well-known (391). There has been considerable research on the effects of antenatal steroids given to women with threatened premature delivery, to accelerate fetal lung maturation. Exogenous glucocortioids such as dexamethasone, given during a critical period of early development, are detrimental to cardiovascular and renal function and to glucose homeostasis in several species (140, 388, 528) and affect neurocognitive function in juvenile non-human primates (503). Most obstetricians now advocate use of only a single dose of glucocorticoid, although even this has been reported to be associated with elevated blood pressure in some children (405). Young adults exposed to a single course of betamethasone in utero showed no psychological or health-related quality of life effects (118), although there were small effects on glucose tolerance (119). The most recent evidence comes from a Cochrane review of 10 trials, which concluded that there is benefit of multiple versus single doses for neonatal outcomes, without any evidence of either significant benefit or harm at follow-up (396). At present, it appears that the gain from administration of antenatal corticosteroids, preferably as a single dose, outweighs the potential disadvantage.
i. Smoking, Toxins, and Endocrine Disruptor Chemicals
Smoking is well-known to produce a range of effects on the fetus, including overall growth restriction (419) and effects on the development of many organs and systems. There are similar concerns about a range of toxins to which the woman and her fetus or suckling infant may be exposed, including those produced by bacteria (288) or fungi (e.g., Ref. 490) in foodstuffs. The detrimental effects of maternal viral or bacterial infection are also well-known, and possible detrimental effects on the fetus associated with their treatment raise other issues (15, 67) as are those of infestation with parasites such as Plasmodium falciparum, Schistosoma, or tapeworm (435, 557, 610). There is great concern about alcohol consumption in pregnancy, and although the recommended safe limit is low, only high levels of consumption are associated with fetal alcohol syndrome and permanent effects on the offspring (300). Cocaine and other drugs are well-known to have a range of effects on the fetus (41), with consequences for cardiometabolic, neural, and behavioral development in the child.
In many industrial and heavily populated urban areas, exposure to heavy metals, dyes, and airborne microparticles is of concern (189, 561, 597). A well-known example of detrimental effects is the contamination of rice oil with PCBs in Japan in 1968, leading to the hyperpigmentation of Yusho disease and low birth weight (597). Similar concerns relate to exposure to pesticides and other agrichemicals in some rural settings (255). The concentrations of such compounds build up in the mother's body over time, and this may explain the relation between effects on the offspring and birth order (314).
The most dramatic effects of such agents are teratological or increased risk of cancer, especially of the reproductive system, in the offspring (558). Such effects are clearly disruptive of development. Some of these agents produce effects on endocrine (45) and nuclear receptor (346) function, hence the term endocrine disruptor chemicals (EDCs). In some instances they may act via effects on epigenetic control systems regulating endocrine function (548), and this is an area of intense research and debate. By acting in this way, they can induce greater risk of NCDs without producing an overt disruption of development per se. In animals, prenatal exposure to compounds such as bisphenol A can produce effects on offspring adiposity and cardiometabolic control (79, 615). As the effects of an EDC in isolation can be subtle, and as even low environmental levels of several EDCs can act together to affect development, more research on their long-term health effects is needed. A recent report from the Royal College of Obstetricians and Gynaecologists on possible effects during pregnancy (491) seems to demonstrate the importance of the precautionary principle in this regard, but may cause unnecessary anxiety in an area for which evidence is not clear.
VI. BROADER IMPLICATIONS OF THE DOHaD CONCEPT
A. Evolutionary Developmental Biology
For natural selection to produce an evolutionary modification, an environmental change, if sustained, must lead to an alteration in genotype that is both adaptive and heritable. But the principles of developmental plasticity suggest that initially these responses may be transient and induced by epigenetic processes. The issue then is whether such induced epigenetic changes could influence the likelihood of genomic change (296) and how these phenotypic changes get fixed into the genotype by the processes of genetic assimilation (Table 1) (647). Two models of genetic assimilation were proposed by Waddington: the first was simply the uncovering of covert genetic variation and selection against modifying genes which suppressed the exhibition of the desired gene, and the second was that such “fixation” occurs by random mutation. More recently, the possibility of a bias in the pattern of mutation has been proposed based on the greater mutability of methylated cytosine, which may spontaneously convert to thymine (467). The evidence for this possibility has been reviewed elsewhere (32, 33).
Much has been written on the evolution of adaptive developmental plasticity (32, 196, 282, 468, 647), but its implications have not been fully reconciled with classical evolutionary theory. The concepts described above suggest the possibility that while adaptive developmental variants are induced during transient environmental change, and preserved in fluctuating environments, they might become fixed during sustained environmental change. This would provide a new model by which epigenetically initiated but biased mutation could lead to developmentally driven evolutionary change and reconcile evolutionary developmental biology with classical evolutionary biology. Indeed, evidence for such transition is found in cancer cells (233), which can be seen as a form of evolution occurring at a cellular clone level. There are now a range of reports which suggest the operation of such processes, although they are beyond the scope of this review (80, 103, 218, 296, 301, 359, 505, 675, but see Refs. 276, 625).
These concepts have many attractive, albeit unproven, aspects. First they explain how an adaptive process can occur rapidly enough to meet an environmental challenge. In addition, they allow such a process to be induced in several individuals in a population simultaneously. They allow the process to be “tuned” to relevant parts of the genome where the probability of mutation is greater at the site of epigenetic change. This is parsimonious and also protects against the potentially harmful effects of merely random mutations. Lastly, they allow the process of genetic assimilation to be related to the environmental change itself. Because the epigenetic effect increases the likelihood of a mutation, such assimilation would be more likely to occur if the environmental change is sustained over several generations. If that change is transient, the reversibility of the epigenetic processes will permit the phenotype to return to a previous state with no genetic modifications.
B. Biomarkers of Risk
The epidemiological studies that focused on lower birth weight raised the issue of whether phenotypic characteristics could be used to predict individuals at particular risk of later NCDs. However, the usefulness of any biomarker depends on its sensitivity and specificity. The association of both low birth weight and high birth weight with greater risk makes the use of this particular characteristic problematic as the adaptive and nonadapative pathways, which are somewhat reflected in birth weight, involve different ultimate and proximate mechanisms that may confer different risks of later disease. In addition, there are multiple pathways to a specific birth weight: a low-birth-weight fetus at term may result from a slow growth trajectory throughout gestation, a rapid trajectory followed by a period of greater constraint, or simply reflect mild prematurity. It is likely that the longer-term consequences of these developmental trajectories may vary.
Even though measurement of the trajectory of fetal growth throughout gestation is feasible, it is not routinely undertaken. Postnatal growth is, however, routinely measured. This allows such growth to be used as a biomarker of later risk, e.g., it has been suggested that crossing of two growth centiles in the first 6 mo after birth is associated with obesity in children aged 3 yr (585). The problem with the use of such measures is that, as with birth weight, growth trajectory is more a measure of the integrated responses of the infant to a complex set of environmental circumstances, an end result of effects on metabolism, appetite, immune function, emotional, and other influences such as parental bonding, etc., so again infants with similar growth trajectories can have different combinations of these processes in operation and so different later risk.
The broader question thus concerns the nature of any biomarker. Is it a component of a mechanistic pathway from challenge to pathological state, or rather a measure of an epiphenomenon that has occurred in parallel? Very often the answer to this question is not known. Surrogate or proxy biomarkers for a process are not necessarily inferior if the effects on them are large and consistent, for example, the use of cholesterol as a marker of cardiovascular disease risk.
Developmental epigenetics may provide new opportunities and certainly provide the capacity to estimate the relative importance of developmental conditioning to disease risk. Epigenetic marks measured at birth cannot only give a measure of the response which the embryo and fetus made to developmental environment, but they can also provide measures of the setting of homeostatic systems, i.e., the conditioning, which will regulate responses to later challenges. For example, the degree of methylation of one CpG associated with the RXRA gene is related both to aspects of maternal diet in early pregnancy and to child's fat mass 6–9 yr later (229). In this example, the effect size is large, with the level of methylation accounting for 25% or more of the variation in fat mass of the child, larger than any other early life marker of lean or fat mass or current estimates of the fixed genomic contribution to obesity. Confirmation of this concept has come from the work of others using cord blood samples (464, 494) and for the Southampton cohort with child's bone density as an outcome (252).
These studies used a combination of array discovery techniques to give a measure of the genomic regions in which epigenetic changes may be manifest, followed by more in-depth analysis of candidate regions. In addition, they examined whether the epigenetic changes occur at sites known to possess SNPs relevant to disease. As this field progresses, the interaction between fixed genetic variants and epigenetic changes is likely to become more apparent (42).
Finally, biomarkers may or may not be useful as measures of the efficacy of an intervention. This may depend on whether they are reversible or not, usually thus on whether they contribute to part of the causative process or are a proxy for it.
C. Possibilities for Interventions: Proof of Principle in Animals, Opportunities for Humans
Several groups have explored the concept that an intervention given during development, either to the mother or her offspring, can reduce or prevent the detrimental effects of dietary mismatch on her offspring. Interventions that have been tested are dietary, endocrine, or pharmacological. A dietary intervention with folic acid or glycine, given to the pregnant dam at the same time as the unbalanced diet, prevented the effects on the epigenome in the offspring (82), with similar effects on growth and metabolic responses of the offspring (71) and on the transcriptome (366). The effects of folate can also be seen during the adolescent period in the rat (72), which is interesting as its protective effects are mediated after administration of the unbalanced diet of the dam which induced the response. This effect has been explored in more detail for the PEPCK gene (266).
Leptin administration to the newborn rat prevents the hyperphagia, adiposity, and other detrimental effects of mismatch produced by global undernourishment of the dam followed by feeding the pups a high-fat diet (613). It also reversed effects of maternal undernutrition on the expression of candidate genes and associated epigenetic processes (222). There are more recent studies on the effects of neonatal leptin (515), exendin-4 (190), including epigenetic effects (472), and growth hormone (232).
Oxidative stress has been shown to play a role in producing effects of maternal low-protein diets on the peripheral vasculature of the offspring and the potential for antioxidants to reverse this has been explored (502, 593). Because effects can be passed to the F2 generation via effects on the uterine vasculature of the pregnant dam offspring, this gives a further avenue for intervention, explored using an angiotensin receptor antagonist to promote such vascular function (514). Administration of a statin to the offspring can prevent effects on vascular function induced by the maternal low-protein diet, and it restored endothelial function without affecting plasma cholesterol (594), an effect possibly mediated by effects on endothelial progenitor cells. If the statin is given in the second half of pregnancy, it can protect the offspring against the conditioning effects of a maternal high-fat diet (158).
While these experiments only serve as proof of principle, they offer promise that, with the appropriate use of biomarkers of risk, timely interventions during development may have the potential to reduce the effects of developmental environment in priming risk of later cardiovascular and metabolic disease in the offspring.
D. Developmental Conditioning and NCDs: Policy and Preventative Interventions
The developmental component of NCD risk does not conform to the generally accepted model of lifestyle and environmental contributions to the burden of NCDs. These are primarily derived from the relatively effective approaches to reducing the adverse health impacts of smoking, but the interventions that have proven effective in reducing smoking do not easily apply even to changing the high energy nutritional environments that contribute both to evolutionary and developmental mismatch.
Legislation cannot be focused so easily on particular aspects of developmental or adult exposures despite considerable attention to the latter in public health policy. Part of the problem is the assumption that adult obesity and its complications are largely a function of inherited genetic susceptibility plus individual choice in lifestyle (203); at the extreme, this leads to blame culture, even the use of pejorative terms such as “gluttony” and “sloth,” two of the biblical deadly sins. Given the impossibility of returning to a premodern diet and lifestyle, a greater focus on creating resilience by obviating the effects of early life influences may be critical.
The science of developmental conditioning thus offers a different perspective. It suggests that a significant component of the risks of obesity and NCDs has a developmental element that is potentially amenable to preventative intervention and monitoring through the use of epigenetic biomarkers. Thus the failure to recognize the lifecourse nature of NCD risk and to incorporate it into public policy may limit capacity to reduce the escalation in the NCD burden. It is now timely to develop an adequate model of this risk and identify points in the lifecourse at which it is most sensitive to strategies to reduce it (660). The epigenetic epidemiological data reviewed above (229) suggest that the developmental component is a major part of the pathway to disease risk. The experimental and epidemiological data on mismatch reviewed above make a point that is often missed, namely, the nature of the response to the modern obesogenic environment is very much influenced by the early environmental exposures even if there is little evidence in early life of a phenotypic effect of that environmental exposure; it may be that the only evidence of the developmental effect at this age lies in an altered epigenetic biomarker.
Public health policy inevitably favors devoting resources to those with current illness over long-term prevention that may not be manifest until the next generation, if only because the former are likely to be voters, taxpayers, and influential citizens. Regrettably, when looked at from a global perspective, the all-too-frequent philosophy of “women and children last” when it comes to issues such as food security appears to operate (530). As will be evident from the science reviewed in this paper, this is not appropriate.
As yet it is hard to put a figure on the cost implications of ignoring a developmental emphasis. For a hypothetical low-income country, an estimate from the World Bank (8) focused only on low birth weight and a limited set of outcomes, in terms of health, education, and economic productivity. More detailed calculations have been made for the early life consequences in terms of cognitive and noncognitive function (254), but this has not as yet been extended to NCDs. The broad range of parental phenotypic characteristics that influence the offspring's future, as well as their lifecourse and even transgenerational influences, suggest that a human capital analogy may be more insightful (121).
Attention is now being given to when in the lifecourse it would be logical for public policy to promote intervention (Figure 11). Adolescence is the logical place to start if the state of the parent prior to conception is to be addressed. Diet and other lifestyle factors operating periconceptionally affect the development of the embryo, fetus, and child, and these processes start to operate even before a couple know that they have conceived. In addition, in many societies a substantial proportion of pregnancies are unplanned, and many women do not change their diet or their lifestyle when they know that they are pregnant (113). Yet adolescents have received less investment in health than younger children in many societies (521). The emphasis needs to extend beyond smoking, alcohol, contraception, and violence to health literacy itself. This is defined as the degree to which individuals have the capacity to obtain, process, and understand basic health information and services needed to make appropriate health decisions (439) and operates at several levels. Increasing health literacy to the highest level in adolescents would allow them to exercise self-efficacy and to make decisions about lifestyle themselves. This may be a particularly important component of addressing behavioral responses conditioned by earlier experience (161).
E. Human Development in the 21st Century: Technology, Social, and Other Issues
Human reproduction has changed dramatically over the last 50 years. On the one hand, the access to effective contraception has enabled many women to choose when to conceive. Many women now opt to have a family later in their lives, to allow greater financial stability and career development as well as freedom from commitments. As fertility declines during the decade before the menopause, this can lead to difficulty in conceiving for some women. In addition, sperm counts are falling in many countries for a variety of reasons including exposure to environmental chemicals, lifestyle, and smoking (534). In parallel with these changes, ART procedures have increased, not only in usage but in complexity. In addition, techniques such as sperm donation and surrogate pregnancy are now more routine.
These benefits do not come without some health costs. As noted above, first-born children may be at greater risk of NCDs and the relative increase in the proportion of the population who is first born has increased dramatically in Western and some Asian countries. Children of older women may be more likely to suffer diabetes in adult life. And ART itself has been reported to be associated with higher levels of cardiometabolic risk markers in the children, an effect which seems to be independent of the current risk factors in the parents and their reasons for seeking fertility treatment. The underlying mechanisms are not known, but animal studies have shown that even embryo transfer produces such effects on the offspring, as does dietary manipulation during oocyte maturation or even the periconceptional period.
Once again, such observations reinforce the importance of an emphasis on parental lifestyle even before conception. In all societies the issues relate to empowerment and self-efficacy particularly in girls and young women. The key issues were raised in a report to World Health Organisation some years ago (661) but have been little acted upon. They included the following: access to secondary education, continued after marriage/childbirth; delaying first pregnancy until at least 4 yr after menarche; diet before and during pregnancy and suckling; reduction of physical exercise levels in pregnancy; and smoking cessation.
The field of DOHaD has grown rapidly to high prominence in biomedical science, public health, and, to an extent, in public understanding of science. However, many of the fundamental concepts underlying it were articulated in different ways many years ago, and there was much supportive data from early studies that has now resurfaced. In hindsight, it now seems surprising that the DOHaD concept took so long to join mainstream science. There are several reasons for this, related to a slowness to recognize that DOHaD processes operate across the normal range of development and are largely physiological rather than pathophysiological. These processes can influence the risk of later disease and involve multiple pathways. For example, it is now recognized that risk is graded across the range of levels of surrogate markers of development such as size at birth and is not only associated with low birth weight. Moving beyond association studies, insights into developmental epigenetic processes are revealing the mechanisms by which the early life environment can influence phenotypic aspects of the individual. This is indicating the normative nature of the phenomena, revealing underlying mechanisms and providing potential biomarkers of risk. Such markers could be measured in early life and used to indicate the most effective interventions and to monitor their efficacy.
The field of DOHaD has been beset by the problem of the use of metaphor, very often an issue in science when new ground is being broken. We have discussed why the use of the term programming has been unhelpful; it implies that a mechanism has been put in place that will inevitably lead to disease, and that the process is ballistic because, once initiated, its outcome is predetermined. This has raised expectations with respect to associations between the environment in early life and later incidence of disease that are unrealistic. We prefer the less deterministic term conditioning as it makes clearer the concept that early life environment only conditions an individual to respond physiologically to later environmental challenges in a particular way. Other processes may intercede so that the response may ultimately be different. Moreover, it does not carry the implication that disease processes start during development, as if they are pathophysiological from the outset. Development can affect the future of an individual, but this is always conditional.
We have discussed these issues in the context of evolutionary biology in the sense that, as part of normal development, they could be argued to have been selected during evolution as they confer an adaptive advantage. We stress that such advantage need only be manifest in terms of survival to the age of reproductive competence and until successful reproduction has been achieved. Our concept of predictive adaptive responses encapsulates this, because environmental cues in development will only be adaptive until reproduction is complete. The risk of NCD, which usually increases rapidly in the post-reproductive period, may or may not be influenced by the fidelity of the prediction during development of the nature of the environment in the period up to reproduction. Throughout life, but particularly in the post-reproductive period, disease risk increases due to the accumulative effects of inadequate responses to challenges and to the consequences of changes in lifestyle, behavior, and environment, etc. This latter group we have termed “mismatch” to indicate that they represent challenges with an element of evolutionary novelty, which neither the individual's development nor their evolutionary ancestry may have equipped them to meet completely. The mismatch between an individual's physiological phenotype and their current environment can occur at any point in the lifecourse. During development this is linked to the category of nonadaptive effects which we discuss, many of which have arisen through Westernization, climate change, socioeconomic progress, pollution, etc.
Evolutionary biology applied to medicine can therefore assist in emphasizing the complex interaction between proximate (e.g., an immediate challenge from unbalanced diet) and ultimate (e.g., evolved traits) in the etiology of disease. The recent developments in developmental biology, and in the field of evolutionary developmental biology (evo-devo), contribute to our understanding of this interaction. In particular, substantial work in developmental epigenetics demonstrates how developmental plasticity may be influenced by normal environmental factors to establish phenotype. Moreover, the transgenerational passage of epigenetic marks, at least in animals, the extent of epigenetic processes beyond imprinted genes and beyond only the early embryonic period, and the possibility that environmentally induced epigenetic change can affect the probability of mutation are challenging long-held views about the inherited risk of disease. The nature-nurture divide is now widely recognized not just to be artificial but as meaningless, and the genocentric view of life which underpinned neo-Darwinism is now severely challenged. The Central Dogma of the Modern Synthesis, and the genetic determinism which followed, may have come to the end of their scientific usefulness. In DOHaD, a new era of applied developmental physiology is beginning.
We believe that the implications of such new perspectives will be far-reaching. They will have direct relevance to the initiatives badly needed to address the challenge of NCDs globally, by emphasizing the importance of including development at the start of the lifecourse in interventions, and in moving beyond the emphasis on adult lifestyle and personal choice in NCD prevention. This will have implications for the health literacy and the empowerment of women of reproductive age and their partners, of children, adolescents, and young adults. It will require wider appreciation of developmental physiology, not only among health professionals but among policy makers and opinion leaders. It will raise many issues concerning the distinction between human biology and our culture, from ethical to philosophical to ideological. And it will underpin the success of interventions to improve health across the lifecourse, operations which will have substantial financial implications across both low and high income societies.
M. A. Hanson is supported by the British Heart Foundation and P. D. Gluckman by Gravida (National Centre for Growth and Development, New Zealand).
No conflicts of interest, financial or otherwise, are declared by the authors.
We are grateful to Jane Kitcher for her patient preparation of the manuscript and to Felicia Low for additional research.
Address for reprint requests and other correspondence: M. A. Hanson, IDS Building, MP887, Southampton General Hospital, Tremona Road, Southampton, SO16 6YD, UK (e-mail:).
↵1 Since writing this review, sadly David Barker died in August 2013. The authors wish to emphasize the enormous contribution that he made to this field and hope that this review will serve as a tribute to his pioneering work.
- Copyright © 2014 the American Physiological Society