Evolutionary and developmental mismatches are consequences of adaptive developmental plasticity in humans and have implications for later disease risk

Abstract

A discrepancy between the phenotype of an individual and that which would confer optimal responses in terms of fitness in an environment is termed ‘mismatch’. Phenotype results from developmental plasticity, conditioned partly by evolutionary history of the species and partly by aspects of the developmental environment. We discuss two categories of such mismatch with reference primarily to nutrition and in the context of evolutionary medicine. The categories operate over very different timescales. A developmental mismatch occurs when the phenotype induced during development encounters a different environment post-development. This may be the result of wider environmental changes, such as nutritional transition between generations, or because maternal malnutrition or placental dysfunction give inaccurate information about the organism's likely future environment. An evolutionary mismatch occurs when there is an evolutionarily novel environment. Developmental plasticity may involve immediate adaptive responses (IARs) to preserve survival if an environmental challenge is severe, and/or predictive adaptive responses (PARs) if the challenge does not threaten survival, but there is a fitness advantage in developing a phenotype that will be better adapted later. PARs can have long-term adverse health consequences if there is a developmental mismatch. For contemporary humans, maternal constraint of fetal growth makes PARs likely even if there is no obvious IAR, and this, coupled with the pervasive nutritionally dense modern environment, can explain the widespread observations of developmental mismatch, particularly in populations undergoing nutritional transition. Both developmental and evolutionary mismatch have important public health consequences and implications for where policy interventions may be most effective.

This article is part of the theme issue ‘Developing differences: early-life effects and evolutionary medicine'.

1. Introduction

Evolutionary medicine employs an evolutionary perspective to improve the understanding of human health and disease, at the levels of both the individual and the population. Apart from its general interest and hypothesis-generating capacity, it aims to inform healthcare professionals and policy-makers. The fundamental principles of evolutionary medicine [1] are that, first, selection serves to maximize reproductive success (Darwinian fitness) rather than to enhance health or longevity; secondly, because of the complexities of parallel selection for multiple traits underpinned by multiple interacting gene pathways and epistasis, this inherently involves trade-offs; thirdly, even where there is strong selection for a particular trait it can only evolve to the point where the benefits are balanced out by the costs; and fourthly, the consequences of trade-offs may be apparent immediately, or they may be delayed. Indeed, the concept of antagonistic pleiotropy demonstrates how one trait may be adaptive earlier in life but maladaptive later in life [2]. Further, it is now apparent that phenotypic effects can be passed across generations by mechanisms not mediated by changes in fixed genetic sequences, and that this can simultaneously affect multiple members of the population, thus acting much faster than genetic selection.

Evolutionary processes can affect health and disease risk via multiple pathways (reviewed in [1]). One of the challenges of evolutionary medicine is moving from hypothesis formulation to hypothesis testing. Nesse [3] has described how evolutionary hypotheses might be tested, pointing out the importance of consistency between the hypothesis and evolutionary thinking, being sure of the object of explanation, being clear about whether one is dealing with a proximate or an ultimate explanation, and recognizing the value of modelling and of comparative and experimental studies.

Evolution primarily involves multigenerational change which can only be directly observed in rapidly reproducing organisms. Thus, given the long generation time in humans, the possibility of designing evolutionary experiments in humans is highly unlikely. Rather, the likely validity of an evolutionary hypothesis has to be inferred through a number of other ways, such as natural experiments in humans, and demographic or historical patterns. Here, the effects of early development and later lifestyle/environment offer a rich source of material.

It is well recognized from human epidemiological studies, which are supported by experimental observations in animals, that events in fetal and infant life can induce long-term pathophysiological changes in humans [4]. In this paper, we place these observations in an evolutionary context, discussing how developing organisms respond to environmental cues, and the two major pathways by which long-term disease risk may be increased, namely evolutionary mismatch and developmental mismatch. Both pathways involve the organism facing environmental challenges. In the context of developmental plasticity, evolutionary mismatch involves exposure during development to environments with which the organism's lineage has not evolved to cope. By contrast, developmental mismatch arises when the organism in early life uses the mechanisms of adaptive developmental plasticity to develop a phenotype suitable for one environment, but is later exposed to a different environment from that to which its anatomy and physiology have adapted. In relation to developmental mismatch, we also discuss the fitness promoting effects of anticipatory responses in development, and outline the evidence for predictive adaptive responses (PARs) in humans.

2. Developmental responses to environmental cues

A developing organism that is exposed to a particular environmental cue may respond in several ways [5]. When the cue is particularly severe, for example, in terms of deficient nutrient or oxygen provision, it may disrupt the developmental programme and induce teratogenesis. However, a less severe cue may instead adjust the developmental programme and change the relationship between inherited genotype and phenotype expression. This phenomenon is generally termed developmental plasticity [6,7]. In both types of effects, the environmental cue or challenge may have to be present during a critical or sensitive period of development, with teratogenesis usually having a more tightly defined window and being irreversible (e.g. neural tube defects associated with folate deficiency).

Developmentally plastic responses may be reflected in structural change, such as a reduction in nephron number, a regulatory change potentially associated with epigenetic alterations or changes in neurodevelopment and learning behaviour. Developmental plasticity is conserved across taxa, suggesting that it is generally adaptive and thus promotes fitness. In fast-reproducing organisms, and particularly in asexual organisms, this may take the form of bet-hedging, where a range of phenotypes is produced randomly or in response to environmental stressors [8].

By contrast, developmental plasticity in sexually reproducing species is generally directional. It may occur in response to a regularly occurring cue, such as in the meadow vole (Microtus pennsylvanicus), a small rodent that develops with a thick or thin coat of fur in utero according to whether it is born in autumn or spring. This is a response to changes in day length, and this cue is transmitted through the placenta via altered maternal melatonin levels [9]. At the time of birth, the choice of fur thickness is likely of no significance, but it becomes an important survival factor as winter or summer arrives.

Developmental plasticity may also occur in response to more irregular cues, such as in the desert locust (Schistocerca gregaria), which responds to signals indicating crowding that determine whether the offspring develop into the solitary or gregarious morph—two phenotypically distinct forms that differ in the body and brain size, colour, dietary preferences and behaviour [10]. The signals may be received via exposure to a gregarizing agent after hatching, or to other population density cues in the early larval instar stages.

Two important aspects of developmental plasticity need to be noted. Firstly, maintaining developmental plasticity indefinitely across the life course is not possible because it would be too costly; the programme of development is not reversible, and thus plasticity declines with age [11]. Secondly, developmental plasticity can involve trade-offs. For example, a spadefoot toad tadpole senses reduced swimming space as drying up of the pond. This prompts early metamorphosis, leading to an adult toad which can survive to reproduce but which is smaller and more likely to be predated [12]. In mammals, exposure of a developing fetus to a non-teratogenic cue, such as reduced oxygenation or nutrition, may induce an immediate adaptive response (IAR) such as reduced fetal growth. This aids survival on limited resources until birth, even though there are post-natal consequences of reduced fetal growth in the form of poorer survival and increased morbidity. In both of these examples, the adaptation—whether as accelerated metamorphosis or as fetal growth restriction—has immediate survival value but incurs post-natal costs [13].

By contrast, the case of the meadow vole demonstrates that the adaptation for fitness may not become obvious until later in life. In the snowshoe hare, maternal signals of predation risk, linked to predator population density, lead to altered stress responses and attention behaviours in the offspring, conferring a post-natal, but not a prenatal, survival and thus fitness advantage.

Adaptive developmental plasticity can only evolve to respond to circumstances that the lineage has previously experienced, so there is a historical path dependency to what is possible. Thus, evolutionary novelty—that is, exposure to an environment that the lineage has never before been exposed to—creates a different challenge for the organism. As an example, there are natural plant chemicals that are toxic to humans but not to some other species that have evolved to cope with and ingest them [14].

(a). Special considerations in humans

An analysis of developmental plasticity in humans requires consideration of specific aspects of our life-history traits and behaviours. Humans have a relatively long period of dependent development: the fetal, infant and childhood periods last for about 7 years, in comparison to the relatively short period of reproductive competence in females, spanning perhaps 25 years in our evolutionary past. Consequently, humans display a high level of parental investment in their offspring and have relatively low fecundity, although relatively early weaning and alloparental care to meet the demands of post-weaning dependency may permit birth interval to be shorter [15]. Prior to modern times and the development of early weaning, maximum fecundity was likely to be about five to six children, of which only two to four would survive to puberty. Modelling work has shown that fitness in humans is largely determined by the survival rate of offspring that, in turn, are able to reproduce [16].

These considerations demonstrate that the evolutionary relevance of developmental plasticity generally applies to the first four decades in life, especially in women, beyond which little adaptive advantage is gained (or lost). This can account for why non-communicable diseases (NCDs) that tend to be manifest in the post-reproductive age, such as obesity, type 2 diabetes and cardiovascular disease generally do not affect human fitness greatly, although we note that these diseases are now increasingly seen in younger people. At the same time, it has long been recognized, because of the pioneering work of David Barker and others, that there is an association between early life events and the later risk of such NCDs [17,18]. This Developmental Origins of Health and Disease (DOHaD) paradigm involves at least two distinct groups of pathways—evolutionary mismatch and developmental mismatch—which are discussed later.

An additional, important consideration is that while some species such as the termite and beaver have evolved as niche constructors [19], creating a constant environment through their activities and behaviours, modern humans by contrast constantly modify their environment (niche modification) rather than create equilibrium. Humans are uniquely able to exploit a wide range of attributes including advanced cognitive and learning skills, manual dexterity and a social structure that allows knowledge to be distributed within a group, thus facilitating cultural evolution. All of these attributes enable progressive and continuous innovation. This pace of change has accelerated through the exponential increase in the development of technologies [20], exposing contemporary humans to environments that are completely novel in the context of our evolutionary history.

3. Evolutionary mismatch

Evolutionary mismatch occurs when the environments experienced by contemporary humans are markedly different from those encountered throughout most of our evolutionary history [21]. Such marked differences can arise in two ways, the first being exposure to an entirely novel environment. Such evolutionary novelty is well illustrated by scurvy, a disease caused by vitamin C deficiency. Our evolutionary lineage from frugivorous ancestor species meant that a frameshift missense mutation occurring in a key gene of the vitamin C biosynthetic pathway was neutral and not selected against. However, it was not until around 300–400 years ago when humans starting embarking on long oceanic journeys in significant numbers, and later when urban environments contributed to very poor diets among the socioeconomically disadvantaged, that scurvy began to become apparent due to lack of access to dietary vitamin C [22].

The second way by which evolutionary mismatch arises is when an individual is exposed to a particular environment that exceeds the range previously experienced in the lineage, leading to more subtle alterations in physiology and structure. A classic example of this is the high availability of appetite-rewarding energy-dense foods, and the inclusion of processed foods and high-fructose corn syrup as mainstays of the Western diet. As extensively discussed elsewhere [23,24], obesity and its comorbidities such as type 2 diabetes can be largely explained by our physiology not having evolved to cope with the considerable changes in our nutritional environment.

Another example relates to the replacement of human breast milk with bottle feeding using cow's milk-based formula. Cow's milk differs in its energy profile and its composition of micronutrients and bioactive compounds, leading to differential nutrient supply, transfer of immunity and impact on the microbiome. In addition, bottle feeding involves different maternal–infant cues from breast feeding [25]. For the latter, the infant is largely determining when s/he is satiated, whereas with bottle feeding, it is the feeder who largely determines when feeding ceases. This therefore has implications for the development of satiety. Bottle feeding with cow's milk-based formula is a novel exposure during a time of significant plasticity, and its impact is increasingly recognized in terms of greater risk of later obesity, altered immune function, and potentially altered neurodevelopment and behavioural development [26–28].

Other, more subtle evolutionary mismatches may explain the impact of maternal obesity and gestational diabetes on the offspring, who are themselves at greater risk of obesity [29]. These maternal characteristics were likely to have been rare in our evolutionary past due to the nature of pre-modern diets, and also because common concomitants of untreated gestational diabetes are fetal macrosomia, which would have been fatal for both mother and infant, and severe hypoglycaemia in the neonate, which without modern medical treatments would potentially lead to infant mortality. Modern epidemiological studies show that rates of gestational diabetes are low until the population has undergone a nutritional transition to the energy-dense modern Westernized diet [30,31]. At a physiological and cellular level, gestational diabetes leads to increased trans-placental transfer of glucose because the placenta has not evolved to limit the transfer of glucose to the fetus; this in turn leads to increased fetal insulin release and adipogenesis [32]. The mere presence of greater numbers of adipocytes in the offspring is itself associated with greater risk of later life obesity and type 2 diabetes.

There may be other forms of evolutionary mismatch that have impacts on later development. Emerging evidence from studies involving in vitro fertilization and other reproductive technologies, where the early embryo is exposed to very unusual conditions, suggests that some DOHaD-like effects may result [33]. There is also increasing evidence that being born by Caesarean section may have long-term immunological effects, perhaps mediated through changes in the microbiome [34].

4. Developmental mismatch

The concept of developmental mismatch differs in distinct ways from that of evolutionary mismatch, although by definition, the processes of developmental plasticity operate in the context of the evolution of the species. Thus, the adverse effects of exposure to infant formula rather than breastmilk, or to high maternal blood glucose levels in gestational diabetes, result from developmental effects of evolutionary novelty. Developmental mismatch involves the developing organism responding to environmental cues in ways that evolved to be adaptive. The cue may lead to an IAR and incur trade-offs as discussed earlier; it may lead to a delayed response, i.e. a PAR; or it may lead to both an immediate and a delayed response [13]. It is important to note that risk of a maladaptive outcome is only increased if a delayed response occurs and is then accompanied by a developmental mismatch—that is, the delayed response has stemmed from a faulty prediction.

IARs represent an outcome of the fetus being exposed to a concurrent maternal cue and needing to respond in order to promote its ability to survive until birth. In early gestation, the response to reduced fetal access to oxygen and/or nutrients is characterized by symmetrical growth restriction. In late gestation, the response produces asymmetrical fetal growth restriction [35] by which the fetus reduces its somatic growth, protecting vital organs such as the brain and heart at the expense of other organs, to maximize its chances of survival. Indeed, such fetal growth restriction is often associated with increased insulin sensitivity in the neonate [36]; this may reflect the need to receive the benefit of lipid-rich breast milk and lay down fat quickly, which are possible mechanisms for metabolic buffering of the brain in the case of later gastrointestinal insults, particularly around weaning [37]. However, there are costs to fetal growth restriction: after birth, there are greater risks of infant death and childhood morbidity, and therefore reduced fitness. Nevertheless, as discussed later, fetal growth restriction, being an IAR, is usually also associated with a PAR, which paradoxically will lead to insulin resistance appearing later in childhood [36,38]. There are examples in animals in which prenatal stress may lead to accelerated growth post-natally [39], which may similarly represent the compensatory effects of a PAR following an IAR.

Other forms of IAR are mild prematurity in response to maternal nutritional and stressor cues during the later stages of pregnancy, where the fetus may slightly shorten its gestational experience so as to be born into a more robust post-natal nutritional environment. Similarly, premature delivery in response to infection demonstrates the comparatively greater safety in being born early; in spite of a high risk of infant death, the fetus is able to minimize exposure to the infectious agent and survive.

(a). Predictive adaptive responses

PARs, also known as anticipatory maternal effects [40], differ from IARs in that the developing fetus or infant responds to a cue by changing its developing physiology to better adapt to the predicted, but not certain, future environment [41,42]. Classic ecological examples of PARs are the meadow vole and the desert locust responding to regular and irregular cues, as discussed earlier. In both cases, the prediction is likely to be reasonably certain. With respect to the human, the fetus or infant responds to maternally mediated cues that predict the environment in which it will grow until it is reproductively competent.

An appropriate interpretation of PARs and their implication for evolutionary thinking requires an understanding of several important features. Firstly, PARs are often, although not invariably, associated with an IAR. Secondly, it has been shown that predictions are not required to be absolutely accurate for PARs to have evolved, especially when the timescale of environmental shifts is about the same as the species' generation time [43]. Thirdly, it is not necessary that PARs have functional symmetry. That is, the consequences of incorrectly predicting a nutritionally rich environment are not necessarily the same as that of incorrectly predicting a nutritionally constrained environment. As discussed later, humans are biased towards predicting constrained environments.

PARs would have evolved in and been particularly important for humans, because of the high investment that is made in every pregnancy. Indeed in every species, exposure of the mother to a stressor will prompt decisions on whether the fetus needs to be sacrificed to protect maternal fitness, or whether further fetal investment is required to maintain its fitness, even at potential cost to the mother. There are numerous circumstances under which considerations of the trade-off between maternal and offspring fitness may be required, as discussed elsewhere [40,44,45]. In humans, the low fecundity and the very high level of maternal investment in every pregnancy suggest that there would be some degree of preference for protecting both offspring fitness and maternal fitness.

This maternal–offspring fitness trade-off can be observed in certain natural experiments. In response to famine, humans unlike many other mammals (but see [46]) do not completely suppress fertility. In the Great Leap Forward famine of China in the 1970s, fertility was only reduced by about 50% [47]. Similarly, births still occurred in the concentration camps of the Nazi Holocaust. Thus, we argue that PARs occur in humans and other species where the life-history traits suggest that protection of offspring fitness is given some priority from early in development.

A basic principle of PARs is that the developing organism changes its physiology to match the predicted future environment. For example, in the desert locust, the solitary form that develops in the absence of evidence of crowding has morphological, physiological, and behavioural characteristics adapted to ample food availability without heavy investment in mobility. Had the developing locust received signals predicting a crowded environment, it would have adopted the migratory form with a very different morphology and metabolic physiology, and greater investment in flight ability [10]. In the case of humans, with a long gestational period, it has been proposed that fetal responses involve a level of inertia [48]. This is supported by a study of fetal sheep exposed to maternal undernutrition, in which a permanent resetting of the growth channel only occurred with prolonged undernutrition [49].

It is assumed that forms of homeorhetic change in both the placenta and the mother provide the fetus with metabolic or hormonal signals that indicate the likely nature of the post-natal environment. We suggest that PARs may also affect basic life-history traits. For example, being born with a lower birth weight is associated with an accelerated rate of maturation [50]. These clinical observations accord with conceptual arguments proposed by Nettle [51] that children who are born in stressful circumstances are likely to live ‘shorter and faster’ lives.

Importantly, the adaptive advantage conferred by PARs is associated with childhood survival and adult reproduction, not later health or longevity. This, therefore, necessitates a focus on the period of childhood and adolescence survival. In humans, the most common cues that induce PARs are likely to be early nutritional cues. If the developing fetus predicts a post-weaning environment of poor nutrition, there is advantage in reduced somatic growth, and a tendency to lay down fat when possible—that is, to develop a degree of insulin resistance. There is extensive experimental and clinical evidence that poor nutrition in early life leads to obesity and altered satiety [52], altered maturation [50] and insulin resistance in children [53]. The observation that insulin resistance only occurs after weaning even though the cues occur early in life is noteworthy. Thus, an infant born with fetal growth restriction due to placenta dysfunction (an IAR) may be insulin hypersensitive then develop insulin resistance at the age of two to three as a PAR.

PARs induced by nutritional cues may occur in isolation from IARs. This is perhaps best demonstrated by the work of Godfrey et al. [54], who showed that epigenetic changes at birth were associated with maternal food intakes in early pregnancy, and that these epigenetic changes in turn were associated with level of adiposity in the children 6–9 years later. Importantly, the epigenetic alterations involved pathways that were biologically plausible.

(b). Malprediction

Developmental mismatch—that is, where PARs induce maladaptive outcomes or evidence of chronic disease later in life—does not arise unless the prediction is inaccurate because the later life environment is different from that predicted in early life. Thus, for example, if a low nutrition environment is predicted in utero and the same nutritionally scarce environment is encountered post-natally, the child may grow to be somatically small, as reflected in stunting, but there are no other long-term consequences beyond those associated with stunting itself. By contrast, if a low nutrition environment is predicted but a nutritionally replete environment is experienced, a developmental mismatch occurs and the risks of NCDs including type 2 diabetes and obesity are increased [23].

There are several ways by which erroneous predictions are made and lead to a developmental mismatch. A major reason is when maternal pathophysiology and/or maternal behaviour gives rise to a false cue. As further discussed later, a degree of maternal constraint of fetal growth operates in all pregnancies, resulting in the fetus predicting a slightly worse post-natal environment than currently exists, which may err on the safe side in terms of later fitness [55]. In addition, maternal illness, an unbalanced maternal diet or placental compromise limiting nutrient transfer to the fetus may falsely signal malnutrition. Alternatively, there may be a significant change in the post-natal environment compared to that in which the mother had lived during pregnancy. This may be due to migration, a rapid nutritional transition within a region or country, or simply the child adopting dietary or other behaviours quite different from the mother. As studies in rodents have shown [52,56], PARs do not induce pathophysiology, but rather change the sensitivity of the organism to the post-natal environment, potentially making mismatch and its adverse health consequences more likely. This amplification effect has also been well demonstrated in a large human cohort study, which showed that at each birth weight range, the relative risk of later developing type 2 diabetes was raised by the number of unhealthy post-natal lifestyle behaviours adopted [57].

Critically, the human life history makes it likely that most fetuses will be exposed in some degree to cues that prompt predictions of lower post-weaning nutritional availability. Maternal constraint describes the situation where fetal growth does not reach its full genetic potential, and is instead inherently limited to ensure birth through the pelvic canal [58]. Maternal constraint is particularly important in humans for two main reasons: the first is the relatively narrow human pelvis compared to that of other primates, itself an outcome of the evolution of bipedalism [59]; the second is the proportionally very large head of the human fetus, which is born in a relatively altricial state compared to other primates. Clinical data support the notion that maternal constraint operates in all humans. As shown by Vasak et al. [55], the lowest levels of neonatal morbidity are observed at birth weights much higher than the median, at around the 80th birth weight centile, suggesting that size at birth is constrained below that at which maximum perinatal survival can be achieved. An important ramification of maternal constraint is that its effects are compounded by the excessive nutritional environment we now face, placing virtually all humans at risk of malprediction and hence developmental mismatch [35,58].

Several critiques of the PAR model have previously been made [60–62], and these appear to have arisen from misunderstandings of the associated conceptual framework. At a broader level, there has been a conflation of developmental and evolutionary mismatch. More specifically, there has been a failure to recognize that, first, the consequences of any developmental cue on maternal and offspring fitness depend on the life-history traits of that species [40]; second, that prediction does not have to be extremely accurate to confer adaptive advantage and therefore to evolve; third, that PARs in humans are about survival to reproductive maturity, and associated life-history traits; and fourth, that the life-history traits of humans differ markedly from highly fecund and polytocous species usually studied experimentally, and therefore that strategies derived from the latter may not apply to humans. Furthermore, critiques that assume a need for symmetry between predicting future high nutritional levels and low nutritional levels ignore the phenomenon of maternal constraint. Most importantly, PARs remain covert when developmental mismatch does not arise; they, therefore, have no overt physiological consequences unless there had been a concomitant IAR. The distinction between PARs and IARs is critical for appropriate interpretation of the model [42].

(c). Predictive adaptive responses in humans

Evidence for the presence of PARs in humans requires identification of a natural experiment in which a fitness effect, reflected in a survival difference, can be demonstrated. There also needs to be biological plausibility, and the observations should be compatible with experimental data.

Studies using two distinct manifestations of the challenge of severe acute malnutrition in infancy, known as kwashiorkor and marasmus, provide such an example. Both syndromes coexist in the same population under similar nutritional stresses [63]. Infants with marasmus tend to show severe wasting in response to malnutrition, whereas those with kwashiorkor become oedematous and are more prone to infection, with fatal outcomes. Therefore, untreated kwashiorkor has a very high mortality compared to untreated marasmus. When acutely ill, kwashiorkor infants tend to downregulate protein and lipid turnover more than do marasmic infants, and thus have a large retention of apparently mobilizable nutritional stores [64,65]. These differences in metabolic patterns suggest that marasmic infants are metabolically thrifty, whereas kwashiorkor infants are unable to use their nutritional stores for survival—a characteristic that has been termed metabolically profligate. These differences persist beyond the severe episode of malnutrition itself, and are demonstrable several months after full nutritional rehabilitation.

The persistent differences raise the question of whether kwashiorkor and marasmic infants also differed metabolically prior to their malnutrition episode. In association with other colleagues, we, therefore, studied a cohort of Jamaican children who had been diagnosed with kwashiorkor or marasmus, and for whom birth weights were available. Data analysis revealed that, as hypothesized, kwashiorkor children were born with significantly higher birth weights (333 g, 95% confidence interval 217, 449, p < 0.001) [66]. When these children were reassessed as young adults, those who were marasmic were more likely to be of shorter stature, and, compared to kwashiorkor survivors, or community or birth weight controls, to have impaired glucose intolerance associated with deficiencies in insulin sensitivity and β-cell function [67]. Epigenetic analysis of skeletal muscle biopsies identified significant differences in methylation over 133 CpG loci. These loci included genes implicated in glucose metabolism, body size and body composition [68], lending biological plausibility to the findings.

Collectively, the studies on the Jamaican survivors of acute childhood malnutrition support the hypothesis that the marasmic individuals had responded to intrauterine cues by predicting that they will live in a malnourished post-natal environment, and thus developed a metabolically thrifty phenotype. However, when exposed to the post-natal high nutrition levels now increasingly prevalent in Jamaica, these individuals are developing glucose intolerance at a relatively young adult age, which might be partially explained by epigenetic alterations [69]. By contrast, kwashiorkor children failed to predict a nutritionally deficient post-natal life and so developed with a more metabolically profligate metabolic phenotype; then, when faced with malnutrition, they were unable to adequately downregulate their metabolism and mobilize their nutritional stores, which increased their mortality risk. However, as adults they have become better matched to the more nutritionally rich environment, and thus were less likely to get type 2 diabetes and other comorbidities.

Because untreated kwashiorkor is associated with a very high rate of infant mortality, there is a clear fitness difference between developing a phenotype likely to become marasmic versus kwashiorkor in response to infant undernutrition, and maintaining the phenotypic difference through to later life, during which time further episodes of malnutrition would have been more likely historically [66]. This fitness difference therefore strongly supports the PAR hypothesis, as follows: the lower birth weight in marasmic children reflects an IAR; the changed metabolic profile of these infants is a PAR enabling metabolic thrift in response to severe childhood malnutrition, and hence greater chance of survival; and the greater risk of adult obesity and diabetes in the marasmus survivors who are now facing a nutritionally rich environment represents a developmental mismatch. This illustrates the interaction between the IAR, the PAR and then the developmental mismatch.

5. Summary and conclusion

The particular evolved characteristics of the human mean that offspring fitness is relatively conserved relative to maternal fitness under situations of maternal stress. This may be manifest as either an IAR and/or a PAR. PARs in isolation have no long-term consequences unless there is a developmental mismatch. In the context of contemporary humans, the operation of maternal constraint makes the induction of PARs likely even if there is no obvious IAR, and this, coupled with the pervasive nutritionally dense modern environment, can explain the widespread observations of developmental mismatch, particularly in populations undergoing nutritional transition [70].

On the other hand, evolutionary mismatch arising from cues such as infant feeding with cow's milk-based formula, or exposure to gestational diabetes or severe maternal obesity, is likely to induce offspring changes through different mechanisms such as by increasing fat cell numbers. Both, paradoxically, lead to adult offspring at greater risk of obesity, diabetes and other comorbidities, which accounts for a U-shaped curve between birth weight and the risks of disease in adulthood [71–73].

Although distinct in origins, timescale and mechanisms, both developmental mismatch and evolutionary mismatch have important public health consequences, but also have clear implications for where policy interventions may be most efficacious [74,75]. For example, measures such as reducing the rates of gestational diabetes and promoting breastfeeding can minimize the prevalence of evolutionary mismatch, while promoting prudent maternal nutrition from prior to conception and throughout pregnancy is likely to reduce the risk of developing erroneous PARs. Optimal prenatal and preconceptional health is critically important in reducing the risks of disease in later life, and ensuring that children's diets are healthful will reduce the risk of developmental mismatch.

Data accessibility

This article has no additional data.

Authors' contributions

P.D.G. and M.A.H. developed the conceptual framework; F.M.L. contributed to interpretation. P.D.G. and F.M.L. wrote initial draft, M.A.H. provided critical revisions; all authors reviewed and approved final version.

Competing interests

M.A.H. is the theme issue guest editor; P.D.G. and F.M.L. have no competing interests.

Funding

M.A.H. is supported by the British Heart Foundation, the NIHR Southampton Biomedical Research Centre and the EU Horizon 2020 programme LifeCycle.