| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Review Article |
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONEPIDEMIOLOGIC EVIDENCE OF...ANIMAL MODELS OF METABOLIC...POTENTIAL MECHANISMSARE SPECIFIC MECHANISMS...SUGGESTIONS TO GUIDE FUTURE...CONCLUSIONREFERENCES |
|---|
Key Words: Metabolic imprinting • nutrition • development • early nutrition • chronic disease • obesity • cardiovascular disease • hypertension • type 2 diabetes mellitus • animal models • epidemiology • metabolism
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONEPIDEMIOLOGIC EVIDENCE OF...ANIMAL MODELS OF METABOLIC...POTENTIAL MECHANISMSARE SPECIFIC MECHANISMS...SUGGESTIONS TO GUIDE FUTURE...CONCLUSIONREFERENCES |
|---|
Ongoing success in enhancing life expectancy and the desire to improve the quality of life at older ages have focused increased attention on diseases with adult clinical onset. The predicted global increase in the proportion of persons >65 y of age (1), the ineffectiveness of treatment options for many diseases of old age, and rising medical costs each argues strongly for a greater focus on preventive strategies.
But at what age is it appropriate to begin targeting preventive approaches? Perinatal nutritional status is recognized to have a profound and persistent influence on neurologic development and cognitive function. Similarly, classic epidemiologic studies of survivors of the Dutch famine of 1944–1945 suggest that perinatal nutrition influences adult body mass index (BMI), often considered a proxy for body composition. More recent evidence that links low birth weight to an increased risk of chronic diseases in old age has generated renewed interest in preventive approaches initiated in the perinatal period.
Specifically, data from human epidemiologic studies and animal models strongly suggest that perinatal nutrition affects predispositions to adult obesity, cardiovascular disease (CVD), hypertension, and type 2 diabetes (see references 2–9 for reviews). For some, a focus on the perinatal period seems misplaced. Adults with modifiable risk factors may misconstrue discussions of predetermined risk as added excuses for not altering current behavior. Alternatively, the possibility of significantly reducing long-term risk via focused interventions in early life suggests an efficacious complement to prevention strategies undertaken at later life stages.
Understanding the underlying biology, however, is a prerequisite for the design of prospective clinical studies of the effects of perinatal nutritional status on adult disease and for the possible identification of interventions to improve health throughout peoples' lifetimes. As is reviewed below, such interventions may be particularly important in populations that experience high rates of low birth weight. Whatever the public health implications of data that link perinatal nutrition to adult disease may be, exploring the underlying mechanisms is in itself important. The underlying biology of these putative relations is at best only generally understood and is amenable to investigation. Accordingly, although salient findings from epidemiologic and animal studies linking early nutritional experiences to the expression of metabolic alterations in adulthood will be summarized, the principle focuses of this review are possible underlying biological mechanisms and the identification of experimental approaches that may help test mechanistic hypotheses.
Links between nutrition and common morbidities
Adult obesity, CVD, hypertension, and type 2 diabetes are exceptional because of their high prevalence and complex interrelatedness. In the United States, data from the third National Health and Nutrition Examination Survey (NHANES III) indicate that one-third of US adults were overweight in 1991 (10). Recent estimates are even higher (11). Overweight puts individuals at elevated risk of CVD, hypertension, and insulin resistance and hyperinsulinemia. CVD is the most common cause of death in industrialized nations, especially among older age groups. One-third of US men and 10% of women develop CVD by age 60 y. CVD causes 70% of all deaths beyond the age of 75 y (12). Obesity and hypertension are established risk factors for CVD (13). More than 10% of Americans aged 65–74 y have type 2 diabetes (14). Hyperinsulinemia associated with type 2 diabetes and obesity-associated insulin resistance is a common component of syndrome X, which also encompasses dislipidemia, hypertension, and central obesity (15). Despite the common coexistence of those conditions, most epidemiologic studies that link perinatal nutrition to them, and animal studies that ascertain causality between them and specific nutritional conditions, consider obesity, CVD, hypertension, and type 2 diabetes individually. This review therefore treats each outcome individually, but also explores basic underlying biological mechanisms that may account for the concomitant occurrence of these morbidities.
Metabolic imprinting
The term metabolic imprinting is used in this review to describe the basic biological phenomena that putatively underlie relations among nutritional experiences of early life and later diseases. The term is intended to encompass adaptive responses to specific nutritional conditions early in life that are characterized by 1) a susceptibility limited to a critical ontogenic window early in development, 2) a persistent effect lasting through adulthood, 3) a specific and measurable outcome (that may differ quantitatively among individuals), and 4) a dose-response or threshold relation between a specific exposure and outcome. Although previous workers have used the term programming to refer to the long-lasting effects of early nutritional experiences, imprinting more effectively conveys the salient features detailed in the first 2 parts of the definition adopted for this review. There is a clear historical precedent for this term. Konrad Lorenz (16) chose imprinting to refer to the setting of certain animal behaviors that resulted from early experience. Central to his definition was the fact that imprinting may only occur during "a narrowly-defined period in the individual's life" (the critical period) and, moreover, that the imprinted behavior "cannot be `forgotten'!". These coincide with the first 2 parts of the definition of metabolic imprinting. The last 2 parts of the definition narrow the biological phenomena under consideration to specific relations appropriate for mechanistic characterization.
It is useful to distinguish metabolic imprinting from other biological phenomena that share similar qualities. Consider, for example, the developmental processes that enable the immune system to distinguish self from nonself. Early in development, fetal exposure to self antigens results in immunologic self-tolerance. Exposure to specific nonself antigens during this period results in persistent tolerance of them as well. Fetuses have an exceptional ability for tolerance induction, and this ability is reduced dramatically with age (17), suggesting a graded critical window that, although maximally sensitive in utero, extends into adulthood. Hence, the immunologic memory conveying tolerance to self and nonself antigens shares many of the hallmarks of imprinting phenomena.
Another relevant example is hormonal imprinting. If day-old male rats are injected with supraphysiologic concentrations of insulin, a permanent decrease in insulin responsiveness results that is correlated with a defect in insulin binding to hepatic membrane receptors. Such experiments suggest that early exposure to a hormone plays a role in the ontogeny of its cell-surface receptors (18), resulting in a sort of endocrinologic memory. Injecting adult rats with insulin has no such persistent effects, indicating that the critical window for imprinting insulin receptors does not extend into adulthood in rats.
|
EPIDEMIOLOGIC EVIDENCE OF METABOLIC IMPRINTING |
|---|
|
TOPABSTRACTINTRODUCTIONEPIDEMIOLOGIC EVIDENCE OF...ANIMAL MODELS OF METABOLIC...POTENTIAL MECHANISMSARE SPECIFIC MECHANISMS...SUGGESTIONS TO GUIDE FUTURE...CONCLUSIONREFERENCES |
|---|
Body mass index
We will review the imprinting of BMI first because of its important implications for the other outcomes of interest. Among the most compelling epidemiologic evidence of metabolic imprinting of BMI is the retrospective study of young male army draftees who were exposed perinatally to famine conditions during the Dutch famine of 1944–1945. Ravelli et al (20) studied 300000 men aged 19 y and found that, compared with those born in non-famine-stricken control areas, men who were exposed in utero to famine conditions at some time during the first 2 trimesters experienced an 80% higher prevalence of overweight (P < 0.0005). Those exposed to famine for some period during the last trimester or the first 5 postnatal months experienced a 40% lower prevalence of overweight than in those not exposed (P < 0.005).
Others reported a high positive correlation between birth weight and adult BMI. These associations have supported speculations that birth weight is a proxy for adequacy of fetal nutrition, and thus that fetal nutrition modulates the risk of adult overweight. The high heritability of birth weight (21) and the well-established postnatal tracking of body size raise the question of whether maternal obesity may confound the association between birth weight and later body size. To address this question, Curhan et al (22) analyzed data from the US Nurses' Health Study. They reported that the women in high-birth-weight categories had increased odds of being in the highest adult BMI category. The odds were unaffected by adjustment for the respondent's characterization of their mother's figure at age 50 y, supporting the imprinting hypothesis. But recall errors and respondent biases limit the value of this approach. In a more robust study of 2880 monozygotic twins, Allison et al (23) found no correlation between intrapair birth weight differentials and BMI differentials at 
40 y of age. These data appear contrary to the imprinting hypothesis. Because monozygotic twins are genetically identical, this design controls completely for the influence of genetics on birth weight. However, these results may be biased by the many social forces that tend to artifactually increase the similarity between monozygotic twins. Interestingly, the intrapair birth weight differential and later intrapair differentials in both height and weight were highly correlated, suggesting metabolic imprinting of body size but not BMI.
Whatever the mechanism, the possibility of a biological link between birth weight and adult BMI complicates the numerous studies that link birth weight to other adult outcomes such as CVD, hypertension, and insulin resistance. Acknowledging the strong correlation between overweight and these other outcomes, many authors adjust for BMI when evaluating the associations of outcomes of interest with birth weight. There are 2 reasons for this. First, one may wish to reduce the error variance in outcome by using BMI as a covariate. However, doing so may lead to misleading interpretations (24) because BMI does not follow the same distribution in different birth weight categories. Second, because adult BMI may be an intervening variable between birth weight and adult outcome, one may wish to correct for this to determine the magnitude of the birth weight effect that is not mediated by BMI. However, this would require that the variance in BMI be partitioned into 2 components, one associated with and the other independent of birth weight (Figure 1
). Although this conceptual model has theoretical appeal, it is unclear how such partitioning could be estimated.
|
), one may assume that the component of BMI not associated with birth weight is most strongly related to lifestyle factors that would predict adult CVD risk. Hence, when comparing 2 individuals of the same adult BMI (ie, adjusting for BMI), the person with low birth weight would be expected to have a greater component of BMI not associated with birth weight (Figure 2). Accordingly, this person's BMI-associated risk of CVD would be higher. This rather simplistic model disregards potential interactions between birth weight and adult BMI and CVD risk, but nonetheless illustrates the potential for BMI adjustment to lead to incorrect interpretations. Thus, although it may be informative to analyze such relations by adjusting for BMI, the problems with such an approach suggest that associations adjusted and not adjusted for BMI should be reported. This has seldom been done.
|
In summary, data from several epidemiologic studies suggest that adult BMI may be imprinted both prenatally and postnatally. The most compelling data, drawn from the retrospective Dutch famine study, suggest that the critical window for BMI may extend throughout the perinatal period. The relation between birth weight and later BMI, whether reflective of imprinting or not, complicates the many epidemiologic studies reporting associations between birth weight and adult outcomes associated with BMI.
Cardiovascular disease
The evidence for early nutritional influences on later CVD was reviewed recently by Leon and Ben-Shlomo (28). The first indication that perinatal circumstances might affect susceptibility to coronary heart disease (CHD) in adult life arose from ecologic studies (29–31). These showed a strong correlation between birthplace and later CHD risk. Specifically, high rates of perinatal mortality appeared to be predictive of elevated adult CHD mortality. The simplest explanation of this relation is that of "continuity of circumstance." In other words, the same general poor environment that leads to high perinatal mortality may also act at later ages to predispose a population to elevated CHD mortality, as suggested by Ben-Shlomo and Smith (32).
Stronger links between perinatal nutrition and adult CVD were suggested by retrospective cohort studies of fetal and infant growth (based on records of birth weight and weight at 1 y of age) and adult incidence of CVD morbidity and mortality. Much of that evidence was derived from records that were collected by an exceptionally thorough midwifery system operating in Hertfordshire, United Kingdom, from 1911 to 1930. Osmond et al (33) linked mortality data to those records and tracked the individuals still living in the area. Analyses of the Hertfordshire data on >15000 individuals indicated that CVD mortality was inversely related to birth weight in both men and women, and that it was similarly related to weight at age 1 y in men but not in women (33). However, because those data at 1 y of age were not adjusted for birth weight, it is unclear whether the relations with weight at age 1 y indicate postnatal imprinting of CVD susceptibility or merely reflect the primary relation with birth weight. The relations between birth weight and CVD mortality in the Hertfordshire cohort were corroborated in part by morbidity data on a subsample of men (n = 290) who still lived in Hertfordshire. Fall et al (34) found an inverse relation between weight at age 1 y and prevalence of CHD, but no such trend was found with birth weight. This is inconsistent with an inverse relation between CHD mortality and birth weight.
Barker et al (35) reported relations between birth weight and CVD mortality in a cohort of 1586 men born in Sheffield, United Kingdom, from 1907 to 1924. CVD mortality was inversely related to birth weight in this cohort. Furthermore, ponderal index at birth (given as oz/in3) was inversely associated with CVD mortality.
As reported by Leon and Ben-Shlomo (28), the inverse relation between birth weight and adult CHD was confirmed in 4 of 5 retrospective studies: in 121000 women aged 30–55 y in the US Nurses' Health Study (36); in 517 men born in Mysore, India, between 1934 and 1954 (37); in 1335 men born between 1920 and 1924 in Uppsala, Sweden (38); and in 1258 men aged 45–59 y in Caerphilly, South Wales (39). However, concern remains regarding the inferences that may be drawn collectively from them. One source of uncertainty is the potential for biases imposed by the relatively high proportion of each birth cohort that was lost to follow-up as a result of migration or nonparticipation (19). This is a valid concern only if the relation between fetal growth and CVD differs between those who migrated and those who did not. Although such a difference is not intuitive, it is plausible. For example, a recent study suggests that these associations may be confounded by correlations between maternal CVD risk factors and probability of migration. Specifically, in a cohort of individuals born in Oxfordshire, United Kingdom, between 1965 and 1986, risk of migration was positively related to social class and inversely related to maternal systolic blood pressure (SBP). Also, the data suggested an increased risk of migration in the group with the lowest birth weight (40).
Another potential confounding variable acknowledged in all of those studies is socioeconomic status (SES), consistent with the alternative continuity of circumstance hypothesis. SES is associated with both birth weight and CVD risk. Although adjusting for SES does not alter the findings of most studies, such adjustments may often be inadequate, raising the possibility of significant residual confounding. For example, several studies that purported to adjust for social class at birth were unable to replicate well-established relations between social class and birth weight (19). Such unexpected results suggest that these studies may not have used a meaningful measure of social class. Conversely, in the Mysore, India, study (37), in which a broad range of well-defined social classes was represented, prevalence of CHD was correlated highly with social class. When the analyses were adjusted for social class, the effect of birth weight on CHD was no longer significant.
The relation between birth weight and adult BMI suggests another potential confounding factor in the association between birth weight and adult CVD risk. As discussed earlier, without adjustment for BMI, one would expect a positive correlation between birth weight and adult CVD because of the correlation between birth weight and adult BMI. Conversely, adjustment for adult BMI is expected to have the opposite effect. Note that inverse correlations between birth weight and adult CVD were found both in studies of CVD mortality (in which no BMI adjustment is possible) and in studies of CVD prevalence (some of which use BMI adjustment). Also, in one study that reported BMI-unadjusted and -adjusted data (39), the findings were not affected by the adjustment, supporting the theory that the birth weight–BMI association does not result in a spurious link between birth weight and adult CVD.
Although these methodologic issues are acknowledged, the preponderance of data suggests an inverse association between birth weight and adult CVD risk, and thus supports the prenatal imprinting of metabolic risk factors for CVD. The qualitative and quantitative similarities of this association among diverse cohorts underscore the generalizability of the association and likely basic biological underpinning. There is, however, little epidemiologic data suggesting that imprinting of adult CVD risk occurs in the early postnatal period.
Blood pressure
The association between fetal growth and adult blood pressure is well documented and suggestive of metabolic imprinting. A recent compilation of numerous studies supports the hypothesis that adult SBP is related inversely to birth weight over a wide range of birth weights (41). The earliest report of this relation focused on a 1946 British birth cohort. In that retrospective study, 3322 individuals were followed up to assess the influence of "social and family factors, smoking and body mass" at 36 y of age (42). Social class and birth weight were both correlated inversely with SBP at age 36 y; surprisingly, the authors did not emphasize the significance of the independent association with birth weight. In a later report, Barker et al (43) studied 449 English men and women identified in birth records from 1935 to 1943. In that sample, adult SBP was correlated inversely with birth weight and directly with placental weight. Further studies have corroborated this inverse correlation between birth weight and adult SBP in different populations, including young Croatian adults (44), elderly British women (45), 50-y-old Swedish men (46), and American women aged 30–55 y (22). However, in a retrospective study of 32580 Israeli adolescents, Seidman et al (47) found a slight positive correlation between birth weight and SBP at age 17 y.
Generally, adjustment for current BMI strengthens the already significant inverse association between birth weight and adult blood pressure. Thus, it is unlikely that this effect only reflects tracking of body size. Nonetheless, analyzing the data by current weight or BMI category has led to disparate results. For example, Leon et al (46) found that the birth weight–blood pressure correlation was strongest in men in the upper tertile of BMI. These authors suggested that excess adiposity in adulthood potentiates the detrimental infuence of lower birth weight on blood pressure. But others have not confirmed this interaction. Other reports state that the inverse correlation between birth weight and SBP is maintained across all current BMI categories; however, inspection of those data often indicates otherwise. Examination of several studies that categorized the data by current BMI suggests that the middle BMI category shows this relation most robustly.
The greatest weakness of this relation is its potential confounding by factors that are difficult to measure, and therefore to adjust for. First, because birth weight and SBP are associated positively and negatively, respectively, with SES, a large component of the birth weight–SBP relation could be mediated by an effect of continuity of circumstance via SES. Many studies include a proxy for SES in their models and find that the relation between birth weight and adult SBP remains significant and is often unattenuated by this adjustment. However, considering the difficulty in assessing SES, especially over a lifetime, residual confounding is likely (19).
Another potential confounding factor is maternal blood pressure. The simplest explanation for a relation between birth weight and adult blood pressure is that maternal blood pressure affects infant birth weight via some physiologic mechanism during gestation and also influences the offspring's adult blood pressure genetically. This possibility has been discounted on the basis of data showing that infant birth weight is unrelated to the blood pressure of pregnant women at their first antenatal visit (43). However, a recent study (48) found that maternal 24-h ambulatory blood pressure measurements at different stages of pregnancy are correlated inversely with infant birth weight. This finding suggests that genetic endowment remains a potential confounding factor between birth weight and blood pressure.
To summarize, retrospective studies in diverse populations have found that birth weight is inversely correlated with adult blood pressure. Although each of the studies has some weaknesses, together they support a biological link between intrauterine growth and adult blood pressure. It remains unclear whether this connection reflects imprinting of blood pressure or some other phenomenon. No epidemiologic evidence currently suggests the imprinting of blood pressure postnatally.
Impaired glucose tolerance and type 2 diabetes
Several large, retrospective cohort studies have identified an inverse correlation between birth weight and measures of glucose tolerance in adulthood. This literature was reviewed recently by McKeigue (49). The first published investigation of this relation measured plasma insulin and glucose concentrations (fasting and after an oral glucose load) in 408 men aged 64 y from the Hertfordshire birth cohort (50). After adjustment for current BMI, odds ratios for impaired glucose tolerance (IGT) or diabetes declined monotonically from 6.6 to 1 as birth weight category increased from 
2.5 to 
4.3 kg. This group later found a similar relation in another English cohort of 266 men and women aged 50 y (51). The odds ratios for IGT or newly diagnosed type 2 diabetes were 3.5-fold higher in men and 12-fold higher in women with birth weights 2.5 kg relative to those whose birth weights were >3.4 kg. Furthermore, intravenous insulin tolerance tests conducted in 103 members of this cohort showed that adult insulin resistance was likewise related inversely to ponderal index (kg/m3) at birth (the relation with birth weight was not significant) (52).
The relation between birth weight and adult glucose tolerance was corroborated further in a cohort of 413 Mexican American men and women with a mean age of 31 y from the San Antonio Heart Study (53). Specifically, fasting insulin concentration was related inversely to birth weight in women (P = 0.001), but no such relation was found in men. In a study of 1179 Pima Indians 20–39 y of age, McCance et al (54) related birth weight to IGT. Unlike others, they found a strong U-shaped association between birth weight and IGT and type 2 diabetes. Age-adjusted prevalence of type 2 diabetes in the lowest and highest birth weight categories was twice that in the middle birth weight categories. This relation appeared to hold across quartiles of age and across tertiles of current BMI. Furthermore, unlike previous studies, the association was of similar magnitude in men and women. Recently, Lithell et al (55) reported an association between birth weight and IGT in a cohort of 1333 Swedish men aged 50–60 y. They found that glucose concentrations 60 min after intravenous glucose infusion were significantly inversely correlated with birth weight. Moreover, after adjustment for current BMI, significant inverse relations were found between birth weight and fasting serum insulin and glucose concentrations.
These compelling studies are characterized by the same potential for confounding that is common to studies relating birth weight to the other outcomes that have been discussed. First, there is the problem of adjusting for current BMI. In all cases, such adjustments markedly enhanced the relation with birth weight, and in some studies, the unadjusted relation was not statistically significant. As discussed earlier, adjustment for a variable that may be in the causal pathway may lead to inappropriate conclusions. Also, it seems questionable that SES is adjusted for adequately in most studies. Several studies that used some proxy measure of SES to adjust for its effects reported no effect of the adjustment on the relation between birth weight and adult glucose tolerance. However, only the San Antonio study replicated the well-established inverse relation between SES and glucose intolerance (53). This suggests that other researchers either did not measure SES adequately or had too homogeneous a sample to detect its effect. Furthermore, genetics may also explain the association between birth weight and type 2 diabetes or IGT. For example, a fetus's genetic tendency for insulin resistance may reduce fetal growth directly and subsequently predispose the individual to glucose intolerance throughout life. In contrast with what is observed in maternal gestational diabetes, reduced fetal weight was shown in a rat model of streptozocin-induced fetal diabetes (56). This demonstration is consistent with this alternative explanation.
Other approaches to investigating potential prenatal imprinting of glucose tolerance have supported the inverse and in some cases U-shaped relations between birth weight and adult IGT. For example, Pettitt et al (57) studied offspring of Pima Indian women who were diabetic during the index pregnancy. To control for genetic influences on diabetes, the prevalence of diabetes in the 10–24-y-old offspring from diabetic pregnancies was compared with that in the offspring of prediabetic women (who developed diabetes at some time after the index pregnancy). Even after the mother's age of onset of diabetes was controlled for, the age-adjusted prevalence of diabetes in offspring of a diabetic pregnancy was twice that in offspring of prediabetics. These results are consistent with those of McCance et al (54) in suggesting that prenatal overnutrition may persistently increase an individual's susceptibility to glucose intolerance. In a recent retrospective study of 702 adults from the Dutch famine cohort, Ravelli et al (58) tested the hypothesis that prenatal undernutrition causes IGT in adulthood. They found that plasma glucose concentrations 2 h after a glucose load were higher (P = 0.006) in individuals who had been exposed to famine prenatally than in individuals born before the famine. Of individuals with prenatal famine exposure, the highest 2-h glucose concentrations were found in those who were exposed during late gestation. Because these individuals also had the lowest mean birth weight, these findings are consistent with the studies linking low birth weight to adult IGT.
Hence, as with the other outcomes discussed here, evidence suggests an association between prenatal nutritional environment and type 2 diabetes and IGT in adulthood. Specifically, studies indicate that either prenatal undernutrition or overnutrition may impair glucose tolerance in adulthood. This apparent discordance may be consistent with distinct imprinting mechanisms operating at both extremes of prenatal nutritional sufficiency. It is interesting that most studies found this relation to be much stronger in women than in men. Because alternative hypotheses about the mechanism underlying this association have not been examined adequately, it is unclear whether the insulin axis is imprinted in utero. No epidemiologic evidence suggests postnatal imprinting of type 2 diabetes or IGT.
To summarize, there is substantial epidemiologic evidence that birth weight is linked to adult BMI, CVD, blood pressure, and glucose tolerance. However, inherent weaknesses of long-term retrospective studies and insufficient consideration of alternative hypotheses undermine inferences that attribute these associations to intrauterine metabolic imprinting. In all of the outcomes, the evidence for prenatal imprinting is stronger than that for postnatal imprinting. This, of course, does not refute the occurrence of postnatal effects, but may simply reflect the lack of a simple and universally recorded index of postnatal nutrition analogous to birth weight. Many of the weaknesses of epidemiologic data can be addressed by the controlled conditions of animal models. Accordingly, the numerous animal models used to examine metabolic imprinting phenomena merit consideration. The following discussion focuses on investigations that model the nutritional exposures and outcomes most relevant to the epidemiologic observations that have been discussed.
|
ANIMAL MODELS OF METABOLIC IMPRINTING |
|---|
|
TOPABSTRACTINTRODUCTIONEPIDEMIOLOGIC EVIDENCE OF...ANIMAL MODELS OF METABOLIC...POTENTIAL MECHANISMSARE SPECIFIC MECHANISMS...SUGGESTIONS TO GUIDE FUTURE...CONCLUSIONREFERENCES |
|---|
The effect of the timing of prenatal energy restriction on adult body composition was probed in greater depth as several workers sought to model the relations between perinatal restriction and adult BMI suggested by the Dutch famine data (20). Indeed, several groups found that when rats were energy restricted during the first 2 trimesters and then allowed free access to food, their offspring became 10–20% heavier and fatter than the progeny of nonrestricted controls (60, 61). Curiously, most studies showed this effect only in male offspring. The increased adiposity of those animals appeared to be attributable to adipocyte hypertrophy rather than to hyperplasia (62). Jones et al (63) postulated that the increased adiposity seen in the adult offspring of prenatally restricted rats could be because of the compensatory hyperphagia that follows a period of restriction rather than because of the restriction per se. Accordingly, the offspring of 2 groups of dams that were energy restricted during the first 2 trimesters were studied (63). In one group (RA), mothers were allowed free access to a stock diet from the third trimester through lactation, whereas the others (RP) were pair fed (with a never-restricted control group) from the third trimester onward to prevent compensatory hyperphagia. At 200 d of age, the RA male progeny were heavier and fatter than the male progeny of the nonrestricted controls. However, the mean body weight and composition of the RP progeny were similar to those of the controls. These findings suggest that maternal compensatory hyperphagia was the cause of the offsprings' later elevated adiposity. Unfortunately, it is unclear whether the effect of the dams' compensatory hyperphagia was mediated gestationally or during the suckling period. As in the other studies, the increased adiposity in the RA males correlated with increased adipocyte size, not number.
Litter manipulation in rodents has also been used extensively to study the long-term effects of postnatal under- and overnutrition. In this model, pups are redistributed shortly after birth to either small or large litters, resulting in overnutrition or undernutrition during the suckling period, respectively. This model provided some of the earliest evidence for a critical window in the imprinting of body size and body composition. Widdowson and McCance (64) showed that rats undernourished during the suckling period followed a permanently diminished growth trajectory, whereas those undernourished to a similar extent after weaning quickly compensated for the period of slowed growth and resumed the growth trajectory of their normally nourished littermates. Subsequently, several studies showed that relative to rodents raised in normal-sized litters, those suckled in small litters were heavier and fatter as adults and those suckled in large litters were smaller and leaner throughout life (65, 66). The persistent effects on weight and adiposity ranged from 10% to 40% higher or lower than control values. Studies focusing on changes in adult rodent adipocyte morphology associated with diverse litter sizes concluded that the effect on adiposity was mediated via changes in adipocyte number rather than size (67–69). This finding is opposite that found in offspring of rodents restricted during gestation.
In another model examining the long-term effects of postnatal nutrition on later adiposity and adipocyte morphology, Lewis et al (70) artificially fed baboons formulas of different energy densities to effect normal nutrition, 31% overnutrition, or 40% undernutrition throughout the suckling period. At 5 y of age, female baboons that had been overnourished in the preweaning period were 40% heavier than their normally fed counterparts. The increased weight was correlated with elevated adiposity; the mean weight of 10 different fat depots in the 5-y-old postnatally overnourished female baboons was 5 times that of the normally nourished females. Although no body weight effect was seen in the males at 5 y of age, the mean weight of the 10 adipose depots in the postnatally overnourished males was 3 times that of the control males. In both males and females, the increased adiposity was primarily attributable to an increase in adipocyte size, not number, in contrast with the association found in rats suckled in small litters.
Cumulatively, these animal studies suggest that nutritional status during the early postnatal period has profound and persistent effects on body weight and adiposity. On the other hand, fetuses seem to be resistant to metabolic imprinting of weight and adiposity resulting from intrauterine undernutrition. It would be interesting to test in an appropriate animal model the putative long-term effects of overnutrition that are limited to the prenatal period, as is suggested by the epidemiologic relations between birth weight and later adiposity.
Cardiovascular disease
Although many of the established risk factors for CVD, such as obesity, hypertension, and insulin resistance, have been considered in this review as separate metabolic imprinting outcomes, abnormal lipid profiles have not. Hypercholesterolemia and elevated ratios of LDL to HDL are well-accepted major causal risk factors for CVD (13). Hence, to avoid overlap with other areas, this section will focus on studies suggesting metabolic imprinting of blood lipid profiles or lipid metabolism.
Most of the epidemiologic evidence for metabolic imprinting of CVD is based on birth weight, implying the importance of prenatal nutrition. Unfortunately, the vast majority of the animal model studies related to metabolic imprinting of CVD are based on postnatal nutritional treatments or on combined prenatal and postnatal treatments. In one study that examined the long-term effects of prenatal nutrition on adult plasma lipids, Lucas et al (71) provided rats a protein-restricted diet during pregnancy, lactation, or both and cross-fostered the offspring to independently assess the effects of exposure to the deficient diet during these 2 periods. Plasma total cholesterol, HDL-cholesterol, and triacylglycerol concentrations were all 20–50% lower in the 6-mo-old offspring of animals exposed to the low-protein diet during both gestation and suckling and in offspring of animals exposed to the low-protein diet during suckling only than in controls. Only the decrease in plasma triacylglycerol concentrations was significant in the group exposed to a low-protein diet during gestation only. These results, of course, are contrary to expectations based on the inverse relation between birth weight and adult CVD in humans.
In 1972, Reiser and Sidelman (72) published a highly controversial study of the hypothesis that ingestion of diverse concentrations of cholesterol in mother's milk imprints cholesterol metabolism in the progeny. In their small rat study, the dams' diets were varied to affect milk cholesterol concentrations. Serum cholesterol concentrations in male offspring fed a high-fat, high-cholesterol postweaning diet (cholesterol-challenge diet) were related inversely (through 32 wk of age) to the cholesterol concentration in their mothers' milk. At 32 wk of age, the serum cholesterol concentration of male offspring who had consumed the highest concentration of milk cholesterol during the suckling period was only 60% of that of control animals whose mothers had been fed a stock diet during lactation. In a later study that attempted to replicate this result, Kris-Etherton et al (73) obtained elevated milk cholesterol concentrations by feeding a high-fat, high-cholesterol diet to dams from day 18 of pregnancy though lactation. At 26 wk of age, the male progeny of these dams had 25% higher serum cholesterol concentrations than control animals. Note that these studies are not necessarily contradictory, because the former varied the dams' diet only postnatally, whereas the latter applied the dietary treatment starting from late gestation.
Results of subsequent studies of the persistent effects of perinatal maternal fat and cholesterol intakes were generally consistent with the findings of Kris-Etherton et al. Rats whose mothers were fed a high-fat diet from late gestation through lactation showed greater elevations in serum cholesterol than did controls in response to a cholesterol-challenge diet in adulthood (74, 75). Strictly speaking, however, the study of Reiser and Sidelman, in which the dams' diets were varied during lactation only, has apparently not been repeated.
Hahn (76) attempted to identify metabolic alterations that underlie metabolic imprinting of cholesterol homeostasis. He studied long-term effects of litter size manipulations in rodents on hepatic and adipose tissue enzyme activities. At 60 and 300 d of age, male rodents who were suckled in small litters (SL; 4/litter) had elevated plasma cholesterol concentrations and plasma insulin concentrations that were nearly twice as high as animals suckled in large litters of 18 (LL). Also, the activity of hepatic ß-hydroxy-ß-methylglutaryl-CoA (HMG-CoA) reductase (the rate-limiting enzyme in cholesterol biosynthesis) was 50% higher in LL than in SL animals at 60 but not at 240 d of age. The elevated hepatic HMG-CoA reductase activity of the LL animals at 60 d of age is intriguing given their decreased plasma insulin relative to the SL animals. A decrement in plasma insulin would be expected to down-regulate HMG-CoA reductase expression (77).
Rodents are generally considered a poor model for human CVD because of their relatively low serum LDL concentrations and their resistance to diet-induced atherosclerosis (78). Therefore, one might ask whether rodent data suggesting metabolic imprinting of cholesterol metabolism are relevant to metabolic imprinting of CVD in humans. Transgenic mouse models of atherogenesis have shown that although normal rodent lipoprotein metabolism differs quantitatively from that of humans, the biochemistry of atherogenesis is qualitatively similar (78). Hence, the rodent studies discussed here are likely to be relevant to metabolic imprinting of CVD in humans.
Mott et al (79) examined the long-term effects of preweaning cholesterol intake on cholesterol homeostasis in baboons. In this model, infant baboons were either breast-fed by their mothers or fed one of several formulas that varied in cholesterol concentration. After weaning, animals were assigned to 1 of 4 diets that varied in cholesterol content and ratios of polyunsaturated to saturated fat. In adulthood, few differences were found in serum cholesterol concentrations or cholesterol metabolism between the groups fed different concentrations of cholesterol during infancy. Mean serum triacylglycerol and HDL-cholesterol concentrations in the adult baboons were 10–15% lower in the mother-fed group than in all of the formula-fed groups (P = 0.05). Also, animals fed by their mothers had a 15% higher ratio of VLDL + LDL to HDL (P = 0.08), indicating a more atherogenic lipid profile. Accordingly, on necropsy at 8 y of age, the breast-fed baboons tended to have more extensive atherosclerotic lesions than did their formula-fed counterparts, but this trend was not statistically significant.
Cumulatively, the results of those studies indicated that adult cholesterol metabolism may be influenced by prenatal and early postnatal nutritional manipulations. Whereas the strongest epidemiologic data suggest prenatal imprinting of CVD susceptibility, most animal models have investigated the persistent effects of postnatal treatments. One study that did examine the long-term effects of prenatal restriction indicated an apparently protective effect against CVD risk factors (71). This was unexpected because of the inverse association found in humans between birth weight and adult CVD morbidity and mortality. Mott et al's (79) results in artificially fed baboons suggest that heretofore unidentified factors in mother's milk, rather than cholesterol, may imprint cholesterol metabolism during the suckling period. Such unknown factors may explain the inconsistent results found in different investigations of milk cholesterol intake during the suckling period and adult serum cholesterol homeostasis in rodents.
Blood pressure
Relatively few animal studies have examined the association between perinatal nutritional status and adult blood pressure. One study that explored this relation used unilateral uterine artery ligation in guinea pigs to create intrauterine growth retardation in half of the littermates (80). The intrauterine growth-retarded group included animals whose birth weight was reduced by 20% relative to their normal littermates. When the animals were 3–4 mo old, heart rate, but not blood pressure, was related inversely to birth weight. An analysis of 12 pairs of littermates showed a significant correlation between intrapair birth weight differentials and intrapair adult blood pressure differentials (r2 = 0.34, P < 0.05).
Langley and Jackson (81) reported a long-term effect of maternal gestational protein intake on the blood pressure of female progeny in the rat model. From before conception to delivery, dams were fed diets containing 6%, 9%, 12%, or 18% protein by weight. After delivery, all dams were provided a stock diet. At 9 wk of age, SBP in the female offspring was correlated inversely with the dams' protein intake during pregnancy (r2 = 0.27, P < 0.005). When combining blood pressure data from all ages (9, 12, 15, and 21 wk), SBP tended to increase with decreasing percentage of maternal protein intake. In a later study that used a similar design, Lucas et al (71) found that gestational protein restriction in rats (8% compared with 20% protein by wt) had the opposite effect on blood pressure in both male and female offspring at 6 mo of age. Recently, Woodall et al (82) studied the effect of maternal energy restriction (to 30% of ad libitum intake) during pregnancy on SBP of offspring. SBP was measured in the progeny at 30, 40, 48, and 56 wk of age, and was 5% higher (P < 0.05) in the restricted animals at every age except for 40 wk.
Unfortunately, none of those studies in rats reported correlations between birth weight and blood pressure at older ages to compare with relevant epidemiologic studies. A major common weakness of the dietary restriction studies is that cross-fostering was not used to distinguish the long-term effects of prenatal and residual postnatal exposure to maternal dietary restriction. Although restricted dams were allowed ad libitum access to control diets after delivery, milk quantity and quality may be reduced after prolonged and severe dietary restriction (83). Hence, without cross-fostering restricted progeny to unrestricted dams, it is unclear whether some of the long-term effects on blood pressure are mediated by diet during the early suckling period. But overall, with the exception of the study by Lucas et al (71), these studies are consistent in their support of the general hypothesis that prenatal undernutrition causes a permanent elevation in blood pressure.
Glucose tolerance
Numerous animal studies have explored the relation between perinatal nutrition and adult indicators of glucose tolerance. Hoet's group (84, 85) studied the effects of maternal low-protein diets on the glucose metabolism of adult progeny. Pregnant rats were fed diets containing 20% or 8% protein. After birth, the progeny of dams fed the low-protein diet were fostered by control dams to limit the period of restriction to gestation only. Growth of these recuperated animals was similar to that of the control group through 80 d of age. Nonetheless, a glucose tolerance test at 70 d of age revealed IGT correlated with blunted insulin secretion in the female, but not the male, recuperated animals. More recent studies with this same model showed that glucose tolerance is elevated in young recuperated rats but lower in older recuperated rats relative to controls (86). Accordingly, enhanced glucose tolerance and a 100% increase in adipocyte insulin receptor number were found in male recuperated rats at 6 wk of age (87). In vitro studies also showed increased glucose uptake in skeletal muscle in 3-mo-old male recuperated animals, which correlated with elevated concentrations of glucose transporter 4 in the plasma membrane fraction of the tissue and elevated insulin binding to glucose transporter 4 (88).
Long-term effects of prenatal exposure to hyperglycemia have been modeled by streptozocin-induced gestational diabetes in rats. Streptozocin was administered to pregnant rats on the first day of gestation (89), causing dramatic hyperglycemia throughout gestation. The female offspring of these animals had IGT at 80 d of age. This persistent effect of gestational hyperglycemia was later confirmed in both male and female offspring (90). Because birth weight is elevated dramatically in progeny of streptozocin-treated dams, these results appear consistent with the epidemiologic data suggesting a U-shaped relation between birth weight and adult IGT.
Several groups have investigated the long-term effects on glucose metabolism associated with various postnatal nutritional states. For example, rodents suckled in litters of different sizes have persistent alterations in fasting serum insulin, but not in glucose concentrations (69, 76). Fasting serum insulin concentrations in adulthood had a dose-response relation to nutritional adequacy during the suckling period: they were 2-fold higher in animals suckled in small litters and reduced by half in animals suckled in large litters relative to those in normal litters (65). Also, glucose tolerance tests conducted at 18 wk of age showed a hyperinsulinemic response in SL animals and a blunted insulin response in LL animals relative to controls suckled in normal-size litters (65). Patel's group (91, 92) used gastrostomy feeding to study persistent effects of feeding a high-carbohydrate diet throughout the suckling period. Compared with animals artificially fed a high-fat diet with a composition similar to that of rat milk, the high-carbohydrate diet group had 80% higher plasma insulin concentrations at 54 and 100 d of age. They also had IGT at 78 d of age (91) and altered insulin release from isolated pancreatic islets at 250 d of age (92). In another model, rats fed a low-protein diet during the postweaning period (from 3 to 6 wk of age) showed a normoglycemic hypoinsulinemic response to a glucose tolerance test at 12 wk of age (93). This suggests increased tissue sensitivity to insulin resulting from early postweaning protein deprivation. Thus, the critical window for the imprinting of insulin sensitivity apparently extends beyond the suckling period in some mammals.
Despite the abundance of data from animal models in this category, the implications of these studies in assessments of epidemiologic associations between prenatal nutrition and later glucose tolerance are unclear. Protein undernutrition of pregnant rats results in altered glucose tolerance in the progeny. The decreased glucose tolerance of older recuperated animals is generally consistent with the association between low birth weight and IGT in adult humans. Unfortunately, these studies have not reported differences in birth weight between controls and gestationally restricted animals. However, other studies in rats subjected to a similar degree of gestational protein restriction reported a significant decrease in progeny birth weight (94). It would be interesting to determine whether gestational energy restriction alone (which clearly results in reduced birth weight) has long-term effects on the glucose metabolism of progeny that are similar to those resulting from protein restriction of pregnant dams. In any event, it is clear from these various animal models that perinatal nutritional status can profoundly affect glucose tolerance and insulin sensitivity in adulthood.
Overall, the data from animal models of metabolic imprinting support the observed epidemiologic associations. The specific relations between perinatal treatments and adult outcomes in these models are not always concordant with expectations based on the epidemiologic literature. However, because most of the animal literature is based on rodent models, differences in perinatal developmental stages between rodents and humans make such comparisons difficult. Nonetheless, the general premise that biological mechanisms exist that result in the perinatal nutritional state imprinting metabolism is upheld. It is appropriate, therefore, to consider the nature of potential mechanisms that account for these putative effects of early diet.
|
POTENTIAL MECHANISMS |
|---|
|
TOPABSTRACTINTRODUCTIONEPIDEMIOLOGIC EVIDENCE OF...ANIMAL MODELS OF METABOLIC...POTENTIAL MECHANISMSARE SPECIFIC MECHANISMS...SUGGESTIONS TO GUIDE FUTURE...CONCLUSIONREFERENCES |
|---|
Generating a usefully concise list of potential mechanisms of metabolic imprinting requires establishing inclusion criteria for candidate mechanisms. The first 2 criteria derive from the definition of metabolic imprinting. First, it is an adaptive process. Accordingly, only mechanisms consistent with adaptive responses to early nutritional environments are considered. (In the context of this review, note that adaptive responses to nutritional conditions during development may in fact prove maladaptive if the organism's nutritional environment is later modified or if its homeostatic conditions are sufficiently changed at later life stages.) Second, imprinting occurs during a limited period of susceptibility. Hence, only mechanisms operating within a limited ontogenic period were included. For these reasons, responses to clearly pathologic insults are not considered. For example, perinatal exposure to toxins that permanently alter metabolism by compromising liver function would not be considered to result in metabolic imprinting. The last criterion stipulates that a potential mechanism must explain a specific putative imprinting process, and not simply be a reflection of the outcome. For example, whereas some have proposed altered cell size as a mechanism of perpetuating the effect of early nutrition, cell size is in most cases likely to be indicative of, rather than a determinant of, cellular metabolism. Nonetheless, this last criterion is the most difficult to apply because almost any biological mechanism may, on close study, be broken down into a subset of component mechanisms. Hence, candidate mechanisms are most likely limited in their scope by incomplete knowledge of upstream component mechanisms.
Others have proposed mechanisms that may be consistent with these criteria. For example, on the basis of studies showing that perinatal undernutrition in rats permanently reduces pancreatic islet vascularization, Hoet's group (95) suggested that metabolic effects of early nutrition may result not only from alterations at the cellular level but also from changes in organ structure. Considering the central role the endocrine system plays in regulating metabolism, Csaba's (96) theory of "hormonal imprinting" may be relevant to metabolic imprinting. His studies suggested that the concentrations of hormones and hormone analogues present at critical developmental periods permanently affect the hormonal responses to specific stimuli, tissue sensitivity to specific hormones, or both. Dorner (97) put forth the concept that abnormal nutritional environments during the period of brain differentiation could lead to persistent metabolic disturbances. The idea that early nutrition may, for example, "set" the hypothalamus and thereby regulate appetite, growth, and behavior has since been considered by several authors.
Several years ago, Lucas (8) put forth the most complete list of potential mechanisms by which perinatal nutrition might persistently affect an organism's structure or function. He proposed that such nutritional programming could occur via 4 general mechanisms: 1) undernutrition at a critical stage may fail to fuel a growth process, 2) the early nutrient environment may permanently alter gene expression, 3) the early nutrient environment may affect clonal selection, or 4) the early nutrient environment may affect cell number.
Lucas's list of mechanisms and the suggestions of others discussed above are a useful starting point, but are often not sufficiently specific to guide experimental characterization. For example, "failing to fuel a growth process" is conceptually too broad, and "permanently altering gene expression," though certainly central to metabolic imprinting, underlies nearly all candidate mechanisms. Admittedly, formulating a useful and concise list of specific potential mechanisms of metabolic imprinting is challenging because of the interrelatedness and nonexclusive characteristics of the likely candidates. On the basis of the criteria outlined above, we propose the following mechanisms as among the most useful to consider in future research: 1) induced variations in organ structure, 2) alterations in cell number, 3) clonal selection, 4) metabolic differentiation, and 5) hepatocyte polyploidization.
The physiologic and molecular bases of these 5 potential mechanisms are reviewed first. Then these mechanisms are considered in the context of examples of putative metabolic imprinting phenomena from the animal and human literature. Last, some experimental approaches are proposed to determine which of these mechanisms may play dominant roles in mediating metabolic imprinting in specific animal models.
Organ structure
This mechanism is restricted to gross morphologic alterations occurring during organogenesis to distinguish it from other candidate mechanisms that may result in altered organ structure via changes at the cellular or subcellular level. Morphologic outcomes consistent with this potential mechanism are, for example, alterations in organ vascularization or innervation and changes in the juxtaposition of different cell types within an organ. As such, altered organ structure is the only mechanism of metabolic imprinting considered in this review that operates above the cellular level.
How might altered organ structure occurring during organogenesis permanently affect the organism's metabolism, and thus lead to imprinting? The most obvious examples pertain to the ability of individual cells to generate and respond to external signals within the organism. If patterns of organ vascularization are permanently affected by nutrition during organogenesis, this could affect the cells' responses to blood-borne nutrients or hormonal signals. Or, within the liver, for example, the specific juxtaposition of hepatocytes, endothelial cells, and Kupfer cells determined during organogenesis could have a persistent influence on hepatic metabolism.
During gastrulation, the 3 embryonic germ layers, the endoderm, mesoderm, and ectoderm, are formed. The subsequent process of organogenesis involves inductive interactions among those different layers to direct changes in cell position, shape, and differentiation that result in the formation of the organ rudiments. For example, the mammalian liver forms from a mutually inductive interaction between endoderm and mesoderm (98). Organogenesis begins very early in human embryonic development. By 5 wk of gestation, a human embryo develops rudiments of most organs, and by 8 wk organogenesis is nearly completed (99). However, the endpoint of organogenesis is difficult to define. Cellular differentiation continues into postnatal development and transforms the fetal organ rudiments into functioning organs.
During the limited period of organogenesis, the fates of cells depend on externally derived inductive signals from adjacent cells and from morphogen gradients that originate at more distant sites. Hence, it is reasonable to expect that the local concentrations of diverse nutrients, their metabolites, or both modulate the end result of this process. For example, many cell-cell interactions critical to inductive interactions and cell migration are mediated through proteoglycans at the cell surface. Imbalances in macronutrients from which these components are derived could affect these processes, resulting in permanent alterations in organ structure. As another example, retinoic acid, a derivative of vitamin A, serves an important regulatory role during normal organogenesis. It binds to ligand-activated transcription factors that then regulate the expression of genes involved in morphologic development (100). Thus, local concentrations of this nutrient-derived transducer assist in the establishment of organ structure.
Cell number
Of all the potential metabolic imprinting mechanisms, a permanent alteration in cell number is the simplest and easiest to put forward. During development, organs can increase in size either by increasing the number of cells (hyperplasia) or by increasing cell size (hypertrophy). Different tissues go through diverse, limited periods of hyperplastic and hypertrophic growth. The rate of cellular proliferation is directly dependent on nutrient availability (the supply of building blocks) and may be indirectly dependent on the organism's general nutritional status via hormonal signals that control cellular proliferation. Nutrient limitations or excesses during critical periods of hyperplastic growth that affect rates of cell division may lead to permanent changes in cell number, regardless of later nutrient availability.
This phenomenon has been examined most exhaustively for the central nervous system. For example, in a rat model, Winick and Noble (101) showed that severe malnutrition during the period of brain cell hyperplasia resulted in permanent deficits in brain cell number, whereas malnutrition during later periods of brain cell hypertrophy did not have permanent effects. Conversely, if nutrient delivery or the hormonal milieu is conducive to rapid cellular proliferation, the organ may have a permanently increased number of cells. Because an organ's metabolic activity is limited by cell number, a permanent effect on cell number could permanently affect metabolism.
Multinucleation of muscle tissue may be viewed as a special category of hyperplastic growth and hence warrants comment in this section. During myogenesis, myoblasts proliferate, align, leftand fuse to form myotubes. These multinucleated syncytia then develop into muscle fibers. Once a myoblast fuses with a myotube it no longer proliferates (102). Hence, the rate of myoblast hyperplasia during the limited period of myogenesis determines the degree of nucleation within muscle fibers. Although the total number of myocytes may not be affected, the metabolic activity of the myocytes will be limited by their degree of nucleation, analogous to an effect on cell number.
Clonal selection
Cellular proliferation in all organs involves the initial multiplication of a finite population of founder cells. Founder cells in specific organs are not necessarily identical to each other. Although cellular proliferation generally precedes terminal differentiation, successive generations of cells undergo limited differentiation in the early proliferative stage. As cellular proliferation proceeds, genetic and epigenetic modifications likely occur within individual cells that distinguish them from others in subpopulations of rapidly dividing cells. Distinguishing characteristics may offer selective advantages to specific clones competing for the available nutrient supply. Thus, for example, an incorrect base pairing during DNA replication may result in subtle effects on a cell's metabolism. This change would then be passed to all its progeny.
Such subtle differences among proliferating cells are the basis for clonal selection. Clonal selection is comparable with Darwinian evolution operating at the cellular level, but rather than through survival of the fittest, clonal selection works by disproportionate population growth of the most rapidly proliferating cells. Hence, 2 similar, heterogeneous populations of rapidly dividing cells may develop very distinct metabolic characteristics as a result of diverse microenvironmental conditions. If, for example, the nutrient environment is deficient in structural fatty acids, cells with a slightly more efficient or more active lipogenic pathway could disproportionately populate a tissue. Thus, it is possible that variations in nutritional status during development could lead to permanent alterations in organ or tissue metabolism via this mechanism.
Metabolic differentiation
The most complex and probably most developmentally relevant potential mechanism of metabolic imprinting is through an effect on metabolic differentiation. Greengard (103) defined enzymic differentiation as "the process whereby, in the course of prenatal or early postnatal development, the different organs of an animal acquire their characteristic, quantitative pattern of enzymes." Metabolic differentiation represents the process whereby individual cells develop a stable quantitative pattern of basal and inducible gene expression. As such, metabolic differentiation pertains not only to enzymes, but also to transcription factors, hormones, hormone receptors, transmembrane transporters, and other elements. Also, metabolic differentiation stresses the importance of the ontogeny of a tissue's capacity for adaptive as well as basal gene expression.
Metabolic differentiation should be distinguished from cellular differentiation. Cellular differentiation is the developmental process by which multicellular organisms develop a limited number of distinct and stable cell types (99). For the most part, developmental biologists have focused on the problem of cellular differentiation from a morphologic perspective, and it is from this perspective that the differentiated state is recognized most often as distinct and stable. Thus, although a human adult may have trillions of cells, there are only 200 different cell types, and a differentiated cell is generally committed to its specific cell type. Differentiated cells may be characterized as being one type or another based on structural features, or in some instances by qualitative protein markers. A fundamental feature of metabolic differentiation is that quantitative as well as qualitative differences in gene expression may distinguish one cell from another. For subtle metabolic differences among cells of a similar type to be important, a cell's quantitative pattern of gene expression, once established during development, must be stable. In other words, cellular memory, or the faithful transmission of determined states to progeny cells, must be maintained quantitatively (104).
Generally speaking, cellular differentiation is characterized by the stable ability to express a limited number of genes, constitutively or in response to given stimuli. This stability is maintained through epigenetic mechanisms. Epigenetics is defined as the study of "mitotically and/or meiotically heritable changes in gene function that cannot be explained by changes in DNA sequence" (105). Epigenetic phenomena play important roles in mammalian development. Several reviews (106–109) discuss potential molecular mechanisms that underlie epigenetic phenomena such as genomic imprinting (whereby the maternal or paternal allele of specific genes are stably repressed) and X chromosome inactivation (which mediates X chromosome dosage compensation in mammals). It is likely that the same epigenetic mechanisms that mediate these well-studied phenomena also contribute to metabolic differentiation. Three epigenetic mechanisms of metabolic differentiation that might mediate metabolic imprinting are discussed below: 1) the autoregulatory pattern of DNA binding proteins, 2) the persistent modulation of chromatin structure, and 3) DNA methylation. Note that these epigenetic mechanisms are not independent of one another, but likely function together to maintain the vast array of differentiated tissues that characterize higher organisms (Figure 3).
|
This simple positive feedback mechanism is a common strategy of multicellular organisms. For example, many of the gene regulatory proteins that function combinatorially to establish body patterning during Drosophila development stimulate their own transcription (110). Also, the products of the genes regulating vertebrate muscle cell differentiation, such as MyoD, activate their own transcription as well as that of numerous downstream genes (111). Thus, through positive-feedback autoregulation and combinatorial control, a relatively small number of gene regulatory proteins can perpetuate a vast number of differentiated cell types. To the extent that early nutrition influences the developmental cascade that establishes cell-specific patterns of gene regulatory proteins, imprinting may occur by this mechanism.
Chromatin structure
DNA in the eukaryotic nucleus is packaged in a highly compact configuration with histone proteins and other DNA binding proteins. This protein-DNA complex is called chromatin. Two general types of chromatin are identified: 1) euchromatin, an open configuration allowing rapid gene transcription; and 2) heterochromatin, the closed and inactive configuration. The fundamental unit of chromatin structure is the nucleosome, which consists of about 200 base pairs of DNA wrapped around a histone octamer. Neighboring nucleosomes assemble into higher-order structural configurations corresponding to more condensed and presumably transcriptionally inactive DNA.
Chromatin structure is highly correlated with gene expression during development (104). In differentiating chicken erythrocytes, both the 
- and ß-globin genes are maintained in an open configuration, whereas in all other cell types these loci are sequestered within regions of inactive chromatin (112, 113). Analogously, in differentiating oligodendrocytes, up-regulation of myelin-associated glycoprotein, a component of the myelin sheath, appears to involve chromatin remodeling (114). X chromosome inactivation in female mammals also suggests an important function of chromatin in inhibiting gene expression. Within each cell, the active X chromosome is predominantly euchromatic, whereas the inactive X chromosome is almost completely heterochromatic (115). Also, tissue-specific regulation of several genes has been correlated with alterations in local chromatin structure. For example, in chickens, the liver-specific nutritional regulation of malic enzyme correlates with DNAase-hypersensitive regions (indicating open chromatin configuration) upstream of the transcription start site that are not found in other tissues (116).
For chromatin structure to play a role in cell memory, and hence metabolic differentiation, the specific chromatin structure within a given region of DNA must be stably propagated from progenitor cells to progeny. It is not obvious how this might occur. The histones and other DNA binding proteins associated with the nucleosome must detach from a region of DNA before the DNA polymerase can replicate it. Several models have been proposed to explain how specific regions of DNA are maintained in a stable, specific chromatin configuration (117, 118).
One such model is based on histone modification. Histone proteins may undergo several posttranslational modifications, such as acetylation, ubiquitination, and phosphorylation. Histone hyperacetylation is associated with regions of open chromatin configuration (119). Hence, histone acetylation could serve as an epigenetic marker to identify regions of DNA that are to be maintained in a transcriptionally active configuration. Although histones detach from DNA as the
