Intergenerational Transmission of Glucose Intolerance and Obesity by In Utero Undernutrition in Mice

Protocols were approved by the Joslin Diabetes Center Institutional Animal Care and Use Committee. Virgin ICR females (6–8 weeks old, parental generation F₀) were caged with ICR males (Fig. 1A). Pregnancy was dated by vaginal plug (day 0.5). Pregnant F₀ females were housed individually with ad libitum access to Purina 9F chow. On pregnancy day 12.5, F₀ females were assigned to either control (C) or undernutrition (UN) groups randomly. Food intake of F₀-UN dams was restricted from day 12.5 until delivery by 50% compared with that consumed by F₀ controls (25). Litters were equalized to eight pups per dam by removing both the heaviest and lightest mice in the litter, thus retaining those with birth weight closest to the median for each litter. Offspring of F₀ mice were designated as first-generation offspring (F₁) (Fig. 1A). After delivery, all F₀ mothers received chow ad libitum. F₁ pups nursed freely and were weaned at 3 weeks onto 9F chow, provided ad libitum.

C and UN females from the F₁ generation were mated at age 2 months with nonsibling F₁-C or F₁-UN males to generate four experimental groups (Fig. 1A). After confirmation of pregnancy, females were caged individually and fed ad libitum with no dietary manipulation during pregnancy. Pups were designated as second-generation offspring (F₂): C♀-C♂ (both parents are controls); C♀-UN♂ (female control, male UN); UN♀-C♂(female UN, male control); and UN♀-UN♂ (both parents UN). F₂ mice were weaned at 3 weeks onto 9F chow ad libitum; only males were followed for metabolic analysis given the higher prevalence of diabetes in UN-F₁ males (100 vs. 30% in females, as described [25]).

Intraperitoneal glucose (2 g/kg weight) and insulin tolerance (1 unit/kg) tests were performed in unrestrained conscious mice after a 16- and 4-h fast, respectively. Intraperitoneal tolbutamide tolerance (1 mg/kg) was assessed after a 4-h fast.

Islets were isolated from 4-month-old mice following intraductal collagenase infusion as previously described (25). For indicated experiments, diazoxide and tolbutamide (Sigma) were added during incubation (final concentration 250 and 200 μmol/l, respectively).

Total RNA was isolated using RNeasy (Qiagen) and cDNA was synthesized using oligo-dT primers (Advantage; Clontech) for real-time quatitative PCR with SybrGreen detection (ABI Prism 7700; Applied Bioscience). Full list of primer sequences is available in supplementary Table 1 (available at http://dx.doi.org/10.2337/db08-0490).

Methylation of the Sur1 promoter was analyzed by PCR (MethylScreen; Orion Genomics) (27) (supplementary methods), while methylation of differentially methylated regions at the Pref1 locus was assessed by Southern blotting (28).

Body composition was analyzed by dual-energy X-ray absorptiometry (Hologic, Waltham, MA) (29).

Blood samples from adult mice were collected from the tail vein, while trunk blood was collected from ED16.5 fetuses after decapitation. Blood glucose was measured with Glucometer Elite (Bayer, Elkhart, IN). Insulin, leptin (enzyme-linked immunosorbent assay; Crystal Chem), triglycerides, glycerol (GTO-Trinder Triglycerides assay; Sigma), and nonesterified fatty acids (Half-micro test; Roche) were measured on 2- to 5-μl serum samples.

Data are expressed as means ± SE. Statistical analysis was performed using a two-tailed t test or ANOVA (Statview). P ≤ 0.05 was considered significant.

As previously described, 50% global caloric restriction of F₀ dams during the last week of gestation blunted feto-maternal growth and resulted in 15% decrease in offspring birth weight (F₁-UN) (25). In pregnancies of F₁ mice, both maternal food intake and feto-maternal weight gain (supplementary Table 2) were similar among the four experimental groups. Despite this, birth weight was modestly, but significantly, reduced in C♀-UN♂ and UN♀-UN♂ offspring compared with C♀-C♂ (Fig. 1B); UN♀-C♂ mice had normal birth weight compared with controls. Litter size and length of gestation were similar in all groups (supplementary Table 2). After delivery, all four F₂ groups had similar weights to 40 weeks (Fig. 1C).

Since feto-maternal nutritional status during pregnancy might account for observed differences in birth weight, we determined both maternal and fetal serum concentrations of major macronutrients contributing to fetal growth. In F₀-UN females, caloric restriction resulted in 34% reduction in serum glucose and 80% reduction in triglycerides by day 16.5 of gestation (P < 0.05, Table 1). Likewise, serum insulin and leptin were reduced in F₀-UN dams by 77 and 94%, respectively (P < 0.05) (Table 1). Maternal dysregulation of fuel metabolism was also accompanied by alterations in fetal metabolism. Fetal glucose and leptin levels were decreased by 45 and 63%, respectively, in ED16.5 F₁-UN fetuses (P < 0.05) (Table 2), while triglycerides, free fatty acids (FFAs), glycerol, and insulin did not differ from controls (not shown).

By contrast, during the F₁ generation pregnancy, maternal glucose, triglycerides, FFAs, insulin, and leptin did not differ between groups at pregnancy day 16.5 (Table 1). However, serum glycerol was increased by 49 and 59% in dams of C♀-UN♂ and UN♀-C♂ pregnancies, respectively (P < 0.05) compared with C♀-C♂ (Table 1). Despite no major differences in maternal metabolism, we observed some differences in ED16.5 F₂ fetuses. Glucose levels were increased by 66% in UN♀-UN♂ fetuses (P < 0.05), and insulin increased by 41% in C♀-UN♂ fetuses (P < 0.05) (Table 2). There were no significant differences in UN♀-C♂ fetuses.

Glucose levels in F₂ males did not differ in either fed (Fig. 2A) or fasting conditions (supplementary Fig. 1A) up to 10 months of age. Likewise, serum levels of triglycerides, glycerol, and FFAs were also similar among the four F₂ groups at 2 and 4 months (not shown). However, UN♀-UN♂, but not C♀-UN♂ or UN♀-C♂ mice, developed hyperinsulinemia in the fed state between 4 and 8 months of age (P < 0.05) (Fig. 2B). Fasting insulin was similar in all F₂ groups (supplementary Fig. 1B). To better define glucose homeostasis, we performed intraperitoneal glucose tolerance testing in all F₂ mice. Strikingly, glucose tolerance was impaired (relative to C♀-C♂) in C♀-UN♂, UN♀-C♂, and UN♀-UN♂ mice between 4 and 6 months of age (P = 0.03, P = 0.003, and P = 0.007, respectively) (Fig. 3A).

We next explored potential mechanisms responsible for glucose intolerance in F₂ offspring of UN parents. UN♀-UN♂ mice were clearly insulin resistant (assessed by insulin tolerance) by age 6 months (P < 0.001) (Fig. 3B). By contrast, C♀-UN♂ and UN♀-C♂ mice exhibited normal insulin sensitivity.

Since glucose intolerance may also reflect dysregulation of insulin secretion, we assessed glucose-stimulated insulin secretion in vivo and ex vivo. We first measured serum insulin with fasting and 30 min after a glucose load (glucose tolerance test) (Fig. 4A); glucose-stimulated insulin secretion was reduced in F₂ offspring from F₁-UN parents compared with C♀-C♂ mice at age 4 months (P < 0.05 for all comparisons) (Fig. 4A). Likewise, ex vivo glucose-stimulated insulin secretion was also impaired in islets from 4-month-old UN♀-C♂ and UN♀-UN♂ mice (P < 0.05 for both) but not C♀-UN♂ mice (Fig. 4B). Thus, reduced glucose-stimulated insulin secretion may also contribute to impaired glucose tolerance (IGT) in F₂ mice.

One potential candidate underlying insulin secretory dysfunction in F₂ mice is altered expression and/or function of the β-cell ATP-dependent K⁺ channel, composed of Sur1/Kir6.2 subunits. We assessed the functional capacity of this complex using in vivo challenge with the sulfonylurea tolbutamide (1 mg/kg i.p.). Tolbutamide induced a 15% increase in serum insulin after 30 min in C♀-C♂ mice (P < 0.05) (Fig. 4C) (supplementary Fig. 2A). In contrast, F₂ offspring from UN parents had impaired tolbutamide response (Fig. 4C) (supplementary Fig. 2A). Paralleling in vivo data, tolbutamide stimulated insulin release from freshly isolated C♀-C♂ islets by fivefold (Fig. 4D), while responses in all other groups were significantly reduced, with only minor increments in C♀-UN♂ and UN♀-UN♂ islets and no effect in UN♀-C♂ islets (Fig. 4D). We also incubated islets with diazoxide, a potassium channel stimulator. Diazoxide tended to reduce insulin release from C♀-C♂ islets (>50%, P = 0.1) (supplementary Fig. 2B). Interestingly, response in UN♀-C♂ islets was nearly normal, but diazoxide had only minimal effect in C♀-UN♂ and UN♀-UN♂ islets (5–10% decrease) (supplementary Fig. 2B). Together, these data suggest that dysfunctional potassium channel activity may contribute to impaired insulin secretion in F₂ mice.

To determine whether alterations in expression might contribute to dysfunction of potassium channels, we assessed Sur1 and Kir6.2 expression in isolated islets. Sur1 expression was reduced by 23% (P < 0.05) in F₁-UN islets (Fig. 4E). Likewise, Sur1 expression was reduced by 33% in C♀-UN♂ (P < 0.05), by 56% in UN♀-C♂(P < 0.05), and by 38% in UN♀-UN♂ islets (P < 0.05) (Fig. 4E). By contrast, Kir6.2 expression was normal in both F₁ and F₂ islets (Fig. 4F).

Since epigenetic regulation at the Sur1 locus might be mechanistically linked to differential expression in F₂ offspring of LBW mice, we assessed DNA methylation within CpG islands at the Sur1 promoter using methylation-sensitive PCR (27); we did not detect differential methylation in either F₁ (Supplementary Fig. 3) or F₂ islets (not shown).

We have previously shown that F₁-UN males develop increased adiposity as early as 2 months of age (26). In the F₂ generation, both maternal lineage offspring groups developed increased adiposity by 4 months (P < 0.05) (Fig. 5A). Interestingly, adiposity in UN♀-UN♂mice is largely visceral in location (Fig. 5A). In parallel, serum leptin is increased in UN♀-UN♂ mice (P < 0.05) (Fig. 5B). We observed no differences in epididymal adipocyte size distribution or histology (not shown) or in adipose expression of genes regulating differentiation and/or metabolic function, including peroxisome proliferator–activated receptor γ, CCAAT/enhancer binding protein α, fatty acid binding protein 4, and Glut4, from any F₂ group at 4 months of age (supplementary Table 3).

Since obesity in UN♀-UN♂ mice implicates a contribution from epigenetic factors, we determined expression of imprinted genes known to regulate adipocyte development and/or function and thus potentially contributing to obesity, including preadipocyte factor 1 (Pref1) (30), Necdin (31), and paternally expressed gene 1 (Peg1) (32). Pref1, an inhibitor of adipogenesis expressed from the paternally inherited chromosome, was reduced in epididymal fat from F₁-UN males (P < 0.05) and from both paternal F₂ lineages, C♀-UN♂ and UN♀-UN♂ (P < 0.01) (Figs. 5C and D). This effect appeared restricted to Pref1 since expression of Necdin and Peg1, also paternally expressed, was not statistically different from controls in either F₁ or F₂ offspring (Figs. 5C and D). In addition, expression of retrotransposon-like 1 (Rtl1) and maternally expressed 2 (Gtl2), two genes also regulated by the same imprinting signals as Pref1 (28), did not differ among groups in F₁ and F₂ offspring (not shown).

To determine whether changes in Pref1 gene expression were related to differential methylation at this imprinted locus, we assessed DNA methylation by Southern blotting. We did not detect any alterations in DNA methylation at the Pref1 differentially methylated region, the intergenic differentially methylated region, or the Gtl2 promoter region (supplementary Fig. 4A and B).

Human data indicate that metabolic phenotypes associated with exposure to maternal undernutrition can also be observed in second-generation offspring. Conceptually, transgenerational inheritance of disease risk may be mediated by nongenomic mechanisms, including either 1) epigenetic mechanisms (16) or 2) other broader indirect mechanisms associated with parental physiology (rev. in 6). First, alterations in nutrition during development can alter epigenetic marks, thus regulating gene expression through DNA methylation (12,17–19) and/or histone modifications (17,20,21). Interestingly, such epigenetic modifications may progress with aging during postnatal life, in association with metabolic phenotypes, as recently observed at the pancreatic duodenal homeobox-1 and GLUT4 locus in rodent models of intrauterine programming (33–35). If these epigenetic changes occur in the germ line, they can be inherited through meiosis (22–24), thus providing a plausible explanation for intergenerational effects, transmitted via either maternal or paternal lines.

In addition, other indirect biological processes may influence phenotypes in subsequent generations. For example, uterine size is reduced in girls that are born small (F₁) and remain short (36); this may influence fetal growth and reduce weight in their progeny (F₂) (rev. in 6). Therefore, physical constraints may alter birth size through the maternal lineage. Furthermore, maternal metabolism may also influence cross-generational phenotypes (37). Maternal undernutrition during pregnancy (F₀) increases risk for developing diabetes and obesity in her offspring (F₁). When these high-risk adult F₁ females become pregnant, the metabolic stress of pregnancy may result in hyperglycemia and/or overt gestational diabetes that may, in turn, contribute to fetal hyperinsulinemia, obesity, and increased diabetes risk in F₂ offspring (37). For these two examples, intergenerational transmission of phenotypes would occur exclusively through the maternal lineage, as opposed to epigenetic mechanisms.

Since it is difficult to dissect the relative parental contributions and the molecular mechanisms that lead to F₂ offspring outcomes in humans, we have utilized our mouse model of maternal undernutrition during pregnancy to address these important questions. We have previously shown that global caloric restriction during the last week of gestation in pregnant females (F₀) impairs fetal growth, resulting in 15–20% reduction in birth weight, and the development of both obesity and glucose intolerance in F₁ offspring (25,26). These alterations are multifactorial in origin, with glucose intolerance mediated, in part, by impaired glucose-stimulated insulin secretion (25). We now demonstrate that metabolic phenotypes are also observed in the F₂ generation: 1) reduced birth weight progresses from F₁ to F₂ through the paternal, but not the maternal, line (Fig. 1B); 2) F₂ offspring of both F₁-UN males and females develop IGT with aging, in parallel with β-cell dysfunction (Figs. 3A and 4A and B) and, in UN♀-UN♂, insulin resistance (Figs. 2B and 3B); and 3) obesity progresses through the maternal, but not the paternal, lineage (Fig. 5A).

Maternal undernutrition (F₀) during the last week of gestation results in reduced birth weight in both F₁ (25) and F₂ offspring of F₁-UN males (C♀-UN♂ and UN♀-UN♂) but not offspring of F₁-UN females (UN♀-C♂) (Fig. 1B). Such patrilineal inheritance of reduced birth weight may be attributed to alterations in fetoplacental gene expression and/or function, most likely mediated by epigenetic modifications. While we did not detect major metabolic dysregulation in C♀-UN♂ and UN♀-UN♂ mice at embryonic day 16.5 (Tables 1 and 2), it is likely that more subtle or time-dependent effects on placental function account for paternally mediated reductions in birth weight in this setting.

While experimental (9,11,38,39) and human (40,41) data show similar paternal transmission patterns of birth weight, population-based studies also indicate that intergenerational aggregation of LBW can also occur through maternal lineages (40,41). Such results may reflect species differences and complex influences on human birth weight (e.g., genetics, maternal size). Furthermore, human population data generally exclude complicated pregnancies (e.g., gestational diabetes) and thus may also exclude the impact of fetal “overgrowth” due to increased nutrient supply in the context of previous maternal LBW.

F₁-UN males exhibit moderate hyperglycemia and IGT with aging and dysregulated glucose-stimulated insulin secretion (25). Additionally, all F₂ offspring of F₁-UN parents develop glucose intolerance by age 4 months (Fig. 3A). Therefore, intergenerational progression of glucose intolerance can derive from both the maternal and paternal lines. While maternal and grandmaternal inheritance of diabetes has been demonstrated in rats (9,11,38,39), here we present the first experimental evidence for transgenerational transmission of IGT also through the paternal lineage. In agreement, paternal LBW in humans has recently been linked to risk of metabolic syndrome in both offspring and grandoffspring (7).

We previously demonstrated that impaired glucose-stimulated insulin secretion is an early key phenotype in F₁-UN males (25). Multiple metabolic adaptations lead to β-cell dysfunction in F₁-UN mice, including altered glucokinase, hexokinase-1, and ATP-dependent K⁺ channel activities (25) (J.C.J.-C., unpublished data). We now find that glucose-stimulated insulin secretion is also impaired in islets of F₂ mice from UN parents and is likely a major contributor to glucose intolerance (Fig. 4A and B). Additionally, we show that potassium channel activity is also altered in islets from F₂-UN offspring both in vivo and in vitro (Fig. 4C and D; supplementary Fig. 2A and B). Thus, it is interesting that expression of Sur1 is reduced by 30% in islets from both F₁-UN mice and islets from F₂ C♀-UN♂, UN♀-UN♂, and UN♀-UN♂ mice (Fig. 4E), indicating that dysregulation of Sur1 gene expression and function can be transmitted to F₂ through both maternal and paternal lines. This effect was specific to Sur1, since Kir6.2 gene expression was normal in both F₁ and F₂ islets (Fig. 4F). Together, these data suggest that dysregulation of Sur1 expression may alter potassium channel function (42) and ultimately contribute to whole-body glucose intolerance in our model. In agreement, genetic ablation of Sur1 results in metabolic phenotypes similar to those of our mouse model. Sur1^−/− mice remain euglycemic for the majority of their lifespan, displaying moderate glucose intolerance with aging (43,44). Isolated islets from Sur^−/− mice show impaired glucose-stimulated insulin secretion and lack of responsiveness to tolbutamide in vitro (43).

Which mechanism(s) is responsible for reduced Sur1 expression in both F₁ and F₂ islets? Since decreased Sur1 gene expression progresses to F₂ offspring through the paternal lineage, we hypothesized that reductions in Sur1 expression would be mediated, in part, by altered DNA methylation; however, we did not observe differences in methylation of the Sur1 proximal promoter region (supplementary Fig. 2). A potential limitation of our approach is that we utilized DNA from islets (a mixture of β-, α-, δ-, and PP cells), potentially masking β-cell–specific differences. Alternatively, reduced expression of Sur1 in islets from F₁ and F₂ mice might also be due to 1) altered methylation of another genomic region(s), 2) altered histone modification, or 3) changes in binding/expression of other transcription factors regulating Sur1. Future experiments will focus on potential mechanisms mediating decreased expression of Sur1.

An additional factor likely to contribute to impaired glucose tolerance in UN♀-UN♂ mice is insulin resistance, observed by age 6 months (Figs. 2B and 3B). Of note, insulin resistance arises when both parents were growth restricted in utero, suggesting that interface between these two lineages is necessary and sufficient to induce insulin resistance. While the specific mechanisms underlying insulin resistance in UN♀-UN♂ mice remain under investigation, it is interesting to note that UN♀-UN♂ mice also display increased visceral fat by age 4 months (Fig. 5A). Therefore, it is tempting to speculate that insulin resistance may develop secondary to progressive accumulation of visceral fat in UN♀-UN♂.

Childhood obesity is linked to parental obesity (45). Although it is difficult to dissect the relative contribution of shared environmental factors to these phenotypes in humans, maternal metabolic dysregulation during pregnancy may be a key contributor (46,47). We similarly demonstrate that adiposity phenotypes progress to F₂ offspring through the maternal lineage (Fig. 5A), implicating subtle maternal metabolic dysfunction during pregnancy and/or maternal epigenetic effects. It is important to note that despite increased adiposity in these groups, body weight remained similar to that of controls (Fig. 1C). Therefore, offspring of F₁-UN females do not have overt obesity but have a shift in body composition (higher fat mass with concomitantly reduced lean body mass) (26). On the other hand, although paternal transmission alone is not sufficient to drive adiposity in C♀-UN♂ mice, paternal effects may contribute to the increased abdominal fat accumulation in UN♀-UN♂ mice as compared with UN♀-C♂ mice.

We next evaluated potential mechanisms mediating increased adiposity in UN♀-C♂ and UN♀-UN♂ mice. First, food intake was similar in all groups (not shown). Second, expression of adipocyte differentiation markers, including peroxisome proliferator–activated receptor γ, CCAAT/enhancer binding protein α, fatty acid binding protein 4, or Glut4, was similar in all F₁ and F₂ groups (supplementary Table 3) (E.I., J.C.J.-C., M. Woo, A.C., J. DeCoste, M. Vokes, M. Liu, S. Kasif, A.-M. Zavacki, R. Leshan, M. Myers, M.E.P., unpublished data). Similarly, we observed no major differences in adipocyte size distribution among F₂ groups (not shown). Additional studies will be required to determine whether alterations in adipocyte metabolic function or in systemic energy homeostasis also contribute to adiposity.

Since obesity in UN♀-UN♂ mice implicates a partial role for epigenetic factors, it is very interesting that Pref1 expression was significantly decreased in F₁-UN, C♀-UN♂, and UN♀-UN♂ mice (Fig. 5C and D). Patrilineal inheritance of Pref1 expression may be explained by the fact that it is expressed from the paternally inherited chromosome; altered expression would be predicted to be inherited from F₁-UN males but not F₁-UN females. Mice lacking functional Pref1 have increased adiposity (48); conversely, mice overexpressing Pref1 have reduced fat content (49) (A.F.-S., unpublished data). Thus, decreased Pref1 expression may contribute, in part, to development of obesity in F₁-UN, UN♀-C♂, and UN♀-UN♂ mice. While our data support a potential role of Pref1 in the development of adiposity in our mouse model, it has to be noted that its expression has been determined in epididymal fat, a depot without direct parallel in humans. Thus, although many adipose genes are similarly expressed in both rodents and humans (50), extrapolation of Pref1 gene expression data from our model to humans must be undertaken with caution.

Imprinting of Pref1 is under complex control by both paternal and maternal alleles (28). We did not detect changes in methylation of the differentially methylated regions that regulate imprinting at this locus in fat from 4-month-old mice of either F₁ or F₂ generations (supplementary Fig. 4). These data suggest that, as with Sur1, altered expression of Pref1 is due to either changes in expression or function of an upstream transcription factor, histone modification, or another yet-unknown genomic region that is epigenetically modified. Thus, additional studies are required to define the potential role of epigenetics in adipose development in our model.

In conclusion, we demonstrate that maternal undernutrition during pregnancy programs reduced birth weight, glucose intolerance, and obesity in first- and second-generation offspring, even despite ad libitum feeding during second pregnancy. Different aspects of these phenotypes are transmitted via the maternal lineage (obesity), the paternal lineage (reduced birth weight), or both (glucose intolerance). Sex differences in transmission of phenotypes implicate complex mechanisms: 1) matrilineal inheritance of disease is multifactorial and includes metabolic, epigenetic, and mitochondrial mechanisms; and 2) patrilineal inheritance is primarily due to epigenetic mechanisms. While we do not yet understand the complex array of molecular mechanisms associated with fetal programming of disease, such studies will be of great importance for the design of future therapeutic interventions aimed to prevent and/or modulate adult phenotypes, not only in LBW humans but also in their children and grandchildren.

The authors gratefully acknowledge support for this study from the American Diabetes Association (Research Award, to M.E.P.), the Graetz Family Fund (to M.E.P.), the Marie Curie International Reintegration Grant (MIRG-CT-2007-046542, to J.C.J.-C.), and the Diabetes Endocrinology Research Center (P30-DK36836).

No potential conflicts of interest relevant to this article were reported.