Research Article

In utero undernourishment perturbs the adult sperm methylome and intergenerational metabolism

See allHide authors and affiliations

Science  15 Aug 2014:
Vol. 345, Issue 6198, 1255903
DOI: 10.1126/science.1255903

Structured Abstract

Introduction

The rapid global rise in metabolic disease suggests that nongenetic environmental factors contribute to disease risk. Early life represents a window of phenotypic plasticity important for adult metabolic health and that of future generations. Epigenetic inheritance has been implicated in the paternal transmission of environmentally induced phenotypes, but the mechanisms responsible remain unknown.

Embedded Image

In utero undernourishment alters the adult germ cell methylome. Undernourishment during PGC reprogramming results in hypomethylation of discrete loci in adult sperm. These regions are enriched in nucleosomes and are low-methylated regions. Although partially resistant to blastocyst reprogramming, differential methylation does not persist in the next generation. However, dysregulated expression of genes neighboring DMRs is observed in F2 offspring.

Rationale

We investigated the role of DNA methylation in epigenetic inheritance in an established murine model of intergenerational developmental programming. F1 offspring of undernourished dams (UN) have low birth weight and multiple metabolic defects. Metabolic phenotypic inheritance to the F2 generation is observed through the paternal line, even though the F1 mice did not experience postnatal environmental perturbation. The timing of nutritional restriction coincides with methylation reacquisition in F1 male primordial germ cells (PGCs). Therefore, we assessed F1 sperm whole-genome methylation using immunoprecipitation of methylated DNA, combined with high-throughput sequencing, followed by independent validation. We characterized the regions susceptible to methylation change and investigated the legacy of such methylation change in the phenotypic development of the next generation.

Results

In UN mice, 111 regions are hypomethylated relative to control sperm, and these changes are validated by bisulfite pyrosequencing. Methylation differences span multiple CpGs, with robust absolute changes of 10 to 30% (relative reduction ~50%). The absolute methylation level is consistent with differentially methylated regions (DMRs) being “low-methylated regions,” known to be enriched in regulatory elements. Indeed, luciferase assays suggest a role for these DMRs in transcriptional regulation. Hypomethylated DMRs are significantly depleted from coding and repetitive regions and enriched in intergenic regions and CpG islands. They are also enriched in nucleosome-retaining regions, which suggests that, at some loci, paternal germline hypomethylation induced by in utero undernutrition is transmitted in a chromatin context. DMRs are late to regain methylation in normal male PGCs. This may render them particularly susceptible to environmental perturbations that delay or impair remethylation in late gestation.

Except for imprinted loci, gene-associated male germline methylation has generally been thought to be largely erased in the zygote,although recent studies suggest that resistance to reprogramming is more widespread. Indeed, 43% of hypomethylated DMRs persist and thus have the potential to affect development of the next generation. We show that differential methylation is lost in late-gestation F2 tissues, but considerable tissue-specific differences in expression of metabolic genes neighboring DMRs are present. Thus, it is unlikely that these expression changes are directly mediated by altered methylation; rather, the cumulative effects of dysregulated epigenetic patterns earlier in development may yield sustained alterations in chromatin architecture, transcriptional regulatory networks, differentiation, or tissue structure.

Conclusion

Prenatal undernutrition can compromise male germline epigenetic reprogramming and thus permanently alter DNA methylation in the sperm of adult offspring at regions resistant to zygotic reprogramming. However, persistence of altered DNA methylation into late-gestation somatic tissues of the subsequent generation is not observed. Nonetheless, alterations in gamete methylation may serve as a legacy of earlier developmental exposures and may contribute to the intergenerational transmission of environmentally induced disease.

The nutritional sins of the mother…

Prenatal exposures of a mother can affect the health of her offspring, but how? Radford et al. found that the male progeny of undernourished pregnant mice had altered DNA chemistry in their sperm. In addition, the offspring displayed compromised metabolic health. The specific affected genes not only lost DNA methylation but also lacked the normal sperm DNA packaging factors (protamines) and instead were enriched in nucleosomes. Thus, when subjected to a suboptimal prenatal environment, offspring feel the effects of the maternal assault.

Science, this issue p. 10.1126/science.1255903

Abstract

Adverse prenatal environments can promote metabolic disease in offspring and subsequent generations. Animal models and epidemiological data implicate epigenetic inheritance, but the mechanisms remain unknown. In an intergenerational developmental programming model affecting F2 mouse metabolism, we demonstrate that the in utero nutritional environment of F1 embryos alters the germline DNA methylome of F1 adult males in a locus-specific manner. Differentially methylated regions are hypomethylated and enriched in nucleosome-retaining regions. A substantial fraction is resistant to early embryo methylation reprogramming, which may have an impact on F2 development. Differential methylation is not maintained in F2 tissues, yet locus-specific expression is perturbed. Thus, in utero nutritional exposures during critical windows of germ cell development can impact the male germline methylome, associated with metabolic disease in offspring.

The rapid global rise in the incidence of diabetes, obesity, and cardiovascular disease suggests that nongenetic environmental factors are major contributors to disease risk. Epidemiological data and animal models have demonstrated that early life represents a window of phenotypic plasticity critically important for later adult metabolic health (1). The impact of the early-life environment has been observed to extend over multiple generations in both human populations and animal models (28). There are at least two potential mechanisms mediating such non-Mendelian phenotypic inheritance: alterations in the parental metabolic milieu that induce fetal developmental exposures in the second generation and epigenetic inheritance. The latter is strongly implicated when paternal transmission of environmentally induced phenotypes is observed because rodent males, present solely at breeding, contribute to the future generation only through the sperm. Although a role for histone modifications and/or RNA has been proposed (4), the epigenetic mechanism(s) responsible for intergenerational inheritance of environmentally induced phenotypes remains unknown.

Paternal transgenerational epigenetic inheritance of altered DNA methylation has been demonstrated previously: for example, in rodents exposed to the endocrine disruptor vinclozolin (9) and in mice with variable methylation at the Agouti viable yellow, Avy, and Axin-Fused, AxinFu, alleles formed by insertion of IAPs (intracisternal A particles) into or near endogenous genes (10). In addition to repeat-mediated cis-acting effects, other endogenous loci that have an inherent epigenetic vulnerability to environmental conditions may contribute to intergenerational phenotypes and play an important role in the developmental origins of health and disease. Furthermore, recent studies have suggested that resistance to zygotic DNA methylation reprogramming extends beyond imprinted domains (1113), which raises the possibility that gametic methylation may play a larger role than previously recognized in early development. A key unanswered question is whether an altered in utero environment or nutritional insult might affect the DNA methylation profile of adult germ cells.

Our aim was to investigate the role of DNA methylation in epigenetic inheritance in an established in utero murine model of intergenerational developmental programming (3). To produce the most robust phenotype, the maximum caloric restriction that does not cause significant fetal loss was chosen (Fig. 1A). This regime is largely incompatible with successful pregnancy in inbred mouse strains. Consequently, we used the outbred ICR strain, which also allowed us to better model the human population. In this model, F1 offspring of undernourished dams have low birth weight as well as early-life adiposity, reduced muscle stem cell number and function, impaired pancreatic function, and progressive glucose intolerance (1416). Inheritance of significantly reduced birth weight and glucose intolerance to the F2 generation is observed through the paternal line in the absence of any further environmental perturbation (fig. S1, D to H) (3). The period of experimentally induced nutritional restriction in this model (day 12.5 to 18.5 of pregnancy) coincides with the reacquisition of methylation in male primordial germ cells as they are epigenetically reprogrammed (17). The dynamics of such methylation changes have been best studied at imprinting control regions (ICRs). However, we have already excluded a substantial perturbation of methylation at ICRs in this model (18). Thus, we now assess the whole-genome distribution of methylation in F1 sperm, using immunoprecipitation of methylated DNA combined with high-throughput sequencing (MeDIP-seq) (1921), followed by independent validation by bisulfite sequencing.

Fig. 1 Total methylation is stable in UN sperm, with significant locus-specific changes.

(A) Experimental design. F1 generation: Dams were randomized on pregnancy day 12.5 to control (C) or undernutrition (UN) groups, and UN food intake was restricted to 50%. Postnatal litters were equalized to eight pups, and animals were fed ad libitum. F2 generation: Control F1 females mated at age 2 months with nonsibling control or UN males and fed ad libitum to produce: CC, both parents controls; CU, control dam, UN sire. (B) Independent sperm DNA samples were quantified and pooled in equimolar ratios to generate two pools per condition. Control pools: n = 8, five litters. UN pools: n = 8, four litters. After MeDIP-seq, two independent C versus UN comparisons identified DMRs where methylation fold change >1.5× and binomial P < 0.0001 in both independent biological replicates. (C) Mass spectrometry quantification of control and UN sperm 5-methyl-cytosine (above) and 5-hydroxymethyl-cytosine (below). E14 embryonic stem cells (ESCs) are shown for comparison. (D) Heat map of 111 hypomethylated DMRs (left) and 55 hypermethylated DMRs (right). Hypermethylated DMRs did not validate.

Experimental design and metabolic phenotype

Mature sperm was isolated from F1 male mice fed standard chow, ad libitum, at 3 months of age, before the onset of glucose intolerance or any discernible metabolic phenotypes (14). These F1 males, previously exposed to experimental undernutrition in utero (UN), were smaller at birth (UN 1.34 ± 0.025 g, controls 1.65 ± 0.028 g, P < 0.0001) and at 3 months of age (UN 41.4 ± 0.82 g, controls 44.2 ± 0.94 g, P = 0.04) (fig. S1, B and C). Blood glucose and white adipose tissue mass at the time of sperm collection were not different between UN and control mice (fig. S1C). We bred F1 control and UN males with control females before sperm isolation; offspring of these pregnancies were designated as CC (F2 offspring of control males) and CU (F2 offspring of UN males) (Fig. 1A). The F2 offspring of F1 sperm donors were harvested at embryonic day 16.5 (E16.5). A contemporaneous adult cohort of F2 CU mice demonstrated, at 8 months of age, metabolic phenotypes similar to those previously observed (3), including reduced muscle mass and increased adiposity, with no difference in overall body or brain weight (fig. S1D). Furthermore, this CU cohort also showed glucose intolerance, particularly in the first-phase response to a glucose challenge (fig. S1E), as was previously observed. Pyruvate tolerance tests suggest that increased gluconeogenesis may contribute to this glucose intolerance (fig. S1F).

To assess whether a metabolic phenotype is discernible at E16.5 in the F2 generation, we examined lipid metabolism. There is an overall trend toward increased lipid abundance, particularly for saturated fatty acid–conjugated triglycerides (fig. S1G). This is associated with a significant increase in expression of genes involved in lipid oxidation in E16.5 CU liver, such as PPARα, Pgc1α, and Pgc1β, and a trend toward down-regulation of genes involved in lipid synthesis, including Scd1, Srebp1, and Dgat1 (fig. S1H), likely secondary to the increased hepatic lipid abundance at E16.5. Together, these data suggest that CU individuals have altered metabolism even in utero.

Hypomethylation of discrete loci in F1 adult sperm of males undernourished in utero

To confirm the purity of F1 sperm samples, bisulfite sequencing of imprinting control regions was carried out (fig. S2A). Independent sperm DNA samples were pooled in equimolar ratios to make two pools for each condition, each pool comprising four individuals from four independent litters (Fig. 1B), which minimized outcomes that might be associated with interindividual genetic differences. Mass spectrometry analysis of F1 sperm DNA demonstrates that in utero nutrition does not affect the total level of DNA methylation or hydroxymethylation (Fig. 1C). It is also notable that the level of hydroxymethylation in sperm is only 2.1% of that observed in embryonic stem cells (Fig. 1C). Consequently, only the genomic distribution of DNA methylation was analyzed further.

We assessed the genome-wide distribution of sperm methylation by MeDIP-seq (Fig. 1B). This approach is most suited to the detection of robust regional changes in DNA methylation, as it offers near-unbiased genome-wide coverage with underrepresentation of low-density mC/mCG (22), which minimizes the possible influence of single-nucleotide variants and allows identification of clusters of differentially methylated cytosines. Optimization of antibody specificity was carried out to ensure that there was no cross-sampling of hydroxymethylated or unmethylated cytosine (fig. S2B) (see materials and methods). Sequencing of antibody-enriched samples generated a total of 322.6 million mappable reads for control and 301.8 million for UN libraries. We conducted two independent comparisons between the control and UN pools, using the MEDIPS package (23) (see materials and methods for more details). Loci with a methylation change >1.5-fold and a binomial P < 0.0001 in both of the independent comparisons were selected for further study and clustered into 166 differentially methylated regions (DMRs), of which 111 were hypomethylated and 55 hypermethylated in UN relative to control sperm (Fig. 1D).

Bisulfite pyrosequencing validation of MeDIP-seq DMRs

To independently validate regions of altered methylation using a different technology, we used bisulfite pyrosequencing assays on 32 regions, using an expanded panel of sperm samples: 12 control males from five litters and 11 UN males from four litters. Twenty-four hypomethylated regions and eight hypermethylated regions were randomly chosen for validation, distributed throughout the range of fold change and P values. No significant difference in methylation was found at any hypermethylated DMR, which suggests that these regions may be false-positives (table S1). In contrast, significant loss of methylation was confirmed at 17 of the hypomethylated regions in the expanded panel of F1 UN sperm samples (Fig. 2 and Table 1). The validation rate of the nonrepetitive, unique hypomethylated regions was 90%. Differences in methylation at these loci span multiple CpGs, with robust absolute changes of 10 to 30%, a relative reduction of up to 50% (Fig. 2). Moreover, these differences are remarkably consistent among individual animals from multiple independent litters, which indicates that they are unlikely to be caused by genetic variation (fig. S3). The bisulfite sequencing data show identical absolute levels of methylation in the two replicate pools assessed by MeDIP-seq (fig. S3). Furthermore, the absolute methylation level (generally, under 50% in both groups) is consistent with these DMRs being “low-methylated regions,” previously shown to be enriched in regulatory elements (24). Together, these data demonstrate that discrete loci in the adult male germ line are susceptible to changes in methylation as a result of nutritional stress in utero.

Fig. 2 Bisulfite mutagenesis validation of hypomethylated DMRs in an expanded panel of F1 males’ sperm.

Seventeen genomic regions validated (Table 1). Data plotted: means ± SEM. C: n = 12, five litters; UN: n = 11, four litters. *P < 0.05, **P < 0.01, ***P < 0.001; unpaired two-tailed t test.

Table 1 Validation of hypomethylated DMRs by bisulfite pyrosequencing.

Absolute methylation level calculated by bisulfite mutagenesis combined with pyrosequencing in C and UN F1 sperm (n ≥ 11, ≥4 litters), F2 E16.5 brain and liver (n ≥ 12, ≥3 litters) at hypomethylated DMRs. DMRs at nonrepetitive, distinct loci are indicated by an asterisk (*). Blastocyst methylation level was extracted from (12) and (26).

View this table:

DMRs are not distributed randomly through the genome

We examined the distribution of unique and repetitive elements among DMRs. Hypomethylated DMRs are significantly depleted from coding regions but enriched in intergenic regions and CpG islands (Fig. 3). Repetitive elements are significantly depleted from hypomethylated DMRs (χ2 P < 0.0001) with underrepresentation of long interspersed nuclear elements (LINEs) (χ2 P = 0.001) and short interspersed nuclear elements (SINEs) (χ2 P < 0.0001) and no significant enrichment of IAPs (Fig. 3).

Fig. 3 DMRs are enriched in intergenic nonrepetitive regions and CpG islands.

(Top) Relative distribution (%) of F1 sperm methylated regions among distinct sequence and repetitive elements genome-wide (left) and among the F1 UN sperm hypomethylated DMRs (right). Distinct regions are significantly enriched (χ2 P < 0.0001), whereas LINEs and SINEs are significantly depleted from hypomethylated DMRs (χ2 P = 0.001; χ2 P < 0.0001, respectively), relative to all methylated regions detected in F1 sperm. (Middle) Relative distribution (%) of methylated regions among coding and noncoding sequences. Exons are significantly depleted (χ2 P = 0.036), and intergenic regions are significantly enriched (χ2 P = 0.0012) among hypomethylated DMRs. (Bottom) Relative distribution (%) of methylated regions detected by MeDIP-seq among CpG islands (CGI) and CGI shores. CGIs are significantly enriched among hypomethylated DMRs (χ2 P < 0.0001).

The predominance of hypomethylated DMRs is striking. This is consistent with in utero undernutrition during the final third of gestation impairing the reacquisition of methylation in developing F1 male primordial germ cells (PGCs). The nutritional insult experienced by the fetus worsens with increasing gestation as maternal energy reserves are depleted. Therefore, we hypothesize that the likelihood of remethylation being disrupted by in utero undernutrition increases toward term. Analysis of the temporal dynamics of methylation reprogramming in normal PGCs (25) suggests that this is indeed the case. In normal male PGCs, whole-genome methylation is progressively reduced from E6.5 to E13.5, with evidence of remethylation by E16.5 (Fig. 4A, gray bars). In contrast, those DMRs found to be hypomethylated in adult UN sperm (green bars) exhibit a distinct temporal pattern of reprogramming. These DMRs have significantly lower methylation levels at E16.5 in normal male PGCs (χ2 P < 0.0001) (Fig. 4A), which suggests that these regions are late to remethylate and may be susceptible to environmental perturbations that delay or impair remethylation at this stage. In normal adult sperm, methylation has largely been regained, but a minority of regions retain low methylation levels (26). UN-associated hypomethylated DMRs are enriched in these low methylated regions (χ2 P < 0.0001) (Fig. 4A).

Fig. 4 DMRs regain methylation late during PGC reprogramming and retain nucleosomes in mature sperm.

(A) Methylation level of hypomethylated (green) and hypermethylated (red) DMRs in our data set versus the whole genome (gray) in normal PGCs (25) and mature sperm from adult males (26). Hypermethylated DMRs act as an additional negative control because they did not validate. E13.5 and E16.5 are male PGCs. E6.5 and E11.5 are mixed-sex PGCs (25). (B) Nucleosome enrichment (27) at five representative hypomethylated DMRs.

During spermiogenesis, 99% of histones are exchanged for protamines, but nucleosomes are particularly retained in regions of high CpG density and low DNA methylation (27). Given the low methylation level of our DMRs, we assessed whether these regions are also enriched in nucleosomes. In mature sperm, 23 out of 111 (21%) hypomethylated DMRs retain nucleosomes (Fig. 4B). Bootstrap resampling of randomly selected regions from the background methylome demonstrates that this is a significant enrichment, P < 0.0001, and a feature of low methylated regions (see fig. S4 for details). This suggests that, at some loci, paternal germline hypomethylation induced by in utero undernutrition is transmitted in a chromatin context.

The developmental legacy of germline DMRs in late gestation of the F2 generation

With the exception of imprints, it has been thought that gene-associated methylation in the male germ line is largely reprogrammed in the zygote by active DNA demethylation (17). However, recent studies suggest that resistance to DNA methylation reprogramming extends beyond imprinted domains (1113). Indeed, 43% of our hypomethylated DMRs are resistant to zygotic reprogramming (26), which suggests that differential methylation in the paternal germ line may persist into the early embryo and may affect the development of the next generation (Table 2).

Table 2 Of hypomethylated DMRs, 43% are resistant to zygotic demethylation.

Hypomethylated DMRs susceptible (<20% methylation) or partially resistant (>20% methylation) to blastocyst reprogramming (28).

View this table:

To determine whether altered F1 sperm methylation persisted as a “memory” of sperm compromise in F2 offspring, we bred F1 control and UN males with control females. Offspring were designated as CC (F2 offspring of control males) and CU (F2 offspring of UN males) as noted above (Fig. 1A). Using liver and brain samples from late-gestation (E16.5) CC and CU embryos, we analyzed DNA methylation at validated germline DMRs. Differential methylation has been lost in F2 E16.5 brain and liver (Table 1 and Fig. 5) and is therefore not a long-term heritable memory of a compromised germ line. These data indicate that any functional consequences of germline DMRs are likely to be established early in development and/or linked to associated, but currently unknown, regulatory effects that may persist despite DNA remethylation in later development.

Fig. 5 Analysis of methylation at F1 sperm DMRs in F2 brain and liver at E16.5.

F2 E16.5 CC and CU brain and liver methylation of F1 sperm previously validated hypomethylated DMRs, measured by bisulfite pyrosequencing. Data are presented as means ± SEM. Brain per condition n = 16, ≥3 litters; liver per condition n = 12, three litters.

Analysis of validated DMRs in publicly available data sets (28) indicates that these loci have cell type–specific enrichment of histone modifications and transcription factor binding, characteristic of a role in cis regulation of transcription (table S2). To assess the function of a randomly selected subset of six DMRs, we conducted luciferase reporter assays in neural stem cells (29) and NIH3T3 cells in culture, using methylation-stable regions (non-DMRs) validated by pyrosequencing as additional negative controls. No significant enhancer function could be attributed to any of the regions tested in either cell type. In contrast, in vectors designed to assess a negative influence on transcription, such as an enhancer blocking or silencer function, when inserted in both the forward and reverse orientation, five out of six regions showed significantly suppressed reporter activity in neural stem cells as did three out of six regions in NIH3T3 cells (Fig. 6A). Taken together, the data suggest that these germline DMRs may play cell-specific regulatory roles in the modulation of transcription.

Fig. 6 Developmental legacy of altered UN sperm methylation in the F2 generation.

(A) Luciferase assay for a negative effect on transcription in 46C neural stem cells (29) (left) and NIH3T3 cells (right). Sequences were inserted between the promoter and enhancer of the control pGL3 vector. The pGL3 promoter vector (lacking an enhancer) was used as a positive control. Two regions validated by pyrosequencing as having unaffected F1 sperm methylation were used as negative controls. Control 1: MMU2:77723600–77723900, control 2: MMU17:87639700-87640000. Data are plotted as means ± SEM, normalized to activity of the control pGL3 vector with no insert. One-way analysis of variance (ANOVA), Dunnett’s post hoc test **P < 0.001, ***P < 0.0001. (B) F2 E16.5 brain expression of genes neighboring F1 sperm DMRs. Data are plotted as means ± SEM. MiR-715 expression normalized to SnoRNA 202; all other expression normalized to Hprt. Hprt and SnoRNA202 were unaffected. Unpaired two-tailed t test: Gmf4983, P = 0.0004; C1qtnf6, P = 0.049; Sstr3, P = 0.02; Tacc2, P = 0.0018; Tfap2c, P = 0.015; and Tbc1d30, P = 0.006. Per condition n = 16, ≥3 litters. (C) F2 E16.5 liver expression of genes neighboring F1 sperm DMRs. Data are plotted as means ± SEM. Normalized as for (B). Unpaired two-tailed t test: Ppp2r5c variant1, P = 0.03; Kcnip1, P = 0.011. Per condition n = 12, three litters. (D) F2 pancreatic expression at 4 months. Per condition n ≥ 5, * P < 0.05, unpaired two-tailed t test (3). (E) Tolbutamide (200 μM)–stimulated insulin secretion, freshly isolated 4-month-old islets; n ≥ 4, ≥2 isolations. **P < 0.01, unpaired two-tailed t test (3). (F) Diazoxide (250 μM) inhibition of insulin secretion, freshly isolated 4-month-old islets; n = 4 per group, ≥2 isolations. *P < 0.05, unpaired two-tailed t test (3). (G) Fold change in serum insulin 30 min after intraperitoneal glucose bolus (1 mg/kg). **P < 0.01, n ≥ 8, unpaired two-tailed t test (3).

To assess this possibility, we examined expression of genes neighboring the 17 germline DMRs, using quantitative polymerase chain reaction (qPCR) in liver and brain of F2 CC and CU fetuses at E16.5. Genes associated with DMRs 15 and 16 were not expressed in these tissues. Eight DMRs showed significant tissue-specific differences in expression of neighboring genes (Fig. 6, B and C). In contrast, no change in expression was found at 12 genes not associated with DMRs (fig. S5). Because methylation differences are not observed in E16.5 tissues of these same F2 offspring (Fig. 5), it is unlikely that these expression changes are directly mediated by alterations in methylation. Rather, the cumulative effects of dysregulated epigenetic patterns earlier in development may yield sustained alterations in chromatin architecture, transcriptional regulatory networks, cell type, or tissue structure.

Several affected genes, including Sstr3, C1qntf6, Tbc1d30, Kcnj11, and Sur1, are candidate contributors to the F2 phenotypes, given their known roles in glucose tolerance and metabolism (3036). For example, the DMR9 lies within the Kcnj11 gene, immediately downstream of Sur1. These genes encode the two subunits of the pancreatic β-cell adenosine triphosphate (ATP)–dependent K+ channel, which are necessary for the physiological control of insulin secretion (34, 36). Furthermore, polymorphisms at these loci are associated with type 2 diabetes (35, 37). In pancreatic islets isolated from 4-month-old CU mice (F2 generation), expression of Sur1 is reduced by 33% (P < 0.05) (Fig. 6D) (3). The function of β-cell ATP-dependent K+ channels in controlling insulin secretion can be assessed by measuring the response to treatment with agents that inhibit and activate these channels, such as sulfonylureas and diazoxide, respectively, or through the insulin secretory response to a glucose challenge. Freshly isolated 4-month-old CU pancreatic islets demonstrate impaired insulin secretion in response to the sulfonylurea tolbutamide (Fig. 6E) and absence of suppression of insulin secretion to diazoxide (Fig. 6F) (3). Furthermore, basal insulin secretion before diazoxide challenge was significantly reduced (3) (Fig. 6F). Consistent with this, CU individuals secrete significantly less insulin during glucose tolerance testing (3) (Fig. 6G). These data strongly suggest impaired function of ATP-dependent K+ channels in the adult CU pancreas and implicate this in the altered glucose tolerance observed in CU individuals (fig. S1E). Further work will be required to delineate the precise relation between compromised F1 germline reprogramming at these loci and F2 phenotypes.

Discussion

Our data indicate that nutritional perturbations during in utero development can alter male germline methylation at discrete loci. In turn, some of these DMRs are associated with differential transcript expression during offspring embryonic life. Our findings contrast with the largely negative data of Carone et al. in which no significant changes were observed in sperm DNA methylation after dietary protein restriction in adult males (4). Disparities may be due to the use of caloric rather than protein restriction, strain differences, or the greater number of individuals assessed in our analysis. Alternatively it may be due to differences in the timing of the nutritional insult, because of Carone and colleagues’ imposed protein restriction during adult life. By contrast, the nutritional perturbation in our model occurs exclusively during late prenatal life, precisely when male PGCs in the developing embryo are undergoing reestablishment of their epigenetic profile. At this time, PGCs may be particularly vulnerable to epigenetic perturbation. It is notable that intergenerational phenotypic inheritance caused by endocrine disruptors associated with altered sperm DNA methylation also involves prenatal exposure (9, 38). However, recent data have suggested that a high-fat diet during adult life might alter sperm DNA methylation, which indicates that the adult germline methylome may be more susceptible to environmental conditions than previously thought (8).

Our experiment was designed to minimize detection of single CpG methylation differences, which we a priori hypothesize to be more likely to be due to genetic differences. Our results indicate that robust germ cell methylation changes do occur after in utero undernourishment at regions partially resistant to zygotic reprogramming. However, persistence of altered DNA methylation into late-gestation somatic tissues of the subsequent generation was not observed. Nonetheless, gene expression is altered in these F2 offspring at regions of F1 germline differential methylation. Such differences in gene expression could reflect the impact of altered methylation during early development, with subsequent transcriptional patterns that persist despite DNA remethylation in later gestation. Alternatively, altered F2 expression may be the cumulative result of multiple locus-specific defects in germline chromatin state. Further work will be required to explore these possibilities.

Recent work in cultured cells demonstrates that regional methylation levels can be a secondary consequence of changes in DNA binding factors (24). Thus, it is possible that the germline DMRs identified in our study are secondary to other chromatin perturbations. Consistent with this, we observed enrichment of nucleosome occupancy at DMRs. Further studies are required to examine whether these represent regions of vulnerability in the sperm genome. Histone modifications and small RNA molecules are known to be required for multigenerational gene-silencing effects in Caenorhabditis elegans (39, 40), an animal that lacks DNA methylation, and such mechanistic processes may also be involved in mammals. Indeed, there is evidence that sperm-borne microRNAs play an important role in early mammalian development (41), and the early-life environment may have the potential to alter the abundance of some sperm microRNAs (42).

Conclusion

Data presented here serve as a proof of principle that undernutrition during prenatal life, even when followed by normal postnatal nutrition, can compromise male germline development and epigenetic reprogramming and so permanently alter DNA methylation in the germ line of the adult offspring. Alterations in adult gamete methylation may serve as a legacy of earlier developmental exposures that may contribute to the intergenerational transmission of environmentally induced disease.

Supplementary Materials

www.sciencemag.org/content/345/6198/1255903/suppl/DC1

Materials and Methods

Figs. S1 to S5

Tables S1 to S6

References (4352)

References and Notes

  1. Acknowledgments: We are grateful to members of the Ferguson-Smith and Patti labs for useful discussions of this work and to K. Tabbada and D. Oxley for expert technical assistance with Illumina sequencing and mass spectrometry. Work was supported by grants from the U.K. Medical Research Council; Wellcome Trust; and European Commission FP7, EpiGeneSys and EpiHealth (to A.C.F.-S.), from the Pediatric Endocrine Society and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, NIH (to E.I.), and from the American Diabetes Association, the Graetz Foundation, and NIH P30DK036836 (to M.-E.P.). W.R. is a consultant for Cambridge Epigenetix. The MeDIP-Seq data sets have been deposited in GEO with accession number GSE58747. A.H.F.M.P. and S.E. are funded by the Novartis Research Foundation.
View Abstract

Stay Connected to Science

Navigate This Article