Introduction

The use of prenatal restricted diets in rodent models can be considered to be a counterpart of the Dutch Hunger Winter syndrome observed in humans after the Second World War, as malnutrition during pregnancy leads to an effect on the health of the progeny in their adult life [1, 2]. It has already been well documented that suboptimal in utero environments in humans lead to long-term consequences for gene expression, DNA methylation, and phenotype [3]. These consequences of prenatal factors acting on the developing fetus are referred to as fetal programming [4]. In rodent models, nutrient-deficient prenatal diets have been widely studied with different modifications in terms of regimen type, time of exposure, and so on [5]. Our previous study applied a calorie-restricted diet [6], which was found to have greater impact on the lipid profile in the F0 dams than in their progeny. The hepatic expression level of the key lipid metabolism gene (Fasn) was also altered, while its methylation remained unchanged in response to the applied diet. This is in agreement with the earlier study of Lee et al. [7], in which the hepatic Fasn mRNA level also increased in response to the maternal calorie-restricted diet.

Early programmed changes during prenatal life may have consequences not only in adulthood, but can also be transmitted to the next generation. Such transgenerational inheritance has already been studied for calorie-restricted diets and their effect on the appetite [5]. The authors of that paper performed a meta-analysis of 35 studies in rats and mice, concluding that the transgenerational effects of restricted diets during pregnancy on progeny appetite were not significant, but that the reduced protein content in the diet significantly increased food intake in female offspring. Moreover, a comparison of obesogenic versus low-nutrition (protein-deficient) prenatal diets in terms of diabetes risk in the F2 generation was recently made by Hanafi et al. [8], who found that malnutrition had a greater transgenerational effect on F2, where glucose homeostasis was significantly altered simultaneously with overexpression of the glucose-sensing genes in fetal tissues. In a previous prenatal calorie-restricted diet study, we found changes in the expression profile of genes involved in epigenetic mechanisms (such as Dnmt1, Mecp2, Hdac1, and Sin3a), but the differences in the three generations were observed in the case of two genes [9]. Thus, it should be taken into account that the food deficiency can have an effect on DNA methylation and on histone modification profiles. Several studies have shown that modifications of epigenotype can be one of the mechanisms linking prenatal exposures and metabolism later in life [4].

In this study, we thus hypothesized that the use of a calorie-restricted diet during pregnancy in the rat can induce changes in hepatic DNA methylation and consequently lead to alterations in gene expression. To exclude the effect of postnatal feeding factors we performed the analysis for female fetuses at the age of 19^th days. Moreover, we assumed that the differences programmed in F1 can be inherited by the next generation. The aim of this study was to determine the effects of prenatal restricted diet on DNA methylation profile, as well as the transcript level of selected genes in the liver, in the female progeny of three generations of rats.

Material and methods

Results

Targeted bisulfite NGS approach

On average, 75% of the reads successfully underwent the trimming procedure. Of this group, 25% of reads mapped to the region captured by molecular probes with an average depth coverage of 100 reads and more than 85% of region being covered by at least 40 reads. Conversion efficiency at least 99% across all samples.

After rejecting cytosines with low coverage, we examined the methylation level differences in 108,019 positions located in 3458 CGIs. We obtained statistically significant differences in 394 positions (adjusted p-value (q-value) < 0.05) in close proximity to 374 genes. Ninety-eight percent of the indicated differences were lower than 10% (Figure A in S2 File). Gene clustering analyses, based on KEGG, revealed 23 significantly enriched pathways (Table C in S2 File). More detailed analysis indicated that among others, 22 genes are involved in oxidative phosphorylation processes (KEGG pathway ID: rno00190) associated with five mitochondrial membrane complexes (Table D in S2 File). However, the differences between the C and the R groups in the methylation level of cytosines varied from -3% to +3%. Thus, according to manufacturer’s recommendations, only seven cytosines which showed a difference of more than 10% between the studied groups (Fig 1) were selected for further analysis. They were located in the proximity of the Usp30 (q-value = 0.0009), Grb10 (q-value = 0.0285), Pld1 (q-value = 0.0324), St6galnac5 (q-value = 0.0255), Oxct2b (q-value = 0.0003), Khk (q-value = 0.0285), and Acsl4 (q-value = 0.0480). The mean read’s coverage from the tNGS data for these selected cytosines reached from 33.5 for Pld1 gene to 72.3 for Oxct2b gene (Table E in S2 File). Moreover, we also checked to which molecular pathways these genes have been assigned (according to KEGG database)–Table F in S2 File.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. DNA methylation differences (≥ 10%) between the control and restricted groups.

Results from targeted bisulfite next-generation sequencing of 1748 genes in the fetal liver. The boxes indicate the lower and upper quartiles, with the median marked by the horizontal line. Whiskers indicate 1.5 of the interquartile range. Outlying observations are marked with diamonds.

https://doi.org/10.1371/journal.pone.0215471.g001

Transgenerational DNA methylation analysis

The hepatic level of the seven cytosines selected from NGS was next determined in three generations of female fetuses. For all the studied cytosines, with the exception of the Grb10 gene (where it was not possible to design primers near the cytosine), the amplicons covered the cytosine indicated by NGS, as well as the neighboring cytosines in the CGI. For Grb10, the sequenced fragment starts 217 bp upstream of the selected cytosine. The results showed that only two of the studied genes in F1—Usp30 (CpG4, p = 0.008—Fig 2A) and St6galnac5 (CpG1, p = 0.034 and CpG2, p = 0.049—Fig 2B) maintained a significant difference between the study groups, but in the case of the St6galnac5 gene, the observed differences had the inverse proportions (with a decreased level in the restricted group). There were no differences in F2, while in F3 only the cytosines near Oxct2b significantly differed between the C and R groups (CpG2, p = 0.011; CpG3, p = 0.011 and CpG4, p = 0.040—Fig 2C). For Grb10, Pld1, Khk, and Acsl4, no difference was observed in any of the studied generations (Figure B in S2 File). Moreover, we performed additional analysis to check whether the level of methylation differed between generations independently from the applied dietary regimen and found such differences for Grb10 and St6galnac5 genes. The Tukey test showed that in all indicated cytosines the methylation level was lower in F1 generation as compared to F2 and F3, but there was no dietary group differences (Table G and Figure C in S2 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Transgenerational study of the cytosines (CpG) methylation levels by pyrosequencing.

The underlined CpGs are these selected from NGS, while the nonunderlined CpGs are the adjacent to the CpG of interest. Values are means ± SDs. A: Usp30; B: St6galnac5; C: Oxct2b genes. A single star indicates significant differences: p = 0.034 for CpG1 and p = 0.049 for CpG2 in St6galnac5; p = 0.011 for CpG2, p = 0.011 for CpG3, and p = 0.040 for CpG4 in Oxct2b. A double star means p = 0.008 for CpG4 in Usp30.

https://doi.org/10.1371/journal.pone.0215471.g002

Discussion

We performed targeted bisulfite NGS to determine the methylation level of over 1700 genes mostly involved in energy homeostasis pathways, and we found that over 300 cytosines showed very small though significant differences between the control and restricted groups. The clustering analysis indicated 23 detailed molecular pathways involving these genes. In the majority of these pathways, these genes were not directly functionally related to each other, with the exception of the oxidative phosphorylation process. In this pathway, 22 genes had altered DNA methylation—those encoding enzymes for NADH dehydrogenases, succinate dehydrogenases, ubiquinol-cytochrome c reductases, cytochrome c oxidases, pyrophosphatase, and ATP synthases. The differences were small (± 3%), but it cannot be excluded that such minor differences may affect metabolism, especially when they are involved in the same processes. The oxidative phosphorylation process that takes place in mitochondria is a key pathway where adenosine triphosphate (ATP) is produced as the main source of energy for cell functioning, as well as for the maintenance of metabolic homeostasis [14]. It can thus be speculated that altered methylation of genes involved in this process can change their expression levels, leading to a disruption of energy balance. Since small differences in a majority of genes were observed for other metabolic pathways, we anticipated that the use of the restricted diet during pregnancy did not dramatically alter DNA methylation in the offspring. In a previous study by Ogawa et al. [15], it was found that a prenatally experience calorie-restricted diet induced a wide range of changes in the hepatic methylome in the fetal liver of mice. It should be mentioned that Ogawa et al. used a different experimental approach in which the biological samples from individual animals were pooled and the methylation level was established using methylated DNA immunoprecipitation with microarray detection (Me-Dip-ChiP). They found that multiple genes involved in glucocorticoid signaling, insulin resistance, PPAR targets, cholesterol and fatty acid metabolism pathways, and immune response processes had altered methylation levels in their promoter regions. Of these, the imprinted Grb10 gene (growth factor receptor-bound protein 10), involved in insulin and mTOR signaling pathways, was changed by the restricted maternal diet. This was also seen in our study, where NGS analysis showed reduced methylation in the restricted group. Ogawa et al. [15] also observed methylation differences in the Acsl6 (long-chain-fatty-acid-CoA ligase 6) gene, which is involved in fatty acid metabolism, whereas our NGS study found such differences in Acsl4, which belongs to the same gene family. Another study employing a 50% reduction in the calorie intake of pregnant mice was performed by Ganguly et al. [16]. To determine the effect of the pregnant dams’ diet on the developing fetuses, the authors focused on transplacental glucose transport processes and studied the Glut1 and Glut3 genes. They noted that, in the restricted group, the methylation level in the promoter region of Glut3 was significantly increased. Moreover, this led to enhanced binding of methyl-CpG-binding protein (Mecp2), together with histone deacetylases (Hdac2); consequently, a reduced level of placental transcription of Glut3 gene was observed. This demonstrates that gene expression was altered by changed epigenetic mechanisms induced by the dietary factor. In our earlier study, we did not observe any such correlation for Fasn, a key lipid metabolism gene [6]. It should be mentioned that, in this study, the seven differently methylated cytosines indicated by NGS were also analyzed by pyrosequencing, and only one of them (near the St6galnac5 gene) had such a correlation, though with a very small difference between the studied groups. This shows that the technological approach must be considered when drawing conclusions. Since minor effects on DNA methylation could be expected from the experimental procedures, the targeted methods applied here for mainly energy homeostasis genes should be extended to global DNA methylation analysis in the future. Moreover, we found that the methylation level for some cytosines in Grb10 and St6galnac5 genes in the first generation was lower than in F2 and F3, in both the C and R groups. A similar tendency, though not significant, was observed for the other examined genes. It can thus be speculated that unknown factors acting on both control and restricted groups affected the methylation profile in F1.

The DNA methylation results obtained here are consistent with our previous results on one-carbon metabolism, in which we analyzed the mothers of the fetuses examined in this work [11]. We showed that food deprivation leads to reduced S-adenosylmethionine concentrations in pregnant dams, but that methylation potential did not differ between the control and restricted groups. Moreover, the transcript levels of the genes from the one-carbon metabolism pathway were not altered.

It seems that a prenatal calorie-restricted diet acts more effectively on the gene expression profile than on epigenetic mechanisms. In the study of Ogawa et al. [15], the global hepatic expression profile, in response to the prenatal restricted diet, was established in mothers and fetuses using a microarray approach. Upregulation was observed in 1941 and 1335 genes in mothers and progeny, respectively. Slightly more genes were downregulated in the liver (2301 in the mothers and 1626 in the offspring). A 20% restriction in diet during the preconception and entire pregnancy periods in rat progeny was recently studied by Ramírez-López et al. [17]. They examined different tissue expression profiles and, in the liver of female offspring, found decreased expression levels of genes like Pparα and Pparγ (key genes involved in adipogenesis), Acaca and Fasn (which encode the de novo lipogenic enzymes), and Insig1 and Hmgcr (important liver-specific cholesterol biosynthesis regulators). In consequence, the lipid profile in the female offspring was altered, with increased levels of plasma LDL concentration. This is in agreement with our previous data, where the LDL fraction was enhanced in the four-week old female offspring of mothers undergoing a 50% calorie restriction during pregnancy [6].

Transgenerational changes programmed by calorie-restricted diets have already been reported. Ponzio et al. [18] noted increased blood pressure in three generations of rats from the restricted group, while Araminaite et al. [19] found elevated body weight and shortened longevity in a male offspring of the F2 generation, indicating a sex-dependent effect of the maternal caloric restriction diet. Concerning the gene expression profile, Hoile et al. [20] employed a protein-restricted diet and concluded that only a tiny group of genes had the same expression profile in three generations, showing the transgenerational effect. Protein restriction was also studied by Burdge et al. [21], and a reduced level of DNA methylation in the promoter region of Pparalfa and Gr genes was seen in the F1 and F2 generations following elevated hepatic transcript level of these genes. In a study of mouse maternal undernutrition during the final week of pregnancy, changes in the expression of lipid and fatty acid metabolism genes were found in the F1 and F2 generations, while the transcript level of the imprinted genes was not altered [22]. In the present study, we found the test diet to have no effect on the hepatic transcript level of the seven studied genes in the first generation. It is thus unlikely that the differences observed in F2 and F3 were caused by the experimental diet. This is in agreement with our previous study [11], where no transgenerational effect was found in terms of hepatic one-carbon cycle metabolites and gene expression profiles. Interestingly, we observed that all the studied genes were overexpressed in the restricted group from F1, while they were underexpressed in F2. This suggests that other factors that affected gene expression might have occurred during F1 or F2 development.

Conclusions

In summary, we postulate that use of the calorie-restricted diet during rat pregnancy did not dramatically alter the DNA methylation of the genes involved in energy homeostasis, and transgenerational inheritance was not observed. Moreover, changed CpG methylation was not accompanied by alterations in transcription level. However, the effect of some minor differences in the methylation level of strictly functionally related genes cannot be excluded. We speculate that overall food deficiency activates regulatory processes that optimize the intrauterine environment for the developing fetuses. These processes stimulate the use of stored nutrients and prevent deregulation of DNA methylation. It must be also borne in mind that we did not perform genome-wide methylation analysis, so changes in other groups of genes were possible.

Supporting information

Acknowledgments

We wish to thank Shinichi Tomizawa for his support in global DNA methylation analysis and to Klaudia Rutkowska for technical support in pyrosequencing.