Undernourishment in utero induced metabolic disorders in F1 offspring

We established a mouse model as described by previous studies [14] and bred F1 and F2 generations (Fig. 1A) [15]. F1 offspring experiencing undernourishment in utero (UN) were smaller from birth to 12 weeks of age for both males (n ≥ 18) and females (n ≥ 23, Fig. 1B, C). The blood glucose was higher in the UN group (n ≥ 11) than that in the control (n ≥ 12) at 8 weeks of age (Additional file 1: Fig. S1A, B). However, it was similar between the control (n ≥ 13) and UN groups (n ≥ 18) at 12 weeks of age (Additional file 1: Fig. S1C, D). We further tested the glucose (GTT) and insulin tolerance (ITT) and found that both GTT and ITT were significantly altered by undernourishment in utero for female UN offspring (n = 10, p < 0.05, Student’s t-test, Fig. 1D, E). Disturbed GTT and ITT were also observed for male UN offspring (n = 10, Additional file 1: Fig. S1E, F) at 8 weeks of age.

Fig. 1

Effects of undernutrition in utero on metabolism of UN female. A The schedule of breeding. F1 and F2 were produced as shown. UN, undergoing undernourishment in utero; Con, the control group; CC, female control mated with male control; UC, female UN mated with male control. B and C The variation of body weight from birth to 12 weeks (n female ≥ 17, n male ≥ 17) of age. D and E GTT and ITT were tested at 8 weeks of age for females. F–H The plasma concentrations of leptin, adiponectin, and insulin were examined, and the statistical difference was tested by Student’s t test. I The L/A ratio (leptin/adiponectin) was calculated. *p < 0.05; **p < 0.01. Average data are showed as mean ± SD

Leptin, insulin, and adiponectin are three of the most important hormones regulating energy metabolism and body weight in mammals. Abnormal leptin, adiponectin, and insulin levels are associated with obesity, diabetes, and other metabolic disorders [16, 17]. We investigated the plasma concentrations of adiponectin, insulin, and leptin, and found that the plasma adiponectin concentration was significantly lower in female UN offspring compared with the control (UN 2.811 ± 0.2798 mg/l n = 8, control 5.663 ± 0.6312 mg/l n = 5, p = 0.0006, Student’s t-test, Fig. 1F). The plasma concentrations of leptin and insulin in female UN offspring (n = 8) were similar to the control (n = 9, leptin p = 0.9153; insulin p = 0.9288, Student’s t-test, Fig. 1G, H). The leptin/adiponectin ratio (L/A ratio) is a better indicator of metabolic diseases, such as obesity and diabetes mellitus [18]. The L/A ratio in F1 female UN offspring was about 2-fold of that in the control (p = 0.0012, Student’s t-test, Fig. 1I). For male UN offspring, the L/A ratio in UN was higher than that in control (Additional file 1: Fig. S1G), but the concentration levels of leptin, adiponectin, and insulin were not significantly affected (n = 8, Additional file 1: Fig. S1H, I, and J).

Intergenerational metabolic disorders induced by undernourishment in utero via female germline

To confirm the inheritance of metabolic disorders via female germline, we bred F1 control and UN females with control males; offspring of these were marked as CC (F2 offspring of control females) and UC (F2 offspring of UN females) (Fig. 2A). The body weight of UC females (n ≥ 14 from 5 litters) at birth and at 12 weeks of age (Fig. 2A) was similar to CC (n ≥ 12 from 5 litters). We further tested ITT and GTT, and they were significantly altered in UC females (n = 10 from 5 litters) compared with CC (n = 10 from 5 litters, p < 0.05, Fig. 2B, C). The plasma concentrations of adiponectin (p = 0.0301) and insulin (p = 0.0039) in female UC offspring (Student’s t-test, n = 12 from 6 litters) were significantly decreased compared with CC (n = 10 from 6 litters, Fig. 2D, E). The plasma concentration of leptin was similar between CC and UC (Fig. 2F). In addition, the L/A ratio increased 1.43-fold in female UC offspring compared to CC (p = 0.1539, Fig. 2G).

Fig. 2

The intergenerational transmission of perturbed metabolism in UC and CU females. A The average body weight changes of female UC. B and C The GTT and ITT changes for female UC, X-axial is the time point when glucose level was tested. D–F The plasma concentrations of leptin, adiponectin, and insulin. G The L/A ratio changes of female UC. The statistical difference was tested by Student’s t test, and * p < 0.05; **p < 0.01. Average data are showed as mean ± SD

For the UC males, the GTT, ITT, and body weight were not affected (n ≥ 10 from ≥5 litters, Additional file 1: Fig. S2A, B and C). Although the plasma concentrations of insulin, adiponectin, and leptin were similar between CC (n = 5 from 4 litters) and UC males (n = 8 from 4 litters, Additional file 1: Fig. S2 D, E, and F), the L/A ratio in UC males was significantly higher than that in CC (Additional file 1: Fig. S2G).

Global hypo-methylation in UN oocytes of F1 offspring exposed to undernourishment in utero

Considering the role of DNA methylation in the inheritance of metabolic disorders, we investigated the effects of undernourishment in utero on global methylation in F1 UN adult oocytes. A total of 100 MII oocytes were pooled to make two pools for each group, each pool comprising seven individuals from five independent litters. Base-resolution methylome was obtained by bisulfite sequencing (BS-seq) for single cell. C (cytosine) methylation was evaluated using Bismark at the sequencing depth ≥ 5 and q-value ≤ 0.01. Global C methylation (mC) level was lower in UN oocytes than that in control (Fig. 3A). According to the context, mC has three subtypes: mCG, mCHG, and mCHH (H = A, C or T). The proportion of mCG in three types was significantly decreased in UN oocytes compared with the control (p < 0.00001, Additional file 1: Fig. S3A).

Fig. 3

Global hypo-methylation in F1 UN oocytes. A Global C hypo-methylation was observed in UN oocytes, two repeats for each group. B Global hypo-methylation was also observed for CG sites in UN oocytes. C Global methylation level and methylation difference. The circles from outer to inner respectively represent chromosomes, the control methylation level in oocytes, the methylation difference between groups, the methylation level in UN oocytes, and graph legends. D–K The CG methylation level at different regions. *p < 0.05, ****p < 0.0001. Average data are showed as mean ± SD

mCG has a more important role in regulating gene expression and biological function; thus, we further analyzed CG methylation in oocytes. Hypo-methylation of CG was observed in oocytes of F1 generation (Fig. 3B), and this was distributed at all chromosomes (Fig. 3C). Further analysis in genomic features and gene regions showed that significant hypo-methylation was mainly distributed in promoters, exons, UTR5s, upstream 2kb regions of TSS (transcriptional start site), and downstream 2kb regions of TES (transcriptional end site) (Fig. 3D–K, and Additional file 1: Fig. S3B, C). These results suggest that the global methylation in UN oocytes is decreased by undernourishment in utero.

Effects of undernourishment in utero on the DMRs methylation level in oocytes of F1 generation

To better understand the effects of undernourishment in utero on genomic methylation in UN oocytes, we scanned the differentially methylated regions (DMRs) between UN and control oocytes at the number of CG ≥ 4 and the absolute difference of methylation ≥ 0.2. We identified a total of 2320 DMRs, of which 943 were hypo-methylated DMRs (hypo-DMRs, 41%) and 1378 were hyper-methylated DMRs (hyper-DMRs, 59%) in UN oocytes compared with control (Fig. 4A). And DMRs were located at all chromosomes (Fig. 4B). Furthermore, we examined the distribution of DMRs in genomic features including repeats and gene regions. Hyper-DMRs were significantly depleted from repeat regions, CGI-shores, UTR3s, UTR5s, and promoters, but enriched in exons and introns (Fig. 4C). Notably, hypo-DMRs were significantly enriched in CGIs, exons, and introns, but depleted from the other regions (repeat regions, CGI-shores, UTR3s, UTR5s, and promoters, Fig. 4C). These results indicate that DMRs are not randomly distributed throughout the genome in UN oocytes.

Fig. 4

DMRs methylation analysis in F1 UN oocytes. A Identified DMRs in UN oocytes including hyper-DMRs and hypo-DMRs. B DMRs was distributed at all chromosomes. The circles from outer to inner respectively represent chromosomes, statistical hyper-DMRs value log5 (|area Stat|), heat map of transcriptional elements (TE) and repeats, heat map of genes, statistical hypo-DMRs value log5 (|area Stat|), and graph legend. C Distribution of DMRs at different genomic regions, the left panel is the distribution of elements in oocytes; the middle panel is the distribution of hyper-DMRs in different elements, and the right panel is the distribution of hypo-DMRs in different elements. *p < 0.05; **p < 0.01, ***p < 0.001

DNA methylation patterns in promoters play an essential role in regulating gene expression. We identified 278 genes with DMRs in promoters, including 118 genes with hypo-DMRs and 160 genes with hyper-DMRs (Additional file 1: Table S1). To further understand the roles of DMRs in biological process, we analyzed the KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment of genes with DMRs in promoters in UN oocytes using KOBAS [19]. These genes were significantly (corrected p < 0.05) enriched in 10 KEGG pathways (Additional file 1: Table S2). The KEGG analysis showed that the genes with differentially methylated promoters were significantly enriched in metabolic pathways (22 genes, Table S3): 10 genes with hypo-DMRs, and 12 genes with hyper-DMRs. Some of genes enriched in metabolic pathways were relative to lipid synthesis, glucose metabolism, insulin secretion, mitochondrial function, fatty acid metabolism, insulin resistance, diabetes, and obesity, for instance Aacs (acetoacetyl-CoA synthetase), Acsl6 (Long-chain acyl-CoA synthetase 6), Adcy8 (Adenylate cyclase), Bcat2 (branched-chain aminotransferase 2), Coq5 (Coenzyme Q5), Got1/2 (aspartate aminotransferase 1/2), Isyna1 (Inositol-3-Phosphate Synthase 1), Pde4d (cAMP phosphodiesterase 4d), Pde10a (cAMP phosphodiesterase 10a), Pde8b (cAMP phosphodiesterase 8b), Plpp2 (phospholipid phosphatase 2), Shmt2 (serine hydroxymethyltransferase 2), Tkt (transketolase), car8 (carbonic anhydrase 8), and Pigl (phosphatidylinositol glycan anchor biosynthesis class L). These results indicate that these genes with DMRs in the promoter regions may play a crucial role in the metabolic disorders of F2 females.

Methylation level of the DMRs located at the promoters of metabolic-related genes in UN oocytes

To further confirm the methylation of the DMRs located at the promoters of metabolic-related genes in UN oocytes, we tested the methylation levels of some DMRs using BS: 4 hyper-DMRs located at the promoters of Hgsnat (H-DMR), Gsto2 (G-DMR), Car8 (C-DMR), Pde4d (P4-DMR), and 3 hypo-DMRs located at the promoters of Gnas (Gn-DMR), Pde8b (P8-DMR), and Acsl6 (A-DMR). A total of 80-120 oocytes were used for each DMR analysis. For hyper-DMRs, the methylation levels of H-DMR, G-DMR, and P4-DMR in UN oocytes were significantly higher than that in the control (Fig. 5A, B, and C). But the methylation level of C-DMR was significantly lower in UN oocytes compared to the control (Fig. 5D). For hypo-DMRs, the methylation levels of A-DMR and Gn-DMR were significantly lower in UN oocytes (Additional file 1: Fig. S4A, B). The methylation levels of P8-DMR were similar between UN and control oocytes (Additional file 1: Fig. S4C). These indicate that some regions were false-positives.

Fig. 5

Confirmation of methylation changes of selected DMRs in F1 UN oocytes. A P4-DMR. B G-DMR. C H-DMR. D C-DMR. E Adiponectin. *p < 0.05; ***p < 0.001. Black circle: methylated CG site; white circle: unmethylated CG site

Martinez et al. found that in utero malnutrition altered the methylation level of 5′ UTR region of Lxra in F1 sperm which may play a key role in the metabolic disorders of F2 [20]. Therefore, we further examined the methylation level of Lxra, but it was not significantly affected in UN oocytes (Additional file 1: Fig. S4D).

Adiponectin and Leptin are essential for the onset of metabolic disorders, such as diabetes and obesity, and their expressions are regulated by the methylation level in promoters. The global methylation sequencing results showed that the methylation levels in promoters of Leptin and Adiponectin were higher in UN oocytes compared with the control (Additional file 1: Fig. S5A and B). But just 3 CpG sites were covered for Adiponectin and 4 CpG sites were covered for Leptin in the genomic sequencing results. To more accurately assess the methylation of Leptin and Adiponectin in UN oocytes, we validated it by BS. We found that DNA methylation level in the promoter of Leptin in UN oocytes was similar to that of the control (Fig. Additional file 1: S4E). But the DNA methylation level in the promoter of Adiponectin was significantly higher in UN oocytes compared with the control (Fig. 5E).

It is known that the methylation pattern in oocytes is different from somatic cells. To exclude somatic cell contamination from our result, we examined the DNA methylation status of paternally imprinted gene H19 and maternally imprinted gene Igf2r in oocytes using COBRA. All the samples for H19 were undigested by both of Taqα I and Rsa I, and the samples for Igf2r were digested by both of Taqα I and Bstu I (Additional file 1: Fig. S4F, G). And the BS results also showed that DNA methylation level of H19 in oocytes was low (Additional file 1: Fig. S4H). These data suggest that our samples are not contaminated by somatic cells.

Altered DMRs methylation in F1 oocytes is transmitted to F2 tissues

After fertilization, there is another round of reprogramming, and the tissue-specific methylation patterns are established during development. To assess whether the methylation changes in UN oocytes can be maintained in F2, we examined DNA methylation levels of some DMRs in UC tissues using pyrosequencing. Compared with the CC group, the methylation level of C-DMR was low in UC livers, but there was no statistical difference besides CpG site 1 (Fig. 6A). The methylation level of G-DMR in UC livers was similar to that in CC (Fig. 6B). The methylation levels of hyper-DMRs, H-DMR, and P4-DMR were significantly higher in UC livers compared with CC (Fig. 6C, D). The methylation level of Adiponectin in UC livers was not significantly different compared with CC (Fig. 6E). However, it was significantly higher in UC adipose tissues compared with CC (Fig. 6E). The methylation of Leptin in livers and adipose was examined using BS, and it was similar between CC and UC (Additional file 1: Fig. S5C). These results suggest that methylation changes in UN oocytes are partly maintained in UC tissues. The published data sets indicate that the validated DMRs-related genes are involved in metabolism. For example, Pde4d inactivates cAMP to inhibit insulin secretion [21], and Pde4d expression is regulated by Leptin [22]. Car8 decreases insulin level via reducing Glp-1 level [23]. Adiponectin can reduce the sensitivity of insulin [16]. Therefore, we further examined the expression of these genes in tissues and found that the mRNA expressions of Car8, Gsto2, and Pde4d were significantly higher in UC livers than that in CC (n = 10 from 6 litters, Fig. 6F). It is unlikely that the changed expressions of Gsto2 and Pde4d are directly mediated by methylation because the methylation of G-DMR and P4-DMR was higher in UC livers (Fig. 6C and E). But the increased expression of Car8 may be mediated by the decreased methylation in livers. The altered expression of Adiponectin and Hgsnat were not observed in UC livers (Fig. 6F). In adipose, the expression of Car8, Hgsnat, Pde4d, and Gsto2 was not observed (Additional file 1: Fig. S6), but the expression of Adiponectin was significantly reduced in UC (Fig. 6G) which may be mediated by the higher methylation in the promoter (Fig. 6E). The significantly increased expression of Leptin in livers and adipose (Fig. 6F and G) did not coincide with its methylation level at promoter, and this suggests that more factors are involved in regulating Leptin expression [18], such as the altered expression of Pde4d in livers. These results suggest that the intergenerational perturbed metabolism induced by undernourishment in utero is, at least partly mediated by methylation changes in UN oocytes.

Fig. 6

DMRs methylation and mRNA expression in F2 UC female tissues. A–E DMRs methylation levels in liver and adipose were examined using pyrosequencing, X-axial is the CpG sites which were tested. F The relative expression of genes related to DMRs was examined using qPCR in liver. G The relative expression of Leptin and Adiponectin in adipose. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Data are showed as mean ± SD

The altered DNA methylation of validated DMRs is observed in F2 oocytes

During embryo development, the genome undergoes a reprogramming process including DNA demethylation and remethylation, and tissue-specific methylation patterns are re-established. As shown in Figs. 5 and 6, the methylation patterns of validated DMRs and Adiponectin in control oocytes are different (lower) from tissues, and the methylation patterns are tissue-specific. These suggest that the altered methylation of validated DMRs in UN oocytes partially escape the reprogramming during F2 embryo development. So, we wanted to know whether the altered methylation levels of validated DMRs could be, at least partially, maintained in oogenesis. To assess this possibility, we examined the methylation of validated DMRs in UC oocytes using BS. Significant difference of average number of oocytes was not observed between UC and CC groups (Additional file 1: Fig. S7A), but average pups per litter in CC group was significantly higher than that of UC group (Additional file 1: Fig. S7B). Furthermore, we found that the methylation levels of Gn-DMR and P8-DMR in UC oocytes were similar to the CC oocytes (Additional file 1: Fig. S7). But the methylation levels of hyper-DMRs, P4-DMR, G-DMR, and H-DMR were significantly higher in UC oocytes than those in CC oocytes (Fig. 7A, B, and C), and the methylation of C-DMR in UC oocytes was lower than that in CC oocytes (Fig. 7D). Compared with CC, the methylation level of Adiponectin was also significantly higher in UC oocytes (Fig. 7E). These results suggest that a part of the altered methylation of validated DMRs in UN oocytes resists the reprogramming during oogenesis, and that may be transmitted to F3.

Fig. 7

Analysis of DMRs methylation in F2 UC oocytes. To assess the transgenerational inheritance of methylation changes, we examined the DMRs methylation levels in UC oocytes using BS. A P4-DMR. B G-DMR. C H-DMR. D C-DMR. E Adiponectin. **p < 0.01, ***p < 0.001. Black circle: methylated CG site; white circle: unmethylated CG site