Insulin Therapy for Gestational Diabetes Mellitus Does Not Fully Protect Offspring From Diet-Induced Metabolic Disorders

Gestational diabetes mellitus (GDM), defined as glucose intolerance first diagnosed during pregnancy, affects up to 15% of pregnancies in the world (1). GDM is associated with adverse consequences not only during fetal development, such as stillbirth, visceromegaly, and macrosomia, but also later in life (2,3). Accumulating evidence suggests that GDM, independent of maternal obesity and genetic background, predisposes offspring to metabolic disorders in later life, such as obesity, impaired glucose tolerance, and diabetes (4–6). Longitudinal studies of GDM offspring indicate that the maternal glucose level is a strong predictor of altered carbohydrate metabolism during childhood, which can be extended into adulthood (7,8).

The therapeutic management of GDM is critical to minimize these complications. Glycemic control is the cornerstone of GDM management (9). Randomized trials have confirmed that the therapy for GDM confers immediate benefits, such as reduction of perinatal complications and prevalence of macrosomia (10,11). However, whether the therapy for GDM confers long-term metabolic benefits to offspring is still unclear (12,13). Importantly, the appropriate offspring follow-up period for assessing the effects of GDM therapy is still debatable. Offspring enrolled in most follow-up studies were prepubertal (age 5–10 years), yet the long-term effect of GDM on offspring metabolic disorders or its reduction through therapy might not be evident until adolescence or adulthood (12,14).

In addition, offspring sex also has a profound effect. Sexual dimorphism in the response to insult in utero presents with an uneven disease susceptibility: although both sexes can be affected, one is more susceptible (15). Metabolic differences exist between male and female fetuses (15), and differing sensitivities to maternal hyperglycemia may result in sex-specific disease risks later in life (6,16). Evidence from epidemiological studies demonstrates that the effect of therapy for GDM also differs with fetal sex in infancy and childhood (11,16). Therefore, fetal sex may influence the effect that therapy for GDM has on offspring long-term health. Moreover, social and environmental factors could also confound the follow-up results. Thus, whether therapy for GDM is a modifiable risk factor for offspring metabolic disorders remains unknown.

Mechanistically, epigenetic modifications, such as DNA methylation, histone modification, and noncoding RNAs, provide a plausible link between environmental exposures early in development and the susceptibility to diseases later in life (17). DNA methylation, the most studied epigenetic modification, can alter the status of gene expression and be inherited mitotically in somatic cells (17,18), which provides a potential mechanism by which environmental effects on the epigenome could have long-term effects on gene expression. Results from animal and human studies support that intrauterine hyperglycemia could result in altered fetal DNA methylation patterns and subsequent changes in the risk of developing disease (6,19). In epidemiological and experimental studies, glycemic control improved GDM neonatal outcomes. However, no studies investigated whether these positive effects were accompanied by a favorable restoration of DNA methylation (20), which may be critically important for the susceptibility to disease later in life. Thus, the possibility that fetal metabolic programming in GDM can have long-term effect on health that may in turn be modified by therapy for GDM remains open to question.

Because excluding the confounders and analyzing the underlying mechanisms in humans is difficult, we established a mouse model of diabetes in the latter half of pregnancy to mimic human GDM with a high incidence during the third trimester of gestation (21). We treated maternal hyperglycemia with insulin and evaluated the pancreatic islet β-cell function in offspring. We also addressed whether lifestyle factors in adulthood, such as a high-fat diet (HFD), may increase the risk of developing metabolic disorders in offspring. Finally, we performed genome-wide DNA methylation sequencing in offspring pancreatic islets and assessed the alterations of candidate genes that may contribute to metabolic phenotypes in offspring.

The Zhejiang University Animal Care and Use Committee approved all animal protocols. All experiments were performed with Institute of Cancer Research (ICR) mice (22), which were purchased from Shanghai SLAC Laboratory Animal Co. (Shanghai, China). Virgin ICR females (age, 6–8 weeks; weight, 26–28 g) were mated with normal ICR males. Pregnancy was dated by the presence of a vaginal plug (day 0.5). Pregnant females were randomly assigned to control (Ctrl), GDM, or GDM + insulin therapy (INS) groups. On day 6 and day 12 of pregnancy, GDM and INS dams were fasted 8 h and received a streptozotocin (STZ) injection (100 mg/kg i.p.) (Sigma-Aldrich, St. Louis, MO) (23,24) (Fig. 1A). Control pregnant mice received an equal volume of citrate buffer. Blood glucose level was measured via the tail vein 48–72 h after the second STZ injection, and diabetes was defined as a blood glucose level between 14 and 19 mmol/L (6).

INS dams were treated with recombinant insulin (Novolin R; Novo Nordisk, Bagsvaerd, Denmark) by miniosmotic pumps (Alzet model 1007D; Durect, Cupertino, CA) at a dose of 0.35 IU/day. Pumps were implanted in INS dams on day 14.5 or 15 of pregnancy posterior to the scapulae under Avertin (Sigma-Aldrich) anesthesia (0.1 mL/20 g body wt). Control and GDM dams were implanted with pumps containing normal saline. To maintain stable glycemic levels, INS dams received another injection of 0.1 units of long-acting insulin (Levemir; Novo Nordisk) ∼1 h before the fed state (darkness) during late gestation. Only mice with nearly normal blood glucose levels in INS group were included in the further study.

At birth, litter size was equalized to eight. Pups from the GDM and INS groups were fostered by normal females until the age of 3 weeks. Offspring were designated as Ctrl-F1, GDM-F1, and INS-F1. At 8 weeks of age, offspring were divided into two groups either receiving a normal chow diet (NCD) or HFD (60% energy as fat; D12492; Research Diets, New Brunswick, NJ) until 20 weeks (Fig. 1A).

Intraperitoneal glucose tolerance tests (2 g/kg body wt) and insulin tolerance tests (ITT) (0.8 unit/kg body wt) were performed in unrestrained conscious mice after a 16- and 4 h-fast, respectively. Serum insulin level was assessed at overnight fasted state and 30 min after the glucose injection (Crystal Chem, Downers Grove, IL).

Serum cholesterol, triglyceride, nonesterified fatty acids, HDL, and LDL were assayed using a biochemical analyzer (TBA120FR; Toshiba, Tokyo, Japan). Serum leptin was determined by radioimmunoassay (North Institute, Beijing, China). HOMA-insulin resistance (IR) was calculated as follows: fasting serum insulin concentration (μU/mL) multiplied by fasting blood glucose level (mg/dL) divided by 405 (25).

Pancreatic islets were isolated from 20-week-old mice as previously described (26). For the indicated experiment, 10 islets (size-matched for each batches) were incubated for 1 h at 37°C in modified Krebs-Ringer bicarbonate buffer containing 2.8 mmol/L glucose and then incubated in the modified Krebs-Ringer bicarbonate buffer containing indicated glucose or various chemical compounds, including 200 μmol/L tolbutamide (Sigma-Aldrich), 250 μmol/L diazoxide (Sigma-Aldrich), 2 μmol/L Bay K8644 (Sigma-Aldrich), and 10 μmol/L nifedipine (Sigma-Aldrich). The supernatant was collected for insulin content assay (Abnova, Taipei, China).

The pancreata of fetal mice at embryonic day 17 was directly digested in 2 mg/mL collagenase with shaking incubation at 37°C for 25 min. After recovering overnight in RPMI medium, fetal islets were handpicked under a stereomicroscope and randomly separated into 5.6 mmol/L, 16.7/5.6 mmol/L, and 16.7 mmol/L glucose groups (Fig. 6A).

Male offspring fed the NCD were chosen at 20 weeks of age for methylated DNA immunoprecipitation sequencing (MeDIP-seq) analysis. Genomic DNA of pancreatic islets was extracted from nine mice per group and pooled by each of the three. The MeDIP-seq was performed as described previously (27). Briefly, DNA was sonicated to obtain the DNA fragments (200–700 base pairs). Sonicated genomic DNA was denatured and immunoprecipitated with anti–5-methylcytosine antibody (#28692; Cell Signaling, Danvers, MA). Illumina libraries were then created from the captured DNA by TruSeq DNA LT Sample Prep Kit (Illumina, San Diego, CA). The samples were sequenced on the Illumina HiSeq 3000 system (Illumina).

Total RNA from isolated pancreatic islets was extracted by using the RNeasy Micro Kit (Qiagen, Valencia, CA). cDNA was synthesized using oligo-deoxythymidylic acid and random primers (TaKaRa, Dalian, China) for real-time quantitative PCR with SYBR green detection (TaKaRa). The primer sequences are provided in Supplementary Table 1.

Genomic DNA was extracted from islets by using the TIANmap Micro DNA kit (Tiangen, Beijing, China). Bisulfite was converted using the EpiTect bisulfite kit (Qiagen) according to the manufacturer’s instructions. The methylation status of gene promoter regions was analyzed by pyrosequencing (28). In brief, pyrosequencing primers were designed by Qiagen PyroMark Assay Design 2.0 software (Qiagen) (Supplementary Table 2). The specificity of each PCR product was checked by agarose gel analysis. Pyrosequencing was conducted on a PyroMark Q24 pyrosequencer (Qiagen) by using PyroMark Gold Q24 reagents (Qiagen), and quantification of methylated and unmethylated alleles were performed by PyroMark Q24 software (Qiagen).

Fetal islets were identified by detecting insulin with immunofluorescence as previously described (29). Isolated fetal islets were incubated with antibody against insulin. All immunofluorescence images were acquired by laser scanning confocal microscope (Zeiss, Jena, Germany). The antibodies used are described in Supplementary Table 3.

The protein was extracted from mouse islets as described before (29). Samples were separated using 8% SDS-PAGE. Western blots were performed using polyvinylidene fluoride membrane and the antibodies listed in Supplementary Table 3. Protein bands were visualized by the enhanced chemiluminescence system (Pierce, Rockford, IL).

The identical amount of STZ was injected in pregnant mice as described above. Only a glucose level of <7.5 mmol/L was considered as an STZ-injected nondiabetic model. Offspring were fostered by normal females until 3 weeks old. Mice were fed the HFD, and glucose tolerance test, and ITT were performed as described above. Islets from the STZ-injected nondiabetic offspring were isolated for further analysis.

All data are shown as mean ± SEM. Statistical analysis were performed by one-way ANOVA, followed by least significant difference post hoc test and two-tailed Student t test, as described in the table and figure legends, using SPSS 17.0 software. P < 0.05 was considered statistically significant.

We established a mouse model of hyperglycemia during the midlate stage of pregnancy. GDM dams were averaging a 328.6 mg/dL plasma glucose level after two STZ injections. Blood glucose level declined and averaged 129.5 mg/dL during insulin therapy in the INS group (Fig. 1B). Gestational length, litter size, and birth weight were similar among the three groups (Supplementary Table 4 and Supplementary Fig. 1).

Notably, increased body weight and most metabolic abnormalities were seen in GDM-F1 adult offspring (Fig. 1C and D and Table 1). Insulin therapy was associated with normal weight trajectory, serum glucose, and lipid metabolism in adult offspring (Fig. 1C and D and Table 1). However, these associations were observed only in NCD offspring. When fed the HFD after 8 weeks of age, a significant increase in body weight (Fig. 1C), fasting insulin levels, and lipid levels was seen in INS-F1 males (Table1). Females in HFD INS-F1 group only exhibited increased body weight compared with HFD Ctrl-F1 females (Fig. 1D).

Insulin therapy for GDM resulted in a distinct rescue of glucose intolerance in INS-F1 male offspring, with only an increase in glucose levels at 60 min after injection (Fig. 1E). But strikingly, the HFD abolished this protection (Fig. 1E). Results of the ITT showed only GDM-F1 male mice exhibited significant impaired insulin sensitivity with aging in the NCD group (Supplementary Fig. 2B and Fig. 1G). However, when challenged with the HFD, not only GDM-F1 males developed much more serious insulin intolerance, but INS-F1 males also exhibited a pronounced impairment of insulin tolerance compared with controls (Fig. 1G). In female offspring, only GDM-F1 females showed an elevation of glucose level at 30 and 120 min after the insulin injection (Fig. 1H).

Insulin secretion defects could also contribute to glucose intolerance. We assessed glucose-stimulated insulin secretion (GSIS) in vivo and in vitro. In vivo, GSIS was reduced in both male and female GDM-F1 offspring (Fig. 2A–D). In the INS-F1 group, only males exhibited lower insulin levels in response to the glucose injection (Fig. 2A and B). In vitro, insulin secretory response to 5.6 mmol/L glucose was similar among all groups (Fig. 2E and F); however, islets of GDM-F1 males or females showed impaired insulin secretion when exposed to 16.7 mmol/L glucose (Fig. 2E and F). Defective insulin response to high glucose (16.7 mmol/L) was also seen in INS-F1 males (Fig. 2E). No significant difference was found in INS-F1 females in glucose tolerance, insulin sensitivity, or GSIS (Figs. 1F and H and 2C, D, and F).

GDM altered the methylation of 220 upstream2k (8.11%), 201 downstream2k (7.41%), 77 5′ untranslated region (UTR) (2.84%), 55 3′UTR (2.03%), 711 coding sequence (26.2%), and 1,448 intron element-associated genes (53.37%) (Fig. 3B and Supplementary Table 5) in islets of GDM-F1 male offspring, respectively. Global cytosine methylation status was also altered in INS-F1 offspring, and the methylation of 258 upstream2k (8.37%), 237 downstream2k (7.69%), 81 5′UTR (2.63%), 70 3′UTR (2.27%), 745 coding sequence (24.17%), and 1,690 intron element-associated genes (54.83%) was changed, respectively (Fig. 3B and Supplementary Table 6). In GDM-F1, there were 1,326 element-associated hypermethylated genes, and 326 were also hypermethylated in INS-F1 islets (Fig. 3C), and there were 1,350 element-associated hypomethylated genes, with 493 also hypomethylated in INS-F1 (Fig. 3D).

KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis showed that the differentially methylated genes in GDM-F1 and INS-F1 mainly encoded the ion channels in islet β-cell and are involved in insulin secretion. These genes selected for validation were Abcc8 (encoding sulfonylurea receptor 1, Sur1, belonged to ATP-binding cassette superfamily of transporters), Cacna1c (Cav1.2, encoding one subunit of L-type Ca²⁺ channels with widespread expression in mouse, rat, and human islets β-cells), Cacna1e (Cav2.3, encoding the R-type Ca²⁺ channel and expressed in rodent and human), and Cacna1g (Cav3.1, encoding T-type Ca²⁺ current and mainly expressed in NOD mouse, rat, and human) (Fig. 3E). MeDIP-seq data showed that these candidate genes displayed hypermethylation status in GDM and INS offspring islets compared with controls.

The mRNA and protein levels of Abcc8, Cav1.2, and Cav2.3 were all significantly lower in GDM-F1 and INS-F1 males (Fig. 4A–G). In addition, HFD exposure downregulated Cav1.2 expression in all groups, but GDM-F1 and INS-F1 males showed dramatically decreased expression of Cav1.2 after HFD feeding (Fig. 4B, D, and F). The similar alteration of gene expression was also seen in GDM-F1 females (Supplementary Fig. 3A–C). But in INS-F1 females, only Cav2.3 showed decreased expression (Supplementary Fig. 3C). There was no significant difference in Cav3.1 expression among the groups (data not shown).

Further, insulin secretion responses to ion channel agonists and inhibitors were tested in the isolated pancreatic islets from male offspring. Tolbutamide, a K_ATP channel blocker, stimulated insulin release by 5.52-fold in NCD control islets and 5.05-fold in HFD control islets, respectively, but tolbutamide responses were significantly reduced in GDM-F1 and INS-F1 groups (Fig. 4H). Diazoxide, a K_ATP channel agonist, significantly inhibited insulin secretion in control islets but did not have such a significant effect on GDM-F1 and INS-F1 (Fig. 4I). Bay K8644, an L-type Ca²⁺ channels agonist, stimulated insulin secretion by 25.18-fold in NCD controls, but islets from GDM-F1 and INS-F1 mice showed decreased insulin secretion by 13.13-fold and 16.35-fold in response to Bay K8644, respectively (Fig. 4J). In addition, HFD exposure reduced the responses to Bay K8644 in control islets, but a more significant decreased insulin secretion was observed in HFD-fed GDM-F1 and INS-F1 islets (Fig. 4J). Nifedipine, an L-type Ca²⁺ channels inhibitor, significantly suppressed insulin secretion in control islets, but nifedipine revealed a minor decrement of insulin secretion in INS-F1, and a minimal suppression in GDM-F1 (Fig. 4K). The inhibitory effect of nifedipine on insulin secretion was not significantly affected by HFD exposure, although insulin secretion showed trends to increased levels in GDM-F1 and INS-F1 HFD males (Fig. 4K).

Pancreatic islets of 20-week-old offspring were isolated, and pyrosequencing was used to analyze the methylation status of 10 cytosine phosphate guanine (CpGs) of the Abcc8 promoter, CpGs of Cav 1.2 promoter, and 11 CpGs of the Cav 2.3 promoter. The CpGs of Abcc8, Cav1.2, and Cav2.3 showed significantly higher DNA methylation status in islets of GDM-F1 and INS-F1 males (Fig. 5A–C). Compared with GDM-F1 males, altered DNA methylation status in Cav1.2 and Cav2.3 was improved to a different degree in INS-F1 males (Fig. 5B and C). Further, we found that the effect of GDM treatment on DNA methylation in the three target genes with a sex-specific difference (Fig. 5 and Supplementary Fig. 3). Notably, maternal glycemic control was associated with a distinct restoration of DNA methylation levels in Abcc8 and Cav1.2 in INS-F1 females (Supplementary Fig. 3D and E) and a moderate hypermethylated level in Cav2.3 promoter regions (Supplementary Fig. 3F). In addition, HFD feeding caused a significantly higher DNA methylated level at Cav1.2 promoter regions in GDM and INS offspring (Fig. 4B and Supplementary Fig. 3E), but the effect of the HFD on DNA methylation was not found in Abcc8 and Cav2.3.

We collected islets from normal mice at embryonic day 17 to verify whether the high-glucose environment directly affected islet gene expression and DNA methylation. Treating fetal islets with high glucose (16.7 mmol/L) for the entire 6 days significantly increased Abcc8, Cav1.2, and Cav2.3 promoter DNA methylation levels and reduced their gene expression compared with physiological glucose level (5.6 mmol/L) (Fig. 6C and F–H). In the 16.7/5.6 mmol/L group, the increased Abcc8, Cav1.2, and Cav2.3 DNA methylation levels and reduced gene transcription persisted during subsequent culture at 5.6 mmol/L glucose (Fig. 6C and F–H). Moreover, we examined DNA methylation enzyme (Dnmts) and demethylation enzyme (TETs) expression in fetal islets after high glucose exposure. We found that the Dnmt1, but not Dnmt3a/3b, expression level was upregulated in the 16.7 mmol/L and 16.7/5.6 mmol/L groups (Fig. 6D) and that TET2 and TET3, but not TET1, expression levels were also downregulated in 16.7 mmol/L and 16.7/5.6 mmol/L groups (Fig. 6E).

STZ, a cytotoxic agent, is targeted to islet β-cells to induce diabetes in specific species (30). In contrast to adult pancreatic β-cells, STZ had no cytotoxic effect on fetal proislets (31). Although STZ does cross the placenta to a limited extent, it is evident that the fetal pancreas does not concentrate this cytotoxic agent (32). Furthermore, STZ’s half-life is very short (5–15 min) in vivo (30), but differentiation of pancreatic endocrine cells occurs after day 15 of gestation in the mouse (33), implying that a low dose of STZ may not directly act for offspring metabolic dysfunction and alterations in gene expression. To further evaluate STZ’s effect on offspring, we collected the nondiabetic pregnant mice administered STZ alone. No significant difference was found with respect to glucose tolerance or insulin sensitivity in NCD or HFD offspring between mice administered STZ and the control group (Supplementary Fig. 4). Quantitative PCR analysis also showed no significant changes in ion channel expression in the group injected with STZ (Supplementary Fig. 5).

In this study, we provide the first experimental evidence addressing the effects of insulin therapy on the long-term metabolic health of GDM offspring. Insulin therapy resulted in a significant improvement of metabolic disorders in offspring fed the NCD. But importantly, in response to the HFD, the offspring developed significantly exacerbated glucose intolerance, obesity, and insulin resistance. These abnormalities were more obvious in males than in females, suggesting that males might be more vulnerable to the adverse environment. The findings indicate that predisposition to metabolic disorders still persisted in offspring even with efficient insulin therapy for GDM and was significantly enhanced by the HFD challenge.

For the offspring with insulin therapy for GDM, impaired glucose tolerance (IGT) is an early key phenotype with a major contribution from β-cell dysfunctions. In vivo and in vitro experiments confirm that the offspring exhibited reduced GSIS. Epigenetic modifications provide a plausible link between the intrauterine environment and alterations in gene expression that may lead to a disease phenotype (17). In accordance with the sex differences of phenotypes, the alterations in gene expression and DNA methylation were also more obvious in males than that in females. In the male offspring, MeDIP-seq data of pancreatic islets showed hypermethylated status of genes that regulate insulin secretion, including ion channel genes Abcc8, Cav1.2, and Cav2.3. Consistently, downregulated expression of Abcc8, Cav1.2, and Cav2.3 and impairment of K_ATP and L-type Ca²⁺ channels that mediate insulin secretion were observed. However, the female offspring only exhibited higher DNA methylation and lower expression of Cav2.3, suggesting that at least at the promoter regions of the three target genes, the epigenetic modification of the male fetus might be more sensitive to the intrauterine hyperglycemia than the female fetus.

All of the Abcc8, Cav1.2, and Cav2.3 ion channel are important but with different functions in pancreatic islets. In agreement, Abcc8 encodes a regulatory subunit of the K_ATP channel, coupling the blood glucose level to membrane electricity activity and insulin secretion (34). Abcc8^−/− mice display moderate glucose intolerance, and isolated islets from Abcc8^−/− mice show impaired first-phase insulin secretion and response to tolbutamide (35). Cav1.2, encoding a subunit of the L-type Ca²⁺channel, functions as an important role in the Ca²⁺ entry pathway (36). Pancreatic β-cell–selective Cav1.2 ablation decreases the whole-cell Ca²⁺ current by only 45% but almost abolishes first-phase insulin secretion and causes glucose intolerance (37). In contrast, Cav2.3 mediates the second phase of insulin secretion, involved in recruiting insulin granules from pools that are not immediately available to release insulin when glucose loaded (38,39). Isolated islets from Cav2.3^−/− mice exhibited reduced insulin content but normal GSIS, implying that the Cav2.3-deficiency may be offset by other compensatory mechanisms (40). This may explain the sex-different phenotype in offspring after insulin therapy for GDM. Compared with the obvious glucose intolerance of male offspring, the female offspring with decreased expression of Cav2.3 alone showed normal glucose tolerance and GSIS.

Further, in vitro culture confirmed the effect of short high-glucose exposure on Abcc8, Cav1.2, and Cav2.3 gene expression and DNA methylation in fetal islets. Our animal model, together with in vitro culture, provides evidence that early development is sensitive to the extrinsic factors (41) and that short exposure to intrauterine hyperglycemia is sufficient to persistently affect ion channel gene expression and DNA methylation. Although direct transfer of our experimental results to the human situation warrants caution, it is important to recognize that freedom from symptoms is one of the major difficulties with GDM, and the pregnant woman is usually unaware that she has GDM until it is diagnosed at routine prenatal screening (42), suggesting that the fetus might already be exposed to the adverse intrauterine environment and exhibit adaptive changes in the epigenome (43).

In addition, the in vitro experiment showed that the altered gene expression of DNA methyl-writing and methyl-erasing enzymes persisted during subsequent normal glucose culture, indicating that other deleterious factors induced by hyperglycemia may also contribute to the sustained epigenetic alterations. Maternal glucose can freely permeate the placenta, and glucose excursions not only cause fetal hyperglycemia but also induce fetal hyperinsulinemia and oxidative stress (44,45). Although insulin therapy for GDM normalized maternal blood glucose level, whether the insulin intervention reversed other adverse factors is unknown. If not, these factors may persist to affect fetal development and epigenetic modifications (45).

Conditions in the postnatal environment are also important cues for inducing adult metabolic diseases (46). In our study, HFD exposure exacerbated glucose intolerance. But importantly, this HFD-induced prediabetic state in GDM offspring, whether with insulin therapy or not, was more severe than that observed in control offspring, suggesting that early fetal insult could impair the ability to adapt to an HFD. Extrinsic factors can also affect the epigenome postnatally (46–48). We consistently found that the HFD caused a higher DNA methylation level of Cav1.2. However, compared with control offspring, the pre- and postnatal factors act synergistically to induce a more significant hypermethylated level of Cav1.2 in the offspring with insulin therapy for GDM, which may partly contribute to the exacerbated glucose intolerance after HFD exposure.

An additional factor likely to be responsible for the exacerbated glucose intolerance is insulin resistance. Defective insulin secretion and action are two major results of diabetes (49,50). Notably, even with insulin therapy for GDM, insulin resistance arose when the offspring were challenged with the HFD in adulthood. Although the underlying mechanisms are still unknown, it is interesting to note that these offspring in HFD group displayed significant obese phenotypes, suggesting that insulin resistance might be associated with the overweight and lipid metabolism disorders.

In summary, our study provides novel experimental evidence about the effects of insulin therapy for GDM on the long-term health of the offspring, revealing that these offspring are still susceptible to metabolic disorders, especially exposed to an adverse postnatal environment. Although our finding was generated in a mouse model, it is important to recognize that even with efficient insulin therapy for GDM, follow-up and lifestyle interventions are still necessary to the offspring during postnatal life. Further, the results show that short exposure to maternal diabetes during early development is sufficient to cause persistent alterations in DNA methylation and expression of genes that regulate insulin secretion, suggesting a methylation-mediated epigenetic mechanism for GDM-induced intergenerational glucose intolerance. These altered epigenetic markers not only partly explained the susceptibility in GDM offspring, but importantly, will also hopefully allow for the early diagnosis and therapy for individuals with a propensity for adult-onset disease. In light of our results, efficacious screenings and more early interventions should be administrated in GDM patients. More importantly, further elucidation of the molecular events that enable, before glycemic control, to result in offspring protection may lead to the development of new approaches for reducing the fetal-originated adult diseases.

Funding. This work was supported by the Special Fund for the National Key Research and Development Plan grant (no. 2017YFC1001300), the National Natural Science Foundation of China (no. 31671569, no. 81490742, no. 31471405, and no. 31571556), the Municipal Human Resources Development Program for Outstanding Young Talents in Medical and Health Sciences in Shanghai (no. 2017YQ047), and the Fundamental Research Funds for the Central Universities.

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

Author Contributions. H.Z. and B.C. designed and performed experiments and analyzed data. H.Z. and G.-L.D. wrote and edited the manuscript. Y.C., Y.Z., and Y.-S.Y. contributed to study design, conducted experiments, and assisted with the data analysis. Q.L. and Y.J. contributed to the discussion and edited the manuscript. J.-Z.S. contributed to the study design and discussion and edited the manuscript. G.-L.D. and H.-F.H. designed and supervised the research, contributed to discussion, and edited the manuscript. H.-F.H. is the guarantor of this work and, as such, had full access to all data in the study and takes responsibility for the integrity and accuracy of data analysis.